(M1030-01-05) Use of Technical Replicate Precision Acceptance for qPCR- and ddPCR-based Validations and Sample Testing

Purpose: Typically, duplicate or triplicate technical replicates are used to establish an average concentration when reporting a sample. Additionally, guidance in the industry is to apply a coefficient of variation (%CV) criteria around the Ct values of standards or samples. We examined the impact of using %CV on the log2 scaled Ct values, and also sought to explore the relationship between technical replicate precision, sample acceptance, overall assay precision, and study repeat rates when testing in duplicate or triplicate.

Methods: Using a mock data set, we looked at the impact of using a %CV on Ct values on the potential interpolated values. This was contrasted against using either a %CV on the interpolated values of each standard point, or by using a fixed acceptable Ct range. Using mock data sets and validation data sets, we performed precision analysis on the technical replicates and on overall precision of the accepted sample results by calculating the %CV. We also tested whether accepting higher intra-replicate precision compared to inter-assay precision impacted the overall precision performance of the method. A study sample set, tested in triplicate was also analyzed. We compared the sample failure rate in triplicate testing compared to duplicate testing using every possible duplicate combination generated from the triplicate results.

Results: At low Ct values (15) we found that using a 2% CV on Ct values results in a 1.5 Ct standard deviation which corresponds to a potential 139.69% CV on the interpolated value. This %CV increased with Ct value: a 2% CV criteria on an average Ct of 40 results in a 4.0 Ct standard deviation which corresponds to a potential 4,689.27% CV on the interpolated values. In contrast, using a 25% CV criteria on the interpolated values, or a 0.6 Ct range for triplicate values (0.5 Ct range for duplicate values) maintains consistent variability across the quantitative range of a qPCR assay. The intra-replicate %CV also impacts the potential results and status of a validation. When sample acceptance is wider than the expected assay precision (e.g. 1 Ct range between samples, 25% precision criteria) we found that out of 10 mock validations, 8 had sample dilution levels that would not pass the precision criteria. This failure rate was eliminated by bringing the intra-replicate precision into parity with the overall assay precision criteria. We also found the same effect on real validation data, with both overall accuracy and precision impacted by the criteria behind sample acceptance. For the impact of duplicate versus triplicate testing, we found that using triplicate testing greatly reduced the need for sample repeats. Triplicate testing resulted in an approximately 3% sample failure rate due to intra-replicate precision. Duplicate testing increased the sample retest rate to over 10% for the study examined. While this retest rate is much higher, the number of samples that can be tested per plate offer advantages for the change in technical replicates, with a 1,000 sample study estimated to take 39 plates using triplicate testing, reduced to 27 plates with duplicate testing.

Conclusion: The use of %CV on Ct values should not be used in PCR assays. Either the replicate interpolated values should be compared, or a Ct range should be applied to accept average sample values. Based on the simplicity of the Ct range, we recommend using this methodology for accepting sample replicates. Additionally, %CV of replicate values should be identical or stricter than the overall or inter-assay precision requirements of the method to ensure the best possibility of success. Duplicate or triplicate testing should be determined based on the expected type of study that the method will be used for. If samples are precious where a repeat may not be possible or extremely limited, a triplicate testing strategy should be employed where the individual sample repeat rate is reduced. If samples are readily available, a duplicate testing model likely will reduce the overall number of plates and shorten expected testing timelines.