Epidemiology and Prevention

Transcription

1 Epidemiology and Prevention Associations of Pregnancy Complications With Calculated Cardiovascular Disease Risk and Cardiovascular Risk Factors in Middle Age The Avon Longitudinal Study of Parents and Children Abigail Fraser, MPH, PhD; Scott M. Nelson, MBChB, PhD; Corrie Macdonald-Wallis, MSc; Lynne Cherry, PhD; Elaine Butler; Naveed Sattar, MBChB, PhD*; Debbie A. Lawlor, MBChB, MSc, PhD* Background The nature and contribution of different pregnancy-related complications to future cardiovascular disease (CVD) and its risk factors and the mechanisms underlying these associations remain unclear. Methods and Results We studied associations of pregnancy diabetes mellitus, hypertensive disorders of pregnancy, preterm delivery, and size for gestational age with calculated 10-year CVD risk (based on the Framingham score) and a wide range of cardiovascular risk factors measured 18 years after pregnancy (mean age at outcome assessment, 48 years) in a prospective cohort of 3416 women. Gestational diabetes mellitus was positively associated with fasting glucose and insulin, even after adjustment for potential confounders, whereas hypertensive disorders of pregnancy were associated with body mass index, waist circumference, blood pressure, lipids, and insulin. Large for gestational age was associated with greater waist circumference and glucose concentrations, whereas small for gestational age and preterm delivery were associated with higher blood pressure. The association with the calculated 10-year CVD risk based on the Framingham prediction score was odds ratio 1.31 (95% confidence interval, ) for preeclampsia and 1.26 (95% confidence interval, ) for gestational diabetes mellitus compared with women without preeclampsia and without gestational diabetes mellitus, respectively. Conclusions Hypertensive disorders of pregnancy and pregnancy diabetes mellitus are independently associated with an increased calculated 10-year CVD risk. Preeclampsia may be the better predictor of future CVD because it was associated with a wider range of cardiovascular risk factors. Our results suggest that pregnancy may be an important opportunity for early identification of women at increased risk of CVD later in life. (Circulation. 2012;125: ) Key Words: cardiovascular disease risk factors long term follow-up longitudinal cohort study prediction pregnancy Cardiovascular disease (CVD) is the leading cause of death in women, accounting for a quarter of deaths in both high-income and low- and middle-income settings. 1 It is increasingly recognized that women experiencing common pregnancy-related complications such as gestational diabetes mellitus (GDM), 2,3 preeclampsia, 4 intrauterine growth retardation, 5 and preterm delivery 6,7 are at increased risk of future CVD. Therefore, it has been suggested that pregnancy offers an opportunity to identify women at risk of future CVD However, whether these pregnancy complications have separate, independent effects on future cardiovascular risk, and, if so, the manner in which their relative and absolute associations differ from each other remain unclear. Such information is important for exploring whether there are common underlying pathways between these conditions and future cardiovascular risk and for considering the most efficient methods for using pregnancy complications to target preventive initiatives in women. Editorial see p 1336 Clinical Perspective on p 1380 It is well established that women with GDM are at increased risk of developing diabetes mellitus later in life, 2 and diabetes mellitus is an established cardiovascular risk factor. 11 Several studies have reported on associations of pregnancy-related complications with other cardiovascular risk factors postpartum, such as increased lipid concentra- Received May 19, 2011; accepted January 18, From the Medical Research Council Centre for Causal Analyses in Translational Epidemiology, School of Social and Community Medicine, University of Bristol, Bristol (A.F., C.M.-W., D.A.L.); School of Medicine, University of Glasgow, Glasgow (S.M.N.); and Institute of Cardiovascular and Medical Sciences, British Heart Foundation Glasgow Cardiovascular Research Centre, University of Glasgow, Glasgow (L.C., E.B., N.S.), United Kingdom. *Drs Sattar and Lawlor contributed equally to this work. Correspondence to Abigail Fraser, MPH, PhD, Medical Research Council Centre for Causal Analyses in Translational Epidemiology, University of Bristol, Oakfield House, Oakfield Grove, Bristol BS8 2BN, United Kingdom. Abigail.Fraser@bristol.ac.uk 2012 American Heart Association, Inc. Circulation is available at DOI: /CIRCULATIONAHA

2 1368 Circulation March 20, 2012 Table 1. Characteristics of Eligible Women Included and Excluded From Analyses Excluded n Included in Analyses* P Age at birth of index child, y, mean (SD) 29.3 (4.7) (4.4) 0.01 Manual social class, % (n) 13.8 (92) (349) 0.01 University level education, % (n) 16.7 (118) (678) 0.09 Prepregnancy BMI, kg/m 2, mean (SD) 22.7 (3.8) (3.3) 0.08 Parity 3, % (n) 5.2 (36) (139) 0.15 No smoking during pregnancy, % (n) 84.2 (627) (2937) 0.19 Pregnancy diabetes mellitus, % (n) None 96.3 (660) 96.1 (3279) GDM 0.4 (3) 0.5 (18) Pregestational diabetes mellitus 0.4 (3) 0.4 (13) Glycosuria 2.9 (17) 3.0 (106) HDP, % (n) None 83.6 (660) 83.9 (2868) Gestational hypertension 14.2 (115) 14.1 (479) Preeclampsia 2.3 (19) 2.1 (69) Preterm birth, % (n) 6.4 (52) (145) 0.03 Size for gestational age, % (n) Small 11.1 (50) 8.1 (278) Appropriate 79.3 (356) 81.8 (2793) Large 9.6 (43) 10.1 (345) Follow-up measures Age, y, mean (SD) 47.8 (4.6) (4.4) 0.13 BMI, kg/m 2, mean (SD) 27.0 (5.7) (5.1) Waist circumference, cm, mean (SD) 85.4 (13.0) (12.0) SBP, mm Hg, mean (SD) (12.2) (12.5) 0.92 DBP, mm Hg, mean (SD) 71.8 (8.0) (8.1) 0.62 Glucose, mmol/l, mean (SD) 5.29 (0.87) (1.01) 0.95 Insulin, mu/l, mean (SD) 4.98 (1.94) (1.82) 0.05 Proinsulin, pmol/l, mean (SD) 6.04 (1.84) (1.71) 0.03 Triglycerides, mmol/l, mean (SD) 0.95 (1.56) (1.53) 0.07 HDL, mmol/l, mean (SD) 1.46 (0.40) (0.39) 0.19 LDL, mmol/l, mean (SD) 2.97 (0.82) (0.80) 0.70 C-reactive protein, mg/l, mean (SD) 1.13 (3.15) (3.10) 0.41 Nonsmokers, % (n) 85.4 (626) (2560) BMI indicates body mass index; GDM, gestational diabetes mellitus; HDP, hypertensive disorders of pregnancy; SBP, systolic blood pressure; DBP, diastolic blood pressure; HDL, high-density lipoprotein; and LDL, low-density lipoprotein. *Included are women who contributed to any analysis; therefore, the number varies between 2877 women included in analyses with the use of smoking at 18 years after pregnancy and 3416 women who contributed to at least 1 exposure/outcome pair analyzed. tions, insulin resistance, increased levels of inflammatory markers, higher prevalence of metabolic syndrome, and vascular dysfunction (reviewed in Retnakaran 12 ). However, most studies have either limited sample sizes, 12 have measured single cardiovascular risk biomarkers, or have measured these shortly after pregnancy, with few exceptions (eg, Lauenborg et al, 13 Magnussen et al, 14 Romundstad et al, 15 and Gunderson et al 16 ). Therefore, it is uncertain whether such associations persist in the longer term but before actual CVD events. In the present study, we examine and compare associations of pregnancy diabetes mellitus (pregestational diabetes mellitus, GDM, and glycosuria), hypertensive disorders of pregnancy (HDP) (gestational hypertension and preeclampsia), size for gestational age, and preterm delivery with a wide range of cardiovascular risk factors and the calculated 10-year risk of CVD based on the Framingham risk score in women in early middle age. In doing so, we seek to identify specific pathways linking pregnancyrelated complications with future cardiovascular health. Methods The Avon Longitudinal Study of Parents and Children (ALSPAC) is a prospective population-based birth cohort study that recruited

3 Fraser et al Pregnancy Complications and CVD Risk 1369 Figure 1. Venn diagram of pregnancy-related complications. SGA indicates small for gestational age; LGA, large for gestational age; and HDP, hypertensive disorders of pregnancy. some pregnant women resident in Avon, United Kingdom, with expected dates of delivery April 1, 1991 to December 31, 1992 ( A total of women with a live singleton birth consented to have their obstetric data abstracted from medical records. Ethical approval for this study was obtained from the ALSPAC Law and Ethics Committee and the local Research Ethics Committee. Pregnancy Complications Research midwives, using a standard protocol, abstracted information on clinical diagnoses of GDM and glycosuria for the index pregnancy from the antenatal, pregnancy, and postnatal medical records of all women, as described previously. 17 Information on glycosuria (recorded as none, trace,,,, or more) was abstracted from the records of each antenatal clinic visit made by the woman (median number, 14 per woman). The practice in the United Kingdom at the time was for all women to be offered urine tests for glycosuria at each antenatal clinic visit. Universal screening of women with a random or fasting blood glucose level or with an oral glucose tolerance test was not undertaken, and diagnostic tests for GDM (most commonly an oral glucose tolerance test) were only undertaken in women with established risk factors (family history, previous history of GDM or macrosomic birth, South Asian ethnicity) or glycosuria. Glycosuria was defined as a record of at least (equal to 13.9 mmol/l or 250 mg/100 ml) on at least 2 occasions at any time during the pregnancy. 17 Although glycosuria has been suggested to be a poor indicator of GDM, we have previously shown its face validity by demonstrating a clear association with birth size and macrosomia. 17 In addition, at recruitment to the study, women were asked about existing diabetes mellitus, any previous history of GDM, whether they had ever been diagnosed with high blood pressure/ hypertension, and, if so, whether this was restricted to blood pressure in previous pregnancies. Using these data, we classified women into 1 of 4 mutually exclusive categories: no evidence of glycosuria or diabetes mellitus (hereafter referred to as healthy women ); pregestational diabetes mellitus before the pregnancy; GDM (ie, a diagnosis in the medical records of GDM in any woman with no history of pregestational diabetes mellitus); and glycosuria (ie, glycosuria on 2 occasions in women with no evidence of pregestational diabetes mellitus or GDM). Figure 2. Median (interquartile range) Framingham cardiovascular disease (CVD) score by pregnancy-related complications (A, pregnancy diabetes; B, HDP; C, birth size; D, gestational age). IQR indicates interquartile range; HDP, hypertensive disorders of pregnancy; SGA, small for gestational age; AGA, appropriate for gestational age; and LGA, large for gestational age.

4 1370 Circulation March 20, 2012 Table 2. Multivariable Associations of Pregestational Diabetes Mellitus, Gestational Diabetes Mellitus, and Glycosuria With Cardiovascular Risk Factors Measured 18 Years After the Index Pregnancy Risk Factor No Glycosuria or Diabetes Mellitus (n 3061) Glycosuria (n 99) Mean Difference (95% CI) Gestational Diabetes Mellitus (n 17) Mean Difference (95% CI) Pregestational Diabetes Mellitus (n 10) Mean Difference (95% CI) BMI, kg/m 2 *(n 3364) Age-adjusted mean (SE) (0.96) (1.07) (1.54) (1.71) Model (0.42, 2.43) 5.25 (2.89, 7.60) 0.24 ( 3.01, 2.53) Model ( 0.47, 0.86) 1.07 ( 2.64, 0.50) 1.49 ( 3.32, 0.34) Model ( 0.49, 0.84) 1.37 ( 2.94, 0.21) 1.70 ( 3.53, 0.12) Waist circumference, cm (n 3358) Age-adjusted mean (SE) (2.26) (2.52) (3.62) (4.01) Model (0.99, 5.70) (7.64, 18.71) 0.89 ( 7.40, 5.62) Model ( 0.98, 2.53) 0.17 ( 3.98, 4.32) 3.43 ( 8.28, 1.42) Model ( 1.10, 2.40) 0.87 ( 5.04, 3.31) 4.26 ( 9.10, 0.58) SBP, mm Hg* (n 3364) Age-adjusted mean (SE) (2.34) (2.61) (3.75) (4.16) Model (0.01, 4.90) 5.12 ( 0.62, 10.86) 5.90 ( 0.85, 12.65) Model ( 0.83, 3.92) 0.18 ( 5.44, 5.80) 4.82 ( 1.74, 11.38) Model ( 0.73, 3.87) 0.97 ( 6.46, 4.52) 3.60 ( 2.76, 9.97) DBP, mm Hg* (n 3364) Age-adjusted mean (SE) (1.54) (1.72) (2.47) (2.74) Model ( 0.01, 3.20) 0.54 ( 3.23, 4.31) 0.06 ( 4.38, 4.49) Model ( 0.54, 2.60) 2.50 ( 6.20, 1.21) 0.60 ( 4.92, 3.73) Model ( 0.47, 2.60) 3.26 ( 6.91, 0.39) 1.25 ( 5.48, 2.99) Glucose, mmol/l Age-adjusted mean (SE) 4.93 (0.17) 5.26 (0.19) 7.51 (0.27) (0.31) Model (0.17, 0.52) 2.58 (2.18, 2.99) 7.13 (6.58, 7.67) Model (0.11, 0.46) 2.35 (1.94, 2.76) 6.82 (6.32, 7.32) Model (0.10, 0.46) 2.25 (1.83, 2.67) 6.78 (6.28, 7.28) HDL, mmol/l Age-adjusted mean (SE) 0.79 (0.07) 0.70 (0.08) 0.61 (0.12) 0.87 (0.13) Model ( 0.16, 0.01) 0.17 ( 0.35, 0.01) 0.10 ( 0.13, 0.34) Model ( 0.13, 0.02) 0.01 ( 0.19, 0.16) 0.13 ( 0.08, 0.33) Model ( 0.13, 0.02) ( 0.17, 0.17) 0.14 ( 0.07, 0.35) LDL, mmol/l Age-adjusted mean (SE) 1.29 (0.15) 1.33 (0.17) 1.15 (0.24) 0.34 (0.27) Model ( 0.14, 0.17) 0.14 ( 0.51, 0.23) 0.86 ( 1.35, 0.37) Model ( 0.14, 0.18) 0.23 ( 0.60, 0.13) 0.97 ( 1.41, 0.53) Model ( 0.14, 0.17) 0.27 ( 0.64, 0.10) 1.00 ( 1.44, 0.55) CI indicates confidence interval; BMI, body mass index; model 1, adjustment for age at measurement; model 2, additional adjustment for prepregnancy BMI, education (yes/no university level), parity, and smoking during pregnancy; model 3, additional adjustment for hypertensive disorders of pregnancy, preterm delivery, and size for gestational age; SBP, systolic blood pressure; DBP, diastolic blood pressure; HDL, high-density lipoprotein; and LDL, low-density lipoprotein. *n for exposure categories 3487, 105, 18, and 14 for no diabetes mellitus/glycosuria, glycosuria, gestational diabetes mellitus, and pregestational diabetes mellitus, respectively. Obstetric data abstractions also included every measurement of systolic and diastolic blood pressure and proteinuria entered into the medical records and the corresponding gestational age and date at the time of these measurements. These measurements were obtained in routine clinical practice by trained midwives and obstetricians. The median number (interquartile range) of blood pressure measurements in pregnancy was 14 (11, 16), and that of urine measurements was 11 (10, 14). We applied the International Society for the Study of Hypertension in Pregnancy 18 criteria to all of the clinic data to determine women with preeclampsia and those with gestational hypertension. According to these criteria, preeclampsia was defined as a systolic blood pressure 140 mm Hg or a diastolic blood pressure 90 mm Hg, measured on at least 2 occasions after 20 weeks of gestation, with proteinuria, diagnosed if the protein reading on dipstick testing (Albustix; Ames Company, Elkhart, IN) was at least 1 (30 mg/dl), occurring at the same time as the elevated blood pressure. 18 Gestational hypertension was defined as the same pattern of elevated blood pressure but without proteinuria. 18 Thus, all women were categorized into 1 of 3 mutually exclusive

5 Fraser et al Pregnancy Complications and CVD Risk 1371 Table 3. Multivariable Associations of Pregestational Diabetes Mellitus, Gestational Diabetes Mellitus, and Glycosuria With Cardiovascular Risk Factors Measured 18 Years After the Index Pregnancy No Glycosuria or Diabetes Mellitus (n 3061) categories of no HDP, gestational hypertension, or preeclampsia. Fifty-six women without HDP, 42 with gestational hypertension, and 9 with preeclampsia reported having hypertension before pregnancy. When these women were excluded from the analysis, results were unchanged from those presented here. Gestational age and birth weight were ascertained from obstetric records. Size for gestational age was categorized as small Glycosuria (n 99) Gestational Diabetes Mellitus (n 17) Pregestational Diabetes Mellitus (n 10) Insulin, mu/l Age-adjusted geometric mean (95% CI) 6.70 (5.34, 8.40) 8.47 (6.59, 10.89) (7.27, 14.87) 5.58 (3.61, 8.62) Model (1.13, 1.42) 1.55 (1.17, 2.04) 0.83 (0.57, 1.20) Model (1.06, 1.33) 1.12 (0.86, 1.46) 0.74 (0.52, 1.05) Model (1.06, 1.33) 1.10 (0.84, 1.43) 0.73 (0.52, 1.04) Proinsulin, pmol/l Age-adjusted geometric mean (95% CI) 4.58 (3.74, 5.62) 5.63 (4.49, 7.06) 7.70 (5.58, 10.64) 3.92 (2.65, 5.80) Model (1.11, 1.37) 1.67 (1.30, 2.15) 0.85 (0.61, 1.19) Model (1.05, 1.27) 1.24 (0.98, 1.57) 0.76 (0.56, 1.05) Model (1.04, 1.27) 1.21 (0.95, 1.53) 0.76 (0.56, 1.04) Triglycerides, mmol/l Age-adjusted geometric mean (95% CI) 0.64 (0.54, 0.75) 0.71 (0.59, 0.85) 0.71 (0.55, 0.91) 0.55 (0.41, 0.75) Model (1.02, 1.21) 1.11 (0.91, 1.35) 0.87 (0.67, 1.13) Model (0.99, 1.17) 0.96 (0.79, 1.16) 0.82 (0.64, 1.06) Model (0.99, 1.16) 0.93 (0.77, 1.13) 0.81 (0.62, 1.04) C-reactive protein, mg/l Age-adjusted geometric mean (95% CI) 1.20 (0.78, 1.84) 1.38 (0.86, 2.22) 1.31 (0.66, 2.58) 2.38 (1.05, 5.44) Model (0.92, 1.44) 1.07 (0.63, 1.81) 1.96 (0.97, 3.96) Model (0.82, 1.25) 0.57 (0.35, 0.94) 1.55 (0.80, 2.98) Model (0.83, 1.25) 0.56 (0.34, 0.92) 1.57 (0.81, 3.02) CI indicates confidence interval; model 1, adjustment for age at measurement; model 2, additional adjustment for prepregnancy body mass index, education (yes/no university level), parity, and smoking during pregnancy; and model 3, additional adjustment for hypertensive disorders of pregnancy, preterm delivery, and size for gestational age. These ratios are interpreted as percentage (relative) differences; for example, the results from model 3 for proinsulin show a 15% increase in those with glycosuria compared with those with no pregnancy diabetes mellitus. (SGA) (a birth weight 10th percentile of birth weight for gestational age); appropriate (AGA) (between the 10th and 90th percentiles for gestational age); and large (LGA) ( 90th percentile for gestational age) with the use of the study population centiles. Using sex-specific centiles yielded virtually the same classification for SGA, AGA, and LGA. Preterm birth was defined as 37 weeks of gestation. Table 4. Associations of Pregestational Diabetes Mellitus, Gestational Diabetes Mellitus, and Glycosuria With the Predicted 10 Years CVD Risk Based on the Framingham Score No Glycosuria or Diabetes Mellitus (n 3061) Glycosuria (n 99) Gestational Diabetes Mellitus (n 17) Pregestational Diabetes Mellitus (n 10) OR for predicted CVD event in next 10 y based on Framingham score (n 2172) Mean score, % (SE) 3.69 (0.06) 4.11 (0.33) 5.52 (0.75) 5.37 (0.99) Model (0.97, 1.29) 1.43 (1.04, 1.96) 1.56 (1.03, 2.37) Model (0.97, 1.25) 1.26 (0.95, 1.68) 1.56 (1.07, 2.26) Model (0.98, 1.25) 1.23 (0.93, 1.25) 1.55 (1.07, 2.24) CVD indicates cardiovascular disease; OR, odds ratio; CI, confidence interval; model 1, unadjusted; model 2, additional adjustment for prepregnancy body mass index, education (yes/no university level), parity, and smoking during pregnancy; and model 3, additional adjustment for hypertensive disorders of pregnancy, preterm delivery, and size for gestational age. These ratios are interpreted as percentage (relative) differences; for example, for model 3 CVD events, pregestational diabetes mellitus is associated with a 55% increase in the predicted odds of an event.

6 1372 Circulation March 20, 2012 Table 5. Multivariable Associations of HDP With Cardiovascular Risk Factors Measured 18 Years After the Index Pregnancy Gestational Hypertension (n 438) Mean Difference (95% CI) Preeclampsia (n 61) Mean Difference (95% CI) No HDP (n 2688) Risk Factor BMI, kg/m 2 *(n 3364) Age-adjusted mean (SE) (0.95) (0.96) (1.13) Model (1.84, 2.82) 4.26 (3.07, 5.46) Model (0.24, 0.91) 1.01 (0.20, 1.82) Model (0.25, 0.92) 1.09 (0.28, 1.90) Waist circumference, cm (n 3358) Age-adjusted mean (SE) (2.24) (2.27) (2.66) Model (3.97, 6.27) 8.74 (5.90, 11.58) Model (0.61, 2.38) 1.86 ( 0.30, 4.02) Model (0.63, 2.40) 1.91 ( 0.26, 4.08) SBP (mm Hg)* (n 3364) Age-adjusted mean (SE) (2.25) (2.29) (2.68) Model (8.23, 10.55) (7.40, 13.09) Model (7.17, 9.49) 8.27 (5.45, 11.09) Model (7.15, 9.47) 8.31 (5.48, 11.14) DBP (mm Hg)* (n 3364) Age-adjusted mean (SE) (1.50) (1.52) (1.78) Model (4.35, 5.89) 6.50 (4.61, 8.39) Model (3.71, 5.25) 5.29 (3.41, 7.16) Model (3.71, 5.26) 5.47 (3.59, 7.36) Glucose, mmol/l Age-adjusted mean (SE) 4.92 (0.19) 5.08 (0.19) 5.43 (0.23) Model (0.06, 0.26) 0.51 (0.26, 0.75) Model ( 0.02, 0.18) 0.37 (0.13, 0.62) Model ( 0.03, 0.15) 0.31 (0.09, 0.53) HDL, mmol/l Age-adjusted mean (SE) 0.83 (0.07) 0.74 (0.07) 0.70 (0.09) Model ( 0.12, 0.05) 0.12 ( 0.22, 0.03) Model ( 0.08, 0.01) 0.05 ( 0.14, 0.04) Model ( 0.09, 0.01) 0.05 ( 0.14, 0.04) LDL, mmol/l Age-adjusted mean (SE) 1.33 (0.15) 1.36 (0.16) 1.42 (0.18) Model ( 0.05, 0.11) 0.10 ( 0.10, 0.29) Model ( 0.06, 0.10) 0.07 ( 0.13, 0.27) Model ( 0.06, 0.10) 0.08 ( 0.12, 0.27) HDP indicates hypertensive disorders of pregnancy; CI, confidence interval; BMI, body mass index; model 1, adjustment for age at measurement; model 2, additional adjustment for prepregnancy BMI, education (yes/no university level), parity, and smoking during pregnancy; model 3, additional adjustment for pregnancy diabetes mellitus; SBP, systolic blood pressure; DBP, diastolic blood pressure; HDL, high-density lipoprotein; and LDL, low-density lipoprotein. *n for HDP, gestational hypertension, and preeclampsia 3044, 504, and 76, respectively. Mothers Follow-Up Assessment A total of 4376 women attended the follow-up clinic. Mean follow-up time was 18 years (range, years). Weight and height were measured with the subjects in light clothing and without shoes. Weight was measured to the nearest 0.1 kg with the use of Tanita scales. Height was measured to the nearest 0.1 cm with a Harpenden stadiometer. Waist circumference was measured twice to the nearest 1 mm at the midpoint between the lower ribs and the pelvic bone with a flexible tape. The mean of the 2 measures is used here. Blood pressure was measured while the women were lying down with the use of an Omron M6 monitor (Omron Healthcare UK Ltd, Milton Keynes, UK). Two

7 Fraser et al Pregnancy Complications and CVD Risk 1373 Table 6. Multivariable Associations of HDP With Cardiovascular Risk Factors Measured 18 Years After the Index Pregnancy readings of systolic and diastolic blood pressure were recorded on each arm, and the mean of these 4 readings was used here. Women were asked to fast for at least 8 hours before attending the clinic. Blood samples were obtained, centrifuged, separated, and frozen at 80 C within 30 minutes. Plasma glucose was measured by automated enzymatic (hexokinase) method with coefficient of variation of 3%. Plasma insulin was measured by an enzyme-linked immunosorbent assay (Mercodia, Uppsala, Sweden) that does not cross-react with proinsulin, and proinsulin was also measured by an enzyme-linked immunosorbent assay (Mercodia) that is a solid-phase 2-site enzyme immunoassay for the quantification of human proinsulin. The coefficients of variation are as follows: within assay, 3.2%; between assay, 5.2%; and total assay, 6.1%. There is no cross-reactivity with insulin or C-peptide. Lipids were measured by automated analyzer with enzymatic methods. C-reactive protein was measured by automated particle-enhanced immunoturbidimetric assay (Roche UK, Welwyn Garden City, UK). Other Variables Maternal age at delivery and parity were obtained from obstetric records. Information on prepregnancy weight and height, maternal smoking in pregnancy, maternal education, and household social class was based on questionnaire responses. Maternal education was categorized as below or above university level. The highest parental occupation was used to allocate the children to family social class groups (classes I [professional/managerial] to No HDP (n 2688) Gestational Hypertension (n 438) Preeclampsia (n 61) Insulin, mu/l Age-adjusted geometric mean (95% CI) 6.51 (5.19, 8.16) 7.68 (6.10, 9.66) 8.57 (6.55, 11.23) Model (1.11, 1.25) 1.32 (1.13, 1.53) Model (1.02, 1.15) 1.12 (0.97, 1.29) Model (1.02, 1.15) 1.12 (0.97, 1.29) Proinsulin, pmol/l Age-adjusted geometric mean (95% CI) 4.47 (3.64, 5.48) 5.16 (4.20, 6.35) 5.93 (4.65, 7.56) Model (1.09, 1.22) 1.33 (1.16, 1.52) Model (1.01, 1.12) 1.15 (1.01, 1.30) Model (1.01, 1.12) 1.13 (1.00, 1.29) Triglycerides, mmol/l Age-adjusted geometric mean (95% CI) 0.63 (0.54, 0.74) 0.68 (0.58, 0.81) 0.70 (0.58, 0.84) Model (1.04, 1.14) 1.11 (1.00, 1.23) Model (1.01, 1.10) 1.04 (0.94, 1.16) Model (1.01, 1.10) 1.05 (0.94, 1.16) C-reactive protein, mg/l Age-adjusted geometric mean (95% CI) 1.14 (0.74, 1.74) 1.43 (0.93, 2.22) 1.90 (1.14, 3.16) Model (1.13, 1.41) 1.67 (1.26, 2.21) Model (0.96, 1.19) 1.21 (0.93, 1.58) Model (0.96, 1.19) 1.25 (0.96, 1.64) HDP indicates hypertensive disorders of pregnancy; CI, confidence interval; model 1, adjustment for age at measurement; model 2, additional adjustment for prepregnancy body mass index, education (yes/no university level), parity, and smoking during pregnancy; and model 3, additional adjustment for pregnancy diabetes mellitus. These ratios are interpreted as percentage (relative) differences; for example, the results from model 3 for proinsulin show a 7% and 13% increase in those with gestational hypertension and preeclampsia compared with those with no HDP, respectively. V [unskilled manual workers], according to the 1991 British Office of Population and Census Statistics classification). Maternal smoking in pregnancy was categorized as follows: never smoked; smoked before pregnancy or in the first trimester and then stopped; and smoked throughout pregnancy. Information on diabetes mellitus and CVD diagnosed during follow-up was collected by a questionnaire completed 18 years after the index pregnancy. Women reported having been told they had a heart attack, heart failure, angina, and/or stroke. CVD Framingham Risk Score The 10-year CVD Framingham risk score (ie, the risk of a CVD event, expressed as a percentage) was calculated with information on age, total cholesterol, high-density lipoprotein (HDL) cholesterol, systolic blood pressure, diabetes mellitus (based on fasting glucose), treatment for hypertension obtained at the follow-up assessment, and smoking status reported on a questionnaire completed 18 years after pregnancy with the use of the equations given on the study Web site. 19,20 In a modified version of the score, we removed diabetes mellitus from the equation. This was done to examine whether the increased risk of developing diabetes mellitus, particularly in women with GDM, was the main driver of results. The correlation of this version with the main risk score was Statistical Analysis Values of insulin, proinsulin, triglycerides, and C-reactive protein were log-transformed to make the data follow normal distribu-

8 1374 Circulation March 20, 2012 Table 7. Associations of HDP With the Predicted 10 Years CVD Risk Based on the Framingham Score No HDP (n 2688) Gestational Hypertension (n 438) Preeclampsia (n 61) OR for predicted CVD event in next 10 years based on Framingham score (n 2172) Mean score, % (SE) 3.55 (0.06) 4.58 (0.15) 5.09 (0.41) Model (1.21, 1.39) 1.42 (1.19, 1.69) Model (1.19, 1.35) 1.31 (1.11, 1.53) Model (1.19, 1.35) 1.30 (1.11, 1.52) CVD indicates cardiovascular disease; HDP, hypertensive disorders of pregnancy; OR, odds ratio; CI, confidence interval; model 1, unadjusted; model 2, additional adjustment for prepregnancy body mass index, education (yes/no university level), parity, and smoking during pregnancy; and model 3, additional adjustment for pregnancy diabetes mellitus. These ratios are interpreted as percentage (relative) differences; for example, model 3 results for CVD events would be interpreted as 27% and 30% increases in the predicted odds of an event. tions. Linear regression models were used to assess associations of diabetes mellitus/glycosuria, HDP, size for gestational age, and preterm delivery with outcomes. In the basic model (model 1), we adjusted for maternal age at follow-up. In the fully adjusted model (model 2), we also included prepregnancy body mass index (BMI), maternal education, parity, and smoking during pregnancy. In model 3, we also mutually adjusted each exposure of interest for the remaining 3 pregnancy complications to examine whether associations remained (except for models with HDP as the exposure, in which we only adjusted for diabetes mellitus/ glycosuria). Regression coefficients are mean differences in the outcome in each category of exposure compared with the reference category that has a null value of 0. When the outcomes were log-transformed (insulin, proinsulin, triglycerides, and C-reactive protein), coefficients were back-transformed (exponentiated) to obtain ratios of geometric means for each exposure category in relation to the reference category that has the null value of 1, as were the values of the 95% confidence intervals. We derived the odds of CVD by dividing the Framingham risk score by (1 score). We then used the natural log of the odds in multivariable linear regression models and back-transformed (exponentiated) coefficients to obtain the odds ratio of having a CVD event over 10 years, with a null value of 1 in the reference category. We examined whether replacing maternal education with household social class or whether using the modified score without diabetes mellitus made a difference to results. Finally, we repeated analyses excluding 31 women who reported having CVD on the follow-up questionnaire and restricting analyses to include only nulliparous women (n 1614). Dr Fraser had full access to and takes full responsibility for the integrity of the data. All authors have read and agree to the manuscript as written. Results Of the 4376 women who attended the clinic (the eligible cohort), 3364 (for BMI and blood pressure) and 3224 (blood measures) were included in analyses. Characteristics of women included in this study compared with women in the eligible cohort who were excluded from analyses because of missing data are presented in Table 1. Women included in the analyses were slightly older at the birth of the index child and were less likely to belong to a manual social class and to have a preterm birth. They were less likely to smoke and had lower mean BMI, waist circumference, insulin, proinsulin, and triglyceride levels at follow-up. No other differences were found. In the whole ALSPAC cohort (n ), 45% of women had at least 1 pregnancy-related complication. Of the eligible cohort (n 1663), 38% had at least 1 pregnancy-related complication, and of the women included in analyses, 35.8% (n 1204) had at least 1 pregnancy-related complication. A total of 1002 women (29.8%) had 1 complication, 175 (5.2%) had 2, 26 (0.8%) had 3, and 1 woman had all 4. The prevalence of each pregnancy-related complication is given in Table 1, and Figure 1 shows the overlap between pregnancy-related complications. Four women with GDM also had preeclampsia, and an additional 4 had gestational hypertension. In women with pregestational diabetes mellitus, equivalent numbers were 4 and 1, and for women with glycosuria, equivalent numbers were 17 and 2. All 13 women with pregestational diabetes mellitus were diagnosed before the age of 29 years, and 2 were treated by diet only. Thirty-one women (1.3%) were diagnosed with CVD during follow-up. Overall, in the cohort, future risk of CVD based on the 10-year Framingham score was low (as would be expected given that the mean age of participants is 48 years), with a median predicted risk of 3.0% (interquartile range, 2.2% to 4.2%), with some evidence that this varied by pregnancy-related complications (Figure 2). Associations of pregnancy diabetes mellitus with cardiovascular risk factors are presented in Tables 2 through 4. Women with glycosuria and GDM had a greater mean BMI and waist circumference 18 years after the index pregnancy compared with women without pregnancy diabetes mellitus in the basic age-adjusted model (model 1), but associations were attenuated to the null in the confounder-adjusted model (model 2). Glycosuria and GDM were also associated with higher levels of insulin, proinsulin, and triglycerides in model 1. In model 2, associations for insulin and proinsulin persisted, although confidence intervals for GDM included the null. Women with pregestational diabetes mellitus, GDM, and glycosuria all had higher fasting glucose levels compared with women without pregnancy diabetes mellitus. The greatest mean difference was in women with pregestational diabetes mellitus (mean difference compared with women without pregnancy diabetes mellitus: 6.82 mmol/l), then in women with GDM (2.35 mmol/l), and then in women with glycosuria (0.28 mmol/l), when adjusted for potential confounders (model 2). Women with pregestational diabetes mellitus also had lower low-density lipoprotein cholesterol compared with women without pregnancy diabetes mellitus. No evidence of associations of pregnancy diabetes mellitus with blood pressure or HDL cholesterol was found. There was evidence of increased calculated risk of a CVD event over 10 years in women with glycosuria, GDM, and pregestational diabetes mellitus in model 1; on adjustment for potential confounders, the odds ratio was 1.10 in women with glycosuria, 1.26 in women with GDM, and 1.56 in women with pregestational diabetes mellitus, with confidence inter-

9 Fraser et al Pregnancy Complications and CVD Risk 1375 Table 8. Multivariable Associations of Size for Gestational Age With Cardiovascular Risk Factors Measured 18 Years After the Index Pregnancy Risk Factor SGA (n 262) Mean Difference (95% CI) AGA (n 2630) LGA (n 332) Mean Difference (95% CI) BMI, kg/m 2 *(n 3364) Age-adjusted mean (SE) (0.99) (0.96) (1.01) Model ( 0.97, 0.29) (1.45, 2.60) Model ( 0.50, 0.34) ( 0.15, 0.62) Model ( 0.59, 0.25) ( 0.12, 0.65) Waist circumference, cm (n 3358) Age-adjusted mean (SE) (2.32) (2.25) (2.36) Model ( 2.66, 0.29) (4.06, 6.74) Model ( 1.90, 0.33) (0.70, 2.74) Model ( 2.12, 0.12) (0.72, 2.76) SBP, mm Hg* (n 3364) Age-adjusted mean (SE) (2.42) (2.34) (2.46) Model (0.81, 3.88) ( 0.30, 2.50) Model (1.16, 4.18) ( 1.71, 1.05) Model (0.46, 3.41) ( 1.74, 0.95) DBP, mm Hg* (n 3364) Age-adjusted mean (SE) (1.59) (1.54) (1.61) Model (0.99, 3.01) ( 0.21, 1.62) Model (1.17, 3.16) ( 1.10, 0.71) Model (0.75, 2.70) ( 1.05, 0.74) Glucose, mmol/l Age-adjusted mean (SE) 4.93 (0.20) 5.00 (0.19) 5.23 (0.20) Model ( 0.19, 0.06) (0.12, 0.35) Model ( 0.20, 0.05) (0.04, 0.27) Model ( 0.17, 0.05) ( 0.05, 0.15) HDL, mmol/l Age-adjusted mean (SE) 0.82 (0.08) 0.79 (0.07) 0.73 (0.08) Model ( 0.02, 0.08) ( 0.11, 0.02) Model ( 0.01, 0.08) ( 0.06, 0.03) Model ( 0.01, 0.09) ( 0.06, 0.03) LDL (mmol/l) Age-adjusted mean (SE) 1.32 (0.16) 1.35 (0.15) 1.42 (0.16) Model ( 0.13, 0.07) ( 0.02, 0.17) Model ( 0.14, 0.06) ( 0.06, 0.03) Model ( 0.15, 0.05) ( 0.04, 0.14) SGA indicates small for gestational age; AGA, appropriate for gestational age; LGA, large for gestational age; CI, confidence interval; BMI, body mass index; model 1, adjustment for age at measurement; model 2, additional adjustment for prepregnancy BMI, education (yes/no university level), parity, and smoking during pregnancy; model 3, additional adjustment for pregnancy diabetes mellitus, hypertensive disorders of pregnancy, and preterm delivery; SBP, systolic blood pressure; DBP, diastolic blood pressure; HDL, high-density lipoprotein; and LDL, low-density lipoprotein. *SGA, n 274; AGA, n 2751; LGA, n 339. vals for glycosuria and GDM spanning the null. Overall, adjusting for HDP, preterm birth, and size for gestational age (model 3) did not substantially change results. Associations of HDP with cardiovascular risk factors are presented in Tables 5 through 7. Both gestational hypertension and preeclampsia were associated with greater BMI, waist, systolic and diastolic blood pressure, insulin, proinsulin, and triglycerides and lower HDL cholesterol in both the basic and confounder-adjusted models. In the confounder-adjusted models, point estimates for the asso-

10 1376 Circulation March 20, 2012 Table 9. Multivariable Associations of Size for Gestational Age With Cardiovascular Risk Factors Measured 18 Years After the Index Pregnancy ciations of gestational hypertension and preeclampsia with HDL cholesterol, insulin, and triglycerides were of similar magnitude, but confidence intervals for preeclampsia (but not for gestational hypertension) included the null (because of the small number of women with preeclampsia). Preeclampsia was also associated with higher glucose and C-reactive protein, whereas associations for gestational hypertension were weaker. No associations with lowdensity lipoprotein cholesterol were noted. The calculated risk of a CVD event over 10 years was elevated in women with HDP compared with those without (models 1 and 2). Adjustment for pregnancy diabetes mellitus did not alter results (model 3). Associations of size for gestational age with outcomes are presented in Tables 8 through 10. Mothers of LGA babies had higher mean BMI, waist circumference, glucose, insulin, proinsulin, triglycerides, and C-reactive protein and lower HDL cholesterol levels than women with an AGA baby in the age-adjusted model (model 1). When adjusted for confounders, only the associations with waist circumference and glucose remained. The association with glucose was attenuated in model 3 after adjustment for other pregnancy-related SGA (n 262) AGA (n 2630) LGA (n 332) Insulin, mu/l Age-adjusted geometric mean (SE) 6.73 (5.35, 8.1) 6.88 (5.48, 8.64) 7.58 (5.97, 9.62) Model (0.91, 1.06) (1.03, 1.18) Model (0.92, 1.06) (0.93, 1.06) Model (0.91, 1.06) (0.93, 1.06) Proinsulin, pmol/l Age-adjusted geometric mean (SE) 4.68 (3.79, 5.78) 4.68 (3.82, 5.75) 5.27 (4.25, 6.53) Model (0.93, 1.07) (1.06, 1.20) Model (0.94, 1.07) (0.96, 1.08) Model (0.93, 1.06) (0.96, 1.08) Triglycerides, mmol/l Age-adjusted geometric mean (SE) 0.63 (0.53, 0.74) 0.65 (0.55, 0.76) 0.71 (0.60, 0.84) Model (0.92, 1.02) (1.04, 1.15) Model (0.92, 1.02) (0.99, 1.09) Model (0.91, 1.01) (0.99, 1.09) C-reactive protein, mg/l Age-adjusted geometric mean (SE) 1.31 (0.84, 2.04) 1.31 (0.84, 2.04) 1.41 (0.90, 2.22) Model (0.94, 1.26) (1.03, 1.34) Model (0.96, 1.26) (0.85, 1.08) Model (0.95, 1.25) (0.85, 1.09) SGA indicates small for gestational age; AGA, appropriate for gestational age; LGA, large for gestational age; CI, confidence interval; model 1, adjustment for age at measurement; model 2, additional adjustment for prepregnancy body mass index, education (yes/no university level), parity, and smoking during pregnancy; and model 3, additional adjustment for pregnancy diabetes mellitus, hypertensive disorders of pregnancy, and preterm delivery. These ratios are interpreted as percentage (relative) differences; for example, the results from model 3 for proinsulin show a 2% increase in women who delivered an LGA baby compared with those who delivered an AGA baby (although CIs include the null value). complications and was due to the adjustment for pregnancy diabetes mellitus (results not shown). Women with an SGA baby had higher systolic and diastolic blood pressure compared with women with an AGA baby in both models 1 and 2. Adjustment for pregnancy diabetes mellitus, HDP, and preterm birth did not alter results (model 3). An initial association of LGA with the calculated 10-year CVD risk was attenuated to the null when adjusted for confounders, whereas SGA was associated with a 10% increased risk in model 2. Preterm birth was associated with systolic blood pressure and more weakly with diastolic blood pressure, but these associations were attenuated on adjustment for other pregnancy-related complications. Specifically, the attenuation was due to the adjustment for HDP (results not shown). When the analysis was restricted to women without HDP, there was no longer any evidence of an association between preterm birth and systolic or diastolic blood pressure (results not shown). There was no strong evidence of associations of preterm birth with any of the other cardiovascular risk outcomes or with the calculated 10-year CVD risk (Tables 11 through 13).

11 Fraser et al Pregnancy Complications and CVD Risk 1377 Table 10. Associations of Size for Gestational Age With the Predicted 10 Years CVD Risk Based on the Framingham Score SGA (n 262) AGA (n 2630) LGA (n 332) OR for predicted CVD event in next 10 y based on Framingham score (n 2172) Mean score, % (SE) 3.94 (0.20) 3.65 (0.06) 4.12 (0.18) Model (0.96, 1.15) (1.04, 1.21) Model (1.01, 1.19) (0.93, 1.08) Model (0.99, 1.16) (0.92, 1.06) CVD indicates cardiovascular disease; SGA, small for gestational age; AGA, appropriate for gestational age; LGA, large for gestational age; OR, odds ratio; CI, confidence interval; model 1, unadjusted; model 2, additional adjustment for prepregnancy body mass index, education (yes/no university level), parity, and smoking during pregnancy; and model 3, additional adjustment for pregnancy diabetes mellitus, hypertensive disorders of pregnancy, and preterm delivery. These ratios are interpreted as percentage (relative) differences. Additional Analyses When the modified Framingham score without diabetes mellitus was used, the odds ratio of CVD in women with GDM compared with women without pregnancy diabetes mellitus was 0.88 (95% confidence interval, ; model 3), suggesting that the increased risk of diabetes mellitus after GDM is the main driver of their increased risk of CVD. In women with glycosuria, the odds ratio was 1.10 (95% confidence interval, ). In comparison, results for other pregnancy-related complications were unchanged. Results were unchanged from those presented in the tables when 31 women who reported having been diagnosed with CVD during follow-up were excluded from analyses. Results were also comparable to those presented when analyses were restricted to the 1614 nulliparous women in our sample (results available from authors on request). Replacing maternal education with household social class also did not alter results. Discussion In this general population of women, more than one third of women experienced at least 1 pregnancy-related complication. Pregestational diabetes mellitus, GDM, and glycosuria were all associated with higher glucose concentrations 18 years after pregnancy even when we controlled for potential confounders, including prepregnancy BMI. Pregnancy diabetes mellitus was associated with higher glucose, GDM and glycosuria were also associated with higher insulin and proinsulin, and glycosuria was associated with higher triglyceride levels. Similarly, mothers of LGA babies had higher glucose levels than mothers of AGA babies, in agreement with the established linear association of maternal glycemic status and the risk of delivering an LGA infant. 21 In comparison, both gestational hypertension and preeclampsia were associated with a greater number of Table 11. Multivariable Associations of Preterm Birth With Cardiovascular Risk Factors Measured 18 Years After the Index Pregnancy Preterm Risk Factor Not Preterm (n 3085) (n 139) Mean Difference (95% CI) BMI, kg/m 2 *(n 3364) Age-adjusted mean (SE) (0.97) (1.04) Model ( 1.21, 0.49) Model ( 0.60, 0.53) Model ( 0.71, 0.42) Waist circumference, cm (n 3358) Age-adjusted mean (SE) (2.27) (2.44) Model ( 3.43, 0.57) Model ( 2.25, 0.72) Model ( 2.51, 0.48) SBP, mm Hg* (n 3364) Age-adjusted mean (SE) (2.34) (2.52) Model ( 0.10, 4.03) Model (0.20, 4.22) Model ( 0.42, 3.52) DBP, mm Hg* (n 3364) Age-adjusted mean (SE) (1.54) (1.66) Model ( 0.23, 2.48) Model ( 0.07, 2.58) Model ( 0.46, 2.16) Glucose, mmol/l Age-adjusted mean (SE) 4.95 (0.19) 5.10 (0.20) Model ( 0.02, 0.31) Model ( 0.01, 0.32) Model ( 0.03, 0.27) HDL, mmol/l Age-adjusted mean (SE) 0.80 (0.07) 0.86 (0.08) Model (0.004, 0.13) Model (0.002, 0.13) Model (0.01, 0.14) LDL, mmol/l Age-adjusted mean (SE) 1.34 (0.15) 1.27 (0.16) Model ( 0.21, 0.06) Model ( 0.21, 0.06) Model ( 0.23, 0.05) CI indicates confidence interval; BMI, body mass index; model 1, adjustment for age at measurement; model 2, additional adjustment for prepregnancy BMI, education (yes/no university level), parity, and smoking during pregnancy; model 3, additional adjustment for pregnancy diabetes mellitus, hypertensive disorders of pregnancy, and size for gestational age; SBP, systolic blood pressure; DBP, diastolic blood pressure; HDL, high-density lipoprotein; and LDL, low-density lipoprotein. *Women with term and preterm deliveries: n 3219 and n 145, respectively.

12 1378 Circulation March 20, 2012 Table 12. Multivariable Associations of Preterm Birth With Cardiovascular Risk Factors Measured 18 Years After the Index Pregnancy Not Preterm (n 3085) Preterm (n 139) Insulin, mu/l Age-adjusted mean (SE) 6.79 (5.41, 8.52) 6.93 (5.43, 8.85) Model (0.92, 1.13) Model (0.94, 1.14) Model (0.93, 1.13) Proinsulin, pmol/l Age-adjusted mean (SE) 4.64 (3.78, 5.70) 4.72 (3.78, 5.88) Model (0.93, 1.11) Model (0.94, 1.12) Model (0.93, 1.11) Triglycerides, mmol/l Age-adjusted mean (SE) 0.64 (0.55, 0.76) 0.64 (0.54, 0.76) Model (0.92, 1.07) Model (0.93, 1.07) Model (0.91, 1.05) C-reactive protein, mg/l Age-adjusted mean (SE) 1.21 (0.79, 1.87) 1.19 (0.75, 1.89) Model (0.81, 1.19) Model (0.83, 1.18) Model (0.78, 1.13) CI indicates confidence interval; model 1, adjustment for age at measurement; model 2, additional adjustment for prepregnancy body mass index, education (yes/no university level), parity, and smoking during pregnancy; and model 3, additional adjustment for pregnancy diabetes mellitus, hypertensive disorders of pregnancy, and size for gestational age. These ratios are interpreted as percentage (relative) differences; for example, the results from model 3 for proinsulin show a 1% increase in women with a preterm delivery (although CIs include the null value). Table 13. Associations of Preterm Delivery With the Predicted 10 Years CVD Risk Based on the Framingham Score Not Preterm (n 3085) Preterm (n 139) OR for predicted CVD event in next 10 y based on Framingham score (n 2172) Mean score, % (SE) 3.72 (0.06) 3.72 (0.26) Model (0.86, 1.07) Model (0.90, 1.10) Model (0.89, 1.09) CVD indicates cardiovascular disease; OR, odds ratio; CI, confidence interval; model 1, unadjusted; model 2, additional adjustment for prepregnancy body mass index, education (yes/no university level), parity, and smoking during pregnancy; and model 3, additional adjustment for pregnancy diabetes mellitus, hypertensive disorders of pregnancy, and size for gestational age. These ratios are interpreted as percentage (relative) differences. cardiovascular risk factors: BMI, waist circumference, systolic and diastolic blood pressure, insulin, proinsulin, triglycerides, and HDL cholesterol. Mothers of SGA babies had higher systolic and diastolic blood pressure compared with mothers of AGA babies, as did mothers who delivered before term, which is in agreement with intrauterine growth restriction and preterm delivery resulting in part from placental insufficiency due to hypertension. No other associations between preterm delivery and cardiovascular risk factors were found in our study. Although others have reported associations between preterm birth and CVD, 22,23 in agreement with the present study, this primarily reflected elective preterm birth due to placental dysfunction rather than spontaneous preterm birth. It was notable that the positive association of preeclampsia with future blood pressure was similar to that of gestational hypertension with future blood pressure, but, in general, for other risk factors the magnitudes of association appeared stronger for preeclampsia than for gestational hypertension. Furthermore, our results suggest that preeclampsia could be a stronger marker of future CVD than GDM because it is associated with a greater number of cardiovascular risk factors, whereas GDM is linked to greater glycemia later in life, as would be expected. Our results also suggest that the mechanisms underlying these associations are different because pregnancy diabetes mellitus and HDP were associated with different cardiovascular risk factors. GDM and LGA were associated primarily with fasting glucose and insulin, whereas HDP were associated with blood pressure, lipids, and insulin but not glucose. Moreover, the increased calculated risk of CVD in women with GDM is explained by their increased risk of developing diabetes mellitus, whereas this was not the case for women with HDP. The fact that associations of both pregnancy diabetes mellitus and HDP with cardiovascular risk factors and the calculated risk of CVD were not attenuated in models in which there was mutual adjustment for them also suggests largely independent pathways of risk. However, the prevalence of these 2 exposures (pregnancy diabetes mellitus and HDP) was low. Limitations and Strengths In our study, women are still not at an age in which CVD events are common. Therefore, we were not able to directly assess the value of adding information on pregnancy-related complications to existing risk scores. Moreover, the number of women experiencing 1 pregnancy complication was limited. Larger studies with longer follow-up could be useful in determining whether HDP, pregnancy diabetes mellitus, or both could be valuable additions to CVD risk prediction scores for women and whether there are interactions between the different complications in their associations with CVD risk. Additional limitations include the lack of information on maternal metabolic control during pregnancy; our lack of knowledge about exactly how GDM was diagnosed; and