SPINE Volume 32, Number 1, pp 82–92

©2007, Lippincott Williams & Wilkins, Inc.

Herbal Medicine for Low Back Pain

A Cochrane Review

Joel J. Gagnier, ND, MSc, PhD(cand.),*† Maurits W. van Tulder, PhD,‡ Brian Berman, MD,§
and Claire Bombardier, MD¶

Study Design. A systematic review of randomized con- by two individuals (J.J.G. and M.W.T.). Disagreements
trolled trials. were resolved by consensus.
Objectives. To determine the effectiveness of herbal Results. Ten trials were included in this review. Two
medicine compared with placebo, no intervention, or high-quality trials utilizing Harpagophytum procumbens
“standard/accepted/conventional treatments” for non- (Devil’s claw) found strong evidence for short-term im-
specific low back pain. provements in pain and rescue medication for daily doses
Summary of Background Data. Low back pain is a standardized to 50 mg or 100 mg harpagoside with an-
common condition and a substantial economic burden in other high-quality trial demonstrating relative equiva-
industrialized societies. A large proportion of patients lence to 12.5 mg per day of rofecoxib. Two moderate-
with chronic low back pain use complementary and alter- quality trials utilizing Salix alba (White willow bark) found
native medicine (CAM) and/or visit CAM practitioners. moderate evidence for short-term improvements in pain
Several herbal medicines have been purported for use in and rescue medication for daily doses standardized to 120
low back pain. mg or 240 mg salicin with an additional trial demonstrat-
Methods. The following databases were searched: ing relative equivalence to 12.5 mg per day of rofecoxib.
Medline (1966 to April 2003), Embase (1980 to April 2003), Three low-quality trials using Capsicum frutescens (Cay-
Cochrane Controlled Trials Register (Issue 1, 2003), and enne) using various topical preparations found moderate
Cochrane Complementary Medicine (CM) field Trials Reg- evidence for favorable results against placebo and one
ister. Additionally, reference lists in review articles, guide- trial found equivalence to a homeopathic ointment.
lines, and in the retrieved trials were checked. Random- Conclusions. Harpagophytum procumbens, Salix alba,
ized controlled trials (RCTs), using adults (⬎18 years of and Capsicum frutescens seem to reduce pain more than
age) suffering from acute, subacute, or chronic nonspe- placebo. Additional trials testing these herbal medicines
cific low back pain. Types of interventions included herbal against standard treatments will clarify their equivalence in
medicines defined as a plant that is used for medicinal terms of efficacy. The quality of reporting in these trials was
purposes in any form. Primary outcome measures were generally poor; thus, trialists should refer to the CONSORT
pain and function. Two reviewers (J.J.G. and M.W.T.) statement in reporting clinical trials of herbal medicines.
conducted electronic searches in all databases. One re- Key words: systematic review, herbal medicine, low
viewer (J.J.G.) contacted content experts and acquired back pain. Spine 2007;32:82–92
relevant citations. Authors, title, subject headings, publi-
cation type, and abstract of the isolated studies were
downloaded or a hard copy was retrieved. Methodologic Low back pain (LBP) and related disability are major
quality and clinical relevance were assessed separately public health problems across industrialized nations. As
a result, the past 15 years have seen an intensive research
effort to identify effective treatment and management
From the *Institute of Medical Science, Faculty of Medicine, University strategies for LBP.1 The 1-month prevalence of LBP is
of Toronto, Toronto, Ontario, Canada; †Ottawa St. Medical Centre,
Windsor, Ontario, Canada; ‡Institute for Research in Extramural reported to be between 35% and 37%, with a lifetime
Medicine, EMGO, VU University Medical Center, Amsterdam, The prevalence between 70% and 85%, peaking between 45
Netherlands; §Complementary Medicine Program, University of and 59 years of age.2,3 In the United States, back pain is
Maryland, College Park, MD; and ¶Institute for Work and Health,
Toronto, Ontario, Canada. the most common cause of disability in those under 45
Supported by the Cochrane Collaborative Back Review Group, the years of age4 and back sprains and strains represent
Canadian Institutes of Health Research, and the Natural Health Prod- about one fourth of work-related injuries resulting in lost
ucts Directorate, the latter two public funding bodies within Health
Canada (to J.J.G. with a postgraduate fellowship). workdays.5 Low back pain is the second most frequent
Adapted from Gagnier JJ, van Tulder M, Berman B, et al. Herbal cause of work absence in industrialized nations6 and is a
medicine for low back pain. Cochrane Database of Systematic Reviews frequent reason for visits to a physician.3,7 It has been
2006;2:CD004504.
The device(s)/drug(s) is/are FDA-approved or approved by correspond- estimated that in the United States back pain has associ-
ing national agency for this indication. ated direct costs of 20 billion dollars and indirect costs of
Federal funds were received in support of this work. No benefits in any between 75 and 100 billion dollars.4 In the United King-
form have been or will be received from a commercial party related
directly or indirectly to the subject of this manuscript. dom, costs associated with LBP are estimated to be over
One of the authors (C.B.) is coordinating editor of the Cochrane Back 500 million pounds.8 This amounts to a substantial pro-
Review Group. Editors are required to conduct at least one Cochrane ductivity and economic burden for healthcare systems in
review. Dr. Bombardier is not the first author of the review. Any editor
who is a reviewer is excluded from editorial decisions on the review in many industrialized countries.1
which they are contributors. Therefore, this involvement does not seem The traditional treatment of LBP includes medication,
to be a source of conflict of interest in the Back Review Group. tissue stimulation (e.g., TENS, ultrasound), rest, and or-
Address correspondence to Joel J. Gagnier, ND, MSc, Institute of Med-
ical Science, Faculty of Medicine, University of Toronto, Toronto, On- thotics (e.g., braces).9 Although systematic reviews sug-
tario, Canada; E-mail: j.gagnier@utoronto.ca gest that few of these have enough evidence to suggest

82
 Herbal Medicine for LBP • Gagnier et al 83

benefit, it does appear that acute LBP can be effectively Methods

managed by encouraging activity, reassurance, and Criteria for Considering Studies for This Review
short-term symptom control (analgesics or nonsteroidal
anti-inflammatory drugs [NSAIDs]).10 Effective manage- Types of Studies. Only randomized controlled trials (RCTs)
ment of chronic LBP consists of exercise therapy, behav- were included.
ioral treatment, and multidisciplinary treatment pro- Types of Participants. Adults (older than 18 years of age)
grams.11 Also, altering beliefs about back pain that are suffering from acute (less than or equal to 6 weeks), subacute
correlated with impairment or disability is one of the (⬎6 weeks and ⬍12 weeks) or chronic (⬎12 weeks) NSLBP.
major challenges in LBP management.4 LBP was defined as pain localized to the area between the costal
Research in complementary and alternative medicine margin or the 12th rib to the inferior gluteal fold. Nonspecific
(CAM) has blossomed in the past 10 years. Rigorous LBP indicates that no specific cause is detectable, such as infec-
literature is growing steadily and is subsequently clarify- tion, neoplasm, metastasis, osteoporosis, rheumatoid arthritis,
fracture, inflammatory process, or radicular syndrome.32
ing the validity of these techniques. Specifically, the num-
ber of randomized trials of complementary treatments Types of Interventions. For the purpose of the present review,
has been doubling approximately every 5 years12 and a herbal medicine was defined as a whole or part of a plant that
currently, the Specialized Trial registry of the Cochrane is used for medicinal purposes with an administration route
Collaboration Complementary Medicine field contains being oral (taken by ingestion) or topical. This definition does
over 6,500 randomized and controlled clinical trials. In not include smoked plant substances (e.g., Cannabis sativa),
individual chemicals derived from plants, or synthetic chemi-
addition, CAM-teaching institutions are now beginning
cals based on constituents of plants. However, Cannabis sativa,
to teach principles of evidence-based medicine and clin- or additional plants which can be smoked, will be considered
ical epidemiology.13,14 These initiatives are well placed, herbal medicines for this review if orally ingested. Various
given the large number of visits to CAM practitioners.15 forms of oral herbal medicine include: standardized extracts
A recent population survey in Canada found that 2.2 (encapsulated or tablet form), tinctures (alcohol, glycerin, etc.),
million women and 1.6 million men visited a CAM prac- dried herb (encapsulated or tablet form), raw whole herb, in-
titioner during 1998/1999.15 Of these, it was found that fusion (e.g., tea), and decoction (e.g., boiled down tea). Various
more than 26% of patients who reported chronic pain forms of topical herbal applications include: ointments, essen-
visited an alternative practitioner during the previous tial oils, creams (petroleum or glycerin based), powders, plas-
year, compared with only 15% who did not report ters, and poultices. Opioids will be excluded from this review
chronic pain. Those who reported back pain had a higher given that they bridge the definitions of herbal medicine and
analgesic.
percentage of visits to alternative practitioners than any
other pain condition. More specifically, 37% of the in-
Types of Outcome Measures
dividuals who reported back pain visited an alternative
practitioner compared with only 17% of the entire pop- 1. Pain intensity, e.g., visual analogue scale (VAS), numer-
ulation.15 Similar percentages have been found in sur- ical rating scale (NRS), and proportion of pain-free pa-
veys conducted in the United States.16 –18 tients.
Several herbal medicines have been reported as treat- 2. Back pain specific functional status expressed by vali-
ments for varying types of pain. These include: Cam- dated instruments, e.g., Roland Disability Questionnaire
(RDQ) and Owestry Disability index (ODI).
phora molmol, Capsicum frutescens, Salix alba, Male-
3. Overall improvement (% subjective improvement,
luca alternifolia, Angelica sinensis, Aloe vera, Thymus NRS).
officinalis, Menthe peperita, Arnica montana, Curcuma 4. Return to work or work status (% of population, num-
longa, Tancaetum parthenium, Harpagophytum proc- ber of days of absenteeism).
umbens, and Zingiber officinicalis.19 Many of these
herbs have been subject of extensive biochemical re- Search Strategy for Identification of Studies. The follow-
search leading to delineation of their pharmacology and ing databases were searched:
physiologic effects.20 For example, the mechanism of C.
1. MEDLINE (1966 to July 2005).
frutescens is partially related to its ability to deplete sub-
2. EMBASE (1980 to July 2005).
stance P.21 S. alba is a platelet inhibitor and analgesic, 3. Cochrane Controlled Trials Registry (Issue 3, 2005).
and H. procumbens has analgesic and anti-inflammatory 4. Clinical Evidence (January 2005).
properties.22 In addition, some of these herbal species
have been clinically tested for the relief of symptoms of The search strategy used for spinal disorders for both MEDLINE
LBP.22–31 and EMBASE was that recommended by the Cochrane Back Re-
view Group33,34 (Appendix A, available for viewing online
Given the large public health and economic burden
through ArticlePlus only). Terms for LBP, RCTs, and herbal med-
LBP causes and the large number of such sufferers who icine interventions were combined. Reference lists in review arti-
regularly visit CAM practitioners, a systematic review of cles, guidelines, and in retrieved trials were reviewed for addi-
these practices is warranted. The objectives were to de- tional trials. Additionally, individuals with expertise in herbal
termine the effectiveness of herbal medicine compared medicine and LBP were contacted by J.J.G. to isolate additional
with placebo, no intervention, or other interventions in trials. Non-English articles were translated and discussed by J.J.G.
the treatment of nonspecific LBP (NSLBP). and M.W.T. following the same procedures described below.
84 Spine • Volume 32 • Number 1 • 2007

Study Selection. One reviewer (J.J.G.) conducted the elec- Table 1. Methodologic Quality Criteria and
tronic searches using the above search strategy in all databases. Operational Definitions
Two reviewers (J.J.G., M.W.T.) selected studies based on title,
abstract, and keywords. Studies that met the inclusion criteria Internal Validity Criteria
(see types of studies, participants, interventions, and outcomes 1. Was the method of randomization adequate?
2. Was the treatment allocation concealed?
above) were included in the review. If it was not clear from title 3. Were the groups similar at baseline regarding the most important
and abstract if a study fulfilled the inclusion criteria, a full copy prognostic indicators?
was retrieved for final selection. A consensus method was used 4. Was the patient blinded to the intervention?
to resolve disagreements about inclusion of studies. 5. Was the care provider blinded to the intervention?
6. Was the outcome assessor blinded to the intervention?
7. Were cointerventions avoided or similar?
Methodologic Quality Assessment. Methodologic quality 8. Was the compliance acceptable in all groups?
was independently assessed by two reviewers (J.J.G., M.W.T.). 9. Was the dropout rate described and acceptable?
Given one of the reviewers’ (J.J.G.) familiarity with the litera- 10. Was the timing of the outcome assessment in all groups similar?
11. Did the analysis include an intention-to-treat analysis?
ture, studies were not blinded for authors, institution, and jour-
nal. The 11 items reflecting internal validity, together with op- Operational Definitions of Internal Validity Criteria
1. A random (unpredictable) assignment sequence. Examples of
erational definitions, recommended in the update of the adequate methods are computer-generated random no. table and
method guidelines for systematic reviews by the Cochrane Back use of sealed opaque envelopes. Methods of allocation using date
Review Group were used to assess methodologic quality.33 of birth, date of admission, hospital numbers, or alternation should
Each criterion can be scored as “yes,” “no,” or “don’t know.” not be regarded as appropriate.
2. Assignment generated by an independent person not responsible for
The score of “yes” reflects the fulfillment of that criterion. The determining the eligibility of the patients. This person has no
scoring of “no” reflects lack of fulfillment of that criterion. The information about the persons included in the trial and has no
scoring of “don’t know” reflects an impossibility to determine influence on the assignment sequence or on the decision about
if this criterion was fulfilled or not. A trial was considered high eligibility of the patient.
quality if more than 50% of internal validity items scored pos- 3. In order to receive a “yes,” groups have to be similar at baseline
regarding demographic factors, duration and severity of complaints,
itively, i.e., scores of 6, 7, 8, 9, and 10 of 11 (Table 1). percentage of patients with neurologic symptoms, and value of main
outcome measure(s).
Data Extraction. Two reviewers (J.J.G., M.W.T.) extracted 4. The reviewer determines if enough information about the blinding is
the data from each trial using a standardized form. The follow- given in order to score a “yes.”
5. The reviewer determines if enough information about the blinding is
ing data were extracted from each study: recruitment, charac- given in order to score a “yes.”
teristics of the study population (age, gender), setting (e.g., 6. The reviewer determines if enough information about the blinding is
year, country of origin), duration of LBP (acute, subacute, or given in order to score a “yes.”
chronic), previous treatment for LBP, number of patients ini- 7. Cointerventions should either be avoided in the trial design or
similar between the index and control groups.
tially recruited, number of patients randomized, number of 8. The reviewer determines if the compliance to the interventions is
dropouts or withdrawals, duration of intervention, type of acceptable, based on the reported intensity, duration, and no. and
herbal medicine used (plant name and form of delivery and frequency of sessions for both the index intervention and control
dosage), standardization information (e.g., percentage of active intervention(s).
constituent per delivery unit), characteristics of the control in- 9. The no. of participants who were included in the study but did not
complete the observation period or were not included in the
tervention (type and duration), types of outcome measures, analysis must be described and reasons given. If the percentage of
summary statistics, timing of outcome assessments, compli- withdrawals and dropouts does not exceed 20% for short-term
ance, adverse effects due to intervention, and author’s conclu- follow-up and 30% for long-term follow-up and does not lead to
sions as to the effectiveness of the intervention. substantial bias a “yes” is scored. (Note: These percentages are
arbitrary, not supported by the literature.)
10. Timing of outcome assessment should be identical for all
Clinical Relevance. The clinical relevance of each study was intervention groups and for all important outcome assessments.
independently assessed by two reviewers. The following five 11. All randomized patients are reported/analyzed in the group they
questions were used: were allocated to by randomization for the most important moments
of effect measurement (minus missing values) irrespective of
1. Are the patients described in detail so that you can decide noncompliance and cointerventions.
whether they are comparable to those that you see in
your practice?
2. Are the interventions and treatment settings described 2. Moderate evidence: consistent findings among multiple
well enough so that you can provide the same for your low-quality RCTs and/or one high-quality RCT.
patients? 3. Limited evidence: one low-quality RCT.
3. Were all clinically relevant outcomes measured and re- 4. Conflicting evidence: inconsistent findings among multi-
ported? ple RCTs.
4. Is the size of the effect clinically important? 5. No evidence from trials: no RCTs.
5. Are the likely treatment benefits worth the potential
Sensitivity analyses were carried out exploring the results when
harms?
definitions of high quality trials were 40% and 60% fulfillment
of internal validity criteria. Also, the validity criteria that are
Data Analysis. Because of insufficient data and clinically het- scored as “don’t know” was assumed to be positive for an
erogeneity, a qualitative analysis was conducted using a rating
additional sensitivity analysis. Subgroup analyses were con-
system consisting of 5 levels of evidence:
ducted separately for acute LBP (lasting ⬍6 weeks), subacute
1. Strong evidence: consistent findings among multiple LBP (lasting from 6 –12 weeks), and chronic LBP (lasting
high-quality RCTs. longer than 12 weeks).
 Herbal Medicine for LBP • Gagnier et al 85

Table 2. Included Studies

Study Methods Participants

Frerick et al 27 RCT with two groups; period: 320 subjects with chronic nonspecific low back pain divided equally between capsicum plaster group
3 wk and placebo group

Ginsberg and Famaey28 RCT with 40 patients assigned to 40 patients with acute mechanical low back pain were assigned to either the Rado-Salil (a
1 of 2 groups; period 14 days capsicum-containing cream) group (N ⫽ 20) or a placebo group (N ⫽ 20). Each patient was also
given 45 paracetamol 250 mg tablets. No other analgesic, anti-inflammatory drug, or physical
treatment was allowed during the 12-wk period. Method of patient selection: clinical examination,
standard radiologic examination of the lumber spine, routine laboratory tests

Keitel et al 29 RCT with two groups; no report 154 patients were randomly allocated to a placebo plaster group (N ⫽ 77) and a capsicum plaster
of randomization method; group (N ⫽ 77). A total of 132 patients completed the study with data being available for the
period: 1 plaster per day at intention-to-treat (ITT) analysis on 150 patients (P ⫽ 0.002). A total of 22 exclusions occurred due
maximum pain site for 4 to 12 to premature discontinuation of the treatment (N ⫽ 19), failure to meet the inclusion criteria (N ⫽
hr for 3 wk 2), or unauthorized concurrent treatment (N ⫽ 1). Inclusion criteria: subjective back pain rating of
5 or more on an 11-grade visual analogue scale as well as a duration of back pain for a minimum
of 3 mo at enrollment. Exclusion criteria: alcohol abuse, drug dependence, forms of specific back
pain, concomitant systemic inflammatory rheumatic condition, no concurrent therapy for back pain

Stam et al 31 RCT with two groups (no 161 subjects who were randomly allocated to either group. A total of 6 subjects were lost to follow-
placebo); randomization was up (SLR ⫽ 2; CCC ⫽ 4). 21 subjects met all per protocol criteria. Inclusion criteria: between 18
performed using RCODE and 65 yr of age, acute attack of low back pain within previous 72 hr, free from back pain during
software (version 3.4) in the previous 3 mo, at least moderately painful limitation of movement on physical examination.
blocks of four; period: 7 days Exclusion criteria: radicular symptoms, pain above T12, rheumatoid arthritis, ankylosing spondylitis,
known hypersensitivity to treatment compounds, use of analgesics other than paracetamol during
the treatment period, use of NSAIDs during the treatment period, receiving other treatment for
acute low back pain, pregnancy, more than 96 hr elapsed since onset of pain, including washout
for analgesic and/or NSAIDs
Chrubasik et al 22 RCT with two groups; patients 118 patients were allocated to a Harpagophytum (H) group (N ⫽ 59) and a placebo (P) group (N ⫽
were placed in groups by 59). 109 completed the trial (H, N ⫽ 54; P, N ⫽ 55). Inclusion criteria: between 18 and 75 yr of
random no. allocation; period: age, at least 6 mo of low back pain not attributable to identifiable causes, suffering from acute
4 wk increases in pain, and who were expected to require at least 4 wk of symptomatic treatment.
Exclusion criteria: participation in other clinical studies or had done so in past 30 days, pregnancy,
lactation, insufficient contraception, difficulties with language or cooperation, known allergy to
proposed trial medication, history of drug or alcohol abuse, requirement of psychotherapeutic
agents, or a serious organic illness affecting any of the organ systems
Chrubasik et al 23 RCT with 3 groups; period: 4 wk 197 patients allocated to Harpagophytum at 600 mg (N ⫽ 65), or 1,200 mg (N ⫽ 66) or matched
placebo (N ⫽ 66). Inclusion criteria: between 18 and 75 yr of age, 6 mo of nonspecific low back
pain, a current exacerbation of their complaint that was effecting both rest and movement, which
was giving rise to pain greater than 5 on a 1–10 VAS and was expected to require at least 4 wk
of symptomatic treatment. Exclusion criteria: current or recent participation in any other clinical
study, serious organic illness effecting any organ system, a history of drug or alcohol abuse or
requirement for psychotherapeutic agents, pregnancy (actual or possible), or lactation, known
allergy to any the proposed trial medications, difficulties with language or anticipated cooperation
Chrubasik et al 24 RCT with three groups; no report Patients were recruited from the Haifa area in Israel between May and November. 210 patients were
of randomization method; randomized into three groups (N ⫽ 70 in each group). 191 completed the trial (P, N ⫽ 59; 120;
period: 4 wk N ⫽ 67; 240; N ⫽ 65). Inclusion criteria: between 18 and 75 yr of age, at least 6 mo of intermittent
low back pain that was not attributable to identifiable causes, a current exacerbation of their
complaint at rest and with movement that caused pain of at least 5 of 10 on a VAS and that was
expected to require at least 4 wk of treatment. Exclusion criteria: participation in other clinical
studies or had done so in past 30 days, pregnancy, lactation, insufficient contraception, difficulties
with language or cooperation, known allergy to proposed trial medication, history of drug or
alcohol abuse, requirement of psychotherapeutic agents

Chrubasik et al 25 Open RCT with two groups 228 individuals divided equally in to two groups (N ⫽ 114 per group). Inclusion criteria: between 18
comparing an herbal medicine and 80 yr of age, 6 at least 6 mo nonspecific low back pain. Exclusion criteria: recent trauma, a
(Salix alba) to a synthetic history of cancer or risk factors for spinal infection, unexplained weight loss or recent fever or
anti-rheumatic (rofecoxib); chills, pain exacerbated by being supine or severe nocturnal pain, perineal anesthesia, recent
period: 4 wk onset of bladder dysfunction or severe progressive neurologic deficit in the lower extremity,
recent participation in other clinical trial, serious organic illness affecting any organ system, a
history of drug or alcohol abuse or requirement for psychotherapeutic drugs, pregnancy or
lactation, known allergy to salicylates, difficulties with language or expected cooperation
Chrubasik et al 26 RCT with two groups; period: 88 subjects allocated to Harpagophytum procumbens (N ⫽ 44) group or rofecoxib (N ⫽ 44). Inclusion
6 wk criteria: between 45 and 75 yr of age, at least 6 mo of nonspecific low back pain, current
exacerbation of complaints for 8 wk that was affecting both rest and movement, was causing pain
of at least 5 of 10 on a visual analogue scale, and was judged to require symptomatic treatment
for 6 wk. Exclusion criteria: red flags for low back pain, participation in any other clinical study
within the last 30 days, serious organic illness affecting any organ system, a history of drug or
alcohol abuse or requirement of psychotherapeutic drugs, pregnancy, or lactation, known allergies
to trial medication, and anticipated difficulties with language or cooperation

Krivoy et al 30 35 subjects randomized to two 51 subjects with 19 in the salix alba group, 16 in a placebo group, and 16 in an acetylsalicylate
groups and a further 16 acted group. Inclusion criteria: acute exacerbations of chronic low back pain, stable ischemic heart
as controls; period: 4 wk disease. Exclusion criteria: NSAID use
86 Spine • Volume 32 • Number 1 • 2007

Table 2. Continued
Interventions Outcomes Notes

Topical plaster containing an ethonolic Outcomes: Arhus Low Back Pain Rating Scale, global Adverse effects: 14 participants in each group. GI: 8 in the Devil’s
extract of cayenne pepper assessment of efficacy by patient and investigator, claw group, 9 in the Vioxx group, which tended to be more
standardized to 22 ␮g/cm2 of global assessment of safety by patient and severe. Two serious adverse events occurred in the Devil’s
capsaicinoids or placebo plaster investigator claw group but were judged as being unrelated to the study
medication. Circulatory and laboratory variables were not
affected by either treatment
Rado-Salil ointment (containing 17.64 mg Outcomes: pain evaluation on a 10-cm linear scale, Adverse events: Pruritus, 1 in placebo, 1 in Rado-Salil group.
ethylsalicylate, 26.47 mg duration of confinement to bed, muscular reflex Local erythema and burning, 3 in the Rado-Salil group
methylsalicylate, 8.82 mg contracture evaluation by the physician on a scale
glycosalicylate, 8.82 mg salicylic acid, of 0–4, and spine mobility by determination of
4.41 mg camphor, 55.14 mg menthol, Schober’s index, the finger to floor distance, the
and 15.44 mg capsicum oleoresin per degree of lumbar extension, global appreciation of
1g) in the form of a 40-g stick applied treatment by patient and physician
as needed or a placebo (containing
only the excipient with three times
the amount of lavendula and
bergamot essences) matched for
appearance
Topical plaster-type application of Primary outcome measure: Arhus Low Back Rating Adverse events: A total of 24 adverse events were reported
Capsicum frutescens (Cayenne) Scale. Secondary outcome measures: global (C ⫽ 15; P ⫽ 9). Most of these were warmth and itching
containing 12 mg of capsaicinoids assessment of efficacy and tolerance by physician locally. The C group had 5 cases of severe adverse events
per plaster. Matched placebo plaster and patient (inflammatory contact eczema, urticaria, minute hemorrhagic
spots, and fasciculation or dermatitis) and the P group had
two such cases (fasciculation or allergic dermatosis). A total
of 16 patients withdrew because of adverse events (C ⫽ 10;
P ⫽ 6). Also, 95.9% of the C group and 48.7% of the P group
experienced sensations of warmth locally. Pruritus was
mentioned in 45.9% of the C group and 31.6% of the P group
Spiroflor SLR homeopathic gel (SLR) Primary outcome: reduction in visual analogue scale Adverse events: Approximately 12% of SLR and 26% of the CCC
group (N ⫽ 83) or a Cremor Capsici (VAS) scores for pain (100-mm scale) and the group experienced an adverse event. Adverse drug reactions
Compositus FNA, the capsici proportion of treatment success (a VAS reduction were reported by 4% of the SLR group and 24% of the CCC
oleoresin gel (CCC) group (N ⫽ 78). of at least 80%). Secondary outcome measures: group. A total of four adverse drug reactions in the CCC group
Each of the gels was applied at 3 g proportion of subjects using paracetamol, no. of and none in the SLR group were considered “severe.” A total
per day nights with disturbed sleep, duration of absence of 8 patients in CCC group and 0 in the SLR group withdrew
from work, and overall assessments of treatment due to adverse drug reactions
efficacy/usefulness by the general practitioners
(GP) and the patients
Oral form of Harpagophytum procubens Primary: cumulative requirement for Tramadol (an Adverse events: A total of 4 adverse effects occurred in the
(Devil’s claw) standardized to a oral opiate-based analgesic) over the last 3 wk of Harpagophytum group with only two of these potentially being
dosage of 50 mg harpagoside per day the study period. Secondary: no. of pain-free due to the treatment (i.e., repeated coughs and tachycardia). A
or 2,400 mg of the crude extract. patients based on a 5-point visual rating scale and total of 10 adverse events occurred in the placebo group
Matched placebo the Arhus Low Back Rating Scale

Harpagophytum Extract WS 1531 600 mg Primary outcome: proportion of pain free patients Adverse events: A total of 4 adverse effects occurred in the
(50 mg harpagoside), 1,200 mg (100 without tramadol for at least 5 days during the Harpagophytum group with only two of these potentially being
mg harpagoside) last week of treatment. Secondary outcomes: due to the treatment (i.e., repeated coughs and tachycardia). A
Arhus Low Back Rating Scale, percentage total of 10 adverse events occurred in the placebo group
requiring tramadol, verbal pain ratings

Extract of dry willow bark (Salix alba): Primary outcome: the proportion of patients who Adverse events: one adverse reaction (exanthem, swollen eyes,
120 mg salicin, 240 mg salicin. responded to treatment by being pain free without pruritus) could be attributed to the 120-mg willow bark extract
Matched placebo tramadol for at least 5 days during the last week group. A total of 2 patients in the 240-mg group reported
of treatment. Secondary outcome: Arhus Low short-lasting adverse events (dizziness attributed to tramadol,
Back Rating Scale scores dizziness, and fatigue). These patients dropped out for
seemingly unrelated reasons. A total 6 adverse events were
reported in the placebo group, including 3 attributable to
tramadol (dizziness/headache, dizziness/vomiting/diarrhea, dry
mouth) and the 3 others reported mild abdominal pain, 2 of
whom dropped out on the first day of the trial
A proprietary extract of Salix alba called Pain on a VAS, modified Arhus Low Back Rating Adverse events: 23 in the Salix alba group (13 of gastrointestinal
Assalix at 4 capsules per day Scale, its pain component and the total pain origin, 5 cutaneous allergy, remaining undefined) and 27 in the
providing a total of 240 mg of salicin index, physician and patient-rated success and rofecoxib group (17 GI effects, 1 asthma, rest undefined). GI
per day, or a single 12.5-mg tablet of the acceptability of the treatment on a verbal adverse events were judged as more severe and caused more
rofecoxib per day scale (very good, good, moderate, poor) withdrawals in the rofecoxib group

Harpagophytum procumbens in a Primary outcome: proportion of patients who Adverse effects: 14 participants in each group. GI: 8 in the Devil’s
proprietary aqueous extract called recorded “no pain” without using tramadol for at claw group, 9 in the Vioxx group, which tended to be more
Doloteffin (standardized to contain 60 least 5 days in the final week or treatment. severe. Two serious adverse events occurred in the Devil’s
mg harpagoside) or 12.5 mg rofecoxib Secondary and other outcomes: proportion of claw group but were judged as being unrelated to the study
per day patients in whom the averaged daily pain scores medication. Circulatory and laboratory variables were not
in the 6th week had decreased by 20% to 50% of affected by either treatment
the average in the first week; the percentage
change from baseline of a modified Arhus Low
Back Rating Scale; the percentage change from
baseline on the Health Assessment Questionnaire;
tramadol requirement
786.48 mg twice per day of an ethanol Primary outcome: platelet aggregation Adverse events: none reported
extract of the bark of salix
daphnoides (240 mg salicin content
per day), matched placebo, or 100 mg
acetylsalicylate
 Herbal Medicine for LBP • Gagnier et al 87

Results rofecoxib. Of those trials using White willow bark ex-

tract: Chrubasik et al 25 used an extract containing 0.153
Study Selection
mg salicin per mg and made comparisons between daily
A total of 295 references were isolated from MEDLINE,
dose of 120 mg salicin, 240 mg of salicin and a matched
EMBASE, Cochrane Controlled Trials Registry, Clinical
placebo, Chrubasik et al 26 used S. alba containing a daily
Evidence searches and through contacting content experts. dose of 240 mg salicin and compared it with 12.5 mg
After reviewing abstracts, 266 citations were excluded ac- rofecoxib, and Krivoy et al 30 used a daily dose of S. alba
cording to predetermined exclusion criteria. A total of 29 containing 240 mg salicin compared with placebo and
papers were retrieved in full, of which 19 were excluded 100 mg acetylsalicylate. The four trials using topical C.
due to improper study design (not an RCT). A total of 10 frutescens preparations used: a topical plaster applica-
citations met inclusion and exclusion criteria.22–31 Details tion containing 11 mg of capsaicinoids per plaster,29 a
of the included trials are outlined in Table 2. plaster containing an ethonolic extract of cayenne pep-
Ten studies were included in this review. Three studies per standardized to 22 ␮g/cm2 of capsaicinoinds,31 a gel
used an oral form of the herbal species H. procumbens called Cremor Capsici Compositus FNA, which contains
(Devil’s claw),22–24 3 used oral S. alba (White willow 100 mg of capsici oleoresin (BPC), 10 g of glycol salicy-
bark),25–27 and 4 used topical C. frutescens (Cayenne).28 –31 late, 1 g of methylnictinate, and a combined 1 g of his-
Four studies compared various oral herbal medicines tamine hydrochloride, sorbitol, methylprahydroxyben-
with placebo.22–23,25,27 Two studies compared oral zoate, triethanolamine, lanette wax, stearic acid, and
herbal medicines to standard pain medications.24,26 purified water,30 or another gel, Rado-Sailil, containing
Three studies compared topical herbal medicines to pla- 17.64 mg ethysalicylate, 26.47 mg methylsalicylate, 8.82
cebo.28 –29,31 One study compared a topical herbal med- mg glycosalicylate, 8.82 mg salicylic acid, 4.41 mg cam-
icine to a topical homeopathic medicine.30 phor, 55.14 mg methanol, and 15.44 mg capsici oleo-
The three studies using H. procumbens included pa- resin per 1 g.28
tients with acute exacerbations of chronic NSLBP.22–24 Nine of the 10 studies reported adverse events associ-
Similarly, the three trials using S. alba preparations in- ated with the study medication. One trial did not report
cluded homogeneous populations with acute episodes of any information regarding adverse events. All details of
chronic NSLBP.25–27 One of the trials using a topical C. each trial are reported in the table of included studies.
frutescens ointment included patients with acute me- Conflict of interest was assessed in an informal way by
chanical LBP.28 Two additional trials using a Capsicum discussion between the 2 reviewers. We looked at fund-
pain plaster included participants with chronic ing sources (public vs. private) and whether an author
NSLBP.29,31 One trial using a topical Capsicum oint- was employed by a private pharmaceutical/nutraceutical/
ment included a sample of patients with either newly herbal medicine manufacturer. Two trials reported no
occurring acute LBP or acute episodes of chronic LBP.30 conflicts of interest.23,28 A conflict of interest was
Three of these studies23–24,31 used a relatively un- deemed possible in five trials.22,24 –25,29 –31 In three trials,
known LBP scale, the Arhus index, which was designed an author was employed by a pharmaceutical compa-
to monitor outcomes of clinical trials of LBP. The Arhus ny,29 –31 one trial was funded by a professional acade-
index is a back pain specific index, including physical my,25 one trial was funded by a pharmaceutical compa-
impairment, pain, and disability scores, which are ny,24 and for one trial the experimental herbal medicine
summed into a total score.35 The pain scale includes back was supplied by a company.22 For the additional two
pain and leg pain and scores range from 0 to 60 and is trials,26,27 conflict of interest was determined to be un-
rated by the patient. The disability scale consists of a likely.
questionnaire that asks about 15 daily tasks, and scores
range from 0 to 30. The physical impairment score is Methodologic Quality
obtained by scoring on a deep knee bend, a modified Several ratings for the fulfillment or not of methodologic
Schober’s test, a low back strength test, and a measure of quality criteria for individual studies varied between re-
analgesic use with a total combined score from 0 to 40. viewers initially but were resolved without need for a
The higher the scores, the more physical impairment, third part assessor. The agreement between the two re-
pain, and disability. This test takes approximately 15 viewers (J.J.G., MVT) regarding total methodologic
minutes to complete. It has been shown to be a valid and quality scores for individual trials was good, as indicated
reliable measure of LBP.35 by a Cohen’s kappa statistic of 0.66. Contact with one
Of those trials using H. procumbens: Chrubasik et primary author clarified one “don’t know” into a “yes”
al 22 used a standardized dosage of 50 mg harpagoside for a single trial.24
per day or 2,400 mg of the crude extract, Chrubasik et The mean score for methodologic quality of all in-
al 23 used daily dosages of the proprietary extract WS cluded studies was 6.6 (median, 7; range, 5–9) (Table 3).
1531 at 600 and 1,200 mg crude herb, which was the Using a cutoff point of 6 of 11 criteria, 8 of the 10 trials
equivalent of 50 and 100 mg harpagoside, and Chru- (80%) were of high quality.22–26,29 –31
basik et al 24 used a proprietary extract of Doloteffin, The main methodologic shortcomings of the Harp-
containing a daily dose of 60 mg harpagoside, or 12.5 mg agophytum trials included a lack of reporting of alloca-
88 Spine • Volume 32 • Number 1 • 2007

Table 3. Methodologic Quality of Included Trials

Ginsberg
Chrubasik Chrubasik Chrubasik Chrubasik Krivoy Chrubasik and Keitel Stam Frerick
Criteria et al 22 et al 26 et al 23 et al 24 et al 30 et al 25 Famaey28 et al 29 et al 31 et al 27

1. Was the method of randomization Y Y Y Y DK Y DK DK Y Y

adequate?
2. Was the treatment allocation DK DK DK DK DK Y DK DK Y Y
concealed?
3. Were the groups similar at Y Y Y N DK Y DK DK Y DK
baseline regarding the most
important prognostic indicators?
4. Was the patient blinded to the Y Y Y Y Y N Y Y Y DK
intervention?
5. Was the care provider blinded to Y Y Y Y Y N DK DK Y DK
the intervention?
6. Was the outcome assessor Y Y Y Y Y N Y DK Y DK
blinded to the intervention?
7. Were cointerventions avoided or DK DK DK DK DK Y DK Y DK DK
similar?
8. Was the compliance acceptable DK DK DK DK DK DK DK Y N Y
in all groups?
9. Was the dropout rate described Y Y Y Y Y Y Y Y Y Y
and acceptable?
10. Was the timing of the outcome Y Y Y Y DK Y Y Y Y Y
assessment in all groups similar?
11. Did the analysis include an N Y Y Y Y N Y Y Y Y
intention to treat?
Y, yes; N, no; DK, don’t know.

tion concealment, compliance rates, controls for cointer- suffering from acute exacerbations of chronic LBP of
ventions, and if the withdrawal or dropout rate was greater than 6 months.
acceptable during the follow-up period. Of the Salix tri-
als, one was an open-label trial and the additional two
Chronic LBP
did not report allocation concealment, compliance rates,
controls for cointerventions, or if the withdrawal or 50 mg Harpagoside Dose. A total of 325 patients were in-
dropout rate was acceptable during follow-up period. cluded in two 4-week trials, which tested extracts of H. proc-
The Capsicum trial by Stam et al 31 had excellent meth- umbens standardized to 50 mg harpagoside (H) per day versus
odologic quality. The additional Capsicum trials failed placebo.23,24 Both trials found a significant increase in the
to report the type of randomization that was done, allo- number of pain-free patients in the 50 mg H group (9%–17%)
cation concealment, similarity of baselines, outcome as- over placebo (2%–5%). One trial found that, for patients tak-
sessor, and investigator and participant blinding, if ing 50 mg H, the percentage with no pain or mild LBP increased
cointerventions were comparable, and if compliance was over the 4-week period (from 2% in week 1 to 24% in week 4),
acceptable. Details regarding the fulfillment of specific whereas the percentage with unbearable or severe pain de-
criteria for each individual trial are contained in the creased over the 4 weeks (from 59% in week 1 to 35% in week
4).23 Although tramadol consumption was decreased in both
Table 3.
trials in the 50 mg H group, compared with placebo, this did
Clinical Relevance not reach statistical significance in one trial23 and the other trial
There were several disagreements on ratings of clinical did not perform a statistical test on this measure.22 Both trials
relevance between reviewers (J.J.G., MVT), but these used the Arhus index, a back pain specific index, including
were easily resolved during the consensus meeting. Sev- physical impairment, pain, and invalidity scores, which are
summed into a total score.32 The overall Arhus improved by
eral trials met all five clinical relevance criteria.22,25,28 Of
21% in both the 50 mg group and the placebo group with no
the trials testing H. procumbens, two trials did not meet difference between groups. The pain subscale was significantly
items four and five.23–24 Of the S. alba trials, one did not improved in favor of the 50 mg H group in both trials (median
meet item 1,27 one did not meet items 4 and 5,26 and it change for those with current LBP of 43%, 23; median change
was not possible to tell if one trial fulfilled or not items 4 of 34%, 22), which was a greater improvement than that of the
and 5.27 Of the C. frutescens trials, two did not meet item additional 100 mg H group in one trial (median change for
1,28,29 two did not meet items 4 and 5,30,31 and for one those with current LBP of 37%, 23).
trial determination of fulfillment or not was not possible Therefore, these findings, when considered together with
for items 4 and 5.29 methodologic quality, suggest that there is strong evidence that
50 mg harpagoside per dose of an aqueous extract of H. proc-
Herbal Medicine for LBP umbens per day reduces pain more than placebo in the short-
H. procumbens (Devil’s claw) versus placebo. Three tri- term in patients with acute episodes of chronic NSLBP. Long-
als were identified, all of which included participants term treatment data are not yet available.
 Herbal Medicine for LBP • Gagnier et al 89

100 mg Harpagoside Dose. A total of 197 patients were required rescue medication (i.e., tramadol) than the placebo
included in one 4-week trial, which tested H. procumbens stan- group, they did not provide any data.
dardized to 100 mg harpagoside per day versus placebo.23 The Therefore, the results of these trials together with their
number of pain-free patients for at least 5 days in the fourth methodologic quality, suggest that there is moderate evidence
week of treatment was significantly higher (n ⫽ 10) than in the that 240 mg salicin dose of an extract of S. alba reduces pain
placebo (n ⫽ 3) or lower dose (50 mg H) groups (n ⫽ 6). Half more than placebo and 120 mg dose of S. alba for the treatment
of the pain-free patients in the 100 mg H group had a neuro- of acute episodes of chronic NSLBP in the short term.
logic deficit at the start of the trial. The change from baseline in
the overall Arhus index, the pain index, invalidity index, and C. frutescens Versus Placebo
physical impairment index were not different between the three Two trials were identified, one with acute LBP,28 al-
groups. The percentage of patients with none or mild LBP in- though the actual duration of LBP was not described,
creased over the 4-week period, whereas the percentage with and one with chronic LBP, greater than 3 months.29
unbearable or severe pain decreased.
Therefore, these results, when considered with methodologic Acute LBP. In the trial by Ginsberg and Famaey,28 40
quality, suggest that there is moderate evidence that 100 mg subjects with acute mechanical LBP were given either a
harpagoside per dose of an aqueous extract of H. procumbens salicylate and Capsicum-containing cream called Rado-
leads to a higher number of pain-free patients for at least 5 days in Salil (N ⫽ 20) or a placebo cream containing bergamont
the fourth week of treatment of acute episodes of chronic NSLBP. and lavender (N ⫽ 20) for a period of 14 days. At day 3,
Superiority of the higher dose has not been shown. there was an improvement in pain score in the Rodo-Salil
group by almost 2 cm on the VAS, which was signifi-
S. alba Versus Placebo cantly better than the placebo group. By day 14, the
Two trials were identified, both of which included par- improvement increased to 3.79 cm, which was also sig-
ticipants suffering from acute exacerbations of chronic nificantly greater than the placebo group. Additionally,
LBP of greater than 6 months. both patients and physicians rated the effect of Rado-
Salil more favorably than the placebo group.
Chronic LBP
These results together with the methodologic quality
120 mg Salicin Dose. A total of 210 patients were included in of this trial suggest that there is limited evidence that
a 4-week trial, which tested two doses of S. alba, standardized Rado-Salil cream reduces pain more than placebo for the
to either 120 mg of 240 mg salicin (S) per day, against placebo treatment of newly occurring episodes of acute NSLBP in
(n ⫽ 70 for each group; 25). The number of pain-free patients the short term (14 days).
for at least 5 days in the fourth week of treatment increased
Chronic LBP. Keitel et al 29 used 154 patients with acute
from baseline in the placebo (n ⫽ 4), 120 mg S group (n ⫽ 15),
and the 240 mg S group (n ⫽ 27) with the trend for dose being episodes of chronic NSLBP, who were randomly allo-
significant. The number of patients requiring relief medication cated to a placebo plaster group (N ⫽ 77) and a Capsi-
(tramadol) during each week decreased to 33 during week 4 for cum plaster group (N ⫽ 77) for a period of 3 weeks. A
the placebo group, 10 for the 120 mg S group, and 3 for the 240 reduction in pain by at least 30% was achieved in 60.9%
mg S group with the trend for dose being significant. The total of the Capsicum group and 42.1% in the placebo group
Arhus index, pain index, invalidity index, and physical impair- and a reduction in pain by at least 50% in 35.1% and
ment index did not change from baseline for the placebo group 17.1%, respectively. The total Arhus score improved in
but improved in the 120 mg and 240 mg salicin groups, with the Capsicum group by 38.5% and by 28% in the pla-
the trend for dose being significant and the 240 mg S group cebo group (P ⫽ 0.002). Physician global ratings of effi-
showing superiority over the 120 mg S group for the total
cacy were considered “excellent” or “good” in 75.7% of
Arhus index score (P ⬍ 0.02) and the pain index (P ⬍ 0.001).
Therefore, these results, when considered with the method-
the Capsicum group and 47.4% of the placebo group.
ologic quality of this trial, suggest that there is moderate evi- After treatment 13.5% of the Capsicum group and 6.6%
dence that 120 mg salicin dose of an extract of S. alba leads to of the placebo group were completely symptom-free.
more pain-free patients for the treatment of acute episodes of Compliance was 90.6% in the Capsicum group and
chronic NSLBP in the short term. 88.1% in the placebo group.
In the study by Frerick et al,27 320 participants with
240 mg Salicin Dose. A total of 261 patients were included in chronic NSLBP were randomly allocated to a placebo
two trials.25,27 Results for the Chrubasik et al trial25 are re- plaster group (N ⫽ 160) and a Capsicum plaster group
ported above but can be summarized as the 240 mg S group: (N ⫽ 160) for a period of 21 days. The intention-to-treat
increasing the number of pain-free patients during 5 days of the analysis included 319 case records and the (N ⫽ 159
fourth week of treatment, decreasing the number of patients Capsicum, N ⫽ 160 placebo), and the per-protocol anal-
requiring relief medication, and improving the Arhus index
ysis yielded 249 cases (N ⫽ 120 Capsicum, N ⫽ 129
scores with a trend for greater improvements with higher dose
for all outcomes, and significant differences between the 120
placebo). The total Arhus scale score decreased by 33%
mg and 240 mg S groups for the total Arhus index and the pain in the Capsicum group and by 22% in the placebo group
index of the Arhus scale. The additional trial by Krivoy et al,30 (P ⬍ 0.001). For the ITT analysis, the Arhus compound
which was designed to test platelet aggregation of S. alba ex- pain score (total out of 30) decreased by 42% (20.0 –
tract, did not employ clinically relevant outcomes. Although 11.5 of 30) in the Capsicum group and 31% (19.7–13.6
the authors stated that less patients in the 240 mg S group of 30) in the placebo group (P ⬍ 0.001). A reduction in
90 Spine • Volume 32 • Number 1 • 2007

pain by at least 30% was achieved in 67% of the Capsi- nonsteroidal anti-inflammatory drugs and/or tramadol
cum group and 49% in the placebo group and a reduc- was 10% for the S. alba group and 13% for the rofe-
tion in pain by at least 50% in 45% and 24%, respec- coxib group. Approximately 90% of physicians and pa-
tively. The reduction in current pain was 49% in the tients rated either treatment as effective and close to
Capsicum group and 37% in the placebo group (P ⬍ 100% rated either treatment as acceptable.
0.001). The Arhus subscale for physical impairment de- Therefore, the methodologic quality and results of
creased from baseline by 21% in the Capsicum group this trial suggest that there is moderate evidence that
and 10% in the placebo group (P ⬍ 0.001). The disabil- there are no differences in effectiveness between a
ity subscale decreased by 35% the Capsicum group and 240-mg salicin dose of an extract of S. alba and 12.5-mg
by 22% in the placebo group (P ⫽ 0.001). The Capsicum rofecoxib per day in treatment of acute episodes of
treatment was rated as either excellent or good by inves- chronic NSLBP in the short term.
tigators in 74% of cases compared with 36% for the
placebo group. Compliance was reported as being very C. frutescens Versus Homeopathic Treatment
good or good in both groups, with the Capsicum group Acute and Chronic LBP. The trial by Stam et al 31 used
compliance being 91% and placebo group 93%. 161 subjects who were a mixed group of new acute LBP
Therefore, these results together with the low method- patients and acute episodes of chronic LBP. Participants
ologic quality of these trials suggest that there is moderate were randomly allocated to either a Spiroflor SLR ho-
evidence that a plaster of C. frutescens reduces pain and meopathic gel (SLR) group (n ⫽ 83) or a Cremor Capsici
improves function more than placebo for the treatment of Compositus FNA, the capsici oleoresin gel (CCC) group
acute episodes of chronic NSLBP in the short term. (n ⫽ 78) for a period of 7 days. Each of the gels was
H. procumbens Versus Rofecoxib applied at 3 g per day. Both groups showed a significant
reduction in pain on the VAS scale with a decrease of
Chronic LBP. A total of 88 patients with acute episodes 38.2 mm in the SLR group and 36.6 mm in the CCC
of chronic NSLBP were included in a 6-week trial, which group. In the SLR group, 50% of subjects reported that
tested H. procumbens standardized to 60 mg harpago- treatment was 80% effective and 18% reported total
side per day versus 12.5 mg rofecoxib per day.24 There (100%) effectiveness. In the CCC group, this was 55%
were no statistically significant differences in the number and 15%, respectively. There were also no differences in
of pain-free patients for at least 5 days in the sixth week proportion of subjects using paracetamol, the propor-
of treatment in the 60 mg H group (10 of 44) than in the tion of subjects still unable to work at the end of the
rofecoxib group (5 of 44; P ⫽ 0.257). The number of study, and overall efficacy.
patients with improvements in pain scores was not dif- Therefore, there is moderate evidence that there are no
ferent between the two groups. The number of patients statistically significant or clinically relevant differences in
consuming rescue medication (tramadol) decreased from effectiveness between Spiroflor SLR homeopathic gel (SLR)
baseline in both groups but did not differ between groups and Cremor Capsici Compositus FNA gel.
at week 6. At the end of 6 weeks, there were no differ-
ences between groups for current LBP, scores on the Ar- Sensitivity Analysis
hus pain index, invalidity index, functional index, or the The definition of high quality given above is somewhat
total score for the Arhus index. The health assessment arbitrary. Therefore, the cutoff of 50% for the internal va-
questionnaire improved in both groups during the lidity item score was modified to 40% and 60% to deter-
6-week period with no differences between groups. mine if conclusions of this review changed. With a 40%
These findings considered together with the method- cutoff (ⱖ5 of 11), all 10 trials included in this review were
ologic quality of these studies suggest that there is mod- high quality. With a 50% cutoff (ⱖ6 of 11), all trials using
erate evidence that there are no statistically significant or Harpagophytum extracts were of high quality22–24 and one
clinically relevant differences in effectiveness between a low quality and,29 3 C. frutescens trials were high quali-
60-mg daily harpagoside dose of an aqueous extract of ty29 –31 and one of low quality.28 With a 60% cutoff for
H. procumbens and 12.5 mg rofecoxib per day in the high quality (OR ⱖ7 of 11), 2 of the Harpagophytum trials
treatment of acute episodes of chronic NSLBP in the were of high quality23–24 with 1 being of low quality,22
short-term. none of the S. alba trials were of high quality,25–27 and 1 of
the Capsicum trials was of high quality30 with the three
S. alba Versus Rofecoxib others being of low quality.28 –29,31
Chronic LBP. A total of 228 subjects with acute episodes
Discussion
of chronic NSLBP were included in a trial,26 which tested
S. alba standardized to provide a daily dose of 240 mg Methodologic Quality
salicin (S) against 12.5 mg per day of rofecoxib in a A total of 10 RCTs were included in this review with 3
4-week trial. Both the rofecoxib and the 240 mg S groups using H. procumbens (Devil’s claw), 3 using S. alba
improved on the pain scale (by 44% in both groups), the (White willow bark), and 4 using C. frutescens (Cay-
Arhus scale invalidity index, pain index, and physical enne). The sensitivity analysis for methodologic quality
impairment index. The percentage of patients requiring revealed that the Harpagophytum trials were generally
 Herbal Medicine for LBP • Gagnier et al 91

of high quality, the S. alba trials were generally of mod- transient gastrointestinal complaints. Large observational
erate quality, and the Capsicum trials were generally of studies are needed to explore the relative safety of these
low quality regardless of cutoff. This analysis does not herbals to standard medications, such as NSAIDs.
affect the conclusions of this paper. This review has several strengths, including the com-
Although the reporting quality in these trials was prehensive search strategy, the inclusion of only the high-
poor, it is known that methodologic quality is not di- est-quality trial designs, and use of suggested methods
rectly related to reporting quality.36 Therefore, the meth- for systematic reviews of interventions for LBP. One ob-
odologic quality of trials with poor reporting remains un- vious drawback of this review is that many of the trials
clear. The present authors did not contact all trial authors included were authored by the same trialists (Chrubasik
to clarify methodologic aspects of trials that were inade- and colleagues). It is possible that the results may be
quately reported in the published manuscripts. systematically biased in some way. It is imperative that
trials of these herbal medicines be repeated by other re-
Efficacy search groups and in different settings.
The results of these 10 trials suggest that specific herbal The qualitative analysis used here may be regarded as a
medicines may be effective for acute episodes of chronic strength and drawback. That is, although it would have
NSLBP in terms of short-term (4 – 6 weeks) improvement been incorrect to statistically combine data from heteroge-
in pain and functional status. These herbal medicine in- neous trials, the qualitative method used does not provide
terventions improved both pain and functional mea- information on the size of the treatment effect. Without this
sures. Seven of the trials included in the review were quantitative data, it is hard to determine whether these
placebo controlled while three trials were comparative. herbal interventions cause clinically significant effects in pa-
There is insufficient evidence to make definitive conclu- tients with NSLBP. Quantitative analyses were precluded
sions regarding those trials comparing herbal medicine by incomplete reporting in these trials. Recent evidence sug-
interventions with standard treatments. Two of the com- gests that reporting of clinical trials, irrespective of the in-
parative trials used Vioxx as a comparator and the other tervention, is poor.37 Specifically, RCTs of herbal interven-
used a homeopathic topical preparation. Given the re- tions have been found to report less than half of the
cent findings regarding severe adverse effects of Vioxx required information as outlined by the CONSORT state-
and it being subsequently taken off the retail market, ment.38 An extension of the CONSORT statement for the
additional trials testing these herbal medicines against reporting of RCTs of herbal medicine interventions has
standard treatments (acetaminophen, NSAIDs) that are been developed and should be referred to when reporting
still readily available. such trial.39,40 These guidelines will aid trialists in planning,
Although 8 of these studies were considered to have implementing, and reporting controlled clinical trials.
homogeneous LBP populations, statistical pooling was Another drawback of this review results from the
not possible due to heterogeneous interventions (3 trials known heterogeneity of herbal medicine products. That
used H. procumbens, 3 S. alba, and 2 C. frutescens) and is, herbal medicines often vary the type of preparation
lack of reporting of sufficient raw data. Therefore, we (liquid, VS dried, VS topical) and, thus, in the amount of
could not provide quantitative evidence of efficacy of chemical constituents per dose. These variations influ-
herbal medicine in general or of any of the 3 individual ence the pharmacokinetics and therefore the relative ef-
herbal medicines used in these trials. Instead, a best- ficacy of these products.
evidence synthesis was undertaken as described above.
Long-term efficacy (e.g., return to work, recurrence) was Conclusion
not assessed in any of these trials and therefore remains Implications for Practice
to be determined. The following herbal medicines have strong evidence for
Given the overwhelming evidence that conflicts of in- the short-term treatment of acute episodes of chronic
terest may bias trial results, we assessed the potential for NSLBP: an aqueous extract of H. procumbens at a stan-
conflict of interest in these trials. We determined that a dardized daily dosage of 50 mg harpagoside, an extract
conflict of interest was a possibility in five trials included of S. alba at a standardized dosage of 240 mg salicin per
in this review. It should be noted that, for each of the day. The following herbal medicines have moderate evi-
herbal medicines examined in this review, at least one dence for the short-term treatment of acute episodes of
trial did not have a conflict if interest. At this point, it is chronic NSLBP: an aqueous extract of H. procumbens at
not possible to determine the specific influence of such a standardized daily dosage of 100 mg harpagoside, an
conflicts on trial results or this review. extract of S. alba at a standardized dosage of 120 mg
This review highlights research that, when combined, salicin per day, and a plaster of C. frutescens.
indicates that there are at least two herbal medicines that
have good evidence for the short-term treatment of acute Implications for Research
episodes of NSLBP. These interventions are reported to Additional high-quality trials must be done to determine
have very few side effects, but more research is required to if H. procumbens standardized to 100 mg harpagoside
more extensively explore the safety of these herbals. The and S. alba standardized to 120 mg salicin are effective in
adverse effects appear to be primarily confined to mild, the treatment of LBP. Also, more trials are needed to
92 Spine • Volume 32 • Number 1 • 2007

more accurately determine the efficacy of topical C. fru- 9. Cherkin DC, Deyo RA. Non-surgical hospitalization for low back pain: is it
necessary? Spine 1993;18:1728 –35.
tescens in any kind of LBP. Additional trials testing these 10. van Tulder M, Koes B. Low back pain and sciatica: acute. Clin Evid 2002;
herbal medicines against standard treatments (acetamin- 8:1156 –70.
ophen, NSAIDs) will clarify their equivalence in terms of 11. van Tulder M, Koes B. Low back pain and sciatica: chronic. Clin Evid
2002;8:1171– 87.
efficacy and effectiveness. The quality of reporting in
12. Vickers A. Recent advances: complementary medicine. BMJ 2000;321:683–6.
these trials was generally poor; thus, trialists should refer 13. Mills EJ, Hollyer T, Guyatt G, et al. Teaching evidence-based complemen-
to the CONSORT statement in designing and reporting tary and alternative medicine: 1. A learning structure for clinical decision
clinical trials of herbal medicines. changes. J Altern Complement Med 2002;8:207–14.
14. Sierpina V. Progress notes: a review of educational developments in CAM.
Altern Ther Health Med 2002;8:104 – 6.
15. Millar WJ. Patterns of use: alternative health-care practitioners. Health Rep
Key Points 2001;13:9 –22.
16. Eisenberg DM, Davis RB, Ettner SL. Trends in alternative medicine use in the
● The following herbal medicines have strong evi- United States, 1990 –1997: results of a follow-up national survey. JAMA
dence for the short-term treatment of acute epi- 1998;280:1569 –75.
17. Astin JA, Pelletier KR, Hansen E. Complementary and alternative medicine
sodes of chronic nonspecific low back pain: an use among elderly persons: one-year analysis of a Blue-Shield Medicare Sup-
aqueous extract of Harpagophytum procumbens plement. J Gerontol: Biol Sci Med Sci 2000;55:M4 –9.
at a standardized daily dosage of 50 mg harpago- 18. Druss BG, Rosenheck RA. Association between use of unconventional ther-
apies and conventional medical services. JAMA 1999;282:651– 6.
side, an extract of Salix alba at a standardized dos- 19. Blumenthal M, ed. The Complete German Commission E Mono-graphs: Thera-
age of 240 mg salicin per day. peutic Guide to Herbal Medicines. Austin, TX: American Botanical Council, 1998.
● The following herbal medicines have moderate 20. Mills S, Bone K. Principles and Practice of Phytotherapy: Modern Herbal
Medicine. Edinburgh: Churchill Livingstone, 1999.
evidence for the short-term treatment of acute epi- 21. Cordell GA, Araujo OE. Capsaicin: identification, nomenclature, and phar-
sodes of chronic nonspecific low back pain: an macotherapy. Ann Pharmacother 1993;27:330 – 6.
aqueous extract of Harpagophytum procumbens 22. Chrubasik S, Zimpfer CH, Schutt U, et al. Effectiveness of harpagophytum procum-
bens in the treatment of acute low back pain. Phytomedicine 1996;3:1–10.
at a standardized daily dosage of 100 mg harpago- 23. Chrubasik S, Junck H, Breitschwerdt H, et al. Effectiveness of Harpagophy-
side, an extract of Salix alba at a standardized dos- tum extractWS1531 in the treatment of exacerbation of low back pain: a
age of 120 mg salicin per day, and a plaster of randomized, placebo-controlled, double blind study. Eur J Anaesthesiol
1999;16:118 –29.
Capsicum frutescens. 24. Chrubasik S, Eisenberg E, Balan E, et al. Treatment of low back pain exac-
● Additional high-quality trials must be done to erbations with willow bark extract: a randomized double-blind study. Am J
determine if Harpagophytum procumbens stan- Med 2000;109:9 –14.
25. Chrubasik S, Kunzel O, Model A, et al. Treatment of low back pain with a
dardized to 100 mg harpagoside and Salix alba herbal or synthetic anti-rheumatic: a randomized controlled study. Willow
standardized to 120 mg salicin are effective in the bark extract for low back pain. Rheumatology 2001;40:1388 –93.
treatment of low back pain. 26. Chrubasik S, Model A, Black A, et al. A randomized double-blind pilot study
comparing Dolotefin and Vioxx in the treatment of low back pain. Rheuma-
● More trials are needed to more accurately deter- tology 2003;42:141– 8.
mine the efficacy of topical Capsicum frutescens in 27. Frerick H, Keitel W, Kuhn U, et al. Topical treatment of chronic low back
any kind of low back pain. pain with a Capsicum plaster. Pain 2003;106:59 – 64.
28. Ginsberg F, Famaey JP. A double-blind study of topical massage with Rado-
● Additional trials testing these herbal medicines Salil Ointment in mechanical low back pain. J Int Med Res 1987;15:148 –53.
against standard treatments (acetaminophen, 29. Keitel W, Frerick H, Kun U, et al. Capsicum pain plaster in chronic non-
NSAIDs) will clarify their equivalence in terms of specific low back pain. Arzneimtelforschung 2001;51:896 –903.
30. Krivoy N, Pavlotzky E, Chrubasik S, et al. Effects of salicis cortex extract on
efficacy and effectiveness. human platelet aggregation. Planta Med 2001;67:209 –12.
31. Stam C, Bonnet MS, van Haselen RA. The efficacy and safety of a homeopathic
gel in the treatment of acute low back pain: a multicentre, randomized, double-
blind comparative clinical trial. Br Homeopath J 2001;90:21– 8.
Appendix available online through Article Plus. 32. Waddell G. Low back pain: a twentieth century health-care enigma. Spine
1996;21:2820 –5.
33. van Tulder M, Furlan A, Bombardier C, et al. Updated guidelines for sys-
References tematic reviews in the Cochrane Collaboration Back Review Group. Spine
2003;12:1290 –9.
1. Mounce K. Back pain [editorial]. Rheumatology 2002;41:1–5. 34. van Tulder MW, Assendelft WJJ, Koes BW. Method guidelines for system-
2. Papageorgiou AC, Croft PR, Ferry S, et al. Estimating the prevalence of low atic reviews in the Cochrane Collaboration Back Review Group for Spinal
back pain in the general population: evidence from the South Manchester Disorders. Spine 1997;22:2323–30.
Back Pain Survey. Spine 1995;20:1889 –94. 35. Manniche C, Asmussen K, Lauritsen B, et al. Low back rating scale: valida-
3. Andersson GBJ. Epidemiological features of chronic low back pain. Lancet tion of a tool for assessment of low back pain. Pain 1994;57:317–26.
1999;354:581–5. 36. Huwiler-Muntener K, Juni P, Junker C, et al. Quality of reporting of random-
4. Borkan J, Reis S. Talking about pain: a patient-centered study of low back ized trials as a measure of methodologic quality. JAMA 2002;287:2801– 4.
pain in primary care. Soc Sci Med 1995;40:977– 88. 37. Moher D, Jones A, Lepage L, et al. Use of the CONSORT statement and
5. Department of Labor: Bureau of Labor Statistics. Survey of Occupational Inju- quality of reports of randomized trials: a comparative before and after eval-
ries and Illnesses. Department of Labor, Bureau of Labor Statistics, 1993. uation. JAMA 2001;285:1992–5.
6. Praemer A, Furner S, Rice DP. Musculoskeletal Conditions in the United 38. Gagnier JJ, DeMelo J, Boon H, et al. Reporting quality of controlled inter-
States. Rosement IL: American Academy of Orthopaedic Surgeons; 1992. vention trials of herbal medicine. Am J Med 2006;119:E1–E11.
7. Coste J, Delecoeuillerie G, Cohen de Lara A, et al. Clinical course and prog- 39. Gagnier JJ, Boon H, Rochon P, et al. Improving the reporting of randomized
nostic factors in acute low back pain: an inception cohort study in primary controlled trials of herbal medicine interventions: an elaboration of the
care practice. BMJ 1994;308:577– 80. CONSORT statement. Ann Intern Med 2006;144:364 –7.
8. Little P, Smith L, Cantrell T, et al. General practitioners’ management of 40. Gagnier JJ, Rochon P, Boon H, et al. Improving the reporting of randomized
acute back pain: a survey of reported practice compared with clinical guide- controlled trials of herbal medicine interventions: explanation and elabora-
lines. BMJ 1996;312:485– 8. tion. J Clin Epidemiol 2006;59:1134 – 49.