Annex 5.3 PICO 1 - How To Test (HBV) Diagnostic Accuracy of Tests To Detect Hepatitis B Surface Antigen: A Meta-Analysis and Review of The Literature

Download as pdf or txt
Download as pdf or txt
You are on page 1of 78

Annex 5.

PICO 1 - How to test (HBV)


Diagnostic accuracy of tests to detect hepatitis B
surface antigen: a meta-analysis and review of the
literature

Ali Amini,* Helen Kelly,* Debi Boeras, Wen Chen, Jane Falconer, Helen Kelly,
Weiming Tang, Joseph Tucker, Olivia Varsaneux, Rosanna Peeling (Team lead)
London School of Hygiene and Tropical Medicine team
*Co-leaders of this review

September 2015

Page | 188
1. Executive summary
Background: Rapid diagnostic tests are potentially useful tools for the diagnosis of hepatitis B
surface antigen (HBsAg) globally, particularly in low-resource areas. Expansion for global use
depends on their performance characteristics clinically in the field, ultimately with the aim
being to reach low-resource settings and offer cost–efficient screening as an alternative to
laboratory tests.
Objectives: The purpose of this review was to identify quantitative evidence on the clinical
sensitivity1 and specificity of available in vitro diagnostics (hereafter referred to as assays)
used to detect hepatitis B antibody, synthesize the evidence, and inform models.
Methods: Two reviewers independently assessed the quality and extracted data for
estimating accuracy. Meta-analysis was performed. We further performed stratified estimates
based on individual products, HIV status, specimen type, study setting and design.
Results: Thirty-three studies were included using an EIA reference standard. The overall
pooled clinical sensitivity and specificity of rapid HBsAg tests were 90.0% (95% CI: 89.1, 90.8)
and 99.5% (95% CI: 99.4, 99.5), respectively, compared to laboratory-based immunoassay
reference standards. Pooled specificity was comparable and less heterogeneous.
Pooled sensitivity in studies of HIV-positive was lower than in known HIV-negative
patients; 72.3% (95% CI: 67.9, 76.4) compared to 92.6% (95% CI: 89.8, 94.8), respectively.
Pooled sensitivity and specificity in blood donors were 91.6% (95% CI: 90.1, 92.9) and 99.5%
(95% CI: 99.5, 99.9), respectively.
Samples using whole blood specimens (venous or capillary) were 91.7% (95% CI: 89.1,
93.9) and 99.9% (95% CI: 99.8, 99.9) sensitive and specific compared to serum.
Results were comparable for studies performed prior to the past ten years as those
performed since 2005. Estimates of assay sensitivity demonstrated significant heterogeneity
not entirely corrected by sub-analysis, although studies using whole blood specimens (venous
or capillary), and the same reference standard (CMIA) were more robust (tau-squared <1 in all
cases).
The overall pooled clinical sensitivity and specificity of laboratory-based HBsAg tests
were comparable, with 88.9% (95% CI: 87.0, 90.6) and 98.4% (95% CI: 97.8, 98.8) sensitivity
and specificity, respectively, compared to immunoassay state-of-the-art chemiluminescent
microparticle enzyme immunoassays.
Conclusions: Assays for HBsAg detection such as rapid diagnostic tests (RDTs), including those
performed on serum/plasma and capillary whole blood specimens, have good sensitivity and
excellent specificity compared to a reference standard comprising laboratory-based methods
of HBsAg detection. Improvement in sensitivity, or development of innovative testing
strategies could potentially enhance their use as first-line screening globally. Caution in HIV-

1
Unless otherwise specified, sensitivity and specificity refer to clinical and not analytical sensitivity for tests.

Page | 189
positive individuals is important, while the reassuring accuracy of capillary whole blood
specimens compared to plasma/serum further facilitates use in settings where phlebotomy is
not available.

2. Background
An estimated 240 million individuals worldwide1 are chronically infected with hepatitis B virus
(HBV) and there are an estimated four million acute HBV infections each year. Twenty per
cent to 30% of those with chronic hepatitis B infection will develop cirrhosis2 or
hepatocellular carcinoma,3 leading to approximately 650 000 deaths each year.4 However,
most individuals with chronic HBV infection are not aware of their serostatus, contributing to
delayed diagnosis and complications from advanced disease.5 HBV testing is critically
important in order to refer infected individuals to HBV treatment and care, to refer uninfected
individuals for vaccination, and to mobilize prevention and control efforts.
In March 2015, the World Health Organization (WHO) published the first guidelines
for the prevention, care, and treatment of individuals with chronic HBV infection.5 These
guidelines focused on assessment for treatment eligibility, initiation of first-line therapies,
switching and monitoring. These initial guidelines did not include testing recommendations,
and in particular which tests to use. Given the large burden of HBV in low- and middle-income
settings where there are limited or no existing HBV testing guidelines, there is a substantial
need for HBV testing guidelines.
Chronic HBV infection is defined as persistence of hepatitis B surface antigen (HBSAg)
for at least six months. However, interpretation of HBV serologies is complex. The serologies
most frequently used for HBV testing include HBsAg, total anti-HBc, and anti-HBs. HBV
screening includes both the one-test (e.g. HBsAg) and two-test strategies (e.g. HBsAg followed
by hepatitia B core antibody [HBcAb] or nucleic acid testing [NAT]).
Detection of HBsAg can include rapid diagnostic tests (RDTs) or immunoassays. Rapid
tests developed for screening include solid-phase assays, flow-through, agglutination and
lateral-flow. The majority, however, are immunochromatographic assays. Immunoassays use
different methods for detection of HBsAg using polyclonal or monoclonal anti-HBs antibodies.
Labelling to measure antigen–antibody complexes can include radioactive compounds,
enzymes with a change in colour in solution, or substances emitting light.
Advances in HBV detection technology create new opportunities for enhancing screening,
referral, and treatment. Previous systematic reviews on hepatitis B infection have focused on
immunological responses,6 surveillance of cirrhosis,7 and treatment.8 Existing systematic
reviews9‒11 on hepatitis B testing focused on point-of-care (POC) tests and included tests with
unclear reference standards or those not appropriate for assessing operational diagnostic
accuracy in the field.

3. Objectives
The purpose of this review was to identify quantitative evidence on the sensitivity and
specificity of assays used to detect hepatitis B antibody, synthesize the evidence and inform
models.

Page | 190
To our knowledge, this is the first study exclusively comparing the clinical
performance of both RDTs and laboratory-based immunoassays, in addition to addressing the
question of accuracy in the context of HIV specifically.

PICO 1 Among persons identified for hepatitis B testing, what is the diagnostic
accuracy of available assays for detecting HBsAg?

P Persons identified for HBV testing

I Rapid diagnostic tests or immunoassays for HBsAg detection

C Reference standard comprising a laboratory-based enzyme immunoassay


(EIA)2 (one or more HBsAg enzyme immunoassays, with or without
neutralization to confirm)

O Diagnostic accuracy (sensitivity, specificity, TN, TP, FN, and FP; positive
predictive value, negative predictive value)

As a subanalysis, we also analysed data for studies comparing the accuracy of HBsAg assays
against a nucleic-acid amplification test (NAT) reference standard. This is important given the
importance of reducing transmission during the seroconversion period and in the diagnosis of
occult hepatitis B, where HBsAg may not be detectable and which is more common in HIV
coinfection. [Results in Annex 9.2]

4. Methods
Search strategy and identification of studies

Literature search strategies were developed by a medical librarian with expertise in systematic
review searching. Our search algorithm consisted of the following components: hepatitis B,
diagnostic tests and diagnostic accuracy (see Annex 1). We searched MEDLINE (OVID
interface, 1946 onwards), EMBASE (OVID interface, 1947 onwards), the Cochrane Central
Register of Controlled Trials (Wiley interface, current issue), Science Citation Index Expanded
(Web of Science interface, 1970 onwards), Conference Proceedings Citation Index-Science
(Web of Science interface, 1990 onwards), SCOPUS (1960 onwards), Literatura Latino-
Americana e do Caribe em Ciências da Saúde (LILACS) (BIREME interface) and WHO Global
Index Medicus. The search was supplemented by searching for ongoing studies in WHO’s
International Clinical Trials Registry.
In addition to searching databases, we contacted individual researchers, experts
working in the field and authors of major trials to address whether any relevant manuscripts
are in preparation or in press. The references of published articles found in the above

2
For convenience we shall refer to all laboratory-based immunoassays for HBsAg detection (ELISA, MEIA, ECLIA,
CMIA) as EIAs as most have some form of enzymatic amplification.

Page | 191
databases were searched for additional pertinent articles. The review was registered in
PROSPERO and reported in accordance with PRISMA guidelines.
Study selection proceeded in three stages. First, titles/abstracts were screened by a
single reviewer (AA/HK) according to standard inclusion and exclusion criteria. Second, all
articles identified for full manuscript review were obtained and assessed independently by the
two reviewers (AA and HK) against inclusion criteria. Papers were accepted or rejected and
reasons for rejection were specified. Third, discrepancies were resolved by discussion
between review authors, with several studies resolved by a third independent reviewer (RP).

Selection criteria

Types of studies
We included case–control, cross-sectional, cohort studies and randomized trials with a
primary purpose of evaluating HBsAg tests published until May 2015. We excluded:
conference abstracts, comments or review papers; studies with primary aims other than
evaluation of both sensitivity and specificity of HBsAg detection; studies related to disease
prevalence, drug resistance, genotyping, sequencing, or non-diagnostic purposes; studies that
focus on detection of anti-HBsAg (antibody to hepatitis B antigen); articles in languages other
than English.

Participants
We included studies using original data from patient specimens in defined populations. We
included all age groups, settings, countries and specimen types, notably whole blood (venous
and capillary), plasma or serum. Saliva specimens were considered, but no suitable studies
were identified. We excluded studies using commercial reference panels or clinical panels not
sourced by authors given applicability and bias concerns from unknown sampling in unclear
populations.

Index tests
Studies utilizing commercially available HBsAg tests were eligible for inclusion. We excluded:
in-house developed tests; laboratory-based immunoassays which are no longer commercially
available. We did not, however, exclude rapid tests based on current commercial availability,
in keeping with the methodology in recent systematic reviews. We did however sub-
categorize more recent studies between 2005 and 2015.

Reference standard
The reference standard for definite diagnosis of hepatitis B is complicated, given the different
viral kinetics of HBsAg and HBV DNA. We included studies using an established commercially
available immunoassay as a reference standard for HBsAg detection. Studies using a NAT
reference standard were included as a supplementary secondary analysis. For studies
comparing immunoassays, we only included studies using chemiluminescent microparticle

Page | 192
immunoassays (CMIAs) as the reference standard, given the generally accepted higher
analytical sensitivity of these assays. For studies of rapid tests, in keeping with previous
systematic reviews, we did not limit based on type of immunoassays.

Data extraction
Information on the following variables were independently extracted by the two review
authors (AA, KH): first author, total sample size, country (and city) of sampling, specimen type
(oral fluid, capillary [finger-prick] whole blood, venous whole blood, etc.), eligibility criteria,
reference standard, manufacturer, raw cell numbers (true positives, false negatives, false
positives, true negatives), HIV coinfection, sources of funding, reported conflicts of interest.

Assessment of methodological quality


Study quality was evaluated using the QUADAS-2 tool12 and the STARD checklist.13 QUADAS
includes domains to evaluate bias in the following categories: risk of bias (patient selection,
index test, reference standard, flow and timing); applicability concerns (patient selection,
index test, reference standard).

GRADE summary of finding tables


The GRADE system was used to rate the strength of evidence of each body evidence as high,
moderate, low, or very low on the basis of aggregate quality, consistency, precision,
directness, and reporting bias. We did not downgrade directness because the outcome
evaluated was diagnostic accuracy (an intermediate outcome), as the PICO focuses on
diagnostic accuracy. Summary of finding tables were prepared. We did not perform a formal
assessment of publication bias.

Data analysis and synthesis


Standard methods for meta-analysis of diagnostic accuracy were used. For each study we
calculated sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR)
and 95% confidence intervals (CI). Likelihood ratios are metrics and are calculated using a
combination of sensitivity and specificity. Positive likelihood ratio (PLR) is the ratio of
sensitivity (1−specificity), whereas the negative likelihood ratio (NLR) is the ratio of specificity
(1−sensitivity). When a diagnostic test has no discriminating capability, both likelihood ratios
equal 1.
We conducted meta-analysis pooling data using the DerSimonian–Laird bivariate
random effect model (REM) to calculate pooled sensitivity, specificity and other related
indices. The REM was more suitable than the fixed-effects model (FEM) given significant
heterogeneity in studies found, as it takes into account variability both within studies (random
error) and between studies (heterogeneity). Statistical heterogeneity was measured using the
random-effect variance (tau-squared).
Meta-regression was performed to investigate the source of heterogeneity within
included studies (inverse variance weighted). We analysed sub-groups according to the
following characteristics: study year (2005–2015); geographical area; individual tests; sample

Page | 193
type; patient type (blood donors); patient HIV status. We did not formally test for publication
bias. Where meta-regression was not possible for covariates, we performed descriptive
statistics.
Meta-analysis of the collected data was conducted using the software: Meta-Disc© version
1.4.7. Statistical analysis was performed using Meta-Disc 1.4 for Windows (XI Cochrane
Colloquium, Barcelona, Spain). QUADAS-2 analysis was performed using Microsoft Excel.

5. Results

Study selection

A total of 11 589 citations were identified and 6575 duplicates were removed. Each of the
5014 titles was examined according to pre-specified inclusion and exclusion criteria. A total of
33 research studies were included in the final primary analysis (Fig. 1), with studies comparing
both rapid diagnostic tests (RDTs)14–42 and enzyme immunoassays43–46 against an immunoassay
reference standard. Of these, 19 studies14–31, 47 were also included in a recent systematic
review by Khuroo et al.;11 eight papers from that study were excluded as they were
conference abstracts or letters to editors,48–50 foreign language articles,51, 52 or evaluated
reference panels.53 Two reports by WHO and the International Consortium for Blood Safety
(ICBS) were also excluded as they were not published in peer-reviewed journals and were
case–control studies constructed using reference panels from populations of affected
individuals. Our search identified 11 additional articles32–42 comparing RDTs against the EIA
reference standard not found in the previous review. Six articles exclusively assessed accuracy
in cohorts of HIV-positive individuals.20, 36–38, 54, 55
Studies evaluating laboratory-based immunoassays for HBsAg detection as the index
test all used state of the art CMIAs as the reference standard. Seven studies were included in
the supplementary analysis assessing diagnostic accuracy of HBsAg assays against a nucleic-
acid based reference standard.47, 54–59

Page | 194
Fig. 1. PRISMA flow diagram outlining study selection examining diagnostic accuracy of
HBsAg assays in our systematic review

Page | 195
Study characteristics

Immunoassay reference standard [Table 1a; 1b]


Overall, the 33 included articles originated in 23 countries: Australia,35 Western Europe16, 18,
Western Africa;15, 25, 28, 34, 36, 38‒40 Eastern Africa;19, 20, 26, 30, 41 South Africa;38 South-eastern Asia;24,
32
Eastern Asia;25, 27, 31, 43 Southern Asia14, 17, 21, 22, 29, 42 and the United States.23 Thirty-three RDTs
were evaluated using serum (fresh or frozen), plasma, venous and capillary whole blood.
Sample sizes varied from 25 to 3928 individuals (mean, 631). Sixty-three data points were
generated from the articles. All had bivariate data for analysis; 38% of data points were
generated from case–control studies, while the remainder were from cross-sectional cohort
studies. Prevalence of hepatitis B in populations tested ranged from 1.9% to 84%.
Populations studied included: general screening of healthy volunteers and blood
donors; screening of at-risk populations such as pregnant women, incarcerated adults; and
patients from hepatitis cohorts, including confirmed hepatitis B.
The majority of studies were performed in laboratory settings, with some studies
performed in the field. A mixture of serum, plasma, capillary and whole blood was used for
RDTs, with serum or plasma used for EIAs. Seven studies assessed performance using capillary
or venous whole blood specifically.15, 16, 20, 24, 28, 37, 40
In terms of tests used, only the Determine HBsAg and BinaxNOW had more than 3
data points.

NAT reference standard [Table 8; Annex 9.2]


Overall, the 7 articles originating in 4 countries (Iran, Nigeria, Uganda, South Africa),
contributed 18 data points and included serum or plasma samples from 3304 individuals, with
sample size ranging from 74 to 950. All studies were either cross-sectional, cohort or case–
control studies. All had data for bivariate analysis.

Page | 196
Table 1a. Study characteristics – EIA vs EIA

Study Location Sample Study Setting Sample Assay under evaluation Reference standard
3 3
[Author, Year] [Country, City] size design [Type, Brand] [Type, Brand]

Liu, 2014 China 250 CC Hospital patients; outpatients Serum ECLIA, Cobas CMIA, Architect HBsAg
(preselected based on CMIA quantitative ELISA, Wantai
results)

Peng, 2011 China 498 CC Hospital patients Serum ELISA, KHB CMIA, Architect HBsAg
(preselected based on S/CO from KHB
screen)

Geretti, 2010 Ghana, Kumasi 838 CS – CSQ HIV clinic (1/3 on lamivudine) Serum CMIA, Architect HBsAg CMIA, Architect/ Liaison
*
CMIA, Liaison Ultra EIA, Murex v3
EIA, Murex v3

Ol, 2009 Cambodia 120 CS – CSQ Blood donors (rural community) Serum ELISA, Monolisa CMIA, Architect HBsAg

Viet, 2012 Vietnam 119 CS – CSQ Blood donors (rural community) Serum EIA, Monolisa Ultra CMIA, Architect HBsAg

*Reactive all three assays OR reactive in one assay with neutralisation


CMIA: chemiluminescent microparticle enzyme immunoassay; ECLIA: electrochemiluminescent immunoassay; EIA: enzyme immunoassay; ELISA: enzyme-linked
immunosorbent assay; CC: case–control; CS: cross-sectional; CSQ: consecutive patients

3
Abbreviated names for table clarity – full product names in Annex

Page | 197
Table 1b. Study characteristics – RDT vs EIA

Study Location Sample Study Setting Sample RDT under Reference test
[Author, Year] [Country, City] size design evaluation [Type, Brand]
[Type, Brand]

Mvere, 1996 Zimbabwe 206 CS Blood bank S Dipstick (PATH) EIA, Auszyme
SimpliRed

Sato, 1996 Japan 462 CC Hospital S Dainascreen EIA, Auszyme


Serodia

Abraham, 1998 India, Vellore 50 CC – Panel Hospital patients (Multiply transfused; chronic liver disease; preop and S QuickChaser EIA, Auszyme or Hepanostika
antenatal patients) Virucheck
400 CS –Screen

Oh, 1999 Korea 250 CC – Panel Blood donor panel S Genedia EIA, Cobas Core
Serodia

Kaur, 2000 India 2754 CS – CSQ Hospital surgery patients; blood donors; patients ruling out HBV S Hepacard EIA, Ortho 3rd generation

Lien, 2000 Viet Nam 328 CC High-risk volunteers; pregnant women; patients with other infectious SP Dainascreen EIA, Monolisa
diseases (including 10 with HIV); preselected HBsAg pos (101), HBsAg Determine MEIA for discordant
neg (99) Serodia

Raj, 2001 India, Vellore 999 CS Hospital laboratory samples (emergency preop screening; antenatal S Hepacard EIA, Auszyme
women in labour; haemodialysis; urgent donor screening) MEIA, AxSYM v2

Clement, 2002 Belgium 942 CC Hospital - patients with biopsy-proven HBV; healthy volunteers from a WB, S BinaxNOW MEIA, AxSYM v2
vaccine evaluation trial; blood donors

Lau, 2003 USA 1011 CS – CSQ Hepatology clinics S fresh Binax NOW EIA, ETI-MAK2

827 CS Incarcerated offenders S frozen

625 CS – CSQ Chinese community health fair (random patients); known HBV-positive WB
patients (liver clinic)

Akanmu, 2006 Nigeria, Lagos 101 CS – CSQ Blood donors (male) WB Binax NOW ELISA, Monolisa

Page | 198
36 Chronic liver disease

Nyirendra, 2008 Malawi 194 CS – CSQ Hospital Hospital patients including 152 HIV+ P Determine EIA, Bioelisa
Neutralisation (positives)

Lin, 2008 China 671 CC Blood donors (500); Clinical specimens HBsAg+ (171) S Determine EIA, Hepanostika Ultra
DRW
Guinea 579 CC Blood donors (491); Stored positives (88) SP

Randrianirina, Madagascar 200 CC Not specified S Cypress EIA, AxSYM


2008 Determine
Hexagon
Virucheck

Ola, 2009 Nigeria 25 CS - CSQ Medical clinic WB AMRAD GWHB ELISA, Wellcozyme Kit

55 Blood donors S Biotec Latex

Khan, 2010 Pakistan 57 CC NS S Accurate ELISA, 2nd generation


Onecheck

Davies, 2010 Malawi 75 CS – CSQ HIV-positive adults (ART naive) S Determine EIA, Biokit
Neutralization (positives)

Bjoerkvoll, 2010 Cambodia 1200 CS – CSQ General screen – blood donors (rural) S ACON EIA, Monolisa Ultra*

Viet Nam 1200

Geretti, 2010 Ghana, Kumasi 838 CS – CSQ HIV clinic (1/3 on lamivudine) S Determine CMIA, Architect/ Liason
VIKIA EIA, Murex v3

Hoffman, 2012 South Africa 973 CS – CSQ HIV-positive adults (ART naïve) – antenatal or primary care WB (cap) Determine ELISA, AxSYM

Bottero, 2013 France, Paris 2472 CS – CSQ General screening (health-care centres) [general population WB (ven) Determine ELISA, Monolisa Ultra
prevention, screening, vaccination] Neutralization (positives)
3922 QUICK PROFILE

3928 VIKIA

Chameera, 2013 Sri Lanka 50 CS Hospital (surgical, other) S Cortez EIA, Surase B-96 (TMB)

Page | 199
Onsite

Franzeck, 2013 Tanzania, 272 CS – CSQ HIV clinic (ART naive) WB (ven) Determine EIA, Murex v3
Ifakara P Neutralization (positives)

Chevaliez, 2014 Unclear 558 CC Chronic hep B (known mutants, blood donors); HBsAg negative (mix, SP DRW v2.0 CMI, Architect
including HIV, 34; HCV, 48)

408 CS Acute hepatitis

802 CS Pregnant - pregnant women at delivery

Erhabor, 2014 Nigeria, Sokoto 130 CC Blood donors SP ACON ELISA, HBsAg Ultra

Gish, 2014 Australia, 297 CS – CSQ At risk health fairs, outreach; Vietnamese (72%) S Nanosign EIA, Quest Diagnostics
Melbourne

Honge, 2014 Bissau 438 CS – CSQ HIV clinic - mixed ART/ naive S VEDA LAB CLIA, Architect

Liu, 2014 China 250 CC Hospital patients; outpatients S Intec One Step CMIA, Architect
(preselected based on CMIA quantitative results)

Mutocheluh, Ghana 150 CS – CSQ Blood donors P Abon ELISA, Human Gesellschaft
2014 Acull-Tell
Core TM
Rapid care
Wondfo

Upretti, 2014 Nepal 347 CS – CSQ Children – pre- and post vaccination; mothers (8) S SD Bioline EIA, Surase B-96 (TMB)

Njai, 2015 Gambia 178 CS Hepatitis patients CHB carriers (study 3), incl 3 coinfected HIV S Determine CMIA (quantitative),
(treatment naive) Architect
203 Espline

773 CS – CSQ General community screen WB Determine EIA (DBS), AxSym +


Neutralisation
476 VIKIA

*Validated random sample with CMIA, Abbott


CC: case–control study; CS: cross-sectional study; CSQ: consecutive patients; S: serum; P: plasma; WB (cap): capillary whole blood; WB (ven): venous whole blood

Page | 200
Assessment of the quality of the studies

The methodological quality of included studies4 is summarized [Fig. 2, Table 2].

Patient selection
We judged 15 studies to be at “high risk of bias”. Of these, 10 were case–control studies.
Others with a high risk of bias included studies in blood donors and highly selected
populations, such as patients with known hepatitis B.
Applicability was judged to be “high risk” in 8 studies, notably those published over
ten years ago or with tests which are no longer commercially available.

Index test
We judged 7 studies as high risk of bias, and 14 as unclear, with the most common reason
being a lack of reported blinding while reading test results.

Reference standard
We judged 5 studies as high risk of bias, with 17 unclear; the most common reason was a lack
of reported blinding interpreting reference tests, or utilisation.

Flow and timing


Bias was predominantly due to lack of reported flow and timing. Although the majority of
studies did not specificity the exact time differences between performance of the index and
reference assays, we can assume they were low risk as they were on the same sample.

4
Studies refers to either entire articles or individual sub-studies within a single publication using different
patients, methods, index tests or reference standard.

Page | 201
Table 2. Risk of bias and applicability according to QUADAS-2 domains for individual studies

Risk of bias Applicability

Patient Index test Reference Flow and Patient Index test Reference
selection standard timing selection standard

Abraham (CC) High Low Low Low Low High Unclear

Abraham (CS) Low Low Low Low Low High Unclear

Akanmu (Blood) Low Low Low Low Low Unclear Low

Akanmu (Liver) Low Low Low Low Low Unclear Low

Bjoerkvoll Low Low Low Low Low Low Low

Bottero Low High Low Low Low Low Low

Chameera Low High Unclear Low Low Low Low

Chevaliez (CC) High Unclear Unclear Low Low Low Low

Chevaliez (Hepatitis) High Unclear Unclear Low Low Low Low

Chevaliez (Pregnant) High Unclear Unclear Low Low Low Low

Clement High Unclear High Low Low Unclear Low

Davies Unclear Low Unclear High Low Low Low

Erhabor High Low Unclear Low Low Low Low

Franzeck Low Low Low Low Low Low Low

Geretti Low Unclear Unclear Unclear Low Low Low

Gish Low Unclear Unclear Low Low Low Low

Hoffman Low Unclear Low Unclear Low Low Low

Honge Low Low Unclear Low Low Low Low

Kaur Low Unclear Unclear Low Low High Unclear

Khan High Low Low Low Low Low Low

Lau (Hepatology clinic) Low Low Low Low Low Unclear Low

Lau (Prison) Low Low Low Low Low Unclear Low

Lau (Screen + Known) Low High Low Low Low Unclear Low

Lien Low High Low Low Low Low Low

Lin High Unclear Unclear Low Low Low Low

Liu High Unclear Low High Low Low Low

Mutocheluh High Low Low Low Low Low Low

Mvere Low Low Low Low Low High Unclear

Njai (1) Low Low Low Unclear Low Low Low

Njai (2) Low Low Low Unclear Low Low Low

Njai (3) High Low Low Low Low Low Low

Page | 202
Risk of bias Applicability

Patient Index test Reference Flow and Patient Index test Reference
selection standard timing selection standard

Nyirendra Low Low Unclear High Low Low Low

Oh High High High Low High High Unclear

Ol Low Unclear Unclear High Low Low Low

Ola BD Low High Unclear Low Low High Low

Ola Clinic Low Low Unclear Low Low Low Low

Peng High Low Unclear High Low Low Low

Raj Unclear Unclear High High Low High Low

Randrianirina High Unclear High Low Low Low Low

Sato Unclear High Low Low Unclear High Unclear

Upretti High Low High High High Low Low

Viet Low Unclear Unclear High Low Low Low

Fig. 2. Risk of bias and applicability summary according to QUADAS-2 domains presented as
percentages across included studies

FLOW AND TIMING Low High


QUADAS-2 Domain

REFERENCE STANDARD

INDEX TEST

PATIENT SELECTION

0% 20% 40% 60% 80% 100% 0% 20% 40% 60% 80% 100%
Proportion of studies with low, high or unclear Proportion of studies with low, high, or unclear
RISK of BIAS CONCERNS regarding APPLICABILITY

Page | 203
Diagnostic accuracy

Overall clinical performance of assays against an immunoassay reference

Pooled test accuracy for RDTs compared to EIAs


A total of 21 studies14‒36, 38‒43 contributing 63 data points evaluated 25 brands of RDTs using 15
EIA reference assays, with 36 919 total samples, including serum, plasma, venous and capillary
whole blood. Sample sizes ranged from 25 to 3983 (mean 586). Sensitivities ranged from 50%
to 100% with overall pooled sensitivity of 90.0% (95% CI: 89.1, 90.8). Specificities ranged from
69% to 100%, with overall pooled specificity of 99.5% (95% CI: 99.4, 99.5). Pooled PLR and NLR
were 117.5 (95% CI: 67.7, 204.1) and 0.095 (95% CI 0.067, 0.136), respectively, with tau-
square 3.89, 1.72, respectively, suggestive of significant heterogeneity between studies [Fig. 3,
Table3].

Pooled test accuracy for EIAs compared to other immunoassays


Five studies36, 43‒46 contributed 8 data points evaluating 8 EIAs with reference to state-of-the-
art immunoassays alone, using a total 3751 serum samples, with sample sizes ranging from
119 to 838 (mean, 469). Studies were performed in China, Ghana, Cambodia and Viet Nam.
Sensitivities ranged from 73% to 100%% with overall pooled sensitivity of 88.9% (95% CI: 87.0,
90.6). Specificities ranged from 88% to 100%, with overall pooled specificity of 98.4% (95% CI:
97.8, 98.8). Pooled PLR and NLR were 46.76 (95% CI: 12.86, 170.03) and 0.041 (95% CI: 0.013,
0.134), respectively, with tau-square 2.95, 2.46, respectively, suggestive of significant
heterogeneity between studies [Fig. 4, Table 3].

Overall clinical performance of assays against a NAT reference

Pooled test accuracy for RDTs compared to NAT


Three articles47, 57, 59 contributed 9 data points evaluating 7 RDTs with a NAT reference, using a
total 1710 serum or plasma samples, with sample sizes ranging from 74 to 950 (mean 190). Of
note, 1440 were from the same 240 patients in one case–control study evaluating 6 tests.
Only one study (Nna) 57 used plasma. Sensitivities ranged from 38% to 99% with overall pooled
sensitivity of 93.3% (95% CI: 91.3, 94.9). Specificities ranged from 94% to 99%, with overall
pooled specificity of 98.1% (95% CI: 97.0, 98.9). Pooled PLR and NLR were 39.42 (95% CI:
22.148, 70.185) and 0.051 (95% CI: 0.009, 0.275), respectively, with tau-square 0.2163, 6.264,
respectively. [Fig. 5, Table 9; Fig. 8 are in the Annex]

Pooled test accuracy for EIAs compared to NAT


Five articles54‒56, 58, 59 contributed 9 data points evaluating EIAs with a NAT reference, using a
total 1594 samples, with sample sizes ranging from 74 to 240 (mean, 177). Sensitivities ranged
from 38% to 98%% with overall pooled sensitivity of 75.7% (95% CI: 72.1, 79.1). Specificities
ranged from 70% to 98%, with overall pooled specificity of 86.1% (95% CI: 83.8, 88.2). Pooled
PLR and NLR were 7.234 (95% CI: 4.441, 11.758) and 0.296 (95%CI: 0.192, 0.458), respectively,

Page | 204
with tau-square 0.3748, 0.3299, respectively suggesting acceptable interstudy heterogeneity.
[Fig. 6, Table 9; and Fig. 9 are in the Annex]

Clinical performance of assays in HIV-positive and negative individuals

Pooled test accuracy for RDTs in HIV-positive individuals


Five articles19, 20, 36‒38 contributed 6 data points evaluating 3 RDTs with an EIA reference, using
a total 3434 samples from 2566 patients, with sample sizes ranging from 75 to 838.
Sensitivities ranged from 62% to 100% with overall pooled sensitivity of 72.3% (95% CI: 67.9,
76.4). Specificities ranged from 99% to 100%, with overall pooled specificity of 99.8% (95% CI:
99.5, 99.9). Pooled PLR and NLR were 192.63 (95% CI: 77.4, 479.17) and 0.288 (95%CI: 0.217,
0.381), respectively, with tau-square 0.3838, 0.0585, respectively [Table 4 and 5; Fig. 11,
Annex 9.3].

Pooled test accuracy for RDTs in HIV-negative individuals


One article40 contributed 4 data points evaluating 3 RDTs with an EIA reference, using a total
1624 samples from 997 patients, with sample sizes ranging from 175 to 773. Sensitivities
ranged from 88% to 95% with overall pooled sensitivity of 92.6% (95% CI: 89.8, 94.8).
Specificities ranged from 93% to 100%, with overall pooled specificity of 99.6% (95% CI: 99.0,
99.9). Pooled PLR and NLR were 79.449 (95% CI: 11.575, 545.315) and 0.082 (95%CI: 0.053,
0.125), respectively, with tau-square 2.9668, 0.0803 respectively [Table 4 and 5; Fig. 12, Annex
9.3].

Pooled test accuracy for EIAs in HIV-positive individuals


One article36 contributed 3 data points evaluating 3 EIAs with an EIA reference. 838 samples
were tested with each index test. Sensitivities ranged from 97% to 99% with overall pooled
sensitivity of 97.9% (95% CI: 96.0, 99.0). Specificities ranged from 99% to 100%, with overall
pooled specificity of 99.4% (95% CI: 99.0, 99.7). Pooled PLR and NLR were 167.26 (95% CI:
95.135, 294.07) and 0.022 (95% CI: 0.012, 0.043), respectively, with tau-square <0.005,
<0.005, respectively [Table 4; Fig. 14, Annex 9.3.]

Clinical performance of RDTS compared to EIAs in other stratified subgroups

Pooled test accuracy for RDTs using whole blood (capillary or venous)
Seven studies15, 16, 20, 24, 28, 37, 40 contributed 11 data points evaluating 5 RDTs with an EIA
reference, using a total 13731 samples, with sample sizes ranging from 25 to 3928 (mean
722). Sensitivities ranged from 75% to 100% with overall pooled sensitivity of 91.7% (95% CI:
89.1, 93.9). Specificities ranged from 99% to 100%, with overall pooled specificity of 99.9%
(95% CI: 99.8, 99.9). Pooled PLR and NLR were 346.64 (95% CI: 157.598, 762.42) and 0.089
(95%CI: 0.058, 0.136), respectively, with tau-square 0.8124, 0.2367, respectively [Table 5; Fig.
18, Annex 9.3.2].

Pooled test accuracy for RDTs in studies using a case-control design

Page | 205
Ten articles14, 18, 22, 24, 25, 27, 30, 31, 33, 34 contributed 21 data points evaluating 13 RDTs with an EIA
reference, using a total 7258 samples, with sample sizes ranging from 50 to 698 (mean 345).
Sensitivities ranged from 50% to 100% with overall pooled sensitivity of 96.7% (95% CI: 96.0,
97.3). Specificities ranged from 91% to 100%, with overall pooled specificity of 99.3% (95% CI:
99.0, 99.5). Pooled PLR and NLR were 105.16 (95% CI: 48.038, 230.212) and 0.028 (95%CI:
0.010, 0.076), respectively, with tau-square 2.2261, 4.8632, respectively. Of note, one study22
had significantly lower sensitivity for both index tests evaluated, with otherwise sensitivities
ranging from 90% to 100% in remaining studies [Table 5; Fig. 19, Annex 9.3.3].

Pooled test accuracy for RDTs used in blood donors


Seven articles25‒28, 32, 34, 39 contributed 19 data points evaluating 15 RDTs with an EIA reference,
using a total 6881 samples, with sample sizes ranging from 25 to 1200 (mean, 362).
Sensitivities ranged from 50% to 100% with overall pooled sensitivity of 91.6% (95% CI: 90.1,
92.9). Specificities ranged from 86% to 100%, with overall pooled specificity of 99.5% (95% CI:
99.3, 99.). Pooled PLR and NLR were 89.219 (95% CI: 32.782, 242.818) and 0.106 (95%CI:
0.055, 0.204), respectively, with tau-square 3.8171, 1.8505, respectively. [Table 5; Fig. 20,
Annex 9.3.4]

Pooled test accuracy for RDTs published before and after 2005
Twenty-one articles15‒17, 19, 20, 22, 25, 28, 30, 32‒43 contributed 44 data points evaluating 26 RDTs with
an EIA reference, using a total 25 261 samples, with sample sizes ranging from 25 to 3928
(mean 574). Sensitivities ranged from 50% to 100 % with overall pooled sensitivity of 86.4%
(95% CI: 85.2, 87.5). Specificities ranged from 69% to 100%, with overall pooled specificity of
99.4% (95% CI: 99.2, 99.5). Pooled PLR and NLR were 84.657 (95% CI: 43.553, 164.553) and
0.126 (95%CI: 0.087, 0.183), respectively, with tau-square 4.0986, 1.2712, respectively.
Nine articles published before 200514, 18, 21, 23, 24, 26, 27, 29, 31 contributed 19 data points
evaluating 10 RDTs with an EIA reference, using a total 25 253 samples, with sample sizes
ranging from 25 to 3928 (mean 1122). Sensitivities ranged from 77% to 100% with overall
pooled sensitivity of 96.9% (95% CI: 96.0, 97.7). Specificities ranged from 97% to 100%, with
overall pooled specificity of 99.7% (95% CI: 99.6, 99.8). Pooled PLR and NLR were 265.5 (95%
CI: 106.1, 664.5) and 0.056 (95%CI: 0.033, 0.095), respectively, with tau-square 2.72, 0.91,
respectively. [Table 5; Figs 21 and 22, Annex 9.3.4]

Pooled test accuracy for RDTs by brand


Stratifying by test brand did not eliminate statistical heterogeneity. Data for all 50 brands used
[Table 6] demonstrate heterogeneous results for sensitivity, with more robust specificity as
previously noted.

Determine HBsAg was evaluated in the most studies, with only one published before 2008.
Ten articles16, 19, 20, 24, 25, 30, 36, 37, 40, 41 contributing 12 data points evaluated against an EIA
reference, using a total 7553 samples, with sample sizes ranging from 75 to2472. Sensitivities
ranged from 56% to 100% with overall pooled sensitivity of 90.8% (95% CI: 88.9, 92.4).

Page | 206
Specificities ranged from 69% to 100%, with overall pooled specificity of 99.1% (95% CI: 98.9,
99.4). Excluding one particularly anomalous study,41 the lowest sensitivities and specificities
would be 69% and 93%, respectively. Pooled PLR and NLR were 239.24 (95% CI: 17.139,
33339.4) and 0.077 (95%CI: 0.035, 0.168), respectively, with tau-square 20.17, 1.556,
respectively. [Table 5, 6]
BinaxNOW HBsAg was evaluated in three articles, (15, 18, 23) all published before
2007, contributing 6 data points evaluating against an EIA reference, using a total 3550
samples, with sample sizes ranging from 36 to 1011. Sensitivities ranged from 94% to 100%
with overall pooled sensitivity of 97.6% (95% CI: 96.2, 98.6). Specificity was 100% in all studies,
with overall pooled specificity of 100% (95% CI: 99.7, 100). Pooled PLR and NLR were 221.21
(95% CI: 36.160, 1354.1) and 0.045 (95%CI: 0.016, 0.128), respectively, with tau-square 3.53,
1.20, respectively. [Tables 5, 6]
VIKIA HBsAg was also evaluated in three articles, 16, 36, 40 all published after 2010,
contributing 3 data points evaluating against an EIA reference, using a total 5242 samples,
with sample sizes ranging from 476 to 3928. Sensitivities ranged from 71% to 97% with overall
pooled sensitivity of 82.5% (95% CI: 77.5, 86.7). Specificities ranged from 99.8% to 100%, with
overall pooled specificity of 99.9% (95% CI: 99.8, 100). Pooled PLR and NLR were 1072.3 (95%
CI: 376.082, 3057.2) and 0.108 (95%CI: 0.026, 0.458), respectively, with tau-square <0.005,
1.472, respectively. [Tables 5, 6]
Serodia HBsAg was also evaluated in three articles24, 27, 31 all published before 2000,
contributing 3 data points evaluating against an EIA reference, using a total 1040 samples.
Sensitivities ranged from 71% to 97% with overall pooled sensitivity of 82.5% (95% CI: 77.5,
86.7). Specificities ranged from 99.8% to 100%, with overall pooled specificity of 99.9% (95%
CI: 99.8, 100). Pooled PLR and NLR were 284.91 (95% CI: 71.42, 1136.6) and 0.045 (95%CI:
0.029, 0.069), respectively, with tau-square <0.005, <0.005, respectively. [Tables 5, 6]

Pooled accuracy of RDTs evaluated against CMIA reference


Five articles33, 36, 38, 40, 43 contributed 9 data points evaluating 6 RDTs against specifically a CMIA
reference, using a total 4513 samples, with sample sizes ranging from 178 to 838 (mean 501).
Sensitivities ranged from 62% to 100% with overall pooled sensitivity of 80.4% (95% CI: 77.9,
82.6). Specificities ranged from 93% to 100%, with overall pooled specificity of 99.0% (95% CI:
98.6, 99.3). Pooled PLR and NLR were 58.5 (95% CI: 31.3, 109.3) and 0.141 (95%CI: 0.074,
0.268), respectively, with tau-square 0.4375, 0.7337, respectively. [Table 5]

Page | 207
Fig. 3. Forest plots, RDT vs EIA, ordered by [Test, Author]*

Specificity (95% CI)


Abon - Mutocheluh Sensitivity
0.99 1.00) CI) Specificity (95% CI)
(0.96 -(95% Sensitivity (95% CI)
Accurate
Abon - Khan
- Mutocheluh 0.95
Abon - Mutocheluh 0.50(0.74 - 1.00)
(0.28 - 0.72) 0.99 (0.96 - 1.00) Abon - Mutocheluh 0.50 (0.28 - 0.72) Abon
ACON - Bjoerkvoll
Accurate - Khan (Camb)
Accurate1.00 - 0.50
Khan (0.99 - 1.00)
(0.33 - 0.67) 0.95 (0.74 - 1.00) Accurate - Khan 0.50 (0.33 - 0.67) Accur
ACON --Bjoerkvoll
ACON Bjoerkvoll(Viet)
ACON 1.00 0.93(0.99
(Camb) - Bjoerkvoll - 1.00)
(Camb)
(0.86 - 0.98) 1.00 (0.99 - 1.00) ACON - Bjoerkvoll (Camb) 0.93 (0.86 - 0.98) ACON
ACON --Erhabor
ACON Bjoerkvoll ACON 0.91 0.82(0.84
(Viet) - Bjoerkvoll - 0.96)
(Viet)
(0.74 - 0.88) 1.00 (0.99 - 1.00) ACON - Bjoerkvoll (Viet) 0.82 (0.74 - 0.88) ACON
ACON - Erhabor
Acull-Tell - Mutocheluh ACON - Erhabor
0.99 1.00(0.96 (0.88 - 1.00)
- 1.00) 0.91 (0.84 - 0.96) ACON - Erhabor 1.00 (0.88 - 1.00) ACON
Acull-Tell
AMRAD - Ola - MutocheluhAcull-Tell1.00- 0.55
Mutocheluh(0.32
(0.40 - 0.76)
- 1.00) 0.99 (0.96 - 1.00) Acull-Tell - Mutocheluh 0.55 (0.32 - 0.76) Acull-
AMRAD - Ola (BD)
Binax - Akanmu AMRAD 1.00 - Ola 0.95(0.03 (0.76 - 1.00)
- 1.00) 1.00 (0.40 - 1.00) AMRAD - Ola 0.95 (0.76 - 1.00) AMRA
Binax
Binax --Akanmu
Akanmu(CLD) Binax - Akanmu
(BD) 1.00 0.94(0.89(BD)
(0.87 - 0.97)
- 1.00) 1.00 (0.03 - 1.00) Binax - Akanmu (BD) 0.94 (0.87 - 0.97) Binax
Binax - Akanmu (CLD) Binax - Akanmu 1.00 (CLD)(0.40 - 1.00) 1.00 (0.89 - 1.00) Binax - Akanmu (CLD) 1.00 (0.40 - 1.00) Binax
Binax - Clement 1.00 (0.99 - 1.00)
Binax - Clement Binax - Clement 1.00 (0.99 - 1.00) 1.00 (0.99 - 1.00) Binax - Clement 1.00 (0.99 - 1.00) Binax
Binax - Lau (Fresh S) 1.00 (1.00 - 1.00)
Binax - Lau (FreshBinax S) - Lau0.94 (Fresh S) - 0.99)
(0.79 1.00 (1.00 - 1.00) Binax - Lau (Fresh S) 0.94 (0.79 - 0.99) Binax
Binax --Lau
Binax Lau(Frozen
(Frozen S)S) - Lau
Binax 1.00 (Frozen
0.95 (1.00 - 1.00)
(0.86S)- 0.99) 1.00 (1.00 - 1.00) Binax - Lau (Frozen S) 0.95 (0.86 - 0.99) Binax
Binax --Lau
Binax Lau(WB) (WB) Binax - Lau 1.00 (WB)
0.96 (0.99 - 1.00)
(0.89 - 0.99) 1.00 (0.99 - 1.00) Binax - Lau (WB) 0.96 (0.89 - 0.99) Binax
Biotec --Ola
Biotec Ola Biotec - Ola0.86 0.59(0.64 - 0.97)
(0.41 - 0.75) 0.86 (0.64 - 0.97) Biotec - Ola 0.59 (0.41 - 0.75) Biotec
Core TM
Core TM--Mutocheluh
MutocheluhCore TM0.98 0.50(0.94
- Mutocheluh - 1.00)
(0.28 - 0.72) 0.98 (0.94 - 1.00) Core TM - Mutocheluh 0.50 (0.28 - 0.72) Core
Cortez --Chameera
Cortez ChameeraCortez - 1.00 0.60(0.92
Chameera - 1.00)
(0.15 - 0.95) 1.00 (0.92 - 1.00) Cortez - Chameera 0.60 (0.15 - 0.95) Corte
Cypress --Randrianirina
Cypress Cypress 0.96
Randrianirina 0.97(0.91
- Randrianirina - 0.99)
(0.91 - 0.99) 0.96 (0.91 - 0.99) Cypress - Randrianirina 0.97 (0.91 - 0.99) Cypre
Dainascreen
Dainascreen - -Lien LienDainascreen 1.00 1.00- (0.98
Lien(0.97 - 1.00)
- 1.00) 1.00 (0.98 - 1.00) Dainascreen - Lien 1.00 (0.97 - 1.00) Daina
Dainascreen
Dainascreen - -Sato SatoDainascreen 1.00 1.00- (0.99
Sato(0.98 - 1.00)
- 1.00) 1.00 (0.99 - 1.00) Dainascreen - Sato 1.00 (0.98 - 1.00) Daina
Determine
Determine - -Bottero
Bottero Determine 1.00 -0.94
Bottero (0.82
(1.00 - 0.99)
- 1.00) 1.00 (1.00 - 1.00) Determine - Bottero 0.94 (0.82 - 0.99) Deter
Determine
Determine - -Davies Determine
Davies(H+)
(H+) 1.00 -1.00
Davies
(0.93 -(H+)
(0.86 - 1.00)
1.00) 1.00 (0.93 - 1.00) Determine - Davies (H+) 1.00 (0.86 - 1.00) Deter
Determine
Determine - -Franzeck
FranzeckDetermine
(H+) 1.00
(H+) -0.96
Franzeck(0.80
(0.99 (H+)
- 1.00)
- 1.00) 1.00 (0.99 - 1.00) Determine - Franzeck (H+) 0.96 (0.80 - 1.00) Deter
Determine - GerettiDetermine (H+) -0.69
Geretti (H+)
(0.61 - 0.77) 1.00 (0.99 - 1.00) Determine - Geretti (H+) 0.69 (0.61 - 0.77) Deter
Determine - Geretti (H+) 1.00 (0.99 - 1.00)
Determine - Hoffman Determine
(H+) -0.75
Hoffman (0.59 (H+)
- 0.87) 1.00 (0.99 - 1.00) Determine - Hoffman (H+) 0.75 (0.59 - 0.87) Deter
Determine - Hoffman (H+)
Determine - Lien Determine -1.00
1.00 Lien (0.99 - 1.00)
(0.97 - 1.00) 1.00 (0.98 - 1.00) Determine - Lien 1.00 (0.97 - 1.00) Deter
Determine - -Lien
Determine Determine
Lin (China) 1.00 Lin(0.98
-0.99 - 1.00)
(China)
(0.96 - 1.00) 1.00 (0.99 - 1.00) Determine - Lin (China) 0.99 (0.96 - 1.00) Deter
Determine - -Lin
Determine Lin(China)
Determine
(Guinea) 1.00 Lin(0.99
-0.94 - 1.00)
(Guinea)
(0.90 - 0.97) 1.00 (0.99 - 1.00) Determine - Lin (Guinea) 0.94 (0.90 - 0.97) Deter
Determine - Lin
Determine - Njai (CHB)(Guinea) 1.00
Determine -0.95 (0.99
Njai (CHB) - 1.00)
(0.90 - 0.98) 0.93 (0.78 - 0.99) Determine - Njai (CHB) 0.95 (0.90 - 0.98) Deter
Determine - -Njai
Determine Njai(CHB)
Determine
(Screen) 0.93 -0.88
Njai(0.78 - 0.99)
(Screen)
(0.81 - 0.94) 1.00 (0.99 - 1.00) Determine - Njai (Screen) 0.88 (0.81 - 0.94) Deter
Determine - -Njai
Determine (Screen)
Determine
Nyirendra 1.00 -0.56 (0.99
Nyirendra - 1.00)
(0.38 - 0.73) 0.69 (0.62 - 0.76) Determine - Nyirendra 0.56 (0.38 - 0.73) Deter
Determine - -Nyirendra
Determine Determine
Randrianirina 0.69 -0.98 (0.62 - 0.76)
Randrianirina
(0.92 - 1.00) 1.00 (0.97 - 1.00) Determine - Randrianirina 0.98 (0.92 - 1.00) Deter
Dipstick
Determine(PATH) Dipstick 1.00
- Mvere
- Randrianirina (PATH)0.93(0.97-(0.68
Mvere - 1.00)
- 1.00) 1.00 (0.98 - 1.00) Dipstick (PATH) - Mvere 0.93 (0.68 - 1.00) Dipsti
DRW
Dipstick- Lin (China)
(PATH) DRW - Lin
- Mvere 1.00 (China)
0.99 (0.97
(0.98 - 1.00)
- 1.00) 0.99 (0.98 - 1.00) DRW - Lin (China) 0.99 (0.97 - 1.00) DRW
DRW
DRW --Lin Lin(China)
(Guinea) DRW - Lin 0.99 (Guinea)
0.97 (0.93
(0.98 - 0.99)
- 1.00) 1.00 (0.99 - 1.00) DRW - Lin (Guinea) 0.97 (0.93 - 0.99) DRW
DRW
DRW -v2.0Lin (Guinea) DRW
- Chevaliez v2.0
(Hep) 1.00-1.00
Chevaliez
(0.98
(0.99 (Hep)
- 1.00)
- 1.00) 0.99 (0.97 - 1.00) DRW v2.0 - Chevaliez (Hep) 1.00 (0.98 - 1.00) DRW
DRW v2.0 - Chevaliez DRW v2.0 -0.96
(Preg)
DRW v2.0 - Chevaliez (Hep) 0.99Chevaliez (0.81 (Preg)
(0.97 - 1.00)
- 1.00) 0.99 (0.97 - 1.00) DRW v2.0 - Chevaliez (Preg) 0.96 (0.81 - 1.00) DRW
DRW v2.0 - Chevaliez DRW CCv2.0 -0.95 Chevaliez
(0.76 CC- 1.00) 0.98 (0.97 - 0.99) DRW v2.0 - Chevaliez CC 0.95 (0.76 - 1.00) DRW
DRW v2.0 - Chevaliez (Preg) 0.99 (0.97 - 1.00)
Espline - Njai (CHB) Espline - Njai 0.94 (CHB)(0.89 - 0.97) 0.95 (0.82 - 0.99) Espline - Njai (CHB) 0.94 (0.89 - 0.97) Esplin
DRW v2.0
Genedia - Oh - Chevaliez CC
Genedia - Oh0.98 (0.97 - 0.99)
0.98 (0.94 - 1.00) 1.00 (0.96 - 1.00) Genedia - Oh 0.98 (0.94 - 1.00) Gene
Espline - Njai
Hepacard (CHB) Hepacard
- Kaur 0.95
-0.93
Kaur (0.82 - 0.99)
(0.84 - 0.98) 1.00 (1.00 - 1.00) Hepacard - Kaur 0.93 (0.84 - 0.98) Hepac
Genedia -
Hepacard - Raj Oh 1.00
Hepacard -0.83 Raj(0.96 - 1.00)
(0.61 - 0.95) 0.99 (0.98 - 0.99) Hepacard - Raj 0.83 (0.61 - 0.95) Hepac
Hepacard -- Kaur
Hexagon Hexagon1.00
Randrianirina 0.96(1.00
- Randrianirina - 1.00)
(0.89 - 0.99) 0.96 (0.91 - 0.99) Hexagon - Randrianirina 0.96 (0.89 - 0.99) Hexag
Hepacard
Intec - Liu- Raj Intec - Liu 0.99 0.51(0.98 - 0.99)
(0.43 - 0.58) 1.00 (0.95 - 1.00) Intec - Liu 0.51 (0.43 - 0.58) Intec
Nanosign
Hexagon - -Randrianirina
Gish (H+)Nanosign0.96 - 0.74
Gish (H+)
(0.49
(0.91 - 0.91)
- 0.99) 0.98 (0.95 - 0.99) Nanosign - Gish (H+) 0.74 (0.49 - 0.91) Nanos
Intec - Liu - Khan Onecheck
Onecheck 1.00 -0.53
Khan (0.36
(0.95 - 0.69)
- 1.00) 1.00 (0.82 - 1.00) Onecheck - Khan 0.53 (0.36 - 0.69) Onec
Onsite
Nanosign - Gish (H+)Onsite - 0.98
- Chameera Chameera
0.80(0.95 (0.28 - 0.99)
- 0.99) 1.00 (0.92 - 1.00) Onsite - Chameera 0.80 (0.28 - 0.99) Onsite
QUICK
Onecheck - Khan -QUICK
PROFILE Bottero PROFILE
1.00 0.90(0.82 - -Bottero
(0.82 - 0.96)
1.00) 1.00 (1.00 - 1.00) QUICK PROFILE - Bottero 0.90 (0.82 - 0.96) QUIC
QuickChaser
Onsite - Chameera QuickChaser
- Abraham CC1.00 0.90 -(0.92
Abraham
(0.68
- 1.00) CC
- 0.99) 1.00 (0.88 - 1.00) QuickChaser - Abraham CC 0.90 (0.68 - 0.99) Quick
QuickChaser - Abraham QuickChaser
CS 0.77 - Abraham CS
(0.55 - 0.92) 0.99 (0.98 - 1.00) QuickChaser - Abraham CS 0.77 (0.55 - 0.92) Quick
QUICK PROFILE - Bottero 1.00 (1.00 - 1.00)
Rapid care - Mutocheluh Rapid care 0.55 - Mutocheluh
(0.32 - 0.76) 0.99 (0.96 - 1.00) Rapid care - Mutocheluh 0.55 (0.32 - 0.76) Rapid
QuickChaser - Abraham CC 1.00- Upretti
SD Bioline - UprettiSD Bioline 1.00
(0.88 - 1.00)
(0.63 - 1.00) 1.00 (0.99 - 1.00) SD Bioline - Upretti 1.00 (0.63 - 1.00) SD Bi
QuickChaser - Abraham
Serodia - Lien
CS 0.99
Serodia - Lien
(0.98 - 1.00)
0.96 (0.90 - 0.99) 1.00 (0.98 - 1.00) Serodia - Lien 0.96 (0.90 - 0.99) Serod
Rapid care
Serodia - Oh- Mutocheluh
Serodia -0.99 Oh 0.96(0.96 - 1.00)
(0.91 - 0.99) 1.00 (0.96 - 1.00) Serodia - Oh 0.96 (0.91 - 0.99) Serod
SD Bioline
Serodia - Upretti Serodia -1.00
- Sato Sato0.96(0.99 - 1.00)
(0.91 - 0.98) 1.00 (0.98 - 1.00) Serodia - Sato 0.96 (0.91 - 0.98) Serod
Serodia - Lien
SimpliRed - Mvere SimpliRed 1.00 -0.93 (0.98
Mvere - 1.00)
(0.68 - 1.00) 1.00 (0.98 - 1.00) SimpliRed - Mvere 0.93 (0.68 - 1.00) Simpl
SerodiaLAB
VEDA - Oh - Honge VEDA
(H+) LAB 1.00 0.62 (0.96
- Honge - 1.00)
(0.51(H+)
- 0.73) 0.99 (0.98 - 1.00) VEDA LAB - Honge (H+) 0.62 (0.51 - 0.73) VEDA
VIKIA
Serodia- -Bottero
Sato VIKIA - Bottero
1.00 0.96(0.98 (0.90 - 0.99)
- 1.00) 1.00 (1.00 - 1.00) VIKIA - Bottero 0.96 (0.90 - 0.99) VIKIA
VIKIA - Geretti
SimpliRed - Mvere (H+)VIKIA - Geretti
1.00 (H+)
0.71(0.98 (0.62 - 0.78)
- 1.00) 1.00 (0.99 - 1.00) VIKIA - Geretti (H+) 0.71 (0.62 - 0.78) VIKIA
VIKIA
VEDA LAB- Njai- Honge
(Screen)VIKIA
(H+) - Njai 0.99 (Screen)
0.90 (0.79
(0.98 - 0.96)
- 1.00) 1.00 (0.99 - 1.00) VIKIA - Njai (Screen) 0.90 (0.79 - 0.96) VIKIA
Virucheck
VIKIA - Bottero - Abraham Virucheck
CC 1.00 -0.90
Abraham (0.68
(1.00 CC
- 0.99)
- 1.00) 1.00 (0.88 - 1.00) Virucheck - Abraham CC 0.90 (0.68 - 0.99) Viruch
Virucheck
VIKIA - Geretti (H+) Virucheck
- Abraham CS 1.00 -0.79
Abraham (0.54
(0.99 CS
- 0.94)
- 1.00) 0.97 (0.95 - 0.98) Virucheck - Abraham CS 0.79 (0.54 - 0.94) Viruch
Virucheck - Randrianirina Virucheck -0.96 Randrianirina
(0.89 - 0.99) 0.98 (0.94 - 1.00) Virucheck - Randrianirina 0.96 (0.89 - 0.99) Viruch
VIKIA - Njai (Screen) 1.00 (0.99 - 1.00)
Wondfo - Mutocheluh Wondfo - Mutocheluh
0.59 (0.36 - 0.79) 0.99 (0.96 - 1.00) Wondfo - Mutocheluh 0.59 (0.36 - 0.79) Wond
Virucheck - Abraham CC 1.00 (0.88 - 1.00)
Virucheck - Abraham CS 0.97 (0.95 - 0.98)
Virucheck - Randrianirina 0.98 (0.94 - 1.00)
Wondfo - Mutocheluh 0.99 (0.96 - 1.00)
Pooled Sensitivity Pooled Specificity
= 0.90 (0.89 to 0.91)= 0.99 (0.99 to 1.00) Pooled Sensitivity = 0.90 (0.89 to 0.91) Poole
Chi-square
Chi-square = 932.08; df = 62 (p= =651.98;
0.0000)df = 62 (p = 0.0000) Chi-square = 932.08; df = 62 (p = 0.0000) Chi-sq
.4 0.8 0.61 Inconsistency
0.8 1 Inconsistency
(I-square) = 93.3 % (I-square) = 90.5 % 0 0.2 0.4 0.6 0.8 1 0Inconsistency
0.2 (I-square)
0.4 = 93.30.6% 0.8 1 Incon
Specificity Sensitivity Specificity
Pooled Specificity = 0.99 (0.99 to 1.00)
Chi-square = 651.98; df = 62 (p = 0.0000)
0.8 1 Inconsistency (I-square) = 90.5 %

Page | 208
Positive LR (95% CI)
Abon - Mutocheluh 64.00 (8.69 - 471.26)
Accurate - Khan 9.50 (1.37 - 65.71)
ACON - Bjoerkvoll (Camb) 517.87 (129.54 - 2,070.26)
ACON - Bjoerkvoll (Viet) 434.51 (108.57 - 1,739.03)
ACON - Erhabor 10.46 (5.70 - 19.19)
Acull-Tell - Mutocheluh 69.82 (9.55 - 510.36)
AMRAD - Ola 9.32 (0.67 - 129.54)
Binax - Akanmu (BD) 3.72 (0.34 - 41.10)
Binax - Akanmu (CLD) 59.40 (3.74 - 944.36)
Binax - Clement 268.83 (67.41 - 1,072.16)
Binax - Lau (Fresh S) 1,811.52 (113.19 - 28,991.24)
Binax - Lau (Frozen S) 1,448.64 (90.62 - 23,156.72)
Binax - Lau (WB) 1,051.25 (65.81 - 16,793.27)
Biotec - Ola 4.12 (1.39 - 12.18)
Core TM - Mutocheluh 32.00 (7.60 - 134.68)
Cortez - Chameera 53.67 (3.14 - 917.12)
Cypress - Randrianirina 26.35 (10.06 - 69.00)
Dainascreen - Lien 422.20 (26.49 - 6,728.17)
Dainascreen - Sato 608.08 (38.12 - 9,699.86)
Determine - Bottero 4,498.21 (281.12 - 71,976.88)
Determine - Davies (H+) 100.04 (6.34 - 1,578.57)
Determine - Franzeck (H+) 467.38 (29.27 - 7,463.70)
Determine - Geretti (H+) 966.70 (60.39 - 15,473.86)
Determine - Hoffman (H+) 174.94 (64.74 - 472.74)
Determine - Lien 422.20 (26.49 - 6,728.17)
Determine - Lin (China) 959.01 (60.07 - 15,311.15)
Determine - Lin (Guinea) 753.59 (47.21 - 12,030.12)
Determine - Njai (CHB) 14.29 (3.74 - 54.54)
Determine - Njai (Screen) 1,180.48 (73.85 - 18,870.34)
Determine - Nyirendra 1.82 (1.25 - 2.67)
Determine - Randrianirina 214.02 (13.47 - 3,400.91)
Dipstick (PATH) - Mvere 348.00 (21.75 - 5,568.55)
DRW - Lin (China) 120.60 (45.44 - 320.04)
DRW - Lin (Guinea) 771.27 (48.32 - 12,311.21)
DRW v2.0 - Chevaliez (Hep) 74.50 (29.76 - 186.52)
DRW v2.0 - Chevaliez (Preg) 73.38 (30.62 - 175.83)
DRW v2.0 - Chevaliez CC 43.75 (27.08 - 70.69)
Espline - Njai (CHB) 17.85 (4.63 - 68.81)
Genedia - Oh 197.32 (12.43 - 3,133.39)
Hepacard - Kaur 4,996.94 (312.36 - 79,938.56)
Hepacard - Raj 73.30 (39.56 - 135.82)
Hexagon - Randrianirina 26.05 (9.95 - 68.23)
Intec - Liu 73.20 (4.61 - 1,163.40)
Nanosign - Gish (H+) 34.14 (14.80 - 78.75)
Onecheck - Khan 21.03 (1.34 - 329.93)
Onsite - Chameera 69.00 (4.22 - 1,129.19)
QUICK PROFILE - Bottero 347.25 (186.27 - 647.36)
QuickChaser - Abraham CC 54.62 (3.48 - 857.69)
QuickChaser - Abraham CS 146.05 (35.99 - 592.65)
Rapid care - Mutocheluh 69.82 (9.55 - 510.36)
SD Bioline - Upretti 642.22 (40.07 - 10,293.22)
Serodia - Lien 404.24 (25.36 - 6,443.55)
Serodia - Oh 193.30 (12.17 - 3,069.96)
Serodia - Sato 290.53 (41.05 - 2,056.47)
SimpliRed - Mvere 348.00 (21.75 - 5,568.55)
VEDA LAB - Honge (H+) 75.01 (23.99 - 234.59)
VIKIA - Bottero 1,853.68 (463.49 - 7,413.61)
VIKIA - Geretti (H+) 986.53 (61.64 - 15,789.02)
VIKIA - Njai (Screen) 374.40 (52.77 - 2,656.47)
Virucheck - Abraham CC 54.62 (3.48 - 857.69)
Virucheck - Abraham CS 25.07 (13.71 - 45.82)
Virucheck - Randrianirina 52.10 (13.19 - 205.84)
Wondfo - Mutocheluh 75.64 (10.41 - 549.47)

Random Effects Model


Pooled Positive LR = 117.54 (67.69 to 204.12)
Cochran-Q = 658.59; df = 62 (p = 0.0000)
0.01 1 100.0 Inconsistency (I-square) = 90.6 %
Positive LR Tau-squared = 3.8937

Page | 209
Negative LR (95% CI)
Abon - Mutocheluh 0.50 (0.33 - 0.77)
Accurate - Khan 0.53 (0.38 - 0.74)
ACON - Bjoerkvoll (Camb) 0.07 (0.03 - 0.14)
ACON - Bjoerkvoll (Viet) 0.18 (0.13 - 0.26)
ACON - Erhabor 0.02 (0.00 - 0.28)
Acull-Tell - Mutocheluh 0.46 (0.29 - 0.72)
AMRAD - Ola 0.08 (0.02 - 0.36)
Binax - Akanmu (BD) 0.09 (0.03 - 0.26)
Binax - Akanmu (CLD) 0.10 (0.01 - 1.41)
Binax - Clement 0.00 (0.00 - 0.02)
Binax - Lau (Fresh S) 0.08 (0.02 - 0.25)
Binax - Lau (Frozen S) 0.06 (0.02 - 0.16)
Binax - Lau (WB) 0.05 (0.02 - 0.13)
Biotec - Ola 0.48 (0.31 - 0.74)
Core TM - Mutocheluh 0.51 (0.33 - 0.77)
Cortez - Chameera 0.42 (0.16 - 1.09)
Cypress - Randrianirina 0.03 (0.01 - 0.10)
Dainascreen - Lien 0.00 (0.00 - 0.07)
Dainascreen - Sato 0.00 (0.00 - 0.05)
Determine - Bottero 0.07 (0.03 - 0.20)
Determine - Davies (H+) 0.02 (0.00 - 0.30)
Determine - Franzeck (H+) 0.06 (0.01 - 0.27)
Determine - Geretti (H+) 0.31 (0.24 - 0.40)
Determine - Hoffman (H+) 0.25 (0.15 - 0.43)
Determine - Lien 0.00 (0.00 - 0.07)
Determine - Lin (China) 0.01 (0.00 - 0.05)
Determine - Lin (Guinea) 0.06 (0.03 - 0.10)
Determine - Njai (CHB) 0.05 (0.02 - 0.11)
Determine - Njai (Screen) 0.12 (0.07 - 0.20)
Determine - Nyirendra 0.64 (0.43 - 0.94)
Determine - Randrianirina 0.03 (0.01 - 0.09)
Dipstick (PATH) - Mvere 0.09 (0.02 - 0.43)
DRW - Lin (China) 0.01 (0.00 - 0.04)
DRW - Lin (Guinea) 0.04 (0.02 - 0.08)
DRW v2.0 - Chevaliez (Hep) 0.00 (0.00 - 0.04)
DRW v2.0 - Chevaliez (Preg) 0.04 (0.01 - 0.26)
DRW v2.0 - Chevaliez CC 0.05 (0.01 - 0.33)
Espline - Njai (CHB) 0.06 (0.03 - 0.12)
Genedia - Oh 0.02 (0.01 - 0.07)
Hepacard - Kaur 0.07 (0.03 - 0.18)
Hepacard - Raj 0.18 (0.07 - 0.43)
Hexagon - Randrianirina 0.05 (0.02 - 0.12)
Intec - Liu 0.50 (0.43 - 0.58)
Nanosign - Gish (H+) 0.27 (0.13 - 0.57)
Onecheck - Khan 0.49 (0.35 - 0.68)
Onsite - Chameera 0.25 (0.06 - 1.01)
QUICK PROFILE - Bottero 0.10 (0.05 - 0.18)
QuickChaser - Abraham CC 0.12 (0.04 - 0.39)
QuickChaser - Abraham CS 0.23 (0.11 - 0.49)
Rapid care - Mutocheluh 0.46 (0.29 - 0.72)
SD Bioline - Upretti 0.06 (0.00 - 0.82)
Serodia - Lien 0.05 (0.02 - 0.11)
Serodia - Oh 0.04 (0.02 - 0.09)
Serodia - Sato 0.04 (0.02 - 0.09)
SimpliRed - Mvere 0.09 (0.02 - 0.43)
VEDA LAB - Honge (H+) 0.38 (0.28 - 0.51)
VIKIA - Bottero 0.04 (0.01 - 0.11)
VIKIA - Geretti (H+) 0.29 (0.23 - 0.38)
VIKIA - Njai (Screen) 0.10 (0.05 - 0.21)
Virucheck - Abraham CC 0.12 (0.04 - 0.39)
Virucheck - Abraham CS 0.22 (0.09 - 0.52)
Virucheck - Randrianirina 0.04 (0.02 - 0.12)
Wondfo - Mutocheluh 0.41 (0.25 - 0.68)

Random Effects Model


Pooled Negative LR = 0.10 (0.07 to 0.14)
Cochran-Q = 1298.04; df = 62 (p = 0.0000)
0.01 1 100.0 Inconsistency (I-square) = 95.2 %
Negative LR Tau-squared = 1.7188

Page | 210
Key brands of RDT and reference standards used in studies

Study Test brand (manufacturer) Reference test type, brand (manufacturer)

Mvere, 1996 Dipstick (PATH) EIA, Auszyme


SimpliRed

Sato, 1996 Dainascreen HBsAG EIA, Auszyme


Serodia HBsAg

Abraham, 1998 QuickChaser EIA, Auszyme or Hepanostika


Virucheck

Oh, 1999 Genedia EIA, Cobas Core


Serodia

rd
Kaur, 2000 Hepacard EIA, Ortho 3 generation

Lien, 2000 Dainascreen EIA, Monolisa


Determine HBsAg MEIA for discordant
Serodia

Raj, 2001 Hepacard EIA, Auszyme


MEIA, AxSYM v2

Clement, 2002 BinaxNOW MEIA, AxSYM v2

Lau, 2003 BinaxNOW EIA, ETI-MAK2

Akanmu, 2006 BinaxNOW ELISA, Monolisa

Bjoerkvoll, 2010 ACON EIA, Monolisa Ultra*

Lin, 2008 Determine HBsAg EIA, Hepanostika Ultra


DRW HBsAg

Nyirendra, 2008 Determine HBsAG EIA, Bioelisa


Neutralisation (positives)

Randrianirina, 2008 Cypress EIA, AxSYM


Determine HBsAg
Hexagon
Virucheck

Ola, 2009 AMRAD GWHB ELISA, Wellcozyme Kit


Biotec Latex

Davies, 2010 Determine EIA, Biokit; Neutralisation (for all positives)

Geretti, 2010 Determine CMIA, Architect/ Liason


VIKIA EIA, Murex v3

nd
Khan, 2010 Accurate ELISA, 2 generation
Onecheck

Page | 211
Study Test brand (manufacturer) Reference test type, brand (manufacturer)

Hoffman, 2012 Determine ELISA, AxSYM

Bottero, 2013 Determine ELISA, Monolisa Ultra


QUICK PROFILE Neutralisation (for all positive)
VIKIA

Franzeck, 2013 Determine HBsAg EIA, Murex v3


Neutralisation (for all positives)

Chameera, 2013 Cortez EIA, Surase B-96 (TMB)


Onsite

Chevaliez, 2014 DRW v 2 HBsAg CMIA, Architect

Erhabor, 2014 ACON ELISA, HBsAg Ultra

Gish, 2014 Nanosign EIA, Quest Diagnostics

Honge, 2014 VEDA LAB CLIA, Architect

Liu, 2014 Intec One Step CMIA, Architect

Upretti, 2014 SD Bioline EIA, Surase B-96 (TMB)

Mutocheluh, 2014 Abon ELISA, Human Gesellschaft


Acull-Tell
Core-TM
Rapid care
Wondfo

Njai, 2015 Determine HBsAg EIA (DBS), AxSym


VIKIA Neutralisation
Espline CMIA (quantitative), Architect

Page | 212
Fig. 4. Forest plots, EIA vs EIA, ordered by [Test, Author]**

Specificity (95% CI)


Sensitivity (95% CI) Sensitivity (95% CI) Specificity (95% CI)
Architect - Geretti H+ Architect -0.98
Geretti H+ - 1.00) 1.00
(0.94 (0.99 - 1.00) Architect - Geretti H+ 0.98 (0.94 - 1.00)Architect - Geretti H+ 1.00 (0.99 - 1.00)
Cobas - Liu Cobas - Liu 0.97 (0.93 - 0.99) 1.00 (0.95 - 1.00) Cobas - Liu 0.97 (0.93 - 0.99)Cobas - Liu 1.00 (0.95 - 1.00)
KHB - Peng KHB - Peng 0.74 (0.69 - 0.79) 0.88 (0.82 - 0.93) KHB - Peng 0.74 (0.69 - 0.79)KHB - Peng 0.88 (0.82 - 0.93)
Specificity (95% CI)
Liaison Ultra - GerettiLiaison
H+ Ultra0.97- Geretti
(0.93 H+
- 0.99) 0.99 (0.99 - 1.00) Liaison Ultra - Geretti H+ 0.97 (0.93 - 0.99)Liaison Ultra - Geretti H+ 0.99 (0.99 - 1.00)
Architect - -Geretti
Monolisa Ol H+Monolisa1.00 Ol (0.99
-1.00 - 1.00)
(0.95 - 1.00) 0.91 (0.79 - 0.98) Monolisa - Ol 1.00 (0.95 - 1.00)Monolisa - Ol 0.91 (0.79 - 0.98)
Cobas - Liu
Monolisa Ultra- VietMonolisa1.000.99(0.95
Ultra- - 1.00)
Viet
(0.93 - 1.00) 0.91 (0.78 - 0.97) Monolisa Ultra- Viet 0.99 (0.93 - 1.00)Monolisa Ultra- Viet 0.91 (0.78 - 0.97)
Murex v3.0 - Geretti
KHB - Peng Murex
H+ v3.0
0.88 - (0.82
0.99 Geretti
(0.95 H+
- 1.00) 0.99
- 0.93) (0.98 - 1.00) Murex v3.0 - Geretti H+ 0.99 (0.95 - 1.00)Murex v3.0 - Geretti H+ 0.99 (0.98 - 1.00)
Wantai - Liu - GerettiWantai
Liaison Ultra H+ - Liu
0.990.78(0.99(0.71 - 0.84) 1.00
- 1.00) (0.95 - 1.00) Wantai - Liu 0.78 (0.71 - 0.84)Wantai - Liu 1.00 (0.95 - 1.00)
Monolisa - Ol 0.91 (0.79 - 0.98)
Monolisa Ultra- Viet 0.91 (0.78 - 0.97)
Murex v3.0 - Geretti H+ 0.99 (0.98 - 1.00)
Wantai - Liu 1.00 (0.95 - 1.00)
Pooled Sensitivity = Pooled Specificity
0.89 (0.87 to 0.91) = 0.98 (0.98 to 0.99) Pooled Sensitivity = 0.89 (0.87 to 0.91) Pooled Specificity = 0.98 (0.98 to 0.99)
Chi-square = 166.37; Chi-square
df = 7 (p == 88.56;
0.0000) df = 7 (p = 0.0000) Chi-square = 166.37; df = 7 (p = 0.0000) Chi-square = 88.56; df = 7 (p = 0.0000)
2.6 0.4
0.8 1 Inconsistency
0.6 0.8 1 Inconsistency
(I-square) = 95.8 % (I-square) = 92.1 % 0 0.2 0.4 0.6 0.8 1 0Inconsistency
0.2 (I-square)
0.4 = 95.8
0.6 % 0.8 1 Inconsistency (I-square) = 92.1 %
y Specificity Sensitivity Specificity
Pooled Specificity = 0.98 (0.98 to 0.99)
Chi-square = 88.56; df = 7 (p = 0.0000)
0.6 0.8 1 Inconsistency (I-square) = 92.1 %
city

+
**H = HIV positive

Key – Types of EIA and reference standards used in studies

Study Test type, brand (manufacturer) Reference

Geretti EIA, Murex v.3.0 (Abbott) CMIA and EIA (agreement) or neutralization

CMIA, Architect (Abbott)

CMIA, Liaison Ultra (Diasorin)

Liu ELISA, Wantai (Beijing Wantai) CMIA, Architect (Abbott)

ECLIA, Cobas e601 (Roche)

Page | 213
Ol ELISA, Monolisa (bioRad) CMIA, Architect (Abbott)

Viet EIA, Monolisa Ultra (bioRad) CMIA, Architect (Abbott)

Peng ELISA, KHB (Kehua Bio-engineeering Co) CMIA, Architect (Abbott)

Positive LR (95% CI) Negative LR (95% CI)


Architect - Geretti H+ 227.68 (73.59 - 704.41)
Architect - Geretti H+ 0.02 (0.01 - 0.07)
Cobas - Liu 138.80 (8.76 - 2,198.11)
KHB - Peng 6.22 (4.14 - 9.35) Cobas - Liu 0.04 (0.02 - 0.08)
Liaison Ultra - Geretti H+ 169.51 (63.77 - 450.57) KHB - Peng 0.30 (0.25 - 0.36)
Monolisa - Ol 10.15 (4.22 - 24.43) Liaison Ultra - Geretti H+ 0.03 (0.01 - 0.08)
Monolisa Ultra- Viet 10.61 (4.17 - 26.99) Monolisa - Ol 0.01 (0.00 - 0.12)
Murex v3.0 - Geretti H+ 137.61 (57.44 - 329.64) Monolisa Ultra- Viet 0.01 (0.00 - 0.10)
Wantai - Liu 112.40 (7.09 - 1,781.70) Murex v3.0 - Geretti H+ 0.01 (0.00 - 0.06)
Wantai - Liu 0.22 (0.17 - 0.29)

Random Effects Model


Pooled Positive LR = 46.77 (12.86 to 170.03)
Cochran-Q = 100.82; df = 7 (p = 0.0000) Random Effects Model
0.01 1 100.0 Inconsistency (I-square) = 93.1 % Pooled Negative LR = 0.04 (0.01 to 0.13)
Positive LR Tau-squared = 2.9510 Cochran-Q = 238.26; df = 7 (p = 0.0000)
0.01 1 100.0 Inconsistency (I-square) = 97.1 %
Negative LR Tau-squared = 2.4644

Abbreviations used in Forest plots

+
*Camb: Cambodia; Viet: Viet Nam; BD: blood donor study; CLD: chronic liver disease study; Fresh S: fresh serum; Frozen S: frozen serum; WB: whole blood; H : HIV positive;
CHB: chronic hepatitis B cohort; Screen: general screen cohort; Hep: acute hepatitis cohort; Preg: antenatal cohort; CC: case–control study; CS: cross-sectional study
+
**H : HIV positive

Page | 214
Table 3. Summary pooled diagnostic accuracy of HBsAg assays using EIA and NAT reference standards

5
Reference Index test Pooled clinical accuracy Likelihood ratios (REM) Heterogeneity
(Tau-squared)

n Sen Spec PLR NLR PLR NLR


(95% CI) (95% CI) (95% CI) (95% CI)

EIA RDT 63 90.0 99.5 118 0.095 3.89 1.72


(89.1–90.8) (99.4–99.5) (67.7–204) (0.067–0.136)

EIA 8 88.9 98.4 46.8 0.041 2.95 2.46


(87.0–90.6) (97.8–98.8) (12.9–170) (0.013–0.134)

NAT RDT 9 93.3 98.1 39.4 0.051 0.22 6.26


(91.3–94.9) (97.0–98.9) (22.1–70.2) (0.009–0.27)

EIA 9 75.7 86.1 7.23 0.296 0.37 0.33


(72.1–79.1) (83.8–88.2) (4.44–11.8) (0.192–-0.458)

*n = number of data points

5
REM : Random effects model

Page | 215
Table 4. Summary pooled diagnostic accuracy of HBsAg assays in patients with known HIV status

Test type HIV status Pooled clinical accuracy Likelihood ratios (REM) Heterogeneity (Tau-squared)

n Sen Spec PLR NLR PLR NLR


(95% CI) (95% CI) (95% CI) (95% CI)

+6
RDT HIV 6 72.3 99.8 193 0.29 0.384 0.0059
(67.9–76.4) (99.5–99.9) (77.4–497) (0.22–0.38)

HIV– 3 92.6 99.6 79.5 0.08 2.967 0.080


(89.8–94.8) (99.0–99.9) (11.6–545) (0.05–0.13)

+
EIA HIV 3 97.9 99.4 167 0.02 <0.005 <0.005
(96.0–99.0) (99.0–99.7) (95.1–294) (0.01–0.04)

HIV–

6
Three studies on ART-naive patients (Hoffman, Davies, Franzeck); two studies from single article (Geretti) in patients who included 1/3 on lamivudine; with one study (Honge) on a mixture

Page | 216
9.1.1.Table 5. Summary pooled diagnostic accuracy of rapid HBsAg assays stratified by study, patient, index and reference tests

Sub-analysis Pooled clinical accuracy Likelihood ratios (REM) Heterogeneity (Tau-squared)


n Sen Spec LR+ LR- PLR NLR
(95% CI) (95% CI) (95% CI) (95% CI)
Study Pre 2005 19 96.9 99.7 266 0.056 2.72 0.91
(96.0–97.7) (99.6–99.8) (106–665) (0.033–0.095)
Post 2005 44 86.4 99.4 84.6 0.126 4.10 1.27
(85.2–87.5) (99.2–99.5) (43.6–165) (0.087–0.183)
Case–control 21 96.7 99.3 105 0.028 2.23 4.86
(96.0–97.3) (99.0–99.5) (48.0–230) (0.010–0.076)
Patient Blood donors 19 91.6 99.5 89.2 0.106 3.82 1.86
(90.1–92.9) (99.3–99.7) (32.8–243) (0.055–0.204)
HIV+ 6 72.3 99.8 193 0.29 0.384 0.0059
(67.9 –76.4) (99.5–99.9) (77.4–497) (0.22–0.38)
HIV– 4 92.6 99.6 79.5 0.08 2.97 0.080
(89.8–94.8) (99.0–99.9) (11.6–545) (0.05-0.13)
Index test Whole blood 11 91.7 99.9 347 0.089 0.81 0.24
(89.1–93.9) (99.8–99.9) (158–762) (0.058–0.136)
Determine 12 90.8 99.1 239 0.077 20.2 1.56
(88.9–92.4) (98.9–99.4) (17.1–33300) (0.035–0.168)
BinaxNOW 6 97.6 100 221 0.045 3.53 1.20
(96.2–98.6) (99.7–100) (36.1–1350) (0.016–0.128)
VIKIA 3 82.5 99.9 1070 0.108 <0.005 1.472
(77.5–86.7) (99.8–100) (376–3060) (0.026–0.458)
Serodia 3 82.5 99.9 285 0.045 <0.005 <0.005
(77.5–86.7) (99.8–100) (71.4–1140) (0.029–0.069)
Reference test CMIA 9 80.4 99.0 58.5 0.141 0.44 0.73
(77.9–82.6) (99.6–99.3) (31.3–109) (0.074–0.268)

Page | 217
n = number of data points

Page | 218
Table 6. Summary pooled diagnostic accuracy of HBsAg assays by brand

EIA reference NAT reference

Type Brand name Sen (95% CI) Spec (95% CI) Sen (95% CI) Spec (95% CI)

RDT Abon 50.0 (28.2–71.8) 99.2 (95.7–100)

RDT Accurate 50.0 (33.4–66.6) 94.7 (74.0–99.9)

RDT ACON 88.0 (83.4–91.7) 99.4 (99.0–99.7) 92.9 (87.3–96.5) 99.1 (96.6–99.9)

RDT Acull–Tell 54.5 (33.2–75.6) 99.2 (95.7–100)

RDT AMRAD 95.2 (76.2–99.9) 100 (39.8–100)

RDT Atlas 97.5 (92.9–99.5) 97.5 (92.9–99.5)

RDT BINAX 97.6 (96.2–98.6) 99.9 (99.7–100)

RDT Blue Cross 99.2 (95.4–100) 98.3 (94.1–99.8)

RDT Biotec 58.8 (40.7–75.4) 85.7 (63.7–97.0)

RDT Core TM 50.0 (28.2–71.8) 98.4 (94.5–99.8)

RDT Cortez 60.0 (14.7–94.7) 100 (92.1–100) 79.7 (73.1–85.3) 97.2 (94.0–99.0)

RDT Cypress 96.7 (90.7–99.3) 96.3 (90.9–99.0)

RDT Dainascreen 100 (98.7–100) 100 (99.3–100)

RDT Determine 90.8 (88.9–92.4) 99.1 (98.9–99.4)

RDT DIMA 98.3 (94.1–99.8) 99.2 (95.4–100)

RDT Dipstick (PATH) 93.3 (68.1–99.8) 100 (98.1–100)

RDT DRW 98.1 (96.1–99.2) 99.5 (98.8–99.9)

RDT DRW v2 99.3 (97.4–99.9) 98.3 (97.5–98.9)

RDT Espline 93.9 (89.1–97.1) 94.7 (82.3–99.4)

RDT Genedia 98.0 (94.3–99.6) 100 (96.4–100)

RDT Hepacard 90.5 (82.1–95.8) 99.7 (99.5–99.9)

RDT Hexagon 95.6 (89.1–98.8) 96.4 (90.9–99.0)

RDT Intec 50.8 (43.3–58.4) 100 (94.9–100) 99.2 (95.4–100) 97.5 (92.9–99.5)

RDT Nanosign 73.7 (48.8–90.9) 97.8 (95.4–99.2)

RDT Onecheck 52.6 (35.8–69.0) 100 (82.4–100)

RDT Onsite 80.0 (28.4–99.5) 100 (92.1–100)

RDT Quick Profile 90.5 (82.1–95.8) 99.7 (99.5–99.9)

RDT QuickChaser 83.3 (68.6–93.0) 99.5 (98.2–99.9)

RDT Rapid Care 54.5 (32.2–75.6) 99.2 (95.7–100)

RDT SD Bioline 100 (63.1–100) 100 (98.9–100)

RDT Serodia 95.8 (93.4–97.5) 99.8 (99.1–100)

RDT SimpliRed 93.3 (68.1-99.8) 100 (98.1-100)

RDT VEDA Lab 62.3 (50.6-73.1) 99.2 (97.6-99.8)

Page | 219
RDT VIKIA 82.5 (77.5-86.7) 99.9 (99.8-100)

RDT Virucheck 92.3 (86.3-96.2) 97.3 (95.5-98.5)

RDT Wondfo 59.1 (36.4-79.3) 99.2 (95.7-100)

EIA ADVIA 77.4 (65.0–87.1) 97.9 (92.6–99.7)

EIA Architect 97.9 (93.9–99.6) 99.6 (98.7–99.9)

EIA AxSym 56.6 (44.7–67.9) 86.8 (81.5–90.9)

EIA AxSym v2 77.4 (67.0–85.8) 75.0 (66.1–82.6)

EIA Cobas 96.6 (92.8–98.8) 100 (94.9–100)

EIA Elecsys 63.4 (55.2–71.0) 95.8 (92.4–98.0)

EIA KHB 73.8 (68.7–78.6) 88.1 (82.4–92.5)

EIA Liaison 100 (97.6–100) 70.0 (63.1–76.3)

EIA Liaison Ultra 97.1 (92.8–99.2) 99.4 (98.5–99.8)

EIA Monolisa 100 (95.2–100) 91.1 (78.8–97.5)

EIA Monolisa Ultra 98.7 (92.9–100) 90.7 (77.9–97.4)

EIA Murex v3 98.6 (94.9–99.8) 99.3 (98.3–99.8)

EIA VIDAS Ultra 69.0 (58.0–78.7) 94.0 (88.0–97.5)

EIA Wantai 78.2 (71.4–84.0) 100 (94.9–100)

Pooled results and I2 (heterogeneity)


2 2
Test Ref Studies, n Data, nd I Sen, % I Spec, %

ACON NAT 2 2 96.0 0.0

Cortez NAT 2 3 97.5 0.0

ADVIA NAT 1 2 0.0 0.0

AxSym NAT 1 2 95.8 89.1

Elecsys NAT 2 2 0.0 49.2

ACON EIA 2 3 86.8 95.2

Binax EIA 3 6 77.3 25.3

Dainascreen EIA 2 2 0.0 0.0

Determine EIA 10 12 92.9 97.0

DRW EIA 1 2 76.2 79.2

DRW v2 EIA 1 3 71.5 0.0

Hepacard EIA 2 2 51.3 96.6

Quickchaser EIA 1 2 20.7 0.0

Serodia EIA 3 3 0.0 0.0

VIKIA EIA 3 3 93.4 7.3

Virucheck EIA 2 3 61.0 11.9

Page | 220
GRADE

Question: Should RDTs be used to diagnose HBsAg in HIV-negative individuals?

Sensitivity 0.88–0.95 Prevalences 5% 20%

Specificity 0.93–1.00

Outcome No. of studies (no. Study design Factors that may decrease quality of evidence Effect per 1000 patients/year Test
of patients) accuracy QoE
Risk of Indirectness Inconsistency Imprecision Publication pre-test pre-test
bias bias probability of 5% probability of 20%

1 2 3 4
True positives 4 studies Cross-sectional (cohort Serious Not serious Serious Serious None 44–48 176–190 ⨁◯◯◯
1234
(patients with HBsAg) 997 patients type accuracy study) Very low

False negatives 2–6 10–24


(patients incorrectly classified
as not having HBsAg)

1 2 3
True negatives 4 studies Cross-sectional (cohort Serious Not serious Serious Not None 884–950 744–800 ⨁⨁◯◯
5 1235
(patients without HBsAg) 997 patients type accuracy study) serious Low

False positives 0–66 0 –56


(patients incorrectly classified
as having HBsAg)

40
1. Downgraded by one for risk of bias: all studies were prospective cohort studies , although one was assessed as high risk of bias because patients were pre-selected based from known chronic hepatitis
B patients.
2. Although study was not specifically designed in HIV-negative patients, clear testing and results were included.
3. Downgraded by one for inconsistency: unexplained heterogeneity may arise from differences between studies in specimen condition (serum, whole blood), specimen processing (field vs laboratory),
reference tests (CMIA; EIA on dried blood spots) and study population (e.g. known chronic hepatitis B patients, general community screen).
4. Downgraded by one for imprecision: confidence intervals extend below 90% accuracy, with tau-squared for PLR >1 (indicating substantial heterogeneity).
Question: Should RDTs be used to diagnose HBsAg in HIV-positive individuals?

Sensitivity 0.72 (95% CI: 0.68–0.76) Prevalences 5% 20%

Page | 221
Specificity 1.00 (95% CI: 0.99–1.00)

Outcome No. of studies (no. Study design Factors that may decrease quality of evidence Effect per 1000 patients/year Test accuracy
of patients) QoE
Risk of Indirectness Inconsistency Imprecision Publication pre-test pre-test probability
bias bias probability of 5% of 20%

1 2 3 4
True positives 5 studies Cross-sectional (cohort Serious Not serious Serious Serious None 36 (34–38) 145 (136–153) ⨁◯◯◯
1234
(patients with HBsAg) 2566 patients type accuracy study) Very low

False negatives 14 (12–16) 55 (47–64)


(patients incorrectly classified as
not having HBsAg)

1 2 5
True negatives 5 studies Cross-sectional (cohort Serious Not serious Not serious Not None 948 (945–949) 798 (796–799) ⨁⨁⨁◯
6 1256
(patients without HBsAg) 2566 patients type accuracy study) serious Moderate

False positives 2 (1–5) 2 (1–4)


(patients incorrectly classified as
having HBsAg)

1. Downgraded by one for risk of bias: all studies were prospective cohort studies of consecutive patients. Studies used different specimens (serum, 2; capillary whole blood, 1; venous whole blood, 1),
reference standards (CMIA, EIA confirmed by neutralization), and had patients with different ART status (four studies ART naive).
20 36 19 37 38
2. Not downgraded for indirectness: all studies performed in cohorts of consecutive patients in Tanzania , Ghana , Malawi , South Africa and Bissau .
3. Downgraded by one for inconsistency with sensitivities ranging from 62% to 100%: unexplained heterogeneity may arise from differences between studies in specimen type, specimen processing and
had very high sensitivities (100%, 96%) while remainder ‒ had low sensitivities (range 62-70%). Tau-squared <1 for studies
19, 20 36 38
study population. Two Studies
19, 20 36‒38
4. Downgraded by one for imprecision: confidence intervals 67.9–76.4%. Two studies had very high sensitivities (100%, 96%) while remainder had low sensitivities (range 62–70%).
5. Not downgraded for inconsistency: specificities ranged from 99% to 100%, with tau-squared <1
6. Not downgraded for imprecision: narrow confidence interval

Page | 222
Question: Should Determine HBsAg be used to diagnose HBsAg in a global setting?

Sensitivity 0.91 (95% CI: 0.89 to 0.92) Prevalences 5% 20%

Specificity 0.99 (95% CI: 0.99 to 0.99)

Outcome No. of studies (no. of Study design Factors that may decrease quality of evidence Effect per 1000 patients/year Test accuracy
patients) QoE
Risk of Indirectness Inconsistency Imprecision Publication pre-test probability pre-test probability of
bias bias of 5% 20%

True positives 12 studies Cohort & case–control type Serious Not serious Very serious2 Not serious None 45 (44–46) 182 (178–185) ⨁◯◯◯
(patients with HBsAg) 7552 patients studies1 Very low2

False negatives 5 (4–6) 18 (15–22)


(patients incorrectly classified as not
having HBsAg)

True negatives 12 studies Cohort & case–control type Serious Not serious Serious3 Not serious None 941 (940–944) 793 (791–795) ⨁⨁◯◯
(patients without HBsAg) 7552 patients studies Low3

False positives 9 (6–10) 7 (5–9)


(patients incorrectly classified as
having HBsAg)

25 24 30
1. Lin , Lien and Randrianirina used a case–control design
2. Significant heterogeneity across studies for sensitivity; tau-squared 20.2
3. Heterogeneity exists, but with lower clinical impact; tau-squared 1.56

Page | 223
6. Discussion
Study findings

Diagnostic accuracy of HBsAg assays using immunoassay reference

Overall, the diagnostic accuracy of 33 RDTs and 8 EIAs were assessed against an EIA reference
standard. Total numbers of patients included 36,131 (RDT vs EIA, figure 3) and 3751 (EIA vs
EIA, figure 4). Both RDTs and EIAs had similar sensitivity and specificity compared to an EIA
reference standard (Table 3).
Clinical sensitivity estimates for both RDTs and EIAs were characterized by statistical
heterogeneity, whereas specificity estimates were less heterogeneous (Figures 3 and 4). This
applied across brands (Table 6). Heterogeneity can be caused by the use of different reference
standards assays, clinical subgroups within the study population, age (children versus adults),
patient status and stage of disease.
Compared to previous systematic reviews, the pooled clinical sensitivity 90.0% (95%
CI: 89.1–90.8) and specificity 99.5% (95% CI: 99.4–99.5) is slightly inferior for RDTs compared
to an EIA reference standard (Table 3). In particular Results were very heterogeneous in terms
of sensitivity (Table 7). Khuroo et al.11 reported 96.7% sensitivity (95% CI: 95.3, 97.7) and
99.7% specificity (95% CI: 98.6, 99.9). Studies included conference abstracts and studies using
reference panels. Shivkumar et al.60 reported a pooled sensitivity 98.2% (95% CI: 94.7, 99.9)
and pooled specificity 99.9% (95% CI: 99.3, 100).

Table 7. Summary pooled diagnostic accuracy of HBsAg assays compared to previous


reviews.

Review, year Index Clinical Accuracy


Test
n Sen (95% CI) Spec (95% CI)

PICO 1, 2015 RDT 63 90.0 (89.1–90.8) 99.5 (99.4–99.5)

EIA 8 88.9 (87.0–90.6) 98.4 (97.8–98.8)

Hwang, 2008 RDT 98.1 (97.7–98.5) 99.6 (99.2–99.9)

Shivkumar, 2012 RDT 94.8 (90.1–98.2) 99.5 (99.0–100)

Khuroo, 2014 RDT 97.1 (96.1–7.9) 99.9 (99.8–00)

*Sen : sensitivity; Spec : specificity; CI : confidence interval; RDT : rapid diagnostic test; EIA : enzyme immunoassay;
+ –
LR : positive likelihood ratio; LR : negative likelihood ratio; REM : random effects model

When comparing EIAs to newer CMIA (chemiluminescent assays), two standard ELISA/
EIA based assays manufactured in markets with transitioning economies appeared to perform
poorly compared to other assays.4 Of note, data exists for one assay (KHB) using different
signal cut-off ratio’s (S/CO), with improved sensitivity but worse specificity when using the
grey zone (S/CO 0.2–0.99); sensitivity 96.2%, specificity 70.6% compared to sensitivity of
73.8% and specificity of 88.1%.45

Page | 224
Results for EIAs are more specific but less sensitive when used in conjunction with
manufacturers’ neutralisation assays. In one study,56 Liason HBsAg (Diasorin) had sensitivity
100% and specificity 70% compared to NAT, with improved overall accuracy using with a
neutralisation assay (sensitivity 96%, specificity 100%). This shows the critical need to use the
neutralization step to confirm any HBsAg reactivity observed upon initial testing.

Diagnostic accuracy of HBsAg assays using NAT reference


The specificity of RDTs and EIA was lower when using a NAT reference; this is understandable
as the serological assays detect HBsAg whereas NAT detects HBV DNA. The viral kinetics of
HBV DNA and HBsAg are not identical. One limitation of HBV DNA as a diagnostic assay would
be that those on anti-viral treatment, including antiretroviral therapy for HIV (containing
lamivudine or tefonovir) might be non-detectable HBV DNA in the presence of HBsAg.
The pooled clinical sensitivity of RDTs was 93.3% (95% CI 91.3–94.9) and significantly
higher than EIAs 75.7% (95% CI 72.1–79.1) using a NAT reference. It is important to note that
study characteristics varied, with one laboratory based case-control study with pre-selected
patients showing particularly good results for RDTs vs NAT and providing 6 data points in this
analysis.47 The remaining studies in this sub-analysis were all performed in field settings in
resource-limited settings, with poor sensitivity (38–60%).

Sub-analyses
HIV
Our results showed that RDTs may be less sensitive in HIV-positive patients. There was still
heterogeneity in terms of results, with one otherwise good quality review finding that
Determine was 96 % sensitive (95% CI: 80,100) and 100% specific (95% CI: 99,100) in this
cohort.20
The difficulty of accurate diagnosis in HIV patients is possibly explained by an
increased incidence of hepatitis B and in particular occult hepatitis B in this cohort. In
Sudanese HIV-positive ART naïve patients, 27% had detectable HBV DNA, with occult hepatitis
B in 15%.61 Among 495 treatment naïve, HIV-infected adults in Cote-d’Ivoire, 13% were HBsAg
positive, 42% isolated anti-HBc positive, and 10% occult hepatitis B only detected by NAT.62
Median HBV DNA level was lower in those with occult HBV compared to those with CHB.
Immune pressure has also been hypothesized to contribute, with Geretti et al. noting
that discrepant results for RDTs were all mutants in their study. The overlapping surface and
polymerase genes in the HBV genome could imply that RT inhibitors (e.g. lamivudine) can lead
to the emergence of variants carrying mutations of both the polymerase and surface genes,
hence avoiding detection by standard HBsAg assays.

Blood Donors
Pre-transfusion screening of blood donations is a major public health challenge in resource-
limited settings, where prevalence rates for TTIs (transfusion-transmissible infections) are
significant. Screening of individuals with RDTs pre-donation have been adopted in areas with
insufficient laboratory capacity.

Page | 225
Diagnostic accuracy in our review was similar in blood donors compared to the overall
pooled estimates of RDTs, with sensitivity 91.6 % (95% CI: 90.1, 92.9) and specificity 99.5%
(95% CI 99.3, 99.7). Results were very heterogeneous, with one study in particular having very
low sensitivity (~50–60). Two studies comparing EIAs against EIAs in blood donors had higher
accuracy, with pooled sensitivity 99.3 % (95% CI: 96.4, 100) and specificity 90.9 % (95% CI
82.9, 96.0).
A recent multinational assessment accuracy of TTI screening in Africa using both RDTs
and EIAs on an external quality assessment panel found poor overall sensitivity (75.6%) and
specificity (94.5%) for HBsAg detection.63 This was driven by very poor clinical sensitivity
(47.4%) of HBsAg RDTs, which was lower than that for HCV (63.7%) and HIV (72.4%) in this
population. This can be explained by their lower analytical sensitivity, difficulties in transport
and quality assurance, in addition to often studies being performed on smaller scales. In a
Nigerian blood donor study, 10% of 113 HBsAg-negative repeat donors using RDTs were found
to have quantifiable HBV DNA.57 These patients either had acute infection or occult chronic
infection.
In a recent systematic review of studies evaluating RDTs for infectious disease blood
screening in Africa, there was again significant variability in performance.64 RDTs for HBsAg
detection were again identified for suboptimal sensitivities, with questionable suitability,
especially in high prevalence regions. High false negatives could be due to operator error, low
HBsAg levels, assay degeneration or lot variation.

Whole blood
For rapid diagnostic tests, accuracy using whole blood (capillary and venous) was marginally
superior to serum. The accuracy was comparable to that of EIAs using serum; data from the
eight studies (eleven data points) is also less heterogenous (Annex 9.3.2). The significantly
lower sensitivity of RDTs using plasma is possibly explained by the nature of the studies; one
was in a population of blood donors, while the other was initially designed to assess the
accuracy of RDTs in determining HBV and HCV prevalence in a Malawian population with high
HIV-co-infection rates.41 The authors of the latter study hypothesised that local operational
problems or unexpected technical issues were the reason for poor performance in resource-
limited setting. Others have also since suggested that the high HIV-co-infection rate could
have contributed, with suppression of HBV replication using lamivudine containing regiments
potentially hindering affecting detection by RDT.

Study setting – field


Some heterogeneity is explained by location where RDTs are performed. Two different studies
from Malawi, in a predominantly HIV-positive cohort41 or entirely HIV-positive cohort19
produced very different results for the same test (Determine HBsAg). Pooled sensitivity 56%
(95% CI: 38, 73) and specificity 69% (95% CI: 62, 76) were much lower in the field study
compared to the sensitivity 100% (95% CI: 86, 100) and specificity 100% (95% CI: 93, 100) in
the study where samples were returned to the UK.

Page | 226
Analytical sensitivity of different assays as a source of heterogeneity
We were unable to explain heterogeneity of results using different assays. Very few rapid
tests meet required analytical LOD (0.130 IU/mL) required by regulatory authorities, but
because of insufficient data in studies we were unable to stratify using LOD as a source of
heterogeneity. This is important as it has been suggested that false-negative HBsAg RDTs are
associated with lower HBsAg levels, low viral load, HBsAg mutants, or specific genotypes, in
addition to ART exposure where lamivudine and tenofovir are used. (16, 25, 36, 65)
In a recent fields studies in the Gambia,40 the range of serum HBsAg levels quantified
by CMIA that showed reactivity with RDTs in the field was 26.5–27, 320 IU/mL, with a
statistically significant (P = 0.0002) difference in median HBsAg level (875 IU/mL) compared to
false negatives using RDT’s (median 1.2; range 0.8–25.5 IU/mL). Interestingly significantly
more false-negatives were female (P = 0.05) with lower median ALT levels (P = 0.01). The
laboratory-based study from the same publication in a chronic hepatitis B cohort found a
higher range of HBsAg levels (2.8–124,925 IU/mL) in those testing positive with RDTs, with a
significant difference in median HBsAg levels (7, 482 vs 0.40 IU/mL; P <0.0001) and median
ALT (P = 0.01) between true-positives and false negatives. The lower limit of detection may be
explained by differences in methodology, given the setting (laboratory vs field), reference test
(CMIA vs ELISA) and sample type (dried blood spots vs serum). This suggests that subjects with
false negatives have lower HBsAg levels and inactive disease compared to true-positives,
minimising the impact of reduced sensitivity. Unfortunately, in the single study identified also
assessing LSM, 17% (4/23) subjects with false negative results had evidence of fibrosis and
would require antiviral therapy.
Another study (Bottero et al)16 also found significantly lower median HBsAg in false
negatives vs true positives. [19.5 vs 2351 IU/mL; p=0.0001], with only 4 false negative having
HBsAg >10 IU/mL. HBV DNA was usually below 200 IU/mL. False positives occurred in
vaccinate patients (n=7), and one patient with resolved infection and anti-HBs titre.
Interestingly ALL false negatives were HBcAb positive.
Data exists from large studies of analytical sensitivity using reference, seroconversion,
mutant panels.53, 65, 66 These include specimens from individuals with low antigenaemia, such
as early infection. Unfortunately, studies of analytical sensitivity of EIAs are conflicting. One
study suggested that 9 out of 10 EIAs were able to detect HBsAg levels as low as 0.2 IU/mL
irrespective of genotype.67 Another comparing newer EIAs (Advia Centaur; Monolisa Ultra;
Liasion; Vidas Ultra) using reference and mutant panels found a lower limit of detection
<350IU/mL, but with varying sensitivity for mutant detection (37.1%–91.4%).68 The authors
hypothesised that the lack of detection was due to epitope recognition of the anti-HBs assay
reagents in the capture phase and in conjugates. Another study assessing 13 different assays
with mutant panels found a range of LOD (0.011-0.096 IU/mL) and sensitivity (63%–98%) in
mutants. Another study found comparable analytical sensitivity between four EIAs but
significant differences in detection of mutants between assays.69 One blood donor study in
China found a significant difference in sensitivities and mutant detection capabilities amongst
assays used by blood banks, with the urgent recommendation of a list of high sensitivity
assays for blood bank screening.70

Page | 227
In blood donors, studies have found some correlation between HBsAg levels (IU/mL)
and NAT (copies/mL).71 The obvious benefit of more sensitive assays with lower limits of
detection would be improved detection of those with occult hepatitis B or in the early window
period of sero-conversion.
It has been suggested that utilizing “grey zones” in EIAs could improve sensitivity and
allow combination of tests to develop of economic testing strategies.45 Sensitivity improved
from 76–88% to 96–97%, with a further increase to 99% when combining the use of two EIAs.
Studies looking at RDTs using clinical panels have found sometimes conflicting results.
One study found equivalent specificities but significant differences in assay sensitivity
between Uni-Gold™ HBsAg and Determine HBsAg.72
Interestingly, studies in Cambodia and Viet Nam by the same group produced
different results for sensitivity, suggesting uncontrolled variables, such as prozone effect and
genotype variations. The prozone effect may explain why some true positives turned out
negative with rapid tests. Given that specificity is excellent but sensitivity is low suggests that
this is genuine poor performance.

Study strengths and limitations

Significant strengths of this meta-analysis include the global evidence base, rigorous pre-
specified protocol incorporating numerous major scientific databases, in addition to review of
the related literature, notably occult hepatitis B and the impact of NAT. We included studies
performed in a range of settings, with a diverse population. We only included studies with
bivariate data, to minimise bias, measuring clinical sensitivity which are more applicable. We
also included evaluations of both RDTs and EIAs, and as such are able to provide a more
comprehensive meta-analysis. Comparing RDTs with EIAs, we were able to identify an
additional 11 studies not found in previous reviews. Incorporation bias was unlikely as all
participants received both index and confirmatory tests independently. As all studies
administered the same reference standard to all patients, which reduces risk of verification
bias. Our study also excluded articles deemed to be high risk of bias or less applicable, such as
conference abstracts or reference panels; reference panels, included in previous reviews, have
higher accuracy than that of tests used in the field but are not as useful in guiding policy.
Accuracy on seroconversion panels do not necessarily reflect the antibody or antigen
spectrum in the populations studied.
Our study, did, however have a number of limitations. First, we only included studies
in English, which potentially introduces publication (language) bias. We will identify relevant
studies in non-English languages from reference lists to address whether this contributed to a
substantial bias or not. Second, a significant proportion of studies were case–control in design
or used preselected cohorts, which would bias results. For example, in well-conducted studies
from the Gambia,40 those in the community setting had a smaller range of HBsAg levels (26.5–
27, 320 IU/mL) than the study conducted in chronic hepatitis B patients (2.8–124,925 IU/mL).
This is one reason for the reduced sensitivity of the same test (Determine HBsAg) in the
screening cohort (88%; 95% CI: 81, 94) compared to the chronic hepatitis cohort (95%; 95% CI:

Page | 228
90, 98) from the same community. Third, some analyses were based on a small number of
studies, which included few positive samples.
There are number of technical and patient factors that could impact accuracy, which
cannot be addressed based on the currently available literature. Specific to hepatitis B
diagnosis, we were unable to correlate the heterogeneity of sensitivity with different stages,
severity and genotypes infection. This was due to insufficient information in studies,
principally additional serology such as HBeAg, anti-HBc IgM, anti-HBc total and anti-HBs
antibodies. Genetic information has long been suspected to impact on diagnostic accuracy,67‒
70, 73, 74
although a recent study of analytical performance found no difference in the detection
of mutants.70, 73‒75 It should be noted that this study was conducted by authors with significant
conflicts of interest. Mutants themselves are also rapidly evolving, such that the prevalence
and type of specific mutants cannot be determined based on historical data, making studies
difficult to organize. Finally, occult hepatitis should also be considered. The addition of NAT
would be useful to stratify patients’ results, but the lack of sufficient of quality studies (Annex
9.2) is testament to the challenges in conducting advanced laboratory based studies in areas
of high disease prevalence. Reference standards are also imprecise, resulting in overdiagnosis
of clinically insignificant disease, and underestimation of diagnostic accuracy of clinically
relevant disease. The natural history of hepatitis B, notably progression and infectiousness, is
being investigated for correlations of quantitative HBsAg, HbeAg and HBV DNA levels. Low
levels of either antigen may not be significant clinically. Short-term spontaneous fluctuations
in DNA and HBsAg are recognized in those with chronic hepatitis B and add extra challenges to
accurate diagnosis with a “gold standard”.76
Statistical heterogeneity was an obvious issue as is often observed in diagnostic
accuracy reviews. Although we performed stratified analyses to identify potential sources,
none fully explained the heterogeneity observed. Firstly, studies evaluated different RDTs,
with rapid changes in technology for both EIAs and RDTs meaning that analytical sensitivity is
variable among the assays evaluated. Although we pooled based on some RDT brands
(Determine HBsAg, BinaxNOW) there were insufficient studies and this pooling did not
entirely account for heterogeneity. Another potential confounder is changes in manufacturing
processes, including components used to manufacture assays Determine, as one example, has
been produced by Abbott, Inverness Medical and Alere Medical Co. Ltd; as the test has been
commercially available for over 10 years, there will undoubtedly be minor product changes.

7. Conclusion
WHO has emphasized the importance of timely global testing, prevention and treatment of
hepatitis B, with predictions of an increasing prominence as a cause of death globally in years
to come. Although RDTs have limitations, many of which can be addressed through improved
training and quality assurance systems, they are frequently the only viable option for
infectious screening in resource-limited settings. Therefore, additional studies and specific
guidelines regarding the use of RDTs in the context of blood safety and patient screening are
needed. In terms of global uptake, lower costs of these assays and ease of use across a variety
of endemic settings is crucial to achieving goals for control of hepatitis. Worldwide, a

Page | 229
significant proportion of countries are unable to afford quality-assured laboratory-based
testing with enzyme immunoassays; the use of NAT to further reduce the window period of
infection and detect occult hepatitis is beyond reach in many settings at present.
This meta-analysis, along with others, suggests that assays for detection of HBsAg
including RDTs and enzyme immunoassays have the potential to contribute significantly to the
control of hepatitis B globally in endemic areas, which are often include low resource remote
regions. Other benefits of RDTS include easy storage, small sample volumes required with
minimal staff training or additional equipment. Unfortunately, with current issues with poor
clinical and analytical sensitivity and potential difficulties in detection of occult hepatitis B and
mutant variants, a number of cases would be missed. There is also concern that sensitivity is
significantly reduced in HIV-positive patients.
There are numerous difficulties in conducting systematic reviews of the performance
of in vitro diagnostics, particularly in resource-limited settings. There is a significant variation
in terms of quality of studies, most with key parameters missing. Further promotion of current
accepted standards to performing and reporting studies of diagnostic accuracy globally can
help improve the evidence base currently available. Further high quality studies are
desperately needed to assess the accuracy in a variety of settings and support the growing
evidence base for RDTs. Specifically, further studies looking at the impact of different
geographic locations and mutant phenotypes would be invaluable.
From included studies, excellent robust specificity of all assays is reassuring in terms
of ensuring cost–effective initiation of algorithms for further investigation and treatment.
Significant heterogeneity and suboptimal sensitivity of RDTs has to be taken into
consideration as country control programmes consider the trade-off between affordability,
accuracy and accessibility (i.e. ease of use in all levels of the health-care system). The
weighting of these three factors are country specific and could be modelled.

8. References
1. Ott JJ, Stevens GA, Groeger J, Wiersma ST. Global epidemiology of hepatitis B virus
infection: new estimates of age-specific HBsAg seroprevalence and endemicity. Vaccine.
2012;30(12):2212–9.
2. Ganem D, Prince AM. Hepatitis B virus infection – natural history and clinical
consequences. N Engl J Med. 2004;350(11):1118–29.
3. Fattovich G, Stroffolini T, Zagni I, Donato F. Hepatocellular carcinoma in cirrhosis:
incidence and risk factors. Gastroenterology. 2004;127(5 Suppl 1):S35–50.
4. Lozano R, Naghavi M, Foreman K, Lim S, Shibuya K, Aboyans V, et al. Global and regional
mortality from 235 causes of death for 20 age groups in 1990 and 2010: a systematic
analysis for the Global Burden of Disease Study 2010. Lancet. 2012;380(9859):2095–128.
5. Guidelines for the prevention, care, and treatment of persons with chronic hepatitis B
infection. Geneva: WHO; 2015.

Page | 230
6. Fabrizi F, Martin P, Dixit V, Bunnapradist S, Dulai G. Meta-analysis: the effect of age on
immunological response to hepatitis B vaccine in end-stage renal disease. Aliment
Pharmacol Ther. 2004;20(10):1053–62.
7. Thompson Coon J, Rogers G, Hewson P, Wright D, Anderson R, Cramp M, et al.
Surveillance of cirrhosis for hepatocellular carcinoma: systematic review and economic
analysis. Health Technol Assess. 2007;11(34):1–206.
8. Qin XK, Li P, Han M, Liu JP. Xiaochaihu Tang for treatment of chronic hepatitis B: a
systematic review of randomized trials. [Article in Chinese]. Zhong Xi Yi Jie He Xue Bao.
2010;8(4):312–20.
9. Hwang SH, Oh HB, Choi SE, Kim HH, Chang CL, Lee EY, et al. [Meta-analysis for the pooled
sensitivity and specificity of hepatitis B surface antigen rapid tests.] [Article in Korean].
Korean J Lab Med. 2008;28(2):160–8.
10. Shivkumar S, Peeling R, Jafari Y, Joseph L, Pai NP. Rapid point-of-care first-line screening
tests for hepatitis B infection: a meta-analysis of diagnostic accuracy (1980–2010). Am J
Gastroenterol. 2012;107(9):1306–13.
11. Khuroo NS, Khuroo MS. Accuracy of rapid point-of-care diagnostic tests for hepatitis B
surface antigen – a systematic review and meta-analysis. J Clin Exp Hepatol.
2014;4(3):226–40.
12. Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ, Reitsma JB, et al. QUADAS-2:
a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med.
2011;155(8):529–36.
13. Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP, Irwig LM, et al. The STARD
statement for reporting studies of diagnostic accuracy: explanation and elaboration. Ann
Intern Med. 2003;138(1):W1–W12.
14. Abraham P, Sujatha R, Raghuraman S, Subramaniam T, Sridharan G. Evaluation of two
immunochromatographic assays in relation to 'RAPID' Screening of HBsAg. Indian J Med
Microbiol. 1998;16:23–5.
15. Akanmu AS, Esan OA, Adewuyi JO, Davies AO, Okany CC, Olatunji RO, et al. Evaluation of
a rapid test kit for detection of HBsAg/eAg in whole blood: a possible method for pre-
donation testing. Afr J Med Med Sci. 2006;35(1):5–8.
16. Bottero J, Boyd A, Gozlan J, Lemoine M, Carrat F, Collignon A, et al. Performance of rapid
tests for detection of HBsAg and anti-HBsAb in a large cohort, France. . J Hepatol.
2013;58(3):473–8.
17. Chameera EWS, Noordeen F, Pandithasundara H, Abeykoon AMSB. Diagnostic efficacy of
rapid assays used for the detection of hepatitis B virus surface antigen. Sri Lankan J Infect
Dis. 2013;3(2):21–7.
18. Clement F, Dewint P, Leroux-Roels G. Evaluation of a new rapid test for the combined
detection of hepatitis B virus surface antigen and hepatitis B virus e antigen. J Clin
Microbiol. 2002;40(12):4603–6.

Page | 231
19. Davies J, van Oosterhout JJ, Nyirenda M, Bowden J, Moore E, Hart IJ, et al. Reliability of
rapid testing for hepatitis B in a region of high HIV endemicity. Trans R Soc Trop Med
Hyg. 2010;104(2):162–4.
20. Franzeck FC, Ngwale R, Msongole B, Hamisi M, Abdul O, Henning L, et al. Viral hepatitis
and rapid diagnostic test based screening for HBsAg in HIV-infected patients in rural
Tanzania. PLoS One. 2013;8(3):e58468.
21. Kaur H, Dhanao J, Oberoi A. Evaluation of rapid kits for detection of HIV, HBsAg and HCV
infections. Indian J Med Sci. 2000;54(10):432–4.
22. Khan J, Lone D, Hameed A. Evaluation of the performance of two rapid
immunochromatographic tests for detection of Hepatitis B surface antigen and anti HCV
antibodies using ELISA tested samples. . AKEMU. 2010;16(1 S1):84–7.
23. Lau DT, Ma H, Lemon SM, Doo E, Ghany MG, Miskovsky E, et al. A rapid
immunochromatographic assay for hepatitis B virus screening. J Viral Hepat.
2003;10(4):331–4.
24. Lien TX, Tien NTK, Chanpong GF, Cuc CT, Yen VT, Soderquist R, et al. Evaluation of rapid
diagnostic tests for the detection of human immunodeficiency virus types 1 and 2,
hepatitis B surface antigen, and syphilis in Ho Chi Minh City, Vietnam. Am J Trop Med
Hyg. 2000;62(2):301–9.
25. Lin YH, Wang Y, Loua A, Day GJ, Qiu Y, Nadala EC, Jr., et al. Evaluation of a new hepatitis B
virus surface antigen rapid test with improved sensitivity. J Clin Microbiol.
2008;46(10):3319–24.
26. Mvere D, Constantine NT, Katsawde E, Tobaiwa O, Dambire S, Corcoran P. Rapid and
simple hepatitis assays: encouraging results from a blood donor population in Zimbabwe.
Bull World Health Organ. 1996;74(1):19–24.
27. Oh J, Kim TY, Yoon HJ, Min HS, Lee HR, Choi TY. Evaluation of Genedia® HBsAg Rapid and
Genedia® anti-HBs Rapid for the screening of HBsAg and Anti-HBs. Korean J Clin Pathol.
1999;19:114–7.
28. Ola SO, Otegbayo JA, Yakubu A, Aje AO, Odaibo GN, Shokunbi W. Pitfalls in diagnosis of
Hepatitis B virus infection among adults nigerians. Niger J Clin Pract. 2009;12(4):350–4.
29. Raj AA, Subramaniam T, Raghuraman S, Abraham P. Evaluation of an indigenously
manufactured rapid immunochromatographic test for detection of HBsAg. Indian J Pathol
Microbiol. 2001;44(4):413–4.
30. Randrianirina F, Carod JF, Ratsima E, Chretien JB, Richard V, Talarmin A. Evaluation of the
performance of four rapid tests for detection of hepatitis B surface antigen in
Antananarivo, Madagascar. J Virol Methods. 2008;151(2):294–7.
31. Sato K, Ichiyama S, Iinuma Y, Nada T, Shimokata K, Nakashima N. Evaluation of
immunochromatographic assay systems for rapid detection of hepatitis B surface antigen
and antibody, Dainascreen HBsAg and Dainascreen Ausab. J Clin Microbiol.
1996;34(6):1420–2.
32. Bjoerkvoll B, Viet L, Ol HS, Lan NT, Sothy S, Hoel H, et al. Screening test accuracy among
potential blood donors of HBsAg, anti-HBc and anti-HCV to detect hepatitis B and C virus

Page | 232
infection in rural Cambodia and Vietnam. Southeast Asian J Trop Med Public Health.
2010;41(5):1127–35.
33. Chevaliez S, Challine D, Naija H, Luu TC, Laperche S, Nadala L, et al. Performance of a new
rapid test for the detection of hepatitis B surface antigen in various patient populations. J
Clin Virol. 2014;59(2):89–93.
34. Erhabor O, Kwaifa I, Bayawa A, Isaac Z, Dorcas IaS I. Comparison of ELISA and rapid
screening techniques for the detection of HBsAg among blood donors in Usmanu
Danfodiyo University Teaching Hospital Sokoto, North Western Nigeria. J Blood Lymph.
2013;4(2).
35. Gish RG, Gutierrez JA, Navarro-Cazarez N, Giang K, Adler D, Tran B, et al. A simple and
inexpensive point-of-care test for hepatitis B surface antigen detection: serological and
molecular evaluation. . J Viral Hepat. 2014;21(12):905–8.
36. Geretti AM, Patel M, Sarfo FS, Chadwick D, Verheyen J, Fraune M, et al. Detection of
highly prevalent hepatitis B virus coinfection among HIV-seropositive persons in Ghana. J
Clin Microbiol. 2010;48(9):3223–30.
37. Hoffmann CJ, Dayal D, Cheyip M, McIntyre JA, Gray GE, Conway S, et al. Prevalence and
associations with hepatitis B and hepatitis C infection among HIV-infected adults in South
Africa. Int J STD AIDS. 2012;23(10):e10–3.
38. Honge BL, Jespersen S, Te DS, da Silva ZJ, Laursen AL, Krarup H, et al. Hepatitis B virus
surface antigen and anti-hepatitis C virus rapid tests underestimate hepatitis prevalence
among HIV-infected patients. HIV Med. 2014;15(9):571–6.
39. Mutocheluh M, Owusu M, Kwofie TB, Akadigo T, Appau E, Narkwa PW. Risk factors
associated with hepatitis B exposure and the reliability of five rapid kits commonly used
for screening blood donors in Ghana. BMC Res Notes. 2014;7:873.
40. Njai HF, Shimakawa Y, Sanneh B, Ferguson L, Ndow G, Mendy M, et al. Validation of rapid
point-of-care (POC) tests for detection of hepatitis B surface antigen in field and
laboratory settings in the Gambia, Western Africa. J Clin Microbiol. 2015;53(4):1156–63.
41. Nyirenda M, Beadsworth MB, Stephany P, Hart CA, Hart IJ, Munthali C, et al. Prevalence
of infection with hepatitis B and C virus and coinfection with HIV in medical inpatients in
Malawi. J Infect. 2008;57(1):72–7.
42. Upreti SR, Gurung S, Patel M, Dixit SM, Krause LK, Shakya G, et al. Prevalence of chronic
hepatitis B virus infection before and after implementation of a hepatitis B vaccination
program among children in Nepal. Vaccine. 2014;32(34):4304–9.
43. Liu C, Chen T, Lin J, Chen H, Chen J, Lin S, et al. Evaluation of the performance of four
methods for detection of hepatitis B surface antigen and their application for testing
116,455 specimens. J Virol Methods. 2014;196:174–8.
44. Ol HS, Bjoerkvoll B, Sothy S, Van Heng Y, Hoel H, Husebekk A, et al. Prevalence of
hepatitis B and hepatitis C virus infections in potential blood donors in rural Cambodia.
Southeast Asian J Trop Med Public Health. 2009;40(5):963–71.

Page | 233
45. Peng J, Cheng L, Yin B, Guan Q, Liu Y, Wu S, et al. Development of an economic and
efficient strategy to detect HBsAg: application of "gray-zones" in ELISA and combined use
of several detection assays. Clin Chim Acta. 2011;412(23–24):2046–51.
46. Viet L, Lan NT, Ty PX, Bjorkvoll B, Hoel H, Gutteberg T, et al. Prevalence of hepatitis B &
hepatitis C virus infections in potential blood donors in rural Vietnam. Indian J Med Res.
2012;136(1):74–81.
47. Ansari MHK, Omrani MD, Movahedi V. Comparative evaluation of
immunochromatographic rapid diagnostic tests (strip and device) and PCR methods for
detection of human hepatitis B surface antigens. Hepat Mon. 2007;7(2):87–91.
48. Nakata JM, Johnson JM. Evaluation of rapid immunoconcentration assay for HBsAg in a
prison inmate population. Annual Meeting of the American Society for Microbiology,
Anaheim. 1990.
49. Palmer C, Cuadrado R, Koenig E. Multicenter evaluation of the determine[TM] Rapid tests
for the diagnosis of HIV, Hepatitis B Surface Antigen, and Syphilis. Interscience
Conference on Antimicrobial Agents and Chemotherapy, Ft Lauderdale. 1999.
50. Torane V, Shastri J. Comparison of ELISA and rapid screening tests for the diagnosis of
HIV, hepatitis B and hepatitis C among healthy blood donors in a tertiary care hospital in
Mumbai. Indian J Med Microbiol. 2008;26(3):284–5.
51. Cha YJ, Yang JS, Chae SL. [Evaluation of indigenously manufactured
immunochromatographic assay systems for rapid detection of hepatitis B surface antigen
and antibody.] [Article in Korean]. Korean J Lab Med. 2006;26:52–7.
52. Whang DH, Um TH. [Comparison of immunochromatography assays and quantitative
immunoassays for detecting HBsAg and anti-HBs.] [Article in Korean]. Korean J Lab Med.
2005;25:186–91.
53. Maity S, Nandi S, Biswas S, Sadhukhan SK, Saha MK. Performance and diagnostic
usefulness of commercially available enzyme linked immunosorbent assay and rapid kits
for detection of HIV, HBV and HCV in India. Virol J. 2012;9:290.
54. Lukhwareni A, Burnett RJ, Selabe SG, Mzileni MO, Mphahlele MJ. Increased detection of
HBV DNA in HBsAg-positive and HBsAg-negative South African HIV/AIDS patients
enrolling for highly active antiretroviral therapy at a Tertiary Hospital. J Med Virol.
2009;81(3):406–12.
55. Mphahlele MJ, Lukhwareni A, Burnett RJ, Moropeng LM, Ngobeni JM. High risk of occult
hepatitis B virus infection in HIV-positive patients from South Africa. J Clin Virol.
2006;35(1):14–20.
56. Khadem-Ansari MH, Omrani MD, Rasmi Y, Ghavam A. Diagnostic validity of the
chemiluminescent method compared to polymerase chain reaction for hepatitis B virus
detection in the routine clinical diagnostic laboratory. Adv Biomed Res. 2014;3:116.
57. Nna E, Mbamalu C, Ekejindu I. Occult hepatitis B viral infection among blood donors in
South-Eastern Nigeria. Pathog Glob Health. 2014;108(5):223–8.
58. Olinger CM, Weber B, Otegbayo JA, Ammerlaan W, van der Taelem-Brule N, Muller CP.
Hepatitis B virus genotype E surface antigen detection with different immunoassays and

Page | 234
diagnostic impact of mutations in the preS/S gene. Med Microbiol Immunol.
2007;196(4):247–52.
59. Seremba E, Ocama P, Opio CK, Kagimu M, Yuan HJ, Attar N, et al. Validity of the rapid
strip assay test for detecting HBsAg in patients admitted to hospital in Uganda. J Med
Virol. 2010;82(8):1334–40.
60. Shivkumar S, Peeling R, Jafari Y, Joseph L, Pai NP. Rapid point-of-care first-line screening
tests for hepatitis B infection: a meta-analysis of diagnostic accuracy (1980–2010). Am J
Gastroenterol. 2012;107(9):1306–13.
61. Mudawi H, Hussein W, Mukhtar M, Yousif M, Nemeri O, Glebe D, et al. Overt and occult
hepatitis B virus infection in adult Sudanese HIV patients. Int J Infect Dis. 2014;29:65–70.
62. N'Dri-Yoman T, Anglaret X, Messou E, Attia A, Polneau S, Toni T, et al. Occult HBV
infection in untreated HIV-infected adults in Cote d'Ivoire. Antivir Ther. 2010;15(7):1029–
34.
63. Laperche S, Francophone African Group for Research in Blood T. Multinational
assessment of blood-borne virus testing and transfusion safety on the African continent.
Transfusion. 2013;53(4):816–26.
64. Pruett CR, Vermeulen M, Zacharias P, Ingram C, Tayou Tagny C, Bloch EM. The Use of
rapid diagnostic tests for transfusion infectious screening in Africa: a literature review.
Transfus Med Rev. 2015;29(1):35–44.
65. Scheiblauer H, El-Nageh M, Diaz S, Nick S, Zeichhardt H, Grunert HP, et al. Performance
evaluation of 70 hepatitis B virus (HBV) surface antigen (HBsAg) assays from around the
world by a geographically diverse panel with an array of HBV genotypes and HBsAg
subtypes. [Erratum appears in Vox Sang. 2010 May;98(4):581]. Vox Sang. 2010;98(3 Pt
2):403–14.
66. Scheiblauer H, Soboll H, Nick S. Evaluation of 17 CE-marked HBsAg assays with respect to
clinical sensitivity, analytical sensitivity, and hepatitis B virus mutant detection. J Med
Virol. 2006;78 Suppl 1:S66–70.
67. Mizuochi T, Okada Y, Umemori K, Mizusawa S, Yamaguchi K. Evaluation of 10 commercial
diagnostic kits for in vitro expressed hepatitis B virus (HBV) surface antigens encoded by
HBV of genotypes A to H. J Virol Methods. 2006;136(1–2):254–6.
68. Ly TD, Servant-Delmas A, Bagot S, Gonzalo S, Ferey MP, Ebel A, et al. Sensitivities of four
new commercial hepatitis B virus surface antigen (HBsAg) assays in detection of HBsAg
mutant forms. J Clin Microbiol. 2006;44(7):2321–6.
69. Moerman B, Moons V, Sommer H, Schmitt Y, Stetter M. Evaluation of sensitivity for wild
type and mutant forms of hepatitis B surface antigen by four commercial HBsAg assays.
Clin Lab. 2004;50(3–4):159–62.
70. Zhang R, Wang L, Li J. Hepatitis B virus transfusion risk in China: proficiency testing for
the detection of hepatitis B surface antigen. Transfus Med. 2010;20(5):322–8.
71. Martin LA, Stramer SL, Kuhns MC, Schlauder GG. Correlation of improved hepatitis B
surface antigen detection limits with hepatitis B virus DNA nucleic acid test yield in blood
donations. Transfusion. 2012;52(10):2201–8.

Page | 235
72. Austin PM. Comparative sensitivities and specificities of two rapid HBsAg detection
methods and their relationship to a third generation commercial enzyme immunoassay.
NZ J Med Lab Science. 2005;59(2):38–40.
73. Kuhns MC, McNamara AL, Holzmayer V, Lou SC, Busch MP. Frequency of diagnostically
significant hepatitis B surface antigen mutants. J Med Virol. 2007;79(Suppl. 1):S42–S6.
74. Servant-Delmas A, Mercier-Darty M, Ly TD, Wind F, Alloui C, Sureau C, et al. Variable
capacity of 13 hepatitis B virus surface antigen assays for the detection of HBsAg mutants
in blood samples. J Clin Virol. 2012;53(4):338–45.
75. Hirzel C, Pfister S, Gorgievski-Hrisoho M, Wandeler G, Zuercher S. Performance of HBsAg
point-of-care tests for detection of diagnostic escape-variants in clinical samples. J Clin
Virol. 2015;69:33–5.
76. Maylin S, Sire JM, Mbaye PS, Simon F, Sarr A, Evra ML, et al. Short-term spontaneous
fluctuations of HBV DNA levels in a Senegalese population with chronic hepatitis B. BMC
Infect Dis. 2015;15(1):154.

Page | 236
9. Annexes
9.2. 9.1 Search strategy

Ovid Medline search strategy


Searched on 20 April 2015 from 1946 – April week 2 2015.
1 Hepatitis, Viral, Human/ (10382)
2 Hepatitis Viruses/ (1363)
3 Hepatitis Antibodies/ (5082)
4 exp Hepadnaviridae Infections/ (47484)
5 Hepatitis B Antibodies/ (8638)
6 Hepatitis B virus/ (20604)
7 Hepadnaviridae/ (192)
8 Hepatitis B Surface Antigens/ (17007)
9 (heptatitis-b or hep-b or (hepatitis adj5 b) or (hep adj5 b) or hbv).ti,ab. (64488)
10 hbsag.ti,ab. (15146)
11 or/1-10 [HEPATITIS B] (87943)
12 exp Reagent Kits, Diagnostic/ (17747)
13 ((rapid or point of care or near patient or poc or poct or bedside) adj5 (test or tests or
testing or detect* or diagnos* or screen* or kit or kits or assay* or device*)).ti,ab.
(63080)
14 (radt or radts or rdt or rdts).ti,ab. (909)
15 rapid test*.ti,ab. (3400)
16 exp Enzyme-Linked Immunosorbent Assay/ (127391)
17 Immunoassay/ (23237)
18 Immunoenzyme Techniques/ (64864)
19 (enzyme-linked immunosorbent assay or ELISA).ti,ab. (139374)
20 (enzyme adj2 (immunoassay* or immuno-assay* or immunosorbent)).ti,ab. (83849)
21 ((antigen* or antibod*) adj3 detect*).ti,ab. (59427)
22 or/12-21 [RAPID DIAGNOSTIC TESTS] (394724)
23 exp "Sensitivity and Specificity"/ (435087)
24 (diagnos* accura* or sensitiv* or specific* or valid*).ti,ab. (3016884)
25 roc curve.ti,ab. (10226)
26 positive predictive value.ti,ab. (25496)
27 negative predictive value.ti,ab. (20415)
28 or/23-27 [DIAGNOSTIC ACCURACY] (3235789)
29 11 and 22 and 28 (3103)
30 Humans/ (13846846)
31 Animals/ (5442465)
32 30 and 31 (1513142)
33 31 not 32 [ALL ANIMAL STUDIES WHICH DO NOT INCLUDE COMPARISON WITH HUMANS]
(3929323)
34 29 not 33 (2856)

Page | 237
35 limit 34 to english language (2345)

Ovid Embase search strategy


Searched on 20 April 2015 from 1947 – 2015 April 17.
1 hepatitis virus/ (4410)
2 hepatitis antibody/ (2216)
3 exp hepadnaviridae/ (42214)
4 hepatitis B surface antigen/ (27312)
5 (heptatitis-b or hep-b or (hepatitis adj5 b) or (hep adj5 b) or hbv).ti,ab. (94627)
6 hbsag.ti,ab. (22290)
7 or/1-6 [HEPATITIS B] (111321)
8 exp diagnostic kit/ (13384)
9 "point of care testing"/ (5530)
10 ((rapid or point of care or near patient or poc or poct or bedside) adj5 (test or tests or
testing or detect* or diagnos* or screen* or kit or kits or assay* or device*)).ti,ab.
(88003)
11 (radt or radts or rdt or rdts).ti,ab. (1652)
12 rapid test*.ti,ab. (5049)
13 enzyme linked immunosorbent assay/ (229634)
14 immunoassay/ (48491)
15 enzyme immunoassay/ (36845)
16 enzyme linked immunospot assay/ (6789)
17 enzyme multiplied immunoassay technique/ (768)
18 (enzyme-linked immunosorbent assay or ELISA).ti,ab. (204052)
19 (enzyme adj2 (immunoassay* or immuno-assay* or immunosorbent)).ti,ab. (98704)
20 antigen detection/ (18155)
21 antibody detection/ (34389)
22 ((antigen* or antibod*) adj3 detect*).ti,ab. (76053)
23 or/8-22 [RAPID DIAGNOSTIC TESTS] (525203)
24 "sensitivity and specificity"/ (221828)
25 diagnostic accuracy/ (189329)
26 (diagnos* accura* or sensitiv* or specific* or valid*).ti,ab. (4113104)
27 roc curve.ti,ab. (20232)
28 positive predictive value.ti,ab. (37541)
29 negative predictive value.ti,ab. (31612)
30 or/24-29 [DIAGNOSTIC ACCURACY] (4280338)
31 7 and 23 and 30 (4018)
32 human/ (15785497)
33 animal/ (1646303)
34 32 and 33 (404532)
35 33 not 34 [ALL ANIMAL STUDIES WHICH DO NOT INCLUDE COMPARISON WITH HUMANS]
(1241771)

Page | 238
36 31 not 35 (3963)
37 limit 36 to english language (3344)

Page | 239
Web of Science
Search was conducted on the Science Citation Index Expanded (1970–20 April 2015) and the
Conference Proceedings Citation Index-Science (1990–20 April 2015).
1 TOPIC: ("hepatitis-b" OR "hep-b" OR (hepatitis near/5 b) OR (hep near/5 b) OR hbv)
(79,505)
2 TOPIC: (hbsag) (12,160)
3 #2 OR #1 (81,526)
4 TOPIC: ((rapid near/5 test) or (rapid near/5 tests) or (rapid near/5 testing) or (rapid
near/5 detect*) or (rapid near/5 diagnos*) or (rapid near/5 screen*) or (rapid near/5 kit)
or (rapid near/5 kits) or (rapid near/5 assay*) or (rapid near/5 device*)) (77,863)
5 TOPIC: (("point of care" near/5 test) or ("point of care" near/5 tests) or ("point of care"
near/5 testing) or ("point of care" near/5 detect*) or ("point of care" near/5 diagnos*) or
("point of care" near/5 screen*) or ("point of care" near/5 kit) or ("point of care" near/5
kits) or ("point of care" near/5 assay*) or ("point of care" near/5 device*)) (5,974)
6 TOPIC: (("near patient" near/5 test) or ("near patient" near/5 tests) or ("near patient"
near/5 testing) or ("near patient" near/5 detect*) or ("near patient" near/5 diagnos*) or
("near patient" near/5 screen*) or ("near patient" near/5 kit) or ("near patient" near/5
kits) or ("near patient" near/5 assay*) or ("near patient" near/5 device*)) (423)
7 TOPIC: ((poc near/5 test) or (poc near/5 tests) or (poc near/5 testing) or (poc near/5
detect*) or (poc near/5 diagnos*) or (poc near/5 screen*) or (poc near/5 kit) or (poc
near/5 kits) or (poc near/5 assay*) or (poc near/5 device*)) (866)
8 TOPIC: ((poct near/5 test) or (poct near/5 tests) or (poct near/5 testing) or (poct near/5
detect*) or (poct near/5 diagnos*) or (poct near/5 screen*) or (poct near/5 kit) or (poct
near/5 kits) or (poct near/5 assay*) or (poct near/5 device*)) (522)
9 TOPIC: ((bedside near/5 test) or (bedside near/5 tests) or (bedside near/5 testing) or
(bedside near/5 detect*) or (bedside near/5 diagnos*) or (bedside near/5 screen*) or
(bedside near/5 kit) or (bedside near/5 kits) or (bedside near/5 assay*) or (bedside
near/5 device*)) (2,705)
10 TOPIC: (radt or radts or rdt or rdts) (1,406)
11 TOPIC: ("rapid test*") (3,783)
12 TOPIC: ("enzyme-linked immunosorbent assay" or ELISA) (141,435)
13 TOPIC: ((enzyme near/2 immunoassay*) or (enzyme near/2 immuno-assay*) or (enzyme
near/2 immunosorbent)) (85,660)
14 TOPIC: ((antigen* near/3 detect*) or (antibod* near/3 detect*)) (56,976)
15 #14 OR #13 OR #12 OR #11 OR #10 OR #9 OR #8 OR #7 OR #6 OR #5 OR #4 (286,936)
16 TOPIC: ("diagnos* accura*" or sensitiv* or specific* or valid*) (4,557,124)
17 TOPIC: ("roc curve") (12,767)
18 TOPIC: ("positive predictive value") (23,706)
19 TOPIC: ("negative predictive value") (18,947)
20 #19 OR #18 OR #17 OR #16 (4,566,667)
21 #20 AND #15 AND #3 (1,789)
22 #20 AND #15 AND #3 Refined by: LANGUAGES: ( ENGLISH ) (1,720)

Page | 240
Scopus
Search was conducted on 20 April 2015.
TITLE-ABS-KEY (("heptatitis-b" OR "hep-b" OR (hepatitis W/5 b) OR (hep W/5 b) OR hbv OR
hbsag) AND (((rapid OR "point of care" OR "near patient" OR poc OR poct OR bedside) W/5
(tests OR test OR testing OR detect* OR diagnos* OR screen* OR kit OR kits OR assay* OR
device*)) OR radt OR radts OR rdt OR rdts OR "rapid test*" OR "enzyme-linked
immunosorbent assay" OR elisa OR (enzyme W/2 (immunoassay* OR immuno-assay* OR
immunosorbent)) OR ((antibod* OR anigen*) W/3 detect*)) AND ("diagnos* accura*" OR
sensitiv* OR specific* OR valid* OR "roc curve" OR "positive predictive value" OR "negative
predictive value")) AND (LIMIT-TO (LANGUAGE, "English")) (3,605)

Cochrane Central Register of Controlled Trials, Wiley


The search was run on 20 April 2015.
1 MeSH descriptor: [Hepatitis, Viral, Human] this term only
2 MeSH descriptor: [Hepatitis Viruses] this term only
3 MeSH descriptor: [Hepatitis Antibodies] this term only
4 MeSH descriptor: [Hepadnaviridae Infections] explode all trees
5 MeSH descriptor: [Hepatitis B Antibodies] this term only
6 MeSH descriptor: [Hepatitis B virus] this term only
7 MeSH descriptor: [Hepadnaviridae] this term only
8 MeSH descriptor: [Hepatitis B Surface Antigens] explode all trees
9 "hepatitis-b":ti,ab,kw (Word variations have been searched)
10 "hep-b":ti,ab,kw (Word variations have been searched)
11 hepatitis near/5 b:ti,ab,kw (Word variations have been searched)
12 hep near/5 b:ti,ab,kw (Word variations have been searched)
13 hbv:ti,ab,kw (Word variations have been searched)
14 hbsag:ti,ab,kw (Word variations have been searched)
15 #1 or #2 or #3 or #4 or #5 or #6 or #7 or #8 or #9 or #10 or #11 or #12 or #13 or #14
16 MeSH descriptor: [Reagent Kits, Diagnostic] explode all trees
17 (rapid or "point of care" or "near patient" or poc or poct or bedside) near/5 (test or tests
or testing or detect* or diagnos* or screen* or kit or kits or assay* or device*):ti,ab,kw
(Word variations have been searched)
18 radt or radts or rdt or rdts:ti,ab,kw (Word variations have been searched)
19 "rapid test*":ti,ab,kw (Word variations have been searched)
20 MeSH descriptor: [Enzyme-Linked Immunosorbent Assay] explode all trees
21 enzyme near/2 (immunoassay* or immuno-assay* or immunosorbent):ti,ab,kw (Word
variations have been searched)
22 (antigen* or antibod*) near/3 detect*:ti,ab,kw (Word variations have been searched)
23 MeSH descriptor: [Immunoassay] this term only
24 MeSH descriptor: [Immunoenzyme Techniques] this term only

Page | 241
25 "enzyme-linked immunosorbent assay" or ELISA:ti,ab,kw (Word variations have been
searched)
26 #16 or #17 or #18 or #19 or #20 or #21 or #22 or #23 or #24 or #25
27 MeSH descriptor: [Sensitivity and Specificity] explode all trees
28 diagnos* accura* or sensitiv* or specific* or valid*:ti,ab,kw (Word variations have been
searched)
29 "roc curve":ti,ab,kw (Word variations have been searched)
30 "positive predictive value":ti,ab,kw (Word variations have been searched)
31 "negative predictive value":ti,ab,kw (Word variations have been searched)
32 #27 or #28 or #29 or #30 or #31
33 #15 and #26 and #32
The search found 64 trials.

Literatura Latino-Americana e do Caribe em Ciências da Saúde (LILACS) (BIREME interface)


LILACS was searched on 20 April 2015
("hepatitis b" or "hep b" or "hbv" or "hbsag") and ("rapid test$" or "point of care test$" or
"near patient test$" or "poc test$" or poct or "bedside test$" or "rapid detect$" or "point of
care detect$" or "near patient detect$" or "poc detect$" or "bedside detect$" or "rapid
diagnos$" or "point of care diagnos$" or "near patient diagnos$" or "poc diagnos$" or
"bedside diagnos$" or "rapid screen$" or "point of care screen$" or "near patient screen$" or
"poc screen$" or "bedside screen$" or "rapid kit$" or "point of care kit$" or "near patient kit$"
or "poc kit$" or "bedside kit$" or "rapid assay$" or "point of care assay$" or "near patient
assay$" or "poc assay$" or "bedside assay$" or "rapid device$" or "point of care device$" or
"near patient device$" or "poc device$" or "bedside device$" or radt or radts or rdt or rdts or
"enzyme-linked immunosorbent assay" or "antigen$ detect$" or "antibod$ detect$" or elisa or
immunoassay or immunoenzyme or "immuno-assay") and ("diagnos$ accura$" or sensitiv$ or
specific$ or valid$ or "roc curve" or "positive predictive value" or "negative predictive value")
(33)

WHO Global Index Medicus

The database was searched on 22 April 2015.


SUBJECT: (("Hepatitis, Viral, Human" OR "Hepatitis Viruses" OR "Hepatitis B virus" OR
"Hepatitis Antibodies" OR "Hepadnaviridae Infections" OR "Hepatitis B Antibodies" OR
"Hepatitis B Virus" OR "Hepadnaviridae" OR "Hepatitis B Surface Antigens") AND ("Reagent
Kits, Diagnostic" OR "Enzyme-Linked Immunosorbent Assay" OR "Immunoassay" OR
"Immunoenzyme Techniques") AND ("Sensitivity and Specificity")) (478)

Page | 242
Summary data for studies assessing diagnostic accuracy against a NAT-reference standard

Table 8. Study characteristics – RDT/ EIA vs NAT

Study Location Sample size Study Setting Sample Test under evaluation Reference test
[Author, Year] [Country, City] design [Type, Brand] [Type, Brand]

Ansari, 2007 Iran, Urumieh 240 CC Hospital patients S RDT, ACON qPCR
RDT, Atlas
RDT, Blue Cross
RDT, Cortez
RDT, DIMA
RDT, Intec

Khadem-Ansari, Iran, Urumieh 350 CC – CSQ Hospital patients – S ChLIA, Liaison Rt-PCR
2014 referred as ?HBV

Lukhwareni, 2009 South Africa 192 CC HIV cohort – pre S ChLIA, Elecsys qPCR
ART

Mphahlele, 2006 South Africa 167 (HIV+) CC HIV cohort S EIA, AxSYM Nested PCR
128 (HIV–)

Nna, 2014 Nigeria 113 CS Blood donors P RDT, ACON Nested PCR;
(repeat) qPCR for positive

Olinger, 2007 Nigeria, Ibadan 200 CS Hospital patients – S MEIA, AxSYM v2 rtPCR and nested
liver disease, HIV ChLIA, Elecsys PCR
ELFA, VIDAS Ultra

Seremba, 2010 Uganda 74 (HIV-) CS – CSQ Hospital patients - S RDT, Cortez PCR
83 (HIV+) ED, including HIV EIA, ADVIA

qPCR: quantitative PCR; rtPCR: realtime PCR; ChLIA: chemiluminescent immunoassay; CMIA:
chemiluminescent microparticle enzyme immunoassay; ECLIA: electrochemiluminescent immunoassay;
EIA: enzyme immunoassay; ELFA: enzyme-linked fluorescent assay; ELISA: enzyme-linked
immunosorbent assay; MEIA: microparticle enzyme immunoassay; RDT: rapid diagnostic test; rtPCR:
real-time PCR; CC: case–control; CS: cross-sectional; CSQ: consecutive patients; LB: lab-based study

Page | 243
Table 9. Summary pooled diagnostic accuracy of HBsAg assays compared to NAT reference

Test type NAT reference

Sen (95% CI) Spec (95% CI)

RDT 93.3 (91.3–94.9) 98.1 (97.0–98.9)

EIA 75.7 (72.1–79.1) 86.1 (83.8–88.2)

*Sen: sensitivity; Spec: specificity; CI: confidence interval; RDT: rapid diagnostic test; EIA: enzyme immunoassay

Table 10. Summary pooled diagnostic accuracy of HBsAg assays

Test type HIV status NAT reference

n Sen (95% CI) Spec (95% CI)

+
RDT HIV 1 37.5 (22.7–54.2) 97.7 (87.7–99.9)


HIV 2 57.1 (41.0–72.3) 97.2 (93.1–99.2)

+
EIA HIV 3 57.9 (49.8–65.6) 95.8 (92.7–97.8)


HIV 2 83.3 (69.8–92.5) 85.7 (79.2–90.8)

Page | 244
Fig. 5. Forest plots, RDT vs NAT, ordered by [Test, Author]

Sensitivity (95% CI) Specificity (95% CI) Sensitivity (95% CI)


ACON - Ansari ACON - 0.98Ansari (0.94 - 1.00) 0.99 (0.95 - 1.00) ACON - Ansari 0.98 (0.94 - 1.00) ACON - Ansari
ACON - Nna ACON - 0.60Nna (0.36 - 0.81) 0.99 (0.94 - 1.00) ACON - Nna 0.60 (0.36 - 0.81) ACON - Nna
Atlas - Ansari 0.98 (0.93 - 0.99) 0.98 (0.93
Atlas - Ansari - 0.99) Atlas - Ansari 0.98 (0.93 - 0.99) Atlas - Ansari
Specificity (95% CI)
Blue Cross - Ansari Blue Cross 0.99 (0.95 - 1.00) 0.98 (0.94
- Ansari - 1.00) Blue Cross - Ansari 0.99 (0.95 - 1.00) Blue Cross - An
ACON --Ansari
Cortez Ansari Cortez0.99
- 0.98 (0.95
Ansari - 1.00)
(0.94 - 1.00) 0.98 (0.94 - 1.00) Cortez - Ansari 0.98 (0.94 - 1.00) Cortez - Ansari
ACON --Nna
Cortez Cortez0.99
Seremba (HIV-) - 0.55 (0.94
Seremba -(HIV-)
(0.321.00)
- 0.76) 0.94 (0.84 - 0.99) Cortez - Seremba (HIV-) 0.55 (0.32 - 0.76) Cortez - Seremb
Atlas - Ansari
Cortez Cortez0.98
- Seremba (HIV+) - 0.38 (0.93
Seremba -(HIV+)
(0.230.99)
- 0.54) 0.98 (0.88 - 1.00) Cortez - Seremba (HIV+) 0.38 (0.23 - 0.54) Cortez - Seremb
Blue Cross
DIMA - Ansari
- Ansari DIMA -0.98 0.98(0.94
Ansari - 1.00)
(0.94 - 1.00) 0.99 (0.95 - 1.00) DIMA - Ansari 0.98 (0.94 - 1.00) DIMA - Ansari
Cortez- -Ansari
Intec Ansari Intec - 0.98
0.99(0.94
Ansari - 1.00)
(0.95 - 1.00) 0.98 (0.93 - 0.99) Intec - Ansari 0.99 (0.95 - 1.00) Intec - Ansari
Cortez - Seremba (HIV-) 0.94 (0.84 - 0.99)
Cortez - Seremba (HIV+) 0.98 (0.88 - 1.00)
DIMA - Ansari 0.99 (0.95 - 1.00)
Intec - Ansari 0.98 (0.93 - 0.99)
Pooled Sensitivity = 0.93Pooled
(0.91Specificity
to 0.95) = 0.98 (0.97 to 0.99) Pooled Sensitivity = 0.93 (0.91 to 0.95) Pooled Specific
Chi-square = 173.29;Chi-square
df = 8 (p = =0.0000)
5.50; df = 8 (p = 0.7026) Chi-square = 173.29; df = 8 (p = 0.0000) Chi-square = 5.
4 Inconsistency
1 0.6 1 Inconsistency
0.8 (I-square) = 95.4 % (I-square) = 0.0 % 0 0.2 0.4 0.6 0.8 1 0Inconsistency
0.2 (I-square)
0.4 = 95.40.6 % 0.8 1 Inconsistency (I
pecificity Sensitivity Specificity
Pooled Specificity = 0.98 (0.97 to 0.99)
Chi-square = 5.50; df = 8 (p = 0.7026)
8 1 Inconsistency (I-square) = 0.0 %

Key – Types of RDT and reference standards used in studies

Study Test brand (manufacturer) Reference test type, brand (manufacturer)

Ansari ACON (Acon laboratories) QPCR

Atlas (William James House) Roto-GENE 3000 Research (Corbet real time PCR) and kit artus (Hamburg)

Blue Cross (Blue Cross Inc.)

Cortez (Cortez diagnostics)

DIMA (Geseeschaft fur Diagnostika mbH)

Intex (Intec Products Inc.)

Page | 245
Nna ACON (Acon laboratories) Nested PCR

QPCR for positive

Seremba Cortez (Cortez diagnostics) b-DNA (Versant);

PCR, Amplicor for discrepant

Page | 246
Fig. 6. Forest plots, EIA vs NAT, ordered by [Test, Author]

Sensitivity (95% CI) Specificity (95% CI) Sensitivity (95% CI)


ADVIA - Seremba (HIV-) 0.73 (0.50
ADVIA - Seremba - 0.89) 0.98 (0.90
(HIV-) - 1.00) ADVIA - Seremba (HIV-) 0.73 (0.50 - 0.89) ADVIA - Se
ADVIA - Seremba (HIV+) 0.80 (0.64
ADVIA - Seremba - 0.91) 0.98 (0.88
(HIV+) - 1.00) ADVIA - Seremba (HIV+) 0.80 (0.64 - 0.91) ADVIA - Se
AxSym - Mphahlele (HIV-) 0.92
AxSym -Specificity(0.75(HIV-)
Mphahlele - 0.99)
(95% CI) 0.79 (0.70 - 0.87) AxSym - Mphahlele (HIV-) 0.92 (0.75 - 0.99) AxSym - M
AxSym - Mphahlele (HIV+)
AxSym - 0.38 (0.25(HIV+)
Mphahlele - 0.53) 0.93 (0.87 - 0.97) AxSym - Mphahlele (HIV+) 0.38 (0.25 - 0.53) AxSym - M
ADVIA v2
AxSym - Seremba
- Olinger(HIV-)
AxSym v2 0.98
0.77 (0.90
(0.67
- Olinger - 1.00)0.75 (0.66
- 0.86) - 0.83) AxSym v2 - Olinger 0.77 (0.67 - 0.86) AxSym v2
ADVIA - Seremba (HIV+)
Elecsys - LukhwareniElecsys 0.98
(HIV+) -0.59 (0.88 - 1.00)0.98 (0.93
(0.47 - (HIV+)
Lukhwareni 0.71) - 0.99) Elecsys - Lukhwareni (HIV+) 0.59 (0.47 - 0.71) Elecsys - L
AxSym -- Mphahlele
Elecsys Olinger (HIV-) -0.67
Elecsys 0.79 (0.56
Olinger (0.70 - 0.87)0.94 (0.88
- 0.77) - 0.98) Elecsys - Olinger 0.67 (0.56 - 0.77) Elecsys - O
AxSym -- Khadem-Ansari
Liaison MphahleleLiaison
(HIV+) - 1.00
0.93 (0.98
(0.87
Khadem-Ansari - 0.97)0.70 (0.63
- 1.00) - 0.76) Liaison - Khadem-Ansari 1.00 (0.98 - 1.00) Liaison - K
AxSymULTRA
VIDAS v2 - Olinger
- Olinger 0.75 -(0.58
0.69
VIDAS ULTRA (0.66 - 0.83)0.94 (0.88
- 0.79)
Olinger - 0.98) VIDAS ULTRA - Olinger 0.69 (0.58 - 0.79) VIDAS ULT
Elecsys - Lukhwareni (HIV+) 0.98 (0.93 - 0.99)
Elecsys - Olinger 0.94 (0.88 - 0.98)
Liaison - Khadem-Ansari 0.70 (0.63 - 0.76)
VIDAS ULTRA - Olinger 0.94 (0.88 - 0.98)
Pooled Sensitivity = Pooled
0.76 (0.72 to 0.79) = 0.86 (0.84 to 0.88)
Specificity Pooled Sensitivity = 0.76 (0.72 to 0.79) Pooled Sp
Chi-square = 135.22;Chi-square
df = 8 (p ==0.0000)
104.27; df = 8 (p = 0.0000) Chi-square = 135.22; df = 8 (p = 0.0000) Chi-square
Inconsistency
10.6 0.8 (I-square) = 94.1 % (I-square) = 92.3 %
1 Inconsistency 0 0.2 0.4 0.6 0.8 1 0Inconsistency
0.2 (I-square)
0.4 = 94.10.6 % 0.8 1 Inconsiste
ecificity Sensitivity Specificity
Pooled Specificity = 0.86 (0.84 to 0.88)
Chi-square = 104.27; df = 8 (p = 0.0000)
1 Inconsistency (I-square) = 92.3 %

Key – Types of EIA and reference standards used in studies


Study Test type, brand (manufacturer) Reference test type, brand (manufacturer)

Seremba EIA, ADVIA Centaur (Siemens) b-DNA (Versant);


PCR, Amplicor for discrepant

Mphahlele EIA, AxSym (Abbott) Nested PCR (in house);


positive quantified with COBAS Amplicor ()

Olinger MEIA, AxSym v2 (Abbott) RT-PCR


ELFA, VIDAS Ultra (Biomérieux)
ChLIA, Elecsys (Roche)

Lukhwareni ChLIA, Elecsys (Roche) Nested PCR, High Pure Viral Nucleic Acid assay (Roche)

Page | 247
Q-PCR, COBAS TaqMan HBV Test 48 assay ()

Khadem- Ansari ChLIA, Liasison (Diasorin) RT-PCR, Robogene (Corbett)

Page | 248
Summary Receiver Operating Characteristic (SROC) curves

Fig. 7. SROC curves for studies comparing RDTs with EIAs

Sensitivity SROC Curve


1

Symmetric SROC
0.9 AUC = 0.9944
SE(AUC) = 0.0025
Q* = 0.9704
SE(Q*) = 0.0078
0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.2 0.4 0.6 0.8 1
1-specificity

Page | 249
Fig. 8. SROC curves for studies comparing RDTs with NATs

Sensitivity SROC Curve


1

Symmetric SROC
0.9 AUC = 0.9974
SE(AUC) = 0.0012
Q* = 0.9816
SE(Q*) = 0.0053
0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.2 0.4 0.6 0.8 1
1-specificity

Page | 250
Fig. 9. SROC curves for studies comparing EIAs with EIAs

Sensitivity SROC Curve


1

Symmetric SROC
0.9 AUC = 0.9953
SE(AUC) = 0.0037
Q* = 0.9735
SE(Q*) = 0.0124
0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.2 0.4 0.6 0.8 1
1-specificity

Page | 251
Fig. 10. SROC curves for studies comparing EIAs with NAT

Sensitivity SROC Curve


1

Symmetric SROC
0.9 AUC = 0.9379
SE(AUC) = 0.0217
Q* = 0.8748
SE(Q*) = 0.0271
0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.2 0.4 0.6 0.8 1
1-specificity

Forest plots of sub-analyses

Forest plots, analysed by HIV status

Fig. 11. Forest plots of RDTs vs EIAs in HIV-positive patients


Sensitivity (95% CI)
Determine - Davies 1.00 (0.86 - 1.00) Specificity (95% CI)
Determine - Franzeck 0.96 (0.80 - 1.00)
Determine - Geretti 0.69 (0.61 - 0.77) Determine - Davies 1.00 (0.93 - 1.00)
Determine - Hoffman 0.75 (0.59 - 0.87) Determine - Franzeck 1.00 (0.99 - 1.00)
VEDA LAB - Honge 0.62 (0.51 - 0.73) Determine - Geretti 1.00 (0.99 - 1.00)
VIKIA - Geretti 0.71 (0.62 - 0.78) Determine - Hoffman 1.00 (0.99 - 1.00)
VEDA LAB - Honge 0.99 (0.98 - 1.00)
VIKIA - Geretti 1.00 (0.99 - 1.00)

Pooled Sensitivity = 0.72 (0.68 to 0.76)


Chi-square = 30.49; df = 5 (p = 0.0000)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 83.6 % Pooled Specificity = 1.00 (1.00 to 1.00)
Sensitivity Chi-square = 12.45; df = 5 (p = 0.0291)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 59.8 %
Specificity

Fig. 13. Forest plots, RDTs vs EIAs in HIV-negative patients


Sensitivity (95% CI) Specificity (95% CI)
Determine - Njai (CHB) 0.95 (0.90 - 0.98) Determine - Njai (CHB) 0.93 (0.78 - 0.99)
Determine - Njai (Screen) 0.88 (0.81 - 0.94) Determine - Njai (Screen) 1.00 (0.99 - 1.00)
Espline - Njai (CHB) 0.94 (0.89 - 0.97) Espline - Njai (CHB) 0.95 (0.82 - 0.99)
VIKIA - Njai (Screen) 0.90 (0.79 - 0.96) VIKIA - Njai (Screen) 1.00 (0.99 - 1.00)

Pooled Sensitivity = 0.93 (0.90 to 0.95) Pooled Specificity = 1.00 (0.99 to 1.00)
Chi-square = 4.74; df = 3 (p = 0.1916) Chi-square = 19.96; df = 3 (p = 0.0002)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square)
0 = 36.8
0.2% 0.4 0.6 0.8 1 Inconsistency (I-square) = 85.0 %
Sensitivity Specificity

Page | 252
Fig. 14. Forest plots, EIAs vs EIAs in HIV-positive patients
Specificity (95% CI)
Architect - Geretti 1.00 (0.99 - 1.00)
Liaison Ultra - Geretti 0.99 (0.99 - 1.00)
Murex v3.0 - Geretti 0.99 (0.98 - 1.00)

Pooled Specificity = 0.99 (0.99 to 1.00)


Chi-square = 0.51; df = 2 (p = 0.7756)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 0.0 %
Specificity

Sensitivity (95% CI)


Architect - Geretti 0.98 (0.94 - 1.00)
Liaison Ultra - Geretti 0.97 (0.93 - 0.99)
Murex v3.0 - Geretti 0.99 (0.95 - 1.00)

Pooled Sensitivity = 0.98 (0.96 to 0.99)


Chi-square = 0.69; df = 2 (p = 0.7067)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 0.0 %
Sensitivity

Fig. 15. Forest plots, EIAs vs NAT in HIV-positive patients


Sensitivity (95% CI)
ADVIA - Seremba 0.80 (0.64 - 0.91)
AxSym - Mphahlele 0.38 (0.25 - 0.53)
Elecsys - Lukhwareni 0.59 (0.47 - 0.71)

Pooled Sensitivity = 0.58 (0.50 to 0.66)


Chi-square = 16.84; df = 2 (p = 0.0002)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 88.1 %
Sensitivity

Specificity (95% CI)


ADVIA - Seremba 0.98 (0.88 - 1.00)
AxSym - Mphahlele 0.93 (0.87 - 0.97)
Elecsys - Lukhwareni 0.98 (0.93 - 0.99)

Pooled Specificity = 0.96 (0.93 to 0.98)


Chi-square = 3.27; df = 2 (p = 0.1953)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 38.8 %
Specificity

Fig. 16. Forest plots, EIAs vs NAT in HIV-negative patients


Sensitivity (95% CI)
AxSym - Mphahlele (HIV-) 0.92 (0.75 - 0.99)
ADVIA - Seremba (HIV-) 0.73 (0.50 - 0.89)

Pooled Sensitivity = 0.83 (0.70 to 0.93)


Chi-square = 3.37; df = 1 (p = 0.0664)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 70.3 %
Sensitivity

Specificity (95% CI)


AxSym - Mphahlele (HIV-) 0.79 (0.70 - 0.87)
ADVIA - Seremba (HIV-) 0.98 (0.90 - 1.00)

Pooled Specificity = 0.86 (0.79 to 0.91)


Chi-square = 12.71; df = 1 (p = 0.0004)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 92.1 %
Specificity

Page | 253
Fig. 17. Forest plots, RDTs vs NAT in HIV-negative patients
Sensitivity (95% CI)
ACON - Nna 0.60 (0.36 - 0.81)
Cortez - Seremba (HIV-) 0.55 (0.32 - 0.76)

Pooled Sensitivity = 0.57 (0.41 to 0.72)


Chi-square = 0.13; df = 1 (p = 0.7211)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 0.0 %
Sensitivity

Specificity (95% CI)


ACON - Nna 0.99 (0.94 - 1.00)
Cortez - Seremba (HIV-) 0.94 (0.84 - 0.99)

Pooled Specificity = 0.97 (0.93 to 0.99)


Chi-square = 2.62; df = 1 (p = 0.1056)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 61.8 %
Specificity

Forest plots, analysed by sample type

Fig. 18. Forest plots, RDTs vs EIA in whole blood


Sensitivity (95% CI)
Binax - Akanmu 1.00 (0.72 - 1.00)
Binax - Lau 0.96 (0.89 - 0.99)
Determine - Njai 0.88 (0.81 - 0.94)
GWHB - GWHB 0.95 (0.76 - 1.00)
VIKIA - Njai 0.90 (0.79 - 0.96)
Determine - Hoffman 0.75 (0.59 - 0.87)
Determine - Bottero 0.94 (0.82 - 0.99)
Determine - Franzeck 0.96 (0.80 - 1.00)
Determine - Lien 1.00 (0.79 - 1.00)
QUICK PROFILE - Bottero 0.90 (0.82 - 0.96)
VIKIA - Bottero 0.96 (0.90 - 0.99)

Pooled Sensitivity = 0.92 (0.89 to 0.94)


Chi-square = 23.13; df = 10 (p = 0.0103)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 56.8 %
Sensitivity

Specificity (95% CI)


Binax - Akanmu 0.99 (0.96 - 1.00)
Binax - Lau 1.00 (0.99 - 1.00)
Determine - Njai 1.00 (0.99 - 1.00)
GWHB - GWHB 1.00 (0.40 - 1.00)
VIKIA - Njai 1.00 (0.99 - 1.00)
Determine - Hoffman 1.00 (0.99 - 1.00)
Determine - Bottero 1.00 (1.00 - 1.00)
Determine - Franzeck 1.00 (0.99 - 1.00)
Determine - Lien 1.00 (0.97 - 1.00)
QUICK PROFILE - Bottero 1.00 (1.00 - 1.00)
VIKIA - Bottero 1.00 (1.00 - 1.00)

Pooled Specificity = 1.00 (1.00 to 1.00)


Chi-square = 22.85; df = 10 (p = 0.0113)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 56.2 %
Specificity

Page | 254
Forest plots, analysed by study design

Fig. 19. Forest plots, RDTs vs EIA in case–control studies


Sensitivity (95% CI)
Abraham - Virucheck 0.90 (0.68 - 0.99)
Chevaliez 1.00 (0.98 - 1.00)
Clement - Binax 1.00 (0.99 - 1.00)
Erhabor - ACON 1.00 (0.88 - 1.00)
Khan - Accurate 0.50 (0.33 - 0.67)
Khan - Onecheck 0.53 (0.36 - 0.69)
Lien - Dainascreen 1.00 (0.97 - 1.00)
Lien - Determine 1.00 (0.97 - 1.00)
Lien - Serodia 0.96 (0.90 - 0.99)
Lin - Determine 0.99 (0.96 - 1.00)
Lin - Determine 0.94 (0.90 - 0.97)
Lin - DRW 0.99 (0.97 - 1.00)
Lin - DRW 0.97 (0.93 - 0.99)
Oh - Genedia 0.98 (0.94 - 1.00)
Oh - Serodia 0.96 (0.91 - 0.99)
Randrianirina - Cypress 0.97 (0.91 - 0.99)
Randrianirina - Determine 0.98 (0.92 - 1.00)
Randrianirina - Hexagon 0.96 (0.89 - 0.99)
Randrianirina - Virucheck 0.96 (0.89 - 0.99)
Sato - Dainascreen 1.00 (0.98 - 1.00)
Sato - Serodia 0.96 (0.91 - 0.98)

Pooled Sensitivity = 0.97 (0.96 to 0.97)


Chi-square = 230.85; df = 20 (p = 0.0000)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 91.3 %
Sensitivity

Specificity (95% CI)


Abraham - Virucheck 1.00 (0.88 - 1.00)
Chevaliez 0.99 (0.97 - 1.00)
Clement - Binax 0.99 (0.98 - 1.00)
Erhabor - ACON 0.91 (0.84 - 0.96)
Khan - Accurate 0.95 (0.74 - 1.00)
Khan - Onecheck 1.00 (0.82 - 1.00)
Lien - Dainascreen 1.00 (0.98 - 1.00)
Lien - Determine 1.00 (0.98 - 1.00)
Lien - Serodia 1.00 (0.98 - 1.00)
Lin - Determine 1.00 (0.99 - 1.00)
Lin - Determine 1.00 (0.99 - 1.00)
Lin - DRW 0.99 (0.98 - 1.00)
Lin - DRW 1.00 (0.99 - 1.00)
Oh - Genedia 1.00 (0.96 - 1.00)
Oh - Serodia 1.00 (0.96 - 1.00)
Randrianirina - Cypress 0.96 (0.91 - 0.99)
Randrianirina - Determine 1.00 (0.97 - 1.00)
Randrianirina - Hexagon 0.96 (0.91 - 0.99)
Randrianirina - Virucheck 0.98 (0.94 - 1.00)
Sato - Dainascreen 1.00 (0.99 - 1.00)
Sato - Serodia 1.00 (0.98 - 1.00)

Pooled Specificity = 0.99 (0.99 to 1.00)


Chi-square = 85.62; df = 20 (p = 0.0000)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 76.6 %
Specificity

9.3.4 Forest plots, analysed by study setting

Page | 255
Fig. 20. Forest plots, RDTs vs EIA in blood donors
Sensitivity (95% CI)
Abon - Mutocheluh 0.50 (0.28 - 0.72)
ACON - Bjoerkvoll (Camb) 0.93 (0.86 - 0.98)
ACON - Bjoerkvoll (Viet) 0.82 (0.74 - 0.88)
ACON - Erhabor 1.00 (0.88 - 1.00)
Acull-Tell - Mutocheluh 0.55 (0.32 - 0.76)
AMRAD - Ola 0.95 (0.76 - 1.00)
Binax - Akanmu 0.94 (0.87 - 0.97)
Biotec - Ola 0.59 (0.41 - 0.75)
Core TM - Mutocheluh 0.50 (0.28 - 0.72)
Determine - Lin (China) 0.99 (0.96 - 1.00)
Determine - Lin (Guinea) 0.94 (0.90 - 0.97)
Dipstick (PATH) - Mvere 0.93 (0.68 - 1.00)
DRW - Lin (China) 0.99 (0.97 - 1.00)
DRW - Lin (Guinea) 0.97 (0.93 - 0.99)
Genedia - Oh 0.98 (0.94 - 1.00)
Rapid care - Mutocheluh 0.55 (0.32 - 0.76)
Serodia - Oh 0.96 (0.91 - 0.99)
SimpliRed - Mvere 0.93 (0.68 - 1.00)
Wondfo - Mutocheluh 0.59 (0.36 - 0.79)

Pooled Sensitivity = 0.92 (0.90 to 0.93)


Chi-square = 229.22; df = 18 (p = 0.0000)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 92.1 %
Sensitivity

Specificity (95% CI)


Abon - Mutocheluh 0.99 (0.96 - 1.00)
ACON - Bjoerkvoll (Camb) 1.00 (0.99 - 1.00)
ACON - Bjoerkvoll (Viet) 1.00 (0.99 - 1.00)
ACON - Erhabor 0.91 (0.84 - 0.96)
Acull-Tell - Mutocheluh 0.99 (0.96 - 1.00)
AMRAD - Ola 1.00 (0.40 - 1.00)
Binax - Akanmu 1.00 (0.03 - 1.00)
Biotec - Ola 0.86 (0.64 - 0.97)
Core TM - Mutocheluh 0.98 (0.94 - 1.00)
Determine - Lin (China) 1.00 (0.99 - 1.00)
Determine - Lin (Guinea) 1.00 (0.99 - 1.00)
Dipstick (PATH) - Mvere 1.00 (0.98 - 1.00)
DRW - Lin (China) 0.99 (0.98 - 1.00)
DRW - Lin (Guinea) 1.00 (0.99 - 1.00)
Genedia - Oh 1.00 (0.96 - 1.00)
Rapid care - Mutocheluh 0.99 (0.96 - 1.00)
Serodia - Oh 1.00 (0.96 - 1.00)
SimpliRed - Mvere 1.00 (0.98 - 1.00)
Wondfo - Mutocheluh 0.99 (0.96 - 1.00)

Pooled Specificity = 1.00 (0.99 to 1.00)


Chi-square = 78.41; df = 18 (p = 0.0000)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 77.0 %
Specificity

Page | 256
Forest plots, analysed by study year

Fig. 21. Forest plots, RDTs vs EIA for studies after 2005

Sensitivity (95% CI)


Abon - Mutocheluh 0.50 (0.28 - 0.72)
Accurate - Khan 0.50 (0.33 - 0.67)
ACON - Bjoerkvoll (Camb) 0.93 (0.86 - 0.98)
ACON - Bjoerkvoll (Viet) 0.82 (0.74 - 0.88)
ACON - Erhabor 1.00 (0.88 - 1.00)
Acull-Tell - Mutocheluh 0.55 (0.32 - 0.76)
AMRAD - Ola 0.95 (0.76 - 1.00)
Binax - Akanmu (BD) 0.94 (0.87 - 0.97)
Binax - Akanmu (CLD) 1.00 (0.40 - 1.00)
Biotec - Ola 0.59 (0.41 - 0.75)
Core TM - Mutocheluh 0.50 (0.28 - 0.72)
Cortez - Chameera 0.60 (0.15 - 0.95)
Cypress - Randrianirina 0.97 (0.91 - 0.99)
Determine - Bottero 0.94 (0.82 - 0.99)
Determine - Davies (H+) 1.00 (0.86 - 1.00)
Determine - Franzeck (H+) 0.96 (0.80 - 1.00)
Determine - Geretti (H+) 0.69 (0.61 - 0.77)
Determine - Hoffman (H+) 0.75 (0.59 - 0.87)
Determine - Lin (China) 0.99 (0.96 - 1.00)
Determine - Lin (Guinea) 0.94 (0.90 - 0.97)
Determine - Njai (CHB) 0.95 (0.90 - 0.98)
Determine - Njai (Screen) 0.88 (0.81 - 0.94)
Determine - Nyirendra 0.56 (0.38 - 0.73)
Determine - Randrianirina 0.98 (0.92 - 1.00)
DRW - Lin (China) 0.99 (0.97 - 1.00)
DRW - Lin (Guinea) 0.97 (0.93 - 0.99)
DRW v2.0 - Chevaliez (Hep) 1.00 (0.98 - 1.00)
DRW v2.0 - Chevaliez (Preg) 0.96 (0.81 - 1.00)
DRW v2.0 - Chevaliez CC 0.95 (0.76 - 1.00)
Espline - Njai (CHB) 0.94 (0.89 - 0.97)
Hexagon - Randrianirina 0.96 (0.89 - 0.99)
Intec - Liu 0.51 (0.43 - 0.58)
Nanosign - Gish (H+) 0.74 (0.49 - 0.91)
Onecheck - Khan 0.53 (0.36 - 0.69)
Onsite - Chameera 0.80 (0.28 - 0.99)
QUICK PROFILE - Bottero 0.90 (0.82 - 0.96)
Rapid care - Mutocheluh 0.55 (0.32 - 0.76)
SD Bioline - Upretti 1.00 (0.63 - 1.00)
VEDA LAB - Honge (H+) 0.62 (0.51 - 0.73)
VIKIA - Bottero 0.96 (0.90 - 0.99)
VIKIA - Geretti (H+) 0.71 (0.62 - 0.78)
VIKIA - Njai (Screen) 0.90 (0.79 - 0.96)
Virucheck - Randrianirina 0.96 (0.89 - 0.99)
Wondfo - Mutocheluh 0.59 (0.36 - 0.79)

Pooled Sensitivity = 0.86 (0.85 to 0.88)


Chi-square = 680.76; df = 43 (p = 0.0000)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 93.7 %
Sensitivity

Page | 257
Specificity (95% C
Abon - Mutocheluh 0.99 (0.96 - 1.0
Accurate - Khan 0.95 (0.74 - 1.0
ACON - Bjoerkvoll (Camb) 1.00 (0.99 - 1.0
ACON - Bjoerkvoll (Viet) 1.00 (0.99 - 1.0
ACON - Erhabor 0.91 (0.84 - 0.9
Acull-Tell - Mutocheluh 0.99 (0.96 - 1.0
AMRAD - Ola 1.00 (0.40 - 1.0
Binax - Akanmu (BD) 1.00 (0.03 - 1.0
Binax - Akanmu (CLD) 1.00 (0.89 - 1.0
Biotec - Ola 0.86 (0.64 - 0.9
Core TM - Mutocheluh 0.98 (0.94 - 1.0
Cortez - Chameera 1.00 (0.92 - 1.0
Cypress - Randrianirina 0.96 (0.91 - 0.9
Determine - Bottero 1.00 (1.00 - 1.0
Determine - Davies (H+) 1.00 (0.93 - 1.0
Determine - Franzeck (H+) 1.00 (0.99 - 1.0
Determine - Geretti (H+) 1.00 (0.99 - 1.0
Determine - Hoffman (H+) 1.00 (0.99 - 1.0
Determine - Lin (China) 1.00 (0.99 - 1.0
Determine - Lin (Guinea) 1.00 (0.99 - 1.0
Determine - Njai (CHB) 0.93 (0.78 - 0.9
Determine - Njai (Screen) 1.00 (0.99 - 1.0
Determine - Nyirendra 0.69 (0.62 - 0.7
Determine - Randrianirina 1.00 (0.97 - 1.0
DRW - Lin (China) 0.99 (0.98 - 1.0
DRW - Lin (Guinea) 1.00 (0.99 - 1.0
DRW v2.0 - Chevaliez (Hep) 0.99 (0.97 - 1.0
DRW v2.0 - Chevaliez (Preg) 0.99 (0.97 - 1.0
DRW v2.0 - Chevaliez CC 0.98 (0.97 - 0.9
Espline - Njai (CHB) 0.95 (0.82 - 0.9
Hexagon - Randrianirina 0.96 (0.91 - 0.9
Intec - Liu 1.00 (0.95 - 1.0
Nanosign - Gish (H+) 0.98 (0.95 - 0.9
Onecheck - Khan 1.00 (0.82 - 1.0
Onsite - Chameera 1.00 (0.92 - 1.0
QUICK PROFILE - Bottero 1.00 (1.00 - 1.0
Rapid care - Mutocheluh 0.99 (0.96 - 1.0
SD Bioline - Upretti 1.00 (0.99 - 1.0
VEDA LAB - Honge (H+) 0.99 (0.98 - 1.0
VIKIA - Bottero 1.00 (1.00 - 1.0
VIKIA - Geretti (H+) 1.00 (0.99 - 1.0
VIKIA - Njai (Screen) 1.00 (0.99 - 1.0
Virucheck - Randrianirina 0.98 (0.94 - 1.0
Wondfo - Mutocheluh 0.99 (0.96 - 1.0

Pooled Specificity = 0.99 (0.99 to 0.99)


Chi-square = 548.27; df = 43 (p = 0.0000)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 92.2 %
Specificity
Page | 258
Positive LR (95% CI)
Abon - Mutocheluh 64.00 (8.69 - 471.26)
Accurate - Khan 9.50 (1.37 - 65.71)
ACON - Bjoerkvoll (Camb) 517.87 (129.54 - 2,070.26)
ACON - Bjoerkvoll (Viet) 434.51 (108.57 - 1,739.03)
ACON - Erhabor 10.46 (5.70 - 19.19)
Acull-Tell - Mutocheluh 69.82 (9.55 - 510.36)
AMRAD - Ola 9.32 (0.67 - 129.54)
Binax - Akanmu (BD) 3.72 (0.34 - 41.10)
Binax - Akanmu (CLD) 59.40 (3.74 - 944.36)
Biotec - Ola 4.12 (1.39 - 12.18)
Core TM - Mutocheluh 32.00 (7.60 - 134.68)
Cortez - Chameera 53.67 (3.14 - 917.12)
Cypress - Randrianirina 26.35 (10.06 - 69.00)
Determine - Bottero 4,498.21 (281.12 - 71,976.8
Determine - Davies (H+) 100.04 (6.34 - 1,578.57)
Determine - Franzeck (H+) 467.38 (29.27 - 7,463.70)
Determine - Geretti (H+) 966.70 (60.39 - 15,473.86)
Determine - Hoffman (H+) 174.94 (64.74 - 472.74)
Determine - Lin (China) 959.01 (60.07 - 15,311.15)
Determine - Lin (Guinea) 753.59 (47.21 - 12,030.12)
Determine - Njai (CHB) 14.29 (3.74 - 54.54)
Determine - Njai (Screen) 1,180.48 (73.85 - 18,870.34
Determine - Nyirendra 1.82 (1.25 - 2.67)
Determine - Randrianirina 214.02 (13.47 - 3,400.91)
DRW - Lin (China) 120.60 (45.44 - 320.04)
DRW - Lin (Guinea) 771.27 (48.32 - 12,311.21)
DRW v2.0 - Chevaliez (Hep) 74.50 (29.76 - 186.52)
DRW v2.0 - Chevaliez (Preg) 73.38 (30.62 - 175.83)
DRW v2.0 - Chevaliez CC 43.75 (27.08 - 70.69)
Espline - Njai (CHB) 17.85 (4.63 - 68.81)
Hexagon - Randrianirina 26.05 (9.95 - 68.23)
Intec - Liu 73.20 (4.61 - 1,163.40)
Nanosign - Gish (H+) 34.14 (14.80 - 78.75)
Onecheck - Khan 21.03 (1.34 - 329.93)
Onsite - Chameera 69.00 (4.22 - 1,129.19)
QUICK PROFILE - Bottero 347.25 (186.27 - 647.36)
Rapid care - Mutocheluh 69.82 (9.55 - 510.36)
SD Bioline - Upretti 642.22 (40.07 - 10,293.22)
VEDA LAB - Honge (H+) 75.01 (23.99 - 234.59)
VIKIA - Bottero 1,853.68 (463.49 - 7,413.61
VIKIA - Geretti (H+) 986.53 (61.64 - 15,789.02)
VIKIA - Njai (Screen) 374.40 (52.77 - 2,656.47)
Virucheck - Randrianirina 52.10 (13.19 - 205.84)
Wondfo - Mutocheluh 75.64 (10.41 - 549.47)

Random Effects Model


Pooled Positive LR = 84.66 (43.55 to 164.55)
Cochran-Q = 534.07; df = 43 (p = 0.0000)
0.01 1 100.0 Inconsistency (I-square) = 91.9 %
Positive LR Tau-squared = 4.0986

Page | 259
Negative LR (95% CI)
Abon - Mutocheluh 0.50 (0.33 - 0.77)
Accurate - Khan 0.53 (0.38 - 0.74)
ACON - Bjoerkvoll (Camb) 0.07 (0.03 - 0.14)
ACON - Bjoerkvoll (Viet) 0.18 (0.13 - 0.26)
ACON - Erhabor 0.02 (0.00 - 0.28)
Acull-Tell - Mutocheluh 0.46 (0.29 - 0.72)
AMRAD - Ola 0.08 (0.02 - 0.36)
Binax - Akanmu (BD) 0.09 (0.03 - 0.26)
Binax - Akanmu (CLD) 0.10 (0.01 - 1.41)
Biotec - Ola 0.48 (0.31 - 0.74)
Core TM - Mutocheluh 0.51 (0.33 - 0.77)
Cortez - Chameera 0.42 (0.16 - 1.09)
Cypress - Randrianirina 0.03 (0.01 - 0.10)
Determine - Bottero 0.07 (0.03 - 0.20)
Determine - Davies (H+) 0.02 (0.00 - 0.30)
Determine - Franzeck (H+) 0.06 (0.01 - 0.27)
Determine - Geretti (H+) 0.31 (0.24 - 0.40)
Determine - Hoffman (H+) 0.25 (0.15 - 0.43)
Determine - Lin (China) 0.01 (0.00 - 0.05)
Determine - Lin (Guinea) 0.06 (0.03 - 0.10)
Determine - Njai (CHB) 0.05 (0.02 - 0.11)
Determine - Njai (Screen) 0.12 (0.07 - 0.20)
Determine - Nyirendra 0.64 (0.43 - 0.94)
Determine - Randrianirina 0.03 (0.01 - 0.09)
DRW - Lin (China) 0.01 (0.00 - 0.04)
DRW - Lin (Guinea) 0.04 (0.02 - 0.08)
DRW v2.0 - Chevaliez (Hep) 0.00 (0.00 - 0.04)
DRW v2.0 - Chevaliez (Preg) 0.04 (0.01 - 0.26)
DRW v2.0 - Chevaliez CC 0.05 (0.01 - 0.33)
Espline - Njai (CHB) 0.06 (0.03 - 0.12)
Hexagon - Randrianirina 0.05 (0.02 - 0.12)
Intec - Liu 0.50 (0.43 - 0.58)
Nanosign - Gish (H+) 0.27 (0.13 - 0.57)
Onecheck - Khan 0.49 (0.35 - 0.68)
Onsite - Chameera 0.25 (0.06 - 1.01)
QUICK PROFILE - Bottero 0.10 (0.05 - 0.18)
Rapid care - Mutocheluh 0.46 (0.29 - 0.72)
SD Bioline - Upretti 0.06 (0.00 - 0.82)
VEDA LAB - Honge (H+) 0.38 (0.28 - 0.51)
VIKIA - Bottero 0.04 (0.01 - 0.11)
VIKIA - Geretti (H+) 0.29 (0.23 - 0.38)
VIKIA - Njai (Screen) 0.10 (0.05 - 0.21)
Virucheck - Randrianirina 0.04 (0.02 - 0.12)
Wondfo - Mutocheluh 0.41 (0.25 - 0.68)

Random Effects Model


Pooled Negative LR = 0.13 (0.09 to 0.18)
Cochran-Q = 866.48; df = 43 (p = 0.0000)
0.01 1 100.0 Inconsistency (I-square) = 95.0 %
Negative LR Tau-squared = 1.2712

Page | 260
Fig. 22. Forest plots, RDTs vs EIA for studies before 2005

Sensitivity (95% CI)


Binax - Clement 1.00 (0.99 - 1.00)
Binax - Lau (Fresh S) 0.94 (0.79 - 0.99)
Binax - Lau (Frozen S) 0.95 (0.86 - 0.99)
Binax - Lau (WB) 0.96 (0.89 - 0.99)
Dainascreen - Lien 1.00 (0.97 - 1.00)
Dainascreen - Sato 1.00 (0.98 - 1.00)
Determine - Lien 1.00 (0.97 - 1.00)
Dipstick (PATH) - Mvere 0.93 (0.68 - 1.00)
Genedia - Oh 0.98 (0.94 - 1.00)
Hepacard - Kaur 0.93 (0.84 - 0.98)
Hepacard - Raj 0.83 (0.61 - 0.95)
QuickChaser - Abraham CC 0.90 (0.68 - 0.99)
QuickChaser - Abraham CS 0.77 (0.55 - 0.92)
Serodia - Lien 0.96 (0.90 - 0.99)
Serodia - Oh 0.96 (0.91 - 0.99)
Serodia - Sato 0.96 (0.91 - 0.98)
SimpliRed - Mvere 0.93 (0.68 - 1.00)
Virucheck - Abraham CC 0.90 (0.68 - 0.99)
Virucheck - Abraham CS 0.79 (0.54 - 0.94)

Pooled Sensitivity = 0.97 (0.96 to 0.98)


Chi-square = 83.02; df = 18 (p = 0.0000)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 78.3 %
Sensitivity

Page | 261
Specificity (95% CI)
Binax - Clement 1.00 (0.99 - 1.00)
Binax - Lau (Fresh S) 1.00 (1.00 - 1.00)
Binax - Lau (Frozen S) 1.00 (1.00 - 1.00)
Binax - Lau (WB) 1.00 (0.99 - 1.00)
Dainascreen - Lien 1.00 (0.98 - 1.00)
Dainascreen - Sato 1.00 (0.99 - 1.00)
Determine - Lien 1.00 (0.98 - 1.00)
Dipstick (PATH) - Mvere 1.00 (0.98 - 1.00)
Genedia - Oh 1.00 (0.96 - 1.00)
Hepacard - Kaur 1.00 (1.00 - 1.00)
Hepacard - Raj 0.99 (0.98 - 0.99)
QuickChaser - Abraham CC 1.00 (0.88 - 1.00)
QuickChaser - Abraham CS 0.99 (0.98 - 1.00)
Serodia - Lien 1.00 (0.98 - 1.00)
Serodia - Oh 1.00 (0.96 - 1.00)
Serodia - Sato 1.00 (0.98 - 1.00)
SimpliRed - Mvere 1.00 (0.98 - 1.00)
Virucheck - Abraham CC 1.00 (0.88 - 1.00)
Virucheck - Abraham CS 0.97 (0.95 - 0.98)

Pooled Specificity = 1.00 (1.00 to 1.00)


Chi-square = 88.17; df = 18 (p = 0.0000)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 79.6 %
Specificity

Positive LR (95% CI)


Binax - Clement 268.83 (67.41 - 1,072.
Binax - Lau (Fresh S) 1,811.52 (113.19 - 28,
Binax - Lau (Frozen S) 1,448.64 (90.62 - 23,1
Binax - Lau (WB) 1,051.25 (65.81 - 16,7
Dainascreen - Lien 422.20 (26.49 - 6,728.
Dainascreen - Sato 608.08 (38.12 - 9,699.
Determine - Lien 422.20 (26.49 - 6,728.
Dipstick (PATH) - Mvere 348.00 (21.75 - 5,568.
Genedia - Oh 197.32 (12.43 - 3,133.
Hepacard - Kaur 4,996.94 (312.36 - 79,
Hepacard - Raj 73.30 (39.56 - 135.82)
QuickChaser - Abraham CC 54.62 (3.48 - 857.69)
QuickChaser - Abraham CS 146.05 (35.99 - 592.65
Serodia - Lien 404.24 (25.36 - 6,443.
Serodia - Oh 193.30 (12.17 - 3,069.
Serodia - Sato 290.53 (41.05 - 2,056.
SimpliRed - Mvere 348.00 (21.75 - 5,568.
Virucheck - Abraham CC 54.62 (3.48 - 857.69)
Virucheck - Abraham CS 25.07 (13.71 - 45.82)

Random Effects Model


Pooled Positive LR = 265.53 (106.10 to 664.50)
Cochran-Q = 88.07; df = 18 (p = 0.0000)
0.01 1 100.0 Inconsistency (I-square) = 79.6 %
Positive LR Tau-squared = 2.7236

Page | 262
Negative LR (95% CI)
Binax - Clement 0.00 (0.00 - 0.02)
Binax - Lau (Fresh S) 0.08 (0.02 - 0.25)
Binax - Lau (Frozen S) 0.06 (0.02 - 0.16)
Binax - Lau (WB) 0.05 (0.02 - 0.13)
Dainascreen - Lien 0.00 (0.00 - 0.07)
Dainascreen - Sato 0.00 (0.00 - 0.05)
Determine - Lien 0.00 (0.00 - 0.07)
Dipstick (PATH) - Mvere 0.09 (0.02 - 0.43)
Genedia - Oh 0.02 (0.01 - 0.07)
Hepacard - Kaur 0.07 (0.03 - 0.18)
Hepacard - Raj 0.18 (0.07 - 0.43)
QuickChaser - Abraham CC 0.12 (0.04 - 0.39)
QuickChaser - Abraham CS 0.23 (0.11 - 0.49)
Serodia - Lien 0.05 (0.02 - 0.11)
Serodia - Oh 0.04 (0.02 - 0.09)
Serodia - Sato 0.04 (0.02 - 0.09)
SimpliRed - Mvere 0.09 (0.02 - 0.43)
Virucheck - Abraham CC 0.12 (0.04 - 0.39)
Virucheck - Abraham CS 0.22 (0.09 - 0.52)

Random Effects Model


Pooled Negative LR = 0.06 (0.03 to 0.09)
Cochran-Q = 74.14; df = 18 (p = 0.0000)
0.01 1 100.0 Inconsistency (I-square) = 75.7 %
Negative LR Tau-squared = 0.9133
9.3. 9.5 List of tests available

Manufacturer Test Test type

Abbott AxSym HBsAg v2 MEIA

Abbott Architect HBsAg CMIA (Quant)

Abbott Architect HBsAg Qualitative CMIA

Abbott Architect HBsAg Qualitative II CMIA

Abbott Auszyme Monoclonal

Abbott PRISM HBsAg ChLIA

Adaltis EIAgen HBsAg Kit EIA

Siemens ADVIA Centaur HBsAg ChLIA

Siemens ADVIA Centaur HBsAg II ChLIA

Page | 263
BIOKIT Bioelisa HBsAg 3.0

bioMerieux Hepanostika Uni-Form II Microelisa

bioMerieux VIKIA HBsAg Kit

bioMerieux VIDAS HBsAg ELFA

bioMerieux VIDAS HBsAg Ultra ELFA

BIO-RAD Genscreen HBsAg 3.0

BIO-RAD MONOLISA HBsAg Ultra EIA/ ELISA

BIO-RAD MONOLISA HBsAg Plus EIA

nd
BIO-RAD MONOLISA AgHBs (2 Gen)

Dade Behring Enzygnost HBsAg EIA 5.0 EIA

Dade Behring Enzygnost HBsAg EIA 6.0 EIA

Diasorin ETI-MAK 4 HBsAg EIA EIA - ? not available

Diasorin (prev Abbott) Murex HBsAg Version 3 ELISA

Diasorin Liason HBsAg ChLIA

Diasorin Liason XL MUREX HBsAg ChLIA (Quant)

General Biologicals Corp. SURASE B-96, TMB ELISA

MBS SRL Medical Biological HBsAg One Step ELISA


Services

Ortho HBsAg ELISA Test System 3

Ortho Vitros ECi HBsAg

Roche Cobas Core HBsAg II EIA

Roche Elecsys HBsAg ECLIA

Roche Elecsys II HBsAg ECLIA

Siemens Medical Solutions Immulite 2000 HBsAg ELISA

Page | 264
Diagnostics

Beijing Wantai Biological Hepatitis B Virus Surface Antigen ELISA


Pharmacy Enterprise Co., Ltd (HBsAg) ELISA

Page | 265

You might also like