Lesion-based detection of early chemosensitivity using serial static FDG PET/CT in metastatic colorectal cancer

Abstract

Purpose

Medical oncology needs early identification of patients that are not responding to systemic therapy. ¹⁸F-Fluorodeoxyglucose (FDG) positron emission tomography (PET) performed before and early during treatment has been proposed for this purpose. However, the best way to assess the change in FDG uptake between two scans has not been identified. We studied cutoff thresholds to identify responding tumours as a function of the method used to measure tumour uptake.

Methods

The study included 28 metastatic colorectal cancer (mCRC) patients who underwent 2 FDG PET/CT scans (baseline and at day 14 of the first course of polychemotherapy). For 78 tumour lesions, 4 standardized uptake value (SUV) indices were measured: maximum SUV (SUV _max) and mean SUV in a region obtained using an isocontour (SUV _40 %), with each of these SUV normalized either by the patient body weight (BW) or body surface area (BSA). The per cent change and absolute change in tumour uptake between the baseline and the early PET scans were measured based on these four indices. These changes were correlated to the RECIST 1.0-based response using contrast-enhanced CT at baseline and at 6–8 weeks on treatment.

Results

The 78 tumours were classified as non-responding (NRL, n = 58) and responding lesions (RL, n = 20). Receiver-operating characteristic (ROC) curves characterizing the performance in NRL/RL classification using early FDG PET uptake had areas under the curve between 0.75 and 0.84, without significant difference between the indices. The cutoff threshold in FDG uptake per cent change to get a 95 % sensitivity of RL detection depended on the way uptake was measured: −14 % (specificity of 53 %) and −22 % (specificity of 64 %) for SUV _max and SUV _40 %, respectively. Thresholds expressed as absolute SUV decrease instead of per cent change were less sensitive to the SUV definition: an SUV decline by 1.2 yielded a sensitivity of RL detection of 95 % for SUV _max and SUV _40 %. For a given cutoff threshold, the sensitivity was the same whatever the normalization (by BSA or BW).

Conclusion

A 14 % drop of tumour FDG SUV _max, 22 % drop of SUV _40 % or 1.2 drop of SUV _max or SUV _mean after one single course of polychemotherapy predicts objective response in mCRC lesions with a high sensitivity, potentially allowing the early identification of non-responding patients.