|Home | About | Journals | Submit | Contact Us | Français|
Anorexia nervosa (AN) is a condition of severe undernutrition associated with altered regional fat distribution in females. Although primarily a disease of females, AN is increasingly being recognized in males and is associated with hypogonadism. Testosterone is a major regulator of body composition in males, and testosterone administration in adults decreases visceral fat. However, the effect of low testosterone and other hormonal alterations on body composition in boys with AN is not known.
We hypothesized that testosterone deficiency in boys with AN is associated with higher trunk fat, as opposed to extremity fat, compared with control subjects.
We assessed body composition using dual-energy X-ray absorptiometry and measured fasting testosterone, estradiol, insulin-like growth factor-1, leptin, and active ghrelin concentrations in 15 boys with AN and in 15 control subjects of comparable maturity aged 12–19 y.
Fat and lean mass in AN boys was 69% and 86% of that in control subjects. Percentage extremity fat and extremity lean mass were lower in boys with AN (P = 0.003 and 0.0008); however, percentage trunk fat and the trunk to extremity fat ratio were higher after weight was adjusted for (P = 0.005 and 0.003). Testosterone concentrations were lower in boys with AN, and, on regression modeling, positively predicted percentage extremity lean mass and inversely predicted percentage trunk fat and trunk to extremity fat ratio. Other independent predictors of regional body composition were bone age and weight.
In adolescent boys with AN, higher percentage trunk fat, higher trunk to extremity fat ratio, lower percentage extremity fat, and lower extremity lean mass (adjusted for weight) are related to the hypogonadal state.
Anorexia nervosa (AN) is a condition of severe undernutrition and is predominantly a disease in females. However, 5–15% of all patients with this disorder are male (1, 2). Although the disease is uniformly characterized by extreme low weight and loss of body fat, specific patterns of body fat distribution may depend on the underlying hormone deficiency syndromes in this disease. The effect of AN on body composition in adolescent females has been investigated (3, 4), particularly given the disproportionate increase in trunk fat in women recovering from this disorder (5). In addition to decreases in fat and lean mass, girls with AN have a lower percentage of trunk fat than do healthy girls, and this is predicted by concentrations of insulin-like growth factor-1 (IGF-1) (4). However, there are no data regarding regional body composition in males with AN.
Adolescence in boys is characterized by marked increases in lean mass (6)—a consequence of rising concentrations of testosterone, growth hormone (GH), and IGF-1 (7). In addition to an increase in lean mass (8–10), testosterone and GH cause a decrease in fat mass, particularly trunk fat (8, 11, 12). Therefore, low testosterone concentrations and GH deficiency may cause truncal adiposity. Boys with AN are hypogonadal and are likely to have low IGF-1 concentrations, similar to girls with AN (13). However, associations of these hormonal alterations with body-composition measures have not been assessed in boys with AN. In addition, recent data indicate that the orexigenic hormone ghrelin can impair gonadotropin secretion (14), and one study in elderly adults reported an inverse association of ghrelin with fat-free mass (15). We reported high ghrelin concentrations in girls with AN (16, 17). However, the relation between ghrelin and regional body composition in boys with AN is unknown. We therefore examined regional body composition using dual energy X-ray absorptiometry (DXA) in adolescent boys with AN and in healthy adolescent boys of similar maturity. We also evaluated endocrine predictors of these measures of body composition.
We studied 15 adolescent boys with AN (diagnosed by DSM-IV criteria) aged 12–19 y and 15 healthy boys of comparable chronologic age and maturity (as assessed by bone age). The time since diagnosis ranged from 0.5 to 54 mo (median: 4.4 mo), and mean age at diagnosis was 14.7 ± 1.7 y in boys with AN. Healthy boys had no history of eating disorders and were weight stable (within 2 kg) over the 3 mo preceding the study. Boys with AN were referred to the study from eating disorder units in the New England area and the Hospital for Sick Children (Toronto, Canada), and control subjects were recruited within the Partners HealthCare system and through mass mailings to pediatricians, nutritionists, and psychiatrists. All boys with AN and all but 2 control subjects were white (1 black and 1 Asian). Subjects were excluded if they had a history of endocrine or metabolic disorders (other than AN) or were taking medications known to affect metabolism. All subjects with AN were enrolled in active outpatient treatment programs. The study was approved by the Institutional Review Board of Partners Health Care (Boston, MA) and the Research Ethics Board (REB) at the Hospital for Sick Children, Toronto, and informed assent and consent was obtained from all subjects and their parents.
Subjects were admitted to the General Clinical Research Center (GCRC) of Massachusetts General Hospital, Boston, or the Clinical Investigation Unit (CIU) at the Hospital for Sick Children, Toronto, for the study visit. All subjects had a hematocrit > 30%, potassium concentration > 3 mmol/L, and glucose concentration > 50 mg/dL. We measured height in triplicate using a single stadiometer at each site and averaged the readings. We used a single electronic scale at each site to measure body weight in the fasting state and used the formula weight (in kg)/height2 (in m) to calculate body mass index. A single endocrinologist assessed the bone age of subjects using the methods of Greulich and Pyle (18). Vital signs were recorded. A subset of our subjects (5 AN and 7 control subjects) completed a 4-d food record (including 3 weekdays and 1 weekend day) that has been validated for use in teenagers and young adults of both sexes (19, 20). Nutrient intake was calculated by using the Minnesota Nutrition Data System software (version 4.03; nutrient database 31) by GCRC dietitians.
We measured body-composition variables using DXA with a fan-beam technique (version 12.6.1, model QDR 4500; Hologic, Waltham, MA) and pediatric software. The CVs for fat and lean mass by DXA are 1.7% and 2.4%, respectively, for our institution. We also compared DXA-derived weight (fat mass + lean mass + whole-body bone mineral content) with measured weight, and the 2 measurements correlated strongly (r = 0.98).
When DXA is used to assess body-composition variables, measures of fat mass, lean mass, and bone mineral content are generated by system software. In addition, the specific region of interest is predefined by using a computer-generated overlay of default lines, which separate the body into 10 parts that represent the various regions of interest, and subregions are then manually adjusted to match the patient's anatomy. The trunk region or modified truncal region includes the chest, abdomen, and pelvis (21) and is defined by 1) a line drawn parallel to and through the base of the neck (separating the head from the torso), 2) vertical lines drawn through the 2 arm sockets (separating the arms from the chest), and 3) diagonal lines passing through the femoral neck joining the greater trochanter to the thumb on the same side and meeting between the thighs. These lines and a vertical line between the legs also define the extremities for body-composition assessment.
Regional body-composition variables were calculated as follows (4):
Although DXA-derived measures of trunk fat are a good surrogate for intraabdominal fat (22) and therefore a useful measure of regional body composition, as are extremity fat mass and lean mass, DXA measures of trunk lean mass are of limited clinical utility because this is a composite measure of abdominal muscles and viscera. Therefore, we present all details of regional fat mass but only extremity (and not trunk) lean mass.
In addition, we obtained fasting blood samples for measurement of IGF-I, testosterone, sex hormone–binding globulin (SHBG), estradiol, active ghrelin, and leptin concentrations. Dietitians of the GCRC at Massachusetts General Hospital and the Eating Disorder Program at the Hospital for Sick Children administered an exercise questionnaire validated for use in adolescents (23) and derived activity scores (h/wk).
An immunoradiometric assay was used to measure IGF-1 (detection limit: 2.06 ng/mL; intraassay CV: 3.9%) and SHBG (Diagnostic Products Corp, Los Angeles, CA; intraassay CV: 2.8–5.3%; sensitivity: 0.04 nmol/L). Radioimmunoassay was used to measure serum testosterone (Diagnostic Products Corp; intraassay CV: 5.1–9.8%; sensitivity: 4 ng/dL), estradiol (Diagnostic Systems Laboratories, Webster, TX; intraassay CV: 6.5–8.9%; sensitivity: 2.2 pg/mL), leptin (Linco Diagnostics Inc, St Louis, MO; sensitivity: 0.5 μg/L; CV: 3.4–8.3%), and active ghrelin (Linco, St Louis, MO; sensitivity: 7.8 pg/mL; CV: 7.4%). Free testosterone concentrations and bioavailable testosterone were calculated from total testosterone and SHBG concentrations by using the laws of mass action (www.issam.ch/freetesto.htm) (24). Samples were stored at –80 °C until analyzed.
Data are presented as means ± SEs and were analyzed by using the JMP program (version 4; SAS Institute Inc, Cary, NC). We used the Student t test to calculate differences between means when data were normally distributed and performed logarithmic conversions to approximate a normal distribution when data were not normally distributed (necessary for active ghrelin concentrations). A P value <0.05 was considered significant. We also examined differences between groups for regional body composition after adjusting for weight, fat mass, and lean mass by using analysis of covariance (ANCOVA). Simple correlation analysis was used to determine associations between DXA-derived weight and measured weight. To determine independent predictors of regional body-composition measures, we used forward stepwise regression analyses (P = 0.11 for entry into the model and P = 0.10 to leave the model). The groups were considered together, given the significant overlap in most variables and the continuum of weight, to better define associations as they pertain to variations in weight. Variables entered into the regression model were bone age as an index of pubertal maturity (given effects of pubertal status on body composition), weight, and known hormonal determinants of body composition (free testosterone, IGF-1, estradiol, and ghrelin). We also entered 2 interaction terms into the model: weight × free testosterone and weight × bone age, given that weight was the major difference between the groups and given differences in slopes and intercepts for the 2 groups for these associations. Although we assessed leptin concentrations, leptin was not entered into the regression model because leptin is a product of fat cells and is driven by fat rather than a determinant of fat.
The clinical characteristics (but not regional body-composition data) of our subjects were reported previously (25) and are summarized in Table 1. Subjects with AN did not differ from control subjects in chronologic or bone age and, as expected, had lower body mass indexes than the control subjects. Boys with AN had lower total, free, and bioavailable testosterone and estradiol concentrations than did control subjects, but IGF-1, ghrelin, and leptin concentrations did not differ significantly. In the subset for whom food records were available, caloric intake did not differ between AN (2286 ± 359 kcal/d) and control (2031 ± 241 kcal/d) subjects. Of note, all AN boys were in active outpatient treatment programs.
Boys with AN had less fat and lean mass than did healthy adolescent boys (Table 2). Fat mass and lean mass in AN boys were 69% and 86% of that in controls. Although both total trunk mass and total extremity mass were lower in AN than in control subjects, there was preferentially lower fat and lean mass in the extremity than in the trunk, such that extremity fat and lean mass were lower in the AN than in the control subjects, whereas trunk fat only trended lower. As a consequence, percentage trunk fat (subtotal) was higher and percentage extremity fat (subtotal and total) and percentage extremity lean mass (subtotal and total) were lower in AN than in control subjects.
Because body-composition variables can be affected by body weight and total fat mass or lean mass, we used ANCOVA to determine whether differences between groups persisted after controlling for these variables. The adjusted differences between the 2 groups and associated P values after weight was controlled for are shown in Table 2. Boys with AN had lower weight-adjusted percentage body fat, extremity fat, percentage extremity fat (subtotal and total), and percentage extremity lean mass (subtotal and total) than did control subjects. In contrast, weight-adjusted percentage trunk fat (subtotal and total) and the ratio of trunk to extremity fat (trunk/extremity fat ratio) were higher in AN than in control subjects.
We also ran the ANCOVA for differences between groups for trunk and extremity fat measures after controlling for total fat mass. Interestingly, after total fat mass was controlled for, trunk fat was higher by 0.3 ± 0.1 kg (P = 0.02), subtotal percentage trunk fat was higher by 4.2 ± 2.0% (P = 0.049), and total percentage trunk fat was higher by 3.9 ± 1.6% (P = 0.02) in AN than in control subjects. Extremity fat was lower in AN than in control subjects by 0.2 ± 0.1 kg, although this difference was no longer significant between groups (P = 0.10). Percentage extremity subtotal and total fat were lower by 4.2 ± 2.0% (P = 0.049) and 3.1 ± 2.0% (NS), respectively, in AN than in control subjects.
Finally, we ran the ANCOVA for differences between groups for extremity lean measures after controlling for total lean mass. Extremity lean mass was still lower in AN than in control subjects by 0.95 ± 0.2 kg (P = 0.0004), and subtotal and total percentage extremity lean mass were lower by 2.3 ± 0.5% (P < 0.0001) and 2.0 ± 0.4%, respectively (P < 0.0003). Reanalysis of the data after exclusion of the 2 control subjects who were not white yielded similar data, except that fat mass after control for weight was significantly lower by 2.3 ± 0.9 kg in AN than in control subjects (P = 0.02)
Associations with total and bioavailable testosterone were similar to those with free testosterone, and only associations with free testosterone are reported. Data from regression modeling (bone age, weight, free testosterone, estradiol, IGF-1, active ghrelin, weight × bone age and weight × testosterone entered into the model) are reported in Table 3. Percentage trunk fat (subtotal) and the trunk/extremity fat ratio were predicted inversely by free testosterone and positively by bone age, whereas percentage extremity fat (subtotal and total) was predicted positively by free testosterone and inversely by bone age. The inverse association of the trunk/extremity fat ratio with free testosterone after control for bone age is shown in Figure 1. Testosterone and bone age predicted percentage extremity lean mass (subtotal), and bone age and the interaction term testosterone × weight predicted percentage extremity lean mass (total).
We report for the first time differences in body composition in males with AN that differ from those reported in females with AN. Boys with AN had lower fat and lean mass than did control subjects, as expected. However, in addition and in contrast with findings in girls with AN, percentage trunk fat was higher than in control subjects, particularly after adjustment for weight and fat mass. This finding is consistent with a sex-specific dichotomy in the association of extreme low body weight with fat distribution and is likely mediated by low concentrations of testosterone.
Few studies have reported differences in body composition between adolescent boys with AN and healthy adolescents. In our study, boys with AN had less fat and lean mass than did control subjects. We reported similar differences between adolescent girls with AN and healthy adolescent girls (4, 26). However, we observed that lean mass in boys with AN was lower in control subjects than we previously reported in girls with AN (26) (14% lower in AN boys compared with 5% lower in AN girls compared with control subjects). This may be a consequence of markedly lower testosterone concentrations in pubertal AN boys than in control subjects, testosterone being an important determinant of muscle mass (8–10). Many studies in hypogonadal adults have shown an increase in lean mass after testosterone replacement (8–10). Consistent with these data, testosterone was an important and independent predictor of regional lean mass in the present study.
Interestingly, regional differences in body composition between boys with AN and control subjects differed greatly from reported differences between girls with AN and healthy girls (4, 26). For instance, adolescent girls with AN have lower percentage trunk fat than healthy adolescent girls (26). In contrast, we observed higher weight-adjusted percentage trunk fat and lower percentage extremity fat and percentage extremity lean fat in boys with AN than in control subjects. Similarly, the trunk/extremity fat ratio was higher in boys with AN than in control subjects. To summarize, boys with AN have preferentially lower fat and lean mass in the extremities and higher fat mass in the trunk than do control subjects. In contrast, girls with AN have much lower fat mass in the trunk than in the extremities than do control girls.
Although our study was not designed to determine the cause of this sparing of trunk fat in boys with AN and in healthy adolescent boys, we propose that this is a consequence primarily of low testosterone concentrations, and that the difference from girls with AN reflects differences in effects of sex-specific gonadal steroids on body-composition measures. Hypogonadal men have greater waist-to-hip circumferences and subcutaneous fat than do eugonadal men, and a decrease in these measures occurs after testosterone replacement (8, 11, 12). Similar effects in hypogonadal AN boys could result in a relative sparing of trunk fat despite the accompanying weight loss. Consistent with this hypothesis, we observed inverse associations of free testosterone with percentage trunk fat and the trunk/extremity fat ratio on regression modeling with bone age (as an index of pubertal maturity), weight, and other hormones known to affect body composition entered into the model. Positive associations of bone age and inverse associations of free testosterone with these measures on regression modeling suggest that older and more mature boys with lower concentrations of testosterone have the highest percentage trunk fat and the greatest relative abdominal adiposity.
Other hormonal aberrations that can cause relative truncal adiposity include GH deficiency and hypercortisolemia. GH decreases fat mass (particularly trunk fat) and increases muscle mass (27, 28), and conditions of GH deficiency are associated with increased trunk fat (29, 30). We previously reported associations of IGF-1 with trunk fat in adolescent girls with AN (4). However, although IGF-1 concentrations were lower in boys with AN than in control subjects, this difference was not statistically significant, and IGF-1 concentrations did not predict body-composition measures in boys with AN. Given that IGF-1 concentrations are markedly lower in girls with AN than in control subjects, associated with a nutritional resistance to GH effects (13, 31), the relative sparing of IGF-1 concentrations in boys with AN was unexpected, and the reason was unclear. Overall, our data suggest a stronger association of hypogonadism than of the GH-IGF-1 axis with body-composition variables in hypogonadal adolescent males.
Of note, excessive cortisol can cause truncal adiposity and decreased extremity muscle mass (32), and girls with AN have high cortisol concentrations (33). Adult women with AN develop increased trunk fat as they recover, and women with the highest cortisol concentrations have the greatest increases in trunk fat with weight recovery (5). We did not measure cortisol concentrations in our subjects—a study limitation. One can speculate that high cortisol concentrations in boys with AN contributes to both a sparing of trunk fat and a lower extremity lean mass. In fact, we previously reported that girls with AN with the highest cortisol concentrations had the lowest percentage extremity lean mass (26). Further studies are therefore necessary to determine the effects of cortisol on body composition in AN boys.
Inverse associations of ghrelin with lean mass have been reported in healthy men (15). In this study, we found no associations of ghrelin with regional body-composition measures after control for bone age, weight, and other hormones. Leptin concentrations were markedly lower in girls with AN than in control subjects (27% of values in control subjects) (16) and were lower in AN boys in our study (54% of values in control subjects), but not to the extent seen in AN girls. Because leptin is primarily secreted by adipocytes, the fact that leptin concentrations were not as low in AN boys compared with control subjects as reported in AN girls likely reflects the fact that fat mass is not as low in AN boys compared with control subjects as reported in AN girls (26).
Study limitations include the fact that we could not obtain a better measure of the duration of a catabolic state in boys with AN than the time since diagnosis. Duration of illness would have been a better measure, however, our subjects and their parents were mostly unsure of the time of onset of this disorder. In addition, although our AN subjects weighed much less than the control subjects, the AN subjects were in active treatment programs, and, in a subset in whom nutritional intake information was available, caloric intake did not differ between AN and control subjects. It is possible that a more profound catabolic state may have resulted in even more profound differences between the groups for both biochemical and body-composition measures. Even though the AN boys weighed much less than the control subjects, our data may reflect a partially recovered state. Therefore, it will be important to follow subjects over time and determine how regional body composition changes with changing weight.
We thus report unique differences in body composition between boys with AN and control subjects, which are predicted by bone age, testosterone, and weight. Low weight in boys with AN is associated with lower weight-adjusted percentage fat and lean mass in the extremities compared with control subjects and a higher percentage trunk fat and trunk/extremity fat ratio. These body-composition differences are related to lower testosterone concentrations in boys with AN than in control boys.
We thank Ellen Anderson and her Bionutrition team as well as the skilled nursing staff of the GCRC, Massachusetts General Hospital (Boston, MA), and of the Clinical Investigation Unit of the Hospital for Sick Children (Toronto, Canada) for their help in completing this study. We also thank Jeffrey Breu of the Core Laboratory of the Massachusetts Institute of Technology for his help in analyzing the IGF-1, testosterone, SHBG, estradiol, ghrelin, and leptin samples. Most of all, we sincerely thank our subjects, without whose participation this study would not have been possible.
The authors’ responsibilities were as follows—MM: designed and conducted the study, collected and analyzed the data, and wrote the manuscript; KKM and AK: designed and conducted the study and reviewed the manuscript; and DKK, DM, DBH, JC, and SJM: conducted the study and reviewed the manuscript.
2Supported by NIH grants R01 DK 062249, K23 RR018851, and M01-RR-01066.
None of the authors reported a conflict of interest.