The study took place at Kaohsiung Armed Forces General Hospital in Kaohsiung, Taiwan. All study procedures were approved by the Institutional Review Board ofKaohsiung Armed Forces General Hospital, Taiwan (Protocol ID: T1–02). All participants provided written consent for all study procedures. This study was registered at clinicaltrials.gov (identifier NCT03730857) on Feb. 11, 2019, and followed CONSORT guidelines for a randomized trial (S1 Table).
Fig. 1 is a flowchart of the recruitment process. We recruited 101 eligible adult patients with schizophrenia who had had an acute psychotic episode and had been admitted to a psychiatric ward. We excluded 11 eligible patients because they refused to participate in the study. Ninety inpatients participated in this study and were randomly assigned to receive treatment with paliperidone (n = 30), risperidone (n = 30) or olanzapine (n = 30) in an allocation ratio of 1:1:1. Participants were assigned using simple randomization procedures (computerized random numbers) to one of the three treatment groups. The randomization was conducted by a research assistant who was blinded to each patient’s status. This study complied with basic principles of balancing the likelihood of harm, the effectiveness of monitoring, and the potential severity of risk. The participants were required to be willing and able to be hospitalized, and the duration of hospitalization was at least 1 month before follow-up as outpatients.
The inclusion criteria for this study were: (1) age 18–65 years; (2) meeting thediagnostic criteria for schizophrenia according to the Diagnostic and Structural Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR) ; (3) not having received long-acting antipsychotic injections in the preceding 6 months; and4) no major systemic illnesses based on physical examination and laboratory test results. The exclusion criteria were as follows: (1) history of clozapine treatment in the previous 6 months; (2) treatment with cholesterol or triglyceride-lowering medications; (3) substantial risks of suicide or violent behavior; (4) pregnancy or breastfeeding; (5) documented neurological or other medical disorder that could influence the function of the central nervous system; and (6) unstable or critical untreated medical illness.
A triple therapy (paliperidone, risperidone and olanzapine) design was adopted in this study with a run-in period of 3 months. The patients were required to have discontinued all prior use of antipsychotics for at least 7 days before receiving one of the three atypical antipsychotics as a washout period.
The antipsychotic agents were administered with flexible dosing, which could be adjusted on the basis of the clinical judgment of the physicians in charge. The oral formulation was taken once daily and the recommended dosages for the three treatment groups were as follows: 6–12 mg paliperidone, 4–6 mg risperidone, and 10–20 mg olanzapine. Throughout the clinical trial, participants were permitted to use their concomitant medication continuously, including lorazepam (up to 6 mg/day) for sleep disturbance or agitation and biperiden (up to 6 mg/day) for treatment of extrapyramidal side effects. The use of other psychotropic agents and lipid-lowering medications was not allowed during the 3-month study.
The initial evaluation included a comprehensive history comprising the patient’s age at onset, duration of mental illness, educational level (years spent in school), family history of schizophrenia, and history of alcohol consumption and cigarette smoking. The clinical psychopathology of participants was assessed by Positive and Negative Syndrome Scale (PANSS) ratings. The PANSS total score ranges from 30 to 210, with higher scores indicating more severe psychopathology. The PANSS includes 30 items measuring positive symptoms, negative symptoms and general psychopathology . The raters who assessed the PANSS did not know the name of the assigned drug. The participants underwent evaluation of clinical psychopathology and metabolic parameters, including serum lipid levels and body mass index (BMI) at initial assessment (baseline, day 0) and were followed up at weeks 4, 8 and 12. In this study, we used the BMI to examine the individual variability in weight change. The height of each participant was measured at baseline, their weight was measured at each clinical visit, and the two measurements were used to calculate BMI (kg/m2). Fasting morning blood samples for the laboratory investigation were initially drawn at baseline (after completion of drug washout), and then at weeks 4, 8 and 12. Analyses of serum total cholesterol (TC), high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol and triglyceride (TG) were carried out by the standardized, automated, high throughput enzymatic analyses, which achieved within- and between-run coefficients of variation of < 5% .
All numeric variables are presented as mean ± SD and categorical variables are summarized using frequencies and percentages. A series of repeated measures analyses of variance were then applied to evaluate continuous variables among the treatment groups. Post hoc analyses were conducted via Tukey tests.
We also applied generalized estimating equations (GEE) for repeated measures analysis and corrected within-subjectcorrelations, and the statistical model was developed to establish a prognostic prediction model for the therapeutic effect of treatment of schizophrenia by atypical antipsychotics across all four time points (day 0 and weeks 4, 8 and 12) . The therapeutic effect was determined by the mean change in PANSS scores. The GEE method is often used for marginal regression analysis of longitudinal dataand other correlated response data, and is especially suitable for repeated measures studies. In addition to providing more efficient estimators of regression parameters, the main advantage of GEE is the comparison of correlated proportions using simulated data and coverage properties of the confidence interval. The GEE estimates average marginal effects and take into account the within-subject correlations by using an empirical covariance matrix.
We first used GEE linear regression models to evaluate adjusted associations between the lipid variables (total cholesterol, TG, HDL and LDL) and PANSS scores. The following baseline measures (see Table 1) were included as covariates in all adjusted analyses. We conducted a sequence of regression models adjusted for gender, age, age at onset, duration of mental illness (continuous), educational level (continuous), family history of schizophrenia (yes/no), history of alcohol consumption and cigarette smoking (yes/no), current type (categorical) and treatment dosage (continuous) of antipsychotics, and BMI (continuous). We then reduced this model by excluding covariates that did not predict the outcome with p ≤ 0.10, or did not change the estimated association between lipid variables and PANSS score by > 10% when removed from the model. The result of the final model is shown in Tables 3 and 4. The time points (weeks) were entered as a within-subject factor in order to examine whether there was a significant time effect in the repeated measures analysis. The data analyses were performed using SPSS version 21.0 software (SPSS Inc., Chicago, IL, USA/IBM, New York, USA). Statistical significance was defined as p < 0.05 (two-sided).