Henry SG Harrison, Kerensa M Beekman, Rimke S Lagrand, Lawrence A Lavery, Mario Maas, Edgar JG Peters.

ABSTRACT

Background:

Plain radiography is first-line imaging for suspected osteomyelitis in persons with diabetes (DFO); however, its early sensitivity is limited. Serial radiography may improve detection and inform prognosis, yet its utility remains uncertain.

Methods:

We performed a retrospective multi-reader diagnostic accuracy study in the BeBoP cohort of persons with diabetes and clinically suspected DFO who underwent percutaneous bone biopsy. Cases with baseline (T0) and ~6-week follow-up (T1) foot radiographs were included. Four readers (two radiologists and two non-radiologists) rated DFO probability on a 5-point scale at T0, T1, and serial (T0+T1). Histopathology and/or culture of percutaneous bone biopsy defined DFO. Discrimination across conditions was evaluated using ROC analysis, DeLong’s test, and mixed-effects models.

Results:

Thirty-four cases were included; 27/34 (79.4%) had DFO. Multi-reader AUCs were 0.52 (T0), 0.64 (T1), and 0.65 (serial); with no significant differences between AUC for serial versus T1. Mixed-effects models showed higher odds of correct classification for serial versus T0 conditions. Radiologists outperformed non-radiologists at T0, but not serial readings; interrater agreement was moderate (κ = 0.44). Reader experience was not associated with AUC.

Conclusions:

Serial radiography modestly improved accuracy over baseline but did not perform better than single follow-up imaging, and reader experience did not influence performance. Serial radiography should be considered supportive in the multidisciplinary evaluation of suspected DFO.a standalone test for diabetic foot osteomyelitis.