To the Editor In the wake of the recently published ω-3 polyunsaturated fatty acids multisite trial (NEURAPRO) by McGorry et al 1 in JAMA Psychiatry (n = 304), the 3 largest studies of preventive interventions in individuals at ultrahigh risk (UHR) for psychosis have turned out to be negative (Morrison et al, 2 n = 288; McFarlane et al, 3 n = 292), suggesting that it may not currently be possible to prevent psychosis. However, the actual risk for psychosis in these interventional studies was found to be extremely low, even in the control group (11.2% in the McGorry et al 1 study at 1 year, 9% at 2 years in the Morrison et al 2 study, and 2.3% at 2 years in the McFarlane et al 3 study), implying that these negative findings are likely to be secondary to small statistical power.On a conceptual level, these studies challenge the underlying assumption that it is acceptable to meet specific psychometric criteria at intake to be slotted into the UHR category. Actually, the predictive power of the UHR criteria is not fixed priority, but is dependent on the risk enrichment of the samples to which they are applied (ie, "pretest" risk). 4 This is owing to the psychometric characteristics of these UHR instruments: they are good to rule out psychosis but only moderately useful to rule in psychosis. 4 Therefore, most of the observed risk for psychosis is not gained through UHR assessment per se, but is obtained before the assessment (pretest phase) during the recruitment of these individuals. Pretest risk enrichment in those undergoing a UHR assessment remains substantial (15% at 38 months) and highly heterogeneous (9%-24% at 38 months). 4 Pretest risk enrichment is modulated by the type of recruitment strategies adopted to select individuals for UHR assessment. 4 For example, recruiting individuals who were already filtered by adult mental health services is associated with higher pretest risk enrichment, as compared with recruiting individuals through intensive community outreach. 5 Not surprisingly, the lowest transition risk was reported in the study by McFarlane et al,3 which has adopted a "community outreach and education program targeted to teachers, school, and college counsellors." Owing to the heightened pressure to recruit participants into interventional studies, it is possible that a substantial proportion of these samples had been recruited through community outreach, diluting the transition risks.Therefore, it seems expedient to control the pretest risk enrichment of samples recruited into future interventional studies for psychosis prevention by using pretest risk stratification models that have been recently validated in this journal. 5