Homeopathy Meta Analysis

Translation of a Dutch article that appeared as Homeopathie: meer dan placebotherapie? in Geneesmiddelenbulletin, 1996, 30 (3) p. 26-32. It originally appeared in French in la revue Prescrire October 1995; 15 (155): 674-684.

This translation is not an official one and has not been approved by either Geneesmiddelenbulletin or la revue Prescrire. It has been translated by J.W. Nienhuys. Interpolations with garrulous remarks made by the translator are clearly marked. Notes are inserted at the place where they appear, instead of at the end of the article. Diacritical signs are inserted before the letter to which they should be applied.

Footnote (*): based on Aulas, J-J, Hom'eopathie: actualisation 1995 du dossier d'evaluation. La Revue Prescrire 1995; 15: 674-684, under co-responsibility of the editors.

Sometime ago a couple of respectable medical magazines published four comparative clinical researches of homeopathic preparations:

Note 1: Jacobs J, Jimenez LM, Gloyd SS, Gale JL, Crothers D. Treatment of acute childhood diarrhea with homeopathic medicine: a randomized clinical trila in Nicaragua. Pediatrics 1994; 719-725.

Note 2: Lange-de Klerk, ESM de, Blommers J, Kuik DJ, Bezemer PD, Feenstra L. Effect of homoeopathic medicines on daily burden of symptoms in children with recurrent upper respiratory tract infections. BMJ 1994; 309; 1329-1332.

Note 3: Lökken P, Straumsheim PA, Tveiten D, Skjelbred P, Borchgrevink CF. Effect of homoeopathy on pain and other events after acute trauma: placebo controlled trial with bilateral oral surgery. BMJ 1995; 310: 1439-1442.

Note 4: Reilly DT, Taylor MA, Beattie NGM, Campbell JH, McSharry C., Aitchison TC et al. Is evidence for homoeopathy reproducible? Lancet 1994; 344: 1601-1606.)

This formed for our sister magazine Prescrire a reason to collect all relevant research published in 1990-1995 and analyse them carefully in an article of their own. The following is an adapted version of that. After a short explanation of a few homeopathic concepts, the advances in the field of fundamental and pharmacological research are discussed. After that we present a critical discussion of the selected clinical trials. After some information about side effects we answer the question if there are indications for effectivity of homeopathic preparations.

Homeopathy is a system of medical treatments developed by the German physician Hahnemann (1755-1843). Classical homeopathy attributes the effectivity to the following hypothetical principles. The first is is the similia principle: equal things are cured by equal means (similia similibus curantur). One first obtains a complete picture of all complaints of the patient. Next one searches for a preparation that produces in high doses symptoms in healthy persons that correspond with the complaints of the patient. This is called the individuality principle. The regular medical therapy of the 18th and 19th century used toxic remedies. This has stimulated the development of homeopathy. The similia principle was in those times opposed to the allopathic principle, that was supposed to cure by opposites (contraria contrariis curantur).

The second principle of homeopathy is administration of remedies in highly diluted form. Each remedy is first obtained as `original tincture' An original tincture contains 10 g of active ingredient dissolved in 100 g of solvent. After this potentiating takes place: repeated dilution and shaking each time after each dilution. Such a dilution is supposed to have a certain strength. This strength increases with the degree of dilution, which is sometimes expressed in D (decimal dilution) and sometimes in C (centesimal dilution). D1 means a 1:10 dilution, so one part of the original material in nine parts solvent. In some cases the `solvent' is a powder and rather than dissolving, the material and the `solvent' are rubbed together.

After good shaking or rubbing together one tenth of this can be mixed with 9 parts of solvent and one obtains D2. This can go on to dilutions of D200 and higher. C1 means a 1:100 dilution, so 1 part material on 99 parts of solvent.

In a dilution higher than D24 or C12 most probably no molecule of the original remedy can be found back anymore, because with such dilutions the Avogadro number is exceeded. The effect of the original remedy is not lost because an `imprint' of the original material remains in the solvent, whether it is fluid or not.

A remedy that might have a provable pharmacodynamic action, doesn't need to have a therapeutic effect. On the other hand, it is not very probable that something has a therapeutic effect without at least some pharmacodynamic action being known.

Until now, there is no proof of pharmacodynamic activity of high dilutions, even though people have tried to find it, notably the research group of Benveniste. Their efforts led in 1988 to a paper in Nature, in which they seemed to demonstrate a biological effect of ultramolecular dilution of goat anti-immunoglobulin-E-serum, namely a loss of coloring ability of polynuclear basophils.

Note 5: Davenas E, Beauvais F, Amara, J. Oberbaum M, Robinzon B, Miadonna A et al. Human basophil degranulation triggered by very dilute antiserum against IgE. Nature 1988; 333: 816-818.

It concerned a dilution of 10 to the power -120 mol, which is much higher than the number of Avogadro. This article led to a lively polemic. The results described turned out to have been artefact.

Note 6: Draaisma D. Voorbij het getal van Avogadro: de Benveniste-affaire. Kennis en Methoden 1989; 13: 84-107.

In this article it is not possible to go into detail regarding several pharmacological researches of homeopathy. Each research group that publishes about this subject works in its own special field and does not take into account the work of others.

Note 7: Recherche en hom'eopathie 1995. Plaquette de 36 pages 'edit'ee par les laboratoires Boiron `a l'occasion du 5e congr`es de l'Organisation M'edicale Hom'eopathique Internationale, Paris, 20 au 22 octobre 1994.

An example of this is research into the effect of certain molecules on the growth of cancer cells. Other investigations occupy themselves with the influence of dilutions of acetylsalicylic acid on the aggregation of pleteles or with the effect of high dilutions of mercury salts on mice kidneys. Many of these researches are published in serious magazines. As far as known, none of the results is replicated by another independent research group. Sometime ago a meta-analysis was published that underlined the methodological weakness of the of the pharmacological research in the field of homeopathy.

Note 8: Linde K, Jonas WB, Melchart D, Worku F, Wagner H, Eitel F. Critical review and meta-analysis of serial agitated dilutions in experimental toxicology. Hum. Exp. Toxicol 1994; 13: 481-492.

There are very many isolated clinical observations that point to the use of certain homeopathic treatments. These are regularly reported in specialised magazines and in homeopathic congresses. Such observations are not unimportant, but they form no evidence for a specific therapeutic effect of homeopathic treatments. Only clinical investigations that compare the effect of a homeopathic remedy with a placebo can determine a specific homeopathic effect that differs from the natural course of the affliction and from the placebo effect that is inherent to the administration of a remedy.

Adherents of `classical' homeopathy think that the similia principle cannot be tested in controlled clinical investigations because every treatment must be directed to individual symptoms. Adherents of the `modern' homeopathy are of the opinion that certain afflictions can be treated by ingredients that are matched to characteristics of diseases rather than to individual symptoms. `Modern' homeopathic principles can be tested in controlled clinical trials.

Comment by translator. The authors are being here rather mild about `classical' homeopathy. In classical homeopathy the patient is interviewed for about an hour during which time a great number of questions are asked whose only relevance to the complaint is that they bring out a great many subjective or even superstitious attitudes of the patient. The matching of these complaints to the remedy chosen is basically a random process. No research is known that shows that different homeopaths can independently arrive at the same remedy, or that remedies of equal positive effectiveness are found. Such a proof would actually be inconsistent with homeopathic theory, because that theory holds that giving the patient a non-optimal remedy will result not in a cure but in making the patient sicker, because the artificial remedy sickness will merely be added to the patient's sickness, without driving it out. The scientific objections to homeopathy usually focus on the high dilutions, but the essentially random process of matching complaints (not: diagnosis) to treatment is at least as serious from a scientific point of view.

Note 9: Kleijnen J, Knipschild P, Riet, G ter. Clinical trials of homeopathy. BMJ 1991; 302: 316-323 [Correction:]BMJ 1991; 302: 818.

Note 10: Aulas J-J. Hom'eopathie. Etat actuel de l'evaluation clinique. Paris: Editions Frison-Roche, 1991.

discuss the most important clinical investigations of homeopathic remedies, that are published until 1990. The 1991 meta-analysis of all controlled investigations of homeopathy (about 100 in the period 1966-1990) concluded that there were positive indications that homeopathy might work.

Because most investigations had a poor methodological quality, it was insufficient to draw conclusions. The book concluded that a specific effect of homeopathy compared to placebo hadn't been demonstrated. Since then several comparative clinical investigations have been published. Some were about indications that often are treated homeopathically, such as warts, recurrent respiratory tract infections in children, asthma, obesitas and pain. Other investigations were concerned with the treatment of serious symptoms, such as advanced rheumatoid arthritis.

Prescrire publishes regularly analyses of investigations of homeopathic remedies. Below is a survey of selected researches from the period 1990 to the beginning of 1995, followed by a comment. The author searched for double blind randomised and placebo controlled trials. Excluded were open researches, researches with healthy volunteers, and the so-called n=1 trial.

Note 11: Sudan BJL. Abrogation of facial seborrhoeic dermatitis with homeopathic high dilutions of tobacco: a new visible model for Benveniste's theory of `memory of water'. Med Hypotheses 1993; 41: 440-444.

Using Medline (period january 1990 to july 1995) and Embase(period january 1992 to january 1995) a simple search strategy was applied with the key words `homeopathy' and `clinical trial'. Finally 11 articles were found that satisfied the mentioned criteria.

Note 12: Hofmeyer GJ, Postpartum homoeopathic Arnica montana: a potency finding pilot study. Br J Clin Parct 1990; 45: 465-466.

This double blind placebo controlled pilot trial was executed to determine which one of two potencies of Arnica montana, D6 or D30 is suitable for use in greater clinical trials. It was executed in a hospital on Johannesburg among 161 women who just had had a baby and who had during delivery either an episiotomy or a perineal tear. The women were randomised into three groups: 85 women received non-impregnated placebo granules; 37 got granules with Arnica montana D6 and 39 got Arnica montana D30. The physician noted the state of the perineum at the start of the treatment and also in the four following days. He paid special attention to the presence or absence of a hematome larger than 50 mm, to inflammation near the stitches, to a possible center of infection, the presence of oedema and the global healing of the wound. Each patient noted in a diary the severity of the perineal pain, the use of painkillers, her mood, the state of the baby and her judgment about the global effectivity of the treatment.

Over the period of four days the investigators found no statistically significant difference between the three groups as regarding as wound closure criteria. On the contrary, significantly more women in the Arnica D30 group (13%) felt unhappy than in the Arnica D6 group and in the placebo group (2%). Also significantly more women considered the treatment as effective with Arnica D6 (43%) and placebo (46%) than with Arnica D30 (23%). The authors conclude that the D6 potency is probably the most effective to use in the next investigation.

Comment: A specific effect of the homeopathic dilution of Arnica montana has not been proven by this. Because of the small number of patients and the large number of criteria (13, altogether) the observed differences, especially between the two dilutions can be accidental.

Comment by the translator. The one-tailed probability of the unhappiness in the D30 group being so high is 0.0096 (exact test of Fisher; two-tailed and one tailed makes no difference here). However, there is from the point of view of the null hypothesis nothing special about the D30 group, and there is certainly no prediction that the unhappiness should be larger or smaller. So there are six ways in which one of these unhappiness numbers might be extreme. For each of them we should then take as significance margin 0.0085. Moreover in this actual example the two other ways of lumping the data together yield p=0.22 and p=0.35 by the two-tailed exact Fisher test. So the author's conclusion that there might be something to Arnica D6 is only justified if they had some a priori certainty that either Arnica D6 or Arnica D30 would be effective. Naturally the judgement about effectiveness is heavily anti-correlated with unhappiness, so that is less striking. In view of the fact that this seems the best result among 13 computations the only reasonable conclusion is that there is nothing significant about the study - as could be expected from a pilot study. (Significance refers to a computation planned in advance of the experiment, not to selected results from an enormous mass of possible computations done afterwards.)

Note 13: Alibeu JP, Jobert J. Aconit en dilution hom'eopathique et agitation post-op'eratioire de l'enfant. Pediatrie 1990; 44: 619-621.

This is an investigation into the effectivity of Aconitum napellus, a homeopathic remedy that is given in case of agitation that accompanied by fear. In this research 50 children aged 1/2 to 14 year were included that had come to the hospital for several surgical treatments below the navel, like inguinal ruptures, ectopies and phimosis. They were divided into two groups, at random. Agitation in the recovery room was treated with either Aconitum napellus C4 or a placebo. The most important criterium was the degree of calming down in the 15 minutes following administration of the remedy. The authors do not state how agitation was measured. The publication gives the impression that they satisfied themselves with a global subjective judgement.

Of 50 children 47 were analysed, 23 with homeopathy and 24 with placebo. The two groups were comparable in age, kind of surgery and type of anesthesia. The authors report that 19 children in the homeopathy group experienced decrease in agitation, and 8 in the placebo group. One patient in the homeopathy group and 11 in the placebo group didn't experience agitation. Comment. The lack of precisely defined inclusion criteria and precise judgement criteria makes it impossible to accept these results without reservations. What is needed is a new investigation with more children and the methods to judge the agitation should be better described.

Note 14: Andrade LEC, Ferraz MB, Atra E, Castro A, Silva MS. A randomized controlled trail to evaluate the effectiveness of homeopathy in rheumatoid arthritis. Scand J Rheumatol 1991; 20: 204-208.

Altogether 44 patients took part in this double blind placebo controlled trial. They were judged clinically before the start of the trial and after that monthly during the six months the trial lasted. A homeopathic physician determined the remedies to be used, and prescribed them as monotherapy or combination therapy in dilutions C5 to C30. Altogether 21 different remedies were prescribed. The physician could change the treatment depending on the development of the patient's condition, even though he didn't know whether the patient was getting homeopathic remedies or placebo (diluted alcohol). The patient could continue the existing treatment with painkillers or anti-rheumatic drugs, but he or she had to note down the amounts used. Patients and physician gave their global judgement of how the affliction developed during each monthly evaluation. Ten patients of the original group stopped because of ineffectivity: five homeopathy, five placebo. One patient stopped because of another reason. The authors found no statistically significant difference in effectivity and side effects between the two groups.

Comment. This investigation took into account an essential characteristic of `classical' homeopathy, namely the individuality principle. However, continuation of an existing treatment is not according to the principles of `classical' homeopathy. Moreover, giving regular medication makes the interpretation of the results more difficult. It can be concluded that the remedies investigated show no value of homeopathy. The small number of patients makes it difficult to demonstrate small differences between the groups.

Note 15: Gaus W, Wiesenauer M. Wirksamkeitsnachweis eines Hooeopathikums bei chronischer Polyarthritis - Eine randomisierte Doppelblindstudie bei niedergelassenen Aerzten zur Kritik und Ausblick. Akt Rheumatol 1991; 16: 1-9.

Eight physicians (general practitioners, internists and orthopedists took part between July 1987 to October 1989 in a double blind, randomised and placebo controlled investigation among 176 outpatients (ambulatory) with rheumatoid arthritis. In this investigation a homeopathic preparation consisting of a combination of 5 remedies (Rhus toxicodendron, Bryonia cretica, Nux vomica, Berberis vulgaris, Ledum palustre) in a D4 dilution compared to an alcoholic placebo solution. During 12 weeks the patients received 3 to 4 times a day 10 to 20 drops of the verum or the placebo. Each patient noted the pain on a visual analog scale, and also the use of pain killers. In contrast with the investigation mentioned above patients were excluded that used local or systemic corticosteroids. Judgement criteria were: seriousness of pain during the night, at rest or in motion, the duration of the morning stiffness, the seriousness of the inflammatory phenomena, measured with BSE and CRP, the degree of tiredness, the score on the functional index of Lee and the global judgement by physician and patient. This was not conform the original protocol in which the development of pain and the use of painkillers were the most important judgement criteria. This means that only a part of the research protocol was executed.

Only 111 of the original 176 patients were analysed: 58 verum and 53 placebo, so 65 (37%) patients were not judged. Statistical analysis consisted of a comparison by a chi-squared test of the percentage that improved in each group. The homeopathy group showed a significant improvement.

There was much criticism about the great number of people that dropped out of the study, because these weren't analysed. Consequently the authors redid their analysis, this time according to what they thought was the `intention to treat' principle, in which all original patients were included in the final analysis.

All patients that had been lost were counted as those with a negative result, regardless whether they were verum or placebo. This produced 44 improved patients in the verum group and 30 in the placebo group. No improvement was observed in 41 verum and 61 placebo. The difference was statistically significant (one tailed chi squared test, p < 0.01)

Note 16: Gaus W, Wiesenauer M. Wirksamkeitsnachweis eines Hooeopathikums bei chronischer Polyarthritis - Stellungname zur Kritik und Ausblick. Akt Rheumatol 1993; 18: 159-162.

Comment. This investigation provokes many questions. The test used is not suitable to prove effectivity. To do so, one should do an analysis of variances for repeated measurements. The authors do not provide no individual data, so this analysis cannot be made. Also the authors misunderstood the `intention to treat' principle. In such an analysis the data of the dropouts are analysed according to the treatment assigned to them during randomisation, regardless of a later change.

Comment by translator. Assuming the numbers 44 and 30 to refer to the non-dropouts, I find by Fisher's exact test that the null hypothesis (no difference) is rejected, with p=0.0436. A chi-square test with continuity correction yields p=0.05137. This is marginally significant, if the number of improved verum or not improved placebo had been one less then a two-tailed test had not produced a "significant " result. It is not clear if there was a group of not improved and not worsened either; if there was such a group it is not clear why they were counted with "not improved". In other words, the deviation from protocol may very easily have been what was needed to get at a "significant" result. It is also not clear why a one-tailed test was used. The null hypothesis says there should be no difference whatsoever, and the alternative hypothesis doesn't preclude a deleterious effect of the remedy. Altogether it seems the researchers had too much freedom in choosing outcome measures and modes of computation. I don't understand the remark about the analysis of variance.

Note 17: Labrecque M, Audet D, Latulippe LG, Drouin J. Homeopathic treatment of plantar warts. Can Med Assoc J 1992; 146: 1749-1753.

The investigation was about 174 people with one or more warts on the soles of the feet. They were recruted via a university institute for general practioning and through the local newspapers. They were included in a double blind randomised and placebo controlled trial. Half (actually 86, tr.) received during six week a homeopathic treatment that is customary in case of warts, namely Thuja C30, 200 granules per week, Antimonum crudum C7, 5 granules a day and Nitricum acidum C7 200 granules per day. The other half received non-impregnated granules as placebo. The used methodology was good: a strict definition of the inculsion criteria, a double clinical judgement and a very meticulous checking of the observations. The most important criterium was complete disappearance of the warts. The percentage cured patients after 6, 12 and 18 weeks was 5%, 13% and 20% in the homeopathy group and 5%, 135 and 24% in the placebo group.

Comment. In this methodologically well performed trial the effect of the homeopathic treatment didn't differ from the placebo therapy.

Note 18: Kurz C, Nagele F, Zorzi M, Karras H, Enzelsberger H. Bewirkt Homoeopathie eine Verbesserung der Reizblasensymptomatik? Gynaekol geburtshilfliche Rundsch 1993; 33 (suppl 1): 330-331.

This research was about 40 patients that complained about an unstable bladder. This affliction was defined as an urgent need to urinate in combination with pollakiuria (abnormally frequent unrination), nycturia (urniation at night) dysuria (difficult urination) and tenesmus (great urge to urinate without being able to do so). The participants were randomised over two groups. The first group received a placebo for four weeks, and the second Causticum D4, a remedy that is supposed to cause in undiluted form (i.e. a mixture of Ca(OH)2 and K2SO4) the phenomena of unstable bladder. Each patient received a urodynamic test at the beginning and at the end of the trial. The most important outcome parameters were the amount of urine released at each urge, the frequency of urination and the nightly urination. After four weeks a significant improvement of each of these parameters was observed in the verum group. Comment. This trial doesn't satisfy any of the the criteria that a good comparative investigation should satisfy. Inclusion and exclusion criteria are mentioned in very summary form and unclearly defined. The data of the urodynamic tests are not given, and one gets the impression that the authors after the test was over have chosen those parameters that developed favorably in the verum group during the trial. No data are given about the placebo group, and the statistical analysis only refers to how the affliction developed in the verum group. This is simply wrong, because only comparison of effects between groups is acceptable.

Note 19: Werk W, Galland F. Helianthus tuberosus Therapie bei Uebergewicht. Gewichtsreduktion langfristig stabilisieren! Therapiewoche 1994; 44: 34-35.

The investigators selected 166 overweight male patients from 12 general practitioners' practices, each of whom were aged between 23 and 72. The were randomised in a double blind trial over one homeopathy and one placebo group. They had a body mass index (BMI) of 26 to 35 and an increased cardiovascular risk profile. For each patient the length, the weight and the BMI was measured and a biological balance was determined. This happened at the beginning of the the 12 weeks of the trial, and also at the end of weeks 4, 8 and 12. The verum group used Helianthus tuberosus D1. In the beginning of the investigation the patients were in all respects comparable (age, height, weight, BMI). Of the 85 placebo patients 45 completed the trial and of the 81 verum patients 57 did. In the placebo group the weight decreased in average 4.7 kg and the BMI about 1.5. In the verum group the avarage decrease was 7.2 kg and 2.3 BMI. These differences were statistically significant (p < 0.05).Comment. About 40% of the orginal patients dropped out during this three month trial. Analysis was not according to the intention to treat principle, hence such a large dropout destroys the validity of the the trial. Furthermore, Helianthus tuberosus is included in the German homeopathic pharmacopiae, but use of a D1 dilution isn't homeopathy, but rather phytotherapy.

Homeopathy with children with diarrhea (Nicaragua, 1994) (see note 1 above).

In two clinics during 3.5 weeks a double blind placebo controlled trial was performed. During this period all children between 0.5 and 5 years old who paid a visit to the clinics for acute diarrea were included in the trial. Acute diarrhea was defined as at least three unformed stools during the previous 24 hours (the actual average was 7 to 8, tr.). The diarrhea should not have lasted more than one week and the children should not have received more than one doses of an antibiotic, an antiparasitic or an antispasmodic. Each child was examined, and its length, weight and dehydration symptoms were noted. Dehydratation was judged according to WHO guidelines. Children with serious dehydration were excluded from the trial. Judgement of clinical improvement was done by the parents, who noted on a diary quality and frequency of the stools. The most important criterion was the duration of the diarrhea episode, defined as the number of days from inclusion to less than 3 unformed stools per day during two consecutive days. The homeopathic remedies (C30 potency) were prescribed by experienced homeopathic practioners from the US, including the first and last author of the paper. After each stool the child was given by the parents one pellet of remedy (verum or placebo). Altogether 92 children were admitted, of whom 81 (40 verum) in the final analysis. Five (3 verum) turned out not to satisfy the inclusion criteria (mistakes of local personnel, tr.) Four others couldn't be traced back and two children were withdrawn by their parents. All children received standard (WHO) oral hydratation treatment. The parents were asked to continue giving ordinary food and if possible and applicable to continue breast feeding. According to the most important prior established judgement criterion the median duration in the homeopathy group was 2.5 day and in the placebo group 4 days. The difference was significant. The authors claimed that this difference could only be due to the homeopathic remedy because the two groups were comparable.

Comment. This investigation has acceptable methodological quality, but it is not perfect. The lack of precision in the most important judgement criterion, namely the solidity of the stool as judged by the parents can have caused a serious distortion. Also the number of children that had received prior regular medication: 9 in the verum group and 5 in the placebo group. This diffrence is not significant(the groups were too small), it cannot be excluded that this caused some distortion. The authors also didn't take fibre intake into account, which can also be a source of distortion.

Comment by translator. This research has been commented upon by Wallace I. Sampson as well. The Gebu/Prescrire authors stress that there were many sources of variation that could have contributed to the result. These comments are mostly not justified according to me. In the above translation I had to correct a number of inaccuracies of the Dutch text. The main point is that we cannot be certain that the mode of computation was determined by the prior protocol in sufficient and excruciating detail. The authors state "the duration of diarrhea was defined as the number of days after entry into the study until there were less than three unformed stool per day for 2 consecutive days. This definition was used in the preliminary study, and was chosen before data collection as the primary outcome measure for this study." The authors report p=4.8% for this primary outcome, and four other, secondary, outcome measures corresponded to p=3.6%, p=5.4%, p=3.7% and p=30%. As 5% is the magical line, small details in interpretation of the data make the difference between "significant" and "non-significant". One would think that 2.5 days and 4.0 days make quite a difference. If one looks at the averages (3.0 and 3.8) the difference becomes less impressive. The standard deviation in both groups was about 1.8 days, quite large and possibly so because of the factors mentioned by the other commentators: unreliability of parent's judgement, variation in fiber content of diet. The point is that significance or nonsignificance in this case hinges on one or at most very few children having a diarrheal duration of one day less. The 0.3 extra days that the diarrhea had lasted already in the verum group, the 4 extra verum kids that received prior regular treatment may have contributed to this. What matters is the question how much liberty the authors had to take these differences into account, or ignore them, whatever suited them best.

Another point of criticism is the overoptimistic attitude of the authors. They have provided extra treatment for a mild self limiting disease for which there is a good standard therapy (oral rehydration). To administer this extra treatment a trained physician has to spend something like an hour with the patient, and if the effect would be real, it would consist of one less porridge-like stool by day three of treatment. Contrary to what the authors suggest, such an effect would hardly make a difference in the death toll of 5 million of this disease, also because there is simply not enough capacity to do all these homeopathic consultations.

A barely significant result such as this is at most a reason to conduct a larger, better planned investigation. One other critic has taken exception to Jacobs et al. quoting the ill-famed Davenas 1988 Nature paper. Actually the results of that paper are not quoted, but merely the science fiction-like theoretical explanation given in that paper, which bears no relation to the remainder of its contents and is also not supported by other research. For example in the Davenas paper a conjectured quantum mechanical effect that persists for about 10 femtoseconds under special circumstances is invoked (by mere paper citing) to render more plausible a totally different effect after a full day. Rather than forthrightly stating the obvious, namely that homeopathy is completely irrational (a random coupling of random patient complaints to random drug pictures resulting in a remedy with no active ingredients), they pay lip service to the folkloristic tradition that one should present a theoretical explanation. Why they should do so is unclear to me, but I think that pointing to others who have tried to pay lip service to the same tradition is about as innocuous as one can get.

Homeopathy for children with recurrent upper respiratory infections (Netherlands 1994) (see note 2 above)

This trial involved 175 children aged 18 months to 10 years that were brought in either by their family doctor or by publicity in the Dutch press (magazines and newspapers). The trial was double blind placebo controlled and randomized. It was first published as a dissertation.

Condition for admission was that the children should have had the previous year at least three upper respiratory infections, or only two plus otitis media acuta (acute middle ear inflammation). Children who had had adenotomy or tonsillectomy were excluded. Also excluded: children that had a homeopathic treatment during the last six months, or that had some kind of chronic condition, like a congenital heart/lung defect, mental retardation, neurological problems or too short of too light for their age. In the beginning of the trial each child and/or parent was questioned carefully and physical examination was done by a homeopathic physician (the first author, tr.) Then an otolaryngolocical (ear-nose-throat) examination was done by a specialist, and a biological balance sheet was established. During the treatment the parents kept a diary on which were noted tiredness, fever, headache, nasal discharge, ear pain, throat pain. Each week these data were collected during a structured telephonic interview. These data were converted into a total score ranging from 0 to 56.This score, the prescribed antibiotics, adenotomy and/or tonsillectomy were the most important parameters. All parents received written information with advice about hygienics and diet. The homeopathic remedies were individually prescribed in dilitions D6 and higher, and changed if that seemed necessary. The children could get a concurrent regular treatment. Of the 175 children, 170 (86 verum) completed the treatment of six months. The analysis of the results didn't show any significant result: not in prevention of middle ear inflammation, (88% in the verum group, 89% in the homeopathy group), nor in other otolaryngological symptoms. In the homeopathy group 59 antibiotic treatments took place and 77 in the verum group, which doesn't differ significantly. Like wise: adenotomy 16% verum, 21% placebo, tonsillectomy: 5% (both), paracentesis: 9 placebo kids, 10 verum kids, tube in eardrum membrane: 16 homeopathy, 15 placebo.

Comment. This research was of the socalled classical homeopathy type, and also the methodological demands of clinical comparative research were observed. The great number of antibiotics prescribed makes the interpretations of result difficult. On the other hand it was possible that the number of patients was not large enough to bring out a significant difference.

Comment by translator. The above is only a summary of the most important points. Actually each of the children was followed for a whole year. The trial took 5 years to complete, because not many people could be found willing to take part in the trial. The author compared also the number of sickness episodes, the average number of sick days, and from these data one can determine that the placebo group was just a bit sicker in all respects: the number of episodes was 6% larger, the duration of each episode was a 7% larger, and the degree of sickness during each episode was 4% larger. None of those differences were significant. Some of these measures were "almost significant" for example the antibiotic use p=0.09 two tailed, and the total day sum score for the whole year p=0.06. The difference in number of sick days, namely 6, was regarded by the principal author as still the best estimate of the difference, so she did not give up her homeopathic practice, because she thought 6 days was still worth the trouble.

Interestingly the use of antibiotics decreased in both groups by about 60%, compared to the year before treatment was started.

As described, the parents were questioned every two weeks by telephone, and had to keep a diary as well. This must make them focused very much on their child. No amount of prior checking and comparing groups can determine whether the parents in one group react to this attention by exaggerating more than in the other group. I have asked the first author some question regarding her data. She apparently didn't know how to answer them, and she directed me to the co-supervisor, who apparently had done all the computer and statistical work. The co-supervisor didn't answer my letters.

The complaint of the children is one that in any case resolves itself gradually. It is caused by a certain slowness in maturation of the immune system. It is this aspect that contributes to the conviction in the Netherlands that homeopathy works well for upper respiratory tract infections. There must be many success stories of the kind `my child suffered much from this, but after seeing the homeopath the child became much better'. An investigation showed that 45% of all regular physicians believed that homeopathy might work for hayfever or upper respiratory infections. In other words, one might think, homeopathy works, then it should work for this complaint.

The remarks about the number of patient insufficient are based on a misunderstanding of the literature, I think. After the BMJ paper somebody commented on this matter. At issue here is clinical significance. The study population produced a mean score of 2.48 in the placebo group and 2.11 in the verum group. This difference corresponds to a one-tailed p=3.4%, whereas for statistical significance p=2.5% is needed. If one talks about clinical significance, (i.e. worth the effort of the doctor) then one has something different in mind, namely a minimum conjectured difference. Suppose one has a statement that the `real' difference between verum and placebo is, say, 1 in a case like this. If one does a test one would like the result to either reject the null hypothesis with 95% reliability or to reject the alternative hypothesis with 90% reliability (i.e. a power of 90%). This means that the number of patients must be large enough. For a difference this large, the number of patients should be at least 91. As the actual number was 170, the number of patients was quite enough. Things become different if the claim is less spectacular. If only a difference of 0.5 is claimed then one needs 4 times as many patients. The desired number is inversely proportional to the square of the difference. The numbers change again when one assumes different power. Saying after the test that one needs more patients is tantamount to retroactively reducing the claim. As the author already has made public her view that for her a reduction of 47 sick days a year to 41 sick days a year (only a 12.5% reduction for a self limiting disease) was meaningful enough, she is - after the test - changing her bets.

Homeopathy with atopic asthma (Great Britain 1994) (see note 4 above) The design of this trial resembled very much an earlier investigation by the authors.

Note 20: Reilly DT, Taylor MA. McSharry C, Aitchinson T. Is homeopathy a placebo response? Controlled trial of homoeopathic potency, with pollen in hayfever as model. Lancet 1986; ii: 881-886.

They investigated double blind and placebo controlled C30 potencies of allergens. The remedies were individually assigned. The trial consisted of 28 patients older than 16 years with allergic asthma, mainly for house dust. The asthma had to exist for over a year and it should have been confirmed with positive skin tests for certain allergens. The reversibility of the forced expiratory volume in one second (FEV-1) after bronchodilatation had to be more than 15%.

The patients were followed for four weeks by a homeopath and a specialist, before they were entered into the trial, so as to guarantee the stability of the symptoms. They could pursue their ordinary treatment, and they were asked not to avoid contact with allergens during the trial. At the beginning of the observational period and also four weeks later, the balance was made on the basis of allergy tests and lung function examination and bronchial hyper-activity. After the preliminary investigation the participants were subdivided into groups depending on the most important causative allergen and the daily inhaled amount of inhaled glucosteroids (more or less 1 mg prednison per day equivalent). Next they received the test remedy (verum of placebo). The patients judged the seriousness of their symptoms on a visual analog scale (VAS). The noted down their normal use of medication and the severity of their night-time asthma (and morning tightness, tr.) on a scale from 0 to 4. Statistical analysis was according to the intention to treat principle. Notwithstanding this only 24 out of 28 patients were analysed. Only the course of the VAS symptom score produced a statistically significant difference in favor of the verum. The authors state that this difference could not be attributed to a change of regular medication.

Comment. This research has been carried out meticulously, but the small number of patients admits no definitive conclusions. The difference between the groups is wholly caused by 3 placebo patients that worsened and one verum patients that improved. The patients were not allowed antihistamines during the trial (bronchodilators were allowed). This makes the interpretation of the results more difficult. The authors give no information about how often these medicines were used. Afterwards the authors gave more details, which showed that there were no differences between the two groups in bronchodilator use. It would have been interesting to know how this use would have been in the 4 patients that caused the difference between the groups. To prove that their results were reproducible, the authors did a meta-analysis on this, their 1986 investigation and more investigation which was never published in a peer reviewed journal. The method used for this meta-analysis is incorrect. Afflictions with a different mechanisme (allergic rhinitis and atopic asthma) cannot be put together in one meta-analysis. The authors do not mention that the methods used in the three trial were not comparable enough.

Comment by translator. This trial is too small. Especially worrying is that the authors indicate that they planned for 120 patients. Their only reason to stop before 120 was: "We did not recruit this number of patients because of qualification screening and closing recruiment before the pollen season." If the investigators had the possibility to inspect the results before the trial was over and were at liberty to stop or not depending on the results, then this trial is worthless. The method by which the remedy was chosen wasn't at all the way classical homeopathy works, because the actual suspected causative agent was diluted and no subjective symptoms were interpreted. This means that this test had a built in protection against a negative outcome: all homeopaths in the world would have summarily rejected the tests as irrelevant to their own practice. Another problem is the interpretation of VAS score changes for asthma. It is not clear whether changes can be added and whether one big change (say, a patient going from 40 to 5) is equivalent to a 7 patients going down 5. If we use these VAS scores only to assess "improved or not", and we follow the intention to treat, then Fisher's exact test on the 28 patients yields p=13% (two sided). I cannot quite follow the computations of the authors. From their graph (figure 2) I obtain about the same as their 7.2 average VAS reduction in the homeopathy group, but I cannot see how that graph would imply a 7.8 VAS increase for the placebo group. I get about 5.8. Even so, the odds that the verum and placebo change would differ so much (given this population of 24 patients) are about 2 in 1000 (i.e. 0.2%). After the fact it is apparently not difficult to choose methods of computation that yield vastly different results, so one needs strong guarantees that the computation as performed is really the one that was planned. I wrote to the author to provide me with more data, or at least a printed version of the protocol that predated the start of the trial, but I received merely the advice to read the follow up discussion in BMJ.

This was a clinical placebo controlled crossover trial with 20 women and 4 men aged 19-28 in good health who needed to have impacted lower wisdom teeth taken out on both sides. Both extractions were performed by the same surgeon, on the same time of the day, and with two exceptions on the same day of the week. The mean time between two extractions was 27 days (minum 14, maximum 51). Each patient received local anaesthesia and received three hours after each extraction a remedy (once verum, once placebo). Administration was repeated each quarter of an hour for three hours, and then each hour until bedtime. The next day the patient received two more doses with an three hour interval. The homeopathic physician could change the homeopathic treatment depending on the developments in the first 24 hours. The homeopathic remedies (Arnica, Hypericum, Staphysagria, Ledum, Phosphorus, Plantago) were chosen according to the signs, and administered as D30 sucrose tablets. If necessary patients were allowed codeine phospate tablets 25mg, usage was noted. Judgement criteria were development of pain on an analog scale, seriousness of swelling of the face after extraction, maximal distance in which the mouth could be opened and the seriousness of postoperative bleeding. The authors also noted the global preference of the patient for the treatment sequence and the ideas of the patient about homeopathy. Only for the distance that the mouth could be opened there a significant difference (p< 0.05) was found in favour of the verum. (More precisely: on day 3; on day 7 the difference was not significant. Tr.) Comment. This research is performed accurately, took the individuality principle into account and used an appropriate methodology. In this research homeopathy did not have an effect on pain, also not on bleeding, and also not on swelling. The marginally significant difference in ability to open the mouth cannot be a convincing argument for a specific effect of homeopathy, given the problem how to judge this aspect. The value of homeopathy compared to ordinary painkilling or placebo is not proved.

In places where side effects of drugs can be reported not many reports have been received about homeopathic drugs. Some of these don't even register them, or only since recently. An offical national poll could give insight in undesirable effects of homeopathic treatments. We (GEBU) have mentioned before (1993) de effects of alternative cures, homeopathy included. Knowledge about this hasn't increased much during the last five years and consists of incidents. For example 3 patients in Austria reproted side effects.

Note 21: Aberer W, Strohal R. Homoeopathic preparations - severe adverse effects, unproven benefits. Dermatologica 1991; 182: 253.

The first patient had during a bout of flu taken a homeopathic remedy that contained Quincana bark and Ipecacuanha in D4 potency. After three days the patient presented herself with itch and swelling of palms of the hand and soles of the feet, followed by reddening of the skin. The second patient developed a spotty kind of eczema after taking a complex homeopathic remedy with 15 ingredients (no dilution factor known) for weight loss. The third patient took a homeopathic complex of grass pollen (no dilution known) and had to be brought into intensive care with an anafylactic shock. Each of these three patients redeveloped the symptoms when the remedy was administered again. These three examples show that allergic effects can also occur when drugs are taken by mouth (one would of course expect more allergic effects when the stuff is injected or so).

The clinical investigations in this article were published in the period 1990-1995. They all more or less satisfy the methodological demands of placebo controlled trials. Most of them concern `modern homeopathy' in which the individuality principle is not observed. Generally they do not show a better effect of homeopathy than placebo. Some of the investigations are doubtful because of small number of patients or insufficient criteria. These doubts can only be relieved by having other research groups repeat the efforts in better planned trials. Some trials are difficult to interpret because simultaneously a regular treatment is given. At this moment there is no single homeopathic remedy that has shown a specific effect in a double blind randomized placebo controlled trial.

If the natural course of a disease often brings a spontaneous, or temporary improvement of the clinical phenomena, then this will make judgement more difficult. Also the placebo effect, inherent in every medical intervention will complicate judgement. Homeopathic treatment is often given in afflictions with a irregular or favourable natural course, while the attention a homeopathic physician gives the patient certainly contributes to the placebo effect.

High dilutions (> C5) administered sublingually or orally are safe, this much is agreed upon. Serious side effects have been reported both by non-oral and oral use of low dilutions (D4 or less). So care is advised when prescribing homeopathic remedies. Only prescribing homeopathic remedies in case of serious diseases is certainly wrong if there exist effective regular treatments.

With what is known at this moment therapeutic use of homeopathic remedies must be considered and used as a special form of placebo therapy. Nothing indicates that homeopathy offers more.

Homeopathy, more than placebotherapy? (*)