Plain X-ray, triple phase technetium99 polyphosphonate bone scan (BS), magnetic resonance imaging (MRI) and/or computed tomography (CT) have been used to evaluate stress fracture. Multiple method imaging is expensive and may expose patients to unnecessary radiation, to minimal clinical advantage. There is a need to establish guidelines for the most appropriate imaging modality to diagnose and prognosticate stress fracture injury. PURPOSE To identify the imaging modality that best predicts tibial stress fracture severity and recovery time. METHODS A convenience sample of 50 patients with tibial stress fracture was recruited. Subjects were examined acutely on the same or subsequent days via standard AP and lateral radiographs, BS, MRI and CT. A repeat MRI was obtained within 3 days of healing (asymptomatic with hopping). Examinations were graded on 0-4 modality-specific severity scales by 4 reviewers (2 stress fracture-specialist musculoskeletal radiologists, 1 non-specialist musculoskeletal radiologist, and 1 stress fracture-specialist orthopaedic surgeon). The relationships of image grade to clinical severity and time to healing were examined via correlation analyses. The predictive abilities of each imaging modality were examined via multiple regression analyses. The reliability of the grading system was evaluated via repeated measures ANOVA and Cronbach's alpha coefficient. RESULTS Despite moderate to good inter-grader correlations for all imaging modalities (range: r = 0.326 - 0.862, p < 0.05, two-tailed) and robust Cronbach's alpha reliability coefficients for the grading system (range: 0.591 - 0.895), no imaging modality consistently predicted tibial stress fracture clinical injury severity. When only two graders were included in the regression model, however, BS grade predicted clinical severity (R2 = 0.138, p = 0.03). BS grade also predicted time to healing (R2 = 0.241, p =.007), as did a combined radiology score of all imaging modalities (R2 = 0.243, p = 0.03) including all graders. CONCLUSION Tibial stress fracture clinical severity bears poor relationship to the appearance of severity on diagnostic imaging. BS was the most effective imaging modality for tibial stress fracture prognostication. Supported by US Army MRMC, DAMD17-98-1-8519, and BIOLECTRON, Inc. now EBI, Parsippany, NJ.