The predictive role of bone turnover markers for BMD in middle-aged men

Abstract

Measurement of bone turnover markers has been proposed as a potentially valuable clinical laboratory aid in osteoporosis risk assessment. These markers may allow quantitative evaluation of rates of bone loss, and thereby identify persons at risk for osteoporosis at an earlier stage. As far as we know, this is the longest longitudinal study on bone turnover markers conducted in adult men. The objectives of this study were to determine whether markers of bone formation (type I procollagen amino-terminal propeptide, PINP, and carboxy-terminal propeptide, PICP), and of bone resorption (type I collagen carboxy-terminal telopeptide, ICTP), are predictive of changes in lumbar spine and femoral neck BMD over a 5-year period, and to determine the ability of the bone resorption marker urine amino-terminal telopeptide (NTx) to explain the variance in BMD change over the past 5 years in a group of men 35–69 years old. In this group, NTx was the only marker to correlate significantly with BMD changes at the femoral neck (r = −0.21), but not at the spine. The use of the biochemical markers studied to predict change in bone density in adult men in middle-aged years is of very limited value.

Keywords:

• Biochemical markers
• bone metabolism
• bone mineral density
• men

Introduction

The use of biochemical markers of bone turnover as indicators of overall bone metabolism has been suggested as a potentially valuable clinical method in osteoporosis for screening, diagnosis, and monitoring the effects of different interventions [1]. Apart from fractures, dual-energy x-ray absorptiometry (DXA) is the only currently accepted method of diagnosis for osteoporosis. It has been suggested that decreased bone mass and architectural deterioration of bone tissue are related to abnormalities of bone turnover [2]. Biochemical markers reflect small changes in bone turnover of the entire skeleton and in a shorter time frame compared to the months or years that it can take to visualize distinct changes in bone mineral density (BMD) using absorptiometry methods [2]. They might also capture bone properties independent of BMD measurements, like bone remodeling, fragility, trabecular connectivity or architecture [3].

As there is a great need to identify persons at risk for osteoporosis, it would be useful to obtain a quantitative evaluation of rates of bone loss to identify ‘fast bone losers’ in an early stage [4] in order to target interventions to decrease further rapid loss of bone. Population studies showed that biochemical measurements may predict rates of bone loss and occurrence of fractures [4]. An elevated bone turnover in elderly women appears to adversely influence bone mass and fracture risk. It has been suggested that combined biochemical markers and BMD screening might better predict future fractures than BMD alone [5]. Studies in which biochemical markers predicted osteoporotic fractures independent of BMD suggested that the measurement of BMD and the biochemical assessment of bone turnover may be complementary in prediction of bone strength and possible future fractures [6],[7]. Understanding the complex interrelationship between BMD and bone turnover markers could facilitate the creation of better predictive models and interventions for osteoporosis that include these factors. Although osteoporosis in men is a serious and frequent condition, there is a paucity of information in the scientific literature concerning osteoporosis in men.

The aim of this study was to determine whether specific markers of bone formation, including type I procollagen amino-terminal propeptide (PINP) and type I procollagen carboxy-terminal propeptide (PICP), and of bone resorption, type I collagen carboxy-terminal telopeptide (ICTP), are predictive of changes in BMD at the lumbar spine and femoral neck over a 5-year period in a group of men 35–70 years old. PINP and PICP are globular domains at the amino-terminus and carboxy-terminus, respectively, of the type I procollagen molecule, the biosynthetic precursor of type I collagen fibrils. The role of these propeptides is in monomer formation, preventing early aggregation inside the cells and aiding ordered fibrillogenesis. The propeptides can be detected intact in the interstitial fluid of tissue undergoing rapid collagen synthesis and in serum. At each end of type I collagen, after removal of the propeptides, short telopeptide domains provide sites of intermolecular cross-linking in collagen fibrils. An assay for cross-linked peptides based on the amino-terminal telopeptide domain (NTx) is in use as an index of bone resorption [8]. A second objective of the present study was to determine the ability of urinary bone resorption marker NTx to explain the variance in BMD change over the past 5 years.

The hypothesis was that the correlations of markers with either previous or future BMD changes (e.g. either in the prior or subsequent 5 years) would be higher than those found in a previous cross-sectional analysis, including some of the same subjects [9]. The markers, which represent bone turnover activity, were expected to better reflect relatively recent changes in bone mass than a measure taken at one point in time reflecting cumulative lifetime influences on BMD.

Methods

Both prospective and retrospective cohort study designs were used to examine the ability of bone turnover markers to predict change in BMD. The biochemical markers PINP, PICP and ICTP were measured in samples collected in 1992, and urine NTx in samples collected in 1997. DXA data were gathered in 1992 and again in 1997.

Subjects

The current study utilized data available from a larger project, in which twin pairs were selected and recruited from the Finnish Twin Cohort, which is representative of the Finnish population. The Finnish Twin Cohort contains virtually all Finnish sex-matched twin pairs born before 1958. The twin pairs were selected solely on within-pair discordance for a common exposure suspected of influencing musculoskeletal degeneration (e.g. occupational physical loading) [10].

Earlier analyses found the selected subjects to be quite representative of the Cohort from which they were drawn on a number of factors investigated. Twins originate from all social levels and all regions of Finland. No significant differences were observed compared to the referents for level of education, social class, occupational category, outdoor vs indoor work, leisure-time physical activity, history of work-incapacitating neck, shoulder or back pain, smoking status and life satisfaction. The only statistically significant differences observed between study pairs and the base population of twins in the Finnish Twin Cohort were work status and physical loading at work, which is probably due to subjects' selection partly on these characteristics. Subjects were slightly more likely to be working and to have higher physical work demands [10].

Collection of data on biochemical markers began after the original study was underway; thus, the sample size is limited to 240 subjects 35–70 years old. The study sample for this study was composed of monozygotic (MZ) pairs with data available for all markers and BMD. The participants provided informed consent forms prior to study participation. The study protocols were reviewed and approved by the Ethical Committee of the Department of Public Health at the University of Helsinki and the Human Research Ethics Board at the University of Alberta.

A detailed, structured interview including lifestyle factors such as dietary calcium intake, cigarette smoking, alcohol and coffee consumption was also conducted. In Finland it has been estimated that 74% of dietary calcium comes from milk products [11], and none of the subjects reported taking calcium or vitamin D supplements. Quantity, frequency and beverage type of alcohol consumption were recorded and methods have been described in earlier papers [12],[13]. No statistically significant associations were found between BMD and cigarette smoking, coffee intake and alcohol consumption in an earlier analysis, and dietary calcium explained only 1% of the variance of BMD at the femoral neck and 0% at the lumbar spine [14]. Thus, we did not control for the effects of these factors in the present study. Subjects were excluded from analysis if they suffered any chronic kidney or liver disorders (3), or if they had a history of the following conditions or medications in the prior year: thyroid or parathyroid disorders (0); hormone (cortisone or steroid) therapy (4); epilepsy or anti-epilepsy medication (3); any skeletal disease or fracture (21); bed rest of more than 1 month (3); or active cancer (3). Thus, a total of 37 subjects were excluded, leaving 203 (82%) subjects with cumulative marker and BMD data for inclusion in analyses. The twins ranged in age from 35–69 years at baseline (mean 49.7, SD 8.4). Additional characteristics of participants are provided inTable I.

Table I.  Summary characteristics of participants.

Variable	N	Mean (SD)	Min	Max
Age (years)	197	49.7 (8.4)	35	69
5-year change in spine BMD (g/cm^2)	127	−0.02 (0.05)	−0.21	0.11
5-year change in femoral neck BMD (g/cm^2)	128	−0.001 (0.05)	−0.19	0.16
PINP (μg/l)	200	46.8 (15.7)	16	132
PICP (μg/l)	200	141 (42)	66	354
ICTP (μg/l)	201	2.9 (0.9)	2	8
NTX (nM BCE/ mM creatinine)	128	39.3 (16.6)	13	132

BMD and biochemical markers

Data collection included DXA of the lumbar spine and right hip, fasting serum and first void urine samples collected in the morning. Serum and urine specimens were subsequently stored at −20°C at the study site hospital. The samples were then moved to another laboratory and stored at −70°C to await analysis of PINP, PICP, ICTP and NTx. All subjects slept at a hotel adjacent to the testing site the night before the samples were collected.

PINP was determined in serum by radioimmunoassay using the propeptide as an antigen (Orion Diagnostica, Finland). The reference interval in men ranges from 20–76 μg/l. Intra- and inter-assay coefficients of variation (CV) are 4.6–10.3% and 3.1–10.8%, respectively. PICP serum concentrations were analyzed with radioimmunoassay kits (Orion Diagnostica, Finland). The intra- and inter-assay CV were <6%.

NTx was measured in urine using a competition ELISA assay (Osteomark®; Ostex International) and was normalized to urinary creatinine. NTx values were corrected for creatinine to adjust for the wide-ranging dilution of normal urine spot collections. The biologic intra-individual CV for NTx was found to be 22% with a range of 16–33%. The analytic intra-assay CV was <5%, and the analytic inter-assay <8.0%.

ICTP, another marker of collagen type I degradation, was analyzed in serum with a radioimmunoassay using polyclonal antibodies against the C-telopeptide region of type I collagen, which were produced in rabbits (Orion Diagnostica, Finland). The reference interval in men ranges from 1.6–4.6 μg/l. Intra- and inter-assay coefficients of variation were 2.8–6.2% and 4.1–7.9%, respectively.

Bone mineral density was measured with DXA (Lunar DPX, Madison, WI) at the L1–L4 vertebrae and right femoral neck and has been described in detail earlier [14]. The coefficient of variation for BMD measurements was 0.9% for the spine and 1.5% for the femoral neck. National, ethnic mean values for BMD in 20–29 year old men are 1.06 g/cm² for femoral neck and 1.23 g/cm² for lumbar spine. Standard deviations for BMD are 0.14 g/cm² for the femoral neck and 0.15 g/cm² for lumbar spine [15]. The same type of DXA equipment and software for pencil beam densitometers (DPX) were used in the present study at baseline and follow-up.

Data analysis

The STATA statistical package was used for data analyses. The α level was set at 0.05 for determining statistical significance. Participants' characteristics, including age, change in BMD at the various sites, and marker values are summarized inTable I. Because the use of twin pairs violates the assumption of independence required for standard statistical models, except when noted, all standard errors were adjusted for clustering by twinship. The intra-class correlations between the twins for our outcome variables range from 0.17 for ICTP to 0.50 for NTx. This reduces our effective sample size from 126 to 84 – 108. However, as the minimum sample size calculated was 60 (n = L/f² + k + 1, where f² = R²/1 – R², k = number of variables = 2, R² = variance to be declared significant and, therefore, n = L/0.25 + 2 + 1 = 14.17/0.25 + 2 + 1 = 56.68+2 + 1 = 60), even 84 is a reasonable sample size, with >80% power to detect effect sizes as low as 0.31 of one standard deviation for the outcomes.

Pearson coefficients assessed the correlation between spine and femoral neck BMD and marker values. Multiple linear regression analyses were conducted to examine the ability of markers to explain change in BMD, with age, fat-free weight, height and baseline BMD introduced in the model as possible confounding factors. The adjusted R² (AR²) indicated the percent of the variance explained by a covariate. Finally, the effect of familial aggregation or twinship on an outcome was determined by entering indicator variables for each twin pair (save one) into the model, and examining the AR². No adjustment of the standard errors was necessary in those models. One-way ANOVA was used to evaluate differences in NTx levels between the 15% of subjects with the greatest changes in BMD (a gain or loss) vs the remainder of the subjects with lesser degrees of change in BMD at the lumbar spine and femoral neck.

Results

Among the markers studied, NTx, a marker of bone resorption, measured at the follow-up correlated with the change in femoral neck BMD during the previous 5 years (r = −0.21, p = 0.006). NTx explained 3.8% of the change in femoral neck BMD. The other variables (age, fat-free weight, height) did not significantly add to the variance explained in femoral neck BMD, with the exception of baseline femoral neck BMD, which brought the total explained variance to 6.7%. Higher NTx and baseline femoral neck BMD were associated with greater decreases in BMD. Familial aggregation did not significantly add to the explained variance. In the cross-sectional analysis, NTx correlated only with BMD at the femoral neck (r = −0.3, p = 0.001), and explained 8% of the variance (p = 0.001). There was no statistically significant difference found in NTx levels between subjects with the greatest BMD changes and subjects with lesser changes (p = 0.2).

Baseline levels of bone formation markers PINP and PICP, and of the bone resorption marker ICTP, did not significantly correlate with the change in spine or femoral neck BMD in the subsequent 5 years (Table II). Although non-significant (p = 0.06), PINP levels explained 3.4% of the variance in spine BMD change when introduced in the regression model. Age explained an additional 5.9% of the variance. None of the other variables (fat-free weight, height and spine BMD at baseline) was remotely significant. When added to the model, familial aggregation explained 23% of the variance.

Table II.  Pearson correlation coefficients for baseline levels of markers of bone turnover and subsequent change in BMD.

BMD	PINP^1	PICP^1	ICTP^1	NTx^2
Change in spine BMD	0.20	0.20	0.05	0.09
N	126	126	127	127
Change in femoral neck BMD	−0.13	−0.06	−0.01	−0.21(*)
N	127	127	128	128

*Correlation is significant at the 0.05 level (2-tailed).

1 = measurements at baseline.

2 = measurements at follow-up.

N = number of observations.

Markers of bone formation and bone degradation were interrelated. PINP and PICP markers of bone formation measured at baseline correlated with NTx measured at follow-up (r = 0.36, p < 0.01 and r = 0.2, p = 0.03, respectively). ICTP measured at baseline correlated with NTx measured at follow-up (r = 0.2, p = 0.04).

Discussion

The study was conducted over a period of 5 years and, as far as we know, is the longest longitudinal study on biochemical markers of bone turnover conducted in adult men. Among the biochemical markers investigated, NTx was the only marker to correlate significantly with changes in BMD at the femoral neck, but not at the spine. The markers of bone formation, PINP and PICP, and the marker of bone resorption, ICTP, did not correlate significantly with change in spine or femoral neck BMD.

In men, marker levels tend to be highest in the third decade (20–30 years old) corresponding to formation of peak bone mass, decrease rapidly from the third decade until the age of 40 years, and remain stable or decrease slightly between 40 and 60 years [16–18]. Marker levels are lowest in the fifth and sixth decade, with an increase in bone resorption markers in eighth decade [18,19]. Given this and the age of the subjects in the study (35–69 years), a small variation in the marker levels and the bone loss over this age range might be a factor limiting the associations. However, in this study group, NTx values had a wide range (min = 13, max = 132, mean = 39, SD = 16), as did the values for change in BMD (min = −0.21, max = 0.16, mean = 0.01, SD = 0.05). Yet, even when comparing the mean NTx in the subjects with the greatest changes in BMD vs those with lesser changes, there was no statistically significant difference in NTx levels between these groups. The small number of subjects older than 60 years (N = 15) limits our ability to extend the results of this study to prediction in the elderly.

NTx, a marker of bone resorption, measured at follow-up correlated with the change in BMD at the femoral neck during the previous 5 years (r = −0.21, p < 0.05), similar to earlier cross-sectional findings using subjects selected from the same database, demonstrating that among the markers studied only NTx correlated with BMD at the femoral neck (r = −0.3, p < 0.05) [9]. In the present study NTx explained only 3.8% of the variance in change in femoral neck BMD, while in both previous [9] and current cross-sectional analyses, NTx explained 8% of the variance. Another longitudinal study in healthy, elderly, ambulatory male volunteers 76 years and older found a similar correlation between femoral neck BMD and the change in NTx (r = −0.26, p < 0.05) [20], while no correlation between baseline bone markers (NTx and bone alkaline phosphatase (BAP)) and change in femoral neck and spine BMD over a 2-year period was found in a group of osteoporotic men 36–83 years old [21]. Other markers of bone turnover, such as bone resorption markers (ICTP and serum CTx), osteocalcin (OC) and bone alkaline phosphatase (BAP) did not correlate with change in BMD in studies in men 25–86 years old [22] and 42–65 years old, respectively [23].

Although non-significant, the low correlations of PINP with change in BMD at the spine and hip approached significance, as did the correlation of PICP with change in spine BMD. Other longitudinal studies, of shorter duration, in men have also reported that PICP and ICTP were not significantly correlated with change in BMD. Scopacasa et al. (2002) [24] found no relationship between the rate of change in forearm bone mineral content (BMC) and biochemical markers including PICP or ICTP over an interval of 41 months in 123 healthy men 20–83 years old. Also, Yoshimura et al. (1999) [25] found that PICP and ICTP were not significantly correlated with change in femoral neck BMD in a cohort of Japanese 40–79-year-old male residents in the same town, over a 3-year period. As the coefficient of determination for these markers in bone loss prediction was only 5% for lumbar spine and 7% for femoral neck BMD, the authors concluded that these markers cannot predict the change in BMD at the individual level. Thus, our results confirm the findings of other studies conducted in adult men and suggest that the use of current biochemical markers to predict change in bone density in the general male population is of little value.

A limitation of longitudinal studies using DXA measurements is that the measurement error in such follow-up studies can be expected to be greater relative to the magnitude of the BMD measure of change than for a single measure of BMD in a cross-sectional study. In the present study, differences between baseline and follow-up DXA measurements due to methodological variations should be minimal, as the same type of DXA equipment (GE Lunar corp. Madison, WI, USA) and software for pencil beam densitometers (DPX) were used at baseline and follow-up. A possible explanation for the lack of clear association between the markers and BMD change could be that the change in BMD over 5 years was of insufficient magnitude to overcome the dilution of correlations by measurement error in the markers and BMD measurements. The present study, and most others, also are limited by the absence of measures and lack of control of factors, such as insulin-like growth factor-1 or parathyroid hormone, which have been found to correlate with a decline in BMD [22] and the sex-hormone binding globulin, which was associated with hip and spine BMD and CTx marker levels [23]. In any event, in middle-aged men, PINP, PICP and ICTP were not informative of future change in BMD over a 5-year period.

Another possible explanation for the low association between the markers and BMD in our study, and others, could be that biochemical markers are connected to aspects of bone quality rather than to bone mass. They might capture more dynamic bone properties such as bone remodeling or architectural characteristics independent of BMD measurements. Therefore, they could be a contributing factor in risk of fracture, even if bone mass measurements are unchanged or do not predict change in BMD.

In contrast to men, higher correlations between changes in BMD and biochemical markers like PICP, PINP and NTx have been reported in women [26],[27]. However, Bauer et al. (1999) [26] and Garnero et al. (1996) [27] suggested that the use of the available biochemical markers to predict change in bone density is of limited value for individual, untreated elderly women.

The difference in the results from most studies conducted in men, which found weak [28],[29] or no significant relationship between markers of bone remodeling and BMD [16], and those in women which found higher correlations, appear to reflect an influence of gender and hormonal differences. The results of several studies suggest that there are factors related to menopause that could contribute to a greater variation in bone metabolism in postmenopausal women when compared to premenopausal women or men [30],[31]. The biochemical markers express bone turnover in the whole skeleton, of which size is considerably influenced by gender [32],[33]. Also, bone metabolism in men is less affected by the influence of estrogen than in women. Therefore, changes in bone mass in women might be of greater magnitude that those found in men, which might not be detectable through statistical methods as they might also be comparable in size to the measurement error. There are also factors that might add to overall variation between different study results, like age-related changes, daily variations in individual levels of markers, age-related decrease in glomerular filtration or tubular reabsorption and metabolic rates that are not yet known for biochemical markers [18].

In conclusion, the correlations of markers with previous or future BMD changes (over a 5-year period) were not higher than those found in a cross-sectional analysis in men 35 to 69 years old. Baseline PINP, PICP and ICTP marker levels did not predict change in spine and femoral neck BMD in this group, and NTx levels explained a statistically significant, yet quite limited, portion of the variance in change in femoral BMD over the prior 5 years. Thus, we conclude that the usefulness of these markers in predicting age-related change in BMD is of limited value in adult men.

Acknowledgements

We thank Dr Laura Gibbons for the help with statistical analysis. We are grateful for support from the National Institutes of Health, USA, Grant AR 40857; the Work Environment Fund and the Academy of Finland, Finland (Grants 38332 and 42044); the Alberta Heritage Foundation for Medical Research, Canada; the CIHR Canada Research Chairs Program and EURODISC (QLK6-CT-2002-02582). Dr Donescu is supported by the Alberta Provincial CIHR Training Program in Bone and Joint Health.