FEW STUDIES HAVE reported the outcome of eating disorder patients with alcoholism, but some studies have shown that some anorexia nervosa patients with poor outcomes had alcoholism.1–3 According to previous studies, alcoholism may be a major factor in the death of anorexia nervosa patients. We previously reported on the clinical characteristics of eating disorder patients with alcoholism and concluded that their characteristics were more severe than those of alcoholics or patients with eating disorders alone.4–6 Our recent outcome study of eating disorder patients with alcoholism showed that the eating disorder symptoms resolved in a quarter of them and that a quarter had died by 5 years after inpatient treatment.7 However, the study followed up only 76% of the subjects and was considered to have low reliability in regard to the mortality of eating disorder patients with alcoholism. On the other hand, since several studies have reported high mortality rates for alcoholics,8–10 it became necessary to investigate whether the mortality of eating disorder patients with alcoholism is higher than that of alcoholics without eating disorders. In this study, we focused on the mortality of eating disorder patients with alcoholism in comparison with eating disorder patients without alcoholism and alcoholics without eating disorders.
Aims: This study was conducted to clarify the mortality of patients with eating disorders associated with alcoholism. We focused on the mortality rate 6 years after inpatient treatment of patients with eating disorders associated with alcoholism compared with eating disorder patients without alcoholism and alcoholic patients without eating disorders.
Methods: The subjects were 164 female Japanese patients 30 years of age or younger with eating disorders or alcoholism who were inpatients at some time during the period from 1990 to 1998 at the Japanese National Hospital Organization, Kurihama Alcoholism Center. A semi-structured interview concerning alcohol problems, eating problems, psychiatric disorders and other clinical characteristics was conducted at the time of the first admission. A survey concerning survival was conducted in October 2001, and 100% of the patients were followed up.
Results: The mortality of the 47 eating disorder patients with alcoholism, 86 eating disorder patients without alcoholism, and 31 alcoholics without eating disorders was 27.7%, 3.5%, and 19.4%, respectively, at 6 years after inpatient treatment, showing significant differences. On the Kaplan–Meier survival curves, the mortality of the eating disorder patients with alcoholism was significantly higher than that of the patients without alcoholism, but not significantly higher than that of young female alcoholics without eating disorders. The 13 eating disorder patients with alcoholism who had died were five anorexia nervosa and seven bulimia nervosa patients.
Conclusion: The results of the study suggest that comorbid alcoholism is a major factor in the death of anorexia nervosa and bulimia nervosa patients.
The National Hospital Organization, Kurihama Alcoholism Center is the national research and treatment center for alcoholism in Japan, and has a treatment unit for eating disorders. The structure and treatment systems of the hospital for alcoholism and eating disorders are briefly described elsewhere.8,11 The subjects of this study were 164 female patients 30 years of age or younger with eating disorders or alcoholism who had first been examined as inpatients at the Kurihama Alcoholism Center between 1990 and 1998. All of them were subjected to semi-structured interviews concerning alcoholism, eating problems, other psychiatric disorders, and clinical characteristics by two of the authors (K.S., A.T.) at the time of the admissions. We used DSM-IV12 criteria for alcohol abuse and dependence (which in this article will simply be referred to as alcoholism), eating disorders, and other psychiatric disorders in this study. We used DSM-III-R13 criteria for those disorders in 1990–1995, and we re-diagnosed them based on DSM-IV criteria in this study. The subjects included 47 eating disorder patients with alcoholism (EDAL), 86 eating disorder patients without alcoholism (ED), and 31 alcoholics without eating disorders (AL). Because the mean ages of the three groups of patients were very different, the subjects were limited to patients 30 years of age or younger to reduce the differences.
The clinical characteristics of the EDAL group, ED group, and AL group at the time of the first inpatient treatment were compared by the following statistical analysis in Table 1. Mean age at the time of the first admission in the EDAL group, ED group, and AL group was 25.7 years (SD 3.5), 20.8 years (SD 4.1), and 27.0 years (SD 2.4), respectively, and the differences were significant (anova, F = 45.2, d.f. = 2/161, P < 0.001) even though all of the subjects were under 30 years of age. The mean age of onset of the eating disorders in the EDAL group was significantly higher than that in the ED group (t = 1.93, d.f. = 131, P < 0.05), but the mean age of onset of alcoholism in the EDAL group was similar to that in the AL group. The results for marital status showed more marriages and divorces in the EDAL group and AL group than in the ED group, and the difference was significant (Kruskal–Wallis test, corrected H = 39.2, d.f. = 2, P < 0.001). Living situation was not different among the three groups. In terms of occupational status, the ED group contained more students and fewer housewives than the other two groups, there were more unemployed patients in the EDAL group than in the other two groups, and the distributions according to social status were significantly different (H = 7.0, d.f. = 2, P < 0.05).
|ED group||EDAL group||AL group||Significance|
|(n = 86)||(n = 47)||(n = 31)|
|Mean age at first admission (years)||20.8 (4.1)†||25.7 (3.5)||27.0 (2.4)||anova, F = 45.2, d.f. = 2/161, P < 0.001|
|Mean age of onset of eating disorders (years)||17.0 (3.5)||18.3 (4.0)||–||ED vs EDAL, t = 1.93, d.f. = 131, P < 0.05|
|Mean age of onset of alcoholism (years)||–||22.4 (3.4)||23.5 (3.6)||NS|
|Mean body mass index at first admission||18.0 (4.1)||18.8 (4.0)||20.2 (2.0)||anova, F = 3.7, P < 0.05|
|Mean maximum body mass index||22.4 (3.5)||22.6 (3.2)||–||NS|
|Mean minimum body mass index||14.9 (2.6)||15.3 (2.5)||–||NS|
|Diagnosis of eating disorders at first admission (%)||ED vs EDAL, χ2 = 13.6, d.f. = 4, P < 0.01|
|Anorexia nervosa, restricting type||24.4||4.3||–|
|Anorexia nervosa, binge–purging type||26.7||25.5||–|
|Bulimia nervosa, not purging type||9.3||6.4||–|
|Bulimia nervosa, purging type||33.8||44.7||–|
|Eating disorders, not otherwise specified||5.8||19.1||–|
|Symptoms of eating disorders (%)|
|Binge eating more than twice a week||67.4||76.6||–||NS|
|Vomiting more than twice a week||47.7||68.1||–||ED vs EDAL, χ2 = 5.11, d.f. = 1, P < 0.05|
|Laxative abuse more than twice a week||26.7||17||–||NS|
|Other psychiatric disorders (%)|
|Anxiety disorders||36||17||16.1||H = 7.8, d.f. = 2, P < 0.05|
|Presence of any personality disorders||61.6||83||38.7||H = 15.9, d.f. = 2, P < 0.0005|
|Borderline personality disorder||24.4||51.1||19.4||H = 12.5, d.f. = 2, P < 0.005|
|History of substance abuse||5.8||27.7||32.3||H = 16.4, d.f. = 2, P < 0.0005|
|Marital status (%)||H = 39.2, d.f. = 2, P < 0.0001|
|Living situation (%)||NS|
|With husband or cohabitant||3.6||23.4||32.3|
|Occupational status (%)||H = 7.0, d.f. = 2, P < 0.05|
The three groups had various comorbid psychiatric disorders. More of the patients in the ED groups had anxiety disorders than in the EDAL group or the AL group, and the differences were significant (H = 7.8, d.f. = 2, P < 0.05). There were significantly more patients with personality disorders in the EDAL group than in the ED group or the AL group (H = 15.9, d.f. = 2, P < 0.0005), and the EDAL group also contained significantly more patients with borderline personality disorders than the other two groups (H = 12.5, d.f. = 2, P < 0.005). Significantly more patients in the AL group and EDAL group had histories of substance abuse than in the ED group (H = 16.4, d.f. = 2, P < 0.0005).
The mean body mass index (BMI) at the time of the first admission of the AL group was significantly higher than in the EDAL group or ED group (anova, F = 3.7, P < 0.05). The ED group and EDAL group had similar mean maximum BMI and minimum BMI values. The diagnoses of eating disorders at the time of the first admission revealed more patients with bulimia nervosa binge–purging type and more eating disorders not otherwise specified, and fewer with anorexia nervosa restricting in the EDAL group than in the ED group, and the differences in the distribution of the subtypes of eating disorders was significant (χ2 = 13.6, d.f. = 4, P < 0.01). Abnormal eating behaviors were similar in the EDAL group and ED group, but more patients in the EDAL group vomited more than twice a week than in the ED group, and the difference was significant (χ2 = 5.11, d.f. = 1, P < 0.05).
The mortality survey of the 164 subjects was conducted in October 2001 by telephone with questions limited to whether the subjects had survived or died and the cause of death. We were unable to reach 21 patients by telephone, and six patients had died before the survey, but the cause of death was unknown. Since we obtained information on survival and the cause of death (if the patients had died) of these 27 patients from local legal affairs offices with the approval of the Japanese Ministry of Justice, we were able to follow up 100% of the 164 subjects. Twenty-two of the 164 subjects had died by the time of follow up. The mean follow-up period of the 142 living subjects was 5.9 (SD 2.7) years.
We compared the clinical characteristics among the three groups using anova for continuous variables, χ2-test for two categorical variables, and Kruskal–Wallis test for three categorical variables. P < 0.05 was considered statistically significant. Then we compared mortality among the three groups using Kaplan–Meier survival curves and Cox proportional hazard analysis. A Cox proportional hazard analysis adjusted for age at admission was used to calculate hazards ratios for mortality risk among the three groups. The protocol for the research project was approved by the Ethics Committee of the Kurihama Alcoholism Center in 2001, and it conforms with the provisions of the Declaration of Helsinki 1995 (as revised in Edinburgh 2000).
Table 2 compares mortality at follow up between the EDAL group, ED group, and AL group. Mortality in the EDAL group, ED group, and AL group was 27.7%, 3.5%, and 19.4%, respectively. Kruskal–Wallis test with post hoc comparison by U-test with Bonferroni's correction revealed that the mortality rate in the EDAL group was significantly greater than those in the ED and AL groups (corrected H = 41.4, d.f. = 2, P < 0.0001).
|ED group||EDAL group||AL group|
|(n = 86)||(n = 47)||(n = 31)|
|Mean follow-up period (living subjects only: years)||5.4 (2.9)||6.9 (2.7)||6.3 (1.8)|
|Number of deaths||3||13||6|
|Mortality rate (%)†||3.5||27.7||19.4|
|Mean age at death (years)||22.7 (2.5)||30.2 (4.8)||32.2 (2.9)|
|Mean interval between first admission to death (years)||4.3 (3.2)||3.5 (2.9)||3.7 (2.4)|
|Numbers of patients according to cause of death|
|Eating disorder-related disorder||0||1||–|
The cause of death of the 13 EDAL patients was an alcohol-related disorder in six, suicide in six, and an eating disorder-related disorder in one who was choked with a large amount of food by overeating, while among the three ED patients, the cause was suicide in two and heart disease in one, and the cause of death in all six AL patients was an alcohol-related disorder. The eating disorders of the 13 patients in the EDAL group who died had been diagnosed as anorexia nervosa restricting type in one, anorexia nervosa binge–purging type in four, bulimia nervosa not purging type in one, bulimia nervosa purging type in six, and an eating disorder not otherwise specified in one at the time of the first admission; their distribution did not significantly differ from their distribution in the 34 patients in the EDAL group who had survived. Nor were there any differences in other clinical characteristics, such as mean age at the time of the first admission, and mean age at the onset of the eating disorder between patients in the EDAL group who had died and survived.
As shown in Figure 1, Kaplan–Meier curves were computed to compare the three groups in regard to the probability of survival after the first admission. The proportional hazard model revealed statistically significant hazard ratios among the three groups with age at the time of admission as the covariate (χ2 = 12.6, d.f. = 2, P < 0.002). The age-adjusted hazard ratios between the EDAL group and the ED group and between the EDAL group and the AL group were 2.53 (95% confidence range: 1.32–5.61) and 1.23 (95% confidence range: 0.76–2.10), respectively.
This is the first report of a study on the mortality of patients with eating disorders associated with alcoholism, and the 100% follow-up study of large groups showed very high mortality for such disorders. The mortality rate of the EDAL group was not only higher than that in the ED group, but higher than that in the AL group. The results of the study demonstrate that the EDAL group is an ominous subgroup of patients with eating disorders.
The subjects of this study were representative of young Japanese women with alcoholism and eating disorders, because the clinical characteristics of the EDAL group and AL group were similar to those in our previous studies5–7,11 and the clinical characteristics of the ED group were similar to those of Japanese women with eating disorders.14 We previously discussed the reasons for the lower percentage of anorexia nervosa restricting type, higher percentage of unemployment, and high prevalence of borderline personality disorder in the EDAL group.5
Several outcome studies on eating disorders have reported higher mortality for anorexia nervosa than for bulimia nervosa.2,3,15 A review on the mortality of eating disorder patients described an adjusted mortality rate for anorexia nervosa of 0.56% a year,16 and two recent studies reported similar adjusted mortality rates.17,18 One report stated that 16.7% of anorexia nervosa patients had died over a period of 21 years.19 Over half of the EDAL group in this study were patients with bulimia nervosa, and their mortality during the 6-year follow up was 28%: a higher mortality than previously reported for anorexia nervosa. Several studies have reported that many anorexia nervosa patients who had died had had alcoholism or substance use disorders.1–3 No reports have shown any relationship between alcoholism and mortality in bulimia nervosa. In this study, 35.7% of the anorexia nervosa patients and 29.2% of the bulimia nervosa patients in the EDAL group had died by the time of the follow up, and the crude mortality rate was similar in both groups. The results of the study suggested that comorbid alcoholism is a major factor in the death of anorexia nervosa and bulimia nervosa patients.
Alcoholism is generally associated with high mortality. Annual adjusted mortality from alcoholism is said to be 1.6–3.7% in one review,9 and an annual adjusted mortality rate of 4.1% has been reported for female Japanese alcoholics.8 Compared to the previous reports, the mortality rate of the EDAL group in this study was slightly higher than that of female alcoholics in general.
Why is the mortality of the EDAL group so much higher than that of other groups of patients with eating disorders? Several explanations have been proposed to answer this question. First, the EDAL group had severe malnutrition and dehydration caused by their thinness and abnormal eating behaviors, which promoted liver cirrhosis and pancreatitis induced by alcohol. Two of the 13 EDAL patients died from liver cirrhosis or pancreatitis. Second, ‘the sudden alcoholic death syndrome’,20 which is an alcohol-related disorder, occurred among the EDAL group. The sudden alcoholic death syndrome is manifested by hypoglycemia, metabolic acidosis, and kidney hypofunction induced by continuous heavy drinking.20 These symptoms were promoted by abnormal eating behaviors, such as self-induced vomiting and laxative abuse among the EDAL group patients. Four of the 13 EDAL patients probably died of sudden alcoholic death syndrome. The third explanation for the high mortality in the EDAL group is the high percentage of comorbid borderline personality disorder, which is often manifested by impulsive, self-destructive and suicidal behavior. Six of the 13 deceased EDAL patients had died of suicide. One more explanation for the high mortality in the EDAL group is that many patients of the EDAL group lost parental protection because they had left their families. The EDAL group exhibited social behavior similar to that of the AL group, and had high marriage and divorce rates. On the other hand, many in the ED group lived with their parents at the time of the first admission and also at the time of the follow up, and they received parental protection.
There are several limitations to the study. The first limitation is that age, duration of eating disorders and the high percentage of comorbid borderline personality disorder of the EDAL group were not controlled. The second limitation is that the number of subjects was small and the results were obtained in only one hospital. The third limitation is that the Kurihama Alcoholism Center is the national center for alcoholism, not for eating disorders. Therefore, there might be a bias for patients with alcoholism and/or eating disorders. In the future, a multicenter study on a larger number of subjects will be necessary to investigate the mortality of patients with eating disorders associated with alcoholism.
This research was supported by a Grant for Research on Psychiatric and Nervous Diseases from the Japanese Ministry of Health, Labor and Welfare (11 SHI-8, 2000).