Establishing a Reference Interval for Bone Turnover Markers in 637 Healthy, Young, Premenopausal Women From the United Kingdom, France, Belgium, and the United States

Abstract

Robust reference intervals are needed for the interpretation of bone turnover markers in large phase III fracture trials. The objectives of the study were to (1) estimate reference intervals for serum bone alkaline phosphatase (bone ALP), serum procollagen type I N propeptide (PINP), serum β cross-linked C-telopeptides of type I collagen (S-βCTX), and urinary cross-linked N-telopeptides of type I collagen (U-NTX) in healthy young premenopausal women; (2) examine geographical differences on bone turnover markers; and (3) assess factors known to influence bone turnover and test whether these explain any regional differences. We studied 637 eligible women from four countries that participated in the Horizon-PFT study (United Kingdom, France, Belgium, United States). The women were 30–39 yr of age (mean, 34.6 yr), with regular cyclic menses. Subjects completed a medical and lifestyle questionnaire. Two-sided 95% reference intervals were estimated on transformed values and transformed back to the original scale using the proposed methodology of the International Federation of Clinical Chemistry. S-βCTX was significantly higher in France relative to the United Kingdom (p = 0.01), and PINP was higher in France (p < 0.001) and Belgium (p = 0.02) relative to the United Kingdom and significantly higher in France relative to the United States (p < 0.01) by ANOVA. Overall, one could associate low bone turnover markers with nonsmoking, use of a contraceptive pill, exercise, being close to the time of ovulation, and having high 25-hydroxyvitamin D levels. Countries differed by these characteristics, and once allowed for in the statistical model, any country differences were attenuated or removed.

INTRODUCTION

Decreasing BMD, determined by DXA, and increasing age are primary risk factors associated with postmenopausal osteoporosis. More recently, biochemical markers of bone turnover (BTMs) have been proposed as additional factors for use in the prediction of fractures, in women with postmenopausal osteoporosis, and for assessing the effects of antiresorptive treatments in clinical trials., Therefore, comparing the values of BTMs before and during initial treatment to reference intervals can provide an early determination of the effectiveness of the treatment response.

We recently completed a large phase III trial of zoledronic acid in postmenopausal osteoporosis. To better interpret the effect of zoledronic acid on bone turnover, we carried out this study to establish premenopausal reference intervals for BTMs in women from four of the countries that participated in the phase III trial.

The effectiveness of treatment for osteoporosis is determined by the reduction in fractures; however, BMD is often used as a surrogate marker for the effectiveness of treatment for osteoporosis. Monitoring acute changes in bone is difficult with BMD, because changes are small.

Biochemical markers of bone turnover change more rapidly and are sufficiently sensitive to effectively monitor acute changes in bone turnover. A suitable combination of BTMs may be used to monitor the pharmacological effect and potential efficacy of treatment for osteoporosis. In recent years, automated assays for the measurement of BTMs have been developed. It has been shown that changes in biochemical markers of bone formation after 1 mo of anabolic therapy correlated with improvements in bone structure after 22 mo of therapy. Therefore, once reference intervals for BTMs in healthy, young premenopausal women are established, comparison between those levels and those of postmenopausal women can be performed., Most osteoporotic postmenopausal women treated with effective antiresorptive treatments show a reduction in bone turnover, which reaches a plateau within a few weeks or months.,

It has been proposed that the goal of antiresorptive therapy is to reduce BTMs, particularly those related to bone resorption, to within the lower one half of the reference interval for healthy young premenopausal women. However, some have questioned whether acute decreases of resorption markers below the reference interval may be damaging because of insufficient turnover to repair microfractures. Therefore, accurate and precise reference intervals are needed for the lower limit of BTMs in healthy premenopausal women, recognizing that some women will have values lower and higher than the established interval. In many clinical studies reported to date, the results typically have been compared with reference intervals often reported by laboratories on a limited number (30–100) of premenopausal subjects of undefined medical history. More recently, reference intervals determined using datasets from larger cohorts of healthy, premenopausal women (∼200) of a defined age range, such as 30–45 yr, have been reported.

It is important to establish reference intervals that take into account the known determinants of BTMs such as age, sex, and BMI. We designed the study to minimize the effects of factors associated with preanalytical variability and included women 30–39 yr of age who were not obese or underweight. It was unclear whether other factors might be important or whether there might be differences between countries, and therefore, we designed this study as multinational. The countries matched those in the BTM substudy of the Horizon trial.

In this study, we had the following aims: (1) to estimate reference intervals for serum levels of bone alkaline phosphatase (bone ALP), serum procollagen type I N propeptide (PINP), serum β cross-linked C-telopeptides of type I collagen (S-βCTX), and urinary N-telopeptides of type I collagen (NTX) in healthy premenopausal women, 30–39 yr of age; (2) to examine for country differences in any of the reference intervals; and (3) to assess factors reported to influence bone turnover and test whether these explain any country differences.

MATERIALS AND METHODS

Subjects

Seven hundred three healthy, nonpregnant, premenopausal women were recruited as volunteers from four countries that participated in the Horizon-PFT study: France (two centers); Belgium (two centers); United States (five centers), and United Kingdom (two centers) (Fig. 1). The women were 30–39 yr of age (mean, 34.6 yr), each having regular cyclic menses (12/yr) or having taken an oral contraceptive pill (OCP) for at least 1 yr before study entry. Subjects were recruited using poster advertisements or mailings from general practitioners and local family planning clinics.

Figure Figure 1. Flow chart showing subject numbers at each stage of the study. Note: some subjects had multiple protocol deviations and some subjects dropped out of the study after recruitment.

The study was carried out between June 28 and September 26, 2006. Subjects weighed between 45 and 100 kg, each with a BMI between 18 and 29 kg/m² (mean, 23.3 kg/m²). None of the subjects taking part in the study had a chronic disease, cancer, or any medical condition known to affect bone metabolism. In addition, subjects were not taking any drugs known to affect bone metabolism, such as selective estrogen receptor modulators, phytoestrogens, anticonvulsants, calcitonin, or anabolic agents. Subjects with a history of alcoholism or who reported a recent clinical fracture (within 2 yr) were excluded. Pregnant or lactating women and women who had taken depo-provera within 2 yr of the study were also excluded. Subjects were excluded from the calculation of reference intervals if screening showed abnormal laboratory results, including calculated creatinine clearance <60 ml/min, even if the laboratory values were not clinically significant.

All subjects completed a medical and lifestyle questionnaire. The number of lifetime pregnancies and use of OCPs within the last 12 mo were recorded. Subjects were classified as smokers if they were current smokers. Subjects were classified as alcohol consumers if they drank more than one alcoholic beverage per day, on average. Subjects provided information about the frequency and nature of any exercise program in which they regularly participated. Subject's weight and height were recorded. Body mass index (BMI; kg/m²) was calculated as the weight (kg) divided by height (m) squared. Ethical approval was obtained from the local ethics committee, and written informed consent was obtained from each subject before enrollment in the study.

Blood and urine samples

Samples of venous blood were taken from all subjects into serum separator tubes provided by the central laboratory (Covance, Indianapolis, or Geneva). The blood was allowed to clot for 30 min at room temperature and was centrifuged at 2500g for 10 min. Samples were stored at −70°C. Second void urine samples were collected and stored at −70°C until assay. All samples were collected between 8:00 and 10:00 a.m. after an overnight fast. Frozen serum and urine samples were shipped to the central laboratory, where they were stored at −70°C until near the end of the study, when they were shipped to the specialized assay laboratory (Synarc Lyon).

Biochemical markers

The following biochemical markers were measured at the same time point using a single lot of reagents in one batch, according to the manufacturer's protocol and following the specialized assay laboratory quality control procedures.

Bone ALP was measured in serum by immunoenzymatic assay with the Access Immuno-assay System (Beckman Coulter) with an intra-assay CV of 2.3–3.7% and an interassay CV of 4.4–9.8%. The limit of detection (LOD) was 0.07 ng/ml.

PINP was measured in serum with a modular autoanalyzer (Roche, Basel, Switzerland) with an intra-assay CV of 1.2–4.9% and interassay CV of 4.3–6.5%. The LOD was 5 ng/ml.

S-βCTX was measured in serum with a modular autoanalyzer (Roche) with an intra-assay CV of 1.6–3.0% and an interassay CV of 1.3–4.3%. The LOD was 0.01 ng/ml.

Urinary NTX was measured with a Vitros ECi automated analyzer (Ortho Clinical Diagnostics, Rochester, NY, USA) with an intra-assay CV of 1.1–6.7% and an interassay CV of 3.8–6.1%. The LOD was 10 nmol BCE. Reported results were corrected for urine creatinine concentration (mmol, see below) and provided in the tables as the ratio of nmol BCE/mmol creatinine.

Urinary creatinine was measured with a Kone 20 Analyzer with an intra-assay CV of 3.1–3.8% and an interassay CV of 2.1–2.9%. The lower limit of quantification was 0.36 mM.

Serum 25-hydroxyvitamin D [25(OH)D] was assayed using a two-step radio immunoassay by DiaSorin (Stillwater, MN, USA). The procedure involves a rapid extraction of 25(OH)D and other hydroxylated metabolites from serum with acetonitrile. The acetonitrile extract is assayed using an equilibrium radioimmunoassay procedure. The assay recognizes both 25(OH)D₂ and 25(OH)D₃ and reports the combined results only. Intra-assay CV is <10.1%.

Statistical analyses

Because normality could not be assumed, the study was designed on a sample size of 650 that would be sufficient to calculate two-sided 95% normative ranges for the biochemical markers such that the tolerance of the reference intervals was ±0.75% with a probability of 90%. In other words, there is a 90% probability that the lower limit of the normative range is not <1.75% with respect to the target lower limit of 2.5%, and the upper limit of the “normative range” is not >98.25% with respect to the target upper limit of 97.5%. The reference intervals were calculated using parametric methodology for normally distributed data., Centiles were estimated with , where z_p is the pth centile of the N(0,1) distribution, if the primary variable is normally distributed. The study was not powered for any other analyses (multivariate/univariate) of prognostic factors (continues/categorical), which were carried out as exploratory analyses.

To achieve a normal distribution for S-βCTX, bone ALP, and NTX, a parametric transformation of the variables was performed. Following the recommendation of the International Federation of Clinical Chemistry (IFCC), the values were log transformed and statistically standardized before Manly's exponential transformation was applied to remove skewness. After final transformation, the calculated centiles were back transformed to original scale. To achieve normality for PINP, the values were log transformed and statistically standardized to a N(0,1) distribution. The transformation did depend on the data. Confidence bands (90%) for the reference intervals were also calculated.

Pearson correlation coefficients were used to determine the association between biochemical markers and the following variables: age; day of menstrual cycle; height; weight; BMI; days since last menstrual period to time of blood draw; and hours of exercise and hours of routine activity per week. Two-sample t-tests were used to compare bone turnover markers in women close to ovulation compared with those who were not, women who were current smokers versus those who were not, women who had at least one previous pregnancy and those who had not, nonexercisers versus regular exercisers, women who on average consumed more than one alcoholic drink per day versus those who did not, and women who were on an OCP versus those who were not. One-way ANOVA was used to compare the concentration of BTMs with all other variables across the four countries studied. Univariate analyses were performed using SPSS software (version 12.0; SPSS, Chicago, IL, USA). Multivariate analysis was performed using SAS software (version 8.2; SAS Institute, Cary, NC, USA).

RESULTS

Subject characteristics

A total of 752 subjects were screened for entry into the study, and of these, 48 were screen failures, leaving 703 subjects. Of these subjects, 637 did not have any major protocol deviations, had normal safety laboratory values (<[2 × upper limit of normal] or >[0.5 × lower limit of normal]), and had an evaluable blood sample for S-βCTX, PINP, or bone ALP, and 630 had an evaluable urine sample for NTX/Cr. Table 1 shows the subject characteristics.

Table Table 1.. Characteristics of Subjects Used in the Calculation of Reference Intervals

Reference intervals

Two-sided 95% reference intervals were calculated for each biomarker after normalization. The 95% reference intervals and 90% CIs of the upper and lower limits are shown in Table 2.

Table Table 2.. Median, Geometric Mean, and 95% Reference Intervals for Bone Turnover Markers

The lower half of the reference interval (values falling between the lower limit of the reference interval and median) has been proposed as a target for treatment, and this range was 16.3–36.0 ng/ml for PINP, 5.15–8.68 ng/ml for bone ALP, 0.114–0.299 ng/ml for serum βCTX, and 9.22–24.8 nmol BCE/mmol creatinine for urinary NTX/Cr.

Factors influencing markers of bone resorption and formation in univariate analyses

Univariate analyses were performed to study the factors thought to influence bone turnover levels. Serum PINP and S-βCTX were found to differ significantly between counties (Fig. 2). Women in France showed significantly (17.2%, p = 0.01) higher mean levels of S-βCTX (0.34 ng/ml) relative to those living in the United Kingdom (0.29 ng/ml) and significantly higher mean levels (10.7%, 19.3%, and 27.1%, respectively) of PINP (44.5ng/ml) than women in Belgium (40.2 ng/ml), the United States (33.7 ng/ml), and the United Kingdom (35.0 ng/ml). PINP levels were also significantly higher in women living in Belgium than the United Kingdom (16.6%). There were no significant differences in NTX or bone ALP mean concentrations across countries.

Figure Figure 2. Bar charts showing mean levels of bone formation and resorption in France (n = 100), Belgium (n = 177), the United States (n = 169), and the United Kingdom (n = 191) Values are mean ± SE.

S-βCTX and PINP mean levels were 8.3% higher in current smokers compared with nonsmokers (p = 0.019 and p = 0.0013, respectively). PINP levels were 12.1% lower in women taking an OCP compared with those not receiving OCPs (p < 0.001). Mean concentrations of markers of bone formation and resorption were significantly lower (between 8% and 16%) in women who regularly participated in sport or routine activity than those who did not (bone ALP, p = 0.049; PINP, p = 0.047; S-βCTX, p = 0.012; NTX, p = 0.01). Bone ALP was 7.2% lower in women who were close to ovulation (days 6–18 of the menstrual cycle) compared with those at the beginning or end of the menstrual cycle (days 0–5 and 19–28) at the time of sampling (p = 0.007; n = 379). Bone ALP increases were associated with significantly decreased 25(OH)D levels (p = 0.025). Overall, 1.3% of women were 25(OH)D deficient (< ∼30nM), 14% were classified as insufficient (>30 and <60 nM), and 83.9% were sufficient (>60 nM). Women who were deficient in 25(OH)D had significantly lower levels of S-βCTX than those who were insufficient (p = 0.042) or sufficient (p = 0.049), but this observation must be viewed with caution because of the few women in this subpopulation.

By univariate analyses, neither age nor BMI was correlated with BTMs. Furthermore, we observed no significant differences in the mean values of BTM levels between women who regularly consumed alcohol and those who did not nor between women who had a previous pregnancy and those who had not.

Differences in variables between countries

Women living in Belgium were significantly taller than those in both the United Kingdom and the United States (p = 0.001 and 0.041, respectively), despite mean values being very similar. UK women had a significantly higher BMI than those in France (p = 0.025). The number of women who undertook regular exercise or routine activity was significantly lower in Belgium than all other countries (p < 0.001), whereas those living in the United States were more likely to be regular exercisers (p < 0.001). The number of current smokers was significantly higher in France than all other countries (p < 0.001). France had the highest number of women who consumed, on average, more than one drink per day, with the United Kingdom having the next highest, followed by Belgium and then the United States at a distant fourth (p < 0.01; Table 3).

Table Table 3.. Subject Demographics by Country and p Values Obtained by ANOVA

Factors associated with markers of bone turnover (multivariate analyses)

Multivariate analyses were performed to further elucidate differences in BTMs across countries and/or other co-factors (Table 4). S-βCTX, PINP, and bone ALP mean concentrations were significantly lower in women taking OCPs than those who were not (all p < 0.001). A significant negative correlation between 25(OH)D serum levels and PINP (p = 0.01) and bone ALP (p = 0.01) serum levels and a slight trend toward a negative correlation between 25(OH)D and NTX were found. Markers of bone resorption (S-βCTX and NTX) and formation (PINP) were significantly higher in women living in Belgium than other countries (p = 0.009, p < 0.001, and p = 0.037, respectively). Women who were close to ovulation (days 6–18 of the menstrual cycle) at the time of sampling had lower levels of S-βCTX (p < 0.024) compared with those at the beginning or end of the menstrual cycle (days 0–5 and 19–28). A significant positive correlation was observed between BMI and PINP in women living in Belgium (p = 0.022), whereas women living in the United Kingdom showed a negative correlation between BMI and bone ALP levels. For women living in Belgium and France, a significant negative association between age and markers of formation and resorption was observed, namely S-βCTX in Belgium and the United Kingdom, PINP in the United Kingdom, and bone ALP in Belgium and France. Women living in France and Belgium who regularly participated in routine activity showed higher levels of S-βCTX (p = 0.018), PINP (p = 0.22), and NTX (p = 0.024).

Table Table 4.. p Values (and Coefficients) for Factors Whose Multivariate Interactions Gave a Significant Association With Markers of Bone Turnover

DISCUSSION

Reference intervals

Robust reference intervals have been established for markers of bone formation and resorption in healthy premenopausal women living in four different countries. In a previous study, based on a cohort of 157 premenopausal women (age, 35–45 yr) living in Sheffield, UK, very similar reference intervals to those established in this study were estimated (Fig. 3). Although exclusion criteria for entry into this study were not based on bone X-ray or spine DXA data, the reference intervals established are nevertheless comparable to those presented by De Papp et al., based on 237 healthy premenopausal women (age, 28–45 yr) from seven U.S. centers who excluded subjects based on such data (Fig. 3). The excellent consensus between the reference intervals and means established in this study and those previously reported provide strong supportive evidence that the levels of bone turnover observed in this study accurately reflect bone turnover in healthy, premenopausal women. Comparison of the results of this study with those of de Papp et al. show consistent data for the same markers measured with the same assays, although they were generated in two different laboratories. This suggests that the use of newer, automated assays, when conducted at laboratories that participate in an external quality assurance program, will minimize variation of results between laboratories. Under these conditions, the lower and higher reference limits determined in this study could be used by other clinicians for comparisons to the bone biomarker results of their patients, provided the same methodology is used and that the laboratory participates in an external quality assurance program. We do, however, need to recognize that it is unlikely many laboratories will have access to large well-characterized populations to enable them to establish their own reference intervals, and we designed this study to provide reference intervals for a well-defined population that will be useful to clinicians.

Figure Figure 3. Comparison of reference intervals established in this study with previously published reference intervals: Glover et al. 2008 based on 153 premenopausal women 35–45 yr of age in Sheffield, UK, and de Papp et al. 2007 established in 237 premenopausal women 28–45 yr of age from seven sites across the United States. A different assay was used by Glover et al. 2008 for bone ALP giving the results in different units; therefore, a comparison cannot be made for bone ALP.

Factors associated with markers of bone resorption and formation and differences in variables between countries

Several factors were identified that are significantly associated with bone turnover in the whole study population. Significant differences in the levels of PINP and βCTX between individual countries were observed (Fig. 2). Univariate analyses showed women in France displayed significantly higher serum levels of S-βCTX than those living in the United Kingdom and significantly higher serum levels of PINP than women in Belgium, the United States, and the United Kingdom. However, these differences were no longer statistically significant after multivariate analysis, suggesting the differences might be largely influenced by lifestyle factors. Specifically, women in France had a significantly lower mean BMI and a significantly higher proportion of current smokers than the other countries, both of which are known to be associated with higher levels of bone turnover. In contrast, both univariate and multivariate analyses showed women living in Belgium had significantly higher levels of PINP than those living in the United Kingdom (data not shown), suggesting only a small attenuation on PINP differences by lifestyle factors of the subjects.

Previous studies have presented conflicting evidence regarding the existence of any differences between the levels of bone turnover in women living in geographically diverse locations. Blumsohn et al. reported no significant differences in markers of bone formation or resorption between women living in five European centers. Conversely, Cohen et al. observed North American women had higher levels of bone turnover than their German and Spanish counterparts. Differences in levels of bone formation and resorption markers in healthy postmenopausal women were not explained by age, years after hysterectomy, or serum follicle-stimulating hormone (FSH).

In this study population as a whole, age was not significantly associated with bone turnover. We did, however, observe a negative correlation between age and individual markers of bone turnover for individual countries: S-βCTX (Belgium and United Kingdom), PINP (United Kingdom), and bone ALP (France). This is consistent with our findings from a previous study, in which we observed higher markers of bone formation and resorption in women who were 30–34 yr of age than those 35–45 yr of age.

This study showed that women taking OCPs have significantly lower levels of bone formation (PINP, bone ALP) and bone resorption (S-βCTX; Table 4). This finding is consistent with previous studies including the study of Garnero et al. in French premenopausal women and the study of de Papp et al.,, in which the U.S. subject population was stratified by the use of OCPs. The type of OCPs most commonly used may vary between Europe and the United States. The levels of progesterone may differ significantly between pills.

Ovulation had a significant impact on markers of resorption and formation in individual countries. Women living in France, who were close to ovulation at the time of sampling, had significantly higher levels of bone ALP and NTX, whereas those living in the United Kingdom had elevated bone ALP around the time of ovulation (days 6–18 of the menstrual cycle) compared with those at the beginning or end of the menstrual cycle (days 0–5 and 19–28). This is consistent with findings by Gorai et al., who provided evidence for increased bone ALP and S-βCTX around the period of ovulation.

In our study, women who were current smokers had higher levels of circulating S-βCTX and PINP compared with nonsmokers. This is consistent with the previous study of women based in Sheffield, UK, in which bone ALP and urinary NTX were higher in current smokers than nonsmokers. Bjarnason and Christiansen found women who smoke have an increased degradation of estradiol and experience menopause earlier than nonsmoking women, thus suggesting a possible mechanism for increased S-βCTX and PINP observed in our study population.

Markers of bone formation and resorption were significantly lower in women who regularly participated in sport or routine activity than those who did not, irrespective of weight of BMI. A study by Woitge et al. showed that aerobic training resulted in a significant decrease in both bone formation and resorption markers. The effects of exercise on biochemical indices of bone turnover are complex and depend on the type of exercise, the intensity of training, and the time of sampling in relation to the exercise.

The effect of previous pregnancy on BTMs in premenopausal women has not been previously studied. No significant differences in markers of either bone formation or resorption in women who had a previous pregnancy and those who had not were found in this study. We observed no differences in the mean values of BTMs between women who regularly consumed alcohol and those who did not. Alcohol consumption has been shown to decrease BTMs, but the effect persists for only a few hours.

A negative correlation was observed between 25(OH)D levels and bone ALP and PINP levels, consistent with previous reports., Previous studies have shown a negative correlation between 25(OH)D and bone resorption markers, and patients with 25(OH)D insufficiency are known to exhibit a higher levels of bone turnover than those with sufficient 25(OH)D levels. Unexpectedly, women who were vitamin D deficient had lower levels of serum s-βCTX; however, this was based on a small number of subjects with a borderline statistical significance p value.

The observed geographic variations in 25(OH)D levels in this study do not explain the relationship between BTMs and country. Surprisingly, women living in Belgium showed the highest levels of 25(OH)D (Table 3). Mean serum concentrations from subjects at each of the three sites in Belgium were consistently higher than the serum concentrations observed from patients in France, the United Kingdom, or the United States (including sites in San Diego, CA; Dallas, TX; Winston-Salem, NC; Rochester, NY; and Pittsburgh, PA). After we observed these results for Belgian subjects, we confirmed from monitors and site personnel that none of the Belgian sites had programs designed to encourage premenopausal women to take daily supplements of calcium and vitamin D. Possibly, this unexpected result arises from individual levels of sun exposure, skin pigmentation, or dietary supplementation, particularly because sites in France and the United Kingdom were at approximately the same latitude and therefore degree of sunlight. Nevertheless, the serum concentrations of 25(OH)D in Belgian subjects is all the more surprising because these subjects tended to have serum βCTX, PINP, and urinary NTX/Cr values that were higher than those seen in subjects from the United Kingdom or the United States.

Limitations of reference intervals

In establishing a reference interval for each BTM, we assumed young, healthy premenopausal women have an ideal bone turnover that is the target range for women receiving osteoporosis therapy. Heaney proposed that humans alive today have a reduced dietary calcium intake than in earlier human evolution, likely resulting in a higher rate of bone remodeling. For this reason, caution should be applied when assuming young women have optimal bone health.

Limitations and strengths of this study

Dietary calcium intake was not analyzed in this study; therefore, we could not examine the effect of calcium intake on bone turnover and whether any differences in calcium intake were present between the different countries studied.

We did not measure BMD and PTH levels or fully assess calcium homeostasis. Because we were unable to exclude women based on these criteria, some women with abnormal BMD or calcium homeostasis may have been included in the study. Whites comprised 94% of subjects in this study. No attempt was made to compare reference intervals for whites to those for women of a black or Asian origin because the numbers in these two ethnic groups were too small. We recommend that reference intervals be established for different ethnic groups.

With all the prognostic factors examined, the R² for the multivariate analyses are very low, indicating only a small percentage of the variability is explained by lifestyle factors. There would seem to be many other outstanding environmental and genetic factors that could influence BTMs in a healthy premenopausal population.

Nevertheless, the study protocol specified detailed inclusion and exclusion criteria, which, combined with the use of a medical questionnaire, allowed a well-defined and characterized study population to be identified. All samples were obtained at the same time of year and time of day (all samples were taken between the end of June and the end of September and 8:00 and 10:00 a.m.) and from women in a fasting state. Assay consistency was maximized by using a central specialized laboratory and automated assays to reduce human error. Thus, we controlled factors to minimize analytical and biological variability within individuals. The large sample size in this study allowed us to reliably estimate reference intervals for markers of bone turnover. The good agreement between the reference intervals established in this study and previous studies gives us confidence that the intervals we established are representative of levels of bone turnover in healthy, premenopausal women.

Application of reference intervals in the clinical setting

There are a number of cautions to be exercised regarding the use of any reference interval in assessing the efficacy of osteoporosis treatment. Whereas we know that reducing bone turnover either by 50% or more or to levels below the mean of the reference interval accounts for a significant proportion of the antifracture efficacy of antiresorptive treatment, it is unknown if and to what extent reducing bone turnover below the lower limit of the premenopausal reference interval would blunt treatment efficacy or even increase bone fragility. Our study and others showed that low levels of bone turnover should not to be a cause for concern, because 2.5% of young, healthy premenopausal women have levels of bone turnover below the lower limit of any reference interval. Rather, it is the long-term suppression of bone turnover that may result in oversuppression. The establishment of reference intervals as in this study will enable us to address the issue of possible oversuppression in future studies.

In conclusion, we established robust reference intervals for four BTMs based on a large international study of premenopausal women. This study provides information about the characteristics that may influence biochemical markers of bone turnover. Interestingly, we identified differences in BTMs between countries. The differences between countries are small, and therefore, we believe the reference intervals estimated in this study can be applied to each of the countries studied. The reference intervals established were only based on subjects living in four countries and primarily in white women. Levels of bone turnover may differ in other countries and in women of different ethnic origin because many unknown factors may contribute the variability in bone marker levels. We therefore recommend that, in countries not taking part in this study, reference intervals need to be established locally.

The regional differences observed in this study are partly explained by differences in BMI, smoking habit, pregnancies, exercise, and alcohol consumption between countries. After adjusting for other factors, the differences in levels of bone turnover seen between one country and the next are small and therefore not clinically significant. The questionnaire we used in the study may not have captured all sources of variability in levels of bone turnover. Unknown environmental or genetic influences on bone turnover, not accounted for in our analysis, may also contribute to the differences in bone turnover observed between the countries studied. The reference intervals presented in this study will help determine the duration of the effect of osteoporosis treatments on the rates of bone formation and resorption relative to the time of dosing, thus assisting in the monitoring of such therapies in the treatment of postmenopausal osteoporosis.

Acknowledgements

This study was funded by a research grant from Novartis. The authors thank the all the staff at each of the research centres participating in this study for help in the recruitment of subjects. The Principal Investigators at the 12 sites included Dr Steven Boonen, Dr Richard Witvrouw, Dr Mathieu Peeters, Dr Regine Rouzier, Dr Hafid Belhadj-Tahar, Dr David Bell, Dr Richard Eastell, Dr Donald Brandon, Dr R Lamar Parker, Dr Tammi Scholtzhauer, and Dr Jane Cauley.