Introduction.—Dropout rates are high in trials involving psychosocially vulnerable populations, yet little is known about the factors contributing to dropout in this context. A better understanding of these factors would allow for more effective targeting of individuals likely to participate in intervention evaluation trials or for the adaptation of participant retention strategies based on the risk of dropout. Materials and Methods.- The objective is to study the determinants of early attrition (before three months postpartum) in CAPEDP, a randomized perinatal intervention trial targeting psychosocially vulnerable young mothers. This trial compared the effectiveness of the CAPEDP intervention against usual care in terms of improving parenting skills, reducing maternal depressive symptoms at 3 months postpartum, and reducing mental health disorders in the child at 2 years of age. Eligible women had to be first-time mothers, less than 27 weeks into their pregnancy, under 26 years of age, and meet at least one vulnerability criterion (educational level below a high school diploma, intention to raise the child alone, or coverage under universal health insurance or state medical aid). The CAPEDP intervention consisted of home visits conducted by psychologists (6 during the prenatal period, 23 in the first year, and 15 in the second year). Participants were assessed without knowledge of their randomization group during the prenatal period and at 3, 6, 12, 18, and 24 months of the child’s age. The maternal factors studied included sociodemographic factors (age, ethnicity, education level, income, employment, etc.), obstetric factors (use of healthcare services, history of abortion, pregnancy planning, etc.), family-related (marital status, etc.), and psychological (“Edinburgh Postnatal Depression Scale,” “Symptom Checklist 90” [SCL-90], “Vulnerable Attachment Style Questionnaire” [VASQ]). Results.- A total of 438 women were enrolled in nine maternity wards in the Île-de-France region between December 2006 and March 2009; 71 women, or 16% of the participants, withdrew their consent immediately after randomization, before any assessment had been conducted. The type of intervention was not significantly associated with this attrition. Between the baseline assessment and the child’s 3-month mark, 67 women withdrew their consent or were lost to follow-up. Factors predicting early attrition were maternal unemployment during pregnancy (87% of participants lost to follow-up (LTU) versus 73% of participants still followed, p = 0.018), a history of induced abortion (38% versus 25%, p = 0.037), and a more insecure attachment style (mean VASQ insecure score: 35.1 versus 33.6, p = 0.067). Conversely, an unplanned pregnancy (27% versus 40%, p = 0.049) and higher psychological vulnerability scores on the SCL-90 (phobia: 0.4 versus 0.6, p = 0.003; paranoia: 0.7 versus 0.9, p = 0.018; psychotic: 0.3 versus 0.5, p = 0.016) predicted higher retention in the study. Early attrition did not differ between the two randomization arms (16% of PDV in the CAPEDP arm versus 21% in the usual care arm, p = 0.215). Conclusion.- Social vulnerability factors are associated with higher attrition, whereas psychological vulnerability factors appear to be factors promoting retention in the study. The results of the multivariate analysis and the determinants of attrition from the intervention will be presented at the conference. Poster cited in the "Journal of Epidemiology and Public Health" in Supplement 2, Issue 62.