Impact of [18F]FDG-PET and [18F]FLT-PET-Parameters in Patients with Suspected Relapse of Irradiated Lung Cancer

Radiation-induced changes may cause a non-malignant high 2-deoxy-2-[18F]fluoro-d-glucose (FDG)-uptake. The 3′-deoxy-3′-[18F]fluorothymidine (FLT)-PET/CT performs better in the differential diagnosis of inflammatory changes and lung lesions with a higher specificity than FDG-PET/CT. We investigated the association between post-radiotherapy FDG-PET-parameters, FLT-PET-parameters, and outcome. Sixty-one patients suspected for having a relapse after definitive radiotherapy for lung cancer were included. All the patients had FDG-PET/CT and FLT-PET/CT. FDG-PET- and FLT-PET-parameters were collected from within the irradiated high-dose volume (HDV) and from recurrent pulmonary lesions. For associations between PET-parameters and relapse status, respectively, the overall survival was analyzed. Thirty patients had a relapse, of these, 16 patients had a relapse within the HDV. FDG-SUVmax and FLT-SUVmax were higher in relapsed HDVs compared with non-relapsed HDVs (median FDG-SUVmax: 12.8 vs. 4.2; p < 0.001; median FLT-SUVmax 3.9 vs. 2.2; p < 0.001). A relapse within HDV had higher FDG-SUVpeak (median FDG-SUVpeak: 7.1 vs. 3.5; p = 0.014) and was larger (median metabolic tumor volume (MTV50%): 2.5 vs. 0.7; 0.014) than the relapsed lesions outside of HDV. The proliferative tumor volume (PTV50%) was prognostic for the overall survival (hazard ratio: 1.07 pr cm3 [1.01–1.13]; p = 0.014) in the univariate analysis, but not in the multivariate analysis. FDG-SUVmax and FLT-SUVmax may be helpful tools for differentiating the relapse from radiation-induced changes, however, they should not be used definitively for relapse detection.


Introduction
The early and precise diagnosis in patients suspected for lung cancer relapse is essential, as a curative intended treatment might be feasible [1]. The 2-deoxy-2-[ 18 F]fluoro-Dglucose (FDG)-PET/CT is recommended in patients with a clinical suspicion of recurrence of lung cancer after treatment [2,3]. However, a pathological confirmation is not always feasible. Concerns of risks associated with performing a biopsy and possibilities of sampling errors in small tumors rule out a biopsy in some patients. Therefore, a better accuracy Patients were eligible if they met the following inclusion criteria: Histologically confirmed lung cancer, i.e., NSCLC or SCLC; definitive radiotherapy within 2 years and current suspicion of local relapse based on symptoms and/or CT-or FDG-PET/CT warranting (further) FDG-PET/CT for relapse diagnosis.
Patients were recruited prospectively from three university hospitals in the Capital Region of Denmark during January 2015 to 2019. The follow-up ended in May 2020. The study was approved by the local ethics committee (approval number H-4-2014-060) and by institutional review boards. All the patients signed a written informed consent. We have previously published data from this cohort concerning the diagnostic value of visual analysis of PET scans [4].

Imaging
All the patients went through an FLT-PET/CT as well as an FDG-PET/CT within 4 weeks of each other.
FLT-PET/CT was performed at Rigshospitalet on a Siemens Biograph TruePoint TrueV 40 or 64 PET/CT-scanner. FLT (5 mg/kg, max 350 MBq) was injected 60 ± 10 min prior to scanning, according to the study guidelines. Patients were not subject to any restrictions regarding fasting or resting before FLT-PET/CT. FLT-PET were reconstructed using an ordered subset expectation maximization (OSEM) with point spread function modelling (PSF), three iterations, and 21 subsets with a 2 mm full width half maximum (FWHM) Gaussian post-reconstruction filter.
FDG-PET/CT was performed as a routine clinical investigation, according to the local procedures at the referring hospital. Patients fasted at least 4 h before the injection of FDG and rested approximately 60 min between the injection and scan. Images  Table S1.

Image Analyses
PET/CT-scans were reviewed and analyzed on a Mirada Medical Ltd. XD 3.6 workstation by TNC, AGA, and BMF. The two latter readers being a board certified radiologist and nuclear medicine physician, respectively with more than 10 years of experience.
From the radiotherapy-plan, a high-dose irradiated volume (HDV) was defined within the 50% isodose curve (Figures 1 and 2) for each patient. Radiation-induced changes may occur in larger areas than, e.g., the planning target volume, thus to be able to evaluate the potential of FDG-PET/CT and FLT-PET/CT to differentiate radiation-induced changes from relapse, HDV was selected to be more representative of the clinically challenging area. Within HDV, the tracer uptake was quantified by a standardized uptake value (SUV) measuring SUV max and SUV peak . SUV max was defined as an SUV in the voxel with the highest SUV in the region of interest. SUV peak was defined as the highest average of SUV in a 1 cm 3 sphere within the region of interest.
Time from relapse suspicion to FLT-PET/CT; in days (median (IQ)) 23 (21)(22)(23)(24)(25)(26)(27)(28)(29) Time between FLT-PET/CT and FDG-PET/CT; in days (median (IQ)) 6 (3-9) Thirty patients were diagnosed with intrapulmonary relapse within 6 months after inclusion. In 16 patients, the relapse was located within the HDV. Figures 1 and 2 illustrate the radiotherapy plan and FDG-PET/CT and FLT-PET/CT from two patients with relapse within HDV and without relapse, respectively. The relapse was confirmed by histology in eight patients (four HDVs) and by subsequent growth in 11 patients (six HDVs). In twelve patients, the relapse diagnosis was based solely on FDG-PET/CT as decided by a multi-disciplinary conference, of these, six patients had the disseminated disease. Two patients were diagnosed with extra-pulmonary relapse only.  In patients with intrapulmonary relapse, SUV max , SUV peak , and the functional tumor volume were measured in the relapses. Functional tumor volumes were delineated in three different ways: (1) by 50% of SUV max (MTV 50 from FDG-PET/CT (metabolic tumor volume) and PTV 50 from FLT-PET/CT (proliferative tumor volume), as recommended for FDG-PET by the European Association of Nuclear Medicine (EANM) procedure guidelines [24]; (2) by 80% of SUV max (MTV 80 and PTV 80 ) as suggested for post-radiotherapy delineation for better discriminating from the adjacent tissue [25]; and (3) by SUV max > 3.0 (MTV 3.0 and PTV 3.0 ) as this threshold previously has shown promise in the post-treatment setting [26]. In patients with more than one intrapulmonary malignant lesions, the highest SUV max and SUV peak from the lesions were used, and the functional tumor volume was calculated as the sum of all malignant lesions. Diagnostics 2021, 11, x FOR PEER REVIEW 6 of 16

Outcome
The endpoints were (1) relapse and (2) overall survival. Relapses were categorized as (1) within HDV, (2) intrapulmonary, and (3) extra-pulmonary. The relapse status was defined retrospectively as the presence or absence of relapse within 6 months after inclusion. The relapse status was preferably confirmed by histology. If histology was not available or inconclusive, the relapse status was judged by a clinical oncologist, based on subsequent imaging procedures, invasive procedures, histology, and conference decisions within 6 months. The clinical oncologist was blinded for the name and age of the patient, names of the involved physicians, and dates of investigations.
The overall survival was calculated from the day of relapse suspicion to death by any cause.

Statistics
The correlation analyses were performed using Spearman's correlation as the PETparameters were not normally distributed. Differences of PET-parameters in HDVs with relapse and non-relapsed HDVs, and differences of PET-parameters in relapsed lesions within vs. outside of HDV were estimated with independent t-tests or the Welch test, if assumptions for the t-test were not fulfilled. Receiver operating characteristic (ROC) curves was created to explore the diagnostic accuracy of FDG-SUV max and FLT-SUV max . The sensitivity and specificity for relapse diagnosis within HDV with variant SUV max cutoffs were estimated. The diagnostic value of combining FDG-SUV max and FLT-SUV max was explored with different combinations of FDG-SUV max -and FLT-SUV max -cutoffs.
Overall survival was calculated by the Kaplan-Meier method. The univariate and multivariate survival analysis including clinical parameters and PET-parameters from relapsed lesions and HDV were performed by Cox regression. PET-parameters were included both as continuous variables and dichotomized by the median cut-off. The number of included covariates in the multivariate analysis was restricted due to the limited number of patients. PET-parameters with p < 0.150 in the univariate analysis were included in addition to preselected clinical covariates: Sex, age (at time of suspicion), and time from the end of radiotherapy to the suspicion of relapse. The correlation analyses were performed separately on FDG-PET and FLT-PET-parameters to avoid including highly correlated parameters. Statistical analyses were performed in SPSS, version 25.

Results
A total of 63 patients were enrolled in this study. In two patients, the relapse status after 6 months of follow-up was non-confirmed, and they were excluded from further analysis. Accordingly, 61 patients were included in the final analysis. Patients were included based on the relapse suspicion raised on surveillance CT (n = 51) or surveillance FDG-PET/CT (n = 9). Four patients had symptoms of relapse when included, of these, one patient was included without prior CT. Fifty-seven patients had NSCLC, two patients SCLC, and two patients mixed NSCLC/SCLC. Patients were included a median of 7 months (interquartile range (IQ): 5-12) after the end of radiotherapy.

FDG-SUV max and FLT-SUV max
SUV max and SUV peak in HDV were strongly correlated for both FDG and FLT (FDG-SUV: R = 0.958; p < 0.001; FLT-SUV: R = 0.958; p < 0.001). FDG-SUV and FLT-SUV in HDV had a positive and strong correlation (SUV max : R = 0.804; p < 0.001; SUV peak : R = 0.770; p < 0.001), as shown in Figure 3.     Similar results were obtained when the analyses were restricted to include only patients with NSCLC (Supplemental Tables S2 and S3).

FDG-SUVmax and FLT-SUVmax
There was no significant correlation between the time from relapse suspicion to FDG-PET/CT and FDG-SUVmax in recurrent pulmonary lesions (p = 0.512), nor from the suspicion of relapse to FLT-PET/CT and FLT-SUVmax in recurrent lesions (p = 0.670), implying that SUV-estimates were not biased from the time delay between the relapse suspicion and conducting the PET scans.

Relapse Detection
The HDV with relapse had a significantly higher SUVmax compared with non-relapsed HDV (median FDG-SUVmax: 12.  Similar results were obtained when the analyses were restricted to include only patients with NSCLC (Supplemental Tables S2 and S3).
There was no significant correlation between the time from relapse suspicion to FDG-PET/CT and FDG-SUV max in recurrent pulmonary lesions (p = 0.512), nor from the suspicion of relapse to FLT-PET/CT and FLT-SUV max in recurrent lesions (p = 0.670), implying that SUV-estimates were not biased from the time delay between the relapse suspicion and conducting the PET scans.
PET-parameters from pulmonary recurrent lesions are presented in Table 2. Recurrent lesions within HDV had higher SUV and larger functional tumor volumes than lesions with recurrence outside of HDV. However, only significantly so for FDG-PET-parameters (FDG-SUV peak , MTV 3.0 , and MTV 50% ).

Diagnostic Accuracy of FDG-SUV max and FLT-SUV max
ROC-curves showed a moderate diagnostic accuracy for both FDG-SUV max (AUC: 0.873) and FLT-SUV max (AUC: 0.767), as shown in Figure 6. It was not possible from the ROC curves to establish cutoffs with a reasonable balance between sensitivity and specificity for neither FDG-SUV max nor FLT-SUV max . The sensitivity and specificity for variant cutoffs of SUV max are presented in Figure 7. PET-parameters from pulmonary recurrent lesions are presented in Table 2. Recurrent lesions within HDV had higher SUV and larger functional tumor volumes than lesions with recurrence outside of HDV. However, only significantly so for FDG-PET-parameters (FDG-SUVpeak, MTV3.0, and MTV50%).

Diagnostic Accuracy of FDG-SUVmax and FLT-SUVmax
ROC-curves showed a moderate diagnostic accuracy for both FDG-SUVmax (AUC: 0.873) and FLT-SUVmax (AUC: 0.767), as shown in Figure 6. It was not possible from the ROC curves to establish cutoffs with a reasonable balance between sensitivity and specificity for neither FDG-SUVmax nor FLT-SUVmax. The sensitivity and specificity for variant cutoffs of SUVmax are presented in Figure 7.

Diagnostic Value of Combining FDG-SUV max and FLT-SUV max
A superior diagnostic value of combining FDG-SUV max and FLT-SUV max was not obvious due to the high correlation of FDG-SUV max and FLT-SUV max . Combinations of FDG-SUV max and FLT-SUV max with variant cutoffs were explored without an additional diagnostic value. In the subgroup of HDVs with high FDG-SUV max (variant cutoffs), no optimal cutoff could be identified by the ROC-curves for FLT-SUV max and the diagnostic accuracy was low (e.g., HDV with FDG-SUV max ≥ 6.0: AUC: 0.654).

Error in Figure
In the original publication [1], there was a mistake in Figure 6a.

Error in Figure
In the original publication [1], there was a mistake in Figure 6a. Figure 6a was replaced by Figure 6b, and thus Figure 6a,b appeared identical. The corrected Figure 6 appears below. The authors apologize for any inconvenience caused and state that the scientific conclusions are unaffected. This correction was approved by the Academic Editor. The original publication has also been updated.

Conflicts of Interest:
The authors declare no conflict of interest.

Diagnostic Value of Combining FDG-SUVmax and FLT-SUVmax
A superior diagnostic value of combining FDG-SUVmax and FLT-SUVmax was not obvious due to the high correlation of FDG-SUVmax and FLT-SUVmax. Combinations of FDG-SUVmax and FLT-SUVmax with variant cutoffs were explored without an additional diagnostic value. In the subgroup of HDVs with high FDG-SUVmax (variant cutoffs), no optimal cutoff could be identified by the ROC-curves for FLT-SUVmax and the diagnostic accuracy was low (e.g., HDV with FDG-SUVmax ≥ 6.0: AUC: 0.654).

Prognosis
At the time of analysis, 33 patients had died. The median overall survival from relapse suspicion was 41 months. The median survival of patients who had intra-pulmonary relapse was 42 months. The two patients with extra-pulmonary relapse only died 7 and 10 months after relapse suspicion. The median overall survival of patients without relapse could not be estimated due to too few events (n = 12), and was at least 41 months.
In the univariate analysis, PET-parameters in HDV did not show a significant prognos- The univariate survival analysis performed on the patients with intrapulmonary relapse, appointed PTV 50% as the only significant prognostic factor for overall survival (HR: 1.07 pr cm 3 [1.01-1.13]; p = 0.014). MTV 50% , MTV 80% , PTV 80% , and sex were borderline significant. Results from the univariate analyses are shown in Table 3. MTV 50% and MTV 80% were highly correlated (R = 0.785, p < 0.001), as were PTV 50% and PTV 80% (R = 0.531, p = 0.003). Therefore, only MTV 50% and PTV 50% were included in the multivariate analysis. Male sex was independently prognostic for a poorer overall survival. No PET-parameter had an independent prognostic value. Larger MTV 50% and a higher age were borderline significant for poorer overall survival. Results from the multivariate analysis are shown in Table 4. The subgroup analysis on patients with NSCLC yielded similar results (Supplemental Tables S4 and S5).

Discussion
We investigated the impact of semi-quantitative parameters of FDG-PET and FLT-PET in patients suspected for having a relapse of lung cancer after treatment with definitive radiotherapy. The FDG-and FLT-uptake was significantly higher in relapsed HDVs than HDVs without relapse. However, with a considerable overlap of SUVs in relapsed vs. non-relapsed HDVs, neither FDG-SUV max nor FLT-SUV max can be used definitively for distinguishing the relapse from radiation-induced changes. However, the longer after radiotherapy the PET scan was performed, the difference between SUV max in HDVs with relapse vs. HDVs without relapse increased. The prognostic value of FDG-and FLT-PETparameters after radiotherapy was limited.
Results from the few studies addressing PET-parameters in relapses after radiotherapy were similar to ours. Nakajima et al. [6] showed that patients with recurrences after SBRT had a significantly higher FDG-SUV max than patients without relapse. Saga et al. [18] reported that after SBRT, FLT-SUV max in relapsed lesions (range: 2.0-5.9; n = 5) were higher than in non-relapsed lesions (range: 1.3-2.1, n = 3; no statistics performed).
In our data, FDG-and FLT-uptakes in non-relapsed HDVs were lower the longer after radiotherapy PET/CT was performed, whereas FDG-and FLT-uptakes were higher in HDVs with relapse the longer after radiotherapy. PET-parameters may have a higher impact late after radiotherapy, as the difference of tracer-uptake in relapsed vs. nonrelapsed lesions increased with time. In FDG-PET, a reduced FDG-uptake with time was expected following the resolution of radiation-induced changes [7,27]. FLT-PET is less likely to have an unspecific uptake in inflammation [28][29][30]. However, a moderate FLTuptake the first 9 months after radiotherapy, as seen in our study, implies that some degree of non-specific FLT-uptake may be present.
It has previously been shown that the diagnostic accuracy of FDG-SUV max was higher when FDG-PET/CT was performed more than 12 months after SBRT [6]. The sensitivity and specificity were 100% for SUV max > 4.5 for patients who had FDG-PET/CT conducted more than 12 months after SBRT, compared with 87% and 98% for their entire cohort. Hoopes et al. [8] did not support a specificity of 100% of SUV max > 4.5, as FDG-SUV max was 5.1 and 5.9 in two patients without evidence of disease 23 and 26 months after SBRT, respectively [8]. A direct comparison of SUV max from different studies should be taken with precaution. Particular precautions when comparing our study with the study of Nakajima et al. [6] are the different reconstruction methods (OSEM with and without PSF point-spread function, respectively) and different FDG-uptake times (60 vs. 90 min). PSF reconstructions lead on average to higher SUV max than non-PSF reconstructions [31,32]. However, in our study, two patients without relapse had a high FDG-uptake on FDG-PET/CT more than 12 months after the end of radiotherapy: In one patient who had ended SBRT 14 months prior to FDG-PET/CT, FDG-SUV max was 5.9. In another patient who had ended normo-fractionated radiotherapy 21 months prior to FDG-PET/CT, FDG-SUV max was 6.0. FLT-SUV max in the two patients were 2.5 and 2.2, respectively. In both patients, the biopsy was negative and there was no evidence of progression the following 6 months. Accordingly, from our data, the specificity of SUV max > 4.5 was 75% in patients with relapse suspicion more than 12 months after the end of radiotherapy. The diagnostic value of the visual interpretation of FDG-PET/CT after treatment with a curative intent, surgery, and radiotherapy has been addressed in several papers with an estimated sensitivity between 82% and 100%, and specificity between 82% and 98% [2]. A meta-analysis showed a pooled sensitivity and specificity of 90% [33]. After radiotherapy, we have previously demonstrated a higher sensitivity and specificity with visual analysis of FDG-PET/CT (94%, and 71%, respectively) and FLT-PET/CT (69% and 90%, respectively) than any SUV max cut-off [4]. From our data, we cannot support a fixed threshold of neither FDG-SUV max nor FLT-SUV max for the relapse diagnosis.
In this study, we also observed that relapses within HDV had higher SUV max , SUV peak , and larger tumor volumes than relapsed lesions that had not been irradiated previously, particularly for FDG-parameters. A non-malignant FDG-uptake caused by radiationinduced changes [6] could attribute to the differences, though also difficulties in the diagnostic process leading to later diagnosis of relapses within HDVs may cause larger and more aggressive tumor relapses in HDVs.
A limitation to our study was that the pathological confirmation of relapse was not always obtained. To minimize the effect of this limitation, a retrospective compound reference standard for the relapse status was applied, based on the patients' records within 6 months of follow-up. In most patients, the relapse was confirmed by histology or follow-up with subsequent progression, and non-relapse by follow-up without progression. However, in 13/61 patients (six HDVs), the relapse was decided based on FDG-PET/CT by a multi-disciplinary conference, and the clinical decision warranted the start of treatment rather than further confirmation, either due to an obvious outcome of FDG-PET/CT and/or due to fact that the patient was unfit for invasive procedures. FDG-PET/CT is recommended as a second step test when the relapse is suspected after radiotherapy [3], thus excluding patients with an obvious relapse on FDG-PET/CT would have caused the study to be less representative.
Our survival analysis showed that FLT-PET (PTV 50% ) had a prognostic value in patients with relapsed lung cancer after radiotherapy. However, when adjusting for sex, age and time since RT, PTV 50% was not independently prognostic for overall survival. MTV 50% was borderline significant in both univariate and multivariate analysis. Therefore, MTV 50% might be a better prognosticator than PTV 50% , but needs further confirmation.
Previous studies exploring the prognostic value of post-radiotherapy PET-parameters are limited, and most studies have focused on an FDG-change rather than the absolute value of PET-parameters [2]. In contrast to our results, Lopez Guerra et al. [20] showed a prognostic value of FDG-SUV max after normo-fractionated radiotherapy for overall survival with HR 1.27. In addition, Bollineni et al. [21] showed a trend for the prognostic value of survival after SBRT [21]. In studies including lung cancer patients treated with surgery [2], and in studies including cohorts mixed of patients treated with a curative intent, the post-treatment FDG-SUV max also showed a prognostic value [2,34]. FLT-SUV max after the end of carbon-ion radiotherapy has not shown a prognostic value [19]. The above studies address different populations with different stages. However, they have in common that all patients have been treated with a curative intent. Baseline studies have previously shown a prognostic value of FDG-SUV max and MTV [35,36]. Results from baseline FLT-PET studies are not in agreement [16,19,37], and Scheffler et al. [37] showed an independent prognostic value of FDG-SUV max , but not FLT-SUV max . The prognostic value of PET-parameters measured before the treatment may, however, not be applicable on PET-parameters measured after radiotherapy, as the more complex biological tumor characteristics may affect the tracer-uptake. In addition, the lesser treatment options for relapsed cancers affect the survival compared with survival in patients with treatment naïve tumors. The survival analysis in our study was limited by the sample size. Our cohort consisted of patients with different stages, different histologies, different treatment schemes, and accordingly a different prognosis, although all the patients were initially treated with a curative intent. A prognostic value of PET-parameters may be affected by other prognostic factors. However, the small patient number limited possibilities for adjustment. The overall survival in our study was long, also for patients with relapse. PET-parameters measured several years prior to death probably contain a less prognostic value. Approximately half of the patients with relapse were treated with a curative intent. However, the intent or relapse treatment did not have a prognostic value. Therefore, in this study, the overall survival may not be useful as an endpoint. The progression free survival was not selected as an endpoint, as patients were included due to a suspicion of relapse, and therefore, many patients were confirmed with relapse shortly after inclusion.
An early diagnosis of lung cancer relapse is essential, as in some cases, a curative intended treatment might be feasible [1]. In this study, relapsed lesions within HDV had higher FDG-SUV peak and larger MTV, than the relapsed lesions outside of HDV. The same characteristics were applied for the FLT-uptake and PTV, though not statistically significant. The cause is unknown, however, it is possible that radiation-induced changes in the HDV challenge and thus delay the diagnostic decision making. FDG-PET-SUV max and FLT-PET-SUV max can be supportive for distinguishing between radiation-induced changes and relapse, but should not be used definitively. Over time, the differences in FDG-SUV max and FLT-SUV max , respectively in relapsed vs. non-relapsed HDVs increased, thus SUV max may have a higher impact later after radiotherapy.

Conclusions
The FDG-and FLT-uptake was significantly higher in relapsed HDV than HDV without relapse. However, caution should be taken if using SUV for the relapse diagnosis due to a considerable overlap of FDG-SUV max , respectively, FLT-SUV max in relapsed vs. non-relapsed HDVs.
PTV 50% was prognostic for the overall survival in patients with relapse, however, PTV 50% had no independently prognostic value.
Supplementary Materials: The following are available online at https://www.mdpi.com/2075-4 418/11/2/279/s1. Table S1: Data on the PET model and reconstruction used for FDG-PET/CT; Table S2: PET-parameters in intrapulmonary malignant lesions in patients with NSCLC; Table S3: FDG-SUV max and FLT-SUV max in relapsed vs. non-relapsed HDV in patients with NSCLC; Table S4: Prognostic value of PET-parameters and clinical variables in patients with intrapulmonary relapse of irradiated NSCLC; Table S5: multivariate survival analysis in patients with relapsed NSCLC. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available on reasonable request from the corresponding author.