Inactivated Influenza Vaccine During Pregnancy and Risks for Adverse Obstetric Events

Abstract

Objective:

To compare risks for adverse obstetric events between females who did and did not receive trivalent inactivated influenza vaccine during pregnancy.

Method:

This retrospective, observational cohort study was conducted at seven Vaccine Safety Datalink sites. Pregnancies were identified from administrative and claims data using a validated algorithm. Females vaccinated while pregnant from 2002 to 2009 were matched 1:2 with replacement to unvaccinated pregnant females. Using generalized estimating equation method with a Poisson distribution and log link, we evaluated the association of trivalent inactivated influenza vaccine with 13 outcomes. Given our large sample size and multiple comparisons (19 contrasts) a cutoff for significance of P<.005 was selected a priori.

Results:

Our cohort included 74,292 vaccinated females matched on age, site, and pregnancy start date with 144,597 unvaccinated females. We did not observe increased risks within 42 days of vaccination for hyperemesis, chronic hypertension, gestational hypertension, gestational diabetes, proteinuria or urinary tract infection. Using a risk window from vaccination through pregnancy end, we did not observe increased risks after vaccination for proteinuria, urinary tract infection, gestational hypertension, preeclampsia or eclampsia, chorioamnionitis, puerperal infection, venous complications, pulmonary embolism, or peripartum cardiomyopathy. Reduced risk for gestational diabetes after vaccination was detected (adjusted hazard rate ratio 0.88, 95% confidence incidence 0.83–0.93), likely due to healthy vaccine bias or to earlier detection among vaccinees.

Conclusion:

In this large cohort, influenza vaccination during pregnancy was not associated with increased risks for medically attended adverse obstetric events.

Précis

Receipt of trivalent influenza vaccine during pregnancy is not associated with adverse obstetric events.

Introduction

The influenza vaccine is strongly recommended for pregnant women during all trimesters of pregnancy.^1,2 Pregnant women have been targeted for vaccination in order to prevent illness in mothers and their offspring.³ Recent studies have highlighted the effect of maternal vaccination in preventing severe influenza infections in pregnant women and newborns.^4,5

Despite the demonstrated beneficial effects of vaccination, influenza vaccine coverage among pregnant women remains low. In the 2010–2011 influenza season, only half of pregnant women reported receiving an influenza vaccine.⁶ Fears regarding the safety of influenza vaccine during pregnancy remain a persistent barrier to vaccine uptake.^7–9 Prior studies have not suggested any association between trivalent inactivated influenza vaccine (TIV) and maternal or fetal adverse events.^4,10–13 However, only a few previous studies have specifically examined the effect of vaccination on pregnancy complications, such as preeclampsia or gestational diabetes.^12,14–17 These outcomes are of importance for the health of women during pregnancy and beyond. In addition, these and other obstetric outcomes also result in increased risk for adverse birth outcomes.

In prior work, we reported that influenza vaccination during pregnancy was not associated with potential vaccine-related acute adverse events, including allergic and local reactions, thrombocytopenia, seizures and other acute neurologic events.¹⁸ In the current study we expand on this research, utilizing our previously described multisite observational cohort to compare risks for adverse obstetric events between females who did and did not receive TIV during pregnancy.

Materials and Methods

This retrospective, observational cohort study was conducted within the Vaccine Safety Datalink (VSD). The VSD is a collaborative effort between the Immunization Safety Office of the Centers for Disease Control and Prevention (CDC) and managed care organizations (MCOs) to monitor vaccine safety within the United States and includes data on approximately 3% of the U.S. population.¹⁹ A detailed description of how the cohort utilized in this study was assembled has been previously presented.¹⁸

Briefly, data for these analyses came from seven VSD sites (Group Health Cooperative, HealthPartners, Kaiser Permanente Colorado, Kaiser Permanente Northwest, Kaiser Permanente Northern California, Kaiser Permanente Southern California, and Marshfield Clinic Research Foundation). Pregnant women with continuous enrollment in one of the participating VSD sites were identified using a validated algorithm, adapted for use in the VSD by Naleway and colleagues.²⁰ The algorithm uses claims, electronic medical record and birth certificate data to identify pregnancy episodes. Pregnancy outcomes, including live births, stillbirths, and spontaneous abortions were identified by International Classifications of Diseases, Ninth Revision (ICD-9) codes, linkage with birth certificates, or both. Gestational age at pregnancy outcome and pregnancy start dates were estimated from ICD-9 codes, birth certificates (for pregnancies ending in live birth), and timing of routine prenatal procedures (eg, nuchal translucency screening). Receipt of TIV was identified through claims and site-based vaccine registries. Timing for receipt of TIV was classified by gestational week and trimester.

Exclusions from the cohort included multiple gestations, ectopic pregnancies, gestational trophoblastic disease, therapeutic abortions, and pregnancies where the outcome could not be determined from available claims and birth certificate data. Females who received other vaccines while pregnant and females with no outpatient medical visits recorded in our data during pregnancy were also excluded. In addition, females vaccinated during their first or second week of gestation or within 1 week of the end of their pregnancy were excluded. In order to avoid misclassification of postpartum vaccines, females receiving TIV within 1 week of the end of their pregnancy were also excluded.

After applying the above inclusion and exclusion criteria, all pregnant females aged 14–49 years who received TIV from June 1, 2002 to July 31, 2009 were matched 1:2 with replacement to females not vaccinated during pregnancy using a variable optimal matching algorithm.²¹ Match variables included maternal age at pregnancy outcome, estimated pregnancy start date, and site. Unexposed females were assigned an index date that was equal to their TIV-exposed match’s gestational age at vaccination.

Adverse obstetric events are defined as new, prespecified, medically attended pregnancy-related comorbidities or pregnancy complications. For the current analyses, background risks for specific obstetric outcomes vary by gestational week of pregnancy. For example, by definition, preeclampsia has an onset of 20 weeks of gestation or later. However, first trimester exposures that affect placentation may predispose women to be at increased risk for preeclampsia. Thus, as described below, in order to evaluate potential risks associated with vaccination, each obstetric event required attention to both the exposure window and timing of vaccination and the expected timeframe for diagnosis.

All potential adverse obstetric events were identified from ICD-9 codes recorded in maternal electronic health data occurring at inpatient, outpatient, or emergency department visits. An adverse obstetric event was defined as the onset of a new, medically attended pregnancy-related comorbidity or pregnancy complication. In order to reduce the likelihood of including preexisting conditions, events occurring on the day of vaccination (Day 0) were utilized only if diagnosed at an inpatient or emergency department visit. Specific outcomes and windows were selected a priori, based on the demonstrated postvaccination inflammatory response,²² prior vaccine safety studies,^12,23 and pathophysiology of the obstetric event.²⁴ In addition, outcomes of high severity and public health importance were evaluated. Obstetric events included: hyperemesis, chronic hypertension, gestational hypertension, mild preeclampsia, severe preeclampsia or eclampsia, gestational diabetes, proteinuria, urinary tract infection, chorioamnionitis, puerperal infections, venous complications of pregnancy, pulmonary embolus, and peripartum cardiomyopathy. Events were only included if they represented new diagnoses. Strategies to identify and exclude preexisting conditions were applied. (Appendix 1, available online at http://links.lww.com/xxx)

The timing for vaccination exposures and risk windows were selected for each obstetric event accounting for the temporal nature of the reproductive process (Appendix 2, available online at http://links.lww.com/xxx). For outcomes with onset early in pregnancy, only first-trimester vaccination and a 42-day risk window could be used. For events occurring late in pregnancy or postpartum, a 42-day risk window would miss a majority of outcomes and rates would be inversely related to preterm delivery rates.²⁵ Thus, these analyses focused on vaccination occurring at 20 weeks or later and a risk window from vaccination or index date through the end of pregnancy or the peripartum period was used. For venous complications of pregnancy, pulmonary embolism, and peripartum cardiomyopathy, risk windows were extended to include the period from vaccination or index date through 30 days postpartum. For preeclampsia or eclampsia, where disease onset is late in pregnancy but risks may be associated with placentation²⁴, only end of pregnancy windows were evaluated, but vaccine exposures throughout pregnancy were studied. For other outcomes occurring at 20 weeks of gestation or later that were not immediate causes for delivery, exposures at 20 weeks or later with both 42-day and end of pregnancy windows were studied. Finally, for outcomes occurring throughout pregnancy, all exposures were included and both 42-day and end of pregnancy risk windows were evaluated (Appendix 2, available online at http://links.lww.com/xxx).

Data on preexisting conditions, including pulmonary disease, hypertension, diabetes, heart disease, neurologic or rheumatologic conditions, and hypercoagulability were abstracted from automated claims data starting 6 months prior to the last menstrual period (LMP) through the vaccination or index date. Similarly, data on pregnancy complications occurring prior to vaccination, including hemorrhage in early pregnancy, gestational hypertension, hyperemesis gravidarum, gestational diabetes, proteinuria, and obesity complicating pregnancy were abstracted from automated claims data from the LMP through the vaccination or index date. Additional outcome-specific risk factors were also identified through ICD-9 codes. Some conditions were both outcomes and risk factors. For example, proteinuria was classified as an outcome if its onset was after the vaccination or index date but it was a risk factor for preeclampsia and eclampsia if its onset was before the vaccination or index date.

To assess covariates, hospitalizations, emergency department visits, and outpatient and urgent care visits before vaccination or index date were recorded. In the absence of socioeconomic variables at the individual level, we used socioeconomic proxies at the census tract level, defined for each female as the percent of families within their Census tract with incomes below 150% of the federal poverty level.²⁶ Missing census data were imputed using the expectation maximization algorithm.²⁷

Chi-square and median two-sample tests were used to compare baseline characteristics between vaccinated and unvaccinated populations. We report 42-day incidence rates in our exposed and unexposed cohorts per 1,000 pregnancies for these adverse obstetric events: hyperemesis gravidarum, chronic hypertension, gestational hypertension, gestational diabetes, proteinuria and urinary tract infection. Using the generalized estimating equation (GEE) method to account for the matching effect with a Poisson distribution and log link we evaluated the association of TIV exposure with 42-day adverse obstetric events. We first created crude models, and then adjusted for demographic and outcome specific risk factors. Incidence rate ratios with 95% confidence intervals (CI) are presented for adjusted models.

Similarly, we report incidence rates per 1,000 pregnancies after vaccine or index date for these outcomes: gestational hypertension, mild preeclampsia, severe preeclampsia or eclampsia, gestational diabetes, proteinuria, urinary tract infection, chorioamnionitis, puerperal infection, venous complications, pulmonary embolus and peripartum cardiomyopathy. Rates were calculated using the GEE method to account for the matching effect with a Poisson distribution and log link. Time to the end of the pregnancy was incorporated as an offset in the Poisson models. This accounted for differences in gestational duration as spontaneous abortions, stillbirths, and live births were all included in the study cohort. Hazard rate ratios with 95% CIs are presented for adjusted models, based on Cox-regression models.

Most of the adverse obstetric events under investigation (individual or groups of ICD-9 codes) were expected to occur during pregnancy with a background prevalence of at least 1–10 per 1,000 pregnancies (Appendix 2, available online at http://links.lww.com/xxx); 42-day incident rates for these conditions would be expected to be lower, and vary based on the timing in pregnancy of the 42-day window. Given our large sample size and multiple comparisons (13 outcomes, with a total of 19 contrasts) a conservative cutoff for significance of P<.005 was selected a priori.²⁸ With an estimated 74,000 females exposed to TIV while pregnant, and an alpha of 0.005, our analyses had power of 80% to detect an incidence rate ratio of 1.1 for outcomes with an incidence rate of 10 per 1000 pregnancies and a rate ratio of 1.04 for outcomes with an incidence rate of 100 per 1000 pregnancies. Venous complications of pregnancy, pulmonary embolus and peripartum cardiomyopathy were all expected to be rare events, with background incidence of less than 1 per 1000 pregnancies. For these outcomes we had 80% power to detect a hazard rate ratio of 1.4. For analyses restricted to females vaccinated at 20 weeks or greater or females vaccinated in their first trimester, our power was reduced. All analyses were done using SAS 9.2. This study was approved by the Institutional Review Boards from all participating sites.

Results

A total of 807,563 pregnancies occurring between 2002 and 2009 were identified. After applying exclusions, 407,745 pregnancies remained in the study cohort and eligible for matching. Our final cohort was comprised of 74,292 TIV-exposed pregnancies matched on age, site, and estimated pregnancy start date with 144,597 unvaccinated pregnancies (Figure 1).

Figure 1.

Flow of participants through the study. The group receiving trivalent inactivated influenza vaccine during their first trimester was compared with the group that was not vaccinated during the first trimester. *Continuous enrollment starting 6 months prior to pregnancy start, through pregnancy including 2 months postpartum ^†Multiple gestations, ectopic pregnancies, gestational trophoblastic disease, therapeutic abortions and undefined. ^‡Outside of window is defined as within 2 weeks following last menstrual period or within last week of pregnancy. ^§One hundred nineteen pregnancies received trivalent influenza vaccine during consecutive seasons and were retained as two episodes. ^||Control episodes (matched 1:2 controls with replacement, matched sets were retained with at least 1 control. ^¶Four thousand, seven hundred eighty-six trivalent influenza vaccine episodes had 1 control (3%). ^#Unique pregnancies. **Selected pregnancies with replacement.

Pregnant females ranged in age from 14–49 years (mean age 30.8±5.6 years). Females received influenza vaccine throughout pregnancy, including 21,107 (28.4%) in their first trimester, 32,847 (44.2%) in their second and 20,338 (27.4%) in their third. Compared with unvaccinated females, preexisting conditions including hypertension and other heart disease, diabetes, pulmonary and rheumatologic conditions were all significantly more common among vaccinated females (Table 1). Similarly, baseline pregnancy complications, including gestational hypertension and gestational diabetes occurred more frequently among vaccinated females. However, vaccinated females were slightly less likely than unvaccinated females to be hospitalized prior to their vaccination/index date (5.4% compared with 5.7%). Outpatient medical visits were more frequent among TIV-exposed women, both before vaccination (5.1 encounters compared with 4.6 encounters) and after vaccination (7.6 encounters compared with 6.8 encounters). Among pregnancies ending in a live birth, for both vaccinated and unvaccinated groups, mean gestational age at delivery was 39.0 weeks (Table 1).

Table 1.

Baseline Comparisons Between Females Exposed and Unexposed to Trivalent Inactivated Influenza Vaccine During Pregnancy

	Vaccinated n=74,292	Unvaccinated n=144,597	P
Age at pregnancy outcome (y)			.36
Younger than 20	2,753(3.7)	5,402 (3.7)
20–34	54,421 (70.6)	102,389 (70.8)
35 or older	19,118 (25.7)	36,806(25.5)
Medical care during first trimester	65,978 (88.8)	124,806 (86.3)	<.001
Hospitalized prior to vaccine or index date	3983(5.4)	8209 (5.7)	.002
Preexisting comorbidities*
Pulmonary	4,672 (6.3)	6,557 (4.5)	<.001
Hypertension	1,766(2.4)	2,976(2.1)	<.001
Diabetes	1,336 (1.8)	1,865 (1.3)	<.001
Heart disease	883 (1.2)	1,362 (0.9)	<.001
Neurologic	388 (0.5)	669 (0.5)	.83
Rheumatologic	304 (0.4)	466 (0.3)	.06
Hypercoagulability	134 (.2)	210 (0.2)	.05
Any comorbidity	8,570(11.5)	12,953(9.8)	<.001
Preexisting pregnancy complications†
Hemorrhage in early pregnancy	6,577(8.9)	12,831(8.9)	.87
Gestational hypertension	1,083(1.5)	1,775(1.2)	<.001
Excessive vomiting	3,951(5.3)	7,874(5.5)	.21
Gestational diabetes	3,322 (4.5)	5,454(3.8)	<.001
Proteinuria	109(0.2)	214(0.2)	.94
Obesity	4,148(5.6)	7,459(5.2)	<.001
Any pregnancy complication	16,169 (21.8)	30,371(21.0)	<.001
Number of outpatient visits
Before vaccine or index date	5.1 ± 4.1	4.6 ± 4.0	<.001
After vaccine or index date	7.6 ± 4.9	6.8± 4.6	<.001
Poverty‡	13.0 ± .12	14.9± .13	<.001
Gestational age at delivery (wk)§	39.0 ± 1.62	39.0± 1.68	.50

Data are n (%) or mean ±standard deviation unless otherwise specified.

^*Diagnoses at inpatient, outpatient or emergency visits, from 6 months prior to estimated last menstrual period through the vaccination or index date.

^†Diagnoses at inpatient, outpatient or emergency visits from the estimated last menstrual period through the vaccination or index date.

^‡Percent of families in census tract with income below 150% of federal poverty level ± standard error

^§For pregnancies ending in a live birth, mean gestational age ± standard deviation

Using a 42-day window, risks for eight adverse obstetric events were compared between vaccinated and unvaccinated females. Based on a predetermined cutoff for significance of P<.005, we did not observe increased risks for the new onset of hyperemesis, chronic hypertension, gestational hypertension, gestational diabetes, proteinuria or urinary tract infection within 42 days after influenza vaccination. We did note a statistically significant reduced risk for the diagnosis of gestational diabetes (adjusted RR 0.89, 95% CI 0.82–0.96; P=.004) (Table 2).

Table 2.

Incidence Rates and Adjusted Incidence Rate Ratios For Adverse Obstetric Events Occurring Within 42 Days of Receiving Trivalent Inactivated Influenza Vaccine

Outcome	42-Day Events Number of events (Rates per 1000 pregnancies)			Multivariable Analyses*
	Vaccinated	Unvaccinated	P	Adjusted Incidence Rate Ratio (95% CI)	P
Vaccinated in first trimester	n=21,107	n=40,738
Hyperemesis	427 (20.2)	907 (22.3)	.10	.88 (.79–0.99)	.03
Chronic hypertension †‡	11 (.5)	30 (.7)	.32	.67 (0.34–1.34)	.26
Vaccinated at 20 weeks or greater	n=38,038	n=74,192
Gestational hypertension†‡	504 (13.2)	973 (13.1)	.85	.99 (.89–1.11)	.89
Gestational diabetes‡	868 (22.8)	1908 (25.7)	.003	.89 (.82–.96)	.004
Vaccinated in any trimester	n=74,292	n=144,597
Proteinuria†	58 (.80)	153 (1.1)	.048	.73 (.54–.99)	.04
Urinary tract infection	609 (8.2)	1305 (9.0)	.049	.92 (.84–1.02)	.10

CI, confidence interval.

^*All outcomes adjusted for receipt of medical care in the first trimester, hospitalization prior to vaccination or index date, and poverty; additional adjustments were outcome specific

^†Also adjusted for diabetes, gestational diabetes

^‡Also adjusted for obesity

Using a risk window from vaccination or index date through the end of pregnancy, we compared risks for 11 potential adverse obstetric events between TIV-vaccinated and unvaccinated pregnant females. Based on predetermined levels for significance, we did not observe increased risks for the new onset of gestational hypertension, proteinuria, urinary tract infection, puerperal infections, venous complications, pulmonary embolus, or peripartum cardiomyopathy after vaccination. For mild preeclampsia and severe preeclampsia or eclampsia, no increased risks after first-trimester vaccination or vaccination in any trimester were observed. Similar to analyses using a 42-day window, when using an end of pregnancy window, we a found a statistically significant reduced risk for gestational diabetes after influenza vaccination (adjusted hazard rate ratio 0.88, 95% CI 0.83–0.92; P<.0001). In addition, we observed a statistically nonsignificant increased risk for chorioamnionitis after vaccination (adjusted hazard rate ratio 1.08, 95% CI 1.01–1.14; P=.01) (Table 3).

Table 3.

Incidence Rates and Adjusted Hazard Rate Ratios For Adverse Obstetric Events Occurring During Pregnancy or Postpartum Period After Trivalent Inactivated Influenza Vaccination

Outcome	Events After Vaccination Number of events (Rates per 1000 pregnancies)			Multivariable Analyses*
	Vaccinated	Unvaccinated	P	Adjusted Hazard Rate Ratio (95% CI)	P
Vaccinated in first trimester	n=21,107	n=40,738
Mild preeclampsia†‡§	657 (30.0)	1,249 (31.4)	.45	.94 (.86–1.04)	.22
Severe preeclampsia or eclampsia†‡§	364 (16.6)	617 (15.5)	.27	1.05 (.92–1.19)	.51
Vaccinated at 20 weeks or greater	n=38,038	n=74,192
Gestational hypertension†§	1,663(44.3)	3,046(41.9)	.053	1.04(.98–1.11)	.19
Gestational diabetes§	1,773(50.1)	3,925(57.0)	<.0001	.88(.83–.93)	<.001
Chorioamnionitis	1,615(42.9)	2,861(39.0)	.002	1.08(1.02–1.15)	.01
Puerperal infection§	363(9.5)	728(9.8)	.64	.96(.85–1.09)	.55
Venous complications§	21(.55)	63(.85)	.08	.64(.39–1.05)	.08
Pulmonary embolus§	11(.20)	17(.16)	.56	1.23(.58–2.64)	.59
Peripartum cardiomyopathy§	13(.24)	37(.35)	.24	.66(.35–1.24)	.20
Vaccinated in any trimester	n=74,292	n=144,597
Mild preeclampsia†‡§	2,338 (18.2)	4,422 (18.2)	.73	.97 (.93–1.02)	.30
Severe preeclampsia or eclampsia†‡§	1,189 (9.2)	2,293 (9.4)	.44	.95 (.88–1.02)	.14
Proteinuria†	400(5.3)	832(5.8)	.09	.88(.78–1.00)	.04
Urinary tract infection	1,905(25.5)	3,655(26.0)	.57	1.0(.94–1.05)	.91

CI, confidence interval.

Women vaccinated in the first trimester and their unexposed matches had a mean follow-up of 6.3 months; for women vaccinated at 20 weeks or greater and their unexposed matches, mean follow-up was 2.4 months; for women vaccinated in any trimester and their unexposed matches, mean follow-up was 4.1 months.

^*All outcomes adjusted for receipt of medical care in the first trimester, hospitalization prior to vaccination or index date, and poverty; additional adjustments were outcome specific.

^†Also adjusted for diabetes, gestational diabetes

^‡Also adjusted for proteinuria

^§Also adjusted for obesity

Among unexposed females, obstetric events occurred at or near their expected background rates (Appendix 2, available online at http://links.lww.com/xxx). The most common 42-day obstetric event was gestational diabetes, which occurred at a rate of 25.7 per 1,000 unexposed pregnancies. The remaining 42-day outcome rates ranged from 0.7–22.3 per 1,000 pregnancies. Rates for obstetric events among unexposed women occurring after the index date through the end of pregnancy or postpartum ranged from 0.17 per 1,000 pregnant women for pulmonary embolus to 57.1 per 1,000 pregnant females for gestational diabetes.

Discussion

In this large, multisite, observational study, using both 42-day and end of pregnancy risk windows, no concerning risks for adverse obstetric events after influenza vaccination were identified. This study provides needed data for providers and expectant mothers on the safety of influenza vaccine during pregnancy. This study was unique as we focused on pregnancy-related comorbidities that directly affect maternal health, such as gestational diabetes, preeclampsia, and puerperal infections. Although pregnant women report fears that the influenza vaccine will adversely affect their health,^7–9 few studies to date have specifically studied these maternal outcomes.^12,14,16

One strength of this study was our large sample of over 70,000 females vaccinated during pregnancy, including over 20,000 vaccinated during their first trimester. To our knowledge, only one prior study has examined vaccination and risks for adverse obstetric events for a nonadjuvanted influenza vaccine in a U.S.-based cohort. In 2005 Munoz and colleagues compared risks for preeclampsia, gestational diabetes and other maternal health outcomes between 225 vaccinated and 826 unvaccinated women.¹² More recently, several groups have presented data from European and South American cohorts, including risks for gestational diabetes and preeclampsia for nearly 30,000 women who received an adjuvant H1N1 vaccine during pregnancy.^14,15,16,17 These studies did not detect any risks for adverse obstetric events. Our large sample allowed us to provide additional data on outcomes that these previous studies may have been underpowered to address, including specific risks after first-trimester vaccination.

An additional strength of the current analyses was our ability to compare rates for medically attended events between vaccinated and unvaccinated females. In the United States and abroad, passive reporting systems have been used to demonstrate the safety of TIV administered during pregnancy.^13,29,30 However, these systems are prone to underreporting and lack a suitable denominator. For example, Moro and colleagues noted that between 1990 and 2009, only one case of gestational diabetes after TIV in had been reported to the U.S.-based Vaccine Adverse Events Reporting System¹³ In contrast, we reported over 868 new cases of gestational diabetes occurring within 42-days of TIV. While our rate may seem high, we also found 1,908 matched unexposed females were newly diagnosed with gestational diabetes in this same 42-day interval.

It is reassuring that we did not detect any safety signals after TIV vaccination. Nevertheless, our findings on chorioamnionitis merit further discussion. Chorioamnionitis is estimated to affect between 1% and 4% of births and is a major risk for premature delivery, cerebral palsy and sepsis in newborns.³¹ In our unexposed population, the rate for chorioamnionitis was 39 per 1,000 pregnancies, within the upper range of the estimated background prevalence. Among TIV-exposed females in our cohort, the rate for chorioamnionitis was 42.9 per 1,000 pregnancies. After adjustments, the adjusted hazard rate ratio was 1.08 (95% CI 1.02–1.15; P=.01). A potential mechanism for an increased risk for chorioamnionitis after vaccination would be the measurable vaccine-induced inflammatory response.²² However, we doubt that our results represent a true association. First, the rate ratio we report was not statistically significant based on our predetermined cutoff of P<.005. As our study included 19 comparisons, there was approximately a 1 in 6 chance that at least one outcome may be expected to reach a p-value of .01. More importantly, if vaccination were a true cause for chorioamnionitis, we would expect new diagnoses to occur in close proximity to vaccination. In our cohort, the average time between TIV and onset of chorioamnionitis was 71 days (data available upon request). In addition, due to data availability, we were not able to fully adjust our models for important chorioamnionitis risk factors, including prolonged rupture of membranes, nulliparity, number of vaginal examinations, and the presence of genital tract pathogens.

We found a reduced risk for gestational diabetes after influenza vaccination. This finding most likely reflects residual unmeasured confounders and healthy vaccinee bias, observed frequently in nonpregnant populations.³² For example, vaccinated females may be more likely than those unvaccinated to engage in other healthy behaviors placing them at lower risk for diabetes. Compared with unvaccinated females, those vaccinated during pregnancy had more medical encounters both before and after vaccination, possibly reflecting increased health-seeking behaviors. In addition, vaccinated women had more medical encounters prior to their vaccination or index date and thus may have been more likely to have their gestational diabetes detected prior to vaccination. On the other hand, influenza infection is a risk for ketoacidosis in diabetics³³ and antecedent infections are a commonly reported trigger for the onset of type 1 diabetes. Further studies should explore whether influenza infection may result in increased risk for gestational diabetes and whether vaccination could reduce this risk.

Limitations to this observational study should be noted. As previously described¹⁸ factors that were outside the scope of this study but could affect the outcomes of interest, such as smoking status or weight gain during pregnancy, may have differed between vaccinated and unvaccinated females. Second, as detailed in our prior related study¹⁸ it is possible that females vaccinated at pharmacies or at their workplace were misclassified as unvaccinated, potentially biasing our results. In our study, these vaccinations would only be captured if a patient informed their health care provider who then manually entered this into the electronic medical record. Finally, as with any study utilizing electronic health data, misclassification of outcomes was also possible. However, it is reassuring that for many outcomes, the rates observed in unvaccinated women were consistent with published background rates.

The current study expands the existing literature regarding the safety of influenza vaccination during pregnancy. Findings from our research support the current Advisory Committee on Immunization Practices and American College of Obstetricians and Gynecologists recommendations for the administration of influenza vaccine during pregnancy.