key: cord-0919184-eatyv9eu
authors: Vinceti, Marco; Filippini, Tommaso; Rothman, Kenneth J.; Di Federico, Silvia; Orsini, Nicola
title: SARS-CoV-2 infection incidence during the first and second COVID-19 waves in Italy
date: 2021-04-01
journal: Environ Res
DOI: 10.1016/j.envres.2021.111097
sha: b1321e028396ef759240777430ef085985f2d1e5
doc_id: 919184
cord_uid: eatyv9eu

We assessed the relation between Covid-19 waves in Italy, which was severely affected during the pandemic. We evaluated the hypothesis that a larger impact from the first wave (February-March 2020) predicts a smaller peak during the second wave (September-October), in the absence of local changes in public health interventions and area-specific differences in time trends of environmental parameters. Based on publicly available data on province-specific SARS-CoV-2 infections and both crude and multivariable cubic spline regression models, we found that for provinces with the lowest incidence rates in the first wave, the incidence in the second wave increased roughly in proportion with the incidence in the first wave until an incidence of about 500-600 cases/100,000 in the first wave. Above that value, provinces with higher incidences in the first wave experienced lower incidences in the second wave. It appears that a comparatively high cumulative incidence of infection, even if far below theoretical thresholds required for herd immunity, may provide noticeable protection during the second wave. We speculate that, if real, the mechanism for this pattern could be depletion of most susceptible individuals and of superspreaders in the first wave. A population learning effect regarding cautious behavior could have also contributed. Since no area-specific variation of the national policy against the SARS-CoV-2 outbreak was allowed until early November 2020, neither individual behaviours nor established or purported environmental risk factors of Covid-19, such as air pollution and meteorological factors, are likely to have confounded the inverse trends we observed in infection incidence over time.

Coronavirus disease (COVID-19) is a severe and potentially life-threatening disease due to infection with the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which spread globally during 2020 after the initial outbreak late in 2019 in China . Its transmission occurs predominantly through air (Wiersinga et al., 2020) . Therapeutic options to treat the infection and its sequelae are limited (CORIST, 2020; Tsatsakis et al., 2020a; Wiersinga et al., 2020) . Before a vaccine was available, and in the absence of a highly effective therapy, attempts to control the SARS-CoV-2 pandemic during 2020 relied on public-health measures such as donning protective gear, social distancing, and shutdowns that limited population mobility and interaction (Ma and Lipsitch, 2021; Vinceti et al., 2020) . Reduced social interaction lowers R t , and can stem the spread of the virus, but lockdowns cannot be sustained indefinitely. With normal social intercourse, protection from herd immunity is thought to be attainable only at high prevalences of immunity, on the order of at least 50% and likely considerably above that for this virus (Fontanet and Cauchemez, 2020; Omer et al., 2020; Randolph and Barreiro, 2020) .

Italy offers a unique opportunity to investigate how successive waves of the infection may interact (Lipsitch et al., 2020) . Italy was the first nation with widespread population involvement, with a high mortality from COVID-19 in the winter of 2020, followed by a spring and summer with relatively low infection incidence after easing of the tight lockdown that was in effect from March 8 through May 2 Vinceti et al., 2020) . The second wave arrived in late August and progressed to a peak in late October. A characteristic of the response to the pandemic in Italy is that the nonpharmacological interventions, including social distancing, mask use and the closing of schools and retail establishments, were evenly adopted throughout the country. There were no important local differences in attempted control until November 6, 2020, when area-specific policies driven by the local spread of the outbreak began (DPCM, 2020) . This history allows a relatively unconfounded comparison of the first and second waves of Covid-19 epidemiology during the February-October period between Italian communities, assuming that environmental and behavioral risk factors for Covid-19 such as outdoor air pollution, meteorological effects, and social interaction are close to uniform across Italy (Ambika et al., 2021; Copat et al., 2020; Domingo and Rovira, 2020; Filippini et al., 2021; Filippini et al., 2020;  J o u r n a l P r e -p r o o f Marques et al., 2021; Mele et al., 2021; Sidiropoulou et al., 2021; Sunyer et al., 2021; Tsatsakis et al., 2020b) . The vaccination campaign did not begin in Italy until December 27, 2020 (Ministry of Health, 2021).

Here we consider how the two waves of infection interacted within Italian communities during the 2020 February-November period.

We used data available without cost from public sources. The number of newly diagnosed infections with SARS-CoV-2 was reported on the website of the Civil Protection Agency (CPD -Italian Civil Protection Department, 2020) for two time frames: from February 24-May 31, 2020 (first wave), and from September 1-October 31, 2020 (second wave). These newly diagnosed infections correspond to the new positive tests of infection based on quantitative reverse transcription polymerase chain reaction reported by regional Health Services. Daily data flow from all regions of Italy was mandatory. Based on these data and the population data available at the Italian National Institute of Statistics website at January 1, 2020 (ISTAT, 2020b), we computed wave-specific incidence of SARS-CoV-2 infection during the first and second waves. We also used the regional anti-SARS-CoV-2 antibody seroprevalence data made available by the National Institute of Statistics (ISTAT, 2020a).

We assessed the association between first wave incidence and regional seroprevalence data made available in the national survey with a weighted linear regression analysis, weighting observations by the ratio of provincial or regional population size to average size for all Italian provinces or regions. To investigate the relation between first and second wave province-specific incidence rates, we used linear regression to fit a restricted cubic spline model that weighted provincial data by population size and based on three knots at fixed percentiles of first wave distribution. We also calculated a pointwise 95% confidence interval (CI) for the forecasted second wave incidence. We used Stata software (Version-16.1 Stata Inc., J o u r n a l P r e -p r o o f request. Figure 1 shows incidence in the investigated periods. In the first wave, the number of diagnosed cases was 233,019, while in the shorter period of the second wave it was 409,241. Accordingly, the cumulative incidence during the first wave was 387/100,000 as a national average, ranging from 60/100,000 in Calabria region to 943/100,000 in Aosta Valley region. Corresponding figures for the slightly shorter period of the second wave were 679/100,000 for the national average, ranging from 171 (Calabria) to 1571/100,000 (Aosta Valley). The region-specific cumulative incidence and seroprevalence were, as expected, strongly correlated (Supplemental Figure S1 ).

In the spline model we found an inverted U-shaped relation between first wave incidence and the estimated predicted mean of the second wave incidence computed on residents uninfected during the first wave, i.e. the susceptible population ( Figure 2 ). If the first wave incidence was relatively low, the incidence was positively correlated with the incidence in the second wave. If the first wave incidence, however, was greater than about 500-600 cases/100,000, the incidence was negatively correlated with the second wave severity. These findings were similar when we computed the incidence in the second wave among all residents, independently of having been infected during the first wave, rather than just in the susceptible population (Supplemental Figure S2 ). Adding potential confounders to the regression models such as proportion of elderly people, residents living alone, and an indicator of mobility, also did not appreciably change the results (Supplemental Figure S3 ). Based on the relation between estimated seroprevalence and first wave incidence, and the downward trend on the right side of the inverted U distribution in Figure 2, we estimated that the seroprevalence threshold for a protective effect in the second wave is around approximately 2.8%--3.6%. Values above this cutpoint, especially much higher levels, were associated with more limited spread during the second wave.

We saw an inverted U-shaped relation between first wave and second wave incidence. Above 500-600 infection cases/100,000, greater first-wave incidence predicted progressively smaller second-wave incidence. This inflection point corresponded with a population-based anti-SARS-CoV2 antibody prevalence of approximately 3%. The official case numbers from which we computed first and second wave infection rates underestimated the actual incidence, since they mostly comprised symptomatic cases, particularly during the first wave. Nonetheless, these figures were highly correlated with the seroprevalence data at the regional level.

The seroprevalence levels corresponding to these first-wave incidences are much lower than those expected to bring about herd immunity, which for Covid-19 is calculated to be above 50% (Fontanet and Cauchemez, 2020) . The R t and the corresponding level for herd immunity could be affected by heterogeneity of behavior, if high-risk people are more likely to have contracted and spread the disease in the first wave (such as superspreaders, frequent travelers, nursing home and hospital personnel) (Calo et al., 2020; Kault, 2020; Kochanczyk et al., 2020; Lewis, 2021; Ma and Lipsitch, 2021; Signorelli et al., 2020) and therefore play a smaller role in the spread of the virus during a second wave. In fact, individuals most likely to contract and to spread the virus were also most likely to be infected early and after that to show immunity, diminishing R t during later waves on the presumption that immunity persists during the interval between waves (Braun et al., 2020; Long et al., 2020; Wajnberg et al., 2020) . This presumption seems reasonable, as a study from China suggested that the duration of neutralizing antibodies against SARS-CoV-2 persisted for at least nine months (He et al., 2021) . In addition, recent evidence indicates that seroprevalence estimates, usually based on anti-SARS-CoV-2 antibody prevalence as was the case in Italy and other countries (Deeks et al., 2020; ISTAT, 2020a; Pollan et al., 2020) , considerably underestimate the prevalence of effective immunity. T-cell mediated immunity may occur even in anti-SARS-CoV-2 antibody negative individuals (Braun et al., 2020; Sekine et al., 2020) , perhaps from cross-reactivity from other human coronavirus infections Kostoff et al., 2020) , such as those accounting for 20% of upper respiratory tract infections, and might last for years (Braun et al., 2020; Callow et al., 1990; Hope and Bradley, 2021; Jarjour et al., 2021; Ledford, 2021; Neagu et al., 2021; Ng et al., 2016) . The effect of this J o u r n a l P r e -p r o o f unmeasured immunity could depress the R t closer to 1 in populations with high first-wave infection incidence.

Populations of the provinces more severely hit by the epidemic during the first wave may have undergone a more profound behavioral change during the summer and the second wave, by voluntarily adhering to prudent behaviors protective against contracting Covid-19. On the other hand, this theory is unlikely to explain the association we observed, because all of Italy engaged in the same set of publichealth measures adopted to curb the Covid-19 outbreak, including the light and tights lockdowns after the beginning of the outbreak until October 2020, as well as the prescriptions for public-health measures such as systematic mask wearing . It was not until November 6 that the Italian government activated area-specific differential approaches for Covid-19 prevention and control, including bar, restaurant, shop and school closure, and mobility restrictions, adapted to the spread of the Covid-19 outbreak in each region. Furthermore, neither the SARS-CoV2 variant B.1.1.7, a variant first detected in the UK in late 2021, nor the South African and Brazilian variants, were detected in Italy during the period investigated in this study (ISS, 2021) , making it unlikely that there was any confounding related to lineagespecific transmission dynamics of the virus (Davies et al., 2021) . It appears that the severity of the first wave of the Covid-19 outbreak may have been a determinant of the severity of the subsequent wave, even with comparatively low levels of immunity at the population level. 

and past infection with SARS-CoV-2. Cochrane Database Syst Rev

A new threat from an old enemy: Reemergence of coronavirus (Review)

Effects of air pollutants on the transmission and severity of respiratory viral infections

Ulteriori disposizioni attuative del decretolegge 25 marzo 2020, n. 19, convertito, con modificazioni, dalla legge 25 maggio 2020, n. 35, recante «Misure urgenti per fronteggiare l'emergenza epidemiologica da COVID-19

Associations between mortality from COVID-19 in two Italian regions and outdoor air pollution as assessed through tropospheric nitrogen dioxide

Satellite-detected tropospheric nitrogen dioxide and spread of SARS-CoV-2 infection in Northern Italy

COVID-19 herd immunity: Where are we?

COVID19 in Northern Italy: An integrative overview of factors possibly influencing the sharp increase of the outbreak

First results of the seroprevalence investigation on SARS-COV-2. Italian National Institute of Statistics

Populations and families

T Cell Memory: Understanding COVID-19

2020. Superspreaders, asymptomatics and COVID-19 elimination

Super-spreading events initiated the exponential growth phase of COVID-19 with R0 higher than initially estimated

Vaccine-and natural infection-induced mechanisms that could modulate vaccine safety

How 'killer' T cells could boost COVID immunity in face of new variants

Superspreading drives the COVID pandemic -and could help to tame it

Defining the epidemiology of Covid-19 -Studies needed

Clinical and immunological

Back to basics in COVID-19: Antigens and antibodies-Completing the puzzle

Memory T cell responses targeting the SARS coronavirus persist up to 11 years post-infection

Herd immunity and implications for SARS-CoV-2 control

Prevalence of SARS-CoV-2 in Spain (ENE-COVID): a nationwide, population-based seroepidemiological study

Herd immunity: Understanding COVID-19

Robust T cell immunity in convalescent individuals with asymptomatic or mild COVID-19

Unraveling the roles of vitamin D status and melanin during Covid19

High seroprevalence of SARS_COV-2 in Bergamo: evidence for herd immunity or reason to be cautious?

Environment and the COVID-19 pandemic

SARS-CoV-2 pathophysiology and its clinical implications: An integrative overview of the pharmacotherapeutic management of COVID-19

COVID-19, an opportunity to reevaluate the correlation between long-term effects of anthropogenic pollutants on viral epidemic/pandemic events and prevalence

Lockdown timing and efficacy in controlling COVID-19 using mobile phone tracking

Robust neutralizing antibodies to SARS-CoV-2 infection persist for months

Pathophysiology, transmission, diagnosis, and treatment of Coronavirus Disease 2019 (COVID-19): A review