Parity, coronary heart disease and mortality in the old order Amish

Atherosclerosis 254 (2016) 14e19

Contents lists available at ScienceDirect

Atherosclerosis journal homepage: www.elsevier.com/locate/atherosclerosis

Parity, coronary heart disease and mortality in the old order Amish Tarec K. Elajami a, 1, Jennifer Giuseffi b, Maria D. Avila a, Ninel Hovnanians a, Kenneth J. Mukamal b, Nisha Parikh a, Francine K. Welty a, * a b

Cardiovascular Division, Department of Medicine, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, USA Department of Medicine, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 26 June 2016 Received in revised form 18 August 2016 Accepted 7 September 2016 Available online 9 September 2016

Background and aims: Prior data on the association between parity and mortality are limited by the presence of sociodemographic confounders including cultural norms of parity. Our objective was to determine the association between parity and mortality in the Amish, a socioeconomically homogenous group with large numbers of children per family. Methods: We conducted a population-based cohort study among 518 Old Order Amish women enrolled in a cardiovascular awareness program. The mean length of follow-up for mortality was 13.52 years. We determined the adjusted associations between parity and obesity, prevalent coronary heart disease and mortality. Results: The mean number of total births per woman was 6.7 ± 3.6 with a mode of 8. No significant association was observed between parity and all-cause mortality when adjusted for age (HR 1.00 per additional birth; 95% CI 0.96e1.05; p ¼ 0.85) or in multivariate analysis (HR 1.00, 95% CI 0.95e1.05; p ¼ 0.95). There was also no association of parity in age- or multivariable adjusted models with prevalent diabetes, hypertension or coronary heart disease. Despite the lack of effect of parity on mortality, a significant association of ten or more births was observed with higher body mass index (BMI) compared to the referent group of 8e9 total births. Conclusions: In a highly homogeneous population with high rates of parity, no association between overall mortality and parity was observed. Ten or more births were significantly associated with a higher BMI but not with overall mortality. © 2016 Elsevier Ireland Ltd. All rights reserved.

Keywords: Mortality Pregnancy Cardiovascular outcomes Coronary heart disease Women

1. Introduction Pregnancy is associated with several physiologic changes that are beneficial to both mother and fetus. However, some of the physiologic changes of pregnancy, including insulin resistance and dyslipidemia, are established cardiovascular disease (CVD) risk factors [1]. Additional changes that occur during pregnancy include alterations to cardiovascular hemodynamic function, hemostasis and thrombophilia. Whether or not these changes are cumulative over successive pregnancies is not known. Prior studies that have sought to determine whether increased parity is associated with CVD risk or mortality have yielded

* Corresponding author. Cardiovascular Division, Beth Israel Deaconess Medical Center, 330 Brookline Ave, SL-423, Boston, MA, 02215, USA. E-mail address: [email protected] (F.K. Welty). 1 Present address: Mount Sinai Medical Center, 4300 Alton Rd. Miami Beach, FL, 33140, USA. http://dx.doi.org/10.1016/j.atherosclerosis.2016.09.007 0021-9150/© 2016 Elsevier Ireland Ltd. All rights reserved.

conflicting results. Some studies suggest an increased risk of CVD with increasing parity at greater than or equal to five pregnancies and with nulliparity [2e4] while others have found no significant association between number of pregnancies and CVD [5,6]. Most of the existing data regarding the association between parity and mortality is equally inconsistent with a suggestion of increased risk of vascular death with grand-grand multiparity in one study [7] but reduced risk of cardiovascular death in another study [8]. The largest study showed an increased risk of CVD death with low education status and decreased risk with high education status [9]. This variation in results in these different studies may be due to the specific study designs and the populations studied. Although sociodemographic confounders were adjusted for with multivariable statistical models in these previous studies, residual confounding remains a serious potential problem. Specific confounders thought to influence the association between parity and CVD include socioeconomic status, access to health care, diet and tobacco abuse. The inability to fully account for socioeconomic

T.K. Elajami et al. / Atherosclerosis 254 (2016) 14e19

confounders has been a significant limitation of prior studies in this area as has the relative rarity of grand-multiparity in most cohort studies [5]. The relative rarity of grand-multiparity in most populations also severely hampers previous studies of this topic. Our aim was to study the association of parity with mortality within a uniform socioeconomic group with large family sizes to better exclude potential confounding and determine the doseresponse relationship. The Old Order Amish are an exclusive conservative religious group living in rural North America; they follow a strictly defined way of life characterized by strong religious practices, deliberate avoidance of modern technology, a rural, largely farming lifestyle involving vigorous exercise and avoidance of tobacco and alcohol [10,11]. They also have the same educational level. Because of the Amish's uniform socioeconomic and occupational characteristics, absence of additional confounders and large family sizes, they represent an ideal population to extend the prior epidemiologic data on the association between increased parity and mortality. Our a priori hypothesis was that the association between parity and mortality would be attenuated in this population with less confounding of parity by other factors. 2. Patients and methods 2.1. Study population In 1987, 1692 members of the Old Order Amish in central Ohio participated in a CVD awareness community outreach program. The study was approved by the Medical Center of Delaware and Beth Israel Deaconess Medical Center IRBs. Participants underwent lipid screening and height, weight and blood pressure measurements as previously described [12]. The subjects provided self-reported medical histories and medications, the latter used to confirm presence of hypertension and diabetes. Subjects were then followed prospectively for self-reported disease outcomes using selfreported mailed questionnaires and/or home visits. Written informed consent was obtained from each subject, and the study conformed to the ethical guidelines of the 1975 Declaration of Helsinki. We restricted our analyses to the subset of 588 women with available information on parity and a baseline visit and, for generalizability, free of the apolipoprotein (apo) B-67 mutation which causes very low levels of apo B and low-density lipoprotein cholesterol (LDL-C) [12]. 2.2. Parity ascertainment Our primary exposure was number of total births. We also collected information on number of singleton live births, number of stillbirths, maternal age at first birth, maternal age at last birth, and time to conceive from the time of marriage. This information was obtained from genealogic data published in the 1988 and 2005 editions of the Ohio Amish Directories [13,14]. Within the directory, family members are listed by head of household. Included is their date of birth, name of spouse, date of marriage and occupation, the couples' number of children, including stillbirths and their dates of birth. Based on the distribution of parity, we categorized number of births into 7 categories, from 0 to 13 or more, setting the median (8e9 births) as the referent given the distribution portrayed in Fig. 1. The other categories used to compare to the referent were determined as nulliparous, 1e3, 4e5, 6e7, 10e12 and greater than or equal to 13 total births. 2.3. Covariates Individuals reported the presence or absence of CVD including

15

Fig. 1. Distribution of total births among 518 Amish women.

coronary heart disease (CHD), congestive heart failure and cerebrovascular disease. Presence of CHD was defined as angina, myocardial infarction, coronary artery bypass graft surgery and percutaneous transluminal coronary angioplasty. Angina was ascertained with the Rose angina questionnaire. Cerebrovascular disease was defined as patients who have undergone carotid endarterectomy or self-reported stroke. The height and weight of each participant was recorded. The body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters. Lipids were measured from blood samples; approximately 80% of the subjects fasted for at least 10 h. Total cholesterol, LDL-C, high density lipoprotein cholesterol (HDL-C) and triglycerides were measured according to Lipid Research Clinic techniques [15] in the laboratory of Pomerene Memorial Hospital in Millersburg, Ohio; this laboratory participates in the standardization program of the Centers for Disease Control, Atlanta, GA.

2.4. Ascertainment of mortality Cause of death and all-cause mortality were obtained on all participants from the date of entry into the study until 2010 by using published obituaries. Cause of death was determined from medical records, death certificates and/or contact with the spouse or offspring (if the spouse was deceased). Concordance between death certificates, obituaries and family-report was 100%.

2.5. Statistical analysis We present the unadjusted prevalence of cardiovascular risk factors, CHD and demographic factors according to parity and test unadjusted differences with analysis of variance or c2 tests. Cox proportional hazards models were used to relate parity with allcause mortality during follow-up. Crude, age-adjusted and multivariable-adjusted models were constructed with adjustment for age at first birth and time to conceive from marriage; we excluded nulliparous women in these latter models. Parity was examined both in categories and as a continuous variable (in risk per additional birth); models that included parity as a quadratic term showed no evidence of a non-linear association. In secondary analyses, we compared parity to presence of diabetes, hypertension, CHD, BMI, total cholesterol, HDL-C, LDL-C and triglycerides in unadjusted, age-adjusted models and multivariable-adjusted models, with adjustment for age at first birth and time to conceive from marriage, using logistic regression for clinical diagnoses and linear regression for risk factors.

16

T.K. Elajami et al. / Atherosclerosis 254 (2016) 14e19

3. Results 3.1. Characteristics of the cohort A total of 518 women had complete information on parity and vital status. Of these, 204 women died during a mean time of follow up of 13.5 years (SD 5.6). Table 1 shows demographics, lipid levels and CVD risk factors. No participant smoked tobacco, a significant confounder in prior studies. The average number of total births was 6.7 (SD 3.6) with a mode of 8 and a broad range of 0e17 births (Fig. 1). Few women were nulliparous. Neither nulliparity nor increasing parity was associated with diabetes, hypertension or a composite CHD outcome (myocardial infarction, percutaneous transluminal coronary angioplasty or coronary artery bypass surgery) in this socioeconomically homogenous population (Table 1). There was also no association of parity in age- or multivariable adjusted models with prevalent diabetes, hypertension or CHD (multivariable-adjusted p trend ¼ 0.74 for diabetes; p ¼ 0.36 for hypertension; p ¼ 0.24 for CHD). 3.2. All-cause mortality according to parity In prospective analyses, we found no association between parity and mortality (Table 2). There was a weak association of higher parity with lower mortality in unadjusted analyses that was eliminated with adjustment for age. In Cox proportional hazard models, we observed no crude association of parity with mortality (HR per additional birth 0.97, p ¼ 0.10; 95% CI 0.93e1.01) and similar results when adjusted for age (HR 1.00, p ¼ 0.86; 95% CI 0.96e1.05) (Table 2). Addition of a quadratic term revealed no evidence of nonlinearity and no additional risk for increased parity (HR 1.00, p ¼ 0.96; 95% CI 0.96, 1.05). In analyses further adjusted for both age at first birth and time to conceive from marriage, we again found no association of parity with mortality (HR 1.00, p ¼ 0.95; 95% CI 0.95e1.05). 3.3. Age-adjusted association of parity with body mass index We examined associations of parity with putative mediators in secondary analyses. A significant positive association of parity with higher BMI (0.2 kg/m2 ± 0.1 per additional birth) was observed.

Compared to the referent of 8e9 total births, BMI was significantly higher with greater parity in age-adjusted analyses although individual differences were only significantly different than the referent for 10e12 and  13 total births (Table 3). There were no significant associations of triglyceride, LDL-C and HDL-C across categories of parity in unadjusted (Table 1) or age-adjusted analyses (p trend ¼ 0.14 for triglyceride; p ¼ 0.09 for HDL-C; p ¼ 0.84 for LDL-C). The cause of death was documented in 130 of 204 deaths. 52.14% of the documented causes of death were CVD-related including CHD, cerebrovascular disease or congestive heart failure. 4. Discussion In a population of Old Order Amish with high rates of parity and socioeconomic homogeneity, we found no association between parity and mortality in crude, age-adjusted or multivariate models. There was also no association of parity with CHD. Table 4 summarizes the prior studies which have reported on the incidence or prevalence of CHD and CVD, total mortality and mortality from CHD and CVD in women with 4 or more pregnancies. Both the Framingham Heart Study and National Health and Nutrition Examination Survey National Epidemiologic Follow-up Study observed significant increases in CHD in women with 6 or more pregnancies after adjustment for age and educational level [2] while others have found no significant association between number of pregnancies and CVD [5,6]. In Sweden, incident CVD was significantly higher in women with five or more pregnancies after adjustment for socioeconomic factors [3]. In a cohort study of 40,454 Israeli women followed for 37 years, the HR for total mortality was significantly higher in those women with 1 or 5 to 10 þ children compared to the referent of 2e4 after adjustment for multiple factors [16]. In contrast, in Australia, total mortality was decreased with 6 or more births, but there was no significant difference in CHD mortality [17]. Although women in Northern Finland with 10 or more births did not have a significant increase in CVD mortality, they did have a significant increase in mortality from hemorrhagic stroke (four-fold higher) compared to those with 2e4 births after adjustment for age, BMI, smoking, socioeconomic position, age at menarche and age at first birth [6]. In a second Finnish study, all-cause mortality was not significantly elevated in grand multiparous (5 or more deliveries) or grand-grand multiparous women (10 or more deliveries) [18]. This

Table 1 Characteristics of the cohort (n ¼ 518). 8e9

10e12

>13

Total parity

0

1e3

4e5

6e7

Number women

32

68

93

102

122

60

41

Singleton births Age (years) Age at 1st birth Age at last birth Total C (mg/dl) HDL-C (mg/dl) LDL-C (mg/dl) TG (mg/dl) Tobacco use (%) Diabetes (%) Hypertension (%) CHDb (%) BMI (kg/m2)

e 63.1 (13.6) e e 232 (47) 57 (16) 137 (31) 171 (115) 0 14.8 42.3 17.9 28.6 (7.4)

2.1 (0.86) 68.0 (12.4) 27.4 (5.5) 31.8 (5.9) 238 (47) 62 (21) 140 (46) 190 (151) 0 28.3 48.0 25.9 26.3 (6.9)

4.4 (0.68) 65.9 (13.7) 25.0 (3.6) 35.6 (4.3) 229 (44) 64 (18) 137 (38) 139 (85) 0 15.0 46.3 26.5 27.3 (6.2)

6.3 (0.7) 65.2 (12.1) 23.6 (2.8) 37.2 (4.7) 241 (52) 62 (23) 146 (44) 181 (118) 0 17.6 52.2 19.6 26.9 (5.7)

8.2 (0.8) 67.3 (10.6) 23.4 (2.5) 39.1 (3.7) 235 (41) 59 (15) 137 (35) 190 (128) 0 21.5 55.7 36.7 27.7 (6.0)

10.3 (0.8) 66.2 (10.6) 23.0 (2.2) 40.6 (2.8) 223 (45) 56 (15) 129 (39) 202 (188) 0 20.0 50.0 25.5 29.1 (6.1)

13.0 (1.9) 63.7 (10.3) 21.7 (1.7) 42.7 (3.4) 237 (40) 60 (15) 142 (39) 176 (120) 0 10.5 52.6 17.1 29.6 (5.3)

p heterogeneitya

0.29 <0.001 <0.001 0.24 0.19 0.29 0.06 0.37 0.84 0.09 0.17

Results shown are mean (SD) for continuous variables and % for binary variables. Information on diabetes was available on 451 women, hypertension on 444 and composite CHD on 456 women. BMI, body mass index; CHD, coronary heart disease; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; n, number; TG, triglyceride. a p-value for heterogeneity derived from analysis of variance for continuous variables and c2 tests for binary variables. b CHD is a composite endpoint of myocardial infarction, percutaneous transluminal coronary angioplasty or coronary artery bypass grafting.

T.K. Elajami et al. / Atherosclerosis 254 (2016) 14e19

17

Table 2 Hazard ratios and 95% confidence intervals for all-cause mortality according to parity. Total parity

0

1e3

4e5

6e7

8e9a

10e12

13

Number women

32

68

93

102

122

60

41

Number deaths Unadjusted HR 95% CI Age adjusted HR 95% CI Multivariate HRb 95% CI

11 0.93 (0.49,1.80) 1.17 (0.61.2.25) NA

32 1.22 (0.78,1.90) 0.91 (0.58,1.41) 0.99 (0.60,1.61)

38 1.03 (0.68,1.57) 0.84 (0.54,1.28) 0.84 (0.53,1.31)

40 0.94 (0.62,1.42) 1.11 (0.73,1.69) 1.10 (0.72,1.68)

50 e e e

21 0.87 (0.52,1.45) 0.91 (0.55,1.52) 0.86 (0.51,1.45)

12 0.65 (0.35,1.22) 1.05 (0.56,1.98) 0.95 (0.48,1.90)

Per additional birth

p

0.97 (0.93,1.01) 1.00 (0.96,1.05) 1.00 (0.95,1.05)

0.10 0.86 0.95

HR, hazard ratio; CI, confidence interval. a Referent group. b Adjusting for age at first birth and time to conceive.

Table 3 Age-adjusted association of parity with body mass index. Total parity

0

1e3

4e5

6e7

8e9

10e12

13

Number women

32

68

93

102

122

60

41

BMI (kg/m2)

28.2

26.6

27.4

26.8

27.8

29.2a

29.4b

Per additional birth

p

0.2 ± 0.1

0.04

BMI reported in 362 of 518 women. BMI, body mass index. a p ¼ 0.04 compared to referent of 8e9. b p ¼ 0.05 compared to referent of 8e9.

reflected the fact that mortality from diabetes and ischemic heart disease was significantly increased in these women, but mortality from breast, uterine, ovarian, bladder and lung cancer was significantly decreased. In contrast to the higher mortality in Scandinavian studies, CVD mortality was significantly lower in those with 4 or more pregnancies compared to those with less than 4 in a retirement community in Rancho Bernardo, CA, after adjustment for age, BMI, HDL and years since menopause [8]. Two major metaanalyses have been conducted examining parity and mortality. In a meta-analysis of 2,813,418 participants, all-cause mortality increased linearly with 5 or more births compared to the referent of 1 and was lowest at 3 or 4 births [19]. In a second meta-analysis of 994, 810 participants, there was a significant nonlinear inverse association between parity number and CVD mortality with the greatest risk reduction at a parity of 4 [20]. While other studies have documented a potential risk of CVD and mortality with increasing parity and nulliparity, socioeconomic confounders likely play a large role and cannot be completely excluded even after adjustment. To better examine the effect of educational level as a potential confounder, stratification by years of education status was performed in the largest cohort study of 527,964 Norwegian women who were followed for 42 years [9]. The cardiovascular HR for cardiovascular death among women with less than 11 years of education increased linearly with number of births from two to five or more (HR 2.37 [2.01e2.80] to 3.14 [2.19e4.51], p trend ¼ 0.02) whereas noncardiovascular death was not associated with number of births above one (p trend ¼ 0.11). In contrast, women with 11 or more years of education had a significant decline in both cardiovascular and noncardiovascular mortality with increasing number of births (p trend ¼ 0.045 and <0.01, respectively). Our study is unique as the Amish all have the same educational level (8th grade) and are remarkably homogenous with regards to socioeconomic status and modifiable cardiovascular risk factors: they eat similar diets, do not smoke or drink, do not drive, obtain regular exercise through walking and are not exposed to automobile exhaust. Moreover, they have a significantly higher number of births per woman (average 6.7 [SD 3.61] with mode of 8 and maximum of 17 births) than the average American population, which CDC quotes as 3.2 pregnancies [21]. Hence, even grandgrand multiparity confers no socioeconomic or cultural stigma in

the Amish. All these factors allowed us to evaluate the prior theorized notion that increased parity and nulliparity puts women at higher risk of CHD and mortality. Our work suggests that socioeconomic and cultural factors that occur in conjunction with increasing parity are likely major contributors to mortality and CHD. Although we observed no association with mortality, those with 10 or more pregnancies had a significantly higher BMI than those with fewer than 10. Our results differ modestly from those observed by Hardy et al. [6] who observed that mean BMI increased with increasing number of children even at lower levels of parity (from 0 to 4 or more, p for trend ¼ 0.01), an increase also observed in the Rancho Bernardo study [8]. This lack of increase in BMI with lower numbers of births may be due to the active lifestyle of the Amish, which may buffer women with limited degrees of parity from fat accumulation. Nonetheless, as parity increased above seven births, BMI did so as well, presumably reflecting the cumulative effect of weight gain over multiple pregnancies. A strength of our study is that the Amish represent a homogeneous population with high internal validity due to lack of variation in key confounders and high normative parity. Another strength is that the Amish eat very similar diets which include fruits and vegetables from their gardens. In previous analyses of formal dietary records amongst a subset of these women in 1990 by our group, their dietary composition was very similar to that of Minnesotan women at the same time period [12,22]. A limitation is that we did not have information on incident CHD on all participants, limiting our analyses of CHD, diabetes, and hypertension to prevalent, self-reported cases. The number of deaths in some categories of parity was limited, and we cannot exclude small increases or decreases in risk associated with parity. In summary, we evaluated the association of parity with mortality in a homogeneous Amish population with a minimum of confounders. Our results suggest that the higher mortality with increasing parity observed in some studies may be related to socioeconomic factors associated with higher parity. Conflict of interest The authors declared they do not have anything to disclose regarding conflict of interest with respect to this manuscript.

18

Table 4 Effect of parity on outcomes in grand multiparous studies. Study design

Population

Number subjects

Number pregnancies

Follow-up years

Outcome

Results

95% CI

Adjusted for

Ness et al. [2]

Cohort

2357

6 vs. none

28

Increase in CHD

RRa ¼ 1.6

1.1e2.2

Age, education level

Hardy et al. [6]

Cohort

Framingham heart study NHEFS MRC NSHD

4 vs. 1

12 53

Increase in BMI

RRa ¼ 1.5 Increase of 1.2 kg/m2

1.1e1.9 p ¼ 0.01

Behavioral and lifestyle variables

Steenland et al. [5]

Cohort

6 vs. none

8

Death from CHD

RRa ¼ 0.94

0.83e1.08

SES, age, race, BMI, HTN medication, smoking

Koski et al. [7]

Cohort

12,055

10 vs. 2-4

35

Total mortality hemorrhagic stroke CVD mortality

RRb ¼ 1.2 RRb ¼ 4.1 RRb ¼ 1.4

0.9e1.6 1.1e15.6 0.8e2.4

Catov et al. [4]

Cross-sectional

540

5 vs. none

e

CVD prevalence

OR ¼ 2.27

1.0e5.15

Age, BMI, smoking, socioeconomic position, age at menarche and age at first birth; homogenous for ethnic background and religion Statin use, HDL-C, perinatal complications

Parikh et al. [3]

Cohort

1,332,062

5 vs. 2

9.5

Incident CVD

HR ¼ 1.47

1.37e1.57

Hinkula et al. [18]

Cohort

84,244

5

27

SMR ¼ 1.42 SMR ¼ 1.1 SMR ¼ 0.89 SMR ¼ 1.28 SMR:1.81 HR ¼ 0.63

1.29e1.55 1.08e1.13 0.87e0.91 1.11e1.47 1.06e2.90 0.4e0.99

HR:1.18 for 1 child; 1.21 for 5 to 9; 1.49 for 10þ 2.62 1.0 (referent) HR ¼ 0.60 HR ¼ 1.34 HR ¼ 0.44

1.04e1.4 1.09e1.33 1.12e1.99 2.3e2.9

American Cancer Society (CPSII) Northern Finland women Health, Aging and Body Composition Study Swedish population register Finnish women

2533 2977 (51% women) 585,445

1294

4 vs. none

19.3

Cohort

Post menopausal women in Rancho Bernardino Israel

Mortality from: DM IHD All cancers IHD DM CVD mortality

40,454

1 to 10 þ vs. 2 to 4

37

Total mortality

Halland et al. [9]

Cohort

Norway

527,964

1,3,4 and 5 þ vs. 2

42

CVD mortality

Simons et al. [17]

Cohort

Australia

1571

6 vs. 0

16

Total mortality CHD mortality Cancer mortality

Jacobs et al. [8]

Cohort

Dior et al. [16]

3678

10

27

Socioeconomic factors

e

0.43e0.85 0.68e2.66 0.19e1.01

Age, BMI, HDL-C, years post menopause

Age at 1st birth, ethnicity, SES, health & ob conditions and smoking <11 yrs education >11 yrs education Age, BMI, HTN, DM, overweight, social variables, alcohol, smoking, peak expiratory flow, physical disability, selfrated health, a fib.

BMI, body mass index; CHD, coronary heart disease; CVD, cardiovascular disease; DM, diabetes mellitus; HDL-C, high-density lipoprotein cholesterol; HR, hazard ratio; HTN, hypertension; IHD, ischemic heart disease; MRC, medical research council; NHEFS, National Health and Nutrition Examination Survey National Epidemiologic Follow-up Study; NSHD, National Survey of Health and Development; Ob, obstetrical; OR, odds ratio; SES, socioeconomic status; SMR, standardized mortality ratio. a RR indicates Rate Ratio. b RR indicates Relative Risk.

T.K. Elajami et al. / Atherosclerosis 254 (2016) 14e19

Study (Ref)

T.K. Elajami et al. / Atherosclerosis 254 (2016) 14e19

Author contributions All authors contributed to the design and conduct of this study, study analyses, the drafting and editing of the manuscript and its final contents. Acknowledgements We thank the Amish community for their participation. References [1] R.J. Kaaja, I.A. Greer, Manifestations of chronic disease during pregnancy, JAMA 294 (2005) 2751e2757. [2] R.B. Ness, T. Harris, J. Cobb, K.M. Flegal, J.L. Kelsey, A. Balanger, et al., Number of pregnancies and the subsequent risk of cardiovascular disease, N. Engl. J. Med. 328 (1993) 1528e1533. [3] N.I. Parikh, S. Cnattignius, P.W. Dickman, M.A. Mittleman, J.F. Ludvigsson, E. Ingelsson, Parity and risk of later life maternal cardiovascular disease, Am. Heart J. 159 (2010) 215e221. [4] J.M. Catov, A.B. Newman, K. Sutton-Tyrrell, T.B. Harris, F. Tylavsky, M. Visser, et al., Parity and cardiovascular disease risk among older women: how do pregnancy complications mediate the association? Ann. Epidemiol. 18 (2008) 873e879. [5] K. Steenland, C. Lally, M. Thun, Parity and coronary heart disease among women in the American Cancer Society CPS II population, Epidemiology 7 (1996) 641e643. [6] R. Hardy, D.A. Lawlor, S. Black, M.E. Wadsworth, D. Kuh, Number of children and coronary heart disease risk factors in men and women from a British birth cohort, BJOG 114 (2007) 721e730. [7] H. Koski-Rahikkala, A. Pouta, K. Pietilainen, A.L. Hartikainen, Does parity affect mortality among parous women? J. Epidemiol. Community Health 60 (2006) 968e973. [8] M.B. Jacobs, D. Kritz-Silverstein, D.L. Wingard, E. Barrett-Connor, The association of reproductive history with all-cause and cardiovascular mortality in

19

older women: the Rancho Bernardo Study, Fertil. Steril. 97 (2012) 118e124. [9] F. Halland, N.H. Morken, L.A. DeRoo, K. Klungsøyr, A.J. Wilcox, R. Skjærven, Association of women's reproductive history with long-term mortality and effect of socioeconomic factors, Obstet. Gynecol. 126 (2015) 1181e1187. [10] V.A. McKusick, J.A. Hostetler, J.A. Egeland, Genetic studies of the Amish, background and potentialities, Bull. Johns Hopkins Hosp. 115 (1964) 203e222. [11] R.F. Hamman, J.I. Barancik, A.M. Lilienfeld, Patterns of mortality in the the old order Amish. I. Background and major causes of death, Am. J. Epidemiol. 114 (1981) 845e861. [12] F.K. Welty, S.T. Hubl, V.R. Pierotti, S.G. Young, A truncated species of apolipoprotein B (B67) in a kindred with familial hypobetalipoproteinemia, J. Clin. Investig. 87 (1991) 1748e1754. [13] E. Gingerich, Ohio Amish Directory Holmes County and Vicinity, Hiland Printing, Millersburg, Ohio, 1988, 1988 Edition. [14] E. Gingerich, Ohio Amish Directory Holmes County and Vicinity, Hiland Printing, Millersburg, Ohio, 2005, 2005 Edition. [15] Lipid Research Clinics Program, Lipid and lipoprotein analysis, in: Manual of Laboratory Operations, second ed., vol. 1, Government Printing Office, , Washington, 1974. DHEW Publication No. 75e628 (NIH). [16] U.P. Dior, H. Hochner, Y. Friedlander, R. Calderon-Margalit, D. Jaffe, A. Burger, et al., Association between number of children and mortality of mothers: results of a 37-year follow-up study, Ann. Epidemiol. 23 (2013) 13e18. [17] L.A. Simons, J. Simons, Y. Friedlander, J. McCallum, Childbearing history and late-life mortality: the Dubbo study of Australian elderly, Age Ageing 41 (2012) 523e528. €, E. Pukkala, Cause-specific mortality of grand [18] M. Hinkula, A. Kauppila, S. N€ ayha multiparous women in Finland, Am. J. Epidemiol. 163 (2006) 367e373. [19] Y. Zeng, Z.M. Ni, S.Y. Liu, X. Gu, Q. Huang, J.A. Liu, et al., Parity and all-cause mortality in women and men: a dose-response meta-analysis of cohort studies, Sci. Rep. 6 (2016) 19351. [20] H. Lv, H. Wu, J. Yin, J. Qian, J. Ge, Parity and cardiovascular disease mortality: a dose-response meta-analysis of cohort studies, Sci. Rep. 5 (2015) 13411. [21] Center for Disease Control.http://www.cdc.gov/nchs/pressroom/ 08newsreleases/pregnancydrop.html (accessed 20.06.16). [22] D.K. Arnett, B. Ziong, P.G. McGovern, H. Blackburn, R.V. Luepker, Secular trends in dietary macronutrient intake in Minneapolis-St. Paul, Minnesota, 1980-1992, Am. J. Epidemiol. 152 (2000) 868e873.