Sampling strategies for detecting rare impurities: an application in gene therapy products

Detection of rare impurities in drug products presents special challenges. Replication competent adenovirus (RCA) is a rare impurity found in adenovirus-based gene therapy products. Various methods are used for detection of RCAs. We primarily focus on qualitative assays. Acceptance sampling plans for detecting RCAs in batches of gene therapy products are discussed. Assuming that the number of RCA units per patient dose follows a Poisson distribution, operating characteristics (OC) of these sampling plans can be studied. The OC curves display the acceptance probabilities for batches with specific true but unknown level of RCA and can be used to assess the specificity and sensitivity of the test strategies. Application of Bayesian methodologies in the assessment of RCA levels in drug batches is also discussed. Using observed data and prior belief, a 95% credible region for the number of RCA units per patient dose can be constructed. Both classical and Bayesian calculations display the impact of sample size, sampling fraction, and assay quality on the detection of RCA. For better sensitivity, the largest possible sampling fraction that does not interfere with the logistics and the performance of the assay should be used. The choice of sample size will depend on the upper limit of the biologically safe level of RCA, the testing strategy, the desired level of sensitivity and specificity, and also on practical issues.