Rheumatology International

, 27:585

Choosing between NSAID and arnica for topical treatment of hand osteoarthritis in a randomised, double-blind study


  • Reto Widrig
    • Rheumatology Clinic
  • Andy Suter
    • Bioforce AG
  • Reinhard Saller
    • Department of Internal Medicine, Institute of Complementary MedicineUniversity Hospital Zurich
    • Department of Internal Medicine, Institute of Complementary MedicineUniversity Hospital Zurich
Original Article

DOI: 10.1007/s00296-007-0304-y

Cite this article as:
Widrig, R., Suter, A., Saller, R. et al. Rheumatol Int (2007) 27: 585. doi:10.1007/s00296-007-0304-y


The use of topical preparations for symptom relief is common in osteoarthritis. The effects of ibuprofen (5%) and arnica (50 g tincture/100 g, DER 1:20), as gel preparations in patients with radiologically confirmed and symptomatically active osteoarthritis of interphalangeal joints of hands, were evaluated in a randomised, double-blind study in 204 patients, to ascertain differences in pain relief and hand function after 21 days’ treatment. Diagnosis was according to established criteria; primary endpoints were pain intensity and hand function; statistical design was as per current regulatory guidelines for testing topical preparations. There were no differences between the two groups in pain and hand function improvements, or in any secondary end points evaluated. Adverse events were reported by six patients (6.1%) on ibuprofen and by five patients (4.8%) on arnica. Our results confirm that this preparation of arnica is not inferior to ibuprofen when treating osteoarthritis of hands.


OsteoarthritisIbuprofen gelArnica gelRandomised trialGel treatment


Osteoarthritis (OA) is one of the most common joint disorders affecting mature adults and the elderly, in whom the prevalence of hand symptoms ranges from 13% (men) to 26% (women), according to the Framingham Study [1]. The commonest affected joints are the distal interphalangeal (DIP) followed by proximal interphalangeal (PIP) and the metacarpal–phalangeal (MCP) joints, with gradual cartilage destruction associated with development of nodule-like swellings on both sides of the joints (Heberden and Bouchard nodes for DIP and PIP, respectively). Swelling and redness are usually mild but may be accompanied by severe pain and increasing functional impairment [2].

Current treatment is essentially supportive and symptomatic [3], as there is no available therapy to reverse or halt the natural progression of OA. Analgesics and non-steroidal anti-inflammatory drugs (NSAIDs) such as ibuprofen and diclofenac remain the mainstay of drug treatment. Ibuprofen, in use for over forty years, has been shown to relieve OA symptoms effectively [4] and intra-articular steroids and hyaluronic acid derivatives are also used, whilst physiotherapy and other supportive care have been shown to improve function [3]. Topical use of NSAIDs reduces systemic adverse effects, particularly on the gastrointestinal (GI) tract [5] and a meta-analysis of 86 placebo controlled studies (10,160 patients) confirmed the benefits of various topical agents [6], while ibuprofen gel 5% has been shown to be effective in musculoskeletal injuries [7, 8], tendonitis [9], general osteoarthritis [10] and osteoarthritis of the knee and finger joints [11, 12].

Drug penetration and bioavailability of ibuprofen gel 5% in muscle and connective tissue beneath the application site are in the same order of magnitude as after oral administration [13, 14].

Topical phytotherapy preparations such as extracts of willow bark, devil’s claw, capsicum and arnica are also frequently used [15]. Arnica montana L (other names are: arnica, mountain tobacco, medicinal leopard’s bane; mountain alkanet) has been used as a topical agent since at least the sixteenth century [16], and nowadays most frequently for the treatment of contusions and sprains, or for symptomatic relief in OA [17].

We stipulated three pre-conditions for choosing the arnica preparation amongst the numerous available. Firstly, the preparation had to have published evidence from preclinical studies of some anti-inflammatory action, secondly the preparation had to have published evidence of skin penetration and, thirdly it had to be available as gel and should be as similar as possible to ibuprofen gel, to ensure blindness during the study. Our chosen preparation has shown evidence of complete inhibition of the transcription factor NF-kappaB by arnica constituents (sesquiterpene lactone helenalin, 11α,13-dihydrohelenalin and its ester) [18], which is similar to the effects of glucocorticoids on inflammatory cascade [19], as well as skin penetration [20]. In appearance, arnica gel was indistinguishable from ibuprofen.

Materials and methods

Patient selection and blinding

We enrolled patients from 20 clinics (12 general practices, 6 rheumatology clinics, 2 general medicine) in three Swiss cantons (Zurich, St Gallen and Aargau) between May 2003 and March 2004. A total of 204 patients were enrolled, of whom 198 were included in the intention-to-treat (ITT) analysis and a further 24 patients were excluded from the per protocol (PP) analysis (Table 1). Diagnosis was according to established criteria [21]. Patient selection was conducted according to the Osteoarthritis Research Society International (OARSI) guidelines [22] and their recommendations for studying polyarthritis of finger joints, and the European Medicines Agency (EMEA) guidelines on choice of controlled studies and statistical design regarding gel evaluations [23, 24]. The study was conducted according to Good Clinical Practice guidelines and the Declaration of Helsinki, and approved by the relevant Ethics Committees and Swiss regulatory authorities.
Table 1

Disposition, demographics and status of patients in the study


Ibuprofen gel

Arnica gel

Randomized (n)



Excluded at baseline visit (n)



ITT Population (n, %)





Treatment duration <18 or >26 days (n, %)





Use of (excluded) analgesics (n, %)





No per protocol radiological evidence (n, %)





Hand function <5 (HAI) or Pain <4 (VAS) (n, %)





Early intolerance of gel (n, %)





Total exclusions (n, %)





Per protocol population (n, %)





Age (years; mean, SD)





Females (n, % PP population)





Weight (kg; mean, SD)





Affected fingers (n; mean, SD)





Affected joints (n; mean, SD)





Hand algofunctional index (mean, SD)





Pain intensity (VAS; mean, SD)





Painful joints (n; mean, SD)





Morning stiffness intensitya (mean, SD)





Morning stiffness durationb (mean, SD)





No statistical difference between groups in any demographic or status parameters at baseline

aMean value of 1 none, 2 slight, 3 moderate, 4 severe, 5 extreme morning stiffness

bMean value of 1 none, 2 up to 2 min, 3 up to 5 min, 4 up to 10 min, 5 more than10 min

Patients were initially screened and, if considered eligible and consented, they were randomised to one of the two treatment groups (ibuprofen or arnica). The protocol selection criteria are summarised in Table 2. Before being dispensed treatment, patients were fully medically evaluated. Both treatments were to be used in identical dispensing doses of gel (4 cm strip), gently rubbed over the affected joints thrice-daily for 3 weeks, and were asked not to wash hands for one hour after application. Patients were dispensed a preset number of paracetamol tablets (20 × 500 mg tablets) as “escape treatment” (not allowed within 24 h prior to the final evaluation) and asked to return any unused tablets at the end of the treatment course. For patients on long-term low-dose aspirin for cardiovascular conditions, 325 mg daily maximum was allowed.
Table 2

Patient selection

Inclusion criteria

 Patients of either gender

 Age between 18 and 88 years, inclusive

 Osteoarthritis diagnosis according to ARC criteria (modified by Altman) [21]

  (a): pain or stiffness in the hands or fingers on most days during the previous month

  (b) hard tissue enlargement in ≥2 of ten joints (DIP and PIP joints II/III as well as both carpal-metacarpal joints)

  (c) less than two swollen MCP joints

  (d) ≥1 DIP joint with hard tissue enlargement or ≥1 of the joints referred to under (b)

 In addition, the following criteria applied for the worst affected hand

  Pain intensity of on the Visual Analogue Scale (VAS 0–100 mm) of at least 40 mm in the finger joint with the worst possible spontaneous pain in the previous 24  h

  At least 5 points, measured on the Hand Algofunctional Index (HAI)

  Radiological confirmation of osteoarthritis in ≥2 joints with radiographs ≤3 months old and ≥1 painful joint confirmed radiologically

 Discontinuation of all NSAIDs (topical and systemic) >10 days prior to entry

 Discontinuation of all analgesics >3 days prior to entry

 Written Informed consent

Exclusion criteria

 Secondary osteoarthritis (due to causes other than solely degenerative joint disease)

 Trauma to the hand or arm in the previous 2 months

 Residual pain following fracture, dislocation or operation

 Pain and stiffness due to tissue scarring


 Carpal tunnel or other nerve compression syndromes

 Serious conditions, such as cancer, uncontrolled hypertension or heart failure

 Systemic or intra-articular corticosteroids in the previous month and during study

 Damaged skin or serious skin disorders affecting hands

 Allergy to Asteraceae

 Intolerance to paracetamol

Ibuprofen Gel 5% (Optifen® Gel) was supplied by Spirig Pharma Ltd, and Arnica Gel (A. Vogel Arnica Gel®; Arnica montana fresh herbal tincture 50 g/100 g gel; drug-to-extract ratio (DER) of the tincture 1:20) was supplied by Bioforce AG.


The aim of the study was to evaluate the efficacy and tolerability of arnica gel compared to ibuprofen gel 5% in the treatment of OA of the fingers using the approach (non-inferiority) suggested by current guidance on comparisons of gel preparations [23, 24]. Randomisation codes were computer-generated in blocks of four (RanCode, Version 3.6, IDV-Gauting, Germany). Double-blindness was assured by identical packaging (100 ml tubes), as well as gel appearance and consistency. There was a slight difference in odour for the first 30 s after application, after which both were odourless. We considered this of no practical consequence for blindness because patients only received one of the treatments, so they were not aware of the comparator.

The primary efficacy parameters were: (a) reduction in pain, recorded as the most intense in the previous 24 h, in the worst affected finger, on VAS (100 mm) by the patient and; (b) functional capacity of the hand using the validated HAI assessment (Table 3) [25]. The secondary parameters were: (a) number of painful joints (DIPs, PIPs and MCPs) in both hands; (b) intensity of morning stiffness in the worst affected hand; (c) duration of morning stiffness in the worst affected hand; (d) analgesic consumption; (e) patients’ and doctors’ global evaluation of efficacy; (f) patients’ acceptance of the gel treatment. Safety was to be evaluated by adverse event monitoring.
Table 3

Hand algofunctional index

1. Are you able to turn a key in a lock?

2. Are you able to cut meat with a knife?

3. Are you able to cut cloth or paper with a pair of scissors?

4. Are you able to lift a full bottle with one hand?

5. Are you able to clench a fist?

6. Are you able to tie a knot?

7a. For women: Are you able to sew?

7b. For men: Are you able to use a screwdriver?

8. Are you able to fasten buttons?

9. Are you able to write for a long period of time?

10. Would you accept a handshake without reluctance?

Scores: 0 possible without difficulty; 1 possible with slight difficulty; 2 possible with significant difficulty; 3 impossible

Non-inferiority (arnica relative to ibuprofen) was set as the difference being not greater than 12% in each of the two main assessment tools, which corresponds to 3.6 points on HAI (for hand function) and 12 mm on the VAS (for pain), hence a standardised difference (effect size) of 0.66. A standard deviation of 5.5 was derived from the validation data of HAI [25]. The Mann–Whitney statistic P(X < Y) was used as a measure of relevance for the difference, hence: P(X < Y) < 0.5 (inferiority); P(X < Y) = 0.5 (no difference); P(X < Y) > 0.5 (superiority). The Mann–Whitney statistics (for an effect size of 0.66) corresponds to P(X < Y) = 0.322.

Non-inferiority of arnica was to be formally shown when the left margin of the one-sided α-confidence limit of the Mann–Whitney statistic was greater than 0.322. In a first step, H01 was tested (arnica < ibuprofen for combined pain and hand function). If the left margin of the one-sided α-confidence limit were greater than 0.322, then hypotheses H02 and H03 (arnica < ibuprofen for pain and hand function, respectively) were tested, maintaining the global level of significance (closed test principle).

Sample size calculation was based on an assumption that the difference between the two treatments would be not more than −1 point on HAI. Thus, with a standard deviation of 5.5, a one-sided t-test with significance level of alpha = 0.024 and 80% power gives a sample size of 73 patients per group to reject the null hypothesis (inferiority of arnica). Since the power of the Wilcoxon test is slightly inferior to that of t-test (for a normal distribution), a minimum of 80 assessable cases (per protocol) were required. An interim analysis was planned for possible adjustment of sample size after 60 patients were entered, for which the statistical tests were adjusted accordingly and the confidence interval was restricted to 95.2%. No adjustments were required to the sample size during the interim analysis.

An ITT analysis was to be conducted similarly to the PP analysis but including all randomised patients who had used the gel at least once and who had a post-baseline value for comparison. The safety analysis was to be conducted in all patients who used treatment at least once.


On entry, there were no differences between groups regarding any demographic or disease parameter (Table 1).

Pain intensity and hand function were similarly improved in both treatment groups, the differences being within the 95.2% confidence intervals and the inferiority thresholds (Table 4). The generalised Wilcoxon’s test for the treatment difference was above the defined non-inferiority threshold of 0.322. Thus, the first hypothesis that arnica is inferior to ibuprofen in terms of both target parameters (function and pain) was rejected and the two criteria could be tested for the one-sided level α = 0.024, which showed the left limit of the respective two-sided 95.2% confidence regions also greater than 0.322 for the two individual parameters, so the inferiority hypothesis was not proven for the two criteria. Results from the analysis of the ITT population were essentially identical to those of the PP population.
Table 4.


Main endpoints

Pain (VAS)

Hand function (HAI)

Ibuprofen gel (n = 85)

Arnica gel (n = 89)

Ibuprofen gel (= 85)

Arnica gel (= 89)

PP population 

 Day 0: mean (SD)

68.0 (14.2)

66.9 (14.4)

12.1 (3.3)

11.3 (4.0)

 Day 21: mean (SD)

44.2 (20.9)

40.4 (21.5)

7.5 (4.3)

7.1 (4.8)

 Difference: Day 0–Day 21: mean (SD)

−23.9 (23.3)

−26.6 (22.4)

−4.6 (3.1)

−4.3 (3.6)

 Difference: Ibuprofen-Arnica: mean (SD)

3.4 (20.8)

−0.2 (3.4)

 95.2% CI for difference

[−2.9; +9.7]

[−1.3; +0.8]

 Associated inferiority threshold



Primary variable test* P(X < Y) + 0.5*P(Y)



95.2% CI

[0.455; 0.627]

[0.382; 0.554]

Directional test* for Primary Variable


For 95.2% CI


Secondary endpoints

Ibuprofen gel

Arnica gel


[95% CI]

Number of painful joints

−2.5 (3.2)

−3.0 (4.4)


[−0.6; +1.4]

Intensity of morning stiffnessa

−1.0 (0.8)

−0.9 (1.1)


[−0.1; +0.2]

Duration of morning stiffnessb

−0.8 (1.1)

−0.9 (1.3)


[−0.3; +0.3]


Pain (VAS)

Hand function (HAI)

Ibuprofen gel (n = 98)

Arnica gel (= 100)

Ibuprofen gel (= 98)

Arnica gel (= 100)

ITT population

 Difference: Day 0–day 21 (SD)

−22.6 (24.0)

−25.1 (22.5)

−4.2 (3.6)

−4.1 (3.6)

 Difference: Ibuprofen-Arnica (SD)

2.5 (21.3)


−0.1 (3.6)


 95.2% CI for difference

[−3.1; 9.0]

[−1.1; 1.0]

 Primary variable test* P(Y) + 0.5*P(Y)



 95.2% CI

[0.448; 0.610]

[0.394; 0.555]

 Directional test* for Primary variable


 For 95.2% CI

[0.429; 0.574]

PP Per protoco, ITT intention to treat, HAI hand algofunctional index, VAS visual analogue scale

aMean value of 1 none, 2 slight, 3 moderate, 4 severe, 5 extreme morning stiffness

bMean value of 1 none, 2 up to 2 min, 3 up to 5 min, 4 up to 10 min, 5 more than10 min

The results were also similar for both groups for all secondary parameters (Table 4). Mean (±SD) paracetamol consumption as “escape treatment” was 11.3 ± 7.2 and 11.2 ± 6.8 tablets (ibuprofen and arnica, respectively) during the 3 weeks.

Global efficacy evaluation by physicians was very good or good in 56.5% (ibuprofen) and 64.0% (arnica), while patients rated efficacy very good or good in 58.8% (ibuprofen) and 64.1% (arnica) of cases (Table 5). Patients’ acceptance of the gel treatment was 76.5 and 78.7% (very satisfied or satisfied; ibuprofen and arnica, respectively) (Fig. 1).
Table 5

Patients’ and physicians’ evaluations


Ibuprofen gel (n = 85)

Arnica gel (n = 89)


n (%)

n (%)

Patients’ evaluation of efficacya

 Very good efficacy

21 (24.7)

27 (30.3)

 Good efficacy

29 (34.1)

30 (33.7)

 Minor efficacy

25 (29.4)

23 (25.8)

 Unchanged condition

10 (11.8)

9 (10.1)

Patients’ acceptance of topical treatment

 Very satisfied

23 (27.1)

33 (37.1)


42 (49.4)

37 (41.6)


19 (22.4)

17 (19.1)

 Very dissatisfied

1 (1.2)

2 (2.2)

Doctors’ evaluation of efficacya

 Not assessable

0 (0.0)

1 (1.1)

 Very good efficacy

23 (27.1)

27 (30.3)

 Good efficacy

25 (29.4)

30 (33.7)

 Minor efficacy

27 (31.8)

21 (23.6)

 Unchanged condition

10 (11.8)

10 (11.2)

Good = significant improvement, partial symptoms remission; Minor = slight improvement, requires other treatments

aVery good = complete or almost complete remission of all symptoms

Fig. 1

Treatment perceptions by patients and doctors

Both treatments were well tolerated by patients, with 6.1 and 4.8% of patients on ibuprofen and arnica, respectively, reporting a treatment-related adverse event. The skin was the main target organ for events (Table 6). There was one serious event (back trauma due to fall), not related to treatment (arnica).
Table 6

Safety Results


Ibuprofen gel (n = 99)

Arnica gel (n = 105)


n (%)

n (%)

Patients with any adverse event

8 (8.1)

14 (13.3)

Patients with treatment-related eventsa

6 (6.1)

5 (4.8)

Patients with serious eventsb


Number of events

9 (9.1)

15 (14.3)

Number of treatment-related events

7 (7.1)

5 (4.8)

All events’ List

 Skin and Hand

  Skin irritation/itching (hands)



  Increased finger pain






  Scaly skin


  Allergic eczema





  Chest pain




  Back pain











9 (9.1%)

15 (14.3%)

aPossibly, probably or definitely related to treatment

bOne case of back pain due to fall causing withdrawal from treatment


Although a multitude of topical gel preparations for OA of peripheral joints are used by patients and prescribed by doctors, we set out to evaluate whether there was a noticeable difference between a commonly used synthetic NSAID and a traditional phytotherapeutic, when tested by objective methods using validated assessment tools, and according to the perceptions of doctors and patients. Our results demonstrate that topical arnica gel is not inferior to ibuprofen gel, regarding hand functional capacity, pain intensity, number of painful joints, duration and severity of morning stiffness, or paracetamol consumption. Importantly, not only were improvements in all parameters similar, they were also clinically relevant. When blinded to treatment, neither patients nor doctors can distinguish between the effects of these two treatments.

There were some limitations in our study, but these were minor and did not affect the plausibility of the results. Firstly, although similar in every other way (packaging, consistency, appearance, physical properties), blinding of the gels could be argued to have been sub-optimal because they emanated slightly different odours during the first 30 s of application. We feel this cannot have had a major influence in the results, because patients were only given one of the treatments and they were to be used at home. Should we have used a cross-over design (where each individual would be exposed to both gels), this might have had an influence on results. Secondly, despite our best efforts, we lost six patients from follow up for the ITT analysis. One has to consider that patients with chronic, non-life threatening conditions have slightly lower attendance compliance than those with serious illnesses. However, this was a very small number relative to the overall sample (n = 204) and it is highly unlikely that the ITT results would have changed in substance, if their post-baseline data were obtainable.

Consumption of paracetamol was strictly controlled and comparable in both groups. Yet, we were cautious to ensure even the “escape treatment” was discontinued 24 h before the final evaluation, to prevent an analgesic carry-over effect.

A review of topical NSAIDs indicated some superiority relative to placebo, particularly in soft tissue disorders, as well as equivalent efficacy when compared with some orally administered NSAIDs, although the literature was not very robust. The safety profile was acceptable (2% self-limiting skin events). The cost–benefit analysis showed an initial higher cost for topical treatments but also an overall benefit [26].

Two meta-analysis of randomised trials with topical application of NSAIDs, one in osteoarthritis and rheumatoid [5] and one of osteoarthritis [27] showed superiority to placebo in pain reduction with short-term (1–2 weeks) treatment. Topical NSAIDs are now recommended in osteoarthritis to reduce risk of adverse events associated systemic administration and to target local discomfort, particularly to control symptom flare up [28]. The analgesic properties of topical NSAIDs in particular, have been demonstrated with ibuprofen gel 5% (thrice-daily) versus oral ibuprofen (400 mg thrice-daily) [29]. Finally, our study investigated short-term use only, but this is in agreement with studies that suggest short-term, such as during flare ups, is the preferred setting for gel treatment [27].

While the skin route for medication is long established in dermatology, in the last few decades it has established itself in cardiology (nitrate patches) and it is gradually becoming common in rheumatology. In this regard, appropriate testing in methodologically strict studies is the only way to address, which amongst the wide variety of agents offer real benefit.

In summary, our results show that short-term use, up to three weeks, of arnica gel improves pain and function in hand OA, indistinguishably from ibuprofen gel. Similar effects are also observed with regard to number of painful joints, severity and duration of morning stiffness, as well as the perceived efficacy by patients and doctors. Topical application of arnica gel can be regarded as an alternative to ibuprofen gel when treating OA of the hand joints.


RS conceived the overall project plan. RW was the Principal Investigator. The following doctors treated and evaluated patients in the 20 study clinics: D. Kähler (Wil); R. Widrig (St Gallen); D. Geiger (Au); S. Bachmann (Valens); H. Martin (Gossau); M. Burla (Zürich); B. Kindler (Zürich); M. Senn (Zürich); U. Aufdenblatten (Zürich); H. Sajdl (Zürich); K. Grimm (Dietikon); E. Zeley-Zimonyi (Zürich); M. Steigbügel (Wald); U. Pachlatko (Ebmatingen); M. Tejero (Baden); B. Haldemann (Küttigen); H. Maurer (St Gallen); F. Akermann (Buchs); F. Jaggi (Zürich); A. Stevka (Zürich); H. Steinemann (St Gallen). Statistical analysis was conducted by DSH Statistical Services GmbH, Germany.

Copyright information

© Springer-Verlag 2007