Participants

Data was taken from the Mater-University of Queensland study of pregnancy and its outcomes (MUSP), ^{Reference Najman, Bor, O'Callaghan, Williams, Aird and Shuttlewood11} a 21-year prospective birth cohort study that investigates the health and behaviour of mother–child pairs. The baseline female cohort (n=3474) were born between 1981 and 1984 at a major metropolitan public hospital in Brisbane, Australia. Mother–daughter pairs were subsequently reassessed at 6 months, 5, 14, and 21 years after birth. During the 21-year follow-up, over the years 2000–2004, all the original mother–child pairs were recontacted; however, resource constraints limited our ability to interview people residing outside of the greater Brisbane area and the 7-year gap between follow-ups made it difficult to regain contact with some. There were few who refused to participate. No social or other characteristics were used to screen inclusion, although there was a modest over-representation of disadvantage in women lost to follow-up (online Table DS1). At 21-year follow-up, 1988 female children (57.2% of the baseline cohort), now young women, were interviewed. Of these, 61.5% (n=1223) had complete data on lifetime pregnancy outcomes, psychiatric diagnoses at age 21 and covariates from birth, 14 and 21 years. Ethical approval for each phase was obtained from the University of Queensland.

Measurement of psychiatric disorders

At 21-year follow-up, the lifetime version of the Composite International Diagnostic Interview computerised version (CIDI– Auto, version 2.1) ¹² was used to assess psychiatric disorders. The CIDI has been found to have good reliability and validity. ^{Reference Andrews and Peters13} The study reports six DSM–IV ¹⁰ disorders: total affective and anxiety disorders; alcohol, cannabis and other illicit drug use and dependence disorders, as well as nicotine withdrawal and dependence disorders. The age at first onset of disorder and disorders occurring within the previous 12 months were used. Although we had intended to explore associations between pregnancy outcomes and specific psychiatric disorders such as psychosis, only 21 women met DSM–IV criteria for any psychosis. We were, therefore, unable to undertake meaningful analysis of the influence of pregnancy outcomes on psychotic disorders.

Measurement of pregnancy outcomes

Women reported the number of previous pregnancies, abortions, miscarriages and births they had experienced by age 21. We created a variable which incorporated all four possibilities (never pregnant, live birth only, ever had an abortion and ever miscarried) (Fig. 1). Seventeen women had both an abortion and miscarriage: for the analysis, six women who had miscarried two or three times were coded to miscarriage, five women who had had two abortions were coded to abortion, and the remaining six women with one abortion and one miscarriage were randomly assigned to either the abortion or miscarriage group. Those with missing data in any of the questions were excluded.

Fig. 1 Pregnancy outcomes for 280 women reporting a pregnancy at 21 years (n=1223, excludes 943 never pregnant women, includes 17 women who had both an abortion and miscarriage).

Confounders

Confounders included maternal and familial factors, measures related to pre-existing behaviour problems and substance misuse, as well as demographic characteristics of the young women.

Statistical analysis

Chi-squared tests and logistic regression were used to explore the bivariate associations of pregnancy outcomes and each DSM–IV disorder. We then fitted multivariate logistic models on a restricted sample of young women (n=1223) with complete data for each DSM–IV disorder separately. Associations with each disorder were adjusted consecutively for maternal and family factors at baseline and 14-year follow-up (model 1), adolescent behaviour problems at 14 (model 2), adolescent behaviour reported at 21 (model 3) and concurrent demographic factors (model 4). Clearly non-significant factors (P>0.1 or those with unstable estimates as assessed by large standard errors) were removed. Each model was reassessed until only significant effects (P<0.05) remained. This was undertaken separately for each of the six DSM–IV disorders. Unadjusted and adjusted odds ratios (OR) and 95% confidence intervals were used to estimate the association between pregnancy outcomes and each disorder compared with the never pregnant group. All pairwise pregnancy outcome comparisons were undertaken with each of the fully adjusted models.

Sensitivity analyses

A series of sensitivity analyses were done to further explore the timing of the associations and effects of substance misuse, behaviour problems and psychiatric morbidity that possibly preceded pregnancy loss. Factors such as adolescent cigarette smoking, alcohol consumption and anxious/depressed and aggressive/delinquent behaviour were excluded from the model as these are possible causal factors associated with later psychiatric morbidity. Six separate sensitivity analyses were conducted:

1. (a) current (12-month) diagnoses were used instead of lifetime diagnoses;

   and exclusion of women who reported:

2. (b) both a miscarriage and abortion;

3. (c) giving birth and pregnancy loss either as an abortion or miscarriage or both;

4. (d) smoking, drinking alcohol or anxious/depressed problem behaviours at 14 years;

5. (e) aggressive/delinquent behaviours at 14 years; and

6. (f) onset of a particular DSM–IV disorder at or before the age of sexual debut.

Finally, we used inverse probability weighting ^{Reference Hogan, Roy and Korkontzelou20} to determine whether loss to follow-up affected the validity of our findings. This was done for the six DSM–IV diagnoses both lifetime and current (12-month) disorders. Analyses were carried out using SPSS version 15.0 and STATA version 9.0 on Windows.

Associations between pregnancy and psychiatric disorders

The unadjusted odds of having a lifetime psychiatric or substance use disorder were highest among women reporting a pregnancy loss and lowest among women who never had a pregnancy or gave birth; women with a nicotine disorder were the exception. Women with a history of abortion or miscarriage were twice as likely to meet criteria for alcohol, cannabis, other illicit drug use, depressive or anxiety disorder compared with those who were never pregnant or who had given birth (Table 1). Young women who became pregnant regardless of outcome had similarly high odds of having a lifetime tobacco dependency compared with never pregnant women.

Table 1 Rates of psychiatric disorders and substance use disorders at 21, by history of birth, abortion or miscarriage (n=1223)

Lifetime DSM—IV disorders	Never pregnanta (n=943)	Live birth (n=97)	Abortion (n=101)	Miscarriage (n=82)	P b
Nicotine (withdrawal and dependence) (n=198)
Prevalence, %	11.8	26.0	34.7	32.9	<0.001
OR (95% CI)c	1.0	2.6 (1.6-4.3)	4.0 (2.5-6.3)	3.7 (2.2-6.1)
Alcohol (misuse and dependence) (n=214)
Prevalence, %	14.7	16.8	35.6	29.6	<0.001
OR (95% CI)	1.0	1.2 (0.7-2.1)	3.2 (2.1-5.0)	2.4 (1.5-4.1)
Cannabis (misuse and dependence) (n=172)
Prevalence, %	11.0	15.5	31.7	25.6	<0.001
OR (95% CI)	1.0	1.5 (0.8-2.7)	3.7 (2.3-6.0)	2.8 (1.6-4.7)
Other illicit substance used (misuse and dependence) (n=93)
Prevalence, %	5.4	6.3	21.8	17.3	<0.001
OR (95% CI)	1.0	1.2 (0.5-2.8)	4.8 (2.8-8.4)	3.6 (1.9-7.0)
Affectivee (n=344)
Prevalence, %	25.1	28.7	45.5	43.2	<0.001
OR (95% CI)	1.0	1.2 (0.8-1.9)	2.5 (1.6-3.8)	2.3 (1.4-3.6)
Anxietyf (n=418)
Prevalence, %	30.1	41.1	51.5	53.7	<0.001
OR (95% CI)	1.0	1.6 (1.1-2.5)	2.5 (1.6-3.7)	2.7 (1.7-4.2)

Adjustment for confounding

Online Table DS2 shows associations between pregnancy outcomes and a number of potential confounding factors. Pregnancy outcomes were associated with greater social and economic disadvantage, family instability, alcohol and cigarette use, aggressive/delinquent behaviours, early sexual intercourse and poorer educational outcomes at both age 14 and 21. There were only modest differences between women reporting an abortion and those with a history of miscarriage or birth.

Table 2 shows the multivariable logistic regression analyses. Women reporting a pregnancy were twice as likely to develop a tobacco dependency, with the association being slightly attenuated by adjustment for maternal and family factors and adolescent behaviour. Women who reported an abortion had twice the odds of having an alcohol disorder and nearly four times the odds of an illicit drug use disorder (excluding cannabis). Also, those reporting an abortion had an increased risk of affective disorder, although this association bordered significance after adjustment for 21-year demographic factors. Women who reported a miscarriage had a similarly high risk of developing an illicit drug use disorder, but not affective or alcohol disorders, as women reporting an abortion.

Table 2 Associations as odds ratio (OR) (95% CI) for lifetime DSM–IV disorders in young women who reported a birth, abortion or miscarriage (n=1223)

	DSM—IV disorders at 21-year follow-up, OR (95% CI)
Model	Nicotine	Alcohol	Cannabis	Other illicit substance use	Affective	Anxiety
1. Adjusted for early life and family factorsa,b
Live birth	2.7 (1.6-4.5)	1.1 (0.6-1.9)	1.3 (0.7-2.3)	1.4 (0.6-3.3)	1.0 (0.6-1.7)	1.4 (0.9-2.1)
Abortion	4.0 (2.5-6.4)	2.8 (1.8-4.4)	3.3 (2.1-5.4)	5.8 (3.3-10.4)	2.2 (1.4-3.5)	2.0 (1.3-3.1)
Miscarriage	3.6 (2.1-6.2)	2.2 (1.3-3.7)	2.5 (1.5-4.4)	4.1 (2.1-8.2)	2.1 (1.3-3.5)	2.3 (1.4-3.6)
2. Adjusted for previous significant factors and adolescent behavioura,c
Live birth	2.4 (1.4-4.1)	0.8 (0.5-1.5)	1.0 (0.6-2.0)	1.1 (0.4-2.8)	0.9 (0.5-1.5)	1.2 (0.8-1.9)
Abortion	3.3 (2.0-5.4)	2.3 (1.4-3.8)	2.6 (1.6-4.3)	4.4 (2.4-8.0)	1.9 (1.2-3.0)	1.8 (1.2-2.8)
Miscarriage	3.1 (1.8-5.4)	1.9 (1.1-3.2)	2.1 (1.2-3.8)	4.0 (2.0-8.1)	2.0 (1.2-3.4)	2.1 (1.3-3.4)
3. Adjusted for previous significant factors and adolescent factors reported at 21a,c,d
Live birth	2.4 (1.4-4.2)	0.8 (0.4-1.5)	0.8 (0.4-1.4)	1.0 (0.4-2.6)	0.8 (0.5-1.4)	1.1 (0.7-1.8)
Abortion	2.6 (1.6-4.3)	2.0 (1.2-3.3)	1.7 (1.0-2.8)	3.6 (2.0-6.7)	1.7 (1.1-2.7)	1.5 (1.0-2.5)
Miscarriage	2.3 (1.3-4.3)	1.5 (0.8-2.7)	1.5 (0.8-2.6)	2.6 (1.2-5.4)	1.7 (1.0-3.0)	1.7 (1.0-2.9)
4. Fully adjusted for previous significant factors and child's demographic factors reported at 21-year follow-upa,e
Live birth	2.1 (1.2-3.8)	0.9 (0.5-1.7)	0.8 (0.4-1.5)	1.0 (0.4-2.6)	0.7 (0.4-1.2)	1.0 (0.6-1.6)
Abortion	2.5 (1.5-4.2)	2.1 (1.3-3.5)	1.5 (0.9-2.5)	3.6 (2.0-6.7)	1.5 (0.9-2.4)	1.5 (1.0-2.4)
Miscarriage	2.1 (1.2-3.9)	1.7 (0.9-3.1)	1.4 (0.8-2.5)	2.6 (1.2-5.4)	1.4 (0.8-2.4)	1.5 (0.9-2.5)

Sensitivity analyses

In this birth cohort study, DSM–IV diagnoses and pregnancy outcomes were concurrently measured, making the direction of the associations difficult to determine. Therefore a series of six sensitivity analyses were undertaken. When we excluded women with complex pregnancy outcomes (17 participants reporting both an abortion and miscarriage and 68 reporting one or more types of pregnancy loss as well as a birth) there was no substantial change to our results, with the exception of the association between abortion and anxiety disorder, which became significant (OR=1.6, 95% CI 1.1–2.7) when individuals with abortion and miscarriage were removed, although the same was not seen when women reporting at least one birth and pregnancy loss were removed. Exclusion of participants reporting cigarette and alcohol use, anxious/depressed behaviour and aggressive/delinquent behaviour at 14 yielded similar results to those presented in Table 2.

As timing of the pregnancy was not recorded, young women reporting onset of a disorder at or before the age of their debut sexual intercourse were excluded. The majority of young women had onset of DSM–IV depression (55.8%, n=192) and anxiety (87.6%, n=366) before they started having sexual intercourse. Removal of those with early onset strengthened associations between abortion and depression (OR=1.9, 95% CI 1.1–3.3) and all the pregnancy outcomes and anxiety (OR=2.9, 95% CI 1.2–6.7). The remaining results were essentially unchanged from those reported in Table 2 (online Table DS4).

A sensitivity analysis using current (12-month) DSM–IV disorders was conducted to better account for the temporal sequence of pregnancy outcomes, psychiatric and substance use disorders. The findings were not substantially different from Table 2, except the association between abortion and current affective disorder became significant (OR=1.9, 95% CI 1.1–3.2) (online Table DS5).

Attrition

There was significant loss to follow-up between the 14- and 21-year surveys. To determine if loss to follow-up at 21 years affected the validity of our findings, we conducted a sensitivity analysis using inverse probability weights ^{Reference Hogan, Roy and Korkontzelou20} modelling the probability of having missing data at age 21. Relevant variables available at baseline were included in an exploratory regression model to determine whether individuals remaining in the study significantly differed from those lost to follow-up. Loss to follow-up at age 21 was predicted by being the child of a teenage mother (OR=1.8, 95% CI 1.5–2.0) who smoked (OR=1.4, 95% CI 1.3–1.7) and reported anxiety symptoms during pregnancy (OR=1.4, 95% CI 1.2–1.7) (online Table DS1). When weights using factors that predicted missing data at age 21 were included in all adjusted models, the results were virtually the same as those presented here, except the association between lifetime alcohol dependency and miscarriage became significant (OR=1.9, 95% CI 1.1–3.5) (online Table DS5).

Main findings

This study found that young women with a lifetime history of either abortion or miscarriage were at increased risk of tobacco dependence and illicit drug use (excluding cannabis) compared with women who had never been pregnant. Women who gave birth were also at similar increased risk of tobacco dependence. However, it appears that pregnancy and parenting tasks may discourage young women from heavy drug and alcohol use. ^{Reference Jaffee, Caspi, Moffitt, Belsky and Silva21} Women who had an abortion were twice as likely as women who were never pregnant or who gave birth to have an alcohol disorder and current affective disorder. Associations remained robust after adjustment for a range of confounders and a series of sensitivity analyses.

Two prospective longitudinal studies, one from New Zealand ^{Reference Fergusson, Horwood and Ridder2} and the other from Norway, ^{Reference Pedersen3} used standardised psychiatric interviews to examine a range of psychiatric and substance use disorders following abortion. Both found evidence of a causal association between abortion and later psychiatric and substance use disorders after controlling for a wide range of pre-pregnancy factors including pre-existing mental health. The New Zealand birth cohort study followed 506 women and found those reporting an abortion before age 21 had significantly higher rates of depression, anxiety, alcohol and illicit drug use by age 25 compared with women who had never been pregnant or women who had given birth. ^{Reference Fergusson, Horwood and Ridder2} The Norwegian study followed 769 young women from age 15 to 27 and found abortion before age 20 was associated with later nicotine dependence, alcohol, cannabis and other illicit drug use in women having no stable relationship with the biological father. ^{Reference Pedersen3} Our study found that women having an abortion had an increased risk of both lifetime and current DSM–IV diagnoses of alcohol, illicit substance use and current affective disorders but women giving birth had no such increased risk, except for tobacco dependency. Contrary to Pedersen, ^{Reference Pedersen3} we found no association between abortion and cannabis disorder.

Despite this, our findings challenge the existing evidence of a causal link between nicotine, ^{Reference Pedersen3} other illicit drug use, alcohol use disorders and abortion. ^{Reference Fergusson, Horwood and Ridder2} Pairwise comparisons found that pregnancy outcomes were all equally associated with tobacco dependence; abortion and miscarriage exhibited similar associations with lifetime and current other illicit drug and lifetime alcohol use disorders. To date, no population-based study has compared adverse risk of psychiatric and substance use disorders in women reporting miscarriage and induced abortion. This study is the first large longitudinal cohort that has found a similar pattern of increased risk of alcohol and other illicit drug use disorders in women reporting either abortion or miscarriage compared with women who were never pregnant or had given birth. The findings suggest that the poor outcomes reported for women who had an induced abortion may be associated with pregnancy loss rather than simply the experience of abortion. Induced abortion and miscarriage are both stressful life events that have been shown to lead to anxiety, sadness and grief ^{Reference Major, Cozzarelli, Cooper, Zubek, Richards, Wilhite and Gramzow5,Reference Klier, Geller and Ritsher22} and, for some women, serious depression and substance use disorders. ^{Reference Klier, Geller and Ritsher22–Reference Martino, Collins, Ellickson and Klein24} Although we were not able to fully explain the mechanisms behind these associations, a number of possible pathways were explored in sensitivity analyses. Early heavy alcohol, illicit drug use and depression have been linked to risky sexual activity, unwanted pregnancies and pregnancy complications, ^{Reference Martino, Collins, Ellickson and Klein24,Reference Ramrakha, Caspi, Dickson, Moffitt and Paul25} and could be on the causal pathway leading to both pregnancy loss and adverse psychiatric outcomes. After taking these factors into account and removing early onset disorders our findings remained robust. This suggests that our results were unlikely to be explained by reverse causality. Pregnancy loss and substance misuse may have shared risk factors. Unplanned pregnancy commonly co-occurs with individual and social risk factors such as early sexual activity, poor school performance, cigarette smoking, alcohol and illicit drug taking and behaviour problems. ^{Reference Jaffee, Caspi, Moffitt, Belsky and Silva21,Reference Martino, Collins, Ellickson and Klein24,Reference Fergusson, Boden and Horwood26} Although we may have not been able to control for all confounding, associations in our study remained robust after adjustment for a wide range of such potential confounding factors. Finally, pregnancy loss may be directly associated with an increased risk of later substance misuse, particularly if alcohol and illicit drugs are used to decrease emotional responses to the loss. ^{Reference Fergusson, Horwood and Ridder2,Reference Pedersen3,Reference Bradshaw and Slade27–Reference Broen, Moum, Bödtker and Ekeberg29} Future longitudinal studies are needed to better understand the relationship of pregnancy loss with adverse psychiatric outcomes in young women.

Limitations

This study has a number of limitations. Contextual information related to circumstances of the pregnancy loss were not collected, so we could not account for associated factors such as reasons to abort, support received and gestational age at the time of the pregnancy loss. We had no access to medical records to investigate whether specific circumstances surrounding the termination, the timing of termination or other pregnancy-related events would confound associations between pregnancy loss and psychiatric disorders. Concern about possible underreporting of abortion has been raised. ^{Reference Fergusson, Horwood and Ridder2,Reference Reardon and Cougle30} Data on abortion history was collected from a self-reported survey, and Jones & Kost ^{Reference Jones and Kost31} found this method provides a more accurate abortion history than face-to-face interview. Women may have misclassified induced abortions as miscarriages; however, this appears unlikely as the reported rate of miscarriage was no higher than expected. ^{Reference de Vaus32} It is also possible that the use of retrospective recall and concurrent assessment of pregnancy outcomes and DSM–IV disorders introduced some bias in our findings, although sensitivity analysis using more recent onset psychiatric disorders produced the same results as those shown in our main analysis. The high rate of attrition between the 14- and 21-year follow-up raised the possibility of bias in estimation of the magnitude of risk. Weighted analysis, which adjusted for factors associated with attrition at age 21, was not substantially different from the reported findings, which suggests that attrition bias is unlikely to have produced bias in our results. Finally, ours is a study of young women whose ages ranged from 18 to 23 years and therefore our results should not be generalised to older women or women of all ages.

Our study found that young women had similar increased risks of lifetime and current tobacco use, alcohol dependency and illicit substance use, although not cannabis, after an abortion or miscarriage compared with women who had never been pregnant. These associations may relate to common factors associated with pregnancy loss or early pregnancy, rather than being caused by the experience of induced abortion.