Association of a Woman's Own Birth Weight With Subsequent Risk for Gestational Diabetes FREE

Studies from several countries have demonstrated that birth weight is inversely related to subsequent risk for type 2 diabetes mellitus (DM),^{1 - 3} insulin resistance,^{2 ,4} and other features of the metabolic or insulin resistance syndrome.^{1 - 2 ,4} These findings support the ‘"fetal origins" hypothesis—that susceptibility to these chronic adult conditions may be programmed in utero. However, other studies, particularly among younger populations, have yielded conflicting findings,^{5 - 6} and some investigators have argued that the observed relation between chronic disease risk and impaired fetal growth could be confounded by later environmental influences and/or by relative postnatal growth.^{7 - 9} In addition, studies have been recently criticized for failure to adjust adequately for gestational age, maternal chronic disease, and other perinatal factors, which may influence both birth size and future chronic disease risk in the offspring.^{8 - 9}

Gestational DM (GDM) is an important complication of pregnancy that strongly predicts the later development of type 2 DM¹⁰ and shares many features of the insulin resistance syndrome.^{11 - 12} A readily measurable outcome that occurs relatively early in life, GDM is well suited for testing the fetal origins hypothesis in young adult populations. However, to date, only 2 published studies have examined the association between markers of a woman's own fetal growth and her subsequent risk for GDM as a pregnant adult.^{13 - 14} The results of these studies are conflicting, and interpretation of findings is limited by small sample sizes,¹³ failure to evaluate dose-response associations,¹³ lack of information on prepregnancy body mass index (BMI),¹⁴ and possible bias due to self-reported outcomes.¹⁴ In this article, we describe findings from a large, population-based study that addresses these concerns. Specifically, we investigate the relation between markers of a woman's own growth in utero and her subsequent risk for GDM among young New York women.

Data for this case-control study were obtained from 2 large, computerized state databases maintained by the New York State Department of Health: the live birth registry and the New York State hospital discharge records. Eligible subjects were defined as all women who completed a first pregnancy (and delivered a liveborn infant) in Upstate New York between 1994 and 1998, and who were also born in New York State in 1970 or later. The New York Bureau of Biometrics performed all record linkages and supplied the final anonymous data set of linked pregnancy and birth records. The study was approved by the Colorado Multiple Institutional Review Board of the University of Colorado.

Hospital discharge records and vital record information for each 1994-1998 pregnancy were matched using a unique combination of mother's medical record number, hospital number, and infant date of birth. Linked pregnancy records indicating birth of the mother in New York State between 1970 and 1985 were then matched to the mother's own birth records using the mother's full maiden name and date of birth. Matched records were also checked manually to verify accuracy. Of the 95 309 eligible linked pregnancy records, 57 588 (60%) were successfully matched to the women's own birth records, of which 29 924 pertained to a first pregnancy. We excluded women whose first pregnancies were complicated by multifetal gestation (n = 270), by the use of illegal drugs (n = 678), or by preexisting DM, heart disease, essential hypertension, renal disease, or other preexisting chronic or serious acute conditions requiring medication or monitoring (n = 3482). Women could have more than one reason for exclusion. Also excluded were women with pregnancy-induced hypertension (PIH) (n = 2278), a condition strongly associated with insulin resistance,¹⁵ leaving a final sample of 23 395 eligible primigravida women.

Cases were defined as eligible women with a record of GDM on their firstborn's birth certificates and/or a diagnosis of GDM (International Classification of Diseases, Ninth Revision [ICD-9] code 648.0) or abnormal glucose tolerance (ICD-9 code 648.8) on their hospital discharge records. Universal screening for GDM has been widely recommended,^{16 - 17} although not always practiced^{18 - 19} since the 1980s, and diagnosis is generally made based on specific standardized clinical criteria.²⁰ A total of 440 GDM cases (1.9% of all eligible first pregnancies) were identified on linked birth registry and New York Statewide Planning and Research Cooperative System (SPARCS) pregnancy records. Controls were defined as all remaining eligible subjects whose first pregnancies were uncomplicated by GDM (N = 22 955). Among eligible subjects, age at delivery ranged from 12 to 28 years and averaged 21.1 ± 3.5 years.

Growth in utero was measured as the subject's own birth weight and gestational age as recorded on her birth certificate. Birth weight was evaluated alone (as an indicator of absolute birth size) and adjusted for gestational age in weeks (as a measure of relative fetal growth). Implausibly extreme values (eg, birth weights <450 g and gestational ages <22 weeks or >46 weeks) were set to missing. Using New York vital record information, we assessed the relation of GDM risk to other factors characterizing the woman's own birth and experience in utero, including multifetal gestation, birth order, maternal age, and parental educational levels at the time of the woman's birth, and preeclampsia, chronic hypertension, diabetes, or other serious conditions complicating the pregnancy of the woman's mother (including heart or kidney disease, abruptio placentae, and placenta previa). We also evaluated the effects of potential risk factors characterizing the woman's first pregnancy, including her race/ethnicity, age, marital status, educational level, employment and insurance status, participation in welfare programs, onset of prenatal care, alcohol and tobacco use during pregnancy, height, prepregnancy BMI, and pregnancy weight gain, using New York birth registry data.

We used multiple logistic regression analysis to assess the independent effects of a woman's own birth weight, gestational age, and other factors on the development of GDM, and to determine the influence of potential confounders. To evaluate the effects of birth weight adjusted for gestational age on GDM risk, we included gestational age as a covariate in the multiple logistic regression model; we assessed the influence of gestational age adjusted for birth weight in the same manner. Logistic regression was also used to evaluate trends in the effects of birth weight, gestational age, and other explanatory variables on GDM risk. We conducted additional analyses stratified by factors known or suspected to modify the association between fetal growth and GDM or chronic disease risk, including prepregnancy BMI, height, and race/ethnicity.

Primigravida women whose pregnancy records could not be linked to those from their own birth were more likely than matched women to be black (20% vs 12%) and receiving Medicaid (46% vs 38%) and averaged slightly younger at the completion of their first pregnancy (20.7 vs 21.1 years). However, unmatched women did not differ from matched women in other demographic or anthropometric characteristics, in the onset of prenatal care, or in substance use during pregnancy. Most important, matching to a birth record was not associated with the prevalence of GDM either overall or within different categories of age and race/ethnicity.

Age was strongly and positively related to risk for GDM in a woman's first pregnancy, while a woman's educational level was inversely related to her risk for GDM after adjustment for other demographic factors (Table 1). The apparent association of GDM risk to marital and employment status was largely eliminated by adjustment for other demographic factors (Table 1).

Risk for GDM rose with increasing prepregnancy BMI and decreasing prepregnancy height (Table 2). Adjustment for other perinatal factors did not appreciably alter the effect of BMI on risk for GDM, but attenuated the influence of pregnancy weight gain and strengthened the relation of GDM risk to prepregnancy height (Table 2).

Birth weight, both alone and adjusted for gestational age, showed a U-shaped relationship to risk for GDM, with both low and high birth weights demonstrating increases in risk (adjusted odds ratios [ORs] for <2000 g and ≥4000 g, respectively, 2.16 [95% confidence interval {CI}, 1.04-4.50] and 1.53 [95% CI, 1.03-2.27]) (Table 3). Up to 4000 g, risk for GDM decreased progressively with rising birth weight (P for trend = .002). Adjustment for other perinatal factors strengthened this inverse, dose-response relationship and increased the magnitude of the association between lower birth weight and GDM risk (adjusted OR for <2000 g, 4.23; 95% CI, 1.55-11.51; P for trend <.001). In contrast, controlling for other risk factors and specifically, BMI and maternal diabetes, largely eliminated the relation between high birth weight and risk for GDM (OR after adjustment for these 2 factors alone = 1.16, 95% CI, 0.72-1.87) (Figure 1). Excluding women born to mothers whose pregnancies were complicated by diabetes or other conditions did not appreciably change the inverse relation between birth weight and risk for GDM, and the inverse association between birth weight and GDM risk was as strong among women born preterm as among those born at term.

As illustrated in Figure 2, the inverse association of birth weight to risk for GDM was more pronounced among women who were shorter and leaner prior to pregnancy (P for interaction = .02 and .007, respectively). After adjustment for gestational age and prepregnancy weight, birth weight less than 2500 g was associated with a 2.3-fold increase in risk for GDM among women shorter than 162.6 cm (OR, 2.33; 95% CI, 1.24-4.37; P for trend = .001), but with only a modest, nonsignificant increase among taller women (OR, 1.39; 95% CI, 0.67-2.87; P for trend = .49). Similarly, the association between reduced birth weights and subsequent risk for GDM was stronger among women with a prepregnancy BMI of less than 25 than among those with a prepregnancy BMI of 25 or more. Adjusted ORs for low birth weight, respectively, were 2.68 (95% CI, 1.37-5.25; P for trend <.001), vs 1.21 (95% CI, 0.65-2.26; P for trend = .54).

Women born to mothers whose pregnancies were complicated by diabetes were themselves over 3 times as likely to develop GDM in their first pregnancies as were those born to nondiabetic mothers. Adjustment for other factors slightly increased the magnitude of this association (Table 3). Women born following any maternal complication of pregnancy were also at increased risk for developing GDM in their own pregnancies. Although these women were more likely to be born low birth weight (OR for birth weight < 2000 g, 2.61; 95% CI, 2.11-3.23), adjusting for birth size did not alter the relation of GDM risk to maternal pregnancy complications (Table 3), nor did removing women born to diabetic pregnancies (OR, 1.62; 95% CI, 1.11-2.42). However, adjustment for other perinatal factors—in particular BMI and short stature—attenuated this association (Table 3).

We found little evidence for an association between risk for GDM and other factors characterizing the woman's own birth, including maternal educational level (Table 1), gestational age, birth order, multifetal gestation (Table 3), or parental age (data not shown). Similarly, a woman's alcohol or cigarette consumption during pregnancy was not related to her risk of developing GDM (Table 2), nor was the timing of her first prenatal care visit or participation in low-income programs (data not shown). We also found little evidence for a modifying influence of demographic or perinatal factors other than prepregnancy BMI and height on the association between markers of fetal growth and later risk for PIH.

In this study of young, primigravida New York mothers, we found a striking, U-shaped relationship between a woman's risk for GDM in her first pregnancy and her own birth weight as a newborn, both alone and adjusted for gestational age. The association of large birth size to GDM risk was largely explained by prepregnancy BMI and maternal diabetes, but adjustment for these and other potential confounders strengthened the inverse association between birth weight and the subsequent development of GDM. Only 2 published studies have evaluated the association of a woman's birth size to subsequent risk for GDM, with conflicting results. A study of Australian women reported no association, although sample sizes in this matched case-control investigation were small (138 GDM cases), and dose-response associations were not evaluated.¹³ In agreement with our findings, Egeland et al¹⁴ demonstrated a U-shaped relationship between birth weight and GDM risk in their investigation of birth characteristics and self-reported GDM in a cohort of parous Norwegian women (N = 498 GDM cases).

Recent studies of nonpregnant populations have yielded findings consistent with our results. Dabelea et al²¹ noted a U-shaped relationship between birth weight and glucose levels in young Pima Indians that was attenuated by adjustment for current body size. Studies of the same population have also demonstrated a U-shaped association between birth weight and subsequent risk for DM,^{21 - 22} and adjustment for maternal diabetes strongly reduced the association of high birth weight to diabetes.²² Likewise, a recent investigation from the Nurses' Health Study suggested a U-shaped association between birth weight and risk for type 2 diabetes,³ and adjustment for maternal history of diabetes eliminated the apparent association of high birth weight to risk for type 2 diabetes, but strengthened that of low birth weight.³

This study is among the first to demonstrate a clear, dose-response relationship between relative fetal growth and adult risk for GDM, an important complication of pregnancy and a strong predictor of diabetes and other insulin resistance conditions. Our findings meet the criteria of Lucas et al,⁹ who argue that for data to support the fetal origins hypothesis, birth weight should have a significant inverse effect in models unadjusted for current weight or BMI, and that the magnitude of this inverse association should be greater in models including current weight. Some previous investigations have reported an inverse relation between birth weight and glucose intolerance that is apparent only after adjustment for current body size⁹ or that is stronger among heavy² and/or tall²³ individuals, suggesting that greater postnatal change in size may explain or amplify the influence of birth weight. In our study, the inverse relation between birth weight and subsequent GDM risk was apparent prior to adjustment for body size, and this association was stronger among women who were leaner and shorter as adults. These findings indicate that the influence of reduced fetal growth on risk for GDM in this population is neither explained nor amplified by greater postnatal growth, and suggest that the effects of reduced birth size, short stature, and obesity on risk for GDM are independent and may operate in part via separate mechanistic pathways.

Biological mechanisms underlying the association between low birth weight and the later development of GDM may include early metabolic changes. Like type 2 diabetes and the metabolic syndrome,¹⁵ GDM is characterized by insulin resistance,¹¹ glucose intolerance,¹¹ hyperlipidemia,^{24 - 25} and impaired beta cell and endothelial function.^{26 - 27} Low birth weight has been linked to endothelial²⁸ and beta cell²⁹ dysfunction, dyslipidemia,^{4 ,30} insulin resistance,^{2 ,4 ,31} and reduced glucose tolerance^{2 ,31} later in life, and a growing body of experimental research in animals suggests that these changes can be programmed in utero.^{29 ,32 - 34} Such alterations in lipid metabolism and endocrine function, whether resulting from fetal growth restriction or other factors, may compromise a woman's ability to meet the diabetogenic challenges of pregnancy, and thereby increase her likelihood of developing GDM.

Our finding of increased risk of GDM among women who were themselves products of diabetic pregnancies is consistent with previous studies. Maternal diabetes has been associated with elevated risk for obesity and glucose intolerance in both children^{35 - 37} and adults.^{34 ,37} Cohort studies of Pima Indians have shown maternal glucose levels in pregnancy to predict gestational glucose levels in the female offspring.³⁶ Likewise, in a recent study of Norwegian women, Egeland et al¹⁴ noted an elevated risk for GDM among women who were offspring of diabetic pregnancies. The relation of GDM to other maternal complications of pregnancy remains little studied. The observed increase in risk associated with maternal pregnancy complications in this study may be due in part to undiagnosed maternal diabetes, hypertension, or other insulin resistance conditions. Neither adjusting for diabetes and other conditions complicating the pregnancy of the woman's mother nor removing women born to mothers with any pregnancy complications altered the observed inverse association between birth weight and GDM risk.

The reduced weight gain associated with GDM that we observed in this study is consistent with the results of several previous studies and may reflect dietary restrictions commonly recommended in managing GDM^{38 - 39} and obesity in pregnancy.⁴⁰ In the current study, risk for GDM rose with increasing prepregnancy BMI, increasing age, and decreasing prepregnancy height. Older maternal age and higher prepregnancy BMI are established risk factors for GDM.¹⁰ In agreement with our findings, several recent studies have also linked short stature to elevated risk for GDM.^{41 - 44} In a study of pregnant Greek women, Anastasiou et al demonstrated increasing severity of glucose intolerance with decreasing mean height, and a negative correlation between basal insulin resistance and height; adjustment for BMI and other factors did not alter these findings.⁴¹ Likewise, studies of nonpregnant adults have demonstrated associations between short stature and glucose intolerance that are independent of age,^{45 - 48} BMI,^{45 - 48} and malnutrition in utero.⁴⁸

High birth weight was strongly and positively related to adult prepregnancy BMI and height in this study. Although adjustment for height alone did not materially alter the risk estimates associated with birth weight, adjustment for BMI reduced the effect of large birth size on risk for GDM. Because maternal diabetes is associated with high birth weight and subsequent obesity and glucose intolerance in the offspring,^{2 ,22} some of the observed effect of both high birth weight and high BMI on GDM risk may also reflect undiagnosed or unreported maternal diabetes.

We did not observe an elevated GDM risk among minority women, an association commonly reported in other studies,⁴⁹ likely reflecting the very young age of our study population (Table 1). Data from our own and previous epidemiological studies of GDM^{16 ,50} suggest that GDM prevalence among minority women may not surpass that in non-Hispanic white women until the early to mid-20s. We observed an interaction between minority status and age, with GDM rates (per 100 first pregnancies) among minority vs non-Hispanic white women as follows: 0.9 vs 1.3 (age <21 years), 2.5 vs 2.9 (age 21-24 years), 2.6 vs 3.2 (age 25 years or older) (P for interaction = .04). Including multigravida women and women with PIH in the sample yields higher GDM prevalence rates, but a similar cross-over pattern: 1.5 vs 1.8 (age <21 years), 3.4 vs 3.1 (age 21-24 years), 5.3 vs 3.6 (age 25 years or older) for minority vs non-Hispanic white women, P for interaction <.001.

Strengths of this study include the population-based design, the racial and ethnic diversity of the study population, and the large sample sizes, which allowed us to evaluate in detail the potential influence of both low and high birth weights on the later development of GDM and to assess potentially important modifying or confounding factors. Using linked vital record and hospital discharge data also enabled us to increase the sensitivity in detecting accurately coded obstetric and birth outcomes. Since the data on birth characteristics were derived from vital records and thus gathered well in advance of that on pregnancy outcomes, the influence of measurement or diagnostic bias is likely to be negligible.

Nonetheless, the use of retrospective vital record and hospital discharge data has some inherent limitations. We were unable to adjust for certain known and potential risk factors for GDM, including family history of diabetes, dietary intake, and exercise habits. In addition, although we had information on several indexes of socioeconomic status, we lacked detailed data on social class, income, and other socioeconomic factors. However, it is improbable that confounding by unmeasured variables could explain our findings. In previous investigations with more detailed data on these factors,^{3 - 4 ,21 ,51 - 54} adjustment for socioeconomic status, family history of diabetes, and other potential confounders did not appreciably alter the inverse association observed between birth size and diabetes, insulin resistance, or glucose intolerance. A potential source of bias is underreporting of certain pregnancy outcomes in the New York State vital records. In particular, our ascertainment of maternal diabetes (diabetes complicating the pregnancy of the woman's mother) is likely incomplete, based on recent prevalence estimates.^{10 ,55} Since gestational age estimates were based on the last menstrual period, a method that tends to overestimate true gestational age,⁵⁶ preterm birth is also likely to have been underreported. However, any such misclassification is unlikely to be differential with regard to subsequent GDM diagnosis.

It is possible that some women diagnosed with GDM had underlying nongestational DM, although this is unlikely to explain our findings. In a postpartum study of women with a history of GDM, Kjos et al⁵⁷ found that 9% had chronic DM; in our sample of very young primigravida women without a record of PIH or other medical risk factors, the prevalence is likely to be much lower. Gestational DM may itself have been underascertained, leading to misclassification of controls. However, the GDM prevalence rates observed in this study are similar to or higher than rates reported for women of comparable age and race in other large population-based studies of maternal diabetes,^{16 ,50 ,58 - 59} including those in which all subjects were screened for GDM.^{58 - 59} Moreover, when multigravida and women with a record of PIH are included in the sample, GDM prevalence among pregnant women 25 to 28 years of age increases to 3.9% overall, and to 5.3% among minority women, figures in the mid to upper range of usual reported rates.^{18 ,60} Nonetheless, it is likely that GDM ascertainment was incomplete due to inadequate screening and varying diagnostic practices.^{18 - 19} Late or no initiation of prenatal care may also have contributed to underdiagnosis of GDM, although this factor was unrelated to GDM risk in this population. Most important, however, since any resulting misclassification of controls is unlikely to be differentially related to a woman's birth weight, incomplete ascertainment of GDM would be expected only to attenuate the observed association between GDM and birth weight.

Another possible limitation of this study is the inability to link approximately 40% of pregnancy and birth records. However, successful linkage of birth and pregnancy records is unlikely to be differentially related to a woman's birth weight.

In summary, this large, population-based study of New York–born women demonstrated a strong inverse association between a woman's own birth weight and her subsequent risk for GDM, an outcome of considerable clinical importance in younger populations. The results of this study are consistent with the hypothesis that susceptibility to diabetes and related insulin resistance conditions may be programmed in utero. In particular, our findings suggest that early life factors, and in particular, fetal growth, may be important in the etiology of GDM, a major complication of pregnancy and predictor of type 2 diabetes.