We conducted a prospective, population-based cohort study of residents in the region of Aragon, Spain. Participants were all of the users of the Aragon Health Service, >16 years old of age, who had no evidence of previous SARS-CoV-2 infection, confirmed by reverse transcription PCR (RT-PCR), antigen test, or immunoglobulin G test for SARS-CoV-2 infection at any time before December 27, 2020. We included all residents registered in the Aragon Healthcare System Users Registry (AHSUR) who met the eligibility criteria as of December 31, 2020. AHSUR consists of periodically updated basic demographic data from users of the Aragon Healthcare Service, the public healthcare provider in Aragon. AHSUR contains data from 89% of Aragon inhabitants. We based the study on data collected during December 28, 2020–May 31, 2021.

Figure 1

Figure 1. Flowchart of cohort evolution for study of coronavirus disease vaccines in preventing confirmed severe acute respiratory syndrome coronavirus 2 infection, Aragon, Spain, January–May 2021. *Participants vaccinated with the AZ vaccine...

We prospectively followed a cohort of 964,258 people >16 years of age from the general population, corresponding to 72.5% of the population of Aragon; the size and exposure status of the cohort evolved across the study period (Figure 1). We stratified participants’ vaccination exposure by their demographic characteristics (Table 1). As of May 31, 2021, among the participants, 242,142 had been vaccinated with >1 dose of the Pfizer-BioNTech vaccine and 212,419 with 2 doses; 32,522 participants had been vaccinated with at least 1 dose of the Moderna vaccine and 15,660 of them with 2 doses; and 97,492 participants had been vaccinated with 1 dose of the Oxford-AstraZeneca vaccine; 592,102 participants had not yet been vaccinated. We observed differences in the number of Pfizer-BioNTech, Moderna, and Oxford-AstraZeneca vaccines doses administered over the study period, which occurred because of different EMA approval times, vaccine doses available over time, and prioritizing of groups considered for earlier vaccination, specifically persons >75 years old and residents and workers in nursing and residential homes (Table 1). Over the study period, 11,557 (1.2%) participants dropped out of the study; we recorded lost participants by demographic characteristics and causes of withdrawal (Tables 1, 2).

The 592,102 unvaccinated participants had 25,767 SARS-CoV-2 infections and an IR of 1.41/1,000 person-weeks. The 242,142 participants vaccinated with 1 dose of the Pfizer-BioNTech vaccine had 463 infections (IR 0.86) and the 212,419 with 2 doses had 280 infections (IR 0.23). The 32,522 participants vaccinated with 1 dose of the Moderna vaccine had 28 infections (IR 0.31) and the 15,660 with 2 doses had 18 infections (IR 0.21). The 97,492 participants vaccinated with 1 dose of the Oxford-AstraZeneca vaccine had 230 infections (IR 0.55).

SARS-CoV-2 Infection Cumulative Risk Curves

Figure 2

Figure 2. Cumulative risk curves (1 minus the Kaplan-Meier risk) of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection for 3 coronavirus disease vaccines, Aragon, Spain, January–May 2021. A) BioNTech-Pfizer BNT162b2 mRNA,...

For unvaccinated participants, the risk for SARS-CoV-2 infection rose to 2% at day 44 and to 4% at day 154 of follow-up. For participants who received 1 dose of the Pfizer-BioNTech vaccine, the risk rose to 1% at day 40 of follow-up, but remained <1% during the entire follow-up period (120 days) for those with 2 doses (Figure 2, panel A). For participants who received 1 dose of the Moderna vaccine, risk remained <0.5% during the entire follow-up time (120 days) and for participants vaccinated with 2 doses, the risk rose from 0% to 0.5% during days 30–71, then remained at 0.5% until the end of follow-up (day 90; Figure 2, panel B). For participants who received 1 dose of the Oxford-AstraZeneca vaccine, risk rose to 0.9% after 80 days of follow-up (Figure 2, panel C).

In the general population, our findings showed an effectiveness of 3 different vaccines against SARS-CoV-2 infection, but with lower efficacy estimates than from clinical trials and other VE studies. We found 20.8% VE after 1 dose of the Pfizer-BioNTech vaccine and 70.0% after 2 doses; for the Moderna vaccine, these numbers were 52.8% VE after 1 dose and 70.3% VE after 2 doses, and for the Oxford-AstraZeneca vaccine, 40.3% after 1 dose.

For the Pfizer-BioNTech and Moderna vaccines, these values were lower than those in other observational studies, which had ranges of 61.9%–80% VE after 1 dose and 90%–96% VE >7 days after 2 doses (8,9,27–29). These differences could possibly be explained by the population-wide design of our study, which included a higher percentage of elderly persons in the Pfizer-BioNTech–vaccinated group than in the other studies. In contrast, our results showed a higher VE after 2 doses of the Pfizer-BioNTech vaccine than the 65% VE found in another study (30), probably because they used a different approach for estimating VE that included only close contacts of positive cases and assigned every person in the cohort the same observation period and as a result vaccinated and unvaccinated participants most likely experienced similar exposure to SARS-CoV-2.

Our findings indicated a higher VE (52.8%) after 1 dose of the Moderna vaccine than after 1 dose of either the Pfizer-BioNTech or Oxford-AstraZeneca vaccines and similar VEs after 2 doses of both the Moderna and Pfizer-BioNTech vaccines. However, our results did not reach the VE estimates of 83% after 1 dose and 82% after 2 doses of Moderna vaccine found in another study (28). The small sample size in that study, which only included healthcare personnel and other essential workers, might explain these differences in VE. However, as in that study (28), VE after 1 and 2 doses of the Moderna vaccine were also very close.

Safety concerns resulted in the suspension of the Oxford-AstraZeneca vaccine before anyone in our cohort received a second dose, and therefore we estimated VE only after 1 dose (40.3%), similar to the 44% VE after 1 dose of the Oxford-AstraZeneca vaccine in another article (30). In contrast, another study found a VE of 60% against symptomatic disease after a single dose of the Oxford-AstraZeneca vaccine in adults >70 years of age, as expected because of the study’s more severe outcome measures and exclusively elderly population (14).

Cumulative risk curves of SARS-CoV-2 infection show that the cumulative risk of infection in unvaccinated participants rose to 4% at day 154 of follow-up whereas the risk remained <1% during the entire follow-up period (120 days) in fully Pfizer-BioNTech–vaccinated participants, results consistent with those from a nationwide study (8). Risk remained <0.5% in participants vaccinated with 1 dose of the Moderna vaccine during the entire follow-up time (120 days) and <1% during the entire follow-up time (90 days) in fully vaccinated participants. In the participants with 2 doses of the Moderna vaccine, the slight increase in risk from day 30 onwards might be explained by the relatively small number of participants from our cohort who were vaccinated with the second dose and reached long follow-up times (>50 days), which can cause instability of estimates for prolonged follow-up times. For the Oxford-AstraZeneca vaccine, the difference in risk between unvaccinated participants and those vaccinated with 1 dose (2.5% vs. 0.9% at day 80 of follow-up) highlights the VE after 1 dose of the Oxford-AstraZeneca vaccine.

One limitation of our study was losses to follow-up because of administrative leaves from AHSUR. Participants lost to follow-up were statistically different from the studied cohort. Nevertheless, they represent only 1.2% of the initial population, which limited the magnitude of this bias. Timing of vaccine rollout also varied between priority groups, targeted for earlier vaccination, and the general population. This difference may have affected the results by adding more variability, particularly because Pfizer-BioNTech was mostly used in population ≥75 years of age, who were vaccinated earlier, whereas Oxford-AstraZeneca was mostly used in general population, who were vaccinated at a later time.

Research has documented that the proportion of symptomatic infections in vaccinated persons is lower than in unvaccinated ones because vaccination prevents symptoms (28). Therefore, studies based on symptomatic persons (1–7,11,13,14) underestimate the total infection rate in vaccinated persons to a greater extent than in unvaccinated ones and consequently overestimate VE. Our study included all confirmed symptomatic and asymptomatic SARS-CoV-2 infections, and thus it would be expected that VE would be lower than in studies only including symptomatic disease and our VE estimates more relevant to transmission control, because in real-world conditions, symptomatic and asymptomatic infections coexist and both contribute to transmission.

Similarly, according to COVID-19 detection and surveillance guidelines in Spain and Aragon (25,26), tests were administered less frequently to asymptomatic than to symptomatic persons, resulting in underdetection of asymptomatic infections. This bias was reduced because underdetection occurred in both vaccinated and unvaccinated persons but could still lead to overestimating VE. On the other hand, also following the detection program guidelines, tests were administered to close contacts regardless of their vaccination status, which reduced the chance of detection bias in our study. However, routine screenings carried out in nursing and residential homes could have altered our findings if there were more screenings in vaccinated than in unvaccinated participants. The role of dominant variants of concern in the transmission was unknown at the time of our data analyses. The rapid circulation of these variants may have introduced confounding, but it was minimized by including weekly variability, and therefore calculated VE estimates represent a summarized measure against all variants, adjusted by incidence. Practical factors such as hygiene and social distance measures might also have affected the estimates of VE.

Our study shows great strength in statistical power because of the large population cohort and use of a refined methodology. Risk of infection differed between participants according not only to vaccination status but also to the evolution of the epidemic curve. For this reason, we used an approach of weekly repeated measures, adjusted by WCI in each primary care service area.

In conclusion, we found effectiveness against SARS-CoV-2 infection for Pfizer-BioNTech, Moderna, and Oxford-AstraZeneca vaccines to be lower than efficacy estimates from clinical trials and other VE studies. Even if high vaccination coverages are reached in the general population (31,32), effectively minimizing transmission opportunities might be limited, because age groups of persons <12 years of age were not being immunized at the time of our data gathering. Even so, reaching high vaccination coverage is important to decrease SARS-CoV-2 transmission in the general population.