Sexual offending against children is among the most serious of crimes, and so it is imperative to have empirical evidence that prison-based treatment programmes for such offenders are successful in reducing the likelihood of future recidivism. The present study examined the criminal history and post-release outcomes for sexual offenders against children who had attended the Kia Marama Special Treatment Unit (STU; treated group; N = 428) and were followed up for an average of 6.36 years, and a cohort of offenders who were also incarcerated for sexual offending against children but did not attend Kia Marama or a similar STU (untreated group; N =1956) and were followed up for an average of 6.81 years. During the follow-up, rates of recidivism (defined as a charge for a new offence) for the untreated and treated groups, respectively were: 7.52% and 7.24% (sexual); 18.35% and 10.28% (violent); and 38.24% and 32.71% (general). There were significant differences between the groups in terms of offence history: The treated group had more prior sexual offences, more prior sentencing dates, more non-contact offences, was more likely to have had a male victim, a longer sentence duration, and an overall higher static risk score, whereas the untreated group were more likely to be young (< 25 years at release) and had more prior violent convictions. To control for differences in risk level between the two groups, we used stepwise logistic regression to develop predictive models for recidivism, and then tested whether treatment group was related to recidivism. Because the design was unbalanced and groups were heterogeneous, to obtain the best estimate of model coefficients we used a differential weighting factor that compensated for the unequal group sizes. Results showed that the treatment group was associated with a significant reduction in sexual, violent and general recidivism. The estimated coefficient for the treatment group indicated a 29.4% reduction in sexual recidivism (odds ratio [OR] = .706), a 49.3% reduction in violent recidivism (OR = .507) and a 27.6% reduction in general recidivism (OR = .724). This result implies that the expected sexual recidivism rate of the treated group would have been 10.0% rather than 7.2% had they not attended the STU. The present findings update previous results of Marentette (2009) with a more comprehensive sample, and add to the growing evidence for the effectiveness of cognitive-behavioural and relapseprevention- based treatment for sexual offenders against children. Our results provide further evidence of the effectiveness of the Kia Marama STU (see also Beggs & Grace, 2011) and endorse the utility of such programmes for reducing sexual offending against children in New Zealand.