|Home | About | Journals | Submit | Contact Us | Français|
Primiparity has been associated with 3 to 4 mg/dl lower HDL-C concentrations in Black and White adult women that persist several years after delivery.
To examine effects of adolescent pregnancy on blood lipids, an early risk factor for future cardiometabolic diseases.
The National Heart Lung and Blood Institute’s Growth and Health Study is a multi–center prospective cohort that measured fasting blood lipids for 1,013 (513 Black, 500 White) participants at baseline (1987–1988) ages 9–10, and again at follow-up (1996–1997) ages 18–19.
Change in fasting plasma total cholesterol, triglycerides, LDL-C and HDL-C, defined as the difference between baseline and follow-up measurements, was compared among 186 (145 Black, 41 White) primi- or multiparas, 106 (55 Black, 51 White) nulliparous, gravidas versus 721 (313 Black, 408 White) nulligravidas. Fully adjusted multiple linear regression models estimated blood lipid changes among these pregnancy groups adjusted for race, age at menarche, baseline lipids, physical inactivity, BMI, and family socio-demographics.
In the 10–year study period, adolescent paras compared with nulligravidas had greater decrements in HDL-C (mg/dl) [fully adjusted mean (95%CI) group differences in Black: − 4.3 (− 6.7, −2.0);P<0.001, and White: − 4.5(− 8.2, − 0.7); P=0.016] and greater increments in fasting triglycerides (mg/dl) [adjusted mean (95%CI) group differences in Black: 10.4 (3.9, 16.8); P<0.001, and White: 11.6 (− 3.6, 26.8); P=0.167].
Adolescent pregnancy contributes to pro-atherogenic lipid profiles that persist after delivery. Further research is needed to assess whether adolescent pregnancy has implications for future cardiovascular disease risk in young women.
Parity is associated with lower plasma high-density lipoprotein cholesterol (HDL-C) in mid to late life. Yet, evidence from cross-sectional studies is mixed; some report lower HDL-C among grand multiparas and multigravidas (5 or more births or pregnancies),1;2 and others report a graded linear inverse association 3–5 By contrast, longitudinal studies from before to after pregnancy observed lower HDL-C after a first birth (primiparity) compared to never giving birth (nulliparity) or becoming pregnant (nulligravidity).6;7 The mixed findings imply that the impact of parity on HDL-C may depend on the life stage, socio-demographics, or other determinants of parity, as well as limitations in study design.
Cross-sectional studies, conducted in primarily postmenopausal women, reported 4 to 5 mg/dl lower HDL-C associated with higher parity or gravidity (i.e., ≥5 births versus ≤4 births, and 6–8 pregnancies versus 0 pregnancies, respectively),1;2 as well as a graded inverse association with the number of births.3–5;8 These studies are limited by the lack of pre-pregnancy blood lipid measurements and retrospective assessment of other risk factors within decades after the childbearing years.
Longitudinal studies of women of reproductive age (18–30 yr) based on the Coronary Artery Risk Development in Young Adults (CARDIA) study are unique because measurements of fasting lipids were obtained both before and after pregnancy. In CARDIA, a lower mean HDL-C of 3 to 4 mg/dl was found among primiparas compared with nulligravidas up to 8 years after delivery.7 Moreover, the persistent decrement in HDL-C did not increase with the number of births (i.e., similar magnitude of the decrement in HDL-C for primiparas and multiparas). The findings persisted after adjustment for race, prepregnancy BMI and HDL-C, smoking habit, and socio-demographics as well changes in body weight, waist girth, alcohol intake, and physical activity.7 These longitudinal data support the hypothesis that a first pregnancy is associated with lasting biologic effects on maternal metabolism.
Adolescence is a critical period for development of cardiovascular disease (CVD) risk factors, including the predisposition to obesity.9–11 Blood lipid changes during adolescence coincide with hormonal changes related to the onset of puberty, including decreased total cholesterol and increased triglyceride concentrations.12–14 Excessive weight gain during adolescence could be particularly harmful because of enhanced abdominal fat deposition,15 a strong predictor of future dyslipidemia and insulin resistance.16
Pregnancy during adolescence may have persistent effects that adversely affect future cardiometabolic disease risk. Specifically, pregnant adolescents are more likely than pregnant adults to experience excessive gestational weight gains 17;18 and substantial postpartum weight retention.19 In NGHS, parous compared with nulligravid adolescents gained more overall and central adiposity after pregnancies.20 Thus, adolescent pregnancy may contribute to obesity onset and more atherogenic blood lipids that increase CVD risk later in life.
We sought to determine whether pregnancy has lasting adverse effects on blood lipids among parous (1 or more births) compared with nulligravid (never pregnant) adolescents within the National Heart Lung and Blood Institute’s Growth and Health Study (NGHS), a multi-center, biracial (50% Black) cohort of young females in which blood lipids were measured at ages 9–10 years and again at 18–19 years. In the 10-year NGHS, one third of participants became pregnant during adolescence. The NGHS provides a unique opportunity to examine the natural history of pregnancy during adolescence and its effects on changes in maternal blood lipids within this key developmental period. The prospective cohort design and internal comparison group of nulligravid adolescents allow us to examine the effects of pregnancy on blood lipids apart from other risk factors, including pre-pregnancy lipids.
The NGHS is a 10-year longitudinal observational investigation of the etiologic factors related to the development of risk factors for cardiovascular disease, including obesity, in a cohort of Black and White girls examined annually from childhood (age 9–10 years) up to age 19 years.21;22 Details of cohort recruitment, characteristics, study methods and instruments are described elsewhere.21–23 Briefly, participants were recruited between January 1987 and May 1988 from three centers: 1) University of California at Berkeley, 2) University of Cincinnati Medical Center and Children’s Hospital Medical Center, and 3) Westat, Inc, in Rockville, MD. Participants were recruited by the University of California at Berkeley via census sampling from all public and parochial schools in west Contra Costa County, California, and by the University of Cincinnati/Cincinnati Children’s Hospital Medical Center from public and parochial schools that were racially and socio-economically representative of the greater Cincinnati area in Ohio. Westat Inc. recruited participants from Group Health Association, a health maintenance organization in the Washington, DC area. Institutional Review Boards at each participating study center approved the study. Written, informed consent was obtained from subjects and their parents or guardians for all study procedures.
In 1987–88, 2,379 girls aged 9–10 years (1,213 Black, 1166 White) and their families were enrolled, and 2,094 were re-examined at age 18–19 in 1997–1998 (88% retention). Of 2,094 girls, 471 reported one or more births, 224 reported one or more pregnancies but no births, 84 were missing reproductive history and 1,315 girls reported no pregnancies during the NGHS study period. We selected participants who reported pregnancies or births, or had incomplete information on reproductive history (n=779) for the NGHS Pregnancy study (2002–2005), in which telephone interviews were done to confirm pregnancies and births during the NGHS study period. We also requested permission to abstract pregnancy medical records and to obtain copies of their children’s birth certificates. To evaluate completeness or reporting, we used data collected (i.e., birthdates for their infants) from NGHS Wave-II (n=2,054) in 1998–2001 to ascertain whether additional births had occurred that had not been reported during the NGHS study period.24 For parous adolescents, visits were scheduled at least 4 months postpartum.
For this analysis we selected 1,013 participants (531 Black, 500 White; 43% of the original cohort) who had provided fasting blood specimens analyzed for lipids at both enrollment (1987–88) and follow-up examinations (1997–98), triglycerides ≤400 mg/dl (1 person excluded), as well as pregnancy history during the NGHS study period. Among 1,013 adolescents, 186 (145 Black, 41 White) experienced one or more births (38 multiparas), 106 were pregnant but did not give birth, and 721 were never pregnant during the NGHS study period. Participants not included were more likely to have enrolled in the Cincinnati, OH site and to have parents with less education, and lower family incomes (P<0.01) than the analytic sample.
Blood specimens were collected in the morning after a 12-hour overnight fast. Total and HDL-cholesterol were determined using the Cholesterol CHODPAP method (Bochringer–Mannheim diagnostics). Blood triglycerides were analyzed enzymatically using a commercially available method (Abbott A-Gent Triglycerides Reagent Set). Quantitative assays of blood lipids were performed at the John Hopkins University’s Lipoprotein Analytic Laboratory, a participant in the Centers for Disease Control (CDC) NHLBI Lipid Standardization Program. Low density lipoprotein cholesterol (LDL-C) was calculated using the Friedewald formula: LDL-Cholesterol = Total Cholesterol – HDL-cholesterol – (Triglycerides/5). Lipid changes were calculated as the difference between visit 10 and baseline measurements to ensure consistency of lipid measures prior to menarche, and because lipids were not measured at each exam. Because LDL-C could not be calculated for triglycerides above 400 mg/dl, we excluded 1 participant at follow-up (age 18–19 years).
Body weight was measured to the nearest 0.1 kg with calibrated Healtho-meter electronic scales (Model 482, Sunbeam Products, Inc, Maitland, FL) and participants wearing only a paper hospital gown or large NGHS standard T-shirt. Two weight measurements were obtained and a third was taken if the first two differed by more than 0.3 kg. Height was measured to the nearest 0.1 cm in socks, using custom-made portable stadiometers. A third measurement was taken if the first two were more than 0.5 cm apart. 21
Participants were asked if they had ever been pregnant, were currently pregnant, number of times they had been pregnant, number of births and birth dates of their children annually from visit 6 (age 14–15;1990–1991) through visit 10 (age 18–19;1997–1998). A birth was defined as a live or still birth, abortion, or fetal death for any pregnancy lasting 20 weeks or more that was delivered at least 4 months prior to visit 10. A pregnancy that did not end in a birth was defined as a spontaneous or therapeutic abortion, ectopic or molar pregnancy, or miscarriage < 20 weeks gestation. The Pregnancy study conducted in 2002–2004 and the NGHS Wave-II from 1998–2001 augmented information on pregnancies and births.
Participants were classified as: 1) primi- or multiparous, 2) gravid and nulliparous, or 3) nulligravida. Among parous adolescents, we calculated age at first delivery and dichotomized them as less than 16, or ≥16 years at first delivery.
Annually, participants were asked if they had started their menstrual cycles. Thus, age at menarche was determined from participants’ self-declaration. Questions about oral contraceptive use were asked annually starting at visit 3 (age 11–12). Questions on Norplant and Depo-provera were added to the questionnaires at visits 9 and 10 (ages 17–19) when the contraceptive products came on the market. We categorized girls as never, past, or current users of hormonal contraceptives at the last NGHS visit. Current users at ages 18–19 were classified by type of hormonal contraceptives (i.e., oral or Norplant/Depo-provera).
Parents and guardians provided information on race, age, family composition, maximum parental education, employment, and household income at the baseline examination. Girls provided information on dietary intake and physical activity patterns annually. Methods for collecting dietary intake and physical activity information were validated using actual observed eating and activity behavior as previously described.25 The 3-day food record was selected due to greater accuracy than 24-hour recall or 5-day food frequency methodologies.26 The Nutrition Coordinating Center (NCC) at the University of Minnesota and the Cincinnati Children’s Hospital Medical Center coded the food records from years 1 and 10, respectively, including number of meals and snacks, and estimated nutrients using Version 11 of the NCC nutrient database as detailed elsewhere.21 The nutrition questionnaire also asked participants about their frequency of eating fast foods, and dieting to lose weight.
Physical activity was self-reported via questionnaire and a 3-day diary for the same period as the diet was recorded. The participants were asked to respond to the following statement, “I am physically active, that means I get lots of exercise”, by choosing from among one of three responses: “never or almost never;” “sometimes;” “usually or always.” Self–perception of physical activity responses were dichotomized with the “never or almost never” response as the referent, and “sometimes” and “usually or always” responses combined into another group. A measure of sedentary behavior, girls reported the number of hours of television or video watched per week. Methodology to collect physical activity data and quantify the scores has been previously described.27 If dietary intake or physical activity was missing at baseline (n=70), we used data from the next available visit up to year 4.
Baseline differences in characteristics of participants and their parents were described by race using chi-square statistics for categorical variables (clinic site, household income, parental education, BMI) and by comparison of means for continuous variables using t-tests (fasting plasma lipids, age, height, dietary intake, television and video viewing, and age at menarche). Bivariate associations between race and reproductive characteristics at the end of follow–up were also examined using t-tests and chi-square statistics. Within each race, blood lipids, age at menarche, height at age 18–19 years, and hormonal contraceptive use categories were examined across pregnancy groups using F-statistics from analysis of variance. Blood lipids were similar for primi- and multiparas, and because few adolescents gave birth two or more times (n=38; 4 White, 34 Black), we combined them into one group (one or more births; parous) for each race. All p-values presented are for two-sided tests; statistical significance was defined to be P<0.05.
Unadjusted and multivariable adjusted means (95%CI) and mean group differences in plasma lipid changes among pregnancy groups were estimated from linear regression models. Statistically significant p-values and confidence intervals were corrected for multiple comparisons of paras or nulliparas to nulligravidas using the Dunnett’s procedure in SAS for Windows 9.1.3 (SAS Institute Inc., Cary, NC, USA). Covariates evaluated as potential confounders based on a priori hypotheses included race, baseline measurements, age at menarche, parental education, household income, height at age 18–19 years, and lifestyle behaviors (baseline physical activity and dietary patterns). Covariates were excluded as confounders if they were not associated with the dependent variables independent of the other covariates. Effect modification by race within pregnancy group associations for each lipid was evaluated by introduction of appropriate cross-product terms into the models (significance P<0.10). Adjusted least square means for fasting lipid changes among pregnancy groups (nulligravida, referent) were obtained from race-specific linear regression models. We examined mean plasma lipid changes stratified by race because of the small number of parous White adolescents that limited statistical power to detect effect modification. We also obtained the race-specific results for this study so that our findings could be compared with previous study findings in adults (Black and White) reporting race-specific longitudinal changes in plasma lipids (from before to after pregnancy) associated with parity.
Fully adjusted means (95%CI) for changes in fasting lipids among pregnancy groups were adjusted for the relevant baseline plasma lipid measurement, age, weight and height (BMI) at age 9–10, clinic, age at menarche, parental education, household income, and physical inactivity. Height at age 18–19 years did not affect pregnancy group estimates of lipid changes. We examined BMI change and hormonal contraceptive use as potential mediators of the pregnancy-association with blood lipid changes. Giving birth during adolescence was associated with use of progesterone only contraception (Depo-provera/Norplant) at follow-up; 25–35% of parous versus fewer than 11% of non-parous adolescents. We also conducted a sensitivity analysis that excluded 89 adolescents using Depo-provera or Norplant to assess pregnancy effects without exposure to progesterone-only hormonal contraception (HC).
Baseline characteristics that varied by race include lower parental education and income, lower fasting plasma triglycerides, higher fasting plasma HDL-C, body weight, height, dietary intake (Kcal as fat), and physical inactivity and greater percentages of overweight and obesity among Black versus White girls. Age at menarche was 10 months later on average for White versus Black girls (P<0.001)(Table 1). Reproductive status varied by race (Table 2); Black compared to White adolescents were more likely to become pregnant and/or give birth (28% versus 8% respectively; P<0.001), and more likely to be currently using Norplant or Depo-provera in 1997–1998 (P<0.001).
Baseline and follow-up blood lipids, and BMI were mostly similar by subsequent number of pregnancies and births during adolescence (Table 3). However, Black primi- or multiparas had lower plasma HDL-C and higher triglycerides at follow-up than nulligravidas. Among Whites, primiparas had the highest mean BMI at follow-up. Age at menarche did not differ by pregnancy groups for the Black NGHS participants, but White nulligravidas reached menarche at slightly older ages. Overall, pregnancy groups in both races differed in hormonal contraceptive use (P<0.001). Paras were more likely to be past or current users of hormonal contraceptives, including Norplant or Depo-provera at follow-up (P<0.001). Height differences were significant among paras versus nulligravidas at baseline among Black adolescents and at follow-up among White adolescents. We combined primiparas and multiparas together within each race group for multivariable analyses because of the very small numbers of multiparas in each race group.
In multivariable models, there was no evidence of effect modification by race (Table 4) in the association of pregnancy groups with change in fasting lipids; all race interactions P>0.25, although our power was limited due to the small number of parous whites. We describe lipid changes stratified by race to describe within race characteristics of blood lipids related to parity and gravidity because of the much smaller sample of parous white participants.
Unadjusted mean (95%CI) changes in fasting lipids during the 10-year period were similar among pregnancy groups, except for lower HDL-C and higher triglycerides in primi- or multiparas than nulligravidas. In black and white races, respectively, unadjusted mean group differences in HDL-C were − 4.2 mg/dl lower (P=0.002) and − 4.1 mg/dl (P=0.044) lower in paras than nulligravidas. Triglyceride mean group differences from unadjusted models were 12.0 mg/dl higher among Black primi- or multiparas (P=0.003), and 13.0 mg/dl higher (P=0.127) among White primi- or multiparas than nulligravidas for specific race groups.
In multivariable models fully adjusted for covariates (age, clinic sites, family income, parental education, age at menarche, baseline lipid measurement, BMI at age 9–10, and physical inactivity), mean group differences in HDL-C were slightly higher; − 4.3 mg/dl (−6.7 to −2.0) and −4.5 mg/dl (−8.2 to −0.7) for parous Black and White adolescents, respectively, (P<0.001 and 0.016). Mean group differences for fasting triglycerides were attenuated to 10.4 mg/dl (3.9 to 16.8) and 11.6 (−3.6 to 26.8), respectively, for Black and White paras (P<0.001 and 0.167). Adjustment for BMI change moderately attenuated pregnancy group differences in blood lipids, but HDL-C remained significantly lower for Black (P=0.010) and White paras (P=0.018). Triglycerides increments were also attenuated, but remained significant for Black paras versus nulligravidas (P=0.030). Finally, Norplant, Depo-provera or other hormonal contraceptive use attenuated mean pregnancy group differences in HDL-C among Black (P=0.033) and White girls (P=0.124), but had minimal impact on triglyceride changes. In the sensitivity analysis where progesterone only users were excluded, results (data not shown) remained similar to those for the full sample. There was no evidence for differences in LDL-C, or total cholesterol changes among pregnancy groups in fully adjusted multivariable models. Among paras within race groups, associations did not vary by age at first birth.
Our study findings show that both Black and White parous adolescents experienced greater decrements in HDL-C (4.3 to 4.5 mg/dl) and greater increments in fasting triglycerides (10.4 to 11.6 mg/dl) after pregnancy compared with lipid changes for nulligravid adolescents during the same 10-year study period. These differences in HDL-C and triglycerides for parous adolescents remained significant after adjustment for BMI and lipid measurements at age 9–10 years, age at menarche, family socio-demographics, and lifestyle behaviors, except for triglycerides in White adolescents. The strength of these associations was similar among Black and White adolescents, despite fewer pregnancies in Whites.
Excess weight gain and increased use of hormonal contraception after pregnancy appeared to modestly mediate the association between adolescent pregnancy and pro-atherogenic lipid profiles, primarily for attenuation of differences in HDL-C among the Black adolescents. Yet, both HDL-C and triglyceride differences among parous versus nulligravid participants remained statistically significant after controlling for overall adiposity gains, except for triglyceride differences among Whites. Our sensitivity analysis, in which we excluded hormonal contraceptive users, showed that our findings remained robust among participants never using hormonal contraceptives.
Our mean concentrations for LDL-C and total cholesterol in the NGHS are comparable to national estimates for LDL-C and total cholesterol, respectively, in adolescent females aged 12–17 years; 93.5 mg/dL and 165.9 mg/dL for black, and 89.8 mg/dL and 165.4 mg/dL for white.28 Another longitudinal study of lipid changes in females (20% Black) followed from age 9 to 18 years reported total cholesterol decrements of 19 mg/dl and triglycerides increments of 15 mg/dl, but did not report whether the adolescents had given birth.13 The lipid changes observed within the NGHS cohort are consistent with the previous study, particularly for the White females, but NGHS Black females showed modest decrements in fasting triglycerides.
Our finding that a first birth is associated with lower mean HDL-C independent of weight gain among adolescents is consistent with previous findings for adult women. CARDIA reported that primiparity compared with nulligravidity was associated with 3 to 4 mg/dl lower plasma HDL-C within 2 to 8 years after delivery among both Black women and White women aged 20–31 years.6;7 The HDL-C decrements in CARDIA women persisted after the first birth, and were not greater with subsequent births controlling for changes in adiposity and lifestyle behaviors (i.e., physical activity, alcohol intake).7
In contrast to CARDIA findings, parous adolescents in NGHS also showed higher fasting triglycerides than nulligravidas, although statistical significance was not reached for White adolescents possibly due to the smaller sample of parous adolescents. Implications of our findings are that a more atherogenic lipid profile at younger ages could influence the long-term risk of cardiometabolic diseases in adulthood, 12 and fetal programming in future pregnancies.29 Lower HDL-C levels associated with primiparity represent a 6 to 12 percent greater risk of CVD during midlife.30 Higher parity also has been directly associated with greater risk of CVD in older women, although residual confounding remains an issue.3;31
Black–white differences in physical maturation and the overall pattern of adolescent growth are well known.22;32 Black females reach menarche earlier, and have greater peak velocities in growth, followed slower growth in late adolescence than White females.22 Pregnant adolescents tend to accrue more subcutaneous fat in central locations compared with adult women,33;34 particularly younger, growing pregnant adolescents.18 Previously, paras compared to nulligravid NGHS adolescents had greater increases in both overall and central adiposity.20 However, weight gain did not explain the lower HDL-C or greater triglycerides among paras in our analysis. The specific mechanism for the lipid changes is unclear, but insulin resistance does not explain our findings because the lower HDL-C and higher TG remained after controlling for adiposity. Moreover, parity is not associated with increased incidence of type 2 diabetes after pregnancy in longitudinal studies,35;36 except among women with a history of gestational diabetes mellitus (GDM),36 which is uncommon in females <20 years of age.
Limitations include fewer White than Black parous adolescents, variable ages for deliveries, later maturation of White girls, and the tendency for Black adolescents to become pregnant at younger ages. We also did not have sufficient numbers of multiparas to assess whether decrements in HDL-C showed a threshold effect or a monotonic trend with higher order births. We adjusted for age at menarche, baseline lipid measurements, and socio-demographic covariates to minimize these differences, but they may still be influential. Hormonal contraceptive use during follow-up was a consequence of prior pregnancies, and appeared to mediate rather than confound our findings. Although we did not assess blood glucose and insulin in our models, adjustment for baseline BMI and changes in BMI during the 10-year period accounted for these metabolic characteristics which may result from excess fat deposition. Adolescents may have under-reported pregnancies ending in miscarriage or abortions which would bias our findings toward the null hypothesis.
The study strengths include the large, community-based sample of Black and White girls that provides an internal comparison group of never pregnant adolescents to evaluate the direct effects of pregnancy on adolescent blood lipid profiles independent of growth in stature, maturation (age at menarche), and secular trends. Blood lipid measurements were obtained prospectively via standard research methodology both before and after pregnancies.
Our findings are potentially important because adolescence has been identified as one of the “critical periods” of growth and development that set the stage for future adult chronic disease, including diabetes and cardiovascular diseases.37 Excessive fat deposition during adolescence may lead to persistent obesity,38 elevated insulin, atherogenic lipids and higher blood pressure levels into young adulthood.39 Relevant to our findings, HDL-C and triglycerides are important predictors of future cardiovascular disease and possibly, diabetes in adulthood.40 Further, in Black women, earlier age at a first birth (<20 years) has been associated with increasing rates of coronary heart disease.31;41 Our findings show that pregnancy at an early age results in lowering of HDL-C and raising of triglycerides that is not explained by pregnancy-related weight retention. Pregnancy during adolescence may have even greater adverse effects on women’s future cardiometabolic health in mid life.
Pregnancy during adolescence or adulthood exerts lasting pro-atherogenic effects independent of weight gain. Future investigation is needed into the possible roles of lactation and central obesity as influencing the return of HDL-C and triglycerides to preconception levels, as well as prevention of long-term cardiometabolic diseases later in life. The demonstration of cardiovascular disease in early life gives credibility to risk factor examination of children and the need for beginning prevention and screening, particularly among parous adolescents.37 Evaluation of maternal lipid profiles among postpartum adolescents may identify those who would benefit from early lifestyle interventions, including adolescents who are not obese. Comprehensive behavioral interventions for postpartum adolescents could promote more favorable maternal blood lipids and glucose tolerance prior to conception, as well as newborn health in future pregnancies.42
Funding: National Institutes of Health, Bethesda, MD; N01-HC-55023-26 from the National Heart, Lung, and Blood Institute, R01 HD39304 from the National Institute of Child Health and Human Development, K01 DK059944 from the National Institute of Diabetes, Digestive and Kidney Diseases, and R01 MH57897 from the National Institute of Mental Health and the National Institute of Diabetes, Digestive and Kidney Diseases.
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.
The authors declare no conflicts of interest.