We thank O’Reilly et al for their interest in our article.1 To answer the specific questions posed in their letter, we used published data from the Medicare dataset on mortality risks for patients with vertebral compression fracture (VCF) stratified by kyphoplasty (BKP), vertebroplasty (VP), and nonsurgical management (NSM).2 As described in that study, all outcomes, including death, were measured from the time of diagnosis of the incident vertebral fracture. Hence, the “clock started” at the same time for all patients. O’Reilly et al also suggested that we should have generated survival curves and hazard ratios ourselves rather than relying on data from another article. That would certainly have been the case had we relied on an article authored by other investigators. However, because our group analyzed the data and authored the original article, we, in fact, were able to calculate the summary survival curves and hazard ratios ourselves. Moreover, in that article, we adjusted the data for a multitude of variables, including fracture location. The authors of the letter highlighted a potential bias favoring survival for patients in the augmentation procedure group. Belying that is the fact that previous sensitivity analyses of the Medicare VCF population have demonstrated improved survival risks for the augmentation over the NSM group that were still observed even when comparing all patients who survived at 1 year.3,4
O’Reilly et al further suggested that the Current Procedural Terminology (CPT) codes for VP 22520–22522 were missing from the analysis and that CPT code 22289 should not have been used for vertebral augmentation. As described by Ong et al,2 the VP codes were used to identify these patients. CPT 22289 was also used to identify BKP procedures before 2006.3,4 This was the code that insurance carriers had required for BKP reimbursement during the period in question. This would have also only applied to 1 year (2005) of 10 years of data (2005–2014) from which the survival curves and hazard ratios were determined.2 O’Reilly et al also queried about what spine fusion codes were used. So as not to become tedious, we refer the authors of the letter back to the original study details.2
Without question, we acknowledge the limitations of this analysis of nonrandomized observational data and the biases present that our group previously attempted to adjust for by using propensity adjustment strategies. Indeed, we point out these limitations directly in the discussion of the underlying article, stating, “Using large claims-based datasets inherently equates to a heterogeneous population being analyzed retrospectively.” However, it is surprising to us that the authors of the letter do not recognize that the mortality benefit is biologically plausible. First, NSM carries its own risks, and kyphotic posture is associated with an elevated risk of mortality.5 The immobility caused by vertebral fractures is also very well-known to lead to increased mortality that rivals or exceeds that of hip fractures.6 Moreover, opioid treatments for NSM of compression fracture pain were widespread from 2005 to 2014, and these medications are themselves associated with disability and increased risk of death.7
To explore this area further, our group also performed a systematic review and meta-analysis on the mortality outcomes of patients with osteoporotic vertebral fractures treated with vertebral augmentation compared with those treated with NSM that has been recently published.8 The pooled hazard ratio (HR) across 7 studies was 0.78 (95% CI, 0.66–0.92; P = .003) in favor of augmentation. Although heterogeneity was high with an I2 of 68%, the result remained robust with sensitivity analysis. Moreover, the lower hazard for mortality has also been independently reported in large Taiwanese (n = 7097; HR, 0.72; 95% CI, 0.56–0.92; P = .008) and German studies (n = 3607; HR, 0.57; 95% CI, 0.48–0.70; P < .001).9,10
Although we believe the mortality benefit of augmentation is supported by the available evidence, biologically plausible, additional high-quality evidence is required. We look forward to better designed and adequately powered randomized controlled trials of vertebral augmentation and additional meta-analyses of individual patient data from randomized trials to further examine clinically relevant outcomes, including mortality.
Indicates open access to non-subscribers at www.ajnr.org
References
- © 2020 by American Journal of Neuroradiology