Object. In a randomized controlled trial comparing surgery and prolonged conservative treatment for sciatica of 6-12 weeks' duration, more than one-third of patients assigned to conservative treatment underwent surgery. The objective of the present study was to evaluate whether MRI at baseline could have predicted this delayed surgery. Methods. Independently evaluated qualitative and quantitative MRI findings were compared between those patients who did and those who did not undergo surgery during follow-up in the conservative care group. In addition, area under the receiver operating characteristic (ROC) curve analysis was used to assess how well MRI parameters discriminated between those who did and those who did not undergo delayed surgery (0.5-0.7 poor discrimination, ≥ 0.7 acceptable discrimination). Results. Of 142 patients assigned to receive prolonged conservative care, 55 patients (39%) received delayed surgery. Of the 55 surgically treated patients, 71% had definite nerve root compression at baseline compared with 72% of conservatively treated patients (p = 0.76). Large disc herniations (size > 50% of spinal canal) were nearly equally distributed between those who did and those who did not undergo surgery (25% vs 21%, p = 0.65). The size of the dural sac was smaller in the patients who underwent surgery (101.2 vs 122.9 mm2, p = 0.01). However, the size of the dural sac discriminated poorly between those who did and those who did not undergo delayed surgery (area under ROC curve, 0.62). Conclusions. In patients who suffered from sciatica of 6-12 weeks' duration, MRI at baseline did not distinguish between patients who did and those who did not undergo delayed surgery. Clinical trial registration no.: ISRCTN26872154 (http://www.controlled-trials.com/ISRCTN/).