Strengths and limitations

To our knowledge, this is the first study to explore long-term (i.e. from pregnancy until age 18 of their study child) trajectories of depressive symptoms among mothers with lifetime eating disorders and current eating disorder symptoms. This study had a number of strengths. We used data from a large general population cohort of mothers, which allowed us to avoid biases associated with clinical samples and investigate symptom presentations that might be common in the general population. Owing to the large size of our sample we could also investigate differences in patterns of depressive symptoms across eating disorder diagnoses, which have not been explored in previous literature. Finally, our multilevel analytical approach allowed us to make efficient use of the longitudinal nature of our sample and minimise the impact of losses to follow-up.

Our study also had a number of limitations. First, our main exposure variable was based on a self-reported measure of lifetime eating disorder, which could have resulted in some degree of misclassification or recall bias. However, a number of studies have used this definition^{Reference Kothari, Solmi, Treasure and Micali11,Reference Nicholls, Viner, Doll, Davies, O'Connor and Sullivan17} as there is evidence that this approach yields acceptable levels of sensitivity and specificity in community surveys.^{Reference Keski-Rahkonen, Sihvola, Raevuori, Kaukoranta, Bulik and Hoek12} Our prevalence estimates are also in line with those previously observed in the literature.^{Reference Smink, Van Hoeken and Hoek18} the use of lifetime data allows the investigation of relatively uncommon disorders such as eating disorders, which are understudied in population-based cohorts.

Women were asked whether they had ever had anorexia or bulimia nervosa. For this reason, we might have missed women with atypical presentations, binge eating disorder – which was not considered as a full eating disorder in DSM-IV – or individuals with cases of undiagnosed anorexia nervosa and bulimia nervosa (i.e. that women might have been unable to recognise as such). To tackle this limitation, we also used a measure of current symptoms, as opposed to diagnoses and not only did we find comparable results, but also observed that the two exposures were strongly and independently associated with our outcome.

Because eating disorders are relatively uncommon, despite our large sample size and the use of lifetime diagnoses, our main exposure categories were not very large, particularly for those with both anorexia and bulimia nervosa resulting in broad 95% CIs often overlapping between groups. A previous study^{Reference Morgan, Lacey and Sedgwick2} found that women with a history of both anorexia and bulimia nervosa reported twice as many cases of postnatal depression compared with women with bulimia nervosa only, lending support to our finding of greater symptom severity in this group.

The associations we observed in our sensitivity analyses could be the result of reverse causation, so that women with greater depressive symptoms expressed more concerns about their image and eating. Similarly, given the self-reported and retrospective nature of the exposure, it was not possible to disentangle the relative contribution of eating disorders and lifetime depression to the outcome, nor the direction of this association. Regardless, we believe that this association offers important insight on the coexistence of these symptoms in pregnancy and might help to identify women at greater risk for both conditions in the short and long term.

Comparison with previous literature

Our findings are largely consistent with those of previous investigations using a trajectory-based modelling approach, finding that women with current or lifetime history of eating disorders have greater depressive symptoms compared with the general population.^{Reference Easter, Solmi, Bye, Taborelli, Corfield and Schmidt5,Reference Micali, Simonoff and Treasure8} These studies, however, also showed that depressive symptoms in women with eating disorders decrease over the antenatal and perinatal period.^{Reference Easter, Solmi, Bye, Taborelli, Corfield and Schmidt5,Reference Micali, Simonoff and Treasure8}

In line with these findings, we observed that women in the top quintile of current body image and eating concerns in pregnancy experienced some degree of symptom improvement over time (see Supplementary Figure 1), although these were not sustained over time. Our findings suggest that these previously observed improvements in depressive symptoms over the antenatal and perinatal periods could be temporary. Alternatively, we suggest that these discrepancies between our findings and those of previous investigations could have emerged because of differences in exposure definition.^{Reference Easter, Solmi, Bye, Taborelli, Corfield and Schmidt5,Reference Micali, Simonoff and Treasure8}

First, we did not define eating disorders as current or past because we would not have had enough statistical power to investigate patterns of depression in women who reported anorexia nervosa or bulimia nervosa to have occurred ‘recently’ (see Supplementary Appendix 1). Nevertheless, to better understand the role of current eating disorder symptoms, we also explored trajectories of depressive symptoms according to current body image and eating concerns while adjusting for lifetime eating disorder. Here, we found that both were independently associated with greater depressive symptoms. This suggests that both lifetime diagnoses and current body image concerns could help to identify women at increased risk of depression not just in pregnancy, but also over the life-course.

Second, we did not group eating disorders according to presence or absence of self-reported lifetime diagnosis of ‘severe depression’. Given the retrospective nature of the questions included in ALSPAC, we would not have been able to exclude that depression was a consequence of eating disorders. Hence, stratifying for depression could have masked the effect of the exposure on the outcome. Nevertheless, future studies should further explore mechanisms that explain these patterns of comorbidity, as they could provide important insights on the aetiology of these conditions across the life-course.

Interpretation of findings

The finding that mothers with a lifetime history of eating disorders – regardless of their diagnosis – continue to experience greater depressive symptoms up to 18 years after pregnancy could indicate that a considerable number of women might not fully recover from eating disorders, since these are highly comorbid with depression. It is known that, among women treated for eating disorders, around a third do not fully recover.^{Reference Keel and Brown19} A study by Bardone-Cone and colleagues found that women who achieve only partial recovery from eating disorders scored similarly to women with active eating disorders on a number of eating disorder, body image and psychopathology scales (including depression and anxiety) with greater scores compared with healthy controls and women who had achieved full recovery.^{Reference Bardone-Cone, Harney, Maldonado, Lawson, Robinson and Smith20}

However, we also found that lifetime eating disorders were still associated with greater depressive symptoms, regardless of body image and eating concerns in pregnancy. This association is consistent with the hypothesis that depressive symptoms might be a risk factors for onset of eating disorders and persist after recovery. Supporting this hypothesis, a study of adolescents who had achieved a 3-year recovery from anorexia nervosa (defined as weight restoration, stable periods, and absence of eating disorder symptoms) found that the latter, compared with healthy controls, still reported greater obsessive–compulsive, anxious and depressive traits.^{Reference Holtkamp, Muller, Heussen, Remschmidt and Herpertz-Dahlmann21}

The potential persistence of depressive symptoms in mothers after eating disorder recovery is of note, as the former have been associated with negative outcomes in the offspring. A previous study has shown that children of women with eating disorders had greater odds of mental health difficulties in early childhood and that these associations were largely mediated by the presence of depression and anxiety in pregnancy.^{Reference Micali, Stahl, Treasure and Simonoff22} Although little is known about the long-term psychopathological outcomes of children of mothers with eating disorders, postnatal depression has consistently been associated with offspring developmental delays,^{Reference Deave, Heron, Evans and Emond23} higher rates of psychopathology^{Reference Flouri, Ruddy and Midouhas24} and obesity.^{Reference Lampard, Franckle and Davison25} A recent study has also shown that children of mothers with more severe and persistent maternal depression,^{Reference Netsi, Pearson, Murray, Cooper, Craske and Stein26} have adverse long-term outcomes, including greater depression and poorer academic achievement. This evidence suggests that children of women with eating disorders could be at greater risk of a number of adverse outcomes by way of the latter experiencing greater and more persistent depressive symptoms. Future studies should explore these associations.

Finally, it is worth noting that for all women, regardless of eating disorder diagnosis, depressive symptoms increased when their offspring was 18 years of age. Recent nationwide UK data from the 2014 Adult Psychiatry Morbidity Survey found that the prevalence of depression was highest in women aged 16–24 years, decreased between 25 and 44 and then increased again between the ages of 45 and 54.²⁷ The reasons for the increase in depression between the ages of 45 and 54 years are poorly understood, but they likely involve a complex interplay of social, psychological and biological factors. The menopause, for example, occurs between the ages of 44 and 55 years, with an average age in the UK of 51 years. Depressive symptoms have been found to increase during the menopausal transition (around 3 years prior to the menopause).^{Reference Freeman, Sammel, Boorman and Zhang28} More research on risk factors for midlife depression in women are thus warranted.

To conclude, from a public health point of view, identifying pregnant women who have had an eating disorder in the past or might be experiencing concerns with their body image in pregnancy could help to also detect women at risk for depression. This might help to reduce the burden of postnatal depression, with benefits for both women and, potentially, their offspring. High levels of stigma perceived by people with eating disorders and difficulties in identifying eating disorders in general practice represent known barriers to treatment access, particularly in pregnancy.^{Reference Bye, Shaw, Bick, Easter, Kash-McDonald and Micali29} A recent systematic review suggests that presence of mental health comorbidity is perceived by people with eating disorders as facilitating help-seeking.^{Reference Ali, Farrer, Fassnacht, Gulliver, Bauer and Griffiths30} In the UK, current guidelines from the National Institute for Health and Care Excellence (NICE) recommend using the EPDS or the Whooley questionnaire as part of a general discussion about woman's mental health and well-being during pregnancy and in the early postnatal period.³¹ As recent evidence confirms, women who disclosed depressive symptoms using these measures were also more likely to report other comorbid mental health disorders, including eating disorders.^{Reference Howard, Ryan, Trevillion, Anderson, Bick and Bye32} Hence, identifying depression in women with current or past eating disorders could provide them with a gateway to accessing more intensive perinatal care, as recommended by NICE.³³ This could also help to prevent postnatal relapse of the eating disorder and long-term depressive symptoms identified in this study.

It is of critical importance to increase awareness of eating disorders in pregnancy and to train clinicians and midwives to recognise and support women with eating disorders in this crucial phase of their lives. The antenatal and early postnatal periods can be a time when eating disorder symptoms temporarily improve^{Reference Taborelli, Easter, Keefe, Schmidt, Treasure and Micali34} and women might be more open to change, hence they could be a key moment for the identification of these conditions with long-term benefits for these women.