Can semiquantitative measurements of SUVmax and cut-off values differentiate colorectal malignant from benign lessions?

OBJECTIVE

We investigated maximum standardized uptake value (SUVmax) and cut-off values for differentiation of malignant and benign lesions in colorectal cancer (CC) as multiple studies have questioned their validity. We also investigated more extended indices using common semi-quantification analysis in incidental colorectal findings (ICF).

SUBJECTS AND METHODS

Fluorine-18-fluoro deoxy glucose positron emission tomography/computed tomography in 25 patients with a total of 30 focal ICF was retrospectively analysed using dedicated software. Method variability was tested through application of three common threshold-based lesion delineation techniques as well as a partial-volume correction (PVC). Lesion SUVmax, SUVmean, metabolically active volume (MAV) and mean total lesion glycolysis (TLG) were thereby extracted along with PVC corrected values (cSUVmean, cTLG) and SUVpeak.

RESULTS

In all lesions, SUVmax was >5 and SUVmean≥2.7. Malignant SUVmax values (mean±SD: 16.5±6.2) were overall significantly higher than benign levels (9.8±3.6). There was a substantial overlap with values in polyps/adenomas (14.4±7.7). Both SUVpeak and SUVmean showed similar characteristics. Malignant MAV and TLG showed more distinct levels. Though different segmentation methods introduced variations, largest in MAV (-58.6%-141.5%), and PVC generally increased measures significantly by a factor of 1.2-2.7, neither changed relative levels much. SUVmax values were inadequate for aetiological differentiation of ICF, which also precludes a clinically significant cut-off value. The same applies to SUVpeak and SUVmean while TLG measures may be more indicative.

CONCLUSION

Semi-quantitative measurements of SUVmax and cut-off values proved inadequate for differentiating colorectal malignancies from benign findings. While integrated measures, e.g. cTLG, are potentially better indicators of disease severity and extent, more optimal segmentation and PVC methods are required.