MRI vs. Plain X-Ray Imaging for the Evaluation of Chronic Lower Back Pain
Benefits in NNT
None were helped (improved diagnosis leading to better therapies and less disability)
Harms in NNT
None were harmed (MRI scan)
View As:
Efficacy Endpoints
Back-related disability measured by the modified Roland questionnaire, Medical Outcomes Study 36-item Short From Health Survey (SF-36), pain preference scores, satisfaction, costHarm Endpoints
NoneNarrative
380 patients with low back pain for > 1 year were randomized to receive either lumbar spine radiographs versus a lumbar spine MRI. The patient groups appear to be well randomized. The primary hypothesis was that a validated back pain disability scale would demonstrate better scores in the MRI patients because of accurate diagnoses and better management.Scores were statistically unchanged at 12 months. Both groups improved (had decreases in scores) that were most substantial in the first 3 months. There were many more diagnoses in the MRI group, however none were consequential (i.e. no diagnosis of metastasis or abscess or spinal cord compression). There were no adverse events in either group. 10 patients in the MRI group versus 4 patients in the radiograph group underwent lumbar spine operations (p = 0.09).
There were no differences between the groups in “pain bothersomeness”, pain frequency, SF-36 score (functional health and well-being), and physical functioning. The MRI group had a mean cost of $2380 versus $2059 for the radiograph strategy, which includes consults, physical therapy, etc. Patients were equally satisfied in both groups; however they and their physicians were slightly more reassured if an MRI was performed.
This randomized controlled trial suggests that patients with low back pain without pathologic features are not helped by having an MRI. This strategy results in increased cost, possibly more procedures, and no change in patient long-term pain, disability, or functional status. The phenomenon of reassurance via diagnostic testing seems apparent in these data.