Improving accuracy of 18F-fluorodeoxyglucose PET computed tomography to diagnose nodal involvement in non-small cell lung cancer: utility of using various predictive models.

Abstract

To determine predictive models (PM) that could improve the accuracy for identifying metastatic regional nodes in non-small cell lung cancer based on both PET and CT findings seen on 18F-FDG PET CT. Three hundred thirty-nine biopsy-proven NSCLC patients who underwent surgical resection and had a staging 18F-FDG PET CT were enrolled. PET parameters obtained were (1) presence of visual PET positive nodes, (2) SUVmax of nodes (NSUV), (3) ratio of node to aorta SUVmax (N/A ratio) and (4) ratio of node to primary tumour SUVmax (N/T ratio). CT parameters obtained were (1) short-axis diameter and (2) Hounsfield units (HU) of PET-positive nodes. PET and CT parameters were correlated with nodal histopathology to find out the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and overall accuracy. Different PM combining these parameters were devised and the incremental improvement in accuracy was determined. Visual PET positivity showed sensitivity, specificity, PPV, NPV and accuracy of 72.4, 76.1, 30.1, 95.1 and 75.6, respectively. PM2 which combined visual PET positivity, NSUV and HU appears more clinically relevant and showed sensitivity, specificity, PPV, NPV and accuracy of 53.5, 96.5, 68.9, 93.6 and 91.2, respectively. PM6 which combined visual PET positivity, NSUV, N/A ratio and HU showed the maximum PPV (80.0%), specificity (98.3%) and accuracy of (91.9%). PM combining parameters like nodal SUVmax, N/A ratio, N/T ratio and HU values have shown to improve the PPV, specificity and overall accuracy of 18FDG PET CT in the preoperative diagnosis of nodal metastases.