Letters

Homoeopathy versus placebo in perennial allergic rhinitis

Statistics in study were flawed

Study shows dissociation between objective and subjective responses to homoeopathy in allergic rhinitis

Study shows double standards in evaluation of homoeopathy

Results of study were not convincingly in favour of homoeopathy

Did patients really have allergic rhinitis?

Authors' reply

Statistics in study were flawed

EDITORTaylor et al come to the conclusion that their study of homoeopathy versus placebo in perennial allergic rhinitis "has failed to confirm our original hypothesis that homeopathy is a placebo."¹ Unfortunately, the statistics do not prove that.

The basis for the study was a prestudy power calculation that required 120 patients to prove the hypothesis with a 5% significance and an 80% power.² In fact, the authors only recruited 51 patients but analysed the results as if they had the required number. Their only conclusion was that they did not have enough data to make a conclusion.

If we accept the availability of only 51 patients at the outset, what are the relevant calculations? The power calculation is only 43%, and to maintain the power calculation at 80% the P value becomes 34%. The only conclusion is that the trial is not able to prove anything.

Barry Miller, consultant anaesthetist. 
Royal Oldham Hospital, Oldham OL1 2JH barry.miller{at}bigfoot.com

Competing interests: None declared.

Study shows dissociation between objective and subjective responses to homoeopathy in allergic rhinitis

EDITORTaylor et al show the apparent dissociation between objective responses to homoeopathy (domiciliary nasal peak flow) and subjective responses (nasal symptoms) after four weeks in 50 patients with allergic rhinitis.¹ Few randomised controlled studies have measured domiciliary peak nasal inspiratory flow rate in allergic rhinitis, which makes these results all the more intriguing.

In one of those studies, of 38 patients with allergic rhinitis, nasal symptom scores showed significant (P<0.01) correlations with morning and evening domiciliary nasal peak flow after four weeks of treatment.² The mean overall improvement in domiciliary nasal peak flow was 25 l/min in response to four weeks of antihistamine, which is comparable to the magnitude of the homoeopathy peak flow response (20 l/min).

Perhaps a longer period of homoeopathy or a different dose might have resulted in a subjective treatment response in patients with allergic rhinitis. Moreover, we need to know how homoeopathy compares to conventional drug treatments such as intranasal corticosteroids and antihistamines, given their proved long term efficacy on symptoms in allergic rhinitis.³

Brian J Lipworth, professor of allergy and respiratory medicine. 
Asthma and Allergy Research Group, Department of Clinical Pharmacology, Ninewells Hospital, Dundee DD1 9SY b.j.lipworth{at}dundee.ac.uk

Competing interests: The Asthma and Allergy Research Group has received funding from Aventis, AstraZeneca, Schering Plough, and GlaxoWellcome, which make intranasal corticosteroids and antihistamines.

Study shows double standards in evaluation of homoeopathy

EDITORThe appearance of yet another high quality randomised trial in allergy raises the important question of whether homoeopathy should be treated any differently from conventional treatments in healthcare systems that are ostensibly evidence based.¹ It also brings to mind an example of the unforeseen way that double standards can rebound on those who refuse to accept any positive results for homoeopathy.

The homoeopathy meta-analysis by Linde et al was generally positive and also found a positive result in a subgroup of the most formally rigorous trialsthose with quality scores 70%.² More recently, Jüni et al compared 25 quality scales, including Linde et al's, by using them to rate a sample of 17 trials of low molecular weight heparin or standard heparin in the prevention of deep vein thrombosis during surgery.³ Trials rated as high quality with Linde et al's scale showed greater benefit from low molecular weight heparin, reversing the findings of the original meta-analysis from which the sample of 17 heparin trials was taken.

Jüni et al attribute the range of results obtained to the content of the scales themselves. Accepting this explanation, one of the authors of the earlier heparin meta-analysis has singled out Linde et al's quality scale for attack: it not only "maximally disconfirmed" his original findings but also achieved a "scientific impossibility"a positive result for homoeopathy.⁴

As usual, an alternative explanation is possible. The content of Linde et al's scale is similar to that of others on test, such as Jadad et al's scale,⁵ that did not reverse the earlier heparin meta-analysis. The main difference is the exceptionally high cut-off point of Linde et al's scale relative to its median score of 50% for the 17 heparin trials: only three reached 70%. Compare this with Jadad et al's scale, where the median score was 60%; nine of the 17 trials were rated as high quality, because the cut- off point was also set at 60%.

If cut-off points are as important as they seem to be here, Vandenbroucke's implication that Linde et al's scale is intrinsically unreliable because it overturned his heparin conclusions is as questionable as his assertion that homoeopathy can never work. It would seem that the original heparin meta-analysis fell at a hurdle designed to trip up homoeopathyone that homoeopathy sailed over.

Michael Emmans Dean, doctoral student. 
Department of Health Sciences and Clinical Evaluation, University of York, York YO10 5DD organon{at}lineone.net

Competing interests: None declared.

Results of study were not convincingly in favour of homoeopathy

EDITORBefore taking the results of Taylor et al's study as an opportunity to speculate about how homoeopathy might work we should first take a careful look at the study and state that the results were negative, the meta-analysis was (or may be) flawed, and there was no homoeopathy at all.¹

All three preceding trials used the change in visual analogue scale scores as the sole main end point, with significant results. In this trial the calculation of sample size was also done for only one main end pointthe change in the visual analogue scale score. The P value was 0.82, which means that the trial was by no means able to reproduce the positive results of the others. The discrepancy with additional "objective" data is interesting but not positive for homoeopathy. What is to be the interpretation of this discrepancy in the light of the preceding trial, in which the discrepancy was just the other way round?

The meta-analysis is founded on four trials with different indications, different treatments, flaws in design and analysis, and a significant heterogeneity in treatment effects. The heterogeneity in treatment effects is almost spectacular in a series of only four relatively small trials, considering the low power of these statistical tests. In such heterogeneous situations, researchers are advised to refrain from doing meta-analyses because these can lead to grossly misleading results.

The patients in this study underwent allergic testing according to standards of orthodox medicine. The allergens were chosen on the basis of standards of orthodox medicine. There was no homoeopathy at all. The only part of the study reminding readers of homoeopathy is the dilution procedure. This whole scenario has nothing to do with the usual practice of homoeopathy, and if the trial had been perceived to be negative this would be the unanimous justification by homoeopaths.

Do we learn anything from this study that is convincingly in favour of homoeopathy? My answer is: No.

Jürgen Windeler, head. 
Department of Evidence Based Medicine, Medical Advisory Service of Social Sickness Funds (MDS), 45116 Essen, Germany j.windeler{at}mds-ev.de

Competing interests: None declared.

Did patients really have allergic rhinitis?

EDITORCan we really believe that the objective improvement in Taylor et al's randomised controlled trial of homoeopathy versus placebo in perennial allergic rhinitis resulted from the administration of 1 g of lactose-sucrose globules (impregnated with either a 30c dilution of the allergen or placebo) at intervals of eight hours for only one day?¹

Examination of the baseline clinical characteristics given in table 1 could provide an explanation. Eight patients in the homoeopathy group and 12 in the placebo group had previously been ineffectively treated with topical steroids, while three of the homoeopathy group and five of the placebo group had been effectively treated with topical steroids. In the homoeopathy group eight had had immunotherapy (three effectively), and in the placebo group five had had immunotherapy (two effectively); but immunotherapy has been impossible in the United Kingdom since 1986. In the homoeopathy group five had had surgery (one with benefit), and in the placebo group six had had surgery (two with benefit). Thirty five of the patients were allergic to mites, but 10 to house dust, presumably not to mites, which is most unusual.

Topical steroids are effective in true allergic rhinitis, confirmed by the presence of many eosinophils in the mucosal smear or blown specimen. Absence of eosinophils goes with lack of response to topical steroids, so the diagnosis of allergic rhinitis in this group is in doubt. It would seem that these results were obtained in a miscellaneous group of volunteers, dominated by 36 women.

H Morrow Brown, emeritus consultant allergist. 
Derby DE22 1HT harry{at}morrow-brown.freeserve.co.uk

Competing interests: None declared.

Authors' reply

EDITORThe authors of these letters and the other rapid responses to our paper¹ have made colourful contributions to this debate. In responding we will try to avoid opinion and concentrate on substance.

Our study was indeed underpowered, but Miller misunderstands the implications of this: the increased risk from a smaller than desirable study is of false negative results, not false positive results. That may explain the result of the visual analogue scale score in this study. In weighing this up readers should note that both the homoeopathy and placebo groups "worked" symptomatically on average; it was not neither group (a common error in interpreting placebo controlled trials).

The data did reproduce the previous responses of useful symptom reductions on average in the visual analogue scale measure with homoeopathy, but this time there was also a relative average decrease in the measure with the placebo. These subjective changes began in both groups during the single blind placebo run-in (figure), and this may have further reduced the subsequent power of the study. Our discussion speculated on the origin of this possible "zeal" factor.

View larger version (23K): [in this window] [in a new window]   Average change each day for each patient compared with his or her baseline average. An average improvement is evident in both groups, beginning during the single blind placebo run-in and continuing thereafter. These data contrast with the objective results shown in the paper, which improved only in the homoeopathy group

Whatever the causes of this symptom reduction, this study also predefined the objective measure as one of two main outcomes. The objective and subjective responses were not uncoupled in the homoeopathic group as Lipworth remarks (because both measures improved on average) but rather were uncoupled only in the placebo group, which showed a subjective but not objective average change.

Dean's analysis suggests that if the trends in the four trials had been in favour of placebo some of the protests might be more muted. Windeler dismisses this trial, and likewise our overview (we accept that it is not a meta-analysis). But a scientific dilemma of this order will not be solved by data-free opinion, only by data. Our experiments were a planned series of studiesthat is, a body of workand are best considered as such. The fundamental coherence of motive, method, and model is well described in the papers. In their responses to these papers some authors do not seem to realise that allergic asthma, allergic hay fever, and allergic perennial rhinitis are all manifestations of the atopic syndrome.

Windeler's comments on what is orthodox medicine may need revision: homoeopaths discovered pollen as the cause of hay fever² and first introduced pollen treatment in rhinitis.³ Just as it would be wrong to use these homoeopathic roots to dismiss orthodox allergen sensitisation, so it would be wrong to use an a priori belief in the implausibility of homoeopathic action to dismiss the results of experimental testing.

In dismissing our results Morrow Brown argues that our patients could not have had perennial rhinitis, citing the previous failures of orthodox treatments. His experience varies widely from that of many patients and general practitioners, who will agree with Lipworth's remarks in his rapid response that he sees "many patients with allergic rhinitis who clearly do not benefit from conventional treatment."⁴ All patients met generally accepted current diagnostic criteria (including results of skin testing). Are we to redefine these criteria rather than accept the results of this trial? And should we also do this for asthma and hay fever because of the positive results in our previous trials?

It is true that the trials are not optimal in terms of the day to day best practice of homoeopathy. This was not the point of our inquiry or the yardstick for our results. We designed the trials for clarity, simplicity, and internal validity to answer the basic question we had posed: Does a homoeopathically prepared dilution show a positive effect over and above its placebo effect? Our data take us nearer to that question's answer.

Morag A Taylor, research associate. 
David Reilly, honorary senior lecturer in medicine. 
davidreilly1{at}compuserve.com University Department of Medicine, Glasgow Royal Infirmary, Glasgow G31 2ER

Robert H Llewellyn-Jones, lecturer. 
Department of Psychological Medicine, University of Sydney, New South Wales 2000, Australia

Charles McSharry, principal immunologist. 
University Department of Immunology, Western Infirmary, Glasgow G11 6NT

Tom C Aitchison, senior lecturer in statistics. 
Department of Statistics, University of Glasgow, Glasgow G12 8QQ

Competing interests: MAT's salary was partly paid by the Blackie Foundation Trust, British Homoeopathic Association, and Scottish Homoeopathic Research and Education Trust administered by Glasgow University. She was reimbursed for attending a symposium organised by the Blackie Foundation Trust. DR began this research programme before using homoeopathy or developing education. He uses homoeopathy in clinical care. He accepts occasional lecture and teaching fees but has no consultancy work. He has declined all direct industry grants for research and has used intermediary regulatory organisations to ensure independence.