Efficacy of high resolution magnetic resonance imaging in preoperative local staging of rectal cancer

Efficacy of high resolution magnetic resonance imaging inpreoperative local staging of rectal cancerINTRODUCTION: To assess the efficacy of magnetic resonance imaging (MRI)with a phased-array coil for preoperative T and N staging and prediction of thecircumferential resection margin involvement in patient with rectal cancer. Thesepatients had no preoperative radiochemotheraphy.METHODS: In this retrospective study, 30 patients with histologically provenprimary rectal cancer were preoperatively evaluated by MRI at 1.5 T, with a phasedarraycoil. Patients were staged preoperatively using magnetic resonance imaging byan abdominal radiologist and a radiology resident. Two observers independentlyscored, on three occasions, the tumour stage, lymph node stage and measured thedistance to the mesorectal fascia. The sensitivity, spesificity, the negative predictivevalue and the positive predictive value of MRI findings were measured 95%confidence interval. The area under the curve was measured for each result.Agreement between radiologic staging of tumor, local lymph nodes, andcircumferential resection margin involvement with pathologic reporting and alsoagreement between two observers were assessed by means of the Kappa statistic.RESULTS: In T staging the accuracy rate of high resolution magnetic resonance(HRMR) was 47-67%, overstaging was 10-21%, and understaging was 13-43%. Inthe prediction of extramural invasion with HRMR, the sensitivity was 79-89%, thespecificity was 72-100%, the PPV was 85-100%, the NPV was 73-86%, the areaunder the curve was 0.89 for first observer and, 0.81 for second observer. Predictinglymph node metastasis, the sensitivity 58%, the specificty 50-55%, the PPV was 44-47%, the NPV was 64-66%, the area under the curve is 0.54 of first observer, 0.57 forsecond observer. When the cut off value was selected as 1 mm, the sensitivity ofHRMR was 38-82%, the specificity was 73-82%, the PPV was 33-42%, and NPV was79-81% in the prediction of the CRM involvement. Statistically, there was goodcorrelation between pathologic and radiologic tumor staging (k=0.73 for first observer,k=0.63 for second observer). There was poor correlation between pathologic andradiologic lymph node staging (k=0.07 for first observer, k=0.13 for second observer).Statistically, there was fair correlation between pathologic and radiologic reporting ofCRM involvement (k=0.25 for first observer, k=0.20 for second observer). Butcorrelation between two observers was moderate for tumor staging, substantial fornode staging and reporting of CRM involvement.CONCLUSION: Preoperative HRMR provides good predictive data for tumorstaging but poor prediction of lymph node status because of the reactive lypmh nodesand poor staging in CRM involvement.(Key words: Rectum ; Rectal cancer; Magnetic resonance imaging; Local staging)