Patients at 98 centres in 12 countries throughout Europe were enrolled in the study between July 2004 and June 2005. The study protocol complied with the Declaration of Helsinki and Good Clinical Practices (GCP) guidelines and was approved by the Institutional Review Board (IRB)/Independent Ethics Committee (IEC) at the individual study centres. Prior to study enrolment, informed written consent was obtained from all patients.
Eligible for inclusion in the study were male and female patients aged 18 to 65 years with a diagnosis of schizophrenia (DSM-IV-TR criteria) treated within a community mental health or hospital-based outpatient centre, for whom a change of antipsychotic medication was indicated due to lack of tolerability and/or symptom control as judged by the clinician.
Key exclusion criteria included acute psychotic symptoms requiring hospitalisation, risk of committing suicide, a diagnosis of schizoaffective disorder, bipolar disorder, depression with psychotic symptoms, or organic brain syndrome. Patients were also excluded if they were considered treatment resistant, had had a significant psychoactive substance use disorder within 3 months prior to screening, or had any history of neuroleptic malignant syndrome, epilepsy, seizures, abnormal electroencephalogram, severe head injury, or stroke.
The design and methodology of the STAR study have been presented in detail elsewhere; hence only a summary will be provided here .
The STAR study was designed as a multi-centre, randomised, naturalistic, open-label study to compare aripiprazole treatment with SOC treatment (limited to a choice of olanzapine, quetiapine or risperidone). Eligible patients were randomised to receive treatment with aripiprazole or SOC for a period of 26 weeks. Patients randomised to the SOC group received one of three selected atypical antipsychotics: olanzapine, quetiapine, or risperidone, based on the investigator's/clinician's judgment as to the optimal treatment for the individual patient and taking into account the patients' previous response to antipsychotic medication.
A cross-titration period of up to 14 days was permitted following the randomisation, with complete discontinuation of any pre-study antipsychotic medication by Day 15. During the 26-week open-label treatment phase, study visits occurred at Weeks 2, 4, 8, 12, 18 and 26 to assess the effectiveness of study treatment.
Patients randomised to the aripiprazole group received a starting dose of 15 mg/day followed by a daily dose of 10–30 mg/day, as judged by the investigator, administered orally once daily independent of meals. Patients randomised to the SOC group received either olanzapine, starting at 10 mg once daily followed by a daily dose of 5–20 mg/day; quetiapine, starting with 50 mg/day on Day 1, 100 mg/day on Day 2, 200 mg/day on Day 3, and 300 mg/day on Day 4, then followed by an individually titrated dose of 800 mg/day maximum; risperidone, starting with 2 mg/day on Day 1, 4 mg/day on Day 2, then a daily dose of 2–16 mg/day; or dosages according to the approved local labelling for each of the three medications.
Benzodiazepines and anticholinergics were allowed during the study if deemed necessary by the investigator. Other psychotropic medications such as antidepressants and mood stabilisers were allowed if the patient was already receiving these medications when entering the study. No additional antipsychotics were allowed during the study.
Full details of the Investigator Assessment Questionnaire (IAQ) and Clinical Global Impression (CGI) Improvement (CGI-I) and Severity (CGI-S) assessments have been published elsewhere . Briefly, the clinical effectiveness of the study treatments was assessed through the IAQ. The primary effectiveness variable in the trial was the mean total score of the 10-item IAQ at Week 26, last observation carried forward (LOCF).
Secondary effectiveness variables included the CGI-I and CGI-S scales. IAQ, CGI-I and CGI-S scores were obtained at Weeks 2, 4, 8, 12, 18, and at the end of Week 26.
Outcomes research assessments
The STAR study outcomes research evaluation comprised the following tools: the Preference of Medicine (POM) questionnaire, the Quality of Life Scale (QLS), the Client Socio-demographic and Service Receipt Inventory – European version (CSSRI-EU), EuroQoL-5D (EQ-5D), Impact of Weight on Quality of Life (IWQoL-Lite), and the Arizona Sexual Experience Scale (ASEX). This paper will report and discuss the findings of the ASEX analysis; the other analyses will be reported separately.
Arizona Sexual Experience Scale (ASEX)
The ASEX is a 5-item scale to assess sexual dysfunction amongst psychiatric patients . The ASEX rates sex drive, arousal, vaginal lubrication/penile erection, ability to reach orgasm, and satisfaction from orgasm. Male and female patients are assessed separately. Possible Total Score ranges from 5 to 30, with the higher score indicating more patient sexual dysfunction. The scale could be either self-administered or clinician-administered at the discretion of the treating clinician. The scale is applicable to patients regardless of the availability of a sexual partner. The ASEX was administered at baseline and at Weeks 8, 18, with a final assessment at the end of Week 26 or at the time of premature discontinuation.
The safety variables of interest for this paper are the change from baseline in serum prolactin levels throughout the study, and the proportion of patients with potentially clinically significant elevations in prolactin levels. Prolactin serum levels were assessed by the Abbott AxSYM system prolactin assay – a Microparticle Enzyme Immunoassay (MEIA) for the quantitative measurement of serum prolactin. The normal range for this test was defined as 1.6–18.8 mg/dL for men and 1.4–24.2 mg/dL for women. The Statistical Analysis Plan for the STAR study defines a clinically significant lab value for prolactin as any value above the upper limit of normal. In addition to serum prolactin, the safety analysis comprised the incidence of adverse events (AEs), including treatment-emergent AEs, AEs leading to discontinuation, treatment-emergent serious AEs, and treatment-emergent EPS-related AEs. Additional safety endpoints included the proportion of subjects with potentially clinically relevant vital signs/laboratory findings as defined by the Food and Drug Administration Division of Neuropharmacological Drug Products 
Sample size and power
It was estimated that 556 patients would have to be randomised to obtain 500 evaluable patients (250 on aripiprazole and 250 on SOC), assuming that 90% of the randomised patients would be evaluable in the primary endpoint. This sample size would yield 95% statistical power with a true difference in the primary endpoint of 2. This assumed a standard deviation of 6.2 and a 2-sided t-test at the 0.05 level of significance. This sample size would also allow for meaningful analysis of secondary endpoints for which it is difficult to obtain reliable estimates of variance, and for exploratory analysis to estimate the effect size of aripiprazole relative to the individual SOC items.
The safety sample comprised all randomised patients who took at least one dose of study medication. The effectiveness sample included all patients in the safety sample who had at least one post-baseline effectiveness evaluation (IAQ or CGI). The outcome research sample, which included assessments of quality of life and medication preference, was comprised of all patients in the safety sample who had at least one post-baseline outcome research evaluation (including ASEX). LOCF was the primary analysis used for the effectiveness research sample (IAQ, CGI-I, CGI-S), whereas observed cases (OC) was the primary analysis used for the outcomes research sample (ASEX).
The results from the ASEX evaluation were subject to analysis of covariance (ANCOVA) with respect to the mean change from baseline in ASEX Total Score. The ANCOVA model included treatment group (aripiprazole, SOC) and gender as main effects, with the baseline score as a covariate. Each additional OC week and each LOCF week was analysed similarly. For the analysis of LOCF data, country was also included as a main effect within the model. In addition, the baseline ASEX Total Score was analysed with an ANCOVA model including treatment group and gender (and country for the LOCF data set).
The proportions of patients with potentially clinically relevant laboratory test values during the 26-week period were summarised by treatment group. The proportions of patients with potentially clinically relevant metabolic parameters during the 26-week period were analysed using a Cochran-Mantel-Haenszel test controlling for the presence or absence of baseline abnormality.
The change from baseline in serum prolactin was analysed using an ANCOVA model which included treatment group as main effect and baseline value as the covariate. Baseline serum prolactin levels were analysed with an ANCOVA model which included treatment group.
Correlation between changes in prolactin and in ASEX Total score were calculated in a post-hoc analysis using Spearman rank-order partial correlation controlling for antipsychotic taken and gender. The antipsychotic taken was preferred in this analysis over the treatment group as the effect on prolactin may be different for each antipsychotic. In another post-hoc analysis, the change from baseline in ASEX Total score (LOCF) was analysed using the ANCOVA model as described above, but with on treatment CGI-Improvement added as covariates to assess whether the ASEX advantage is specific and not the result of the patients overall improvement. Although such an analysis may suggest a residual treatment effect on the ASEX Total Score that can not be attributed to the patient's global improvement, any analysis that adjusts for post-randomization covariates must be considered speculative in nature.