|Home | About | Journals | Submit | Contact Us | Français|
To identify associations between fertility treatment use (assisted reproductive technologies, ovulation induction, and artificial insemination) and subsequent infant feeding practices.
The Upstate KIDS population-based cohort enrolled mothers who delivered live births, sampling on fertility treatment and plurality.
New York, USA between 2008-2010 (excluding New York City).
Data regarding singletons and one randomly selected infant among twins were used.
Mothers reported breastfeeding and formula feeding practices at 4, 8, and 12 months postpartum. Modified Poisson regression was used to compare risks for feeding practices by mode of conception. Marginal structural models were used to estimate the controlled direct effects of fertility treatment on feeding, independent of preterm birth.
Among 4,591 mothers, 1,361 (30%) conceived using fertility treatments. Mothers who used fertility treatments were less likely to breastfeed until twelve months postpartum (aRR 0.58, 95%CI 0.40, 0.86) and more likely to provide formula (aRR 1.22, 95%CI 1.11, 1.33), solids (aRR 1.17, 95%CI 1.05 1.31), and juice (aRR 1.44, 95%CI 1.00, 2.06) by four months than mothers who did not conceive with treatments. Fertility treatment remained associated with breastfeeding cessation and formula feeding in mediation analyses, suggesting preterm birth does not fully explain these associations.
Women who conceived with fertility treatments were less likely to breastfeed later in infancy and more likely to provide formula, solids, and juice earlier in infancy. As our analyses accounted for confounding and preterm birth, other contributing factors may include difficulties feeding twins or workplace breastfeeding accommodations.
There is increasing interest in the growth and development of children conceived by fertility treatments.(1) Infant feeding may contribute to both growth and development, yet few studies describe differences in feeding practices by mode of conception. Differences may stem from the increased risks of preterm birth among those conceived with assisted reproductive technologies (ART)(2)—as this outcome may influence growth or nutritional needs in infancy and perhaps, subsequent physician advice.(3) Alternatively, difficulties breastfeeding, socioeconomic factors, or anxieties that may be unique to or more prevalent among parents using fertility treatments, may also be associated with infant feeding.(4,5,6)
Current feeding guidelines in the United States (U.S.) recommend exclusive breastfeeding in the first six months of life and then continued breastfeeding until 12 months while solid foods and small amounts of juice are added into the infant's diet.(7,8,9) Despite these recommendations, results from the Infant Feeding Practices Study II suggest that almost 40% of mothers in the U.S. introduce solid foods before four months; the prevalence was lowest among mothers exclusively breastfeeding (24%) compared to those using mixed feeding (50%).(10) Many studies explore the influence of feeding on infant growth, but the inclusion of infants conceived with fertility treatments is seldom addressed and among studies that examine the growth of these infants, few comment on feeding.(11,12) Further study of infant feeding practices by mode of conception is warranted because feeding is an important mechanism through which infant health and growth is influenced and there is an emerging body of literature reporting that breastfeeding differences, in particular, do exist.(4, 11, 13-15) However, limited inference can be made as to why these differences exist, owing to small sample sizes, incomplete control for confounding by socioeconomic factors, and inappropriate adjustment for potential causal intermediates like preterm birth.
Therefore, our objective was to compare feeding practices during infancy by mode of conception using data from a U.S. population-based cohort: Upstate KIDS (2008-2010). We hypothesized that differences in feeding do exist, but may largely be explained by confounding or the higher prevalence of poorer birth outcomes among those conceiving with treatments. By identifying differences in feeding and potential contributing factors, we can better understand if tailoring breastfeeding or nutritional counseling for parents who conceived with fertility treatments would be beneficial.
Upstate KIDS Study is a population-based birth cohort established to study fertility treatment and child development.(16) The cohort included live births from New York State (NY; New York City was excluded) between 2008 and 2010 and sampled on fertility treatment exposure. All mothers who conceived with treatments and all mothers of multiples were invited to participate. Singleton live births conceived without treatments were frequency matched on perinatal care region of delivery to a random sample of births conceived with treatments (3:1). Study data were compiled from vital records, hospital discharge data, and written questionnaires completed postpartum. Although all mothers of multiples were invited to enroll, these analyses included information from mothers of singletons and one randomly selected infant from twin sets, as feeding was presumed to be similar within a set of twins (n=4,971). We next excluded mothers who did not complete the baseline questionnaire (four months postpartum) on which feeding practices were first queried (n=380; leaving a final N=4,591). All participants provided informed consent and study procedures were approved by the New York State Department of Health and University of Albany Institutional Review Boards (#07-097 and #08-179, respectively).
Our primary exposure was any fertility treatment used in the index pregnancy, compared to no treatments. Treatment information came from questionnaires completed four months postpartum; birth certificate data were used when this was missing. Assisted reproductive technologies (ART) and ovulation induction/medications (OI) with or without additional procedures (e.g., intrauterine insemination, IUI) were included in the fertility treatment group. Secondary analyses separating ART and OI/IUI were completed. Self-reported ART exposure was concordant with the Society for Assisted Reproductive Technology - Clinical Outcome Reporting System (SART).(17)
Our dichotomous outcomes of interest included the following feeding practices during infancy (assessed via paper questionnaires at 4, 8, and 12 months postpartum): any breastfeeding at a given time point, the provision of any formula at a time point, and providing solid foods or juices at time point. Breastfeeding questions did not distinguish between physical breastfeeding or providing expressed breast milk. Solid food categories included cereal in a bottle; other cereals; fruit or vegetables; finger foods; pureed table food (four months postpartum only); or meat, eggs, cheese, and dairy. Mothers were queried about introducing juice into their baby's diet on each questionnaire, but not the type or amount.
Maternal, paternal, and infant characteristics, as well as feeding behaviors, were compared by mode of conception using chi-square or Fisher exact tests as appropriate. For regression analyses, we used a modified Poisson regression with robust error variances to estimate risk ratios (RR) and 95% confidence intervals (CI), as described by G. Zou.18 Models were run for each time point within each outcome of interest. We also ran several independent regression models in sensitivity analyses: the first models limited our exposed groups to those who used only ART treatments (n=649) or those who used only OI/IUI (n=712); the next models added daycare initiation to assess residual confounding by maternal return to the workplace and feeding practices of daycare providers (all types of providers); we also ran models limiting the population to mothers submitting all questionnaires no later than 2 months after their intended completion dates (n=843 removed) in order to assess the ideal scenario that feeding information was obtained near the times specified on questionnaires; and lastly, we adjusted for whether or not another pregnancy occurred within the year after enrollment in a model for breastfeeding at twelve months postpartum.
To address Upstate KIDS unique sampling strategy as well as loss to follow-up, we used sampling and missingness weights in our models. We created a stabilized inverse probability weight for missingness for each outcome at each time point. We then multiplied the missingness weight with a sampling weight designed to correct for the study's sampling on region, fertility treatment, and plurality in order to make our data more representative of New York State during 2008-2010.
By using directed acyclic graphs (DAGs) and our knowledge of the literature, we selected the following potential confounders a priori: maternal age, race and ethnicity, post-pregnancy body mass index (BMI), education, morbidities (any diabetes; hypertension; cardiovascular disease; celiac disease; irritable bowel syndrome; Crohn's disease; eating disorders; hepatitis B or C; polycystic ovarian syndrome, PCOS; and/or human immunodeficiency virus, HIV), number of prior live births, history of seeking medical help for fertility/getting pregnant, marital status, private insurance status, daycare initiation by 8 months, and paternal age, race/ethnicity, body mass index, education, and morbidities (not including hepatitis, PCOS, or HIV). Data are from a combination of maternal report and vital records; variables were categorized as presented in our tables. All potential confounders were included in the final models, with the exception of maternal or paternal morbidities, for which we used a 10% change-in estimate criterion and found that only maternal morbidities was a confounder in our data. Adjustment for daycare was performed as a secondary analysis due to missing information.
We further examined the role of preterm birth in infant feeding, because it is unknown whether the higher prevalence of preterm birth associated with fertility treatments explains any feeding differences. As fertility treatment may affect breastfeeding through mechanisms that include preterm birth (that is mediation, rather than confounding), we did not select on or adjust for preterm birth.(19) For similar reasons, we did not exclude twins or infants with congenital disorders that may lie on a causal pathway between fertility treatment use and breastfeeding.(20,21)
In these analyses, we examined the impact of preterm birth (less than 37 weeks gestational age, from vital records) using a mediation analysis. We estimated the controlled direct effects of fertility treatment on infant feeding, independent of preterm birth, using marginal structural models(22); mediation weights were multiplied by our combined missingness and sampling weight. We obtained effect estimates and percentile confidence intervals after bootstrapping.(22) The DAG used to inform our mediation analyses is provided as supplemental Figure 1. SAS 9.4 was used for statistical analyses (SAS Institute, Cary, NC).
Of the 4,591 mothers included in analyses, 30% (n=1,361) conceived with fertility treatments. Women who conceived with treatments were more likely to be older, non-Hispanic white, and married or living as married, and to have morbidities, fewer prior pregnancies, completed more years of education, and private insurance than women who did not conceive with treatments (Table 1). Fathers of infants conceived with fertility treatments were more likely to be older and non-Hispanic white and to have a higher BMI, morbidities, and completed more years of education than other fathers. Infants conceived with fertility treatments were more often born preterm and at lower birth weights. Among the 571 women who conceived with fertility treatment and linked to the SART within New York State, the leading cause of infertility was female factor (38%, includes diminished ovarian reserve, endometriosis, PCOS, tubal factor, and uterine factor) followed by male factor infertility (19%).
Breastfeeding was more prevalent among mothers conceiving with fertility treatments at four months postpartum (54% versus 48%), but less prevalent by twelve months (18% versus 22%; Table 2). Mothers who used treatments were more likely to provide formula at later time points. The provision of solid foods was similar throughout infancy, but mothers conceiving with treatments were slightly less likely to provide solids at four months and more likely to do so at eight. Mothers who conceived with treatments provided juice less during infancy than other mothers.
Among women who were not breastfeeding at 4 months (n=2008), 22% indicated that they had never attempted to breastfeed. Reasons for this included “did not want to breastfeed” (65%) and “not allowed due to a medical condition for which breastfeeding is not advised” (13%), while 21% did not provide a reason. Women who were not breastfeeding at 4 months gave several reasons for not currently breastfeeding including the infant having difficulty latching or suckling (15%), inability to establish a milk supply (18%), infection or pain in the breasts (6%), other (35%), and 25% did not provide a reason. These reasons differed by fertility treatment status (p<0.001); women who conceived with treatment were less likely to skip this question and more likely to have stopped due to an inability to establish a milk supply (25% versus 15% in the non-treatment group). Reporting problems with milk supply was slightly more common among mothers who conceived with ART than among those who conceived with OI/IUI (27% versus 22%, respectively). Among the mothers not breastfeeding because of difficulty with milk supply, the prevalence of preterm birth was similar and high for both the any fertility treatment and no treatment groups (30%). Information on breastfeeding at 4 months was available among 88% of the mothers who conceived with fertility treatment and linked to SART; the prevalence of breastfeeding among those with known causes of infertility was 59% for those with multiple female factor infertility, 58% for male factor, and 52% for both “female factor, one cause” and “multiple factors, male and female.”
After adjusting for confounding, mothers conceiving with fertility treatments were less likely to breastfeed until twelve months postpartum (adjusted RR 0.58, 95% CI 0.40, 0.86) than those who did not use these treatments (Table 3); no differences in breastfeeding were identified before 8 months. Use of fertility treatment was associated with providing formula earlier in infancy (aRR 1.22, 95% CI 1.11, 1.33), as well as solids and juice.
The patterns of breastfeeding and formula and solids initiation remained consistent when limiting our exposed group to those who used only ART treatments (Table 3). Comparatively, the effects for breastfeeding reached statistical significance at eight rather than twelve months among OI/IUI-exposed and the evidence for earlier introduction of solids was imprecise (Table 3). In other sensitivity analyses, any fertility treatment use continued to be associated with breastfeeding cessation later in infancy and with earlier provision of formula and solids in models that additionally adjusted for daycare initiation or limited to mothers submitting all questionnaires near their intended completion dates. The effect of early juice initiation was still suggested, but unsurprisingly, did not reach statistical significance in the daycare adjusted model. Finally, fertility treatment continued to be associated with breastfeeding cessation at twelve months in our secondary model that adjusted for any pregnancy since enrollment (aRR 0.59, 95% CI 0.40, 0.88).
In mediation analyses, we observed significant controlled direct effects of fertility treatment on breastfeeding cessation at twelve months, independent of preterm birth (Table 4). Evidence for a controlled direct effect of treatment on formula feeding remained as well after accounting for preterm birth.
To our knowledge, this study is the first to identify associations between fertility treatment and a variety of feeding practices throughout infancy among a U.S. population. These analyses suggest that fertility treatment is associated with breastfeeding cessation later in infancy and earlier provision of formula, solids, and juices. We found that the experiences of fertility treatment may still impact feeding when preterm birth is ruled out as a contributing factor and after adjustment for confounding. Families formed by fertility treatment are unique not only in their exposure to these procedures, but also in their healthcare needs and decision making processes—before and after childbirth. There are likely many reasons for associations between assisted conception and infant feeding, such as physical difficulties breastfeeding, feeding multiple infants, a biologic mechanism through which fertility treatment affects infant health, poor workplace or work schedule accommodation for breastfeeding mothers, differential impact of legislation or hospital policy, and personal or sociocultural beliefs. While mothers who used fertility treatments had different feeding practices, regardless of the reasons, our data indicate that breastfeeding prevalence can be improved among all mothers. The overall prevalence in our study was 50% and 33% at four and eight months postpartum, respectively; the Healthy People 2010 target was 50% at six months.(23)
We identified multiple studies that comment on fertility treatment and breastfeeding practices.(11-15, 24-27) Most report no differences in breastfeeding prevalence across modes of conception (12, 24-27), but small sample sizes and use of unadjusted analyses for comparisons (e.g., chi-square tests) are problematic. Furthermore, fertility treatment classifications, definitions of breastfeeding duration, and selection criteria across studies are quite heterogeneous. To our knowledge, McDonald and colleagues provide the only estimates of association between mode of conception and breastfeeding from an adjusted regression model (accounting for factors such as maternal morbidities and neighborhood and hospital characteristics).(13) Among their Canadian population-based cohort, reproductive assistance was associated with increased exclusive breastfeeding at discharge, but the researchers adjusted for factors that may not act as confounders (e.g., gestational age at delivery) and were not able to adjust for some confounders (e.g., education, BMI).(13) Ultimately, the prevalence of breastfeeding at discharge by mode of conception varies across studies, but several do find that mothers who conceived with fertility treatments were more likely to stop breastfeeding in early infancy compared to mothers who did not conceive with treatments.(11,14,15)
Previous researchers have hypothesized that more frequent cesarean deliveries without a trial of labor might account for decreased breastfeeding among women who used ART, but our data do not support this.(4) We saw cesarean without labor was a more common mode of delivery among women who conceived with fertility treatments (44%) than those who did not conceive with treatments (32%; Table 1). Compared to other modes of delivery, the prevalence of exclusive breastfeeding at discharge was lowest among women delivering via a cesarean without labor among both women who conceived with treatments and those who did not (56% exclusive breastfeeding in the treatment group versus 49% in the non-treatment group). However, women who conceived with fertility treatments were more likely than women who did not conceive with treatments to exclusively breastfeed at discharge across all modes of delivery (66% breastfeeding in the treatment group versus 64% in the non-treatment group for cesarean with labor and 75% versus 63% for vaginal delivery, respectively). This higher prevalence of breastfeeding at discharge across modes of delivery generally persisted after further stratifying on preterm birth status. After adjusting for confounding, we did find a negative association between fertility treatment and exclusive breastfeeding at discharge (aRR 0.88, 95%CI 0.78, 0.997). As previously noted, milk supply problems were more prevalent among women conceiving with fertility treatment, especially via ART—which suggests that the relationship between fertility treatment drugs and milk production should be explored. However, it is important to note that factors like plurality or preterm birth may also influence mode of delivery, breastfeeding at discharge, and milk production. Findings from previous studies,(11,14,15) in conjunction with ours, largely suggest that differences in breastfeeding by mode of conception may be a function of breastfeeding maintenance rather than initiation.
ART procedures are associated with epigenetic modifications of genes related to growth and metabolism(28,29) as well as not achieving catch-up body weight in infancy.(12) Perhaps feeding practices are altered to compensate for the unique health and growth of infants conceived with these technologies. We hypothesized that observed differences in feeding were not due to mode of conception per se, but rather the higher proportion preterm infants and low birth weight born to couples using ART(2) and resulting concerns about nutrition-based growth trajectories among parents and physicians.
The American Academy of Pediatrics Committee on Nutrition believes “there is a paucity of data on what to feed the preterm infant after hospital discharge, especially if the goal is to achieve “catch-up” growth.”(3) Preterm infants may have gastric and intestinal problems that impact feeding, but generally human milk is tolerated and preferred over formula.(3) Formula may be introduced when milk intake to support growth is inadequate, but this in turn may lead to decreased breastfeeding and milk production.(30) We did not ask mothers if solid foods or formula were introduced for nutritional supplementation or to promote infant weight gain. However, after appropriately accounting for preterm birth in our mediation analyses, there were direct effects of fertility treatment on breastfeeding and formula feeding. This suggests that among mothers who conceived with fertility treatments, preterm birth does not fully account for the feeding practices we observed.
It is also possible that mothers who conceived with fertility treatments returned to work earlier or returned to jobs with poorer breastfeeding accommodation than their peers who did not conceive with treatments. Previous studies have identified socioeconomic differences between women who do and do not use fertility treatments(6,31); it follows that job type and schedule may differ as well, though differences in postpartum work intentions or work environments by mode of conception are understudied. During recruitment for our study, NY had legislation that allowed mothers to breastfeed in public and required workplace accommodations for expressing milk, but the extent to which these regulations were enforced is unknown.(32) We were limited in evaluating this directly, as information regarding job type or return to part- versus full-time work was unavailable, but our secondary analyses adjusting for daycare initiation produced consistent results. This may suggest that an earlier return to work is not driving our observations.
Around the time that Upstate KIDS recruited mothers, NY also began initiatives to improve breastfeeding rates and standardize hospital policy, including initiating the Breastfeeding Quality Improvement in Hospitals Learning Collaborative, passing the Breastfeeding Mothers’ Bill of Rights law, and developing the New York State Model Hospital Breastfeeding Policy.(33,34) Between the beginning and end of our study, statewide breastfeeding measures improved, but the proportion of infants receiving formula in the first two days of life was still well above the National average in 2011 (29% versus 19%).(35,36) The impact and implementation of these policies may have differed across hospitals and the prevalence of deliveries from assisted conceptions likely varied with hospital type as well; it is possible that breastfeeding differed by mode of conception in part due to a causal pathway through the type of hospital in which a woman received care. Sampling of singletons in Upstate KIDS was matched by perinatal care region of birth(16) and hospital-specific information was unavailable—ultimately limiting our ability to examine associations between hospital policy and breastfeeding duration.
These analyses revealed associations between fertility treatment and breastfeeding after controlling for confounding and preterm birth, but these observations can be attributed to a variety of factors. Ultimately, additional research is needed to untangle sociocultural influences on infant feeding that are associated with fertility treatment from potential biological mechanisms through which treatment affects infant health and subsequently, feeding practices. Infant feeding warrants consideration as an important mechanism through which fertility treatment affects infant health and growth.
Mothers who conceived with fertility treatments were at risk for stopping breastfeeding by late infancy and providing formula, solids, and juices earlier than mothers who did not conceive with these treatments. Some of these associations could be influenced by differential feeding of preterm infants, though we still observed significant controlled direct effects of treatment independent of preterm delivery. Further studies can help determine if differential guidance for infant feeding by mode of conception is warranted.
Capsule: Women who conceived with fertility treatments were more likely to stop breastfeeding and provide formula to their infants earlier. A higher prevalence of preterm birth does not fully explain these associations.
This research was supported by the Intramural Research Program of the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD; contracts #HHSN275201200005C, #HHSN267200700019C). We thank all the members of SART for providing clinical information to the SART Clinic Outcome Reporting System database for use by patients and researchers.
Role of the Funder/Sponsor: The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Disclosures: None of the authors has any financial, personal, political, professional, religious, or intellectual conflicts of interest to disclose.