Do C-Reactive Protein Measurements Predict Treatment Failure in Native Vertebral Osteomyelitis? A Multisite Retrospective Cohort Study

Abstract Background The management of native vertebral osteomyelitis (NVO) can be challenging due to imperfect prognostic tools to predict treatment outcome. It is common practice to utilize serial serum C-reactive protein (CRP) to help decide treatment response, although limited evidence supports this practice. We evaluated whether CRP assessment predicts treatment failure in NVO. Methods We conducted a multisite retrospective cohort study at the University of Vermont Health Network among adults with NVO managed with at least 6 weeks of antibiotics and in whom there were at least two documented CRP values at least a week apart. The primary outcome was microbiologic or clinical failure of antibiotic therapy within 6 months of the initial diagnosis. Results Among 628 patients identified by the initial screening criteria, 143 met inclusion criteria. In univariate analyses, median last CRP differed between patients with treatment success and failure (8.9 mg/dL vs 21.0 mg/dL, p < 0.01) as did median percent decrease from first to last CRP (88.9% vs 76.9% reduction, p = 0.03). However, in multivariate analyses adjusting for Staphylococcus aureus infection and epidural abscess we observed that there were only weak and non-significant associations between CRP values and treatment failure: median last CRP (OR 1.02, p = 0.06) and median percent difference CRP (0.99, p = 0.11). Conclusion Serial CRP measurements were only weakly statistical associated with treatment failure in NVO. This common clinical practice is likely of little value.