Abstract
AIMS
Indirect comparison (IC) and direct comparison (DC) between aripiprazole and olanzapine for schizophrenia were conducted to compare their efficacy and safety. The objective was to determine the usability of IC and consistency of results delivered by the two comparisons. Factors that might influence the inconsistency of results were also investigated.
METHODS
ICs and DCs were conducted using the change from baseline of the Positive and Negative Syndrome Scale (PANSS) total score as an efficacy endpoint and the dropout rate was selected as a safety endpoint. Placebo and risperidone were used as common comparators for ICs.
RESULTS
A literature search identified 20 articles. The efficacy analysis gave results on the mean difference in PANSS change (95% CI) of −5.72 (−10.22, −1.22) in ICs using placebo as a common comparator and −7.41 (−15.96, 1.14) in DCs. When using risperidone as a common comparator, it was −9.15 (−20.12, 1.82). In rate ratio analysis of the all cause dropout rate, the IC result was 1.17 (0.83, 1.65) using placebo as a common comparator and 1.56 (0.57, 4.26) using risperidone as a common comparator. Both analyses gave consistent results between ICs and DCs. A slightly lower estimated value was observed in ICs using placebo.
CONCLUSIONS
This study demonstrated that ICs between olanzapine and aripiprazole can deliver results consistent with those of DCs. It is also suggested that the selection of a common comparator is important when control group bias is suspected in the data set.
Keywords: indirect comparison, antipsychotic, schizophrenia, meta-analysis
WHAT IS ALREADY KNOWN ABOUT THIS SUBJECT
Conducting indirect comparisons (ICs) is one method to resolve the lack of direct comparison (DC) data between targeted pharmaceuticals.
Some researchers conducting ICs in various therapeutic areas reported that IC and DC yielded similar results. However, the majority did not discuss in detail the reasons why some ICs gave results different from DCs.
WHAT THIS STUDY ADDS
This study demonstrated that ICs between olanzapine and aripiprazole can deliver results consistent with those of DCs when using the change from baseline in the PANSS total score and all cause dropout rate in patients with schizophrenia
It should be emphasized that selection of the common comparator is important when control group bias is suspected in the data set.
Introduction
In the field of pharmaceutical development and medical practice, it is well known that published sources can frequently provide useful data on alternative interventions. It is also recognized that a meta-analysis using published data is a useful approach to achieve this purpose. However, sometimes there are too few publications describing the results of direct comparisons (DCs) between targeted pharmaceuticals. Data on drugs under clinical development or newly approved are usually limited and DC data are rarely available.
One method to resolve the lack of DC data is conducting indirect comparisons (ICs) to assess the efficacy and/or safety of drugs, since numerous IC results are available in various therapeutic areas. Donegan et al. [1] found that 43 articles published from 1992 to 2007 compared interventions indirectly. On the other hand, Song et al. [2] assessed the appropriateness of statistical IC methods used in 88 publications from 2000 to 2007. The majority of reports assert that IC is useful for assessing data when DC data are insufficient and supplementary data are required.
Sauriol et al. [3] reported the results of ICs between new generation antipsychotics. Only one DC report on olanzapine (OLZ) and risperidone (RIS) was available at the time of comparison because OLZ had recently been launched. They therefore conducted an IC of efficacy and safety between OLZ and RIS incorporating study data using haloperidol (HAL) as a common comparator [3]. Rabinowitz et al. conducted DC of dropout rates for first and second generation antipsychotic drugs and investigated how a broad range of design features effect dropout rate [4].
In this paper, IC and DC of the efficacy and safety of aripiprazole (ARP), a currently new antipsychotic, and OLZ, considered the gold standard antipsychotic agent after long term clinical experience, were conducted. The results were then used to determine the utility of ICs and consistency between IC and DC results. Some researchers conducting ICs in various therapeutic areas reported that IC and DC yielded similar results. However, the majority did not discuss in detail the reasons why some ICs gave results different from DCs. Therefore, we also investigated factors that may contribute to inconsistencies between the results of ICs and DCs.
Methods
Study selection
A literature search was conducted in PubMed and Embase, using the key words ‘aripiprazole,’ ‘olanzapine,’ ‘risperidone,’ ‘placebo’ and ‘schizophrenia.’ The search was limited to ‘randomized controlled trials’ and conducted in July 2012. All literature published from January 1990 to June 2012 was searched. After screening the search results, reports using similar doses, similar diagnostic criteria (i.e. diagnosed by either the DSM or ICD), and similar dosing periods ranging from 4 to 26 weeks were selected. The quality of the reports was evaluated based on the Jadad and Moore [5] score and those with scores of ≥3 were selected for this analysis. One of us (TK) initially selected the literature and extracted all of the data. The literature was independently searched by another author (MH), who also independently confirmed each value.
Outcome parameters
The primary efficacy endpoint for this analysis was the change from baseline in the Positive and Negative Syndrome Scale (PANSS) [6]. Differences between ARP groups and OLZ groups were evaluated to compare the results of ICs and DCs. As common comparators for ICs, placebo (PLB) or RIS were selected. We also investigated whether the results of ICs were affected by the selection of a common comparator.
The dropout rate was selected as the safety endpoint. ICs between ARP and OLZ were also conducted from the safety aspect using this endpoint and the result was compared with that of DCs. As a common comparator, RIS or PLB were used to investigate differences in IC results which might be caused by the common comparator. The dropout rate was defined in three ways: all cause dropout rate, safety reason dropout rate and efficacy reason dropout rate.
Stastistical methods
We first carried out meta-analyses using head-to-head DC data reported in the literature between two of ARP, OLZ, RIS or PLB using Review Manager software version 5. Mean difference analysis was conducted to assess the change from baseline in PANSS and rate ratio analysis with the random effect model was used to assess the dropout rate. In conducting ICs for each analysis, we followed Bucher et al.'s method [7] using DC meta-analysis data obtained using Review Manager which included ARP/OLZ vs. RIS or ARP/OLZ vs. PLB.
 |
 |
where D1 − D2 is the mean difference in the change from baseline in the PANSS obtained by DC between drug 1 or drug 2 and the common comparator, SE1, SE2 are the standard error of the mean difference in the change from baseline in the PANSS obtained by DC between drug 1 or drug 2 and the common comparator, DIC is the mean difference in the change from baseline in the PANSS between drug 1 and drug 2 obtained by IC and SEIC is the standard error of the mean difference in the change from baseline in the PANSS between drug 1 and drug 2 obtained by IC.
The rate ratio of the dropout rate between two interventions was also obtained following Bucher et al.'s method [7] using DC meta-analysis data obtained using Review Manager which included ARP/OLZ vs. RIS or ARP/OLZ vs. PLB:
 |
 |
where R1/R2 is the rate ratio of the dropout rate obtained by DC between drug 1 or drug 2 and the common comparator, SE1, SE2 are the standard error of the rate ratio of the dropout rate obtained by DC between drug 1 or drug 2 and the common comparator, RIC is the rate ratio of the dropout rate between drug 1 and drug 2 obtained by IC and SEIC is the standard error of the rate ratio of the dropout rate between drug 1 and drug 2 obtained by IC.
Results
Eligible studies and characteristics
The literature search identified 20 studies that fulfilled the selection criteria (population I). Eleven studies included a PLB arm (five studies in comparison with ARP and six in comparison with OLZ), and seven studies included a RIS arm (two studies in comparison with ARP and five in comparison with OLZ). Three DC studies between ARP and OLZ were also included in population I (Table 1).
Table 1.
Summary of studies used in the current analysis
| Reference # | Population II* | Diagnosis | Age (years) | Intervention (daily dose, mg) | n | Dosing duration (weeks) | PANSS total score at baseline | PANSS change from baseline | All cause dropout | Efficacy reason dropout | Safety reason dropout |
|---|---|---|---|---|---|---|---|---|---|---|---|
| ARP vs. PLB | |||||||||||
| [11] | Yes | Schizophrenia or schizoaffective disorder (DSM-IV) | 18 to 65 | ARP (15) | 102 | 4 | 98.5 | −15.5 | 34 | 5 | 9 |
| ARP (30) | 102 | 99.0 | −11.4 | 42 | 15 | 8 | |||||
| Hospitalized for acute relapse | PLB | 106 | 100.2 | −2.9 | 48 | 15 | 17 | ||||
| [12] | Yes | Schizophrenia or schizoaffective disorder (DSM-IV) | 18 to 65 | ARP (20) | 101 | 4 | 94.4 | −14.5 | 40 | 9 | 11 |
| ARP (30) | 101 | 92.6 | −13.9 | 34 | 8 | 8 | |||||
| Hospitalized for acute relapse | RIS (6) | 99 | 94.9 | −15.7 | 37 | 8 | 8 | ||||
| PLB | 103 | 95.7 | −5.0 | 51 | 17 | 17 | |||||
| [13] | Yes | Schizophrenia (DSM-IV) | ≥18 | ARP (2) | 93 | 6 | 90.8 | −8.2 | 42 | 22 | 2 |
| Acute relapse, documented worsening of schizophrenia within the previous 3 months and required inpatient hospitalization | ARP (5) | 92 | 92.2 | −10.6 | 45 | 13 | 1 | ||||
| ARP (10) | 94 | 90.9 | −11.3 | 41 | 14 | 4 | |||||
| PLB | 88 | 90.9 | −5.3 | 44 | 20 | 6 | |||||
| [14] | No | Chronic schizophrenia (DSM-IV) | ≥18 | ARP (15) | 155 | 26 | 81.22 | −2.04 | 84 | 42 | 16 |
| PLB | 155 | 83.12 | 1.78(week 6) |
110 | 76 | 13 | |||||
| [15] | No | Schizophrenia (DSM-IV) | 13 to 17 | ARP (10) | 99 | 6 | 93.7 | −28.6 | 16 | NA | NA |
| ARP (30) | 97 | 94.9 | −26.7 | 18 | |||||||
| PLB | 98 | 95.0 | −21.2 | 10 | |||||||
| ARP vs. RIS | |||||||||||
| [16] | Yes | Schizophrenia or schizoaffective disorder (DSM-IV), and hospitalized due to acute relapse | 18 to 65 | ARP (15) | 49 | 4 | 85.1 | −19.6 | 11 | 5 | 2 |
| RIS (6) | 34 | 84.6 | −21.1 | 10 | 2 | 0 | |||||
| [17] | Yes | Schizophrenia, acute schizophrenia-like psychotic disorder, or schizoaffective disorder (ICD−10) | ARP (12–30) | 21 | 8 | 88.5 | −18.4 | 11 | 10 | 1 | |
| RIS (3–12) | 20 | 93.6 | −24.7 | 5 | 4 | 0 | |||||
| OLZ vs. PLB | |||||||||||
| [18] | Yes | Schizophrenia (DSM-IV) | 18 to 65 | OLZ (15) | 93 | 6 | NA | −31.5 | 24 | 4 | 7 |
| PLB | 87 | −12.6 | 22 | 15 | 3 | ||||||
| [19] | Yes | Schizophrenia (DSM-IV) | ≥18 | OLZ (10) | 105 | 6 | 94.9 | −18.4 | 60 | 24 | 8 |
| PLB | 105 | 93.6 | −8 | 73 | 39 | 5 | |||||
| [20] | Yes | Schizophrenia (DSM-IV) | ≥18 | OLZ (10) | 126 | 6 | 93.3 | −18.1 | 40 | 16 | 5 |
| PLB | 120 | 93.9 | −2.8 | 76 | 54 | 5 | |||||
| [21] | Yes | Primary diagnosis of schizophrenia (DSM-IV) | 18 to 75 | OLZ (15) | 122 | 6 | 96.3 | −28.7 | 39 | 9 | 8 |
| PLB | 114 | 95.8 | −16 | 45 | 18 | 10 | |||||
| [22] | No | Prodromal syndrome based on COPS | 12 to 45 | OLZ (5–15) | 30 | 8 | 66.83 | −6.73 | 11 | 4 | 1 |
| PLB | 29 | 63.26 | −1.66 | 8 | 4 | 1 | |||||
| [23] | No | Schizophrenia (DSM-IV-TR) | 13 to 17 | OLZ (2.5–20) | 72 | 6 | 95.3 | −21.3 | 23 | 10 | 5 |
| PLB | 35 | 95.5 | −8.8 | 20 | 18 | 0 | |||||
| OLZ vs. RIS | |||||||||||
| [24] | Yes | Schizophrenia, schizophreniform disorder, or schizoaffective disorder (DSM-IV) | 18 to 65 | OLZ (10–20) | 166 | 28 | 96.3 | −28.1 | 70 | 23 | 16 |
| RIS | 165 | 95.7 | −24.9 | 87 | 27 | 26 | |||||
| [25] | Yes | Schizophrenia or schizophreniform disorder, at least of moderate severity (DSM-IV) | ≥18 | OLZ (10–20) | 32 | 30 | 94.7 | −28.2 | 15 | 6 | NA |
| RIS (4–8) | 33 | 88.9 | −16.3 | 21 | 11 | ||||||
| [26] | No | Chronic schizophrenia or schizoaffective disorder (DSM-IV) and suboptimal response to previous treatment | 18 to 60 | OLZ (10–40) | 39 | 14 | 91.0 | −4 | NA | 4 | 2 |
| RIS (4–16) | 41 | 89.5 | −2 | 2 | 0 | ||||||
| [27] | No | First or second psychotic episode of schizophrenia, schizophreniform, or schizoaffective disorder (DSM-IV) | 16 to 28 | OLZ (5–30) | 18 | 6 | 90.7 | −15.1 | NA | NA | NA |
| RIS (1–8) | 24 | 90.4 | −15 | ||||||||
| [28] | No | Schizophrenia, schizoaffective, and schizophreniform disorder (DSM-IV) | 16 to 40 | OLZ (2.5–20) | 133 | 12** | 74.3 | −14.3 | 91 | 15 | 14 |
| First episode of psychotic illness and continuously ill for at least 1 month and no more than 5 years | RIS (0.5–4) | 133 | 73.0 | −13.7 | 95 | 12 | 13 | ||||
| ARP vs. OLZ | |||||||||||
| [29] | Yes | Schizophrenia (DSM-IV) in acute relapse and demonstrated previous response to antipsychotic drugs (other than clozapine) | 18 to 65 | ARP (15–30) | 347 | 6** | 70.7 | −24.6 | 104 | 30 | 37 |
| OLZ (10–20) | 344 | 77.9 | −29.5 | 77 | 25 | 18 | |||||
| [17] | Yes | Schizophrenia, acute schizophrenia-like psychotic disorder, or schizoaffective disorder (ICD−10) | ARP (12–30) | 21 | 8 | 88.5 | −18.4 | 11 | 10 | 1 | |
| OLZ (10–20) | 17 | 104.3 | −33.4 | 2 | 2 | 0 | |||||
| [30] | No | Schizophrenia (DSM-IV-TR) | 18 to 65 | ARP (10–30) | 285 | 8** | 95.0 | −22.2 | 143 | 31 | 27 |
| OLZ (10–20) | 281 | 95.7 | −26.8 | 120 | 10 | 26 | |||||
Studies in patients with acute exacerbation of schizophrenia are population II.
Interim assessment time point data in long term dosing were used.
Eight reports were excluded from the separately defined population II because those studies involved chronic, adolescent, or prodromal patients (Table 1). As a result, population II included 12 studies that evaluated the efficacy and safety of schizophrenic patients with acute exacerbation. Population II was expected to provide lower heterogeneity because chronic, adolescent and prodromal patients are known to have a different response rate to antipsychotics from that of patients with acute phase schizophrenia. When more than one dose was used in a report, the highest dose arm was selected as long as it was within the US Food and Drug Administration (FDA) approved level.
Direct and indirect analysis of changes in the PANSS total score
Results in population I
First, DCs between ARP and PLB and then between OLZ and PLB were conducted using the population I data set. Subsequently, those DC data were applied for IC between ARP and OLZ using PLB as a common comparator. Table 2 indicates favourable results for OLZ in this analysis, with a mean difference (95% CI) of the change from baseline in the PANSS total score of −6.04 (−9.74, −2.34). When using RIS as a common comparator, the point estimate value was similar to that of PLB, with a mean difference of −5.90 (−13.79, 1.99) between ARP and OLZ. The results of DCs using data from three studies included in population I showed a similar point estimate value with a mean difference of −5.06 (−7.85, −2.28).
Table 2.
Results of DC and IC
| Parameter | Population | DC | IC Common comparator PLB | IC Common comparator RIS |
|---|---|---|---|---|
| Mean difference (95% CI) | ||||
| Change from baseline in PANSS total score | I | −5.06 (−7.85, −2.28) | −6.04 (−9.74, −2.34) | −5.90 (−13.79, 1.99) |
| II | −7.41 (−15.96, 1.14) | −5.72 (−10.22, −1.22) | −9.15 (−20.12, 1.82) | |
| Rate ratio (95% CI) | ||||
| All cause dropout | I | 1.28 (1.01, 1.62) | 1.16 (0.86, 1.57) | 1.40 (0.52, 3.79) |
| II | 1.87 (0.68, 5.18) | 1.17 (0.83, 1.65) | 1.56 (0.57, 4.26) | |
| Safety reason dropout | I | 1.45 (0.84, 2.51) | 0.65 (0.33, 1.29) | 2.26 (0.48, 10.54) |
| II | 2.03 (1.19, 3.46) | 0.52 (0.24, 1.12) | 3.15 (0.68, 14.49) | |
| Efficacy reason dropout | I | 2.16 (0.97, 4.80) | 1.72 (1.06, 2.77) | 2.66 (0.97, 7.26) |
| II | 1.84 (0.57, 5.90) | 2.13 (1.15, 3.94) | 3.17 (1.12, 8.93) | |
Results in population II
IC analysis of the mean difference between ARP and OLZ in population II was conducted similar to that in population I (Table 2). The mean difference between ARP and OLZ was −5.72 (−10.22, −1.22) using PLB as a common comparator and −9.15 (−20.12, 1.82) using RIS as a common comparator. DC between those two interventions in population II provided a mean difference result of −7.41 (−15.96, 1.14), which is in between the estimated value of two ICs obtained in these conditions.
Direct and indirect analysis of the all cause dropout rate
Results in population I
As in the efficacy analysis, DCs between ARP and PLB followed by those between OLZ and PLB were conducted using the population I data set. The DC data were then used for ICs between ARP and OLZ using PLB as a common comparator. The results of the ICs are shown in Table 2. The rate ratio (95% CI) of 1.16 (0.86, 1.57) of the all cause dropout rate was in favour of OLZ. When using RIS as a common comparator, the result was similar to that when PLB was the common comparator, with a value of 1.40 (0.52, 3.79). DC analysis of this population gave a rate ratio value of 1.28 (1.01, 1.62), which is in between the two IC values described above.
Results in population II
IC analysis of the dropout rate in population II was also conducted (Table 2). The result of analysis with PLB as the common comparator was 1.17 (0.83, 1.65) and that with RIS as the comparator was 1.56 (0.57, 4.26). The two analyses thus gave similar point estimate results. DCs of population II yielded results similar to the ICs, with a value of 1.87 (0.68, 5.18).
Direct and indirect analyses of the safety reason dropout rate
Results in population I
Analysis of the safety reason dropout rate was conducted similar to that of the all cause dropout rate. The result using PLB as the common comparator was 0.65 (0.33, 1.29). On the other hand, opposite results were obtained with RIS as the common comparator, which favoured OLZ with a value of 2.26 (0.48, 10.54) (Table 2). DC gave a result of 1.45 (0.84, 2.51). This point estimate value was in between the two IC values, and the three analyses did not yield the same conclusion in this dropout rate analysis.
Results in population II
When the dropout rate analyses were carried out in population II, the IC results when using PLB and RIS as common comparators also differed, as in population I, with a value of 0.52 (0.24, 1.12) and 3.15 (0.68, 14.49), respectively (Table 2). The DC result was 2.03 (1.19, 3.46).
Direct and indirect analysis of efficacy reason dropout rate
Results in population I
In population I, the efficacy reason dropout rate of ARP and OLZ assessed using PLB as the common comparator was 1.72 (1.06, 2.77) and that using RIS was 2.66 (0.97, 7.26) (Table 2). DC analysis yielded an intermediate point estimate value of 2.16 (0.97, 4.80).
Results in population II
The IC result of the efficacy reason dropout rate using the population II data set (Table 2) with PLB as the common comparator was 2.13 (1.15, 3.94) and that using RIS was 3.17 (1.12, 8.93). DC gave a smaller point estimate value of 1.84 (0.57, 5.90). The results therefore varied among the three analyses.
Summary of studies analyzed
Mean change in the PANSS total score
The mean change in the total PANSS score in population II is shown in Figure 1. The result in the ARP group was −22.9 and that in the OLZ group −28.2 when the control group received RIS and two studies were used for each calculation. In placebo-controlled studies, the result was −12.2 and −24.2 in the ARP and OLZ group when three and four studies were used for calculation, respectively. RIS controlled studies thus yielded a greater change in the PANSS total score compared with placebo controlled ones.
Figure 1.
Mean PANSS total score change in each data set. Open bars, OLZ; dotted bars, RIS, filled bars, ARP; hatched bars, PLB
Mean all cause dropout rate
Figure 2 shows the mean all cause dropout rate in population II. The rate in the ARP group was 37.4% and that in the OLZ group 34.5% in RIS controlled studies when the same studies as for the investigation of the PANSS total score change were used for calculation. When using the same PLB controlled studies as for the PANSS total score change, the all cause dropout rate in the ARP group was 39.5% and that in the OLZ group was 35.8%. Therefore no major difference was found in the dropout rate between the two groups regardless of the comparator.
Figure 2.
Mean all cause dropout rate in each data set. Open bars, OLZ; dotted bars, RIS, filled bars, ARP; hatched bars, PLB
Discussion
ARP is an atypical antipsychotic agent that was approved by the FDA in 2002 for the treatment of schizophrenia. OLZ is also an atypical antipsychotic agent approved by the FDA in 1996. It has been recognized as the gold standard treatment for schizophrenia after 15 years of clinical experience accumulated after its launch.
We compared the efficacy and safety of ARP and OLZ using both DC and IC methods, as well as attempting to determine the factors that may influence the results of DC and IC. As summarized in Figures 3 and 4, our results showed that ICs yield results similar to DCs in both efficacy analysis using a continuous parameter (the change from baseline in the PANSS total score) and safety analysis using binary data (the all cause dropout rate). On the other hand, the results of the efficacy and safety reason dropout rate gave inconsistent results between DCs and ICs. When a common efficacy or safety endpoint is not available, either DC or IC is often used to evaluate them based on the efficacy or safety reason dropout rate. This study shows that those items do not always provide consistent results in DC and IC. One reason for this may be varying definitions of reasons for dropout among studies. Some studies included seven categories and others four categories, meaning that some reasons were merged. In addition, the definitions of ‘lack of efficacy’ and ‘safety concerns’ differ among studies. Commonly defined endpoints should be used to obtain consistent results between DCs and ICs, since this study showed that the change from baseline in the PANSS total score and the all cause dropout rate gave consistent results between DCs and ICs.
Figure 3.
Mean difference in the change from baseline in PANSS total score. □: Population I; ▪: population II; DC: direct comparison; IC (PLB): indirect comparison with PLB as a common comparator; IC (RIS): indirect comparison with RIS as a common comparator
Figure 4.
Rate ratio of dropout rate. □: Population I; ▪: population II; DC: direct comparison; IC (PLB): indirect comparison with PLB as a common comparator; IC (RIS): indirect comparison with RIS as a common comparator
Population II in this study came from the literature in which patients had similar background data such as age, clinical status and therapeutic dosage, and study designs were similar. It was expected that population II would yield more consistent results between DCs and ICs in both efficacy analysis and safety analysis. However, when we used PLB as a common comparator in the analyses of the change from baseline in the PANSS total score and the all cause dropout rate, the results of ICs differed slightly from those of IC using RIS as a common comparator and from those of DCs. ICs using PLB tended to give smaller point estimate values.
We investigated potential factors that could influence the results of PLB controlled antipsychotic trials and would in turn affect IC results. In many clinical trials of antipsychotic agents conducted in the 1990s, it was difficult to use a PLB due to ethical issues. After the 1990s, PLB use became common in clinical trials, especially in some regions including the USA and EU, and it has become a standard requirement of many regulatory authorities. Kemp et al. [8] reported that the PLB effect on efficacy evaluations had gradually increased in antipsychotic clinical trials. It was reported that the change from baseline in the PANSS total score in PLB arms observed in clinical trials conducted between 1993 and 1996 was approximately in a range of −4 to 4. Subsequently, it increased to −10 in a trial conducted in 2004 and has tended to increase annually [8]. PLB effects also vary among trials even when the same intervention is used and they are conducted by the same sponsor, with the same design and in the same year. In this report, the majority of PLB arm data included in population II showed a limited reduction in the change in the PANSS total score, although the literature from which the data were taken was published between the mid-1990s and mid-2000s, and some of the reports were included in Kemp et al.'s investigation [8]. Investigators should take the absolute values of efficacy parameter changes in common comparators into account when conducting ICs because they may cause inconsistent results among data sets used for the analysis.
Secondly, we investigated control group bias in the studies used for analyses. Woods et al. [9] reported that the degree of efficacy improvement in antipsychotic medication trials varied between active controlled and PLB controlled studies. The degree of improvement in the Brief Psychiatric Rating Scale score in antipsychotic medication trials reported from 1997 to 2000 was nearly double in active controlled compared with PLB controlled trials. The mean PANSS total score change from baseline in our data set was −28.2 and −22.9 in the OLZ group and ARP group, respectively, in RIS controlled studies in population II, while the mean in PLB controlled studies was −24.2 and −12.2, respectively. Two active controlled DC studies of OLZ and ARP found a mean change of −31.5 in the OLZ and −21.5 in the ARP group. Based on this result, we suspect that a control group bias exists in our data set. It is suggested that ICs using PLB give slightly different results from those using RIS as a common comparator because of control group bias. It should be noted that the DC results we used were obtained from meta-analysis of two active-controlled studies and therefore should have yielded results comparable to ICs with the active comparator.
Kemmler et al. [10] reported a higher dropout rate in PLB controlled than in active controlled trials of antipsychotic agents. Since we used the dropout rate as a parameter for IC analysis, the impact of this higher dropout rate should be investigated. Our data set did not show a marked difference in the dropout rate between PLB controlled and active controlled trials, and it is not clear why IC using PLB showed a smaller point estimate result in the analysis of the all cause dropout rate.
In pharmaceutical development, a PLB arm is frequently used to investigate the efficacy and safety of drugs, especially in dose finding studies. However, it is less common in confirmatory studies, which usually include an active comparator arm. This research shows the potential of ICs to compare the efficacy and safety of experimental drugs with those of approved drugs using a PLB or active agent as a common comparator, meaning that if dose-finding data with a PLB or data from a trial with an active comparator are available, they can be used to investigate the efficacy and safety of a new experimental agent. This will allow investigators to compare new agents with the current gold standard treatment without conducting head-to-head trials. As a result, these methods can potentially be used to shorten the total development period by eliminating the need for confirmatory studies with approved drugs.
This report also highlights the importance of common comparator selection. If bias, such as an increase in the PLB effect over time or control group bias, is known, researchers should determine the effects of that bias.
In conclusion, this study demonstrated that IC between OLZ and ARP can deliver results consistent with those of DC when using the change from baseline in the PANSS total score and all cause dropout rate in schizophrenia patients. It should be also emphasized that selection of the common comparator is important when control group bias is suspected in the data set.
Competing Interests
All authors have completed the Unified Competing Interest form at http://www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare no support from any organization for the submitted work, no financial relationships with any organizations that might have an interest in the submitted work in the previous 3 years and no other relationships or activities that could appear to have influenced the submitted work.
Supporting Information
Additional Supporting Information may be found in the online version of this article at the publisher's web-site:
Flow diagram of assessment of studies identified in the systematic review
Resultant search data. Tabulated study data for analyses
References
- 1.Donegan S, Williamson P, Gamble C, Tudur-Smith C. Indirect comparisons: a review of reporting and methodological quality. PLoS ONE. 2010;5:e11054. doi: 10.1371/journal.pone.0011054. doi: 10.1371/journal.pone.0011054. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Song F, Loke Y, Walsh T, Eastwood A, Altman DG. Methodological problems in the use of indirect comparisons for evaluating healthcare interventions: survey of published systematic reviews. BMJ. 2009;338:b1147. doi: 10.1136/bmj.b1147. doi: 10.1136/bmj.b1147. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Sauriol L, Laporta M, Edwardes M, Deslandes M, Ricard N, Suissa S. Meta-analysis comparing newer antipsychotic drugs for the treatment of schizophrenia: evaluating the indirect approach. Clin Ther. 2001;23:942–956. doi: 10.1016/s0149-2918(01)80082-5. [DOI] [PubMed] [Google Scholar]
- 4.Rabinowitz J, Levine SZ, Barkai O, Davidov O. Dropout rates in randomized clinical trials of antipsychotics: a meta-analysis comparing first-and second-generation drugs and an examination of the role of trial design features. Schizophr Bull. 2009;35:775–788. doi: 10.1093/schbul/sbn005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Jadad AR, Moore RA, Carroll D, Jenkinson C, Reynolds DJ, Gavaghan DJ, McQuay HJ. Assessing the quality of reports of randomized clinical trials: is blinding necessary? Control Clin Trials. 1996;17:1–12. doi: 10.1016/0197-2456(95)00134-4. [DOI] [PubMed] [Google Scholar]
- 6.Kay SR, Fisbein A, Opler LA. The Positive and Negative Syndrome Scale (PANSS) for schizophrenia. Schizophr Bull. 1987;13:261–276. doi: 10.1093/schbul/13.2.261. [DOI] [PubMed] [Google Scholar]
- 7.Bucher HC, Guyatt GH, Griffith LE, Walter SD. The results of direct and indirect treatment comparisons in meta-analysis of randomized controlled trials. J Clin Epidemiol. 1997;50:683–691. doi: 10.1016/s0895-4356(97)00049-8. [DOI] [PubMed] [Google Scholar]
- 8.Kemp AS, Schooler NR, Kalali AH, Alphs L, Anand R, Awad G, Davidson M, Dubé S, Ereshefsky L, Gharabawi G, Leon AC, Lepine JP, Potkin SG, Vermeulen A. What is causing the reduced drug-placebo difference in recent schizophrenia clinical trials and what can be done about it? Schizophr Bull. 2010;36:504–509. doi: 10.1093/schbul/sbn110. doi: 10.1093/schbul/sbn110. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Woods SW, Gueorguieva RV, Baker CB, Makuch RW. Control group bias in randomized atypical antipsychotic medication trials for schizophrenia. Arch Gen Psychiatry. 2005;62:961–970. doi: 10.1001/archpsyc.62.9.961. [DOI] [PubMed] [Google Scholar]
- 10.Kemmler G, Hummer M, Widschwendter C, Fleischhacker WW. Dropout rates in placebo-controlled and active-control clinical trials of antipsychotic drugs. Arch Gen Psychiatry. 2005;62:1305–1312. doi: 10.1001/archpsyc.62.12.1305. [DOI] [PubMed] [Google Scholar]
- 11.Kane JM, Carson WH, Saha AR, McQuade RD, Ingenito GG, Zimbroff DL, Ali MW. Efficacy and safety of aripiprazole and haloperidol versus placebo in patients with schizophrenia and schizoaffective disorder. J Clin Psychiatry. 2002;63:763–771. doi: 10.4088/jcp.v63n0903. [DOI] [PubMed] [Google Scholar]
- 12.Potkin SG, Saha AR, Kujawa MJ, Carson WH, Ali M, Stock E, Stringfellow JS, Ingenito G, Marder SR. Aripiprazole, an antipsychotic with a novel mechanism of action, and risperidone vs placebo in patients with schizophrenia and schizoaffective disorder. Arch Gen Psychiatry. 2003;60:681–690. doi: 10.1001/archpsyc.60.7.681. [DOI] [PubMed] [Google Scholar]
- 13.Cutler AJ, Marcus RN, Hardy SA, O'Donnell A, Carson WH, McQuade RD. The efficacy and safety of lower doses of aripiprazole for the treatment of patients with acute exacerbation of schizophrenia. CNS Spectr. 2006;11:691–702. doi: 10.1017/s1092852900014784. [DOI] [PubMed] [Google Scholar]
- 14.Pigott TA, Carson WH, Saha AR, Torbeyns AF, Stock EG, Ingenito GG. Aripiprazole for the prevention of relapse in stabilized patients with chronic schizophrenia: a placebo-controlled 26-week study. J Clin Psychiaty. 2003;64:1048–1056. doi: 10.4088/jcp.v64n0910. [DOI] [PubMed] [Google Scholar]
- 15.Findling RL, Robb A, Nyilas M, Forbes RA, Jin N, Ivanova S, Marcus R, McQuade RD, Iwamoto T, Carson WH. A multiple-center, randomized, double-blind, placebo-controlled study of oral aripiprazole for treatment of adolescents with schizophrenia. Am J Psychiatry. 2008;165:1432–1441. doi: 10.1176/appi.ajp.2008.07061035. [DOI] [PubMed] [Google Scholar]
- 16.Chan HY, Lin WW, Lin SK, Hwang TJ, Su TP, Chiang SC, Hwu HG. Efficacy and safety of aripiprazole in the acute treatment of schizophrenia in Chinese patients with risperidone as an active control: a randomized trial. J Clin Psychiatry. 2007;68:29–36. doi: 10.4088/jcp.v68n0104. [DOI] [PubMed] [Google Scholar]
- 17.Hatta K, Sato K, Hamakawa H, Takebayashi H, Kimura N, Ochi S, Sudo Y, Asukai N, Nakamura H, Usui C, Kawabata T, Hirata T, Sawa Y. Effectiveness of second-generation antipsychotics with acute-phase schizophrenia. Schizophr Res. 2009;113:49–55. doi: 10.1016/j.schres.2009.05.030. [DOI] [PubMed] [Google Scholar]
- 18.Corrigan MH, Gallen CC, Bonura ML, Merchant KM. Effectiveness of the selective D4 antagonist sonepiprazole in schizophrenia: a placebo-controlled trial. Biol Psychiatry. 2004;55:445–451. doi: 10.1016/j.biopsych.2003.10.004. [DOI] [PubMed] [Google Scholar]
- 19.Marder SR, Kramer M, Ford L, Eerdekens E, Lim P, Eerdekens M, Lowy A. Efficacy and safety of paliperidone extended-release tablets: results of a 6-week, randomized, placebo-controlled study. Biol Psychiatry. 2007;62:1363–1370. doi: 10.1016/j.biopsych.2007.01.017. [DOI] [PubMed] [Google Scholar]
- 20.Davidson M, Emsley R, Kramer M, Ford L, Pan G, Lim P, Eerdekens M. Efficacy, safety and early response of paliperidone extended-release tablets (paliperidone ER): results of a 6-week, randomized, placebo-controlled study. Schizophrenia Res. 2007;93:117–130. doi: 10.1016/j.schres.2007.03.003. [DOI] [PubMed] [Google Scholar]
- 21.Meltzer HY, Cucchiaro J, Silva R, Ogasa M, Phillips D, Xu J, Kalali AH, Schweizer E, Pikalov A, Loebel A. Lurasidone in the treatment of schizophrenia: a randomized, double-blind, placebo-and olanzapine-controlled study. Am J Psychiatry. 2011;168:957–967. doi: 10.1176/appi.ajp.2011.10060907. [DOI] [PubMed] [Google Scholar]
- 22.Woods SW, Breier A, Zipursky RB, Perkins DO, Addington J, Miller TJ, Hawkins KA, Marquez E, Lindborg SR, Tohen M, McGlashan TH. Randomized trial of olanzapine versus placebo in the symptomatic acute treatment of the schizophrenic prodrome. Biol Psychiatry. 2003;54:453–464. doi: 10.1016/s0006-3223(03)00321-4. [DOI] [PubMed] [Google Scholar]
- 23.Kryzhanovskaya L, Schulz C, McDougle C, Frazier J, Dittmann R, Robertson-Plouch C, Bauer T, Xu W, Wang W, Carlson J, Tohen M. Olanzapine versus placebo in adolescents with schizophrenia: a 6-week, randomized, double-blind, placebo-controlled trial. J Am Acad Child Adolesc Psychiatry. 2009;48:60–70. doi: 10.1097/CHI.0b013e3181900404. [DOI] [PubMed] [Google Scholar]
- 24.Tran PV, Hamilton SH, Kuntz AJ, Potvin JH, Andersen SW, Beasley C, Tollefson GD. Double-blind comparison of olanzapine versus risperidone in the treatment of schizophrenia and other psychotic disorders. J Clin Psychopharmacol. 1997;17:407–418. doi: 10.1097/00004714-199710000-00010. [DOI] [PubMed] [Google Scholar]
- 25.Gureje O, Miles W, Keks N, Grainger D, Lambert T, McGrath J, Tran P, Catts S, Fraser A, Hustig H, Andersen S, Crawford AM. Olanzapine vs risperidone in the management of schizophrenia: a randomized double-blind trial in Australia and New Zealand. Schizophr Res. 2003;61:303–314. doi: 10.1016/s0920-9964(02)00226-8. [DOI] [PubMed] [Google Scholar]
- 26.Volavka J, Czobor P, Sheitman B, Lindenmayer JP, Citrome L, McEvoy JP, Cooper TB, Chakos M, Lieberman JA. Clozapine, olanzapine, risperidone, and haloperidol in the treatment of patients with chronic schizophrenia and schizoaffective disorder. Am J Psychiatry. 2002;159:255–262. doi: 10.1176/appi.ajp.159.2.255. [DOI] [PubMed] [Google Scholar]
- 27.van Bruggen J, Tijssen J, Dingemans P, Gersons B, Linszen D. Symptom response and side-effects of olanzapine and risperidone in young adults with recent onset schizophrenia. Int Clin Psychopharmacol. 2003;18:341–346. doi: 10.1097/00004850-200311000-00005. [DOI] [PubMed] [Google Scholar]
- 28.McEvoy JP, Lieberman JA, Perkins DO, Hamer RM, Gu H, Lazarus A, Sweitzer D, Olexy C, Weiden P, Strakowski SD. Efficacy and tolerability of olanzapine, quetiapine, and risperidone in the treatment of early psychosis: a randomized, double-blind 52-week comparison. Am J Psychiatry. 2007;164:1050–1060. doi: 10.1176/ajp.2007.164.7.1050. [DOI] [PubMed] [Google Scholar]
- 29.Fleischhacker WW, McQuade RD, Marcus RN, Archibald D, Swanink R, Carson WH. A double-blind, randomized comparative study of aripiprazole and olanzapine in patients with schizophrenia. Biol Psychiatry. 2009;65:510–517. doi: 10.1016/j.biopsych.2008.07.033. [DOI] [PubMed] [Google Scholar]
- 30.Kane JM, Osuntokun O, Kryzhanovskaya LA, Xu W, Stauffer VL, Watson SB, Breier A. A 28-week, randomized, double-blind study of olanzapine versus aripiprazole in the treatment of schizophrenia. J Clin Psychiatry. 2009;70:572–581. doi: 10.4088/jcp.08m04421. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Flow diagram of assessment of studies identified in the systematic review
Resultant search data. Tabulated study data for analyses