Sensitivity, Specificity and Predictive Values of Molecular and Serological Tests for COVID-19: A Longitudinal Study in Emergency Room

Abstract

Background: We assessed the sensitivity, specificity and positive and negative predictive value (PPV and NPV) of molecular and serological tests for the diagnosis of SARS-CoV-2 infection.

Methods: A total of 346 patients were enrolled in the emergency room. We evaluated three Reverse Transcriptase-real time PCRs (RT-PCRs) including six different gene targets, five serologic rapid diagnostic tests (RDT) and one ELISA. The final classification of infected/non-infected patients was performed using Latent Class Analysis combined with clinical re-assessment of incongruous cases.

Results: Out of these, 24.6% of patients were classified as infected. The molecular test RQ-SARS-nCoV-2 showed the highest performance with 91.8% sensitivity, 100% specificity, 100.0% PPV and 97.4% NPV respectively. Considering the single gene targets, S and RdRp of RQ-SARS-nCoV-2 had the highest sensitivity (94.1%). The in-house RdRp presented the lowest sensitivity (62.4%). The specificity ranged from 99.2% for in-house RdRp and N2 to 95.0% for E. The PPV ranged from 97.1% of N2 to 85.4% of E and the NPV from 98.1% of S to 89.0% of in-house RdRp. All serological tests had < 50% sensitivity and low PPV and NPV. VivaDiag IgM (RDT) had 98.5% specificity, with 84.0% PPV, but 24.7% sensitivity.

Conclusion: Molecular tests for SARS-CoV-2 infection showed excellent specificity, but significant differences in sensitivity. Serological tests have limited utility in a clinical context.

Keywords: RT-PCR; SARS-CoV-2; accuracy; diagnosis; serology.

Figure 1

Study Flow chart. The white arrows indicate the study flow, up to the final classification of 85 subjects with the infection (red box) and 261 without the infection. The two grey arrows point to the two boxes resuming all the available details and outcomes of the two groups. ER, emergency room; NEG, negative; POS, positive; GP, general practitioner.

Figure 2

Table chart on main demographic, clinical, laboratory and imaging characteristics of the study population. CRP, C reactive protein; RR, respiratory rate; WBC, white blood cells; pO2, partial pressure of oxygen; SaO2, oxygen saturation.

Figure 3

Sensitivity (filled squares), specificity (filled rhombus) and predictive values of three molecular tests and of single gene targets, according to the final classification (Infected = 85, not infective = 261). “Target S and RdRp (kit)” corresponds to the RealQuality RQ-SARS-nCoV-2 assay (cod. RQ-130, AB Analitica, Italy), targeting the spike protein gene (S) and the RNA-dependent RNA polymerase gene (RdRp). “Target N1 and N2” corresponds to the CDC 2019-Novel Coronavirus (2019-nCoV) Real-Time RT-PCR Diagnostic Panel, targeting the nucleocapsid protein gene (N), regions N1 and N2. “Target E and RdRp” corresponds to the in-house RT-PCR protocol, targeting the envelope protein gene (E) and the RdRp. +, positive result; −, negative result; Se, sensitivity; Sp, specificity; PPV, positive predictive value; NPV, negative predictive value.

Figure 4

Visual representation of the 70 records with at least one discordant gene target result.

Figure 5

Sensitivity (filled squares), specificity (filled rhombus) and predictive values of six antibody tests according to the final classification (Infected = 85, not infective = 261).

Figure 6

Test accuracies using a subset of 140 patients with a symptom duration of ≥ 7 days. Sensitivity (filled squares), specificity (filled rhombus) and predictive values of six antibody tests according to the final classification.