["Criminal Behaviour and Mental Health, Volume 31, Issue 6, Page 421-435, December 2021. ", "\nAbstract\n\nBackground\nVarious studies have demonstrated that people who do not complete sex offender treatment have higher recidivism rates than completers or untreated controls. The mechanism behind this association, however, remains unclear. One explanation is that responsivity issues are more prevalent in high‐risk offenders, making treatment failure and recidivism more likely, especially if treatment intensity is insufficient to match criminal needs. In addition, dropout may have a negative effect on offenders by increasing feelings of frustration or self‐doubt.\n\n\nAims\nTo retest for a link between treatment discontinuation and recidivism by sex offenders and analyse the mechanisms mediating the link between treatment discontinuation and their recidivism. We hypothesise that non‐completion has a negative effect on recidivism after controlling for a priori differences in risk of recidivism.\n\n\nMethods\nWe analysed data from all sex offenders released from Bavarian prisons between 2004 and 2015 who had participated in prison‐based social‐therapeutic treatment and compared non‐completers (n = 100) to offenders ending treatment as planned (n = 428). Criminal risk, offence characteristics, and treatment‐related propensities were measured by a comprehensive assessment instrument, including the Static‐99. Recidivism data were obtained from Federal Official Records with an average time‐at‐risk of 9.43 years (SD = 3.29).\n\n\nResults\nTreatment non‐completion was significantly associated with recidivism after controlling for criminal risk, offence characteristics and treatment duration. This was mainly accounted for by men who had spent at least two years in treatment before premature termination and return to an ordinary prison setting.\n\n\nConclusions\nOur findings aid understanding the potentially negative impact of discontinuing social‐therapeutic treatment for male sex offenders; termination after substantial commitment may be particularly problematic. Future research should encompass large enough sample sizes to study details behind staff or prisoner decisions to return the prisoner to ordinary prison conditions and to mitigate harms by loss of privileges.\n\n"]