EMPIRICAL ARTICLE
Highly Increased Risk of Type 2 Diabetes in patients with Binge Eating Disorder and Bulimia Nervosa Anu Raevuori, MD, PhD1,2,3,4* Jaana Suokas, MD, PhD4,5 Jari Haukka, PhD1,4 Mika Gissler, MD, PhD6,7 Milla Linna, MD1 Marjut Grainger, Student4 Jaana Suvisaari, MD, PhD4,8
ABSTRACT Objective: We aimed to examine the prevalence and incidence of type 2 diabetes (T2D) in a large patient cohort treated for binge eating disorder (BED), bulimia nervosa (BN), and anorexia nervosa. Method: Patients (N 5 2,342) treated at the Eating Disorder Unit of Helsinki University Central Hospital over the period up to 16 years were compared with matched general population controls (N 5 9,368) in three stages: before entering to the treatment for an eating disorder, after the entrance until the end of the study period, and combined any time before, during, and after the treatment. The study population was linked with the oral TSD medication data of 17 years from The Medical Reimbursement Register. Data were analyzed using conditional and Poisson regression models. Results: Before entering to the treatment for eating disorders, the risk of T2D was substantially increased in patients compared with controls (OR 6.6, 95% CI
Introduction Eating disorders are often associated with significant physical morbidity, which frequently relates to Accepted 6 July 2014 Supported by grant 259764 (to A.R.) from Academy of Finland and by Finnish Cultural Foundation grant. All other authors indicate no potential conflicts of interest. *Correspondence to: Anu Raevuori, Department Public Health, Hjelt Institute, PO Box 41 (Mannerheimintie 172), University of Helsinki, FIN-00014 Helsinki, Finland. E-mail: [email protected] 1 Department of Public Health, Hjelt Institute, University of Helsinki, Finland 2 Department of Adolescent Psychiatry, Helsinki University Central Hospital, Helsinki, Finland 3 Department of Child Psychiatry, Faculty of Medicine, University of Turku, Finland 4 Department of Mental Health and Substance Abuse Services, National Institute for Health and Welfare, Helsinki, Finland 5 Department of Psychiatry, Helsinki University Central Hospital, Finland 6 Information Department, National Institute for Health and Welfare, Helsinki, Finland 7 Nordic School of Public Health, Gothenburg, Sweden 8 Department of Social Psychiatry, Tampere School of Public Health, Finland Published online 25 July 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/eat.22334 C 2014 Wiley Periodicals, Inc. V
International Journal of Eating Disorders 48:6 555–562 2015
4.0–10.7). At the end of the study period, the lifetime prevalence of T2D was 5.2% among patients, 1.7% among controls (OR 3.4, 95% CI 2.6–4.4), and in male patients, it was significantly higher compared with females. Of those treated for BED, every third had T2D by the end of the study period (OR 12.9, 95% CI 7.4– 22.5), whereas the same was true for 4.4% of those with BN (OR 2.4, 95% CI 1.7–3.5). Discussion: Our findings provide strong support for the association between T2D and clinically significant binge eating. Disturbed glucose metabolism may contribute to the onset and maintenance of C 2014 Wiley Periodicals, BED and BN. V Inc. Keywords: eating disorders; diabetes mellitus type 2; diabetes mellitus; glucose intolerance (Int J Eat Disord 2014; 48:555–562)
their symptoms such as binge eating, purging behavior, starvation, weight fluctuations or obesity. On the other hand, both somatic and psychiatric comorbidity may provide clues for etiological features of these disorders. Association of eating disorders with diabetes is recognized in the scientific literature, but the majority of the studies1 focus on eating disorders in young females with type 1 diabetes (T1D). The far more common type 2 diabetes (T2D) along with the reverse phenomenon, diabetes in individuals with eating disorders has attracted substantially less research attention.2–4 T2D is characterized by an array of dysfunctions defined by hyperglycemia, that is, elevated blood sugar, resulting from the combination of resistance to the action of insulin hormone, inadequate insulin secretion, and inappropriate secretion of glucagon, an antagonist hormone to insulin. Individuals with both T2D and T1D may intentionally omit prescribed insulin for the purpose of weight control, and the current diagnostic manuals5,6 specify omitting/reducing insulin as a potential purging method in criteria for anorexia nervosa (AN) and/or bulimia nervosa (BN) in 555
RAEVUORI ET AL.
individuals with T1D. Respectively, individuals with T2D have been reported to intentionally cut down oral diabetes medications,7 resulting in poor glycemic control, and increased risk of long-term complications, such as vascular and neuropathic damage. Research suggests that up to 10–40% of individuals with T2D may meet the criteria for an eating disorder,8–11 binge eating disorder (BED) being the most common eating disorder, followed by BN.4,8,9,11 However, not all studies support such an elevated prevalence4,12 or any difference compared with age-, sex, and weight-matched nondiabetic population.13–15 Along with BED, also nighteating syndrome, characterized by excessive food consumption at night or after evening meal, has been reported to relate to T2D.14 In addition, binge-eating in individuals with T2D is common also in the absence of a diagnostic eating disorder, and is reported to associate with higher rates of extreme obesity, depressive symptoms, and impaired quality of life.16 Notably, binge eating appears to be an independent risk factor for T2D, evidence indicating that in vast majority of the cases, binge eating precedes the onset and is linked with significantly earlier age at the diagnosis of T2D.4,8,9 Tendency toward weight gain and deviant glucose metabolism associate with disordered eating,17 which in turn predisposes to clinical eating disorders.18 According to current understanding, development of these tendencies begin early in life through metabolic programming where both prenatal malnutrition as well as abundant nutrition have long-lasting adverse effects on offspring metabolism later in life.19,20 Evidence further suggests that BED may increase the risk of T2D, as well as other components of the metabolic syndrome, independently of obesity.2 Increased oxidative and inflammatory stress due to binge eating have been proposed as the underlying mechanisms.21 On the other hand, among some nondiabetic patients with AN, pretreatment glucose responses have been shown to be disturbed,22 and compared with AN patients with normal glucose response, those with disturbed response have been reported to have higher fear of eating and lower success in re-feeding. It has been assumed that in a subpopulation of AN patients, disturbance in a glucose metabolism is secondary to a disease process of AN.23 However, the impaired glucose tolerance might as well play an uncharacterized role in the onset of AN.22 Finally, substantially elevated rates of gestational diabetes24 and polycystic ovary syndrome25 have been observed in patients with BN, which further support these individuals’ vulnerability to T2D. 556
Only one large scale study of the comorbidity of T2D and eating disorders has been published: out of 16 mental disorders, de Jonge et al.4 reported that the strongest associations with subsequent T2D were observed for BED and BN. The study included diverse general population samples worldwide, and relied on diagnostic interviews on mental disorders and self-reported diagnosis of T2D. Of DSM-IV eating disorders, BN and eating disorders not otherwise specified/BED were included, but not AN. Underdiagnosing of T2D is, however, common even in Western countries, and in other parts of the world the situation is likely to be worse, which might have led to significant underestimates of the true prevalences of T2D. The magnitude of the risk may neither be generalizable from population-based samples to patient samples, which tend to include cases of longer duration and increased severity. In this study, we aimed to examine prevalence, incidence, and risk of T2D in a large Finnish patient cohort of individuals treated for BED, BN, and AN during 16 years’ time period. To separate the potential effect of the treatment system to detect the T2D among patients compared with controls (detection bias), and to assess new onset T2D cases among participants, we explored the potential associations in three stages. First, we assessed the period prevalence of T2D accounting for the time before treatment of an eating disorder; second, we assessed the incidence among those who were free of T2D after entering to eating disorder treatment until the end of the study period to yield the information of new onset cases of T2D. Finally, to provide an overview, we examined lifetime prevalence in the aggregate any time before, during, or after the treatment. We hypothesized that compared with control individuals, the risk of T2D would be increased in patients with BED in all stages; Increased in patients with BN after entering to the treatment, and decreased in those with AN in all three stages.
Method Study Participants We identified all patients (N 5 2,342) treated in the Eating Disorder Unit at the Helsinki University Central Hospital during January 1, 1995 to December 31, 2010. Most treatment was provided at the outpatient level, but day patient and inpatient treatments were also available. For each patient, four controls (N 5 9,368) matched for age, sex, and place of residence were selected from the Central Population Register. The individual follow-up International Journal of Eating Disorders 48:6 555–562 2015
TYPE 2 DIABETES IN EATING DISORDERS
period of each patient and the control individuals extended from the first day at treatment in the Eating Disorder Unit until the end of the study period or death (N 5 61 patients, N 5 69 controls). Eating disorder diagnosis was set by the attending psychiatrist at the Eating Disorder Clinic on the patient’s arrival using the ICD-10 criteria.6 The participants included patients with ICD-10 diagnoses of F50.0, F50.1, F50.2, F50.3, and F50.8 indicating AN, atypical AN, BN, atypical BN, and BED, respectively. As current definitions for eating disorders in DSM-55 have been relaxed compared with ICD-10 or DSM-IV criteria, we used broader definitions for AN and BN: AN (N 5 911 patients) included ICD-10 AN (F50.0) and ICD-10 atypical AN (F50.1). Atypical AN diagnosis was used for patients in whom one or more key features of AN, such as amenorrhea or weight loss leading to a body weight less than 85% of that expected, was absent but who otherwise presented a fairly typical clinical picture. BN (N 5 1,260 patients) included ICD-10 BN (F50.2) and ICD-10 atypical BN (F50.3). Atypical BN diagnosis was used in patients among whom one or more key criteria listed for BN were absent, but who otherwise present a fairly typical clinical picture. BED (N 5 171 patients), coded as F50.8, that is, “Other eating disorder” in ICD-10, was diagnosed in the Eating Disorder Unit according to the DSM-IV research criteria.
prescription was used as an outcome variable and logarithm of cumulated person-years as offset term. Followup was stopped at death that was treated as censoring event. Results are reported as rate ratios (RRs) with 95% confidence intervals (CIs). Period and lifetime prevalences were modeled using conditional logistic regression model, conditioned with the matching group. Period prevalence covered all existing and new onset cases before the treatment; the lifetime prevalence, respectively, included all cases for the whole study period. Odds ratios (ORs) with 95% CIs are reported. All analyses were carried out using R software.26 Ethical Considerations Ethical permits were obtained from the Ethics committee of National Institute of Health and Welfare (Dnro THL/184/6.02.00/2011), and the permission of the use of the data of The Social Security Institution was obtained. The study was conducted according to the Helsinki Declaration. Data handling was performed according to the Finnish data protection legislation and the rules of National Institute of Health and Welfare. The authors did not have access to the personal identification data.
Results General
Assessment of T2D This study utilized the registries of the Social Insurance Institution of Finland. The institution provides social security benefits to Finnish residents, including medical care. It maintains a Medical Reimbursement Register, which includes data on all redeemed reimbursed prescriptions. Individuals in these registries can be identified by a personal identification number, which allows for linkages between the registries. The study population was linked to the computerized data of oral (tablet) medication prescriptions against T2D (Anatomical Therapeutic Chemical [ATC] code A10B; blood glucose lowering drugs, excluding insulins) obtained from The Medical Reimbursement Register. These medication data were available from January 1st 1995 to December 31st 2011. Medical reimbursement covers all diabetes medications, and having at least one redeemed prescription of oral TSD medication was utilized as a proxy for T2D. The time of the onset of T2D was defined as the first day of the redeemed prescription of T2D medication. Statistical Analyses To assess the occurrence of new onset T2D cases since beginning of the eating disorder treatment, incidence of T2D was analyzed using Poisson regression model where occurrence of the first redeemed oral T2D medication International Journal of Eating Disorders 48:6 555–562 2015
Characteristics of the study sample by diagnostic groups are presented in the Table 1. Participants with BN made up the largest patient group (54%), followed by those with AN (39%), and those with BED (7%). The mean age of patients with eating disorders at the start of the follow-up differed significantly (p < .001) between diagnostic groups. Proportion of males across eating disorder categories also differed significantly from 2.3% in BN to 12.9% in BED (p < .001). Lifetime prevalence of T2D for the whole study period was higher in men compared with women across eating disorder categories and among their control groups (Table 1), which is in line with Finnish epidemiological data of T2D.27 However, the respective sex differences appeared substantially larger among patient groups compared with control groups. Lifetime prevalences of T2D in control groups also comply with epidemiological data; steeply rising total prevalence from 0.9% in AN control group to 4.1% in BED control group is explained by the increasing mean age of the control participants across groups. At the onset of the treatment for eating disorders, mean ages of patients with both an ED and T2D were significantly higher compared with overall mean ages among patients in each eating disorder group 557
RAEVUORI ET AL. TABLE 1.
Study sample by diagnostic groups with lifetime prevalence figures of T2D at the end of the study period
Women: Mean age, years (SD, range) Men: Mean age, years (SD, range) Total: Mean age, years (SD, range) Total N participants with ED (% within a diagnostic group) N (%) participants ever treated as inpatient for ED Participants with ED and T2D (N, positive/ negative for T2D, % positive) Matched control participants with T2D (N, positive/negative for T2D, % positive)
Women Men Total (women and men) Women: Mean age, years (SD, range) Men: Mean age, years (SD, range) Total Women Men Total
Anorexia Nervosa (N 5 911)
Bulimia Nervosa (N 5 1,260)
Binge Eating Disorder (N 5 171)
F50.0 Anorexia Nervosa
F50.2 Bulimia Nervosa
F50.1 Atypical Anorexia Nervosa
F50.3 Atypical Bulimia Nervosa
F50.8 Other Eating Disorder, DSM-IV BED
23.6 (6.7, 15.9–64.8) 26.2 (7.7, 18.0–49.1) 23.8 (6.8, 15.9–64.8) 867 (95.2%) 44 (4.8%)
26.0 (7.3, 17.2–62.2) 32.8 (12.3, 18.2–64.5) 26.2 (7.5, 17.2–64.6) 1,231 (97.7%) 29 (2.3%)
36.5 (10.4, 17.8–64.2) 41.1 (11.1, 19.5–61.3) 37.0 (10.6, 17.8–64.2) 149 (87.1%) 22 (12.9%)
322 (35.3%)
134 (10.6%)
7 (4.1%)
5/862 (0.6%) 36.9 (12.6, 20.2–50.1) 2/42 (4.5%) 46.6 (3.6, 44.0–49.1) 7/904 (0.8%) 31/3,437 (0.9%) 3/173 (1.7%) 34/3,610 (0.9%)
51/1,180 (4.3%) 38.0 (10.6, 18.4–62.3) 5/24 (17.2%) 44.7 (14.5, 27.2–64.5) 56/1,204 (4.4%) 96/4,828 (1.9%) 5/111 (4.3%) 101/4,939 (2.0%)
46/103 (30.9%) 41.2 (11.2, 20.9–64.2) 12/10 (54.5%) 46.2 (8.5, 35.7–61.3) 58/113 (33.9 %) 21/575 (3.5%) 7/81 (8.0%) 28/656 (4.1%)
Notes: ED, eating disorder; T2D, type 2 diabetes. Age information from participants is from the beginning of the treatment at the Eating Disorder Unit at the Helsinki University Central Hospital. TABLE 2. Prevalence status and conditional logistic regression models of T2D in the cohort at the entrance to the Eating Disorder Unit, that is, period prevalence accounting for the time period before beginning of the treatment and lifetime prevalence at the end covering the whole study period N (Positive/Negative for T2D, % Positive) Patients
Controls
Period prevalence accounting for the time before the treatment for an eating disorder All eating disorders Total 45/2,297 (1.9%) 31/9,337 (0.3%) Anorexia nervosa Total 1/910 (0.1%) 3/3,641 (0.08%) Bulimia nervosa Total 18/1,242 (1.5%) 13/5,027 (0.3%) Binge eating disorder Total 26/145 (15.2%) 15/669 (2.2%) Lifetime prevalence covering the whole study period All eating disorders Total 121/2,221 (5.2%) 163/9,205 (1.7%) Women 102/2,145 (4.5%) 148/8,840 (1.6%) Men 19/76 (20.0%) 15/365 (3.9%) Anorexia nervosa Total 7/904 (0.8%) 34/3,610 (0.9%) Women 5/862 (0.6%) 31/3,437 (0.9%) Men 2/42 (4.5%) 3/173 (1.7%) Bulimia nervosa Total 56/1,204 (4.4%) 101/4,939 (2.0%) Women 51/1,180 (4.1%) 96/4,828 (1.9%) Men 5/24 (17.2%) 5/111 (4.3%) Binge eating disorder Total 58/113 (33.9%) 28/656 (4.1%) Women 46/103 (30.9%) 21/575 (3.5%) Men 12/10 (54.5%) 7/81 (8.0%)
OR (95% CI)
p Value
6.57 (4.04–10.7) 1.33 (0.14–12.80) 5.83 (2.80–12.10) 8.78 (4.32–17.90)
Highly Increased Risk of Type 2 Diabetes in patients with Binge Eating Disorder and Bulimia Nervosa Anu Raevuori, MD, PhD1,2,3,4* Jaana Suokas, MD, PhD4,5 Jari Haukka, PhD1,4 Mika Gissler, MD, PhD6,7 Milla Linna, MD1 Marjut Grainger, Student4 Jaana Suvisaari, MD, PhD4,8
ABSTRACT Objective: We aimed to examine the prevalence and incidence of type 2 diabetes (T2D) in a large patient cohort treated for binge eating disorder (BED), bulimia nervosa (BN), and anorexia nervosa. Method: Patients (N 5 2,342) treated at the Eating Disorder Unit of Helsinki University Central Hospital over the period up to 16 years were compared with matched general population controls (N 5 9,368) in three stages: before entering to the treatment for an eating disorder, after the entrance until the end of the study period, and combined any time before, during, and after the treatment. The study population was linked with the oral TSD medication data of 17 years from The Medical Reimbursement Register. Data were analyzed using conditional and Poisson regression models. Results: Before entering to the treatment for eating disorders, the risk of T2D was substantially increased in patients compared with controls (OR 6.6, 95% CI
Introduction Eating disorders are often associated with significant physical morbidity, which frequently relates to Accepted 6 July 2014 Supported by grant 259764 (to A.R.) from Academy of Finland and by Finnish Cultural Foundation grant. All other authors indicate no potential conflicts of interest. *Correspondence to: Anu Raevuori, Department Public Health, Hjelt Institute, PO Box 41 (Mannerheimintie 172), University of Helsinki, FIN-00014 Helsinki, Finland. E-mail: [email protected] 1 Department of Public Health, Hjelt Institute, University of Helsinki, Finland 2 Department of Adolescent Psychiatry, Helsinki University Central Hospital, Helsinki, Finland 3 Department of Child Psychiatry, Faculty of Medicine, University of Turku, Finland 4 Department of Mental Health and Substance Abuse Services, National Institute for Health and Welfare, Helsinki, Finland 5 Department of Psychiatry, Helsinki University Central Hospital, Finland 6 Information Department, National Institute for Health and Welfare, Helsinki, Finland 7 Nordic School of Public Health, Gothenburg, Sweden 8 Department of Social Psychiatry, Tampere School of Public Health, Finland Published online 25 July 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/eat.22334 C 2014 Wiley Periodicals, Inc. V
International Journal of Eating Disorders 48:6 555–562 2015
4.0–10.7). At the end of the study period, the lifetime prevalence of T2D was 5.2% among patients, 1.7% among controls (OR 3.4, 95% CI 2.6–4.4), and in male patients, it was significantly higher compared with females. Of those treated for BED, every third had T2D by the end of the study period (OR 12.9, 95% CI 7.4– 22.5), whereas the same was true for 4.4% of those with BN (OR 2.4, 95% CI 1.7–3.5). Discussion: Our findings provide strong support for the association between T2D and clinically significant binge eating. Disturbed glucose metabolism may contribute to the onset and maintenance of C 2014 Wiley Periodicals, BED and BN. V Inc. Keywords: eating disorders; diabetes mellitus type 2; diabetes mellitus; glucose intolerance (Int J Eat Disord 2014; 48:555–562)
their symptoms such as binge eating, purging behavior, starvation, weight fluctuations or obesity. On the other hand, both somatic and psychiatric comorbidity may provide clues for etiological features of these disorders. Association of eating disorders with diabetes is recognized in the scientific literature, but the majority of the studies1 focus on eating disorders in young females with type 1 diabetes (T1D). The far more common type 2 diabetes (T2D) along with the reverse phenomenon, diabetes in individuals with eating disorders has attracted substantially less research attention.2–4 T2D is characterized by an array of dysfunctions defined by hyperglycemia, that is, elevated blood sugar, resulting from the combination of resistance to the action of insulin hormone, inadequate insulin secretion, and inappropriate secretion of glucagon, an antagonist hormone to insulin. Individuals with both T2D and T1D may intentionally omit prescribed insulin for the purpose of weight control, and the current diagnostic manuals5,6 specify omitting/reducing insulin as a potential purging method in criteria for anorexia nervosa (AN) and/or bulimia nervosa (BN) in 555
RAEVUORI ET AL.
individuals with T1D. Respectively, individuals with T2D have been reported to intentionally cut down oral diabetes medications,7 resulting in poor glycemic control, and increased risk of long-term complications, such as vascular and neuropathic damage. Research suggests that up to 10–40% of individuals with T2D may meet the criteria for an eating disorder,8–11 binge eating disorder (BED) being the most common eating disorder, followed by BN.4,8,9,11 However, not all studies support such an elevated prevalence4,12 or any difference compared with age-, sex, and weight-matched nondiabetic population.13–15 Along with BED, also nighteating syndrome, characterized by excessive food consumption at night or after evening meal, has been reported to relate to T2D.14 In addition, binge-eating in individuals with T2D is common also in the absence of a diagnostic eating disorder, and is reported to associate with higher rates of extreme obesity, depressive symptoms, and impaired quality of life.16 Notably, binge eating appears to be an independent risk factor for T2D, evidence indicating that in vast majority of the cases, binge eating precedes the onset and is linked with significantly earlier age at the diagnosis of T2D.4,8,9 Tendency toward weight gain and deviant glucose metabolism associate with disordered eating,17 which in turn predisposes to clinical eating disorders.18 According to current understanding, development of these tendencies begin early in life through metabolic programming where both prenatal malnutrition as well as abundant nutrition have long-lasting adverse effects on offspring metabolism later in life.19,20 Evidence further suggests that BED may increase the risk of T2D, as well as other components of the metabolic syndrome, independently of obesity.2 Increased oxidative and inflammatory stress due to binge eating have been proposed as the underlying mechanisms.21 On the other hand, among some nondiabetic patients with AN, pretreatment glucose responses have been shown to be disturbed,22 and compared with AN patients with normal glucose response, those with disturbed response have been reported to have higher fear of eating and lower success in re-feeding. It has been assumed that in a subpopulation of AN patients, disturbance in a glucose metabolism is secondary to a disease process of AN.23 However, the impaired glucose tolerance might as well play an uncharacterized role in the onset of AN.22 Finally, substantially elevated rates of gestational diabetes24 and polycystic ovary syndrome25 have been observed in patients with BN, which further support these individuals’ vulnerability to T2D. 556
Only one large scale study of the comorbidity of T2D and eating disorders has been published: out of 16 mental disorders, de Jonge et al.4 reported that the strongest associations with subsequent T2D were observed for BED and BN. The study included diverse general population samples worldwide, and relied on diagnostic interviews on mental disorders and self-reported diagnosis of T2D. Of DSM-IV eating disorders, BN and eating disorders not otherwise specified/BED were included, but not AN. Underdiagnosing of T2D is, however, common even in Western countries, and in other parts of the world the situation is likely to be worse, which might have led to significant underestimates of the true prevalences of T2D. The magnitude of the risk may neither be generalizable from population-based samples to patient samples, which tend to include cases of longer duration and increased severity. In this study, we aimed to examine prevalence, incidence, and risk of T2D in a large Finnish patient cohort of individuals treated for BED, BN, and AN during 16 years’ time period. To separate the potential effect of the treatment system to detect the T2D among patients compared with controls (detection bias), and to assess new onset T2D cases among participants, we explored the potential associations in three stages. First, we assessed the period prevalence of T2D accounting for the time before treatment of an eating disorder; second, we assessed the incidence among those who were free of T2D after entering to eating disorder treatment until the end of the study period to yield the information of new onset cases of T2D. Finally, to provide an overview, we examined lifetime prevalence in the aggregate any time before, during, or after the treatment. We hypothesized that compared with control individuals, the risk of T2D would be increased in patients with BED in all stages; Increased in patients with BN after entering to the treatment, and decreased in those with AN in all three stages.
Method Study Participants We identified all patients (N 5 2,342) treated in the Eating Disorder Unit at the Helsinki University Central Hospital during January 1, 1995 to December 31, 2010. Most treatment was provided at the outpatient level, but day patient and inpatient treatments were also available. For each patient, four controls (N 5 9,368) matched for age, sex, and place of residence were selected from the Central Population Register. The individual follow-up International Journal of Eating Disorders 48:6 555–562 2015
TYPE 2 DIABETES IN EATING DISORDERS
period of each patient and the control individuals extended from the first day at treatment in the Eating Disorder Unit until the end of the study period or death (N 5 61 patients, N 5 69 controls). Eating disorder diagnosis was set by the attending psychiatrist at the Eating Disorder Clinic on the patient’s arrival using the ICD-10 criteria.6 The participants included patients with ICD-10 diagnoses of F50.0, F50.1, F50.2, F50.3, and F50.8 indicating AN, atypical AN, BN, atypical BN, and BED, respectively. As current definitions for eating disorders in DSM-55 have been relaxed compared with ICD-10 or DSM-IV criteria, we used broader definitions for AN and BN: AN (N 5 911 patients) included ICD-10 AN (F50.0) and ICD-10 atypical AN (F50.1). Atypical AN diagnosis was used for patients in whom one or more key features of AN, such as amenorrhea or weight loss leading to a body weight less than 85% of that expected, was absent but who otherwise presented a fairly typical clinical picture. BN (N 5 1,260 patients) included ICD-10 BN (F50.2) and ICD-10 atypical BN (F50.3). Atypical BN diagnosis was used in patients among whom one or more key criteria listed for BN were absent, but who otherwise present a fairly typical clinical picture. BED (N 5 171 patients), coded as F50.8, that is, “Other eating disorder” in ICD-10, was diagnosed in the Eating Disorder Unit according to the DSM-IV research criteria.
prescription was used as an outcome variable and logarithm of cumulated person-years as offset term. Followup was stopped at death that was treated as censoring event. Results are reported as rate ratios (RRs) with 95% confidence intervals (CIs). Period and lifetime prevalences were modeled using conditional logistic regression model, conditioned with the matching group. Period prevalence covered all existing and new onset cases before the treatment; the lifetime prevalence, respectively, included all cases for the whole study period. Odds ratios (ORs) with 95% CIs are reported. All analyses were carried out using R software.26 Ethical Considerations Ethical permits were obtained from the Ethics committee of National Institute of Health and Welfare (Dnro THL/184/6.02.00/2011), and the permission of the use of the data of The Social Security Institution was obtained. The study was conducted according to the Helsinki Declaration. Data handling was performed according to the Finnish data protection legislation and the rules of National Institute of Health and Welfare. The authors did not have access to the personal identification data.
Results General
Assessment of T2D This study utilized the registries of the Social Insurance Institution of Finland. The institution provides social security benefits to Finnish residents, including medical care. It maintains a Medical Reimbursement Register, which includes data on all redeemed reimbursed prescriptions. Individuals in these registries can be identified by a personal identification number, which allows for linkages between the registries. The study population was linked to the computerized data of oral (tablet) medication prescriptions against T2D (Anatomical Therapeutic Chemical [ATC] code A10B; blood glucose lowering drugs, excluding insulins) obtained from The Medical Reimbursement Register. These medication data were available from January 1st 1995 to December 31st 2011. Medical reimbursement covers all diabetes medications, and having at least one redeemed prescription of oral TSD medication was utilized as a proxy for T2D. The time of the onset of T2D was defined as the first day of the redeemed prescription of T2D medication. Statistical Analyses To assess the occurrence of new onset T2D cases since beginning of the eating disorder treatment, incidence of T2D was analyzed using Poisson regression model where occurrence of the first redeemed oral T2D medication International Journal of Eating Disorders 48:6 555–562 2015
Characteristics of the study sample by diagnostic groups are presented in the Table 1. Participants with BN made up the largest patient group (54%), followed by those with AN (39%), and those with BED (7%). The mean age of patients with eating disorders at the start of the follow-up differed significantly (p < .001) between diagnostic groups. Proportion of males across eating disorder categories also differed significantly from 2.3% in BN to 12.9% in BED (p < .001). Lifetime prevalence of T2D for the whole study period was higher in men compared with women across eating disorder categories and among their control groups (Table 1), which is in line with Finnish epidemiological data of T2D.27 However, the respective sex differences appeared substantially larger among patient groups compared with control groups. Lifetime prevalences of T2D in control groups also comply with epidemiological data; steeply rising total prevalence from 0.9% in AN control group to 4.1% in BED control group is explained by the increasing mean age of the control participants across groups. At the onset of the treatment for eating disorders, mean ages of patients with both an ED and T2D were significantly higher compared with overall mean ages among patients in each eating disorder group 557
RAEVUORI ET AL. TABLE 1.
Study sample by diagnostic groups with lifetime prevalence figures of T2D at the end of the study period
Women: Mean age, years (SD, range) Men: Mean age, years (SD, range) Total: Mean age, years (SD, range) Total N participants with ED (% within a diagnostic group) N (%) participants ever treated as inpatient for ED Participants with ED and T2D (N, positive/ negative for T2D, % positive) Matched control participants with T2D (N, positive/negative for T2D, % positive)
Women Men Total (women and men) Women: Mean age, years (SD, range) Men: Mean age, years (SD, range) Total Women Men Total
Anorexia Nervosa (N 5 911)
Bulimia Nervosa (N 5 1,260)
Binge Eating Disorder (N 5 171)
F50.0 Anorexia Nervosa
F50.2 Bulimia Nervosa
F50.1 Atypical Anorexia Nervosa
F50.3 Atypical Bulimia Nervosa
F50.8 Other Eating Disorder, DSM-IV BED
23.6 (6.7, 15.9–64.8) 26.2 (7.7, 18.0–49.1) 23.8 (6.8, 15.9–64.8) 867 (95.2%) 44 (4.8%)
26.0 (7.3, 17.2–62.2) 32.8 (12.3, 18.2–64.5) 26.2 (7.5, 17.2–64.6) 1,231 (97.7%) 29 (2.3%)
36.5 (10.4, 17.8–64.2) 41.1 (11.1, 19.5–61.3) 37.0 (10.6, 17.8–64.2) 149 (87.1%) 22 (12.9%)
322 (35.3%)
134 (10.6%)
7 (4.1%)
5/862 (0.6%) 36.9 (12.6, 20.2–50.1) 2/42 (4.5%) 46.6 (3.6, 44.0–49.1) 7/904 (0.8%) 31/3,437 (0.9%) 3/173 (1.7%) 34/3,610 (0.9%)
51/1,180 (4.3%) 38.0 (10.6, 18.4–62.3) 5/24 (17.2%) 44.7 (14.5, 27.2–64.5) 56/1,204 (4.4%) 96/4,828 (1.9%) 5/111 (4.3%) 101/4,939 (2.0%)
46/103 (30.9%) 41.2 (11.2, 20.9–64.2) 12/10 (54.5%) 46.2 (8.5, 35.7–61.3) 58/113 (33.9 %) 21/575 (3.5%) 7/81 (8.0%) 28/656 (4.1%)
Notes: ED, eating disorder; T2D, type 2 diabetes. Age information from participants is from the beginning of the treatment at the Eating Disorder Unit at the Helsinki University Central Hospital. TABLE 2. Prevalence status and conditional logistic regression models of T2D in the cohort at the entrance to the Eating Disorder Unit, that is, period prevalence accounting for the time period before beginning of the treatment and lifetime prevalence at the end covering the whole study period N (Positive/Negative for T2D, % Positive) Patients
Controls
Period prevalence accounting for the time before the treatment for an eating disorder All eating disorders Total 45/2,297 (1.9%) 31/9,337 (0.3%) Anorexia nervosa Total 1/910 (0.1%) 3/3,641 (0.08%) Bulimia nervosa Total 18/1,242 (1.5%) 13/5,027 (0.3%) Binge eating disorder Total 26/145 (15.2%) 15/669 (2.2%) Lifetime prevalence covering the whole study period All eating disorders Total 121/2,221 (5.2%) 163/9,205 (1.7%) Women 102/2,145 (4.5%) 148/8,840 (1.6%) Men 19/76 (20.0%) 15/365 (3.9%) Anorexia nervosa Total 7/904 (0.8%) 34/3,610 (0.9%) Women 5/862 (0.6%) 31/3,437 (0.9%) Men 2/42 (4.5%) 3/173 (1.7%) Bulimia nervosa Total 56/1,204 (4.4%) 101/4,939 (2.0%) Women 51/1,180 (4.1%) 96/4,828 (1.9%) Men 5/24 (17.2%) 5/111 (4.3%) Binge eating disorder Total 58/113 (33.9%) 28/656 (4.1%) Women 46/103 (30.9%) 21/575 (3.5%) Men 12/10 (54.5%) 7/81 (8.0%)
OR (95% CI)
p Value
6.57 (4.04–10.7) 1.33 (0.14–12.80) 5.83 (2.80–12.10) 8.78 (4.32–17.90)
