OBJECTIVES: Recurrent events are common in clinical trials of chronic diseases. Due to the ongoing nature of chronic conditions, and consequently clinical trials, patient follow-up can be intermittent and information regarding the time of symptom development is often unreported or poorly reported. Motivated by a large randomised trial in overactive bladder (OAB) syndrome, interest lies in evaluating the duration of treatment effect defined as the time to patient-reported return of symptoms following repeat injection of botulinum toxin in patients with interval censored data. METHODS: Poly-Weibull models were fitted in a Bayesian framework in order to obtain posterior predictive distributions from which to sample unreported event times for interval censored data. To further account for the correlation between repeated events within the same individual, we incorporate a shared frailty term. We applied this methodology to a clinical trial of patients receiving a maximum of 3 repeated injections of botulinum toxin for OAB over a 5 year follow-up period. To compare repeated injection, restricted mean survival time (RMST) was calculated. Performance of Bayesian prediction models were assessed for varying proportions of missing data, and misspecification of distributional form using simulation studies. RESULTS: Bayesian flexible parametric frailty models found that there may be a small cumulative effect of botulinum toxin in patients with OAB (RMST: 0.93, 95%CrI: 0.81,1.05; RMST: 0.98, 95%CrI: 0.75, 1.24 for active injection 1, 2 and 3 respectively) but this difference was not of clinical or patient importance. Simulation studies found that Bayesian prediction models generally perform well with up to 50% of interval censored data, but care should be taken when selecting distributional form. CONCLUSIONS: With an increasing need to assess the time to symptom recurrence in chronic conditions, and the difficulties faced with intermittent follow-up, the use of a flexible Bayesian framework would appear to be advantageous.