The recent contribution by Lippi et al., in a rebuttal to a previously published paper by de La Matta et al., on this journal,1,2 addressed the analytical performance of an RT-PCR swab in COVID-19 pandemic. The effectiveness of a test for COVID-19 diagnosis has many pitfalls, a sound estimation of test accuracy must be interpreted bearing in mind that these analyses do not possess a definitive reference standard able to diagnose or even rule out a SARS-CoV2 infection. Moreover, infection prevalence plays a role of utmost importance. In their comment to de La Matta et al.’s paper, the authors did not consider the role of virus prevalence in the tested population, which significantly affect the final outcome of a swab test.3 In Italy, our calculations reached a power of 86.65% (IC95 = 76.08%–94.88%) for the first wave of pandemic (March–June 2020) and a power of 76.77% (IC95 = 64.33%–81.88%) in the second wave (Sept–Dec 2020), to detect true positive SARS-CoV2 infected subjects (Software STATA Lab v 12.1).
Actually, the conclusions forwarded by de La Matta et al., referred to scenarios of very low prevalence. The inability of a molecular RT-PCR swab to establish a proper cutoff of negativity, likewise other viruses such as HIV, for which an effective anti-viral therapy is available, set the questionable item of RT-PCR performance on Bayesian parameters such as virus prevalence and analytical performance. It is very hard to set a diagnostic negativity in the complete absence of a quantitative cutoff. Some authors have outlined that it is quite embarrassing that scant studies addressed the prevalence of false negative results for SARS-CoV2 molecular assays.4 Contrarily to the authors,1 is particularly odd to evaluate the quality performance of a RT-PCR test if the threshold under which no infectivity can be forecast (as occurring for HIV) is not clearly established.5 The magnification of false negative reported by Lippi et al., cannot depend on the simple categorization in asymptomatic or mild symptomatic, as the lack of true reference standards makes very burdensome to achieve a correct PPV. A recent pooled analysis of 16 studies (3818 patients) estimated a sensitivity of 87.8% (95% CI 81.5%–92.2%) for an initial reverse-transcriptase PCR test, which is far from the 95% reported by the authors.1 The increase in false negative results cannot affect the prevalence of SARS-CoV2 infection without having any standard able to set a correct cutoff of virus negativity in the nasal mucosa.
Conflict of interestThe author states he has no conflict of interest.