Table 7

Summary of main findings

Test methodClassification by serology testResults per 1000 patients tested (95% CI)
5% prevalence10% prevalence20% prevalence
Population: SARS-CoV-2 infected
ELISA (IgG or IgM):    
 9 studies, 766 samples. Pooled sensitivity 84.3% (95% CI 75.6% to 90.9%)Correctly classified as infected42 (38 to 45)84 (76 to 91)169 (151 to 182)
Incorrectly classified as uninfected8 (5 to 12)16 (9 to 24)31 (18 to 49)
LFIA (IgG or IgM):    
 11 studies, 2660 samples. Pooled sensitivity 66.0% (95% CI 49.3% to 79.3%)Correctly classified as infected33 (25 to 40)66 (49 to 79)132 (99 to 159)
Incorrectly classified as uninfected17 (10 to 25)34 (21 to 51)68 (41 to 101)
CLIA (IgG or IgM):    
 2 studies, 375 samples. Pooled sensitivity 97.8% (95% CI 46.2% to 100%)Correctly classified as infected49 (23 to 50)98 (46 to 100)196 (92 to 200)
Incorrectly classified as uninfected1 (0 to 27)2 (0 to 54)4 (0 to 108)
Population: not infected with SARS-CoV-2
ELISA (IgG or IgM):    
 6 studies, 1109 samples. Pooled specificity 97.6% (95% CI 93.2% to 99.4%)Correctly classified as uninfected931 (884 to 941)882 (837 to 891)784 (744 to 792)
Incorrectly classified as infected19 (9 to 66)18 (9 to 63)16 (8 to 56)
LFIA (IgG or IgM):    
 11 studies, 2874 samples. Pooled specificity 96.6% (95% CI 94.3% to 98.2%)Correctly classified as uninfected918 (896 to 933)869 (849 to 884)773 (754 to 786)
Incorrectly classified as infected32 (17 to 54)31 (16 to 51)27 (14 to 46)
CLIA (IgG):    
 9 studies, 2804 samples. Pooled specificity 97.8% (95% CI 62.9% to 99.9%)Correctly classified as uninfected929 (598 to 949)880 (566 to 899)782 (503 to 799)
Incorrectly classified as infected21 (1 to 352)20 (1 to 334)18 (1 to 297)

Quality of evidence and practical implications: Pooled sensitivity and specificity should be interpreted with caution. Accuracy might have been over-estimated in most studies owing to bias arising from patient selection or how index and reference tests were performed, or both. Estimates of sensitivity and specificity were inconsistent between studies (heterogeneity was important). Estimates might have limited applicability to outpatient settings and for testing at the point of care. Point-of-care LFIAs consistently had the lowest sensitivity and specificity. The poorest performance was seen with commercial LFIA kits; these tests should not be used for clinical purposes. Clinical studies designed to overcome the weaknesses are urgently needed.

RT-PCR=reverse transcriptase polymerase chain reaction; ELISA=enzyme linked immunosorbent assay; LFIA=lateral flow immunoassay; CLIA=chemiluminescent immunoassay; Ig=immunoglobulin; SARS-CoV-2=severe acute respiratory syndrome coronavirus 2; covid-19=coronavirus disease 2019.

Index test: serology tests to detect immunoglobulins to SARS-CoV-2. Target condition: covid-19, reference standard: RT-PCR, Studies: predominantly case-control design diagnostic test accuracy.