Background and Aims: Relapse rates for psychosocial substance use disorder (SUD) treatments are high, and dropout is a robust predictor of relapse. This study aimed to estimate average dropout rates of in-person psychosocial SUD treatments and to assess predictors of dropout. Design: A comprehensive meta-analysis of dropout rates of studies of in-person psychosocial SUD treatment. Studies included randomized controlled trials (RCTs) and cohort studies. Setting: Studies conducted anywhere in the world that examined SUD treatment and were published from 1965 to 2016, inclusive. Participants/cases: One hundred and fifty-one studies, 338 study arms and 299 dropout rates including 26 243 participants. Measurements: Databases were searched for studies of SUD treatment that included an in-person psychosocial component. Meta-analyses and meta-regressions were conducted to estimate dropout rates and identify predictors of dropout, including participant characteristics, facilitator characteristics and treatment characteristics. Pooled estimates were calculated with random-effects analyses accounting for the hierarchical structure of study arms nested within studies. Findings: The average dropout rate across all studies and study arms was 30.4% [95% confidence interval (CI) = 27.2–33.8 and 95% prediction interval (PI) = 6.25–74.15], with substantial heterogeneity (I2 = 93.7%, P < 0.0001). Studies including a higher percentage of African Americans and lower-income individuals were associated with higher dropout rates. At intake, more cigarettes/day and a greater percentage of heroin use days were associated with lower dropout rates, whereas heavier cocaine use was associated with higher dropout rates. Dropout rates were highest for studies targeting cocaine, methamphetamines and major stimulants (broadly defined) and lowest for studies targeting alcohol, tobacco and heroin, although there were few studies on methamphetamines, major stimulants and heroin. Programs characterized by more treatment sessions and greater average session length were associated with higher dropout rates. Facilitator characteristics were not significantly associated with dropout. Conclusions: On average, approximately 30% of participants drop out of in-person psychosocial SUD treatment studies, but there is wide variability. Drop-out rates vary with the treated population, the substance being targeted, and the characteristics of the treatment.