The initial dataset comprised 1,779,448 index cases from 1,535,288 unique households, and 4,110,051 contacts. For those households in which transmission patterns were likely different from household transmission in the community, and to maintain comparability to the earlier analysis, we applied exclusions in 2 steps, first at the household level, and then at the individual level. For the first step we applied overlapping criteria, excluding households with index cases in children (n = 1,265,196 persons), with residents tested in pillar 1 (n = 99,627), with co-primary index cases (n = 785,669), and with residents who were vaccinated ahead of the national rollout (before January 2021; n = 194,160), leaving 1,035,271 index cases or households and 2,818,017 contacts. For the second step, we excluded persons with indeterminate vaccine status, after which there were 606,720 index cases and 1,440,269 contacts in the analysis dataset. Characteristics of contacts excluded through specific criteria compared with those excluded for indeterminate vaccination status of household members were similar in age, IMD, and geography (Appendix Table 1); the main difference appeared to be an increased proportion of secondary cases in households with multiple index cases.

The median age of index cases was 35 (IQR 25–50) years, and of contacts, 32 (IQR 13–52) years. We restricted index cases to those >16 years of age and eligible for vaccination, whereas we included all contacts. The proportion of female participants was 50.5% in index cases and 49.4% in contacts. We included 122,423 (8.5%) secondary cases among contacts (Appendix Table 2). During February–May 2021, proportions of secondary cases were highest in household contacts of unvaccinated index cases. However, proportions increased in contacts of index cases vaccinated with ChAdOx1 and BNT162b2 vaccines over time. From June onwards, these were highest among contacts of index cases who had received ChAdOx1 vaccine. Proportions of secondary cases among contacts of index cases receiving mRNA-1273 vaccine were consistently low, although data were sparse.

We estimated a strong protective effect for ChAdOx1 vaccine (HR 0.54, 95% CI 0.41–0.69) (Table) and a very strong effect for BNT162b2 vaccine (HR 0.19, 95% CI 0.13–0.28) in the February–May period, which persisted under different inclusion criteria and models. However, we estimated that in June–September the ChAdOx1 vaccine had minimal effect except in model D (Table), which included only unvaccinated contacts, and the protective effect of BNT162b2 vaccine was attenuated (HR 0.74, 95% CI 0.72–0.76). Estimates for mRNA-1273 vaccine, for which there were data only in June–September, were strongly protective.

Figure

Figure. Adjusted hazard ratios for household contacts of COVID patients becoming secondary cases by calendar time and time since vaccination of the index case. A) Data for contacts of index cases...

We estimated the effect of index case vaccination over time and according to time since index case vaccination (Figure). For BNT162b2 vaccine, we noted strong protective effects up to 2 months after the index case was vaccinated, but hazard ratios were attenuated toward the null from 2–3 and 3–4 months after vaccination, particularly in August and September. The pattern of decreasing effect with time since index case vaccination was less clear for ChAdOx1 vaccine; we estimated little protective effect even soon after vaccination from June onwards. Estimates crossed the null to indicate an apparent increased risk of being a secondary case in August–September for contacts of index cases who had received ChAdOx1 vaccine, compared with contacts of unvaccinated index cases.

This analysis provides evidence that persons who have received >2 doses of vaccine >14 days before testing positive for SARS-CoV-2 infection had a reduced likelihood for onward transmission to household contacts in February–May 2021, the period in which Alpha variant was dominant. Of note, this reduction was smaller for BNT162b2 vaccine and all but disappeared for ChAdOx1 vaccine during the subsequent June–September 2021 period, when the Delta variant dominated England. ChAdOx1 vaccine showed reduction in onward transmission only in household contacts who were completely unvaccinated. This finding is consistent with observations of reduced vaccine effectiveness against symptomatic disease for confirmed cases of Delta infection compared with Alpha (2,10–12).

mRNA-1273 vaccine played a smaller role in the initial vaccination rollout of primary courses, so we could only analyze the effects of the vaccine during the latter half of the study period. However, we noted a large and sustained reduction in household transmission in addition to the direct protection of the vaccines in preventing index cases initially and therefore potential for transmission. We also demonstrated a differential reduction in the likelihood of household transmission that is based on the type of vaccine. During the period of Delta dominance, mRNA-1273 vaccine had the greatest association with reduced likelihood of household transmission, followed by BNT162b2 vaccine and then ChAdOx1 vaccine. This finding was consistent with observations of higher vaccine effectiveness with mRNA-1273 and BNT162b2 vaccines compared with ChAdOx1 vaccine for the prevention of symptomatic infection in confirmed Delta variant infections (2). This triangulation with observations from other analyses of protection from disease (2) provides further evidence of a differential effect by vaccine for Delta variant cases.

Our work supplements the growing body of evidence related to waning of vaccine-driven protection. Previous analyses have shown reduced protection of persons from symptomatic infection (10,11). Our findings support another dimension to the waning effectiveness of vaccines: an attenuation in the reduction of onward transmission from index cases vaccinated with Ch4dOx1 and BNT162b2 vaccines to their household contacts (13). We observed this effect for vaccinated index cases who tested positive for COVID-19 >2 months after receiving their second dose. Hazard ratios for becoming a secondary case increased among household contacts of those case-patients, compared with the contacts of index cases who had a shorter interval between their vaccination and confirmed infection. The protective effect against onward transmission conferred by ChAdOx1 vaccine was initially smaller than that for BNT162b2 vaccine and diminished to no effect as time since second dose increased. The protective effect conferred by BNT162b2 vaccine, despite also reducing over time, did not fully disappear. Waning among index cases receiving mRNA-1273 vaccine could not be analyzed in this way because this vaccine was rolled out later in England.

This waning effect was more consistently observed from July 2021 onward, before which the effect was present but less apparent. Early July 2021 demarcates the transition between periods of Alpha and Delta variant dominance in England, suggesting a potential intrinsic difference between the variants and related protection from the vaccine (13).

Because vaccination programs have evolved internationally to provide boosters to vulnerable groups, we may observe a temporary reversal of the waning protection against onward transmission from cases to others. However, this effect requires further evaluation; factors such as time interval between doses, prevalent variants, and vaccine types could all influence any potentially observed effects. Such evaluation would be informative for future assessments of the overall effects of booster programs and therefore influence decision making.

The HOSTED analysis benefits from its very large number of cases and their contacts obtained from data for all reported test-confirmed cases in England; the dataset provides substantial sample sizes that enable large-scale monitoring of secondary infections without specialized data collection. In addition, we conducted several sensitivity analyses, which produced consistent and convergent results. However, the tradeoff is that passive surveillance systems are reliant on testing behaviors instead of targeted case-finding. In passive surveillance, chains of onward household transmission may be missed where contacts did not test, and the secondary attack rate would be underestimated as a result. These missed chains of transmission would be disproportionately present in households where access to testing is lower because of circumstance or choice. This result might explain our observation that contacts of index cases vaccinated with ChAdOx1 with a vaccination-infection interval of >3 months since their second dose had an apparent increase in risk compared with contacts of unvaccinated index cases in August–September 2021. Although vaccination might not reduce onward transmission, increased infectivity caused by vaccination does not appear biologically plausible. This increased risk for infection in the contacts of vaccinated index cases likely indicates differential case ascertainment or risk behaviors between vaccinated and unvaccinated groups, demonstrating a limitation of passive surveillance data.

Without specialized data collection, we were limited by the lack of information on symptoms and case severity, both of which may affect the likelihood of onward transmission (14). As such, our analyses treated all test-confirmed infections uniformly; the granularity that symptom and severity data provide could have highlighted subpopulations for which the vaccine-induced reduction of household transmission was more or less noticeable than for the overall population.

The logic underpinning how cases in a household are classified as index or secondary cases is rooted in the intervals between specimen dates, an approach already established in published analyses. However, the use of symptom onset dates may be more optimal, with symptom onset dates potentially being less affected by behavior and health service access than specimen collection dates.

In conclusion, receipt of 2 doses of mRNA vaccines among index cases reduced transmission to unvaccinated household contacts during a period in which the Delta variant dominated, compared with the Alpha-dominant periods, during which a single dose of mRNA or ChAdOx1 vaccine provided equivalent reductions in transmission. We also observed waning of this protective effect over time since the last dose of vaccine administered. These findings highlight the potential variation in protection from vaccines in relation to transmission effects that are caused by emerging variants and waning protection, and the importance of monitoring the effects of these factors for future SARS-CoV-2 vaccine strategies.