Effect sizes and heterogeneity of effect sizes on continuous measures.

aThe FDA statistical review noted that many participants violated study protocol, often as a result of taking non-allowed medications. The numbers in parentheses represent data from only participants who did not violate the study protocol.bAs data were not presented clearly for two of three trials, we opted to not treat the single study with clearly reported data as representative of the three aripiprazole studies; thus, we provide no summary effect size calculation for sexual functioning.c“ns” indicates no statistically significant difference versus placebo; data were not reported in a more detailed manner.dOFC had a greater increase in prolactin versus fluoxetine (0.31 nmol/l, ppp-values, we did not calculate an effect size.eOFC had a greater increase in mean total nonfasting cholesterol versus fluoxetine (0.30 mmol/l, pp = 0.004). Given the inexact p-value of the OFC versus fluoxetine comparison, we did not calculate an effect size.fHedges' g includes effects from all three trials, whereas the raw unit analysis includes only change on the MADRS, which was not used in the McIntyre et al. [84] study.gIt was reported that 150 mg of quetiapine was superior to placebo, with an associated p-value of 0.095, from which a d = 0.20 was calculated. For the 300-mg dose, the p-value was reported as hThere were “no apparent differences among the treatment groups” according to the El-Khalili et al. [83] clinical trial registry report. Given the lack of clarity regarding sexual functioning data, we provide no summary effect size calculation for sexual functioning.i“ns” indicates no statistically significant difference versus placebo; data were not reported in a more detailed manner.jThe primary efficacy end point in the Mahmoud et al. [44] study was 4 wk, so data in parentheses indicate data from this a priori end point rather than from the 6-wk end point that was emphasized much more heavily in the study publication.kHedges' g pools data from Mahmoud et al. [44] and Reeves et al. [77]. We provide no summary estimate of mean differences on the MADRS, as only the small Reeves et al. [77] study reported these data.lMahmoud et al.'s primary end point was 4 wk [44], but these data are presented at 6 wk. These results may be inflated relative to the primary end point, given that the effect favoring risperidone on the HAM-D was smaller at week 4 than at week 6.mProlactin levels were apparently not measured in these trials.nPooled raw units are for MADRS only.oThese data are pooled across the SDS, Q-LES-Q, SF-36 Mental Component Summary, and SF-36 Physical Component Summary.CGI-S, Clinical Global Impressions–Severity; CSFQ, Changes in Sexual Function Questionnaire; IDS-SR, Inventory of Depressive Symptomatology–Self Report; QoL, quality of life; SF-36 MCS, SF-36 Mental Component Summary; SF-36 PCS, SF-36 Physical Component Summary; SFI, Sexual Function Inventory.