By the time children reach preschool age, racial/ethnic disparities in obesity are present. Specifically, the prevalence of obesity is 16·5 % among Hispanic/Latinx, 11·6 % among Black, 7·0 % among Asian and 9·9 % among White American children aged 2–5 years(Reference Ogden, Fryar and Hales1). Higher prevalence among racial/ethnic minorities suggests a need to examine risk and protective factors for obesity early in life. While numerous studies have identified early-life biological and behavioural factors that influence obesity risk throughout the lifespan, the literature on racial/ethnic differences in maternal feeding factors during infancy has largely focused on breast-feeding status and age of introduction to solids(Reference Baidal, Locks and Cheng2,Reference Dixon, Peña and Taveras3) . Compounded by the fact that the majority of early childhood obesity prevention studies in general have been conducted with higher socio-economic, non-ethnically diverse samples, there remains a knowledge gap in the field when it comes to understanding the emergence of racial/ethnic disparities in childhood obesity rates(Reference Baidal, Locks and Cheng2,Reference Grossman, Bibbins-Domingo and Curry4) .
Parental feeding behaviours, beyond those directly related to infant dietary patterns (e.g. mode of infant feeding and early introduction of solids), influence obesity risk during early life(Reference Mitchell, Farrow and Haycraft5,Reference Schwartz, Scholtens and Lalanne6) . For example, when faced with concerns about child underweight or overweight, caregivers often assume more control over child nutrition(Reference Clark, Goyder and Bissell7). However, the decisions parents make about feeding their children have immediate and long-lasting implications for child growth and development(Reference Fisher and Birch8). For example, maternal control of feeding can affect weight gain acceleration or deceleration in children as early as infancy(Reference Farrow and Blissett9). Infants naturally regulate their energy intakes, but parents’ behaviours can override hunger and satiety cues(Reference DiSantis, Hodges and Johnson10). Pressured feeding (i.e. encouraging infants to finish their bottle), ‘bottle propping’ (i.e. giving infants a bottle by leaning it on a pillow, blanket or other support) and giving infants a bottle to go to bed with have been independently found to be associated with increased obesity risk(Reference Dewey, Goldberg, Prentice and Prentice11,Reference Kimbro, Brooks-Gunn and McLanahan12) . However, responsive feeding, generally defined as developmentally appropriate responses to the infant’s hunger and satiety cues, can have a positive effect on infant weight gain. Although feeding in the first year of life necessitates a great deal of parental control, infants do best when they are allowed to self-regulate their food consumption(Reference Farrow and Blissett9).
Independent of potential confounders, Black and Hispanic/Latinx children have demonstrated increased odds of rapid infant weight gain, greater maternal control of infant feeding, bottle propping and pressured feeding compared with White children(Reference Perrin, Rothman and Sanders13,Reference Taveras, Gillman and Kleinman14) . Differences in risk and protective factors have been attributed to socio-economic status, culture, acculturation and a history of disadvantage and discrimination(Reference Dixon, Peña and Taveras3,Reference Caprio, Daniels and Drewnowski15) . Studies have also demonstrated how feeding styles in Black and Hispanic/Latinx parents influence infant energy intake and infant size, and how maternal characteristics can be associated with infant feeding styles(Reference Gross, Fierman and Mendelsohn16,Reference Thompson, Adair and Bentley17) . Racial/ethnic differences in early-life risk factors for obesity require further understanding and evaluation as they may contribute to the high prevalence of obesity among minority preschool-age children and may thus signify important areas for intervention.
The Infant Feeding Questionnaire (IFQ), developed by Baughcum et al., assesses maternal feeding practices that may promote self-regulation and other healthy eating habits in children to prevent obesity(Reference Baughcum, Powers and Johnson18). Specifically, the IFQ was designed to assess nurture-based feeding practices adopted by mothers when feeding their infants during the first year of life as well as the beliefs that guide such practices.
Breast-feeding is less common among low-resourced women(Reference Cohen, Alexander and Krebs19), who are also more likely to smoke or be smoke exposed(Reference Carswell, Ward and Vander Weg20). The Baby’s Breath study created and tested an intervention for decreasing smoke exposure of low-income pregnant women and their newborns. The purpose of this study was to use the IFQ to examine racial/ethnic differences in maternal infant feeding practices and beliefs at 6 months postpartum in a sample at high risk for early obesity.
Methods
Study design
This secondary data analysis cross-sectionally examines data collected for Baby’s Breath, a randomised control trial aimed at limiting environmental tobacco smoke exposure among pregnant women in the state of Rhode Island(Reference Risica, Gavarkovs and Parker21,Reference Risica, Parker and Jennings22) . Participants were recruited from prenatal clinics that largely serve low-income women in the Providence area. Women were eligible if they spoke English, were at least 18 years of age, had access to a working telephone and VCR/DVD player and were either a current smoker, recently quit smoker or a non-smoker with current exposure to second-hand smoke from someone in their household. Women who were pregnant with multiples or further than 16 weeks’ gestation at the time of recruitment were not eligible to participate. Additional recruitment details along with intervention details have been published elsewhere(Reference Risica, Gavarkovs and Parker21). Briefly, participants were randomised to receive newsletters and videos during pregnancy and the first 6 months of the postpartum period aimed at either smoking cessation and avoidance or other healthy pregnancy topics. Any information regarding infant feeding was given to both experimental groups. Informed written consent was obtained from all participants, and they were compensated for their time. The analytic sample for this analysis includes participants for which maternal feeding practices and beliefs were assessed at 6-month postpartum (n 343).
Measures
Main exposure
The primary exposure in this analysis was mother’s racial/ethnic background. At baseline (16 weeks’ gestation), participants reported whether or not they identified as Hispanic or Latina. They were then asked if they identified as American Indian or Alaskan Native, Asian, Black or African American, Caucasian or White, Native Hawaiian or Pacific Islander, multiracial and/or other. The Department of Finance method of racial/ethnic classification, in which Hispanic/Latinx is deemed a mutually exclusive racial category, was used, such that Hispanic/Latina mothers were classified as Hispanic/Latina regardless of their racial identity(Reference Mays, Ponce and Washington23). Given the sample distribution, we examined race/ethnicity as four-level categorical variable in our analyses, with categories non-Hispanic White, Hispanic/Latina, non-Hispanic Black and other (which included multiracial individuals).
Outcomes
The outcomes of interest were maternal infant feeding practices and beliefs self-reported by mothers at 6 months postpartum using the IFQ(Reference Baughcum, Powers and Johnson18) (see Table 1). The IFQ consists of twenty items, for which factor analysis resulted in the following seven factors: (1) concern about the child undereating or becoming underweight, (2) concern about infant’s hunger, (3) awareness of infant’s hunger and satiety cues, (4) concern about the infant becoming overweight or overeating, (5) feeding the infant on a schedule, (6) using food to calm the infant’s fussiness and (7) social interaction during feeding. Items are assessed using a 5-point Likert scale. The response options for the 12 items assessing maternal feeding practices or child eating behaviours include 0 = never, 1 = rarely, 2 = sometimes, 3 = often and 4 = almost always. The response options for the remaining eight items, which assess maternal beliefs, are 0 = disagree a lot, 1 = disagree a little, 2 = no strong feelings either way, 3 = agree a little and 4 = agree a lot. Higher scores thus indicate a higher frequency of the specified maternal feeding practice or a higher agreement with the maternal belief statement. In addition to the continuous scales, we also dichotomised the items to report the proportion of mothers who commonly (i.e. sometimes, often or almost always) engaged in each practice and agreed (either a little or a lot) with each belief statement.
Items in the Infant Feeding Questionnaire
* Response options were 0 = never, 1 = rarely, 2 = sometimes, 3 = often, 4 = almost always.
† Factors are: (1) concern about the child undereating or becoming underweight, (2) concern about infant’s hunger, (3) awareness of infant’s hunger and satiety cues, (4) concern about the infant becoming overweight or overeating, (5) feeding the infant on a schedule, (6) using food to calm the infant’s fussiness and (7) social interaction during feeding.
‡ Item reverse scored when calculating factor score.
§ Response options were 0 = disagree a lot, 1 = disagree a little, 2 = no strong feelings either way, 3 = agree a little, 4 = agree a lot.
Consistent with Baughcum et al., factor scores were created by calculating the mean score of the items included in each factor(Reference Baughcum, Powers and Johnson18). If any item was missing for factors consisting of just two or three items (i.e. factors 2, 4, 5, 6 and 7), the factor score was considered missing. Factors with four items (i.e. factors 1 and 3) were considered missing if more than one item was missing; if only one item was missing, the missing item score was replaced with the mean score of the other items in that factor before calculating the factor score. Also consistent with Baughcum et al., the first item ‘Did you let him/her eat whenever he/she wanted to?’ was reverse scored because of negative loading on factor 5 (feeding infant on a schedule)(Reference Baughcum, Powers and Johnson18). Cronbach’s α scores for these factors in the Baby’s Breath sample were similar to those reported by Baughcum et al. (2001)(Reference Baughcum, Powers and Johnson18).
Other measures
Socio-demographic and other sample characteristics were collected at baseline via survey, with the exception of parity, which was abstracted from hospital records. Participant age was reported as a continuous variable while all other variables were categorical. Marital status was dichotomised as either married/cohabitating or not married/cohabitating. Parity was dichotomised as either multiparous (previous live births) or not. For employment status, participants indicated whether they were employed full time, employed part time, not employed, a student or other. Participants reported their highest level of education which was categorised as less than high school diploma, high school diploma or general education degree, or at least some college or technical school. As this was a low-income sample, annual household income was categorised as less than $10 000 per year, between $10 000 and $30 000 per year or more than $30 000 per year. Finally, we assessed breast-feeding status in the postpartum period via survey by asking participants if they had initiated breast-feeding after birth, as well as whether or not they were exclusively breast-feeding, exclusively formula feeding or mixed feeding (i.e. partially breast-feeding) at both 12 weeks and 6 months postpartum.
Statistical analyses
Bivariate associations between race/ethnicity and socio-demographic characteristics and breast-feeding status were analysed using χ 2 tests (with Fisher’s exact tests used when there were cell values less than five), while associations with continuous scores from the IFQ were analysed using ANOVA. We then constructed multivariable regression models to assess associations between racial/ethnic background with IFQ factor scores adjusting for potential confounders. Potential confounders, informed by the literature, included age, marital status, parity, education, household income and breast-feeding status. Multiple imputation was used for missing covariate data which was assumed to be missing at random. The statistical significance level was set at 0 05 for all procedures, with analyses performed using Stata 15.0 (Stata: Software for Statistics and Data Science, 2020).
Results
The racial/ethnic composition of the sample was 39 % non-Hispanic White (n 134), 28 % Hispanic/Latina (n 95), 13 % Black (n 46) and 20 % other (n 68). More than half (56 %) of the mothers who were classified as other identified as multiracial; American Indians/Alaskan Natives (22 %) and Asians (16 %) were also represented in this group (See Table 2). The average age of participants at baseline was 24·0 (sd 5·0) years, just over half (53 %) of the participants were married or cohabitating and half (51 %) had no prior children. Employment status and highest level of education varied across the sample, with 47 % of participants not employed and 37 % not having received a high school diploma. Three-quarters of the sample (74 %) had an annual household income of $30 000 or less, with 42 % of the overall sample having an annual household income less than $10 000. This distribution slightly over-represents non-White populations in comparison to the overall population in 2015 as 57 % non-Hispanic White, 7 % non-Hispanic Black, 25 % Hispanic, 12 % other or mixed race exclusive of these groups. (PRAMS) and 13 % with less than 12 years of education, though the proportion married was similar(Reference Monteiro, Cooper and Viner-Brown24).
Sample demographics, overall and by race/ethnicity
Ns may not add to the total sample size due to missing values; GED, general education degree.
* Includes fifteen individuals identifying as American Indian/Alaskan Native, eleven Asian, one Native Hawaiian or Other Pacific Islander, three other and thirty-eight multiracial; P-value is for χ 2 test, Fisher’s exact used when there were cell values <5.
Demographic differences were observed across race/ethnicity with Hispanic/Latina and Black participants less likely to be married or cohabitating and more likely to be multiparous compared with non-Hispanic White children. Differences across race/ethnicity were also observed for breast-feeding status. While more than half of the sample (61 %) initiated breast-feeding after birth, initiation was highest among Hispanics/Latinas (79 %, P < 0·001) compared with the other groups. However, by 12 weeks only 22 % of the sample was still breast-feeding, with only 6 % of the sample exclusively breast-feeding and no differences by race/ethnicity. By 6 months, only 10 % of the sample was still breast-feeding (3 % exclusively). Of the mothers exclusively breast-feeding at 6 months, all but one was non-Hispanic White.
Table 3 displays the sample averages for maternal feeding practices and behaviours at 6 months as well as differences in means across racial/ethnic categories. While nearly half of the mothers in this sample (48 %) reported (either sometimes, often or almost always) worrying about whether or not their infant was eating enough, only 11 % of mothers worried that their infant would become underweight. Mothers commonly endorsed at least sometimes, often or almost always putting cereal in their infant’s bottle so that they would stay full (46 %) or sleep longer (43 %), and half (48 %) of the sample reported at least sometimes, often or almost always feeling that their infant wanted more than just formula and/or breast milk prior to being 4 months old. The large majority of mothers (92–99 %) felt agreed that both they and their infant knew when he or she was hungry and full. However, 37 % at least sometimes, often or almost always worried that their infant was eating too much and 22 % agreed with the statement that they worried that their infant would become overweight. Most mothers at least sometimes, often or almost always allowed their infant to eat whenever he or she wanted (84 %) and reported that at least sometimes, often or almost always feeding was the first thing they did when their infant was fussy (76 %). Most mothers also reported that they at least sometimes, often or almost always held their infant when giving a bottle (96 %) and at least sometimes, often or almost always talked to him or her while feeding (93 %).
Maternal feeding practices and beliefs at 6 months, overall and by race/ethnicity
* Includes those identifying as multiracial.
† n (%) represents those agreeing either a little or a lot to belief statements and those who either sometimes, often or almost always practised behaviours in practice statements.
‡ Responses were recorded using the following scale: 0 = disagree a lot, 1 = disagree a little, 2 = no strong feelings either way, 3 = agree a little, 4 = agree a lot; response options for all other items were: 0 = never, 1 = rarely, 2 = sometimes, 3 = often, 4 = almost always.
§ Item was reverse scored; P-value is for ANOVA test.
Racial/ethnic differences in several feeding practices and behaviours were observed. Racial/ethnic minority mothers more frequently put cereal in their infant’s bottle so the infant would stay full longer (P = 0·032) and stated their infant wanted more than just formula or breast milk prior to 4 months (P = 0·019) than non-Hispanic White mothers. Allowing the infant to eat whenever he/she wanted (P = 0·023) was also more frequently reported by racial/ethnic minority mothers than non-Hispanic White mothers, however, so too was only allowing the infant to eat at set times (P < 0·001).
Table 4 shows the multivariable linear regression models for each IFQ factor. In bivariate analyses, we found racial/ethnic differences for factors 2 (concern about infant’s hunger), 5 (feeding infant on a schedule) and 7 (social interaction with the infant during feeding). These associations persisted after adjustment for maternal age, marital status, parity, education, household income and breast-feeding status at 6 months. After adjustment, Hispanic/Latina and Black mothers were more concerned about their infant’s hunger (β: 0·40 (95 % CI 0·09, 0·71) and β: 0·61 (95 % CI 0·22, 0·99), respectively) than non-Hispanic White mothers. They were also more likely to feed on a schedule (β: 0·70 (95 % CI 0·37, 1·02) and β: 0·60 (95 % CI 0·20, 1·00), respectively), as were mothers classified as other compared with non-Hispanic White mothers (β: 0·60 (95 % CI 0·25, 0·95)). Compared with non-Hispanic White mothers, Hispanic/Latina mothers and mothers classified as other reported less social interaction with their infant during feeding (β: –0·23 (95 % CI –0·43, –0·04) and β: –0·26 (95 % CI –0·47, –0·04), respectively).
Multivariable adjusted associations of race/ethnicity and Infant Feeding Questionnaire factors
* P < 0 05.
Referent group for race/ethnicity is non-Hispanic White, for marital status is not married/cohabitating, for multiparous is no, for education is some college and for breast-feeding at 6 months is formula only; age and income are continuous.
In the adjusted models, we additionally observed racial/ethnic differences for factors 1 (concern about infant undereating or becoming underweight) and 4 (concern about infant overeating or becoming overweight). Mothers classified as Hispanic/Latina or other reported greater concern about their infant undereating or becoming underweight than non-Hispanic White mothers (β: 0·25 (95 % CI 0·04, 0·46) and β: 0·24 (95 % CI 0·01, 0·46), respectively), while Hispanic/Latina and Black mothers reported greater concern about their infant overeating or becoming overweight than non-Hispanic White mothers (β: 0·25 (95 % CI 0·01, 0·49) and β: 0·31 (95 % CI 0·01, 0·61), respectively).
Discussion
In this secondary data analysis with a high-risk sample, racial/ethnic minority mothers reported feeding practices and beliefs that demonstrated greater concern about their infant’s hunger, as well as the amount of food their infant consumed (i.e. concern about under- and overeating) and his or her weight status (i.e. concern about the infant becoming underweight or overweight) compared with non-Hispanic White mothers at 6 months postpartum. Racial/ethnic minority mothers were also more likely to feed their infant on a schedule and less likely to socially interact with their infant while feeding, than White mothers. Observed differences were independent of maternal age, marital status, parity, education, household income and breast-feeding status.
The findings from this study are consistent with the previous literature on racial/ethnic differences in maternal feeding practices in early childhood(Reference Perrin, Rothman and Sanders13,Reference Taveras, Gillman and Kleinman14) . Perrin et al. found that racial/ethnic minority mothers, especially Hispanics/Latinas, were more likely to encourage their infant to finish feedings(Reference Perrin, Rothman and Sanders13). Similarly, in this study, racial/ethnic minority mothers expressed greater concern regarding underweight or undereating. However, unlike Perrin et al., feeding as a first response to a crying or fussy baby did not differ by race/ethnicity in this sample. Our finding that racial/ethnic minority mothers were more likely to feed their infant on a schedule is consistent with findings by Taveras et al., in which Hispanic/Latina and Black mothers had nearly two times the odds of engaging in restrictive feeding practices compared with White mothers(Reference Taveras, Gillman and Kleinman14). This study advances the literature by examining and observing differences in additional maternal feedings practices and beliefs (e.g. concern about infant overeating or becoming overweight and social interaction during feeding), across race/ethnicity. Our findings suggest that mothers, especially racial/ethnic minorities, may simultaneously be concerned about their infant under- and overeating or becoming under- or overweight, and that social interaction during feeding, which may foster responsive feeding, varies across groups.
Differences with regard to a mother’s concern about her infant being hungry or becoming underweight may explain why racial/ethnic minority infants are more likely to be introduced to solids early(Reference Perrin, Rothman and Sanders13,Reference Taveras, Gillman and Kleinman14) . Further underscoring this point, we found that racial/ethnic minority mothers were more likely to mix cereal in their infant’s bottle so he or she would stay full longer and felt as though their infant wanted more than just formula and/or breast milk prior to 4 months of age. This is an important topic for future intervention efforts to consider as early introduction to solids (i.e. prior to 4 months of age) and may increase obesity risk(Reference Daniels, Mallan and Fildes25,Reference Pearce, Taylor and Langley-Evans26) . The presence of significant differences across race/ethnicity in regard to feeding on a schedule may be another important intervention target as responsive feeding has long been recommended to foster the development of infant self-regulation, though more research explicitly measuring dimensions of appetite regulation (e.g. hunger, satiety and satiation) is needed(Reference DiSantis, Hodges and Johnson10,Reference Hurley, Cross and Hughes27) . Our findings also suggest that racial/ethnic minority mothers are relatively concerned about their infant’s weight status and/or weight gain trajectory. Thus, interventions targeting maternal feeding should address concerns about both infant’s undereating or becoming underweight and overeating or becoming overweight.
In targeting maternal infant feeding practices and beliefs, interventions must consider the contextual factors that shape those practices and beliefs. In addition to being influenced by factors related to socio-economic status and family structure, maternal infant feeding practices are also shaped by cultural influences(Reference Pak-Gorstein, Haq and Graham28). For example, some racial/ethnic minorities prefer chubby infants and toddlers and frequently do not identify their overweight children as being overweight(Reference Chatham and Mixer29). Cultural preferences in which heavier infants are perceived as healthier infants may shape the way mothers feed their children. Cultural preferences like these can also be exacerbated by life experiences like food insecurity; for example, one study described how Latina immigrant mothers not only needed to resist indulgent feeding practices but they also need to resist the pressure to give in that arises from the scarcity they faced during their own childhood(Reference Pineros-Leano, Tabb and Liechty30). Cultural preference and experiences like these likely contribute to the higher rates of racial/ethnic minority mothers engaging in feeding practices that increase obesity risk such as encouraging their infant to finish their bottle, supplemental feeding, bottle propping, immediately feeding their infant when crying and putting their infant to bed with a bottle(Reference Kimbro, Brooks-Gunn and McLanahan12,Reference Perrin, Rothman and Sanders13) . Without consideration for cultural preferences and beliefs, intervention efforts aimed at promoting certain maternal feeding behaviours (e.g. nutrition counselling) are unlikely to be successful.
The findings from this study also highlight additional practices and beliefs interventions, particularly interventions targeted at high-risk low-income samples, should consider. Mothers in this sample frequently reported worrying about whether the amount of food their infant was eating was sufficient or healthy. Prior research has found that low-income mothers are more concerned about their infant’s hunger than high-income mothers and that they may find it difficult to withhold food from their children even when they have just eaten(Reference Baughcum, Powers and Johnson18,Reference Jain, Sherman and Chamberlin31) . The majority of mothers in this sample also reported that feeding their infant was the first thing they would do when he or she was fussy. Low-income mothers may thus require additional support around picking up on infant hunger and satiety cues and calming without feeding. Low-income mothers are less likely to believe that infants know their own hunger and satiety than high-income mothers(Reference Gross, Mendelsohn and Fierman32); thus, interventions require teaching or other encouragement of mothers about the ability of infants to self-regulate.
Taken together, the findings from this study may be used to inform maternal infant feeding intervention efforts aimed at high-risk populations. Additional efforts in this area are paramount as maternal feeding practices during infancy have been associated with later feeding practices, eating behaviours and obesity risk(Reference Koletzko, Von Kries, Monasterolo, Koletzko, Decsi and Molnar33–Reference Taveras, Scanlon and Birch35). For example, infants who are encouraged to finish their bottles have been shown to be about twice as likely to eat all of the food on their plate at 6 years old than those who were rarely encouraged to finish their bottle during early infancy(Reference Li, Scanlon and May34). Further, young children from both racial/ethnic minority backgrounds and low socio-economic status households have a disproportionate risk for obesity and are more likely to be exposed to feeding practices associated with weight gain(Reference Weden, Brownell and Rendall36).
While the present study expands the literature on racial/ethnic differences in the ways in which mothers feed their infants, it is not without limitations. First, as a cross-sectional secondary data analysis, we were limited in the analyses we could conduct and variables we could adjust for. Also, the fact that mothers self-reported their feeding practices means our results may be biased by mothers’ perception. Additional research using more robust study designs and measures would further expand the literature on this topic and allow investigations of changes over time. Second, we classified race/ethnicity using the Department of Finance method(Reference Mays, Ponce and Washington23). Racial/ethnic classification is a complex task due to the heterogeneity and social and cultural complexity of groups, and method of classification can have important implications for the results. While prior studies used the rarest group method to classify multiracial individuals(Reference Mays, Ponce and Washington23), the majority of multiracial participants in this sample did not further specify their racial/ethnic background; thus, we included individuals identifying as multiracial in the ‘other’ category. Results from this group must be interpreted with caution due to extreme within-group heterogeneity. Finally, the study sample was comprised of low-income and smoke-exposed women living in the Providence area, RI; thus, our results may not be generalisable to broader populations. While the implications of smoke exposure on the relationship between race/ethnicity and maternal infant feeding are unknown, racial/ethnic differences in maternal infant feeding practices are likely more pronounced in samples that represent different levels of socio-economic status given associations between race/ethnicity, socio-economic status and infant feeding behaviours.
Conclusions
This study revealed differences in maternal infant feeding practices and beliefs across race/ethnicity among low-income mothers at 6 months postpartum. These differences were not explained by mothers’ age, marital status, parity or educational level, nor by household income or breast-feeding status. Given the evidence linking maternal feeding practices during infancy to childhood obesity risk, interventions are needed to further promote responsive feeding among mothers identifying as racial/ethnic minorities and address their concerns regarding their infant’s hunger, amount of food being eaten and weight status.
Acknowledgements
Acknowledgements: We would like to acknowledge the mothers who participated in this study. Financial support: The intervention in this study was funded by the National Heart, Lung, and Blood Institute (NHLBI; grant number R01 HL070947-O1A). Authorship: T.A. conducted the analysis and led manuscript preparation, C.L. drafted the introduction and A.A. and P.M.R. helped formulate the research questions and guided the analysis. All authors contributed to the interpretation of the findings and editing of the manuscript. Ethics of human subject participation: This study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving research study participants were approved by the institutional review board of Brown University. Written informed consent was obtained from all subjects/patients.
Conflict of interest:
There are no conflicts of interest.
Open access
