Clin Chem Lab Med 2020; 58(8): 1282–1290

Chenbin Li, Mingting Peng*, Ji Wu, Zhongli Du, Hong Lu and Wenbin Zhou

Long-term biological variation estimates

of 13 hematological parameters in healthy
Chinese subjects
https://doi.org/10.1515/cclm-2019-1141 10 parameters differed significantly between genders, ren-
Received November 4, 2019; accepted December 16, 2019; previously dering partitioning of CVG data between genders. No signifi-
­published online February 19, 2020
cant differences were detected between most BV estimates
Abstract and recently published estimates representing a Europid
population.
Background: The complete blood count (CBC) is a basic Conclusions: Most BV estimates in BIVAC-compliant stud-
test routinely ordered by physicians as a part of initial ies are similar. The turnover time of blood cells and age
diagnostic work-up on their patients. To ensure safe clini- distribution of participants should be considered in a CBC
cal application of the CBC, reliable biological variation BV study. Our study will contribute to global BV estimates
(BV) data are needed to establish analytical performance and future studies.
specifications. Our aim was to define the BV of CBC para-
Keywords: biological variation; complete blood count;
meters using a rigorous protocol that is compliant with
hematology analyzer; index of individuality; reference
the Biological Variation Data Critical Appraisal Checklist
change value.
(BIVAC) provided by the European Federation of Clinical
Chemistry and Laboratory Medicine.
Methods: Blood samples drawn from 41 healthy Chinese
subjects (22 females and 19 males; 23–59 years of age) Introduction
once monthly for 6 consecutive months were analyzed
using an ABX Pentra 80 instrument. The instrument was The complete blood count (CBC) is a basic test routinely
precisely calibrated. All samples were analyzed in dupli- ordered by physicians as a part of an initial diagnostic
cate for 13 CBC parameters. The data were assessed for work-up on their patients. To assure the safe and valid
outliers, normality, and variance homogeneity prior to clinical interpretation of test results, objective analytical
nested ANOVA. Gender-stratified within-subject (CVI) and performance specifications (APSs) are needed in clinical
between-­subject (CVG) BV estimates were calculated. laboratories [1]. Three models to establish APS have been
Results: The number of remaining data for each subject was identified by the European Federation of Clinical Chem-
442–484 after removing outliers. No significant differences istry and Laboratory Medicine (EFLM), of which models
existed between female/male CVI estimates. Except for leu- 1 and 2 are recommended [2]. Although it has been pro-
kocytes, neutrophils, and lymphocytes, the mean values of posed that the APS for hemoglobin (Hb), platelets (PLT),
and neutrophil (Neu) leukocytes should be based on clini-
*Corresponding author: Prof. Mingting Peng, National Center cal outcomes (model 1) [3], no reliable clinical outcomes
for Clinical Laboratories, Beijing Hospital, National Center of are available in practice and biological variation (BV) is
Gerontology, Beijing Engineering Research Center of Laboratory the most reliable source of useful information in defining
Medicine, Beijing, P.R. China; Institute of Geriatric Medicine, Chinese APS for CBC (model 2) [4].
Academy of Medical Sciences, Beijing 100730, P.R. China; and
Several studies involving the BV of the CBC have been
Graduate School of Peking Union Medical College, Chinese Academy
of Medical Sciences, Beijing 100730, P.R. China, included in the online 2014 BV database. The designs of
E-mail: mtpeng@nccl.org.cn these studies differed as follows: inclusion and exclusion
Chenbin Li, Zhongli Du, Hong Lu and Wenbin Zhou: National criteria, duration of study, and type of statistical model.
Center for Clinical Laboratories, Beijing Hospital, National Center The observations reported in the previous studies showed
of Gerontology, Beijing Engineering Research Center of Laboratory
a substantial difference in BV results. Specifically, there
Medicine, Beijing, P.R. China; and Institute of Geriatric Medicine,
Chinese Academy of Medical Sciences, Beijing,
were differences in the within-subject BV (CVI) of white
P.R. China. https://orcid.org/0000-0003-2434-7140 (C. Li) blood cells (WBC; 9.4%–17.3%) [5–10] and the CVI of PLT
Ji Wu: National Center for Clinical Laboratories, Beijing, P.R. China (6.6%–13.7%) [5, 6, 8, 10–12]. In the absence of an ideal
 Li et al.: Long-term biological variation estimates of hematological parameters 1283

standardized approach to estimate the BV, the EFLM rested 10–15 min in a sitting position. All samples were collected
­Biological Variation Working Group (BVWG) was estab- into K3EDTA anti-coagulated 2-mL tubes (Greiner Bio-One Vacuette®;
Kremsmunster, Austria) and analyzed within 2 h of collection. Sam-
lished and the Biological Variation Critical Appraisal
ples were drawn from all subjects once monthly for 6 consecutive
Checklist (BIVAC) was proposed [13]. In addition, the months. Using this protocol, the pre-analytical variance was consid-
BVs of CBC parameters were defined by the BVWG [14]; ered to be negligible.
however, the study period was 10 weeks, which does not
cover one turnover period for erythrocytes (approximately
4 months) [15–17]. This feature may impact the BV estima- Analytical procedure
tion of CBC parameters. Additionally, nearly all published
studies involving BV of CBCs were based on Caucasian/ All specimens were analyzed in duplicate using one ABX Pentra
White populations [6–8, 10–12, 18–21]. Indeed, only two 80 hematology analyzer (Horiba ABX SAS; Montpellier, France) in
order to calculate the analytical variation. To minimize inter-batch
studies have determined the BV among Asian popula-
analytical variation, these specimens were assayed using one lot of
tions, both of which were short-term studies (1 [22] or reagents throughout the entire study. Prior to first-batch specimen
3 days [23]). The duration of the studies was too short to analysis, instrument calibration was performed using ABX Minocal
generate reliable BV estimates. Therefore, a long-term (Horiba ABX SAS). Internal quality controls (IQC) were performed
study estimating BV is warranted. daily using ABX Difftrol Control (Horiba ABX SAS) with two differ-
The aim of this study was to estimate the components ent concentrations to provide consistent determination during the
course of the study. Each assay was performed by one analyst.
of long-term (6 months) BV, the index of individuality (II),
reference change values (RCVs), APS, and the number of
samples required to estimate the homeostatic set points
(NHSPs) for 13 CBC parameters in a cohort of healthy Statistical analysis
Chinese subjects by applying a strictly designed protocol
The homogeneity of within-subject data was assessed using average
compliant with the BIVAC, as proposed by the EFLM BVWG.
of the duplicate results of each sample from one subject by Bartlett
and Cochran tests, respectively. In heterogeneity cases, outlier data
were excluded until homogeneity was achieved. The Dixon-Reed cri-

Materials and methods terion was used to detect outliers between subjects. To confirm that
all subjects were in a steady state, linear regression was performed
on the mean group value during the entire study period for each
Participants parameter.
The Shapiro-Wilk test was used to verify the normality of
Forty-three healthy volunteers (21 males and 22 females; age range, between- and within-subject data. If the data were not normally
23–59 years; median age, 43 years) were recruited for our study. The distributed, the data were log-transformed prior to re-evaluation of
study protocol was approved by the Beijing Hospital Ethics Commit- normality.
tee and all individuals gave written informed consent to participate. According to the current recommendations of BIVAC [13], a
The exclusion criteria were as follows: surgery within 6 months; blood nested ANOVA was used to provide an accurate estimate of the CVI
transfusion and donation 4 months before or during the current study; and between-subject biologic variation (CVG). Analytical variation
pregnant or anticipating pregnancy within 6 months; breastfeeding; (CVA) was calculated from the duplicate measurements of each sam-
excessive consumption of alcohol (average alcohol consumption >60 ple. The 95% confidence intervals (CIs) for BV estimates were calcu-
g ethanol/day for men and >40 g for women); tobacco abuse (average lated as described by Roraas [24] and Burdick [25].
cigarette consumption ≥20 cigarettes/day); body mass index (BMI) Female and male data were analyzed separately. The signifi-
>30.0 kg/m2; and use of any medications. At the pre-study screening, cance of differences in CVI and CVG was assessed in subgroups by
all individuals were deemed healthy based on health questionnaires. estimating the overlap in the 95% CI. When the 95% CIs of CVI of
A CBC and routine biochemistry parameters (fasting serum glucose, females and males overlapped, it was determined that no significant
lipids, and liver and kidney function) were measured to exclude sub- difference existed between female and male BV data, and CVIs were
tle blood cell abnormalities and chronic diseases. During the study, reported for all subjects and these estimates were used in the applica-
the subjects were requested to maintain normal lifestyle habits. All tion of the BV data. When the 95% CIs of the mean values of females
subjects were asked to fast, cease smoking cigarettes and consump- and males did not overlap, the lower of the two CVG estimates was
tion of alcohol, and refrain from heavy exercising from 8:00 p.m. on used to calculate the APS.
the day before blood was to be drawn. CVI and CVG data were used to calculate the desirable APSs for
imprecision (IAPS), bias (BAPS), total error (TEAPS), the IIs, the RCVs, and
the NHSPs using the following equations [26, 27]:
Specimen collection
IAPS = 0.50 × CVI(1)

The blood samples were collected between 8:00 and 9:00 in the
morning by one experienced phlebotomist after the participants B APS = 0.25 × (CVI2 + CVG2 )1/2(2)
1284 Li et al.: Long-term biological variation estimates of hematological parameters

TEAPS = (1.65 × IAPS ) + B APS(3) lymphocytes (Lym), the mean values of other parameters
in females and males differed significantly, rendering par-
II = (CVI2 + CVA2 ) / CVG(4) titioning of the CVG data between genders.
The 95% CIs for CVI estimates of all parameters,
RCV = 2 1/2 × Z × (CVA2 + CVI2 )1/2(5) except the mean cell hemoglobin concentration (MCHC)
and basophil (Bas [×109/L]), overlapped with the BIVAC-
NHSP = (Z × (CVA2 + CVI2 )1/2 /D)2 ,(6) compliant studies reported by Coskun et al. [14], which
showed there were no significant differences in CVI esti-
where D is the allowed percentage deviation from the true homeo-
static set point, CVA denotes analytical variation, and Z is 1.96 for a
mates between the two studies. Our 95% CIs for CVG esti-
p-value <0.05. We calculated NHSPs associated with 5%, 10%, and mates of all parameters, except Bas (×109/L), overlapped
20% deviations from the true homeostatic set points. with the CVG estimates reported by Coskun et al. [14],
Data analysis was performed using SPSS (version 16.0; SPSS which also showed that there were no significant differ-
IBM, Chicago, IL, USA) and Microsoft Excel 2007 (Microsoft Corpora- ences in CVG estimates between the two studies.
tion, Redmond, WA, USA).
The 95% CIs of CVI estimates for all parameters,
except red blood cells (RBC), Hb, and mean cell hemo-
globin (MCH), overlapped with the CVI estimates of

Results BIVAC-compliant studies reported by Buoro et al. [28–30]

(medium-term BV), which showed there were no signifi-
cant differences in CVI between the two studies. Our 95%
Main characteristics of participants CIs of CVG estimates for all parameters overlapped with
those of Buoro et al. [28–30], which also showed that there
Thirty-seven subjects completed all six scheduled collec- were no significant differences in CVG estimates between
tions, four subjects completed five collections, and two the two studies.
subjects completed one collection. Thus, 41 of the 43 sub- As shown in Table 1, the CVs of the IQC for all para-
jects (22 females and 19 males) were included in the study. meters, except the mean corpuscular volume (MCV) and
The mean ages of the females and males were 44 years (age MCHC, were higher than the CVA based on duplicate tests.
range, 23–59 years) and 38 years (age range, 23–59 years),
respectively. There were no significant differences in age
and BMI between males and females. Indices derived from BV

The APSs for hematological parameters are shown in

Outlier removal Table 2. The CVAPS of all parameters, except MCV, MCH,
and MCHC, and BAPS of all parameters, except MCHC, Lym
Two hundred and forty-two fresh blood samples from 41 (×109/L), and eosinophils ([Eos] ×109/L), and the RCV of
participants were collected and measured in duplicate. all parameters, except PLT, MCH, and MCHC, were higher
The number of replicates, samples, and subjects identi- than the corresponding values reported by Coskun et al.
fied as outliers by Bartlett and Cochran tests is shown in [14].
Supplementary Table A. In total, 2.2% of obtained data The IIs for nine of 13 parameters were <0.60, and for
were excluded from the final analysis. All subjects were in the remaining parameters IIs ranged from 0.63 to 1.50. The
a steady state during the study (p > 0.05). NHSP (within 5% of the actual value) was 2 in the eryth-
rocyte group, but much higher in the leukocyte group
(exceeding 100 for the Bas count). In the leukocyte group,
Comparison of BV data between the current particularly for Bas (×109/L), widening the target range
study and BIVAC-compliant studies to 20% still required measurement of seven samples to
derive the estimates of homeostatic set points.
The mean, CVA, CVI, and CVG of all parameters are shown
in Table 1. The CVI of the CBC parameters displayed a
constant overlap of 95% CIs between females and males; Discussion
thus, there were no significant gender differences in
CVI estimates. Therefore, we reported the CVI estimates In the experimental design of this study, several key ele-
for all subjects. With the exception of WBC, Neu, and ments of BV for the CBC, such as the number of subjects,
Table 1: Biological variation estimates for hematological parameters with 95% CIs, accompanied by the corresponding BV estimates in two BIVAC-compliant studies.

Parameters Sex Mean CV% This study Coskun et al. Buoro et al.
(95% CI) (IQC)
CVA CVI CVG CVA CVI CVG CVA CVI CVG
(95% CI), % (95% CI), % (95% CI), % (95% CI), % (95% CI), % (95% CI), % (95% CI), % (95% CI), % (95% CI), %

WBC, ×109/L All 6.03 1.84 1.48 11.61 23.05 1.49 NA 16.53 1.5 11.1 15.0
(5.88–6.16) (1.36–1.63) (10.57–12.88) (18.72–29.74) (1.38–1.62) (12.83–22.08) (1.3–1.7) (9.62–13.18) (10.9–22.5)
M 6.08 NA 1.47 10.40 23.85 NA 7.96 NA NA NA NA
(5.88–6.29) (1.30–1.69) (9.06–12.19) (17.79–35.59) (7.02–9.19)
F 5.97 NA 1.50 12.56 22.96 NA 12.82 NA NA NA NA
(5.78–6.15) (1.34–1.71) (11.09–14.48) (17.35–33.23) (11.51–14.46)
RBC, ×1012/L All 4.49 1.79 0.70 3.23 8.95 0.66 2.77 NA 0.64 1.8 9.1
(4.45–4.53) (0.64–0.77) (2.93–3.59) (7.31–11.50) (0.61–0.72) (2.55–3.04) (0.56–0.74) (1.5–2.1) (6.9–13.1)
M 4.81 NA 0.93 3.12 6.10 NA NA 5.58 0.6 1.4 7.8
(4.77–4.86) (0.82–1.07) (2.71–3.67) (4.52–9.13) (4.01–9.40) (0.5–0.75) (1.1–1.9) (5.2–14.9)
F 4.22 NA 0.75 3.26 6.11 NA NA 6.98 0.66 2.4 7.1
(4.19–4.26) (0.67–0.85) (2.87–3.77) (4.62–8.84) (5.15–10.65) (0.56–0.80) (1.9–3.0) (5.0–12.1)
Hb, g/L All 140.72 1.29 0.60 3.16 11.00 0.58 2.74 NA 0.52 2.0 8.2
(139.30–142.15) (0.55–0.66) (2.87–3.51) (9.00–14.11) (0.53–0.63) (2.52–3.00) (0.46–0.60) (1.7–2.4) (6.3–11.9)
M 153.45 NA 0.60 2.86 5.94 NA NA 5.81 0.44 1.3 6.3
(152.13–154.77) (0.53–0.69) (2.49–3.36) (4.42–8.88) (4.08–9.57) (0.37–0.56) (1.1–1.7) (4.2–12.1)
F 130.12 NA 0.59 3.46 8.47 NA NA 6.57 0.59 2.3 6.8
(128.70–131.54) (0.53–0.67) (3.05–3.99) (6.45–12.19) (4.83–10.01) (0.48–0.69) (1.9–2.9) (4.8–11.7)
HCT, % All 0.417 1.73 1.11 3.12 10.22 0.63 2.82 NA 0.60 2.4 7.4
(0.412–0.420) (1.02–1.22) (2.82–3.48) (8.36–13.12) (0.59–0.69) (2.59–3.09) (0.53–0.70) (2.1–2.9) (5.6–10.7)
M 0.452 NA 1.17 2.82 5.25 NA NA 5.46 0.58 2.2 5.7
(0.449–0.456) (1.03–1.35) (2.43–3.34) (3.88–7.88) (3.87–9.07) (0.48–0.73) (1.8–2.9) (3.8–11.1)
F 0.386 NA 1.04 3.40 7.62 NA NA 5.51 0.62 2.6 6.2
(0.382–0.390) (0.93–1.18) (2.98–3.94) (5.79–10.98) (4.03–8.41) (0.53–0.76) (2.1–3.2) (4.3–10.7)
PLT, ×109/L All 223.83 2.76 2.74 6.74 23.71 1.80 7.22 NA 1.09 7.22 15.4
(218.89–228.77) (2.51–3.01) (6.08–7.53) (19.39–30.42) (1.67–1.96) (6.63–7.91) (0.97–1.27) (6.27–8.52) (11.5–22.7)
M 211.57 NA 3.08 6.76 25.54 NA NA 7.23 NA NA NA
(204.17–218.98) (2.72–3.56) (5.78–8.05) (19.20–37.91) (4.93–12.42)
F 233.76 NA 2.49 6.71 21.96 NA NA 17.42 NA NA NA
(227.35–240.18) (2.22–2.83) (5.87–7.80) (16.80–31.51) (12.90–26.71)
MCV, fL All 92.71 0.65 0.74 0.69 4.61 0.18 0.72 3.96 0.19 0.9 3.9
(92.32–93.10) (0.68–0.81) (0.58–0.81) (3.77–5.93) (0.16–0.19) (0.66–0.79) (3.15–5.33) (0.17–0.22) (0.8–1.1) (3.0–5.6)
M 94.05 NA 0.76 0.50 3.13 NA NA NA 0.17 0.9 3.2
(93.65–94.44) (0.67–0.88) (0.30–0.68) (2.36–4.64) (0.14–0.21) (0.7–1.1) (2.2–6.3)
Li et al.: Long-term biological variation estimates of hematological parameters

F 91.55 NA 0.73 0.83 5.41 NA NA NA 0.21 1.0 4.2

(90.93–92.16) (0.65–0.83) (0.68–1.01) (4.13–7.82) (0.18–0.25) (0.8–1.2) (3.0–7.2)
MCH, pg All 31.51 0.93 0.70 0.69 4.94 0.78 0.75 0.77 0.3 4.7
(31.39–31.64) (0.64–0.77) (0.58–0.81) (4.03–6.37) (0.73–0.85) (0.65–0.86) (0.68–0.89) (0.0–0.5) (3. 6–6.9)
M 31.93 NA 0.73 0.69 3.30 NA NA 2.9 0.72 0.3 3.5
1285

(31.79–32.07) (0.64–0.84) (0.53–0.87) (2.48–4.90) (2.22–5.15) (0.60–0.91) (0.0–0.6) (2.3–6.7)

Table 1 (continued)
1286

Parameters Sex Mean CV% This study Coskun et al. Buoro et al.
(95% CI) (IQC)
CVA CVI CVG CVA CVI CVG CVA CVI CVG
(95% CI), % (95% CI), % (95% CI), % (95% CI), % (95% CI), % (95% CI), % (95% CI), % (95% CI), % (95% CI), %

F 31.13 NA 0.67 0.69 4.69 NA NA 6.11 0.81 0.2 5.6

(30.95–31.32) (0.59–0.77) (0.55–0.85) (3.56–6.86) (4.55–9.31) (0.68–0.99) (0.0–0.6) (3.9–9.9)
MCHC, g/L All 337.86 1.11 1.02 0.54 1.07 0.79 0.97 1.59 0.77 0.8 1.8
(337.39–338.33) (0.94–1.12) (0.34–0.70) (0.85–1.40) (0.73–0.86) (0.87–1.09) (1.13–2.69) (0.68–0.89) (0.6–1.0) (1.3–2.6)
M 339.30 NA 1.13 0.32 0.77 NA NA NA 0.74 0.8 0.9
(338.67–339.90) (1.00–1.30) (0.00–0.64) (0.53–1.20) (0.61–0.93) (0.5–1.1) (0.6 –2.0)
F 336.66 NA 0.92 0.67 1.18 NA NA NA 0.8 0.9 2.1
(336.00–337.32) (0.82–1.05) (0.47–0.86) (0.87–1.73) (0.67–0.96) (0.6–1.2) (1.4–3.6)
Neu, ×109/L All 3.22 2.51 1.70 15.43 28.62 1.88 NA NA 1.9 14.6 24.1
(3.13–3.32) (1.56–1.87) (14.03–17.13) (23.20–36.97) (1.74–2.05) (1.7–2.2) (12.7–17.2) (17.9–35.4)
M 3.18 NA 1.72 14.21 27.68 NA 11.60 16.68 NA NA NA
(3.05–3.31) (1.52–1.99) (12.37–16.69) (20.53–41.44) (10.20–13.44) (11.80–28.39)
F 3.26 NA 1.94 16.32 29.96 NA 20.09 27.81 NA NA NA
(3.13–3.39) (1.73–2.21) (14.39–18.85) (22.63–43.36) (18.05–22.65) (20.17–42.79)
Lym, ×109/L All 2.11 4.5 2.12 11.08 21.71 2.40 9.81 2.7 11.0 23.0
(2.06–2.15) (1.95–2.33) (10.07–12.31) (17.62–28.02) (2.21–2.61) (9.00–10.77) (2.4–3.1) (9.5–13.0) (17.2–34.2)
M 2.14 NA 2.09 10.77 25.77 NA NA 23.66 NA NA NA
(2.05–2.22) (1.84–2.41) (9.36–12.67) (19.23–38.43) (17.1–39.9)
F 2.08 NA 2.15 11.31 18.06 NA NA 20.36 NA NA NA
(2.03–2.14) (1.92–2.45) (9.96–13.08) (13.56–26.24) (15.17–31.62)
Mono, ×109/L All 0.44 9.19 7.26 15.40 27.69 4.95 NA NA 3.9 13.4 22.2
(0.42–0.45) (6.67–7.97) (13.85–17.25) (22.40–35.83) (4.58–5.39) (3.5–4.5) (11.6–15.9) (11.5–32.7)
M 0.47 NA 6.71 14.24 24.44 NA 11.07 17.01 NA NA NA
Li et al.: Long-term biological variation estimates of hematological parameters

(0.45–0.49) (5.93–7.74) (12.19–16.92) (17.99–36.78) (9.67–12.88) (12.34–29.47)

F 0.41 NA 7.79 15.57 29.65 NA 15.33 29.78 NA NA NA
(0.39–0.42) (6.95–8.86) (13.49–18.22) (22.39–42.92) (13.68–17.40) (22.53–46.92)
Eos, ×109/L All 0.17 16.25 8.67 16.15 54.65 11.01 NA 70.5 9.5 15.6 61.5
(0.16–0.18) (7.94–9.55) (14.41–18.24) (44.67–70.17) (10.07–12.13) (59.4–100.6) (8.3–11.1) (13.0–19.1) (45.7–90.5)
M 0.18 NA 8.23 13.58 48.43 NA 10.11 NA NA NA NA
(0.17–0.20) (7.22–9.57) (11.33–16.51) (36.34–71.96) (7.77–12.26)
F 0.16 NA 9.05 14.58 60.77 NA 14.83 NA NA NA NA
(0.15–0.17) (8.06–10.32) (12.42–17.29) (46.56–87.11) (12.42–17.62)
Bas, ×109/L All 0.035 21.28 18.57 18.85 43.71 16.65 11.36 22.10 12.2 12.8 33.9
(0.034–0.037) (16.99–20.48) (15.93–22.07) (35.43–56.48) (15.24–18.35) (7.66–13.24) (16.82–29.93 (10.7–14.1) (10.1–16.1) (25.5–49.5)
M 0.040 NA 17.82 19.26 42.32 NA NA NA NA NA NA
(0.037–0.043) (15.65–20.70) (15.09–24.19) (32.34–65.55)
F 0.032 NA 18.94 17.87 42.02 NA NA NA NA NA NA
(0.030–0.034) (16.82–21.67) (13.87–22.30) (31.73–60.82)

NA, not available; IQC, internal quality control.

 Li et al.: Long-term biological variation estimates of hematological parameters 1287

number of samples, number of replicates, and the feasi-

1
1

1
1

1
No (20%)

1
Coskun et al.

2
1
2
3
4
bility of conducting studies, were taken into account, as
follows: (1) The number of subjects was between 20 and
40 in nearly all previously published studies involving

1
1

1
1

1
No (10%)

3
1

6
4
6
9
16
BV for the CBC [6, 8, 10, 11, 18, 20, 23, 31, 32]. Recently,
Braga et al. [4] concluded that a minimum of 10 subjects
is sufficient to obtain a good BV estimate for each iden-
tified subgroup. Therefore, we enrolled 19 males and 22

1
2

1
2

1
No (5%)

11
2

22
16
23
35
63
females to participate in the study. (2) Based on the review
by Braga et al. [4], the study duration must be neither too
short (a few days) nor too long (years) to eliminate influ-
Table 2: The APS for hematological parameters based on biological variation, accompanied by the corresponding BV estimates in BIVAC-compliant study.

0.61
0.62

0.26
0.47

0.18
II

0.48
0.50

1.00

0.70
0.48
0.65
0.14
0.51
ence by additional causes. Based on the turnover time
for CBC parameters (i.e. RBC ~4 months), a study period
3.47
8.00

3.00
7.76

2.06
TEAPS, %

22.40
7.89

20.60

32.50
28.00
33.20
41.40
55.80 of 6 months was used in our study. (3) The number of
samples from each individual was 5–12 and the number of
measurements for each sample was 1–2 in previously pub-
lished reports [6, 8, 10, 11, 18, 20, 23, 31, 32]. Considering
0.47
1.34

0.75
1.61

1.01
BAPS, %

4.59
1.56

2.55

5.08
6.65
5.07
17.80
6.21

that the CVA was calculated with all replicates for the same
subject, the number of measurements for each sample
was two in our study. According to the model of the power
0.49
1.41

0.38
1.37

0.36
IAPS, %

3.98
1.39

3.61

5.80
4.91
5.54
5.06
5.68

estimated by Roraas et al. [24], at least six samples are

required for each individual when the power is defined as
IAPS, the desirable APSs for imprecision; BAPS, the desirable APSs for bias; TEAPS, the desirable APSs for total error.

>80% with a ratio of CVA to CVI < 2 and 41 subjects. In our

This study

1
1

1
No (20%)

1
2
1

3
2
3
4
7

study following the aforementioned design, the power of

all parameters was 1.00, except MCHC, with a power of
0.91 (Supplementary Table B), which indicates that the
1
1

1
No (10%)

1
6
1

10
5
12
13
27

power of our study design was higher.

Calibration and quality control are essential to ensure
the accuracy of hematology analyzer. The Clinical Labo-
ratory Improvement Amendments (CLIA §493.1255 Stand-
1
2

1
No (5%)

1
22
2

38
20
45
52
108

ard) [33] and National Health Standard of China (WS/T

347—2011) [34] require (re)calibration at least once every
6 months and when there is major preventive maintenance
1.50
0.63

0.30
II

0.54

0.32
0.51
0.54

0.33

0.54
0.52
0.70
0.34
0.63

or replacement of critical parts, or when control materials

reflect an unusual trend or shift, or are outside of the labo-
3.20
9.18

2.72
RCV, %

8.92

2.80
32.44
9.16

20.17

43.03
31.27
47.19
50.81
73.35

ratory’s acceptable limits. In our study, we performed one

calibration prior to first-batch specimen analysis due to
the duration time being no more than 6 months, no major
0.68
4.10

1.41
TEAPS, %

4.29

1.37
16.03
4.39

11.30

20.86
15.23
19.93
27.57
27.06

preventive maintenance or repair service, and no analyti-

cal run considered out of control.
In previous BV studies [7, 12, 20, 21, 23, 31, 32, 35], CVA
was derived from the laboratory IQC data generated by
0.24
1.53

0.84
BAPS, %

1.68

0.80
6.45
1.73

5.74

8.13
6.09
7.22
14.25
11.51

use of commercial sample materials. Fraser et al. [8] and

Braga and Panteghini [4] suggested that CVA is estimated
as the average variance between duplicate measurements
0.27
1.56

0.35
1.58

0.35
IAPS, %

5.81
1.62

3.37

7.72
5.54
7.70
8.08
9.43

of the analyte. Roraas et al. [24] and Braga et al. [36] pro-
posed that CVA derived from the laboratory IQC data is not
necessarily transferable to clinical samples. In our study,
Mono, ×109/L
RBC, ×1012/L
WBC, ×10 /L

Neu, ×109/L
Lym, ×109/L

Bas, ×109/L
Eos, ×109/L
Parameters

PLT, ×109/L

MCHC, g/L

the CVs from the IQC of 12 CBC parameters were greater

9

MCH, pg
Hb, g/L

MCV, fL
HCT, %

than the CVs from duplicate measurements of clinical

samples, except the MCV, which indicates that the derived
1288 Li et al.: Long-term biological variation estimates of hematological parameters

parameters (such as II, RCV, and NHSP) would be over- menstrual cycle [43, 44]. When the BVs of the WBC sub-
estimated. Therefore, we support the use of a duplicate populations for females are estimated, the influence of the
analysis to calculate CVA. female menstrual cycle should be considered.
The previously published BV data were compared The CVGs for PLT in the male group of the present study
with the data listed in the online 2014 BV database. We were higher than the CVGs reported by Coskun et al. [14].
reviewed the 15 articles cited in the BV database following Based on the findings of Biino et al. [45], the PLT count was
the standards in the BIVAC and found the studies had a stable for males in the 18–49-year age range, whereas the
low-quality score. Twelve articles [6–11, 37–42] received a PLT count was significantly decreased in subjects >50 years
D score because the instruments used were obsolete or the of age. The age range in the Coskun et al.’s study [14] and
methods are no longer used routinely. The estimates from the present study was 20–36 and 23–59 years, respectively.
these studies were considered unsuitable for application in The difference in age range might account for the differ-
clinical practice. Two articles [23, 32] were scored as a C due ence in CVGs. If we calculated the CVG using data from
to the absence of outlier removal, normality assessments, males <36 years of age, the CVG for PLT decreased from
and variance homogeneity examination, which affect the 25.54% to 13.66%. Therefore, the age distribution of par-
accuracy of BV estimates (over- or under-estimation). One ticipants might significantly influence the CVG estimates.
article [12] was scored as a C for lack of CIs around esti- The II has been used to investigate the utility of con-
mates of CVI and inappropriate methods of outlier analy- ventional population-based reference intervals (RIs). The
ses. Therefore, we used PubMed and Google Scholar to IIs of nine of 13 CBC parameters were <0.6, which suggests
retrieve relevant literature without any time limits, using that conventional RIs are of very limited utility. In these
the following keywords: biological variation (variability); cases, the RCV is a better parameter for monitoring of lon-
hematology; complete blood cell; leukocytes/WBC; eryth- gitudinal changes in test results in an individual [46]. The
rocytes/RBC; hemoglobin/Hb/Hgb; hematocrit/HCT/Htc; IIs of two of nine parameters <0.60 in our study are not in
and platelet/PLT. As unique inclusion/exclusion criteria, agreement with those reported by Coskun et al. [14]. Addi-
the recruited studies should have an explicit aim of the tionally, the II of MCHC was >1.4, which makes RI a useful
experimental assessment of CBC BV components. Nine- tool for test interpretation. Such is not the case for II of
teen articles were recruited [6–12, 14, 23, 28–30, 32, 37–42]. MCHC in the Coskun et al.’s study [14]. Additional research
These articles were reviewed by the quality indicators (QIs) is warranted to further clarify the differences.
included in the BIVAC. Four articles [14, 28–30] were con-
sidered to be grade A, which met the QIs of the BIVAC and
were reliable estimations. Therefore, we used the BV esti-
mation of these studies as the comparators. Conclusions
Our study showed that the CVIs for RBC, Hb, and
HCT were higher than those reported in two studies by In conclusion, the present study has provided an estimate
Coskun et al. [14] and Buoro et al. [28–30] (Supplemen- of the BV and the main indices of healthy Chinese adults
tary Figure A). This finding can be explained by the dif- following an experimental protocol with high power
ferent duration of the studies. The duration of our study and a rigorous statistical approach complying with the
was 6 months and longer than the two studies. Addition- BIVAC. With some exceptions, the 95% CIs of BV in our
ally, the CVIs of 10 weeks’ duration by Coskun et al. [14] study for CBC parameters overlapped with the 95% CIs
were also higher than those of 5 weeks’ duration by Buoro of the recently published BIVAC-compliant BV studies of
et al. [28–30], which followed the same trend. Within our a Europid population, which showed that no differences
study period of 6 months, we covered one turnover period exist between the BVs of the two populations. The BV data
for erythrocytes, and the CVIs for the erythrocyte group for erythrocyte parameters (RBC, Hb, and HCT) and PLT in
tests might be more reliable. Although the 95% CIs of CVI this study, however, were higher than those in the BIVAC-
between genders were overlapped for all CBC parameters compliant studies published because of the differences
in the current study, the CVIs for WBC, Neu (×109/L), Mono associated with the study period and age distribution.
(×109/L), and Eos (×109/L) in the females were higher than Our study will provide reference and aid for the global BV
those in the males. Significant differences in the CVI esti- database by EFLM and future studies on this subject.
mates between genders for leukocytes, Mono, Neu, and
Eos were also evident in the Coskun et al.’s study [14]. Author contributions: All the authors have accepted
This finding can be explained by the cyclic variations in responsibility for the entire content of this submitted
WBC and the WBC sub-populations during the female manuscript and approved submission.
 Li et al.: Long-term biological variation estimates of hematological parameters 1289

Research funding: National key research and d ­ evelopment 13. Aarsand AK, Roraas T, Fernandez-Calle P, Ricos C, Diaz-Garzon
program, No. 2017YFC0910003. National Natural Science J, Jonker N, et al. The biological variation data critical appraisal
checklist: a standard for evaluating studies on biological varia-
Foundation of China, No. 81772254. National Special
tion. Clin Chem 2018;64:501–14.
Project of Science and Technology Basic Work of Ministry 14. Coskun A, Carobene A, Kilercik M, Serteser M, Sandberg S,
of Science and Technology of China, No. 2013FY113800. Aarsand AK, et al. Within-subject and between-subject bio-
Employment or leadership: None declared. logical variation estimates of 21 hematological parameters in
Honorarium: None declared. 30 healthy subjects. Clin Chem Lab Med 2018;56:1309–18.
15. Glader B. Destruction of erythrocytes. In: Greer JP, Foerster J,
Competing interests: The funding organization(s) played
Lukens JN, Rodgers GM, Paraskevas F, Glader B, editors. Win-
no role in the study design; in the collection, analysis, and trobe’s clin hematol, 2nd ed. Philadelphia: Lippincott Williams
interpretation of data; in the writing of the report; or in the and Wilkins, 2004:249–65.
decision to submit the report for publication. 16. Lu SJ, Li F, Yin H, Feng Q, Kimbrel EA, Hahm E, et al. Platelets
generated from human embryonic stem cells are functional
in vitro and in the microcirculation of living mice. Cell Res
2011;21:530–45.
References 17. Pillay J, den Braber I, Vrisekoop N, Kwast LM, de Boer RJ,
Borghans JA, et al. In vivo labeling with 2H2O reveals a human
1. Falay M, Senes M, Korkmaz S, Zararsiz G, Turhan T, Okay M, neutrophil lifespan of 5.4 days. Blood 2010;116:625–7.
et al. Biological variation of peripheral blood T-lymphocytes. 18. Winkel P, Statland BE, Saunders AM, Osborn H, Kupperman H.
J Immunol Methods 2019;470:1–5. Within-day physiologic variation of leukocyte types in healthy
2. Panteghini M, Sandberg S. Defining analytical performance subjects as assayed by two automated leukocyte differential
specifications 15 years after the Stockholm conference. Clin analyzers. Am J Clin Pathol 1981;75:693–700.
Chem Lab Med 2015;53:829–32. 19. Jones AR, Twedt D, Swaim W, Gottfried E. Diurnal change of
3. Ceriotti F, Fernandez-Calle P, Klee GG, Nordin G, Sandberg S, blood count analytes in normal subjects. Am J Clin Pathol
Streichert T, et al. Criteria for assigning laboratory measurands to 1996;106:723–7.
models for analytical performance specifications defined in the 1st 20. Sennels HP, Jorgensen HL, Hansen AL, Goetze JP, Fahrenkrug J.
EFLM Strategic Conference. Clin Chem Lab Med 2017;55:189–94. Diurnal variation of hematology parameters in healthy young
4. Braga F, Panteghini M. Generation of data on within-subject males: the Bispebjerg study of diurnal variations. Scand J Clin
biological variation in laboratory medicine: an update. Crit Rev Lab Invest 2011;71:532–41.
Clin Lab Sci 2016;53:313–25. 21. Sudmann-Day AA, Piehler A, Klingenberg O, Urdal P. Six-day
5. Costongs GM, Janson PC, Bas BM, Hermans J, Brombacher stability of erythrocyte and reticulocyte parameters in-vitro: a
PJ, van Wersch JW. Short-term and long-term intra-individual comparison of blood samples from healthy, iron-deficient, and
variations and critical differences of haematological laboratory thalassemic individuals. Scand J Clin Lab Invest 2015;75:247–53.
parameters. J Clin Chem Clin Biochem 1985;23:69–76. 22. Yang D, Zhou Y, Yang C. Daytime biological variation of hema-
6. Dot D, Miro J, Fuentes-Arderiu X. Within-subject biological vari- tological parameters in a healthy Chinese population. Int J Lab
ation of hematological quantities and analytical goals. Arch Hematol 2016;39:e37–40.
Pathol Lab Med 1992;116:825–6. 23. Tang H, Jing J, Bo D, Xu D. Biological variations of leuko-
7. Dot D, Miro J, Fuentes-Arderiu X. Biological variation of the cyte numerical and morphologic parameters determined by
leukocyte differential count quantities. Scand J Clin Lab Invest UniCel DxH 800 hematology analyzer. Arch Pathol Lab Med
1992;52:607–11. 2012;136:1392–6.
8. Fraser CG, Wilkinson SP, Neville RG, Knox JD, King JF, MacWalter 24. Roraas T, Petersen PH, Sandberg S. Confidence intervals and
RS. Biologic variation of common hematologic laboratory quan- power calculations for within-person biological variation: effect
tities in the elderly. Am J Clin Pathol 1989;92:465–70. of analytical imprecision, number of replicates, number of sam-
9. Looker AC, Sempos CT, Liu KA, Johnson CL, Gunter EW. Within- ples, and number of individuals. Clin Chem 2012;58:1306–13.
person variance in biochemical indicators of iron status: effects 25. Burdick RK, Graybill F. Confidence intervals on variance compo-
on prevalence estimates. Am J Clin Nutr 1990;52:541–7. nents, 1st ed. New York: Marcel Dekker, Inc., 1992.
10. Statland BE, Winkel P, Harris SC, Burdsall MJ, Saunders AM. 26. Fraser CG. Biological variation: from principles to practice.
Evaluation of biologic sources of variation of leukocyte counts Washington, DC: AACC, 2001.
and other hematologic quantities using very precise automated 27. Harris EK. Effects of intra- and interindividual variation on the
analyzers. Am J Clin Pathol 1978;69:48–54. appropriate use of normal ranges. Clin Chem 1974;20:1535–42.
11. Maes M, Scharpe S, Cooreman W, Wauters A, Neels H, Verkerk 28. Buoro S, Carobene A, Seghezzi M, Manenti B, Pacioni A, Ceriotti
R, et al. Components of biological, including seasonal, varia- F, et al. Short- and medium-term biological variation estimates
tion in hematological measurements and plasma fibrinogen of leukocytes extended to differential count and morphology-
­concentrations in normal humans. Experientia 1995;51:141–9. structural parameters (cell population data) in blood samples
12. Pineda-Tenor D, Laserna-Mendieta EJ, Timon-Zapata J, obtained from healthy people. Clin Chim Acta 2017;473:147–56.
­Rodelgo-Jimenez L, Ramos-Corral R, Recio-Montealegre A, et al. 29. Buoro S, Seghezzi M, Manenti B, Pacioni A, Carobene A, Ceriotti
Biological variation and reference change values of common F, et al. Biological variation of platelet parameters deter-
clinical chemistry and haematologic laboratory analytes in the mined by the Sysmex XN hematology analyzer. Clin Chim Acta
elderly population. Clin Chem Lab Med 2013;51:851–62. 2017;470:125–32.
1290 Li et al.: Long-term biological variation estimates of hematological parameters

30. Buoro S, Carobene A, Seghezzi M, Manenti B, Dominoni P, 38. Holzel WG. Intra-individual variation of analytes in serum from
Pacioni A, et al. Short- and medium-term biological variation patients with chronic liver diseases. Clin Chem 1987;33:1133–6.
estimates of red blood cell and reticulocyte parameters in 39. Holzel WG. Influence of hypertension and antihypertensive
healthy subjects. Clin Chem Lab Med 2018;56:954–63. drugs on the biological intra-individual variation of electrolytes
31. Van Wyck DB, Alcorn H, Jr., Gupta R. Analytical and biological and lipids in serum. Clin Chem 1988;34:1485–8.
variation in measures of anemia and iron status in patients 40. Gallagher SK, Johnson LK, Milne DB. Short-term and long-term
treated with maintenance hemodialysis. Am J Kidney Dis variability of indices related to nutritional status. I: Ca, Cu, Fe,
2010;56:540–6. Mg, and Zn. Clin Chem 1989;35:369–73.
32. Zhang P, Tang H, Chen K, Chen Y, Xu D. Biological variations 41. Borel MJ, Smith SM, Derr J, Beard JL. Day-to-day variation in
of hematologic parameters determined by UniCel iron-status indices in healthy men and women. Am J Clin Nutr
DxH 800 hematology analyzer. Arch Pathol Lab Med 1991;54:729–35.
2013;137:1106–10. 42. Ahluwalia N, Lammi-Keefe CJ, Haley NR, Beard JL. Day-to-day
33. Clinical Laboratory Improvement Amendments – State variation in iron-status indexes in elderly women. Am J Clin Nutr
Operations Manual. Available at: https://www.cms.gov/ 1993;57:414–9.
Regulations-and-Guidance/Guidance/Manuals/downloads/ 43. Mathur S, Mathur RS, Goust JM, Williamson HO, Fudenberg HH.
som107ap_c_lab.pdf. Accesed: 4 Dec. 2019. Cyclic variations in white cell subpopulations in the human
34. Ministry of Health of the People’s Republic of China. Guide- menstrual cycle: correlations with progesterone and estradiol.
line for the calibration of blood cell assays (WS/T 347—2011). Clin Immunol Immunopathol 1979;13:246–53.
­Beijing: Chinese Standard Press, 2011. 44. Bain BJ, England JM. Variations in leucocyte count during men-
35. Lacher DA, Barletta J, Hughes JP. Biological variation of hematol- strual cycle. Br Med J 1975;2:473–5.
ogy tests based on the 1999–2002 National Health and Nutrition 45. Biino G, Santimone I, Minelli C, Sorice R, Frongia B, Traglia M,
Examination Survey. Natl Health Stat Report 2012;54:1–10. et al. Age- and sex-related variations in platelet count in Italy:
36. Braga F, Dolci A, Mosca A, Panteghini M. Biological variability of a proposal of reference ranges based on 40987 subjects’ data.
glycated hemoglobin. Clin Chim Acta 2010;411:1606–10. PLoS One 2013;8:e54289.
37. Costongs GM, Bas BM, Janson PC, Hermans J, Brombacher PJ,
van Wersch JW. Short-term and long-term intra-individual vari-
ations and critical differences of coagulation parameters. J Clin Supplementary Material: The online version of this article offers
Chem Clin Biochem 1985;23:405–10. supplementary material (https://doi.org/10.1515/cclm-2019-1141).