To our knowledge, there have been no studies published that have examined peptic ulcer disease and heart disease as risk factors for periprosthetic fracture in THR patients. Our observation of an association between peptic ulcer disease and periprosthetic fractures in patients with THR is interesting in light of (1) recently described associations between the medication used for treatment for peptic ulcer disease and the risk of osteoporotic fractures in general populations (Laine 2009
) and (2) our recent observation of association of peptid ulcer disease with periprosthetic fractures after primary TKR (Singh and Lewallen 2011
). The magnitude of increase in risk of periprosthetic fracture was similar for primary and revision THR, at about 50%, supporting the robustness of this association. Also, our finding of this association in three different cohorts (primary TKR, primary THR, and revision THR) indicates that this is a true association.
A recent editorial identified 3 key studies that provided evidence of increased fracture risk with the use of proton-pump inhibitors (Richards and Goltzman 2008
). Both the dose and the duration of proton pump inhibitors were associated with higher risk of hip fracture in patients in the General Practitioners’ Research Database (Yang et al. 2006
). Use of a proton-pump inhibitor for 7 or more years was associated with double the risk of osteoporotic fracture and a 5-times higher risk of hip fracture, and use for 5 or more years was associated with double the risk of hip fracture (Targownik et al. 2008
). Another study found that use of a proton-pump inhibitor in the previous year was associated with increased fracture risk (18%, 45%, and 60% higher odds for overall fracture, hip fracture, and spine fracture risk, respectively) (Vestergaard et al. 2006
). It has been hypothesized in previous studies that this effect may be mediated by lower calcium absorption due to hypochlorhydria related to proton-pump inhibitors. On the other hand, use of histamine-2 receptor antagonists was associated with increased fracture risk in one study (Yang et al. 2006
), but reduced fracture risk in another study (Vestergaard et al. 2006
). Due to the lack of availability of data on use of medication prior to and after THR in the Total Joint Registry, we were unable to test the hypotheses of whether the observed association was due to the disease or whether it was due to one of its treatments (such as use of proton-pump inhibitor).
The association between heart disease and periprosthetic fractures in patients with primary THR adds to our current knowledge. Recent publications have reported significant associations between cardiovascular disease and lower bone mineral density in the NHANES sample (Broussard and Magnus 2008
) and between beta-blockers and fragility fractures in postmenopausal women (Sosa et al. 2011
). Other authors have reported an association between postoperative use of statin and lower risk of revision (Thillemann et al. 2010
), and between postoperative loop diuretic use and higher risk of revision due to deep infection and periprosthetic fracture (Thillemann et al. 2009
), which raises questions related to the underlying mechanisms of the association between heart disease and periprosthetic fractures in the present study. Are these associations in this study related to the medications used for treatment of peptic ulcer disease and heart conditions? If so, which ones? Is the risk related to disease or to its severity? One recent genetic study in twins provided clues to the link between cardiovascular disease and hip fracture risk (Sennerby et al. 2009
). People with heart disease had a higher risk of subsequent hip fracture. Increased risks in co-twins without an index diagnosis suggested that genetic factors may have a role in the association between cardiovascular disease and osteoporotic fractures. Our findings are in agreement with key findings from these earlier studies in larger cohorts that were not limited to arthroplasty.
The present study also provides information on the frequency of comorbidities in patients undergoing primary or revision THR. Patients had a mean of 1 comorbid condition in either cohort. This is not surprising, considering that the mean age of both cohorts was about 65 years.
We also observed that point estimates for several other comorbidities were similar to those for peptic ulcer disease, but not statistically significantly. For example, peripheral vascular disease and other diseases in primary THR cohort did not reach statistically significant levels of association with the fracture risk. Similarly, peripheral vascular disease, heart disease, and cancer in revision THR cohort did not reach statistically significance.
The study has some limitations. As with other registry studies, some patients may have been lost to follow-up—despite the close follow-up in our joint registry (visit to clinic, mailed questionnaire, telephone follow-up)—which would be expected to lead to underestimation of fractures. Thus, actual estimates of periprosthetic fractures may be higher. In addition, patients with higher comorbidity are also more likely to die during the follow-up, which reduces their risk of having a fracture, with death as the competing risk.
Censoring of observations at the time of death or fracture ensured that time for which a patient was observed in this study was only for those at risk of fracture (i.e. living patients at risk). Since the cohort grew over several decades and follow-up information was obtained through multiple sources, not just clinic visits, we were unable to calculate actual loss to follow-up. The Mayo Clinic provides both primary and specialty care to the local population and tertiary specialty care to those referred for THR; thus, generalization of these findings to all settings may not be possible. Residual confounding is possibly related to study design (non-randomized), although we attempted to control for several important variables. The strengths of our study include the large sample size, the use of prospective data from an institutional total joint registry, the ability to control for several important variables (BMI, ASA class, implant fixation), and the use of multivariable-adjusted estimates. Our estimates were quite robust, in that the associations were similar in the 2 cohorts (primary THR and revision THR) and the estimates for the significant associations were minimally attenuated after multivariable adjustment.