We aimed to determine the prevalence and evolution of SARS-CoV-2 antibodies in the general population aged 18 years and older during the 13-months study period. The increase in SARS-CoV-2 antibody prevalence, from 25.1% in April 2021 to 92.3% in March 2022, was the consequence of the vaccination campaign during this period. These rates were found to be consistently lower in comparison to the seroprevalence rates from the Belgian blood donors study [2] between April and December 2021. It is unlikely that salivary test we used explains this difference, as it was validated against COVID-19 PCR and paired serum/saliva samples with 95.1% sensitivity. Rather, our results highlight the value of general population studies to complement the scope of national serological surveillance.

Importantly, this study allowed assessing the trend in prevalence of SARS-CoV-2 antibodies in both vaccinated and unvaccinated people from the general population. In vaccinated people, the seroprevalence was high throughout the study period, but increased slightly as time progressed. However, a seroprevalence of 81.3% should be interpreted with caution due to the very small number of vaccinated participants in the pilot phase period. From the next period onwards, the number of vaccinated participants was high enough to give reasonably precise estimates. Here, the prevalence of SARS-CoV-2 antibodies in fully vaccinated people increased from 92.5 to 99.9% between May 2021 and March 2022. The lower rates found in the beginning probably result from the vaccine campaign starting with older people and people with chronic morbidities, who may have a lower immune response to vaccination. The increase in immune response over time is possibly due to multiple exposures to the antigen as time progressed (through COVID-19 infection or vaccine, including booster). Among the unvaccinated people, the prevalence of SARS-CoV-2 antibodies increased during wave 1, but was variable across time in wave 2 and wave 3, and lower than expected. Again, the low number (N = 100) of unvaccinated people in these periods call for caution in interpreting the results. The low antibody prevalence among unvaccinated people could derive from them adhering more strickly to the sanitary measures to prevent infection(e.g. lock-down, tele-working, mask wearing). Indeed, the proportion of participants who reported a history of a COVID-19 infection was lower than expected from the official COVID-19 infection statistics.

Among fully vaccinated people, seropositivity was significantly lower in those with a chronic disease, more particularly a neurological disease or a transplant. This result needs further investigation, as our study relies on self-report and a limited number of affected individuals. However, other findings support our results. For instance, patients with multiple sclerosis (a neurological affection), receiving disease-modifying therapies showed a reduced humoral immunity after SARS-CoV-2 vaccination [23].In addition, the seropositivity results in our group of participants with a transplant (58.3%) were remarkably similar to the SARS-CoV-2 anti-Spike seroprevalence of 52.4% found in a study among renal transplant patients [24, 25].

Social factors were also influencial. First, antibodies were more often present in people with a bachelor diploma who had higher seropositivity rates compared to those with a lower education. A possible explanation is that higher socio-economic status is associated with a better health status and behaviors, hence a stronger immune system. However, this was not confirmed among people with a master degree and above. Second, vaccinated people living with others had higher seropositivity rates compared to those living alone. Possibly they have a greater chance to be exposed to the virus. Studies have shown that living with children for instance increases the risk of SARS-CoV-2 infection [26, 27].

Finally, although both types of anti-COVID vaccines (nucleic-acid and viral-vectored) have demonstrated their effectiveness and their association with antibody development, some studies showed a higher seroprevalence among people who received a nucleic-acid vaccine compared to those with a viral-vectored vaccine [28]. This was also observed in our study. Seroprevalence was lower in the virus-vectored vaccine group compared to those having a nucleic-acid vaccine, whether delivered in a basic vaccination scheme or as a booster.

Among the unvaccinated people, seropositivity rate was lower in those with a O blood type compared to in those with an non-O blood type. A systematic review and meta-analysis indicated that blood group A may be a risk factor for COVID-19, whereas blood group O appears to be somewhat protective [29]. To what extent and how this relates with our findings remains unclear.

Seroreversion occurred in only 40 study participants. This low number may be related to the surge of Delta and Omicron variants of the virus between wave 2 and 3, resulting in many reinfections, hence few seroreversions.

Clearly, the time since the latest vaccination was an important predictor of seroreversion. Seroreversion was also much higher among the partially or unvaccinated people compared to people who were fully vaccinated. This confirms findings that antibodies developed following vaccination or following a mix of vaccination and COVID-19 infection were more robust and waned less rapidly than those developed after natural infection only [30, 31].

Our study has some important limitations regarding serological surveillance. First, we opted to detect antibodies in saliva, while serum-based methods are the preferred reference for seroprevalence studies. Still, our in-house SARS-CoV-2 RBD IgG ELISA test on saliva had shown high sensitivity and specificity. Unfortunately it made no distinction between antibodies from natural infection and from vaccination. Furthermore, the outcome reported in this study was dichotomous (presence or absence of antibodies), which had an impact on the level of analyses (less precision), but also on the interest of the study to participants. Indeed, providing the test result to the participants was initially an important incentive, but their motivation for follow-up decreased as the level of protection was unknown, the vaccine roll-out was fast and the epidemic was on a decline. Finally, the saliva collection was executed by the participants themselves, without supervision. The method to collect saliva (Oracol®) is designed for self-use and much effort had been put in giving clear instructions in a leaflet and a video. Nevertheless, it appeared that 17% of the initial swap samples down to 9% of the wave 3 swap samples did not contain enough saliva to be analysed.

The study also bears important strengths. It is a population-based probability sample including residents from 317 of the 581 Belgian municipalities. Even though non-response and drop-out were substantial and biases are inevitable, the use of post-stratification weights with both the national register and the exhaustive Belgian vaccination record database as auxiliary data sets, ensured that results were as representative as possible of the Belgian population aged 18 years and above.

Furthermore, the questionnaires that accompanied the three data collection waves allowed to gather extensive information from the participants in many different domains: socio-demographic information, health related factors, health behaviors, COVID-19 infection, vaccination status, etc. Additionally, over 90% of the study participants agreed that their saliva samples could be stored in a biobank and that their results could be linked with administrative databases for further research.