Impact of baseline SARS-CoV-2 antibody status on syndromic surveillance and the risk of subsequent COVID-19—a prospective multicenter cohort study

Effective and durable host immunity directed against severe acute respiratory syndrome (SARS-CoV-2) is key to the long-term control of the current coronavirus disease 2019 (COVID-19) pandemic. In consequence, the degree and duration of protection against re-infection in those with specific antibodies against SARS-CoV-2 are currently being debated [1]. Documented cases of re-infection (mean interval between infections was 106 days) are increasing and alternative avenues for immunity to SARS-CoV-2 have been proposed [2]. However, recent evidence suggests that neutralizing antibodies against SARS-CoV-2 are consistently detectable for at least 9 months and offer protection against clinically relevant re-infection [3‐5]. The most compelling evidence comes from a UK study, where—among 12,000 healthcare workers (HCWs) with a follow-up of 6 months—those with detectable anti-spike antibodies at baseline were less likely to have SARS-CoV-2 detected in a subsequent nasopharyngeal swab [4]. However, this study has not specifically assessed the frequency of COVID-19 specific symptoms among participants.

This HCW cohort study prospectively evaluated the risk of SARS-CoV-2 infection and the occurrence of COVID-19 symptoms among participants with and without SARS-CoV-2 anti-nucleocapsid antibodies at baseline.

Between 22 June and 20 October 2020, we recruited 4812 HCW from 17 institutions across Northern and Eastern Switzerland, corresponding to 28% of the eligible HCW population (n = 17,060). Of the 4812 HCW, 78% were female and median age was 38.9 years; most worked as nurses (47%) or physicians (17%) (Table 1). These figures were similar to the characteristics of the eligible population, where 76% were female, median age was 40 years, and 40% worked as nurses and 15% as physicians [6].

A total of 5318 NPS were performed during follow-up, including 3391 (64%) were PCR and 1879 (36%) were antigen tests; 2712 individuals reported having at least one NPS performed (i.e., NPS group). Seropositive participants were less likely (67/144, 47%) to undergo NPS testing than seronegative participants (2645/4668, 57%) (P = 0.02). Conversely, the mean number of NPS per person (among those with at least one NPS) did not differ significantly between seropositive and seronegative HCWs (1.8 vs. 2.0 tests, P = 0.34). Also, the proportion of antigen tests (vs. PCR) was similar between seropositive (39%) and seronegative HCWs (36%) (P = 0.69).

In total, 550 of 2712 participants in the NPS group reported at least one positive NPS during follow-up. Of 67 seropositive participants in the NPS group, only three (4.5%) received at least one positive NPS result during follow-up, compared with 547 (20.7%) of the 2645 seronegative participants (Fig. 1). This translates into a RR of 0.22 (95% CI 0.07 to 0.66, P = 0.002) for a positive NPS after positive baseline serology. Including HCW who did not undergo NPS testing (i.e., analysis of full cohort), the corresponding RR was 0.18 (95% CI 0.06 to 0.55, P < 0.001). In Cox regression, this translated into a hazard rate of 0.16 (95% CI 0.05 to 0.51, P < 0.001) for the full cohort (Fig. 2). The three cases with presumable re-infection after positive baseline serology were all diagnosed in January 2021 after a follow-up (i.e., time from baseline serology to second positive SARS-CoV-2 test) of 198, 200, and 220 days. One of the three HCWs was asymptomatic at time of re-infection. Post hoc measurement revealed positive anti-S antibodies at baseline for all three HCW. For detailed characteristics, see Table 2.

Among 4812 participants, 2879 HCW (59.8%) reported at least one symptom during follow-up, whereas the others (n = 1933) remained without symptoms during follow-up. Symptoms were reported by 79/144 (54.9%) of the initially seropositive participants and by 2800/4668 (60%) of the initially seronegative participants (P = 0.25) (Fig. 1). The total number of reported symptoms was not significantly different in baseline seropositive compared to seronegative individuals (median of 5 vs. 6 symptoms, P = 0.30). Out of 15 different symptoms, ten were reported less frequently by participants who were seropositive at baseline, although the difference was statistically significant only for impaired olfaction/taste (RR 0.33, 95% CI 0.15–0.73, P = 0.004), chills (RR 0.59, 95% CI 0.39–0.90, P = 0.01), and limb/muscle pain (RR 0.68 95% CI 0.49-0.95, P = 0.02) (Fig. 3, Table 3).

For the analysis of symptom specificity, we included 532 episodes with positive and 1527 with negative test results. Almost all (14 out of 15) symptoms were less frequently reported during episodes with a positive NPS (compared to episodes with negative NPS result). The symptoms which discriminated best between episodes with positive and negative NPS were impaired olfaction/taste (OR 22.2 95% CI 17.1–29.1, P < 0.001), limb/muscle pain (OR 4.7 95% CI 3.8–5.9, P < 0.001), and weakness (OR 4.4 95% CI 3.1–6.4, P < 0.001) (Additional file 2: Figure S2, Additional file 3: Table S1).

For validation of swab results, we cross-checked self-reported NPS results for a subgroup of participants (from the largest participating institution). We found that 150 out of 174 presumable positive NPS were indeed documented in the database of the division of occupational medicine. The remaining HCWs most likely had a positive NPS outside of their working place. On the other hand, none of the randomly selected 175 HCWs reporting only negative NPS results was found to have a positive NPS in the database.

In this prospective cohort of over 4800 HCW followed during the second wave in Switzerland, we demonstrate that the presence of anti-nucleocapsid SARS-CoV-2 antibodies at baseline not only reduced the risk of positive nasopharyngeal SARS-CoV-2 tests, but also the occurrence of COVID-19 specific symptoms such as loss of smell and limb or muscle pain. The follow-up of almost 8 months, the large sample size, the systematic collection of symptoms, and the excellent questionnaire response rate are among the strengths of the study.

These results add to the mounting evidence that specific antibodies protect against subsequent SARS-CoV-2 infection. In the study of Lumley et al., the adjusted incidence rate ratio for seropositive HCW to have a positive PCR (median follow-up 6 months) was 0.11 compared to those without antibodies [4]. Within a cohort of over 3.2 million US patients, Harvey et al. found a ratio of 0.10 of positive PCRs among those with vs. those without positive antibody test at baseline (for tests performed > 90 days after baseline) [8]. Among over 43000 people with a positive antibody test (median follow-up 4 months) from Qatar, the estimated efficacy of natural infection against re-infection was above 90% [9]. Also, data from a population-based study (> 500,000 people) conducted in Denmark suggest that protection after natural SARS-CoV-2 infection was 80% after 6 months (but only 47% for adults aged 65 years or older) [10]. A retrospective propensity-score matched cohort study from Western Switzerland found a 94% reduction in the hazard of having a positive SARS-CoV-2 test for seropositive individuals [11]. In a large HCW cohort from England, a previous history of SARS-CoV-2 infection was associated with a 84% lower risk of infection (median follow-up 7 months) [12]. Likewise, in a university student population from the USA undergoing repeat mandatory testing, an 84% protection from SARS-CoV-2 infection was found [13]. Also, in more than 3000 prospectively followed male US Marine recruits, subsequent infection was found about one fifth compared with seronegative individuals [14]. Although our median follow-up of 8 months is among the longest compared with other studies, our point estimate (RR 0.22) for protection from re-infection with SARS-CoV-2 is perfectly in line with these results. Our findings are further supported by a modeling study of the progression over 250 days of neutralizing antibodies following infection or vaccination and their protective role against symptomatic SARS-CoV-2 infection [15].

Previous studies have mainly looked at the incidence of confirmed SARS-CoV-2 infection to assess the risk of re-infection [8‐10]. This approach has the inherent limitation that people who are seropositive might be less likely to undergo NPS testing, as seen in our data. In addition to the collection of NPS test results, we therefore used a second approach to assess the protective effect of seropositivity on subsequent COVID-19. Using weekly symptom frequency from diaries as proxy for COVID-19 allows comparing incidence of COVID-19 irrespective of whether patients undergo NPS testing or not. Of course, this signal is being diluted by infections caused by other respiratory viruses, especially for symptoms like coryza, sore throat, cough, or fever. However, certain symptoms such as loss of taste or smell and myalgias have already been shown by others to be more specific for COVID-19 [16, 17]. The fact that exactly these symptoms occurred less frequently among those with antibodies at baseline supports our finding of a reduced risk for SARS-CoV-2 re-infection in this group. A limitation of this approach is that some participants might suffer from persisting COVID-19 symptoms (i.e., long-COVID). In particular, loss of smell has been shown to persist for several weeks in a certain proportion of COVID-19 patients [18]. If we had excluded seropositive participants who already reported this symptom at baseline, the effect would have even been more pronounced.

Worryingly, re-infections with phylogenetically different SARS-CoV-2 strains are increasingly being reported [19]. A recent report documented severe re-infection with the “new variant” VOC-202021/01 8 months after documented primary infection, even in the presence of anti-nucleocapsid antibodies and in the absence of overt immunosuppression [20]. We also observed three nurse assistants with presumable SARS-CoV-2 re-infection approximately 300 days after first infection and about 6 months after documented seroconversion. Post hoc testing revealed that all three re-infected cases had positive anti-spike titers at baseline. Of note, one of the nurses with highest anti-S did not report any symptoms at time of the test. Although we do not have any sequencing data for these particular viral strains, re-infections occurred in January 2021, when the proportion of the B.1.1.7 (alpha variant) was estimated to account for less than 20% of all SARS-CoV-2 isolates in Switzerland and when B.1.617.2 (delta variant) was not circulating at all [21]. Importantly, recent data suggest that protection from the SARS-CoV-2 delta variant, which is currently the predominant strain in most Western countries, might be clearly below 80% after natural infection [22]. Continuing follow-up of our and other cohorts will reveal the long-term protective effect of specific antibodies against SARS-CoV-2 and its emerging variants. Another open research question regards the long-term protective effect of vaccine-induced immunity, which might be higher compared to natural infection [15, 23].

A limitation of our study is the fact that nasopharyngeal SARS-CoV-2 testing was not routinely performed. Therefore, asymptomatic carriage of SARS-CoV-2 cannot be excluded in those with antibodies at baseline. We also used SARS-CoV-2 anti-nucleocapsid antibody titers and not anti-spike antibodies, which have been shown to correlate better with virus neutralization, to define seropositivity. Post SARS-CoV-2 infection, anti-N antibodies are detected equally [24] or even more frequently [25] than anti-S antibodies. We therefore suggest that our results represent, or somewhat underestimate, the true protective effect mediated by anti-S antibodies although all three re-infected cases had also detectable anti-spike antibodies at baseline. Swab results and symptoms were self-reported. Because validation of some of the swab results showed mostly consistent results, we consider these self-reported data to be highly reliable. Furthermore, we cannot definitely confirm that the three seropositive HCWs with positive SARS-CoV-2 NPS were indeed re-infected with a new strain. However, the long latency between the episodes, new onset of symptoms (two cases), and a negative PCR between episodes (one case) strongly support our hypothesis of re-infection (rather than persistence of viral RNA for more than 6 months). Another shortcoming of this study is that SARS-CoV-2 specific cellular immunity was not evaluated. This is mainly due to the fact that these measurements are still very time-consuming and cost-intensive in a large population. Thus, data are scarce so far. However, specific T cells most likely also contribute decisively to protection against SARS-CoV-2 infection [26]. Immunity mediated by specific T cells can be present even if there have never been signs of disease and antibodies are absent [27, 28]. In consequence, measuring antibodies alone, such as in our study, underestimates protection against COVID-19 in a population. Finally, we acknowledge that the diagnostic sensitivity of SARS-CoV-2 antigen tests is only around 80%; because 36% of NPS were indeed done by antigen test, the effect size observed in our study might be slightly overestimating the true protective effect.

We conclude that anti-nucleocapsid antibodies acquired after natural infection convey an approximately 80% protection against symptomatic SARS-CoV-2 infection, at least for a period of 8 months and in a setting of unvaccinated HCW where “new variant” mutations were not widely present at the end of follow-up. Syndromic surveillance for specific COVID-19 symptoms allows estimating the probability of SARS-CoV-2 re-infection irrespective of whether participants undergo NPS testing or not.