SEARCH

SEARCH BY CITATION

Keywords:

  • Cohort;
  • data linkage;
  • epidemiology;
  • prenatal alcohol;
  • stillbirth

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Exposed and comparison groups
  6. Frequency and probabilistic matching
  7. Statistical analysis and covariates
  8. Results
  9. Discussion
  10. Conclusion
  11. Disclosure of interest
  12. Contribution to authorship
  13. Details of ethics approval
  14. Funding
  15. Acknowledgements
  16. References
  17. Supporting Information

Please cite this paper as: O’Leary C, Jacoby P, D’Antoine H, Bartu A, Bower C. Heavy prenatal alcohol exposure and increased risk of stillbirth. BJOG 2012;119:945–952.

Objective  To investigate the association between heavy prenatal alcohol exposure and stillbirth.

Design  Data linkage cohort study.

Setting  Western Australia (WA).

Population  The exposed cohort included mothers with an alcohol-related diagnosis (International Classification of Diseases, ninth/tenth revisions) recorded in health data sets and all their offspring born in WA (1983–2007). Mothers without an alcohol-related diagnosis and their offspring comprised the comparison cohort.

Methods  Exposed and comparison mothers were identified through the WA Data Linkage System. Odds ratios for stillbirth at 20 + weeks of gestation were estimated by logistic regression, stratified by Aboriginal status.

Main outcome measures  The proportion of stillbirths at 20 + weeks of gestation is presented per 1000 births, as well as adjusted odds ratios (aOR) and 95% confidence intervals (95% CI), and population-attributable fractions.

Results  Increased odds of stillbirth were observed for mothers with an alcohol-related diagnosis at any stage of their life for both non-Aboriginal (aOR 1.36; 95% CI 1.05–1.76) and Aboriginal (aOR 1.33; 95% CI 1.08–1.64) births. When an alcohol diagnosis was recorded during pregnancy, increased odds were observed for non-Aboriginal births (aOR 2.24; 95% CI 1.09–4.60), with the highest odds of Aboriginal stillbirth occurring when an alcohol diagnosis was recorded within 1 year postpregnancy (aOR 2.88; 95% CI 1.75–4.73). The population-attributable fractions indicate that 0.8% of non-Aboriginal and 7.9% of Aboriginal stillbirths are the result of heavy alcohol consumption.

Conclusions  Prevention of heavy maternal alcohol use has the potential to reduce stillbirths. The lack of an association between exposure during pregnancy and Aboriginal stillbirth in this study needs further investigation.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Exposed and comparison groups
  6. Frequency and probabilistic matching
  7. Statistical analysis and covariates
  8. Results
  9. Discussion
  10. Conclusion
  11. Disclosure of interest
  12. Contribution to authorship
  13. Details of ethics approval
  14. Funding
  15. Acknowledgements
  16. References
  17. Supporting Information

The reductions in stillbirth that occurred in high-income countries in the last century have not continued and stillbirth in western countries has not decreased since the 1980s.1,2 The occurrence of stillbirth varies across countries and across populations within countries, with the lowest proportion reported in Norway (2.2/1000)2 and a much higher proportion occurring in Western Australia (7.5/1000),1 indicating that reductions in stillbirth can still be achieved in some jurisdictions.2

The early pregnancy environment, including maternal factors in minority populations and lifestyle factors such as maternal alcohol use, have been highlighted as top priorities for research into stillbirth.2 A recent systematic review of risk factors for stillbirth identified three modifiable risk factors, overweight and obesity, maternal age >35 years, and maternal smoking, that should be targeted in prevention efforts.3 The high correlation between maternal smoking and alcohol consumption was overlooked in the studies included in the review of stillbirth, and prenatal alcohol exposure was not considered as a potential confounding factor in the analyses.4–7 The prevalence of maternal alcohol consumption during pregnancy is high, ranging from around 30%8,9 to 50%10 or higher,11,12 so if an association between prenatal alcohol exposure and stillbirth exists then the contribution to the number of stillbirths may be substantial.

The evidence supporting an association between prenatal alcohol exposure and stillbirth is conflicting. Increases in stillbirth have been reported in mothers consuming the equivalent of three drinks per day (approximately UK 4 units),13 binge drinking during pregnancy,14 and having five or more drinks per week (approximately 7–8 UK units).15,16 However, some studies have reported no increase in stillbirth from maternal alcohol abuse,17 and other studies have found stillbirth increases in abstinent women compared with those who drink.18,19

Further reductions in stillbirth will require strategies to address known risk factors. Research investigating the association between prenatal alcohol exposure and stillbirth is essential to confirm whether strategies targeting this factor could help to reduce stillbirth. This study investigates the association between heavy prenatal alcohol exposure in women diagnosed with an alcohol-use disorder and stillbirth (at 20+ weeks of gestation), in a population-based cohort.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Exposed and comparison groups
  6. Frequency and probabilistic matching
  7. Statistical analysis and covariates
  8. Results
  9. Discussion
  10. Conclusion
  11. Disclosure of interest
  12. Contribution to authorship
  13. Details of ethics approval
  14. Funding
  15. Acknowledgements
  16. References
  17. Supporting Information

The population at-risk comprised women with a birth recorded on the WA Midwives Notification System between 1983 and 2007 and all their offspring whose births were recorded on the Midwives Notification System. All births with a recorded birthweight of ≥400 g or a gestational age of ≥20 weeks are reported to the Midwives Notification System within 2 days following birth (statutory notification). The information from the Midwives Notification System is linked with the Hospital Morbidity Data set and the Registry of Births, Deaths, and Marriages to ensure complete ascertainment.20

Exposed and comparison groups

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Exposed and comparison groups
  6. Frequency and probabilistic matching
  7. Statistical analysis and covariates
  8. Results
  9. Discussion
  10. Conclusion
  11. Disclosure of interest
  12. Contribution to authorship
  13. Details of ethics approval
  14. Funding
  15. Acknowledgements
  16. References
  17. Supporting Information

All mothers with an alcohol diagnosis (International Classification of Diseases, ninth revision [ICD-9]/ICD-10) (see Supplementary material, Table S1) recorded from 10 years of age and older were identified on one or more of the following data sets: the WA Hospital Morbidity data set (all inpatient admissions), Mental Health Outpatient data set, or identification of alcohol as either the primary or secondary drug problem in the Perth-based, government Drug and Alcohol Services data set. The presence of an alcohol-related diagnosis was used as a proxy for heavy maternal alcohol consumption and as a proxy for heavy prenatal alcohol exposure when the diagnosis was recorded during pregnancy. These women and their offspring are referred to as the exposed group.

The comparison cohort included a random selection of mothers with a birth recorded on the Midwives Notification System who did not have an alcohol-related diagnosis identified on the Hospital Morbidity data set, Mental Health Outpatient or Drug and Alcohol Services data sets described above. Selection of the comparison group also excluded women with an ICD-8 diagnostic code for alcoholic psychosis (291.0–291.3; 291.9), alcoholism (303.0–303.2; 303.9), or ‘accidental’ poisoning (E860) occurring in the Hospital Morbidity data set from 1970 to 1982 and the Mental Health Outpatient data set from 1966 to 1982.

Frequency and probabilistic matching

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Exposed and comparison groups
  6. Frequency and probabilistic matching
  7. Statistical analysis and covariates
  8. Results
  9. Discussion
  10. Conclusion
  11. Disclosure of interest
  12. Contribution to authorship
  13. Details of ethics approval
  14. Funding
  15. Acknowledgements
  16. References
  17. Supporting Information

A random selection of the frequency-matched records was made at a ratio of approximately 4:1 exposed:comparison groups for non-Aboriginal mothers and approximately the frequency matching for Aboriginal mothers should read 2 to 3:1. The frequency matching on maternal age was at 1-year intervals, with the exception of some teenage mothers and mothers older than 40 years of age where matching may have required larger year intervals. These women and all their children whose births were recorded on the Midwives Notification System are the comparison group.

Data from the administrative health data sets were linked by the WA Data Linkage Branch, using probabilistic matching.21 The method of linkage between the four core data sets (Midwives Notification System, Hospital Morbidity, Mental Health Outpatient and Drug and Alcohol Services data sets) is based on six passes using the unit medical record number (unique only to metropolitan public hospitals), surname, first given name, initial, date of birth, sex and address as the principal matching fields, with manual checking undertaken for possible matches, approximately 1.9% of all links. These data were then linked with mortality data from the WA Registry of Births, Deaths and Marriages. After linkage, de-identified data files were provided to the researchers after being ascribed a unique identifier for each individual. The WA linked database, in particular the data sets used in this study, is consistent with or exceeds international benchmarks with evaluations of the performance of matching estimating the proportion of mismatches to be 0.11%.21

Statistical analysis and covariates

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Exposed and comparison groups
  6. Frequency and probabilistic matching
  7. Statistical analysis and covariates
  8. Results
  9. Discussion
  10. Conclusion
  11. Disclosure of interest
  12. Contribution to authorship
  13. Details of ethics approval
  14. Funding
  15. Acknowledgements
  16. References
  17. Supporting Information

Maternal alcohol diagnosis was coded into (1) a binary variable (yes/no) and (2) the timing of recording of the alcohol-related diagnosis in relation to pregnancy (Table 1). Maternal alcohol diagnosis was coded to reflect the timing of the diagnosis in relation to pregnancy using hierarchical coding with the order as follows: (i) any record during pregnancy; and for women without a diagnosis during pregnancy the groups were (ii) if no record during pregnancy, any record within 1 year prepregnancy, (iii) if no record during pregnancy, any record within 1 year postpregnancy, (iv) if no record during pregnancy, any record more than 1 year prepregnancy, and (v) only record(s) more than 1 year postpregnancy (Table 1). The pregnancy period was estimated by subtracting gestational age22 at birth from date of birth to give the date of conception. Where a mother has more than one admission with an alcohol-related diagnosis, this method of coding classifies the timing in the order stipulated above, such that a mother with diagnoses during pregnancy and >1 year after pregnancy will be classified in category (i) ‘during pregnancy’ group. Similarly a mother with diagnoses >1 year prepregnancy and ≤1 year postpregnancy will be classified in category (iii), within 1 year postpregnancy (Table 1).

Table 1.   Coding of timing of alcohol-related diagnosis in relation to pregnancy
 (i) During(ii) ≤1 year prepregnancy(iii) ≤1 year postpregnancy(iv) >1 year prepregnancy(v) >1 year postpregnancy
(i) During pregnancyYesPossiblePossiblePossiblePossible
(ii) ≤1 year prepregnancyNoYesPossiblePossiblePossible
(iii) ≤1 year postpregnancyNoNoYesPossiblePossible
(iv) >1 year prepregnancyNoNoNoYesPossible
(v) >1 year postpregnancyNoNoNoNoYes

For mothers with an alcohol diagnosis recorded during pregnancy we examined whether the timing of the diagnosis within the gestational period affected the pregnancy outcome. We ran a regression analysis with pregnancy diagnoses coded as within 3 months before the birth, 3–6 months before, and >6 months before the birth.

The proportion of stillbirth for 20+ weeks of gestation was calculated separately for Aboriginal and non-Aboriginal children, stratified by exposure groups. All analyses were stratified by Aboriginal status. This was undertaken in recognition of the complex health issues23 and socio-economic disadvantage of Aboriginal mothers24 and the disproportionate levels of alcohol-related health and social problems.25 Results are presented per 1000 births and the net excess per 1000 births (proportion in the exposed group minus the proportion in comparison group) is calculated for the exposed group.

Odds ratios for stillbirth were calculated for the exposed offspring relative to the comparison offspring, stratified by Aboriginal status, using a logistic regression incorporating Generalised Estimating Equations, which take into account the correlation between pregnancies to the same mother.26 Generalised Estimating Equation analyses were conducted using SPSS version 17.0 (SPSS Inc., Chicago, IL, USA). Results are presented as odds ratios (OR) with 95% confidence intervals (95% CI).

All analyses were adjusted for the factors used in frequency matching (maternal age and year of birth) and stratified by Aboriginal status. Other potential confounders included maternal demographic characteristics; maternal illicit drug use (any ICD9/10 code for illicit drugs present on either the hospital morbidity or Mental Health Outpatient data sets, or illicit drug use recorded on the and Drug and Alcohol Services data set), any mental health diagnoses (ICD9/10 codes on the Hospital Morbidity and Mental Health Outpatient data sets), parity (0, 1, 2, 3+) and marital status (married, never married, separated/divorced/widowed). Maternal smoking during pregnancy (yes/no) has been collected on the Midwives Notification System since 1998 so all analyses were re-run for the birth years 1998–2007 adjusting for smoking. We also examined the relationship between maternal smoking and stillbirth entering alcohol into the model as a covariate. The population-attributable fraction for stillbirth occurring in mothers with an alcohol diagnosis and 95% CI were calculated using the method published by Natarajan et al.27 for the significant results of any alcohol diagnosis by Aboriginal status and during pregnancy for non-Aboriginal mothers using whole population numbers for the comparison numerators and denominators identified through the Midwives Notification System.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Exposed and comparison groups
  6. Frequency and probabilistic matching
  7. Statistical analysis and covariates
  8. Results
  9. Discussion
  10. Conclusion
  11. Disclosure of interest
  12. Contribution to authorship
  13. Details of ethics approval
  14. Funding
  15. Acknowledgements
  16. References
  17. Supporting Information

A higher percentage of non-Aboriginal than Aboriginal mothers were 25 years or older (58.8% versus 40.3%) and married (non-Aboriginal comparison mothers 86.5% versus exposed 73.9%; Aboriginal comparison mothers 65.5% married versus 58.5% exposed) (Table 2). Mothers with an alcohol diagnosis and Aboriginal comparison mothers were of higher parity than non-Aboriginal comparison mothers. A higher percentage of exposed non-Aboriginal and Aboriginal mothers had a mental health diagnosis or an illicit drug diagnosis than the respective comparison mothers. Maternal smoking during pregnancy was highest in exposed mothers, non-Aboriginal (50.0%) and Aboriginal (67.5%), and twice as common in Aboriginal comparison mothers (45.2%) than non-Aboriginal comparison mothers (20.1%).

Table 2.   Maternal characteristics by alcohol exposure
 Non-AboriginalAboriginal
Exposed (n = 13 919)Comparison (n = 40 377)Exposed (n = 9910)Comparison (n = 20 877)
  1. *Single includes separated, widowed, or divorced.

  2. **Smoking data collected since 1998.

Maternal age (years)
<201671 (12.0)4827 (12.0)2601 (26.2)5497 (26.3)
20–244090 (29.4)11 823 (29.3)3287 (33.2)6983 (33.4)
25–294115 (29.6)11 987 (29.7)2263 (22.8)4784 (22.9)
30–342733 (19.6)7907 (19.6)1238 (12.5)2540 (12.2)
35–391088 (7.8)3228 (8.0)460 (4.6)952 (4.6)
40+222 (1.6)605 (1.5)61 (0.6)121 (0.6)
Marital status
Married10 250 (73.9)34 885 (86.5)5760 (58.5)13 596 (65.5)
Never married3087 (22.3)5101 (12.6)3867 (39.3)6816 (32.8)
Single*527 (3.8)355 (0.9)215 (2.2)358 (1.7)
Parity
05383 (38.7)19 793 (49.0)2421 (24.4)6225 (29.8)
14234 (30.4)12 725 (31.5)2158 (21.8)4882 (23.4)
22435 (17.5)5343 (13.2)1885 (19.0)3825 (18.3)
3+1867 (13.4)2516 (6.2)3446 (34.8)5945 (28.5)
Illicit drug use2911 (20.9)554 (1.4)933 (9.4)440 (2.1)
Mental Health—Any7369 (52.9)3686 (9.1)2944 (29.7)2472 (11.8)
Maternal smoking**n = 5595n = 16 346n = 3615n = 8442
Yes2797 (49.8)3283 (20.1)2439 (67.5)3816 (45.2)

The overall proportion of stillbirth at 20 + weeks of gestation for the non-Aboriginal cohort was 6.9/1000 and for the Aboriginal cohort 14.6/1000 (results not shown in table) and were higher in exposed non-Aboriginal and Aboriginal mothers than in the respective comparison mothers (Table 3). The net excess proportion of stillbirth in exposed non-Aboriginal pregnancies was 2.9/1000 and for Aboriginal pregnancies it was 4.3/1000. The proportion of stillbirth for unexposed Aboriginal mothers was around twice that for unexposed non-Aboriginal mothers; 13.3/1000 versus 6.1/1000, respectively (Table 3).

Table 3.   Stillbirth mortality 20+ weeks of gestation for exposed and comparison groups by timing of alcohol diagnosis in relation to pregnancy and Aboriginal status
Stillbirth ≥20 weeksNon-Aboriginal (n = 54 296)Aboriginal (n = 30 787)
 Total nStillbirth nProportion/1000 birthsNet excess*Total nStillbirth nProportion/1000 birthsNet excess*
  1. *Net excess proportion per 1000 births (Proportion–comparison proportion for stillbirth in Table 2).

Exposed13 9191269.1 (7.5–10.6)2.9 (1.2–4.7)991017417.6 (15.0–20.1)4.3 (1.3–7.3)
Comparison40 3772486.1 (5.4–6.9) 20 87727713.3 (11.7–14.8) 
Timing of alcohol diagnosis
During pregnancy530815.1 (4.7–25.5)9.0 (−1.5 to 19.4)8401214.3 (6.3–22.3)1.0 (−7.2 to 9.2)
≤1 year prepregnancy879910.2 (3.6–16.9)4.1 (−2.6 to 10.8)6581319.8 (9.1–30.4)6.5 (−4.3 to 17.2)
≤1 year postpregnancy472714.8 (3.9–25.7)8.7 (−2.2 to 19.6)5021937.9 (21.2–54.5)24.6 (7.8–41.4)
>1 year prepregnancy37764010.6 (7.3–13.9)4.5 (1.1–7.8)16582615.7 (9.7–21.7)2.4 (−3.8 to 8.6)
>1 year postpregnancy8262627.5 (5.6–9.4)1.4 (−0.7 to 3.4)625210416.6 (13.5–19.8)3.4 (−0.2 to 6.9)

The net excess proportion of stillbirth varied according to the timing of exposure and ethnicity (Table 3). For non-Aboriginal children the highest net excess stillbirth was observed in children of mothers with an alcohol-related diagnosis recorded during pregnancy (9.0/1000; 95% CI – 1.5 to 19.4), or >1 year prepregnancy (4.5/1000; 95% CI 1.1–7.8). Examination of the timing of the diagnosis during each trimester of pregnancy showed the net excess proportion doubled to 18.9/1000 (95% CI 0.5–3.6) when a diagnosis was recorded during the first trimester (results not shown in table). The numbers of stillbirths in late pregnancy were too small to calculate the proportions. The highest net excess cases of Aboriginal stillbirth occurred when a maternal diagnosis was recorded within 1 year postpregnancy (24.6/1000; 95% CI 7.8–41.4). When a diagnosis was recorded during pregnancy, the net excess for Aboriginal pregnancies was 1.1/1000 (95% CI – 7.2 to 9.2) (Table 3) increasing slightly when an alcohol diagnosis was recorded in the first trimester (1.2/1000; 95% CI – 9.4 to 11.9) (results not shown in table).

The results of the logistic regression analyses are shown in Table 4. In the adjusted model, the odds of stillbirth were increased for non-Aboriginal children in the analyses of ‘any’ alcohol diagnosis (adjusted odds ratio [aOR] 1.36; 95% CI 1.05–1.76) and elevated in all categories of timing of exposure with the exception of > 1 year postpregnancy. The highest odds ratios were when a diagnosis was recorded during pregnancy (aOR 2.24; 95% CI 1.09–4.60), within 1 year postpregnancy (aOR 2.22; 95% CI 1.02–4.85), and > 1 year prepregnancy (aOR 1.62; 95% CI 1.14–2.31). Adjusting for maternal smoking during pregnancy in the model restricted to 1998–2007 births did not change the odds, with the exception of a reduction in odds for the timing of within 1 year prepregnancy (aOR 1.46; 95% CI 0.94–2.27). When the timing of the alcohol diagnosis within pregnancy was examined, a fourfold increased odds ratio was observed when a diagnosis was recorded within 3 months of the birth (aOR 3.72; 95% CI 1.60–8.66), which was maintained following adjustment for maternal smoking during pregnancy (Table 4).

Table 4.   Odds ratios for stillbirth at 20 +  weeks of gestation by alcohol exposure and Aboriginal status
MortalityNon-AboriginalAboriginal
Adjusted for frequency matching variables*Adjusted**Adjusted for smoking***Adjusted for frequency matching variables*Adjusted**Adjusted for smoking***
  1. *Adjusted for frequency matching variables maternal age, year of birth.

  2. **Adjusted for frequency matching variables and marital status, parity, illicit drugs, any mental health.

  3. ***Analyses restricted to birth years 1998–2007 and adjusted for previous variables plus smoking during pregnancy.

Stillbirth ≥20 weeks
No alcohol diagnosis1.001.001.001.001.001.00
Any alcohol diagnosis1.48 (1.19–1.84)1.36 (1.05–1.76)1.47 (1.00–2.15)1.33 (1.09–1.63)1.33 (1.08–1.64)1.26 (0.89–1.77)
Timing of alcohol diagnosis
No alcohol diagnosis1.001.001.001.001.001.00
During pregnancy2.40 (1.18–4.89)2.24 (1.09–4.60)2.24(1.02–4.91)1.05 (0.59–1.87)0.95 (0.52–1.76)0.95 (0.43–2.11)
≤1 year prepregnancy1.69 (0.88–3.26)1.60 (0.82–3.13)1.16 (0.46–2.90)1.49 (0.86–2.60)1.51 (0.86–2.65)1.18 (0.50–2.77)
≤1 year postpregnancy2.39 (1.11–5.11)2.22 (1.02–4.85)2.02 (0.75–5.43)2.79 (1.72–4.54)2.88 (1.75–4.73)3.22 (1.67–6.23)
>1 year prepregnancy1.75 (1.25–2.45)1.62 (1.14–2.31)1.46 (0.94–2.27)1.16 (0.77–1.74)1.14 (0.75–1.73)1.15 (0.69–1.92)
>1 year postpregnancy1.21 (0.91–1.61)1.12 (0.80–1.58)1.26 (0. 67–2.38)1.28 (1.01–1.63)1.30 (1.01–1.66)1.09 (0.61–1.92)

For Aboriginal stillbirth at 20 + weeks of gestation increased odds were observed for analyses of ‘any’ alcohol diagnosis (aOR 1.33; 95% CI 1.08–1.64) and when a diagnosis was recorded within 1 year postpregnancy (aOR 2.88; 95% CI 1.75–4.73) (Table 4). No increased odds were observed when a diagnosis was recorded during pregnancy (aOR 0.95; 95% CI 0.52–1.76). Adjusting for maternal smoking during pregnancy did not substantially alter the odds.

The population-attributable fractions were calculated for significant results. For ‘any’ alcohol diagnosis the population-attributable fractions were 0.8% (95% CI 0.4–1.3%) for non-Aboriginal stillbirth and 7.9% (95% CI 4.0–12.2%) for Aboriginal stillbirth. The fraction for non-Aboriginal stillbirth when an alcohol diagnosis was recorded during pregnancy was 0.11% (95% CI 0.005–0.27%).

An association between maternal smoking during pregnancy and stillbirth was evident only in the univariate analysis for non-Aboriginal pregnancies (OR 1.55; 95% CI 1.10–2.17) and the association decreased once the alcohol variable was entered into the model (OR 1.27; 95% CI 0.90–1.78) and adjusted (aOR 1.14; 95% CI 0.8–1.62) (results not shown in table). There was no association between maternal smoking and stillbirth in the Aboriginal cohort (univariate OR 1.08; 95% CI 0.80–1.47) or after alcohol was added to the model (OR 1.04; 95% CI 0.76–1.43) (results not shown in table).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Exposed and comparison groups
  6. Frequency and probabilistic matching
  7. Statistical analysis and covariates
  8. Results
  9. Discussion
  10. Conclusion
  11. Disclosure of interest
  12. Contribution to authorship
  13. Details of ethics approval
  14. Funding
  15. Acknowledgements
  16. References
  17. Supporting Information

The presence of an alcohol diagnosis, indicating heavy prenatal alcohol exposure, is associated with increased odds of stillbirth in both non-Aboriginal and Aboriginal mothers. The odds of stillbirth were elevated for both non-Aboriginal mothers with any alcohol-use disorder (aOR 1.36; 95% CI 1.05–1.76) and Aboriginal mothers (aOR 1.33; 95% CI 1.08–1.64). In the relationship between the timing of the alcohol diagnosis and stillbirth from 20 weeks of gestation the results varied between the non-Aboriginal and Aboriginal mothers. Increased odds ratios for non-Aboriginal stillbirths were observed when an alcohol diagnosis was recorded during pregnancy (aOR 2.24; 95% CI 1.09–4.60). The odds ratio increased to almost four-fold when an alcohol diagnosis was recorded in the last 3 months of the pregnancy, supporting the biological plausibility of the observed association. Although Aboriginal stillbirth showed a different association with timing of alcohol diagnosis with increased odds when a diagnosis was recorded within 1 year postpregnancy (aOR 2.88; 95% CI 1.75–4.73) but not when a diagnosis was recorded during pregnancy.

These results strengthen the evidence that heavy prenatal alcohol exposure increases the risk of stillbirth, which further highlights the importance of preventing heavy alcohol consumption by pregnant women. The findings are among the highest odds ratios reported for prenatal alcohol exposure13–16 and other known modifiable risk factors,1,3 and are similar to the two-fold to three-fold increased risk of stillbirth reported for women drinking at levels ranging from five or more drinks/week15,16 to more than five drinks/occasion (binge drinking).14 Importantly, the estimates of effect remained when the analyses were adjusted for maternal smoking during pregnancy, which is consistent with the previous studies.14–16 However, as maternal smoking data were only available from 1998 onwards the smaller sample size resulted in the confidence intervals overlapping for some timing groups.28 When we ran the analysis using maternal smoking as the exposure variable, the association between smoking and stillbirth disappeared when the alcohol variable was entered as a covariate. These findings indicate that future research investigating the association between smoking and stillbirth needs to take into account maternal alcohol consumption during pregnancy. The lack of an association between Aboriginal mothers’ smoking and stillbirth is probably a result of the high prevalence of smoking in the comparison Aboriginal mothers.

To our knowledge this is the first study to examine the odds of stillbirth in mothers using the presence of an alcohol-related diagnosis as a proxy for heavy alcohol consumption. The exposed mothers in this cohort represent a heavy-drinking, high-risk population which, by virtue of their alcohol diagnosis, differ from other women who consume alcohol at heavy levels but who have not been identified with an alcohol-use disorder by health services. Exposed mothers have a higher risk profile than comparison mothers including a higher proportion smoking during pregnancy, with a diagnosis of illicit drug use, and a mental health diagnosis.

To assess the public health impact of heavy maternal alcohol use and stillbirth we have calculated the population attributable fraction, which gives an indication of how many stillbirths are attributable to heavy maternal alcohol use and so, theoretically, preventable. Assuming that the association is causal the population-attributable fractions indicate that for non-Aboriginal mothers the proportion of stillbirths attributable to any maternal alcohol-related diagnosis is 0.8%, whereas for Aboriginal mothers 7.9% of stillbirths are the result of heavy alcohol consumption. The population-attributable fractions reported here represent the fraction for women with an alcohol diagnosis and should be viewed as a conservative estimate of magnitude of effect of prenatal alcohol exposure on stillbirth, because increased risk of stillbirth has been documented for women drinking at relatively low levels (five or more drinks/week).15,16 The lower fraction observed when a diagnosis was recorded during pregnancy for non-Aboriginal mothers (0.11%) and the lack of an association in Aboriginal mothers suggest that there has been under-recognition of alcohol-related problems occurring during pregnancy.

This is supported by the two-fold to three-fold increase in the odds of stillbirth for both non-Aboriginal and Aboriginal mothers when an alcohol diagnosis was recorded within 1 year postpregnancy. Aboriginal mothers are less likely to access antenatal services and for those who do attend antenatal services their first visit is later in pregnancy and they have fewer sessions than non-Aboriginal mothers,29 which limits the opportunity for an alcohol diagnosis to be recorded during pregnancy. Also, the high proportion of stillbirths occurring in Aboriginal comparison mothers may reflect unrecognised/unrecorded alcohol problems or confounding by the high prevalence of other health and social problems.23–25 Unadjusted confounding may have contributed to our findings, particularly as we are unlikely to have fully accounted for other possible confounding factors in this group of high-risk mothers.

The strengths of this study are that it is a large, population-based cohort of high-risk mothers who have an alcohol diagnosis and all their births recorded on the Midwives Notification System. The exposed and comparison cohorts have been identified through accessing multiple data sources in the WA data linkage system, which has been demonstrated as a valid means of identifying women admitted to hospital for a health-related condition.30 This methodology overcomes loss-to-follow-up and recall bias, enabling longitudinal follow up of both the mothers and their offspring. Women in this cohort would have had overt alcohol-related problems in order for the alcohol diagnosis to have been recognised and recorded given the lack of routine questioning by health professionals of pregnant women31 and mental health patients32,33 about their alcohol consumption, so we are confident that the women who received an alcohol diagnosis were drinking heavily. The prevalence of alcohol diagnoses recorded during pregnancy is similar to the findings of a recent New South Wales data-linkage study.34 However, the lack of routine questioning of pregnant mothers about their alcohol consumption has resulted in a lower prevalence of alcohol-use disorders in non-Aboriginal women in this study compared with the reported population estimate of 5% for these disorders in Australian women.35

There is a high percentage of women in WA drinking heavily, with the highest rates occurring during the childbearing years,36,37 and alcohol contributes to a substantial proportion of hospital bed-days for both Aboriginal and non-Aboriginal women in WA.36 However, few Aboriginal mothers access the metropolitan drug and alcohol services for which we had data and we were not able to access data for Aboriginal-specific drug and alcohol and Mental Health Outpatient services so this may have exacerbated the under-recognition of alcohol-related problems in Aboriginal mothers during pregnancy.

These results indicate that more emphasis should be given to identifying heavy alcohol consumption and alcohol-related problems in women of childbearing age and pregnant women followed by implementation of prevention31 and early intervention38 strategies to reduce the risk of stillbirth.

Conclusion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Exposed and comparison groups
  6. Frequency and probabilistic matching
  7. Statistical analysis and covariates
  8. Results
  9. Discussion
  10. Conclusion
  11. Disclosure of interest
  12. Contribution to authorship
  13. Details of ethics approval
  14. Funding
  15. Acknowledgements
  16. References
  17. Supporting Information

Mothers with an alcohol-related diagnosis, indicating heavy alcohol consumption, have increased odds of having a stillbirth. The increased odds of stillbirth for non-Aboriginal mothers when an alcohol diagnosis was recorded during pregnancy imply a causal association. The lack of an association between an alcohol diagnosis during pregnancy and stillbirth in Aboriginal mothers may be the result of under-ascertainment of drinking during pregnancy or unadjusted confounding and requires further investigation. These results indicate that prevention of alcohol misuse during pregnancy has the potential to reduce stillbirth in developed nations. Future research investigating the association between maternal smoking during pregnancy and stillbirth needs to also examine alcohol exposure.

Contribution to authorship

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Exposed and comparison groups
  6. Frequency and probabilistic matching
  7. Statistical analysis and covariates
  8. Results
  9. Discussion
  10. Conclusion
  11. Disclosure of interest
  12. Contribution to authorship
  13. Details of ethics approval
  14. Funding
  15. Acknowledgements
  16. References
  17. Supporting Information

All authors have contributed to the study and approved the final version of this manuscript. The original cohort design was developed by CO’L and CB; CO’L designed this study, analysed the data, and drafted the manuscript; PJ provided statistical advice for this paper; CO’L, CB and AB provided expertise on alcohol and pregnancy; AB provided expertise on maternal alcohol use disorders; HD’A provided expertise on Aboriginal issues. CO’L has access to all study data.

Details of ethics approval

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Exposed and comparison groups
  6. Frequency and probabilistic matching
  7. Statistical analysis and covariates
  8. Results
  9. Discussion
  10. Conclusion
  11. Disclosure of interest
  12. Contribution to authorship
  13. Details of ethics approval
  14. Funding
  15. Acknowledgements
  16. References
  17. Supporting Information

Ethics approval for the conduct of this study was granted by the Princess Margaret Hospital Research Ethics Committee (#1244/EP), the WA Confidentiality of Health Information Committee, WA Department of Health (#2006/18) (now called the Health Research Ethics Committee), the WA Aboriginal Health Information Ethics Committee (#134-04/06) and the Curtin Human Research Ethics Committee (#39/2010).

Funding

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Exposed and comparison groups
  6. Frequency and probabilistic matching
  7. Statistical analysis and covariates
  8. Results
  9. Discussion
  10. Conclusion
  11. Disclosure of interest
  12. Contribution to authorship
  13. Details of ethics approval
  14. Funding
  15. Acknowledgements
  16. References
  17. Supporting Information

This study was supported by an Australian National Health and Medical Research Council (NHMRC) Public Health (Australia) Fellowship (594451) (CO’L), NHMRC programme grant number 572742, and an NHMRC Research Fellowship (353628) (CB). CO’L was also supported by infrastructure grants from Curtin University and the Western Australian Drug and Alcohol Office.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Exposed and comparison groups
  6. Frequency and probabilistic matching
  7. Statistical analysis and covariates
  8. Results
  9. Discussion
  10. Conclusion
  11. Disclosure of interest
  12. Contribution to authorship
  13. Details of ethics approval
  14. Funding
  15. Acknowledgements
  16. References
  17. Supporting Information

The authors thank the staff of the WA Data Linkage Unit for access to the WA Data Linkage System and for their assistance in obtaining the data, and the WA Health Data Custodians for access to the core health data sets.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Exposed and comparison groups
  6. Frequency and probabilistic matching
  7. Statistical analysis and covariates
  8. Results
  9. Discussion
  10. Conclusion
  11. Disclosure of interest
  12. Contribution to authorship
  13. Details of ethics approval
  14. Funding
  15. Acknowledgements
  16. References
  17. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Exposed and comparison groups
  6. Frequency and probabilistic matching
  7. Statistical analysis and covariates
  8. Results
  9. Discussion
  10. Conclusion
  11. Disclosure of interest
  12. Contribution to authorship
  13. Details of ethics approval
  14. Funding
  15. Acknowledgements
  16. References
  17. Supporting Information

Table S1. International Classification of Disease, tenth revision (ICD-10) and ICD-9 codes for alcohol diagnoses.

FilenameFormatSizeDescription
BJO_3333_sm_TableS1.pdf58KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.