Hiroyuki Kimura, MD, Department of Psychiatry, Nagoya University Graduate School of Medicine, Nagoya, Aichi 466-8550, Japan. Email: email@example.com
Abstract The clinical characteristics differentiating late-onset anorexia nervosa (AN) from typical pubertal onset AN remain unclear. The purpose of the present study was to examine these differences in a retrospective analysis. A total of 149 female AN patients was divided into two groups: a peak-onset AN group (n = 125) in which onset occurred between the ages of 15 and 24 years, and a late-onset AN group (n = 24) in which onset occurred at the age of ≥25 years. A logistic regression analysis was conducted with this classification as the target variable and five clinical factors as explanatory variables for the clinical characteristics at the time of initial examination. Body mass index (BMI) at the time of presentation was identified as a possible factor affecting classification as peak-onset or late-onset AN. In addition, a negative linear correlation was detected between age of onset and BMI at the time of initial examination. The results suggest that BMI at the time of the initial examination is an important clinical characteristic to differentiate peak-onset AN and late-onset AN.
Since the time Morton first described eating disorders as a medical condition in the late 17th century,1 puberty has been considered the most common period for the onset of these disorders. For example, in the diagnostic criteria for eating disorders produced by Feighner et al. in 1972, age at onset <25 years was established as a diagnostic criterion.2 Therefore, the pathology of eating disorders has been hypothesized from the biological, physiological, and social changes that occur in puberty. However, the existence of late-onset anorexia nervosa (AN) has been documented.3 For example, case reports on late-onset AN have been seen sporadically since the 1970s,4–8 and compared with AN occurring in puberty (peak-onset AN).
Some authors have noted that late-onset AN is related to various separation experiences that occur in middle and old age (e.g. death of a spouse or independence of children). From an investigation of 50 cases of anorexia tardive with patients ranging in age from 21 to 80 years, Dally described a relationship between this condition and marriage, birth, and spouses, and emphasized its difference to peak-onset AN.9
Further findings on peak-onset and late-onset AN are described in the following. A very similar clinical picture of peak-onset AN and late-onset AN was reported in terms of time from onset until consultation with a clinician, weight at time of onset, maximum weight, minimum weight, and frequency of overeating and vomiting.10 In contrast another comparative study found that late-onset AN patients had lower weight at the time of initial examination than peak-onset AN patients.11
Other comparisons have indicated that late-onset AN patients have a worse prognosis than peak-onset patients.3,12 However, there is currently no consensus of opinion on the differences in clinical characteristics between peak-onset and late-onset AN.
The authors reported a patient with extremely low bodyweight (body mass index [BMI]: 10.8 kg/m2) at the time of presentation, and focused on mid-life crisis in four patients diagnosed with late-onset AN.13 This led us to question the general differences in clinical characteristics between peak-onset and late-onset AN at the time of initial examination.
Steinhausen reviewed the factors that are generally viewed to be clinical characteristics of eating disorders, and identified age at onset, period until treatment, period of treatment in hospital, amount of weight loss, vomiting, overeating and overuse of purgatives, developmental problems, parent–child relationship, chronicity, hysteric personality, and compulsive personality as factors affecting course and outcome.14
Thus, in the present study we conducted a retrospective comparison of the available clinical characteristics at the time of initial examination in cases of peak-onset and late-onset AN.
Four hundred women who were initially examined in the psychiatry department of Nagoya Daini Red Cross Hospital between 1 January 1998 and 31 December 2003, and who fulfilled the criteria for eating disorders on the DSM-IV,15 were potential subjects for the study. Four hundred of 406 patients identified during that period as having eating disorders were female. Because the overwhelming majority of such patients were female, the subjects of the present analysis were limited to women. After eliminating 251 bulimia nervosa patients, the remaining 149 AN patients were divided into two groups based on age of onset, and a comparison was conducted of these two groups. Although some studies had used an age of onset of >30 years for the diagnosis of late-onset eating disorders, in the present study the age of 25 years was adopted, the same as that used by Feighner et al.,2 Boast et al.,11 and others, and the subjects were divided into two groups: peak-onset AN (onset between 15 and 24 years), and late-onset AN (onset at age ≥25). Of the 149 female AN patients, 125 (age 15–24 years, mean age 17.8 ± 2.5 years) had peak-onset AN, and 24 (age 25–50 years, mean age 29.9 ± 5.9) had late-onset AN.
Patient information was obtained at the time of initial examination in the psychiatry department. Height and weight were not obtained by interview but by measurement on a bodyweight and height scale. BMI was calculated from a measurement of the height and weight. The difference between maximum and minimum BMI was obtained at the time of initial examination by asking the patient (and/or family member) her maximum and minimum bodyweight up to that time. The diagnosis and assessment at initial examination were made through a patient interview by a psychiatrist experienced in treating eating disorders. Although subclinical states on adolescence of late-onset AN was evaluated carefully, there is a limitation due to the hidden medical history on eating disorders.
According to factors suggested by Steinhausen, five factors including BMI, period from onset until the initial examination, subtypes of binge-eating/purging or restricting, difference between maximum and minimum BMI and birth order14 were extracted from medical records for the present study. Duration of illness was investigated by detailed clinical interviews with patients and their parents. The onset of illness was defined when patients first met the AN criteria in DSM-IV. The binge-eating/purging type and restricting type were diagnosed with DSM-IV criteria. Difference between maximum and minimum BMI was taken from the maximum BMI until the initial examination and minimum BMI during illness.
The present study was approved under the guideline for epidemiological studies by the Ethical Committee of Nagoya University Graduate School of Medicine.
Nagoya Daini Red Cross Hospital is located in west central Nagoya (population approx. 2.1 million). The hospital plays a central role in the community with a focus on emergency medical care. In the area of eating disorders, the hospital's psychiatry department conducts outpatient treatment with a focus on psychological education, hospital treatment for low bodyweight,16 and psychological interviews by a clinical psychotherapist. The hospital functions as a treatment center for eating disorders in the city of Nagoya.
The dependent variable was the classification of the two groups of peak-onset and late-onset AN. A logistic regression analysis was then carried out with five probable independent variables of BMI at the time of initial examination; period from onset until the initial examination; subtypes of binge-eating/purging or restricting; difference between maximum and minimum BMI; and birth order. All statistical analysis was done with SPSS for Windows version 11.5 (SPSS Japan, Tokyo, Japan).
The logistic regression analysis including five factors showed that BMI at the time of initial examination had the largest impact on classification as peak-onset AN or late-onset AN (odds ratio 1.26, P = 0.052). The other four factors were not significantly associated with their classification (period from onset until the initial examination: odds ratio 0.90, P = 0.27; subtype of binge-eating/purging or restricting: odds ratio 1.09, P = 0.89; difference between maximum and minimum BMI: odds ratio 0.93, P = 0.51; birth order: odds ratio 0.82, P = 0.67). BMI at the time of initial examination was close to being significantly lower in the late-onset AN group than in the peak-onset AN group (Table 1).
Table 1. Comparison of five factors in peak-onset and late-onset AN
AN, anorexia nervosa; BMI, body mass index; BMI (max–min), difference between maximum BMI until the initial examination and minimum BMI during illness; BMI (initial exam), BMI at the time of initial examination; period, periods from onset until the time of initial examination.
Mean age (years)
17.8 ± 2.5
29.9 ± 5.9
6.7 ± 2.5
7.2 ± 2.8
BMI (initial exam)
13.4 ± 1.9
12.6 ± 2.0
2.8 ± 3.7
2.2 ± 2.0
Because logistic regression indicated that BMI at the time of initial examination was a factor demarcating the two groups of peak-onset AN and late-onset AN, it was important to analyze the correlation between the age of onset and BMI at the time of initial examination for all patients. BMI was investigated with age at onset in a simple regression analysis. A significant correlation was seen between age at onset and BMI at the time of initial examination (r2 = 0.064, P = 0.0018), but the contribution of age at onset to BMI at the time of initial examination was approximately 6.0%. (Common regression formula: y = 15.6 − 0.12x; r2 = 0.064), where y is BMI at the time of initial examination and x is age at onset.) The number of patients with age at onset in the 40s and 50s was only one for each, and their BMI was considerably low among all patients; therefore the data for these two patients were excluded. Even though these two patients were excluded there was still a significant correlation (P = 0.026; Fig. 1).
The present study showed that BMI at the time of the initial examination had the largest influence on classification for peak-onset AN and late-onset AN among the five factors (BMI at the time of initial examination, period from onset until the initial examination, subtypes of binge-eating/purging or restricting, difference between maximum and minimum BMI, and birth order). Moreover, BMI decreased as age of onset increased. Thus, BMI at the time of the initial examination is an important clinical characteristic to differentiate peak-onset AN and late-onset AN.
The late-onset AN patients might not be able to face their painful intrapsychic anxiety in adolescence, and they might be in false adaptation by formulating a severe pathological denial for a long time. When it becomes impossible to maintain their denial they must suddenly face their accumulated anxiety. This situation might induce more rapid and larger BMI reduction in late-onset AN patients.
Boast et al. studied 186 AN patients including a peak-onset AN group with onset 15–19 years of age, and a late-onset AN group with onset >25 years.11 Their results were similar to the present findings, with the late-onset group having a lower bodyweight than the peak-onset group at presentation. According to several follow-up studies, the clinical significance of BMI at presentation has been shown to be a predictive factor for a poor outcome.17–19 Therefore, the low BMI at presentation in the current late-onset AN patients suggests the possibility that late-onset AN has a poorer prognosis than peak-onset AN, although this possibility needs to be investigated in future studies.
Joughin et al., in contrast, compared 427 AN patients consisting of an early onset group in which onset was at 15–19 years of age, and a late-onset group in which onset was after the age of 30, and reported a very similar clinical picture between the two groups.10 In their study the age of division between peak-onset and late-onset AN was different to the present one. In addition they used bodyweight rather than BMI at the time of initial examination, although simple bodyweight has a different meaning than BMI, which takes into account the factor of height.
Although the present study found no significant differences between the late-onset AN and peak-onset AN groups with regard to period from onset until the initial examination, binge-eating/purging type or restricting type, difference between maximum and minimum BMI, and birth order, some concerns should be taken into account.
First, Beck et al. reported that the time from onset to presentation of late-onset AN is longer than that for peak-onset AN due to poor insight into the disease.20 It has been reported that AN is not readily noticeable for a long period of time when the onset occurs after marriage.21 However, these studies lacked direct comparisons of peak-onset AN and late-onset AN groups. The present study involved both peak-onset and late-onset AN patients, and therefore had an advantage in being able to compare them within the study.
Second, recent studies have indicated no difference in terms of birth order in relation to AN,22 although several studies showed that AN is more prevalent among firstborns23 while others found that it is more prevalent in children born second or later.24 Regarding the relation between birth order and age of onset of AN, no difference was found even in a study that compared AN with onset before the age of 14 and AN with onset after that age.25 The present results provide further evidence that there is no significant difference in birth order between peak-onset AN and late-onset AN.
Interestingly, clinical features of late-onset AN are similar to those of peak-onset AN except for BMI at the time of initial examination. This may be related to the common psychopathology of late-onset AN and peak-onset AN.
In conclusion, BMI at the time of initial examination is an important clinical characteristic to differentiate peak-onset AN and late-onset AN. This was a retrospective study and was unable to adequately investigate outcome or to consider factors such as personality traits, early experience, and social support, which remain to be investigated.
The authors would like to express their appreciation to N. Sugimoto, Sanwa Kagaku Kenkyusho Contact Center, and Y. Shirasawa, Department of Internal Medicine and Pathophysiology, Nagoya City University Graduate School of Medical Sciences, for their assistance in statistical analysis.