Diagnostic accuracy estimates for COVID-19 RT-PCR and Lateral flow immunoassay tests with Bayesian latent class models

Abstract

The objective of this work was to estimate the diagnostic accuracy of RT-PCR and Lateral flow immunoassay tests (LFIA) for COVID-19, depending on the time post symptom onset. Based on the cross-classified results of RT-PCR and LFIA, we used Bayesian latent class models (BLCMs), which do not require a gold standard for the evaluation of diagnostics. Data were extracted from studies that evaluated LFIA (IgG and/or IgM) assays using RT-PCR as the reference method. The cross-classified results of LFIA and RT-PCR were analysed separately for the first, second and third week post symptom onset. The SeRT-PCR was 0.695 (95% probability intervals: 0.563; 0.837) for the first week and remained similar for the second and the third week. The SeIgG/M was 0.318 (0.229; 0.416) for the first week and increased steadily. It was 0.755 (0.673; 0.829) and 0.927 (0.881; 0.965) for the second and third week, respectively. Both tests had a high to absolute Sp, with point median estimates for SpRT-PCR being consistently higher. SpRT-PCR was 0.990 (0.980; 0.998) for the first week. The corresponding value for SpIgG/M was 0.962 (0.905; 0.998). Further, Sp estimates for each test did not differ between weeks. BLCMs provide a valid and efficient alternative for evaluating the rapidly evolving diagnostics for COVID-19, under various clinical settings and for different risk profiles.

Bayesian latent class models to estimate diagnostic test accuracies of COVID‐19 tests

“…Therefore, in a Bayesian framework, we rely on prior information for the specificities of all three tests similar to a previous analysis with two tests and four populations. 6 We assume that they are close to 100%. Further technical details about the model and Markov Chain Monte Carlo simulation, including the code to reproduce the results, are available on (https://github.com/shartn/BLCM-COVID19).…”

Performance of Three Tests for SARS-CoV-2 on a University Campus Estimated Jointly with Bayesian Latent Class Modeling

“…While this approach does not make test results more accurate per se , it does reduce the risk of bias associated with erroneously assuming that a gold standard is without error. This approach has been applied in some cases for molecular tests for SARS-CoV-2, resulting in differences relative to estimates that relied on a gold standard ( 22 – 24 ). For example, in a meta-analysis comparing RT-PCR testing of nasopharyngeal and saliva samples, allowing for imperfections in both types of tests resulted in higher estimates of specificity and narrower uncertainty about sensitivity ( 23 ).…”

“…In this study, we applied BLCM to a data set from a SARS-CoV-2 testing program in a university setting during October 2020. A unique feature of this data set is that it includes both RT-PCR and antigen tests, which have not been compared in previous BLCM analyses for SARS-CoV-2 that we are aware of ( 22 – 24 ). Another unique feature of this data set is that the majority of subjects were tested during surveillance screening and were not suspected of being infected at the time of testing.…”

Performance of three molecular tests for SARS-CoV-2 on a university campus estimated jointly with Bayesian latent class modeling

“…This method involves joint estimation of the sensitivity and specificity of each type of test used, by virtue of considering the possibility that any given test result could have been erroneous for some, all, or none of the tests used. This approach has been applied in some cases for molecular tests for SARS-CoV-2, resulting in differences relative to estimates that relied on a gold standard (22–24). For example, in a meta-analysis comparing RT-PCR testing of nasopharyngeal and saliva samples, allowing for imperfections in both types of tests resulted in higher estimates of specificity and narrower uncertainty about sensitivity (23).…”

“…In this study, we applied BLCM to a data set from a SARS-CoV-2 testing program in a university setting during October 2020. A unique feature of this data set is that it includes both RT-PCR and antigen tests, which have not been compared in previous BLCM analyses for SARS-CoV-2 that we are aware of (22–24). Another unique feature of this data set is that the majority of subjects were tested for surveillance screening and were not suspected of being infected at the time of testing.…”