Bone Parameters and Risk of Hip and Femur Fractures in Patients on Hemodialysis

Discussion

We detected several findings with respect to hip fractures in this large national period prevalent cohort of patients on HD. First, we found that the rate of hip fractures did not change significantly in patients on HD from 2000 to 2013. Second, there was a secular trend toward increasing iPTH from 2000 to 2005 and a progressive trend from 2000 to 2013 toward lower serum phosphorus. Third, lower time–averaged concentrations of iPTH (particularly <273 pg/ml) and lower levels of serum calcium (<8.7 mg/dl) were associated with higher risk for hip fractures. Fourth, among elderly white patients, there were no clear relationships between the risk factors studied and risk for hip fractures, with the possible exception of low iPTH.

The finding of stability in the incidence of hip fractures between 2000 and 2013 is important in that it updates published data that found a substantial increase in the rate of fractures in patients on HD from the 1990s to approximately the mid-2000s (2,6,7). Although it remains unclear why the incidence of hip fractures had previously increased, it is reassuring to note that the rate has stabilized. Moreover, because the proportion of patients >80 years of age has increased and because age is a determinant of hip fracture risk, it is possible that the age-adjusted incidence of hip fracture might have decreased over time.

The trend toward higher serum concentrations of iPTH from 2000 to approximately 2005 seems consistent with a decreasing focus during that time period on intense iPTH suppression (11). This trend became more standardized with the 2009 Kidney Disease Improving Global Outcomes guidelines (12), which liberalized iPTH targets. The previous Kidney Disease Outcomes Quality Initiative PTH target of 150–300 ng/ml (13) was raised to two to nine times the upper limit of the PTH assay (12). It should be noted that PTH assays have changed over time, and at least some of the secular change could be explained by these changes. The decreasing trend in serum phosphorus during the period of observation may reflect a greater level of attention that had been placed on phosphate management, perhaps reflecting studies indicating harmful effects associated with higher phosphorus levels (14–16).

We found that patients who had time-averaged iPTH <273 pg/ml had a greater risk for hip fractures, even after adjustment for clinical factors, medications, and laboratory tests. This relationship might help explain the longitudinal trend in hip fractures, which increased substantially from 1992 to the early 2000s (2,6,7) and then, stabilized up to 2013, which we showed in this study. We found the mean iPTH to be 197.5 pg/ml in 2000. To the extent that 2000 was representative of iPTH values over the preceding decade, the relative suppression of iPTH during that period could explain the increasing rates of fractures. From the relatively low mean iPTH value in 2000, we found that mean iPTH values increased to approximately 300 pg/ml by 2004 and that, in parallel, the rate of hip fractures stabilized during this time period. The stabilization in fracture rate might reflect the relatively large increase in mean iPTH values to above the higher risk threshold of 273 pg/ml.

One possible explanation for greater fracture risk with lower levels of iPTH could be low–turnover bone disease, which may be found with lower iPTH values (17) and might predispose to bone fragility. Although studies relating PTH values and bone morphology have yielded inconsistent results, most studies have found a fairly high positive predictive value for lower PTH levels and low–bone turnover morphology; Barreto et al. (17) found that iPTH<150 pg/ml had an 83% positive predictive value for predicting low bone turnover. The relationship between bone histomorphology and fracture risk is less clear, because the literature is generally composed of studies too small in sample size to detect sufficient events of hip fracture. A recent study evaluated bone quality as assessed by material and mechanical properties. The primary finding was that “bone with low turnover had microstructural abnormalities such as lower cancellous bone volume and reduced trabecular thickness” (18). We did not have access to tests, such as bone alkaline phosphatase, vitamin D levels, or fibroblast growth factor-23, that might have provided information on bone health.

It might be expected that lower time–averaged iPTH values would be related to treatment with active vitamin D (or more recently, cinacalcet) and that, therefore, the association of lower iPTH with higher fracture risk should extend to higher doses of active vitamin D being a risk factor as well. However, accounting for intravenous vitamin D, we did not find this to be true. After adjustment, there was no relationship between active intravenous vitamin D dose and fracture risk. This divergence is a surprising finding that might suggest a hypothesis that patients with spontaneously low to normal iPTH levels (rather than those suppressed with treatment) may be a subset with greater bone fragility. One reason for lower unsuppressed iPTH concentrations could be reduced sensitivity to lower serum calcium levels. Our finding that lower serum calcium, like lower iPTH, was an independent risk factor for hip fractures may lend some support to this hypothesis. Indeed, in Figure 7, we show a general trend toward greater fracture risk when lower iPTH, vitamin D, and serum calcium are combined. However, absence of information regarding the use of oral calcimimetics or postsurgical parathyroidectomy leaves the potential for iatrogenic low calcium and low iPTH to be contributory.

Our findings with respect to lower iPTH being associated with hip fracture risk are externally consistent with a study by Coco and Rush (19), which found that lower PTH concentrations (<195 pg/ml) predicted hip fracture risk in patients on HD. In contrast, other studies failed to find PTH to be a predictor of fractures (20–22). Previous studies of bone parameters and fracture risk generally studied periods before 2000, were usually on the basis of single laboratory values, and were of much smaller sample sizes. Our study may have benefitted from a more contemporary view, a 1-year time-averaged view on risk exposure given the known variability of PTH (23), and a large sample size, although we also recognize limitations of our study below.

In previous studies, we and others found a secular trend toward increased hip fractures in patients on dialysis from 1992 to approximately 2005, which with additional analyses, occurred in patients of white race or those >65 years of age (6). Furthermore, there seemed to be less correlation between histomorphic bone turnover and iPTH in black patients (24). Taken together, this knowledge led to a hypothesis that, in this subgroup of patients, who often are of a phenotype representative of osteoporosis or bone fragility, oversuppression of iPTH could be particularly harmful (6). However, we found no consistent significant relationship between lower iPTH and higher hip fracture risk in this subgroup.

Our study has certain strengths, including large sample size, the contemporary dataset, and the use of time–averaged risk exposures. However, there are certain limitations as well. First, the observational design prevents understanding of causality in relationships. Second, although the dataset used was a clinical repository, we cannot assure absolute capture of all fracture events. Third, although the cohort seems to be representative of typical dialysis populations, the results are not necessarily generalizable to all United States dialysis populations. Fourth, for the year 2000 cohort, duration of dialysis was unknown in 53.1% of patients (Table 2). Fifth, we did not have information on parathyroidectomy.

A more important limitation is that we lack information on oral medications for bone and mineral metabolism management, particularly cinacalcet. This agent, US Food and Drug Administration approved in 2004, is used in treatment in approximately 20% of patients on dialysis (25). Because it is an effective treatment for hyperparathyroidism and has been shown to reduce fracture risk (26), it could be an important exposure variable. Future studies should seek to incorporate information on cinacalcet treatment where information on treatment is available.

In conclusion, there seemed to be a stabilization of hip fracture rates at the end of the first decade of this millennium. Furthermore, we affirm an association between lower levels of iPTH and serum calcium and greater risk for hip fractures among patients on HD. We suggest a hypothesis that patients with lower levels of iPTH in the absence of high–intravenous vitamin D doses may be at higher risk for hip fracture; this hypothesis requires additional evaluation, particularly with differentiation for patients with spontaneous low iPTH levels versus iatrogenic nonvitamin D parathyroid suppression therapy (e.g., calcimimetics or postparathyroidectomy). These findings support evaluating long-term safety of therapeutic strategies for hyperparathyroidism in patients with ESRD.