LAUSR

449 and alcohol, tobacco or other substance use at or before cohort entry), socioeconomic deprivation, rurality, and the time period of index hospitalization. We included quadratic time interaction terms to determine associations by year of follow-up. There were 5,169 women with anorexia nervosa in the cohort, including 227 who died during follow-up. Mortality was higher for women with anorexia than no anorexia (3.24 vs. 0.38 per 1,000 person-years). In adjusted models, anorexia was associated with 2.47 times the risk of death compared with no anorexia (95% CI: 2.01-3.04). Women with three or more anorexia admissions had 4.05 times the risk of death over time (95% CI: 2.85-5.75). Anorexia nervosa was associated with 9.01 times the risk of death at 5 years (95% CI: 7.28-11.16), 7.18 times the risk at 10 years (95% CI: 6.07-8.51), and 2.90 times the risk at 20 years (95% CI: 2.16-3.89), but was not significantly associated with mortality at 25 years of follow-up (HR=1.47, 95% CI: 0.88-2.45). Anorexia nervosa was associated with death from suicide (HR=4.90, 95% CI: 1.93-12.46), pulmonary disease (HR=3.49, 95% CI: 1.77-6.89), diabetes and other endocrine disease (HR=7.58, 95% CI: 1.89-30.42), liver and other digestive disease (HR=3.27, 95% CI: 1.33-8.06), and shock and organ failure (HR=3.59, 95% CI: 1.23-10.49). Among pulmonary causes, anorexia was most strongly associated with death due to pneumonia (HR=8.19, 95% CI: 2.78-24.14). The cause of death was specified as anorexia nervosa for five patients (2.2%). There was no long-term association with death from cardiovascular or other causes. Risk of death was particularly elevated for diabetes and pneumonia, disorders that may be underappreciated conditions associated with anorexia nervosa. While it is plausible that severe calorie restriction has effects on pancreatic and lung function, it is also known that women with type 1 diabetes are at greater risk of developing eating disorders. Diabetic women with anorexia nervosa sometimes manipulate their insulin injections to control weight, increasing the risk of hyperglycemic episodes, diabetic ketoacidosis, and life-threatening complications such as diabetic coma. Women with anorexia nervosa may be at risk of pneumonia due to food aspiration. The elevated risk of pneumonia mortality may also be due to a reduced immune response to bacterial infections, leading to delayed diagnosis or treatment and more severe pulmonary infections. Suicide was also a leading cause of death. Anorexia nervosa frequently clusters with depression, anxiety, and personality disorders, as well as substance use. Alcohol use in particular is associated with a high risk of suicide attempt in patients with anorexia nervosa. However, some data suggest that mortality rates are elevated even in women with anorexia nervosa who do not have psychiatric comorbidities. In the present study, anorexia nervosa was associated with greater mortality even after adjusting for depression and anxiety, suggesting that at least some of the pathways linking anorexia nervosa with mortality are independent of comorbid mental disorders. In contrast to the frequent involvement of the cardiovascular system in acute anorexia nervosa, cardiovascular disease was not a leading cause of death in this analysis. In a prior study of 6,009 Swedish women, anorexia nervosa was similarly more strongly associated with suicide, respiratory and endocrine-related causes than cardiovascular death. It may be that low weight due to decreased calorie intake mitigates damage to the cardiovascular system. This study has limitations. We assessed severe anorexia nervosa requiring hospitalization, not milder anorexia adequately managed in outpatient settings. We did not have information on anorexia relapse or recovery status, body mass index, physical activity, or nutrition. Cause of death data were partially missing before 2006. We used a comparison group comprised of fertile women. Our results may therefore differ from studies using the general population as a reference group. The long-term role of anorexia nervosa in mortality has yet to be fully appreciated. In this study with 29 years of follow-up, anorexia nervosa hospitalization was associated with an increased risk of death up to 20 years later and was strongly associated with mortality due to diabetes, pneumonia and suicide. As the risk of death was most pronounced in the first two decades, earlier interventions to treat anorexia nervosa may have greatest potential for reducing harm. To improve survival and reduce morbidity, better documentation of the impact of anorexia nervosa over the life course is needed.