Rapid Bone Loss Is Associated with Increased Levels of Biochemical Markers

Authors


  • Presented in part at the 25th European Symposium on Calcified Tissues, Harrogate, U.K., April 26, 1997.

Abstract

We examined associations of biochemical markers of bone turnover with rapid bone loss, as measured by changes in bone mineral density (BMD). To improve the precision of bone loss estimates, calcaneal BMD was measured up to eight times over a long interval (13 years) among postmenopausal women (mean age = 62 years at baseline). Women with fractures during the previous year, and users of corticosteroids, active vitamin D, bisphosphonates or calcitonin were excluded to avoid potential transient effects on marker levels. Among the remaining 354 women, markers were measured for 100 women with the fastest BMD loss (rapid loss group; mean = 2.2%/year) and 100 with the slowest loss (mean = 0.4%/year). Two markers of bone formation, serum bone alkaline phosphatase (Alkphase-B; BAP) and osteocalcin (NovoCalcin; OC), and two markers of bone resorption, urinary creatinine-corrected free deoxypyridinoline (Pyrilinks-D; DPD) and free pyridinolines (Pyrilinks; PYD), were measured. In separate logistic regression models, each of the markers was strongly associated with rapid loss: the odds of rapid loss increased by 1.8 to 2.0 times for each 1.0 standard deviation (SD) increase of the marker. For BAP levels 2 SD above the mean, the probability of rapid bone loss was 80%; in contrast, the probability was only 20% at 2 SD below the mean. The other markers yielded similar results. We conclude that these markers are associated with rapid bone loss; this relationship appears to be continuous, with progressively greater risk of rapid bone loss with increasing levels of biomarkers. Prospective studies that include the entire distribution of bone loss rates are warranted to confirm these findings.

INTRODUCTION

Low bone mineral density (BMD) is an important risk factor for osteoporotic fractures. Evidence suggests that rapid reductions in bone mass as measured by changes in BMD, or increased levels of biochemical markers of bone turnover, may also increase fracture risk independent of current BMD.(1–10) Some data also suggest that people with rapid bone turnover may exhibit greater response to treatment.(11–14) Thus, people with high turnover or rapid BMD decreases might be targeted for treatment to improve the efficacy of fracture prevention.

Biochemical markers might relate to fracture risk by providing an indication of bone quality, such as losses of trabecular connectivity or an increase in the number of resorption sites, which weaken the bone structure.(3,15) Alternatively, markers may provide an indication of the rate of decline in bone density. Rates of change are difficult to measure precisely over periods of less than several years because BMD measurement errors are similar in magnitude to the average yearly change. This measurement “noise” may explain the inability of some studies to demonstrate an association between markers and rates of change in BMD which might be observed in a longer study.

The current study utilizes rates of change in BMD measured over an average of 13 years among postmenopausal women in the Hawaii Osteoporosis Study, together with serum and urine samples collected at the end of the follow-up, to examine whether biochemical markers are associated with long-term, rapid bone loss.

MATERIALS AND METHODS

The analyses reported here are based on data from the Hawaii Osteoporosis Study (HOS), a longitudinal study of bone loss and fractures.(16,17) Biochemical markers were measured using samples of serum and urine collected at the end of follow-up, at the time of the final measurement of bone density. Participants in the HOS were derived from the population-based Honolulu Heart Program (HHP). All men of Japanese ancestry were invited to participate in the HHP in 1965; 8,006 out of 11,136 eligible men participated. A random subsample of men who had participated in the HHP, and their wives, was invited to participate in the HOS beginning in 1981. All subjects were of Japanese ancestry, but only data on the wives are presented here. These community-dwelling women (n = 1105; participation rate = 72%) ranged in age from 43 to 80 years, and 99% were postmenopausal at the time of the first examination. Additional HOS examinations which included measurements of calcaneus bone density, were initiated in November 1981, August 1982, August 1983, October 1984, February 1987, August 1989, and January 1992. Anteroposterior and lateral spine radiographs were obtained for a 50% random sample of participants at the first examination, and lateral radiographs were obtained for all participants at later examinations. A total of 721 women were examined at HOS examination 8 between 1992 and 1994; 538 of these women provided serum and urine samples. Additional details regarding recruitment and examination of participants in the HOS were reported previously.(16,17)

Bone density was measured at the calcaneus using a custom single-photon absorptiometer at examinations prior to 1990, and a single-energy X-ray (SXA) densitometer (OsteoAnalyzer, Norland, Fort Atkinson, WI, U.S.A.) thereafter. The densitometers were cross-calibrated by measuring people on both machines. The calcaneus measurements utilize the average of nine scan rows, which represents a 2.7 cm long region of lowest bone density. Regions of interest on subsequent scans were matched to the initial scan. Machine calibration was performed daily. Rates of change in bone density were calculated by linear regression of all BMD measurements against time.

Biochemical assays

Women with new vertebral fractures identified on radiographs at the end of follow-up, and women with nonspine fractures during the previous 12 months were excluded to avoid potential effects of fracture healing on marker levels. Women who used corticosteroids, active forms of vitamin D, bisphosphonates, or calcitonin were also excluded. Among the 538 women who provided serum and urine samples at HOS examination 8, 354 women remained after the exclusions for recent fractures and medication use. These 354 women were sorted by rates of change in BMD and divided into three groups: group 1, 100 women with the most rapid declines in BMD; group 2, 100 women with the slowest bone loss rates; and group 3, all other women. In the present study, biochemical markers were analyzed only for the 200 women with rapid and slow bone loss (groups 1 and 2, respectively).

Serum and urine samples were collected at the time of the eighth HOS examination clinic visit (1992–1994). Participants were asked to refrain from eating foods containing fat or gelatin within 12 h of their clinic visit. Serum samples were stored at −70°C, and urine was stored at −20°C until analysis, at which time they were shipped with dry ice to Metra Biosystems (Mountain View, CA, U.S.A.), where biochemical analyses were performed. Serum measures included bone alkaline phosphatase (BAP; Alkphase-B, Metra Biosystems) and osteocalcin (OC; NovoCalcin, Metra Biosystems). Urinary analyses included free deoxypyridinoline (DPD; Pyrilinks-D, Metra Biosystems), and free pyridinolines (PYD; Pyrilinks, Metra Biosystems). Creatinine-corrected values were calculated by dividing DPD and PYD by urinary creatinine (Cr) measured using a standard colorimetric assay (DPD/Cr and PYD/Cr). Assays were performed without knowledge of the rate of bone loss.

Statistics

The associations of biochemical markers with rapid bone loss were explored using logistic regression analysis, with rapid bone loss as the outcome (dependent) variable, and biochemical markers as predictor variables, adjusting for sample collection time and age. Adjustment for sample collection time was performed because diurnal fluctuations of as much as 30% have been reported for some resorption markers, and differences in sample collection times might introduce additional variability in marker levels which are unrelated to long-term rates of bone loss.(18–20) In models that used markers as continuous variables, odds ratios were calculated as the natural antilogarithm of the regression coefficient for the marker after first multiplying by the standard deviation (SD) of the marker; the SD was based on all women in this study. In other models, women were categorized as having marker levels in the highest, middle, or lowest third; odds ratios for each category were calculated as the natural antilogarithm of the regression coefficient. Odds are defined as p/(1 − p), where p is the probability of the event (in this case, rapid bone loss). When the outcome is rare (less than 5%), the odds ratio estimates the relative risk. In this study, half of the women were selected to have rapid bone loss, so the odds ratio overestimates the true relative risk.

RESULTS

The mean interval between the first and last BMD measurement was 13.4 years (SD = 1.0, range = 9.8–14.8), and the mean time of serum and urine collection was 11:40 a.m. (SD = 2.1 h). One unreported case of Paget's disease was identified in the rapid loss group and was excluded from subsequent analyses, leaving 100 women in the slow loss group and 99 in the fast loss group. Among women in the fast loss group, BAP levels were 28% higher, OC was 21% higher, PYD/Cr was 16% higher, and DPD/Cr levels were 17% higher on average, compared with the slow loss group. Differences in these markers between the fast and slow loss groups were highly significant (Table 1).

Table Table 1. Distributions of Variables by Category of BMD Loss Rate
original image

Correlations between the variables used in this study are presented in Table 2. The strongest correlation was observed between the two resorption markers, PYD/Cr and DPD/Cr (r = 0.88). However, significant correlations (r ≈ 0.4) were also observed between all other marker combinations. Correlations with bone loss rate were not examined because the sample was selected to exclude values in the middle of the range.

Table Table 2. Correlations Between Biochemical Markers
original image

All four markers were strongly associated with rapid loss in age-adjusted logistic regression models. The odds of rapid loss increased by 1.8 to 2.0 times per SD for each of the markers (Table 3). The magnitudes of association were similar for both resorption and formation markers. Age was not significantly associated with fast bone loss. When sample collection time was dropped from these models, the odds ratios decreased by 12% for DPD/Cr, 6% for PYD/Cr, 3% for BAP, and 0.5% for OC (data not shown).

Table Table 3. Associations of Individual Biochemical Markers with Rapid Bone Loss in Logistic Regression Models
original image

Next, women were categorized into three groups of equal size based on marker levels (Table 4), to examine whether the odds of rapid loss increased progressively with increasing marker levels, or if the relationship between markers and rapid loss was primarily due to the effects of marker values at one end of the distribution. Women in the highest third of BAP levels had an odds of rapid loss 4.7 times greater than the reference group (women in the lowest third of BAP). Women in the middle third had 1.8 times the odds compared with the lowest third. The trend of increasing risk of fast loss with increasing category of BAP was highly significant and indicated that the odds of rapid bone loss increased by a factor of 2.2 for each progressive increase in marker category (Table 5). Thus, using the analysis from Table 5, women in the middle third have 2.2 times the odds of rapid loss compared with women in the lowest third, and women in the highest third have 2.22 = 4.8 times the odds of the lowest third, confirming a progressive dose-response relationship. Similar results were observed for the other markers.

Table Table 4. Associations of Biochemical Marker Categories (Thirds) with Rapid Bone Loss in Logistic Regression Models
original image
Table Table 5. Associations of Biochemical Marker Thirds as an Ordinal Variable with Rapid Bone Loss in Logistic Regression Models
original image

To help illustrate the potential application of these findings, the probability of rapid bone loss was calculated from the logistic regression models over the range for each marker (Fig. 1). Women with BAP values equal to the mean of the entire sample had approximately a 50% probability of rapid bone loss, because the sample was selected so that half the women had rapid loss rates. Women with BAP values 2 SD below the mean had a 20% probability of rapid bone loss; thus, one out of five women with these BAP levels had rapid loss. At the other extreme, women with BAP values 2 SD above the mean had an 80% probability of rapid loss—four out of five such women had rapid bone loss. Women with values 1 SD above and 1 SD below the mean had probabilities of 67% and 33%, respectively, of rapid loss. The results were similar for other markers. Combining two marker measurements (such as one formation and one resorption marker, or two resorption markers) did not significantly improve odds ratios or probabilities of rapid loss (data not shown).

Figure FIG. 1.

Probability of rapid bone loss as a function of biochemical marker levels. The probability of rapid loss was calculated from the corresponding logistic regression equation for each marker (Table 3). At each level, from left to right, the markers were: BAP (solid bars), OC (diagonal hatch increasing from left to right), DPD (cross-hatched bars), and PYD (diagonal hatch decreasing from left to right). The means and SDs were based on all 199 samples (see Table 1).

DISCUSSION

Biochemical markers and bone loss rates have been associated with increased risk of hip and other osteoporotic fractures.(1–10) In some studies, the association of markers with fracture risk was independent of bone density, suggesting that the markers may reflect defects in bone architecture (such as reduced trabecular connectivity or excess bone remodeling sites) or increased rates of bone loss. Levels of both formation and resorption markers increase after menopause, and increased levels of either formation or resorption markers are generally considered to reflect the increased rate of bone turnover associated with postmenopausal bone loss.(21–24) Most studies that examined resorption markers found that higher levels were associated with increased fracture risk,(1,3,4,6,9) although at least one study found an increase in risk with lower resorption as measured by the carboxy-terminal telopeptide of type I collagen (ICTP).(10) Regarding markers of bone formation, some studies have found increased risk with high formation marker levels,(2,5) or nonsignificant associations with fracture risk,(1,5,6,10) whereas two studies reported significantly increased fracture risk when either osteocalcin(6) or procollagen I carboxy-terminal extension peptide (PICP)(10) levels were low.

In the current study, we examined associations of markers with rapid bone loss. At least two studies have reported that rapid bone loss increased the risk of subsequent fractures independent of the initial bone density.(7,9) For example, Sklarin et al.(7) reported that the risk of hip fracture increased 1.5 times for every increase in hip bone loss rate of 1.4% per year. The true association may be even stronger than this because bone loss measurement errors over the relatively short period of 3.5 years may have attenuated the true association in that study.

A strong association was observed between biochemical markers and rapid bone loss during the previous 13 years in this study. The average rate of bone loss was 1.8%/year faster in the rapid loss group, compared with the slow loss group. Increased levels of all four markers were associated with rapid bone loss, suggesting that both formation and resorption markers provide information on the bone turnover rate, which is in turn related to the bone loss rate. Combining two marker measurements did not significantly improve the association between high turnover and rapid bone loss, suggesting that measurement of only one marker is sufficient.

Some previous studies have also reported associations between markers (or bone turnover measured by radioisotope uptake) and bone loss rate,(4,8,12,13,25–32) whereas others have failed to find a significant association for certain markers.(26,27,29–31) Most studies had small samples sizes (fewer than 100 subjects) and short follow-up (typically less than 3 years), and some used total alkaline phosphatase or urinary hydroxyproline, which are not specific to bone. Furthermore, osteophytes and other sources of interference may obscure changes in spinal BMD.(30) Thus, differences in the types of markers, length of follow-up, skeletal measurement site, time since menopause, or other factors may be responsible for the lack of significant associations in some studies.

Although prospective fracture data were not available in this study, the clinical significance of using markers to detect rapid bone loss can be estimated using published data. The rapid loss group experienced an average of 7.3–1.5 = 5.8 mg/cm2/year greater bone loss than the slow group (Table 1). Over a period of 13 years (the length of this study), this difference in bone loss would correspond to more than 1 SD in bone density. Thus, while both groups have increased fracture risk (due to progressive bone loss), the increase is more than twice as great for the rapid loss group, because each SD decrease in bone density approximately doubles fracture risk.(33,34) Furthermore, the difference in fracture risk would increase progressively with time as long as the differences in loss rate persist.

Markers might complement or predict fracture risk better than measured rates of change in BMD (independent of initial BMD). For example, increased marker levels (as indicators of bone turnover) may represent an increase in the number of bone remodeling sites or trabecular perforations; such architectural defects may weaken bone beyond that predicted from measured decreases in BMD. Another possibility is that markers may provide a more representative indication of overall skeletal bone loss than would be obtained by measuring rates of change in BMD at a single skeletal site.

Advantages of this study include the long duration of follow-up and the relatively large sample size. The long observation period would yield stable, precise estimates of BMD rates of change; this may explain why biochemical markers of bone turnover were associated with rapid bone loss in the current study, whereas some previous studies with shorter follow-up periods failed to find significant associations. One limitation of the present study is that samples were collected at different times during the day. Although we adjusted for time of sample collection, statistical adjustment may not fully correct for the effects of sample collection time on marker levels. Furthermore, day-to-day variations, renal clearance, and other physiologic processes may have introduced additional variability in marker levels. These sources of variability may have caused the observed odds ratios to underestimate the true associations; stronger associations might have been observed if multiple samples had been available. Another limitation is that bone loss rates were measured prior to obtaining the serum and urine samples. In theory, this may not be important if markers level and loss rates are relatively stable over long periods. One prospective study found that average PYD and DPD levels did not change during the first 10 years after menopause.(32) However, prospective studies are needed to evaluate whether the associations of markers with future loss might be different than the relationship with past loss.

We conclude that markers of bone formation and resorption are strongly associated with rapid bone loss; this relationship appears to be continuous, with progressively greater probability of rapid bone loss with increasing levels of biomarkers. The associations of these markers with rapid bone loss are similar in magnitude to the associations of markers(1–3,5,6) and BMD(33,34) with fracture risk. Additional studies are warranted to confirm the associations reported here. Ideally, such studies should be of prospective design and should include the full distribution of loss rates instead of only fast and slow losers.

Acknowledgements

The authors thank the staff of the Hawaii Osteoporosis Center, especially Dee Tenma and Kristen Freeman, for providing technical and logistical support, and Mr. Dean Jenkins and Dr. Robert Cannon of Metra Biosystems, Inc. for their thoughtful comments on the manuscript. This study was funded in part by the National Institutes of Health, National Institute on Aging (grant #AG10412), Metra Biosystems, Inc., and the Hawaii Osteoporosis Foundation.

Ancillary