**3. Discussion**

In the present study, we investigated the role of 18F-FDG-PET/CT for monitoring functional tumor response in comparison to morphological imaging and its ability to predict PFS and DSS in patients with advanced, radioiodine refractory DTC undergoing Lenvatinib treatment. Our results demonstrate that functional imaging is able to evaluate tumor response in a more differentiated manner than morphological imaging only. In our series, all patients who responded to therapy showed a decline in all PET-parameters except SUVmean, whereas a lack of functional tumor response was associated with a worse outcome (PFS and DSS). PET-parameters SUVpeak, SUVmax, MTV and TLG at 3 and 6 month follow-ups were significantly higher in patients with disease progression and could serve as additional markers for monitoring early tumor response and outcome.

The current standard for monitoring treatment response and progression in clinical trials is the change in tumor size assessed by RECIST. To date, the current guidelines do not give specific recommendations for response monitoring in patients with DTC outside clinical trials and response assessment by PERCIST is not mentioned [19]. Since clinical patient management and treatment planning depend on response monitoring through imaging, clinical decisions may vary based on different imaging modalities.

18F-FDG-PET/CT combines functional and morphological imaging. Measurement of glucose metabolism varies less than tumor size measurements and can better distinguish between active tumor and post-therapeutic changes [20,21]. Tumor response evaluation using 18F-FDG-PET/CT showed promising results for several cancer entities, such as breast, lung and pancreatic cancer [17,21–23]. In addition, many previous studies have shown the important role of 18F-FDG-PET/CT in staging and follow-up of patients with advanced, metastatic DTC [24–26] and its valuable role in patients with metastatic DTC under treatment with TKIs [27,28]. However, to date the number of studies in patients with DTC undergoing Lenvatinib treatment is limited and no standardized treatment assessment has been proposed so far.

Tumor FDG-uptake can be measured in various ways. Firstly, we used the single-lesion method according to Wahl et al. [29], which was shown to be superior to the five-lesion method [22,30]. Secondly, in accordance to Fendler et al., we used peak standardized uptake value corrected for body weight (SUVpeak) instead of lean body mass corrected SUV (SULpeak) as proposed by PERCIST 1.0, because the main objective is the percentage change of SUV from baseline to follow-up imaging and should therefore not be a significant confounder [31]. Riedl et al. could also show that response classification was unchanged when SUVmax was used instead of SULpeak in patients with metastatic breast cancer [22]. Thus, we measured the SUVpeak of the most active lesion, which may differ in consecutive scans. Furthermore, we assessed the PET-parameters SUVmean, SUVmax, MTV and TLG to identify possible other measures since technical methods for quantitative measurement are under continuous improvement [32].

Novel cancer therapies, such as TKI treatment, are cytostatic rather than cytotoxic and therefore may not result in a significant decrease of tumor size [33,34]. Additionally, certain metastasis localizations such as bone metastasis do not frequently show morphological changes after therapy [22]. Thus, differentiation of tumor response categorization in CR, PR, SD and PD in these patients by morphological imaging is limited. However, FDG-uptake in tumor cells is known to correlate with disease prognosis [35] and, as precision therapy is evolving, the current monitoring of treatment response does not seem to have been adjusted accordingly.

Whereas the majority of patients in our study were categorized as SD by RECIST 1.1 on CT at both follow-up times, our data show that patients were categorized in a more differentiated manner by mPERCIST using 18F-FDG-PET. This finding is in line with the study of Riedl et al., who reported that patients with metastatic breast cancer with SD and PD by RECIST were frequently classified worse by PERCIST [22].

Based on these differences in categorization of patients, Kaplan-Meier analysis demonstrated significant distinction between DC and PD. Tumor response by mPERCIST was significantly correlated with PFS and DSS, whereas tumor response by morphological imaging showed no significant correlation. In our study, responses determined by using the RECIST 1.1 criteria at 3 and 6 months was found to be statistically significant for DSS, but the association was lower compared to the mPERCIST response, indicating that mPERCIST

detects progression more precisely than RECIST. All other PET-parameters (SUVmean/max, MTV and TLG) showed no significant association with PFS and DSS. These findings are consistent with the results reported by Riedl et al. [22] and could subsequently lead to earlier change of therapy in patients considered SD by RECIST but do not show therapy response by mPERCIST. The association of metabolic response with survival benefit has already been shown in several other solid tumors such as breast cancer [36].

In this study, a decline of nearly all PET-parameters was found in patients with DC. Studies on the role of the PET-parameter SUVmax in DTC patients are limited. In line with our data, Carr et al. showed that patients with metastatic DTC or medullary thyroid carcinoma treated with Sunitinib showing DC had a significant decline in average and mean percentage change SUVmax compared to patients with progressive disease [27]. A significant decline in SUVmax could also be shown in patients with radioiodine refractory DTC treated with Apatinib, whereas in DTC patients treated with Vandetanib, no correlation between SUVmax and DC could be found [37]. To date, SUVmax is the most commonly used semiquantitative PET-parameter due to its simple application but there are more and more studies suggesting the use of SUVpeak alternatively [38]. Since SUVpeak is measured in a larger VOI than the single-pixel SUVmax, it appears to be more robust to image noise.

The volume-related PET-parameters MTV and TLG could also show promising results in metastatic DTC patients. The study by Manohar et al. showed that these parameters can be used for dynamic risk stratification regarding PFS [39]. TLG appeared promising in some cancers such as colorectal cancer and brain tumors, but not in others such as sarcomas [40–43]. Data from Lee et al. showed that in patients with pancreatic cancer, TLG was an independent prognostic factor for predicting recurrence-free and overall survival, whereas Benz et al. reported that TLG was less accurate in predicting tumor response in sarcomas compared to SUVmean and SUVmax [23,40]. The same was shown for MTV, which was found to be useful for treatment response assessment in non-small cell lung cancer (NSCLC) and pancreatic cancer [23,44]. Furthermore, it was demonstrated to be an independent prognostic factor for DSS in patients with cervical cancer treated by radical surgery and could independently predict survival in patients with locally advanced squamous cell cervical carcinoma [45,46]. Indeed, in our analysis, volume related PETparameters (MTV and TLG) were found to be useful tools to distinguish DC from PD but failed to provide prognostic value in terms of PFS and DSS. SUVmean showed the weakest correlation with tumor response in our study. In contrast, Werner et al. identified a SUVmean of less than 4.0 before treatment in medullary thyroid carcinoma as a predictor of longer PFS [20]. Consequently, all PET-parameters except for SUVmean seem to be a useful tool and may be evaluated alongside SUVpeak as suggested in PERCIST 1.1 criteria [29].

The percentage change of Tg-levels from baseline to 3 month- and baseline to 6 month follow-up 18F-FDG-PET/CT showed no significant difference between DC and PD patients. One possible explanation for the missing correlation between Tg and 18F-FDG-PET/CT might be the different de-differentiation level of the thyroid cancer patients treated with Lenvatinib.

To our knowledge, this is the first study attempting to address the impact of using different imaging modalities and therefore different evaluation criteria for treatment response in patients with advanced, radioiodine refractory DTC undergoing Lenvatinib treatment. However, the role of 18F-FDG-PET/CT and the selection of the optimal PET-parameters for monitoring of functional tumor response in patients with advanced, radioiodine refractory DTC undergoing Lenvatinib treatment will have to be verified in prospective trials in larger patient cohorts.

Our study has several limitations. Due to the retrospective design of the study, the span between baseline imaging and initiation of treatment and the intervals of follow-up imaging varied between patients and follow-up imaging at 3 months was not available in all patients. Furthermore, due to the small cohort, the rarity of the disease and heterogeneity in patient cohort (histology, stage of disease), statistical power of the analysis is limited.
