In this large, population-based study of older Americans with a hip, clinical vertebral, or wrist/forearm fracture, we found that the risk for death was substantial and often exceeded the risk for second fracture after a first hip, clinical vertebral, or wrist/forearm fracture. However, despite the substantial mortality observed in this post-fracture population, the risk for subsequent fracture was substantial, and prescription osteoporosis treatment fell within an approximate range considered cost-effective in the U.S by the NOF (24
). Our results also suggest that the increased fracture risk associated with age is largely offset by a corresponding increase in mortality among older fracture patients. This resulted in a risk of death that was highly variable, but a risk for a subsequent fracture that was much less so. Practicing clinicians may find this study useful as they consider the merits of prescribing fracture prevention therapy for their older patients with fracture.
Our results are compatible with prior literature showing a substantial mortality after hip and vertebral fractures (25
). As we found, this risk has been shown to vary over time, with a greater risk both for mortality and second fracture occurring within the first 6–24 months after the first fracture (28
). Several studies examining mortality risk over time have found the highest risk of death within six months following hip fracture (27
). It was for this reason that we conducted our sensitivity analysis restricting eligible participants to those who survived 6 months after their first fracture, since this is likely to be a more relevant decision-making time for many physicians, particularly those who may not have provided care for the patient while hospitalized for fracture.
Many reports of second fracture and mortality following an initial fracture have differed from ours with respect to both methodology and presentation of results. Importantly, a common statistical method known as censoring removes patients from further consideration at the time of death in contrast to our approach, which was to treat death as a competing risk (28
). We found that the censoring approach may substantially inflate the apparent risk for second fracture, which gives a more favorable impression of the benefits of osteoporosis treatment. The use of survival analysis, while appropriate for determining the efficacy of a therapy in older populations where there is a high rate of drop out due to death, may not be optimal for economic analyses or for clinical decision making for individual patients, where the competing risk method more accurately reflects the absolute risk of both fracture and death.
In further considering whether a patient will benefit or not from treatment, based upon the competing risk for death, we showed that the risk for death often exceeded the risk for second fracture. For that reason, whether it might be reasonable to not
provide osteoporosis medications to some individuals who experience a fragility fracture is debatable. In the extreme, we found that individuals age 85 and older with dementia who experience a hip fracture have a 7-fold greater risk for death (81%) than a subsequent fracture (12%) in the next five years. Many patients, physicians and policymakers may question whether it is worth treating such individuals for osteoporosis. A similar uncertainty exists for older individuals regarding receipt of other preventive testing and services such as cancer screening, where the number needed to screen with mammography for women ages 75–85 ranges from several hundred to greater than 1,000, depending upon life expectancy (30
). However, the NNTs we observed for almost all demographic and fracture groups are comparable to the NNTs for accepted secondary prevention strategies for aspirin after myocardial infarction (MI) (NNT = 150), statins after MI (NNT = 94), or beta blockers after MI (NN= 54) (31
). Furthermore, the NNTs that we calculated are broadly similar to those observed (NNT = 8 – 64) from the pivotal osteoporosis medication randomized clinical trials in post-menopausal women (most with a 3-year time horizon) (32
As a complementary consideration, the National Osteoporosis Foundation (NOF) has recommended prescription treatment for anyone in the U.S. with a 10-year risk of major fracture (hip, clinical vertebral, wrist, or humerus) that exceeds 20% (24
). In our analysis, a majority of subgroups had a 5-year
risk for second fracture that exceeded 20%. Therefore, treatment might be reasonably inferred to be cost-effective in light of the NOF recommendations. Moreover, the NOF recommends treatment for all persons with a hip or vertebral fracture, irrespective of BMD or 10-year risk of fracture. Despite this recommendation for prescription osteoporosis treatment, adhering to guidelines for preventive services requires time and energy on the part of the physician, particularly for complex patients with multiple comorbidities who each have individualized goals and treatment preferences (36, 37).
The strengths of our study include a large, population-based sample that is representative of the entire U.S. fee-for-service Medicare population. Because of the large sample size, we had precision to stratify by a number of important demographic factors and various common medical comorbidities. Moreover, we were able to provide data for men and African Americans, who have typically been understudied in fracture epidemiology. Despite these strengths, our study must be interpreted in the context of some limitations. Importantly, this analysis focused on risks did not take into account the substantial morbidity and impact on health-related quality of life that is typically associated with fractures (17
); considering this aspect of fracture prevention makes the need for osteoporosis treatment even more compelling. Additionally, we did not have access to medical records to confirm fractures or comorbidities. However, we used published, claims-based algorithms that have been shown to have high positive predictive values to identify incident fractures compared to a gold standard of medical record review (20
). Any mis-classification of first fractures (e.g. incidence vertebral fracture) would lead to an under-estimation of the 5 year fracture risk following a certain clinical vertebral fracture, which would strengthen our conclusions with respect to the appropriateness of pharmacologic therapy. Information regarding comorbidities was derived from administrative data, which despite their high validity (33
), lack precision on the severity of these conditions. Likewise, other important factors such as body mass index or bone mineral density are not available in this type of data. Until 2006 (the end of our study period), medications were not covered for Medicare beneficiaries, and it is possible that some individuals received treatment for osteoporosis which lowered the observed risk for subsequent fracture. Our sensitivity analysis restricted to patients who never had a DXA or an osteoporosis diagnosis (as a proxy for persons likely not treated with an osteoporosis medication) addressed this and supported the robustness of our results. We assumed that osteoporosis medications would reduce the risk of clinical fractures by 30% in our NNT calculations; the actual effectiveness of individual drugs may vary and also will be impacted by sub-optimal adherence. Finally, although every individual had five years of follow-up time after their fracture (unless they died), the time horizon relevant for treatment, and comparisons with fracture risk derived from the fracture risk assessment tool (35
), may need to extend beyond five years.
In conclusion, despite a substantial risk of death after a hip, clinical vertebral, or wrist/forearm fracture, the risk for a subsequent fracture at the same or another site was generally high enough to warrant preventive treatment with prescription osteoporosis medications. This is consistent with existing national recommendations largely motivated by cost-effectiveness considerations. These results are likely to be useful to clinicians managing patients post-fracture and provide support for treating older individuals with prescription medications, irrespective of age or comorbidities, unless there is a very high expectation of short term mortality. As with all medical services, particularly preventive services for older individuals, this decision-making process must be shared between the physician, the patient and caregivers to maximize quality of life. Mitigating fracture risk should remain an important goal of that process and may be informed by our study results.