Prospective assessment using 18F-FDG PET/CT as a novel predictor for early response to PD-1 blockade in non-small-cell lung cancer

Anti-programmed death-1 (PD-1) blockade is a standard treatment for advanced non-small-cell lung cancer (NSCLC). However, no appropriate modality exists for monitoring its therapeutic response immediately after initiation. Therefore, we aimed to elucidate the clinical relevance of 18F-FDG PET/CT versus CT in predicting the response to PD-1 blockade in the early phase. This prospective study included a total of 54 NSCLC patients. 18F-FDG PET/CT was performed at 4 weeks and 9 weeks after PD-1 blockade monotherapy. Maximum standardized uptake values (SULmax), metabolic tumor volume (MTV), and total lesion glycolysis (TLG) were evaluated. Among all patients, partial metabolic response and progressive metabolic disease after PD-1 blockade were observed in 35.2% and 11.1% on SULmax, 22.2% and 51.8% on MTV, and 27.8% and 46.3% on TLG, respectively, whereas a partial response (PR) and progressive disease (PD), respectively, based on RECIST v1.1 were recognized in 35.2% and 35.2%, respectively. The predictive probability of PR (MTV: 57.9% vs. 21.1%, p = 0.044; TLG: 63.2% vs. 21.1%, p = 0.020) and PD (MTV: 78.9% vs. 47.3%, p = 0.002; TLG: 73.7% vs. 21.1%, p = 0.007) detected based on RECIST at 4 weeks after PD-1 blockade initiation was significantly higher using MTV or TLG on 18F-FDG uptake than on CT. Multivariate analysis revealed that metabolic response by MTV or TLG at 4 weeks was an independent factor for response to PD-1 blockade treatment. Metabolic assessment by MTV or TLG was superior to morphological changes on CT for predicting the therapeutic response and survival at 4 weeks after PD-1 blockade.

The concordance rate of tumor response between confirmed overall objective response (OOR) by RECIST and that at 4 and 9 weeks after PD-1 blockade treatment are shown in Fig. 1. The concordance rate of PR (MTV: 57.9% vs. 21.1%, p = 0.044, TLG; 63.2% vs. 21.1%, p = 0.020) and PD (MTV: 78.9% vs. 47.3%, p = 0.002; TLG: 73.7% vs. 21.1%, p = 0.007) detected based on RECIST at 4 weeks after PD-1 blockade initiation was significantly higher in MTV or TLG than in CT (Fig. 1A). SUL max and SUL peak were significantly superior to CT in the concordance rate of PR at 4 weeks (63.2% vs. 21.1%, p = 0.020) whereas they were inferior to CT in the concordance rate of PD at 4 weeks (10.5% vs. 47.3%, p = 0.029) (Fig. 1A). Although the concordance rate of PR (94.7% vs. 73.6%, p = 0.179) and PD (92.8% vs. 71.4%, p = 0.325) confirmed based on RECIST at 9 weeks was higher in MTV or TLG than in CT, it was not statistically significant (Fig. 1B). Next, concordance rates according to treatment lines (first-line and second or more line settings) and histological type were also examined (Fig. C, online only). The concordance rate of PR and PD in the 18 F-FDG uptake (SUL max , MTV and TLG) at 4 weeks tended to be high in patients with non-AC compared to those with AC, without statistical significance (Figs. C1, C3). Moreover, the concordance rate of PR in the 18 F-FDG uptake (SUL max , MTV and TLG) at 4 weeks exhibited a significantly higher in first-line setting than in second line or more (Figs. C5, C7).
Out of 54 patients, 2 (3.7%) patients experienced pseudoprogression. One patient with confirmed PR based on RECIST, had PMD by MTV and TLG, PD by CT scan, and SMD by SUL max and SUL peak at 4 weeks after firstline pembrolizumab, because of markedly increased primary site. However, PMR was observed by MTV, TLG, SUV, and SUV on 18 F-FDG PET at 9 weeks, similar to PR by CT scan. Although the other patient with confirmed PR based on RECIST also experienced psuedoprogression within 4 weeks after nivolumab initiation as second line setting, the objective response by CT at 4 and 9 weeks exhibited SD, that by MTV and TLG showed SMD at 4 weeks and PMR at 9 weeks, and that by SUL max and SUL peak depicted PMR at 4 and 9 weeks.
Survival analysis according to 18 F-FDG uptake. The median follow-up period for all patients was 296 days (range 75-741). Forty patients experienced disease recurrence, and 21 died. The median PFS and OS were 174 days and not reached, respectively. Kaplan-Meier curves of PFS and OS according to CT and 18 F-FDG uptake at 4 and 9 weeks after PD-1 blockade among all patients are shown in Fig. 2. A significant difference in PFS and OS was identified between PMD and non-PMD defined according to 18 F-FDG uptake by MTV and TLG at 4 and 9 weeks, but not at 4 weeks but 9 weeks on PET by SUL max (Fig. 2). Next, the outcome of 38 patients with SD on CT scan at 4 weeks after PD-1 blockade initiation were analyzed according to metabolic response by 18 F-FDG uptake (PMD vs. non-PMD) (Fig. 3). A significant difference in PFS and OS was identified between PMD and non-PMD based on MTV (Fig. 3A) and TLG (Fig. 3B) at 4 and 9 weeks, but not on PET by SUL max (Fig. 3C). Results of the univariate and multivariate analyses are presented in Table 2. In multivariate analysis, MTV and TLG on 18 F-FDG uptake at 4 weeks after PD-1 blockade were confirmed as independent predictive factors.

Discussion
This prospective study compared CT from PET for the assessment of early response after PD-1 blockade monotherapy in advanced NSCLC patients. The concordance rate of PR and PD using RECIST at 4 weeks after PD-1 blockade was significantly higher using 18 F-FDG uptake by MTV or TLG than by using morphological changes on CT. In addition, PMD by MTV or TLG at 4 weeks could significantly predict worse survival after PD-1 blockade administration. At least 9 weeks after its administration, 18 F-FDG uptake by MTV or TLG accurately predicted the tumor response confirmed based on RECIST and survival after PD-1 blockade, compared to the morphological assessment by CT. In this study, we also found that the concordance rate of PR and PD detected based on RECIST at early phase after PD-1 blockade tended to be higher in patients with non-AC or first-line setting. In particular, metabolic response by MTV and TLG could differentiate responder from non-responder in 38 patients with SD on CT at 4 weeks after PD-1 blockade administration.
Recently, Park et al. 16 retrospectively evaluated early response assessment after immunotherapy using 18 F-FDG PET/CT in 24 advanced NSCLC patients. They presented the case of 5 patients with CMR or PMR who had a clinical benefit after two or three cycles of ICI treatment, whereas none of the 14 patients with PMD experienced any clinical benefit 16 . It was speculated that 5 patients with SMD needed meticulous follow-up because of varying clinical benefits 16 . Castello et al. 17 prospectively compared morphological and metabolic responses at 8 or 9 weeks after PD-1 blockade using 18 F-FDG PET/CT in 35 NSCLC patients. Although they assessed the metabolic response by the SUV value, 3 (75%) of 4 patients with PR had PMR or CMR, 14 (87%) of 16 patients with PD exhibited PMD, and SMD was observed in 4 (26%) of 15 patients with SD 17  www.nature.com/scientificreports/ metabolic response by PET could discriminate those with longer survival 17 . 18 F-FDG uptake based on PERCIST or immunotherapy-modified PERCIST accurately reflects the overall metabolic response and survival after PD-1 immunotherapy in NSCLC patients 4,14,18 . Currently, PD-L1 is considered a rough biomarker for the therapeutic prediction of PD-1 blockade, and promising markers such as TMB or tumor-infiltrating lymphocytes fail to identify the progression to ICI treatment. Thus, the presence of responders and their progression should be identified as early as possible after PD-1 blockade to predict long-term responders. In our study, 18 F-FDG uptake on PET yielded a higher predictive value www.nature.com/scientificreports/ to achieve PR at 4 weeks in first-line setting or histology with non-AC compared to CT. However, it remains unclear about its detailed mechanisms, thus, further investigation is warranted to elucidate the results of our study by using large -sample size. Rossi et al. 19 compared between 18 F-FDG and CT-based criteria as response assessment at 8 weeks after nivolumab in 48 NSCLC patients and reported a low overall concordance between CT-based and PET-based responses, but PMR assessed by PET predicted longer OS than CT-based PR. In their study, neither MTV nor TLG but SUV peak was used to assess the metabolic response, and early response was not investigated, as we did in this study. Our results indicated that the concordance rate by SUL max was apparently inferior to that by MTV or TLG, and there was no significant difference between metabolic response by SUL max and morphological response by CT. However, MTV or TLG was identified as a significant marker to predict the tumor response and survival in the early phase, such as 4 weeks after PD-1 blockade, compared to CT. Thus, SUL max on PET may not be suitable for early detection of ICI response. The present study includes the patients receiving PD-1 blockade as first-line and second-line or more setting, however, previous investigations discussed the therapeutic monitoring of 18 F-FDG PET after immunotherapy in patients with previously treated NSCLC 4,13,16 . The added value of our study demonstrated that the tumor response detected by early PET is useful for the therapeutic prediction of first-line pembrolizumab, moreover, the therapeutic significance of monitoring by early PET is also different according to the histological types of NSCLC. Nowadays, PD-1 blockade is generally established treatment as first-line setting against the advanced NSCLC without any drive gene mutations. Compared to previous studies reporting the monitoring by PET after ICIs administration, the data of first-line pembrolizumab in our study is helpful for our clinical practice. Moreover, our analysis was on a per patient basis and not on a per lesion basis. As several lesions in the same patients may depict different 18 F-FDG uptake, the therapeutic monitoring by 18 F-FDG uptake based on a per patient basis would be helpful for evaluating the response and survival of ICIs treatment.
Although several investigations demonstrated that baseline 18 F-FDG PET could predict the outcome of PD-1 immunotherapy in NSCLC patients, it is difficult to discriminate responders from non-responders or PD from non-PD patients by baseline 18 F-FDG uptake [20][21][22] . Because 18 F-FDG uptake was assessed according to different PET machines in individual institutions, it may not be consistent at baseline. Therefore, we did not explore the efficacy and survival of PD-1 blockade based on baseline 18 F-FDG uptake.
Recently, Lopci et al. 23 described the new guidelines of 18 F-FDG PET imaging during immunotherapy treatment in patients with solid tumors. To identify pseudoprogressive patients after immunotherapy, the refinement of standard response evaluation guidelines is needed, thus, immune-related response criteria (irRC), immune RECIST (iRECIST) and immune-modified RECIST (irRECIST) for solid tumors were proposed 23 . However, there are no established immunotherapy guidelines to categorize the therapeutic response of patients with pseudoprogressive disease (PPD). The interpretation of 18 F-FDG PET should take into account PPD during ICIs treatment. In the current study, we found two patients with PPD (3.7%; 2 of 54 patients). SUL max and SUL peak were useful to detect the exact response at 4 weeks after ICIs administration in one of 2 patients with PPD, whereas, it was difficult to identify the correct response at early phase by MTV and TLG on PET. In these two patients, 18 F-FDG PET imaging could detect the true response based on RECIST at 9 weeks after PD-1 blockade administration. However, it remains unclear whether 18 F-FDG uptake on PET could be helpful for the detection of PPD. In our study, the response evaluation by PET was investigated based on the RECIST, however, there are several new immunotherapy guidelines such as irRC, iRECIST and irRECIST. Although we tried to analyze our data using iRECIST, the results of response evaluation after immunotherapy were not different between RECIST and iRECIST (data not shown). Therefore, further investigation is warranted to establish a new response guideline during immunotherapy aside from the current guidelines.
There are several limitations in our studies. In this study, firstly, the SULs were not harmonized between the PET scanners. However, as these were devices of almost the same generation of the same manufacturer, we believe that changes before and after treatment of SULs, MTV, and TLG were captured to some extent in each facility. Our results warrant for large-scale multi-center research on harmonizing SULs of each PET scanner. Second, our sample includes heterogeneous populations such as AC and non-AC or first-line and second-line settings. As the concordance rate of objective response was different according to histology and treatment lines, further study should be focused on the selected patients. In the present study, 6 (20.6%) of 29 patients with AC yielded positive epidermal growth factor (EGFR) mutation and one patients had anaplastic lymphoma kinase (ALK)-echinoderm Figure 1. Concordance rate between Response by RECIST and tumor response at 4 weeks (A) and 9 weeks (B) after PD-1 blockade. (A) Among the 19 patients with PR based on RECIST, PMR at 4 weeks after PD-1 blockade by SUL max , SUL peak , MTV, and TLG was observed in 12 (63.2%), 12 (63.2%), 11 (57.9%), and 12 (63.2%), respectively, and CT at 4 weeks confirmed PR in 4 of 19 patients (21.1%). In the 19 patients with PD based on RECIST, PMD at 4 weeks by SUL max , SUL peak , MTV, and TLG was noted in 2 (10.5%), 2 (10.5%), 15 (78.9%), and 14 (73.7%), respectively, whereas CT at 4 weeks identified PD in 9 of 19 patients (47.3%). The predictive probability of PR and PD according to RECIST at 4 weeks after PD-1 blockade administration was significantly higher in MTV and TLG than in CT, whereas, the predictive probability of SD after its treatment was significantly higher in CT than in MTV and TLG. (B) Moreover, PMR at 9 weeks by SUL max , SUL peak , MTV, and TLG was observed in 15 (78.9%), 13 (68.4%), 18 (94.7%), and 18 (94.7%) of 19 patients with PR according to RECIST, respectively, and CT at 9 weeks confirmed PR in 10 of 14 patients (73.6%). Among the 19 patients with PD according to RECIST, PMD at 9 weeks by SUL max , SUL peak , MTV, and TLG was identified in 7 (50.0%), 6 (42.8%), 13 (92.8%), and 13 (92.8%), respectively, and CT at 9 weeks displayed PD in 10 of 14 patients (71.4%). The predictive probability of PR and PD according to RECIST at 9 weeks after PD-1 blockade administration was significantly higher in CT and SUL max than in MTV and TLG. *Statistically significant difference. Kaplan-Meier curve of PFS and OS according to CT and 18 F-FDG uptake at 4 and 9 weeks after PD-1 blockade initiation in all patients (n = 54). A significant difference in PFS, but not in OS, was noted between PD and non-PD defined according to CT at 4 and 9 weeks (A). A significant difference in PFS and OS was identified between PMD and non-PMD defined according to the 18 F-FDG uptake by SUL max (B) and SUL peak (C) at 9 weeks, but not at 4 weeks. A significant difference in PFS and OS was identified between PMD and non-PMD defined according to 18  www.nature.com/scientificreports/ microtubule-associated protein-like 4 (EML4) fusion gene. Five of 7 patients with these driver mutations were confirmed as PD in second-line setting (data not shown). The AC patients harboring EGFR mutations are low sensitive to ICs and were identified as lower 18 F-FDG uptake on PET than wild-type. 24 This may bias the different www.nature.com/scientificreports/ concordance rate between AC and non-AC. Finally, it is unknown if morphological assessment with RECIST is suitable modality to evaluate response to ICI therapy 4,13 . The aim of our study is to compare PET with CT in the assessment of tumor response at early phase after PD-1 blockade initiation. However, we believe that metabolic tumor response by PET may be better to predict the response and outcome of ICIs compared to CT. Moreover, the optimal detection by CT scan in PPD after ICIs initiation is difficult, but, the therapeutic monitoring by 18 F-FDG uptake on PET also seemed to yield some limitations for the early detection of PPD. Further study is warranted to develop the modality to confirm the presence of PPD at early phase after ICIs initiation.
In conclusion, metabolic assessment by MTV or TLG was useful in predicting an early therapeutic response and survival after PD-1 blockade, compared to morphological changes on CT, specially, in patients with non-AC or first-line setting. 18 F-FDG uptake may be a promising biomarker for predicting the therapeutic efficacy of ICIs; thus, it may contribute to individualized treatment planning in clinical practice.

Methods
Patients. This prospective study enrolled advanced NSCLC patients who received PD-1 blockade monotherapy at multiple institutions between January 2019 and October 2020. The inclusion criteria were (a) pathologically confirmed NSCLC; (b) candidate for PD-1 blockade monotherapy such as nivolumab, pembrolizumab or atezolizumab in first-, second-or more lines; (c) performance status (PS) on the Eastern Cooperative Oncology Group of 0-2; (d) 18 F-FDG PET/CT imaging scheduled within 4 weeks before the first cycle of PD-1 blockade monotherapy, and (e) possessing adequate organ functions. The exclusion criteria were (a) evidence of concurrent cancer, (b) uncontrolled diabetes mellitus, (c) interstitial pneumonia or pulmonary fibrosis, and (d) active infection requiring antibiotic therapy. Baseline 18 F-FDG PET/CT was performed as part of the disease evaluation workup before initiating PD-1 blockade monotherapy. Post-treatment PET/CT was needed at 4 and 9 weeks after the first cycle of PD-1 blockade. The protocol required that both pre-and post-treatment PET-CT be performed using the same scanner (Fig. A, online only).
This study was approved by the institutional review board (Saitama Medical University) and conducted according to the Declaration of Helsinki. All patients provided written informed consent before participation and were able to withdraw from the study at any time. This trial was registered in Japan Registry of Clinical Trials (jRCTs031180036) on 01/11/2018. PET imaging and data analysis. Patients fasted for at least 6 h before 18 F-FDG administration for PET, performed using a PET/CT scanner. Three-dimensional data acquisition was initiated 60.0 ± 8.1 min after FDG injection. Attenuation-corrected transverse images obtained with 18 F-FDG were reconstructed with the condi-  18 F-FDG uptake at 4 and 9 weeks in 38 patients with SD on CT scan at 4 weeks after PD-1 blockade initiation. A significant difference in PFS and OS was noted between PMD and non-PMD based on MTV (A) and TLG (B) at 4 and 9 weeks, but not SUL max at 4 and 9 weeks (C). A significant difference in the PFS and OS according to SUL peak at 9 weeks was observed between PMD and non-PMD, but no at 4 weeks (D).  CT for initial staging was performed with an intravenous contrast medium, and board-certified radiologists interpreted the images. We used RAVAT software (Nihon Medi-physics Co. Ltd., Japan) on a Windows workstation to semi-automatically calculate the maximum SUL (SUL max ), SUL peak , MTV, and TLG, defined as MTV multiplied by SUL mean , of each lesion using SUL thresholds obtained by the SUL in the liver VOI. Each threshold was defined as the average of 1.5 × SUL (SUL mean ) plus 2 × SD of SUL in the liver 26,27 . These SUL thresholds were the optimum values for generating a 3D VOI in which the whole tumor mass was completely enclosed in all cases using the CT image as the reference. Regions of activity other than tumors, including myocardium,  www.nature.com/scientificreports/ gastro-intestinal tracts, kidneys, and urinary tracts, were eliminated manually according to the orientation provided by the board-certified nuclear medicine physician. In this study, SULs between facilities and between devices were not harmonized.
Efficacy assessment. The confirmed tumor response was assessed according to the findings in CT imaging at 9 weeks interpreted using the Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST v. 1.1). PET-based tumor response was defined according to the PET Evaluation Criteria in Solid Tumors (PERCIST) guidelines 27 : complete metabolic response (CMR), complete resolution of 18 F-FDG uptake within the target lesion; partial metabolic response (PMR), ≥ 30% decrease in 18 F-FDG uptake in the target tumor; progressive metabolic disease (PMD), ≥ 30% increase in 18 F-FDG uptake in the target tumor or the advent of new 18 F-FDG avid lesions; stable metabolic disease (SMD), neither CMR, PMR, nor PMD.
Statistical analysis. Statistical analyses were performed using Student's t-test and χ 2 test for continuous and categorical variables, respectively. Statistical significance was set at p < 0.05. Correlations between SUL max , MTV, and TLG on 18 F-FDG uptake were analyzed using Pearson's rank test. Univariate and multivariate analyses of the relationship between scoring by 18 F-FDG uptake and different variables were performed using logistic regression analysis. Progression-free survival (PFS) was defined as the time from initial immunotherapy to disease progression or death; OS was defined as the time from initial immunotherapy to death from any cause. The Kaplan-Meier method was used to estimate survival as a function of time, and survival differences were analyzed using log-rank test. The progression of disease for survival analysis was defined as progression in imaging. Metabolic responses at 4 and 9 weeks after the injection of PD-1 blockade were evaluated according to the response on PET 25 . All statistical analyses were performed using GraphPad Prism software (v.8.0; GraphPad Software, San Diego, CA, USA) and JMP 14.0 (SAS Institute Inc., Cary, North Carolina, USA).

Data availability
The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.