European Journal of Clinical Nutrition (2003) 57, 260–265 ß 2003 Nature Publishing Group All rights reserved 0954–3007/03 $25.00 www.nature.com/ejcn

ORIGINAL COMMUNICATION Bone resorption in anorexia nervosa and rehabilitated patients ˜a1, V Di Mattei1, L Rossi1, A Polito1, M Cuzzolaro2 and F Branca1* S Valtuen 1

Human Nutrition Unit, National Institute for Food and Nutrition Research, Rome, Italy; and 2Department of Neuropsychiatry, University ‘La Sapienza’, Rome, Italy Objective: To assess the impact of anorexia nervosa and that of nutritional rehabilitation on bone resorption. Design: Cross-sectional, observational study. Setting: Rome, Italy Subjects: Twenty-eight female patients affected by anorexia nervosa (AN, BMI
17.0 kg=m2), 18 females rehabilitated from anorexia nervosa and weight-stable for at least 6 months (RE, BMI  18.5 kg=m2) and 34 age- and sex-matched healthy controls (CO, BMI  18.5 kg=m2). Among AN patients, 16 were affected by the ‘restrictive’ (ANr) and 12 by the ‘purging’ type (ANp) of anorexia nervosa. Method: Body weight, height and skeletal diameters were measured on each individual. The skeletal mass (SKM) was predicted from the skeletal diameters of the elbow, wrist, knee and ankle, using the equation of Martin. Twenty-four-hour urinary excretion of pyridinium crosslinks of collagen (pyridinoline (Pyd) and deoxypyridinoline (Dpd)) and creatinine was assessed by reversed-phase HPLC with fluorimetric detection after solid-phase extraction and by the Jaffe´-method with deproteinization, respectively. Results: Twenty-four-hour urinary output of Pyd and Dpd was not significantly different between AN and CO when expressed in absolute values, but AN showed higher bone resorption than CO when Pyd and Dpd excretion was adjusted by either creatinine (P < 0.0000) or the SKM (P < 0.05). Within the AN group, urinary excretion of both cross-links was significantly and consistently higher in ANp compared with ANr (P < 0.05). However, these differences disappeared when crosslink output was adjusted either by urinary creatinine or SKM. RE subjects showed no differences in bone resorption with the AN group despite weight gain, being crosslink excretion consistently elevated compared to controls (Pyd: P < 0.01 by creatinine and P < 0.05 by SKM; Dpd: P < 0.01 by creatinine and P < 0.05 by SKM). Conclusion: Bone resorption is elevated in anorexia nervosa and different strategies for low-weight maintenance do not seem to have a differential impact. Increased bone resorption persists in subjects with past diagnosis of anorexia nervosa despite rehabilitation lasting more than 6 months. This finding indicates that bone mass and turnover should be monitored in anorexia nervosa patients and ex-patients well beyond recovery of normal body mass. Further investigation is warranted to examine the long-term effect of such prolonged increase in bone turnover at a young age. ˜ a is supported by a Marie Curie Fellowship from the European Commission (Contact no. HPMF-CTSponsorship: Dr. Valtuen 1999-00192). European Journal of Clinical Nutrition (2003) 57, 260 – 265. doi:10.1038=sj.ejcn.1601527 Keywords: anorexia nervosa; ‘purging’ anorexia; restrictive anorexia; pyridinoline; deoxypyridinoline; nutritional rehabilitation

*Correspondence: F Branca, Human Nutrition Unit, National Institute for Food and Nutrition Research, Via Ardeatina 546, 00179 Roma, Italy. E-mail: [email protected] Guarantor: Dr. Francesco Branca. Contributors: SV was responsible for the analysis of the data and the drafting of the paper; VDM was responsible for the set-up of the database and the performance of creatinine and collagen crosslink assays; LR was responsible for the performance of crosslink assays and

preliminary data analysis; AP was responsible for the multidisciplinary project on metabolic adaptation, vitamin status and bone health in anorexia nervosa, of which the present paper is a part; MC was responsible for subject selection and clinical assessment; FB was responsible for the design of the bone health study. Received 20 March 2002; revised 3 May 2002; accepted 8 May 2002

Bone resorption in anorexia nervosa S Valtuen˜ a et al

Introduction Anorexia nervosa (AN) is often associated with low bone mass, low bone density and increased risk of osteoporotic fractures (Andersen et al, 1995; Bachrach et al, 1990; Castro et al, 2000; Lucas et al, 1999). This risk could be mediated by the uncoupling between bone formation and resorption. Previous studies have found an increased bone resorption in AN that is not adequately compensated by an increase in bone formation (Lennkh et al, 1999; Stefanis et al, 1998), but an insufficient bone formation in the presence of ‘normal’ bone resorption has also been described in anorexia patients (Soyka et al, 1999). Although both mechanisms of bone loss would lead to osteoporosis in the end, knowledge about which dominates the osteoporotic process in AN is essential to identify the best therapeutic strategy. Common features of the disease that contribute to the development of osteoporosis include oestrogen deficiency, glucocorticoid excess, malnutrition and low calcium intake (Andersen et al, 1995; Bachrach et al, 1990; Castro et al, 2000; Lucas et al, 1999; Rigotti et al, 1984). However, patients with AN may not constitute a homogeneous group as far as bone metabolism is concerned, since there is an additional factor associated with the low-weight maintenance strategy adopted that may influence bone metabolism. The ‘purging’ or ‘binge eating’ type of AN is characterized, among others, by frequent vomiting, a well-recognized cause of transitory acid – base unbalance. Given the demonstrated sensitivity of bone to release calcium and potassium in order to compensate plasma variations of pH, it may well be that bone turnover is differentially affected in the ‘purging’ and ‘restrictive’ types of AN (Goldhaber & Rabadjija 1987; Val˜a et al, 2001; Wachman & Bernstein, 1968). The only tuen data to this respect has been recently reported by Zipfel et al (2001), who observed that patients with the binge-eating=purging type of AN (n ¼ 8) show lower bone mineral density and increased bone resorption compared with the restricting type patients (n ¼ 7), serum phosphorus levels being significantly associated with urinary excretion of deoxypyridinoline (Dpd). A second question deserving attention is the impact of nutritional rehabilitation on bone mass and metabolism. The first study published on ex-anorexia patients reported lower bone density (BMD) than expected even years after body weight had been restored (Ward et al 1997), whereas a recent retrospective work shows that BMD is completely recovered if weight is regained (Valla et al, 2000). As far as bone turnover is concerned, Hotta et al (2000) describe a rapid increase of plasma osteocalcin in eight severely malnourished AN patients after 5 weeks of i.v. hyperalimentation. However, by the end of the therapy, bone resorption remained elevated, as assessed by the urinary excretion of Dpd. In this cross-sectional, observational, pilot study, we aimed to investigate the bone resorption status in AN, and whether it may differ between ‘purgative’ and ‘restrictive’

261 anorexia patients. A second aim was to assess the impact of nutritional rehabilitation on bone metabolism.

Subjects and methods Twenty-eight female patients between 16 and 37 y of age (mean ¼ 24.1  5.4 y) with AN who had been recruited among outpatients attending the Ambulatory for Eating Disorders at the University of Rome ‘La Sapienza’ for previous research studies (Polito et al, 1998; 2000) were included. All met the criteria of DSM IV-R Criteria (American Psychiatric Association, 1994) for AN, had a body mass index (BMI)
17.0 kg=m2 and were weight stable for the 3 months prior to test. Sixteen had been classified as ‘restrictive’ in the absence of purge – binge habits (mean age ¼ 24.6  6.2 y), whereas 12 were considered of the ‘purging’ type of AN for recurrent vomiting (mean age ¼ 23.5  4.2 y). Also, 18 recently rehabilitated females (RE, BMI  18.5 kg=m2) with a previous history of AN of at least 8 months (median and interquartile range: 2 (4) y) who had been weight stable for at least 6 months (range ¼ 6 – 48 months) were included in the study. Thirty-four healthy, normal-weight (BMI  18.5 kg=m2) females within the same age range as patients were recruited from the general population to serve as controls (CO). All subjects had a complete medical history and medical exam, including a gynaecological history (consumption of oral contraceptives, presence of menses) and an anthropometric evaluation (weight, height and skeletal diameters). The totality of AN patients were amenorrhoeic and all CO had normal menses. None of the AN patients was or had been on estrogenic therapy of any kind during the 6 months prior the study. Mean duration of illness established from the start of amenorrhea was 2 (3) y. Out of the 18 RE subjects, 11 had normal menses, five were amenorrhoeic, and two were on oral contraceptives. The protocol was approved by the Ethics Committee for Clinical Investigation in Humans of the University of Rome ‘La Sapienza’ and all subjects gave written informed consent prior to enrolment.

Anthropometrics Height was measured with a wall-mounted Holtain stadiometer to the nearest 0.1 cm according to Lohman et al (1988). Weight was measured without shoes and in light clothes to the nearest 0.01 kg on an electronic balance with digital read-out (K-Tron P1-SR). BMI was calculated as weight (in kg) divided by height (in m) squared (kg=m2). Skeletal diameters of the elbow, wrist, knee and ankle were measured with Harpenden spreading calipers to the nearest 1 mm. The standard error of measurements for the skeletal diameters ranged from 0.09 to 0.2 cm in our laboratory (Polito et al, 1998). Skeletal mass (SKM) was predicted using the equation P of Martin (1991): SKM (kg) ¼ 0.601074S( bi)2, where S European Journal of Clinical Nutrition

Bone resorption in anorexia nervosa S Valtuen˜ a et al

262 is height in cm and bi are the individual skeletal diameters in cm.

Collection of urine samples Timed 24 h urine samples were collected in tightly controlled conditions during the day the subjects attended the National Institute for Food and Nutrition Research for a 24 h calorimetric testing. The collection was performed in two timeframes within a day (from 8 am to 8 pm and from 8 pm to 8 am). In subjects with normal menses, urine was sampled from the 6th to the 12th day of the menstrual cycle to avoid changes in the composition of body fluids due to sexual hormones. All subjects were on a 4 day meat-free diet prior to the collection day. Urine samples were collected in polyethylene bottles containing 10 ml of 6 N HCl as a preservative, sampled and stored at 7 20 C until the analysis. Freeze – thawing was avoided.

Biochemical analysis in urine samples Collagen cross-links were assayed in urine by reversed-phase HPLC with fluorimetric detection after solid phase extraction, as described elsewhere (Pratt et al, 1992). The interassay coefficient of variation was < 8%. Creatinine was determined by Jaffe` method with deproteinization (CV < 4%) (Spierto et al, 1979).

Statistical analysis Statistical analysis were performed using the statistical program Statistica for Windows, release 4.5. Since conditions for use of the parametric ANOVA test were not met by most variables (normality and homogeneity of variances), threegroup comparisons were performed with the non-parametric Kruskal – Wallis ANOVA test. Post-hoc and two-group comparisons were assessed using the Mann – Whitney U-test for numerical, continuous variables. Results are presented as medians (interquartile range), unless otherwise noted. All analyses were two-tailed. The level of significance was set at P < 0.05.

Results AN patients showed significantly lower body weight and BMI than CO or RE subjects, but the three groups were comparable regarding age, stature and SKM assessed by anthropometry (Table 1). SKM, however, represented a higher percentage of total body weight in AN patients than in either RE or CO. Within the AN group, the ‘purging’ (ANp) and ‘restrictive’ (ANr) subgroups were not significantly different regarding age, weight, height, BMI and SKM, expressed either in absolute values or as percentage of body weight (data not shown). The RE group had achieved normal BMI, but still had significantly lower body weight than CO. European Journal of Clinical Nutrition

Table 1 Anthropometric characteristics of controls (CO), rehabilitated subjects (RE) and anorexia patients (AN)a Variable

CO (n ¼ 34)

AN (n ¼ 28)

Age (y) Height (cm) Weight (kg) BMI (kg=m2) SKM (kg) SKM (% BW)

24.5 163.2 53.7 20.4 6.4 11.6

23.0 165.1 42.6 15.8 6.0 13.9

(6.5) (11.0) (8.6) (2.3) (1.3) (1.7)

(8.0) (10.1) (1.6){{ (1.6){§ (1.3) {§ (2.7)

RE (n ¼ 18) 25.0 160.5 51.4 19.4 5.9 11.6

(8.5) (8.8) (4.9)* (2.0)** (1.0) (2.0)

P-valueb NS NS < 0.0001 < 0.0001 NS < 0.0001

a

Results are expressed as median (interquartile range). Significance for trend by the Kruskal – Wallis ANOVA test. BMI, body mass index; SKM, skeletal mass; BW, body weight. Significantly different from controls: *P > 0.05; **P < 0.01; {P < 0.0000. Significantly different from rehabilitated subjects: { P < 0.0001; §P < 0.0000. b

Given that the normal muscle=bone mass ratio is unlikely to be preserved in AN due to severe protein-energy malnutrition (and probably in RE subjects), traditional methods for the standardization of crosslink excretion may not be appropriate to compare AN, RE and CO subjects. Therefore, in addition to creatinine, the anthropometric estimation of SKM was considered for standardization. This is supported by the significant correlation found between SKM and total body bone mineral density (TBBMD) assessed by DXA (r ¼ 0.57; P < 0.0000) in 25 AN patients, 27 RE subjects and 28 CO of the same characteristics as our study population (unpublished data). *SKM was also correlated with TBBMC within RE subjects (r ¼ 0.57; P < 0.01) and within CO (r ¼ 0.56; P < 0.01), although it was not within AN patients. Urinary output of Pyd and Dpd for the AN, RE and CO groups is depicted in Table 2. AN subjects showed increased excretion of pyridinium crosslinks (in nmol=mmol creatinine) compared with CO in day (Pyd: 66.5 (60.0) vs 44.0 (23.9), P < 0.0001; Dpd: 17.0 (18.0) vs 11.6 (5.2), P < 0.0001), overnight (Pyd: 81.6 (53.6) vs 46.3 (21.2), P < 0.0001; Dpd: 20.5 (11.8) vs 13.1 (4.3), P < 0.0001) and 24 h urine collections (Pyd: 154.2 (92.8) vs 94.6 (43.3), P < 0.0000; Dpd: 36.9 (27.7) vs 24.2 (8.8), P < 0.0000). When crosslinks were expressed as nmol=kg SKM, only differences between groups in overnight (P < 0.05) and 24 h collections (P < 0.05) remained significant. Within the AN group, urinary excretion of Pyd and Dpd was significantly and consistently higher in ANp compared to ANr (P < 0.05), but such differences disappeared when crosslink excretion was adjusted by either SKM or creatinine (Table 3). In RE subjects, excretion of bone resorption markers was no different from AN patients (Table 2), but both Pyd and Ddp (nmol=mmol creatinine) were elevated compared with controls in day (Pyd: 57.4 (49.2), P < 0.01; Dpd: 14.6 (12.8), P < 1074), overnight (Pyd: 81.9 (81.1), P < 0.01; Dpd: 18.0 (14.7), P < 0.0001) and 24 h urine collections (Pyd: 132.5 (112.5), P < 0.01; Dpd: 35.8 (25.8), P < 0.0001). As for the AN group, when crosslinks were standardized by SKM only differences between RE and CO in overnight (P < 0.05) and 24 h collections (P < 0.05) remained significant (Table 3). There was no difference in the circadian pattern of crosslink

Bone resorption in anorexia nervosa S Valtuen˜ a et al

263 Table 2 Urinary excretion of pyridinium collagen crosslinks in controls (CO), rehabilitated subjects (RE) and anorexia patients (AN)a Variable

CO (n ¼ 34)

Pyd=day (nmol=12 h) Pyd=day (nmol=mmol creatinine) Pyd=day=SKM (nmol=kg) Dpd=day (nmol=12 h) Dpd=day (nmol=mmol creatinine) Dpd=day=SKM (nmol=kg) Pyd=night (nmol=12 h) Pyd=night (nmol=mmol creatinine) Pyd=night=SKM (nmol=kg) Dpd=night (nmol=12 h) Dpd=night (nmol=mmol creatinine) Dpd=night=SKM (nmol=kg) Pyd=24 h (nmol=24 h) Pyd=24 h (nmol=mmol creatinine) Pyd=24 h=SKM (nmol=kg) Dpd=24 h (nmol=24 h) Dpd=24 h (nmol=mmol creatinine) Dpd=24 h=SKM (nmol=kg) Pyd=Dpd

211.6 44.0 33.9 51.7 11.6 8.5 213.3 46.3 32.5 61.0 13.1 9.5 407.8 94.6 59.9 113.2 24.2 16.7 3.7

AN (n ¼ 28)

(45.7) (23.9) (22.7) (26.0) (5.2) (4.1) (80.7) (21.2) (17.7) (22.5) (4.3) (3.5) (215.9) (43.3) (31.8) (46.2) (8.8) (5.2) (1.0)

a

206.6 66.5 46.3 56.8 17.0 9.6 263.2 81.6 41.9 66.4 20.5 12.2 459.2 154.2 75.7 128.7 36.9 22.7 3.8

(135.3) (60.0){ (30.8) (37.9) { (18.0) (6.6) (124.7) (53.6){ (20.4)* (41.7) { (11.8) (6.4)* (173.7) (92.8)§ (38.6)* (55.0) § (27.7) (14.6)* (0.9)

RE (n ¼ 18) 253.6 57.4 46.3 62.2 14.6 10.1 292.8 81.9 44.2 67.7 18.0 11.2 520.9 132.5 85.1 138.7 35.8 23.8 4.0

(213.7) (49.2)** (30.8) (51.4) (12.8)* (7.6) (219.7) (81.1)** (37.1)* (42.7) { (14.7) (8.0)* (465.7) (112.5)** (70.7)* (77.9) { (25.8) (11.2)* (0.5)

P-valueb NS < 0.001 NS NS < 0.001 NS NS < 0.0001 0.046 NS < 0.0001 0.013 NS < 0.0001 0.045 NS < 0.0001 0.016 NS

b

Results are expressed as medians (interquartile ranges). Significance for trend by the Kruskal – Wallis ANOVA-test. Pyd, pyridinoline; Dpd, deoxypyridinoline. Significantly different from controls: *P < 0.05; **P < 0.01; {P < 0.001; {P < 0.0001; § P < 0.0000.

Table 3 Urinary excretion of pyridinium collagen crosslinks in anorexia patients affected by the restrictive (ANr) and purging (ANp) type of anorexia nervosa a Variable Pyd=day (nmol=12 h) Pyd=day (nmol=mmol creatinine) Pyd=day=SKM (nmol=kg) Dpd=day (nmol=12 h) Dpd=day (nmol=mmol creatinine) Dpd=day=SKM (nmol=kg) Pyd=night (nmol=12 h) Pyd=night (nmol=mmol creatinine) Pyd=night=SKM (nmol=kg) Dpd=night (nmol=24 h) Dpd=night (nmol=mmol creatinine) Dpd=night=SKM (nmol=kg) Pyd=24 h (nmol=24 h) Pyd=24 h (nmol=mmol creatinine) Pyd=24 h=SKM (nmol=kg) Dpd 24 h (nmol=24h) Dpd=24 h (nmol=mmol creatinine) Dpd 24 h=SKM (nmol=kg) Pyd=Dpd

ANr (n ¼ 18) 196.1 63.5 35.6 47.4 14.7 8.5 220.2 75.3 39.1 61.5 19.4 10.9 418.4 154.2 73.4 115.5 36.2 20.5 3.9

(85.2) (45.7) (22.8) (19.2) (13.4) (4.4) (111.6) (54.9) (21.4) (44.2) (22.5) (8.1) (150.6) (102.9) (35.2) (51.7) (39.4) (10.1) (0.8)

ANp (n ¼ 12) 284.9 70.0 47.3 79.4 24.5 12.8 305.2 82.1 52.7 86.1 21.4 13.8 567.7 157.0 101.1 154.7 43.7 23.7 3.6

(217.8) (77.9) (38.6) (42.9) (19.0) (8.3) (184.8) (67.7) (34.3) (58.4) (10.4) (11.0) (347.5) (140.09) (56.9) (77.7) (28.2) (13.4) (0.7)

P-value

b

0.048 NS NS 0.034 NS NS 0.004 NS 0.026 0.021 NS NS 0.020 NS NS 0.023 NS NS NS

a

Results are expressed as medians (interquartile ranges). Significance for trend by the Kruskal – Wallis ANOVA test. Pyd, pyridinoline; Dpd, deoxypyridinoline.

b

excretion between the AN, RE and CO groups (data not shown).

Discussion We found an increased bone resorption in amenorrheic female patients affected by AN compared with healthy,

normal-weight, normally menstruating, age-matched controls. This is consistent with data from Stefanis et al (1998), who report an approximately 50% higher excretion of Dpd=mmol creatinine in second-void morning urine samples (SVS) for patients with AN. Differences of similar magnitude were observed in the present study in day, night and 24 h urine collections. In contrast, Soyka et al (1999) do not find differences in SVS-Dpd excretion=mmol creatinine between AN patients and controls. Discrepant data may be explained by the fact that subjects (AN and CO) in the Soyka study were substantially younger (mean age between 15 and 16 y) than in Stefanis’ (23 – 28 y) and ours (24 – 26 y). Urinary Dpd is a marker of growth in normal children (Branca et al, 2002) and adolescents (Rauch et al, 1996) which experiences a dramatic increase around puberty (Marowska et al, 1996) to progressively decrease until reaching adult values by age 20 y (Vesper et al, 2002). Thus, menstruating controls in the Soyka study were likely to show higher growth velocity than amenorrheic AN patients both being in pubertal ages, and therefore higher Dpd excretion if bone resorption in AN patients had not been increased for other reasons. Unfortunately, growth rates were not reported. Especially interesting is the observation that beta-CTX may be more suitable to monitor bone resorption in AN patients than alpha-CTX, given the major amount of cross-laps excreted in urine coming from old bone in AN compared with controls (de la Piedra et al, 1999). Our data does not support the hypothesis of differential bone resorption between ANp and ANr patients suggested by Zipfel et al (2001). In their sample of patients, the binge eating=purging type showed decreased bone formation and increased bone resorption compared with the restrictive type European Journal of Clinical Nutrition

Bone resorption in anorexia nervosa S Valtuen˜ a et al

264 of AN patients, but the duration of amenorrhea, which is a well-known factor contributing to the severity of bone damage in AN patients (Baker et al, 2000), was also higher in the former group. This was not the case in our ANp and ANr patients, but our data is not sufficient to exclude the possibility of a differential bone metabolism in the two types of AN and this hypothesis should be further explored. On the other hand, controversy exists regarding the capacity of restoring bone mass and normal bone turnover in AN following nutritional rehabilitation. BMD appears to increase slowly with weight gain and the regularization of the menstrual cycle, but it remains lower than that observed in female age-matched populations even after long-term recovery from AN (Andersen et al, 1995; Ward et al, 1997). Although it has been hypothesized that potential for recovery of BMD may be intact for many years after menarche (Valla et al, 2000; Castro et al, 2000), even if at slow rate and with measurable effects only several months after the improvement of nutritional status (Jagielska et al, 2001), in a big population-based retrospective cohort study overall fracture risk (for hip, vertebrae and forearm) was 2.5 times higher (95% CI: 2.0 – 3.9) for ex-AN females than for controls later in life (Lucas et al, 1999). Regarding bone resorption markers, the only data available comes from AN patients undergoing short-term nutritional therapy for weight gain. In a 2-month refeeding study, excretion of free Dpd at week 8 was not different from baseline in 15 AN females. Similarly, urinary excretion of C-terminal telopeptide of collagen type I (u-CTX) did not change significantly in AN patients on a 5 week i.v. hyperalimentation (Hotta et al, 2000). The same applies for u-CTX in another small group of nine AN women after refeeding, although the non-significant decrease in u-CTX was of 40% from baseline. Serum-CTX did indeed decrease significantly, and its circadian variation was restored (Caillot-Augusseau et al, 2000). We observe that urinary excretion of Pyd and Dpd was higher in nutritionally rehabilitated ex-AN patients that were weight stable for at least 6 months than in healthy controls and no different from AN patients. This is consistent with the idea that bone resorption could remain elevated after weight recovery and restoration of menses (in 11 patients out of 18), although the cross-sectional nature of this study limits the statement. If these results are confirmed in larger, prospective studies, elevated bone resorption could be an additional explanation (besides reduced BMD) for the higher risk of fracture observed in ex-AN subjects (Garnero et al, 1996a,b 1999). In addition, it would suggest that antiresorptive therapy may be of use in the treatment of ANinduced osteoporosis. In our study we did not observe differences in the diurnal pattern of collagen crosslinks in AN, RE and CO. This indicates that the number of bone remodelling units activated rather than the duration of osteoclast activation seems to be involved in the increased bone resorption. Further and more detailed physiopathological investigations are warranted. Present uncertainties about bone resorption status European Journal of Clinical Nutrition

in AN and rehabilitated patients, besides the small sample sizes considered in the studies published, may come from some methodological inaccuracies. First, second morning urine samples were collected for biochemical analysis, whereas higher differences in urinary crosslink excretion between anorexia patients and controls may be observed in overnight collections, as suggested by the present study. Only the latter would reflect the early morning peak of the circadian variation in crosslink excretion described in normal adults and children (Fincato et al, 1993; McLaren et al, 1993). In addition, that circadian variation, observed as well for plasma and serum C-telopeptides of collagen type-I, seems to be lost in AN and recovered with weight regain (Caillot-Augusseau et al, 2000), which would limit the use of spot urine samples for comparisons between AN patients before and after rehabilitation, RE subjects and controls. Second, standardization of resorption markers by creatinine as a surrogate of skeletal mass is optimal to study healthy, normal-weight subjects, but inappropriate to compare them with individuals in which the standard relationship between muscle mass and skeletal mass is lost. This typically happens in growing children and in patients with chronic nutritional disturbances, such as the severely mal˜a, 1999). In the nourished and the obese (Kehayias & Valtuen absence of a direct body composition method to estimate SKM, we used an anthropometric prediction for the standardization of urinary crosslinks in addition to creatinine. This decision is supported by the significant correlation found between SKM and TBBMD assessed by DXA in AN patients, RE subjects and CO of the same characteristics as our study population. SKM was also correlated with TBBMC within RE subjects and within CO, but it was not within AN patients. However, this methodological problem is less relevant when two groups of AN patients of similar age and anthropometric characteristics, such as ANr and ANp in the present study, are to be compared. Indeed, no differences were found between ‘restrictive’ and ‘purging’ anorexia patients in bone resorption regardless of the method of standardization, whereas significance was lost for day differences in Pyd and Dpd excretion between CO and AN=RE patients when SKM was used for crosslink standardization instead of creatinine.

Conclusion Female patients with AN show an increased bone resorption compared with healthy controls, but no differences in this variable are observed between the ‘purgative’ and the ‘restrictive’ type of AN. Bone resorption rates remain elevated after 6 months of nutritional rehabilitation. This indicates that careful monitoring of bone mass and turnover should be carried out in anorexia patients during and after the course of rehabilitation. Further investigation is warranted to clarify the pathogenesis of the increased turnover and to ascertain the long-term consequences of the bone disorder.

Bone resorption in anorexia nervosa S Valtuen˜ a et al

Acknowledgements We would like to thank Donatella Ciarapica for her help in performing the 24 h urine collections, subject recruitment and anthropometric measurements.

References American Psychiatric Association (1994): Diagnostic and Statistical Manual of Mental Disorders, 4th edn. Washington, DC. Andersen AE, Woodward PJ & LaFrance N (1995): Bone mineral density of eating disorder subgroups. Int. J. Eat. Disord. 18, 335 – 342. Bachrach LK, Guido D, Katzman D, Litt IF & Marcus R (1990): Decreased bone density in adolescent girls with anorexia nervosa. Pediatrics 86, 440 – 447. Baker D, Roberts R & Towell T (2000): Factors predictive of bone mineral density in eating-disordered women: a longitudinal study. Int. J. Eat. Disord 27, 29 – 35. ˜a S, Golden M & Robins S (2002): Urinary collagen Branca F, Valtuen cross-links as biochemical markers of growth: an evaluation of biological variables. Ann. Nutr. Metab. 46, 80 – 87. Caillot-Augusseau A, Lafage-Proust MH, Margaillan P, Vergely N, Faure S, Paillet S, Lang F, Alexandre C & Estour B (2000): Weight gain reverses bone turnover and restores circadian variation of bone resorption in anorexic patients. Clin. Endocrinol. (Oxf.) 52, 113 – 121. Castro J, Lazaro L, Pons F, Halperin I & Toro J (2000): Predictors of bone mineral density reduction in adolescents with anorexia nervosa. J. Am. Acad. Child Adolesc. Psychiatr. 39, 1365 – 1370. ˜oz de la Piedra C, Calero JA, Traba ML, Asensio MD, Argente J & Mun MT (1999) Urinary alpha and beta C-telopeptides of collagen I: clinical implications in bone remodeling in patients with anorexia nervosa. Osteoporos. Int. 10, 480 – 486. Fincato G, Barticci F, Rigoldi M, Abbiati G, Colombo M, Bartolini O, Brandi M & De Leonardis V (1993): Urinary excretion of pyridiniline and deoxypyridinoline: circadian rhythm in healthy premenopausal women. J. Interdiscpl. Cycle Res. 24, 72 – 78. Garnero P, Hausherr E, Chapuy M, Marcelli C, Grandjean H, Muller C, Cormier C, Breart G, Meunier P & Delmas P (1996a): Markers of bone resorption predict hip fracture in elderly women: The EPIDOS prospective study. J. Bone Miner. Res. 11, 1531 – 1538. Garnero P, Sornay-Rendu E, Chapuy M & Delmas P (1996b): Increased bone turnover in late postmenopausal women is a major determinant of osteoporosis. J. Bone Miner. Res. 11, 337 – 349. Garnero P, SornayRendu E, Duboeuf F & Delmas P (1999): Markers of bone turnover predict postmenopausal forearm bone loss over 4 y: the OFELY study. J. Bone Miner. Res. 14, 1614 – 1621. Goldhaber P & Rabadjija L (1987): H þ stimulation of cell-mediated bone resorption in tissue culture. Am. J. Physiol. 253, E90 – 98. Hotta M, Fukuda I, Sato K, Hizuka N, Shibasaki T & Takano K (2000): The relationship between bone turnover and body weight, serum insuline-like growth factor (IGF) I, and serum IGF-binding protein levels in patients with anorexia nervosa. J. Clin. Endocrinol. Metab. 85, 200 – 206. Jagielska G, Wolan´ czky T, Komender J, Tomaszewicz-Libudzic C, Przedlacki J & Ostrowski K (2001): Bone mineral content and bone mineral density in adolescent girls with anorexia nervosa-a longitudinal study. Acta Phychiatr. Scand. 104, 131 – 137. ˜a S (1999): Neutron activation analysis deterKehayias JJ & Valtuen mination of body composition. Curr. Opin. Clin. Nutr. Metab. Care 2, 453 – 463.

Lennkh C, de Zwaan M, Bailer U, Strnad A, Nagy C, el-Giamal N, Wiesnagrotzki S, Vytiska E, Huber J & Kasper S (1999): Osteopenia in anorexia nervosa: specific mechanisms of bone loss. J. Psychiatr. Res. 33, 349 – 356. Lohman TC, Roche AF & Martorell R (1988): Antropometric Standardization Reference Manual. Chicago, IL: Human Kinetics. Lucas AR, Melton LJ, Crowson CS & O’Fallon WM (1999): Long-term fracture risk among women with anorexia nervosa: a populationbased cohort study. Mayo Clin. Proc. 74, 972 – 977. Marowska J, Kobylinska M, Ukaszkiewicz J, Taajko A, RymkiewiczKluczynska B & Lorenc R (1996): Pyridinium crosslinks of collagen as a marker of bone resorption rates in children and adolescents: normal values and clinical application. Bone 19, 669 – 677. Martin AD (1991): Antropometric assessment of bone mineral. In: Antropometric Assessment of Nutritional Status, ed. JH Himes, pp 184 – 196. Minneapolis, MN: Wiley-Liss. McLaren A, Isdale A, Whitings P, Bird H & Robins S (1993): Physiological variations in the urinary excretion of pyridinium crosslinks of collagen. Br. J. Rheumatol. 32, 307 – 312. Polito A, Cuzzolaro M, Raguzzini A, Censi L & Ferro-Luzzi A (1998): Body composition changes in anorexia nervosa. Eur. J. Clin. Nutr. 52, 655 – 662. Polito A, Fabbri A, Ferro-Luzzi A, Cuzzolaro M, Censi L, Ciarapica D, Fabbrini E & Giannini D (2000): Basal metabolic rate in anorexia nervosa: relation to body composition and leptin concentrations. Am. J. Clin. Nutr. 71, 1495 – 1502. Pratt DA, Daniloff Y, Duncan A & Robins SP (1992): Automated analysis of the pyridinium crosslinks of collagen in tissue and urine using solid-phase extraction and reversed-phase high-performance liquid chromatography. Anal. Biochem. 207, 168 – 175. Rauch F, Rauch R, Woitge H, Seibel M & Schonau E (1996): Urinary immunoreactive deoxypyridinoline in children and adolescents: variations with age, sex and growth velocity. Scand. J. Clin. Lab. Invest. 56, 715 – 719. Rigotti NA, Nussbaum SR, Herzog DB & Neer RM (1984): Osteoporosis in women with anorexia nervosa. New Engl. J. Med. 311, 1601 – 1606. Soyka LA, Grinspoon S, Levitsky LL, Herzog DB & Klibanski A (1999): The effects of anorexia nervosa on bone metabolism in female adolescents. J. Clin. Endocrinol. Metab. 84, 4489 – 4496. Spierto FW, MacNeil ML & Burtis CA (1979): The effect of temperature and wavelength on the measurement of creatinine with the Jaffe` procedure. Clin. Biochem. 12, 18 – 21. Stefanis N, Mackintosh C, Abraha HD, Treasure J & Moniz C (1998): Dissociation of bone turnover in anorexia nervosa. Ann. Clin. Biochem. 35, 709 – 716. Valla A, Groenning IL, Syversen U & Hoeiseth A (2000): Anorexia nervosa: slow regain of bone mass. Osteoporos. Int. 11, 141 – 145. ˜a S, Sette S & Branca F (2001): Influence of Mediterranean Valtuen diet and Mediterranean lifestyle on calcium and bone metabolism. Int. J. Vit. Nutr. Res. 71, 189 – 202. Vesper HW, Demers LM, Eastell R, Garnero P, Kleerekoper M, Robins SP, Srivastava AK, Warnick GR, Watts NB & Myers GL (2002): Assessment and recommendation on factors contributing to variability of preanalytical urinary pyridinoline and deoxypyridinoline. Clin. Chem. 48, 220 – 235. Wachman A & Bernstein DS (1968): Diet and osteoporosis. Lancet 1, 958 – 959. Ward A, Brown N & Treasure J (1997): Persistent osteopenia after recovery from anorexia nervosa. Int. J. Eat. Disord. 22, 71 – 75. ¨ we B, Beumont PJ, Kasperk C & Herzog W Zipfel S, Seibel MJ, Lo (2001): Osteoporosis in eating disorders: a follow-up study of patients with anorexia and bulimia nervosa. J. Clin. Endocrinol. Metab. 86, 5227 – 5233.

265

European Journal of Clinical Nutrition