Value of CT to detect radiographically occult injuries of the proximal femur in elderly patients after low-energy trauma: determination of non-inferiority margins of CT in comparison with MRI

Purpose To determine the margins of non-inferiority of the sensitivity of CT and the sample size needed to test the non-inferiority of CT in comparison with MRI. Materials and methods During a 2-year period, elderly patients with suspected radiographically occult post-traumatic bone injuries were investigated by CT and MRI in two institutions. Four radiologists analyzed separately the CT and MRI examinations to detect post-traumatic femoral injuries. Their sensitivities at CT (SeCT) and MRI (SeMRI) were calculated with the reference being a best valuable comparator (consensus reading of the MRI and clinical follow-up). ROC analysis followed by an exact test (Newcombe's approach) was performed to assess the 95% confidence interval (CI) for the difference SeCT-SeMRI for each reader. A sample size calculation was performed based on our observed results by using a one-sided McNemar's test. Results Twenty-nine out of 102 study participants had a post-traumatic femoral injury. SeCT ranged between 83 and 93% and SeMRI ranged between 97 and 100%. The 95% CIs for (SeCT-SeMRI) were [- 5.3%, + 0.8%], (p(R1) = 0.1250), [- 4.5%; + 1.2%] (p(R2) = 0.2188), [- 3.4%; + 1.1%] (p(R3) = 0.2500) to [- 3.8%; + 1.6%] (p(R4) = 0.3750) according to readers, with a lowest limit for 95% CIs superior to a non-inferiority margin of (- 6%) for all readers. A population of 440 patients should be analyzed to test the non-inferiority of CT in comparison with MRI. Conclusion CT and MRI are sensitive for the detection of radiographically occult femoral fractures in elderly patients after low-energy trauma. The choice between both these modalities is a compromise between the most available and the most sensitive technique.