Introduction

Vaccination is an important tool to control the COVID-19 pandemic. In Portugal COVID-19 vaccines started to be deployed on December 27, 2020, prioritising individuals with a higher risk of exposure and/or COVID-19 complications [1]. On February 3, 2021,mass vaccination was implemented starting with those aged 80 or more years and then by decreasing age groups [1]. The first vaccine to be approved and administered was mRNA Pfizer/BioNTech (Comirnaty) (21/12/2020), followed by Moderna (Spikevax) (06/01/2021), AstraZeneca (Vaxzevria) (29/01/2021) and Janssen vaccine (COVID-19 Vaccine Janssen) (11/03/2021) [2].

The Vaxzevria vaccine was first given to those aged less than 65 years [3], but the recommendation changed in April 2021, to be administrated to those aged more than 60 years old [4]. A single-dose Janssen vaccine was recommended to all males aged 18 and more years old and females aged 50 or more years [5]. In early October 2021, vaccine coverage in the Portuguese population reached a milestone of 84% (complete schedule) and was estimated to be higher than 95% for those aged 65 and more years [6].

Coinciding with the start of the vaccination campaign, Portugal faced the third wave of COVID-19. In January 2021, the 14-day incidence rate reached the peak of 1,667 cases per 100,000 inhabitants and the transmissibility (Rt) was over 1 [7]. Given the health care rupture, and to control the epidemic, a lockdown was implemented on January 15^th, 2021. At this point, the predominant circulating variant was the B.1.7.1 (alpha variant), which was progressively replaced by the B.1.617.2 (delta variant) that became predominant in June 2021. Data on the variants of concern (VOC) surveillance system indicate that by the end of May 2021 the proportion of circulating variants were 87.7% and 4.8%, for alpha and delta variants respectively [8]. By the end of June 2021, these figures were 9.8% and 88.2%, for alpha and delta variants, respectively [9]. Along with the end of the lockdown and the increasing frequency of the Delta variant, an increasing trend of the epidemic activity was observed in all Portuguese regions affecting mainly young adults and adolescents [7]. There were several changes in the national testing strategy over time, but until October 13^th, 2021, the recommendation was still in place to test vaccinated persons in the same circumstances as for non-vaccinated (suspected or contact with COVID-19, regular testing in high risk of exposure or high-risk populations, regular testing in specific settings such schools and workplaces).

Post-licencing COVID-19 vaccine studies are needed to measure the protection provided by vaccination in real-world conditions. They add to the information provided by clinical trials, as larger and more heterogeneous populations are included. Early evaluations focused mainly on the performance of Comirnaty, and Vaxzevria adenoviral vaccines, showing high short-term effectiveness against SARS-CoV-2 infections and severe outcomes in the general population and older adults [10, 11]. VE against COVID-19-related hospital admissions and mortality in older adults populations varied between 94–97% for the completed vaccination schedule in December 2020—April 2021 [10–13]. Following the emergence of the Delta SARS-CoV-2 VOC, studies showed a decrease in VE for both mRNA and Vaxzevria vaccines for symptomatic infections in the general population [14–16], but sustained VE estimates against hospitalizations [14–17] and COVID-19-related mortality [15].

Immunological studies suggest a decrease of IgG and neutralizing antibody titers 3–6 months after completion of the two-dose schedule [18, 19]. Continued monitoring of VE over time is important to provide input on a potential vaccine waning effect, as has been observed with vaccines against other respiratory pathogens [20, 21] and COVID-19 vaccines [16].

In the context of multiple vaccines and variant epidemiological setting, it is important to assess the protection conferred by the COVID-19 vaccines in use. Moreover, monitoring over time allows detection of potential decay of VE and may be important for the decision on the vaccination strategy.

This study aimed to estimate the effectiveness of COVID-19 vaccines against symptomatic infections, COVID-19-related hospitalizations and deaths in Portuguese adults aged 65 years old or more from February-September 2021

Methods

Results

Participants

Database extraction on 16^th September 2021 contained 2,117,002 individuals aged 65 to 79 years and 923,450 individuals aged ≥80 years. After excluding, according to previously described criteria, cohorts of sizes 1,414,909 aged 65 to 79 years (S1 Fig) and 470,023 aged ≥80 years were constituted (S2 Fig).

Distribution of vaccines varied by age cohort. For the 65 to 79 years cohort, the Corminaty vaccine was administrated to 644,959 individuals (45.6%) and the second most frequent vaccine was from Vaxzevria (499,770 persons, 35.3%). Spikevax (115,217, 8.1%) and Janssen (40,484, 2.9%) were less frequently administered in this age cohort (S1 Fig). The majority of adults aged 80 and more years, were vaccinated with the Corminaty (379,285, 80.7%). The other vaccines were also available for this age group but were less frequently used (Spikevax: 56,037 individuals, 11.9%; Vaxzevria: 8,509 individuals, 1.8%; Janssen: 2,191 individuals, 0.5%) (S2 Fig).

The unvaccinated were older than those vaccinated with Vaxzevria in the 80 and more yo cohort (median of 86 vs 82, respectively). In both cohorts, the proportion of individuals without any chronic condition was higher in the unvaccinated group than in the vaccinated with Vaxzevria or mRNA vaccines (65 to 79 yo cohort: 48.9% vs 22.9% or 29.3% or 22.7%; 80 and more yo cohort: 37.2% vs 11.2% or 14.3% or 10.2%) (S1 and S2 Tables).

During the study period, in the 65 to 79 years cohort, a total of 4,282 symptomatic infections, 338 hospitalizations, and 172 deaths were observed (Table 1), compared to a total of 3,078 symptomatic infections, 854 hospitalizations, and 764 COVID-19-related deaths observed in the 80 and more years cohort (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Symptomatic SARS-CoV-2 infection, COVID-19 related hospitalizations and deaths, incidence, hazard ratios and vaccine effectiveness by mRNA (Comirnatyor Spikevax) and Vaxzevria vaccination status for individuals aged 65–79 years, Portugal, March–September 2021 (n = 1,414,909).

https://doi.org/10.1371/journal.pone.0274008.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Symptomatic SARS-CoV-2 infection, COVID-19 related hospitalizations and deaths, incidence, hazard ratios and vaccine effectiveness by mRNA (Comirnaty or Spikevax) vaccination status for individuals aged ≥80 years, Portugal, February–September 2021 (n = 470,023).

https://doi.org/10.1371/journal.pone.0274008.t002

Discussion

Main results indicate that VE varied according to the outcome and vaccine type. VE estimates against symptomatic infections range from 43% to 65% in the 65–79 years cohort and 53% in the 80 and more years cohort. These estimates were lower than VE estimated against severe COVID-19 outcomes–hospitalization and death (higher than 89% in the 65–79 years cohort and higher than 76% in the 80 and more years cohort). These results suggest a high level of vaccine protection against severe COVID-19 outcomes in the 65 and more years population that is more prone to COVID-19 related complications.

Compared to other studies we found some differences. A living review, which included studies published up to 25 August 2021, allowed a meta-analysis and indicated that VE against symptomatic infection was 75.7% (95%CI: 69.3 to 80.8) [27]. The differences between our overall VE estimates and other studies might be explained by different study periods and consequently different epidemiological and virological contexts.

Our study comprises a period of high incidence observed at the beginning of the vaccination campaign, corresponding to the third COVID-19 peak in January-February 2021 (for the 80 and more years cohort), but also the fourth wave associated with the circulation of SARS-CoV-2 Delta variant during May-September 2021. As described in the living review, when estimating VE comparing Delta vs Alpha, estimates tended to be 10 to 20% lower for less severe COVID-19 infection but not different for severe outcomes [27].

Considering VE against COVID-19 related death, we could find fewer studies that focused on this severe outcome, as the majority reported VE against severe outcomes which included hospitalizations or death and thus are not comparable with our results [14, 25]. Results from one study conducted in the population aged 60 or more years in Scotland, reported high VE against of 90% (95% CI: 84 to 94) for Vaxzevria vaccine and 87% (95% CI: 77 to 93) for Corminaty vaccine [28]. Our VE estimates against death are in line with these estimates and also support high protection conferred by both vaccines when considering this fatal outcome.

For the 60–79 years cohort, we were able to estimate VE for the different outcomes and two types of vaccines: mRNA (Comirnatyand Spikevax) and adenovirus-based vaccine (Vaxzevria). A similar analysis was not possible for the 80 and more cohort or the Janssen vaccine, as the proportion of vaccinated individuals was too low. For the 65–79 years cohort and considering symptomatic infection, VE estimates indicate lower protection conferred by the adenovirus vaccine (Vaxzveria VE of 43%; 95%CI 37 to 49) vs mRNA (Comirnaty or Spikevax) VE of 65% with 95%CI of 62 to 68. Other brands or type-specific vaccine studies have also reported similar lower VE against symptomatic infection of Vaxzevria when compared to Corminaty vaccine [29, 30].

After complete vaccination schedule, we observed a decay of the VE with time, particularly against symptomatic infection. To highlight that we observed a decaying the 80 and more years cohort, where VE waining was observed for both symptomatic infection and hospitalizations.

The drop over time of VE against symptomatic infection has also been reported by several authors [14–16] but the decay of VE against hospitalization was not reported in that period [14–17]. In part, this result may be explained by an insufficient time of follow-up. Our results indicate a significant decay of VE against hospitalizations in the 80 and more years cohort, from 83% (the peak achieved 14–41 days after complete vaccination) to 63% (after 124 days of complete vaccination).

One valid argument is that this decay may be attributable to the increased circulation of the Delta variant which could have increased the effect on the older aged cohort. However, waning of vaccine effect with time cannot be ruled out. As observed by Tartof et al. [16], the decay on VE against symptomatic infection was observed in all population (even in younger-aged cohorts with less than <45 years old) and this pattern was observed even before the circulation of Delta variant. This study was not focused on VE against specific variants and as such, this hypothesis needs confirmation.

The present study has some limitations, namely related to the datasets used and their quality. For instance, the main dataset used to link data was the NHSU, which contains the unique health number attributed to each individual residing in Portugal. However, the NHSU database could have update issues, and occasional and temporary registrations of NHS users, such as immigrants and asylum seekers, that could artificially increase the number of individuals in a given cohort. To overcome this limitation, several restriction criteria were included, and from the initial extraction, a total of 1,178,975 registries were excluded from the analysis. Final cohort age and sex distribution were comparable to National Statistics estimates for individuals aged 65 and more years [31]. In terms of the prevalence of chronic conditions, final cohorts were also comparable to National estimates (INS 2019).

Another potential limitation is related to information bias. Differential testing guidelines between vaccinated and unvaccinated may affect the opportunity of infection ascertainment in vaccinated and unvaccinated. In Portugal, during the study period, testing guidelines were not different for vaccinated and non-vaccinated individuals. Nevertheless, we cannot rule out the hypothesis that vaccinated individuals could have had a differential (due to acceptance) testing performance if asymptomatic or if they had any contact with the COVID-19 case. Under this hypothesis, there is a potential bias due to different infection ascertainment in vaccinated versus unvaccinated, which might have contributed to overestimating VE against infection. This potential bias is less probable for severe outcomes such as hospitalization. Nevertheless, when we compared the frequency of testing during 2021 between the groups compared in this study we did not observe relevant differences.

In what concerns the COVID-19 hospitalizations outcome, the delay in the update of information in the registries regarding discharge diagnosis might contribute to underrepresenting this specific outcome. Although we didn´t find any reason for having a different delay in the registry of the discharge information between vaccinated and non-vaccinated, there is a possibility of a differential bias if vaccinated had a shorter hospitalization stay when compared to non-vaccinated.

Finally, there is potential for residual confounding bias. Adjustment for confounding was made for several variables, most of them very relevant from a clinical and epidemiological point of view. Nevertheless, residual confounding bias could remain mainly due to the absence of information on non-pharmaceutical protective measures such as the use of facemask, hand washing, and social distancing, all of which could be associated with both vaccine uptake and risk of infection.

Besides the impossibility of excluding the presence of selection or information bias driven by health-seeking behavior, our results are generally in line with results reported in studies performed elsewhere. Finally, as our cohort comprises all the Portuguese NHS users older adults population, we consider our results to represent a description of the direct protection conferred by the COVID-19 vaccines used within the target population, and a good approach to monitoring VE along time and waning of the vaccine protection.

Supporting information