Background
Influenza may cause severe illness and death in pregnant women. During the A(H1N1) influenza pandemic in 2009, pregnant women with influenza had higher risk of hospitalization than non-pregnant individuals with influenza [
1,
2]. Less is known about the effects of maternal influenza infection on the fetus. Maternal infections may increase the risk of pre-eclampsia [
3], a major cause of intrauterine growth restriction and preterm birth (PTB), but there are few studies on influenza and risk of pre-eclampsia [
4‐
7]. High rates of PTB were observed among pregnant women hospitalized with influenza during the 2009 pandemic [
8‐
10], but these studies did not include a comparison group of pregnant women without influenza. Several other studies have included both pregnant women with and without influenza such that appropriate comparisons of the risk of adverse birth outcomes like PTB, low birth weight, or fetal death could be made [
7,
11‐
18]. However, the number of studies on each outcome is limited, and results are inconsistent. Thus, in a recent systematic review, no firm conclusions could be drawn regarding maternal influenza in relation to the main outcomes PTB, small for gestational age (SGA) birth, and fetal death [
19]. Moreover, in most studies, the influenza cases consist of women who sought medical care, thus women with mild infection are unlikely to be included among the cases. Since a substantial proportion of infected individuals develop mild illness or remain asymptomatic [
20], studies on the impact of mild influenza infection on pregnancy complications and outcomes are needed.
The Norwegian Influenza Pregnancy Cohort (NorFlu) Study is a population-based cohort of women who were pregnant during the pandemic in 2009. Using this cohort, we studied maternal influenza in relation to the risk of pre-eclampsia and adverse birth outcomes among the unvaccinated participants.
Discussion
In this study, we investigated potentially adverse effects of maternal influenza on birth outcomes in a cohort of women that were pregnant during the influenza pandemic in 2009. The influenza cases consisted primarily of women with mild illness. The majority of cases were not medically attended, and only a small proportion of the cases were hospitalized. Neither influenza in pregnancy nor detection of maternal antibodies against A(H1N1)pdm09 at the time of delivery were significantly associated with risk of pre-eclampsia, PTB, or SGA birth. However, detection of antibodies was associated with more than 100 g lower 10th percentiles of birth weight, both overall and among women with pregnancy start before the main wave of the pandemic.
The NorFlu Study is a population-based cohort, providing a unique opportunity to study the impact of mild influenza in pregnancy. Extensive information was gathered on the participants. Through linkage with the MBRN, a national registry with data of high quality, we had complete and accurate information on the outcomes we studied. In order to capture both the milder cases of influenza and those that were medically attended, we combined information from different sources. Moreover, blood samples from a large number of women were tested for antibodies against A(H1N1)pdm09, an objective indicator of infection in unvaccinated individuals that is not influenced by recall. Serology may detect both symptomatic and asymptomatic infections, and serological surveys have been used to estimate the cumulative incidence of A(H1N1)pdm09 infection [
28].
The main limitation of our study is the possible misclassification of the exposures. Due to limited capacity during the pandemic, laboratory testing of suspected A(H1N1)pdm09 influenza cases was restricted [
21]. Less than 1/5 of the cases in our study, and only about 1/3 of the medically attended cases, were laboratory confirmed. Approximately half of the influenza cases were based on self-reported illness. However, the participants in our study were pregnant at the time of the pandemic, so we would expect them to recall illness more accurately than the general population. Moreover, since the questionnaire was completed prior to birth, misclassification does not depend on the outcomes we studied. Furthermore, we required that all the influenza cases, both the self-reported and the medically attended, were ill during the main wave of the pandemic, thereby reducing the likelihood that their illness was caused by infectious agents other than A(H1N1)pdm09, since this was by far the dominating respiratory virus accounting for influenza-like illness during the main wave of the pandemic in Norway [
29].
Given the unpredictable nature of both influenza pandemic occurrence and pregnancy, we did not have the opportunity to collect any pre-pandemic samples. Thus, we could not identify the infected individuals according to an increase in HI-titer. Influenza seropositivity is commonly defined as HI-titer ≥40 [
28], which is associated with a 50% reduced risk of influenza infection [
30]. However, a substantial proportion of individuals infected during the pandemic did not become seropositive [
31,
32], and the use of HI-titer of 40 as a threshold has been found to lead to an underestimate of the cumulative incidence of influenza infection during the pandemic [
31]. Moreover, the prevalence of antibodies against A(H1N1)pdm09 among Norwegian women aged 20 to 40 years was very low before the 2009 pandemic [
25]. Therefore, detection of antibodies (HI-titer ≥10) after the pandemic is probably a good proxy for infection in the unvaccinated women in our cohort. However, a small proportion in our cohort may have had cross-reactive antibodies against A(H1N1)pdm09 from previous influenza infections. On the other hand, due to HI-titer waning [
33], some of the infected women may have had undetectable levels of antibodies as the mean time between the pandemic and blood sampling was 6 months. However, this time interval was adjusted for in the analyses.
Few of the women in our study experienced adverse birth outcomes. We observed 41 cases of PTB among the 1258 women included in the analyses, corresponding to a rate of 3.3%. Among all the women who participated in the study, the rate of singleton PTB was 3.1%, which is lower than the average rate of 5.3% observed among live-born singletons in Norway in 2008 [
34]. The lower rate may be explained by the NorFlu participants being more health conscious than the general population. This is supported by the high proportion of participants using folic acid during pregnancy (according to their record in the Medical Birth Registry) compared to all women in Norway giving birth in 2010 (71.5% vs. 27.0%) [
35]. The participants in the NorFlu study may not be representative of the general population, but this does not necessarily result in biased effect estimates. A study on self-selection in MoBa found lower prevalence of several risk factors, pregnancy complications, and adverse birth outcomes among participants than among all women giving birth in Norway [
36]. However, this study found no evidence of bias in exposure-outcome associations. Recruitment to NorFlu was based on the experience from MoBa and followed the same procedures, thus we expect this to be true for the NorFlu study also.
Even though the information on the outcomes was almost complete in this study, and only a handful of the participants were lost to follow-up, we did not have complete information on the exposures. More than 10% of the unvaccinated women did not return the influenza questionnaire and were therefore excluded. Among these women, the rate of singleton PTB was 6.0%, considerably higher than among the women included in the study. Furthermore, women without a blood sample who were excluded from the analyses with HI-titer as the main exposure, had a higher rate of singleton PTB than women included in these analyses (4.6% vs 2.4%). This indicates that the included participants tend to be healthier than the excluded participants. It is possible that influenza infection has a more harmful effect on birth outcomes for less healthy individuals with certain underlying conditions. In that case, the selection will probably have resulted in an underestimation of the effects of influenza during pregnancy.
In previous studies, mild A(H1N1)pdm09 influenza during pregnancy was not associated with mean birth weight [
7,
16] or increased risk of PTB [
14‐
16], SGA birth [
7,
14,
16], or birth weight < 2500 g [
7,
14‐
16]. This is in correspondence with our results. Risks of hospitalization and death due to influenza are highest in the third trimester [
37], but whether the timing of influenza exposure during pregnancy is of importance in relation to birth outcomes is not clear. In a recent study from Canada, no increased risk of PTB was observed for women with medically attended pandemic influenza in their first or second trimester [
18]. Influenza in the third trimester was associated with significantly increased risk of PTB, but only in the subgroup of women belonging to an influenza risk group. Since recruitment to our study started in February 2010, the majority of the participants were in their first trimester during the main wave of the pandemic. Thus, we could not study whether the impact of influenza differs with trimester of exposure.
Influenza and detection of A(H1N1)pdm09 antibodies were not significantly associated with risk of pre-eclampsia in our study. As far as we are aware, only two studies have previously investigated A(H1N1)pdm09 influenza during pregnancy and risk of pre-eclampsia [
6,
7]. In both studies, women with A(H1N1)pdm09 influenza were compared to women with suspected influenza who tested negative for A(H1N1)pdm09. In accordance with our results, no significant differences in the proportion with pre-eclampsia were observed in these studies.
Only a few, mainly older, studies have used serology to identify women infected with influenza during pregnancy, thus also capturing milder and asymptomatic influenza cases [
38‐
41]. When comparing infected and uninfected mothers, none of these studies found significant differences in mean birth weight [
39‐
41] or proportion with low birth weight [
38]. Detection of antibodies was not associated with median birth weight in our study. However, the 10th percentile of birth weight was more than 100 g lower for women with HI-titer ≥10 than women with undetectable levels of antibodies. In the subgroup with birth weight in the lowest decile, HI-titer was not associated with gestational age. Thus, the difference in birth weight does not seem to be a result of lower gestational age among those with HI-titer ≥10. Possibly, influenza infection may have a direct effect on intrauterine growth. This is supported by randomized clinical trials (RCTs) on maternal influenza vaccination [
42,
43]. In RCTs, a higher risk of adverse birth outcomes among unvaccinated women can be attributed to the higher incidence of influenza infection. In line with our results, the RCTs found that women who received influenza vaccine had lower risk of low birth weight compared to women in the control group, although a significant difference in mean birth weight was also observed. In our study, a high proportion of women with HI-titer ≥10 and birth weight in the lowest decile belonged to an influenza risk group (19.2%). This may indicate that influenza infection has a stronger effect on birth weight for mothers with underlying conditions. Whether the offspring of less healthy pregnant women are more susceptible to the harmful effects of influenza infection should be further studied.