Dual-time-point [18F]FDG-PET/CT as a prognostic biomarker in patients with pleural mesothelioma undergoing immunotherapy Authors: Solfrid Thunold, Eivor Hernes, Lars Tore Gyland Mikalsen, Saima Farooqi, Åsa Kristina Öjlert, Roslyn J Francis, Anna K Nowak, Weronika Maria Szejniuk, Søren Steen Nielsen, Susana Cedres, Marc Simo Perdigo, Jens Benn Sørensen, Carin Meltzer, Åslaug Helland, Eirik Malinen, Vilde Drageset Haakensen Published: 2026 ABSTRACT Background[18F]FDG PET/CT provides a non-invasive assessment of tumour glucose metabolism. While conventional PET/CT captures a single time point, dual-time-point PET/CT (DTP-PET) evaluates metabolic dynamics by acquiring images at two defined times after injection. This study investigated whether the change in [18F]FDG uptake between 60 and 120 min—retention index (RI)—predicts clinical outcomes in patients with pleural mesothelioma (PM) receiving immunotherapy. Patients from the NIPU trial (NCT04300244) underwent DTP-PET at baseline (n = 50) and/or week-5 (n = 45), with 42 completing both. Peak SUV was measured in a 2-cm spherical volume of interest centred on the most avid lesion at 60 and 120 min, and RI was calculated as the percentage change between the two. Survival (OS, PFS) was analysed using Kaplan–Meier curves and Cox proportional hazards models based on tertiles of RI and SUV60min. Objective response and disease control were defined by mRECIST and iRECIST. Group comparisons used the Wilcoxon rank-sum test.ResultsAt week-5, RI tertiles showed stepwise separation for both OS (p = 0.038) and PFS (p = 0.031), with the lowest tertile associated with the most favourable outcomes. SUV60min tertiles were also significantly associated with OS (p = 0.042) and PFS (p = 0.043), though OS exhibited a non-monotonic pattern in which the middle tertile had the poorest survival. Objective responders displayed significantly lower RI and SUV60min at week-5. Baseline RI tertiles were not associated with OS or PFS.ConclusionsWeek-5 [18F]FDG DTP-PET suggests prognostic value in PM patients receiving immunotherapy. Both RI and SUV60min were associated with objective response. RI showed a consistent stepwise association with survival, whereas SUV60min showed a non-monotonic relationship for OS. These findings are exploratory and require validation in larger cohorts. Further studies are warranted to explore underlying biological mechanisms.Trial registrationClinicalTrials.gov NCT04300244. Registered 2020-03-09. https://clinicaltrials.gov/study/NCT04300244Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s13550-026-01469-x. Background Pleural mesothelioma (PM) is an aggressive malignancy that lacks reliable non-invasive biomarkers, a challenge that has become increasingly relevant with the introduction of immune checkpoint inhibition (ICI) [1–3]. The prognostic factors relevant to ICI are complex and can differ from those established for chemotherapy [4]. While previous studies on PM have demonstrated that high [18F]fluorodeoxyglucose (FDG) uptake is associated with a poor prognosis [5, 6], our earlier analysis of baseline [18F]FDG positron emission tomography (PET)/computer tomography (CT) scans in the NIPU study did not confirm this association in PM patients receiving immunotherapy [7]. Similar findings have been reported in other malignancies [8, 9], highlighting the importance of gaining a deeper understanding of the biological mechanisms that govern the uptake characteristics of [18F]FDG relevant to ICI. [18F]FDG-PET/CT is a non-invasive imaging modality that provides spatially resolved insights into tumour glucose metabolism. While conventional single-time-point PET captures only a snapshot, dual-time-point [18F]FDG-PET (DTP-PET) acquires images at two defined intervals post-injection, enabling assessment of uptake dynamics. [18F]FDG DTP-PET has demonstrated prognostic value in malignancies, including breast, non-small cell lung cancer (NSCLC), pancreatic cancer, and lymphoma, where a high increase in [18F]FDG from early to delayed imaging has been associated with worse outcome [10–17]. However, its utility in PM and in other malignancies treated with ICIs remains largely unexplored. The therapeutic efficacy of ICI depends on the presence of non-cancerous cells within the tumour microenvironment (TME), particularly immune cells [18–20], which also contribute to overall [18F]FDG uptake [21, 22]. Immune cells, especially myeloid-derived populations, exhibit high glycolytic activity. Consequently, the [18F]FDG signal within a tumour may reflect contributions from both cancer cells and immune cells, each expected to display distinct temporal uptake patterns. Therefore, metrics derived from [18F]FDG DTP-PET may capture metabolic dynamics of both cancer and immune cells, providing a non-invasive, whole-tumour approach for assessing ICI-induced metabolic alterations and treatment response. This study investigates whether [¹⁸F]FDG DTP-PET, quantified as the difference in [18F]FDG uptake between 60 and 120 min using the retention index (RI), can predict outcomes in patients with PM undergoing immunotherapy. We primarily evaluated week-5 [18F]FDG DTP-PET as an early-on-treatment prognostic marker in PM patients receiving immunotherapy, with baseline DTP-PET as a secondary exploratory analysis. Methods Participants In the NIPU trial (NCT04300244) [23, 24], 118 patients were randomised 1:1 to nivolumab and ipilimumab ± UV1 telomerase peptide vaccination in 2nd line. The current analysis includes a predefined DTP-PET sub-study cohort from the Oslo and Aalborg centres. 240 mg of nivolumab was administered every second week and 1 mg/kg ipilimumab every sixth week until progression, intolerable side effects, or for a maximum of two years. Patients randomised to the UV1 vaccine (arm A) received 8 vaccinations during the first 3 months of treatment. Histology subtype classifications were collected from the local pathology report. Programmed cell-death ligand 1 (PD-L1) expression was centrally assessed at the Department of Pathology, Oslo University Hospital, using immunohistochemistry on archived Formalin-fixed paraffin-embedded tumour tissue with a mouse monoclonal antibody for staining (405.9A11, Cell Signaling Technology, Danvers, USA). The trial was approved by the regional ethics committee (20/47804) and the site ethics committees, and was conducted following the Declaration of Helsinki of the World Medical Association and the ICH E6 for Good Clinical Practice. All patients provided written informed consent. [18F]FDG-PET/CT Patients underwent an [18F]FDG PET/CT examination at baseline (median 12 days prior to treatment initiation) and again five weeks after treatment initiation (week-5). After administering 3.0-3.5 MBq/kg [18F]FDG, a standard 60-minute post-injection scan from the skull base the mid-thigh was performed (60 min scan), followed by an additional 120-minute post-injection scan limited to the thoracic region (120 min scan) (Fig. 1). Images were obtained on Siemens Biograph mCT40 (Oslo) and Biograph mCT64 (Aalborg) with a scan speed of 2 min/bed position at the 60-minute scan and 3 min/bed position at the 120-minute scan for the Biograph mCT40, and with a scan speed of 0.6–1 mm/s at the 60-minute and 120-minute scans for the Biograph mCT64. Images were reconstructed using time-of-flight and point spread function modelling (TOF + PFS), 2 iterations, 21 subsets, and a Gaussian filter of 4 mm (Oslo) and 5 mm (Aalborg), on a 400 × 400 matrix. Only patients included in Oslo and Aalborg were eligible for this analysis because [¹⁸F]FDG DTP-PET was implemented at these two centres, and both scanners were accredited by the European Association of Nuclear Medicine Research Ltd (EARL) harmonisation programme and complied with the EARL Standards 2 specifications [25, 26]. Fig. 1Timeline of injection of [18F]FDG and protocol of the dual-time-point PET protocol. The scans were performed at baseline and repeated after 5 weeks of treatment. 60-minute scans were whole-body scans, while 120-minute scans only included the thoracic region, as indicated. Abbreviations: FDG, fluorodeoxyglucose; PET, positron emission tomography Timeline of injection of [18F]FDG and protocol of the dual-time-point PET protocol. The scans were performed at baseline and repeated after 5 weeks of treatment. 60-minute scans were whole-body scans, while 120-minute scans only included the thoracic region, as indicated. Abbreviations: FDG, fluorodeoxyglucose; PET, positron emission tomography For quantitative analysis, only lesions that were within the field of view on both the 60-minute and 120-minute images were included. The most FDG-avid lesion was identified on the baseline 60-minute scan and selected as the target lesion for analysis. One lesion per patient was analysed; in all cases, the selected lesion was located within the thorax. A spherical volume of interest (VOI) with a fixed diameter of 2 cm was positioned over the region demonstrating the highest uptake intensity to derive peak standardised uptake value (SUVpeak), in line with the PET Response Criteria in Solid Tumours (PERCIST) recommendation to use SUVpeak as a quantitative metric [27], and was recorded as SUV60min. A slightly larger VOI than the standard PERCIST SUVpeak VOI was used to reduce potential inter-reader variability and enhance measurement robustness, particularly important in pleural mesothelioma, which is characterised by irregular growth patterns and challenging anatomical delineation. The identical VOI (same size, shape, and anatomical position) was then copied to the corresponding anatomical location on the baseline 120-minute scan and on the week 5 scans, to ensure consistent sampling across time points. In a few cases where tumour thickness was insufficient to accommodate a 2 cm spherical VOI, the VOI diameter was reduced to ensure complete inclusion within tumour tissue and to minimise contamination from adjacent structures. PET analyses were conducted by two investigators; S.T. identified the target lesion and defined the VOIs, and E.H. reviewed and verified target lesions, VOI placements and measurements. RI was calculated as follows: Retention index (RI) = (( SUV120min – SUV60min) / SUV60min) x 100. Radiological assessment A CT with intravenous contrast was performed at baseline and every 6 weeks for the first 12 months following randomisation (± 7 days) and every 12 weeks thereafter (± 7 days), with additional scans performed as clinically indicated. The local site performed a radiological assessment according to the modified Response Criteria in Solid Tumours (mRECIST) [28] for mesothelioma and the immune RECIST (iRECIST) [29]. Objective response was defined as patients whose best overall response was complete response (CR) or partial response (PR) as compared to stable disease (SD) and progressive disease (PD). Disease control included patients with CR, PR, or SD compared to PD as their best overall response. Statistics Statistical analyses were performed in R version 4.2.2 (2022-10-31). Due to the skewed distribution of the data, non-parametric methods were employed. Spearman’s rank correlation (ρ) was performed for associations between PET metrics (RI, SUV60min, and SUV120min). Comparisons of PET metrics across clinical/pathological subgroups (objective response, disease control, PD-L1, histology) were tested with the Wilcoxon rank-sum test. Patients were stratified into tertiles (T1 low, T2 intermediate, T3 high) of RI and SUV60min. Objective response was summarised across week-5 PET tertiles and according to whether values increased or decreased from baseline. Survival was analysed with Kaplan–Meier curves with a log-rank test across tertiles; when the log-rank test was significant, pairwise unadjusted log-rank tests were performed. Cox proportional hazards (PH) models with tertiles as a categorical predictor (T1 as reference) provided hazard ratios (HRs) with two-sided 95% CIs; Cox p-values for the two reported comparisons (T2 vs. T1 and T3 vs. T1) were unadjusted. PET variables were modelled as tertiles in Cox PH analyses due to non-linear relationships, which precluded robust continuous modelling. Progression-free survival (PFS) was defined as the interval from randomisation until progression or death by any cause, and overall survival (OS) as the interval from randomisation until death by any cause. As a landmark sensitivity analysis, survival was additionally analysed from the date of the week-5 PET scan for week-5 SUV60min and RI. All tests were two-sided with α = 0.05. Results Fifty-three patients with EARL2-accredited [¹⁸F]FDG DTP-PET scans were eligible for this analysis, with 50 undergoing baseline DTP-PET, 45 at week-5, and 42 completing DTP scans at both time points; all patients had a standard baseline [¹⁸F]FDG PET/CT scan, even if delayed imaging (DTP) was not acquired. Among the eight patients without week-5 DTP-PET, two died before the examination and one did not undergo imaging due to clinical deterioration. The remaining scans were not performed due to technical or logistical reasons, or were excluded due to lack of EARL2 accreditation. The patient cohort with [¹⁸F]FDG DTP-PET scan at baseline was predominantly male (84%) with a median age of 71 years, epithelioid histology (82%), and PD-L1 negative status (68%). PR was observed in 20% of patients, with no complete responses. Full baseline demographics are provided in Table 1. At baseline, RI correlated moderately with SUV120min (ρ = 0.52) but only weakly with SUV60min (ρ = 0.23). At week-5, correlations were stronger for both SUV60min (ρ = 0.57) and SUV120min (ρ = 0.72), explaining 32% and 52% of the shared variance, respectively (Supplementary Figure S1). RI tertiles were associated with OS and PFS with stepwise separation at week-5 (OS p = 0.038; PFS p = 0.031, Fig. 2), but not on baseline scans (OS p = 0.47; PFS p = 0.67; Supplementary Fig. S2). For tertiles identified at week-5, median OS was 23.6 months (95% CI, 6.5–30.5) in T1 (low), 13.6 (4.2–17.7) in T2 (intermediate), and 7.2 (5.3–14.8) in T3 (high). Median PFS was 7.0 months (1.7–15.4) in T1, 2.9 (1.6–7.1) in T2, and 3.9 (1.4–5.4) in T3. In unadjusted pairwise comparisons, OS differed for T2 vs. T1 (p = 0.044) and T3 vs. T1 (p = 0.016), and PFS differed for T3 vs. T1 (p = 0.014). See Supplementary Table S1. Fig. 2Kaplan–Meier curves at week-5 stratified by SUV60min and RI tertiles. Panel A: SUV60min tertiles; Panel B: RI tertiles. Within each panel, PFS (left) and OS (right) are shown with risk tables. Groups are labelled T1 (Low), T2 (Intermediate), and T3 (High), defined by ascending rank tertiles across the week-5 cohort for SUV60min and RI separately. Log-rank p-values compare all three tertiles. SUV60min denotes [¹⁸F]FDG uptake at 60 min post-injection. Abbreviations: SUV, standardised uptake value; RI, retention index; PFS, progression-free survival; OS, overall survival; FDG, fluorodeoxyglucose Kaplan–Meier curves at week-5 stratified by SUV60min and RI tertiles. Panel A: SUV60min tertiles; Panel B: RI tertiles. Within each panel, PFS (left) and OS (right) are shown with risk tables. Groups are labelled T1 (Low), T2 (Intermediate), and T3 (High), defined by ascending rank tertiles across the week-5 cohort for SUV60min and RI separately. Log-rank p-values compare all three tertiles. SUV60min denotes [¹⁸F]FDG uptake at 60 min post-injection. Abbreviations: SUV, standardised uptake value; RI, retention index; PFS, progression-free survival; OS, overall survival; FDG, fluorodeoxyglucose Table 1Patient baseline demographics and characteristicsBaseline DTP, N = 50Week 5 DTP, N = 45 Age – median (IQR) 71 (67, 75)71 (67, 74) Sex - N (%) Female8 (16%)5 (11%)Male42 (84%)40 (89%) Treatment arm- N (%) +UV125 (50%)21 (47%)-UV125 (50%)24 (53%) ECOG Performance status – N (%) ECOG 013 (26%)14 (31%)ECOG 137 (74%)31 (69%) PD-L1 status – N (%) PD-L1 negative34 (68%)30 (67%)PD-L1 positive8 (16%)7 (16%)Missing8 (16%)8 (18%) Histology – N (%) Epithelioid41 (82%)39 (87%)Non-epithelioid9 (18%)6 (13%) RI – median (IQR) 18 (9, 27)19 (10, 25)SUV60 min – median (IQR)9.8 (6.4, 13.0)8.7 (4.7, 10.7)SUV120 min – median (IQR)11.6 (6.8, 15.0)9.8 (5.2, 13.1)Baseline demographics and characteristics for patients with a baseline [¹⁸F]FDG DTP-PET (N = 50) and those with a week-5 [¹⁸F]FDG DTP-PET (N = 45); 42 patients underwent both scans. The week-5 cohort partially overlapped with the baseline cohort, as three patients did not undergo baseline DTP-PET. SUV60min and SUV120min denote [¹⁸F]FDG uptake at 60 and 120 min, respectively. Abbreviations: UV1, UV1 telomerase peptide vaccine; ECOG, Eastern Cooperative Oncology Group performance status; PD-L1, Programmed Cell-Death Ligand 1; SUV, standardised uptake value; FDG, fluorodeoxyglucose; DTP-PET, dual-time-point positron emission tomography Table 1Patient baseline demographics and characteristicsBaseline DTP, N = 50Week 5 DTP, N = 45 Age – median (IQR) 71 (67, 75)71 (67, 74) Sex - N (%) Female8 (16%)5 (11%)Male42 (84%)40 (89%) Treatment arm- N (%) +UV125 (50%)21 (47%)-UV125 (50%)24 (53%) ECOG Performance status – N (%) ECOG 013 (26%)14 (31%)ECOG 137 (74%)31 (69%) PD-L1 status – N (%) PD-L1 negative34 (68%)30 (67%)PD-L1 positive8 (16%)7 (16%)Missing8 (16%)8 (18%) Histology – N (%) Epithelioid41 (82%)39 (87%)Non-epithelioid9 (18%)6 (13%) RI – median (IQR) 18 (9, 27)19 (10, 25)SUV60 min – median (IQR)9.8 (6.4, 13.0)8.7 (4.7, 10.7)SUV120 min – median (IQR)11.6 (6.8, 15.0)9.8 (5.2, 13.1)Baseline demographics and characteristics for patients with a baseline [¹⁸F]FDG DTP-PET (N = 50) and those with a week-5 [¹⁸F]FDG DTP-PET (N = 45); 42 patients underwent both scans. The week-5 cohort partially overlapped with the baseline cohort, as three patients did not undergo baseline DTP-PET. SUV60min and SUV120min denote [¹⁸F]FDG uptake at 60 and 120 min, respectively. Abbreviations: UV1, UV1 telomerase peptide vaccine; ECOG, Eastern Cooperative Oncology Group performance status; PD-L1, Programmed Cell-Death Ligand 1; SUV, standardised uptake value; FDG, fluorodeoxyglucose; DTP-PET, dual-time-point positron emission tomography Patient baseline demographics and characteristics Baseline demographics and characteristics for patients with a baseline [¹⁸F]FDG DTP-PET (N = 50) and those with a week-5 [¹⁸F]FDG DTP-PET (N = 45); 42 patients underwent both scans. The week-5 cohort partially overlapped with the baseline cohort, as three patients did not undergo baseline DTP-PET. SUV60min and SUV120min denote [¹⁸F]FDG uptake at 60 and 120 min, respectively. Abbreviations: UV1, UV1 telomerase peptide vaccine; ECOG, Eastern Cooperative Oncology Group performance status; PD-L1, Programmed Cell-Death Ligand 1; SUV, standardised uptake value; FDG, fluorodeoxyglucose; DTP-PET, dual-time-point positron emission tomography At week-5, SUV60min tertiles also showed overall differences (OS p = 0.042; PFS p = 0.043). Median OS was 20.2 months (7.0–27.7) in T1, 6.9 (4.2–14.6) in T2, and 17.3 (6.2–23.0) in T3; median PFS was 7.1 (1.6–15.4), 3.4 (1.8–5.5), and 2.7 (1.4–5.4) months in T1, T2, and T3, respectively. In unadjusted pairwise comparisons, OS differed for T2 vs. T1 (p = 0.020), and PFS differed for T2 vs. T1 (p = 0.028) and T3 vs. T1 (p = 0.041); no other pairs were significant. Baseline SUV60min tertiles, in contrast, were not significantly associated with OS or PFS. See Fig. 2 and Supplementary Table S1. For context, survival analyses of baseline SUV have been reported by Thunold et al. (2024) [7], based on a broader cohort with available baseline imaging, regardless of whether dual-time-point scans were available. Results from univariate Cox PH models at week-5 were broadly consistent with the unadjusted pairwise log-rank tests (Supplementary Figure S3). In landmark sensitivity analyses, with survival time defined from the date of the week-5 PET scan, both KM-curves and univariable Cox PH analyses yielded comparable results for OS and PFS (Supplementary Figures S4 and S5). For PFS, eight patients were excluded from the landmark analyses due to radiological progression prior to the week-5 PET scan. To further explore the biological correlates of the PET metrics, we analysed their associations with treatment response and histopathological features at week-5 (Fig. 3). Patients who achieved an objective response had significantly lower metabolic activity and retention compared with non-responders, as reflected by all three PET metrics (RI: p = 0.0077; SUV60min: p = 0.0015; SUV120min: p = 0.00085). At week-5, 7 of 9 objective responders were in the lowest SUV60min tertile, with 1 in the intermediate tertile and 1 in the highest tertile. For RI, 6 of 9 responders were in the lowest tertile and 3 in the intermediate tertile, with none in the highest tertile. Within these tertiles, responders were mostly found among patients who exhibited a decrease in SUV60min and RI from baseline (Supplementary Table S2). None of the week-5 PET metrics differed according to disease control, histology (epithelioid vs. non-epithelioid) or PD-L1 status. For completeness, we also assessed baseline RI in relation to objective response, disease control, histology, and PD-L1. Baseline RI was higher in PD-L1–positive than PD-L1–negative tumours (p = 0.04); otherwise, no baseline associations were observed, including RI vs. objective response (p = 0.14; Fig. 4). Fig. 3Comparison of RI, SUV60min, and SUV120min by clinical and pathological subgroups at week-5. Panels: top left Objective Response; bottom left Disease Control; top right histology (epithelioid vs. non-epithelioid); bottom right PD-L1 (positive vs. negative; 8 missing). SUV60min and SUV120min represent [¹⁸F]FDG uptake at 60- and 120-minute post-injection, respectively. Abbreviations: DCR, disease control rate; RI, Retention Index; SUV, standardised uptake value; PD-L1, programmed cell death-ligand 1; FDG, [¹⁸F]fluorodeoxyglucose Comparison of RI, SUV60min, and SUV120min by clinical and pathological subgroups at week-5. Panels: top left Objective Response; bottom left Disease Control; top right histology (epithelioid vs. non-epithelioid); bottom right PD-L1 (positive vs. negative; 8 missing). SUV60min and SUV120min represent [¹⁸F]FDG uptake at 60- and 120-minute post-injection, respectively. Abbreviations: DCR, disease control rate; RI, Retention Index; SUV, standardised uptake value; PD-L1, programmed cell death-ligand 1; FDG, [¹⁸F]fluorodeoxyglucose Fig. 4Distribution of RI by clinical and pathological subgroups at baseline. Panels: (A) Objective Response, (B) Disease Control, (C) histology (epithelioid vs. non-epithelioid), (D) PD-L1 (negative vs. positive). Eight patients had missing PD-L1 status. Abbreviations: RI, retention index; PD-L1, programmed cell death-ligand 1 Distribution of RI by clinical and pathological subgroups at baseline. Panels: (A) Objective Response, (B) Disease Control, (C) histology (epithelioid vs. non-epithelioid), (D) PD-L1 (negative vs. positive). Eight patients had missing PD-L1 status. Abbreviations: RI, retention index; PD-L1, programmed cell death-ligand 1 Re-running the analyses in the epithelioid subset yielded similar results. Discussion In this prospective imaging sub-study of patients with PM receiving nivolumab–ipilimumab +/- UV1 telomerase vaccine, on-treatment (week 5) [¹⁸F]FDG DTP-PET was prognostic, whereas baseline DTP-PET showed no significant associations with survival. Both RI and SUV60min at week 5 were associated with OS and PFS, and patients achieving an objective response had significantly lower metabolic activity and retention across all PET metrics (RI, SUV60min, SUV120min). At week-5, [¹⁸F]FDG DTP-PET demonstrated a distinct prognostic pattern. RI showed a monotonic association with survival, with higher values associated with shorter PFS and OS. In contrast, SUV60min exhibited a non-monotonic pattern for OS: the intermediate tertile (T2) had the poorest OS, whereas OS in the highest tertile (T3) was closer to the lowest tertile (T1); PFS was similar between T2 and T3. Despite strong correlations between week-5 RI and SUV measures, RI’s monotonic association with survival versus SUV60min’s non-monotonic pattern may indicate that RI provides complementary prognostic information beyond early uptake alone. The apparent superiority of SUV60min tertile T3 over T2 for OS at week-5 is not readily explained. One possible explanation is an early [18F]FDG “flare” after starting ICI, occasionally observed in responders, which could transiently increase SUV uptake and result in apparently higher SUV values, despite favourable outcome [30]. In our cohort, however, the only patient with an objective response in T3 actually exhibited a decrease in SUV60min from baseline to week-5. The interpretation of response by mRECIST is inherently uncertain, with a risk of misclassification of responders and non-responders [31]. The observed non-monotonic pattern may therefore reflect a combination of biological heterogeneity, e.g., changes in tumour–immune balance, or small-number variability, highlighting the need for replication in larger cohorts. Consistent with this, the distribution of objective responders across week-5 SUV60min and RI tertiles highlights heterogeneous metabolic trajectories. Most responders were in the lowest tertiles, but a few had higher SUV60min or intermediate RI. One patient classified as a pseudo-progressor had an RI in the upper range of T2, supporting that RI alone may not reliably distinguish pseudo-progression from true progression [32]. These findings indicate that, although low SUV60min and RI generally predict objective response, individual metabolic trajectories can vary. Notably, no clear associations were observed between week-5 PET metrics and disease control, likely because SD in the context of immunotherapy represents a heterogeneous group, which attenuates differences compared with objective response [33, 34], further compounded by the limitations of mRECIST [31]. Malignant lesions typically show increased retention of [18F]FDG over time, whereas benign processes usually plateau or decline [35, 36]. The sustained retention in malignant cells is driven by complex biological mechanisms, including increased glucose transporter expression in the cell membrane, enhanced intracellular phosphorylation by hexokinase, and reduced dephosphorylation. The latter, caused by low glucose-6-phosphatase activity, limits the ability of phosphorylated [18F]FDG to be dephosphorylated and transported out of the cell, leading to entrapment since it cannot proceed through subsequent metabolic steps [17, 36, 37]. Although immune cells within TME can contribute to the [18F]FDG uptake [21], their uptake patterns are less well studied. Mononuclear cells have demonstrated a more rapid washout of [18F]FDG compared to cancer cells, potentially related to relatively higher glucose-6-phosphatase activity. How the tumour-to-immune cells ratio within the TME affects the [18F]FDG uptake dynamics remains unclear. A low post-treatment RI could be consistent with a relatively more favourable metabolic state, for example, reflecting a lower burden of viable tumour cells and/or a greater relative contribution from immune and other non-malignant cells. Conversely, a persistently high RI may suggest ongoing dominant tumour metabolism despite immunotherapy. These interpretations are exploratory and remain speculative, as we did not directly characterise the TME or measure tumour cell death; further studies integrating DTP PET with detailed tissue analyses are required to clarify the underlying mechanisms. Previous [18F]FDG DTP-PET studies in other malignancies, none involving ICI, have linked a high baseline RI to poorer outcomes [10–16]. Studies have also linked biological characteristics of tumours with [18F]FDG DTP-PET [38, 39]. In one preoperative NSCLC study, a pronounced increase in [18F]FDG uptake between early and delayed scans was associated with PD-L1 expression and CD8 + tumour-infiltrating lymphocytes (TILs) [40]. Since PD-L1 expression and CD8 + TILs have been linked to a better immunotherapy response in NSCLC, a high RI might similarly serve as a marker of immunotherapy sensitivity. In our study, baseline RI was likewise higher in PD-L1 positive cases. However, this finding should be interpreted with caution due to the high proportion of patients with unknown PD-L1 status, the small number of PD-L1-positive cases and the uncertain relationship between PD-L1 and immunotherapy response in PM [2]. At baseline, RI was not significantly associated with survival. Although baseline RI tertiles showed numerical separation in OS (T1 > T3), this did not reach statistical significance, likely reflecting limited power and biological heterogeneity. Objective responders also tended to have lower baseline RI without reaching statistical significance. Overall, these findings are consistent with RI being potentially more informative as an on-treatment marker than at baseline, although interpretation is limited by small sample size. This study has several limitations that should be considered when interpreting these findings. The small sample size and limited number of events reduce statistical power and limit the generalisability of the findings. As such, all analyses should be considered exploratory and hypothesis-generating, and require confirmation in larger cohorts. Continuous variables were not included in Cox PH models due to non-linear behaviour, and tertile-based categorisation has inherent limitations. Multivariable adjustment was not performed due to the limited sample size and low number of events in certain subgroups. The interpretation of PD-L1 association is limited by a higher proportion of missing data and low number of PD-L1-positive cases. PET analyses were based on a single, manually defined VOI over the most FDG-avid lesion at baseline, which was copied to subsequent scans. While this approach improves consistency within the study, it may be difficult to reproduce and does not capture inter-lesional heterogeneity, which may be particularly relevant in PM given its complex tumour growth pattern and known heterogeneity. Biological interpretation of RI remains speculative, as tissue-based analyses of the tumour microenvironment were not included. Finally, the clinical implementation of dual-time-point PET is limited by increased scan time and logistical demands, and larger prospective studies are needed to evaluate its added value over standard imaging, as well as feasibility and cost–benefit considerations. Nevertheless, our findings suggest that [18F]FDG DTP-PET after initiation of immunotherapy may provide prognostic information and capture metabolic changes reflecting early immune responses. As this study focused solely on SUV and RI, future investigations could incorporate additional PET metrics, such as metabolic tumour volume (MTV) and total lesion glycolysis (TLG). Radiomics and machine-learning approaches could further characterise the complex spatial and temporal patterns of tumour metabolism [41–43]. Integrating dual-time-point PET with tissue-based analyses of the TME, including molecular and cellular features, could provide deeper insights into the biological mechanisms underlying early metabolic changes in PM treated with immunotherapy. Conclusions Week-5 [18F]FDG DTP-PET may provide prognostic information in patients with PM undergoing immunotherapy. RI tertiles were associated with survival in a stepwise manner, with low RI linked to more favourable outcomes. SUV60min tertiles were also associated with survival, with patients in the lowest tertile having the best outcome, while the middle tertile showed unexpectedly worse outcomes than the highest tertile for OS. Patients achieving objective response had lower RI and SUV values at week-5, suggesting an association with treatment response. Overall, on-treatment DTP-PET appeared to be more prognostic than baseline scans, potentially reflecting treatment-induced metabolic alterations. These exploratory findings require confirmation in larger, independent cohorts, and studies integrating [18F]FDG DTP-PET with tumour microenvironment analyses are needed to elucidate the underlying biological mechanisms. Supplementary Information Supplementary Material 1. Supplementary Material 1.