- Email: [email protected]
Can homeopaths detect homeopathic medicines? A pilot study for a randomised, double-blind, placebo controlled investigation of the proving hypothesis
British Homeopathic Journal (2001) 90, 126–130 ß 2001 Nature Publishing Group All rights reserved 0007–0785/01 $15.00 www.nature.com/bhj
ORIGINAL PAPER
Can homeopaths detect homeopathic medicines? A pilot study for a randomised, double-blind, placebo controlled investigation of the proving hypothesis A Vickers1, R McCarney1*, P Fisher1 and R van Haselen1 1
Academic Unit, Royal London Homoeopathic Hospital, London, UK
Homeopaths believe that a medicine, which causes a particular symptom in a healthy volunteer, will cure a similar symptom in a sick patient. From this phenomenon, it is possible to deduce a hypothesis: homeopaths should be able to distinguish a homeopathic medicine from a placebo by taking both and observing their effects. If true, this would support an effect of homeopathic medicines different from that of placebo. If false, it casts serious doubt on the contemporary homeopathic knowledge base and on homeopathic pathogenetic trials (HPTs) as currently practised. The study design was a double-blinded, crossover trial. It consisted of a 1-week study medication period, a 2-week washout period and a further 1-week on study medication. Bryonia in a 12c potency was chosen as the trial medication. Participants were recruited in the UK from the Faculty of Homeopathy and the Society of Homeopaths and were currently healthy, aged 18 or over with at least three years’ clinical experience of homeopathy. Of the 500 recruitment packs despatched, we received 88 responses (17.6%). Seventy homeopaths were randomised of whom 50 completed the trial. In the main analysis 60% correctly identified the bottle containing Bryonia (n ¼ 40; 95% confidence interval 43%, 75%; P ¼ 0.27). There was evidence of an order effect: subjects were much more likely to think they received active Bryonia in the first rather than the second period. In this study a promising trend was observed that symptoms reported by some homeopaths may not be completely attributable to placebo. A multi-national, large-scale trial will be required to investigate this phenomena with adequate statistical power. British Homeopathic Journal (2001) 90, 126–130. Keywords: homeopathy; randomised controlled trial; homeopathic pathogenetic trial; provings
Introduction Homeopathy is based on the principle of ‘let like be cured by like’: a patient presenting with a particular set of symptoms is prescribed a medicine that would cause those symptoms if given to a healthy individual. This information is obtained from ‘homeopathic *Correspondence: R McCarney, Academic Unit, Royal London Homoeopathic Hospital, Great Ormond Street, London WC1N 3HR, UK. E-mail: [email protected] Received 12 October 2000; revised 10 January 2001; accepted 1 February 2001
pathogenetic trials’ (HPTs), more commonly known as ‘provings’. Whilst some HPTs involve the administration of pharmacological doses, most contemporary studies are conducted with homeopathically prepared medicines. Many of the medicines used in homeopathy are diluted beyond the point, which according to Avogadro’s law, they could contain molecules of the original source material. Such medicines have never been demonstrated unequivocally to cause specific symptoms, rather than non-specific placebo symptoms, in healthy volunteers. The most rigorous overview of drug provings to date is that of Dantas and Fisher, who
Can homeopaths detect homeopathic medicines A Vickers et al
127
undertook a systematic review of drug provings published in the UK.1 The authors retrieved 45 reports of drug provings of 44 different medicines in 779 volunteers. The methodological quality of the reports was very low. The number of symptoms reported per subject was related to the methodological quality of the drug proving with only one of the four best studies reporting more proving symptoms in subjects than controls. Two studies published since this review was undertaken have similarly failed to find strong evidence of a drug proving effect.2,3 One trial of a homeopathic prophylactic did find statistically significant differences between placebo and active treatment for the number of adverse events reported by subjects.4 However, the reported analysis was unplanned and may have been data driven. Typically contemporary HPTs administer homeopathic substances to a small number of practising homeopaths or students of homeopathy. From this practice it is possible to generate a hypothesis: healthy homeopaths can distinguish that medicine from placebo by taking both and observing the effects. If this hypothesis is true, it supports an effect of homeopathic medicines greater than that of placebo. If it is false, it casts serious doubt on the contemporary homeopathic knowledge base and on HPTs as currently practised. This study was designed as a pilot to test the feasibility of a larger scale trial. The main objective was to test this design to determine resource implications and levels of recruitment. The formal question asked in the pilot was: can healthy homeopaths distinguish Bryonia in a 12c potency from placebo by taking both preparations and observing their effects?
Methods The study design was a double-blinded, cross-over trial. It consisted of a 1-week study medication period, a 2-week washout period and a further 1-week on study medication. Ethical approval was sought and obtained from the University College London and University College London Hospital joint committee. Participants were recruited from the Faculty of Homeopathy (FHom), which represents registered health professionals who practice homeopathy, and the Society of Homeopaths (SocHom), which represents non-medically qualified homeopaths. Information about the study was mailed to all 300 full Members of the Faculty of Homeopathy and a random sample of 200 Society of Homeopath members. Subjects were able to request an interview with an experienced research nurse or senior researcher if they wanted to ask questions or required further information before deciding whether to join the trial. Participants were practising homeopaths aged 18 or over with at least 3 years’ clinical experience of homeopathy. Exclusion criteria included self-report of
any perceptible illness; concurrent use of any form of medical, homeopathic or herbal drugs; any anticipated major life change events during the trial. Any homeopaths reporting illness or medication use were further investigated. Those judged by the study doctor to have a ‘stable’ condition or who had a ‘stable’ use of medication were not excluded. As a general rule, chronic medication use was allowed if of 3 months or greater duration; for example the use of hormone replacement therapy would be considered ‘stable’ whereas the use of analgesics and anti-emetics on an infrequent basis for migraine was considered ‘unstable’. Participants were sent two bottles of indistinguishable pills (one containing verum and the other placebo, in a random order), an instruction sheet, a card to report any adverse events to the study team, a consent form, a questionnaire and a ‘Freepost’ return envelope. The adverse events card asked participants to describe any symptoms that they felt may have been caused by the study medication, which required medical treatment. The instruction sheet detailed how to take the pills and some typical symptoms of Bryonia. Advice on filling in the questionnaire was given by using some hypothetical scenarios. Participants were instructed to take one pill, three times a day from bottle 1 for the first week. Then after a 2-week wash-out period, one pill, three times a day was to be taken from bottle 2 for the fourth week. In the event of a ‘proving reaction’ participants were instructed to stop taking pills from the appropriate bottle. Proving reactions were to be individually determined by each homeopath, which included reference to the information sheet detailing typical Bryonia symptoms. In addition, subjects could refer to any homeopathic literature they deemed relevant. The questionnaire asked which bottle the participant thought contained Bryonia; with what level of confidence they gave the answer to the first question (high=low=purely a guess). After consulting a number of homeopaths, Bryonia was chosen as the trial medication as it is a wellknown and commonly prescribed medicine. Moreover, recent (currently unpublished) provings were said by the study sponsor to have determined Bryonia proving symptoms to a high degree of validity. The trial medication, randomised and blinded in individual two-bottle (verum and placebo) sets, was supplied by Clinical Research HomInt. It was freshly prepared according to the German Homeopathic Pharmacopoeia (HAB) by Deutsche Homo¨opathie-Union (DHU) in Karlsruhe, Germany. Placebo was prepared in an identical fashion except that distilled water was used instead of the mother tincture. Treatment allocation was fully concealed from the researcher sending medication to participants. Clinical Research HomInt held the code. Placebo and active remedy were tested for indistinguishability by asking seven independent (non-study) homeopaths to state whether the contents of seven bottles were the same British Homeopathic Journal
Can homeopaths detect homeopathic medicines A Vickers et al
128
or different to a reference sample on the basis of appearance, smell and taste. Mean validity was 57% suggesting that the medication and placebo were indistinguishable, and full double-blindness was achieved. As the trial was designed as a pilot, no formal sample size calculations were made. The target of 80 enquiries was determined by resource constraints. The number of participants written to, randomised, and providing outcome data were recorded as well as the number requesting more information about the trial or requiring a telephone interview. Proportions and 95% confidence intervals were calculated for number of participants providing outcome data, and for level of confidence (high, low and purely a guess) and response (whether bottle 1 or 2 was said to contain Bryonia). A binomial comparison was made between the proportion of correct responses to the proportion expected by chance (50%). A P-value and 95% confidence interval was calculated. The main analysis excluded subjects who answered ‘purely a guess’ to the second question; missing data was taken as a ‘low confidence’ response. Two secondary analyses included all subjects and only those respondents with high confidence in their answer. Logistic regression analysis was undertaken to see if professional affiliation, number of years in practice or level of confidence predicted a correct response. Backward stepwise regression where a P-value of 0.05 is the criterion for keeping a variable in the model was conducted. All statistical analyses were conducted using the statistical software package Stata1 (Stata Corporation, TX). Complete, independent, blinded, double-entry of data was undertaken using automated consistency and logical checks. An independently appointed study monitor conducted an audit at the study centre at the end of the trial to check the consent forms, case report forms and outcome questionnaires of each patient. A monthly audit also took place to ensure the confidentiality and integrity of databases and paper records.
88 responses (17.6%). The response rate was higher amongst FHom members (n ¼ 71, 23.6%) than amongst SocHom members (n ¼ 16, 8.0%). A total of 70 homeopaths were randomised, of which 50 completed the trial. One FHom member requested a telephone interview and 16 FHom members required a telephone interview to clarify medication use or to clarify other data supplied. Of the 50 homeopaths returning data, 31 were male and 19 were female. The majority of homeopaths completing the trial were FHom members (n ¼ 42); the remaining eight were SocHom members. Ten of the Faculty homeopaths said that their judgement of which bottle contained Bryonia was ‘purely a guess’ and so were excluded from the main analysis. From this we can deduce that in a definitive trial which included only medically qualified homeopaths (such as the members of the FHom), we would expect about 10% (95% confidence interval 7.5%, 14.5%) of those sent a recruitment mailing to provide data for analysis. In the main analysis (excluding those who answered purely a guess — see Table 2), 60% responded correctly (n ¼ 40; 95% confidence interval 43%, 75%; P ¼ 0.27). In addition, analysing the responses of all participants (n ¼ 50), 56% responded correctly (95% confidence interval 41%, 70%, P ¼ 0.58). For those reporting ‘high confidence’ in their answer (n ¼ 15), 53% responded correctly (95% confidence interval 27%, 79%; P ¼ 1). Homeopaths reporting that their response was purely a guess answered 40% correctly. In the logistic regression model, neither years in practice, affiliation (FHom or SocHom) or level of confidence were predictive of a correct response. One further analysis looked at the possibility of an order effect; that whether the order in which the Table 2 Main analysis: participants who indicated their response was more than ‘purely a guess’ Level of confidence Response
High confidence n (%)
Low confidence* n (%)
Total n (%)
Results
Incorrect Correct Total
7.(46.7%) 8.(53.3%) 15.(100%)
9.(36%) 16.(64%) 25.(100%)
16.(40%) 24.(60%) 40.(100%)
Participant flow through the trial is given in Table 1. Of the 500 recruitment packs despatched, we received
*Includes one participant who did not indicate level of confidence.
Table 1 Participant flow through the trial Event Sent recruitment pack Replied Replied set at 100% Ineligible* Randomised Randomised set at 100% Withdrawn{ Completed trial
Faculty of Homoeopathy n.(%)
Society of Homoeopaths n.(%)
Total n.(%)
300 71.(23.7%) 71.(100%) 14.(19.7%) 57.(80.3%) 57.(100%) 15.(26.3%) 42.(73.7%)
200 16.(8.0%) 16.(100%) 3.(18.8%) 13.(81.2%) 13.(100%) 5.(38.5%) 8.(61.5%)
500 88.(17.6%) 88.(100%) 18.(20.5%) 70.(79.5%) 70.(100%) 20.(28.6%) 50.(71.4%)
*Self report of perceptible illness, concurrent use of medication, homeopathic or herbal drugs, anticipated major life change during trial. Lost to follow-up.(failed to return questionnaire), concomitant disease, eligibility change or withdrew consent.
{
British Homeopathic Journal
Can homeopaths detect homeopathic medicines A Vickers et al
129 Table 3 Order effect by level of confidence Level of confidence High Low Purely a guess Total
Number in category 15 25.* 10 50
Number stating bottle1contained Bryonia (%) 10 16 7 33
(66%) (64%) (70%) (66%)
*Includes one responder with missing data.
medications were presented had an effect on participant response. The proving literature3,5 indicates that more symptoms tend to be reported in the first weeks. This would make participants more likely to decide that verum was taken during the first phase. The analysis (see Table 3) found that irrespective of the treatment received, more than half of the whole sample stated bottle 1 contained Bryonia (n ¼ 33 of 50 (66%); 95% confidence interval 51%, 79%; P ¼ 0.03). The order effect remained significant even if those stating ‘purely a guess’ were excluded. Three participants spontaneously reported adverse events. These included: violent stomach cramps; tiredness; a cough; increased mucous production; muscular pain in the sacroiliac region; and muscle spasms in the lumbar region. All of these participants were willing to continue with the study and the symptoms soon abated in each case. Two of them were uncertain as to whether their symptoms were related to the trial medication; the other participant felt that the trial medication was the probable cause of her stomach cramps. In all three cases the symptoms were reported during, or in the washout period following, the Bryonia phase. Two of the three subjects completed the trial and subsequently correctly guessed which bottle contained the Bryonia.
Discussion The main purpose of this study was to pilot an innovative HPT in order to determine the feasibility of a larger trial. The trial was not resource intensive — after minor modifications to the questionnaire, no participant requested or required a telephone interview — and recruitment rates were adequate, if moderate. Recruitment was higher amongst FHom members (10.5%) than amongst SocHom members (4.0%). This may be because in general the medically qualified (FHom) homeopaths have a more ‘scientific’ background, and so may be more appreciative of the need to conduct such trials. The symptoms reported by the different subjects during the trial were unrelated. This suggests that if these were indeed caused by the study medicine, they were idiosyncratic proving reactions, rather than adverse events caused by the trial medication.
The main analysis and logistic regression model were underpowered and so we cannot rule out a significant effect of these variables. The results of the trial are encouraging therefore and clearly warrant the conduct of a full scale, definitive trial. The power calculation for the definitive trial needs to be approached with caution as the model is very sensitive to small changes. In the following examples, power is 90%, a ¼ 0.05 and an analysable response to the mailing of 10.5% in all cases. For an expected correct response rate of 60% we would need a sample size of 259 with an initial mailing of 2500. For an expected correct response rate of 57.5% we would need a sample size of 463 and an initial mailing of 4500, and for an expected correct response rate of 55% we would need a sample size of 1047 and an initial mailing to 10 000 homeopaths. Considering the low numbers in the pilot it would be appropriate to base the power calculation on this more conservative estimate for the number of correct responses. Such a study may be feasible, but only on a multinational basis. As there is no currently validated instrument available to screen subjects more likely to be sensitive to any remedy under investigation, the current approach is relatively inefficient. Some homeopaths may be able to identify from experience which homeopathic remedies they are likely to be sensitive to (that is previous proving or curative reactions). This is not incompatible with homeopathic theory in which the concept of individual sensitivity and idiosyncrasy is important. A modification to the current design involving the administration of these ‘individualised’ remedies on a cross-over basis is currently being developed. Though the trial was underpowered for the main analyses, preliminary conclusions can be drawn. The 95% confidence interval for correct guesses excludes the possibility that all, nearly all or even most homeopaths can distinguish a particular homeopathic medicine from placebo. The upper limit of the confidence interval for proportion of correct guesses was 75%. Given that we expect a rate of 50% by chance, we can estimate that at most half of the homeopaths taking part in an HPT will have symptoms that they can distinguish from placebo. However, the pilot study casts some doubt on contemporary HPT methodology where data from all or nearly all participants are used in analyses. The hypothesis tested in this HPT required the use of a cross-over design, in which the volunteers compared verum with placebo. In view of the observed strong order effect we considered the use of a parallel group design in the subsequent definitive trial. This would involve asking a somewhat different research question such as ‘can homeopaths determine whether a given bottle contains an ultramolecular dilution or placebo?’ Order effects have been observed in previous HPTs3,5 and may seem to be a structural problem in HPTs of cross-over or time series design. British Homeopathic Journal
Can homeopaths detect homeopathic medicines A Vickers et al
130
However, cross-over trials are insensitive to pure order effects6 because, as illustrated by this trial, the overall effect will be the average of the effect in the two strata (verum=placebo and placebo=verum). These findings therefore strengthen the argument in favour of the cross-over design because it does not lead to bias and is efficient. In this study a promising trend was observed that symptoms reported by some homeopaths may not be completely attributable to placebo. A multi-national, large scale trial will be required to investigate this phenomena with adequate statistical power. Also, the administration of ‘individualised’ homeopathic remedies on a double-blind basis could be explored to see if an ‘efficiency gain’ can be achieved.
Acknowledgements The authors gratefully acknowledge the help of Clinical Research HomInt (in particular Dr M Heger), who
British Homeopathic Journal
provided funding for the study and supplied the study medication.
References 1 Dantas F, Fisher P. A systematic review of homoeopathic pathogenetic trials (provings) published in the United Kingdom from 1945 to 1995. In: Ernst E, Hahn EG (eds). Homoeopathy: a Critical Appraisal. Oxford: Butterworth Heinemann, 1998, Chap 5, pp 69 – 97. 2 Walach H, Hieber S, Hieber E. Effects of belladonna 12cH and 30cH in healthy volunteers. In: Bastide M (ed). Signals and Images. Dordrech: Kluwer Academic Publishers, 1997. 3 Koster D, van Haselen R, Jansen G, Dicke M. Homeopathic medicine test according to the crossover design: from theory to practice. Br Hom J 1998; 87: 181 – 189. 4 Attena F, Toscano G, Agozzino E, Del Giudice N. A randomized trial in the prevention of influenza-like syndromes by homeopathic management. Revue d’Epidemiologie et de Sante Publique 1995; 43: 380 – 382. 5 Clover AM, Jenkins S, Campbell AC, Jenkins MD. Report on a proving of pulsatilla 3. Br Hom J 1980; 69: 134 – 147. 6 Senn S. Cross-over Trials in Clinical Research. Chichester: John Wiley and Sons, 1993.
ORIGINAL PAPER
Can homeopaths detect homeopathic medicines? A pilot study for a randomised, double-blind, placebo controlled investigation of the proving hypothesis A Vickers1, R McCarney1*, P Fisher1 and R van Haselen1 1
Academic Unit, Royal London Homoeopathic Hospital, London, UK
Homeopaths believe that a medicine, which causes a particular symptom in a healthy volunteer, will cure a similar symptom in a sick patient. From this phenomenon, it is possible to deduce a hypothesis: homeopaths should be able to distinguish a homeopathic medicine from a placebo by taking both and observing their effects. If true, this would support an effect of homeopathic medicines different from that of placebo. If false, it casts serious doubt on the contemporary homeopathic knowledge base and on homeopathic pathogenetic trials (HPTs) as currently practised. The study design was a double-blinded, crossover trial. It consisted of a 1-week study medication period, a 2-week washout period and a further 1-week on study medication. Bryonia in a 12c potency was chosen as the trial medication. Participants were recruited in the UK from the Faculty of Homeopathy and the Society of Homeopaths and were currently healthy, aged 18 or over with at least three years’ clinical experience of homeopathy. Of the 500 recruitment packs despatched, we received 88 responses (17.6%). Seventy homeopaths were randomised of whom 50 completed the trial. In the main analysis 60% correctly identified the bottle containing Bryonia (n ¼ 40; 95% confidence interval 43%, 75%; P ¼ 0.27). There was evidence of an order effect: subjects were much more likely to think they received active Bryonia in the first rather than the second period. In this study a promising trend was observed that symptoms reported by some homeopaths may not be completely attributable to placebo. A multi-national, large-scale trial will be required to investigate this phenomena with adequate statistical power. British Homeopathic Journal (2001) 90, 126–130. Keywords: homeopathy; randomised controlled trial; homeopathic pathogenetic trial; provings
Introduction Homeopathy is based on the principle of ‘let like be cured by like’: a patient presenting with a particular set of symptoms is prescribed a medicine that would cause those symptoms if given to a healthy individual. This information is obtained from ‘homeopathic *Correspondence: R McCarney, Academic Unit, Royal London Homoeopathic Hospital, Great Ormond Street, London WC1N 3HR, UK. E-mail: [email protected] Received 12 October 2000; revised 10 January 2001; accepted 1 February 2001
pathogenetic trials’ (HPTs), more commonly known as ‘provings’. Whilst some HPTs involve the administration of pharmacological doses, most contemporary studies are conducted with homeopathically prepared medicines. Many of the medicines used in homeopathy are diluted beyond the point, which according to Avogadro’s law, they could contain molecules of the original source material. Such medicines have never been demonstrated unequivocally to cause specific symptoms, rather than non-specific placebo symptoms, in healthy volunteers. The most rigorous overview of drug provings to date is that of Dantas and Fisher, who
Can homeopaths detect homeopathic medicines A Vickers et al
127
undertook a systematic review of drug provings published in the UK.1 The authors retrieved 45 reports of drug provings of 44 different medicines in 779 volunteers. The methodological quality of the reports was very low. The number of symptoms reported per subject was related to the methodological quality of the drug proving with only one of the four best studies reporting more proving symptoms in subjects than controls. Two studies published since this review was undertaken have similarly failed to find strong evidence of a drug proving effect.2,3 One trial of a homeopathic prophylactic did find statistically significant differences between placebo and active treatment for the number of adverse events reported by subjects.4 However, the reported analysis was unplanned and may have been data driven. Typically contemporary HPTs administer homeopathic substances to a small number of practising homeopaths or students of homeopathy. From this practice it is possible to generate a hypothesis: healthy homeopaths can distinguish that medicine from placebo by taking both and observing the effects. If this hypothesis is true, it supports an effect of homeopathic medicines greater than that of placebo. If it is false, it casts serious doubt on the contemporary homeopathic knowledge base and on HPTs as currently practised. This study was designed as a pilot to test the feasibility of a larger scale trial. The main objective was to test this design to determine resource implications and levels of recruitment. The formal question asked in the pilot was: can healthy homeopaths distinguish Bryonia in a 12c potency from placebo by taking both preparations and observing their effects?
Methods The study design was a double-blinded, cross-over trial. It consisted of a 1-week study medication period, a 2-week washout period and a further 1-week on study medication. Ethical approval was sought and obtained from the University College London and University College London Hospital joint committee. Participants were recruited from the Faculty of Homeopathy (FHom), which represents registered health professionals who practice homeopathy, and the Society of Homeopaths (SocHom), which represents non-medically qualified homeopaths. Information about the study was mailed to all 300 full Members of the Faculty of Homeopathy and a random sample of 200 Society of Homeopath members. Subjects were able to request an interview with an experienced research nurse or senior researcher if they wanted to ask questions or required further information before deciding whether to join the trial. Participants were practising homeopaths aged 18 or over with at least 3 years’ clinical experience of homeopathy. Exclusion criteria included self-report of
any perceptible illness; concurrent use of any form of medical, homeopathic or herbal drugs; any anticipated major life change events during the trial. Any homeopaths reporting illness or medication use were further investigated. Those judged by the study doctor to have a ‘stable’ condition or who had a ‘stable’ use of medication were not excluded. As a general rule, chronic medication use was allowed if of 3 months or greater duration; for example the use of hormone replacement therapy would be considered ‘stable’ whereas the use of analgesics and anti-emetics on an infrequent basis for migraine was considered ‘unstable’. Participants were sent two bottles of indistinguishable pills (one containing verum and the other placebo, in a random order), an instruction sheet, a card to report any adverse events to the study team, a consent form, a questionnaire and a ‘Freepost’ return envelope. The adverse events card asked participants to describe any symptoms that they felt may have been caused by the study medication, which required medical treatment. The instruction sheet detailed how to take the pills and some typical symptoms of Bryonia. Advice on filling in the questionnaire was given by using some hypothetical scenarios. Participants were instructed to take one pill, three times a day from bottle 1 for the first week. Then after a 2-week wash-out period, one pill, three times a day was to be taken from bottle 2 for the fourth week. In the event of a ‘proving reaction’ participants were instructed to stop taking pills from the appropriate bottle. Proving reactions were to be individually determined by each homeopath, which included reference to the information sheet detailing typical Bryonia symptoms. In addition, subjects could refer to any homeopathic literature they deemed relevant. The questionnaire asked which bottle the participant thought contained Bryonia; with what level of confidence they gave the answer to the first question (high=low=purely a guess). After consulting a number of homeopaths, Bryonia was chosen as the trial medication as it is a wellknown and commonly prescribed medicine. Moreover, recent (currently unpublished) provings were said by the study sponsor to have determined Bryonia proving symptoms to a high degree of validity. The trial medication, randomised and blinded in individual two-bottle (verum and placebo) sets, was supplied by Clinical Research HomInt. It was freshly prepared according to the German Homeopathic Pharmacopoeia (HAB) by Deutsche Homo¨opathie-Union (DHU) in Karlsruhe, Germany. Placebo was prepared in an identical fashion except that distilled water was used instead of the mother tincture. Treatment allocation was fully concealed from the researcher sending medication to participants. Clinical Research HomInt held the code. Placebo and active remedy were tested for indistinguishability by asking seven independent (non-study) homeopaths to state whether the contents of seven bottles were the same British Homeopathic Journal
Can homeopaths detect homeopathic medicines A Vickers et al
128
or different to a reference sample on the basis of appearance, smell and taste. Mean validity was 57% suggesting that the medication and placebo were indistinguishable, and full double-blindness was achieved. As the trial was designed as a pilot, no formal sample size calculations were made. The target of 80 enquiries was determined by resource constraints. The number of participants written to, randomised, and providing outcome data were recorded as well as the number requesting more information about the trial or requiring a telephone interview. Proportions and 95% confidence intervals were calculated for number of participants providing outcome data, and for level of confidence (high, low and purely a guess) and response (whether bottle 1 or 2 was said to contain Bryonia). A binomial comparison was made between the proportion of correct responses to the proportion expected by chance (50%). A P-value and 95% confidence interval was calculated. The main analysis excluded subjects who answered ‘purely a guess’ to the second question; missing data was taken as a ‘low confidence’ response. Two secondary analyses included all subjects and only those respondents with high confidence in their answer. Logistic regression analysis was undertaken to see if professional affiliation, number of years in practice or level of confidence predicted a correct response. Backward stepwise regression where a P-value of 0.05 is the criterion for keeping a variable in the model was conducted. All statistical analyses were conducted using the statistical software package Stata1 (Stata Corporation, TX). Complete, independent, blinded, double-entry of data was undertaken using automated consistency and logical checks. An independently appointed study monitor conducted an audit at the study centre at the end of the trial to check the consent forms, case report forms and outcome questionnaires of each patient. A monthly audit also took place to ensure the confidentiality and integrity of databases and paper records.
88 responses (17.6%). The response rate was higher amongst FHom members (n ¼ 71, 23.6%) than amongst SocHom members (n ¼ 16, 8.0%). A total of 70 homeopaths were randomised, of which 50 completed the trial. One FHom member requested a telephone interview and 16 FHom members required a telephone interview to clarify medication use or to clarify other data supplied. Of the 50 homeopaths returning data, 31 were male and 19 were female. The majority of homeopaths completing the trial were FHom members (n ¼ 42); the remaining eight were SocHom members. Ten of the Faculty homeopaths said that their judgement of which bottle contained Bryonia was ‘purely a guess’ and so were excluded from the main analysis. From this we can deduce that in a definitive trial which included only medically qualified homeopaths (such as the members of the FHom), we would expect about 10% (95% confidence interval 7.5%, 14.5%) of those sent a recruitment mailing to provide data for analysis. In the main analysis (excluding those who answered purely a guess — see Table 2), 60% responded correctly (n ¼ 40; 95% confidence interval 43%, 75%; P ¼ 0.27). In addition, analysing the responses of all participants (n ¼ 50), 56% responded correctly (95% confidence interval 41%, 70%, P ¼ 0.58). For those reporting ‘high confidence’ in their answer (n ¼ 15), 53% responded correctly (95% confidence interval 27%, 79%; P ¼ 1). Homeopaths reporting that their response was purely a guess answered 40% correctly. In the logistic regression model, neither years in practice, affiliation (FHom or SocHom) or level of confidence were predictive of a correct response. One further analysis looked at the possibility of an order effect; that whether the order in which the Table 2 Main analysis: participants who indicated their response was more than ‘purely a guess’ Level of confidence Response
High confidence n (%)
Low confidence* n (%)
Total n (%)
Results
Incorrect Correct Total
7.(46.7%) 8.(53.3%) 15.(100%)
9.(36%) 16.(64%) 25.(100%)
16.(40%) 24.(60%) 40.(100%)
Participant flow through the trial is given in Table 1. Of the 500 recruitment packs despatched, we received
*Includes one participant who did not indicate level of confidence.
Table 1 Participant flow through the trial Event Sent recruitment pack Replied Replied set at 100% Ineligible* Randomised Randomised set at 100% Withdrawn{ Completed trial
Faculty of Homoeopathy n.(%)
Society of Homoeopaths n.(%)
Total n.(%)
300 71.(23.7%) 71.(100%) 14.(19.7%) 57.(80.3%) 57.(100%) 15.(26.3%) 42.(73.7%)
200 16.(8.0%) 16.(100%) 3.(18.8%) 13.(81.2%) 13.(100%) 5.(38.5%) 8.(61.5%)
500 88.(17.6%) 88.(100%) 18.(20.5%) 70.(79.5%) 70.(100%) 20.(28.6%) 50.(71.4%)
*Self report of perceptible illness, concurrent use of medication, homeopathic or herbal drugs, anticipated major life change during trial. Lost to follow-up.(failed to return questionnaire), concomitant disease, eligibility change or withdrew consent.
{
British Homeopathic Journal
Can homeopaths detect homeopathic medicines A Vickers et al
129 Table 3 Order effect by level of confidence Level of confidence High Low Purely a guess Total
Number in category 15 25.* 10 50
Number stating bottle1contained Bryonia (%) 10 16 7 33
(66%) (64%) (70%) (66%)
*Includes one responder with missing data.
medications were presented had an effect on participant response. The proving literature3,5 indicates that more symptoms tend to be reported in the first weeks. This would make participants more likely to decide that verum was taken during the first phase. The analysis (see Table 3) found that irrespective of the treatment received, more than half of the whole sample stated bottle 1 contained Bryonia (n ¼ 33 of 50 (66%); 95% confidence interval 51%, 79%; P ¼ 0.03). The order effect remained significant even if those stating ‘purely a guess’ were excluded. Three participants spontaneously reported adverse events. These included: violent stomach cramps; tiredness; a cough; increased mucous production; muscular pain in the sacroiliac region; and muscle spasms in the lumbar region. All of these participants were willing to continue with the study and the symptoms soon abated in each case. Two of them were uncertain as to whether their symptoms were related to the trial medication; the other participant felt that the trial medication was the probable cause of her stomach cramps. In all three cases the symptoms were reported during, or in the washout period following, the Bryonia phase. Two of the three subjects completed the trial and subsequently correctly guessed which bottle contained the Bryonia.
Discussion The main purpose of this study was to pilot an innovative HPT in order to determine the feasibility of a larger trial. The trial was not resource intensive — after minor modifications to the questionnaire, no participant requested or required a telephone interview — and recruitment rates were adequate, if moderate. Recruitment was higher amongst FHom members (10.5%) than amongst SocHom members (4.0%). This may be because in general the medically qualified (FHom) homeopaths have a more ‘scientific’ background, and so may be more appreciative of the need to conduct such trials. The symptoms reported by the different subjects during the trial were unrelated. This suggests that if these were indeed caused by the study medicine, they were idiosyncratic proving reactions, rather than adverse events caused by the trial medication.
The main analysis and logistic regression model were underpowered and so we cannot rule out a significant effect of these variables. The results of the trial are encouraging therefore and clearly warrant the conduct of a full scale, definitive trial. The power calculation for the definitive trial needs to be approached with caution as the model is very sensitive to small changes. In the following examples, power is 90%, a ¼ 0.05 and an analysable response to the mailing of 10.5% in all cases. For an expected correct response rate of 60% we would need a sample size of 259 with an initial mailing of 2500. For an expected correct response rate of 57.5% we would need a sample size of 463 and an initial mailing of 4500, and for an expected correct response rate of 55% we would need a sample size of 1047 and an initial mailing to 10 000 homeopaths. Considering the low numbers in the pilot it would be appropriate to base the power calculation on this more conservative estimate for the number of correct responses. Such a study may be feasible, but only on a multinational basis. As there is no currently validated instrument available to screen subjects more likely to be sensitive to any remedy under investigation, the current approach is relatively inefficient. Some homeopaths may be able to identify from experience which homeopathic remedies they are likely to be sensitive to (that is previous proving or curative reactions). This is not incompatible with homeopathic theory in which the concept of individual sensitivity and idiosyncrasy is important. A modification to the current design involving the administration of these ‘individualised’ remedies on a cross-over basis is currently being developed. Though the trial was underpowered for the main analyses, preliminary conclusions can be drawn. The 95% confidence interval for correct guesses excludes the possibility that all, nearly all or even most homeopaths can distinguish a particular homeopathic medicine from placebo. The upper limit of the confidence interval for proportion of correct guesses was 75%. Given that we expect a rate of 50% by chance, we can estimate that at most half of the homeopaths taking part in an HPT will have symptoms that they can distinguish from placebo. However, the pilot study casts some doubt on contemporary HPT methodology where data from all or nearly all participants are used in analyses. The hypothesis tested in this HPT required the use of a cross-over design, in which the volunteers compared verum with placebo. In view of the observed strong order effect we considered the use of a parallel group design in the subsequent definitive trial. This would involve asking a somewhat different research question such as ‘can homeopaths determine whether a given bottle contains an ultramolecular dilution or placebo?’ Order effects have been observed in previous HPTs3,5 and may seem to be a structural problem in HPTs of cross-over or time series design. British Homeopathic Journal
Can homeopaths detect homeopathic medicines A Vickers et al
130
However, cross-over trials are insensitive to pure order effects6 because, as illustrated by this trial, the overall effect will be the average of the effect in the two strata (verum=placebo and placebo=verum). These findings therefore strengthen the argument in favour of the cross-over design because it does not lead to bias and is efficient. In this study a promising trend was observed that symptoms reported by some homeopaths may not be completely attributable to placebo. A multi-national, large scale trial will be required to investigate this phenomena with adequate statistical power. Also, the administration of ‘individualised’ homeopathic remedies on a double-blind basis could be explored to see if an ‘efficiency gain’ can be achieved.
Acknowledgements The authors gratefully acknowledge the help of Clinical Research HomInt (in particular Dr M Heger), who
British Homeopathic Journal
provided funding for the study and supplied the study medication.
References 1 Dantas F, Fisher P. A systematic review of homoeopathic pathogenetic trials (provings) published in the United Kingdom from 1945 to 1995. In: Ernst E, Hahn EG (eds). Homoeopathy: a Critical Appraisal. Oxford: Butterworth Heinemann, 1998, Chap 5, pp 69 – 97. 2 Walach H, Hieber S, Hieber E. Effects of belladonna 12cH and 30cH in healthy volunteers. In: Bastide M (ed). Signals and Images. Dordrech: Kluwer Academic Publishers, 1997. 3 Koster D, van Haselen R, Jansen G, Dicke M. Homeopathic medicine test according to the crossover design: from theory to practice. Br Hom J 1998; 87: 181 – 189. 4 Attena F, Toscano G, Agozzino E, Del Giudice N. A randomized trial in the prevention of influenza-like syndromes by homeopathic management. Revue d’Epidemiologie et de Sante Publique 1995; 43: 380 – 382. 5 Clover AM, Jenkins S, Campbell AC, Jenkins MD. Report on a proving of pulsatilla 3. Br Hom J 1980; 69: 134 – 147. 6 Senn S. Cross-over Trials in Clinical Research. Chichester: John Wiley and Sons, 1993.