62 patients could be included in the intention-to-treat analysis. There were no group differences on the primary outcome measures of pain (P=0.79) and bruising (P=0.45) at day four. Swelling and use of analgesic medication also did not differ between arnica and placebo groups. Adverse events were reported by 2 patients in the arnica 6C group, 3 in the placebo group and 4 in the arnica 30C group.

The alpine plant Arnica montana is recommended by homeopathic practitioners for treating injuries on account of its alleged ability to control bruising, reduce swelling and promote recovery¹. Homeopathic arnica is popular with patients undergoing surgery, who hope it will reduce postoperative complications. However, despite favourable anecdotal reports² there is little scientific evidence of its efficacy³. A reduction in the pain associated with routine dental extractions has been reported in placebo-controlled trials of homeopathic arnica^4,5,6, but two double-blind randomized trials demonstrated no effect on pain, swelling or bleeding after surgical removal of impacted wisdom teeth^7,8. Other trials have shown no beneficial effects on postoperative haematomas⁹ or on recovery from hysterectomy¹⁰, acute stroke illness^11,12 or childbirth¹³. Tentatively positive results were reported from a small study in patients with acute trauma¹⁴. A recent trial of homeopathic arnica administered in conjunction with herbal arnica ointment was reported to show no effect on swelling or grip strength but a reduction in pain following carpal tunnel release surgery¹⁵.

Patients received a medication bottle, identifiable only by study number, containing either high (30C) or low (6C) potency homeopathic arnica tablets or indistinguishable placebo. Tablets were to be taken three times daily for seven days preoperatively and fourteen days postoperatively. Patients were advised to refrain from handling the tablets or from eating, drinking, smoking or brushing teeth within 20 minutes of taking the tablets and were asked to suck the tablets rather than simply swallow them. Homeopathic and placebo tablets were supplied by A Nelson & Co Ltd. Medication bottles were labelled with study numbers derived from a computer-generated randomization list in blocks of three by an individual not involved with running the trial. The randomization list was kept in a sealed envelope in a locked drawer until the end of the trial. All patients and investigators, including the surgeon, physiotherapists and data analysts, remained blind to treatment allocation until after data analysis.

All patients were admitted as day cases and received conventional preoperative and postoperative care. The operations were done under local anaesthesia. Afterwards the hand was rested on a palmar plaster splint to maintain the wrist in slight dorsiflexion, allowing the fingers to be gently mobilized within the dressing and the hand elevated in a high sling. Oral analgesic medication, either paracetamol or diclofenac, was routinely prescribed on the hospital discharge form. All patients were seen by the physiotherapist at four, nine and fourteen days post-surgery (or the closest possible day). At day four the splint was removed and digits and wrists were mobilized. A Futura aluminium wrist splint was given to the patients to wear for a further week. Sutures were removed at day fourteen.

The primary outcome measures were pain and bruising, and the secondary measures were swelling and use of analgesic medication. Pain was assessed with the short-form McGill Pain Questionnaire (SF-MPQ)¹⁶ completed by the patient at recruitment (to provide a measure of pain from carpal tunnel syndrome) and on days four, nine and fourteen post-surgery. The SF-MPQ includes a visual analogue scale (VAS)¹⁷ to indicate the intensity of pain and a list of fifteen descriptive words (e.g. stabbing, gnawing, shooting). The VAS is sensitive to changes in pain intensity¹⁸. The pain descriptors are each rated on a 4-point scale (0=none, 1=mild, 2=moderate, 3=severe) yielding a total score ranging from 0 to 45. Patients kept a daily pain diary throughout the trial, including VAS scores; they also recorded use of analgesics and any adverse events. To quantify bruising, a photograph of the patient's wrist at the distal crease was taken by the physiotherapist on days four, nine and fourteen post-surgery under standard lighting conditions. Scanned images were analysed with Adobe Photoshop 4.0 software. For each patient, frames representative of normal skin (thenar zone) and of the bruised areas (operative site) were selected. The distribution of red and blue pixels within each frame was calculated. This information, displayed as a histogram of the number of pixels (y-axis) against an increasing scale of colour brightness from 0 to 255 (x-axis), enabled an objective and quantitative comparison of the colour of the bruised area with the colour of the normal skin. This method was developed for objective measurement of burn scar hypertrophy¹⁹ and has been successfully used to assess healing at skin graft donor sites²⁰. For each patient the extent of bruising was also assessed independently by two plastic surgeons blind to treatment allocation on a 3-point scale (0=none, 1=mild—moderate, 2=severe) as a check on the validity of colour separation analysis in assessing bruising.

To quantify swelling, wrist circumference was measured at the distal wrist crease before anaesthetic infiltration and on days four, nine and fourteen post-surgery. Three readings were taken of each measurement. Adherence to the study medication was assessed by tablet counts at the end of the study. Patients were asked to tick boxes in the study diary as a further record of tablet taking. The success of patient blinding was assessed by a question in the study diary on the last day of the trial, asking patients to indicate which treatment they believed they had received (arnica, placebo, don't know).

Results of the primary outcome measures of pain and bruising are presented in Tables ​Tables22 and ​and3.3. Postoperative pain did not differ between the groups at day 4 according to VAS scores (χ²=4.81, d.f.=2, P=0.79) or MPQ descriptor scores (χ²=2.48, d.f.=2, P=0.29). Similarly, bruising did not differ between the groups at day 4 in terms of blue (χ²=1.61, d.f.=2, P=0.45) or red (χ²=3.89, d.f.=2, P=0.14) channel brightness. The null hypothesis was therefore accepted.

This trial was designed to be a methodologically rigorous investigation of the specific effects of two potencies of homeopathic arnica. Attempts were made to select outcome measures that were objective and/or adequately validated. In particular, by employing colour separation analysis it was intended to assess bruising in a more objective quantitative manner than previous studies that have relied only on crude and subjective ratings^{9, 22}.

These results do not support the routine use of homeopathic arnica for preventing or reducing postoperative complications such as bruising, swelling and pain. However, they do not rule out the possibility that individual patients could benefit. Homeopathic practitioners identify specific patient characteristics (e.g. fear of being touched, denial of illness, difficulty sleeping due to a hard bed, anxious dreams²) that help predict who will respond to arnica. This trial did not apply traditional homeopathic principles in this way. However, in one trial of surgical patients where the homeopathic remedy was chosen to match the patient's constitution, arnica was selected in 21 of 24 cases and no differences from placebo were demonstrated⁸.

The use of non-parametric tests reduces the likelihood of detecting statistical differences since they are less powerful than parametric tests particularly with smaller samples. However, a type II error does not appear to account for the results of this trial, where differences between groups are so small that they would have no clinical relevance even if statistically significant. Furthermore, there is no clear trend or consistent pattern that favours any of the interventions, indicating that measurement and random error can explain any perceived differences.

Poor adherence to the trial regimen was seen in over one-third of the sample. However, it seems unlikely that this explains the lack of difference between arnica and placebo since homeopathic practitioners often recommend that a single dose of arnica before and after surgery is sufficient to hasten recovery². From the study diaries it seems that more tablets were missed in the latter days of the trial when patients are likely to have fully recovered.

A final point to consider in attempting to explain the lack of difference between arnica and placebo groups is the surgical model used in this trial. There was little bruising, pain and swelling in any of the groups, so perhaps the skill of the surgeon offered little scope for arnica. However, arnica is reputed to be effective in every case of trauma, however slight²; some of the rigorous trials cited above have tested arnica in bigger operations without significant effects^{7,8,9,10,11,12,13}. Conversely, it might also be argued that major tissue injury is too severe to benefit from the subtle effects of homeopathy; the present choice of surgical model represented a compromise between minor and major trauma.