Comparative effectiveness and safety of homologous two-dose ChAdOx1 versus heterologous vaccination with ChAdOx1 and BNT162b2: a cohort analysis

Small trials have suggested that heterologous vaccination with rst-dose ChAdOx1 and second-dose BNT162b2 may generate a better immune response than homologous vaccination with two doses of ChAdOx1. We used linked data from Catalonia (Spain), where those aged <60 who received a rst dose of ChAdOx1 could choose between ChAdOx1 and BNT162b2 for their second dose. Comparable cohorts were obtained after exact-matching 14,325/17,849 (80.3%) people receiving heterologous vaccination to 14,325/149,386 (9.6%) receiving homologous vaccination by age, sex, region, and date of second dose. Of these, 238 (1.7%) in the heterologous and 389 (2.7%) in the homologous groups developed COVID-19 between 1st June 2021 and 11th October 2021. The resulting hazard ratio (95% condence interval) was 0.61 [ 0.52-0.71 ], favouring heterologous vaccination, with a Number Needed to Treat of 94.9 [ 71.8 - 139.8 ]. The two groups had similar testing rates and safety outcomes. Sensitivity and negative control outcome analyses conrmed these ndings. In conclusion, we demonstrate that a heterologous vaccination schedule with ChAdOx1 followed by BNT162b2 was more ecacious than and similarly safe to homologous vaccination with two doses of ChAdOx1. Most of the infections in our study occurred when Delta was the predominant SARS-CoV-2 variant in Spain. These data agree with previous phase 2 randomised trials.

groups had similar testing rates and safety outcomes. Sensitivity and negative control outcome analyses con rmed these ndings. In conclusion, we demonstrate that a heterologous vaccination schedule with ChAdOx1 followed by BNT162b2 was more e cacious than and similarly safe to homologous vaccination with two doses of ChAdOx1. Most of the infections in our study occurred when Delta was the predominant SARS-CoV-2 variant in Spain. These data agree with previous phase 2 randomised trials.

Rationale
The rapid development of COVID-19 vaccines has allowed remarkable progress in the global ght against the SARS-CoV-2 pandemic. As of 27 th October 2021, around half of the world's population had received at least one dose of a COVID-19 vaccine. While 15 vaccines have been approved for use by at least one authority, most countries have received more than four approved vaccines 1 .
The ChAdOx nCoV-19 (ChAdOx1) and BNT162b2 vaccines were among the rst approved by the European Medicines Agency for emergency use in the European Union. Both vaccines were tested in large phase 3 randomised controlled trials and found to be highly effective against symptomatic SARS-CoV-2 infection when given as two doses 2-5 . Follow-up studies have demonstrated their clinical effectiveness against severe disease, including preventing hospitalisations and mortality, overall 6,7 and in previously under-researched populations 8 .
Spanish guidelines initially recommended the use of ChAdOx1 for people aged younger than 60 years due to the under-representation of elderly people in the initial pivotal trials 2 . Key workers were targeted in this initial stage to maximise the impact of vaccination on community transmission 9 . Despite their e cacy, reports of thrombotic events after the rst dose of adenovirus-based COVID-19 vaccines led to recommendations for heterologous vaccination for those vaccinated with a rst dose of ChAdOx1, i.e., many European authorities recommended the use of BNT162b2 for second doses to avoid further exposure to ChAdOx1. A small randomised controlled trial was rapidly conducted that demonstrated better immunogenicity from heterologous vaccination in Spain 10 , but larger studies on clinical effectiveness and safety are urgently needed 11 . The Spanish authorities allowed citizens previously vaccinated with a rst dose of ChAdOx1 to choose between ChAdOx1 and BNT162b2 for their second dose. The majority chose homologous vaccination with two doses of ChAdOx1 12 . In the absence of phase 3/4 randomised controlled trials, this created a natural experiment for studying the comparative safety and effectiveness of these two vaccination schedules. We leveraged routinely collected health data, including electronic medical records linked to vaccination data and laboratory tests, to study the comparative effectiveness and safety of homologous (two-dose ChAdOx1) and heterologous (ChadOx1 followed by BNT162b2) vaccination.  Figure 2 shows that the matched cohorts had similar timings for vaccination and testing over time.
Between 1st June and 11th October 2021, SARS-CoV-2 infections were recorded for 238 (1.66%) people in the heterologous group, equivalent to an incidence rate of 0.13/1,000 person-years, and 389 (2.72%) people in the homologous group, equivalent to an incidence rate of 0.21/1,000 person-years. These rates are equivalent to a hazard ratio of 0.61 [0.52-0.71], favouring heterologous vaccination ( Figure 3 and Table 2). This was equivalent to an absolute risk reduction (ARR) of 0.011 [0.007-0.014] and a Number Needed to Treat of 94.9 [71.8-139.8] in the study period.
No hospital admissions with COVID-19 were identi ed in the heterologous group, compared with 4 (0.03%) hospitalisations in the homologous group. No deaths were seen in either group.

Safety and sensitivity analyses
Primary analyses of 14,325 people per group found only one venous thromboembolism event (0.007%) and one venous thromboembolism with thrombocytopenia event (0.007%), both in the heterologous group. No myopericarditis events were seen in either group (Table 3).
Back pain episodes were used as a negative control outcome. They were recorded at similar frequencies in  (Table 3).

Discussion
Key ndings This is the rst report to date comparing the safety and effectiveness of homologous vaccination against COVID-19 with two-dose ChAdOx1 and heterologous vaccination with rst-dose ChAdOx1 and second-dose BNT162b2. Our primary analysis included over 28,000 people, over 14,000 per group, exactly matched on age, sex, region, and date of second-dose vaccination. In this rich linked cohort, we found a 40% relative risk reduction of SARS-CoV-2 infection (primary outcome) among those on the heterologous vaccination schedule compared with those on the homologous vaccination schedule, despite similar testing rates in the two groups. A total of 95 people would need heterologous (instead of homologous) vaccination to prevent 1 additional case of COVID-19 in the study period.
No safety concerns were identi ed, with only one event (<0.01%) of venous thromboembolism and one event of venous thromboembolism with thrombocytopenia in the heterologous group in the main analysis. One additional event of myopericarditis was observed in 42,845 people receiving homologous vaccination in the 1:5 matched sensitivity analysis.
Sensitivity analyses using 1:2 and 1:5 matching increased the sample size to >37,000 and >50,000 participants respectively and con rmed the safety and effectiveness ndings. The null association between vaccination schedule and our chosen negative control outcome of back pain supported the robustness of our ndings, ruling out residual confounding.

Research in context
Our ndings that heterologous vaccination was more effective than homologous vaccination against COVID-19 agrees with emerging e cacy evidence based on immunological endpoints. Two small, randomised trials have reported higher immunogenicity, characterised by humoral and cellular responses, from ChAdOx1/BNT162b2 than ChAdOx1/ChAdOx1 10,13 . These results were also corroborated by several cohort studies [14][15][16]  Data on the post-marketing safety of heterologous vaccination schedules remain sparse, particularly for rare safety events, with most evidence from evaluations of reactogenicity 10,13-15,20 . Although reactogenicity endpoints are informative for assessing potential vaccine side effects, these trials are underpowered to study rare safety outcomes. Adverse events related to the ChAdOx1 vaccine include the rare (<1/1000 to ≥ 1/10 000) outcome venous thrombosis and the very rare (<1/10 000) outcome vaccine-induced immune thrombosis with thrombocytopenia syndrome 21,22 . Similarly, myocarditis and pericarditis outcomes possibly associated with BNT162b2 are expected to affect around 10 to 24 people per 10 million fully vaccinated people aged ≥ 30 years 23 . In this study with >28,000 participants, we identi ed one venous thromboembolism event and one venous thromboembolism with thrombocytopenia event. The number of events did not increase much in sensitivity analyses including up to >50,000 participants, suggesting that larger studies are needed to investigate these safety signals for heterologous vaccination schedules.

Strengths and weaknesses
Our study has several limitations. The main limitation is the observational nature of our data. However, exact matching on age, region, and date led to a good balance in all observed confounders, including sociodemographics, comorbidity, and medicines use. Our analysis of a negative control outcome (back pain) suggested comparability of the matched cohorts, including unobserved covariates.
As most of our participants were middle-aged adults aged less than 60 years old, our risk-bene t assessment may not be valid for younger or elderly people. Our sample size was insu cient for studying severe COVID-19 outcomes, including hospitalisation and mortality, or rare safety outcomes.
This study also has strengths. The rich, representative linked dataset used allowed a robust analysis of vaccine exposure and outcomes at speed to inform ongoing international vaccination campaigns. Catalonia has a universal healthcare system and uses a centralised, secure data ecosystem with a long track record of research and high-impact publications 8, 24 . The granularity of these data made it possible to control for confounding and test for residual systematic bias. Linkage to additional data sources on RT-PCR and LFT tests allowed for a comprehensive assessment of testing rates and reliable SARS-CoV-2 infection rates. The data collection period covered a time when most cases of COVID-19 in Catalonia were attributable to the Delta variant of SARS-CoV-2 25 ,which is still the predominant variant worldwide. Our data are therefore highly relevant for ongoing global vaccination strategies and future and current third-dose and booster campaigns.

Conclusions
In conclusion, by leveraging the potential of multiple data sources in parallel, our study con rmed that a heterologous vaccination schedule of ChAdOx1 and BNT162b2 was safe and provided better protection

Study design and data sources
We performed a cohort study based on linked routinely collected data available to the Public Health Secretariat of Catalonia. Vaccine exposure was obtained from the Catalan Shared Clinical Records, a database with vaccine data covering the entire Catalan health system and all its vaccination centres.
Additional linked data were obtained from the Catalan database of reverse transcription polymerase chain reaction (RT-PCR) tests and lateral ow tests (LFT) for SARS-CoV-2, from primary-care electronic health records, and from a population-based administrative hospital admissions data (CMBD-AH for its acronym in Catalan language). Data from these linked databases have previously been used for multiple COVID-19 research studies and include information for nearly 90% of the Catalan population 8 .

Participants, cohorts, and follow-up
For our primary analysis, we included all individuals aged 19-59 years old who received a rst dose of the ChAdOx1 vaccine and a second dose of ChAdOx1 (homologous vaccination) or BNT162b2 (heterologous vaccination). We followed participants from the day they received their second dose of either vaccine until an outcome, death, or the end of the study (13 th October 2021).
We excluded people with a previous SARS-CoV-2 infection identi ed by a positive RT-PCR test or LFT and people assigned to one of the 10% of primary-care practices not contributing to our database.
Each participant receiving heterologous vaccination was matched 1:1 to one person receiving homologous vaccination using exact matching by age, sex, general practice centre, and date of second dose. In a sensitivity analysis, we changed the matching ratio to 1:2 and 1:5 to increase sample size.

Study outcomes
The primary outcome for effectiveness analyses was SARS-CoV-2 infection, de ned by the date of the earliest of a positive RT-PCR test or LFT, regardless of symptoms or clinical diagnosis. We measured the number of tests over time regardless of results as an additional outcome to account for diagnostic effort.
Safety outcomes included venous thromboembolism, venous thromboembolism with thrombocytopenia, and myopericarditis within 21 days after the second vaccine dose, based on ChAdOx1 21 and BNT162b2 23 safety reports. Supplementary Table S2 includes the ICD-10-CM codes (international classi cation of diseases, 10th revision, clinical modi cation) used to ascertain when these events occurred.
We analysed the occurrence of a negative control outcome -low back pain -to identify potential unmeasured confounding. Negative control outcomes are health events not causally associated with the exposure of interest, here vaccination.

Additional covariates
Covariates used for confounding assessment included socio-demographics and clinical features assessed at the time of inclusion (day of the second vaccination), as recorded in primary care electronic health records and linked administrative data: age (in years), sex, area of residence, rurality and socioeconomic status, number of RT-PCR tests or LFT performed, pre-existing comorbidities, and long-term medicine use.
Supplementary Table 1 provides ICD-10-CM codes for comorbidities and Anatomical Therapeutic Chemical Classi cation (ATCC) codes used to identify previous medicine use. We assessed socioeconomic status using a validated deprivation index based on census data (MEDEA deprivation index) 26,27 . Rurality of residence was measured, with rural areas de ned by a population <10,000 inhabitants and a density <150 inhabitants/km2, as per regional guidance.

Statistical analysis
Exact matching (1:1) between heterologous ChAdOx1/BNT162b2 vaccination and homologous two-dose ChAdOx1 was performed using the following variables: age (+/-2 years), sex, general practice centre, and day of the second vaccination (+/-2 days). As a sensitivity analysis, we generated additional study populations and repeated all analyses after matching with 1:2 and 1:5 ratios. We assessed confounding due to known variables by measuring covariate imbalance as the standardised mean difference (SMD) of all covariates listed above. We considered SMD>0.1 to be imbalanced 28 .
We plotted time-to-event Kaplan-Meier estimates strati ed by vaccine exposure (homologous vs heterologous). Absolute risk reduction (ARR) was estimated as the difference in cumulative incidence of  Vaccine uptake and testing rates strati ed by vaccination schedule