The new england journal of medicine Original Article
Antibody Status and Incidence of SARS-CoV-2 Infection in Health Care Workers
S.F. Lumley, D. O’Donnell, N.E. Stoesser, P.C. Matthews, A. Howarth, S.B. Hatch, B.D. Marsden, S. Cox, T. James, F. Warren, L.J. Peck, T.G. Ritter, Z. de Toledo, L. Warren, D. Axten, R.J. Cornall, E.Y. Jones, D.I. Stuart, G. Screaton, D. Ebner, S. Hoosdally, M. Chand, D.W. Crook, A.-M. O’Donnell, C.P. Conlon,
K.B. Pouwels, A.S. Walker, T.E.A. Peto, S. Hopkins, T.M. Walker, K. Jeffery, and D.W. Eyre, for the Oxford University Hospitals Staff Testing Group*
ABSTRACT
BACKGROUND
The relationship between the presence of antibodies to severe acute respiratory syn- drome coronavirus 2 (SARS-CoV-2) and the risk of subsequent reinfection remains unclear.
METHODS
We investigated the incidence of SARS-CoV-2 infection confirmed by polymerase chain reaction (PCR) in seropositive and seronegative health care workers attend- ing testing of asymptomatic and symptomatic staff at Oxford University Hospitals in the United Kingdom. Baseline antibody status was determined by anti-spike (primary analysis) and anti-nucleocapsid IgG assays, and staff members were fol- lowed for up to 31 weeks. We estimated the relative incidence of PCR-positive test results and new symptomatic infection according to antibody status, adjusting for age, participant-reported gender, and changes in incidence over time.
RESULTS
A total of 12,541 health care workers participated and had anti-spike IgG mea- sured; 11,364 were followed up after negative antibody results and 1265 after positive results, including 88 in whom seroconversion occurred during follow-up. A total of 223 anti-spike–seronegative health care workers had a positive PCR test (1.09 per 10,000 days at risk), 100 during screening while they were asymptomatic and 123 while symptomatic, whereas 2 anti-spike–seropositive health care workers had a positive PCR test (0.13 per 10,000 days at risk), and both workers were asymp- tomatic when tested (adjusted incidence rate ratio, 0.11; 95% confidence interval, 0.03 to 0.44; P=0.002). There were no symptomatic infections in workers with anti-spike antibodies. Rate ratios were similar when the anti-nucleocapsid IgG assay was used alone or in combination with the anti-spike IgG assay to determine base- line status.
CONCLUSIONS
The presence of anti-spike or anti-nucleocapsid IgG antibodies was associated with a substantially reduced risk of SARS-CoV-2 reinfection in the ensuing 6 months. (Funded by the U.K. Government Department of Health and Social Care and others.)
n engl j med nejm.org
The authors’ full names, academic de- grees, and affiliations are listed in the Ap- pendix. Address reprint requests to Dr. Eyre at the Microbiology Department, John Radcliffe Hospital, Headley Way, Oxford OX3 9DU, United Kingdom, or at david.eyre@bdi.ox.ac.uk.
*A complete list of members of the Ox- ford University Hospitals Staff Testing Group is provided in the Supplemen- tary Appendix, available at NEJM.org.
This article was published on December 23, 2020, at NEJM.org.
DOI: 10.1056/NEJMoa2034545
Copyright © 2020 Massachusetts Medical Society.
The New England Journal of Medicine
Downloaded from nejm.org on December 24, 2020. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved.
1

2
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection pro- duces detectable immune responses in most cases reported to date; however, the extent to which previously infected people are protected from a second infection is uncertain. Understand- ing whether postinfection immunity exists, how long it lasts, and the degree to which it may pre- vent symptomatic reinfection or reduce its sever- ity has major implications for the SARS-CoV-2 pandemic.
Postinfection immunity may be conferred by humoral and cell-mediated immune responses. Key considerations when investigating postinfec- tion immunity include identifying functional cor- relates of protection, identifying measurable sur- rogate markers, and defining end points, such as prevention of disease, hospitalization, death, or onward transmission.1
The assay-dependent antibody dynamics of SARS-CoV-2 anti-spike and anti-nucleocapsid anti- bodies are being defined.2-6 Neutralizing antibod- ies against the spike protein receptor-binding domain may provide some postinfection immu- nity. However, the association between antibody titers and plasma neutralizing activity is assay- and time-dependent.7-10
Evidence for postinfection immunity is emerg- ing. Despite more than 76 million people infected worldwide and widespread ongoing transmission, reported reinfections with SARS-CoV-2 have been rare, occurring mostly after mild or asymptom- atic primary infection,11-20 which suggests that SARS-CoV-2 infection provides some immunity against reinfection in most people. In addition, small-scale reports suggest that neutralizing anti- bodies may be associated with protection against infection.21 We performed a prospective longitu- dinal cohort study of health care workers to as- sess the relative incidence of subsequent positive SARS-CoV-2 polymerase-chain-reaction (PCR) tests and symptomatic infections in health care work- ers who were seropositive for SARS-CoV-2 anti- bodies and in those who were seronegative.
Methods
Cohort
Oxford University Hospitals offer SARS-CoV-2 test- ing to all symptomatic and asymptomatic staff working at four teaching hospitals in Oxfordshire, United Kingdom. SARS-CoV-2 PCR testing of com-
n engl j med
bined nasal and oropharyngeal swab specimens for symptomatic staff (those with new persis- tent cough, temperature ≥37.8°C, or anosmia or ageusia) was offered beginning on March 27, 2020. Asymptomatic health care workers were invited to participate in voluntary nasal and oropharyn- geal swab PCR testing every 2 weeks and sero- logic testing every 2 months (with some partici- pating more frequently for related studies) beginning on April 23, 2020, as previously de- scribed.5,22 Staff were followed until November 30, 2020. Deidentified data were obtained from the Infections in Oxfordshire Research Database, which has generic research ethics committee, Health Research Authority, and Confidentiality Advisory Group approvals.
Laboratory Assays
Serologic investigations were performed with use of an anti-trimeric spike IgG enzyme-linked im- munosorbent assay (ELISA), developed by the Uni- versity of Oxford,23,24 and an anti-nucleocapsid IgG assay (Abbott). See the Supplementary Appendix, available with the full text of this article at NEJM .org, for details on the assays and PCR tests.
Statistical Analysis
We classified health care workers according to their baseline antibody status. Those with only negative antibody assays were considered to be at risk for infection from their first antibody as- say until either the end of the study or their first PCR-positive test, whichever occurred earlier. Those with a positive antibody assay were con- sidered to be at risk for infection (or reinfection) from 60 days after their first positive antibody result to either the end of the study or their next PCR-positive test, whichever occurred earlier, ir- respective of subsequent seroreversion (i.e., any negative antibody assay occurring later). The 60-day window was prespecified to exclude per- sistence of PCR-positive RNA after the index infection that led to seroconversion, on the basis of earlier reports of RNA persistence for 6 weeks or more.22,25,26 Similarly, we considered only PCR- positive tests occurring at least 60 days after the previous PCR-positive test.
We used Poisson regression to model the in- cidence of PCR-positive infection per at-risk day according to baseline antibody status, adjusting for incidence over time, age, and participant-reported gender. Primary analyses used anti-spike IgG as-
nejm.org
The new england journal of medicine
The New England Journal of Medicine
Downloaded from nejm.org on December 24, 2020. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved.

Antibody Status and SARS-CoV-2 in Health Care Workers
say results, which were expected before the start of the study to be more closely related to neutral- izing activity and protection from infection.7,10 We also investigated anti-nucleocapsid antibody as- say results and a combined model with three base- line antibody statuses (both assays negative, both positive, or only one positive). Sensitivity analyses investigated the effect of different asymptomatic testing rates according to antibody status and dif- ferent follow-up windows (see the Supplementary Appendix).
Results
Baseline Anti-Spike IgG Assays and PCR Testing Rates
A total of 12,541 health care workers underwent measurement of baseline anti-spike antibodies; 11,364 (90.6%) were seronegative and 1177 (9.4%) seropositive at their first anti-spike IgG assay, and seroconversion occurred in 88 workers dur- ing the study (Table 1, and Fig. S1A in the Sup- plementary Appendix). Of 1265 seropositive health care workers, 864 (68%) recalled having had symp- toms consistent with those of coronavirus disease 2019 (Covid-19), including symptoms that pre- ceded the widespread availability of PCR testing for SARS-CoV-2; 466 (37%) had had a previous PCR-confirmed SARS-CoV-2 infection, of which 262 were symptomatic. Fewer seronegative health care workers (2860 [25% of the 11,364 who were seronegative]) reported prebaseline symptoms, and 24 (all symptomatic, 0.2%) were previously PCR-positive. The median age of seronegative and seropositive health care workers was 38 years (interquartile range, 29 to 49). Health care work- ers were followed for a median of 200 days (inter- quartile range, 180 to 207) after a negative anti- body test and for 139 days at risk (interquartile range, 117 to 147) after a positive antibody test.
Rates of symptomatic PCR testing were simi- lar in seronegative and seropositive health care workers: 8.7 and 8.0 tests per 10,000 days at risk, respectively (rate ratio, 0.92; 95% confidence in- terval [CI], 0.77 to 1.10). A total of 8850 health care workers had at least one postbaseline asymp- tomatic screening test; seronegative health care workers attended asymptomatic screening more frequently than seropositive health care workers (141 vs. 108 per 10,000 days at risk, respectively; rate ratio, 0.76; 95% CI, 0.73 to 0.80).
n engl j med
Incidence of PCR-Positive Results According to Baseline Anti-Spike IgG Status
Positive baseline anti-spike antibody assays were associated with lower rates of PCR-positive tests. Of 11,364 health care workers with a negative anti-spike IgG assay, 223 had a positive PCR test (1.09 per 10,000 days at risk), 100 during asymp- tomatic screening and 123 while symptomatic. Of 1265 health care workers with a positive anti- spike IgG assay, 2 had a positive PCR test (0.13 per 10,000 days at risk), and both workers were asymp- tomatic when tested. The incidence rate ratio for positive PCR tests in seropositive workers was 0.12 (95% CI, 0.03 to 0.47; P=0.002). The incidence of PCR-confirmed symptomatic infection in sero- negative health care workers was 0.60 per 10,000 days at risk, whereas there were no confirmed symptomatic infections in seropositive health care workers. No PCR-positive results occurred in 24 seronegative, previously PCR-positive health care workers; seroconversion occurred in 5 of these workers during follow-up.
Incidence varied by calendar time (Fig. 1), reflecting the first (March through April) and second (October and November) waves of the pandemic in the United Kingdom, and was con- sistently higher in seronegative health care work- ers. After adjustment for age, gender, and month of testing (Table S1) or calendar time as a con- tinuous variable (Fig. S2), the incidence rate ratio in seropositive workers was 0.11 (95% CI, 0.03 to 0.44; P = 0.002). Results were similar in analyses in which follow-up of both seronegative and sero- positive workers began 60 days after baseline serologic assay; with a 90-day window after positive serologic assay or PCR testing; and after random removal of PCR results for seronegative health care workers to match asymptomatic test- ing rates in seropositive health care workers (Tables S2 through S4). The incidence of positive PCR tests was inversely associated with anti-spike antibody titers, including titers below the posi- tive threshold (P<0.001 for trend) (Fig. S3A).
Anti-Nucleocapsid IgG Status
With anti-nucleocapsid IgG used as a marker for prior infection in 12,666 health care workers (Fig. S1B and Table S5), 226 of 11,543 (1.10 per 10,000 days at risk) seronegative health care work- ers tested PCR-positive, as compared with 2 of 1172 (0.13 per 10,000 days at risk) antibody-posi-
nejm.org 3
The New England Journal of Medicine
Downloaded from nejm.org on December 24, 2020. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved.

The new england journal of medicine
Table 1. Demographic Characteristics and SARS-CoV-2 PCR Testing for 12,541 Health Care Workers According to SARS-CoV-2 Anti-Spike IgG Status.*
Anti-Spike Seronegative at Baseline and throughout Follow-Up
Characteristic (N=11,276)
Anti-Spike Seronegative at Baseline, Converting to Seropositive† (N=88)
Anti-Spike Seropositive at Baseline (N=1177)
Age—yr
Median (IQR) Gender — no. (%)‡
Male
Race or ethnic group — no. (%)§
Asian
Chinese Role — no. (%) Physician
Medical or nursing student
Physical, occupational or speech therapist Security, estates, or catering staff
Symptoms resembling Covid-19 between February 1, 2020, and baseline serologic assay — no. (%)
≥1 Positive PCR test with symptoms before baseline — no. (%)
Positive PCR during follow-up — no. Symptomatic
38 (29–49)
2900 (25.7)
1719 (15.2) 121 (1.1)
1671 (14.8) 578 (5.1) 342 (3.0) 245 (2.2) 2826 (25.1)
19 (0.2)
106
41 (28–49)
20 (23)
20 (23) 0
4 (5) 6 (7) 7 (8) 3 (3)
34 (39)¶ 5 (6)
17
38 (29–49)
339 (28.8)
287 (24.4) 9 (0.8)
184 (15.6) 36 (3.1) 37 (3.1) 23 (2.0) 810 (68.8)
239 (20.3)
0
Range
16–86
21–67
17–69
Female
8360 (74.1)
68 (77)
835 (70.9)
Other
16 (0.1)
0
3 (0.3)
White
8313 (73.7)
58 (66)
703 (59.7)
Black
425 (3.8)
4 (5)
81 (6.9)
Other
698 (6.2)
6 (7)
97 (8.2)
Nurse or health care assistant
3930 (34.9)
43 (49)
555 (47.2)
Administrative staff
1452 (12.9)
10 (11)
95 (8.1)
Laboratory staff
413 (3.7)
3 (3)
36 (3.1)
Porter or domestic worker
319 (2.8)
0
58 (4.9)
Other
2326 (20.6)
12 (14)
153 (13.0)
≥1 PCR test for symptoms before baseline — no. (%)
857 (7.6)
10 (11)
358 (30.4)
Person-days of follow-up
2,036,358
7121 (while seronegative) 5076 (while seropositive)
152,983
Total
197
26‖
2
Asymptomatic
91
9
2

• Percentages may not total 100 because of rounding. Covid-19 denotes coronavirus disease 2019, IQR interquartile range, and PCR poly- merase chain reaction.
  † Those in whom seroconversion occurred were included in the analysis twice, once while they were at risk for infection and antibody-negative and then subsequently while they were antibody-positive and at risk for reinfection.
  ‡ Gender was reported by the participants. “Other” includes transgender and nondisclosed gender; the categories were combined owing to small numbers.
  § Race and ethnic group were reported by the participants.
  ¶Twenty additional health care workers in whom seroconversion occurred reported symptoms between baseline testing and seroconversion. ‖ All PCR-positive results in workers with seroconversion occurred while they were in the seronegative follow-up group. A single health care
  worker in whom seroconversion occurred first tested PCR-positive while asymptomatic, and is recorded in the asymptomatic category, but also had a further PCR-positive result when symptomatic 8 days later.
  n engl j med nejm.org
  The New England Journal of Medicine
  Downloaded from nejm.org on December 24, 2020. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved.
  4

Antibody Status and SARS-CoV-2 in Health Care Workers
tive health care workers (incidence rate ratio ad- justed for calendar time, age, and gender, 0.11; 95% CI, 0.03 to 0.45; P=0.002) (Table S6). The incidence of PCR-positive results fell with increas- ing anti-nucleocapsid antibody titers (P<0.001 for trend) (Fig. S3B).
A total of 12,479 health care workers had both anti-spike and anti-nucleocapsid baseline results (Fig. S1C and Tables S7 and S8); 218 of 11,182 workers (1.08 per 10,000 days at risk) with both immunoassays negative had subse- quent PCR-positive tests, as compared with 1 of 1021 workers (0.07 per 10,000 days at risk) with both baseline assays positive (incidence rate ratio, 0.06; 95% CI, 0.01 to 0.46) and 2 of 344 workers (0.49 per 10,000 days at risk) with mixed antibody assay results (incidence rate ratio, 0.42; 95% CI, 0.10 to 1.69).
Seropositive Health Care Workers with PCR-Positive Results
Three seropositive health care workers subsequent- ly had PCR-positive tests for SARS-CoV-2 infection (one with anti-spike IgG only, one with anti- nucleocapsid IgG only, and one with both anti- bodies). The time between initial symptoms or seropositivity and subsequent positive PCR testing ranged from 160 to 199 days. Information on the workers’ clinical histories and on PCR and sero- logic testing results is shown in Table 2 and Figure S4.
Only the health care worker with both anti- bodies had a history of PCR-confirmed symptom- atic infection that preceded serologic testing; af- ter five negative PCR tests, this worker had one positive PCR test (low viral load: cycle number, 21 [approximate equivalent cycle threshold, 31]) at day 190 after infection while the worker was asymptomatic, with subsequent negative PCR tests 2 and 4 days later and no subsequent rise in antibody titers. If this worker’s single PCR-posi- tive result was a false positive, the incidence rate ratio for PCR positivity if anti-spike IgG–sero- positive would fall to 0.05 (95% CI, 0.01 to 0.39) and if anti-nucleocapsid IgG–seropositive would fall to 0.06 (95% CI, 0.01 to 0.40).
A fourth dual-seropositive health care worker had a PCR-positive test 231 days after the worker’s index symptomatic infection, but retesting of the worker’s sample was negative twice, which sug- gests a laboratory error in the original PCR re-
n engl j med
sult. Subsequent serologic assays showed waning anti-nucleocapsid and stable anti-spike antibodies.
Discussion
In this longitudinal cohort study, the presence of anti-spike antibodies was associated with a sub- stantially reduced risk of PCR-confirmed SARS- CoV-2 infection over 31 weeks of follow-up. No symptomatic infections and only two PCR-positive results in asymptomatic health care workers were seen in those with anti-spike antibodies, which suggests that previous infection resulting in anti- bodies to SARS-CoV-2 is associated with protec- tion from reinfection for most people for at least 6 months. Evidence of postinfection immunity was also seen when anti-nucleocapsid IgG or the combination of anti-nucleocapsid and anti-spike IgG was used as a marker of previous infection.
The incidence of SARS-CoV-2 infection was inversely associated with baseline anti-spike and anti-nucleocapsid antibody titers, including titers below the positive threshold for both assays, such
nejm.org 5
3
No. of PCR-Positive Results per 10,000 Days at Risk
1
0
Baseline Anti-Spike Antibody Status
2
Negative Positive
April–June
July Aug. 2020 2020 2020
Adjusted Incidence RR, 0.11 (95% CI, 0.03–0.44)
Sept. Oct. Nov. 2020 2020 2020
Days at Risk
Seronegative Seropositive
456,963 307,508 316,141 312,027 332,704 329,469 316 19,474 31,601 34,011 36,824 37,098
Figure 1. Observed Incidence of SARS-CoV-2–Positive PCR Results According to Baseline Anti-Spike IgG Antibody Status.
The incidence of polymerase-chain-reaction (PCR) tests that were positive for SARS-CoV-2 infection during the period from April through November 2020 is shown per 10,000 days at risk among health care workers according to their antibody status at baseline. In seronegative health care workers, 1775 PCR tests (8.7 per 10,000 days at risk) were undertaken in symptom- atic persons and 28,878 (141 per 10,000 days at risk) in asymptomatic per- sons; in seropositive health care workers, 126 (8.0 per 10,000 days at risk) were undertaken in symptomatic persons and 1704 (108 per 10,000 days at risk) in asymptomatic persons. RR denotes rate ratio.
The New England Journal of Medicine
Downloaded from nejm.org on December 24, 2020. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved.

The new england journal of medicine
that workers with high “negative” titers were rela- tively protected from infection. In addition to the 24 seronegative health care workers with a previ- ous positive PCR test, it is likely that other health care workers with baseline titers below assay thresholds, which were set to ensure high speci- ficity,23 had been previously infected with SARS- CoV-2 and had low peak postinfection titers or rising or waning responses at testing.5
Two of the three seropositive health care workers who had subsequent PCR-positive tests had discordant baseline antibody results, a find- ing that highlights the imperfect nature of anti- body assays as markers of previous infection. Neither worker had a PCR-confirmed primary SARS-CoV-2 infection. Subsequent symptomatic infection developed in one worker, and both workers had subsequent dual antibody serocon- version. It is plausible that one or both had false positive baseline antibody results (e.g., from im- munoassay interference27). The health care work- er in whom both anti-spike and anti-nucleocap- sid antibodies were detected had previously had PCR-confirmed SARS-CoV-2 infection; the sub- sequent PCR-positive result with a low viral load was not confirmed on repeat testing and was not associated with a change in IgG response. These results could be consistent with a reexposure to SARS-CoV-2 that did not lead to symptoms but could also plausibly have arisen from undetected laboratory error; although contemporaneous re- testing of the PCR-positive sample was not un- dertaken, samples tested 2 and 4 days later were both negative. If the PCR-positive result is incor- rect, the incidence rate ratio for PCR positivity if anti-spike IgG–seropositive would fall to 0.05. We detected and did not include in our analysis a presumed false positive PCR test in a fourth seropositive health care worker.
Owing to the low number of reinfections in seropositive health care workers, we cannot say whether past seroconversion or current antibody levels determine protection from infection or de- fine which characteristics are associated with reinfection. Similarly, we cannot say whether protection is conferred through the antibodies we measured or through T-cell immunity, which we did not assess. It was not possible to use se- quencing to compare primary and subsequent infections, since only one of the three seropositive health care workers with a subsequent PCR-posi- tive test had PCR-confirmed primary infection
6 n engl j med
nejm.org
The New England Journal of Medicine
Downloaded from nejm.org on December 24, 2020. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved.
Table 2. Demographic, Clinical, and Laboratory Characteristics of Health Care Workers with Possible SARS-CoV-2 Reinfection.
between Timing of PCR, Ct Value, Follow-up Serologic No. of Days Health Care Worker Baseline Serologic Assay Episodes* Clinical Characteristics and Assay Assay
in anti-nucleocapsid Dual antibody serocon- version with a rise IgG titer
1st episode: Ct 36.0 (PHE assay) No rise in antibody 2nd episode: CN 21.2 (Abbott titers and day 4 both negative assay), repeat PCR on day 2
(mild, Covid-19–like 2nd episode: asymptomatic 190 1st episode: symptomatic symptoms)
Anti-nucleocapsid IgG: detected Anti-spike IgG: detected
White female nurse, 55–59 yr of age
Worker 2:
version, with a rise Dual antibody serocon- in anti-spike IgG titer
start of asymptomatic testing) PCR on same sample Ct 19.0 assay), repeat extraction and 1st episode: not done (before 2nd episode: CN 10.6 (Abbott, (Thermo Fisher assay)
assay), repeat PCR on day 2 Ct 24.0 (Altona assay) 1st episode: PCR-negative 2nd episode: CN 12.6 (Abbott
(mild, febrile illness) 1st episode: asymptomatic 2nd episode: symptomatic
shortly after influenza vaccine 1st episode: symptomatic (mild) 2nd episode: asymptomatic when tested (transient myalgia 1 week earlier)
160
199
Anti-spike IgG: not detected
Anti-nucleocapsid IgG: detected
Anti-spike IgG: detected Anti-nucleocapsid IgG: not detected
White female physician, 25–29 yr of age Worker 1:
White female administrator
with patient contact,
50–54 yr of age
Worker 3:

• The number of days between episodes was calculated from the date of symptom onset if the index infection was symptomatic (as it was for health care Workers 2 and 3) or the date of the first clinic attendance if the presumed first episode was asymptomatic with no PCR performed (as it was for Worker 1). Both baseline serologic assays for Worker 1 were repeated and confirmed. A single positive PCR result for Worker 1 was confirmed by repeat nucleic acid extraction. Abbott PCR assay cycle number (CN) values are approximately equivalent to cycle threshold (Ct) values 10 units higher (e.g., CN 21 is approximately equivalent to Ct 31). See Figure S4 for quantitative antibody results. All PCR tests listed were performed at Oxford University Hospitals.

Antibody Status and SARS-CoV-2 in Health Care Workers
and that worker’s original sample was not stored. Our study was relatively short, with up to 31 weeks of follow-up. Ongoing follow-up is needed in this and other cohorts, including the use of markers of both humoral and cellular immunity to SARS-CoV-2, to assess the magnitude and du- ration of protection from reinfection, symptom- atic disease, and hospitalization or death and the effect of protection on transmission.
Health care workers were enrolled in a volun- tary testing program with a flexible follow-up schedule, which led to different attendance fre- quencies. Although health care workers were offered asymptomatic PCR testing every 2 weeks, the workers attended less frequently than that (mean, once every 10 to 13 weeks). Therefore, asymptomatic infection is likely to have been underascertained. In addition, as staff were told their antibody results, “outcome ascertainment bias” occurred, with seropositive staff attending asymptomatic screening less frequently. However, a sensitivity analysis suggests that the differing attendance rates did not substantially alter our findings. Staff were told to follow guidance on social distancing and use of personal protective equipment and to attend testing if Covid-19 symp- toms developed, even if the worker had been previously PCR- or antibody-positive. This is re- flected in the similar rates of testing of symptom- atic seropositive and seronegative health care workers.
Some health care workers were lost to follow- up after terminating employment at our hospi- tals; this was likely to have occurred at similar rates in seropositive and seronegative staff. Not all PCR-positive results from government symp- tomatic testing sites were communicated to the hospital. This is a study of predominantly healthy adult health care workers 65 years of age or young- er; further studies are needed to assess postinfec- tion immunity in other populations, including
children, older adults, and persons with coexist- ing conditions, including immunosuppression.
In this study, we found a substantially lower risk of reinfection with SARS-CoV-2 in the short term among health care workers with anti-spike antibodies and those with anti-nucleocapsid anti- bodies than among those who were seronegative.
The views expressed in this article are those of the authors and not necessarily those of the National Health Service, the Na- tional Institute for Health Research, the Department of Health, or Public Health England.
Supported by the U.K. Government Department of Health and Social Care; the National Institute for Health Research (NIHR) Health Protection Research Unit in Healthcare Associated Infections and Antimicrobial Resistance, Oxford University, in partnership with Public Health England (NIHR200915); the NIHR Biomedical Research Centre (BRC), Oxford; and benefac- tions from the Huo Family Foundation and Andrew Spokes. This study is affiliated with Public Health England’s SARS-CoV-2 Immunity and Reinfection Evaluation (SIREN) study. Dr. Eyre is a Robertson Foundation Fellow and an NIHR Oxford BRC Senior Fellow; Dr. Lumley is a Wellcome Trust Clinical Research Fellow; Prof. Stuart’s work is supported by the Medical Research Council (MR/N00065X/1); Dr. Matthews holds a Wellcome Intermediate Fellowship (110110/Z/15/Z); Dr. Marsden’s work is supported by the Kennedy Trust for Rheumatology Research and by the SGC, a registered charity (no. 1097737) that receives funds from AbbVie, Bayer Pharma, Boehringer Ingelheim, Canada Foundation for Innovation, Eshelman Institute for In- novation, Genome Canada through Ontario Genomics Institute (OGI-055), Innovative Medicines Initiative (EU/EFPIA) (ULTRA- DD grant no. 115766), Janssen, Merck, Darmstadt, Germany, MSD, Novartis Pharma, Pfizer, São Paulo Research Foundation (FAPESP), Takeda, and Wellcome. Prof. Screaton is a Wellcome Trust Senior Investigator with funding from the Schmidt Foun- dation; Dr. Timothy Walker is a Wellcome Trust Clinical Career Development Fellow (214560/Z/18/Z); and Prof. A. Sarah Walker is an NIHR Senior Investigator.
Disclosure forms provided by the authors are available with the full text of this article at NEJM.org.
We thank all Oxford University Hospitals personnel who par- ticipated in the staff testing program and the staff and medical students who ran the program. This study uses data provided by health care workers and collected by the U.K. National Health Service as part of their care and support. We thank additional members of the Infections in Oxfordshire Research Database team: L. Butcher, H. Boseley, C. Crichton, O. Freeman, J. Gear- ing (community), R. Harrington, M. Landray, A. Pal, T.P. Quan, J. Robinson (community), J. Sellors, B. Shine, and D. Waller; and the patient and public panel: G. Blower, C. Mancey, P. McLough- lin, and B. Nichols.
Appendix
The authors’ full names and academic degrees are as follows: Sheila F. Lumley, B.M., B.Ch., Denise O’Donnell, B.Sc., Nicole E. Stoesser, M.B., B.S., D.Phil., Philippa C. Matthews, F.R.C.P., D.Phil., Alison Howarth, Ph.D., Stephanie B. Hatch, Ph.D., Brian D. Marsden, D.Phil., Stuart Cox, Tim James, Ph.D., Fiona Warren, B.Sc., Liam J. Peck, D.Phil., Thomas G. Ritter, B.A., Zoe de Toledo, M.Sc., Laura Warren, David Axten, B.A., Richard J. Cornall, M.D., D.Phil., E. Yvonne Jones, D.Phil., David I. Stuart, Ph.D., Gavin Screaton, B.M., B.Ch., D.Phil., Daniel Ebner, B.Sc., Sarah Hoosdally, Ph.D., Meera Chand, M.B., B.S., Derrick W. Crook, F.R.C.Path., Anne-Marie O’Donnell, M.B., B.S., Christopher P. Conlon, M.D., Koen B. Pouwels, Ph.D., A. Sarah Walker, Ph.D., Tim E.A. Peto, F.R.C.P., Susan Hopkins, F.R.C.P., Timothy M. Walker, M.B., B.S., D.Phil., Katie Jeffery, F.R.C.Path., Ph.D., and David W. Eyre, B.M., B.Ch., D.Phil.
The authors’ affiliations are as follows: Oxford University Hospitals NHS Foundation Trust (S.F.L., N.E.S., P.C.M., S.C., T.J., F.W., L.W., D.A., A.-M.O., K.J.), Nuffield Department of Medicine (S.F.L., D.O., N.E.S., P.C.M., A.H., S.B.H., B.D.M., R.J.C., E.Y.J., D.I.S., G.S., D.E., S. Hoosdally, D.W.C., C.P.C., A.S.W., T.E.A.P., T.M.W.), the National Institute for Health Research (NIHR) Oxford Bio-
n engl j med nejm.org 7
The New England Journal of Medicine
Downloaded from nejm.org on December 24, 2020. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved.

8
n engl j med nejm.org
References

1. Plotkin SA. Vaccines: correlates of vaccine-induced immunity. Clin Infect Dis 2008;47:401-9.
2. Seow J, Graham C, Merrick B, et al. Longitudinal observation and decline of neutralizing antibody responses in the three months following SARS-CoV-2 in- fection in humans. Nat Microbiol 2020;5: 1598-607.
3. Robbiani DF, Gaebler C, Muecksch F, et al. Convergent antibody responses to SARS-CoV-2 in convalescent individuals. Nature 2020;584:437-42.
4. Gudbjartsson DF, Norddahl GL, Mel- sted P, et al. Humoral immune response to SARS-CoV-2 in Iceland. N Engl J Med 2020;383:1724-34.
5. Lumley SF, Wei J, O’Donnell D, et al. The duration, dynamics and determi- nants of SARS-CoV-2 antibody responses in individual healthcare workers. Novem- ber 4, 2020 (https://www.medrxiv.org/ content/10.1101/2020.11.02.20224824v1). preprint.
6. Long Q-X, Tang X-J, Shi Q-L, et al. Clinical and immunological assessment of asymptomatic SARS-CoV-2 infections. Nat Med 2020;26:1200-4.
7. Muecksch F, Wise H, Batchelor B, et al. Longitudinal analysis of serology and neutralizing antibody levels in COVID19 convalescents. J Infect Dis 2020 Novem- ber 3 (Epub ahead of print).
8. To KK-W, Tsang OT-Y, Leung W-S, et al. Temporal profiles of viral load in pos- terior oropharyngeal saliva samples and serum antibody responses during infec- tion by SARS-CoV-2: an observational co- hort study. Lancet Infect Dis 2020;20:565- 74.
9. GeurtsvanKessel CH, Okba NMA, Igloi Z, et al. An evaluation of COVID-19 serological assays informs future diag- nostics and exposure assessment. Nat Commun 2020;11:3436.
10. Wajnberg A, Amanat F, Firpo A, et al. Robust neutralizing antibodies to SARS-
    CoV-2 infection persist for months. Sci- ence 2020;370:1227-30.
11. Tillett RL, Sevinsky JR, Hartley PD, et al. Genomic evidence for reinfection with SARS-CoV-2: a case study. Lancet Infect Dis 2020 October 12 (Epub ahead of print).
12. Van Elslande J, Vermeersch P, Vander- voort K, et al. Symptomatic SARS-CoV-2 reinfection by a phylogenetically distinct strain. Clin Infect Dis 2020 September 5 (Epub ahead of print).
13. Prado-Vivar B, Becerra-Wong B, Gua- dalupe JJ, et al. COVID-19 re-infection by a phylogenetically distinct SARS-CoV-2 variant, first confirmed event in South America. 2020 (https://papers.ssrn.com/ sol3/papers.cfm?abstract_id=3686174). 14. To KK-W, Hung IF-N, Ip JD, et al. COVID-19 re-infection by a phylogeneti- cally distinct SARS-coronavirus-2 strain confirmed by whole genome sequencing. Clin Infect Dis 2020 August 25 (Epub ahead of print).
14. Gupta V, Bhoyar RC, Jain A, et al. Asymptomatic reinfection in two health- care workers from India with genetically distinct SARS-CoV-2. Clin Infect Dis 2020 September 23 (Epub ahead of print).
15. Walker AS, Pritchard E, House T, et al. Viral load in community SARS-CoV-2 cas- es varies widely and temporally October 27, 2020 (https://www.medrxiv.org/content/ 10.1101/2020.10.25.20219048v1). preprint. 17. Mumoli N, Vitale J, Mazzone A. Clini- cal immunity in discharged medical pa- tients with COVID-19. Int J Infect Dis 2020;99:229-30.
16. Zhang K, Yiu-Nam Lau J, Yang L, Ma Z. SARS-CoV-2 reinfection in two patients who have recovered from COVID-19. Prec Clin Med 2020 (https://academic.oup.com/ pcm/advance-article/doi/10.1093/pcmedi/ pbaa031/5901533).
17. Gousseff M, Penot P, Gallay L, et al. Clinical recurrences of COVID-19 symp- toms after recovery: viral relapse, reinfec-
    tion or inflammatory rebound? J Infect 2020;81:816-46.
18. European Centre for Disease Preven- tion and Control. Reinfection with SARS- CoV-2: considerations for public health response. September 21, 2020 (https:// http://www.ecdc.europa.eu/sites/default/files/ documents/Re-infection-and-viral -shedding-threat-assessment-brief.pdf ). 21. Addetia A, Crawford KHD, Dingens A, et al. Neutralizing antibodies correlate with protection from SARS-CoV-2 in hu- mans during a fishery vessel outbreak with a high attack rate. J Clin Microbiol 2020;58(11):e02107-20.
19. Eyre DW, Lumley SF, O’Donnell D, et al. Differential occupational risks to healthcare workers from SARS-CoV-2 ob- served during a prospective observational study. Elife 2020;9:e60675.
20. National SARS-CoV-2 Serology Assay Evaluation Group. Performance charac- teristics of five immunoassays for SARS- CoV-2: a head-to-head benchmark com- parison. Lancet Infect Dis 2020;20: 1390-400.
21. Adams ER, Ainsworth M, Anand R, et al. Antibody testing for COVID-19: a report from the National COVID Scien- tific Advisory Panel. Wellcome Open Re- search. June 11, 2020 (https://wellcome openresearch.org/articles/5-139/v1).
22. Cento V, Colagrossi L, Nava A, et al. Persistent positivity and fluctuations of SARS-CoV-2 RNA in clinically-recovered COVID-19 patients. J Infect 2020;81(3): e90-e92.
23. Chu CM, Leung WS, Cheng VCC, et al. Duration of RT-PCR positivity in severe acute respiratory syndrome. Eur Respir J 2005;25:12-4.
24. Ward G, Simpson A, Boscato L, Hick- man PE. The investigation of interferenc- es in immunoassay. Clin Biochem 2017; 50:1306-11.
    Copyright © 2020 Massachusetts Medical Society.
    The new england journal of medicine
    medical Research Centre (N.E.S., P.C.M., S. Hoosdally, D.W.C., A.S.W., T.E.A.P., D.W.E.), the Kennedy Institute of Rheumatology Research (B.D.M.), the Medical School, University of Oxford (L.J.P., T.G.R., Z.T.), Target Discovery Institute (D.E.), Nuffield Depart- ment of Population Health (A.-M.O., K.B.P., D.W.E.), and the Big Data Institute (D.W.E.), University of Oxford, and the NIHR Health Protection Research Unit in Healthcare Associated Infections and Antimicrobial Resistance at University of Oxford in partnership with Public Health England (N.E.S., P.C.M., S. Hoosdally, D.W.C., K.B.P., A.S.W., T.E.A.P., D.W.E.), Oxford, and the National Infection Service, Public Health England at Colindale, London (M.C., S. Hopkins) — all in the United Kingdom; and the Oxford University Clinical Research Unit, Ho Chi Minh City, Vietnam (T.M.W.).
    The New England Journal of Medicine
    Downloaded from nejm.org on December 24, 2020. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved.