In this 12-year study of over 90 million hospital discharge records, we took advantage of the large sample size and nationwide sampling of NIS to estimate rare incidences of subtrochanteric fragility fractures in the elderly population. In the context of a decline in typical hip fractures in both women and men, we found increasing trends in incidence rate and percentage of subtrochanteric hip fractures only among women. The incidence rates of subtrochanteric fractures in women were twice as high as those in men, although the proportions were similar and low (2% to 4%) in the beginning. The secular trend of increases in both rate and proportion were significant in women but not in men, suggesting that elderly women bear increased burdens of such rare fractures not only in absolute terms but also relative to overall hip fractures. Furthermore, these trends since 1999 were preceded by an increase in prevalence of bisphosphonate use among women since 1996. On the other hand, national uses of other fracture-related medications were similar between men and women.
These findings suggest that the case reports of more than 54 atypical hip fractures in the United States since 2005 may be not anecdotal but rather reflective of systemic changes in the nature of hip fractures. This is in agreement with a recent case-control study confirming the association between atypical hip fractures and bisphosphonate use.(30
) However, this is contradictory to recent reports from post-hoc analyses of clinical trial data(19
) and a cohort study.(20
) In both studies, no significant differences were found between typical and subtrochanteric or diaphyseal hip fractures.
Interestingly, in the study by Black and colleagues, point estimates for the risk ratio of subtrochanteric and diaphyseal fractures from three trials were all above 1.00 in the treatment group and higher in trials with longer exposure to bisphosphonates or with more potent agents such as zoledronic acid. It also should be mentioned that in the only trial where the intertrochanteric fractures were significantly reduced (HORIZON trial), the risk ratio of subtrochanteric and diaphyseal femur fractures was highest in point estimates. Furthermore, while seven subtrochanteric or diaphyseal femur fractures in treatment all were found to result from minimal trauma (a fall from standing height or less), all three in the placebo group were caused by higher impact such as falls from a stair, step, or curb. Although it is hard to discuss the significance of these findings, randomized trials should give some indications as to the tendency of risk changes.
In another study with population-based registry, fracture risks for the hip and subtrochanteric or diaphyseal femur were more common in bisphosphonate users than in nonusers.(20
) This may be an example of confounding by indication often seen in observational studies,(31
) given the tendency to use bisphosphonates in patients with osteoporosis and/or higher risk of hip fractures. In addition, this study did not exclude fractures owing to high-energy trauma. And finally, despite from a national register, fewer than 200 people were determined to be highly adherent with more than 6 years of bisphosphonate treatment.
A recent study of National Hospital Discharge Survey (NHDS) indicates that although there was a decline in hip fracture rate, no increase in closed subtrochanteric fractures was found.(21
) This is not unexpected because estimates for less common procedures and diagnoses tend to be less reliable if the sample size is smaller than a million.(32
) NIS, on the other hand, has been used for studying other rare medical events such as pneumonia hospitalizations complicated by empyema in children with an incidence rate of less than 10 per 100,000.(33
Our estimates of rates and declines for typical hip fractures are consistent with earlier studies using NIS(34
) and Medicare data.(6
) In addition to the diagnosis code, we also incorporated procedure codes to ascertain the incidence of hip fractures and excluded revision surgeries and follow-up visits. Our estimates of subtrochanteric hip fractures were comparable with those from Black and colleagues(19
) and Nieves and colleagues,(21
) although they included femur shaft fractures and had patients from 50 years and older. The reason we did not include the shaft fractures was that we intended to study the proportion of subtrochanteric hip fracture among all hip fractures, and femur shaft fractures represent a different type in that regard. On including the shaft fractures, increasing trends of subtrochanteric and femoral shaft fractures were still observed (unpublished data).
While our results are consistent with the protective effect of bisphosphonates against typical hip fracture, we also showed a temporal correlation between national use of bisphosphonates and delayed increase in subtrochanteric fractures. It remains to be seen whether the increasing incidence should occur specifically among those with relevant medication usage. We noted that women were the predominant users of bisphosphonates and experienced the vast majority of subtrochanteric fractures. If the rise in subtrochanteric fractures were due to an environmental effect, a shift in coding practices, or rising general awareness, it would be seen in men as well.
Our study has strengths related to the national representation, high sensitivity, and contextual analysis for different outcomes. However, there are limitations associated with this ecologic investigation. First, MEPS and NIS may not be representative of the same segment of the US population such that comparing trends in these two data sets has limitations. Second, data were not available or shown regarding other trends in lifestyle factors that also might explain fracture trends, such as physical activity, use of dietary supplements, smoking, and use of other medications such as selective estrogen receptor modulator (SERMs), parathyroid hormone, and estrogen. And finally, the administrative data we used are limited in that closed subtrochanteric fracture were unspecific for those “atypical” fractures without knowing if such characteristic radiographic findings were present as reported in clinical case reports.(14
) We may have overestimated the extent of such fractures by using ICD-9 codes rather than radiographs to identify fractures. However, since subtrochanteric hip fracture is a very rare outcome, claims data may be better suited to address the problem with sufficient power. Our preliminary findings call for a large cohort study with longitudinal data about health outcomes and long-term medication history.
Our data may define the extent of divergence between these two trends. National estimates of hospitalizations for hip fractures declined by more than 30,000 between 1996 and 2007 despite aging of the population. In contrast, there were approximately 2,500 more subtrochanteric hip fractures in 2007 than in 1996. It is obvious that the extent of increase in subtrochanteric hip fractures remains very small compared with the substantial declines in typical or overall hip fractures. Using age-adjusted rates, we estimate that for every 323 reduction in typical femoral neck or intertrochanteric fractures, there was concurrent increase of 3 subtrochanteric fractures. Even after we included closed femoral shaft fractures, which have been implicated following long-term bisphosphonate therapy,(36
) the increase in these hip fractures is still around 10 per 300 typical hip fractures reduced (unpublished data). This estimate was similar to what Black and colleagues reported based on the minimum value of hypothetical relative risk of subtrochanteric or diaphyseal femur fractures in their clinical trial data.
We also confirmed that in the beginning of our study period, subtrochanteric fractures constitute only 2% to 4% of all hip fractures,(37
) a proportion of which increased significantly but modestly among women. Thus, even though increasing trends of subtrochanteric hip fractures were significant, the extent of the problem was minor, and the potential for future increases may be limited. However, such incidences should be monitored on the population level and prevented, if possible, during individual care.
In the United States, where burdens of overall hip fractures are falling, annual national estimates and age-adjusted rates of subtrochanteric hip fractures, albeit very low, are on the rise among women aged 65 years and older. Although these increasing trends were temporally correlated with prevalence in bisphosphonate use among postmenopausal women, this study only fulfills a temporality requirement for a possible causal relationship between bisphosphonate use and changes in the patterns and distribution of hip fractures.