key: cord-0720089-5n4phjno authors: McKeigue, P. M.; McAllister, D.; Robertson, C.; Stockton, D.; Colhoun, H. title: Reinfection with SARS-CoV-2: outcome, risk factors and vaccine efficacy in a Scottish cohort date: 2021-11-24 journal: nan DOI: 10.1101/2021.11.23.21266574 sha: ac936d285d76da7d87e57bbac6c8d29411970704 doc_id: 720089 cord_uid: 5n4phjno Background -- The objective of this study was to investigate how protection against COVID-19 conferred by previous infection is modified by vaccination. Methods -- In a cohort of all 152655 individuals in Scotland alive at 90 days after a positive test for SARS-CoV-2 (confirmed by cycle threshold < 30, or two tests) followed till 22 September 2021, rate ratios for reinfection were estimated with calendar time or tests as timescale. Findings -- Rates of detected and hospitalised reinfection with COVID-19 while unvaccinated were respectively 6.8 (95% CI 6.4 to 7.2) and 0.18 (95% CI 0.12 to 0.25) per 1000 person-months. These rates were respectively 68% and 74% lower than in a matched cohort of individuals who had not previously tested positive. Efficacy of two doses of vaccine in those with previous infection was estimated as as 84% (95 percent CI 81% to 86%) against detected reinfection and 71% (95 percent CI 29% to 88%) against hospitalised or fatal reinfection. The rate of detected reinfection after two doses of vaccine was 1.35 (95% CI 1.02 to 1.78) times higher in those vaccinated before first infection than in those unvaccinated at first infection. Interpretation -- The combination of natural infection and vaccination provides maximal protection against new infection with SARS-CoV-2: prior vaccination does not impair this protection. Although cohort studies have shown that previous SARS-Cov-2 infection confers pprotection against reinfection 31 [1] , few studies have examined how this protection is modified by vaccination before or after first infection. or by 32 the emergence of new variants. 33 In the UK and most other countries those with evidence of previous SARS-CoV-2 infection are included 34 in programmes for vaccination against COVID-19 and are not exempt from mandatory vaccine certification 35 where this is required. This policy has been questioned as an unnecessary use of scarce vaccines, or as having an 36 unfavourable risk-benefit ratio because adverse reactions to vaccine have been reported to be more common in 37 the those with previous infection [2] . The evidence base for these policies depends on comparisons of rates of CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 24, 2021. ; https://doi.org/10.1101/2021.11.23.21266574 doi: medRxiv preprint Electronic Communication of Surveillance in Scotland database (ECOSS). This database captures all tests in 48 Pillar 1 (NHS laboratories testing those with a clinical need and workers in health or social care) and Pillar 49 2 (Lighthouse laboratories testing the wider population from September 2020 onwards). Only the Lighthouse 50 laboratories report Ct values. To exclude possible false-positives, definite first infection was defined as a positive 51 test with Ct < 30 for both N and ORF genes, or a second positive test within fourteen days of the first. 52 Entry date to the cohort at risk of reinfection was 90 days after first testing positive. Exit date was the 53 earliest of date of reinfection (defined as any positive test after entry date), date of death, or end of follow-up 54 period (22 September 2021). 55 Construction of comparison cohort 56 To compare reinfection rates with first infection rates, we constructed a comparison cohort of individuals 57 who had not previously tested positive for SARS-CoV-2. As all test-positive cases cases in Scotland had been 58 sampled in the REACT-SCOT case-control study [4] , for each individual in the cohort at risk of reinfection up to 59 ten controls were available who were sampled from the general population, matched for age, sex and general 60 practice, alive on the day that the case first tested positive, and had not tested positive by that date. As with 61 the cohort at risk of reinfection, this comparison cohort was restricted to those who were still under observation 62 (without having tested positive) at 90 days from the date that they were first sampled as controls. Figure S1 shows that the number of individuals in the cohort at risk of reinfection increased 112 rapidly after mid-December 2020, around 90 days after Pillar 2 testing began. The comparison cohort comprised 1177827 individuals matched for age, sex and general practice to those in 114 the cohort at risk of reinfection, with average follow up of 5.6 months. Of these, 1132118 were unvaccinated 115 when first sampled. The average testing rate while unvaccinated was higher in the cohort with prior infection 116 (0.56 per month) than in the comparison cohort (0.27 per month). In the cohort at risk of reinfection, there were 1070 detected reinfections while still unvaccinated, of which 118 28 were hospitalised or fatal. The corresponding rates per 1000 person-months were 6.8 (95% CI 6.4 to 7.2) 119 for reinfections and 0.18 (95% CI 0.12 to 0.25) for hospitalisation. In the comparison cohort, there were 36488 120 4 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 24, 2021. ; https://doi.org/10.1101/2021.11.23.21266574 doi: medRxiv preprint detected first infections while still unvaccinated of which 1211 were hospitalised or fatal. The corresponding 121 rates per 1000 person-months were 21.1 (95% CI 20.8 to 21.3) for reinfection and 0.7 (95% CI 0.66 to 0.74) for 122 hospitalisation. This equates to a reduction in risk of 68% for detected new infection, and 74% for hospitalisation. Because the testing rate was lower in the comparison cohort than in the cohort at risk of reinfection, this 124 comparison of detected infections is likely to underestimate the protection against new infection that is conferred 125 by previous infection. For the subset of hospitalised unvaccinated cases whose SMR01 records were available, the proportion admitted Table S1 shows how crude testing rates in those who were unvaccinated at time of first infection varied with 140 vaccination status and other covariates. This gives some idea of the likely direction of biases in estimates of the 141 effects of these covariates on reinfection rates. Testing rates were higher in fully vaccinated than in unvaccinated 142 or partly vaccinated time intervals, and the associations with other covariates varied by vaccination status. Testing rates were higher in women than in men, and higher in teachers and health care workers than in other 144 occupations. In unvaccinated individuals testing rates were higher in care home residents than those living 145 independently, higher in older than in younger age groups, and higher in those with clinical risk conditions 146 compared with those with no risk conditions. 148 Figure 1 compares the coefficients estimated with each of the two alternative Cox regression models for 149 detected reinfection. Table S2 tabulates the covariate distributions and coefficients. In the model with calendar timescale, several factors that were associated with higher testing rates -female 151 sex, care home residence, and occupation -were associated with detected reinfection. In the model with tests 152 as timescale these associations were reduced or reversed, consistent with detection bias as the explanation for 153 the effects estimated with calendar timescale. In both models vaccination was associated with lower rates of 154 reinfection. With tests as timescale the efficacy of vaccination against reinfection was estimated as 64% (95 155 percent CI 60% to 68%) for one dose of vaccine and 84% (95 percent CI 81% to 86%) for two doses. With tests as 156 5 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 24, 2021. ; https://doi.org/10.1101/2021.11.23.21266574 doi: medRxiv preprint timescale the estimated efficacy of vaccination was higher (rate ratios were lower) than with calendar timescale, 157 suggesting that the efficacy estimated from the model with calendar timescale may have been biased downwards 158 by lower detection rates in unvaccinated individuals. Teaching and health care occupations were associated with increased risk of detected reinfection but the rate 160 ratios were lower in the model with tests as timescale than in the model with calendar timescale, suggesting that 161 much of this excess may be driven by increased testing. In both models clinical risk conditions were associated 162 with increased risk of reinfection; a possible explanation for this is that clinical risk conditions are associated Table 1 shows that hospitalised or fatal reinfection was associated with older age (median age 54 years in 168 cases versus 38 years in noncases) and clinical risk conditions. The numbers of hospitalised cases were too small 169 for associations with occupation to be estimated reliably. The efficacy of two doses of vaccine against hospitalised 170 reinfection was estimated as 71% (95 percent CI 29% to 88%). The main limitation is that without regular scheduled testing, estimates of association with detected reinfection 192 are subject to ascertainment bias. Because testing rates are lower in unvaccinated than in vaccinated individuals 193 in this cohort, the efficacy of vaccination is likely to be underestimated by a model with calendar timescale. 194 We have attempted to overcome this by comparing two alternative models: a conventional Cox regression with 195 calendar timescale, and a Cox regression with tests as timescale to adjust for differential testing rates. The CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this this version posted November 24, 2021. ; https://doi.org/10.1101/2021.11.23.21266574 doi: medRxiv preprint healthcare provider reported a 32% reduction but the numbers of cases were small and the confidence interval 229 was wide [16] . Neither study reported or adjusted for differences in testing rates. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 24, 2021. ; https://doi.org/10.1101/2021.11.23.21266574 doi: medRxiv preprint 13 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 24, 2021. ; https://doi.org/10.1101/2021.11.23.21266574 doi: medRxiv preprint PF, patient-facing 14 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 24, 2021. ; https://doi.org/10.1101/2021.11.23.21266574 doi: medRxiv preprint 15 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 24, 2021. ; https://doi.org/10.1101/2021.11.23.21266574 doi: medRxiv preprint A Systematic Review of the Protective Effect of Prior SARS-CoV-2 274 Vaccinating people who have had covid-19: Why doesn't natural immunity count in the US? 276 Centers for Disease Control and Prevention. Common Investigation Protocol for Investigating Suspected 278 SARS-CoV-2 Reinfection 268 We thank Jen Bishop, Bob Taylor and David Caldwell for undertaking the data extraction and linkage.