Osteoporos Int (2014) 25:1831–1836 DOI 10.1007/s00198-014-2654-0
ORIGINAL ARTICLE
Waning predictive value of serum adiponectin for fracture risk in elderly men: MrOS Sweden H. Johansson & A. Odén & M. K. Karlsson & E. McCloskey & J. A. Kanis & C. Ohlsson & D. Mellström
Received: 24 October 2013 / Accepted: 11 February 2014 / Published online: 9 May 2014 # International Osteoporosis Foundation and National Osteoporosis Foundation 2014
Abstract Summary Serum adiponectin is a risk factor for fracture. The predictive value attenuates with time in elderly men so that its use for the risk assessment in the long term is questionable. The study underlines the importance of testing the long-term stability of potential risk factors. Introduction High serum adiponectin is associated with an increased risk of fracture in elderly men. The aim of the present study was to determine the impact of adiponectin on the probability of fracture as a function of time. Methods The probability of osteoporotic fracture was computed in 989 elderly men from the MrOS study in Sweden. Baseline data included clinical risk factors for fracture, femoral neck BMD and serum adiponectin. Men were followed for up to 7.4 years with a mean follow up of 5.3 years (range 0.0– 7.4 years). Poisson regression was used to model the hazard function for osteoporotic fracture and death to determine the 10 year probability of fracture. Results During follow up, 124 men sustained one or more osteoporotic fracture. There was a significant interaction between adiponectin and time since baseline (p=0.026) such that the longer time since baseline, the lower the gradient of fracture risk. When using this interaction in the calculation of
H. Johansson : A. Odén : C. Ohlsson : D. Mellström Centre for Bone and Arthritis Research (CBAR), Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden M. K. Karlsson Clinical and Molecular Osteoporosis Research Unit, Department of Orthopedics and Clinical Sciences, Lund University, Skane University Hospital, Malmo, Sweden H. Johansson (*) : A. Odén : E. McCloskey : J. A. Kanis WHO Collaborating Centre for Metabolic Bone Diseases, University of Sheffield, Sheffield, UK e-mail: [email protected]
10-year probability of fracture, the probabilities of osteoporotic fracture varied little over the range of adiponectin values. Conclusion Serum adiponectin is a risk factor for fracture. Nevertheless, the predictive value attenuates with time so that its use for the risk assessment in the long term is questionable. This study underlines the importance of testing the long-term stability of potential risk factors that might be used in fracture risk assessment. Keywords Adiponectin . Elderly men . Probability of fracture . Osteoporotic fracture
Introduction Osteoporosis is characterized by reduced bone mass and disruption of bone architecture, resulting in increased bone fragility and increased fracture risk [1]. The publication of a World Health Organization (WHO) report on the assessment of fracture risk and its application to screening for postmenopausal osteoporosis in 1994 provided diagnostic criteria for osteoporosis based on the measurement of bone mineral density (BMD) and recognized osteoporosis as an established and well-defined disease [2]. The consequences of low BMD reside in the fractures that arise but many risk factors have been identified that contribute to a greater or lesser extent to fracture risk independently of BMD. The advent of FRAX® in 2008 [3] provided a clinical tool for the calculation of fracture probability from several well established risk factors. A major advantage of using fracture probability is that it standardises the output from the multiple techniques and sites used for assessment and also permits the presence or absence of risk factors other than BMD to be incorporated as a single metric. Today FRAX relies on a handful of risk factors for fracture (age, BMI, prior fracture, family history of hip fracture, current smoking, alcohol use, glucocorticoid use, history of
1832
rheumatoid arthritis and femoral neck BMD [4]) and there are more potential risk factors or aspects of the current risk factors that could enhance the prediction of fracture [5, 6]. Adiponectin is a hormone produced mainly by adipocytes. Culture of osteoblast-like cells with adiponectin stimulates osteoblastogenesis and the osteoclast receptor activator of NF-kB ligand (RANKL) pathway while inhibiting its decoy receptor, osteoprotegerin [7]. Adipocytes and osteoblasts share a common progenitor, and there is, therefore, potential for both autocrine and endocrine effects of adiponectin on skeletal metabolism. We have earlier showed that there was a significant association between adiponectin and osteoporotic fracture adjusted for age, BMD, general health and previous fracture (HR per SD 1.32 (95 % CI 1.13–1.54) in an elderly Swedish male cohort, MrOs [8]. The aim of the present study was to determine the predictive value of serum adiponectin over time.
Materials and methods MrOS is a multi-centre, prospective cohort study of elderly men in Sweden, Hong Kong and the United States [9]. The present study is based on data from the Swedish MrOS cohort of older men recruited at medical centres in Gothenburg. Details have been described previously [10, 11]. In brief, men aged 70–81 years were randomly identified using national population registers. To be eligible for the study, men had to be able to walk without aids, provide self-reported data and give written informed consent. There were no other exclusion criteria. The participation rate in MrOS Sweden was 45 %. In the present report, the baseline data in MrOS Sweden Gothenburg was used together with up to 7 years of follow up for fracture for 989 men with data on adiponectin and femoral neck BMD. At baseline, height (in centimeters) and weight (in kilograms) were measured, and BMI was calculated as kilograms per square meter. The international MrOS questionnaire [9] was administered at baseline to collect information about current smoking, number and type of medications, fracture history, family history of hip fracture, history of diseases (rheumatoid arthritis) and the use of alcohol. Use of alcohol was expressed as more than two glasses of alcohol-containing drinks per day, calculated from the reported frequency and amount of alcohol use. Areal bone mineral density (BMD) was measured using the Hologic QDR 4500/A-Delphi (Hologic, Bedford, MA, USA) at the femoral neck. Plasma and serum samples were collected at 8 am after at least 10 hours of fasting and abstinence from smoking. Samples were frozen immediately and stored at −80ºC. Serum adiponectin was measured at baseline in 989 men and analysed with an ELISA-kit (Linco Research, MO,
Osteoporos Int (2014) 25:1831–1836
USA; inter-assay CV 8 %) at the Department of Clinical Chemistry, Sahlgrenska University Hospital. Fracture outcomes were collected up to 1 September 2009, and the site of fracture and time to fracture were retrieved from computerized X-ray archives in Gothenburg using a unique personal registration number held by all Swedish citizens. Fractures reported by the study subject, but not independently confirmed, were not included in the present study. Deaths were documented from the National Cause of Death Register up to the end of 2009. This register comprises records of all deaths in Sweden and is more than 99 % complete. Emigrants were followed up to the day of emigration. Participants were followed until death or fracture. The mean follow up was 5.3 years (range 0.0–7.4 years) after the baseline examination. For fracture outcomes, we included sites considered to be associated with osteoporosis [12], i.e. fractures of the spine, sacrum, ribs, pelvis, humerus, forearm, elbow, hip, other femoral, clavicle, scapula, sternum. Fractures of the skull, face, hands and fingers, feet and toes, tibia and fibula, ankle and patella were excluded. Statistical methods An extension of Poisson regression models [13] was used to study the association between age, the time since baseline (used as a continuous variable), serum adiponectin, other covariates on the one hand and on the other hand, the risk of fracture. In contrast to logistic regression, the Poisson regression utilises the length of each individual’s follow-up period and the hazard function is assumed to be exp (β0 +β1 ·time from baseline+β2 ·current age+β3 ·current variable of interest). The observation period of each participant was divided in intervals of 1 month. One fracture per person was counted, and time to the first fracture was counted. A similar approach was used to examine other predictors of fracture, and a final multivariable model was constructed to determine which predictors had an independent contribution to risk. The distribution of serum adiponectin was not normal and was normalised using a piecewise linear function. The association between predictive factors and risk of fracture were described as a gradient of risk (GR), expressed as the hazard ratio per 1 standard deviation change. In order to study the linearity of the association between serum adiponectin and fracture risk in more detail, a spline Poisson regression model was fitted using knots at the 10th, 50th and 90th percentiles of serum adiponectin. The splines were second order functions between the breakpoints and linear functions at the tails resulting in a smooth curve. Two hazard models were constructed to calculate the 10 year probability. These comprised the risk of osteoporotic fracture and the risk of death both using Poisson regression
Osteoporos Int (2014) 25:1831–1836
[14]. When building the hazard functions on all risk factors comprised in FRAX (age, BMI, prior fracture, family history of hip fracture, current smoking, alcohol use, glucocorticoid use, history of rheumatoid arthritis and femoral neck BMD [4]), only 634 men (64 %) had enough data to contribute. When each single risk factor used in FRAX were tested, just age, previous fracture and BMD has a statistically significant association with future osteoporotic fracture. When adjusted for age, BMD and adiponectin, previous fractures were not significantly associated with future osteoporotic fracture. When 10 year probability was calculated just using age, BMD and adiponectin, the whole cohort (n=989) was used. The hazard functions used to calculate the 10 year probability in this report comprised the risk of osteoporotic fracture and the risk of death using age, time since baseline, femoral neck BMD and adiponectin as covariates. Several additional probability models were used that varied the hazard function of fracture (Table 1). For the simplest model (model A), the hazard function for fracture depended on age, serum adiponectin and femoral neck BMD. Other models included an interaction between time since baseline and adiponectin. Model B assumed that the interaction with time would be constant for 10 years. We have previously described a nonlinear relationship between adiponectin and fracture risk, in that serum adiponectin below 18 mg/ml had no predictive value for fracture risk [8] and this phenomenon was accommodated in model C. When studying the time interaction with a linear model, the hazard ratio fell below 1 after 6.2 years. Since it is implausible that the analyte would be a protective factor for fracture, an additional model was constructed. In model D, the predictive power of adiponectin decreased with time (as in model B) but never with a hazard ratio of less than 1.00. The hazard function for mortality was not changed in the different models, using age, adiponectin and femoral neck BMD with no interactions. As an internal control, we assessed the predictive value of femoral neck BMD for fracture risk and its interaction with time. Two-sided p value was used for all analyses and p0.30), i.e. the GR was the same for all ages. Table 3 compares the 10 year fracture probabilities in men aged 75 years with a BMD T-score of −2.0 SD when adiponectin was set at 5.6 μg/ml (the 10th percentile) and when adiponectin was set at 20.1 μg/ml ( at the 90th percentile). When assuming no interaction between time since baseline and adiponectin (model A), i.e. assuming a constant gradient of risk (GR 1.40) for 10 years, the 10 year probabilities were 21 and 39 % respectively—an increase of 86 % over this range of serum adiponectin. When comparing the same probabilities for the same clinical scenario but incorporating the interaction between time since baseline and adiponectin (model B), the probability for the 90th percentile of adiponectin was only 3 % higher. The difference was slightly greater (6 %) when the GR for adiponectin was allowed to be nonlinear (model C). When GR for adiponectin was assumed not to fall below 1.00, the difference was 29 % (model D).
1834
Osteoporos Int (2014) 25:1831–1836
Table 2 Characteristics of 989 men aged 70–81 years at baseline according to fracture status during follow up Variable
Age (years) BMI (kg/m2) Femoral neck BMD (g/cm2) Adiponectin (μg/ml) Glucocorticoids (current) Current smoking Previous osteoporotic fracture Parental history of hip fracture Rheumatoid arthritis Alcohol 3 or more units per day a
No fracture during follow up (n=865)
Fracture during follow up (n=124)
Significance (two-sided p value)a
Mean±(SD) 75.3±3.2 26.2±3.5 0.86±0.14
Mean±(SD) 75.5±3.2 25.8±3.8 0.80±0.13
0.30
Poisson regression adjusted for age and time since baseline
b
percent of total number of men without a fracture (n) with a positive response
c
percent of total number of men with a fracture (n) with a positive response
The contrasting effect of a constant or decreasing predictive value for adiponectin on fracture probability is shown in Fig. 2 (model A and B). When the gradient of risk (GR) was assumed to be constant over time, fracture probability rose progressively with increasing serum values of adiponectin. When the GR was assumed to decrease with time, the 10 year probabilities were more or less constant over the range of serum adiponectin. In the case of femoral neck BMD, the hazard ratio per SD difference in BMD for fracture was 1.50 (95 % CI 1.22–1.85). There was no significant interaction with time.
Gradient of risk (HR/SD)
2.5
2.0
1.5
1.0
0.5
0.0 0
1
2
3
4
5
6
7
Time since baseline (years)
Figure 1 Gradient of risk per 1 SD (95 % confidence interval) for the association between serum adiponectin and the risk of osteoporotic fracture adjusted for age, BMD and time since baseline
Discussion In this prospective study of community-living older men, we describe a significant association between serum adiponectin and the risk of osteoporotic fracture that was independent of age and BMD. These observations confirm earlier findings of an association between serum adiponectin and the risk of fracture in men [15–17]. The novel feature of the present study was to incorporate these finding into models of fracture probability. When serum adiponectin was assumed to have a constant gradient of risk (i.e. a 40 % increase in fracture risk per 1 SD increase in serum adiponectin) over 10 years, fracture probability increased progressively with increasing values of serum adiponectin. However, our data indicate that the predictive value of serum adiponectin is not constant and wanes progressively with time. The interaction with time had a very marked effect over a 10 year time horizon. Thus, between the 10th and 90th percentile of serum adiponectin, 10 year fracture probabilities were only 3 % higher at the 90th percentile when the interaction with time was not included in the model, whereas the increase was by 86 % when the gradient of risk was assumed to be constant. The present findings have important implications for the incorporation of risk variables into fracture risk prediction models in that they illustrate the need to determine the stability of a risk indicator with time. For this reason, we routinely checked if there was a significant interaction with the variable ‘time since baseline’ and BMD, BMI and clinical risk factors in the meta-analyses used to synthesise FRAX [4]. To our knowledge, this is the first time a variable with a significant
Osteoporos Int (2014) 25:1831–1836 Table 3 10 year probability of osteoporotic fracture (%) for a man aged 75 years with a BMD Tscore of −2.0 SD using different model assumptions (see text)
Serum adiponectin is set at the 10th percentile (5.6 μg/ml) and 90th percentile (20.1 μg/ml)
1835
Model
A B C D
10 year probability of osteoporotic fracture Adiponectin (10th percentile)
Adiponectin (90th percentile)
21 32 31 28
39 33 33 36
% % % %
interaction with time since baseline has been identified and incorporated into a model of fracture probability. If the associations that we describe are reproduced in other male cohorts with a wider age range, serum levels of adiponectin may have a role in the short-term assessment of fracture risk in men but a limited or no role over a 5 to 10 year time horizon. A waning effect with time is not unexpected. Indeed, the risk of a subsequent fracture following a previous fracture appears to decrease with time [18, 19], as does the risk of fall following a fall [20]. We have previously reported a waning of effect of serum vitamin D to predict mortality [21]. Given the inevitability of death, all risk factors for death will have no predictive value with an infinite follow-up time. With shorter observation times, a waning of effect may be due to heterogeneity in the natural history of disease as reported for some of the biochemical markers of bone turnover [22, 23]. In one of these studies with a 5-year follow-up period, the ratio of carboxylated to total serum osteocalcin was a very strong risk factor for fracture over the early period of observation, but the predictive value of the marker decreased progressively with time [22, 24]. In the case of
Figure 2 10 year probability of osteoporotic fracture (%) for a man aged 75 years with a BMD T-score of −2.0
% % % %
Ratio between 90th and 10th percentile
1.86 1.03 1.06 1.29
adiponectin, heterogeneity in natural history might reflect changes in fat mass. For normally distributed variables, the long-term predictive value can be derived from the correlation coefficient between repeated samples over a defined interval of time [25]. The work presented in this article has a number of strengths and weaknesses. A strength with regard to the principal finding is that we were able to test the long-term predictive value of femoral neck BMD which we have shown in previous studies to be stable with time [26]. The hazard ratio per SD was stable with time, providing a type of internal control on which to evaluate the properties of serum adiponectin. The general limitations to the MrOs study have previously been discussed [8]. A significant limitation for the present analysis is the narrow age range of men that were recruited (70–81 years), so that our findings may not be applicable to younger ages or women. The assay that we used did not distinguish high and low molecular weight forms of adiponectin, which may have varying biological activities, but this limitation would be expected to weaken rather than strengthen any association of the analyte with fracture risk. Another limitation is that the mean follow up in this cohort was 5.3 years, and the maximum follow up was 7.4 years from which we calculated the 10 year probability of fracture. This extrapolation in time is, however, robust given that predictive value of serum adiponectin for fracture outcomes was no longer apparent after 6 years of observation. We conclude that there is a significant association between serum adiponectin levels and risk of osteoporotic fracture in elderly men and that the association decreases over time. The attenuation of the predictive value of serum adiponectin with time impairs its utility as a risk factor in the calculation of 10 year probability of fracture. This study underlines the importance of testing the longterm stability of potential risk factors in risk assessment.
Acknowledgments HJ was supported by an ESCEO-AMGEN Osteoporosis Fellowship Award. Amgen had no input into the analysis plan or in the writing of this report.
1836 Conflicts of interest None.
References 1. Consensus development conference: diagnosis, prophylaxis, and treatment of osteoporosis (1993). Am J Med 94 (6):646–650. 2. Assessment of fracture risk and its application to screening for postmenopausal osteoporosis. Report of a WHO Study Group (1994). World Health Organ Tech Rep Ser 843:1–129 3. Kanis JA on behalf of the World Health Organization Scientific Group (2008) Assessment of osteoporosis at the primary healthcare level, Technical Report. WHO Collaborating Centre, University of Sheffield, UK 4. Kanis JA, Oden A, Johnell O, Johansson H, De Laet C, Brown J, Burckhardt P, Cooper C, Christiansen C, Cummings S, Eisman JA, Fujiwara S, Gluer C, Goltzman D, Hans D, Krieg MA, La Croix A, McCloskey E, Mellstrom D, Melton LJ 3rd, Pols H, Reeve J, Sanders K, Schott AM, Silman A, Torgerson D, van Staa T, Watts NB, Yoshimura N (2007) The use of clinical risk factors enhances the performance of BMD in the prediction of hip and osteoporotic fractures in men and women. Osteoporos Int 18(8):1033–1046 5. Leslie WD, Lix LM, Johansson H, Oden A, McCloskey E, Kanis JA (2011) Spine-hip discordance and fracture risk assessment: a physician-friendly FRAX enhancement. Osteoporos Int 22(3):839– 847. doi:10.1007/s00198-010-1461-5 6. Kanis JA, Johansson H, Oden A, McCloskey EV (2011) Guidance for the adjustment of FRAX according to the dose of glucocorticoids. Osteoporos Int 22(3):809–816. doi:10.1007/s00198-010-1524-7 7. Luo XH, Guo LJ, Yuan LQ, Xie H, Zhou HD, Wu XP, Liao EY (2005) Adiponectin stimulates human osteoblasts proliferation and differentiation via the MAPK signaling pathway. Exp Cell Res 309(1):99–109. doi:S0014-4827(05)00252-1 [pii] 10.1016/j.yexcr. 2005.05.021 8. Johansson H, Oden A, Lerner UH, Jutberger H, Lorentzon M, Barrett-Connor E, Karlsson MK, Ljunggren O, Smith U, McCloskey E, Kanis JA, Ohlsson C, Mellstrom D (2012) High serum adiponectin predicts incident fractures in elderly men: osteoporotic fractures in men (MrOS) Sweden. J Bone Miner Res 27(6):1390– 1396. doi:10.1002/jbmr.1591 9. Orwoll E, Blank JB, Barrett-Connor E, Cauley J, Cummings S, Ensrud K, Lewis C, Cawthon PM, Marcus R, Marshall LM, McGowan J, Phipps K, Sherman S, Stefanick ML, Stone K (2005) Design and baseline characteristics of the osteoporotic fractures in men (MrOS) study—a large observational study of the determinants of fracture in older men. Contemp Clin Trials 26(5):569–585. doi:10. 1016/j.cct.2005.05.006 10. Mellstrom D, Vandenput L, Mallmin H, Holmberg AH, Lorentzon M, Oden A, Johansson H, Orwoll ES, Labrie F, Karlsson MK, Ljunggren O, Ohlsson C (2008) Older men with low serum estradiol and high serum SHBG have an increased risk of fractures. J Bone Miner Res 23(10):1552–1560. doi:10.1359/jbmr.080518 11. Mellstrom D, Johnell O, Ljunggren O, Eriksson AL, Lorentzon M, Mallmin H, Holmberg A, Redlund-Johnell I, Orwoll E, Ohlsson C (2006) Free testosterone is an independent predictor of BMD and prevalent fractures in elderly men: MrOS Sweden. J Bone Miner Res 21(4):529–535. doi:10.1359/jbmr.060110
Osteoporos Int (2014) 25:1831–1836 12. Kanis JA, Oden A, Johnell O, Jonsson B, de Laet C, Dawson A (2001) The burden of osteoporotic fractures: a method for setting intervention thresholds. Osteoporos Int 12(5):417–427 13. Breslow NE, Day NE (1987) Statistical methods in cancer research. IARC Sci Publ No 32 Volume II:131–135 14. Kanis JA, Johnell O, Oden A, Johansson H, McCloskey E (2008) FRAX and the assessment of fracture probability in men and women from the UK. Osteoporos Int 19(4):385–397 15. Barbour KE, Zmuda JM, Boudreau R, Strotmeyer ES, Horwitz MJ, Evans RW, Kanaya AM, Harris TB, Bauer DC, Cauley JA (2011) Adipokines and the risk of fracture in older adults. J Bone Miner Res 26(7):1568–1576. doi:10.1002/jbmr.361 16. Araneta MR, von Muhlen D, Barrett-Connor E (2009) Sex differences in the association between adiponectin and BMD, bone loss, and fractures: the Rancho Bernardo study. J Bone Miner Res 24(12): 2016–2022. doi:10.1359/jbmr.090519 17. Kanazawa I, Yamaguchi T, Yamamoto M, Yamauchi M, Yano S, Sugimoto T (2009) Relationships between serum adiponectin levels versus bone mineral density, bone metabolic markers, and vertebral fractures in type 2 diabetes mellitus. Eur J Endocrinol 160(2):265– 273. doi:10.1530/eje-08-0642 18. Johnell O, Oden A, Caulin F, Kanis JA (2001) Acute and long-term increase in fracture risk after hospitalization for vertebral fracture. Osteoporos Int 12(3):207–214 19. Johnell O, Kanis JA, Oden A, Sernbo I, Redlund-Johnell I, Petterson C, De Laet C, Jonsson B (2004) Fracture risk following an osteoporotic fracture. Osteoporos Int 15(3):175–179. doi:10.1007/s00198003-1514-0 20. Johansson H, Oden A, Karlsson MK, Rosengren B, Ljunggren O, McCloskey E, Kanis J, Ohlsson C, Mellstrom D (2013) FRAX predicts future falls in elderly men. MrOs Sweden. JBMR 28((Suppl 1):S166 21. Johansson H, Oden A, Kanis J, McCloskey E, Lorentzon M, Ljunggren O, Karlsson MK, Thorsby PM, Tivesten A, BarrettConnor E, Ohlsson C, Mellstrom D (2011) Low serum vitamin D is associated with increased mortality in elderly men: MrOS Sweden. Osteoporos Int. doi:10.1007/s00198-011-1809-5 22. Luukinen H, Kakonen SM, Pettersson K, Koski K, Laippala P, Lovgren T, Kivela SL, Vaananen HK (2000) Strong prediction of fractures among older adults by the ratio of carboxylated to total serum osteocalcin. J Bone Miner Res 15(12):2473–2478. doi:10. 1359/jbmr.2000.15.12.2473 23. Ivaska KK, Gerdhem P, Vaananen HK, Akesson K, Obrant KJ (2010) Bone turnover markers and prediction of fracture: a prospective follow-up study of 1040 elderly women for a mean of 9 years. J Bone Miner Res 25(2):393–403. doi:10.1359/jbmr.091006 24. Ivaska K, McGuigan F, Obrant K, Gerdhem P, Akesson K (2013) Bone turnover markers and prediction of fractures in a 15-year follow-up study of elderly women. JBMR 28(Suppl 1):s34 25. Kanis JA, Johnell O, Oden A, Jonsson B, De Laet C, Dawson A (2000) Prediction of fracture from low bone mineral density measurements overestimates risk. Bone 26(4):387–391. doi:S87563282(00)00238-6 [pii] 10.1016/S8756-3282(00)00238-6 26. Johnell O, Kanis JA, Oden A, Johansson H, De Laet C, Delmas P, Eisman JA, Fujiwara S, Kroger H, Mellstrom D, Meunier PJ, Melton LJ 3rd, O’Neill T, Pols H, Reeve J, Silman A, Tenenhouse A (2005) Predictive value of BMD for hip and other fractures. J Bone Miner Res 20(7):1185–1194. doi:10. 1359/JBMR.050304
ORIGINAL ARTICLE
Waning predictive value of serum adiponectin for fracture risk in elderly men: MrOS Sweden H. Johansson & A. Odén & M. K. Karlsson & E. McCloskey & J. A. Kanis & C. Ohlsson & D. Mellström
Received: 24 October 2013 / Accepted: 11 February 2014 / Published online: 9 May 2014 # International Osteoporosis Foundation and National Osteoporosis Foundation 2014
Abstract Summary Serum adiponectin is a risk factor for fracture. The predictive value attenuates with time in elderly men so that its use for the risk assessment in the long term is questionable. The study underlines the importance of testing the long-term stability of potential risk factors. Introduction High serum adiponectin is associated with an increased risk of fracture in elderly men. The aim of the present study was to determine the impact of adiponectin on the probability of fracture as a function of time. Methods The probability of osteoporotic fracture was computed in 989 elderly men from the MrOS study in Sweden. Baseline data included clinical risk factors for fracture, femoral neck BMD and serum adiponectin. Men were followed for up to 7.4 years with a mean follow up of 5.3 years (range 0.0– 7.4 years). Poisson regression was used to model the hazard function for osteoporotic fracture and death to determine the 10 year probability of fracture. Results During follow up, 124 men sustained one or more osteoporotic fracture. There was a significant interaction between adiponectin and time since baseline (p=0.026) such that the longer time since baseline, the lower the gradient of fracture risk. When using this interaction in the calculation of
H. Johansson : A. Odén : C. Ohlsson : D. Mellström Centre for Bone and Arthritis Research (CBAR), Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden M. K. Karlsson Clinical and Molecular Osteoporosis Research Unit, Department of Orthopedics and Clinical Sciences, Lund University, Skane University Hospital, Malmo, Sweden H. Johansson (*) : A. Odén : E. McCloskey : J. A. Kanis WHO Collaborating Centre for Metabolic Bone Diseases, University of Sheffield, Sheffield, UK e-mail: [email protected]
10-year probability of fracture, the probabilities of osteoporotic fracture varied little over the range of adiponectin values. Conclusion Serum adiponectin is a risk factor for fracture. Nevertheless, the predictive value attenuates with time so that its use for the risk assessment in the long term is questionable. This study underlines the importance of testing the long-term stability of potential risk factors that might be used in fracture risk assessment. Keywords Adiponectin . Elderly men . Probability of fracture . Osteoporotic fracture
Introduction Osteoporosis is characterized by reduced bone mass and disruption of bone architecture, resulting in increased bone fragility and increased fracture risk [1]. The publication of a World Health Organization (WHO) report on the assessment of fracture risk and its application to screening for postmenopausal osteoporosis in 1994 provided diagnostic criteria for osteoporosis based on the measurement of bone mineral density (BMD) and recognized osteoporosis as an established and well-defined disease [2]. The consequences of low BMD reside in the fractures that arise but many risk factors have been identified that contribute to a greater or lesser extent to fracture risk independently of BMD. The advent of FRAX® in 2008 [3] provided a clinical tool for the calculation of fracture probability from several well established risk factors. A major advantage of using fracture probability is that it standardises the output from the multiple techniques and sites used for assessment and also permits the presence or absence of risk factors other than BMD to be incorporated as a single metric. Today FRAX relies on a handful of risk factors for fracture (age, BMI, prior fracture, family history of hip fracture, current smoking, alcohol use, glucocorticoid use, history of
1832
rheumatoid arthritis and femoral neck BMD [4]) and there are more potential risk factors or aspects of the current risk factors that could enhance the prediction of fracture [5, 6]. Adiponectin is a hormone produced mainly by adipocytes. Culture of osteoblast-like cells with adiponectin stimulates osteoblastogenesis and the osteoclast receptor activator of NF-kB ligand (RANKL) pathway while inhibiting its decoy receptor, osteoprotegerin [7]. Adipocytes and osteoblasts share a common progenitor, and there is, therefore, potential for both autocrine and endocrine effects of adiponectin on skeletal metabolism. We have earlier showed that there was a significant association between adiponectin and osteoporotic fracture adjusted for age, BMD, general health and previous fracture (HR per SD 1.32 (95 % CI 1.13–1.54) in an elderly Swedish male cohort, MrOs [8]. The aim of the present study was to determine the predictive value of serum adiponectin over time.
Materials and methods MrOS is a multi-centre, prospective cohort study of elderly men in Sweden, Hong Kong and the United States [9]. The present study is based on data from the Swedish MrOS cohort of older men recruited at medical centres in Gothenburg. Details have been described previously [10, 11]. In brief, men aged 70–81 years were randomly identified using national population registers. To be eligible for the study, men had to be able to walk without aids, provide self-reported data and give written informed consent. There were no other exclusion criteria. The participation rate in MrOS Sweden was 45 %. In the present report, the baseline data in MrOS Sweden Gothenburg was used together with up to 7 years of follow up for fracture for 989 men with data on adiponectin and femoral neck BMD. At baseline, height (in centimeters) and weight (in kilograms) were measured, and BMI was calculated as kilograms per square meter. The international MrOS questionnaire [9] was administered at baseline to collect information about current smoking, number and type of medications, fracture history, family history of hip fracture, history of diseases (rheumatoid arthritis) and the use of alcohol. Use of alcohol was expressed as more than two glasses of alcohol-containing drinks per day, calculated from the reported frequency and amount of alcohol use. Areal bone mineral density (BMD) was measured using the Hologic QDR 4500/A-Delphi (Hologic, Bedford, MA, USA) at the femoral neck. Plasma and serum samples were collected at 8 am after at least 10 hours of fasting and abstinence from smoking. Samples were frozen immediately and stored at −80ºC. Serum adiponectin was measured at baseline in 989 men and analysed with an ELISA-kit (Linco Research, MO,
Osteoporos Int (2014) 25:1831–1836
USA; inter-assay CV 8 %) at the Department of Clinical Chemistry, Sahlgrenska University Hospital. Fracture outcomes were collected up to 1 September 2009, and the site of fracture and time to fracture were retrieved from computerized X-ray archives in Gothenburg using a unique personal registration number held by all Swedish citizens. Fractures reported by the study subject, but not independently confirmed, were not included in the present study. Deaths were documented from the National Cause of Death Register up to the end of 2009. This register comprises records of all deaths in Sweden and is more than 99 % complete. Emigrants were followed up to the day of emigration. Participants were followed until death or fracture. The mean follow up was 5.3 years (range 0.0–7.4 years) after the baseline examination. For fracture outcomes, we included sites considered to be associated with osteoporosis [12], i.e. fractures of the spine, sacrum, ribs, pelvis, humerus, forearm, elbow, hip, other femoral, clavicle, scapula, sternum. Fractures of the skull, face, hands and fingers, feet and toes, tibia and fibula, ankle and patella were excluded. Statistical methods An extension of Poisson regression models [13] was used to study the association between age, the time since baseline (used as a continuous variable), serum adiponectin, other covariates on the one hand and on the other hand, the risk of fracture. In contrast to logistic regression, the Poisson regression utilises the length of each individual’s follow-up period and the hazard function is assumed to be exp (β0 +β1 ·time from baseline+β2 ·current age+β3 ·current variable of interest). The observation period of each participant was divided in intervals of 1 month. One fracture per person was counted, and time to the first fracture was counted. A similar approach was used to examine other predictors of fracture, and a final multivariable model was constructed to determine which predictors had an independent contribution to risk. The distribution of serum adiponectin was not normal and was normalised using a piecewise linear function. The association between predictive factors and risk of fracture were described as a gradient of risk (GR), expressed as the hazard ratio per 1 standard deviation change. In order to study the linearity of the association between serum adiponectin and fracture risk in more detail, a spline Poisson regression model was fitted using knots at the 10th, 50th and 90th percentiles of serum adiponectin. The splines were second order functions between the breakpoints and linear functions at the tails resulting in a smooth curve. Two hazard models were constructed to calculate the 10 year probability. These comprised the risk of osteoporotic fracture and the risk of death both using Poisson regression
Osteoporos Int (2014) 25:1831–1836
[14]. When building the hazard functions on all risk factors comprised in FRAX (age, BMI, prior fracture, family history of hip fracture, current smoking, alcohol use, glucocorticoid use, history of rheumatoid arthritis and femoral neck BMD [4]), only 634 men (64 %) had enough data to contribute. When each single risk factor used in FRAX were tested, just age, previous fracture and BMD has a statistically significant association with future osteoporotic fracture. When adjusted for age, BMD and adiponectin, previous fractures were not significantly associated with future osteoporotic fracture. When 10 year probability was calculated just using age, BMD and adiponectin, the whole cohort (n=989) was used. The hazard functions used to calculate the 10 year probability in this report comprised the risk of osteoporotic fracture and the risk of death using age, time since baseline, femoral neck BMD and adiponectin as covariates. Several additional probability models were used that varied the hazard function of fracture (Table 1). For the simplest model (model A), the hazard function for fracture depended on age, serum adiponectin and femoral neck BMD. Other models included an interaction between time since baseline and adiponectin. Model B assumed that the interaction with time would be constant for 10 years. We have previously described a nonlinear relationship between adiponectin and fracture risk, in that serum adiponectin below 18 mg/ml had no predictive value for fracture risk [8] and this phenomenon was accommodated in model C. When studying the time interaction with a linear model, the hazard ratio fell below 1 after 6.2 years. Since it is implausible that the analyte would be a protective factor for fracture, an additional model was constructed. In model D, the predictive power of adiponectin decreased with time (as in model B) but never with a hazard ratio of less than 1.00. The hazard function for mortality was not changed in the different models, using age, adiponectin and femoral neck BMD with no interactions. As an internal control, we assessed the predictive value of femoral neck BMD for fracture risk and its interaction with time. Two-sided p value was used for all analyses and p0.30), i.e. the GR was the same for all ages. Table 3 compares the 10 year fracture probabilities in men aged 75 years with a BMD T-score of −2.0 SD when adiponectin was set at 5.6 μg/ml (the 10th percentile) and when adiponectin was set at 20.1 μg/ml ( at the 90th percentile). When assuming no interaction between time since baseline and adiponectin (model A), i.e. assuming a constant gradient of risk (GR 1.40) for 10 years, the 10 year probabilities were 21 and 39 % respectively—an increase of 86 % over this range of serum adiponectin. When comparing the same probabilities for the same clinical scenario but incorporating the interaction between time since baseline and adiponectin (model B), the probability for the 90th percentile of adiponectin was only 3 % higher. The difference was slightly greater (6 %) when the GR for adiponectin was allowed to be nonlinear (model C). When GR for adiponectin was assumed not to fall below 1.00, the difference was 29 % (model D).
1834
Osteoporos Int (2014) 25:1831–1836
Table 2 Characteristics of 989 men aged 70–81 years at baseline according to fracture status during follow up Variable
Age (years) BMI (kg/m2) Femoral neck BMD (g/cm2) Adiponectin (μg/ml) Glucocorticoids (current) Current smoking Previous osteoporotic fracture Parental history of hip fracture Rheumatoid arthritis Alcohol 3 or more units per day a
No fracture during follow up (n=865)
Fracture during follow up (n=124)
Significance (two-sided p value)a
Mean±(SD) 75.3±3.2 26.2±3.5 0.86±0.14
Mean±(SD) 75.5±3.2 25.8±3.8 0.80±0.13
0.30
Poisson regression adjusted for age and time since baseline
b
percent of total number of men without a fracture (n) with a positive response
c
percent of total number of men with a fracture (n) with a positive response
The contrasting effect of a constant or decreasing predictive value for adiponectin on fracture probability is shown in Fig. 2 (model A and B). When the gradient of risk (GR) was assumed to be constant over time, fracture probability rose progressively with increasing serum values of adiponectin. When the GR was assumed to decrease with time, the 10 year probabilities were more or less constant over the range of serum adiponectin. In the case of femoral neck BMD, the hazard ratio per SD difference in BMD for fracture was 1.50 (95 % CI 1.22–1.85). There was no significant interaction with time.
Gradient of risk (HR/SD)
2.5
2.0
1.5
1.0
0.5
0.0 0
1
2
3
4
5
6
7
Time since baseline (years)
Figure 1 Gradient of risk per 1 SD (95 % confidence interval) for the association between serum adiponectin and the risk of osteoporotic fracture adjusted for age, BMD and time since baseline
Discussion In this prospective study of community-living older men, we describe a significant association between serum adiponectin and the risk of osteoporotic fracture that was independent of age and BMD. These observations confirm earlier findings of an association between serum adiponectin and the risk of fracture in men [15–17]. The novel feature of the present study was to incorporate these finding into models of fracture probability. When serum adiponectin was assumed to have a constant gradient of risk (i.e. a 40 % increase in fracture risk per 1 SD increase in serum adiponectin) over 10 years, fracture probability increased progressively with increasing values of serum adiponectin. However, our data indicate that the predictive value of serum adiponectin is not constant and wanes progressively with time. The interaction with time had a very marked effect over a 10 year time horizon. Thus, between the 10th and 90th percentile of serum adiponectin, 10 year fracture probabilities were only 3 % higher at the 90th percentile when the interaction with time was not included in the model, whereas the increase was by 86 % when the gradient of risk was assumed to be constant. The present findings have important implications for the incorporation of risk variables into fracture risk prediction models in that they illustrate the need to determine the stability of a risk indicator with time. For this reason, we routinely checked if there was a significant interaction with the variable ‘time since baseline’ and BMD, BMI and clinical risk factors in the meta-analyses used to synthesise FRAX [4]. To our knowledge, this is the first time a variable with a significant
Osteoporos Int (2014) 25:1831–1836 Table 3 10 year probability of osteoporotic fracture (%) for a man aged 75 years with a BMD Tscore of −2.0 SD using different model assumptions (see text)
Serum adiponectin is set at the 10th percentile (5.6 μg/ml) and 90th percentile (20.1 μg/ml)
1835
Model
A B C D
10 year probability of osteoporotic fracture Adiponectin (10th percentile)
Adiponectin (90th percentile)
21 32 31 28
39 33 33 36
% % % %
interaction with time since baseline has been identified and incorporated into a model of fracture probability. If the associations that we describe are reproduced in other male cohorts with a wider age range, serum levels of adiponectin may have a role in the short-term assessment of fracture risk in men but a limited or no role over a 5 to 10 year time horizon. A waning effect with time is not unexpected. Indeed, the risk of a subsequent fracture following a previous fracture appears to decrease with time [18, 19], as does the risk of fall following a fall [20]. We have previously reported a waning of effect of serum vitamin D to predict mortality [21]. Given the inevitability of death, all risk factors for death will have no predictive value with an infinite follow-up time. With shorter observation times, a waning of effect may be due to heterogeneity in the natural history of disease as reported for some of the biochemical markers of bone turnover [22, 23]. In one of these studies with a 5-year follow-up period, the ratio of carboxylated to total serum osteocalcin was a very strong risk factor for fracture over the early period of observation, but the predictive value of the marker decreased progressively with time [22, 24]. In the case of
Figure 2 10 year probability of osteoporotic fracture (%) for a man aged 75 years with a BMD T-score of −2.0
% % % %
Ratio between 90th and 10th percentile
1.86 1.03 1.06 1.29
adiponectin, heterogeneity in natural history might reflect changes in fat mass. For normally distributed variables, the long-term predictive value can be derived from the correlation coefficient between repeated samples over a defined interval of time [25]. The work presented in this article has a number of strengths and weaknesses. A strength with regard to the principal finding is that we were able to test the long-term predictive value of femoral neck BMD which we have shown in previous studies to be stable with time [26]. The hazard ratio per SD was stable with time, providing a type of internal control on which to evaluate the properties of serum adiponectin. The general limitations to the MrOs study have previously been discussed [8]. A significant limitation for the present analysis is the narrow age range of men that were recruited (70–81 years), so that our findings may not be applicable to younger ages or women. The assay that we used did not distinguish high and low molecular weight forms of adiponectin, which may have varying biological activities, but this limitation would be expected to weaken rather than strengthen any association of the analyte with fracture risk. Another limitation is that the mean follow up in this cohort was 5.3 years, and the maximum follow up was 7.4 years from which we calculated the 10 year probability of fracture. This extrapolation in time is, however, robust given that predictive value of serum adiponectin for fracture outcomes was no longer apparent after 6 years of observation. We conclude that there is a significant association between serum adiponectin levels and risk of osteoporotic fracture in elderly men and that the association decreases over time. The attenuation of the predictive value of serum adiponectin with time impairs its utility as a risk factor in the calculation of 10 year probability of fracture. This study underlines the importance of testing the longterm stability of potential risk factors in risk assessment.
Acknowledgments HJ was supported by an ESCEO-AMGEN Osteoporosis Fellowship Award. Amgen had no input into the analysis plan or in the writing of this report.
1836 Conflicts of interest None.
References 1. Consensus development conference: diagnosis, prophylaxis, and treatment of osteoporosis (1993). Am J Med 94 (6):646–650. 2. Assessment of fracture risk and its application to screening for postmenopausal osteoporosis. Report of a WHO Study Group (1994). World Health Organ Tech Rep Ser 843:1–129 3. Kanis JA on behalf of the World Health Organization Scientific Group (2008) Assessment of osteoporosis at the primary healthcare level, Technical Report. WHO Collaborating Centre, University of Sheffield, UK 4. Kanis JA, Oden A, Johnell O, Johansson H, De Laet C, Brown J, Burckhardt P, Cooper C, Christiansen C, Cummings S, Eisman JA, Fujiwara S, Gluer C, Goltzman D, Hans D, Krieg MA, La Croix A, McCloskey E, Mellstrom D, Melton LJ 3rd, Pols H, Reeve J, Sanders K, Schott AM, Silman A, Torgerson D, van Staa T, Watts NB, Yoshimura N (2007) The use of clinical risk factors enhances the performance of BMD in the prediction of hip and osteoporotic fractures in men and women. Osteoporos Int 18(8):1033–1046 5. Leslie WD, Lix LM, Johansson H, Oden A, McCloskey E, Kanis JA (2011) Spine-hip discordance and fracture risk assessment: a physician-friendly FRAX enhancement. Osteoporos Int 22(3):839– 847. doi:10.1007/s00198-010-1461-5 6. Kanis JA, Johansson H, Oden A, McCloskey EV (2011) Guidance for the adjustment of FRAX according to the dose of glucocorticoids. Osteoporos Int 22(3):809–816. doi:10.1007/s00198-010-1524-7 7. Luo XH, Guo LJ, Yuan LQ, Xie H, Zhou HD, Wu XP, Liao EY (2005) Adiponectin stimulates human osteoblasts proliferation and differentiation via the MAPK signaling pathway. Exp Cell Res 309(1):99–109. doi:S0014-4827(05)00252-1 [pii] 10.1016/j.yexcr. 2005.05.021 8. Johansson H, Oden A, Lerner UH, Jutberger H, Lorentzon M, Barrett-Connor E, Karlsson MK, Ljunggren O, Smith U, McCloskey E, Kanis JA, Ohlsson C, Mellstrom D (2012) High serum adiponectin predicts incident fractures in elderly men: osteoporotic fractures in men (MrOS) Sweden. J Bone Miner Res 27(6):1390– 1396. doi:10.1002/jbmr.1591 9. Orwoll E, Blank JB, Barrett-Connor E, Cauley J, Cummings S, Ensrud K, Lewis C, Cawthon PM, Marcus R, Marshall LM, McGowan J, Phipps K, Sherman S, Stefanick ML, Stone K (2005) Design and baseline characteristics of the osteoporotic fractures in men (MrOS) study—a large observational study of the determinants of fracture in older men. Contemp Clin Trials 26(5):569–585. doi:10. 1016/j.cct.2005.05.006 10. Mellstrom D, Vandenput L, Mallmin H, Holmberg AH, Lorentzon M, Oden A, Johansson H, Orwoll ES, Labrie F, Karlsson MK, Ljunggren O, Ohlsson C (2008) Older men with low serum estradiol and high serum SHBG have an increased risk of fractures. J Bone Miner Res 23(10):1552–1560. doi:10.1359/jbmr.080518 11. Mellstrom D, Johnell O, Ljunggren O, Eriksson AL, Lorentzon M, Mallmin H, Holmberg A, Redlund-Johnell I, Orwoll E, Ohlsson C (2006) Free testosterone is an independent predictor of BMD and prevalent fractures in elderly men: MrOS Sweden. J Bone Miner Res 21(4):529–535. doi:10.1359/jbmr.060110
Osteoporos Int (2014) 25:1831–1836 12. Kanis JA, Oden A, Johnell O, Jonsson B, de Laet C, Dawson A (2001) The burden of osteoporotic fractures: a method for setting intervention thresholds. Osteoporos Int 12(5):417–427 13. Breslow NE, Day NE (1987) Statistical methods in cancer research. IARC Sci Publ No 32 Volume II:131–135 14. Kanis JA, Johnell O, Oden A, Johansson H, McCloskey E (2008) FRAX and the assessment of fracture probability in men and women from the UK. Osteoporos Int 19(4):385–397 15. Barbour KE, Zmuda JM, Boudreau R, Strotmeyer ES, Horwitz MJ, Evans RW, Kanaya AM, Harris TB, Bauer DC, Cauley JA (2011) Adipokines and the risk of fracture in older adults. J Bone Miner Res 26(7):1568–1576. doi:10.1002/jbmr.361 16. Araneta MR, von Muhlen D, Barrett-Connor E (2009) Sex differences in the association between adiponectin and BMD, bone loss, and fractures: the Rancho Bernardo study. J Bone Miner Res 24(12): 2016–2022. doi:10.1359/jbmr.090519 17. Kanazawa I, Yamaguchi T, Yamamoto M, Yamauchi M, Yano S, Sugimoto T (2009) Relationships between serum adiponectin levels versus bone mineral density, bone metabolic markers, and vertebral fractures in type 2 diabetes mellitus. Eur J Endocrinol 160(2):265– 273. doi:10.1530/eje-08-0642 18. Johnell O, Oden A, Caulin F, Kanis JA (2001) Acute and long-term increase in fracture risk after hospitalization for vertebral fracture. Osteoporos Int 12(3):207–214 19. Johnell O, Kanis JA, Oden A, Sernbo I, Redlund-Johnell I, Petterson C, De Laet C, Jonsson B (2004) Fracture risk following an osteoporotic fracture. Osteoporos Int 15(3):175–179. doi:10.1007/s00198003-1514-0 20. Johansson H, Oden A, Karlsson MK, Rosengren B, Ljunggren O, McCloskey E, Kanis J, Ohlsson C, Mellstrom D (2013) FRAX predicts future falls in elderly men. MrOs Sweden. JBMR 28((Suppl 1):S166 21. Johansson H, Oden A, Kanis J, McCloskey E, Lorentzon M, Ljunggren O, Karlsson MK, Thorsby PM, Tivesten A, BarrettConnor E, Ohlsson C, Mellstrom D (2011) Low serum vitamin D is associated with increased mortality in elderly men: MrOS Sweden. Osteoporos Int. doi:10.1007/s00198-011-1809-5 22. Luukinen H, Kakonen SM, Pettersson K, Koski K, Laippala P, Lovgren T, Kivela SL, Vaananen HK (2000) Strong prediction of fractures among older adults by the ratio of carboxylated to total serum osteocalcin. J Bone Miner Res 15(12):2473–2478. doi:10. 1359/jbmr.2000.15.12.2473 23. Ivaska KK, Gerdhem P, Vaananen HK, Akesson K, Obrant KJ (2010) Bone turnover markers and prediction of fracture: a prospective follow-up study of 1040 elderly women for a mean of 9 years. J Bone Miner Res 25(2):393–403. doi:10.1359/jbmr.091006 24. Ivaska K, McGuigan F, Obrant K, Gerdhem P, Akesson K (2013) Bone turnover markers and prediction of fractures in a 15-year follow-up study of elderly women. JBMR 28(Suppl 1):s34 25. Kanis JA, Johnell O, Oden A, Jonsson B, De Laet C, Dawson A (2000) Prediction of fracture from low bone mineral density measurements overestimates risk. Bone 26(4):387–391. doi:S87563282(00)00238-6 [pii] 10.1016/S8756-3282(00)00238-6 26. Johnell O, Kanis JA, Oden A, Johansson H, De Laet C, Delmas P, Eisman JA, Fujiwara S, Kroger H, Mellstrom D, Meunier PJ, Melton LJ 3rd, O’Neill T, Pols H, Reeve J, Silman A, Tenenhouse A (2005) Predictive value of BMD for hip and other fractures. J Bone Miner Res 20(7):1185–1194. doi:10. 1359/JBMR.050304
