SARS-CoV-2 Reinfection and Severity of the Disease: A Systematic Review and Meta-Analysis

Since the discovery of SARS-CoV-2, changes in genotype and reinfection with different variants have been observed in COVID-19-recovered patients, raising questions around the clinical pattern and severity of primary infection and reinfection. In this systematic review, we summarize the results of 23 studies addressing SARS-CoV-2 reinfections. A total of 23,231 reinfected patients were included, with pooled estimated reinfection rates ranging from 0.1 to 6.8%. Reinfections were more prevalent during the Omicron variant period. The mean age of reinfected patients was 38.0 ± 6. years and females were predominant among reinfected patients (M/F = 0.8). The most common symptoms during the first and second infection were fever (41.1%), cough (35.7% and 44.6%), myalgia (34.5% and 33.3%), fatigue (23.8% and 25.6%), and headaches (24.4% and 21.4%). No significant differences of clinical pattern were observed between primary infection and reinfection. No significant differences in the severity of infection were observed between primary infection and reinfection. Being female, being a patient with comorbidities, lacking anti-nucleocapsid IgG after the first infection, being infected during the Delta and Omicron wave, and being unvaccinated were associated with a higher risk of reinfection. Conflicting age-related findings were found in two studies. Reinfection with SARS-CoV-2 suggests that natural immunity is not long-lasting in COVID-19 patients.


Introduction
Severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) is the virus responsible for the infection known as coronavirus disease 2019 (COVID-19) [1]. Reinfection may occur after recovery from the disease. The first published case of SARS-CoV-2 reinfection was documented in a 33-year-old Hong Kong man in August 2020, with two distinct virus strains corresponding to two episodes of infection [2]. In a meta-analysis addressing the risk of SARS-CoV-2 reinfection in discharged patients, the pooled estimated reinfection rate was 0.3% [3]. However, this study was conducted before the Omicron variant wave and did not compare clinical symptoms during the first and second infections. We recently observed that the Omicron variant was associated with an increased risk of reinfection in recovered patients [4]. The length of time between two episodes used to define a SARS-CoV-2 reinfection varies. The US Centers for Disease Control and Prevention uses the time frame of 90 days [5], while the European Centre for Disease Prevention and Control uses 60 days or more [6]. However, a shorter duration has been reported in unvaccinated patients, with a mean duration of 47 days (range 17-65 days) between two episodes of infection [7]. The risk factors for SARS-CoV-2 reinfection with recent variants of concern following a previous infection remain unclear. Therefore, we conducted an updated systematic review and meta-analysis aiming at assessing the proportion of reinfections with SARS-CoV-2 over time. We also compared the clinical patterns and severity of infection in reinfected patients to their status during primary infection. Finally, we aimed to identify risk factors for reinfection.

Eligibility Criteria
All published studies and pre-prints (including case series and prospective cohort studies) were selected according to the following inclusion criteria: (1) Studies reporting on adult COVID-19 patients (≥18 years old) with infection confirmed by a positive RT-PCR (Reserve Transcription Polymerase Chain Reaction); (2) Due to the heterogeneity of duration between two episodes of infection, we included studies reporting on patients that met the following criteria, based on the US CDC Common Investigation Protocol for Investigating Suspected SARS-CoV-2 Reinfection [8]: + Two episodes of infection at ≥90 days interval in patients with negative RT-PCR results between two episodes of infection, including asymptomatic patients; + Two episodes of infection at 60 to 89 days of interval: symptomatic patients confirmed by RT-PCR test with ≥ one intermediate negative RT-PCR test with no other reason or symptomatic patients tested with different genotypes between two episodes of infection; + Two episodes of infection at an interval of <60 days: symptomatic patients confirmed by differences between two episodes of infection in whole genome sequencing in patients infected with different virus genotypes; Only papers in English were included in this work.

Exclusion Criteria
We excluded single-case reports and all studies based on animal experiments. Review papers were excluded, though the reference lists were assessed to identify potentially relevant studies. We also excluded studies conducted only on immunocompromised patients with cancer and acquired immunodeficiency syndrome.

Data Collection Process
Four researchers independently checked the relevant articles and data extraction. All discordant results were discussed to reach a consensus between researchers.
The type of study, country, and setting where studies were performed, the number and proportion of reinfection cases, demographic information (age and sex), comorbidities and vaccination status, the time between two episodes of infection, symptoms, imaging findings, and severity outcomes (hospitalization and transfer to intensive care unit and death) were extracted from the included articles. When available, risk factors for reinfection were reported or calculated.

Assessment for Quality of Studies
The quality of cohort studies was accessed using the Newcastle-Ottawa Scale (NOS) [9], consisting of eight items with a maximum score of nine. The score of the study classifies the quality as good (7)(8)(9), fair (4)(5)(6), and poor (0-3). The quality of the cases series was evaluated using the National Institute for Health and Clinical Excellence (NICE) checklist, allowing the following stratification of series: good quality (score ≥ 7), fair quality (score 4-6), and poor quality (score 0-3) [10].

Statistics
All statistical analyses were performed using the open-source software R [R Core Team] R: A language and environment for statistical computing from the R foundation for statistical computing, based in Vienna, Austria, in 2020. URL: [http://www.r-project.org, accessed on 31 March 2023]. In the present study, the heterogeneity I2 > 50% was identified as representing substantial heterogeneity [11]. The pooled effect size and 95% confidence interval (95% CI) were given using the random effect model. A funnel plot was applied to estimate publication bias. We excluded all studies from the analysis where there was no event in either arm, according to the Cochran guide [12].
The most frequent variant circulating at the time of reinfection in each country was assessed using the CoVariants website (https://covariants.org/, accessed on 31 March 2023) and, when available, the SARS-CoV-2 sequencing data described in each study. Figure 1 shows the search algorithm. A total of 3051 articles were identified from PubMed and Google Scholar. After removing duplicates, 2903 articles were scanned by reading the title and abstract. A total of 237 articles were accessed for the full-text screening. Ultimately, 26 articles met the inclusion criteria for the meta-analysis (Supplementary Table S1).

Study Selection and Quality Assessment
Studies assessed using the NICE criteria were all scored as good quality (eight studies) and those assessed with NOS were scored as either good quality (seven studies) or fair quality (eleven studies) (Supplementary Tables S2 and S3).

Clinical Symptoms
Seven studies reported detailed symptoms for both episodes of infection in 168 reinfected patients [15][16][17]20,22,23,26]. In one of these studies, we included 121/209 patients for whom symptom information was available during both episodes [26]. The most common symptoms during the first and second infection were fever (41.1%), cough (35.7% and 44.6%), myalgia (34.5% and 33.3%), fatigue (23.8% and 25.6%), and headaches (24.4% and 21.4%). The prevalence of other symptoms is presented in Figure 2. Although sore throats tended to be less frequent during the first infection as compared to the second infection, no significant differences in the prevalence of symptoms were observed.  Figure 2. Although sore throats tended to be less frequent during the first infection as compared to the second infection, no significant differences in the prevalence of symptoms were observed. Only one study [14] reported the results of computed tomography chest at first infection with abnormal (ground-glass opacity) in 2/4 patients. Two studies reported abnormal chest CT results [14,15] in 3/7 patients with a second infection.

Severity of Infection
A total of 12 studies reported information on the hospitalization rate for both episodes of infection in 14,136/14,460 patients [14][15][16][17][18]20,23,24,26,28,35,36]. We excluded all studies from the analysis where there was no event of hospitalization in either arm. As Only one study [14] reported the results of computed tomography chest at first infection with abnormal (ground-glass opacity) in 2/4 patients. Two studies reported abnormal chest CT results [14,15] in 3/7 patients with a second infection.

Discussion
An early systematic review, mostly based on Chinese data and describing 466 reinfection cases, concluded with an estimated 15% reinfection rate [39]. This high reinfection rate is probably due to the lack of a clear definition for reinfection, which likely unduly considered discharged patients with a second positive test as being "reinfected" [39]. We already discussed this issue in a previous paper, showing that PCR sensitivity and specificity and sampling methods likely account for an overestimation of the SARS-CoV-2 reinfection rate [40]. In a subsequent study based on 1096 reinfection cases across three continents, a more realistic pooled reinfection rate of 0.7% was found, in line with our 0.9% result on 18,175 reinfected patients on five continents [41]. However, due to the many epidemics of new SARS-CoV-2 variants that continue to occur worldwide, it is likely that the reinfection rate will slightly increase in future studies as a result of a cumulative effect. At our institute, we observed that the prevalence of reinfection among SARS-CoV-2 infection increased from 0.2% during the second epidemic to 6.8% during the fifth wave of epidemic due to the Omicron variant [4]. Most reinfected patients were middle-aged, which was in line with previous results [39]; we did not find significant differences in age between reinfected patients and those with only a primary reinfection, when data was available. Furthermore, data about age as a risk factor for reinfection in studies where it was specifically evaluated were contradictory. Our data suggests that studies conducted on healthcare workers show a higher rate of reinfection, which might be due both to a higher likelihood of testing in this population and a higher occupational risk of exposure to the virus.
The time between two episodes of infection was about five months, as compared to two months in a previous meta-analysis based on 577 cases of reinfection [42]; this disparity likely results from a mechanical effect due to the naturally increased duration of the pandemic, as new studies are conducted.
We confirm the results of colleagues regarding the highest prevalence of reinfection in female patients [42]. This observation may be a result of lower screening rates in males and higher occupational exposure among females.
We found no clear evidence of different clinical patterns between primary and second infection, although our analysis was conducted on a small sub-sample of only 168 patients [40].
Dhillon et al. reported a higher proportion of transfer to the ICU during reinfection [42], which differed from our findings. However, differences were not statistically significant in either study.
Data on vaccination status are too heterogeneous to allow strong conclusions about the effectiveness of vaccination in lowering reinfection rates, though some studies suggest a potential protective effect [20,30]. Nevertheless, one recent study suggests that vaccination has no effect on reinfection rate, regardless of the number of doses provided [43].
We acknowledge that our review has some limitations. Firstly, high heterogeneity was observed in all studies analyzed. These results might have been caused by various factors, including methodology, sample size and, especially, the periods when studies were conducted, given the prevalence different SARS-CoV-2 variants at different stages in the COVID-19 pandemic.

Conclusions
Reinfection with SARS-CoV-2 does occur, suggesting that natural immunity is not longlasting in COVID-19 patients. Female patients seem more likely to experience reinfection than male patients. Thus far, no clear evidence for a difference in severity between the first infection and reinfection has been observed.