Cohort and study design

The CovidCatCentral (CCC) cohort consists of two groups of primary HCWs from three primary care counties in Barcelona, Spain⁴⁵ (Fig. 1, Supplementary Table 2). The first group includes 247 individuals enrolled from September 2020 to January 2021 who had experienced symptomatic SARS-CoV-2 infections prior to recruitment. The second group comprises 200 naïve HCWs recruited after receiving full primary vaccination (March–April 2021). The first group was visited at 11 timepoints (T) while the second was visited at 7 timepoints, between September 2020 and June 2023 (n at T9: 377, n at T10: 392, n at T11: 346) (Table 1). Most of the study participants were female (83.8% at T9, 85.2% at T10, and 86.4% at T11). The mean age of the participants was 48.6 years (standard deviation [SD] 10.7) at T9, 50.0 years (SD 10.7) at T10, and 50.0 years (SD 10.4) at T11. Additionally, 66.8% of participants at T9, 66.3% at T10, and 70.2% at T11, had underlying comorbidities.

The grey plot represents the number of SARS-CoV-2 cases at a given time in Catalonia according to the official data available from IDESCAT (Statistical Institute of Catalonia, https://www.idescat.cat/dades/covid19/?lang=es (accessed on 03 July 2024)). The bottom timeline depicts the main SARS-CoV-2 lineages circulating in Spain at those time intervals according to GISAID. The orange lines represent the three follow-up periods (6 months each) for breakthrough infections in this study. The top grey lines represent the cohort of participants recruited during the first wave of the COVID-19 pandemic (March–April 2020, n = 247) with symptomatic SARS-CoV-2 infection confirmed by rRT-PCR and/or RDT (first-infected), and the cohort recruited in March–April 2021 after completing full primary vaccination (n = 200) without evidence of prior infection (first-vaccinated). T: Timepoint.

SARS-CoV-2 infections and COVID-19 vaccination

Symptomatic SARS-CoV-2 infections were detected by rRT-PCR and/or RDT. To identify potential undiagnosed COVID-19 infections (asymptomatic infections), we used already available serology data and employed a FC analysis of antibody levels to S and N antigens across consecutive study timepoints. We identified 2, 2, 14, 8, 10, 30, 39, and 55 asymptomatic infections within the intervals: T3-T4, T4-T5, T5-T6, T6-T7, T7-T8, T8-T9, T9-T10, and T10-T11, respectively. Using this method, we were able to detect at least 75% of the diagnosed infections in each interval, which indicates we had a minimum of 75% sensitivity to detect the undiagnosed infections (Supplementary Table 1).

At T9 (May 2022, n = 377) (Supplementary Table 3), 333 participants had received the primary series of vaccination, and 245 had received the 1 st booster dose. The median time since last vaccination was 157 days (IQR: 144–171 days). Up to T9, 79.6% of the participants (n = 300) had been previously infected by SARS-CoV-2 according to rRT-PCR, RDT, or serology data, and 20.4% (n = 77) had no evidence of infection. The median time since last infection was 204 days (666.5 IQR). At T10 (November – December 2022, n = 392) (Supplementary Table 4), 89% of participants (n = 350) had received the primary series of vaccination, 67% (n = 264) had received a 3 rd dose (1 st booster), and 10% (n = 41) had received the 2 nd booster dose. The median time since last vaccination was 342.5 days (37 IQR). Up to T10, 89% of the participants (n= 349) had been previously infected by SARS-CoV-2, and 10.9% (n = 43) had no evidence of infection. The median time since last infection was 279 days (251 IQR). At T11 (May – June 2023, n = 346) (Table 1), 14 participants had not received any vaccine doses (4%), 308 participants had received the primary series of vaccination (89%), 241 had received a 3 rd dose (70%), and 72 had received the 2 nd booster dose (21%). The median time since last vaccination was 514 days (234.5 IQR). Up to T11, 94.2% of the participants (n= 326) had been previously infected by SARS-CoV-2, and 5.8% (n = 20) had no evidence of infection but were vaccinated with 3 or 4 doses. The median time since last infection was 360.5 days (219.25 IQR).

Among the vaccinated participants, 141 (36.9%) had vaccine breakthroughs between T8 – T9 (113 diagnosed and 28 detected by serology), 128 (35.3%) had vaccine breakthroughs between T9–T10 (92 diagnosed and 36 detected by serology), and 89 (23.7%) had vaccine breakthroughs between T10 – T11 (54 diagnosed and 35 detected by serology).

Recent infection, hybrid immunity, and booster doses are associated with higher IgG antibody levels

At T9, those with hybrid immunity exhibited higher IgG levels against all measured antigens, corresponding to the Wuhan strain (Beta coefficient [β] anti-S IgG 21.9%, 95% confidence interval [CI]: 52.0-89.6) (Fig. 2, Supplementary Fig. 2). Similarly, the number of prior infections correlated with increased IgG levels against all Wuhan antigens. The number of vaccine doses received was associated with elevated IgG levels against S, S1, and RBD)(β anti-S IgG 67.5%, 95% CI: 51.5-85.2). Receiving three doses, as opposed to two, was linked to higher IgG levels against S antigens (β anti-S IgG 52.5%, 95% CI: 22.1-90.3). In addition, age was positively associated with higher levels of IgG to S2, and individuals with comorbidities had higher levels of IgG against S antigens.

Beta (β) and CI values have been transformed to a percentage for an easier interpretation. Filled dots indicate p <0.05, while unfilled dots indicate non-significant p-values after adjustment for multiple testing by Benjamini-Hochberg. Error bars represent the 95% confidence intervals (CI). ^aAdjusted by sex^b Adjusted by age. ^c Adjusted by age and sex. ^d Adjusted by age, sex, smoking, number of doses and number of infections. ^e Adjusted by age, sex, comorbidities, first exposure type, time since last exposure, number of doses and number of infections. ^f Adjusted by age, sex, comorbidities, smoking, and number of infections. ^g Adjusted by age, sex, comorbidities, smoking, and number of doses. Full-length (Fl), Nucleocapsid (N), Receptor-binding domain (RBD), Spike (S). T: timepoint.

To further explore the role of comorbidities on antibody levels we performed additional analysis grouping comorbidities by type metabolic disorders—diabetes mellitus, dyslipidemia, obesity, hypothyroidism; cardiovascular diseases—hypertension, heart diseases; immunological conditions—immunosuppression, cancer; neurological and mental health: depression and neurological diseases) and assessing their association with antibody levels as compared to individuals without any comorbidities. Among individual comorbidity groups, cardiovascular diseases showed a significant positive association, while neurological and mental health conditions were close to significance (Supplementary Data 1).

Having hybrid immunity was associated with higher IgG levels to S and S1 from Wuhan and RBD from Wuhan and Delta variants at T10 (β anti-S Fl IgG 36.9%, 95% CI: 11.7-67.9) and T11 (β anti-S Fl IgG 37.9%, 95% CI: 8.03–76.2), as well as Wuhan S2 at T10. However, at both timepoints, hybrid immunity did not correlate with IgG levels to Omicron antigens (Fig. 2, Supplementary Fig. 2).

The number of vaccine doses received were positively associated with IgG levels against all S antigens at both T10 (β anti-S Fl IgG 43.7%, 95% CI: 34.1-53.9) and T11 (β anti-S Fl IgG 44.2%, 95% CI: 35.4-53.6), with the exception of S2 from Wuhan at T10. (Fig. 2, Supplementary Fig. 2). Notably, at T10, the number of doses was negatively associated with IgG levels to N. By T11, receiving four doses compared to three was linked to higher IgG levels against all antigens except N (β anti-S Fl IgG 45.5%, 95% CI: 20.9-75.0).

The number of prior infections consistently showed a positive association with IgG levels to S2 and N proteins at all three timepoints (Fig. 2, Supplementary Fig. 2). Additionally, at T10, a recent infection (within 190 days) was linked to higher IgG levels against all antigens compared to no infection (β anti-S1 IgG 79.1%, 95% CI: 46.3-116.9), and higher IgG levels to RBD from Omicron and Delta variants, as well as against Wuhan S1 and N, compared to having an infection more than 190 days prior (β anti-S1 IgG 37.6%, 95% CI: 13.8-66.4). At T11, a recent infection similarly correlated with higher IgG levels against Delta RBD, BA.2 RBD, BQ.1.1 RBD, and Wuhan S1, S2, and N, as compared to no infection (β anti-S1 IgG 59.95%, 95% CI: 23.1-107.8), whereas infections occurring within 190 days associated with higher IgG levels to all antigens except Wuhan RBD compared to infections more than 190 days prior (β anti-S1 IgG 19.1%, 95% CI: 2.99-38.6). Infections between 190- and 380-days prior were associated with elevated IgG levels against Delta RBD, BA.2 RBD, and Wuhan S1 and N at both T10 (β anti-S1 IgG 91.7%, 95% CI: 8.80-237.9) and T11 (β anti-S1 IgG 32.6%, 95% CI: 5.15-67.2), as well as IgG to RBD and S2 at T10, compared to no infection (Fig. 2, Supplementary Fig. 2).

Age was consistently positively associated with IgG levels to S, S1, and S2 antigens at both T10 and T11, with additional associations observed with all Omicron RBDs except BA.2 (Fig. 2, Supplementary Fig. 2). Even after adjusting for comorbidities, number of exposures, and time since the last exposure, age was still found to be associated with higher IgG levels against both Wuhan and Omicron variants (Supplementary Fig. 3). Lastly, smoking was inversely associated with IgG levels to S2 at T10 and with S2 and N at T11.

Having a recent infection is associated with lower risk of Omicron breakthrough infection

The number of previous infections was positively associated with protection against symptomatic infections during the T9–T10 and T10–T11 six-month periods in multivariable Cox models (T9–T10 hazard ratio [HR], 0.55; 95% CI, 0.40-0.76; T10–T11 HR, 0.56; 95% CI, 0.35-0.91) (Fig. 3a, Supplementary Fig. 4a). Similarly, a higher number of previous infections was associated with protection against all infections (symptomatic plus asymptomatic) in both periods according to multivariable logistic regression models (T9–T10 odds ratio [OR], 0.45; 95% CI, 0.69 to 0.29; T10–T11 OR, 0.43; 95% CI, 0.65 to 0.28) (Fig. 3b, Supplementary Fig. 4b). Furthermore, having a recent infection within the previous 190 days was associated with protection against symptomatic infections over both the T9–T10 and T10–T11 periods in multivariable Cox models (T9–T10 HR, 0.15; 95% CI, 0.07-0.31; T10–T11 HR, 0.007; 95% CI, 0.01-0.46) (Fig. 3a, Supplementary Fig. 4a). This protective effect was also observed for all infections (T9–T10 OR, 0.21; 95% CI, 0.46 to 0.09; T10–T11 OR, 0.16; 95% CI, 0.46 to 0.05) (Fig. 3b, Supplementary Fig. 4b).

a. Association with symptomatic infections were estimated using using multivariable Cox regression models. b. Association with asymptomatic infections were estimated using multivariable logistic regression models. The color of the dots represents the P value, where black represents < 0.001, dark grey < 0.01, light grey < 0.05, and white non-significant. Error bars represent the 95% confidence intervals (CI). ^a Adjusted by sex. ^b Adjusted by age. ^c Adjusted by smoking, age and sex. ^dAdjusted by age and sex. ^e Adjusted by age, sex, comorbidities, smoking, and number of doses. ^f Adjusted by age, sex, comorbidities, smoking, and number of infections. ^g Adjusted by age, sex, comorbidities, first exposure type, time since last exposure, number of doses and number of infections. T: timepoint.

In unadjusted models, hybrid immunity was associated with protection in T9 –T10 (HR, 0.4; 95% CI, 0.61 to 0.26; OR, 0.46; 95% CI, 0.78 to 0.26) and T10 – T11 (HR, 0.38; 95% CI, 0.92 to 0.15; OR, 0.32; 95% CI, 0.71 to 0.15) periods, however, this association was lost after adjusting for the number of exposures and time since the last exposure (Fig. 3a,b, Supplementary Fig. 4a,b). To further investigate the role of variant type in hybrid immunity, we performed a stratified analysis distinguishing individuals with hybrid immunity who had a prior Omicron infection from those who did not. Among individuals with hybrid immunity and a previous Omicron infection, there was a statistically significant association with protection in the T9–T10 period, whereas this association was not significant in the T10–T11 period (Supplementary Tables 5, 6). Despite this, we did not detect a significant difference in the association of hybrid immunity and protection between T9-T10, and T10-T11 periods (AIC with interaction: 759.9, AIC without interaction: 759.2). In contrast, for individuals with hybrid immunity without a prior Omicron infection, no significant protective effect was observed in either period. Notably, the interaction between variant of prior infection (Omicron vs non-Omicron) and hybrid immunity was significant in the T9–T10 period (AIC with interaction: 999, AIC without interaction: 1008), indicating that in our cohort, the protective effect of hybrid immunity at T9–T10 was primarily driven by prior Omicron infections. Similarly, we found a statistically significant association between hybrid immunity and protection among those with an exposure in the prior 6 months (period in which Omicron was prevalent) during the T9–T10 interval (HR: 0.37; CI: 0.16-0.84).

On the other hand, comorbidities were associated with a higher risk of symptomatic breakthrough infections during the T10–T11 period (HR, 2.7; 95% CI, 1.02-7.13) (Fig. 3a, Supplementary Fig. 4a). In contrast, age, sex, the number of vaccine doses, and having an infection between 190- and 380-days prior were not found to be associated with protection during any of the examined periods.

Antibody levels to previous VoC correlate with protection against Omicron breakthrough infection

We evaluated the relationship between IgG and IgA levels and protective immunity across the three consecutive time periods (T8–T9, T9–T10, and T10–T11). We measured antibodies to the Wuhan strain: RBD, S Fl, S2 and N for all periods, and S1 for T9–T10 and T10–T11. Additionally, during the T8–T9 period, we measured antibodies targeting the RBD from Delta, Alpha, Beta, and Gamma variants, and during the T10-T11 period, antibodies to RBD from Omicron sub-variants (BA.1.1, BA.2, BA.4/5, BQ.1.1, and XBB). We found an association between antibody levels and protection against symptomatic infections for all tested antibodies except IgG to S2 Wuhan at T9-T10 and T10-T11, and IgA to S2 and S Wuhan at T9-T10 (Fig. 5a). Similarly, we found an association between antibody levels and protection against symptomatic plus asymptomatic infections for all antibodies except IgA to any antigen at T9-T10, and IgG to S2, RBD Wuhan and RBD Delta at T10-T11 (Supplementary Fig. 5). In addition, we did not detect any significant interaction between antibody levels and hybrid immunity status on the risk of infection for antibodies targeting the S protein, suggesting that the type of immunity does not substantially modify antibody-mediated protection (Supplementary Tables 7, 8).

IgG and IgA antibody levels to RBD Wuhan were stratified into tertiles, and survival curves plotted for each level (Fig. 5b). Higher levels of IgG to RBD were associated with increased protection against symptomatic infection during the T8–T9 and T9–T10 periods, but not during the T10–T11 period. Similarly, higher levels of IgA to RBD were associated with protection across all three periods. Nevertheless, the differences between tertiles were most pronounced in the T8–T9 period compared to subsequent periods.

Changes in the correlation of antibodies with protection over time

To investigate how antibody protective effect varied across the three time-periods, we assessed the interaction between the time-period and antibody levels using multivariable logistic regression and Cox models (Supplementary Table 9 and 10, respectively). The association between antibody levels and protection against symptomatic and asymptomatic infections was stronger during the T8–T9 than the T9–T10 period for IgG and IgA to RBD and S (Supplementary Table 9). Similarly, for IgG to RBD from the Delta variant, and RBD, S, S2, and N from Wuhan, the association was stronger during the T8–T9 than the T10–T11 period (Supplementary Table 9). The association between antibody levels and protection against only symptomatic infections was also stronger during the T8–T9 compared to the T9–T10 period for IgG to RBD, S2, S and N, and for IgA to RBD, S, and S2 (Supplementary Table 10). Similarly, for IgG to RBD from Delta variant, and RBD, S, and S2 from Wuhan, the association was stronger during the T8–T9 than the T10–T11 period (Supplementary Table 10). No significant differences were found between the T9–T10 and T10–T11 periods.

Notably, when comparing the protective effect of antibody levels against Wuhan RBD during the T8–T9 period with the protective effect of antibody levels against Omicron RBDs during the T10–T11 period, we observed that the association with protection against symptomatic infections was generally stronger for IgG to Wuhan RBD at T8 than for IgG to Omicron RBD at T10 (Supplementary Table 11a). Similarly, the association with protection against all infections was stronger for IgG against Wuhan RBD at T8 than for IgG against RBD from all tested Omicron sub-variants (BA.1.1, BA.2, BA.4/5, BQ.1.1, XBB) at T10 (Supplementary Table 11b, Fig. 4a, Supplementary Fig. 5). Regarding protection against symptomatic infections, the association of IgG against Wuhan RBD was stronger than that of BQ.1.1 and BA.4/5 RBDs (Supplementary Table 11a, Fig. 4a, Supplementary Fig. 5). This suggests that the decline in the protective effect of antibody levels over time is not due to lower antibody levels against the newer Omicron variants but due to a reduced antibody functionality.

a Forest plot of multivariable Cox Regression models. Models were adjusted by age, comorbidities, hybrid immunity, number of infections, number of doses, sex, smoking, and time since last exposure. The color of the dots represents the P value, where black represents < 0.001, dark grey < 0.01, light grey < 0.05, and white non-significant. Error bars represent the 95% confidence intervals (CI). b, c Kaplan–Meier survival curves of risk of breakthrough infection by tertiles of anti-RBD IgG (b) and IgA levels (c). Tertile T1 corresponds to the lowest antibody levels, whereas T3 denotes the highest. Shaded areas represent the 95% confidence intervals. Full-length (Fl), Nucleocapsid (N), Receptor-binding domain (RBD), Spike (S). Kaplan–Meier curves were compared using the log-rank test. T: tertile.

Neutralizing capacity correlates with protection against Omicron breakthrough infection

We found a positive correlation between antibody levels to S antigens and plasma neutralizing activity at T9 (Fig. 5) and T11 (Supplementary Fig. 6), particularly for the anti-Omicron antibodies at T11. When we analyzed the neutralizing activity against the Wuhan, BA.1, BA.4/5, and BQ.1.1 variants at T9 and its relation to protection against symptomatic or all infections during the T9–T10 period in a subset of participants (n = 113) (Tables 2–3, Supplementary Table 12), we found a positive association between neutralizing activity to Wuhan (D614G) and Omicron BA.4/5 variants and protection against symptomatic infections in a multivariable Cox regression model (Table 2, Supplementary Table 12a). Similarly, higher neutralizing activity against Wuhan and Omicron BA.1 and BA.4/5, were associated with protection against symptomatic and asymptomatic infections in a multivariable logistic regression model (Table 3, Supplementary Table 12b).

Pearson correlation R values and correlation p-values (* p < 0.05, ** p < 0.01, *** p < 0.001, ns) are shown in the heatmap.