SEARCH

SEARCH BY CITATION

Keywords:

  • bone turnover;
  • mortality;
  • bone resorption;
  • BMD;
  • men

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Low BMD, high concentration of 17β-estradiol (17βE2), and decreased level of 25-droxycholecalciferol [25(OH)D] predict mortality. Our hypothesis is that high levels of biochemical bone turnover markers (BTMs) are independent predictors of mortality in home-dwelling men. In 781 men ≥50 yr of age followed up prospectively for 10 yr, we studied the association of BTMs with mortality after adjustment for confounders including BMD, major osteoporotic fractures, and concentrations of 17βE2 and 25(OH)D. Men who died had lower BMD and higher BTM levels. In multivariate models, mortality was higher in men with low BMD (lowest quartile) at the total hip, whole body, and ultradistal radius (HR = 1.49–1.70, p < 0.05). After exclusion of the first 3 yr, higher levels (fourth quartile) of bone resorption markers (free and total deoxypyridinoline and urinary and serum type I collagen C-telopeptide) predicted mortality in multivariate models adjusted for age, BMI, smoking habits, alcohol intake, physical performance and activity, comorbidities, total hip BMD, major osteoporotic fractures, creatinine clearance, season, and concentrations of 17βE2 and 25(OH)D (HR = 1.58–2.44, p < 0.05–0.001). In conclusion, in older community-dwelling men, increased bone resorption markers levels predicted mortality regardless of age and other confounders. Thus, in older men, high bone resorption may reflect poor current health status and poor aging.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Osteoporotic fractures are a major public health problem because of their mortality, morbidity, and costs.(1,2) Postfracture mortality depends on the fracture and the prefracture health status.(1,3) Low BMD predicts mortality regardless of age and confounders.(4–10) In contrast, association between biochemical bone turnover markers (BTMs) and mortality is poorly studied. In the frail elderly, high BTM levels predicted death regardless of age and other predictors of mortality.(11) However, this analysis was performed in men and women jointly, and men constituted 20% of the cohort. Thus, it is not clear whether these data apply to a younger home-dwelling population, especially men. In older men, bone resorption slightly increases, whereas bone formation shows various trends.(12) The scatter of BTM values is large and increases with age.(13) Because mortality reflects health status, the study of the link between BTMs and mortality can improve our understanding of determinants of bone turnover in men.

However, several parameters can interfere in the association between BTMs and mortality. Because BMD and BTM levels are correlated, it is not known which factor is predominant.(12,13) Moreover, increased bone resorption may be associated with an increased risk of fracture in men,(14) and major osteoporotic fractures are associated with higher mortality.(15) Several determinants of bone turnover in men were suggested (deficit of sex steroids and vitamin D, smoking, sedentary lifestyle).(16–18) We showed that elevated 17β-estradiol (17βE2) and low 25-hydroxycholecalciferol [25(OH)D] levels predict death in home-dwelling men.(19) We develop former data that a high 17βE2 level predicts death in critically ill, injured, or surgical patients and in men with severe infection.(20–23) A low level of 25(OH)D predicted all-cause, cardiovascular, and cancer mortality in both sexes.(24–26) However, both 17βE2 and vitamin D are involved in the regulation of bone metabolism in men.(16,17) Because they modify bone turnover and BMD, they can influence the relationship of BMD and BTM with mortality. Concentrations of testosterone and PTH did not predict mortality in our cohort.(19) However, testosterone level predicted mortality in elderly men with a low testosterone concentration.(27,28) Elevated PTH level predicted short-term mortality (<31 mo) in old persons with high morbidity and reduced survival (frail institutionalized old persons, critically ill patients, elderly with low-trauma hip fractures).(29–32)

Thus, our aim was to study the relationship between BTM levels and mortality during a 10-yr follow-up in community-dwelling men ≥50 yr of age, taking into account potential confounders such as lifestyle, physical performance, health status, BMD, hormone concentrations, and major prevalent and major incident fragility fractures.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Cohort

MINOS is a prospective study of male osteoporosis initiated in 1995 as a collaboration between the INSERM and Société de Secours Minière de Bourgogne (SSMB) in Montceau les Mines.(33) Letters inviting participation in the study were sent to a randomly selected sample of 3400 men 50–85 yr of age insured by SSMB. Among 841 men who provided informed consent, 781 men had BMD measurement, lateral radiograph of the spine, blood sampling, and urinary collection at the baseline examination in 1995–1996. Sixty men refused bone densitometry or blood sampling or had radiographs of poor quality. The study was accepted by the local ethics committee and performed in accordance with the Helsinki Declaration of 1975 as revised in 1983. For 182 men who died during the follow-up, dates of death were provided by the SSMB administration. For two men, we did not obtain data on their life status after 5 yr of follow-up. All other survivors were followed up for 10 yr.

At baseline, tobacco smoking was assessed as current smoker, former smoker >25 packet-years, former smoker ≤25 packet-years, or never-smoker. Alcohol intake was assessed as sum of current average weekly intakes of wine, beer, and spirits and divided into quartiles. Current leisure physical activity was calculated on the basis of the overall amount of time spent walking, gardening, and participating in leisure sport activities including seasonal activities. It was assessed in classes: <10, 10–19.99, 20–30, and >30 h/wk. Assessment of comorbidities present at baseline included ischemic heart disease (history of myocardial infarction, angina pectoris, and corresponding treatment), arterial hypertension, type I and II diabetes, Parkinson's disease, history of stroke, pulmonary diseases (chronic bronchitis, chronic obstructive pulmonary disease, silicosis, chronic respiratory insufficiency), prostate cancer, and gastrointestinal and liver diseases (peptic ulcer, ulcerative colitis, chronic pancreatitis, cirrhosis, chronic hepatitis). No patient reported Crohn disease or current cancer of the lung or digestive system. Data on the medical history were dichotomized as yes versus no, self-reported, and not further ascertained. Number of medications taken at the time of recruitment followed a skewed distribution (0–15, median and interquartile range: 2 [1, 5]).

Assessment of fractures

At baseline, major prevalent fractures were observed in 74 men (74 vertebral fractures in 66 men, 3 proximal humerus fractures, 1 fracture of pelvis, 1 fracture of distal femur, 5 fractures of proximal tibia, 2 men with multiple rib fractures). During the follow-up, major incident fractures occurred in 47 men (28 vertebral fractures in 27 men, 5 hip fractures, 7 proximal humerus fractures, 2 fractures of pelvis, 2 fractures of distal femur, 2 fractures of proximal tibia, and 4 men with multiple rib fractures).

Clinical tests

The physical performance score was calculated according to the short physical performance battery (SPPB).(34) It takes into account ability and time necessary to perform the test. The participants stood up and sat down from a hard chair five times.(34,35) Those who did not get up five times were scored unable (score 0). For men who accomplished five stands, time was scored 1–4 according to quartiles (4 for the shortest time). To assess standing balance, men stood with their feet in the side by side position for 10 s with eyes open and then for 10 s with eyes closed.(34,36) The test was scored as follows: unable to stand in the positions (score 0), <10 s with eyes open (score 1), 10 s with eyes open and <5 s with eyes closed (score 2), 10 s with eyes open and 5–9 s with eyes closed (score 3), 10 s in each position (score 4). To test dynamic balance, men performed a 10-step tandem walk forward on a line on the floor.(37) The participant was scored unable when he made <10 steps (score 0). For men who accomplished 10 steps, the time was scored 1–4 according to quartiles (4 for the shortest time). The 10-step tandem walk backward was scored in the same way. The global score was calculated by adding up the four tests (0–16).

Bone mass measurement

BMD was measured at the lumbar spine, hip, and whole body by DXA (QDR-1500; Hologic, Waltham, MA, USA) and at the distal forearm by single energy X-ray densitometry (DTX100 Osteometer).(33) At the forearm, we assessed the distal region including 20 mm of ulna and radius situated proximally to the site where the spacing between the two bones is 8 mm, and the ultradistal radius situated distally to the distal site (subchondral cortical bone is excluded). The CV of BMD was 0.33% for the spine phantom, 0.94% for the hip phantom, 0.62% for the human lumbar spine phantom, and 0.47% for the calibration standard of the forearm. Scans were analyzed manually. Scans with evident positioning error were excluded (three hip and three forearm scans).

Assessment of aortic calcifications

Aortic calcifications were assessed on baseline lateral radiographs of the lumbar spine by a semiquantitative method.(38) Calcific deposits in the abdominal aorta adjacent to the first four lumbar vertebrae were assessed using the 24-point severity scale (aortic calcification score [ACS]) for the posterior and anterior wall of the aorta using the midpoint of the intervertebral space above and below the vertebrae as boundaries. ACS reflects general cardiovascular status.(39) ACS was dichotomized (ACS ≥ 3 versus 0–2) because this threshold was associated with a higher mortality.(40)

Biochemical measurements

Fasting serum collected at baseline in the fasting state at 8:00 a.m. and 24-h urine were stored at −80°C. Serum osteocalcin (OC), bone-specific alkaline phosphatase (BALP), N-terminal extension propeptide of type I collagen (PINP; procollagen Intact P1NP; Orion Diagnostica, Oulunsalo, Finland), and β-isomerized C-terminal telopeptide of collagen type I (β-CTX-I), as well as urinary free and total deoxypyridinoline (DPD) were measured as described previously.(13) Renal creatinine clearance was estimated on the basis of serum creatinine concentration, age, and body weight by the formula of Cockroft and Gault.(41)

Hormone concentrations

Serum 17βE2 and total testosterone were measured by tritiated radioimmunoassay (RIA) after diethyl ether extraction.(42) For testosterone, the detection limit was 0.06 nM, and the interassay CV was 7.8% for 6 nM. For 17βE2, the detection limit was 11 pM. Interassay CV was 12.1% for 21 pM, 6% for 99 pM, and 9.4% for 169 pM. Sex hormone–binding globulin (SHBG) was measured by immunoradiometric assay (IRMA) (125I Coatria; Bio-Mérieux, Marce l'Etoile, France) with a detection limit of 0.5 nM and a CV of 4.1% for 16 nM. Serum 25(OH)D was measured by RIA (Incstar, Stillwater, MN, USA) after extraction with acetonitrile.(43) Intra-assay CV was 6.9% for 25 nmol/ml and 5.9% for 47 nmol/ml. Interassay CVs were 11–13%. Serum intact PTH was measured by immunochemoluminometric assay (Magic Lite; Ciba Corning Diagnostics, Medfield, MA, USA).(43) For a concentration of 4 pM, intra- and interassay CVs were 5% and 7%, respectively.

Statistical analysis

We used SAS 8.2 software (Cary, NC, USA). For unadjusted and age-adjusted comparisons of the survivors and the nonsurvivors, we used a log-rank test. Survival curves were modeled by Kaplan-Meier curves. As we have shown previously, mortality in men in the highest quartile of 17βE2 was slightly lower than in the three lower quartiles during the first 3 yr and higher thereafter.(19) Therefore, all the multivariate models were calculated by Cox model separately for the first 3 yr and for the period thereafter. First, we compared the mortality in four quartiles of BMD and BTM. Then, we modeled the risk of mortality according to BMD (lowest versus three upper quartiles) and BTMs (highest versus three lower quartiles). To adjust for seasonal variation, BTM quartiles were assessed separately in summer and in the three other seasons.(17) Data are presented as HR and 95% CI. The following confounders were studied: age (continuous), BMI <27.7 or >30.3 versus 27.7–30.3 kg/m2, smoking (four groups), log-transformed alcohol intake, professional physical activity (heavy versus light-moderate), leisure physical activity (<10 versus ≥10 h/wk), physical performance score (lowest versus three upper quartiles), history of ischemic heart disease, arterial hypertension, stroke, Parkinson's disease, diabetes mellitus, pulmonary diseases, gastrointestinal and liver diseases, prostate cancer, log-transformed number of drugs, ACS (≥3 versus 0–2), 17βE2 (highest versus three lower quartiles), and 25(OH)D (lowest versus three upper quartiles). Analyses of bone resorption markers were also adjusted for log-transformed estimated creatinine clearance rate. We selected these thresholds because age-adjusted mortality was similar in the three upper quartiles for physical performance score, BMD, and 25(OH)D; in the three lower quartiles for BTM levels and 17βE2; and in the three upper classes for leisure physical activity. Variables were retained in the model if p < 0.15 or if they changed HR value by >0.05.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Bivariate analysis

Nonsurvivors were, at baseline, older, lighter, and shorter (Table 1). They smoked more, had less leisure physical activity, poorer physical performance, more comorbidities, higher ACS, reported more major osteoporotic fractures, and took more medications. The nonsurvivors had lower BMD (except lumbar spine), higher serum levels of BALP and CTX-I, higher urinary excretion of free and total DPD, lower estimated creatinine clearance rate, lower 25(OH)D concentration, and higher levels of 17βE2 and PTH. After adjustment for age, the differences between the men who did or did not die were significant for smoking habits, leisure physical activity, physical performance score, ACS, BMD at distal forearm and ultradistal radius, serum levels of BALP, 17βE2, 25(OH)D, urinary free, and total DPD, and prevalence of ischemic heart disease, diabetes, pulmonary diseases, digestive diseases, and taking more than five medications.

Table Table 1.. Bivariate and Age-Adjusted Comparison of Men Who Died During the Study and Survivors: Baseline Data
Thumbnail image of

Prediction of mortality during the first 3 yr of follow-up

During this period, 30 deaths occurred. Except for age (HR = 1.14/yr; 95% CI: 1.07–1.21; p < 0.001), no other variable (BTM, BMD, hormones, incident or prevalent major osteoporotic fractures, smoking, physical activity and performance, comorbidities) predicted mortality in multivariate models (p > 0.15).

Association between baseline BMD and mortality

Mortality was higher in the first quartile of BMD and similar in the three higher quartiles (Fig. 1). The trend persisted after standardization for age (e.g., for distal forearm [first quartile: 38.77, 24.13, and 23.40 and fourth quartile 26.06 per 1000 person-years, χ2 = 10.12, p < 0.02] and for ultradistal radius BMD [39.03, 27.81, 20.82, and 26.65 per 1000 person-years, χ2 = 8.35, p < 0.05]). For other skeletal sites, age-adjusted mortality did not vary across BMD quartiles. In the multivariate models assessed for the lowest versus three upper quartiles of BMD, BMD of the total hip, whole body, and ultradistal radius predicted mortality (Table 2).

Table Table 2.. Risk of Mortality During the 10-yr Follow-Up According to Baseline BMD: Data Presented as HR and 95%CI for the Lowest Quartile of BMD vs. Three Higher Quartiles
Thumbnail image of
thumbnail image

Figure Figure 1. Survival of men according to baseline BMD. Survival of men from the MINOS cohort during the 10 yr of follow-up according to the quartiles (QI, lowest; QII; QIII; QIV, highest) of BMD presented with Kaplan-Meier curves: (left) total hip BMD and (right) distal forearm BMD.

Download figure to PowerPoint

Association between baseline levels of BTMs and mortality

Mortality was higher in the highest quartile of BTM levels and similar in the three lower quartiles (Fig. 2). The trend persisted after standardization for age (e.g., for BALP [χ2 = 10.66, p < 0.02], free DPD [25.12, 20.80, 25.53, and 37.58 per 1000 person-years, χ2 = 9.52, p < 0.05], total DPD [χ2 = 10.67, p < 0.02], and serum CTX-I [χ2 = 8.62, p < 0.05]). For the levels of OC, P1NP, and urinary CTX-I, age-adjusted mortality did not vary across BMD quartiles.

thumbnail image

Figure Figure 2. Survival of men according to baseline levels of biochemical BTMs. Survival of men from the MINOS cohort during the 10 yr of follow-up according to the quartiles (QI, lowest; QII; QIII; QIV, highest) of the levels of biochemical BTMs presented with Kaplan-Meier curves: (left) BALP and (right) creatinine-adjusted urinary free DPD.

Download figure to PowerPoint

After 3 yr, the highest quartile of BALP and total and free DPD was associated with higher age-adjusted mortality (Table 3). Further adjustment weakened the HR for BALP and increased the HR for bone resorption markers. In the fully adjusted models including total hip BMD (the most predictive skeletal site) and incident major osteoporotic fractures, the highest quartiles of all the bone resorption markers were associated with a significantly higher mortality. Lowest quartile of total hip BMD was associated with borderline increased mortality (HR= 1.47–1.49, p < 0.07). Results for BTMs were similar in models adjusted for BMD of ultradistal radius or whole body; however, in these models, BMD lost statistical significance.

Table Table 3.. Risk of Mortality After 3 yr of Follow-Up (up to 10 yr) According to Baseline Levels of Biochemical BTMs in the MINOS Cohort: Data Presented as HR and 95% CI for the Highest Quartile of Bone Marker vs. Three Lower Quartiles
Thumbnail image of

Association between confounding variables and mortality

In the fully adjusted models excluding the first 3 yr, the following variables were associated with higher mortality: age (HR = 1.10/yr, p < 0.001), diabetes (HR = 2.3–2.9, p < 0.001), ACS ≥ 3 (HR = 1.9–2.1, p < 0.001), highest 17βE2 quartile (HR = 1.9–2.3, p < 0.001), lowest 25(OH)D quartile (HR = 1.6–1.8, p < 0.05), smoking (HR = 1.2–1.3/class, p < 0.05), BMI <27.7 or >30.3 kg/m2 (HR = 1.8–2.1, p < 0.05), heavy professional physical activity (HR = 1.8–2.1, p < 0.05), and low leisure physical activity (HR = 1.3–1.5, p < 0.05). Neither incident nor prevalent major osteoporotic fractures predicted mortality in the multivariate models (HR = 1.1–1.3 and HR =0.8–1.0, respectively, p > 0.15).

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

We showed that increased levels (highest quartile) of bone resorption markers predict long-term all-cause mortality after adjustment for other confounders including BMD, major osteoporotic fractures, age, lifestyle factors, comorbidities, and concentrations of 17βE2 and 25(OH)D in a large cohort of community-dwelling men ≥50 yr of age.

Low BMD is an independent predictor of mortality both in men and women.(5–10) In our cohort, this association was slightly weaker but significant. BMD was predictive mainly in shorter follow-ups.(6,8–10) We assessed all-cause mortality, and the predictive power of BMD may vary according to the cause of death.(7,10) This association may depend on confounders and how they are analyzed (e.g., ACS may better reflect cardiovascular health than self-reported cardiovascular disease). We used confounders in the strongest form (e.g., applied thresholds of leisure physical activity and physical performance score were more significant than the same parameters used as continuous variables).

Sambrook et al.(11) showed that high BTM levels predict mortality in the elderly. However, this analysis was carried out in frail, sick old persons, followed up for <3 yr on average, and was not adjusted for BMD. Specific results for men were not provided. Thus, it was not known whether such an association is present in home-dwelling men and whether it is independent of other predictors of mortality.

Our data showed that bone resorption markers predict mortality more strongly than BMD. This is logical because the risk of death depends more on the current health status that influences current bone turnover, whereas BMD is determined progressively throughout life. Our data raise the possibility that the association between BMD and mortality is partly secondary to the correlation between BMD and BTMs. However, this point needs to be studied. It is less clear why bone resorption markers better predict mortality than bone formation markers, although in the study of Sambrook et al.,(11) serum CTX-I was also a more significant predictor of mortality than P1NP. Bone resorption may be more responsive to systemic factors influencing and reflecting health status, whereas bone formation is influenced both by factors regulating bone formation and by the intensity of prior bone resorption. Moreover, total OC is more stable than intact OC, although probably less sensitive to changes in bone turnover.(44) Circulating P1NP level may be also influenced by collagen synthesis in other tissues.(45,46)

Our data do not explain the association between bone resorption and the risk of death. We adjusted analyses for other predictors and determinants of mortality such as age, diabetes, and cardiovascular health status. We analyzed potential confounders associated with the increased bone resorption and mortality such as smoking, low physical mobility, and major osteoporotic fractures.(11,15,47) We used confounders in the strongest form to reduce residual variability. At recruitment, none of the participants had known bone metastases. However, certain variables were not recorded. For instance, depressive disorders are associated with increased BTM levels and higher mortality.(48,49) It is also possible that the confounding effect of some factors was not detected because of their inaccurate estimation or lack of power.

Several cytokines are potent stimulators of bone resorption, e.g., interleukin-1β, interkeukin-6, or TNFα.(50) These cytokines are involved in the regulation of metabolic processes in the cardiovascular, respiratory, and central nervous systems.(51) Their increased levels also predict mortality.(52,53) The effect of C-reactive protein on bone metabolism is not clear; however, its concentration correlated positively with bone resorption markers in women.(54) Its increased concentration is also associated with higher mortality in chronic obstructive pulmonary disease and in end-stage renal disease.(55,56)

The strengths of our study are prospective design, large cohort, long-term follow-up, complete data on mortality, and a large set of confounders. Our data concern home-dwelling older men. Because the proportional hazard assumption was not met for 17βE2, we analyzed separately the first 3 yr and the period thereafter. Early death exclusion is supposed to eliminate carriers of occult disease that rapidly lead to death. This method has been criticized.(57) However, the early death exclusion can reduce the confounding effect of preexisting diseases that produce an increase in mortality that attenuates over time.(58) In this case, the early mortality exclusion may better show the existing exposure—mortality relation.

Our study has limitations. The volunteers were lower-middle class, home-dwelling men without overt nutritional problems. The cohort is not population based and may not be representative of the French population; thus, our data cannot be extrapolated to other populations. The response rate was 24%, but responders and nonresponders did not differ.(19) Our project was designed to study osteoporosis; thus, some risk factors were not thoroughly recorded. Reliability of self-reported diseases was not verified by formal adjudication. We did not collect data on the duration, severity, or treatment of diseases, nor did we assess incident diseases developing during follow-up. We acknowledge the possibility of unmeasured residual confounding. A major weakness is that only all-cause mortality could be assessed, and we cannot establish à posteriori the causes of death. The system of SSMB was computerized in 2003, and before, causes of death were not noted systematically in the medical records. Because only 23% of the cohort died, the analysis may be underpowered and conclusions might not be definitive. For measurements of BTM, we used the most accurate systems available in 1997–1998. Blood samples were taken at 8:00 a.m. This obviates circadian variation but not day-to-day variation.(59) Some inaccuracy of urinary collection cannot be excluded, mainly in the oldest men. We used a robust and sensitive RIA for 25(OH)D.(43) However, it does not distinguish between vitamin D3 and D2, and it does not distinguish forms 25-hydroxy, 24,25-dihydroxy, and 25,26-dihydroxy. Our major methodological limitation was the measurement of 17βE2 using an RIA that does not provide accurate estimation of low concentrations. However, blood samples were obtained in 1995–1996 and we cannot measure 17βE2 by more sensitive methods such as liquid chromatography/mass spectrophotometry. This probably does not jeopardize the message that men with the highest 17βE2 concentration (>130 pM) have a higher risk of death. In contrast, the assessment of the relation between the 17βE2 level and mortality could be inaccurate for low 17βE2 concentrations.

In community-dwelling men ≥50 yr of age, higher bone resorption predicted mortality regardless of age and other confounders, including BMD, major osteoporotic fractures, lifestyle factors, comorbidities, and concentrations of 17βE2 and 25(OH)D. Thus, in older men, high bone resorption reflects poor current health status and poor aging. However, pathophysiological mechanisms underlying association between increased bone resorption and mortality in older men remain to be studied.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES