Reinfection rates among patients previously infected by SARS-CoV-2: systematic review and meta-analysis.

BACKGROUND: Asymptomatic or symptomatic infection with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) can be followed by reinfection. The protection conferred by prior infection among coronavirus disease 2019 (COVID-19) patients is unclear. We assessed the incidence of SARS-CoV-2 reinfection and the protection effect of previous infection against reinfection.
METHODS: We searched PubMed, EMBASE, Cochrane, Scopus, Web of Science, and ClinicalTrials.gov for publications up until the end date of May 1, 2021. The reinfection rate of recovered patients and the protection against reinfection were analyzed using meta-analysis.
RESULTS: Overall, 19 studies of 1096 reinfection patients were included. The pooled reinfection rate was 0.65% (95% confidence interval [CI] 0.39-0.98%). The symptomatic reinfection rate was a bit lower (0.37% [95% CI 0.11-0.78%], I2 = 99%). The reinfection rate was much higher in high-risk populations (1.59% [95% CI 0.30-3.88%], I2 = 90%). The protection against reinfection and symptomatic reinfection was similar (87.02% [95% CI 83.22-89.96%] and 87.17% [95% CI 83.09-90.26%], respectively).
CONCLUSIONS: The rate of reinfection with SARS-CoV-2 is relatively low. The protection against SARS-CoV-2 after natural infection is comparable to that estimated for vaccine efficacy. These data may help guide public health measures and vaccination strategies in response to the COVID-19 pandemic. High-quality clinical studies are needed to establish the relevant risk factors in recovered patients.

Introduction

The emergence of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which causes coronavirus disease 2019 (COVID-19), has had profound implications, not only on human health but also on collective mental health, the economy, and the social structure of global communities.[ At the time of this writing (June 10, 2021), SARS-CoV-2 has caused >174 million cases of COVID-19, which have led to >3.7 million deaths worldwide.[ Furthermore, reinfection may occur, which is of great importance to public health.

On August 25, 2020, the first case of reinfection was reported in the medical literature; a total of 24 nucleotide differences existed between the viruses identified in the two infections;[ this was followed by the establishment of other cases of reinfection around the world.[ One of the largest studies in the UK reported 304 reinfections in 36,509 recovered patients, or a reinfection rate of 0.8%.[ These cases have aroused widespread concern. Several questions are still unclear. For instance, what is the incidence of reinfection? Are there any differences in incidence by sex or region? How long after initial infection can reinfection be expected?

Vaccines have now been licensed in various countries that show efficacies ranging from 62% to 95%.[ Because there is an urgent need for immunity from SARS-CoV-2, a more comprehensive understanding of the degree of protection provided against SARS-CoV-2 reinfection is critical for guiding the ongoing development of vaccines and the creation and implementation of appropriate interventional strategies. However, to date, evidence for protective efficacy against reinfection has been lacking. To address this gap in the research, we performed a systematic review and meta-analysis in patients previously infected with SARS-CoV-2 using a wide range of pertinent studies.

Methods

This meta-analysis was conducted and reported in accordance with Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) standards.[ The PRISMA checklist is given in [Supplementary Table 1]. The study protocol was pre-registered on International Platform of Registered Systematic Review and Meta-analysis Protocols (INPLASY) with the registration number INPLASY202160104.

Search strategy

A comprehensive search was conducted using searches on PubMed, EMBASE, Cochrane Library, Scopus, Web of Science, and ClinicalTrials.gov up to May 1, 2021. An extensive search strategy was used, intended to retrieve all relevant articles, using both Medical Subject Headings terminology and relevant keywords “coronavirus,” “COVID-19,” “reinfection,” “SARS-COV-2,” “coronavirus disease 2019,” and “severe acute respiratory” [Supplementary Table 2]. In addition, a manual search of World Health Organization reporting and references in the retrieved articles ensured the identification of studies that were not found in the initial literature search. This selection was limited to publications in English.

Study outcomes

According to the definition of the US Centers for Disease Control and Prevention (CDC),[ reinfection is defined as occurring ≥90 days after initial positive testing or ≥45 days with background information supporting contact with confirmed cases or the reappearance of COVID-19–like symptoms. The reinfection rate was determined by dividing the number of reinfected patients by the total number of initially positive patients. Furthermore, the symptomatic reinfection rate was defined by including only symptomatic reinfection patients in the numerator. A high-risk population (HRP) was defined as one that faced a higher burden of virus exposure (e.g., front-line healthcare workers, the residents and staff of care homes and skilled nursing facilities, and older patients with comorbidities), and the reinfection rate in this group was determined by dividing the number of HRP reinfection patients by the total number of initially positive HRP patients. The protection provided by previous infection was measured as one minus risk ratio with a 95% confidence interval (CI) (computed as the infection rate of the initially positive patients vs. the infection rate of the initially negative patients). The comparison between the number of initially positive infections and the number of initially negative infections for each study are shown in [Supplementary Table 3].

Study selection

The studies were chosen using the following inclusion criteria: (1) studies reporting the number of COVID-19 reinfection that met US CDC criteria and (2) original research including cohort, ecological, and cross-sectional studies. The exclusion criteria were (1) studies with patients with Middle East respiratory syndrome coronavirus or other serotypes of SARS-CoV infection and (2) reviews, commentaries, case reports, case series, and non-human studies.

Data extraction and quality assessment

Two independent reviewers (M-YJ and W-WW) extracted data from each eligible study and then cross-checked the results. Any disagreements between reviewers on data extraction were resolved through discussions involving and requiring the consensus of the third reviewer (W-SS). The following information was extracted: first author, publication year, country, study design, interval between two infections, age and sex of the initial infection and reinfection patients, reinfection severity (cycle threshold [CT] value), reinfection clinical manifestation, and hospitalization for reinfection. The Newcastle–Ottawa Scale for cohort study was used to evaluate the risk for bias in cohort studies,[ and the Joanna Briggs Institute critical appraisal tool was used to evaluate the risk for bias for cross-sectional and ecological studies.[

Statistical analyses

Because heterogeneity among the included studies was relatively large due to the various clinical and methodological perspectives in the rate study, we adopted a random-effects model,[ which was used to obtain a pooled estimate and 95% CI for reinfection rate after recovery from COVID-19, and an arcsine transformation was conducted to stabilize the variance.[ Heterogeneity was assessed using Cochran I2 and Q. The heterogeneity test was truncated at significant Cochran Q values (P < 0.1) and I2 > 50%, because an I2 of 30% to 50% has been recommended as a truncation value for moderate heterogeneity.[ A prediction interval for the proportion in a new study is calculated if the arguments prediction and “comb.random” are TRUE.[ The protection provided by previous infection was calculated using a combined Mantel–Hanzsel method with the random-effects model. We used the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) to evaluate the quality of evidence for each study outcome.[

Cumulative meta-analyses were carried out to determine whether the reinfection rate tended to stabilize as the sample size increased. Sensitivity analyses were also performed with the exclusion of each study at each time to test the reliability of the reinfection rate. In addition, subgroup analyses were conducted according to sex (male or female), study design (prospective cohort, retrospective cohort, ecological, or cross-sectional study), continent (Europe, North America, or Asia), and infection interval (≥90 days or ≥45 days).

All statistical analyses were conducted with the statistical software R version 4.0.3 (package “meta”).

Results

Search results and study characteristics

The PRISMA flow chart for the literature selection is shown in Figure 1. Ultimately, 19 of 925 studies with a total of 325,225 COVID-19 patients with initially positive infections were included in the meta-analysis.

Figure 1

PRISMA flow chart illustrating study selection process. PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-analyses.

The studies[ were published between 2020 and 2021 and included 17 cohort studies (3 prospective,[ 14 retrospective[), 1 ecological study,[ and 1 cross-sectional study.[ The characteristics of the included studies are summarized in Table 1. Reinfections occurred across three continents in our study. Among the nine studies reporting age, the reinfection patients ranged from <1 to 99 years old.[ Eight studies mentioned cases of reinfection in HRP, including 30 health care workers and five skilled nursing residents.[ Thirteen studies reported reinfections of 253 asymptomatic and 447 symptomatic patients.[ Six studies reported on 13 hospitalized patients who were reinfected.[ Among the six studies describing CT values,[ one had mean CT values of 32.9,[ and five had CT values of <30.[

Table 1

Baseline characteristics of the included trials.

					Initial positive infection		Reinfection		Clinical manifestation
Study (years)	Country	Study design	Interval between two infections, days	Reinfection/Initial positive infection	Age,years	Female (%)	Age,years	Female (%)	Symptomatic	Asymptomatic	Hospitalization for reinfection, n (%)	CT value (average)
Jeffery-Smith et al[20] (2021)	UK	Retrospective cohort	>90	1/88	NA	NA	NA	NA	0	1	NA	NA
Lumley et al[21] (2021)	UK	Prospective cohort	160–199∗	3/1265	17–69∗	77	25–59∗	100	1	2	NA	24.6
Hansen et al[22] (2021)	Denmark	Retrospective cohort	>90	72/11,068	NA	NA	NA	NA	NA	NA	NA	NA
Abu-Raddad et al[23] (2020)	Qatar	Retrospective cohort	45–129∗	54/133,266	NA	NA	16–57∗	13	23	31	1 (1.9)	28
Harvey et al[24] (2021)	USA	Retrospective cohort	>90	125/41,587	44 ± 18	54.1	NA	NA	NA	NA	NA	NA
Graham et al[8] (2021)	UK	Ecological study	>90	304/36,509	NA	NA	NA	NA	249	55	NA	NA
Soriano et al[26] (2021)	Spain	Retrospective cohort	>90	2/122	NA	NA	NA	NA	NA	NA	1 (50)§	NA
Breathnach et al[27] (2021)	UK	Retrospective cohort	>90	8/10,727	53†	60	55†	100	NA	NA	NA	NA
Zare et al[25] (2021)	Iran	Retrospective cohort	107–234∗	9/4039	64 ± 28	49.8	13–90∗	44.4	9	0	5 (55.6)	NA
Hanrath et al[28] (2021)	UK	Retrospective cohort	162–229‡	0/1038	30–49‡	82.5	NA	0	0	0	0 (0)	NA
Sheehan et al[29] (2021)	USA	Retrospective cohort	90–295∗	62/8845	52 ± 22	52.1	NA	NA	31	31	NA	NA
Qureshi et al[30] (2021)	USA	Retrospective cohort	116 ± 21	63/9119	NA	NA	NA	55.6	19	44	NA	NA
Dubelbeiss et al[31] (2021)	USA	Retrospective cohort	106–151∗	3/45	NA	100	NA	100	0	3	NA	NA
Sanchez-Montalva et al[32] (2021)	Spain	Prospective cohort	>90	3/20	26–37‡	60	NA	NA	0	3	NA	NA
Abu-Raddad et al[33] (2021)	Qatar	Retrospective cohort	>45	129/43,044	28–47‡	20.8	<1–72∗	28.7	NA	NA	NA	32.9
Hall et al[34] (2021)	UK	Prospective cohort	95–297∗	155/8278	19–78∗	82.6	20–68∗	80	78	77	NA	28
Pilz et al[35] (2021)	Austria	Retrospective cohort	212 ± 25	40/14,840	NA	NA	26–55‡	62.5	NA	NA	5 (12.5)||	NA
Mukherjee et al[36] (2021)	India	Cross-sectional study	>102	58/1300	NA	NA	34 ± 11	76.3	32	6	NA	25.1
Cavanaugh et al[37] (2021)	USA	Retrospective cohort	101–110∗	5/25	NA	NA	67–99∗	80	5	0	1 (20)	<30

Data are shown as ∗range, †mean, or ‡lower quartile–upper quartile. §The study by Soriano et al[ reported two cases of reinfection, but only described the first case as a 42-year-old obesity male who suffered pneumonia during the two episodes and required hospitalization. ||Five patients were hospitalized during the second infection, and four of them were also hospitalized during the first infection.CT: Cycle threshold; NA: Not available.

Quality assessment

Of the 17 cohort studies, 7 (41%) were of moderate quality,[ and the quality of the remainder was high; 53% of the cohort studies did not mention an adjustment of confounders.[ The quality of both the cross-sectional study and the ecological study were high.[ These two studies clearly described their study subjects and the setting, and they measured the exposures and outcomes in a valid and reliable way [Supplementary Table 4].

Reinfection rate

Single-study and summary incidences of reinfection are shown in Figure 2. The incidence of reinfection in recovered COVID-19 patients ranged from 0% to 20% across the 19 studies. The pooled reinfection rate was 0.65% (95% CI 0.39–0.98%), with high heterogeneity (I2 = 99%) [Figure 2]. The symptomatic reinfection rate was a bit lower (0.37% [95% CI 0.11–0.78%], I2 = 99%; Supplementary Figure 1A), whereas the reinfection rate was higher in HRP (1.59% [95% CI 0.30–3.88%], I2 = 90%; Supplementary Figure 1B).

Figure 2

Forest plot illustrating the single study and summary incidence of SARS-CoV-2 reinfection. CI: Confidence interval; SARS-CoV-2: Severe acute respiratory syndrome Coronavirus 2.

Protection against reinfection

Protection against reinfection and symptomatic reinfection was 87.02% (95% CI 83.22–89.96%) [Figure 3A] and 87.17% (95% CI 83.09–90.26%) [Figure 3B], respectively.

Figure 3

Forest plot illustrating the protection afforded from initial SARS-CoV-2 infection (1-RR). (A) Protection against reinfection, (B) Protection against symptomatic reinfection. CI: Confidence interval; RR: Risk ratio; SARS-CoV-2: Severe acute respiratory syndrome Coronavirus 2.

Subgroup analyses of reinfection rate

Table 2 summarizes the results of subgroup analyses for reinfection rate in patients with COVID-19. Only three studies reported reinfection rates, which overall were 0.38% (95% CI 0.27–0.51%) in females and 1.77% (95% CI 0.01–6.75%) in males. Among the study designs, the reinfection rate in cross-sectional studies was higher than in other study designs (4.46% [95% CI 3.41–5.65%]). Similarly, the reinfection rates in both North America and infection intervals >90 days were higher than in the remaining subgroups.

Table 2

Results of subgroup analysis of the incidence of reinfection in patients with COVID-19.

Items	No. of studies	Reinfection (n)	Initial positive infection (n)	Incidence, % (95% CI)	I2 (%)	Prediction interval, %(95% CI)
Gender
Male∗	3	221	37,541	1.77 (0.00, 6.75)	99	0.00, 100.00
Female	3	71	17,807	0.38 (0.27, 0.51)	35	0.00, 2.29
Study design
Prospective cohort∗	3	161	9563	1.66 (0.18, 4.58)	95	0.00, 82.05
Retrospective cohort	14	573	277,853	0.35 (0.19, 0.57)	98	0.00, 1.40
Ecological study†	1	304	36,509	0.83 (0.74, 0.93)	–	–
Cross-sectional study†	1	58	1300	4.46 (3.41, 5.65)	–	–
Continent
Europe	10	588	83,955	0.54 (0.22, 0.99)	97	0.00, 2.61
North America	5	258	59,621	0.73 (0.36, 1.25)	94	0.00, 2.90
Asia	4	250	181,649	0.63 (0.21, 1.28)	99	0.00, 5.75
Infection interval
≥90 days	17	913	148,915	0.74 (0.46, 1.08)	97	0.02, 2.48
≥45 days†	2	183	176,310	0.14 (0.00, 0.51)	99	–

Prediction interval was too wide due to severe heterogeneity in very small number of included studies.

Prediction interval could not be calculated due to small number of included studies. CI: Confidence interval; COVID-19: Coronavirus disease 2019.

Cumulative analyses of reinfection rate

Cumulative meta-analyses indicated that with increased sample size, the point estimate gradually stabilized and the CI gradually narrowed, showing that the larger the sample size, the greater the accuracy of the results [Supplementary Figure 2].

Sensitivity analyses of reinfection rate

Excluding each study one by one from the analyses, the results of sensitivity analyses (0.51–0.73%) were in good agreement with the reinfection rate, indicating the robustness of the results [Supplementary Figure 3].

Evidence quality

The GRADE system showed that the quality of protection against symptomatic reinfection was moderate, while the results of protection against reinfection, reinfection rate, symptomatic reinfection rate, and reinfection rate for HRPs had low quality [Supplementary Table 5].

Discussion

To the best of our knowledge, this was the first meta-analysis to investigate the reinfection rate of SARS-CoV-2 in a large population. Our results indicate a relatively low reinfection rate in the general population but a much higher rate in HRPs, and protection against reinfection or symptomatic reinfection was 87%.

Reinfection with the SARS-CoV-2 virus can be attributed to two main causes. The first reason is the decline in immunity over time or the failure of naturally acquired immunity, which results in reinfection with the same virus strain, making people sick or asymptomatic carriers.[ Another reason for this may be viral mutations that can easily lead to reinfection because the previously established naturally acquired immunity may not be effective against the mutant strain.[ Hence, regardless of whether long-term protective immunity is possible for all patients after exposure to COVID-19, it may make them vulnerable to reinfection. It should be recalled that social distancing, the use of masks, hand hygiene, and other preventive measures are very important for recovering patients, particularly those in HRPs who are more exposed to the virus.

Several factors may influence the reinfection rate. First, subjects infected during the first wave of the pandemic did not undergo antibody or polymerase chain reaction testing and were not admitted or hospitalized for treatment (particularly if they had asymptomatic attacks).[ Thus, it was difficult to accurately identify all reinfected individuals. Second, a recent meta-analysis[ indicates that some reinfection cases may appear as false-positive results in the first and/or second infection tests, which may produce an overestimated reinfection rate. Additionally, most of the positive cases may simply be protracted first infections rather than true reinfections due to the relatively high positive retest rate (12.0–32.9%) following the convalescent period.[ Thus, the reinfection rate in some included studies might be an overestimate.

As COVID-19 vaccination programs develop, it is important to note that patients who had SARS-CoV-2 antibodies were excluded from some vaccine studies. Nevertheless, previous infections still had an 87% protective effect during the study period. This is equivalent to or better than the protective effect reported in recent vaccine studies. However, due to differences in study design and study populations, direct comparison is not possible.[ Based on our findings, we believe that in areas where vaccines are rare, the vaccination of patients previously infected with COVID-19 can be delayed to allow HRPs to be vaccinated first. However, the efficacy of the vaccine for previously immune patients is still unclear and may need to be further examined.

Several limitations should be noted. First, the incidence of reinfection might have been overestimated because most included studies lack the gold standard of confirmation (i.e., genetic lineage or clades between initial infection and reinfection). Second, due to the lack of detailed clinical features in most studies that were examined, cases of reinfection cannot be examined in detail, particularly the immune features, which would be of great assistance to our understanding of the protection of natural immunity and virus escape. Finally, subgroup analyses based on disease severity, age, and comorbidities could not be performed due to the lack of specific data. Thus, the results should be interpreted with caution.

Conclusion

The reinfection rate of SARS-CoV-2 is relatively low. It has a similar protective effect against SARS-CoV-2 reinfection as vaccine inoculation. These data may help determine public health measures and vaccination strategies in response to the COVID-19 pandemic. Meanwhile, factors affecting reinfection incidence, such as strains of the virus, patient immune status, or other patient-level characteristics, should be evaluated in future studies to help develop strategies to control and prevent their occurrence.

Funding

This study is funded by grants from the National Natural Science Foundation of China (No. 72074011), the National Key Technology R&D Program of China (No. 2020YFC0840800), and the National Key R&D Program of China (No. 2021YFC2301601).

Conflicts of interest

None.