Questions for This Update

• Key Question 2. What is the risk of reinfection from SARS-CoV-2 among adults with prior SARS-CoV-2 infection?
  1. Does the risk of reinfection vary by patient characteristics (e.g., age, sex, race/ethnicity, and comorbidities), severity of the initial infection, initial antibody levels, SARS-CoV-2 variants, or vaccination status?
  2. Is there a threshold level of detectable anti-SARS-CoV-2 antibodies necessary to confer natural immunity immunity acquired by infection, and if so, does this threshold vary by patient characteristics (for example, age, sex, race/ethnicity, and comorbidities)?
• Key Question 3. What is the duration of protection against reinfection among adults with prior SARS-CoV-2 infection?
  1. Does the duration of protection vary by patient characteristics (e.g., age, sex, race/ethnicity, and comorbidities), severity of the initial infection, initial antibody levels, SARS-CoV-2 variants, or case identification method (e.g., surveillance, symptomatic testing only)?

American College of Physicians (ACP) Practice Points

Based on the findings for this report, the ACP has issued the following practice points:

"Practice Point 1:
Do not use SARS-CoV-2 antibody tests for the diagnosis of SARS-CoV-2 infection.

Practice Point 2:
Do not use SARS-CoV-2 antibody tests to predict the degree or duration of natural immunity conferred by antibodies against reinfection, including natural immunity against different variants."

Statements From Public Health Organizations

The CDC states that, based on experience with other human coronaviruses:

"…the probability of SARS-CoV-2 reinfection is expected to increase with time after recovery from initial infection because of waning immunity and the possibility of exposure to virus variants…. The risk of reinfection may be increased in the future with exposure to SARS-CoV-2 variant virus strains that are not neutralized by immune antisera, such as one recently described in South Africa…. The risk of reinfection also depends on the likelihood of re-exposure to infectious cases of COVID-19. Continued widespread transmission makes it more likely that reinfections will occur." (June 2021)

The World Health Organization (WHO) does not have a recent statement about reinfection.

Future Research and Ongoing Studies

Longer follow-up from the included studies should assess whether protection lasts for periods longer than 7 to 10 months, whether variants that were not prevalent in current studies can evade immunity acquired by previous infection, and whether vaccination adds significant protection among individuals who have been infected. Additional studies are needed to address protection against reinfection in the young, in the elderly, in patients who tested positive for SARS-CoV-2 but had no symptoms, and in immunocompromised patients and those with other comorbidities.

The populations studied to date are also relatively homogenous racially, ethnically, and geographically. Analysis of larger, more diverse cohorts of previously infected individuals could help verify whether the estimates of reinfection rates we found are applicable in other populations, social circumstances, and settings.

Ongoing studies of immune responses and risk of reinfection may address some of these gaps. Three of these studies are examining COVID-19 survivors under the age of 18, contributing to the substantial knowledge gap in pediatric populations. Additionally, three studies will include groups of vaccinated individuals in their cohorts.

This review does not provide cost information.

Limitations and Applicability

In the evidence we reviewed, most investigators described their decision-making regarding study methods, and many conducted sensitivity analyses or alternative cohort analyses to minimize error and detect biases that are inherent in studying immunity acquired by previous infection. Nevertheless, it is not clear that they overcame every challenge. With respect to cohort composition, no feasible study design can ensure that—within the target population—all infected individuals, regardless of symptoms, are identified and allocated appropriately, or that exclusions of individuals who lacked required tests for allocation would not bias the results. Most studies did not perform protocolized follow-up testing designed to capture all incident infections and reinfections. While widely used in the literature, the term "asymptomatic infection" and "asymptomatic reinfection" are poorly defined, and methods to distinguish symptomatic reinfections from virological recurrence without clinical evidence for infection were problematic.51 Study methods and knowledge of SARS-CoV-2 are not sufficiently developed to distinguish which people with "asymptomatic infections" are "pre-symptomatic" on the one hand, or "colonized" on the other. Also, many of the studies used surveillance methods that were not adequate for detection of all asymptomatic infections. Nevertheless, the protection estimates for asymptomatic and symptomatic infection were similar, and the result was not sensitive to these and other potential methodological weaknesses.

Antibody testing has been proposed as a potential marker or correlate of protection against infection. In our analysis, seroconversion or a positive antibody test obtained soon after the onset of infection was strongly associated with protection against reinfection. This finding applies only to people who have had a negative antibody test (e.g., for surveillance in a study setting) and convert to a positive one, or people who have never been infected and develop antibodies during or immediately after a wave. In these situations, the prognostic value of antibody testing was identical to the prognostic value of the more widely used PCR test, which, it should be noted, has additional value because it not only tells us about reinfection risk but also about transmission risk.

A key limitation of this literature is that it does not apply to antibody testing in people and clinical settings when the timing of testing in relation to infection is unknown. Ongoing research may provide better information about the utility of antibody testing in actual practice. Specific gaps in current evidence are whether failure to develop antibodies, antibody titers or levels, the loss of antibodies, or the antibody target (which spike proteins it binds to) provide useful information about reinfection risk. A particularly important gap is how much protection infection confers in immunocompromised people who do or do not develop antibodies (or high titers of antibodies) after infection. Until ongoing research addresses these gaps, our results shed little light on the role of antibody testing in actual practice.

None of the studies could account directly for the behavioral and occupational variables that affect infection risk and might well be unevenly distributed between the positive and negative cohorts. It is also possible that a group of people at higher reinfection risk, perhaps because they engaged in much riskier behavior than most people, were less likely to be recruited, perhaps because they avoided the testing that would make them eligible and countable in these studies. While possible, this and other scenarios that can be imagined seem unlikely and would require that all of the studies suffered from large, undetected confounding. Despite evidence of heterogeneity, our results were consistent across a wide range of methodological diversity, increasing our confidence in the main findings and in the robustness of the results of the antibody-only studies.

Of less concern, but worth noting, results do not address protection conferred by a first infection that occurred between or after high incidence surges. Many of the studies measured reinfections as new cases happening within the second pandemic wave in a particular geographic location. This approach avoids the time confounding that might exist should cases have been considered continuously. That is, because public health restrictions, variants, and other potential confounders changed frequently over time, no study could reasonably account for these changes analytically in continuous time. This means that the degree of protection afforded by immunity acquired from recent infection between waves has not been thoroughly studied.

The results may also be difficult to apply when there is uncertainty about how much time has elapsed since initial infection, as is often the case in clinical practice. Also, as the vast majority of timepoints included in studies were prior to the emergency use authorization of any vaccines in late 2020, the results may be less applicable in populations with high vaccination rates.

All the included studies were conducted in highly developed countries, and our findings may not be as applicable to less-developed countries where exposures may differ due to preventative public health measures not being as widespread or feasible. It is reassuring that the results apply to frail individuals residing in long-term care facilities, but results may also be less applicable to groups that were not well-represented in the studies, especially immunocompromised patients.

Our results do not in any way argue for infection rather than vaccination as a means of obtaining individual or herd immunity. Follow-up studies of protection against reinfection do not include people who died from COVID-19 and do not consider that morbidity from COVID-19 far outweighs any potential advantage conferred by immunity acquired from recent infection. Nor do our results provide evidence that immunity acquired by infection is longer-lasting or in other ways superior to immunity acquired by vaccination.

Despite the noted limitations, the findings provide strong evidence that the immunity afforded by previous infection confers strong protection against reinfection for at least 7 months. At present, recent infection is a reliable marker of protection against symptomatic reinfection with SARS-CoV-2.