The literature contains a lot of conflicting findings regarding diagnostic accuracy and protocols for imaging suspected lower extremity stress fractures. The lack of systematic reviews on this topic has limited the development of evidence-based recommendations for appropriate imaging protocols in cases of suspected lower extremity stress fractures. This study sought to determine the diagnostic accuracy statistics of imaging modalities used to diagnose lower extremity stress fractures and to synthesize evidence-based recommendations for clinical practice. Reported sensitivity and specificity (95% CI) were as follows: For conventional radiography, sensitivity ranged from 12% (0%-29%) to 56% (39%-72%) and specificity ranged from 88% (55%-100%) to 96% (87%-100%). For nuclear scintigraphy (NS), sensitivity ranged from 50% (23%-77%) to 97% (90%-100%) and specificity from 33% (12%-53%) to 98% (93%-100%). For magnetic resonance imaging (MRI), sensitivity ranged from 68% (45%-90%) to 99% (95%-100%) and specificity from 4% (0%-11%) to 97% (88%-100%). For computed tomography, sensitivity ranged from 32% (8%-57%) to 38% (16%-59%) and specificity from 88% (55%-100%) to 98% (91%-100%). For ultrasound, sensitivity ranged from 43% (26%-61%) to 99% (95%-100%) and specificity from 13% (0%-45%) to 79% (61%-96%).
MRI was identified as the most sensitive and specific imaging test for the diagnosis of stress fractures of the lower extremity. When MRI is available, NS is not recommended because of its low specificity, high dosage of ionizing radiation, and other limitations. It is probably that conventional radiographs will result in false negatives upon initial presentation, particularly in the early stages of stress fracture, and in some cases may not reveal an existing stress fracture at any time. A diagnostic imaging algorithm was developed with specific recommendations for cost-efficient imaging of low-risk and high-risk suspected stress fractures.