SEARCH

SEARCH BY CITATION

Keywords:

  • Cerebrovascular disease;
  • epidemiology;
  • family history;
  • ischaemic heart disease;
  • miscarriage;
  • venous thromboembolism

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Please cite this paper as: Smith G, Wood A, Pell J, Hattie J. Recurrent miscarriage is associated with a family history of ischaemic heart disease: a retrospective cohort study. BJOG 2011;118:557–563.

Objective  To determine whether women experiencing recurrent miscarriage were more likely to have a family history of cardiovascular disease.

Design  Retrospective cohort study.

Setting  Women having a first birth in Scotland between 1992 and 2006.

Sample  A total of 74 730 first births were linked to the hospital admission and death certification data for the women’s parents through the women’s birth certificates.

Methods  The incidence of cardiovascular disease in the women’s parents was related to the number of miscarriages experienced before their daughters’ first births using a Cox proportional hazards model.

Main outcome measures  Death or hospital admission of the women’s parents for ischaemic heart disease (IHD), cerebrovascular disease (CVD) or venous thromboembolism (VTE).

Main results  There was an increased incidence of IHD in the parents of women who experienced two miscarriages before their first birth (hazard ratio 1.25, 95% CI 1.04–1.49) and parents of women who experienced three or more miscarriages before their first birth (hazard ratio 1.56, 95% CI 1.14–2.15). Adjustment for the characteristics of the women at the time of the first birth was without material effect. There was no significant association between miscarriage and family history of CVD or VTE. There was no significant association between the number of therapeutic terminations of pregnancy before the first birth and the incidence of any type of cardiovascular disease in the women’s parents.

Conclusions  The parents of women who experience recurrent miscarriage are more likely to experience IHD. Recurrent miscarriage and IHD may have common patho-physiological pathways and genetic predispositions.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Studies in the 1950s and 1960s demonstrated that women who experienced miscarriage were more likely to have ischaemic heart disease (IHD) in later life.1 However, it was unclear if this was independent of the total number of pregnancies, which has itself been identified as a risk factor for IHD.2 We addressed this by studying 129 290 women having first births in Scotland and relating their subsequent experience of IHD to the number of miscarriages and therapeutic terminations of pregnancy experienced before their first birth.3 That study confirmed a specific association between miscarriage and the subsequent risk of IHD. We speculated that the association may indicate common risk factors for miscarriage and IHD, which may be inherited or acquired. In the present study we sought to determine whether there was evidence for common inheritable risk factors for both miscarriage and IHD by studying the relationship between women’s experience of miscarriage before the first birth and their parents’ incidence of cardiovascular disease.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Data and record linkage

We used a linked data set, which has previously been described in detail.4 In brief, we used a national registry of pregnancy outcome data, the Scottish Morbidity Record 02 (SMR02), to obtain information about women having their first birth. We then used identifiers for the mother (first name, unmarried name and date of birth) to link this record to her own birth certificate using a probability-based matching approach. The parents’ names and dates of births were recorded on the birth certificate and this was then used to link the SMR02 pregnancy data to the parents’ records for death (the General Registrar’s Office, GRO) or hospital admission (SMR01). We identified hospital admissions for cardiovascular disease, specifically, ischaemic heart disease (IHD), cerebrovascular disease (CVD) or venous thromboembolism (VTE). We used SMR02 records from 1992 to 2006. The start date was chosen because smoking was only recorded from this year onwards and the end date represented the latest period for which data were available at the time of the linkage. The birth certificate database was maintained in a computerised form from 1967 onwards. Hence, we were able to perform the linkage for women who were born in Scotland from 1967 onwards. Death certificate and SMR01 records were available between 1981 and 2006. Hence, we were able to identify events in the parents during this period. Approval of the study was provided by the Privacy Advisory Committee of the Information Services Division of NHS Scotland.

Definitions

The number of previous miscarriages and therapeutic terminations of pregnancy was self-reported by the women. Hence, this would have included losses occurring in the community (i.e. not requiring hospital admission). However, further details of the losses, such as gestational age, were not available. Hence, miscarriage was defined as previous spontaneous loss before viability. Demographic characteristics were available from the SMR02 for the women at the time of their first births. Postcode of residence at the time of the first birth was used to derive Carstairs socioeconomic deprivation scores.5 Smoking was defined as the smoking status of the woman at the time of first attendance for prenatal care. Maternal height was measured in centimetres and the value used was that documented in each woman’s clinical record. Maternal age was defined as the age of the mother at the time of birth. Marital status was defined as married or unmarried. First singleton livebirth was defined as records where all previous pregnancies had ended in either miscarriage or therapeutic termination of pregnancy and the current pregnancy ended in an infant born weighing 400 g or more between 24 and 43 weeks of gestation, excluding stillbirths and multiple pregnancies. Events in the women’s parents were defined on the basis of International Classification of Disease (ICD) codes listed in the principal position of the hospital record (SMR01) or death certificate (GRO). Ischaemic heart disease was defined as either 410–414 (ICD9) or I20–25 (ICD10); CVD was defined as either 430–438 (ICD9) or I60–I69 or G45 (ICD10); and VTE was defined as either 453.8, 453.9 or 415.1 (ICD9) or I82.8, I82.9, or I26 (ICD10).

Statistical analysis

Continuous variables were summarised by the median and interquartile range and comparisons between groups were performed using the Kruskal–Wallis test. Unadjusted comparisons of categorical data were performed using the chi-square test for trend. The P values for all hypothesis tests were two-sided and statistical significance was assumed at P < 0.05. The aim of the analysis was to estimate the measure of association between miscarriage in the woman and the experience of IHD in her parents. The approach we took was to use miscarriage as the exposure and to use IHD in the parents as the event. The rationale for this is addressed in the Discussion. The relative hazard of cardiovascular disease in the women’s parents was estimated using a Cox proportional hazards model, with parental age as the time scale. The unit of analysis was the parents and the number of previous losses, miscarriage or therapeutic terminations of pregnancy, experienced by their daughters was the exposure. The analyses were clustered on the parental identifier to allow dependence between siblings (some parents had more than one daughter included in the cohort). Covariables with > 5% missing values (height and smoking status) were imputed using multiple imputation by chain equations.6 Five imputed data sets were created by replacing missing values for height and smoking status with simulated values from a set of imputation models constructed from all covariables and all outcome variables (i.e. the Nelson–Aalen estimators for parental age at disease onset and the censoring indicators). Distributions of imputed values were visually checked for comparability with the observed data. Almost identical results were obtained from analyses of the complete cases. All available maternal characteristics were included in multivariable analyses. Categorical covariables were modelled by dummy variables and maternal age and height were treated continuously. The baseline hazard for the Cox model was stratified by maternal age quintiles. Assessment of linearity was performed using fractional polynomials and assessment of the proportional hazards assumption was performed using the Schoenfeld residuals in the observed data (i.e. not the imputed data).7 Where the proportional hazards assumption was violated, analyses were stratified on the given variable. All statistical analyses were performed using the Stata software package (Stata Corporation, College Station, TX, USA), version 10.1.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Our previous linkage and cohort selection resulted in 120 317 pregnancy records from a total of 78 294 women (see previous publication for details of selection process4). From this group, we now excluded 45 580 records which were not first births and seven records with a missing value for the primary exposure (number of prior miscarriages). This resulted in a dataset of 74 730 first births, with data on number of previous miscarriages for the women and data on hospital admission or death of the parents between 1981 and 2006 with cardiovascular disease. The basic characteristics of the women are described in relation to the number of miscarriages experienced before the first birth (Table 1). Women experiencing previous miscarriages tended to be older, more likely to be married and were more likely to smoke. Linkage identified 51 729 fathers of these women and 51 105 mothers of these women (some mothers had daughters in the cohort who had different fathers). Among the fathers, there were records for death or hospital admission for 5345 IHD events, 1454 CVD events and 288 VTE events. Among the mothers, there were records for death or hospital admission for 1908 IHD events, 1000 CVD events and 319 VTE events.

Table 1.   Summary of maternal characteristics (n = 74 730)
 Number of previous miscarriages*P
0123 or more
n = 66 934n = 6 671n = 901n = 224
  1. IQR, interquartile range.

  2. Data are n (%) unless stated otherwise.

  3. *Kruskal–Wallis test or chi-square test for trend, as appropriate.

Age (years)
Median (IQR)20 (18–23)21 (19–24)22 (20–25)23 (20–25)< 0.001
Marital status
Unmarried60 865 (90.9)5915 (88.7)765 (84.9)188 (83.9)< 0.001
Married6069 (9.1)757 (11.4)136 (15.1)36 (16.1)
Socio-economic deprivation category
1 (least deprived)1371 (2.0)125 (1.9)25 (2.8)0 (0.0)0.03
25342 (8.0)546 (8.2)68 (7.6)13 (5.8)
311 569 (17.3)1088 (16.3)135 (15.0)36 (16.1)
418 060 (27.0)1795 (26.9)230 (25.5)69 (30.8)
513 742 (20.5)1420 (21.3)208 (23.1)47 (21.0)
69615 (14.4)972 (14.6)123 (13.7)24 (10.7)
7 (most deprived)7235 (10.8)725 (10.9)112 (12.4)35 (15.6)
Height (cm)
Median (IQR)163 (158–167)163 (159–167)163 (159–167)162 (158–167)0.75
Missing, n (%)13 375 (20.0)1 351 (20.3)196 (21.8)59 (26.3)
Smoking status
Non-smoker29 653 (44.3)2611 (39.1)340 (37.7)74 (33.0)< 0.001
Smoker24 182 (36.1)2796 (41.9)368 (40.8)107 (47.8)
Ex-smoker8237 (12.3)787 (11.8)116 (12.9)18 (8.0)
Missing4862 (7.3)477 (7.2)77 (8.6)25 (11.2)

The risk of each type of event in the parents was then analysed using a Cox proportional hazard model, with the daughters’ experience of miscarriage before the first birth as the main exposure. Univariate analysis demonstrated that there was a significant relationship between the number of miscarriages experienced before the women’s first births and the incidence of IHD in their parents (Table 2). There was a 25% increase in risk among parents of women with two previous losses and a 56% increase in risk among parents of women with three or more losses. Associations were adjusted for the other characteristics of the women at the time of the first birth. Adjustment for the maternal characteristics at the time of the first birth was without material effect. The specificity of relationships was examined by comparison with associations with the daughters’ experience of therapeutic terminations of pregnancy before the first birth. There was no independent and statistically significant association between the number of previous therapeutic terminations of pregnancy and the incidence of IHD in the women’s parents. There was no significant relationship between the number of miscarriages experienced before the first birth and the risk of CVD (Table 3) or VTE (Table 4) in the women’s parents.

Table 2.   Early pregnancy losses before first singleton live birth and parental risk of ischaemic heart disease
 Death or hospital admission of parents due to IHD
UnadjustedAdjusted*
HR (95% CI)PHR (95% CI)P
  1. 95% CI, 95% confidence intervals; HR, hazard ratio; IHD, ischaemic heart disease.

  2. *Adjusted for maternal age in years, marital status, smoking status, socio-economic deprivation category and height and stratified by quintile of maternal age.

Miscarriages
01.00 1.00 
11.02 (0.94–1.10)0.0041.04 (0.96–1.12)0.002
21.25 (1.04–1.49) 1.29 (1.08–1.55) 
3 or more1.56 (1.14–2.15) 1.53 (1.11–2.11) 
Per miscarriage increase1.08 (1.02–1.14)0.0041.09 (1.03–1.15)0.001
Therapeutic terminations of pregnancy
01.00 1.00 
10.91 (0.84–0.99)0.100.96 (0.88–1.04)0.59
20.84 (0.65–1.09) 0.91 (0.70–1.18) 
3 or more1.15 (0.65–2.01) 1.21 (0.68–2.14) 
Per termination of pregnancy increase0.93 (0.87–1.00)0.0470.97 (0.90–1.04)0.40
Table 3.   Early pregnancy losses before first singleton live birth and parental risk of cerebrovascular disease
 Death or hospital admission of parents due to CVD
UnadjustedAdjusted*
HR (95% CI)PHR (95% CI)P
  1. 95% CI, 95% confidence intervals; CVD, cerebrovascular disease; HR, hazard ratio.

  2. *Adjusted for maternal age in years, marital status, smoking status, socio-economic deprivation category and height and stratified by quintile of maternal age.

Miscarriages
01.00 1.00 
11.05 (0.92–1.19)0.831.07 (0.94–1.22)0.57
21.10 (0.80–1.53) 1.17 (0.84–1.62) 
3 or more0.93 (0.52–1.90) 0.94 (0.47–1.87) 
Per miscarriage increase1.03 (0.94–1.12)0.541.05 (0.96–1.15)0.31
Therapeutic terminations of pregnancy
01.00 1.00 
10.94 (0.82–1.09)0.871.00 (0.87–1.16)0.97
20.99 (0.66–1.50) 1.11 (0.73–1.68) 
3 or more0.82 (0.26–2.56) 0.94 (0.30–2.94) 
Per termination of pregnancy increase0.96 (0.85–1.07)0.431.01 (0.90–1.13)0.87
Table 4.   Early pregnancy losses before first singleton live birth and parental risk of venous thromboembolism
 Death or hospital admission of parents due to VTE
UnadjustedAdjusted*
HR (95% CI)PHR (95% CI)P
  1. 95% CI, 95% confidence intervals; HR, hazard ratio; VTE, venous thromboembolism.

  2. *Adjusted for maternal age in years, marital status, smoking status, socio-economic deprivation category and height and stratified by quintile of maternal age.

Miscarriages
01.00 1.00 
10.74 (0.55–1.00)0.140.76 (0.56–1.03)0.19
More than 11.03 (0.57–1.87) 1.07 (0.59–1.95) 
Per miscarriage increase1.03 (0.94–1.12)0.541.05 (0.96–1.14)0.28
Therapeutic terminations of pregnancy
01.00 1.00 
11.00 (0.75–1.34)0.991.05 (0.79–1.41)0.94
More than 10.93 (0.42–2.08) 1.05 (0.47–2.36) 
Per termination of pregnancy increase0.99 (0.79–1.24)0.911.04 (0.83–1.30)0.73

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The main finding of the present study is that there was an increased incidence of IHD in the parents of women who experienced multiple miscarriages before their first birth. There was evidence of a ‘dose-dependent’ relationship with a 25% increased risk in the parents of women experiencing two losses and a 56% increase in the parents of women experiencing three or more losses. The association did not appear to be explained by a confounding effect of the women’s age, marital status, smoking status, socio-economic deprivation or height, as adjustment for these factors had no material effect. Neither was there an association between therapeutic termination of pregnancy and the risk of any type of cardiovascular disease in the women’s parents. We interpret these data as supporting the hypothesis that there may be common inheritable risk factors for both miscarriage and IHD.

Many studies have demonstrated that women who experience complications during pregnancy, such as pre-eclampsia, preterm birth and intrauterine growth restriction, are at increased risk of developing cardiovascular disease in later life.8–11 We have interpreted these observations as indicating that there are common predisposing factors for pregnancy complications and cardiovascular disease, stating that ‘occult cardiovascular, microvascular, or hemostatic dysfunction is manifested in pregnancy complications during reproductive years and in overt cardiovascular disease in later life’.12 We recently reported the relationship between late pregnancy complications and family history of cardiovascular disease. We demonstrated that the parents of women who experienced pre-eclampsia, preterm birth or delivery of a small-for-gestational-age infant during pregnancy had an increased incidence of cardiovascular disease.4 However, a significant proportion of the associations was lost following statistical adjustment for the characteristics of the women at the time of the pregnancy (age, marital status, smoking status, socio-economic deprivation and height).4 We concluded that there were common risk factors for late complications of pregnancy and cardiovascular disease which demonstrated familial aggregation. The attenuation of the association in multivariate analysis led us to conclude that these were, in whole or in part, environmental in nature. In contrast, in the present analysis, adjustment for the same series of maternal characteristics had no material effect on the association between miscarriage and family history of IHD. Hence, we hypothesise that the most plausible explanation for the association between recurrent miscarriage and family history of IHD is the existence of common genetic or epigenetic risk factors for both conditions. Further research should address whether women with a history of recurrent miscarriage have higher rates of carriage of known genetic predisposing risk factors for IHD. Studies should also address whether they differ in relation to other risk factors, such as hypertension, insulin resistance, dyslipidaemia, elevated high sensitivity C-reactive protein and vascular dysfunction, including endothelial dysfunction. It is possible that a history of recurrent miscarriage may be clinically useful in identifying women who would benefit from screening for cardiovascular risk factors.

The lack of association between women’s recurrent miscarriage and their parents’ experience of VTE is consistent with other recent evidence. A number of studies have suggested that there are associations between inherited and acquired thrombophilia and recurrent miscarriage.13 However, two recent large-scale studies failed to demonstrated beneficial effects of anti-coagulant treatment on pregnancy outcome among women with recurrent miscarriage.14,15 These observations suggest that, in a significant proportion of women, recurrent miscarriage is not related to a pro-thrombotic tendency and the current data are consistent with this view. An important caveat to this interpretation is that there were fewer VTE events than IHD. Consequently, the confidence intervals for the associations are also consistent with an association between recurrent losses and family history of VTE of up to a 14% increase per previous miscarriage (Table 4). Similarly, the association with CVD had confidence intervals that were consistent with a comparable relationship to IHD and the negative result should not be interpreted as excluding a significant association.

In the present analysis, we used a Cox proportional hazards model, with age as the time scale, to quantify variation in the incidence of cardiovascular disease in the women’s parents. This approach may be counterintuitive, as events in the parents could have either preceded or followed the experience of miscarriages in the daughters. Moreover, it may seem more intuitive to study miscarriage as the ‘event’ with family history of IHD as the exposure. However, all measures of familial association have weaknesses16 and the approach we employed had two major strengths. First, the analysis of family history of heart disease was limited to the parents. The experience of heart disease in first-degree relatives will depend on the number of siblings. If there are inheritable predispositions to miscarriage, it is possible that the number of siblings a woman has would vary according to her propensity to experience miscarriage. By confining analysis to the parents, we overcame this possible source of bias. Second, women who are older are more likely to have older parents. The older the parents, the more likely they are to have experienced disease. Women who experienced multiple miscarriages were older than those who had not (Table 1). These characteristics could, therefore, lead to an association between miscarriage and IHD through confounding by age. However, by modelling IHD in the parents using a Cox model with age as the time scale, we were able to compare the incidence of IHD independently of variation in the parental age. A more conventional indicator of family history of IHD is self-reported history of IHD in a first-degree relative. However, if we had used such a measure in the present analysis, the results would have been vulnerable to biases related to systematic variation in the number of siblings and the age of the women’s parents in relation to the primary exposure. Hence, although the current approach to assessing family history is different from many previous studies, the method employed addressed two important potential sources of bias.

The strengths of the present analysis are that we were able to study a large number of women and were able to compare associations between previous miscarriage and therapeutic termination of pregnancy. Moreover, all definitions were obtained from registries so these were not open to bias through selective recall, as could occur in a case–control study. However, our study also has a number of weaknesses. The association was relatively modest with an approximate 50% increase in the incidence of IHD in the women’s parents. However, the current study was the result of a previously stated hypothesis3 and the P value for the multivariate analysis was 0.001. Hence, it is very unlikely to be a chance finding. The SMR02 lacked data on the gestational age of previous miscarriage. Hence, we were unable to determine whether the associations were related to first-trimester losses, second-trimester losses, or both. Furthermore, because of the constraints of the data, we only addressed losses before a first pregnancy. Hence, we do not have direct information on women who had losses but no births. Neither do we present information on the association between total number of losses in a woman’s life history. Registry-based analysis relies on correct coding of both exposures and definitions. Hence, we do not have access to information that allows confirmation of either, such as cotinine assay for smoking status or the results of diagnostic tests for the parental experience of disease. Similarly, we lack information on specific causes of recurrent miscarriage, such as balanced chromosomal translocation in the parents, or diagnosis of antiphospholipid syndrome. Furthermore, as previously discussed in detail,4 the linkage process favoured matching records from younger and unmarried women. The current cohort is neither a random nor a representative sample of the Scottish population. Nevertheless, we have addressed an important question for the first time using a large-scale data set and the results suggest a significant association. Moreover, a previous analysis relating birthweight, pre-eclampsia and preterm birth to the mother’s subsequent risk of cardiovascular disease using the same data sources8 has been replicated in multiple subsequent studies in many different countries.9–11

In conclusion, we found that there was an increased incidence of IHD in the parents of women who experienced multiple miscarriages before their first birth. We hypothesise that there are common pathophysiological pathways linking IHD and recurrent miscarriage. This hypothesis makes the prediction that genetic risk factors for IHD will also be associated with an increased risk of recurrent miscarriage.

Disclosure of interests

None of the authors has any form of conflict of interest in relation to this work.

Contribution to authorship

GS had the idea. GS and JH designed the data linkage. JH performed the data linkage and extracted the data. GS and AW performed the data analysis. GS drafted the paper. All authors critically reviewed the manuscript and approved the final version.

Details of ethics approval

The record linkage was approved by the Privacy Advisory Committee of the Information Services Division of NHS Scotland.

Funding

The record linkage was funded by a National Institutes for Health Research (NIHR, UK) Senior Investigator award to GS and this work was conducted with support from the NIHR Cambridge Comprehensive Biomedical Research Centre.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We are grateful to Mr Ian White, Senior Scientist at the MRC Biostatistics Unit, Institute of Public Health, Cambridge for discussions on the statistical analysis.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References