UvA-DARE (Digital Academic Repository)

Transcription

1 UvA-DARE (Digital Academic Repository) Methodological issues in studies of major depression and schizophrenia. Implications for the "Committee for proprietary medicinal products" guidelines Storosum, J.G. Link to publication Citation for published version (APA): Storosum, J. G. (2002). Methodological issues in studies of major depression and schizophrenia. Implications for the "Committee for proprietary medicinal products" guidelines. General rights It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons). Disclaimer/Complaints regulations If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible. UvA-DARE is a service provided by the library of the University of Amsterdam ( Download date: 22 Mar 2019

2 Schizophrenia: Do we really need placebo-controlled studiess? Storosumm JG, Elferink AJA, Zwieten van BJ. Schizophrenia: Do we really need placebo-controlledd studies? European Neuropsychopharmacol 1998;8:

3 61 1 Summary y Objective:: To investigate whether placebo control is necessary to prove efficacy in short-termm studies in schizophrenia. Design:: This study compares the efficacy results of placebo-controlled studies versuss positive controlled studies, that is controlled studies without a placebo control, inn the short-term treatment of chronic schizophrenia. Results.. Concerning mean improvement on the BPRS, the placebo arms showed in twoo cases a worsening, in one case almost no change, and in the remaining studies (6)) the improvement was between 1 and 5%. The percentage mean improvement in thee haloperidol arms of the placebo controlled studies was comparable to the percentagee mean improvement in the corresponding arms of the non-placebocontrolledd studies. The highest percentage responders in the placebo-groups was 43% andd the lowest was 6%. Moreover the responder rates in the atypical antipsychotic andd haloperidol -arms of the non-placebo-controlled studies were in two of the three studiess in the same order of magnitude as the responder rates of the placebo-arms in thee placebo-controlled studies Thee overall dropout rates in the placebo arms was between 48% and 80% and were higherr than the drop out rates in the atypical neuroleptic arms and haloperidol arms of thee placebo-controlled studies. The dropout rates due to an insufficient response in the atypicall neuroleptic arms and haloperidol arms of the non- placebo controlled studies weree lower as compared to corresponding treatment arms of the placebo-controlled studies.. Conclusion:: In contrast to mean improvement on the BPRS, responder rates in the placeboo arms varied from study to study considerable. Responder rates in the atypical antipsychoticc and haloperidol -arms of the non-placebo-controlled studies were in two off the three studies in the same order of magnitude as the responder rates of the placebo-armss in the placebo-controlled studies. These results indicate that placebo controll is necessary. Moreover as responders are a more clinically relevant outcome measuree as compared to mean improvement on a rating scale, placebo-controlled studiess are still needed. However, consensus on responder definition should be agreedd upon. For the moment, alternatives to avoid placebo-controlled studies are inadequatee to demonstrate efficacy in studies with schizophrenic patients.

4 62 2 Introduction n Theree is debate about the use of placebo in clinical trials (Rothman and Michels, 1994).. On the one hand, a placebo-control is necessary to define the "absolute" effectt of a product and as an internal validation of the trial, on the other hand, the usee of placebo raises ethical problems, especially if an effective treatment is availablee and when changes are more or less irreversible in case of a progressive disease.. Also, in placebo-controlled trials adverse effects due to a product can be distinguishedd from disease/disorder related events. In schizophrenia the discussion aboutt the need and feasibility of placebo in short-term studies and/or maintenance studiess has not been settled. This is partly due to ethical concerns and partly because,, after a long time of silence, new products are being developed, using modernn trial methodology. At the same time, the concepts of the disorder, the diagnosticc criteria and efficacy criteria changed considerably. Consequently, historicall information is not very useful which implies that, in principle, placebocontrolledd trials are required for establishing efficacy of a new product. Inn the last few years new products, the "atypical neuroleptics", have been introduced. Thee pivotal clinical trials performed with these compounds incorporated a placeboarmm in many cases. Therefore, the data of these pivotal studies could give an evidencee based (Evidence-Based Medicines Working Group, 1992) answer to the questionn whether placebo-controlled studies are necessary for assessing short term efficacyy in schizophrenia.

5 63 3 Methods.. Thiss study compares the efficacy results of placebo-controlled studies versus those off positive controlled studies, that is controlled studies without a placebo control, in short-termm treatment of chronic schizophrenia. All efficacy studies concerning atypicall neuroleptics recently (from 1992) submitted to the Medicines Evaluation Boardd of the Netherlands from the original registration files were screened. To be includedd the studies had to be randomised, double-blinded, last 6 to 8 weeks and thee patients enrolled in the studies had to be comparable. Furthermore, a baseline andd endpoint BPRS or PANSS had to be available. Arbitrary, the number of patients includedd in the non-placebo controlled studies should be at least 50 per treatment arm.. Thee BPRS score was chosen as outcome parameter for comparing the efficacy resultss of placebo-controlled studies versus those of positive controlled studies. In studiess where the PANSS was used instead of the BPRS, the results on the PANSS weree converted to the BPRS. Thee mean BPRS score and the responder rates were compared between the differentt studies. A responder was defined on basis of the BPRS score (see table I), ass could be found in the original registration files. Thee mean BPRS score and responder rate were based on an intention-to-treat analysiss with the last observation carried forward. This type of analyses counteracts thee bias that may arise from excluding patients who withdrawn. The intention-to-treat efficacyy population consisted of all patients who were randomised to receive doubleblindd treatment, took at least one dose, and provided at least one efficacy evaluation onn therapy. As some of these studies are not published yet, all studies were anonymised.. Anonymised were the investigators, the name of the active compounds andd the exact dosages. Some studies had more than one atypical neuroleptic arm and onee study had several haloperidol-arms. Also to lessen reducibility, only the placebo dataa and data of the reference compound (haloperidol) and investigation compounds inn the dosage arms giving the highest percentage improvement are shown (see also tablee 1). As this is a descriptive qualitative study no formal statistical analyses were performed.. Results s Twelvee studies fulfilled the inclusion criteria of this survey (study 7 had a duration of 40 days);; nine placebo-controlled and three positive-controlled studies, here defined as controlledd studies without a placebo-arm. Table 1 gives an overview of these twelve studies.. In all studies patients diagnosed as schizophrenic were included. In two studiess (no. 8 and 11) patients with schizo-affective disorder were included as well. Figuree 1 shows the mean BPRS score at baseline and at the endpoint over the studies,, for the placebo-arms (figure 1A), best atypical neuroleptic (figure 1B) and for thee best haloperidol-arms (figure 1C) separately. Inn table 2 the lowest and highest value of mean baseline score, mean percentage improvementt from baseline and percentage dropouts of the treatment-arms of all studiess included are shown. Thee lowest mean baseline scores was 31.5 points andd highest mean baseline score wass 42.6 points in the placebo-controlled studies. The lowest mean baseline score

6 64 4 wass 30.1 points and the highest mean baseline score was 42.3 points in the nonplacebo-controlledd studies. The percentage improvement in the haloperidol arms of the placeboo controlled studies were comparable to the percentage improvement in the correspondingg arms of the non-placebo-controlled studies. Figuree 2 shows the responder rates of the placebo-arms, best atypical neuroleptic armss and best haloperidol-arms. In the studies under investigation responders were definedd as at least 20 % reduction on the BPRS total score in studies were the publishedd scale was used and 40% reduction in the other studies. In study 3,4,7 and 88 responders were defined as a 40% reduction on the BPRS total score at endpoint andd a CGI-severity score < 3. The percentage responders in the placebo-groups rangedd from 6% to 43%. Responders rates of the atypical neuroleptics and haloperidol groupss in the placebo controlled studies were higher than those in the non-placebo controlledd studies. Responder rates in the atypical neuroleptic and haloperidol arms in thee non-placebo controlled studies were in the same order of magnitude as the responderr rates in the placebo arms of the placebo-controlled studies. Dropoutt rates due to an insufficient response are given in figure 3. The overall dropout ratess in the placebo arms ranged from 48% to 80% and were higher as those in the atypicall neuroleptic arms and haloperidol arms of the placebo-controlled studies (see tablee 2). The dropout rates due to an insufficient response in the atypical neuroleptic armss and haloperidol arms of the non- placebo controlled studies were lower as comparedd to corresponding treatment arms of the placebo-controlled studies (see figuree 3). Discussion n Concerningg the use of placebo, it is remarkable that the mean improvement in the placebo-groupss was quite small and stable over all nine placebo controlled studies (figuree 1). In two cases there was a worsening, in one case almost no change, and in thee remainder improvement was between 1 and 5%. However the SD's of the mean improvementt were high and skewed. In the placebo-controlled studies haloperidol was usedd as an internal validation of the study analogous to imipramine which is often used forr the same purpose in studies with antidepressants. This means that if under haloperidoll the expected effect is not shown, there can be doubt about the validity of thee study. Thee mean improvement in the haloperidol -arms of the non-placebo-controlled studies didd not differ substantially from those of the corresponding arms in the placebocontrolledd studies (figure 1). This can not be attributed to the lower percentage of drop outss in the haloperidol arms of the non placebo-controlled studies compared to the correspondingg arms of the placebo controlled studies. This should only lead to a higher effectt in the haloperidol arms of the non placebo controlled studies, which was not the case.. So it is likely that in the non placebo-controlled studies the placebo effect was alsoo small and stable. Consequently one might conclude that placebo control is not necessaryy in future trials. Inn addition, the percentage dropouts in the placebo-controlled studies was considerablee higher in the placebo-groups and atypical compound groups compared to thee non-placebo-controlled studies. In conclusion, it seems that there is much to be saidd for abandoning placebo-controlled studies in schizophrenic patients.

7 65 5 However,, if responder rates are used as outcome measure a different opinion is forced upon.. Responder rates indicate that the placebo-response varied considerably from studyy to study. This variation is larger than was expected by the difference in responderr definitions (see table 1) Furthermore,, the responder rates in the atypical antipsychotic and haloperidol -arms of thee non-placebo-controlled studies were in two of the three studies in the same order off magnitude as that of the placebo-arms in the placebo-controlled studies (figure 2). Thee lower dropout rates due to an insufficient response in the non-placebo-controlled studiess (figure 3) compared to the placebo controlled studies should have resulted in higherr response rates, which was not the case. Therefore,, if responders are used as outcome parameter placebo controlled studies remainn necessary. It may be argued that responder rates are less sensitive than the meann BPRS score but one also might argue that when no statistically significant differencess can be found on responder rates, a product is not showing a clinically relevantt effect. Increasing sensitivity for showing treatment differences is not equivalentt to establishing clinical relevance treatment differences. For instance, how shouldd an improvement from 34 to 26 points (study 1, haloperidol group) on the BPRS bee interpreted? Is it an 1 point improvement on 8 of the 18 items of the BPRS or a 2 pointss improvement on 4 of the 18 items of the BPRS? Instead stating that under placeboo responder rate was 30% versus 60% under active treatment arms makes moree sense in terms of clinical relevance. Responder rate is an easily understood measuree which is intuitively acceptable both for the sufferer and for the clinician (Reportt of consensus meeting, 1995). Unfortunately,, at the moment, there is no generally accepted definition of responders inn schizophrenia. AA reduction of a certain percentage in the score on the pivotal severity scale is an arbitraryy but helpful measure that has been widely used to define responders as a measuree of clinical relevance. Kane et all (1988) used as definition for responders at leastt 20% reduction on the BPRS plus either a post-treatment CGI = 3 or a posttreatmentt score of 35 on the BPRS or lower. In the published respiridone studies (Carmann et al., 1996) responders were defined as patients with at least 20% reduction onn the PANSS. Thee patients included in the study of Kane however, were therapy-resistant. It is very questionablee whether extrapolation of this is permissible to non-therapy resistant patients.. In our opinion responders should be defined in advance of a study by experts inn the field of schizophrenia and should depend on the course of the patients included andd the severity of the disorder. Thee use of placebo in schizophrenia may raise ethical problems. Moreover, the dropoutt rate in placebo-controlled studies in schizophrenic patients is unacceptable high.. In general there are three options to avoid placebo-controlled studies. First,, when a dose response effect can be demonstrated, the pharmacological effect of ann product may be considered proven. However, a dose response in schizophrenia withh regard to efficacy is difficult to demonstrated as can be illustrated in a recently publishedd study by Peuskens, In this 8 week double-blind parallel group study thee efficacy of different doses risperidone was compared to that of haloperidol. Table 3 presentss the results of this study. Responders were defined as at least 20% improvementt from baseline on the BPRS.

8 66 6 Concerningg responders," the key efficacy parameter", no statistically significant differencess between treatment groups were found. Treatment with risperidone at the dosee of 4 mg, 8 mg and 16 mg resulted in a significant greater improvement in mean changee from baseline than risperidone 1 mg. However, a dose of 12 mg risperidone didd not show a statistically significant difference in improvement as compared to 1 mg norr did haloperidol 10 mg. A positive dose response effect, whether linear or U- shaped,, could not be demonstrated and although many patients were included in this studyy efficacy was by no means demonstrated. Thus although in theory a dose responsee study is a good model for demonstrating efficacy, for the indication schizophrenia,, the probability of getting a result is low. Second,, placebo-controlled studies could be omitted when it is demonstrated that the neww product is consistently superior to a product that is approved for the same indication,, which is considered the gold standard and which is given in the correct dosee (regimen). In studies with schizophrenia haloperidol is often chosen as the active comparatorr and as such haloperidol is implicitly considered the gold standard. Haloperidoll is an effective compound in the treatment of schizophrenia. However, it is questionablee whether haloperidol can fulfil the role as gold standard as in the studies reportedd here the effect of haloperidol is rather small. This could be due to, for instance,, patients selection or to the dose of haloperidol. Moreover, there is discussion concerningg negative symptoms versus extrapyramidal symptoms (Möller et al., 1994). AA compound that causes less extrapyramidal symptoms as compared to haloperidol cann give an improvement on the negative PANSS score, which can wrongly lead to the conclusionn that this new compound has demonstrated a direct effect on negative symptoms.. Furthermore, the optimal dose of haloperidol is not well-established. Moreover,, there is always a chance that superiority to haloperidol cannot be demonstrated.. Third,, placebo-controlled studies can be omitted when equal efficacy could be demonstrated,, within small margins. The large number of patients per treatment arm necessaryy to demonstrate equal efficacy is often a problem. This number depends on thee difference that is still accepted to considered two product equally effective. Moreover,, it should be ensured that the trial set-up is such that a difference with placeboo indeed could have been shown, if present. This may be extremely difficult in casee of a high or variable placebo-effect. Furthermore, there is the problem that the rolee of haloperidol as the gold standard is questionable, as discussed above. Conclusion n Basedd upon the mean BPRS-improvement, there is much to say for abandoning placebo-controlledd clinical trials in schizophrenia. The mean change from baseline BPRS-scoree under placebo is small and stable, whereas the improvement from baselinee BPRS-score in the haloperidol-arm is significantly better compared to placebo.. Most important this improvement under haloperidol seems to be independent fromm whether a study included a placebo-arm or not. An additional advantage may be thatt the efficiency of these studies is increased due to less higher drop-out rate in nonplaceboo controlled studies compared to placebo-controlled studies. However based uponn responder rates, the conclusion stated above is not endorsed. Responderr rates in the placebo-arms varied considerable from study to study. Moreoverr responder rates in the placebo-arms of the placebo-controlled studies were

9 67 7 inn the same order of magnitude as the responder rates in the haloperidol-arms of the non-placeboo controlled studies. This is unexpected as, given the lower drop-out rate duee to inefficacy in the non-placebo controlled studies, the expected responder rates in thee haloperidol arms should have been higher. Thee discrepancy in effect between mean BPRS improvement and response rates may bee due to the fact that a responder rate is less sensitive in showing treatment differencee than the mean BPRS score. However, an increased sensitivity for showing treatmentt differences is not equivalent to establishing clinical relevance treatment differences.. Whenn no statistically significant differences can be found on responder rates, a product iss not showing a clinically relevant effect. Therefore, placebo controlled studies remain necessary.. Moreover, consensus on responder definition should be agreed upon. Forr the moment alternatives to avoid placebo-controlled studies can not give a solution withh regard to demonstrate efficacy in studies with schizophrenic patients. Acknowledgement: : Thiss paper has been sent to the companies concerned. We are indebted to the companiess for giving permission for publication.

10 68 8 References s Carmann, J., Peuskens, J. and Vangeneugden, A. (1996) Risperidone in the treatment off negative symptoms of schizophrenia: a meta-analysis. Int. Clin. Psychopharmacol. 10, Evidence-Basedd Medicine Working Group. Evidence-based medicine. A new approach too teaching the practice of medicine (1992) JAMA 268, Kane,, J.M., Honigfeld G, Singer J., Meltzer A. and the Clozaril Collaborative Study Groupp (1988) Clozapine for the treatment resistant schizophrenic: a double-blind comparisonn with chlorpromazine. Arch. Gen. Psychiatry 45, Möller,, H.J., van Praag,H.M. Aufdembrinke, B., Baily, P., Barnes, T.R.E., Beck, J., Bentsenn H., Eich, F.X., Farrow, L, Fleischhacker, W.W., Gerlach, J. f Grafford, K. r Hentschel,, B., Hertkorn, A., Heylen, S., Lecrubier,. Y., Leonard, J.P., McKenna, P., Maier,, W., Pederson, V., Rappard, A., Rein, W., Ryan, J., Sloth Nielsen, M., Stieglitz, R.-D.,, Wegener, G., Wilson, J. (1994) Negative symptoms in schizophrenia: considerationss for clinical trials. Psychopharmaocology 115, Peuskens,, J. (1995) Risperidone in the treatment of patients with Chronic schizophrenia:: a multi-national, multicentre, double-blind, parallel-group study versus haloperidol.. Br. J. Psychiatry 166, Reportt of consensus meeting. Clinical relevance of response and improvement in psychopharmacology.. A statement from the European College of Neuropsychophamnacologyy from an ECNP workshop, Jerusalem, October (1995)) Eur Neuropsychopharmacol 5, Rothman,, K.J. and Michels, K.B. (1994) The continuing unethical use of placebo controls.. N. Eng. J. Med. 6,

11 69 9 Tablee 1 Features of the studies included. Studies s s s Diagnosis/C C ourse e DSMM III-R Acute e exacerbatio o n n DSMM III-R Chronic c DSMM UI-R Acute e DSMM III-R Acute e DSMM 111- R\IV V DSMM III- R\IV V DSMM III-R DSMM III-R' 77 weeks DSMM III Probably y chronic c DSMM III-R Acute e DSM-IH-R 3 3 DSMM III-R Chronic c Duration n Double-blind d period d 66 weeks 88 weeks 66 weeks 88 weeks 88 weeks 88 weeks 400 days 400 days 88 weeks 66 weeks 88 weeks Treatmentt groups Dosages s Haloperidoll flexible doses Haloperidoll 20 mg Haloperidoll 15 mg S mg Haloperidoll 8 mg Haloperidoll 16 mg Haloperidoll 16 mg Haloperidoll 15 mg 5 mg Haloperidoll flexible doses Haloperidoll 10 mg,» Baselinee BPRS/ derivedd BPRS' Respondenn definition' Patientss with an 20% totall score. Patientss with an 20% totall score. totall score and a CGIseverityy score S 3. totall score and a CG1- severityy score S 3. totall score. totall score. totall score and a CGIseverityy score 3. totall score and a CGIseverityy score 3. totall score. totall score. totall score. Patientss with an 20% totall score. 11 ITT efficacy population was defined as all patients who were randomised to receive double-blind treatment, took at least one dose, and providedd at least one efficacy evaluation on therapy. 'Studyy 9 was discontinued prematurely because an interim analyses did not show efficacy for the new compound. 'Alsoo including patients with schizoaffective disorders. 11 In studies 3 through 11 the 18 items of the BPRS was scored on a 0-6 point rating scale. In studies 1, 2 and 12 the 1-7 point rating scale wass used.

12 70 0 Tablee 2 Lowest and highest value of mean baseline score, mean percentage improvement from baselinee and percentage dropouts of the treatment-arms. Placebo-controliedd studies Non-placeboo controlled studies arms Haloperidol-arms s arms Halopendol-arms s Baselinee (points) Meann percentage improvement Percentagee dropouts (n=9) ) <-> %+++ 19% 40%% «- 54% (nn = 6) %% <-> 32% 49%% ** 59% (nn = 3) %% <- 39% 18%<-»39% % (nn = 3) 30..!! « %% ++ 30% 24%% «-> 47%

13 71 1 Tablee 3. Diagnosis/ / course e DSMM III-R, chronic c Treatmentt groups Dosages s Risperidonee 1 mg Risperidonee 4 mg Risperidonee 8 mg Risperidonee 12 mg Risperidonee 16 mg Haloperidoll 10 mg) n n Baseline e Improvementt from baseline PANSSS score IV*IV* improvement 90.1(1.18) ) -12.5(1.55)) /2I% 89.6(1.16) ) (1.56)*/31% 89.2(174) ) (1.55)*/30% 90.5(1.20) ) (1.39) /27% 89.8(1.20) ) (1.54)*/28% 88.8(1.10) ) -15.0(1.46)) /25% Responden:: (» 20% reduction off baseline total PANSS) 54% % 63% % 66% % 58% % 61% % 59% % ** significant better than risperidone 1 mg (Pairwisc intergroup comparison (p )

14 72 2 FIGUREE I MEAN BPRS SCORE AT BASELINE AND ENDPOINT AA PLACEBO-arms "11 Baseline Endpoint X X X X o o z z o o Z Z o o Z. Z en n too 25 PS S && 20 pa a ss,s SS io 0 0 J J j j 1 1 BB ATYPICAL ANTIPSYCHOTIC-arms I I 1 r~ ~

15 73 3 Halopcridol-arms s 600 f Placebo-arms s 40 E E OS S II 30 _J J ü ü U o- a. [ee e e zz z z Atypicall Anti psychotic-arms i U U

16 74 4 FIGUREE 3 DROPOUTS DUE TO AN INSUFFICIENT RESPONSE.. Atypical Antipsychotics j Placebo-armss i ii i nh-!! r e e g g CC 'E Ï.K.. Studiess I through 12. Filled bars refer to the arms of the non-placebo controlled studies

17 75 5