Pseudoprogression in neoadjuvant immunotherapy for resectable NSCLC: a PET–CT-based study Authors: Y Pan, J Chen, H Lin, L-B Tang, H-H Yan, J-X Zeng, S Dong, L-X Yan, Y Yang, D-J Li, Y-L Wu Published: 2026 ABSTRACT AbstractBackgroundNeoadjuvant chemoimmunotherapy is standard for resectable non-small-cell lung cancer (NSCLC) but response assessment is challenging due to pseudoprogression, which may lead to unnecessary cancellation of surgery. We characterized pseudoprogression and developed a clinical strategy to distinguish it from true progression after neoadjuvant immune checkpoint inhibitor (ICI) therapy.Patients and methodsWe retrospectively analyzed data from 283 patients with resectable NSCLC receiving neoadjuvant ICI (n = 235) or tyrosine kinase inhibitor (n = 48) therapy from January 2021 to December 2023. Positron emission tomography–computed tomography (PET–CT) scans were acquired at baseline and preoperatively. Responses were evaluated using RECIST 1.1 and European Organisation for Research and Treatment of Cancer (EORTC) criteria. Predictors of pseudoprogression were evaluated using multivariable logistic regression. Model performance was assessed in an independent validation cohort (n = 95).ResultsBy RECIST 1.1, 26 of 283 patients (9.2%) had progressive disease (PD), all in the ICI cohort. By EORTC criteria, the number with PD increased to 71 of 283 (25.1%). Despite PD, all 26 patients underwent R0 resection; complete pathological response (CPR) and major pathological response (MPR) rates were 53.8% and 65.3% compared with 30.0% and 49.0%, respectively, in patients without PD. Among 71 with progressive metabolic disease, 69 (97.2%) achieved R0; CPR and MPR were 38.0% and 54.9%, respectively. Primary-tumor progression had lower MPR than nodal or distant progression (23.1%, 58.2%, and 73.3%, respectively, P = 0.020). Pseudoprogression occurred in 7.2% based on CT, increasing to 16.2% with PET–CT. A ≥53.1% reduction in primary-tumor maximum standardized update value (SUVmax) identified pseudoprogression [area under the curve (AUC) 0.865, sensitivity 94.7%, specificity 74.1%]. A multivariable model combining change in SUVmax, longest diameter change, smoking status, and histology showed excellent discrimination (AUC 0.935) and remained robust in validation (AUC 0.859).ConclusionsImaging-defined progression after neoadjuvant ICI may reflect pseudoprogression, most commonly nodal or distant lesions without primary tumor progression. A validated model incorporating primary-tumor metabolic response and clinical features improves discrimination of pseudoprogression and may support decision making to avoid inappropriate cancellation of potentially curative resection.Key words: neoadjuvant immunotherapy, non-small-cell lung cancer, pseudoprogression, response assessment Introduction Approximately 25% of non-small-cell lung cancer (NSCLC) cases are detected at an early stage through low-dose chest computed tomography (CT) screening.1 The benefit of neoadjuvant chemotherapy is limited, providing only a 5% improvement in 5-year survival compared with surgery alone in resectable disease.2 Recent randomized phase III trials have shown that neoadjuvant immune checkpoint inhibitor (ICI) therapy improves the pathologically documented response (pathological response) rate in patients with resectable NSCLC compared with neoadjuvant chemotherapy.3, 4, 5, 6, 7 Approximately 17%-25% of patients achieve a complete pathological response (CPR), and 30%-48% have a major pathological response (MPR) after a neoadjuvant ICI regimen, leading to improved event free survival (EFS). However, 3.8%-6.8% of patients experience tumor progression after neoadjuvant immunotherapy,3, 4, 5, 6 which is the second leading cause for the cancellation of surgery. Neoadjuvant treatment response is generally assessed using RECIST. However, patients receiving ICIs may present with unique response patterns, such as pseudoprogression, hyperprogression, and dissociated response, which are different from those seen with chemotherapy and targeted therapies. Pseudoprogression is characterized by an initial increase in tumor size or the development of new lesions, followed by a subsequent decrease in tumor burden due to immune cell infiltration within the tumor. This phenomenon has been reported in 2%-8% of patients with advanced NSCLC treated with ICIs.8 In the neoadjuvant setting, pseudoprogression manifests as changes in tumor size observed on CT scans that are inconsistent with the pathological response. Progressive disease (PD) on CT imaging can still show pathological regression. To address this discrepancy, immune-related RECIST have been proposed, requiring continued treatment and additional follow-up to confirm progression after an initial increase in tumor size. However, for early-stage NSCLC, decisions regarding surgery must be made within 6 weeks after the completion of neoadjuvant therapy, leaving no time for subsequent scans. Therefore, accurate assessment of the neoadjuvant immunotherapy response and identification of pseudoprogression are critical to avoid the cancellation of potentially curative surgical resections. [18F]-2-fluoro-2-deoxy-d-glucose (18F-FDG) positron emission tomography–computed tomography (PET–CT) provides information on tumor metabolic activity, which is related to tumor cell proliferation and viability and often precedes morphologic changes. PET–CT has demonstrated value for response evaluation in chemotherapy and targeted therapy settings and is recommended by the National Comprehensive Cancer Network guidelines for restaging after neoadjuvant therapy.9 However, its utility in assessing the response to neoadjuvant immunotherapy remains unclear and pseudoprogression after neoadjuvant ICI treatment is not well characterized. This retrospective study evaluated radiological changes after neoadjuvant ICI treatment and corresponding post-operative pathological responses, using targeted therapy as a comparative cohort. It aims to characterize pseudoprogression manifestations and explore the potential of PET–CT in distinguishing pseudoprogression from true progression in patients with resectable NSCLC. We further confirmed our results in an independent cohort of 95 NSCLC patients. Materials and methods Patient population Patients with resectable NSCLC who received neoadjuvant chemotherapy and immunotherapy followed by surgery between January 2021 and December 2023 at Guangdong Provincial People’s Hospital were included in this study. Patients treated with neoadjuvant targeted therapy combined with definitive resection during this period were enrolled as a comparative cohort. An independent temporal validation cohort of 95 patients treated between January and December 2024 at the same institution was used to validate the model. Eligibility criteria included histologically confirmed NSCLC, clinical stage II-III according to the American Joint Committee on Cancer (eighth edition), and PET–CT scans carried out both at baseline and after neoadjuvant treatment before surgery. Patients who participated in blinded clinical trials with undisclosed treatment regimens were excluded. Details of baseline characteristics, treatment, pathological evaluation, and follow-up data were retrieved from clinical records. The study was conducted in compliance with local and national regulations and received approval from the Ethics Committee of Guangdong Provincial People’s Hospital (KY-Q-2021-222-01). Imaging data A baseline PET–CT scan was acquired within 1 month before the initiation of neoadjuvant treatment and a subsequent PET–CT scan was carried out within 1 week before surgery to assess treatment response. PET–CT imaging was conducted using a uEXPLORER PET–CT scanner (United Imaging Healthcare, Shanghai, China). PET and CT images were acquired in the same session. All patients were instructed to fast for ≥6 h and to avoid strenuous or prolonged exercise for 24 h before the injection of 18F-FDG. Fasting blood glucose levels were required to be ≤11.0 mmol/L. Images were captured 60 min after the injection of 3.7 MBq of 18F-FDG per kilogram of body weight. A CT scan (50 mAs; 5-mm slices) was carried out before PET imaging for attenuation correction and anatomical localization with a dose modulation technique, followed by contrast-enhanced CT (140 mAs; 5-mm slices), unless contraindicated. Finally, total-body PET images were acquired within 5 min duration and three-dimensional PET images were reconstructed. Assessment of radiological response All imaging data were reviewed and analyzed by a senior radiologist who had access to limited generalized clinical information required for image interpretation but was blinded to treatment regimen, clinical stage, pathological results, and survival outcomes to minimize information bias. PET–CT scans obtained before and after neoadjuvant therapy were compared, with 18F-FDG tumor uptake quantified using the maximum standardized uptake value (SUVmax), calculated as the maximum activity concentration of 18F-FDG divided by the injected dose and corrected for the patient’s body weight. The change in SUVmax (ΔSUVmax) was defined as the SUVmax at the time point after neoadjuvant treatment minus the SUVmax at baseline. In addition, metabolic tumor volume (MTV) and total lesion glycolysis (TLG) of the primary tumor were measured and changes in MTV and TLG were calculated using the same approach as SUVmax. Tumor response and metabolic response were evaluated according to RECIST 1.1 and the European Organization for Research and Treatment of Cancer (EORTC) criteria, respectively.10,11 To evaluate interreader reliability, a second senior radiologist independently reviewed a randomly selected subset of 80 cases balanced across treatment cohorts [ICI and tyrosine kinase inhibitor (TKI)], pathological response (MPR and non-MPR), and imaging-defined progression and nonprogression (Supplementary Table S1, available at https://doi.org/10.1016/j.esmoop.2026.108366). Assessment of pathological response The pathological response was evaluated by an experienced lung cancer pathologist (L.X.Y.), who examined hematoxylin and eosin-stained slides of the resected primary tumor and lymph nodes, following the International Association for the Study of Lung Cancer Multidisciplinary Recommendations for Pathologic Assessment of Lung Cancer Resection Specimens After Neoadjuvant Therapy.12 The primary tumor bed components included viable residual tumor, necrosis, and stromal tissue. An MPR is defined as ≤10% residual viable tumor cells, while a CPR is characterized by 0% residual viable tumor cells in both the primary tumor and lymph nodes. Treatment Patients with EGFR/ALK wild-type or unknown genotype tumors were treated with platinum-doublet chemotherapy combined with antiprogrammed cell death protein 1 or antiprogrammed death-ligand 1 (PD-L1) antibodies, administered every 3 weeks for three to four cycles. In contrast, those with confirmed EGFR mutations or ALK rearrangement received EGFR/ALK TKIs as neoadjuvant treatment. Additionally, five patients with other driver gene mutations underwent corresponding TKI therapy, including three with ROS1 mutation, one with MET exon 14 skipping mutation, and one with RET mutation. Lobectomy or pneumonectomy, along with systematic complete mediastinal lymph node dissection, was carried out within 6 weeks of completing the neoadjuvant therapy. Patients with suspected new lesions or imaging-defined progression after neoadjuvant therapy were reviewed by a multidisciplinary team. Surgical resection was considered only when the lesions were judged to be indeterminate rather than unequivocal metastatic progression, the primary tumor remained technically resectable, the patient had no major clinical deterioration, and curative intent surgery was considered feasible after comprehensive multidisciplinary team assessment. After surgery, patients could receive adjuvant TKIs or immunotherapy according to national guidelines. Statistical analysis Continuous data are presented as mean ± standard deviation. Categorical variables were analyzed using the correction for continuity χ2 test, Pearson χ2 test, or Fisher’s precision test. Student’s t-tests were used to compare differences in the parameter variables. Interreader agreement for continuous measurements including primary-tumor SUVmax and longest diameter was assessed using the intraclass correlation coefficient. Agreement for RECIST/EORTC response assessment was evaluated using Cohen’s kappa. Univariate analysis was conducted to identify clinical and radiological factors associated with pathological response, while logistic regression was used to determine independent predictors of MPR. Subsequently, the sensitivity and specificity were explored via receiver operating characteristic (ROC) curve analysis to determine the multivariate logistic regression model that best distinguishes between true tumor progression and pseudoprogression. The model constructed from the training cohort was applied to this validation set and its diagnostic performance was assessed using ROC curve analysis and calibration plots. Model discrimination was quantified by the area under the curve (AUC) and calibration was evaluated using the Hosmer–Lemeshow goodness-of-fit test. All statistical analyses were carried out using the Statistical Package for the Social Sciences Statistics version 26.0 (IBM Corp., Armonk, NY) and R version 4.5.2 (R Foundation for Statistical Computing, Vienna, Austria). All tests were two-sided and P < 0.05 was considered statistically significant. Results Patient characteristics A total of 283 patients (214 male and 69 female) with resectable NSCLC entered the study. The median age was 61 years (range 28-82). The ICI cohort included 235 patients receiving neoadjuvant ICIs and the TKI cohort included 48 patients treated with neoadjuvant TKIs. Baseline smoking status, histology, and sex differed between the ICI and TKI cohorts. The median follow-up time was 32.8 months. Baseline characteristics are shown in Table 1. Table 1Baseline clinical and histopathological characteristicsVariableICI cohortTKI cohortValidation cohortTotal, N2354895Age, years, median (range)61 (28-82)61 (41-78)64 (36-78)Sex Male194 (82.6%)20 (41.7%)85 (89.5%) Female41 (17.4%)28 (58.3%)10 (10.5%)Smoking Ever148 (63.0%)12 (25.0%)70 (73.7%) Never87 (37.0%)36 (75.0%)25 (26.3%)ECOG PS 034 (14.5%)9 (18.8%)26 (27.4%) 1201 (85.5%)39 (81.2%)69 (72.6%)Stage I7 (3.0%)2 (4.2%)5 (5.3%) II51 (21.7%)8 (16.7%)23 (24.2%) III177 (75.3%)38 (79.1%)67 (70.5%) IV0 (0%)0 (0%)0 (0%)PD-L1 expression <1%54 (23.0%)8 (16.6%)32 (33.7%) 1-49%80 (34.0%)22 (45.8%)28 (29.5%) ≥50%63 (26.8%)9 (18.8%)22 (23.2%) NA38 (16.2%)9 (18.8%)13 (13.7%)Histology ADC74 (31.5%)43 (89.6%)34 (35.8%) SCC126 (53.6%)2 (4.2%)52 (54.7%) Other35 (14.9%)3 (6.2%)9 (9.5%)Driver mutation EGFR21 (8.9%)31 (64.6%)7 (7.4%) ALK0 (0%)12 (25.0%)2 (2.1%) Other6 (2.5%)5 (10.4%)7 (7.4%) NA208 (88.6%)0 (0%)79 (83.2%)Values are n (%) unless otherwise specified.ADC, adenocarcinoma; ALK, anaplastic lymphoma kinase; ECOG PS, Eastern Cooperative Oncology Group performance status; EGFR, epidermal growth factor receptor; ICI, immune checkpoint inhibitor; NA, not available; PD-L1, programmed death-ligand 1; SCC, squamous-cell carcinoma; TKI, tyrosine kinase inhibitor. Baseline clinical and histopathological characteristics Values are n (%) unless otherwise specified. ADC, adenocarcinoma; ALK, anaplastic lymphoma kinase; ECOG PS, Eastern Cooperative Oncology Group performance status; EGFR, epidermal growth factor receptor; ICI, immune checkpoint inhibitor; NA, not available; PD-L1, programmed death-ligand 1; SCC, squamous-cell carcinoma; TKI, tyrosine kinase inhibitor. Outcomes Among the 283 patients included, 275 (97.1%) underwent complete (R0) resection, while 5 (1.7%) had microscopic residual disease (R1) and 3 (1.0%) had macroscopic residual disease (R2). Of those who achieved R0 resection, 91 patients (33.1%) attained CPR and 142 (51.6%) had MPR. In the ICI cohort, the CPR and MPR rates were 37.0% (87 patients) and 57.0% (131 patients), respectively. The median EFS for the entire cohort was 57.3 months, with 57.3 months for the TKI cohort (Figure 1A). However, the median EFS was not reached in the ICI cohort. The 2-year EFS rate was 52.7% overall, 48.9% in the ICI group, and 70.8% in the TKI group. The 2-year overall survival (OS) rate for the entire cohort was 63.3%, with 58.3% in the ICI cohort and 87.5% in the TKI cohort (Figure 1B). Pseudoprogression is associated with improved OS in the immunotherapy cohort. (A) EFS for the entire cohort. (B) OS for the entire cohort. (C) EFS in the ICI cohort with and without pseudoprogression defined by EORTC criteria. (D) OS in the ICI cohort with and without pseudoprogression defined by EORTC criteria. In the ICI cohort, patients with pseudoprogression had significantly better OS and numerically better EFS than those without pseudoprogression, indicating that imaging-defined progression during neoadjuvant ICI therapy may often represent pseudoprogression, which is associated with improved outcomes. EFS, event free survival; EORTC, European Organisation for Research and Treatment of Cancer; ICI, immune checkpoint inhibitor; OS, overall survival. Radiological and metabolic progression after neoadjuvant treatment After neoadjuvant treatment, all patients were assessed for surgical candidacy using PET–CT. Twenty-six patients (9.2%) showed tumor enlargement or developed new lesions, and 26 (9.2%) met RECIST 1.1 for PD, all within the ICI cohort. Despite radiological progression, all 26 patients proceeded to R0 resection. Post-operative pathology showed that 17 of 26 patients (65.4%) achieved MPR, including 14 patients (53.8%) with CPR. In patients without radiological progression (n = 257), the CPR and MPR rates were 30.0% (77 of 257) and 49.1% (126 of 257), respectively. There was no significant difference in either the rate of complete resection or pathological response between patients with and without radiological progression (P =1.000 and 0.060, respectively). The 2-year EFS rates were 53.8% compared with 52.1% (P = 1.000). Identification of tumor progression on imaging poses a challenge in determining surgical eligibility after neoadjuvant therapy. To better assess the efficacy of neoadjuvant therapy, integration of metabolic response assessment using PET–CT is warranted. Surprisingly, 71 patients (25.1%) were evaluated as having progressive metabolic disease (PMD) based on EORTC criteria, including 65 in the ICI cohort and 6 in the TKI cohort, a rate markedly higher than with CT alone. R0 resection was not achieved in two patients (0.7%) with PMD because of extensive adhesions (R1 in one and R2 in one), whereas six patients (2.1%) without metabolic progression also failed to achieve R0 resection. Despite being classified as PMD, 27 patients (38.0%) achieved CPR and 39 (54.9%) had MPR. Of them, 38 were in the ICI cohort, and only one patient was in the TKI cohort, who showed an isolated increase in lymph node SUVmax and achieved MPR. Complete resection rate and pathological response were similar in patients with and without metabolic progression (P = 1.000 and 0.530, respectively). The 2-year EFS rates were 43.7% and 55.2 %, respectively (P = 0.122). Table 2 summarizes the pathological response and image progression in both the ICI and TKI cohorts. Table 2Radiological, metabolic, and pathological responses in two cohortsICI cohortTKI cohortRECIST 1.1EORTCRECIST 1.1EORTCPathological responsePDNon-PDPMDNon-PMDPDNon-PDPMDNon-PMDMPR17 (7.2%)115 (48.9%)38 (16.2%)94 (40%)0 (0%)11 (22.9%)1 (2.1%)10 (20.8%)Non-MPR9 (3.8%)94 (40%)27 (11.5%)76 (32.3%)0 (0%)37 (77.1%)5 (10.4%)32 (66.7%)EORTC, European Organisation for Research and Treatment of Cancer; ICI, immune checkpoint inhibitor; MPR, major pathological response; PD, progressive disease; PMD, progressive metabolic disease; TKI, tyrosine kinase inhibitor. Radiological, metabolic, and pathological responses in two cohorts EORTC, European Organisation for Research and Treatment of Cancer; ICI, immune checkpoint inhibitor; MPR, major pathological response; PD, progressive disease; PMD, progressive metabolic disease; TKI, tyrosine kinase inhibitor. These findings indicate that radiological or metabolic progression did not translate into worse surgical or survival outcomes. Given the strong correlation between pathological response and survival, tumor enlargement on imaging may represent either true progression or pseudoprogression, which require completely different clinical management strategies. This underscores the need for accurate diagnostic approaches in the neoadjuvant setting. Radiological and metabolic pattern of progression Further analysis showed that radiological progression by RECIST 1.1 comprised primary tumor progression in five patients (19.2%), isolated regional lymph node progression in seven (26.9%), and distant metastasis in 14 (53.8%). After incorporating 18F FDG–PET imaging, patients with regional lymph node progression increased to 43 (60.6%); primary tumor progression was identified in 13 patients (18.3%) and distant metastasis in 15 patients (21.1%). Regional nodal progression predominantly involved stations 4 (21%), 10 (21%), and 11 to 14 (22%). Among patients with a new distant lesion, nine (56.3%) presented with pulmonary nodules, two (12.5%) had multiple pleural nodules, two (12.5%) had extraregional lymph node involvement (hepatic portal and axillary nodes), one (6.3%) developed an adrenal gland lesion, and one (6.3%) had a bone lesion. The diameters of the new lesions ranged from 0.5 to 3.0 cm with SUVmax 0.9 to 14.8. R0 resection was achieved in 92.3%, 97.7%, and 100% of patients with primary tumor progression, isolated regional lymph node progression, and distant progression, respectively (P = 0.377). The corresponding MPR rates for the three progression patterns were 23.1%, 58.2%, and 73.3%, respectively (P = 0.020). Patients with primary tumor progression had worse EFS and OS than those with isolated regional lymph node or distant progression (EFS P = 0.425; OS P = 0.088) (Supplementary Figure S1A and B, available at https://doi.org/10.1016/j.esmoop.2026.108366), although the difference was not statistically significant. They also had inferior survival compared with patients without imaging-defined progression who did not achieve MPR, with a significant difference in EFS (P < 0.0001) but not in OS (P = 0.213) (Supplementary Figure S1C and D, available at https://doi.org/10.1016/j.esmoop.2026.108366). Patients with isolated regional lymph node or distant progression had EFS and OS similar to those without imaging-defined progression, suggesting that isolated nodal or distant lesions in the absence of primary tumor progression most likely represented pseudoprogression. Giving that MPR among patients with imaging-defined progression was observed exclusively in those receiving neoadjuvant ICI, subsequent analyses were conducted by treatment cohort (ICI and TKI). In the ICI cohort, primary tumor progression was associated with a significantly lower MPR rate than lymph node or distant progression (23.1%, 65.8%, and 71.4%, respectively; P = 0.014), and with worse EFS and OS (EFS P = 0.393; OS P = 0.105) (Supplementary Figure S1E and F, available at https://doi.org/10.1016/j.esmoop.2026.108366), consistent with the findings observed in the overall population. In contrast, in the TKI cohort, patients with PMD had shorter EFS and OS than those without progression (EFS P = 0.150; OS P = 0.616), although the difference was not statistically significant, indicating concordance between imaging and pathological response (Supplementary Figure S2, available at https://doi.org/10.1016/j.esmoop.2026.108366). Among patients with progression, only primary tumor and nodal progression were observed and survival was similar between these two patterns (EFS P = 0.520; OS P = 0.710). A Sankey plot mapping EORTC and RECIST categories to pathological outcomes supports these cohort differences, showing concordance in the TKI and discordance in the ICI (Supplementary Figure S3, available at https://doi.org/10.1016/j.esmoop.2026.108366). This finding indicates that pseudoprogression, a response pattern unique to immunotherapy, may limit the utility of conventional radiological evaluation in the neoadjuvant setting. Pseudoprogression after neoadjuvant ICI treatment Pseudoprogression was defined as PD according to RECIST 1.1 or PMD according to EORTC criteria, followed by MPR in the resected tissue. The overall incidence of pseudoprogression by RECIST was 7.2% (17 patients) and 16.2% by EORTC (38 patients). Among those patients with pseudoprogression by EORTC criteria, reginal lymph node involvement was the predominant presentation, with new nodes in 32 patients (84.2%). Patients with pseudoprogression had significantly better OS (P = 0.039) and numerically favorable EFS (P = 0.125) than those without pseudoprogression (Figure 1C and D). These findings suggest that imaging-defined progression during neoadjuvant ICI therapy may represent pseudoprogression. However, in a supplementary analysis restricted to patients who achieved MPR, EFS and OS did not differ significantly between patients with pseudoprogression and those without imaging-defined progression (Supplementary Figure S4, available at https://doi.org/10.1016/j.esmoop.2026.108366). In the ICI cohort, among patients classified as having PMD, only seven (3.0%) showed primary tumor enlargement in the resected specimen relative to baseline imaging and failed to achieve CPR, indicating true progression. A strong association between the SUVmax of the primary tumor and progression was observed. Among patients classified as having PMD, ROC curve analysis using percentage change in SUVmax versus change in longest diameter showed that a reduction in SUVmax ≥53.1% optimally predicted pseudoprogression (AUC 0.865), outperforming size change (AUC 0.727). The SUVmax threshold demonstrated high diagnostic accuracy, with a sensitivity of 94.7% and a specificity of 74.1% (Figure 2). ROC curves for identifying pseudoprogression. The blue curve represents the longest diameter change (%) of the primary tumor, with an AUC of 0.727, indicating limited diagnostic performance. The red curve corresponds to ΔSUVmax of the primary tumor, where a decrease of ≥53.1% demonstrated an AUC of 0.865. The black curve represents the multivariate combined model, integrating ΔSUVmax of the primary tumor, smoking status, and histology, which achieved the highest diagnostic accuracy with an AUC of 0.935. AUC, area under the curve; ROC, receiver operating characteristic; Se, sensitivity; Sp, specificity; ΔSUVmax, change in maximum standardized uptake value. To distinguish pseudoprogression from nonresponse, clinical and radiological variables, including age, clinical stage, smoking status, PD-L1 status, cycles of treatment, ICI type, change in primary tumor longest diameter, primary tumor ΔSUVmax, ΔMTV and ΔTLG, the number of regional lymph nodes with increased SUVmax, and the number of new regional lymph nodes, were analyzed. In univariate analysis using independent samples t-tests or χ2 tests, significant differences were observed in ΔSUVmax, ΔTLG (%), and longest diameter change (%) of the primary tumor (Supplementary Table S2, available at https://doi.org/10.1016/j.esmoop.2026.108366), smoking status (χ2 = 6.884, P = 0.009), and histology (χ2 = 7.991, P = 0.018). In multivariate logistic regression, histology and ΔSUVmax of the primary tumor were significant independent factors associated with pseudoprogression (Supplementary Table S3, available at https://doi.org/10.1016/j.esmoop.2026.108366). Combining multiple variables in a multivariate model further enhanced diagnostic accuracy, yielding an AUC of 0.935 (Figure 2). We further evaluated volumetric PET parameters in an expanded multivariable analysis. TLG change remained associated with pseudoprogression; however, adding TLG change to the SUVmax-based model resulted in only minimal improvement in discrimination (AUC 0.936 compared with 0.935). Given the limited incremental value and the greater complexity of volumetric segmentation, the SUVmax-based model was retained as the primary model. Detailed TLG analyses are provided in Supplementary Table S3 and Supplementary Figure S5, available at https://doi.org/10.1016/j.esmoop.2026.108366. Temporal validation of the predictive model To assess reproducibility, the model was evaluated in the independent temporal validation cohort (n = 95). In this cohort, 40 patients (42.1%) achieved CPR and 57 (60.0%) had MPR. By RECIST 1.1, only two patients (2.1%) were categorized as having radiological PD. In contrast, EORTC criteria identified 25 patients (26.3%) with PMD, among whom 13 (52.0%) subsequently achieved MPR on pathological evaluation, again indicating a substantial rate of pseudoprogression. The model demonstrated discriminative performance in the validation dataset, with an AUC 0.859 (sensitivity 92.3%; specificity 75.0%). Calibration was excellent overall (Hosmer–Lemeshow P = 0.941), with accurate prediction for probabilities ≤0.3 and mild overestimation beyond this range (Supplementary Figure S6A, available at https://doi.org/10.1016/j.esmoop.2026.108366). ROC analysis confirmed good discriminative ability (Supplementary Figure S6B, available at https://doi.org/10.1016/j.esmoop.2026.108366). These findings support the reproducibility and generalizability of the model across independent patient populations. Discussion CT or PET–CT are routinely used to evaluate the response to neoadjuvant therapy and to determine surgical eligibility in resectable NSCLC. However, our study demonstrated a fundamental limitation of conventional imaging assessment after neoadjuvant immunotherapy. In patients receiving neoadjuvant ICIs, imaging-defined progression did not translate into poor surgical or survival outcomes. Notably, nearly two-thirds of patients classified as having radiological progression still achieved MPR, indicating that apparent progression on preoperative imaging may represent pseudoprogression and should not automatically preclude potentially curative surgery. Pseudoprogression, a distinct response pattern observed exclusively in ICI therapy, occurred in 7.2% of cases in our study based on CT imaging, which is relatively higher than the rates reported in advanced NSCLC.13 PET–CT has been widely regarded as a sensitive tool for assessing pathological responses.14 In the TKI cohort, radiological and metabolic changes were concordant with pathological response and PD rarely achieved MPR, supporting the reliability of conventional imaging assessment in the neoadjuvant TKI setting, consistent with the findings from the phase II prospective study by Blakely et al.15 In contrast, in the ICI cohort, incorporating PET imaging increased the proportion of cases labeled as progression, while more than half of those meeting EORTC criteria for PMD ultimately achieved MPR. Accordingly, the incidence of pseudoprogression increased to 16.2% when PET–CT was used. Pathological evaluation of resected specimens confirmed true tumor progression in only 3% of patients defined as PMD. These findings underscore a key clinical pitfall that PET–CT-defined progression after neoadjuvant immunotherapy may not indicate treatment failure and should not be used in isolation to withhold potentially curative surgery. The predominant manifestation of pseudoprogression in our cohort was isolated regional lymph node involvement, either as newly apparent nodes or increased nodal SUVmax, commonly located in hilar and interlobar stations. Cascone et al.16 reported that nodal enlargement after neoadjuvant ICI therapy may reflect immune-related granulomatous inflammation rather than residual malignancy, a phenomenon not observed with chemotherapy. Similar observations have been reported in other malignancies. In head and neck cancer, metabolically active lymph nodes after neoadjuvant ICI therapy frequently lack viable tumor cells on pathological examination.14,17,18 Our study further extends these findings by demonstrating that pseudoprogression may also present as new distant lesions with diameters >1.0 cm and SUVmax values ≤6.0, making misclassification as true tumor progression particularly likely. To our knowledge, this is the first study to report that pseudoprogression in the neoadjuvant immunotherapy setting may manifest as new lesions in distant sites throughout the body, further complicating radiological assessment. Importantly, when nodal or distant progression occurred in the absence of primary tumor progression, the MPR rate exceeded 40%, strongly supporting pseudoprogression as the underlying mechanism. For PD during immunotherapy, current immune-related RECIST guidelines recommend continuing ICI therapy and performing a subsequent evaluation. However, such an approach is not feasible in the neoadjuvant setting, where the treatment window is short and a surgical decision must be made without delay. This underscores the clinical importance of distinguishing pseudoprogression from true progression on imaging. Although accumulating evidence suggests that metabolic metrics such as SUVmax, MTV, and TLG are associated with immunotherapy response, their role in identifying pseudoprogression has remained limited because they have often been excluded from analysis.14,17,18 Our findings indicate that changes in primary tumor SUVmax after neoadjuvant ICI therapy provide a robust and clinically actionable signal for distinguishing pseudoprogression from true progression. A reduction of ≥53.1% SUVmax identified pseudoprogression with an AUC of 0.865 and outperformed size-based criteria. A simple model integrating primary tumor ΔSUVmax with smoking status and histology further improved diagnostic accuracy, yielding excellent performance in both the discovery and validation cohorts. Collectively, these findings support a pragmatic workflow that integrates progression pattern recognition with quantitative metabolic response to guide surgical decisions after neoadjuvant therapy. By identifying patients in whom imaging-defined progression reflects immune-related changes rather than treatment failure, this strategy helps avoid unnecessary abandonment of surgery and preserves curative intent in appropriate patients receiving neoadjuvant ICI (Figure 3). A flowchart of the progression pattern and quantitative metabolic response to guide surgical decision making after neoadjuvant therapy. CT, computed tomography; ICI, immune checkpoint inhibitor; PET, positron emission tomography; ΔSUVmax, change in maximum standardized uptake value; TKI, tyrosine kinase inhibitor. This study was limited in that it was a single-center retrospective analysis, which may introduce selection bias and limit generalizability. In addition, the TKI cohort and the subgroup of patients with imaging-defined progression were relatively small, resulting in limited statistical power for survival comparisons. Therefore, although the EFS difference between patients with and without metabolic progression did not reach statistical significance, the absolute difference in EFS should be interpreted cautiously and cannot exclude a clinically meaningful survival difference. It should also be noted that pseudoprogression in this study was defined as imaging-defined progression followed by MPR in the resection specimen. Therefore, patients classified as having pseudoprogression were enriched for deep pathological response, which is a well-established favorable prognostic factor after neoadjuvant immunotherapy. Accordingly, the favorable survival observed in the pseudoprogression group should not be interpreted as evidence that pseudoprogression itself confers an independent survival benefit. In a supplementary analysis restricted to patients with MPR, survival did not differ substantially between patients with pseudoprogression and those without imaging-defined progression, supporting the interpretation that favorable outcomes were largely driven by pathological response. No external central pathology review was carried out; therefore, potential interobserver variability in MPR/CPR assessment may have influenced the classification of pseudoprogression. Although we included a temporal validation cohort from the same institution, true prospective, multicenter and geographically independent external validation was not carried out. Therefore, the generalizability and transportability of the predictive model require further validation in independent prospective cohorts from other institutions. Nevertheless, the inclusion of an independent temporal validation cohort provides preliminary support for the robustness and generalizability of the model. To our knowledge, this is the first comprehensive description of pseudoprogression in resectable NSCLC patients undergoing neoadjuvant ICI therapy and directly addresses a critical gap in clinical decision making. In summary, imaging-defined progression after neoadjuvant immunotherapy may reflect pseudoprogression, most commonly involving regional lymph nodes and less often distant lesions. Primary tumor progression represents the highest risk pattern and warrants particular caution. Quantifying primary tumor ΔSUVmax and considering smoking status and histology provide an objective and clinically accessible approach to discriminate pseudoprogression from true progression, thereby improving surgical decision making in the era of neoadjuvant immunotherapy. Acknowledgements The authors thank the staffs at Guangdong Lung Cancer Institute, Guangdong Provincial People’s Hospital and Guangdong Academy of Medical Sciences for their help in this work. Ethical approval The study was conducted in compliance with local and national regulations and received approval from the Ethics Committee of Guangdong Provincial People’s Hospital (KY-Q-2021-222-01). The requirement for written informed consent was waived due to the retrospective nature of the study. Data availability Data are available upon reasonable request. Funding This work was supported by National Natural Science Foundation of China [grant number 82473445 to Y.P.], Guangdong Basic and Applied Basic Research Foundation [grant number 2025A1515011768 to Y.P.]. The funding sources had no role in the preparation of this manuscript. Disclosure YP declares honoraria from AstraZeneca, Bristol Myers Squibb (BMS), Eli Lilly, Merck Sharp & Dohme (MSD), Pfizer, Roche, and Sanofi, outside the submitted work. YLW declares advisory services for AstraZeneca, Boehringer Ingelheim, Novartis, and Takeda; speaker fees from AstraZeneca, Beigene, Boehringer Ingelheim, BMS, Eli Lilly, MSD, Pfizer, Roche, and Sanofi; and grants from AstraZeneca, Boehringer Ingelheim, BMS, Hengrui, and Roche, outside the submitted work. The remaining authors declare no potential conflicts of interest. Contributor Information D.-J. Li, Email: gd_pet@hotmail.com. Y.-L. Wu, Email: syylwu@live.cn. Supplementary data Supplementary Figure S1 Supplementary Figure S2 Supplementary Figure S3 Supplementary Figure S4 Supplementary Figure S5 Supplementary Figure S6 Supplementary Table S1 Supplementary Table S2 Supplementary Table S3 References Associated Data Supplementary Materials Supplementary Figure S1 Supplementary Figure S2 Supplementary Figure S3 Supplementary Figure S4 Supplementary Figure S5 Supplementary Figure S6 Supplementary Table S1 Supplementary Table S2 Supplementary Table S3 Data Availability Statement Data are available upon reasonable request.