Using the SINAN database, we identified 3,298 SARI cases in Ceará during 2013–2018, including 145 cases among pregnant women (Table 1). We linked the SINAN and SINASC databases to identify 61 children born to mothers who had >1 SARI during pregnancy. We used the same databases to identify 122 children born to age-matched pregnant women who did not have recorded SARI during pregnancy.

We observed equal proportions of SARI cases among male and female patients registered in the SINAN database of 3,298 overall SARI cases in Ceará. Children <5 years of age comprised 27%–61% of patients; children <6 months of age comprised 10%–31% of patients. Older adults (7%–32%) and pregnant women (2%–38%) also comprised large proportions of patients. We observed cases of seasonal H1N1 throughout the study period, notably in 2013 (54%), 2014 (75%), 2016 (83%), and 2018 (69%). The highest number of SARI cases occurred in 2018, mostly caused by seasonal H1N1 and influenza B viruses (23%). We observed sporadic cases of seasonal influenza caused by other H1 subtypes in 2015 and 2017 and seasonal H3 subtypes in 2013 and 2017–2018 (Table 1; Figure 3, panels A, B). Influenza death rates varied from 0%–21%; the peak death rate occurred during a season predominated by H3 subtypes.

The median age of pregnant women who had SARI was 26 years (range 15–44 years). Among 145 pregnant women who had SARI, 43 (32%) had laboratory-confirmed influenza. Among those 43 women, 42% had illnesses caused by H1N1, 33% by H3 subtypes, 7% by influenza A viruses without an identified subtype, and 19% by influenza B subtypes (19%). We identified no deaths caused by SARI in pregnant women (Table 2).

To better visualize the relationship between birth outcomes, SARI, and influenza, we overlaid sets of data for 2018 on the same plot (Figure 4, panel A). We found that average birthweight decreased shortly before influenza season. During 2018, birthweight peaked in the first week of the year. By week 15, average birthweight had fallen by ≈40 g (Figure 4, panel A). After the influenza vaccination campaign ended, SARI cases declined and birthweights returned to their yearly averages. For all years of the study, we found lower average gestational scores, which indicates a higher proportion of preterm births, before and during influenza season (Figure 4, panel B).

Each year, average birthweights oscillated by up to 40 g, or 1%–2% of total birthweight (Figure 5). In February, a month associated with worse birth outcomes, 15.5% (8,399/54,311) of children were born prematurely (<37 weeks), whereas in October, a month associated with better birth outcomes, 10.7% (6,552/61,067) of children were born prematurely. These data indicate that circannual oscillations in birth outcomes might be associated with SARI and seasonal influenza in Ceará.

Children born to mothers who had SARI during pregnancy had significantly lower birthweights (p = 0.02), higher risk for prematurity (p = 0.03), shorter gestation times (p<0.01), and lower Apgar scores at 5 minutes after birth than children in the control group (p<0.01). Mothers who had SARI during pregnancy had significantly less formal education than mothers who did not have SARI (p<0.01). Mothers with SARI had a significantly lower number of previous pregnancies (p = 0.01), previous vaginal births (p<0.01), and previous live births (p = 0.01). Mothers with SARI had a higher number of previous cesarean sections (p<0.01). Cesarean deliveries and medical assistance were more frequent in cases versus controls (86.7% vs. 0.8%; p<0.01) (Table 3).

We used multiple logistic regression to identify predictor variables independently associated with SARI during pregnancy. First, we examined 11 significant variables identified by univariate analysis (Table 3), of which 5 showed >40% collinearity. We had an adequate sample size (123 cases) to run a logistic regression for these 5 variables (23,25). The overall model fit showed a χ² value of 23.135 (df = 6; p<0.01). The Cox and Snell test and Nagelkerke test indicated variances between 17.1% and 23.2%. Including predictor variables increased model accuracy from 61% to 68%. We found that birthweight (p = 0.03) and attendance of birth by a physician (p = 0.04) were significantly associated with SARI during pregnancy (Table 4; Figure 6).

We documented 3,298 SARI cases in Ceará, Brazil, during 2013–2018. Cases occurred predominantly in younger children, especially children <6 months of age, as well as older adults and pregnant women. These data are consistent with previous studies showing higher rates of infection among younger populations but increased death rates among older adults (26).

H1N1 was the dominant influenza subtype during seasonal epidemic outbreaks, illustrating the capacity of this strain to recirculate and co-circulate with other seasonal influenza strains (Table 1; Figure 3, panel A). H1N1 caused a high death rate throughout the study. However, in 2015, when seasonal H1 strains predominated in Ceará, we observed a lower overall death rate among influenza patients. These data are consistent with prior literature showing more deaths associated with H1N1 (27). The mortality rate in our study might be attributable to the mistiming of vaccination campaigns, which occurred during and after peak influenza activity in Ceará. The state has high vaccination coverage, suggesting that earlier administration of influenza vaccines might reduce death and disease. Mistiming of immunization schedules also might explain the unusually high disease incidence among infants <6 months of age. The World Health Organization does not recommend immunization in this age group. Consequently, immunization of pregnant mothers is especially crucial for passive immunity against influenza during the first 6 months of life.

In this study, we assessed whether SARI during pregnancy was associated with a higher risk for low birthweight or prematurity. We found statewide correlations between peak influenza activity and nadirs in birthweight and gestational length. Furthermore, we confirmed associations of maternal SARI with low birthweight and preterm birth in matched mother–infant pairs. These associations remained significant when adjusted for confounders by multiple logistic regression. Our findings agree with 2 recently published studies (7,8) showing an association of SARI, including influenza, among pregnant women with low birthweights. Pregnant mothers who had SARI were more likely to require medical assistance during labor than those who did not have SARI.

Our study aligns with earlier reports showing the importance of prevention and adjustment of influenza vaccine campaign schedules to avoid complications of influenza (6,28–31). Previous studies show the importance of the first 1,000 days of life in reducing undernutrition, enteric infections, and risk for metabolic syndrome and cardiovascular diseases (32,33). Neurocognitive, physical, and educational deficits have been well-documented among children exposed in utero or during the first months of life to influenza and other diseases such as enteric infections (10–12,34).

The first limitation of our study is that we analyzed only cases of influenza associated with SARI and did not include cases of mild-to-moderate influenza. However, our analyses of statewide birth outcomes detected substantial periodicity in birthweights and gestational length; poorer outcomes coincided with influenza season. Maternal influenza also might affect other perinatal outcomes, such as medical necessity for caesarean birth. Second, our nested observational descriptive study cannot infer a causal relationship between maternal SARI and adverse birth outcomes. However, the associations were robust to logistic regression adjusted for several potential confounders. In addition, because hospitalization is part of the case definition for SARI, public and private hospitals (but not private clinics) are required to report SARI cases to SINAN. Although many private clinics do report, most reported cases come from public institutions. Thus, we might not have analyzed all SARI cases in Ceará. Finally, our results suggest that asynchronous vaccination schedules might be associated with adverse influenza outcomes in Ceará, but we did not model the extent to which earlier immunization or the use of vaccine strains from the Northern or Southern Hemispheres might mitigate these outcomes. Recent epidemiologic models suggest Ceará is the starting point for influenza transmission from the semiarid region in southern Brazil, hence earlier immunization in Ceará might have substantial benefits for the region and country (5). We did not account for infections with Zika virus as a potential confounder of our findings because the reported Zika incidence was 0 during 2013–2016; however, testing for Zika was not routinely performed during this time period. The state had low Zika incidence during the study: 5.6 cases/100,000 persons in 2017 and 0.2 cases/100,000 persons in 2018 (35).

In conclusion, our results show that late timing of influenza vaccination in Ceará, a populous semiarid state in Brazil with high vaccination coverage, correlates with adverse perinatal outcomes. In addition, we found that mean birthweight and rates of prematurity followed an annual periodicity, suggesting additional associations with seasonal influenza. Finally, we confirmed a robust association of maternal SARI with poor birth outcomes using an observational descriptive study design. Further work is urgently needed to model and study the optimal timing, potential impact, logistics, economics, and implementation of such a diversified national influenza vaccine strategy. Because Ceará is the presumptive starting point for an annual north-to-south pattern of seasonal influenza transmission in Brazil (15), our data indicate earlier timing of national immunization campaigns, ideally before seasonal influenza circulation in Ceará, might provide substantial benefits not only for women and children in the semiarid region but also for Brazil as a country.