key: cord-0832439-inetf5c6 authors: Barchuk, A.; Cherkashin, M.; Bulina, A.; Berezina, N.; Rakova, T.; Kuplevatskaya, D.; Stanevich, O.; Skougarevskiy, D.; Okhotin, A. title: Vaccine Effectiveness against Referral to Hospital and Severe Lung Injury Associated with COVID-19: A Population-based Case-control Study in St. Petersburg, Russia date: 2021-08-23 journal: nan DOI: 10.1101/2021.08.18.21262065 sha: 81c133a0e2425ca33d7a17486675690572f47a04 doc_id: 832439 cord_uid: inetf5c6 Background: Results of a randomised trial showed the safety and efficacy of Gam-COVID-Vac against COVID-19. However, compared to other vaccines used across the globe, the real-world data on the effectiveness of Gam-COVID-Vac, especially against the disease caused by Delta variant of concern, was not available. We aimed to assess the effectiveness of vaccination mainly conducted with Gam-COVID-Vac in St. Petersburg, Russia. Methods: We designed a case-control study to assess the vaccine effectiveness (VE) against lung injury and referral to hospital. Self-reported vaccination status was collected for individuals with confirmed SARS-CoV-2 infection who were referred for initial low-dose computed tomography triage in two outpatient centres in July 3 -- August 9, 2021 in St. Petersburg, Russia. We used logistic regression models to estimate the adjusted (for age, and triage centre) VE for complete (>14 days after the second dose) and partial vaccination (>14 days after the first dose). We estimated the VE against referral for hospital admission, COVID-19-related lung injury assessed with LDCT, and decline in oxygen saturation. Results: In the final analysis, 13,894 patients were included, 1,291 (9.3%) of patients met our criteria for complete vaccination status, and 495 (3.6%) were referred to hospital. In the primary analysis, the adjusted VE against referral to hospital was 81% (95% CI: 68-88) for complete vaccination. The effect of the partial vaccination was not statistically significant. The VE against referral to hospital was more pronounced in women (84%, 95% CI: 66-92) compared to men (76%, 95% CI: 51-88). Vaccine protective effect increased with increasing lung injury categories, from 54% (95% CI: 48-60) against any sign of lung injury to 76% (95% CI: 59-86) against more than 50% lung involvement. A sharp increase was observed in the probability of hospital admission with age for non-vaccinated patients in relation to an almost flat relationship for the completely vaccinated group. Conclusions: COVID-19 vaccination was effective against referral to hospital in patients with symptomatic SARS-CoV-2 infection in St. Petersburg, Russia. This protection is probably mediated through VE against lung injury associated with COVID-19. Vaccination proved to be an effective pharmaceutical control measure during the COVID19 pandemic. More than 4 billion doses have been already administered, but realworld evidence is not equally available for all vaccines used globally. Even 1 . 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 preprint this version posted August 23, 2021. ; https://doi.org/10.1101/2021.08.18.21262065 doi: medRxiv preprint though the results from clinical trials of GamCOVIDVac [1, 2] suggested safety and efficacy similar to other COVID19 vaccines used in populationbased programmes across the globe [3, 4] , this finding was met with scepticism and criticism [5] [6] [7] [8] [9] . Despite adoption of GamCOVIDVac in 49 countries, the independent safety data for this vaccine is available only from San Marino and Argentina [10, 11] . Further lacking is realworld evidence on GamCOVIDVac vaccine effectiveness (VE) from populationbased studies. A casecontrol study design and its modifications are the core methods to assess the effectiveness of vaccination in realworld settings [12] . Multiple casecontrol studies of COVID19 vaccines have been published [13, 14] . The study that explored the VE of vaccines available in Scotland showed diminished protection against Delta VOC infection and hospital admission [15] . The COVID19 pandemic in Russia has claimed more than 500,000 excess deaths by the spring of 2021 [16] while the lack of realworld evidence on the VE became a public health issue in Russia. Vaccination uptake in the country was undermined by vaccine hesitancy [17] , partly driven by the lack of independent research exploring vaccine safety and effectiveness and failures to communicate the balance of benefits and harms of COVID19 vaccination. St. Petersburg is the second most populated city in Russia and the fourth in Europe. More than 45% of the population have contracted the SARSCoV2 infection by the end of April, 2021. However, only 10% received at least one dose of any vaccine by April, 2021 which is below the levels seen in other countries [18] . Low uptake by the spring of 2021 failed to stop the spread of the new Delta VOC in May-June 2021 in St. Petersburg. The spread of the Delta VOC, in turn, caused the rise in vaccination demand in the summer of 2021. As reported by the city government, by July 27, 2021, approximately 27% of the adult population of St. Petersburg received at least one dose of any vaccine, and 19% received two doses. This increase in the vaccination uptake, consequently, has led to an increase in the absolute number of vaccinated individuals who reported to have a breakthrough infection or hospital admission. Such anecdotal evidence has caused further mistrust in the vaccination programme in absence of reliable scientific reports. In this paper we present the first independent assessment of the VE in Russia. We designed a casecontrol study to assess the VE against referral to hospital and lung injury in the individuals who reported symptoms and were referred for initial computed tomography assessment in two outpatient centres in St. Petersburg, Russia. St. Petersburg and other cities in Russia have established a triage service for symptomatic patients with COVID19. Symptomatic patients with confirmed SARSCoV2 (using polymerase chain reaction (PCR) test) are referred to outpatient triage, including brief physical examination and lowdose computed tomography (LDCT). The decision about hospital admission is based on symptoms, i.e. shortness of breath, hypoxia, overall clinical condition, and lung injury assessed with LDCT. We designed a populationbased casecontrol study to determine the VE against hospital referral and COVID19 lung injury in symptomatic patients with confirmed SARSCoV2 infection, who were referred to the LDCT triage. 2 . 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) The copyright holder for this preprint this version posted August 23, 2021. ; https://doi.org/10.1101/2021.08.18.21262065 doi: medRxiv preprint We retrospectively collected individuallevel patient data from two outpatient triage centres of the Medical Institute named after Berezin Sergey (MIBS), a private medical facility contracted by the city government to provide triage service for nearly half of the city districts. All patients referred to LDCT triage underwent brief physical examination, including pulse oximetry. A local computed tomography score (CTscore) was implemented in Russia [19] . It has five gradations (0, 1, 2, 3, 4) which are related to the volume of involved lung segments (0, < 25%, 25-50%, 50-75%, 75-100%). It is used in the country to define the severity of COVID19 lung injury and to triage patients. CTscore of 3 and 4 means more than 50% of lung volume involvement and is often used as an indication for the hospital admission. Patients with CTscore less than 3, i.e. less than 50% lung injury seen on the LDCT, and in the absence of severe symptoms, are normally sent to outpatient treatment or followup. In our primary analysis, cases were all patients referred to hospital after triage starting from July 3, 2021 till August 9, 2021. We used several definitions of the vaccination status. Complete vaccination status was assigned to the patients who had reported receiving two doses at least 14 days before the referral to the LDCT triage. Partial vaccination status was assigned to the patients who failed to meet the above criteria for complete vaccination, but had reported receiving one dose at least 14 days prior to the referral. In the secondary analyses, we used the CTscore [19] and oxygen saturation as the secondary outcomes to assess the VE against COVID19 lung injury. Oxygen saturation was grouped in the ranges according to the NEWS2 score [20] which was wellvalidated in nonCOVID19 settings. It has four gradations (0, 1, 2, 3) which are related to the oxygen saturation in the following categories: >95, 94-95, 92-93, <92. We modeled our study plan following the WHO interim guidance to evaluate COVID19 vaccine effectiveness [21] . We col lected information on all patients referred to the LDCT triage, as most patients were sent for outpatient treatment and outpatient followup. We used unconditional logistic regression for our primary and secondary outcomes to estimate odds ratios (ORs) for vaccination status among cases and controls, which approximates ORs for the outcomes (hospital admission, different levels of lung injury and decline in oxygen saturation) among the vaccinated and nonvaccinated patients. The VE was calculated as 100% × (1 − OR) adjusted for age (continuous variable), sex, and the triage LDCT centre. Our sample size of 495 hospitalized patients (cases), 13,399 nonhospitalized patients (controls), and 1,291 patients with the complete vaccination status (exposure level of 9.3%) provides 80% power to detect an odds ratio of 0.58 (or the VE of 42%) and 1.46 (or the VE of 46%) for hospi talization at the 5% alpha level in our unmatched casecontrol study. All standard errors and confidence intervals were adjusted for heteroskedasticity with the HuberEickerWhite sandwich estimator. 3 . 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 preprint this version posted August 23, 2021. ; In the sensitivity analysis, for the complete vaccination status, we stratified the period after second dose to the period of 14 to 55 days and the period of 56 or more days. Finally, we investigated the relationship between the hospitalization and the thin plate regression spline of the patient age by vaccination status in a semiparametric logistic regression. The Ethics Committee of the MIBS approved the study on June 21, 2020. The Ethics Committee of the Pavlov First Saint Petersburg State Medical University approved the joint study of COVID19 vaccine effectiveness in St. Petersburg on July 15, 2020. All research was performed following the relevant guidelines and regulations. All participants signed the informed consent upon referral to the LDCT triage. The joint study of COVID19 vaccine effectiveness in St. Petersburg was registered at ClinicalTrials.gov (NCT04981405, date of registration -Aug 4, 2021). This publication covers the data collected at the two outpatient centres of the MIBS that contributed to the study data. The study team was entirely responsible for the study design and for collecting and analysing the data. There was no additional funding of the study provided. Overall, 13,894 patients were included in the final analysis. Patients characteristics are presented in Table 1 . Among all pa tients, 1,965 (14.1%) received at least one dose, 1,379 (9.9%) received two doses, 1,291 (9.3%) met our criteria for complete vaccination status, and additionally 449 (3.2%) met our criteria for partial vaccination status. 495 (3.6%) patients were referred to hospital after the LDCT triage. The majority of referred patients (63.1%) had CTscore 3-4, or >50% lung involvement on LDCT. Patients referred to hospital were also older (66.1% were older than 60 years). Only 17 (3.4%) patients who were referred for hospital admission met the criteria for complete vaccination. In the primary analysis, the adjusted VE against referral to hospital admission was 81% (95% CI: 68-88) for complete vaccination ( Table 2 ). The effect of the partial vaccination against referral to hospital was 35% (95% CI: 21-65). Crude and adjusted ORs and VE against lung injury following the LDCT assessment and decline in oxygen saturation is presented in Table 2 . Only a few patients had more than 75% of lung involvement and none in the vaccination group, so we calculated the VE for the combined category, which included patients with more than 50% of lung involvement. The VE increased with the lung injury categories, from 54% (95% CI:48-60) against any signs of lung injury to 76% (95% CI:59-86) against more than 50% lung injury. The VE against different levels of hypoxia was consistent ( Table 2 ). The VE against hospital admission was more pronounced in women (84%, 95% CI: 66-92) compared to men (76%, 95% CI: 51-88) and older age groups (77%, 95% CI: 62-86) ( Table 3) . We observed a sharp increase in the probability of hospital admission with age for nonvaccinated patients in relation to an almost flat relationship between age and the probability of hospital admission for the completely vaccinated group of patients (Figure 1 ). There was no difference in the VE by the LDCT triage centre. Crude VE estimates were lower compared to the adjusted VE. The adjusted VE by the time elapsed from the second dose was 60% (95% CI: 5-83) for the 14-55 day period and 85% (95% CI: 72-92) for the 56+ day period. 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. The copyright holder for this preprint this version posted August 23, 2021. ; https://doi.org/10.1101/2021.08.18.21262065 doi: medRxiv preprint . 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 preprint this version posted August 23, 2021. ; https://doi.org/10.1101/2021.08.18.21262065 doi: medRxiv preprint . 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 preprint this version posted August 23, 2021. ; https://doi.org/10.1101/2021.08.18.21262065 doi: medRxiv preprint This is the first study examining the realworld effectiveness of COVID19 vaccination in Russia, and one of the first studies globally to provide information about the vaccine protective effect against COVID19associated lung injury. Vaccination showed 81% protection against referral to hospital in patients with symptomatic SARSCoV2 infection during the third wave of COVID19 pandemic caused by Delta VOC in St. Petersburg. It is quite similar to the VE against Delta VOC reported for other COVID19 vaccines [15] . Previously, only invitro data about reduced virusneutralizing activity of GamCOVIDVac was published [22] , but it was not clear how laboratory findings convert into clinical practice. The clear strength of this study is the availability of independent lung injury assessment using the LDCT. The radiologists did not have information on the vaccination status, while the computed tomography was applied as a triage method for all symptomatic patients. The protective effect of vaccination was consistent against all grades of lung injury. We did not observe 75% or more lung tissue involvement during the LDCT in the group of fully vaccinated patients, suggesting strong and con sistent protective effect. Protective effect was also observed against the decline in oxygen saturation. It was clear that both secondary outcomes (lung injury and decline in oxygen saturation) are closely related, but the oxygen saturation and LDCT were independently evaluated at the triage centres. Both factors are independently influencing the decision to refer to hospital, thus explaining the greater VE against hospital admission than for CTscore and oxygen saturation alone. These results further support the findings of the primary analysis. Since all the patients in our study had COVID19, our results provide specific information on vaccination impact on disease severity and its clinical course. When breakthrough infections are evolving as a cause of major concern, our data is assuring that the vaccine effect is going beyond the risk of contracting the infection. It shows that vaccination significantly diminishes disease severity and protects the lungs from virusinduced injury. While SARSCoV2 is likely converting from a pandemic virus to an endemic one, the role of vaccination is also evolving from pandemic control to damage control. Our data shows that it is feasible. Infection fatality rate for COVID19 is progressively increasing with age in St. Petersburg [18] . Our data shows that it is an effective way of protecting the most vulnerable elderly persons from devastating consequences of the pandemic. Lung injury is not only observed in COVID19 but also in many other respiratory viral infections [23] . Lung injury leading to pneumonia and acute respiratory distress syndrome is a major cause of morbidity and mortality in these diseases. Lung imaging with computed tomography were used for assessment and risk stratification in COVID19 [24] . However, in the United States and Europe computed tomography was discouraged by the professional societies as the primary imaging modality in mild COVID19 due to high cost and transmission risk with emphasis placed on clinical assessment instead [25] . In Russia, due to less developed primary care and the low cost of the LDCT, it became the primary imaging modality used for COVID19 triage. This offered the unique opportunity to assess the VE against lung injury detected with LDCT. Our study does not provide direct evidence on the VE against SARSCoV2 infection and symptomatic disease. However, some indirect inferences can be made when analysing the characteristics of the overall patient group in our study. By July 27, 2021, the city government reported that at least one dose of vaccine was given to 1,178,266 residents of St. Petersburg (26.5% of 7 . 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 preprint this version posted August 23, 2021. ; the adult population), and 819,768 received two doses (18.5% of the adult population). In our study, among 13,992 symptomatic patients who were referred to the LDCT triage, only 14.1% received at least one dose, and 9.3% received two doses. In the absence of individuallevel data, we can only estimate crude effectiveness against the symptomatic disease, which is around 50%. However, the adjustment for the age and sex in the estimation of the VE against hospital admission in our study moved the point estimate from roughly 66% to 81%. It is likely that our crude indirect estimate of the VE against symptomatic disease is also underestimated. However, if we combine the protective effect against symptomatic disease and hospital admission, it is in line with the findings for other vaccines [13] . The VE against Delta VOC for one dose of vaccine was lower than for two doses, and the effect did not clearly manifest until at least 28 days after the first one [15] . We obtained similar results in our study. The point estimate for partial vaccination was 35%, but the confidence interval for the corresponding OR included unity, suggesting inadequate sample size due to the minimum detectable VE of 70% for the partial vaccination exposure as our power calculations show. It should be noted that in Russia the recommended 21day interval between the first and the second dose of GamCOVIDVac is usually strictly followed. In contrast to another study [13] we did not find any possible indications of waning efficacy. The VE was higher in patients with longer period after the second dose. However, these results should be interpreted with caution, as possible biases can intervene in the stratified analysis of observational data. There are several important limitations of our study. We did not have the information on the vaccine type, so the estimated It is safe to assume that our study approximates the effectiveness of GamCOVIDVac. The effectiveness for EpiVacCorona and KoviVac would be difficult to derive from observational casecontrol studies due to negligible uptake. Another important limitation of our study is possible referral bias. Patients were referred to the triage centres if they had positive PCR tests and disease symptoms, but the decision to refer was left at the physician's discretion. We cannot rule out that the vaccination status was influencing this decision. If this is the case, in our data we would observe more severely ill vaccinated individuals than the unvaccinated ones. Alternatively, the physicians may fully rely on vaccine protection and never refer the vaccinated patients to the triage. However, we observed a considerable number of vaccinated individuals at the triage centres. Therefore, we believe that the referral bias would result in underestimation of the VE. Vaccination could also influence patients' medical care seeking behaviour. We assume that selfassurance on vaccine protection would predominantly prevent patients with mild disease from seeking medical care, and this would also lead to underestimation of VE against severe disease in our study. The study results may be also biased if the triage referred symptomatic patients to hospital based on their vaccination status. To the best of our knowledge, this decision was solely based on physical examination and the LDCT results. The vaccination status was collected for research purposes, was not a part of any official medical records. There is little reason to believe that the vaccine influenced the decision to refer the patient to hospital. 8 . 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) The copyright holder for this preprint this version posted August 23, 2021. ; https://doi.org/10.1101/2021.08.18.21262065 doi: medRxiv preprint Finally, the data we used in our study was collected at the triage centres and not in hospitals. Even though all patients triaged for inpatient treatment were followedup by the LDCT centre's personnel from the moment they were referred to the hospital till hospital admission, we do not know the length or the outcome of the hospital admission itself. It is also possible that the patients triaged for outpatient treatment were later admitted. Our study covers only one point in their disease course and traces only the decisions made in the triage centres. Additional observational studies are needed to assess the vaccination effectiveness against SARSCoV2 infection and COVID19 associated death. Given the limitations of our data, we were only able to estimate the protective effect against lung injury and referral to hospital. However, it is safe to assume that those outcomes pose as a surrogate for COVID19associated death. Our study does not provide any information on the vaccine safety, but at least for GamCOVIDVac, the safety data are available from independent sources [10, 11] . In conclusion, we showed that COVID19 vaccination is effective against referral to hospital in patients with symptomatic SARSCoV2 infection in St. Petersburg, Russia. The protection against hospital admission is probably mediated through pro tection against lung injury associated with COVID19. Realworld evidence on the VE against COVID19 should be integrated into populationbased vaccination programmes to negotiate the balance of benefits and harms of this effective COVID19 con trol measure, that is likely gaining even more importance in the light of gradual ceasing of nonpharmaceutical interventions against COVID19 and calls for a "return to normality". Safety and immunogenicity of an rAd26 and rAd5 vectorbased heterologous prime boost COVID19 vaccine in two formulations: two open, nonrandomised phase 1/2 studies from Russia. The Lancet Safety and efficacy of an rAd26 and rAd5 vectorbased heterologous primeboost COVID19 vaccine: an interim analysis of a randomised controlled phase 3 trial in Russia Safety and efficacy of the ChAdOx1 nCoV19 vaccine (AZD1222) against SARSCoV2: an interim analysis of four randomised controlled trials in Brazil, South Africa, and the UK. The Lancet Safety and efficacy of the BNT162b2 mRNA Covid19 vaccine What do we know about Sputnik V and other Russian vaccines? Safety and efficacy of the Russian COVID19 vaccine: more information needed Data discrepancies and substandard reporting of interim data of Sputnik V phase 3 trial. The Lancet Safety and efficacy of the Russian COVID19 vaccine: more information needed -Authors' reply. The Lancet Data discrepancies and substandard reporting of interim data of Sputnik V phase 3 trial -Authors' reply. The Lancet ROCCA observational study: Early results on safety of Sputnik V vaccine (GamCOVIDVac) in the Republic of San Marino using active surveillance Active surveillance of the Sputnik V vaccine in health workers Vaccine Effectiveness Studies in the Field Effectiveness of the PfizerBioNTech and OxfordAstraZeneca vaccines on Covid19 related symptoms, hospital admissions, and mortality in older adults in England: test negative casecontrol study Effectiveness of CoronaVac among healthcare workers in the setting of high SARSCoV2 Gamma variant transmission in Manaus, Brazil: A testnegative casecontrol study. The Lancet Regional Health -Americas SARSCoV2 Delta VOC in Scotland: demographics, risk of hospital admission, and vaccine effectiveness Tracking excess mortality across countries during the COVID19 pandemic with the World Mortality Dataset COVID19 vaccine acceptance and hesitancy in low and middle income countries, and implications for messaging COVID19 pandemic in Saint Petersburg, Russia: combining surveillance and populationbased serological study data in Chest computed tomography for outcome prediction in laboratoryconfirmed COVID19: a retrospective analysis of 38,051 cases The National Early Warning Score 2 (NEWS2) in patients with hypercapnic respiratory failure Evaluation of COVID19 vaccine effectiveness: interim guidance Neutralizing Activity of Sera from Sputnik VVaccinated People against Variants of Concern Imaging findings in coronavirus infections: SARSCoV, MERSCoV, and SARSCoV2 Performance of LowDose Chest CT as a Triage Tool for Suspected COVID19 Patients The Role of Chest Imaging in Patient Management during the COVID19 Pandemic: A Multinational Consensus Statement from the Fleischner Society We thank Alexandra Vasilieva, Alla Samoletova, and Maria Batygina (European University at St. Petersburg) for the science communication and for administrative support. We also thank Evgeny Bakin for his insightful comments and Ivan Moiseev for