A Retrospective observational cohort study of the effect of antenatal influenza vaccination on birth outcomes in Cape Town, South Africa, 2015‐2016

Abstract Background There are conflicting data concerning the impact of antenatal influenza vaccination on birth outcomes including low birthweight (LBW), preterm birth, small for gestational age (SGA), and stillbirth. Methods We conducted a retrospective observational cohort study of infants born to women residing in Mitchells Plain, Cape Town. Infants were born at 4 health facilities during May 28 – December 31, 2015 and April 15 – December 31, 2016. We performed crude and multivariable logistic regression, propensity score (PS) matching logistic regression, and inverse probability of treatment weighted (IPTW) regression to assess vaccine effectiveness (VE) against LBW, preterm birth, SGA, and stillbirth adjusting for measured confounders. Results Maternal vaccination status, antenatal history, and ≥1 birth outcome(s) were available for 4084/5333 (76.6%) pregnancies, 2109 (51.6%) vaccinated, and 1975 (48.4%) unvaccinated. The proportion LBW was lower in vaccinated (6.9%) vs. unvaccinated (12.5%) in multivariable [VE 0.27 (95% CI 0.07‐0.42)], PS [VE 0.30 (95% CI 0.09‐0.51)], and IPTW [VE 0.24 (95% CI 0.04‐0.45)]. Preterm birth was less frequent in vaccinated (8.6%) than unvaccinated (16.4%) in multivariable [VE 0.26 (0.09‐0.40)], PS [VE 0.25 (95% CI 0.09‐0.41)], and IPTW [VE 0.34 (95% CI 0.18‐0.51)]. The proportion SGA was lower in vaccinated (6.0%) than unvaccinated (8.8%) but not in adjusted models. There were few stillbirths in our study population, 30/4084 (0.7%). Conclusions Using multiple analytic approaches, we found that influenza vaccination was associated with lower prevalence of LBW (24‐30%) and preterm birth (25‐34%) in Cape Town during 2015‐2016.


| Introduc tion
Seasonal and pandemic influenza virus infections are associated with severe disease outcomes in pregnant women and young infants. [1][2][3][4][5][6] Maternal antenatal influenza vaccination reduces the incidence of influenza among pregnant women and their young infants during the first 4-6 months of life. [7][8][9][10][11] In 2012, the World Health Organization Strategic Advisory Group of Experts on Immunization recommended prioritization of pregnant women for influenza vaccination. 12 Since that time, several countries have introduced influenza vaccination programs including programs that target pregnant women 13 ; however, this recommendation has not been widely implemented in sub-Saharan Africa. 14 There have also been conflicting data concerning the potential impact of antenatal influenza vaccination on birth outcomes including stillbirth, preterm birth, and low birth weight. Influenza pandemics have been associated with fetal loss, [15][16][17] preterm birth, 16,18 and infants born small for gestational age 18 particularly following maternal influenza-associated or acute respiratory hospitalization; however, there appears to be limited association between seasonal influenza epidemics and birth outcomes. [19][20][21][22] A small randomized, controlled trial (RCT) of maternal influenza immunization in Bangladesh 23 found that maternal influenza immunization increased mean birth weight following the influenza season and a larger RCT in Nepal 24 found that year-round maternal influenza immunization reduced the proportion of infants born low birth weight, but similar trials in South Africa 9 and Mali 10 found no impact on these measures.
Likewise, observational studies of the impact of maternal influenza vaccination on birth outcomes including stillbirth, preterm birth, low birth weight, or small for gestational age have not provided conclusive evidence of association [25][26][27][28] and may be subject to bias introduced in vaccination or reporting of comorbidities/risk factors when an adverse birth outcome occurs.
Since 2010, South Africa has offered seasonal influenza vaccination at no cost at public health facilities. 29

| Study design and population
In South Africa, most antenatal care is provided at primary health centers and maternal obstetric units. Women with chronic illnesses, prior pregnancy complications including cesarean section, or who develop high-risk conditions during pregnancy are referred to district hospitals or regional referral hospitals for continuation of antenatal care. We conducted a retrospective observational cohort study of infants born to women residing in Mitchells Plain, a suburb of Cape Conclusions: Using multiple analytic approaches, we found that influenza vaccination was associated with lower prevalence of LBW (24-30%) and preterm birth (25-34%) in Cape Town during 2015-2016.

K E Y W O R D S
influenza vaccines, pregnancy outcome, premature birth vaccines were offered at no cost at all public antenatal clinics except Groote Schuur Hospital's antenatal clinic where most high-risk pregnancies are referred for care. Maternal age, parity, HIV status, medical history, number of antenatal visits, tobacco, and alcohol use were abstracted from antenatal records or birth registers.

| Primary outcome measures
We used influenza surveillance from local general practitioners in the WCP (from the Viral Watch Programme 33 ) to monitor influenza activity during the study period. Birth outcomes of interest included stillbirth (fetal death after 20 weeks gestation), mean birthweight, low birthweight (LBW) (birthweight less than 2500 g), preterm birth (birth before 37 weeks gestation), and small for gestational age (SGA) (birthweight <10th percentile for gestational age using WHO fetal growth charts 34 ). Gestational age was determined by the clinician from last menstrual period, antenatal ultrasound, or fundal height at first antenatal visit. Birthweight was recorded at the time of delivery in grams.

| Vaccination campaign and determination of influenza vaccination status
Influenza vaccines are offered in limited supply at no cost in public antenatal clinics. Influenza vaccines are also offered by private general practitioners and at local pharmacies. During the period of this study, we increased influenza vaccine availability at selected clinics to increase coverage as part of a larger study of the effectiveness of maternal influenza vaccination against infant influenza-associated hospitalization. As part of the study, we kept influenza vaccine registers at study clinics to document vaccine uptake and encouraged documentation of influenza vaccination status of pregnant women in antenatal records and birth registers. We reviewed antenatal records, birth registers, and vaccine registers at clinics to determine maternal vaccination status. Women who did not receive antenatal care and those who received care at clinics that did not provide influenza vaccination were considered unvaccinated in all analyses.
Women who had not entered the second trimester prior to the end of the influenza vaccine campaign and who were not listed in clinic vaccine registers were also considered unvaccinated. Women who received influenza vaccination less than 2 weeks prior to delivery or whose date of vaccination was outside the campaign period were excluded from the primary analyses.

| Statistical analysis
We described the characteristics of vaccinated and unvaccinated women and their infants using frequencies for categorical variables and means for continuous variables. We used Pearson Χ 2 and Wald Χ 2 to assess differences in categorical variables, and two-sample t test for differences in means. We performed multivariable logistic regression to assess vaccine effectiveness against low birth weight, preterm birth, small for gestational age and stillbirth adjusting for measured confounders. We calculated vaccine effectiveness (VE) as 1-odds ratio (OR) for all logistic regression analyses.
We also explored other methods to adjust for observational bias using methodology similar to Walsh et al. 35 We used standardized differences to assess the balance of baseline covariates between the vaccinated and unvaccinated groups and considered an absolute standardized difference below 0.10 indicative of a balanced covariate.
We included the following variables in logistic regression propensity score models: maternal age, campaign year, season of birth, site, gravidity, parity, maternal HIV status, non-HIV chronic illness, smoking, alcohol use, and frequency of antenatal care attendance. Non-HIV chronic illness included anemia (n = 246), pregnancy induced hypertension (n = 184), hypertension (n = 120), asthma (n = 118), diabetes (n = 76), obesity (n = 67), psychiatric disorders (n = 20), heart disease (n = 18), seizures (n = 12), and other comorbidities (n = 98). We then developed inverse probability of treatment weights (IPTWs), which weighted vaccinated pregnancies by the inverse of the propensity score and unvaccinated pregnancies by the inverse of one minus the propensity score. To generate adjusted results, we ran propensity score matching logistic regression models and IPTW weighted logistic regression models to assess the average treatment effect in the treated (ATET) for each outcome except birthweight. Complete data were available on 3869/4084 (94.7%) of pregnancies for inclusion in propensity score analyses. We assessed the impact of vaccination on birth weight and adjusted for measured confounders using inverse probability weighted (IPW) linear regression.
Because we had limited data about outcomes of prior pregnancies, we created additional propensity score models for only primigravidae including all listed variables except parity. IPTWs were assigned using the same methodology but with the propensity score developed for primigravidae. Complete data were available for 1218/1286 (94.7%) of primigravidae for propensity score development.

| Vaccination campaigns
Due to manufacturing challenges, there was a delay in receiving Southern Hemisphere influenza vaccines in 2015. 36 The maternal influenza vaccination campaign began on May 14th during week 20, less than three weeks before influenza activity peaked. During May 14th-August 16th, 2015, 6058 pregnant women were vaccinated in clinics serving the catchment area of the study sites. Influenza A(H1N1)pdm09 was the predominant circulating strain, and there was no vaccine mismatch. In 2016, the vaccine campaign began on April 1st during week 14, and 6461 pregnant women were vaccinated before August 25th, 2016. All but 6 (0.1%) women were vaccinated more than two weeks before the peak of the season. Influenza A(H3N2) was the predominant circulating strain and vaccine effectiveness was 18% (95% CI −55% to 57%) due to poor match with the A(H3N2) strain. 33

| Vaccine coverage and baseline characteristics
We collected data on birth outcomes for 5333 pregnancies during the study period ( Figure 2). Maternal vaccination status, antenatal history, and one or more birth outcomes were available for 4084 (76.6%) pregnancies. Among these 4084 women, 2109 (51.6%) were vaccinated and 1975 (48.4%) were unvaccinated (Table 1)

| Birth outcomes
The mean birth weight among live births included in our analyses (n = 4053) was 3150g (95% CI 3133 to 3167) and was statistically significantly higher in vaccinated compared to unvaccinated women using crude linear regression [114g difference (95% CI 80 to 148),  Table 2). Analyses limited to primigravidae also found a significant difference in mean birth weight between vaccinated and unvaccinated women using crude linear regression [104g difference (95% CI 46 to 162), P < 0.001], but no significant difference remained after adjustment in multivariable linear regression (P = 0.22) or IPW regression (P = 0.10)( Table 3).

| Discussion
Using multiple analytic approaches to adjust for bias associated with receipt of influenza vaccination in this observational study, we found that influenza vaccination was associated with lower preva-  was only significantly associated with lower prevalence of stillbirth in the final multivariable logistic regression models. Campaign year was not significantly associated with lower prevalence of preterm birth, low birthweight, or SGA There are several limitations to this analysis including the retrospective design and the data sources. In South Africa, antenatal records and birth registers often lack sufficient data on pregnancies to adequately adjust for prior adverse birth outcomes. Difference in the timing and method of gestational age assessment, and the low proportion of women that initiate ANC during the first trimester, introduce uncertainty in ascertainment of gestational age which may impact assessment of preterm and SGA. Likewise, important predictors of infant weight including maternal education and pre-pregnancy body mass index are not routinely collected or reported. Furthermore, the decision to receive influenza vaccine was not randomized. We identified a high proportion of women with non-HIV chronic illness (24.2%) in this population; however, there was insufficient data to assess birth outcomes according to different disease types or severity of illness. The timing of influenza vaccination both in relation to the gestational age of the infant and the influenza transmission season is also important in understanding the potential impact of vaccination on birth outcomes. We adjusted for season of birth by determining if the infant was delivered before, during, or after the influenza season. This adjustment was balanced in the propensity score algorithm, but the standardized difference (0.323) following propensity score adjustment indicates that IPTW weighting may not have sufficiently adjusted for bias associated with the timing of vaccination.

Ratio in Vaccinated vs
In conclusion, in this retrospective observational study we identified statistically significant associations between antenatal influenza vaccination and reduced risk of low birthweight and preterm delivery. In future studies, it would be valuable to implement cluster randomization of vaccination, prospective data collection including maternal comorbidities and their severity, history of prior pregnancies, maternal BMI, education level, and socioeconomic status.

Disclaimer
The findings and conclusions of this manuscript are those of the authors and do not reflect the official position of the US Centers for Disease Control and Prevention.