Pregnant women have been considered a high-risk group for influenza complications since the 1918 and 1957 pandemics.1,2 In 1997, the Centers for Disease Control and Prevention Advisory Committee on Immunization Practices expanded U.S. recommendations for pregnant women from only those with an underlying high-risk medical condition to include healthy pregnant women in the second or third trimester of pregnancy.3,4 In 2004, the recommendation was further expanded to include pregnant women in any trimester.5 Current European guidelines also recommend the seasonal vaccination of all pregnant women, regardless of trimester.6,7
Evidence supporting the safety of influenza vaccination in early pregnancy is sparse. Assessment of spontaneous abortion is difficult because the outcome is common and not always documented in medical records. Only three studies have assessed the effects of first trimester seasonal influenza vaccination on pregnancy outcomes, and these studies followed-up a combined total of 65 women who received first trimester vaccination.8–10 A fourth study included 650 women who were immunized with seasonal influenza vaccine in the first trimester, but losses occurring before 20 weeks of gestation were excluded from the analysis.11,12 Several studies have assessed pregnancy outcomes after seasonal influenza vaccination in the second or third trimester, and no increased risk of adverse events has been detected.11,13–16 The safety of several pandemic influenza A (H1N1) vaccines administered during pregnancy recently has been evaluated, but these studies also were limited in their ability to examine first trimester administration or spontaneous abortion.17,18 We conducted a case-control study to estimate the relationship between influenza immunization and confirmed spontaneous pregnancy loss occurring at or before 16 weeks of gestational age.
MATERIALS AND METHODS
Participants in this case-control study were enrolled in one of six Vaccine Safety Datalink health care organizations: Group Health Cooperative, Seattle, Washington; HealthPartners, Minneapolis, Minnesota; Kaiser Permanente Colorado, Denver, Colorado; Marshfield Clinic, Marshfield, Wisconsin; Kaiser Permanente Northwest, Portland, Oregon; and Southern California Kaiser Permanente, Pasadena, California.19 The study was approved by the Institutional Review Boards of each organization.
Ambulatory, urgent care, emergency department, and inpatient records were searched for relevant International Classification of Diseases, 9th Revision, Clinical Modification diagnosis codes (634 [spontaneous abortion], 637 [unspecified abortion]) to identify potential cases of spontaneous abortion. Codes assigned from October 25, 2005 to February 4, 2006, and from October 22, 2006 to February 3, 2007, were included to maximize the potential for influenza vaccine exposure in early pregnancy. These dates maximized inclusion of women who conceived in the autumn, when influenza vaccine is commonly administered in the United States (Fig. 1).
All potential cases of spontaneous abortion identified through electronic diagnosis codes were manually reviewed by trained abstractors. Confirmation of spontaneous abortion required medical record documentation of intrauterine pregnancy and evidence of natural or spontaneous fetal demise; cases of therapeutic abortion were excluded. Women were eligible for the study if they were aged 18–44 years at the time of pregnancy loss, had documentation of last menstrual period (LMP) in the medical record, and had continuous enrollment in the health care organization for the 12 months preceding the spontaneous abortion diagnosis. The latter was applied to ensure that previous influenza vaccinations and chronic medical conditions would be captured in the medical record. The requirement for LMP documentation was included to allow individual matching of case group and control group participants by LMP (and thus approximate date of conception). Pregnancy losses through 16 weeks of gestation were included to capture potential events occurring after exposure to influenza vaccine during the first trimester.
Control patients were eligible for inclusion in the study if they were aged 18–44 years at the time of delivery, had continuous enrollment in the health care organization for the 12 months preceding delivery, and had documentation of LMP in the medical record. Potential control group participants were randomly identified using International Classification of Diseases, 9th Revision, Clinical Modification codes specifying delivery assigned to encounters occurring from May 7, 2006 to September 2, 2006, and from May 6, 2007 to September 1, 2007. These periods allowed for conception within a similar timeframe as case group individuals and facilitated matching on LMP. Medical records were then abstracted for all potential control group participants, and only those with confirmation of intrauterine pregnancy and delivery beyond 20 weeks of gestation were eligible for inclusion in the analysis.
Control group participants were initially frequency-matched to case group participants based on LMP strata (2-week intervals) and health care organization. Within each LMP interval, the case group and control group participants were individually matched based on closest LMP. Matching on LMP and organization was performed to ensure that case group and control group participants had similar opportunity for influenza vaccine exposure in early pregnancy.
The exposures of interest were the 2005–2006 and 2006–2007 seasonal trivalent inactivated influenza vaccines, with receipt documented in the medical record. The primary exposure window for case group and control group participants within each matched pair was the 28-day period preceding the date of spontaneous abortion within the matched pair. The 28-day window was chosen based on the known immunologic effects of influenza vaccine that occur within 2–4 weeks after vaccination.20
Case group participants were defined by the natural or spontaneous loss of an embryo or fetus. Date of spontaneous abortion was determined using ultrasound data when available. For those case group participants in whom ultrasound dating was unavailable, date of loss was based on clinical diagnosis. Ultrasound dating provides a more precise determination of date of pregnancy loss than clinical diagnosis, which is dependent on presentation of symptoms of loss, changes in blood test results over time, and health-seeking behavior of the mother. All cases of spontaneous abortion with an ultrasound scan displaying a gestational sac were adjudicated by an obstetrician (M.A.M.) who was blinded to the patient’s vaccination status to determine exact gestational age. For this group, date of spontaneous abortion was calculated as date of LMP plus gestational age (in days) at demise. Date of spontaneous abortion within each matched case-control pair was used as the reference date for analysis. Cases of spontaneous abortion occurring before 5 weeks of gestation were excluded from the final analyses because of increased difficulty in ascertaining accurate date of fetal demise and gestational age.
We performed conditional logistic regression to estimate the association between spontaneous abortion and receipt of influenza vaccine in the 28-day exposure window, adjusting for the following potential confounders: maternal age; parity; maternal diabetes; and previous health care utilization. Age was included in the model as an unrestricted quadratic spline.21 Dichotomous variables were created for parity (none compared with one or more live births) and maternal diabetes (history of type 1 or 2 diabetes compared with no history). Health care utilization was defined as the number of days with an outpatient or inpatient encounter in the year before the LMP, and was included in the model as an unrestricted quadratic spline. Smoking status during pregnancy was examined in a preliminary model and determined not to be a confounder with vaccine exposure in our study population. Additionally, because smoking status was unavailable for at least one member of 31 (13%) matched pairs, we excluded smoking status from the final model to maximize sample size and power. Febrile illness in the first trimester, asthma, and hypertension also were evaluated as potential confounders but were not included in final adjusted models after also being ruled out as confounders.
Our primary analysis included a three-level categorical variable for influenza vaccine exposure relative to the reference date of the matched pair (ie, the date of spontaneous abortion within the matched pair): 1) exposure 1–28 days before the reference date, 2) same-season exposure more than 28 days before the reference date, and 3) unexposed as of the reference date. The unexposed category served as the referent group for both categories 1 and 2.
A secondary analysis was performed examining the association between spontaneous abortion and influenza vaccine receipt relative to pregnancy status (preconception and postconception). This analysis also included a three-level categorical variable for influenza vaccine exposure: 1) exposure after conception and before reference date, 2) same-season exposure before conception, and 3) unexposed as of the reference date. The unexposed category served as the referent group for both categories 1 and 2. For this analysis, date of conception was estimated as LMP plus 14 days, and vaccine receipt before conception included influenza vaccine administered on or before the estimated date of conception. Conditional logistic regression models for this analysis adjusted for maternal age, parity, maternal diabetes, and health care utilization.
When planning the study, we estimated that inclusion of 233 case group and 233 control group individuals in the analysis would provide 80% power to detect an odds ratio (OR) of 2.0 or more, assuming 12.8% vaccine exposure among control group participants overall and α=0.05. It was difficult to estimate vaccination rates within the exposure window; therefore, seasonal estimates of influenza vaccine uptake among pregnant women were used. The proportion of control group participants receiving influenza vaccine was estimated based on data from the Centers for Disease Control and Prevention.22
Univariate P values represent findings from paired t tests and Wilcoxon rank-sum tests for continuous variables or McNemar tests for dichotomous variables. All reported P values were based on two-sided tests for significance, and P<.05 was considered statistically significant; SAS 9.2 was used for analyses. A temporal scan statistic was used to objectively evaluate whether vaccinations were clustered within any segment of the 8-week preconception period among women with spontaneous abortion.23,24
Three hundred eighty-six potential cases of spontaneous abortion were identified electronically; 255 cases were confirmed by medical record review and matched to control group participants. After excluding six pairs with unknown vaccination status, one pair with an invalid LMP, and five pairs with fetal demise at less than 5 weeks of gestation, 243 pairs were included in the final analysis. The number of cases identified at each health care organization ranged from 21 to 62. Two hundred (82%) case group participants had one or more ultrasound examinations; 153 of 200 ultrasound reports provided additional information to confirm the date of fetal demise.
Case group and control group participants were similar in terms of health care utilization, previous spontaneous abortion, history of asthma, hypertension, and parity (Table 1). Women with spontaneous abortion were older than control group participants (mean age at LMP 31.7 years compared with 29.3 years; P<.001) and more likely to have insulin-dependent diabetes mellitus or noninsulin-dependent diabetes mellitus diagnosed (10 case group participants compared with one control group participant; P=.01). The mean pair-wise difference in LMP between case group participants and matched control group participants was −0.41 days (median 0 days, range −12 to 14 days). The mean gestational age at fetal demise was 7.8 weeks (range 5.0–16.6 weeks, median 7.1 weeks) based on ultrasound dating (if available) or date of clinical diagnosis (Fig. 2).
Sixteen percent of case group participants and 13% of control group participants received the same-season influenza vaccine before the pair-specific reference date (Table 2); all vaccinations in our analysis occurred before conception or in the first trimester. In unadjusted matched analyses, case group participants and matched control group participants did not differ in exposure during the 28-day exposure window (unadjusted matched OR 1.10, 95% confidence interval [CI] 0.53–2.29) or in overall exposure to same-season influenza vaccine (vaccination at any time before pair-specific reference date) (unadjusted matched OR 1.29, 95% CI 0.76–2.20).
In the adjusted conditional logistic regression model, the matched OR for vaccine receipt in the primary 28-day exposure window was 1.23 (95% CI 0.53–2.89; P=.63). The OR for exposure more than 28 days before the reference date also was not statistically significantly increased (Table 3). This analysis was repeated using the subset of pairs in which the date of fetal demise was confirmed by ultrasonography (153 of 243 pairs; 63%). The adjusted matched OR for vaccination in the 28-day exposure window for this subset was 1.29 (95% CI 0.41–4.02).
Secondary post hoc analyses were performed to examine the association between spontaneous abortion and the timing of influenza vaccination relative to estimated date of conception. Twenty-two case group participants and 11 control group participants received the influenza vaccine before conception. There were 24 discordant pairs for this analysis, including 17 pairs in which the case group participant was vaccinated before conception and the matched control group participant was not (unadjusted matched OR 2.55, 95% CI 1.06–6.11) (Table 3). In a logistic regression model that adjusted for age, health care utilization, maternal diabetes, and parity, the association between vaccine receipt before conception and spontaneous abortion was not significant (matched OR 2.34, 95% CI 0.86–6.33; P=.10). Among all case group participants, vaccination dates tended to cluster in the week before conception; seven women with spontaneous abortion received the influenza vaccine in this period compared with two control group participants (Fig. 3).
We used a scan statistic to formally evaluate the distribution of vaccination in the preconception period. This analysis included 39 vaccinated women with spontaneous abortion (exposed case group participants). The null hypothesis was that the interval from vaccine receipt to date of conception was randomly distributed, ie, not clustered anywhere within the range of observed intervals. The observed range was vaccination 56 days before conception to 55 days after conception. The scan statistic identified the cluster least likely due to chance as a 3-day window from 2 to 4 days before conception when 6 of the 39 (15.4%) vaccinations occurred (P=.12).
The analysis reported here adds evidence to support the safety of influenza vaccination when administered during the first trimester. We found no statistically significant association between pregnancy loss occurring between 5 and 16 weeks of gestation and influenza vaccination in a 28-day exposure window, predefined based on the period of maximum immune response to the vaccine.
The risk of spontaneous abortion after receiving a seasonal influenza vaccine has not been thoroughly examined in existing literature. Previous studies did not detect an increased risk of spontaneous abortion, but were limited in size or timing of vaccine administration, or excluded spontaneous abortion as an outcome measure.8–12 Pasternak et al25 recently evaluated fetal loss after receipt of pandemic influenza vaccine in a large register-based cohort study. Although this study examined a novel, adjuvanted vaccine, it also found no association between vaccination and spontaneous abortion.
The availability and use of seasonal influenza vaccine vary greatly by season and month, and time is a potential source of confounding in any observational study of influenza vaccine safety in pregnancy. In this analysis, pregnant women were tightly matched based on date of LMP as reported in the medical record. The median pair-wise difference in date of LMP was less than 0.5 days. This limited the potential for differential access to influenza vaccination based on calendar time. Other strengths of the study include the use of geographically diverse populations via six health care organizations across the United States, medical confirmation of pregnancy for all study participants, medical confirmation of spontaneous abortion, and examination and inclusion of potential confounders in the analysis.
We performed a post hoc analysis using an exposure window defined by estimated date of conception (LMP plus 14 days). Although the point estimate was increased, there was no statistically significant association between spontaneous abortion and influenza vaccine receipt before conception when all 243 pairs were included in the analysis. Temporal clustering was greatest in the window 2–4 days before estimated date of conception, but this also was not statistically significant. An increased risk after preconception vaccination could be biologically plausible, because humoral and cell-mediated immune response occurs 7–14 days after vaccination; however, the actual mechanism by which the immune response would result in fetal demise is unknown.26–28 This study was not designed to examine vaccine exposure before conception, and further research could address this question in greater detail.
There are several limitations of this study that should be considered. It is important to note that this study was powered to exclude only a minimum of twofold increase in risk. A smaller increase may not have been identified. We examined only the association between influenza vaccination and spontaneous abortion; therefore, no conclusions can be drawn regarding other adverse pregnancy outcomes. Because we examined exposure within an exposure window, precisely defining the window was critically important. This required accurate determination of the date of spontaneous abortion, which was used to define the 28-day exposure window. Although spontaneous abortion was confirmed by medical record review, determining the exact date of loss can be difficult, because clinical presentation and diagnosis do not always align with timing of fetal demise. Ultrasound reports allowed calculation of gestational age at demise to the day, but these were unavailable for 18% of case group participants and did not provide useful information for another 19% (eg, report of empty uterus on date of clinical diagnosis). However, as noted, results from the primary analysis were similar when restricted to women with ultrasound confirmation.
Misclassification of vaccine exposure is possible if women received influenza vaccines outside the health plans. We could not assess this, but it is likely that most insured pregnant women receive influenza vaccines from health care providers rather than retail outlets or public health clinics. Any vaccines administered outside the health plan would be captured in this study if they were reported to the provider and captured in the medical record. Given the universal recommendation for vaccination, providers either should have administered the vaccine or should have asked about previous same-season receipt, which then would have been documented in the medical record.
Other limitations include the inability to assess differences by vaccine manufacturer, race or ethnicity, and the insufficient power to assess risk in women with chronic medical conditions. As is the case with all observational studies, unmeasured confounding may have occurred; we were unable to examine body mass index and potential exposures such as prescription or recreational drug use or alcohol intake, for example. Finally, this study examined spontaneous abortion identified in the medical setting. Our results may not be representative of loss never brought to medical attention, if those pregnancies differ systematically.
We found no statistically significant increase in the risk of pregnancy loss in the 4 weeks after seasonal inactivated influenza vaccination. However, our sample size was not large enough to rule out any increase in risk. Although the results of this study support the current Advisory Committee on Immunization Practices recommendations for influenza vaccination during all trimesters of pregnancy, further research on the safety of influenza vaccine in pregnant women and other high-risk groups should remain a priority, including further assessment of pregnancy outcomes.
1. Harris JW. Influenza occurring in pregnant women: a statistical study of thirteen hundred and fifty cases. JAMA 1919;72:978–80.
2. Freeman DW, Barno A. Deaths from Asian influenza associated with pregnancy. Am J Obstet Gynecol 1959;78:1172–5.
3. Recommendations for prevention and control of influenza. Ann Intern Med 1986;105:399–404.
4. Prevention and control of influenza: recommendations of the Advisory Committee on Immunization Practices (ACIP). MMWR Recomm Rep 1997;46:1–25.
5. Harper S, Fukuda K, Uyeki T, Cox N, Bridges C. Prevention and control of influenza. MMWR Recomm Rep 2004;53:1–40.
6. Seasonal Flu Programme, Immunisation Branch. Seasonal Flu Plan 2012/13. London (UK): Department of Health; 2012.
7. National Immunisation Advisory Committee. Immunisation guidelines for Ireland. Dublin (Ireland): Royal College of Physicians of Ireland; 2011.
8. Deinard AS, Ogburn P. A/NJ/8/76 influenza vaccination program: effects on maternal health and pregnancy outcome. Am J Obstet Gynecol 1981;140:240–5.
9. Hulfka JF. Effectiveness of polyvalent influenza vaccine in pregnancy. Report of a controlled study during an outbreak of Asian influenza. Obstet Gynecol 1964;23:830–7.
10. Murray DL, Imagawa DT, Okada DM, St Geme JW. Antibody response to monovalent A/New Jersey/8/76 influenza vaccine in pregnant women. J Clin Microbiol 1979;10:184–7.
11. Heinonen OP, Shapiro S, Monson RR, Hartz SC, Rosenberg L, Slone D. Immunization during pregnancy against poliomyelitis and influenza in relation to childhood malignancy. Int J Epidemiol 1973;2:229–35.
12. Heinonen OP, Slone D, Shapiro S. Immunizing agents. In: Kaufman DW, editor. Birth defects and drugs in pregnancy. Boston (MA): Littleton Publishing Sciences Group; 1977.
13. Zaman K, Roy E, Arifeen SE, Rahman M, Raqib R, Wilson E, et al.. Effectiveness of maternal influenza immunization in mothers and infants. N Engl J Med 2008;359:1555–64.
14. France EK, Smith-Ray R, McClure D, Hambidge S, Xu S, Yamasaki K, et al.. Impact of maternal influenza vaccination during pregnancy on the incidence of acute respiratory illness visits among infants. Arch Pediatr Adolesc Med 2006;160:1277–83.
15. Munoz FM, Greisinger AJ, Wehmanen OA, Mouzoon ME, Hoyle JC, Smith FA, et al.. Safety of influenza vaccination during pregnancy. Am J Obstet Gynecol 2005;192:1098–106.
16. Sumaya CV, Gibbs RS. Immunization of pregnant women with influenza A/New Jersey/76 virus vaccine: reactogenicity and immunogenicity in mother and infant. J Infect Dis 1979;140:141–6.
17. Pasternak B, Svanström H, Mølgaard-Nielsen D, Krause TG, Emborg HD, Melbye M, et al.. Risk of adverse fetal outcomes following administration of a pandemic influenza a(h1n1) vaccine during pregnancy. JAMA 2012;308:165–74.
18. Tavares F, Nazareth I, Monegal JS, Kolte I, Verstraeten T, Bauchau V. Pregnancy and safety outcomes in women vaccinated with an AS03-adjuvanted split virion H1N1(2009) pandemic influenza vaccine during pregnancy: a prospective cohort study. Vaccine 2011;29:6358–65.
19. Baggs J, Gee J, Lewis E, Fowler G, Benson P, Lieu T, et al.. The vaccine safety Datalink: a model for monitoring immunization safety. Pediatrics 2011;127(suppl 1):S45–53.
20. Ohfuji S, Fukushima W, Deguchi M, et al.. Immunogenicity of a monovalent 2009 influenza a (H1N1) vaccine among pregnant women: lowered antibody response by prior seasonal vaccination. J Infect Dis 2011;203:1301–8.
21. Greenland S. Dose-response and trend analysis in epidemiology: alternatives to categorical analysis. Epidemiology 1995;6:356–65.
22. Prevention and control of influenza: recommendations of the Advisory Committee on Immunization Practices (ACIP), 2008. MMWR Recomm Rep 2008;57:1–60.
23. Kulldorf M; Information Management Services I. SaTScan v9.1.1: software for the spatial and space-time scan statistics. Available from: http://www.satscan.org/
. Retrieved November 28, 2012.
24. Kulldorff M. A spatial scan statistic. Commun Stat Theory Methods 1997;26:1481–96.
25. Pasternak B, Svanström H, Mølgaard-Nielsen D, Krause TG, Emborg HD, Melbye M, et al.. Vaccination against pandemic A/H1N1 2009 influenza in pregnancy and risk of fetal death: cohort study in Denmark. BMJ 2012;344:e2794.
26. Cox RJ, Brokstad KA, Zuckerman MA, Wood JM, Haaheim LR, Oxford JS. An early humoral immune response in peripheral blood following parenteral inactivated influenza vaccination. Vaccine 1994;12:993–9.
27. Halliley JL, Kyu S, Kobie JJ, Walsh EE, Falsey AR, Randall TD, et al.. Peak frequencies of circulating human influenza-specific antibody secreting cells correlate with serum antibody response after immunization. Vaccine 2010;28:3582–7.
28. Subbramanian RA, Basha S, Shata MT, Brady RC, Bernstein DI. Pandemic and seasonal H1N1 influenza hemagglutinin-specific T cell responses elicited by seasonal influenza vaccination. Vaccine 2010;28:8258–67.