Demographic and Baseline Information
Among the 113 participants assessed for eligibility, 80 were recruited and randomly assigned to one of the groups (Figure 1). Among these patients, 78 completed the 6-week trial, and 71 completed the 12-week trial. At week 6, one patient (2.5%) in the amisulpride group, and one patient in the placebo group dropped out. At week 12, three patients (7.5%) in the amisulpride group and four patients (10.0%) in the placebo group dropped out. As shown in Table 1, at baseline, except for BMI (F = 4.85, p = 0.03), there was no significant difference in any demographic or clinical characteristics (PANSS, RBANS, SANS and CGI scores) between the two groups (all p > 0.05). Therefore, BMI was adjusted for in subsequent statistical analysis. There was no difference in the clozapine dose or serum clozapine levels at baseline between the amisulpride and the placebo groups (p > 0.05). Furthermore, after adjusting BMI and baseline clozapine dose, RM ANOVA showed no group × time effect, time effect, or group effect on serum clozapine levels (all p>0.05), indicating that there was no difference in the change of serum clozapine levels after 12 weeks of treatment. In addition, there was no difference in demographic or clinical characteristics between dropouts and completers (all p > 0.05).
Effect of amisulpride augmentation therapy on PANSS scores
RM MANOVA was first conducted using PANSS subscales and total score as the outcome measurement and BMI as the covariate, and showed a significant group × time effect (Wilks’ lambda F =10.50, p < 0.0001). Then, RM ANOVA showed group × time effects on PANSS total score (Wilks’ lambda F = 11.75, p < 0.0001), positive symptom subscore (Wilks’ lambda F=3.66, p=0.03) and general psychopathology subscore (Wilks’ lambda F = 9.03, p < 0.0001) (Table 2). Next, after adjusting for BMI, age, sex, disease course, baseline PANSS scores and baseline clozapine serum levels, an ANCOVA was used to examine the group difference in PANSS total and subscale scores, at week 6 and week 12, respectively. As shown in Figure 2A-2D, at week 12, amisulpride group displayed lower PANSS total score, positive symptom subscore, and general psychopathology subscore compared with placebo group (p Bonferroni = 0.004, Cohen's d = 0.45; p Bonferroni < 0.0001, Cohen's d = 0.97; p Bonferroni <0.001, Cohen's d = 0.92; respectively). At week 6, amisulpride group displayed lower PANSS total score, positive symptom subscore and general psychopathology subscore compared with placebo group (p Bonferroni = 0.004, Cohen's d = 0.88; p Bonferroni< 0.0001, Cohen's d = 0.99; p Bonferroni = 0.004, Cohen's d = 0.89; respectively).
Effect of amisulpride augmentation therapy on treatment response rate
ITT analysis showed that, at week 12, the response rate of amisulpride group (10 patients, 25%) was higher than that of placebo group (2 patients, 5%) (X2=6.28, p=0.01, OR = 6.33, 95% CI: 1.29-31.12). After adjusting for BMI, age, sex, disease course, and baseline clozapine serum levels, significant between-group differences (B=1.82, Wald's statistics= 4.06, p= 0.04, OR = 6.15, 95% CI: 1.06-36.03).
Effect of amisulpride augmentation therapy on cognitive function
RM MANOVA showed a significant group × time effect (Wilks’ lambda F=4.64; p=0.01) on RBANS scores. After adjusting for BMI, an RM ANOVA was applied for RBANS total and on subscale scores, respectively. As shown in Table 2, there was a group × time effect of RBANS total score and language score (Wilks’ lambda F = 3.54, p = 0.03; Wilks’ lambda F = 5.49, p = 0.006).
Then, after adjusting the baseline score and other clinical covariates, an ANCOVA was applied to examine the group differences in the RBANS total score and language score at week 6 and week 12, respectively. As shown in Figure 3A-3F, at week 12, the amisulpride group displyed higher RBANS total and language scores compared with placebo group (p = 0.01, Cohen's d = 0.41; p < 0.0001, Cohen's d = 0.77). However, only the difference in language score remained significant after Bonferroni correction (p Bonferroni < 0.001). At week 6, there were no between-group differences in RBANS total score or language score (p= 0.12; p = 0.08).
Effect of amisulpride augmentation therapy on SANS and CGI scores
An RM ANOVA was performed on SANS, CGI-S, CGI-I and CGI-E scores, after controlling for BMI as a covariate. As shown in Table 2, there were group × time effects on CGI-S (Wilks’ lambda F = 10.85; p<0.0001), CGI-I (Wilks’ lambda F = 4.16; p = 0.02) and CGI-E (Wilks’ lambda F = 12.17; p < 0.0001) scores. However, there was no significant group × time effect on SANS score (Figure 4A).
Next, an ANCOVA was conducted to examine the group differences in the CGI-S, CGI-I and CGI-E scores at week 6 and week 12, respectivey, after adjusting for baseline scores and other clinical covariates. As shown in Figure 4B-4D, at week 12, the amisulpride group had lower CGI-S, CGI-I and CGI-E scores than the placebo group (p < 0.0001, Cohen's d = 0.91; p = 0.02, Cohen's d = 0.80; p < 0.0001, Cohen's d = 1.06; respectively). However, after Bonferroni correction, only CGI-S and CGI-E scores still showed significant between-group differences (both p Bonferroni< 0.0001). At week 6, the amisulpride group had lower CGI-S and CGI-E scores than the placebo group (p Bonferroni = 0.003, Cohen's d = 0.72; p Bonferroni = 0.01, Cohen's d = 0.73).
Treatment side effects and Safety
As shown in Table 3, after adjusting for BMI as a covariate, RM ANOVA showed no significant group × time effect, main time effect, or group effect on the TESS total score (all p>0.05). In addition, after Bonferroni correction, there were no significant group × time effects on BMI, QT intervals, or laboratory measurements (p Bonferroni> 0.05).
At week 12, the most common adverse effects were mild in both groups, including dry mouth, constipation, extrapyramidal symptoms, gastrointestinal reactions, saliva, hypersomnia, insomnia and headache. There was no significant difference in the incidence of side effects between the amisulpride group and the placebo group (p > 0.05).