Antiviral medication use in a cohort of pregnant women during the 2009-2010 influenza pandemic.

Abstract Preventing influenza-like illness (ILI) during pregnancy with antiviral medication use (AVMU) can mitigate serious health risks to mother and foetus. We report on AVMU in pregnant women in Ontario, Canada, and describe characteristics of AVMU during the 2009–2010 H1N1 pandemic. Rates and risk estimates of AVMU were compared across multiple categories and stratified across ILI infection status. Increased AVMU was observed in women with influenza infections, active smokers, those vaccinated against influenza, and those with pre-existing co-morbidities. Decreased AVMU was observed in women with multiple gestations, and those in neighbourhoods of high immigrant concentrations. Our stratified analysis indicated that the observed patterns differed by ILI infection status. We demonstrated that once infected, women across multiple groups were equally likely to use antiviral medications. In this report we also propose possible clinical explanations for the observed differences in AVMU, which will be useful in planning prevention initiatives for future pandemics.


Introduction
Infl uenza and infl uenza-like illnesses (ILI) present serious threats to pregnant women and their unborn children, including increased risk of complication during pregnancy and even maternal and or foetal death (Carlson et al. 2009). Th e 2009 -2010 infl uenza pandemic confi rmed this and is consistent with the fi ndings of several other infl uenza epidemics (Dom í nguez- Cherit et al. 2010), which have demonstrated disproportionately high rates of complications among infl uenza-infected pregnant women. It is well known that women with co-morbidities are more severely aff ected by ILIs, and thus should be monitored closely by their health care providers (Allen et al. 2006). Currently, vaccination is the most eff ective prevention for ILIs; however, a recent systematic review showed that its effi cacy may not be as high as previously suggested (Osterholm et al. 2012). Alternatives such as antiviral medications are oft en used to treat infl uenza symptoms, and have been shown to be eff ective up to 48 h post infection (Ward et al. 2005).
In non-pandemic periods, physicians are advised to treat ILIinfected pregnant women with antiviral therapy. However, during the 2009 -2010 pandemic the Society of Obstetricians and Gynecologists of Canada updated this recommendation to include providing prophylaxis for those women at risk of an ILI infection (Bozzo et al. 2009). It is easy to see how such recommendations can trigger changes in the distribution of antiviral medication use (AVMU) among patient populations. Yet still little is known about the general distribution of AVMU within this group of women. Studies are few and not granular enough with respect to rare subgroups (Rasmussen et al. 2012;Atkins et al. 2011). We present population-based information on patterns of AVMU seen in pregnant women of Ontario, Canada during the 2009 -2010 infl uenza pandemic.

Materials and methods
Our cohort consists of pregnant women who gave birth in an Ontario hospital to a live or stillborn infant ( Ͼ 20 weeks ' gestation and birth weight: Ͼ 500 g) between November 2nd 2009 and October 31st 2010. Information on these women was collected by the Better Outcomes Registry & Network (BORN) Ontario, a province-wide maternal-child registry, and is described in detail elsewhere (Liu et al. 2012). Th e primary outcome (i.e. self-reported AVMU) was defi ned as the use of Oseltamivir (Tamifl u ® ) or Zanamivir (Relenza ® ) at any time during pregnancy. Records were geocoded and linked (via maternal place of residence postal code) to Statistics Canada ' s Census-level (dissemination area) information on education, concentration of recent immigrant ( Յ 5 years) and median family income quartiles.
Descriptive statistics were used to identify diff erential AVMU across maternal age at time of delivery, parity, smoking status, ILI status anytime during pregnancy, infl uenza vaccination status anytime during pregnancy, multifoetal pregnancies, those with a fi rst-trimester visit, and the presence of any co-morbidities (including asthma, chronic hypertension, insulin-and non-insulin-dependent diabetes, and heart disease). Census-level variables, the type of antenatal care provider and rural area of residence were also included in the analysis. Multivariable-adjusted risk estimates of AVMU were produced using log binomial regression models with a generalised estimating equation to account for clustering at the dissemination area. We then stratifi ed the analysis by ILI status. Co-linearity was checked and we restricted our analysis to only variables with Ͻ 10% missing information. A Preventing infl uenza-like illness (ILI) during pregnancy with antiviral medication use (AVMU) can mitigate serious health risks to mother and foetus. We report on AVMU in pregnant women in Ontario, Canada, and describe characteristics of AVMU during the 2009 -2010 H1N1 pandemic. Rates and risk estimates of AVMU were compared across multiple categories and stratifi ed across ILI infection status. Increased AVMU was observed in women with infl uenza infections, active smokers, those vaccinated against infl uenza, and those with pre-existing co-morbidities. Decreased AVMU was observed in women with multiple gestations, and those in neighbourhoods of high immigrant concentrations. Our stratifi ed analysis indicated that the observed patterns diff ered by ILI infection status. We demonstrated that once infected, women across multiple groups were equally likely to use antiviral medications. In this report we also propose possible clinical explanations for the observed diff erences in AVMU, which will be useful in planning prevention initiatives for future pandemics.
Keywords: Antiviral , antiviral medication , H1N1 , Infl uenza , pregnant women , pregnancy sensitivity analysis was conducted on records with missing outcome information to assess the presence of selection bias. Data analysis was performed using SAS (SAS Institute, Cary, NC) and statistical signifi cance was evaluated with a two-sided p-value of 0.05. Th is study was approved by the Research Ethics Boards of the Ottawa Hospital Research Institute, and the Children ' s Hospital of Eastern Ontario.

Antiviral medication use among pregnant women 553
A total of 2,860 (2.5%) women reported an ILI during pregnancy, of which AVMU was observed in 926 (32.4%) records. Within this sub-cohort only the presence of any maternal health problem showed a statistically signifi cant increase in AVMU (aRR: 1.21; 95% CI: 1.02 -1.46). Th e non-ILI sub-cohort showed similar results to those seen in the full cohort, with the notable exception of a marked increase in AVMU within women who were vaccinated for infl uenza (aRR: 2.81; 95% CI: 2.40 -3.29). Co-linearity observed between variables was negligible (vif: Ͻ 1.7). Potential selection bias in records with missing outcome information could not be ruled out for maternal smoking, the presence of any maternal health problems, the use of an obstetrician and maternal place of residence in a rural area (see Supplementary Tables I -II available

Discussion
Th is study provides the largest and most complete description of AVMU in a population of pregnant women residing in Ontario during the 2009 -2010 H1N1 infl uenza pandemic. With such a large number of participants we were able to provide accurate estimates of descriptors of characteristics associated with the use of antiviral medication for the treatment of an ILI during pregnancy, covering multiple rare demographics.
We observed that all predictors with a signifi cant rise in AVMU (excluding vaccination) are also known to increase susceptibility to ILI (Viboud et al. 2004). However, aft er stratifying by ILI status, only the presence of any maternal co-morbidity remained statistically signifi cant within the ILI stratum, while in the non-ILI stratum these relationships persist to roughly the same magnitude. With respect to the increased AVMU seen in those women vaccinated for infl uenza, it is likely that pregnant women and their health care providers were taking extra precautions during the pandemic. Women received vaccinations and supplemented this with AVMU, thus mitigating the likelihood of having an ILI. Women vaccinated for infl uenza showed an approximate threefold increase in AVMU among the non-ILI stratum, whereas among the ILI population there was no observable diff erence. Th is may be the product of a " healthy vaccinee " eff ect (Virtanen et al. 2000), in that women who are vaccinated are also more likely to seek interventions such as AVM.
While there is no " a priori " reason to assume that AVMU in women with multifoetal pregnancies would be less than their singleton counterparts, it is possible, given that these women are monitored closely at high-risk centres, that they were more concerned about the perceived threat of AVMU during the pandemic period (Mandeville et al. 2013, Van et al. 2010). AVMU in women residing in areas of high recent immigrant concentrations was lower than that in other areas, and could be due to documented access to health care issues experienced within this population (Pottie et al. 2011). Th ese include lack of contact with appropriate health care provider, transportation and other fi nancial issues. Moreover, they may also be generally unfamiliar with the Canadian Health Care system, or are unable to communicate eff ectively in either of Canada ' s offi cial languages. Alternatively, the observed diff erence could also be explained if women residing within this group were less likely to become infected in the fi rst place, due to observed " healthy immigrant " eff ects (Pottie et al. 2011).
Th e data presented in this report were derived from a large population-based birth registry, and is expected to be precise and accurate. However, there are several limitations that cannot be overlooked. Th e main limitation of this study is a lack of timing information on AVMU with respect to onset of illness. Th is prevented any conclusions on whether or not AVMU was used as a preventative treatment or post-exposure prophylaxis. We cannot completely rule out the presence of reporting and recall bias within the data collection process, since the outcome variable was self-reported. Only hospital-based deliveries were included in our study (excluding an estimated 2% of Ontario deliveries occurring within the home). Th is small group of women could potentially show diff ering AVMU patterns; however, our results are not likely to change signifi cantly with their addition due to the relatively small contribution of this group to overall birth numbers.
In summary, it appears that in the absence of an ILI, diff erential AVMU is seen within the demographic groups identifi ed above. However, once a woman is infected with an ILI she is treated with antiviral medication regardless of her demographic profi le and the observed relationships in the ILI population dissipates. In this study we present empirical evidence identifying groups of women where increased and decreased rates of AVMU were observed. Future studies are needed to investigate qualitative aspects of these associations, particularly those related to decreased AVMU in multifoetal and recent immigrant populations.