Antibody levels in a cohort of pregnant women after the 2009 influenza A(H1N1) pandemic: Waning and association with self‐reported severity and duration of illness

Background A population‐based pregnancy cohort was established in Norway to study potential effects of exposure to the 2009 influenza pandemic or pandemic vaccination during pregnancy. Objectives We studied maternal A(H1N1)pdm09‐specific hemagglutination inhibition (HI)‐titer levels and waning in women with influenza‐like illness (ILI) in pregnancy compared to vaccinated women. Moreover, we studied the association between HI‐titers and self‐reported severity and duration of ILI. Methods HI‐titers against the pandemic virus were measured in maternal blood samples obtained at birth, 3‐9 months after exposure, and linked with information about pregnancy, influenza and vaccination from national registries and a cohort questionnaire. Results Among 1821 pregnant women included, 43.7% were unvaccinated and 19.3% of these had ILI. HI‐titers were low (geometric mean titer (GMT) 11.3) in the unvaccinated women with ILI. Higher HI‐titers (GMT 37.8) were measured in the vaccinated women. Estimated HI‐titer waning was similar for vaccinated women and women with ILI. Most ILI episodes were moderate and lasted 3‐5 days. Women with ILI reporting specific influenza symptoms such as fever or cough had higher HI‐titers than women without these symptoms. Women who reported being “very ill” or illness duration of >5 days had higher HI‐titers than women reporting less severe illness or illness of shorter duration, respectively. Conclusions Antibody waning was similar in vaccinated women and women with ILI. More severe ILI or longer duration of illness was associated with higher HI‐titers. Most unvaccinated pregnant women with ILI had low HI‐titers, probably due to moderate illness and HI‐titer waning between exposure and sampling.


| INTRODUC TI ON
During the 2009 influenza A(H1N1) pandemic, pregnant women were at increased risk of hospitalization and death due to severe influenza infection. [1][2][3] Following the first reports on this association, a population-based cohort of pregnant women (The Norwegian Influenza Cohort Study, NorFlu) was established in Norway, to study the potential effects of maternal pandemic influenza and vaccination on the women and their children.
In Norway, the main pandemic period occurred between October 1, 2009, to December 31, 2009, peaking in early November. 4,5 A vaccination campaign with the AS03-adjuvanted A(H1N1)pdm09 vaccine (Pandemrix) started 19th October. The vaccine was recommended to pregnant women in their second or third trimester and to groups at high risk of severe influenza. 6 It was mandatory to report pandemic vaccination to the national immunisation register (SYSVAK). 4,5 Approximately 54% of pregnant Norwegian women were vaccinated. 4 In contrast, among 46 000 pregnancies in Norway during the pandemic, only 516 cases of laboratory-confirmed influenza infection were registered. Due to limited laboratory capacity during the pandemic, testing of patients with severe illness was prioritized.
Furthermore, only 8.9% of women who gave birth in 2009 or 2010 were diagnosed with influenza by a physician. 4 However, national influenza surveillance data indicated a clinical attack rate of approximately 30% in the Norwegian population. 5 Data on laboratory-confirmed influenza were limited also for women in the NorFlu pregnancy cohort, but antibodies against the A(H1N1)pdm09 virus, measured as hemagglutination inhibition (HI)titers, were measured in maternal blood samples taken at delivery, 3-9 months after the pandemic period. As the 2009 A(H1N1)pdm09 virus was an antigenically novel virus 7 and the frequency of preexisting antibodies to the virus was low in the Norwegian population prior to the pandemic, 8 high HI-titers might serve as a proxy for infection in unvaccinated individuals.
Antibodies induced by influenza infection or vaccination are known to wane over time, [9][10][11][12][13] and HI-titers have been suggested to decline faster after vaccination with pandemic vaccines than after infection. [14][15][16] Studying antibody waning is important for understanding the longevity of the maternal antibodies. A small study from the 2009 pandemic found no significant difference in HI-titer waning between vaccinated and infected pregnant women, although the HItiters declined at a slightly slower rate in the infected women. 17 As the immune system is altered to tolerate the fetus, pregnancy may influence the immune response to pandemic influenza infection and vaccination. 18 However, studies on influenza vaccination of pregnant women mostly indicate that their immune responses are comparable to those of non-pregnant healthy individuals. 19 Due to the long time interval between the pandemic exposure and blood sampling at delivery, we aimed to estimate HI-titer waning in the unvaccinated pregnant women with ILI during the 2009 pandemic, and compare with the estimated waning in the vaccinated women. In addition, as HI-titers have been reported to be positively associated with influenza symptoms and severity of illness, 20,21 we also studied whether HI-titers were associated with self-reported ILI symptoms, severity or duration of ILI, in this cohort of pregnant women.

| Hemagglutination inhibition (HI) assay
Maternal serum samples were tested in HI assays with the pandemic vaccine virus NYMC X-179A, in twofold dilutions starting at 1:10. 24 The HI-titer was defined as the reciprocal of the highest serum dilution that completely inhibited turkey red blood cell agglutination. The HI-titers were normalized against the human international standard 09/194 (NIBSC, UK). 25 Samples with an HI-titer below the level of detection were assigned a titer of 5 for calculation purposes.

| Pandemic vaccination status
Women registered with one dose of the pandemic vaccine in SYSVAK (date of vaccination registered), or self-reported pandemic vaccination in the questionnaire (month and year of vaccination), were defined as vaccinated.

| Influenza and ILI categories
Women were defined as having "medically attended influenza" if they were registered with either an influenza diagnosis (code R80 in the International Classification of Primary Care-2 26 ) in the KUHR database and/or with laboratory-confirmed A(H1N1)pdm09 infection in MSIS during the main pandemic period. Women who solely selfreported ILI during the pandemic period were defined as having "self-reported ILI." For self-reported ILI, the women reported month and year of disease. Women with either medically attended influenza or self-reported ILI were defined as having "ILI."

| Study population
Among the 3201 women in the cohort, participants were excluded if any of the following data were missing: biological samples, HItiter measurements, questionnaire data or MBRN records ( Figure 1).
Women were also excluded if their pregnancy started after the pandemic period, if their start-date was unknown, or if they were not pregnant when they were vaccinated or had ILI. Women who selfreported vaccination had to be pregnant by October 19, 2009, when the vaccination campaign started. For women with self-reported ILI, month of illness could be no earlier than month of pregnancy start. Excluded women were similar to those who were included in terms of age, parity, and proportion registered in SYSVAK and with a primary care influenza diagnosis. However, for excluded women, the HI geometric mean titer (GMT) was slightly higher (19.9 vs 18.4), the mean interval between the peak pandemic and birth was longer (240 days vs 225 days) and fewer were registered in MSIS (0.6% vs 1.5%).

| Symptoms, severity, and duration of ILI
Self-reported symptoms were extracted from a list of common influenza symptoms in the questionnaire. "Fever > 39°C," "fever < 39°C," or "unmeasured fever" were combined to "fever" in the analysis. The US Centers for Disease Control and Prevention (CDC) case definition of ILI is fever >37.8°C, cough and/or a sore throat in the absence of another known cause. 27 When classifying the women according to the CDC definition, the combined definition of "fever" described above was used instead of the criteria fever >37.8°C. Severity was based on the question "How ill did you feel?" with categories "not very ill," "quite ill," and "very ill." Duration of ILI was based on the question "How long were you ill?" with categories "0-2," "3-5," or "more than 5 days."

| Statistical analysis
Characteristics of unvaccinated and vaccinated women were compared using a t test for continuous variables and a chi-square test for categorical variables.
Linear regression analysis was used to estimate waning of HItiters among unvaccinated (n = 154) women with ILI and vaccinated women without ILI (n = 909). 17 Women who had been both vaccinated and ill were excluded from the comparison. The log2transformed HI-titers were regressed on time since exposure. The  difference between the estimated rates of decline among women with ILI and vaccinated women was evaluated with an interaction term between time and type of exposure. Half-life estimates were calculated from the rates of decline. 28,29 Preliminary analyses of the study population found no significant differences in HI-titers between vaccinated women with or without ILI, possibly due to the high HI-titers induced by the vaccine. 30,31 Analyses of HI-titers in the different ILI categories were therefore restricted to unvaccinated women (n = 796). The association between HI-titers and symptoms, severity and duration were studied in unvaccinated women with ILI (n = 154). For these analyses, we used a Wilcoxon signed rank-sum test. Differences in proportions of women with HI-titers <10 or ≥20 were compared using a chi-square test. To ensure that any observed HI-titer differences were not due to differences in time since exposure, subanalyses were performed for women with or without ILI giving birth ≥210-≤280 days after the pandemic (>60% of the women) and for medically attended vs self-report women giving birth ≥180-≤270 days after the ILI episode (>70% of these women). A P-value < 0.05 was considered statistically significant. Calculations were made using Stata/SE 14.0.

| RE SULTS
A total of 1821 pregnant women were included in the study.   Since medically attended influenza possibly represents cases more likely due to influenza virus and/or more severe illness than self-reported ILI, a subanalysis was performed in unvaccinated women with medically attended influenza (n = 67). No significant HItiter waning was found (P = 0.49), and waning was not significantly different from the estimated waning in vaccinated women (P = 0.98).

| Estimation of HI-titer waning
The estimated HI-titer half-life for the medically attended women was 250 days, slightly longer than for all women with ILI.

| HI-titers in the unvaccinated women
We observed low HI-titers (GMT 11.3) among the unvaccinated women with ILI; however, they had significantly higher HI-titers than women with no ILI (GMT 6.5, P < 0.0001; Figure 3A). Several of the unvaccinated women with laboratory-confirmed influenza also had low titers (11.1% with HI < 10 and 33.3% with HI < 20). 80.1% of the women with no ILI had HI < 10, compared to 50.0% of the women with ILI (P < 0.001). Conversely, 35% of women with ILI had HI-titers ≥20 compared to 9.6% of women without ILI (P < 0.001). Women with medically attended influenza had significantly higher HI-titers (P = 0.001, GMT 14.8) than women with self-reported ILI (GMT 9.2; Figure 3B). Consistent with this, 47.8% of women with medically F I G U R E 2 Estimated waning of HI-titers in pregnant women after influenza-like illness (ILI) or pandemic vaccination. The graph shows the log2-transformed HI-titers regressed on time since exposure with 95% confidence intervals (dashed curves). Time was defined as the interval in days from exposure (ILI or pandemic vaccination) to delivery attended influenza had HI-titers ≥20 compared to 23.0% of women with self-reported ILI (P = 0.001). When analyzing subsets of women with similar time since exposure, the differences in HI-titers were still significant between the previously mentioned groups (data not shown).

| HI-titers and self-reported symptoms, severity and duration of ILI in unvaccinated women
The frequency of self-reported symptoms in unvaccinated women with ILI is listed in Table 2. Women who reported fever, cough, sore throat, shortness of breath or chest pain had significantly higher HI-titers than women without these specific symptoms ( Table 2). The mean number of self-reported symptoms was 5.8. Women who reported >6 symptoms had significantly higher HI-titers (P = 0.02, GMT 13.4) than women with ≤6 symptoms (GMT 9.6; Figure 4A). ILI cases who matched the CDC definition of ILI 27 ("CDC-ILI"; 74.0%), had significantly higher HI-titers (P < 0.0001, GMT 13.8) than the women who did not (GMT 6.5; Figure 4B). Similarly, 43.0% of the CDC-ILI cases had HI-titers ≥20 compared to 7.5% of the women not matching the definition, while 41.2% and 75.0%, respectively, had HI < 10 (P < 0.001).

F I G U R E 3 HI-titers according to influenza-like illness (ILI) status in unvaccinated, pregnant women after the 2009 pandemic (A) HI-titers in women with no ILI or with ILI (B) HI-titers in women with medically-attended influenza and women who self-reported ILI. The graphs
show the geometric mean HI-titers with 95% confidence intervals. *P ≤ 0.001 (Wilcoxon signed rank-sum test) Women reporting the symptom, n (%) TA B L E 2 Frequency of self-reported symptoms in unvaccinated women with influenza-like illness (ILI; n = 153) and HI-titers according to self-report of individual symptoms

If symptom not reported
Most of the unvaccinated women with ILI felt "quite ill" (67.4%). Women who were "very ill" had significantly higher HItiters (GMT 24.8) compared to women who reported less severe ILI (either "not very ill" GMT 8.4, P = 0.002 or "quite ill" GMT 12.2, P = 0.03; Figure 5A). The proportion of women with an HItiter ≥20 increased from 18.8% to 69.2% with increasing severity (P-values < 0.05 for "very ill" vs either "not very ill" or "quite ill"). Most women (64.9%) reported an ILI duration of 3-5 days.
HI-titers were significantly higher in women who reported ILI for more than 5 days (GMT 22.3) than in women with a shorter duration of illness (GMT 8.6, P = 0.04 vs 0-2 days and GMT 9.5, P = 0.001 vs 3-5 days; Figure 5B). Again, there was a significantly higher proportion (P < 0.001) of women with HI-titers ≥20 (62.5%) among women with more than 5 days of ILI than among women with shorter durations. A subanalysis of only the laboratoryconfirmed influenza cases (n = 27) was also performed. HI-titers in these women showed similar associations with severity and duration as in all women with ILI, with the exception of women who reported to be "very ill" (n = 3 and data not shown). However, all subgroups were very small (range n = 3-15).

| D ISCUSS I ON
In this unique cohort of women who were pregnant during the 2009 influenza A(H1N1) pandemic, HI-titers against the pandemic virus were measured in maternal blood samples collected at delivery, 3-9 months after exposure to pandemic infection or vaccination. We found that HI-titers were higher after vaccination with the AS03adjuvanted vaccine than after ILI. HI-titers waned over time, both for vaccinated women and for unvaccinated women with ILI. The estimated HI-titer waning was comparable between these groups.
HI-titers were higher among women reporting either more severe ILI or longer duration of ILI, compared to women with less severe ILI or ILI of shorter duration, respectively.
Immune responses after pandemic vaccination or natural infection with influenza A(H1N1)pdm09 virus may be qualitatively and quantitatively different in non-pregnant individuals. 32 However, few studies have compared HI-titers after pandemic vaccination and infection in pregnant women. In accordance with a Danish study, we found that the AS03-adjuvanted pandemic vaccine had induced significantly higher HI-titers than ILI in pregnant women. 30 This F I G U R E 4 HI-titers according to self-reported symptoms in unvaccinated, pregnant women with influenza-like illness (ILI) during the 2009 pandemic (A) HItiters in women reporting 0-6 symptoms and more than six symptoms (B) HI-titers in women with or without symptoms matching the CDC case definition of ILI. The graphs show the geometric mean HI-titers with 95% confidence intervals. *P < 0.05 (Wilcoxon signed rank-sum test) F I G U R E 5 HI-titers according to self-reported severity and duration of influenza-like illness (ILI) in unvaccinated, pregnant women with influenza-like illness (ILI) during the 2009 pandemic (A) HI-titers in women who were "not very ill," "quite ill," or "very ill" (B) HI-titers in women who were ill for "0-2," "3-5," or "more than 5" days. The graphs show the geometric mean HI-titers with 95% confidence intervals. *P < 0.05 (Wilcoxon signed rank-sum test) finding is in contrast to a US study, where vaccination and infection induced similar HI-titers in pregnant women; however, there a non-adjuvanted vaccine was used, all influenza cases were medically attended, and time since exposure was shorter. 17 Although we found that unvaccinated women with ILI had higher HI-titers than those without ILI, the GMT HI-titer was low, and almost half of the women had HI < 10, including some with laboratoryconfirmed influenza. Low HI-titers after PCR-confirmed pandemic influenza have also been described previously. 33 4,5 Since we believe that the real number of pregnant women with pandemic influenza was higher than reflected in the surveillance registry, 4 we also defined women with a diagnosis of influenza or who self-reported ILI, as "ILI cases" in our study. By restricting the analyses to ILI occurring during the main pandemic peak, the likelihood that ILI was indeed caused by the pandemic virus was increased. 38 According to national surveillance data, the pandemic virus was by far the dominating respiratory virus at this time. 38 As can be expected, medically attended influenza was associated with higher HI-titers as compared to self-reported ILI. This could be due to less misclassification of influenza in this group, but may also be the result of more severe illness in the medically attended women, since high HI-titers were positively associated with severity in these women. Moreover, higher HI-titers were also observed in women who self-reported ILI compared to the women with no ILI, implying that they were infected with influenza. Furthermore, ac- A correlation between HI-titers and disease severity has previously been observed in non-pregnant patients with A(H1N1)pdm09 infection 20 and seasonal influenza. 21 In the latter study, this association was not found for the 2009 pandemic; however, the authors noted that this could be due to too few pandemic cases. In accordance with the positive association between HI-titers and severity, we have also found an association between T-cell and NK-cell responses and ILI symptoms in this pregnancy cohort. 43 Since data on viral load were lacking, we could not explore if high HI-titers in the severely ill women were the result of stimulation from a high viral load, and previous studies have been contradictory. [44][45][46][47][48][49] The strength of this study was that it was conducted within a large, prospective, population-based pregnancy cohort with both vaccinated and unvaccinated women where an extensive combination of data was available: blood samples collected at delivery, data from several national registries and cohort questionnaire data. The questionnaire data provided valuable information about symptoms and illness. Pregnant women generally pay special attention to their health and as most of the women filled out the questionnaire prior to birth, the reports were not biased by the outcome of the birth.
A limitation of our study was that recruitment of the study participants was not feasible until after the pandemic peak, hence pre-pandemic or pre-pregnancy samples could not be obtained and each woman contributed with a single blood sample at delivery. It was therefore not possible to study seroconversion or actual To conclude, in this population-based cohort of women who were pregnant during the 2009 influenza A(H1N1) pandemic, we found that vaccination with the AS03-adjuvanted pandemic vaccine induced much higher HI-titers than ILI. 50% of the vaccinated women had HI-titers ≥20 even 8 months after vaccination.
Despite the observed HI-titer differences between vaccinated women and women with ILI, the estimated HI-titer waning was similar. Knowledge on the longevity of maternal antibodies is of particular relevance for future influenza pandemics or situations where adjuvanted monovalent vaccines are provided to pregnant women, and may be transferable to seasonal influenza. Moreover, our results indicate that more severe illness may induce higher HItiters. Since HI-titers are influenced by both waning and severity, the generally low HI-titers measured among the unvaccinated women with ILI may be explained by moderate illness and the long interval in time between exposure and blood sampling.

ACK N OWLED G EM ENTS
We thank the participants in the Norwegian Influenza Cohort Study (NorFlu). We also wish to thank the staff at Oslo University Hospital,