key: cord-1007852-754eoeoa
authors: Ferreira, L. S.; de Almeida, G. B.; Borges, M. E.; Simon, L. M.; Poloni, S.; Bagattini, A. M.; da Rosa, M. Q. M.; Filho, J. A. F. D.; Kuchenbecker, R. d. S.; Camey, S. A.; Kraenkel, R. A.; Coutinho, R. M.; Toscano, C. M.
title: Modelling optimal vaccination strategies against Covid-19 in a context of Gamma variant predominance in Brazil
date: 2021-11-21
journal: nan
DOI: 10.1101/2021.11.19.21266590
sha: b4be2dc5fe2533d6a0970bbfd79108ea303db287
doc_id: 1007852
cord_uid: 754eoeoa

Brazil experienced moments of collapse in its health system throughout 2021, driven by a timid initial vaccination strategy against Covid-19, combined with the emergence of variants of interest (VOC). Mathematical modelling has been used to subsidize decision-makers in public health planning. Considering the vaccine products available, effectiveness estimates, the emergence of Gamma as the predominant VOC circulating in 2021, and national estimated doses available for the next several months, we developed a Markov-chain mathematical modelling approach to evaluate optimal strategies for Covid-19 vaccination in Brazil in terms of Covid deaths averted. Our main findings are that in order to reach higher vaccination impact in Brazil, Covid-19 immunization strategies should include recovering vaccination coverage rates in high-risk groups reaching higher coverage; expanding vaccination to younger age groups should be considered only after ensuring at least 80% coverage in older age groups; reducing the interval between doses of AZD1222 from 12 to 8 weeks. We also demonstrate that the latter is only feasible if accompanied by an increase in vaccine supply of at least 50% in the next six month period.

Brazil experienced moments of collapse in its health system throughout 2021, driven 2 by a timid initial vaccination strategy against Covid-19, combined with the emergence 3 of variants of interest (VOC). Mathematical modelling has been used to subsidize 4 decision-makers in public health planning. Considering the vaccine products available, 5 effectiveness estimates, the emergence of Gamma as the predominant VOC 6 circulating in 2021, and national estimated doses available for the next several 7 months, we developed a Markov-chain mathematical modelling approach to evaluate 8 optimal strategies for Covid-19 vaccination in Brazil in terms of Covid deaths averted. 9 Our main findings are that in order to reach higher vaccination impact in Brazil, Covid-10 19 immunization strategies should include recovering vaccination coverage rates in 11 high-risk groups reaching higher coverage; expanding vaccination to younger age 12 groups should be considered only after ensuring at least 80% coverage in older age 13 groups; reducing the interval between doses of AZD1222 from 12 to 8 weeks. We also 14 demonstrate that the latter is only feasible if accompanied by an increase in vaccine 15 supply of at least 50% in the next six month period. Introduction 1 To date, Brazil has reached over 600,000 deaths due to Covid-19, ranking 2 second in the world for the absolute number of Covid-19 deaths (1) . The first Covid-3 19 case was identified in the country in March 2020, and the first wave of the disease 4 reached its first peak in July. By October 2020, when more than 5 million confirmed 5 cases and at least 150.000 deaths had been registered in Brazil (2), the first peak of 6 the disease had been overcome. The downward trend that lasted 16 weeks ceased in 7 late November 2020, when a new rise in cases and deaths was observed (3). This 8 coincides with the detection of the variants of concern (VOC) in the country, 9 challenging Brazil's health systems and the government's public response. As the fifth-10 largest world's country, Brazil presented multiple and different epidemic curves 11 according to the transmission of SARS-CoV-2 between regions in 2020, progressively 12 reaching countryside smaller cities and developing a national synchronization process 13 between 2020 and 2021 where outbreaks caused by VOC, such as Delta and Gamma, 14 were important drivers of sustainable transmission. The Gamma variant (P1 lineage 15 or GR/501Y.V3), first identified in Brazil (3), spread throughout the country before the 16 national vaccination campaign, resulting in numerous surges throughout the territory. 17 In the tenth epidemiological week (EW) of 2021, there was an impressive number of 18 40,797 hospital admissions, corresponding to an increase of 192% compared with the 19 worst week of the epidemic in 2020 (EW 28) (4). 20 Covid-19 vaccination rollout started on EW three of 2021 and followed a 21 prioritization framework proposed by WHO (5) , initially targeting health professionals, 22 older adults, and people with comorbidities, in this order. Vaccination started slowly 23 and restrictively in its first weeks; the daily-applied doses 7-day rolling average was 24 . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 21, 2021. ; https://doi.org/10.1101/2021.11.19.21266590 doi: medRxiv preprint around 200 thousand doses a day (6). From January to March, vaccination was 1 jeopardized in different moments by limited vaccine supply. Only after April, the daily 2 number of administered vaccines reached 600 to 700 thousand (6). The current Covid-3 19 vaccines deployed in Brazil are AZD1222 (AstraZeneca/Oxford/Fiocruz), a two-4 dose adenovirus vaccine, currently administered in a 12-week dose-interval; 5 CoronaVac (Sinovac/Butantan), a two-dose inactivated virus vaccine, using a 4-week Care System (SUS) (12). Managed by the Federal Government, together with States 23 and Municipalities, decentralized and with good capillarity, achieved by more than 36 24 thousand rooms in 5,570 Brazilian cities, the PNI has historically been able to deliver 25 . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint Differently from what would be expected in a country with this record on 5 immunization activities, the National Immunization Campaign for Covid-19 in Brazil 6 has faced several new obstacles. These included lack of national coordination and 7 support to evidence-based decision making, inconsistent vaccine supply and 8 availability over time, limited social communication strategies, and widespread social 9 networks misinformation about vaccine safety and efficacy. In addition, the anti-vax 10 movement in the country has been gaining strength throughout the SARS-CoV-2 11 dissemination process. From the operational point of view, the Covid-19 vaccination 12 rollout was fragmented, with different vaccination strategies being adopted as a 13 decentralized process involving state and municipal levels. In a country of continental 14 dimensions experiencing different regional and local pandemic waves, the lack of 15 communication and an integrated response brought additional challenges to 16 vaccination implementation strategies. As such, vaccines were progressively made 17 available to adults without comorbidities, by age, irrespective of vaccine coverages in 18 priority risk populations. One critical point is that, in this age-based strategy, a 19 minimum coverage goal set by age group was not established before the vaccine was 20 made available for younger groups. This progression was based on a temporal 21 criterion (for example, a one-week period designated for each age group), varying by 22 locality. 23 Considering the vaccine products available, effectiveness estimates, the is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 21, 2021. ; https://doi.org/10.1101/2021.11.19.21266590 doi: medRxiv preprint estimated doses available for the next several months, we developed mathematical 1 modelling approaches to evaluate optimal strategies for COVID-19 vaccination in 2 Brazil. This work aims to provide immediate answers to pressing questions and a 3 proposed framework for future analyses to provide decision-makers with modelling 4 evidence to support the decision-making process at a national level regarding best 5 strategies to minimize the burden of Covid-19 deaths in Brazil. 6 We specifically aim to address three programmatic questions: what should be is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 21, 2021. ; https://doi.org/10.1101/2021.11.19.21266590 doi: medRxiv preprint (Pfizer/BioNTech). Since our goal is to estimate optimal strategies considering the 1 interval between vaccine doses, we have not included the Ad26.COV2.S vaccine in 2 the model. Also, the minor frequency of individuals that received or will receive this 3 vaccine (4.5 million people out of 211 million people, representing less than 2% of the 4 population) means that this should not affect our conclusions. 5 We use the number of doses procured by the Brazilian Ministry of Health and The structure of the model is the same as an extended SEIR model (15) , 20 accounting also for asymptomatic, hospitalized, and deceased individuals, thus being is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 21, 2021. ; https://doi.org/10.1101/2021.11.19.21266590 doi: medRxiv preprint CoronaVac, and BNT162b2), and the vaccines are modelled simultaneously (see 1 Figure 1 ). is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 21, 2021. ;

To estimate the number of doses of vaccine that should be allocated for the first 1 or second dose, we use a modified version of the optimization model developed by 2 (16) that accounts for varying production (or deployment) rates and also previously 3 vaccinated individuals with only one dose (see further details on the SM). This 4 optimization model calculates the number of first or second doses used by day, given 5 production rate and the interval between doses, minimizing the number of doses that 6 should be kept in stock while guaranteeing that individuals receive the second dose 7 when recommended. 8 We assumed a single model for the entirety of Brazil. We also assume that after 9 initially targeting high-risk populations, vaccination rollout followed an age-10 prioritization. We considered the number of administered doses by vaccine, over time, 11 as being proportional to the total number of doses of each vaccine made available at 12 the time of vaccination. We considered that the vaccination rate in each age group is 13 proportional to the unvaccinated population in this group (more details in 14 Supplementary material). 15 We limited the analysis of optimal time between doses (and the required 16 supplied doses) only to AZD1222 for the following reasons: 1) there is no evidence to 17 support the use of CoronaVac vaccine in a longer interval than the recommended four is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 21, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 21, 2021. comparability of strategies. Each scenario is then run using 500 combinations of 1 values for each result shown in the next section. What should be the vaccination coverage of an age group before starting 5 vaccination in a younger group? 6 We calculated the excess deaths caused by starting vaccination in a younger 7 age group after a coverage threshold compared to the strategy of only starting to 8 vaccinate a younger age group after fully vaccinating the older age group. As shown 9 in Figure 2 , the lower the vaccination coverage reached in older age groups, the 10 greater the estimated excess of deaths, regardless of the probability of infection. 11 However, we find that the magnitude of the impact is smaller when the probability of 12 infection is lower. Furthermore, it can be seen that at least 90% vaccination coverage 13 is necessary for a minimal excess of deaths to be reached (varying from 10.57 (95%CI: is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 21, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 21, 2021. ; https://doi.org/10.1101/2021.11.19.21266590 doi: medRxiv preprint intervals (8, 9, 10, and 11 weeks) compared to the standard currently recommended 1 12-week interval (Figure 3 ). We assumed a setting without limitation of AZD1222 2 vaccine doses, i.e., we ran the model assuming a number of doses up to ten times 3 higher than the currently available and projected doses of AZD1222. We found that 4 the lower dose-interval of 8 weeks leads to a greater reduction in the number of Covid- is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint This result leads to whether the strategy of using a lower dose-interval for 1 AZD1222 would be effective considering the currently projected vaccine supply until 2 the end of 2021. When considering the scenario of the currently projected AZD1222 3 vaccine supply, we see that the reduction in deaths is negligible, less than one death 4 regardless of the probability of infection (Figure 4) . This, in turn, leads to the next 5 question which this modelling addresses. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint What is the minimum amount of vaccines made available over time that 1 will allow the implementation of the optimal interval between doses? 2 Considering the different dose-intervals for the AZD1222 vaccine, we estimate 3 that the number of vaccine doses administered needs to be increased by at least 50% 4 to avoid supply bottlenecks and allow for an impact on Covid-19 deaths reduction be 5 observed ( Figure 5 ). When comparing these estimates to the ones presented in Figure   6 3, we can observe the different population impact of the strategy when considering a is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 21, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 21, 2021. ; https://doi.org/10.1101/2021.11.19.21266590 doi: medRxiv preprint masks, social distancing and respect for quarantine measures, claiming that these 1 measures would bring enormous losses of economic impact. This, associated with the 2 emergence of new variants of concern (VOC), such as the Gamma variant (26), can 3 represent a major problem. Being flexible and rethinking strategies has been 4 mandatory in a large, plural country without a central plan to fight the epidemic (27). 5 By implementing vaccination on a large scale, so far with good efficacy data for 6 all vaccines with only one dose for the original virus, Brazil focused on vaccinating the 7 greatest number of people with one dose, ensuring some immunity, and spacing out 8 the second dose for the maximum periods stipulated by the manufacturers. Thus, until 9 July 2021, both the AZD1222 and BNT162b2 vaccines were administered with a 12- 10 week interval between doses. In addition, the eagerness to reach a more significant 11 number of vaccinated individuals has led states and municipalities to expand 12 vaccination including early on younger age groups without defining a priori a minimum 13 coverage to be reached in the groups already prioritized, which can also represent a 14 problem. 15 In the context of scarce vaccine supply, it is crucial to ensure that at-risk 16 individuals are adequately protected against Covid-19. Since our model limits to age-17 stratified populations, ignoring other groups, for example, pregnant women and 18 immunosuppressed individuals, we can measure the effect of different thresholds of 19 coverage of older individuals before making vaccine doses available to younger 20 individuals. Figure 2 shows that ensuring a good vaccine coverage of older individuals 21 (at least 90% of coverage) reduces the number of deaths considerably, as expected. 22 However, more important than that, using lower coverages (lower than 80%) as a 23 threshold generates a sharp increase in the number of additional deaths compared to 24 vaccinating the whole population of older individuals beforehand. Thus, the first strong 25 . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 21, 2021. ; https://doi.org/10.1101/2021.11.19.21266590 doi: medRxiv preprint recommendation that this article can make to optimize the ongoing vaccination plan in 1 Brazil is that it is necessary to resume efforts to achieve minimum coverage in older 2 age groups, so that, only after a minimum coverage of 80%, we can advance is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 21, 2021. ; https://doi.org/10.1101/2021.11.19.21266590 doi: medRxiv preprint doses should be in the order of, at least, 50%, independently of the transmission level 1 of the epidemic. This might be achievable as Brazil's AZD1222 producer Fiocruz- 2 Biomanguinhos is upgrading its factory from filling doses to in loco production of active 3 pharmaceutical ingredients. This would, in principle, enable the production of four 4 monthly lots of doses instead of the current three lots. Ideally, increasing vaccine 5 availability by 100%, would result in an ideal scenario in terms of vaccination impact. 6 We did not model an increase beyond that level, as we assumed it to be unrealistic. 7 Thus, one more strong recommendation based on mathematical modelling can be is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 21, 2021. ; https://doi.org/10.1101/2021.11.19.21266590 doi: medRxiv preprint assume that the same vaccination strategies identified as optimal should be used to 1 contain the disease caused by the Delta variant, i.e., reinforcing vaccination coverage 2 of 60+ age groups to at least 80%, reducing the interval between doses of AZD1222 3 from 12 to 8 weeks and upgrading the production capabilities to at least 100%. 4 Mathematical modelling has been extensively used to assist policymakers 5 during the Covid-19 pandemic. The range of scenarios studied includes, but is not 6 limited to, school reopening (28), the effects of lockdown (29), and, of course, (Table 2) . 22 Ferreira et al. (16) considered varying (constant) production rates and have shown 23 that, besides first dose efficacy, knowledge of vaccine production is also an important 24 parameter when considering the optimal interval between doses, whereas the optimal 25 . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 21, 2021. parameters. All authors discussed and agreed on model parameters to be considered. 10 LSF, MEB, SP, and RMC were primarily responsible for model structuring, 11 parametrization, and statistical analyses. All authors participated in the interpretation 12 of the analyses. 13 LSF and GBA drafted the paper. All authors contributed to revising the paper, 14 the tables, and the figures critically and for important intellectual content. 15 All authors approved the final submitted version of the paper. 16 All authors have full access to the data, are accountable for all aspects of the 17 work, and will ensure that questions related to the accuracy or integrity of any part of 18 the work are appropriately investigated and resolved. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint 13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39 . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 21, 2021. ; https://doi.org/10.1101/2021.11.19.21266590 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 21, 2021. ; https://doi.org/10.1101/2021.11.19.21266590 doi: medRxiv preprint . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint 

Boletim epidemiológico especial -Doença pelo Novo Coronavírus -COVID-19

COVID-19 in Brazil: 150,000 deaths and the Brazilian 5 underreporting

despite high seroprevalence. The Lancet

COVID-19 hospital admissions: Brazil's first 9 and second waves compared

World Health Organization. WHO SAGE Roadmap For Prioritizing Uses Of COVID-19 Vaccines In The 11

Context Of Limited Supply

Campanha Nacional de Vacinação contra Covid-19

efficacy and safety of an inactivated 17 whole-virion SARS-CoV-2 vaccine (CoronaVac): interim results of a double-blind, randomised, placebo-18 controlled, phase 3 trial in Turkey

Phase 3 Safety and Efficacy of 20 AZD1222 (ChAdOx1 nCoV-19) Covid-19 Vaccine

BNT162b2 mRNA Covid-19 Vaccine

Interim recommendations for use of the Pfizer-BioNTech COVID-19 vaccine, 11. World Health Organization. Interim recommendations for use of the ChAdOx1-S

COVID-19 (AstraZeneca COVID-19 vaccine AZD1222 Vaxzevria TM , SII COVISHIELD TM

O Programa Nacional de Imunizações (PNI): origens e desenvolvimento

Uptake of pandemic influenza (H1N1)-2009 vaccines in Brazil

An introduction to stochastic modeling

Modeling Infectious Diseases in Humans and Animals. Modeling Infectious Diseases 11 in Humans and Animals

Assessing optimal time between 13 doses in two-dose vaccination regimen in an ongoing epidemic of SARS-CoV-2 | medRxiv

Three-quarters attack 16 rate of SARS-CoV-2 in the Brazilian Amazon during a largely unmitigated epidemic

A seroprevalence survey of anti-SARS-CoV-2 antibodies among 20 individuals 18 years of age or older living in a vulnerable region of the city of São Paulo

Emerging viruses and current strategies for vaccine intervention

Influence of age 1 on the effectiveness and duration of protection in Vaxzevria and CoronaVac vaccines

Single-dose administration and 4 the influence of the timing of the booster dose on immunogenicity and efficacy of

AZD1222) vaccine: a pooled analysis of four randomised trials. The Lancet

Effectiveness of mRNA and ChAdOx1

COVID-19 vaccines against symptomatic SARS-CoV-2 infection and severe outcomes with variants of 8 concern in Ontario

ChAdOx1 vaccine in older adults during SARS-CoV-2 Gamma variant circulation in São Paulo

BNT162b2 mRNA Covid-19 Vaccine in a 14

Nationwide Mass Vaccination Setting

Brazil in the face of 16 new SARS-CoV-2 variants: emergencies and challenges in public health. Rev bras epidemiol

Governing COVID-19 without government in Brazil: Ignorance, neoliberal authoritarianism, 18 and the collapse of public health leadership. Global Public Health

Modeling the impact of school 20 reopening and contact tracing strategies on COVID-19 dynamics in different epidemiologic settings in Brazil