Summary of findings 1. What is the diagnostic accuracy of antibody tests, for the diagnosis of current or prior SARS‐CoV‐2 infection?

Question	What is the diagnostic accuracy of antibody tests, for the diagnosis of current or prior SARS‐CoV‐2 infection?
Population	Adults or children suspected of current SARS‐CoV‐2 infection or who may have had prior SARS‐CoV‐2 infection, including populations undergoing screening for SARS‐CoV‐2 such as asymptomatic contacts of confirmed COVID‐19 cases or community‐based testing
Index test	Any commercially produced test for detecting antibodies to SARS‐CoV‐2, including: laboratory‐based methods enzyme‐linked immunosorbent assays (ELISA) chemiluminescent immunoassays (CLIA) other laboratory‐based methods, rapid tests using a lateral flow format that can be used at the point‐of‐care, including colloidal‐gold based immunoassays (CGIA) fluorescent immunoassays (FIA) alternative formats
Target condition	Detection of: current SARS‐CoV‐2 infection prior SARS‐CoV‐2 infection
Reference standard	Presence of current infection: RT‐PCR alone or combined with clinical diagnosis of COVID‐19 based on established guidelines or combinations of clinical features for RT‐PCR‐negative Presence of prior infection: RT‐PCR alone Absence of infection: pre‐pandemic sources of samples for testing, RT‐PCR‐negative samples from COVID‐suspects, from healthy participants or those with pre‐existing disease
Action	The primary use case for antibody tests is for identification of those with previous infection with SARS‐CoV‐2 (e.g. for seroprevalence purposes or research). Although studies included in this review were conducted prior to the introduction of SARS‐CoV‐2 vaccination programmes, antibody tests used for seroprevalence purposes will also identify those with vaccination‐induced antibody responses. This review was not able to consider whether antibody test accuracy is the same for detecting antibodies resulting from vaccination. The sensitivity of antibody tests is too low early in disease for use as a primary test of diagnosis, but they may have some diagnostic utility two to three weeks after onset of infection, particularly in those who are RT‐PCR‐negative.
Limitations in the evidence
Risk of bias	Participant selection: high risk of bias in 172 studies (99%), primarily because of selection for inclusion based on known disease status (i.e. separate recruitment of confirmed SARS‐CoV‐2 cases and non‐cases) Index test: high risk of bias in 35 studies (22%) because blinded index test interpretation was not implemented or the threshold to define test positivity was determined by analysing the data rather than prespecified Reference standard: high risk of bias in 39 studies (22%) because of inadequate reference standards for confirming absence of infection, e.g. reliance on a single negative RT‐PCR result in people with suspected COVID‐19, or no RT‐PCR testing reported in contemporaneous healthy or other disease non‐COVID‐19 groups, or because serology results in part determined the presence of infection Flow and timing: high risk of bias in 146 studies (84%) because of different reference standards used to verify presence or absence of infection, some participants with no reference standard, exclusions from analyses, and inclusion of multiple samples per participant
Concerns about applicability of the evidence	Participants: high concerns in 171 studies (98%) because participants were unlikely to be similar to those in whom the test would be used in clinical practice, e.g. hospitalised confirmed cases of COVID‐19 or healthy or other disease non‐SARS‐CoV‐2 groups Index test: no studies rated as high concerns for applicability Reference standard: high concerns in 162 studies (93%), primarily because cases were defined based only on RT‐PCR‐positive results and did not consider clinically defined cases
Findings
We included 178 studies evaluating 64,688 samples. 25,724 samples were from people with SARS‐CoV‐2. Seventy‐seven studies (43%) evaluated a single‐test brand and 103 compared the accuracy of two or more assays, for a total of 527 index test evaluations (counting each test brand once per study). These studies included data on 124 commercial antibody assays. SARS‐CoV‐2 cases were mainly hospital inpatients (44%) with small numbers from community settings (8%), hospital outpatients (3%), or emergency departments (3%). Almost half of studies recruited cases from multiple settings (20%) or did not clearly report the source of participants (22%). All studies were conducted prior to the availability of vaccines against SARS‐CoV‐2 and therefore represent antibody response after naturally acquired SARS‐CoV‐2 infection. Most studies reported data for assays targeting IgG alone, IgM alone, the combination of IgG or IgM antibodies, or total antibodies (including IgA). Test evaluations included ELISA assays (31%), CLIA assays (32%) and lateral flow assays (36%). Many studies only applied tests in laboratory settings on plasma or serum. Nearly all studies sampled cases with and without SARS‐CoV‐2 infection separately, and methods for selecting participants were not described. The strength of the relationship of sensitivity with time shows exceptionally high levels of statistical significance (P < 0.0001), with sensitivity reaching its highest value (> 90%) for all target antibodies apart from IgM in the convalescent phase of infection, or week four onwards. Sensitivity for assays targeting IgM alone was highest (at 78%) in week three (15 to 21 days after onset). Pre‐pandemic samples provided the least biased estimate of assay specificity; average specificities were 98.6% or more for all target antibodies. Results according to type of antigen used in the test (nucleocapsid, spike, or both) were variable but suggest any differences in sensitivity by antigen type, especially for IgG, are limited to the first week or two after onset. Some differences in average sensitivities were observed by test technology (marginally higher for CLIA methods), however, heterogeneity in study results, timing of sample collection, and smaller sample numbers in some groups complicates interpretation. Investigations of test performance by brand showed considerable variation in sensitivity between tests, and variability in results between studies evaluating the same test. None of the test brands in our review fully met UK MHRA target performance criteria for sensitivity or specificity. Data for IgA as target antibody are based on smaller numbers of samples but suggest a similar pattern as for other antibodies, with average sensitivity for IgA alone exceeding 80% from week 3 onwards. For asymptomatic participants, a similar effect from time after diagnosis was observed, with lower sensitivity for IgG assays within two weeks of a positive RT‐PCR result, increasing by 14 or more days after positive PCR.
Quantity of evidence	Number of studies	Total participants or samplesa				Total cases
	178	64,688				25,724
		Sensitivity (95% CI) N evaluations (TP/SARS‐CoV‐2 cases)				Specificity (95%CI) N evaluations (TN/non‐SARS‐CoV‐2 cases)
		Week 1	Week 2	Week 3	Convalescent	Pre‐pandemic
Assays targeting IgG alone		27.2 (24.9, 29.7)	64.8 (62.1, 67.4)	88.1 (86.6, 89.5)	89.8** (88.5, 90.9)	98.9** (98.6, 99.1)
		189 (2177/6679)	202 (5883/9078)	190 (4328/5027)	253 (14,183/16,846)	179 (37,385/38,090)
Assays targeting IgM alone		29.5 (25.8, 33.6)	64.6 (60.3, 68.7)	78.3 (74.8, 81.4)	71.2 (65.5, 76.2)	98.6 (98.0, 99.1)
		126 (1770/4492)	122 (3715/5577)	118 (2416/3231)	125 (4683/7124)	83 (14,691/15,126)
Assays targeting either IgG or IgMb		41.1 (38.1, 44.2)	74.9 (72.4, 77.3)	88.0* (86.3, 89.5)	92.9 (91.0, 94.4)	99.2* (98.5, 99.5)
		103 (1593/3881)	96 (2904/3948)	103 (2571/2929)	108 (3206/3571)	68 (8989/9262)
Assays targeting total antibodies		37.7 (31.0, 44.9)	79.4 (74.0, 83.9)	90.9 (87.8, 93.2)	94.3 (92.8, 95.5)	99.8 (99.6, 99.9)
		27 (428/1010)	29 (804/1030)	33 (908/1016)	58 (6652/7063)	45 (12,166/12,207)
Antibody tests for diagnosis of current infection: Numbers applied to a hypothetical cohort of 1000 people, using summary data for the combination of IgG or IgM in week 3 after onset of infection for sensitivity and pre‐pandemic samples (denoted using * above)
Prevalence of current infection	TP (95% CI)	FP (95% CI)	FN (95% CI)	TN (95% CI)	PPV (%)	1‐NPV (%)
1%	9 (9, 9)	8 (5, 15)	1 (1, 1)	982 (975, 985)	53	0.1
2%	18 (17, 18)	8 (5, 15)	2 (2, 3)	972 (965, 975)	69	0.2
5%	46 (46, 47)	8 (5, 14)	4 (3, 5)	942 (936, 945)	85	0.6
Antibody tests for diagnosis of prior infection: Numbers applied to a hypothetical cohort of 1000 people, using summary data for IgG alone during the convalescent phase of infection for sensitivity and pre‐pandemic samples (denoted using ** above)
Prevalence of prior infection	TP (95% CI)	FP (95% CI)	FN (95% CI)	TN (95% CI)	PPV (%)	1‐NPV (%)
20%	180 (177, 182)	9 (7, 11)	20 (18, 23)	791 (789, 793)	95	2.5
50%	449 (443, 455)	6 (5, 7)	51 (46, 58)	494 (493, 496)	99	9.4
*Data applied to hypothetical cohort with current infection. ** Data applied to hypothetical cohort with prior infection.

CGIA: colloidal gold immunoassays
CI: confidence interval
CLIA: chemiluminescence immunoassays
ELISA: enzyme‐linked immunosorbent assays
FIA: fluorescence‐labelled immunochromatographic assays
FN: false negative
FP: false positive
RT‐PCR: reverse transcription polymerase chain reaction
TN: true negative
TP: true positive

^aSamples counted once per study; results per antibody and time period were counted per test evaluated (i.e. could be counted more than once per study)

^bPositive if either IgG‐ or IgM‐positive