Received 13 February 2002; accepted 26 April 2002

Clinica Chimica Acta 322 (2002) 121 132 www.elsevier.com/locate/clinchim Is the predictive power of previous fractures for new spine and non-spine fractures associated with biochemical evidence of altered bone remodelling? The EPOS study P. Vergnaud a, M. Lunt b,c, C. Scheidt-Nave d, G. Poor e, C. Gennari f, K. Hoszowski g, A. Lopes Vaz h, D.M. Reid i, L. Benevolenskaya j, S. Grazio k, K. Weber l, T. Miazgowski m, J.J. Stepan n, P. Masaryk o, F. Galan p, J. Bruges Armas q, R. Lorenc g, S. Havelka n, R. Perez Cano p, M. Seibel d, G. Armbrecht r, S. Kaptoge b, T.W. O Neill c, A.J. Silman c, D. Felsenberg r, J. Reeve b, *, P.D. Delmas a a U. INSERM 403, Lyon, France b Strangeways Research Laboratory, Institute of Public Health, Cambridge CB1 8RN, UK c ARC Epidemiology Unit, Manchester, UK d University of Heidelberg, Heidelberg, Germany e National Institute of Rheumatology and Physiotherapy, Budapest, Hungary f Institute of Clinical Medicine, University of Siena, Siena, Italy g PKP Hospital, Warsaw, Poland h Hospital de San Joao, Oporto, Portugal i Department of Medicine and Therapeutics, University of Aberdeen, Aberdeen, UK j Institute of Rheumatology, Moscow, Russia k Clinical Hospital, Zagreb, Croatia l Department of Medicine, University Hospital, Graz, Austria m Academy of Medicine, Szczecin, Poland n Institute of Rheumatology, Charles University, Prague, Czech Republic o Institute of Rheumatic Diseases, Piestany, Slovakia p Department of Medicine, University of Sevilla, Seville, Spain q Hospital de Santo, Azores, Portugal r University Klinikum Benjamin Franklin, Berlin, Germany Received 13 February 2002; accepted 26 April 2002 Abstract Background: In the European Prospective Osteoporosis Study (EPOS), a past spine fracture increased risk of an incident fracture 3.6 12-fold even after adjusting for BMD. We examined the possibility that biochemical marker levels were * Corresponding author. Department of Medicine (Box 157), Addenbrooke s Hospital, Hills Road, Cambridge CB2 2QQ, England, UK. Tel.: +44-1223-741617; fax: +44-1223-741618. E-mail address: j.reeve@srl.cam.ac.uk (J. Reeve). 0009-8981/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved. PII: S0009-8981(02)00164-X

122 P. Vergnaud et al. / Clinica Chimica Acta 322 (2002) 121 132 associated with this unexplained BMD-independent element of fracture risk. Methods: Each of 182 cases in EPOS of spine or non-spine fracture that occurred in 3.8 years of follow-up was matched by age, sex and study centre with two randomly assigned never-fractured controls and one case of past fracture. Analytes measured blind were: osteocalcin, bone-specific alkaline phosphatase, total alkaline phosphatase, serum creatinine, calcium, phosphate and albumin, together with the collagen cross-links degradation products serum CTS and urine CTX. Most subjects also had bone density measured by DXA. Results: Cases who had recent fractures did not differ in marker levels from cases who had their last fracture more than 3 years previously. No statistically significant effect of recent fracture was found for any marker except osteocalcin, which was 17.6% lower in recent peripheral cases compared to unfractured controls ( p < 0.05) and this was independent of BMD. Conclusion: Past fracture as a risk indicator for future fracture is not strongly mediated through increased bone turnover. D 2002 Elsevier Science B.V. All rights reserved. Keywords: Osteoporosis; Fractures; Biochemical markers; Bone turnover; Epidemiology; Bone densitometry 1. Introduction There is now clear evidence that low bone density predicts both vertebral and nonvertebral fractures in postmenopausal women [1] and men over the age of 50 [2]. Furthermore, evidence from several cohort studies and trials has shown that a previous fracture is a strong risk factor for a future fracture, which is independent of bone density [3,4]. While the biomechanical basis for the effect of low bone density on fracture risk is easily understandable, in contrast, it remains conjectural why a past fracture should predict future fractures independently of bone density. One possibility might be that conventional bone densitometry does not easily represent key factors determining the structural strength of the bones measured. It has been postulated, for example, that strength (or toughness) of human bones might be quite strongly dependent on properties not measurable by densitometry, such as the accumulation of microcracks [5], small variations in average mineralisation density [6] or the rate of initiation of new cycles of bone remodelling at the organ level or their duration at the BMU level. Increased remodelling might increase bone fragility because in cancellous (trabecular) bone, in particular, remodelling causes transient trabecular thinning as well as a finite risk of trabecular perforation. In line with this suggestion, there is now increasing evidence that several biochemical markers of increased bone turnover predict hip [7,8] or other non-spine fractures [9 11] as well as spine fractures [11]. However, results with the bone formation markers have been less consistent than with the newer resorption markers and some studies have suggested that osteocalcin in particular may even be reduced in fracture cases [12]. We decided to study biochemical markers of bone turnover in a large cohort study of osteoporotic fractures, the European Prospective Osteoporosis Study (EPOS). This was to investigate the mechanism underlying the increased risk of a new fracture attributable to a documented previous fracture rather than to the component of risk associated with low bone density. We have undertaken a case control study of biochemical markers in men and women who suffered incident spine and non-spine fractures and compared the values obtained with those from controls matched for investigational centre, age and gender. A second set of matched controls were included with baseline prevalent, but not incident, fractures so that any effect of the recent status of a fracture could be monitored. The large majority of subjects also had bone densitometry using DXA of the spine or hip. The aim of the study was to see if clear differences emerged in biomarker levels between incident and prevalent fracture cases on the one hand and controls in the other. If this was the case, it would allow us to postulate a biologic pathway linking the so far unexplained fracture risk associated with a previous fracture to abnormalities of bone turnover. 2. Materials and methods 2.1. Subjects and X-rays The subjects followed up were those who had participated in the EVOS prevalence survey and this

P. Vergnaud et al. / Clinica Chimica Acta 322 (2002) 121 132 123 is described in detail elsewhere [13 17]. In brief, participating centres aimed to recruit 600 subjects aged 50 to 79 (50 males and females in each 5-year age band) from population-based registers. Subjects were invited to attend for an interview administered questionnaire and a spinal radiograph. Participating centres undertook annual postal follow-up to determine the interval occurrence of a peripheral limb fracture. At the second scheduled follow-up, 29 of the centres invited the responders to attend for a follow-up radiograph and 17 of these 29 centres also invited their subjects to provide a single blood and urine sample at approximately the time of the second X-ray allowing their participation in the present study. Lateral spinal radiographs of the dorsal and lumbar spine followed a structured protocol including film focus distance, breathing technique and film centring as detailed at the baseline film [15,18]. Each subject gave written informed consent to their participation in this study in the manner required by their centre s Research Ethics Committee. 2.2. Collection of biochemistry samples All subjects who replied to the second EVOS questionnaire were sent a letter inviting them to attend for marker studies. Subjects not replying to the first letter of invitation were sent a reminder 4 weeks after the first invitation to attend. On the day before the appointment for phlebotomy, subjects were instructed to follow their normal diet with the exception that they were requested to take as little dairy produce, fat and alcohol as possible. If it was possible to offer a morning appointment (which was strongly preferred), the subject was told that the best results could be achieved if the sample was taken after fasting since the evening before. However, if the subject had to have breakfast before the appointment, they were instructed to have a light meal with as little dairy produce and fat as possible. For afternoon appointments, lunch was instructed to be as small as possible and to contain no dairy produce or fat. Urine samples could be collected in one of two ways: as an overnight urine including the first morning void, or (in the case of morning appointments only) as a 2-h fasting morning urine. The quantity of blood required from each subject was 8 ml. Ideally, centrifugation (10 min at 1000 g) occurred within 1 h of collection, or if the blood was kept at 4 jc, a maximum time lapse of 3 h was accepted. The serum was aliquoted into 3 1ml samples and each tube labelled with freezer resistant labels (tubes, labels and pens provided centrally) before the samples were frozen in temporary storage at 20 jc to 70 jc). The phlebotomy entry in the Biochemistry Record Book was then made immediately. The urine sample provided by the subject was mixed and 2 1 ml aliquots of urine placed into identical labelled tubes and frozen. In the record book, the date and time the subject was bled and the time the sample was centrifuged were recorded. A note was made whether the subject collected an overnight sample of urine or a 2-h fasting morning urine sample. It was also recorded whether or not the subject fasted for 5 h or more prior to the taking of blood. Finally, the temperature at which the blood samples were stored prior to dispatch, the number of aliquots of serum collected and the date on which the samples were sent to Lyon were all recorded. Samples were all couriered by air, packed in dry ice, inspected for their state of preservation and temperature and, if these were satisfactory, stored at 80j or lower on receipt in Lyon. 2.3. Analysis of samples Homogeneity of analysis was of vital importance in achieving validity. Therefore, the analysis of samples was undertaken in a single laboratory in Lyon, France. The Lyon laboratory assayed the samples without knowing the case control status of any subject. The biochemical markers analysed were: in serum intact parathyroid hormone (PTH); 25-hydroxy-vitamin D [25(0H)VitD]; albumin; creatinine; osteocalcin; total and bone alkaline phosphatase; serum C-terminal peptides of type 1 collagen (CTS) [19] and in the urine C-terminal peptides of type 1 collagen (CTX) [20] and creatinine. 2.4. Ascertainment of spine fracture/deformity Subjects in the study were classified as having one or more incident spine deformities if such a deformity was detected as described below. Subjects with a prevalent spine deformity were identified as described

124 P. Vergnaud et al. / Clinica Chimica Acta 322 (2002) 121 132 by O Neill et al. [15] using the McCloskey Kanis method [15,21] applied to the first X-ray, if they had no further deformity on the second film. All X-rays were captured as digital images on receipt of the film and then all assessable vertebral body images were immediately quantitated by placing six points to define the anterior, mid- and posterior vertebral body heights of the vertebral body in question. A qualitative diagnostic reading was also performed at this time. When the follow-up film had been quantitated, the results were compared with the results from the first film and all vertebral bodies showing an absolute change in the anterior/posterior or mid-body/ posterior ratio greater than 15% identified. Also, all vertebral bodies with one of these ratios less than 75% were identified. All film pairs scoring positive for one or more vertebrae by either of these criteria were set aside for review (50% of the data set). In addition, for three centres, all films were reviewed without exception, to be certain that this selection procedure did not miss any clinical fractures. Then all film pairs were reviewed side-by-side on the same screen with image enhancement (magnification, contrast) by a single experienced radiologist who made a clinical judgement as to whether there had been a fracture event in any spinal body between the two films [18]. When appropriate, the points defining the vertebral body dimensions were adjusted. Then the morphometric readings were analysed after adjusting the images to the same magnification [22]. Our approach was modelled on our pilot study [23]. Deformed vertebrae on the second film were identified by the McCloskey Kanis criteria [15,21]. To qualify as an incident deformity, a deformity on the second film had to have lost 4 mm in height in at least one of its measured dimensions and be reduced by at least 20%. Any subject who satisfied either the morphometric criteria or was judged clinically to have an incident vertebral deformity was included as a case. 2.5. Ascertainment of non-spine fractures The procedures for identifying cases of nonvertebral fracture have been previously described, together with the results of a successful validation study, in which estimates of false positive and false negative rates of ascertainment were calculated [24]. Briefly, at intervals of 12 24 months (6 in Germany), questionnaires that asked about occurrence of fractures since the previous questionnaire were completed. Each questionnaire included a diagrammatic mannikin on which the subject was asked to indicate the location of any recorded fracture. Validation was by reference to X-rays and X-ray reports where possible, failing that by reference to medical notes or interview with the subject. Duplicate reports were eliminated by reference to notes and X-rays, as well as by cross-examination of the subject by the study nurse. 2.6. Bone densitometry In all centres, most or all subjects were also asked to agree to bone densitometry of the spine, hip or both regions of interest, using dual X-ray absorptiometry as previously described [25]. Densitometers in the study were cross-calibrated with the European Spine Phantom (ESP, definitive version) as previously described [26,27]. 2.7. Choice of subjects for the case control analysis All subjects with one or more documented incident spine or non-spine fracture occurring since the inception of the study were selected. There were defined as cases suffering recent fractures. For each case, three controls were selected by the study statistician who were matched to the cases by age, study centre and gender. Two controls were selected who had no history of fracture and the third control on the basis of a prevalent fracture at baseline (spine or non-spine, as appropriate) but no incident fracture during the study (cases of pre-study fracture). The study numbers of the nearly 1000 subjects so selected were submitted to the first author for biochemical analysis, but the Lyon biochemistry centre as a whole was blinded as to the fracture status of the subjects. 2.8. Statistical analysis The analysis was undertaken in two stages. First, we tested to see whether there was an effect of the occurrence of fracture in the recent past. We did this by comparing cases of pre-study fracture with cases of recent fracture. Next, we compared cases of recent fracture with their other two, unfractured, controls.

P. Vergnaud et al. / Clinica Chimica Acta 322 (2002) 121 132 125 To estimate the difference between recent cases and controls for each of the analytes, random effects models were used. Random effects models do not assume that observations are independent, as a generalised linear model would, but that the observations occur in clusters (in this case, the case control groups, i.e. investigational centres). Thus, differences in outcome between centres in age, sex and standardised procedures within the individual centres for handling the samples are adjusted for, and only the within-cluster differences between cases and controls are estimated. Since age, gender and centre did not vary within case control groups (due to the matching), differences in analyte levels due to differences in these variables were estimated between case control groups. It should be noted that in these types of analysis, it is only the differences between cases and controls that are measured with optimal accuracy. Although reported, the difference between centres cannot be regarded as reliable because subjects for analysis in each centre were chosen according to the subjects with a recent fracture in each centre, rather than being representative of the centre. Table 1 (a) Types of incident peripheral fractures (only first fracture is given) Fracture type ICD code Men Women Total Neck of femur 811 0 1 1 Tarsal/metatarsal 814 2 14 16 Rib/sternum 815 6 6 12 Humerus 816 1 4 5 Radius and ulna 824 4 25 29 Tibia and fibula 826 4 7 11 Other non-limb 2 5 7 Other limb 8 13 21 Total 27 75 102 (b) Ages and genders of recent fracture cases Number Male vertebral fracture cases Female vertebral fracture cases Male nonvertebral fracture cases Female nonvertebral fracture cases Mean age (S.D.) 18 68.8 (9.7) 62 69.1 (6.9) 27 68.1 (8.9) 75 67.3 (7.8) 3. Results 3.1. Cases identified A total of 80 subjects (62 female) were identified as cases of recent spine fracture of whom 22 had multiple fractures. There were 135 recent non-spine fractures in 102 subjects (75 female). Table 1 shows the numbers and the distribution of fracture types seen in the recent fracture cases, together with their age and gender distribution. Only one suitable control could be matched to 8 subjects, 2 controls to a further 8 but 3 controls were found for each of the remaining 166. 3.2. Descriptive biochemical results Table 2 shows the means and interquartile ranges of the biochemical analyte data in fracture cases and age- and sex-matched controls. In all groups, the results needed a logarithmic transformation in order to normalise the distributions. The data are grouped according to fracture status. It should be noted that the controls are not representative of the generality of EPOS populations because they were selected to match the cases for age and sex. 3.3. Comparisons between groups 3.3.1. Recent fracture cases compared with pre-study fracture cases When the recent fracture cases were compared with the previous-fracture controls, there was no significant effect of recent fracture compared with previous fracture, in either direction. Thus, for the analytes listed in Table 2, the p values for the between group differences ranged from 0.09 to 0.76, with the two resorption markers giving p values of 0.60 and 0.63, and the two formation markers p values of 0.76 and 0.89. In men only, the range of p values was from 0.06 to 0.81 and in women only from 0.08 to 0.98. When the vertebral and nonvertebral fractures were considered separately, the closest to a significant difference was in the serum calcium of the nonvertebral fractures ( p = 0.06), which was regarded as being nonsignificant because of multiple testing.

126 Table 2 (a) Levels of analytes in men in the various fracture groups: median (interquartile range) Vertebral Peripheral Recent fracture Pre-study fracture No fracture Recent fracture Pre-study fracture No fracture Osteocalcin (ng/ml) 17.0 (14.5, 21.8) 18.2 (10.6, 25.2) 17.0 (10.0, 22.4) 14.1 (10.9, 18.5) 16.5 (11.6, 23.1) 17.8 (12.5, 23.9) Bone alkaline phosphatase (U/l) 18.4 (14.5, 20.3) 17.3 (14.7, 26.4) 18.1 (13.3, 27.1) 17.3 (13.2, 20.4) 15.1 (13.4, 19.1) 17.8 (14.5, 21.6) Serum CTS (ng/ml) 38 (23, 50) 33 (21, 45) 30 (15, 52) 24 (15, 43) 27 (16, 39) 32 (21, 43) Corrected urine CTX (mg/mol creatinine) 188 (134, 226) 152 (96, 190) 160 (80, 299) 157 (99, 209) 159 (84, 233) 164 (109, 232) 25(OH) vitamin D (ng/ml) 13.9 (10.7, 22.4) 14.3 (6.3, 20.6) 17.9 (9.8, 23.4) 17.4 (13.7, 25.7) 25.3 (17.3, 28.6) 18.8 (12.5, 23.8) Ca 2+ (mg/l) 93 (92, 95.5) 94.0 (92.0, 96.0) 95.0 (93.0, 96.0) 94.0 (91.0, 96.0) 95.0 (93.0, 97.0) 94.0 (91.0, 95.0) Phosphate (mg/l) 32 (29.5, 34) 31.0 (29.0, 35.0) 34.5 (29.0, 41.0) 34.5 (29.0, 36.3) 32.0 (28.0, 41.0) 32.0 (28.5, 35.5) Total alkaline phosphatase (UI/l) 51 (42.5, 61) 51.0 (43.5, 74.5) 54.5 (40.0, 68.5) 46.0 (33.8, 54.3) 39.0 (32.5, 52.0) 42.0 (36.0, 54.5) Serum creatinine (mg/l) 10.3 (8.8, 11.7) 10.6 (9.7, 12.3) 9.9 (8.4, 13.3) 10.2 (8.8, 11.8) 10.2 (8.5, 12.1) 10.0 (8.7, 11.4) Albumin (g/l) 48 (46, 50.5) 48 (46, 53) 49.5 (47, 52) 48 (46, 51) 49 (45.5, 52.5) 48 (47, 51.5) (b) Levels of analytes in women in the various fracture groups: median (interquartile range) Vertebral Peripheral Recent fracture Pre-study fracture No fracture Recent fracture Pre-study fracture No fracture Osteocalcin (ng/ml) 22.2 (16.0, 28.9) 21.8 (15.3, 29.9) 19.1 (15.4, 26.8) 20.0 (14.5, 27.3) 20.5 (15.1, 27.9) 22.3 (15.0, 28.4) Bone alkaline phosphatase (U/l) 21.2 (17.2, 27.4) 20.8 (16.9, 24.0) 20.8 (16.3, 26.1) 20.2 (15.9, 24.6) 20.2 (16.3, 26.0) 20.9 (16.8, 26.2) Serum CTS (ng/ml) 36 (24, 56) 36 (23, 54) 34 (25, 50) 30 (18, 48) 34 (19, 50) 33 (23, 53) Corrected urine CTX (Ag/mM creatinine) 222 (137, 330) 243 (154, 321) 218 (144, 298) 219 (128, 307) 264 (151, 346) 233 (153, 343) 25(OH) vitamin D (ng/ml) 18.9 (12.4, 26.1) 16.6 (11.7, 24.2) 16.0 (11.9, 21.6) 15.4 (9.9, 22.0) 15.4 (10.5, 22.2) 18.8 (12.6, 24.4) Ca 2+ (mg/l) 95.0 (93.0, 98.0) 96.0 (93.0, 99.0) 95.0 (92.0, 97.0) 95.0 (93.0, 97.0) 95.0 (93.0, 97.0) 95.0 (93.0, 97.0) Phosphate (mg/l) 37.0 (32.5, 38.5) 37.0 (33.0, 41.0) 36.0 (33.0, 39.0) 36.0 (33.0, 39.0) 36.0 (33.0, 42.0) 37.0 (33.0, 40.0) Total alkaline phosphatase (UI/l) 59.0 (45.5, 69.0) 57.0 (44.0, 71.0) 52.0 (42.0, 64.8) 51.5 (42.0, 60.0) 52.0 (44.0, 67.0) 54.0 (46.0, 64.5) Serum creatinine (mg/l) 8.6 (7.6, 10.0) 8.1 (7.1, 9.3) 9.0 (7.9, 10.1) 9.0 (7.8, 10.1) 8.9 (7.6, 10.2) 8.4 (7.5, 9.6) Albumin (g/l) 49.0 (45.0, 52.0) 48 (47, 50) 49 (46, 51) 49 (47, 53) 48 (47, 50) 49 (46, 51) P. Vergnaud et al. / Clinica Chimica Acta 322 (2002) 121 132

P. Vergnaud et al. / Clinica Chimica Acta 322 (2002) 121 132 127 3.3.2. Recent fracture cases compared with neverfractured controls We therefore proceeded to compare fracture cases with controls without making any assumption of an effect of recent fracture to transiently increase marker levels. The random effects case control analysis applied to the data in Table 2 showed no statistically significant effect ( p > 0.18) of recent fracture on any marker except osteocalcin, which was 17.6% lower in recent peripheral (but not spinal) cases than controls ( p < 0.05). A similar, nonsignificant trend was noted for bone-specific alkaline phosphatase. The study had a posteriori power of >95% to detect a difference of 0.2 population S.D. between cases and controls for all markers. 3.4. Adjusting for possible confounding by BMD BMD was measured on 61 recent vertebral fracture cases, 71 recent peripheral fracture cases and 405 controls. Using random effects models, BMD was greater at the spine in normal never-fractured controls than in either recent cases or pre-study fracture cases, for both vertebral and peripheral groups. BMD was also reduced at the femoral neck in recent and prestudy vertebral fracture cases, but not peripheral fracture cases. There was no significant difference in BMD between recent and pre-study fracture cases at any site (see Table 3). When the random effects modelling was re-run including both BMD and biochemical marker data, both recent and pre-study peripheral fracture cases had significantly lower osteocalcin levels, after adjusting for BMD at the proximal femur (femoral neck or trochanter). Otherwise, the results were similar to those obtained when no adjustment for BMD was made. In summary, the only marker showing a significant BMD-independent association with fracture was osteocalcin and that association was restricted to non-spine fractures. 4. Discussion The principal purpose of this study was to allow us to capture the biochemical correlates of a prior fracture, in order to test the hypothesis that the large component of fracture risk that is associated with prior fracture and is also independent of BMD [3] can be attributed to high bone turnover and by implication non-site directed remodelling of bone. We could not confirm that there was any general effect of any marker to be positively associated with previous spine fracture, Indeed, as observed by Åkesson et al. [12], peripheral fractures within the last 4 years were Table 3 Effects of age, sex and investigational centre on the log means of biochemical analytes and BMD values Increase with age per decade (95% CI) Excess in women (95% CI) Between centre differences ( p) Log of osteocalcin (ng/ml) 8.4% (3.2%, 13.9%) 28.5% (17.7%, 40.4%) 0.0087 Log of bone alkaline 1.2% ( 2.9%, 5.5%) 19.0% (10.5%, 28.2%) 0.0333 phosphatase (U/l) Log of serum CTS (ng/ml) 13.9% (4.6%, 23.9%) 25.8% (8.0%, 46.4%) < 0.0001 Log of corrected urine CTX 2.5% ( 4.0%, 9.5%) 50.4% (33.3%, 69.6%) < 0.0001 (Ag/mM Creatinine) Log of 25(OH) vitamin D (ng/ml) 18.4% ( 23.2%, 13.3%) 14.8% ( 23.7%, 5.0%) < 0.0001 Log of Ca 2+ (mg/l) 0.3% ( 0.7%, 0.1%) 0.9% (0.1%, 1.7%) 0.0024 Log of phosphate (mg/l) 1.9% ( 3.6%, 0.1%) 8.6% (5.1%, 12.2%) < 0.0001 Log of total alkaline 3.2% ( 0.5%, 7.0%) 13.8% (6.6%, 21.6%) < 0.0001 phosphatase (UI/l) Log of serum creatinine (mg/l) 6.6% (4.1%, 9.2%) 17.0% ( 20.5%, 13.3%) 0.0001 Log of albumin (g/l) 0.9% ( 1.7%, 0.2%) 0.2% ( 1.5%, 1.1%) < 0.0001 Spine BMD 0.004 g/cm 2 ( 0.046, 0.039) 0.165 g/cm 2 ( 0.231, 0.093) 0.9545 Neck BMD 0.043 g/cm 2 ( 0.057, 0.030) 0.103 g/cm 2 ( 0.126, 0.078) 0.0068 Trochanter BMD 0.026 g/cm 2 ( 0.040, 0.012) 0.141 g/cm 2 ( 0.164, 0.117) 0.0075

128 P. Vergnaud et al. / Clinica Chimica Acta 322 (2002) 121 132 associated with lower, not higher, osteocalcin levels even after adjusting for BMD. Although raised marker levels have been reported for up to 1 year following fracture [28], we found no association of high marker levels with either recent or less recent prevalent fractures, and there was no contrast between the two groups of fracture cases in this respect. It is necessary to conclude that under the conditions of the present study, markers of high bone turnover were not capable of identifying the same BMD-independent component of future fracture risk as a documented previous fracture. We do not believe that this study s retrospective design invalidates our general conclusion that there are no grounds for believing bone turnover at the high end of the age- and sex-matched distribution to be a major risk factor for fracture, unless it also leads to reduced BMD. EPOS began as a prevalence study nearly a decade ago when biochemical markers were of less interest since they were then less predictive of bone loss. Biochemical markers are now well established as tools in the investigation of the determinants of bone loss [29]. A wealth of data exists to show that markers of bone formation and bone resorption are capable of predicting bone loss rates in groups of untreated postmenopausal women, in a variety of disorders leading to secondary osteoporosis and also in noncompliant patients who are not taking anti-osteoporosis medications. The degree of precision with which bone loss can be predicted in the individual has been the subject of considerable debate [30], and consensus has yet to be reached on the place of biochemical markers in the clinical diagnosis of osteoporosis, as distinct from the management of the patient assigned to treatment [31,32]. Studies of markers using fracture as the principle end point have been fewer than those relating markers to bone density [33]. Riis et al. [10] performed a longterm follow-up study in a postmenopausal group of women and found that biochemical predictors of fast (vs. slow) bone loss also predicted fractures with an odds ratio of about 2. The EPIDOS and Rotterdam hip fracture studies have been used to test the predictive power of markers for this important fracture and an independent effect of markers particularly those of bone resorption was found [7,8,34], which was largely independent of and additive to that of BMD [33]. The OFELY study, which was also prospective, found that urinary CTX, but not other markers, was significantly predictive of non-spine fractures, of which only 14% were hip fractures, suggesting that the prediction of fracture extended to other fractures besides hip fractures [9]. Ross and his colleagues in the Hawaii Osteoporosis Study [11] found that bone alkaline phosphatase and serum CTX were predictive of spine and non-spine fractures occurring in the next 2.7 years. Under-carboxylated osteocalcin, thought to be a marker of poor vitamin K status and possibly also poor vitamin D status, was found to be predictive of hip fracture in institutionalised women by Szulc et al. [35] and of non-spine fractures by Luukinen et al. [36]. In general, these prospective studies have supported the concept that high bone turnover in postmenopausal women leads to accelerated bone loss and a consequent increase in risk of fracture, with the best markers capturing an element of fracture risk that is largely independent of BMD, either because of the know inaccuracies of BMD or for other reasons [33]. However, Weel et al. [37] have found a substantial contrast between, on the one hand wrist fractures and on the other hip and humerus fractures, with the biochemical markers being much more predictive of the latter (socalled frailty) than the former ( activity ) group of fractures. Thus, when groups of patients are stratified by bone resorption levels, an association of markers with hip and non-spine fractures is apparent, at least in some studies. However, the SOF study [38] showed no relationship between marker levels and fracture rates. We undertook a retrospective rather than a prospective study from necessity because funding for further follow-up of our subjects after venesection was denied. We therefore undertook first a careful exploration of the data to see if we could detect an effect of a recent fracture on any marker level by comparing levels in those who had fractures within the last 3.8 years with those who had fractures previously. Although effects of recent fracture on marker levels have been reported, these have tended to be maximal during the healing phase, i.e. in the first year after fracture [28], in the present study, we could detect no significant effect of the timing of the last fracture on any marker and we concluded that undertaking the present analysis on retrospective data was unlikely to cause confounding of our results. In considering the contrast between our results and those of Garnero on hip fracture [7], we only had one

P. Vergnaud et al. / Clinica Chimica Acta 322 (2002) 121 132 129 case of hip fracture in this younger population. Our results are concordant with the recently updated results of the Rotterdam study [37]. The association of high bone turnover markers, more specifically with hip and other frailty fractures, might in part be secondary to vitamin D deficiency osteopathy and secondary hyperparathyroidism; it will be noted that our cases and controls had similar 25(OH) vitamin D levels so that increased bone turnover driven by vitamin D deficiency leading directly to secondary hyperparathyroidism and osteopathy could not have contributed importantly to the fractures we observed. This study has a number of limitations. In a multicentre study where samples have to be collected from subjects living in the community, it is necessary to consider whether the conditions under which the samples were taken or stored could have contributed to the similarities between the results we observed in cases and controls. The protocol used for sample collection and storage was as well-specified as was practical and contemporary records were kept of actual practice in the centres. However, it was found not to be possible for the centres to use precisely identical protocols to collect the samples. Therefore, we could not completely control the times of day at which samples were taken because for practical reasons nurses had often to travel between several patients homes on any particular morning, so in part negating the complete control over diurnal rhythms affecting several markers [30] that we would ideally have liked to impose. Therefore, the statistically significant differences between the centres (as shown in Table 3) may be partly due to differences in sample handling and subject selection (as described in Materials and methods) and cannot be ascribed solely to biological differences between groups of subjects. The between centres difference might also be due to seasonal variations in bone resorption that reflect rhythms in serum vitamin D and PTH concentrations [39] and geographic variations [40]. Finally, the control data cannot be used for comparing centres for normal analyte values because the controls for analysis were selected specifically to match the cases as to age and gender. The laboratory assays were chosen for their robustness in the face of nonideal sampling times in relation to separation and chilling of sera. However, we have no reason to suspect that this would have caused any systematic alteration in the ratios of the results we obtained in the fracture cases and the controls. This is because cases and controls were matched for centre and, in the case of the vertebral fractures at least, the nurses drawing the blood had no knowledge of the fracture status of the subjects. This was not a study of the ability of biochemical markers to predict fractures. For that purpose, the study protocol should have included venesection at recruitment into the EVOS study, at the time of the first X-ray. This was considered at the start up meeting for the study, but resource constraints at the time made it impractical. We went ahead with this retrospective study because we wanted to probe the biological mechanisms underlying the strong component of bone fragility which is tracked by a previous fracture, independently of measured bone density. In two companion papers on the EPOS study, we have found that both the presence and the shape of a baseline vertebral fracture are strong determinants both of the risk of an incident spine [41] or hip fracture [42] and in the case of spine fractures of its severity as assessed by the volume loss of the affected vertebral body [43]. Since the component of risk tracked by a previous fracture appeared not to be associated significantly with increases in bone turnover markers in these EPOS subjects, alternative explanations must be sought for bone fragility in those subjects who suffered non-traumatic fractures. The possibility exists that some individuals have less fragile (or tougher) bone than others at a given level of BMD. The basis for such a difference might be biochemical, such as subtle differences in the bone collagen as have been postulated to result from the Sp1 polymorphism [44], differences in cross-linking between cases and controls [45], or degradation in some individuals of the ideal structural anisotropy or connectivity of the cancellous bone of the spine. In cortical fracture sites, we have shown elsewhere that it is possible for localised osteoporosis (in this case, in one quadrant of the femoral neck) to co-exist with apparently normal bone in the other three quadrants in cases of hip fracture [46 48]. Furthermore, van der Linden et al. [49] have found that the mechanical consequences of microscopic defects in trabecular structures can be severe, even if they are temporary as occurs during the resorption phase of bone remodelling. In conclusion, we have found that biochemical markers of bone turnover, as well as 25(OH) vitamin

130 P. Vergnaud et al. / Clinica Chimica Acta 322 (2002) 121 132 D levels measured after fracture are not associated significantly with recent previous fracture, except for a weak tendency for markers of bone formation to be reduced in cases of previous non-spine fractures. Ongoing studies, such as the OFELY study, will answer the question as to whether biochemical markers have a place, alongside fracture history, bone densitometry and critical examination of spine X-rays, in the identification of individuals who will fracture in future. Meanwhile, to understand how bone fragility develops in some individuals more than others at a given level of bone density, increased attention should be directed towards biomechanical studies of microstructure, crack propagation, crack repair through remodelling and other means by which bone becomes less mechanically tough with age [5,6]. These should be conducted alongside studies of the relationship of bone s toughness to the biochemical composition, including cross-linking, of its organic and mineral components and of the determinants of the composition of the matrix, both genetic and environmental. Acknowledgements The study was financially supported by European Union Concerted Action Grants under Biomed-1 (BMH 1CT 931448 and 920182), and also EU grants C1PDCT925102, ERBC1PDCT 930105 and 940229. The central coordination was also supported by the UK Arthritis Research Campaign, the Medical Research Council (G9321536), and the European Foundation for Osteoporosis and Bone Disease. The EU s PECO program linked to BIOMED 1 funded in part the participation of the Budapest, Warsaw, Prague, Piestany, Szczecin and Moscow centres. Data collection from Zagreb was supported by a grant from the Wellcome Trust. The central X-ray evaluation was generously sponsored by the Bundesministerium fur Forschung and Technologie, Germany. The remaining funding was provided by or through the following centers: Radiological Evaluation Center: Department of Radiology and Nuclear Medicine, Free University, Berlin, Germany (DF,GA); Co-ordination and Data Evaluation Centers: University Institute of Public Health, Cambridge, U.K (JR, ML) and ARC Epidemiology Unit, University of Manchester, U.K (AJS); Participating Investigative Centers: Institute of Rheumatology, Moscow, Russia (LIB); Hospital de Angra do Herismo, Azores, Portugal (JBA); Charles University, Prague, Czech Republic (JJS); PKP Hospital, Warsaw, Poland (KH); Clinical Hospital, Zagreb, Croatia (SG); Hospital de San Joao, Oporto, Portugal(ALV); Institute of Rheumatic Diseases, Piestany, Slovakia (PM); Academy of Medicine, Szczecin, Poland (TM); University of Siena, Italy (CG); National Institute of Rheumatology and Physiotherapy, Budapest, Hungary (GP); University of Aberdeen, U.K.(DMR); University of Heidelberg, Germany (CS-N, MS); University Hospital, Graz, Austria (KW); We would like to thank the following individuals: Aberdeen, UK: Rita Smith; Cambridge and Harrow, UK: Anna Martin, Judith Walton; Truro, UK: Mrs Joanna Parsons; Oviedo, Spain: Manuel Naves Diaz, J. Bernardino Diaz Lopez, Ana Rodriguez Rebollar. References [1] Marshall D, Johnell O, Wedel H. Meta-analysis of how well measures of bone mineral density predict occurrence of osteoporotic fractures. Br Med J 1996;312:1254 9. [2] Lunt M, Felsenberg D, Reeve J, Benevolenskaya L, Cannata J, Dequeker J, et al. Bone density variation and its effects on risk of vertebral deformity in men and women studied in 13 European centres: the EVOS study. J Bone Miner Res 1997;12: 1883 94. [3] Klotzbuecher CM, Ross PD, Landsman PB, Abbott III TA, Berger M. Patients with prior fractures have an increased risk of future fractures: a summary of the literature and statistical synthesis. J Bone Miner Res 2000;15:721 7. [4] Felsenberg D, Lunt M, Armbrecht G, Benevolenskaya L, Bhalla A, Bruges Armas J, et al. Rates and determinants of vertebral fracture incidence in European men and women. J Bone Miner Res 1999;14(Suppl 1):s159 and errata (abstract). [5] Burr D, Forwood M, Fyhrie D, Martin R, Schaffler M, Turner C. Bone microdamage and skeletal fragility in osteoporotic and stress fractures. J Bone Miner Res 1997;12:6 15. [6] Currey J, Brear K, Zioupos P. The effects of ageing and changes in mineral content in degrading the toughness of human femora. J Biomech 1995;29:257 60. [7] Garnero P, Hausherr E, Chapuy MC, Marcelli C, Grandjean H, Muller C, et al. Markers of bone resorption predict hip fracture in elderly women: the EPIDOS prospective study. J Bone Miner Res 1996;11:1531 8. [8] van Daele PLA, Seibel MJ, Burger H, Hofman A, Grobbee DE, van Leeuwen JPTM, et al. Case control analysis of bone resorption markers, disability, and hip fracture risk: the Rotterdam study. BMJ 1996;312:482 3. [9] Garnero P, Sornay-Rendu E, Claustrat B, Delmas PD. Bio-

P. Vergnaud et al. / Clinica Chimica Acta 322 (2002) 121 132 131 chemical markers of bone turnover, endogenous hormones and the risk of fractures in postmenopausal women: the OFELY study. J Bone Miner Res 2000;15:1526 36. [10] Riis SBJ, Hansen AM, Jensen K, Overgaard K, Christiansen C. Low bone mass and fast rate of bone loss at menopause equal risk factors for future fracture. A 15 year follow-up study. Bone 1996;19:9 12. [11] Ross PD, Kress BC, Parson RE, Wasnich RD, Armour KA, Mizrahi IA. Serum bone alkaline phosphatase and calcaneus bone density predict fractures; a prospective study. Osteoporos Int 2000;11:76 82. [12] Åkesson K, Ljunghall S, Jonsson B, Sernbo I, Johnell O, Gärdsell P, et al. Assessment of biochemical markers of bone metabolism in relation to the occurrence of fractures: a retrospective and prospective population-based study of women. J Bone Miner Res 1995;10:1823 9. [13] O Neill TW, Cooper C, Algra D, Pols HAP, Agnusdei D, Dequeker J, et al. Design and development of a questionnaire for use in a multicentre study of vertebral osteoporosis in Europe: the European vertebral osteoporosis study (EVOS). Rheumatol Eur 1995;24:75 81. [14] O Neill TW, Marsden D, Silman AJ, on behalf of the European Vertebral Osteoporosis Study Group. Differences in the characteristics of responders and nonresponders in a prevalence survey of vertebral osteoporosis. Osteoporos Int 1995; 5:327 34. [15] O Neill TW, Felsenberg D, Varlow J, Cooper C, Kanis JA, Silman AJ, et al. The prevalence of vertebral deformity in European men and women: the European vertebral osteoporosis study. J Bone Miner Res 1996;11:1010 7. [16] O Neill TW, Silman AJ, Naves Diaz M, Cooper C, Kanis J, Felsenberg D, et al. Influence of hormonal and reproductive factors on the risk of vertebral deformity in European women. Osteoporos Int 1997;7:72 8. [17] O Neill TW, Varlow J, Felsenberg D, Johnell O, Weber K, Marchant F, et al. Variation in vertebral height ratios in population Studies. J Bone Miner Res 1994;9:1895 907. [18] Felsenberg D, Wieland E, Hammermeister C, Armbrecht G, Gowin W, Raspe H, et al. Pravalenz der vertebralen Wirbelkorperdeformationen bei Frauen und Mannern in Deutschland. Med Klin 1998;93(Suppl 2):31 4. [19] Rosenqvist C, Fledelius C, Christgau S, Pedersen BJ, Bonde M, Christiansen C. The serum CTxk one step ELISA. The first application of monoclonal antibodies for measurement in serum of bone-related degradation products from C-terminal telopeptides of type 1 collagen. Clin Chem 1998;44: 2281 9. [20] Bonde M, Qvist P, Fledelius C, Riis BJ, Christiansen C. Immunoassay for quantifying type 1 collagen degradation products in urine evaluated. Clin Chem 1994;40:2022 5. [21] McCloskey EV, Spector TD, Eyres KS, Fern ED, O Rourke N, Wasikaran S, et al. The assessment of vertebral deformity: a method for use in population studies and clinical trials. Osteoporos Int 1993;3:138 47. [22] Lunt M, Johnell O, Gowin W, Armbrecht G, Felsenberg D, Reeve J. A statistical method to minimise magnification errors in serial vertebral radiographs. Osteoporos Int 2001:909 13. [23] Weber K, Lunt M, Felsenberg D, Lauermann T, Gowin W, Wieland E, et al. Measurement imprecision in digital vertebral morphometry of spinal radiographs obtained in the EPOS study: consequences for the identification of prevalent and incident deformities. Br J Radiol 1999;72:957 66. [24] Ismail AA, O Neill TW, Cockerill W, Finn JD, Cannata JB, Hoszowski K, et al. Validity of self-report of fractures: results from a prospective study in men and women across Europe. Osteoporos Int 2000;11:248 54. [25] Lunt M, Felsenberg D, Adams J, Benevolenskaya L, Cannata J, Dequeker J, et al. Population-based geographic variations in DXA bone density in Europe: the EVOS study. Osteoporos Int 1997;7:175 89. [26] Pearson J, Dequeker J, Henley M, Bright J, Reeve J, Kalender W, et al. European semi-anthropomorphic spine phantom for the calibration of bone densitometers: assessment of precision, stability and accuracy. Osteoporos Int 1995;5:174 84. [27] Kalender WA, Felsenberg D, Genant HK, Fischer M, Dequeker J, Reeve J. The European Spine Phantom a tool for standardization and quality control in spinal bone mineral measurements by DXA and QCT. Eur J Radiol 1995;20:83 92. [28] Ingle BM, Hay SM, Bottger HM, Eastell R. Changes in bone mass and bone turnover following distal forearm fracture. Osteoporos Int 1999;10:399 407. [29] Stepan JJ. Prediction of bone loss in postmenopausal women. Osteoporos Int 2000;11(Suppl 6):s45 54. [30] Hannon R, Eastell R. Preanalytic variability of biochemical markers of bone turnover. Osteoporos Int 2000;11(Suppl 6): s30 44. [31] Delmas PD, Eastell R, Garnero P, Seibel MJ, Stepan J. The use of biochemical markers of bone turnover in osteoporosis. Osteoporos Int 2000;11(Suppl 6):s2 17. [32] Delmas PD. Markers of bone turnover for monitoring treatment of osteoporosis with anti-resorptive drugs. Osteoporos Int 2000;11(Suppl 6):s66 76. [33] Garnero P. Markers of bone turnover for the prediction of fracture risk. Osteoporos Int 2000;11(Suppl 6):s55 65. [34] Chapurlat RD, Garnero P, Bréart G, Meunier PD, Delmas PD. Afternoon-sampled serum type 1 collagen breakdown product (serum CTX) predicts hip fracture in elderly women: results of the EPIDOS study. Bone 2000;27:283 6. [35] Szulc P, Chapuy MC, Meunier PJ, Delmas PD. Serum undercaroxylated osteocalcin is a marker of the risk of hip fracture: a three year follow-up study. Bone 1996;18:487 8. [36] Luukinen H, Käkönen SM, Petterson K, Koski K, Laippala P, Lövgren T, et al. Strong prediction of fracture among older adults by the ratio of carboxylated to total serum osteocalcin. J Bone Miner Res 2000;15:2473 8. [37] Weel AEAM, Seibel MJ, Hofman A, ven Leeuwen J, Pols HAP. Which fractures are associated with high bone resorption in elderly women? The Rotterdam Study. J Bone Miner Res 1999;14(Suppl 1):s160 (abstract). [38] Bauer DC, Black DM, Ensrud K, Ovist P, Williams EN. Serum Markers of bone turnover and fractures of the hip and spine: a prospective study. J Bone Miner Res 1999;14:S147. [39] Woitge HW, Knothe A, Witte K, Schmidt-Gayk H, Ziegler R, Lemmer B, et al. Circannual rhythms and interactions of vita-

132 P. Vergnaud et al. / Clinica Chimica Acta 322 (2002) 121 132 min D metabolites, parathyroid hormone, and biochemical markers of skeletal homeostasis: a prospective study. J Bone Miner Res 2000;15:2443 50. [40] Cohen FJ, Eckert S, Mitlak BH. Geographic differences in bone turnover: data from a multinational study in healthy postmenopausal women. Calcif Tissue Int 1998;63:277 82. [41] Lunt M, O Neill TW, Felsenberg D, Reeve J, Kanis JA, Cooper C, et al. Characteristics of a prevalent vertebral deformity predict subsequent vertebral fracture: results from the European prospective osteoporosis study (EPOS). Bone, 2002, in press. [42] Ismail AA, Cockerill W, Cooper C, Finn JD, Abendroth K, Parisi G, et al. Prevalent vertebral deformity predicts future hip though not distal forearm fracture: results from the European prospective osteoporosis study. Osteoporos Int 2001;12: 85 90. [43] Reeve J, Lunt M, Felsenberg D, Scheidt-Nave C, Poor G, Parisi G, et al. Determinants of the size of incident vertebral deformities in European men and women in the 6th 9th decades of age: the European prospective osteoporosis study (EPOS). J Bone Miner Res, 2002, in press. [44] McGuigan FEA, Armbrecht G, Smith R, Felsenberg D, Reid DM, Ralston SH. Prediction of osteoporotic fracture by bone densitometry and COL1A1 genotyping: a prospective, population based study in men and women. Osteoporos Int 2001;12:91 6. [45] Bailey A, Sims T, Ebbesen E, Mansell J, Thomsen J, Li M. Age-related changes in the biochemical properties of human cancellous bone collagen: relationship to bone strength. Calcif Tissue Int 1999;65:203 10. [46] Bell K, Loveridge N, Power J, Garrahan N, Stanton M, Lunt M, et al. Structure of the femoral neck in hip fracture: cortical bone loss in the inferoanterior to superoposterior axis. J Bone Miner Res 1999;14:112 20. [47] Bell KL, Loveridge N, Jordan GR, Power J, Constant CR, Reeve J. A novel mechanism for the induction of the increased cortical porosity in cases of intra-capsular hip fracture. Bone 2000;27:297 304. [48] Bell K, Loveridge N, Power J, Rushton N, Reeve J. Intracapsular hip fracture: increased cortical remodelling in the thinned and porous anterior region of the femoral neck. Osteoporos Int 1999;10:248 57. [49] van der Linden JC, Homminga J, Verhaar JAN, Weinans H. Mechanical consequences of bone loss in cancellous bone. J Bone Miner Res 2001;16:457 65.