Predictive Performance of MRI for Antibiotic Treatment Failure of Pyogenic Vertebral Osteomyelitis: A Validation Study

Sugihiro Hamaguchi; Sei Takahashi; Yuji Endo; Yohei Nakamoto; Tetsuro Aita; Toru Naganuma; Hiroaki Nakagawa; Toshihiko Takada

doi:10.7759/cureus.47933

Predictive Performance of MRI for Antibiotic Treatment Failure of Pyogenic Vertebral Osteomyelitis: A Validation Study

Cureus. 2023 Oct 29;15(10):e47933. doi: 10.7759/cureus.47933. eCollection 2023 Oct.

Authors

Sugihiro Hamaguchi^{1

2}, Sei Takahashi^{1

3}, Yuji Endo⁴, Yohei Nakamoto¹, Tetsuro Aita¹, Toru Naganuma^{1

3}, Hiroaki Nakagawa^{1

2}, Toshihiko Takada⁵

Affiliations

¹ Department of General Internal Medicine, Fukushima Medical University, Fukushima, JPN.
² Center for Innovative Research for Communities and Clinical Excellence (CiRC2LE), Fukushima Medical University, Fukushima, JPN.
³ Futaba Emergency and General Medicine Support Center, Fukushima Medical University, Fukushima, JPN.
⁴ Department of Orthopedic Surgery, Fukushima Medical University, Fukushima, JPN.
⁵ Department of General Medicine, Shirakawa Satellite for Teaching and Research (STAR), Fukushima Medical University, Shirakawa, JPN.

Abstract

Introduction: Intravenous antibiotics are the primary treatment of choice for pyogenic vertebral osteomyelitis (PVO). Surgical intervention is required when the initial antibiotic treatment fails but is often difficult to perform, especially in older adults with multiple comorbidities, because of the reduced physical activity. The size of the infection signal in the spinal bone on magnetic resonance imaging (MRI) at the time of diagnosis was reported to have a high predictive accuracy for antibiotic treatment failure. However, the sample size was too small for this result to be adopted in clinical practice. Thus, we conducted a validation study of the previous research using a larger sample size.

Methods: We conducted a retrospective review of electronic medical records of patients admitted to the orthopedic department of a university hospital with a diagnosis of PVO between 2006 and 2021, and consecutively included patients without planned PVO surgery on admission and with a sagittal view of T1-weighted spinal MRI at the time of diagnosis. The index test was the percentage involvement of the affected areas in one motion segment on sagittal MRI. We also evaluated other MRI findings, such as bone destruction, segmental instability, epidural abscesses, and multiple sites for their predictive accuracy for antibiotic treatment failure.

Results: A total of 82 participants were eligible for the analysis. The presence of ≥90% affected area of one motion segment had a sensitivity of 16.7% and a specificity of 70.3% for future antibiotic treatment failure, resulting in poor predictive performance, with positive (LR+) and negative likelihood ratios of 0.56 and 1.19, respectively. The area under the receiver operating characteristic curve for a 10% increase in the affected area was 0.48. Among the other MRI findings, the presence of bone destruction had a significantly higher predictive accuracy (LR+ 3.11, 95% confidence interval 1.30-7.42).

Conclusion: An infection signal ≥90% on a T1-weighted MRI of one spinal motion segment did not show sufficient predictive performance for antibiotic treatment failure. Spinal bone destruction had a mild-to-moderate predictive accuracy.

Keywords: antibiotic treatment failure; bone destruction; magnetic resonance imaging; predictive performance; pyogenic vertebral osteomyelitis.