Objective: To evaluate and compare different PET- and CT-based imaging criteria for early response assessment and PFS prediction in patients with gastroenteropancreatic neuroendocrine tumors (GEP-NET) under peptide receptor radionuclide therapy (PRRT).

Materials and Methods: This single-center retrospective study included n=178 patients with well-differentiated (G1 and G2) GEP-NETs treated with at least two consecutive cycles of PRRT with ¹⁷⁷Lu-DOTA-TATE who underwent somatostatin receptor (SSTR) PET/CT at baseline and after two cycles of PRRT (follow-up). The category differences between RECIST 1.1 (morphological criteria), mCHOI (functional criteria), and the multidisciplinary tumor board assessment (MDT) as the reference standard were evaluated. This included an analysis of patients’ histories, physical examinations, laboratory results (NSE and CgA), and imaging findings (PET/CT). Subsequently, patients were categorized into two clinically relevant groups: responders for partial remission (PR) and stable disease (SD) and non-responders for progressive disease (PD) and mixed response (MR) to PRRT therapy. The predictive value of the different response categories in Krenning and SSTR-RADS (molecular criteria) for progression-free survival (PFS) were also analyzed.

Results: RECIST categorized 84% patients as SD and 7% as PR, but mCHOI (53% SD, 33% PR) and MDT (72% SD, 21% PR) showed a category shift towards PR (figure 1). The PFS of patients with PD according to RECIST 1.1 (p < 0.001) and mCHOI (p < 0.001) was significantly shorter compared to patients with SD and PR. This correlates with the results of the MDT, where patients with PD and MR had significantly shorter PFS than patients with SD and PR (p < 0.001, figure 2). Regression models in PFS using mCHOI (AIC 611.78, concordance index 0.69) displayed a non-significantly higher concordance to MDT (AIC 605.1, concordance index 0.73) than models using RECIST 1.1 (AIC 615.75, concordance index 0.67). mCHOI trended to a non-significantly higher agreement with MDT for classification of patients into responders (PR/SD) and non-responders (PD/MR) compared to RECIST 1.1 (mCHOI/MDT: κ = 0.59, 95%-CI 0.36-0.81; RECIST 1.1/MDT: κ = 0.44, 95%-CI 0.17-0.72). PFS was significantly longer in patients with a Krenning score of 4 at baseline compared to patients with a Krenning score of 3 (p=0.01, figure 3). SSTR-RADS score of 3, 4 and 5 showed no significant difference for prediction of PFS (figure 3).

Conclusion: The data indicate that mCHOI demonstrates robust performance for predicting PFS and assessing response in well-differentiated GEP-NETs undergoing PRRT with higher agreement to MDT compared to RECIST 1.1. This approach is more accurate than RECIST 1.1 in classifying patients as responders and non-responders. Furthermore, the use of Krenning for PFS prediction is indicated for this patient cohort.

Clinical Relevance Statement: Despite multiple limitations, RECIST 1.1 has been established as the standard for response assessment of NETs under PRRT (1), as no significant advantages were found when using either CHOI or RECIST 1.1 criteria (2, 3). However, the data presented here demonstrate improved concordance of mCHOI with MDT. Both, RECIST 1.1 and mCHOI offer comparable diagnostic value for predicting PFS. At baseline, the Krenning score can be used to estimate long-term therapy response. Consequently, in addition to RECIST 1.1, further PET- and CT-based imaging criteria can be recommended for response assessment and PFS prediction in well-differentiated GEP-NETs undergoing PRRT.

Image/Figure:
Click to View

Author