The widespread use of bisphosphonates for osteoporosis is estimated to have reduced hip fracture rates by as much as 30%1. As more patients have used these drugs for longer durations, it has also been noted that a small number of patients have developed atypical femoral fractures, which were sometimes associated with delayed healing. Negative media attention regarding these rare events has, in one Australian report, resulted in a voluntary decline in bisphosphonate use sufficient to cause a projected seventy additional hip fractures and fourteen deaths in that country alone2.
Although an absolute cause-and-effect relationship between atypical femoral fractures and bisphosphonates has not been established, the probability that these events represent rare adverse drug reactions has resulted in safety notifications and requirements for drug label warnings in the United States, Canada, and Australia3-5. In addition, the literature suggests that the incidence of atypical fractures increases with increasing duration of therapy6. It is also possible that lack of awareness and underreporting may mask the true frequency of these fractures7.
In their very ambitious article, Edwards et al. attempt not only to come closer to establishing the actual incidence of atypical femoral fractures but also to establish whether bisphosphonates are implicated in their causality as indicated by accepted statistical criteria. They combined an examination of the United States Food and Drug Administration (FDA) Adverse Event Reporting System (FAERS) database from 1996 to 2011 with an examination of the medical literature from 1990 to 2012 for cases consistent with atypical femoral fractures. They also considered historical FDA, European, Canadian, and Australian safety notifications regarding bisphosphonates.
Their results indicate that atypical femoral fractures are indeed associated with bisphosphonate use and suggest that bisphosphonates are causally involved in these fractures. In addition, the application of their data mining methods suggests that a proportionally greater role in the causality of these fractures can be attributed to bisphosphonates than to other drugs or comorbidities. The authors are even able to stratify the results according to the specific bisphosphonate by analyzing the actual instances of fracture in the published case reports, although they are careful to point out that these numbers may disproportionately relate to the number of individuals in the population taking a given drug as well as to how long each type of drug has been in general use.
The far-reaching scope of this article and its attempt to combine so many sources of information represent an admirable effort to tie many things together, although there are several problems with its format. Not the least of these is that the terms “atypical femoral fracture” and “nonhealing femoral fracture” are sometimes used interchangeably and sometimes distinguished from one another. In fact, although only nonhealing femoral fractures are mentioned in the title, the main portions of the article clearly dwell more on atypical femoral fractures than on their state of healing. In addition, as the authors point out, there are potential flaws in the studies quoted, including possible reporting biases in the original reports and selective reporting of findings. The results section also incorporates a rather lengthy review of basic-science literature summarizing the possible mechanisms that might be operant in the development of atypical femoral fractures after long-term bisphosphonate administration. The authors state that they have done this to indicate coherence of the data with the Bradford-Hill criteria, although exactly how the data cohere is somewhat confusing to this reader. Nevertheless, reading the review of research studies dealing with the various effects of bisphosphonates on bone metabolism is a useful exercise for anyone involved in the treatment of osteoporosis and its potential side effects, although it lengthens the report considerably.
From the perspective of a bone pathologist, who usually does not examine these parameters in the course of histopathological diagnosis, it is curious that although the published literature contains many reports on micro-cracks, increased collagen cross-linking, increased cortical mineralization, decreased tissue heterogeneity, and deceased remodeling associated with bisphosphonate administration, it is almost devoid of histological studies of atypical femoral fractures. Despite this, the scant case reports illustrating the pathology of these fractures have important implications, both for the pathogenesis of these fractures and for the reasons for their sluggish healing, although those implications are beyond the scope of this perspective. Finally, although suppressed remodeling may age bone matrix in viable osteons, the literature also does not dwell heavily on the role of interstitial lamellae, which comprise nearly one-third of the cortical mass, are mainly nonviable, and thus remodel more slowly than viable osteons even in normal cortex8.
In summary, although the report by Edwards et al. presents more evidence that bisphosphonates have a causal role in atypical femoral fractures than has been shown in previous meta-analyses, it remains true that the aggregate benefit of bisphosphonate usage is two orders of magnitude greater than the risk of atypical fractures. These societal benefits, of course, provide no consolation to those patients who sustain these rare events. As the authors point out, much more work is needed in this area; as my colleague Dr. Edward DiCarlo points out, perhaps when we refer to a safety signal we should call it a danger signal.