Background Stress fractures (SFX) are classified by localisation (low- and high-risk) and severity, which can be determined by imaging modalities (low- and high-grade).

Objective The objective was to determine whether a combined analysis of risk-localisation and image-grading in SFX allows a more accurate prediction of return-to-sports-time (RTST).

Design Retrospective study. In a blinded read three independent specialists rated imaging-data (MRI and bone scintigraphy) as low- or high-grade SFX. SFX detection, risk-classification and RTST were determined by an interdisciplinary truth-panel providing a reference standard. Two-sided Wilcoxon's rank sum test and Kruskal-Wallis test were used for group comparisons.

Setting Patients from an athlete boarding school, an Olympic training centre and from professional teams were included.

Patients 50 consecutive athletes (male, n=20; female, n=30; mean age, 23.6 years) with SFX receiving imaging were included (track, n=18; long distance running, n=16; handball, n=11; other, n=5). In all athletes follow-up data was complete until full recovery.

Interventions All patients received standardised treatment with adaptation to sport and injury-site with special focus on safe return-to-full activity.

Main outcome measurements Estimation of RTST depending on risk-classification and image-based grading.

Results In our study 21/50 SFX at a high-risk localisation had a mean RTST of 135 days (d) compared to 119 d for low-risk sites (p=0.18). RTST was significantly longer (p=0.01) in imaging-based high-grade lesions (mean=143 d) than in low-grade SFX (mean=96 d). Analysis of only high-risk SFX showed no difference in RTST (p=0.58) for high- and low-grade SFX (mean, 131 d vs 140 d). In contrast the difference was significant for low-risk SFX (p=0.004; mean, low-grade=61 d vs high-grade=153 d).

Conclusion In low-risk SFX the significant difference in RTST between low- and high-grade lesions allow a more accurate estimation of the healing time needed. As a consequence, both risk localisation and image grading of SFX have to be considered for the prediction of RTST.