Five Clinical Trials in Schizophrenia
This dataset combines the data that were collected in five double-blind randomized clinical trials. In these trials, the objective was to examine the efficacy of risperidone to treat schizophrenia. Schizophrenia is a mental disease that is hallmarked by hallucinations and delusions (American Psychiatric Association, 2000).
In each trial, the Clinical Global Impression (CGI; Guy, 1976), the Brief Psychiatric Rating Scale (BPRS; Overall and Gorham, 1962), and the Positive and Negative Syndrome Scale (PANSS; Singh and Kay, 1975) were administered. These instruments are clinical rating scales that are routinely used to
Age-Related Macular Degeneration Trial. Scatter plot of the change in visual acuity after 52 weeks (T) against the change in visual acuity after 24 weeks (S). The full and dashed lines result from regressing T on S in the groups that were administered interferon-a and placebo, respectively.
assess symptom severity in patients with schizophrenia (Mortimer, 2007). The patients in the different trials were administered risperidone or an active control (e.g., haloperidol, levomepromazine, or perphenazine) for four to eight weeks. The main endpoints of interest were the change in the CGI score (= CGI score at the end of the treatment — CGI score at the start of the treatment), the change in the PANSS score, and the change in the BPRS score.
A total of 2,128 patients participated in the five trials (1,591 patients received risperidone and 537 patients were given an active control). The patients were treated by a total of N = 198 psychiatrists. Each of the psychiatrists treated between n* = 1 and 52 patients.
Figure 2.4 shows the histograms of the changes in the CGI, BPRS, and PANSS scores, respectively. As can be seen, the CGI score is a categorical outcome (score range [1, 7]) whilst the BPRS and PANSS scores can be considered (semi)continuous outcomes that are approximately normally distributed. Figure 2.5 shows scatter plots of the different combinations of the endpoints. Especially the change in the PANSS and the change in the BPRS scores were highly correlated, i.e., r(BPRS, PANSS) = 0.9597 and 0.9644 in the active control and the experimental treatment groups, respectively (when the information of both treatment groups is combined, r(BPRS, PANSS) = 0.9634). The correlations between the changes in the CGI and the changes in BPRS/PANSS scores were moderate in both treatment groups, i.e., r(CGI, BPRS1) = 0.7394 and 0.7393 in the active conTABLE 2.1
Age-Related Macular Degeneration Trial. Overview of the variables in the dataset.
Age-Related Macular Degeneration (ARMD) Trial. First five observations in the dataset.
trol and the experimental treatment groups and r(CGt, PANSS) = 0.7532 and 0.7503 in the active control and the experimental treatment groups, respectively. When the information of both treatment groups is combined, r(CGI, BPRS) = 0.7394 and r(CGI, PANSS) = 0.7522.
The means of the changes in the active control and the risperidone treatment groups were 3.4465 and 3.2050 for the CGI endpoint (Дcar = -0.2415, t = -3.254, p = 0.0012; two-sided t-test), -6.6461 and -8.9383 for the BPRS endpoint (ABPrs = -2.2923, t = -3.407, p = 0.0007), and -11.4719 and -16.0151 for the PANSS endpoint (ДPanSS = -4.5432, t = -3.876, p = 0.0001), respectively. Note that a larger positive change in the CGI score indicates deteriorated mental health (more schizophrenic symptoms), whereas a larger negative change in the PANSS or BPRS score indicates improved mental health (less schizophrenic symptoms). There was thus strong evidence that the true differences of the means for the three endpoints in both treatment groups differed significantly from zero (in favor of risperidone).
In the reminder of this book, this dataset will be referred to as the Schizophrenia dataset. In the subsequent chapters, different combinations of
Five Clinical Trials in Schizophrenia. Histograms of the change in the CGI scores (upper left figure), the change in the BPRS scores (upper right figure), and the change in PANSS scores (bottom figure).
Note: CGI = Clinical Global Impression, BPRS = Brief Psychiatric Rating Scale, and PANSS = Positive and Negative Syndrome Scale.
the endpoints (CGI, BPRS, and PANSS) in various forms (binary, continuous) will be considered as the candidate surrogate and the true endpoint. The binary endpoints BPRS/PANSS/CGI reflect the presence or absence of clinically relevant change in schizophrenic symptomatology. Clinically relevant change is defined as a reduction of 20% or more in the BPRS/PANSS scores,
Five Clinical Trials in Schizophrenia. Overview of the variables in the dataset.
i.e., 20% reduction in posttreatment scores relative to baseline scores, or a change of 3 points in the original CGI scale (Kane et al., 1988; Leucht et al., 2005).
The dataset is included in the R library Surrogate (where it can be accessed using the command data(Schizo)) and it can be downloaded from http: //ibiostat.be/online-resources (file Schizo.txt). Table 2.3 provides an overview of the variables that are included in the Schizophrenia dataset. The dataset is organized in the “wide” format (i.e., each row contains the data of a single patient).
The data were provided by the Risperidone Study Group (Peuskens et al., 1995).