Pooled analysis of laboratory-based SARS-CoV-2 antigen immunoassays

To the Editor,

More than three years after the abrupt emergence of the still ongoing coronavirus disease 2019 (COVID-19) pandemic, the diagnostic approach to suspected SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2) acute infection remains almost entirely based on laboratory testing, as extensively reviewed elsewhere [1]. After an initial period exclusively characterized by molecular testing (i.e., SARS-CoV-2 RNA detection) as the only available weapon in our armamentarium, the current strategies for diagnosing acute SARS-CoV-2 infections have been recently supplemented by new options, endorsed by both the World Health Organization (WHO) [2] and the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) [3], and encompassing detection and/or measurement of SARS-CoV-2 antigens (Ag), especially those belonging to the nucleocapsid protein (N). These antigens can be assayed in respiratory specimens by means of rapid diagnostic tests (RDTs) or laboratory-based immunoassays, while either approach carrying obvious advantages and drawbacks, as summarized in Table 1. In particular, laboratory-based immunoassays have the remarkable benefits of higher diagnostic accuracy and throughput, combined with provision of quantitative test results and better evidence of clinical validation, which is often lacking in many SARS-CoV-2 Ag-RDT, especially for many low-price devices marketed in pharmacies and/or supermarkets.

To this end, joint efforts of the IFCC Task Force on COVID-19 and IFCC Working Group on SARS-CoV-2 variants have enabled to publish a series of critical literature reviews and meta-analyses, which have provided summary statistics of diagnostic accuracy of most of the currently commercially available laboratory-based SARS-CoV-2 immunoassays [4], [5], [6], [7], [8], [9], [10]. In this short report we aim to provide a brief overview of all such reports, compounded by a pooled analysis of their individual diagnostic performance. Briefly, the individual data of the meta-analyses were used for constructing a cumulative 2 × 2 table, which allowed the estimation of the pooled diagnostic performance of seven laboratory-based SARS-CoV-2 immunoassays currently available in the diagnostic market. The calculation of diagnostic accuracy has encompassed the generation of Summary Receiver Operating Characteristic Curves (SROC), agreement, diagnostic sensitivity, diagnostic specificity, negative (NPV) and positive (PPV) predictive values. Data were pooled using the Mantel-Haenszel test and a random effects model, while χ² test and I² statistic were used for calculating the heterogeneity. The statistical analysis was carried out using Meta-DiSc 1.4 (Unit of Clinical Biostatistics team of the Ramón y Cajal Hospital, Madrid, Spain) [11]. The analysis was performed in accordance with the Declaration of Helsinki, under the terms of all relevant local legislations.

The main characteristics of the seven immunoassays and the relative meta-analyses included in this study are summarized in Supplementary Table 1. Overall, 58 individual studies and 28,916 respiratory samples could be included in our pooled analysis, the results of which are shown in Table 2 and Supplementary Figure 1. Overall, the area under the curve (AUC) and accuracy were 0.978 and 0.92, with 0.74 sensitivity and 0.98 specificity. The PPV and NPV were instead 0.93 and 0.91, respectively. Importantly, for the four laboratory-based SARS-CoV-2 immunoassays for which sufficient data were available in respiratory samples with high viral load (21 pooled studies; 1,570 respiratory samples), the diagnostic sensitivity increased to 0.87 (95%CI, 0.85–0.89) (Supplementary Figure 2).

Taken together, the results of this pooled analysis show that, with values of accuracy, PPV and NPV higher than 0.90, laboratory-based immunoassays will ostensibly play a major role in the future clinical and laboratory management of the COVID-19 pandemic, the burden of which is now further aggravated by transformation into a “tripledemic” due to combination of COVID-19 with concomitant “waves” of common flu and Respiratory Syncytial Virus (RSV), thus virtually overwhelming healthcare capacity and responsiveness in many regions worldwide [12].

Importantly, the minimum performance criteria set by the WHO for diagnostic sensitivity (i.e., ≥80) and diagnostic specificity (i.e., ≥97) could only be met for the second statistics, whereby the cumulative sensitivity of these immunoassays remained inferior to the threshold set by the WHO (i.e., 0.74; with values <0.80 in 3/7 immunoassay). Nonetheless, the evidence that the diagnostic sensitivity increased to 0.87 in samples with high viral load, and that the WHO minimum performance criterion was hence nearly met by virtually all the laboratory-based immunoassays included in this analysis would enable to conclude that their usage may allow to reliably capture individuals with higher viral load in respiratory specimens, who are at major risk of spreading the virus (i.e., the so-called “super-spreaders”) and experiencing unfavourable disease progression and higher risk of long-COVID.

In conclusion, although we proffer that laboratory-based immunoassays are not reliable and accurate enough for completely replacing molecular detection of SARS-CoV-2 RNA for purposes of diagnosing acute infections to date, they could be more pragmatically employed for screening and identifying patients with higher SARS-CoV-2 viral load, especially in all circumstances where contagion is more likely to occur (i.e., mass gatherings, indoor meetings without using physical protective measures like glasses, face masks or social distancing), or where high infectivity is more likely to generate substantial harm in secondary cases (i.e., fragile and more vulnerable individuals such as those older, immunocompromised or bearing important medical conditions).