Predictive value of 18F‐FDG PET/CT and serum tumor markers for tumor mutational burden in patients with non‐small cell lung cancer

Abstract Purpose To investigate the correlations between metabolic parameters (MPs) of 18F‐fluorodeoxyglucose (FDG) uptake on positron emission tomography/computed tomography (PET/CT), serum tumor markers (STMs), and tumor mutational burden (TMB) in patients with non‐small cell lung cancer (NSCLC). Materials and Methods In this retrospective study, we enrolled 129 patients with NSCLC (males, 78; females, 51) who underwent baseline TMB and STM tests and 18F‐FDG PET/CT scans before treatment between March 2018 and September 2022. Patients were categorized into TMB‐high (TMB ≥10 mutations/Mb; n = 27 [20.9%]) and non‐TMB‐high (TMB <10 mutations/Mb; n = 102 [79.1%]) groups. Binary logistic regression analyses were performed to determine independent predictors of TMB‐high. Univariate and multivariate linear regression analyses were performed to determine independent predictors of TMB level on a log scale. Subgroup analyses for adenocarcinoma (ADC), ADC with EGFR+, ADC with EGFR−, and squamous cell carcinoma (SCC) were performed. Results For ADC, all MPs (SULpeak, SULmax, SULmean, MTV, and TLG) were significantly higher in the TMB‐high group than the non‐TMB‐high group; smoker (odds ratio [OR] = 27.08, p = 0.018), EGFR+ (OR = 0.03, p = 0.033), KRAS+ (OR = 7.98, p = 0.083), high CEA (OR = 33.56, p = 0.029), and high CA125 (OR = 13.68, p = 0.030) were independent predictors of TMB‐high; and all MPs showed significant positive linear correlations with TMB on a log scale, with SULpeak as an independent predictor. However, no significant correlation was observed for SCC. Conclusion MPs and STMs can predict the TMB level for patients with ADC, and may serve as potential substitutes for TMB with increased value and easy implementation in guiding immunotherapy through noninvasive methods.


| INTRODUCTION
Immune checkpoint inhibitor (ICI) therapy has revolutionized the treatment of advanced non-small cell lung cancer (NSCLC).Unfortunately, the response rate to ICI monotherapy is only 15%-20% for unselected patients. 1,2redictive biomarkers remain a major challenge for clinical applications.Tumor mutational burden (TMB), defined as the total number of somatic mutations per coding area of a tumor genome, has been acknowledged as a promising biomarker of response to ICIs.A higher TMB is associated with higher response rates and longer progression-free survival. 3However, TMB has not been widely available clinically because of difficulties in detection, including invasive procedures followed by pathological assessment, the inability to obtain sufficient specimens sometimes or obtain specimens repeatedly, and expensive detection methods for next-generation sequencing.Therefore, it is important to explore noninvasive surrogate biomarkers for TMB.
As a functional noninvasive imaging modality based on glucose metabolism, 18 F-fluorodeoxyglucose (FDG) positron emission tomography/computed tomography (PET/CT) is a significant imaging modality.Previous studies have reported that baseline metabolic parameters (MPs) of NSCLC may be predictive factors for response to immunotherapy. 4 Serum tumor markers (STMs) from peripheral blood have the advantages of being noninvasive, convenient, and inexpensive and are potentially useful in the early diagnosis, treatment monitoring, and prognosis evaluation of NSCLC. 5,6STMs can serve as ideal tumor-associated antigens for inducing effective tumor immunity. 7ome retrospective analyses have also demonstrated that pretreated STMs are prognostic predictors of immunotherapy. 8,9Therefore, we hypothesized that there may be links between MPs, STMs, and TMB.
To date, few studies have explored the correlations between MPs, STMs, and TMB, [10][11][12] and the results remain controversial.Haghighat-Jahromi et al. confirmed an independent correlation between the maximum standardized uptake value (SUV max ) and TMB in various tumor types. 10Ono et al. reported that SUV max and carcinoembryonic antigen (CEA) were independent predictors of TMB in NSCLC. 12However, Moon et al. found no significant relationship between metabolic features and TMB in lung cancer. 11Additionally, these studies had some limitations: (a) subgroup analyses of pathological type and driver gene mutation type were not performed; (b) the time interval of PET/CT scan, STM test, and biopsy/surgery were not determined; (c) the cutoff of TMB-high was different; and (d) limited MPs and STMs were studied.
To complement the deficiency in the extant literature, we aimed to further explore the correlations of MPs and STMs with TMB.Detailed analyses of different pathological types and driver gene mutation types were performed.We expect to provide new insights for potential substitutes to TMB which has clinical utility.

| Patient selection
We enrolled 2597 patients with lung cancer who underwent PET/CT examinations with data retrieved from the PET/CT Center data system between March 2018 and September 2022.The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Institutional Review Board of the Institutional Ethics Committee of our hospital.The inclusion criteria were as follows: (a) pathologically confirmed NSCLC; (b) baseline TMB data of the primary tumor tissue on biopsy/surgery smoker (odds ratio [OR] = 27.08,p = 0.018), EGFR+ (OR = 0.03, p = 0.033), KRAS+ (OR = 7.98, p = 0.083), high CEA (OR = 33.56,p = 0.029), and high CA125 (OR = 13.68,p = 0.030) were independent predictors of TMB-high; and all MPs showed significant positive linear correlations with TMB on a log scale, with SUL peak as an independent predictor.However, no significant correlation was observed for SCC.

Conclusion:
MPs and STMs can predict the TMB level for patients with ADC, and may serve as potential substitutes for TMB with increased value and easy implementation in guiding immunotherapy through noninvasive methods.

K E Y W O R D S
18 F-FDG PET/CT, immunotherapy, non-small cell lung cancer, serum tumor marker, tumor mutational burden samples obtained before treatment, excluding TMB data of metastases and lymph nodes; (c) baseline PET/CT scan performed before treatment; (d) time interval between biopsy/surgery and PET/CT scan was within 1 month to ensure an acceptable temporal correlation between imaging and genomic evaluation; and (e) measurable and describable tumor lesions on PET/CT.When a patient had multiple primary lung cancers, the lesion undergoing biopsy/ surgery for TMB test was included in the analyses.
Finally, 129 patients who underwent TMB tests and PET/CT scans before treatment were enrolled in the study.Clinical and demographic characteristics were retrospectively collected from medical records.We further collected baseline STM data of the enrolled patients, and the time interval between biopsy/surgery and the STM test was within 1 month.Ultimately, 98 of the 129 enrolled patients had available STM data.

| FDG PET/CT acquisition
All patients underwent a whole-body PET/CT scan as per the standard protocol of our hospital.Head-to-groin imaging was performed using an integrated PET/CT device (Discovery 690; GE Healthcare).Patients were asked to fast for at least 6 h before the scan, and blood glucose level (normal, <145 mg/dL) was determined.Patients were intravenously injected with 18 F-FDG at a dose of 3.70-4.44MBq/kg.Three-dimensional (3D) whole-body PET/ CT scans with CT were acquired approximately 50-70 min after 18 F-FDG injection.PET images were obtained in 3D mode at a rate of 2 min per frame from head to groin (generally, 7-8 beds location) and reconstructed using the VPFX-S algorithm, with two iterations, 24 subsets, and a 4 mm Gaussian postfilter.Spiral CT scans were performed using a standardized protocol (tube voltage, 120 kV; tube current, 150 mA; slice thickness, 3.75 mm; and rotation speed, 0.5 s) without intravenous contrast administration.All patients underwent a breath-hold thoracic spiral CT scan at 120 kV, 205 mA, 0.5 s per tube rotation, and a thickness of 5 mm after the PET/CT scan.Images were reconstructed from a thickness of 1.25 mm and 0.8 mm space.

| Image analysis
PET images were analyzed using PETVCAR (PET Volume Computerized Assisted Reporting), an automated segmentation software system on an Advantage Workstation (version 4.6; GE Healthcare).The volume of interest (VOI) of the whole tumor was autocontoured and segmented using a 3D cube.The VOI was placed by one radiologist (Q.Z.) and verified by another experienced radiologist (X.L.T.) to ensure that the 3D cube contained all FDG PET-positive areas of the primary tumor in the axial, coronal, and sagittal planes.Disagreements were resolved by consultation.The two radiologists were blinded to the clinical data of all patients when placing the VOI.
PETVCAR is an automated segmentation software system that uses an iterative adaptive algorithm to detect the threshold level.It separates the target volume from the background tissue by weighting SUV max and SUV mean within the target volume with a weighting factor of 0.5. 13ETVCAR was used to calculate the following parameters of the primary tumor: SUL max , SUL mean , SUL peak , metabolic tumor volume (MTV), and total lesion glycolysis (TLG).SUL max was defined as the maximum SUL derived from the single voxel with the highest uptake; SUL mean was calculated as the sum of SUL in each voxel divided by the number of voxels within the target volume; SUL peak was defined as the mean SUL value measured in a 1 cm 3 volume spherical region of interest centered around the hottest point of the tumor; MTV represented the volume of the contoured tumor tissue with active FDG uptake; and TLG was obtained by multiplying SUL mean by MTV.

| Serum tumor markers assessment
Based on clinical applications, STMs, including CEA, carbohydrate antigen 125 (CA125), cytokeratin 19 fragment (Cyfra21-1), and squamous cell-associated antigen (SCC-Ag), were tested using immunoelectrochemistry according to institutional standard laboratory procedures.Upper limits of normal were 5.0 ng/mL for CEA, 35.0 U/ mL for CA125, 3.3 ng/mL for Cyfra21-1, and 1.5 ng/mL for SCC-Ag, respectively.Records were obtained from the database and categorized as high or normal based on the upper limits.

| Histologic and genomic analysis
The approach used for obtaining pathological specimens was determined according to the patient's condition: 26 patients underwent surgery, 67 underwent CT-guided lung biopsy, and 36 underwent bronchofibroscopy biopsy.Next-generation sequencing was used to detect the TMB level, epidermal growth factor receptor (EGFR) mutation status, and Kirsten rat sarcoma viral oncogene homolog (KRAS) mutation status.Records were obtained from the official reports of the Pathology Department at our hospital.TMB was measured as mutations per megabase pair.Patients were categorized into TMB-high (≥10 mutations/ Mb) and non-TMB-high (<10 mutations/Mb) groups.Currently, there is no generally accepted cutoff value to define TMB-high.In this study, TMB-high was defined as ≥10 mutations/Mb as this category is consistent with those used in several randomized trials in lung cancer 3,14,15 and adopted by the Food and Drug Administration for refractory solid tumor approval of anti-PD-1 therapy. 16In addition, it is also the minimum threshold developed by a multistakeholder group convened by the Friends of Cancer Research to coordinate a standard cutoff across clinical trials. 17,18

| Statistical analysis
Descriptive statistics were used to analyze the differences between the TMB-high and non-TMB-high groups.Categorical data were presented as numbers (percentages).The Kolmogorov-Smirnov test was used to analyze the distribution normality of continuous data.Normally distributed data were presented as mean ± standard deviation, whereas non-normally distributed data were presented as median (range).Comparisons between the TMB-high and non-TMB-high groups were evaluated using Pearson's chi-squared test or Fisher's exact test for categorical variables and independent samples t-test or Mann-Whitney U test for continuous variables.Binary logistic regression analyses with backward stepwise selection were performed to determine independent predictors of TMB-high.Univariate and multivariate linear regression analyses were performed to determine the correlative factors of log 10 -transformed TMB, denoted as TMB (lg).TMB, MTV, and TLG were required to transform into the shifted log scale in the linear regression analyses because of the highly skewed distribution.Significant variables in the univariate linear regression were included in the multivariate linear regression model with backward stepwise selection.
Notably, there was significant collinearity among PET parameters.Significant explanatory variables in multivariate linear regression analyses were identified using a backward stepwise elimination method.Variables with collinearity were not retained.The final regression model was checked to make sure each variable made a significant contribution.Multicollinearity analyses showed that the variance inflation factors for all enrolled variables were below 5.0.Finally, through model selection, three variables (SUL max , SUL mean , and TLG) were excluded, and two variables (SUL peak and MTV) were included to construct the multivariate linear regression analyses.
Subgroup analyses for adenocarcinoma (ADC), ADC with EGFR+, ADC with EGFR−, and squamous cell carcinoma (SCC) were performed.Statistical analyses were conducted using SPSS (version 25.0;IBM Corp.).Statistical significance was set at p < 0.05.

| Patient characteristics
Based on the inclusion criteria, 129 patients with NSCLC, all of whom had baseline TMB and PET/CT scans before treatment, were enrolled in the present study.Among the 129 enrolled patients (  1 and 2.

TMB-high
Tables 1-4 demonstrate the differences between the TMBhigh and non-TMB-high groups for NSCLC, ADC, ADC with EGFR−, and SCC, respectively.Subgroup analysis of ADC with EGFR+ was not performed because only one patient had TMB-high in this subgroup.
TMB-high was associated with male, SCC, smoker, and EGFR− for NSCLC, and male, smoker, EGFR−, and KRAS+ for ADC (all p < 0.05).TMB-high was additionally associated with high CA125, Cyfra21-1, and SCC-Ag for NSCLC, and high CEA, CA125, and Cyfra21-1 for ADC (all p < 0.05).All MPs (SUL peak , SUL max , SUL mean , MTV, and TLG) were significantly higher in the TMB-high group than in the non-TMB-high group for NSCLC and ADC (all p < 0.05).In the subgroup analysis of ADC with EGFR−, TLG (p = 0.048) was significantly higher in the TMB-high group than in the non-TMB-high group.However, no significant correlation was observed for SCC.

| Correlations of MPs and STMs with TMB on a log scale
Owing to the highly skewed distribution of TMB, linear regression analysis was used to identify independent predictors of TMB on a log scale.T A B L E 1 Clinicopathological characteristics of NSCLC.

| DISCUSSION
Correlations between MPs, STMs, and TMB remain poorly understood.This study revealed different findings between ADC and SCC.For ADC, pretreated MPs and STMs were generally higher in the TMB-high group than in the non-TMB-high group; all MPs showed significant positive linear correlations with TMB (lg), and SUL peak was an independent predictor of TMB (lg), including subgroups of ADC with EGFR+ and EGFR−.However, no significant correlation was observed for SCC.We identified that MPs and STMs can predict the TMB level for patients with ADC, and may serve as potential substitutes for TMB with increased value and easy implementation in guiding immunotherapy through noninvasive methods.
The present study had the following improvements over previous studies: (a) subgroup analysis of the EGFR mutation was performed for the first time, as we believed it was important that patients with EGFR− were the target candidates for immunotherapy; (b) correlations between SUL parameters and TMB were analyzed for the first time.SUL parameters eliminate the influence of adipose tissue compared to SUV parameters.PERCIST 1.0 also recommends SUL peak as the most important quantitative indicator of therapeutic response for immunotherapy; (c) CA125 and SCC-Ag were studied for the first time, which enriched the blood-based biomarkers; (d) TMB-high was defined as TMB ≥10 mutations/Mb as it may be the optimal and universal cutoff to guide clinical applications; and (e) we set a 1-month interval between PET/CT scan, STM test, and biopsy/surgery to ensure an acceptable temporal correlation.A longer time lapse may result in a poor correlation.
This study demonstrated that TMB-high in NSCLC was associated with male, smoker, and SCC, which was consistent with the result of previous studies. 11,19,20orrelations of MPs and STMs with TMB only existed in ADC, and no significant correlation was observed for SCC.The reasons for the different results between ADC and SCC remain unclear.In our previous publication, we reported that baseline SUL max and SUL peak were positively correlated with the degree of pathological regression to neoadjuvant immunotherapy in NSCLC (SUL max , p = 0.036; SUL peak , p = 0.058), and a large majority of the patients (80.6%, 29/36) had SCC. 4 Therefore, we speculated that there may be links between MPs and immunotherapy biomarkers in SCC.However, no significant correlation was observed for SCC in this study.One reason could be the limited number of patients.Further studies involving more patients are required.Another possibility is that MPs correlate with programmed death ligand 1 (PD-L1) expression instead of TMB in SCC.Previous studies have reported a significant positive correlation between SUV max and PD-L1 expression in SCC. 21,22It has been proven that PD-L1 expression is independent of TMB, and both can independently predict immunotherapy efficacy. 23The predictive effect of PD-L1 expression and TMB, alone or in combination, and their correlations with tumor metabolism deserve further study.The subgroup analysis of ADC with EGFR+ was not performed because only one patient had TMB-high in this subgroup.
A negative correlation was observed between EGFR+ and TMB, which may explain why EGFR+ patients had a poor response to ICIs and even underwent hyperprogression. 24,25Among the 53 ADC patients with EGFR+ in this study, only one patient had TMB-high (TMB, 15 mutations/Mb), which was contradictory to the result of the negative correlation.However, the patient also had KRAS+.KRAS+ was found to be a predictor of TMB-high and positively correlated with TMB in the univariate linear analysis.Previous studies have also reported that patients with KRAS+ had better immunotherapy efficacy. 26herefore, KRAS+ may be an important predictor of TMBhigh, which needs further verification.
We found that all MPs (SUL peak , SUL max , SUL mean , MTV, and TLG) were significantly higher in the TMB-high group than in the non-TMB-high group, and all MPs presented significant positive correlations with TMB.However, the exact mechanism between FDG uptake and TMB remains unclear.The mechanism of FDG uptake involves glucose metabolism, angiogenesis, hypoxia, and the mammalian target of rapamycin (mTOR) signaling pathway. 27A previous study showed that carbohydrate metabolism had a positive correlation with TMB, 28 and another study showed that mutations of the mTOR pathway were associated with increased TMB. 29 Thus, the correlation between high MPs and high TMB may reflect increased glucose metabolism and activation of the mTOR pathway.Another alternative explanation for the correlation may be based on an innate immune response triggered by tumors with higher TMB.FDG is actively entrapped in the increased infiltration of immune effector cells such as tumor-infiltrating lymphocytes and tumor-associated macrophages. 30In the presence of the activated mTOR pathway, pathway enrichment was involved in antitumor immunity, enhanced antigen presentation, and increased infiltration of immune effector cells. 29Moreover, tumor aggressiveness may be another intrinsic link between MPs and TMB.It is generally believed that tumors with high FDG uptake have a higher likelihood of aggressiveness, 31 and the phenotypic features related to aggressiveness may become more prominent as the genetic mutation rate increases. 32he molecular pathogenesis of lung cancer involves various oncogenes, tumor suppressor genes, signaling pathway components, and other cellular processes. 33hese molecular processes lead to the release of various mutated or overexpressed proteins associated with lung cancer into body fluids, known as STMs, which are potential noninvasive biomarkers for diagnosis, monitoring, and prognosis assessment. 34The correlation between STMs and TMB remains poorly understood.Ono et al. reported that CEA level was an independent predictor of TMB. 12 The present study also found that high CEA and CA125 levels were independent predictors of TMB-high.The reappearance of CEA in patients with cancer indicates that certain genes are reactivated by the malignant transformation of cells.CEA can serve as an ideal tumor-associated antigen for inducing effective tumor immunity. 7he present study had several limitations.First, this was a preliminary, exploratory, and single-center retrospective study with a small number of patients, although approximately 4 years of data were included.Second, we only studied EGFR and KRAS mutations.Other driver gene mutations were not included because of the limited number of patients.Third, not all patients had available STM data.Fourth, the spatial heterogeneity of tumors might influence TMB detection.Finally, exploring the F I G U R E 1 Independent predictors of TMB-high by binary logistic regression analysis for NSCLC and ADC.TMB, tumor mutational burden; NSCLC, non-small cell lung cancer; ADC, adenocarcinoma; SCC, squamous cell carcinoma; EGFR, epidermal growth factor receptor; KRAS, Kirsten rat sarcoma viral oncogene homolog; CEA, carcinoembryonic antigen; CA125, carbohydrate antigen 125; SUL, standardized uptake value corrected by lean body mass.EGFR, epidermal growth factor receptor; KRAS, Kirsten rat sarcoma viral oncogene homolog; MTV, metabolic tumor volume; NA, not applicable; NSCLC, non-small cell lung cancer; SCC, squamous cell carcinoma; SCC-Ag, squamous cell-associated antigen; SUL, the standardized uptake value corrected by lean body mass; TLG, total lesion glycolysis; TMB, tumor mutational burden.
F I G U R E 2 Independent predictors of TMB on a log scale by multivariate linear regression analysis for NSCLC, ADC, ADC with EGFR+, and ADC with EGFR−.TMB, tumor mutational burden; NSCLC, non-small cell lung cancer; ADC, adenocarcinoma; SCC, squamous cell carcinoma; EGFR, epidermal growth factor receptor; KRAS, Kirsten rat sarcoma viral oncogene homolog; CEA, carcinoembryonic antigen; SUL, standardized uptake value corrected by lean body mass.direct relationship between MPs, STMs, and immunotherapy response is a future step.Therefore, prospective studies are necessary to validate our findings.

| CONCLUSION
MPs and STMs can predict the TMB level for patients with ADC, and may serve as potential substitutes for TMB with increased value and easy implementation in guiding immunotherapy through noninvasive methods.We highlight the findings for ADC patients with EGFR−, as they are the target candidates for immunotherapy.

F I G U R E 3
Correlations between SUL peak and TMB for NSCLC, ADC, and SCC on a log scale.(A) Scatter diagram showing the distribution of SUL peak and TMB (lg) for ADC and SCC.(B,C) Positive correlations between SUL peak and TMB (lg) for NSCLC (B) and ADC (C).(D) No significant correlation between SUL peak and TMB (lg) for SCC.TMB, tumor mutational burden; NSCLC, non-small cell lung cancer; SUL, standardized uptake value corrected by lean body mass; ADC, adenocarcinoma; SCC, squamous cell carcinoma.
Table 5 demonstrates the univariate linear regression analyses for NSCLC, ADC, ADC with EGFR+, ADC with EGFR−, and SCC, respectively.All MPs showed significant positive linear correlations with TMB (lg) for NSCLC, ADC, ADC with EGFR+,