Showing papers by "David R. Gandara published in 2023"

Convergence of Precision Oncology and Liquid Biopsy in Non-Small Cell Lung Cancer.

Abstract: This review article illuminates the role of liquid biopsy in the continuum of care for non-small cell lung cancer (NSCLC). We discuss its current application in advanced-stage NSCLC at the time of diagnosis and at progression. We highlight research showing that concurrent testing of blood and tissue yields faster, more informative, and cheaper answers than the standard stepwise approach. We also describe future applications for liquid biopsy including treatment response monitoring and testing for minimal residual disease. Lastly, we discuss the emerging role of liquid biopsy for screening and early detection.

Quality-of-life outcomes and risk prediction for patients randomized to Nivolumab/Ipilimumab vs Nivolumab on LungMAP-S1400I.

Abstract: BACKGROUND An important issue for patients with cancer treated with novel therapeutics is how they weigh the effects of treatment on survival and quality-of-life (QOL). We compared QOL in patients enrolled to SWOG-1400I, a substudy of the LungMAP biomarker-driven master protocol. METHODS SWOG S1400I was a randomized phase III trial comparing nivolumab/ipilimumab vs nivolumab for treatment of immunotherapy-naïve disease in advanced squamous cell lung cancer. The primary endpoint was the MDASI-LC severity score at Week-7 and Week-13 with a target difference of 1.0 points, assessed using multivariable linear regression. A composite risk model for progression-free and overall survival was derived using best-subset selection. RESULTS Among 158 evaluable patients, median age was 67.6 years and most were male (66.5%). The adjusted MDASI-LC severity score was 0.04 points (95%-CI, -0.44 to 0.51, p=.89) at Week-7 and 0.12 points (95%-CI, -0.41 to 0.65, p=.66) at Week-13. A composite risk model showed that patients with high levels of both appetite loss and shortness-of-breath had a 3-fold increased risk of progression or death (HR = 3.06, 95%-CI, 1.88-4.98, p<.001) - and that those with high levels of both appetite loss and work limitations had a 5-fold increased risk of death (HR = 5.60, 95%-CI, 3.27-9.57, p<.001) - compared to those with neither risk category. CONCLUSIONS We found no evidence of a benefit of ipilimumab added to nivolumab compared to nivolumab alone for QOL in S1400I. A risk model identified patients at high risk of poor survival, demonstrating the prognostic relevance of baseline patient-reported outcomes even in those with previously-treated advanced cancer.

The effect of ctDNA tumor fraction (TF) on overall survival and concordance between tissue genomics and ctDNA in Lung-MAP.

Abstract: 9035 Background: Discordance in genomic profiling between tumor tissue and ctDNA can occur due to inter-patient variability in shed tumor DNA, and intra-patient variability in sources of cell-free DNA (tumor heterogeneity, metastatic variants, CHIP). TF is an aneuploidy-based estimation of tumor-specific DNA in circulation. We evaluated the impact of TF levels on overall survival (OS), mutation concordance and tumor mutation burden in tissue (tTMB) and ctDNA (bTMB) in the LUNGMAP screening study. Methods: Eligible advanced NSCLC pts had blood drawn ≤30 days of fresh tissue biopsy with no intervening therapies. Genomic profiling was done by FoundationOne CDx and FoundationOne Liquid CDx. ctDNA TF was quantified via aneuploidy using deviations in genome-wide SNP coverage or somatic allele frequencies if under aneuploidy LoD. High TF was defined as ≥10%. Concordance among driver genes, cancer genes of interest (CGoI: T P53, PIK3CA, STK11, KEAP1, RB1, PTEN, ATM, BRCA1/2), and variant subtype were assessed by the percentage positive agreement (PPA) and positive predictive value (PPV) with 95% confidence intervals (CI). OS by TF used a Cox model and log-rank test. PPA and TF among CGoIs was compared with a Wilcoxon test. Correlation between tTMB and bTMB used Lin’s coefficient (LC). Results: 167 pts met inclusion criteria; 161 had TF data with 51 (32%) ≥10% (high TF). High TF was associated with worse OS (HR: 2.06 [1.43-2.95], p<0.001). The PPA and PPV ranged from 50%-100% and 60%-100%, respectively for driver genes. The PPA was numerically larger for high versus low TF for most drivers (p=0.18) and significantly larger for all mutations and SNVs combined (Table). For CGoI, the distribution of PPA differed between high and low TF (p=0.03). The median PPA for CGoI (25th%ile,75th%ile) was 56% (50,87) for low TF and 100% (100,100) for high TF. PPV for CGoI did not differ by TF (p=0.20). High TF was associated with higher levels of TMB and bTMB (p<0.01 for both). TMB LC (CI) for low and high TF were 0.46 (0.27,0.61) and 0.69 (0.53,0.81). Conclusions: High TF was associated with worse OS, better PPA among CGoI, and better LC for t/bTMB. TF may improve the clinical utility of mutations detected by liquid biopsies and should be reported and considered in interpreting these results. [Table: see text]

Abstract 5955: In Silico dissection of immune infiltrate signatures that are detected by DetermaIO, a predictor of response to immune therapy

Abstract: Background: The composition of tumors includes not only malignant but also immune, stromal and other cell types. Understanding this dynamic tumor immune microenvironment (TIME) is important to guide treatment and develop novel therapies and markers. We have previously validated immuno-oncology gene expression signatures (DetermaIO (DTIO) and a 101-gene algorithm), that predict efficacy of checkpoint inhibitors (ICI) based on distinguishing immunomodulatory (IM), mesenchymal stem-like (MSL), and mesenchymal (M) phenotypes. In this study we used TCGA and clinical cohorts to identify immune infiltrate populations within these defined TIME spaces and their association with ICI treatment. Methods: We derived novel human immune infiltrate signatures from a translation of murine ImmGen cell populations and a search for conserved co-expression of immune markers across multiple tumors. In total, 20 tumors from TCGA were employed for derivation and analysis encompassing 7163 unique samples. These novel signatures were compared to published immune infiltrate signatures and then their association with ICI efficacy and each other assessed in three cohorts treated with ICI therapy, IMvigor210 and an additional bladder cohort, comprising 272 and 89 patients with censored outcome results, and a melanoma cohort (N=105). Results: The ImmGen analysis created 35 immune cell signatures and pan-tumor conserved co-expression of immune markers created eight signatures. The co-expression signatures often contained a mixed population of cell-type markers, though largely dominated by either myeloid or lymphoid markers. These signatures showed highly reproducible proportions of samples with strong expression between train and test TCGA sets. Most immune signatures had their highest representation in IM and DTIO+ tumors, however there was also consistent identification of presumptive immune infiltrate presence in MSL, M and DTIO negative cases. Two of the conserved co-expression signatures, one comprised of B-cell markers, and the other of T cell and other lymphoid markers, were associated with ICI efficacy in IMvigor210 and validated in the other “real-world” bladder cohort (B-cell: OR=0.8, p=0.022, T lymphoid: OR=0.7, p=0.005). Both signatures also had significant association with outcome in the cohort with clinical response outcomes, being strongest in patients after treatment had initiated. Conclusions: These cell-type signatures may be identifying novel immune infiltrate populations that co-exist within the tumor immune microenvironment and are potentially predictive of ICI response. The two signatures were not independent of DTIO in either cohort, suggesting that the 27-gene algorithm DTIO largely incorporates this information. This analysis begins to dissect the complex physiology of the tumor immune microenvironment that mediates response to immune therapy. Citation Format: Brian Z. Ring, Catherine T. Cronister, Robert S. Seitz, Douglas T. Ross, Brock Schweitzer, David R. Gandara. In Silico dissection of immune infiltrate signatures that are detected by DetermaIO, a predictor of response to immune therapy. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5955.

Circulating tumor DNA (ctDNA) dynamics and survival outcomes in patients (pts) with advanced non-small cell lung cancer (aNSCLC) and high (>50%) programmed cell death-ligand 1 (PD-L1) expression, randomized to cemiplimab (cemi) vs chemotherapy (chemo).

Abstract: 9022 Background: While ctDNA has emerged as a promising tool for monitoring response on therapy, there is a paucity of data from prospective, randomised Phase 3 studies to establish clear criteria for the clinical application of ctDNA. In the EMPOWER-Lung 1 study, first line (1L) cemi monotherapy improved overall survival (OS) vs platinum chemo in pts with aNSCLC, PD-L1 ≥50%, and no EGFR, ALK, or ROS1 aberrations and an acceptable risk/benefit profile. We performed personalized tumor-specific analysis of ctDNA from pts treated in the EMPOWER-Lung 1 study to determine the optimal time point and magnitude of ctDNA variation that are associated with OS and progression-free survival (PFS). Methods: Tumor tissue next-generation sequencing was performed to identify tumor-specific DNA variants. ctDNA levels were monitored using personalized pt specific probe sets (Natera, Foundation Medicine) in the plasma at baseline, end of week 3 (W3) and W9. Endpoints included overall response rate (ORR) complete or partial response (CR/PR), stable disease (SD) and progressive disease (PD) per RECIST 1.1, OS and PFS. Association between ctDNA decrease and clinical endpoints was tested in 3 groups of pts defined by sensitivity analysis: no-decrease, any decrease (molecular response, MR), and complete clearance (complete MR, cMR) of ctDNA. Results: ctDNA analysis was performed on samples from 175 pts (chemo n = 89; cemi n = 86). Baseline characteristics, ORR, OS, and PFS were comparable to those of the intention to treat population. At W9, cMR/MR was associated with clinical response to cemi (CR/PR: 97%, SD:58%, PD:17%), but not with clinical response to chemo, as it was detected in the majority of chemo pts across all RECIST categories (CR/PR: 100%, SD:90%, PD:80%). In the cemi arm, ctDNA cMR was associated with the longest survival, with median OS (mOS) not-reached compared to pts with MR (mOS of 29 months) and pts with no-decrease (mOS of 8 months), as well as statistically significant hazard ratios (cMR vs MR: W9 HR = 8, 95% CI 1.8-35, p = 0.0056; cMR vs no-decrease: W9 HR = 25, 5.7–110, p = 0.000021 and W3 HR = 5.9, 1.4-25, p = 0.017). Delayed MR/cMR (i.e., MR/cMR at W9 but not at W3) was observed in 8/66 pts (12%), and transient MR (i.e., MR at W3 but not W9) in 9/66 pts (14%), with available ctDNA data at both timepoints. Conclusions: This is the largest data set to date correlating ctDNA levels with efficacy outcomes from a randomized clinical trial comparing chemo with immunotherapy in 1L aNSCLC. The results indicate that lack of treatment-induced decrease in ctDNA may identify pts with inferior OS benefit from cemi monotherapy as early as 3 weeks following initiation of therapy and can inform future use of early ctDNA response assessment in prospective interventional studies. Clinical trial information: NCT03088540 .

Phase II study of docetaxel and trametinib in patients with KRAS mutation positive recurrent non-small cell lung cancer (NSCLC) (SWOG S1507, NCT-02642042).

Abstract: PURPOSE Efficacy of MEK inhibitors in KRAS+ NSCLC may differ based on specific KRAS mutations and co-mutations. Our hypothesis was that docetaxel and trametinib would improve activity in KRAS+ NSCLC and specifically in KRAS G12C NSCLC. PATIENTS AND METHODS S1507 is a single arm phase II assessing the response rate (RR) with docetaxel plus trametinib in recurrent KRAS+ NSCLC and secondarily in the G12C subset. The accrual goal was 45 eligible patients with at least 25 with G12C mutation. The design was 2-stage design to rule out a 17% RR, within the overall population at the 1-sided 3% level and within the G12C subset at the 5% level. RESULTS Between July 18, 2016 and March 15, 2018, 60 patients were enrolled with 53 eligible and 18 eligible in the G12C cohort. The RR was 34% (95%CI- 22-48) overall and 28% (95%CI- 10-53) in G12C. Median PFS and OS were 4.1 and 3.3 months and 10.9 and 8.8 months, overall and in the subset, respectively. Common toxicities were fatigue, diarrhea, nausea, rash, anemia, mucositis, neutropenia. Among 26 patients with known status for TP53 (10+ve) and STK11 (5+ve), OS (HR:2.85, 95%CI 1.16-7.01) and RR (0% vs. 56%, p = 0.004) were worse in patients with TP53 mutated versus wild type cancers. CONCLUSIONS RRs were significantly improved in the overall population. Contrary to pre-clinical studies, the combination showed no improvement in efficacy in G12C patients. Co-mutations may influence therapeutic efficacy of KRAS directed therapies and are worthy of further evaluation.

Tumor mutational burden (TMB) measurement from an FDA-approved assay and real-world overall survival (rwOS) on single-agent immune checkpoint inhibitors (ICI) in over 8,000 patients across 24 cancer types.

Abstract: 2503 Background: There is controversy around the applicability of TMB across cancer types and reliability between assays from different manufacturers. The KEYNOTE 158 trial supported FDA approval of the Foundation Medicine test (FoundationOne CDx) at TMB ≥ 10 mut/Mb as a solid tumor companion diagnostic (CDx) for single-agent pembrolizumab in 2nd+ line. Using a large real-world dataset with validated rwOS endpoint data, we evaluated the clinical validity of the TMB measurement by the FDA approved test in over 8,000 patients across 24 cancer types who received single agent ICI. Methods: Following a prespecified analysis plan, this study used the nationwide (US-based) de-identified Flatiron Health-Foundation Medicine clinico-genomic database (FH-FMI CGDB), with data originating from approximately 280 US cancer clinics (~800 sites of care). The overall cohort included patients from 24 cancer types of interest with adv/metastatic disease treated with single-agent anti-PD(L)1 therapy in the FH network between 1/1/2011 - 9/30/2022. This study used the TMB algorithm from the FDA-approved test supporting solid tumor CDx and composite mortality variable validated against the national death index. Relative hazards of death by TMB interval were assessed using Cox PH models adjusted for ECOG PS, prior treatment, sex, age, opioid rx pre-therapy, genetic ancestry, and socioeconomic assessment. Multi-solid tumor cohort Cox models were additionally adjusted for MSI status, with baseline hazards stratified by cancer type. Results: 8,440 patients from 24 cancer types met inclusion criteria for the overall cohort. Adjusting for aforementioned factors, the hazards of death {HR, (95% CI)} across the entire cohort, relative to TMB < 5, was 0.95 (0.89 – 1.02) for TMB 5-10, 0.79 (0.73 – 0.85) for TMB 10-20, and 0.52 (0.47 – 0.58) for TMB 20+. Adjusted Cox models comparing TMB ≥ 10 vs. TMB < 10 were pre-specified for cancer types with at least 15 death events in each group: breast 0.51 (0.35 - 0.76), colorectal 0.37 (0.27 - 0.49), cancers of unknown primary 0.52 (0.30 - 0.90), endometrial 0.38 (0.26 - 0.55), gastric 0.46 (0.34 - 0.62), head & neck 0.49 (0.33 - 0.73), melanoma 0.55 (0.43 - 0.71), non-small cell lung 0.76 (0.71 - 0.83), small cell lung 0.69 (0.44 - 1.09), and urothelial 0.63 (0.53 – 0.75). Additional individual tumor types, stratification by MSI status, and time to next treatment associations to be presented at conference. Conclusions: Across a heterogeneous cohort, and within individual cancer types with sufficient power, elevated TMB using the TMB measurement based on the FDA-approved test associated with more favorable rwOS on ICI than for similar patients with lower TMB levels. Completed and ongoing randomized controlled trials are encouraged to report out subgroup analyses comparing ICI vs. comparator arm by TMB level.

Abstract 2159: A pretreatment blood-based proteomic biomarker for enhanced decision-making in non-small cell lung cancer

Abstract: Introduction: Treatment modality selection for metastatic NSCLC patients (immunotherapy alone vs. combination of immunotherapy with chemotherapy) relies mainly on determining the programmed cell death 1 (PD-L1) expression levels in the tumor. However, available assays are only moderately predictive. Here we set to develop PROphet®, a plasma proteomics-based predictive model for informing treatment decisions for NSCLC patients receiving immune checkpoint inhibitor (ICI)-based therapy. Methods: Pre-ICI plasma samples were collected in 12 centers in a clinical trial (NCT04056247) from 367 advanced-stage NSCLC patients. Clinical benefit (CB) was assessed at 12 months based on the occurrence of progression-free survival (PFS) until this time point. Deep proteomic profiling of the plasma samples was performed using aptamer-based technology. A novel machine learning model was developed to determine the CB probability for each patient, and the performance was successfully evaluated on an independent validation set, followed by training and prediction over the entire cohort using cross-validation. The resulting PROphet® score (positive or negative) was determined by setting the median CB rate probability as a threshold. The patients were divided into four subgroups based on their PD-L1 expression level combined with their PROphet® score prediction, and the overall survival (OS) was examined for each subgroup. Results: The PROphet® computational model was evaluated in a blinded manner on a subset of 85 patients and displayed strong predictive capability with area under the curve (AUC) of the receiver operating characteristics (ROC) plot of 0.78 (p-value = 5.00e-05), outperforming a PD-L1-based predictive model (AUC = 0.62; p-value 2.76e-01). When combining PROphet® score with PD-L1 expression levels, four different outcome patterns were identified: (i) Patients with PD-L1 ≥50% and PROphet® negative score, who displayed significantly longer OS when treated with ICI-chemotherapy combination therapy compared to ICI monotherapy and may consider combination therapy. (ii) Patients with PD-L1 ≥50% and PROphet® positive score, who benefit similarly from either treatment modalities, and may consider monotherapy to avoid the potential toxicity of the combination therapy. (iii) Patients with PD-L1<50% and PROphet® negative score, who do not benefit from either treatment modalities and may consider chemotherapy alone or treatment beyond standard of care. (iv) Patients with PD-L1<50% and PROphet® positive score who benefit from combination therapy. Conclusions: Altogether, the PROphet® model, when combined with PD-L1 test, stratifies the patients into four subgroups, providing additional resolution to the PD-L1 biomarker currently used to guide treatment selection. Furthermore, the model accurately predicts CB at 12 months based on proteomic analysis of a pre-treatment plasma sample. Citation Format: Michal Harel, Petros Christopoulos, Coren Lahav, Itamar Sela, Nili Dahan, Niels Reinmuth, Ina Koch, Alona Zer, Mor Moskovitz, Adva Levy-Barda, Michal Lotem, Hovav Nechushtan, Rivka Katzenelson, Abed Agbarya, Mahmoud Abu-Amna, Maya Gottfried, Ido Wolf, Ella Tepper, Yanyan Lou, Raya Leibowitz, Adam P. Dicker, David Gandara, David P. Carbone. A pretreatment blood-based proteomic biomarker for enhanced decision-making in non-small cell lung cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 2159.

A novel decision-making tool for first-line treatment selection in metastatic non-small cell lung cancer based on plasma proteome profiling

Abstract: : Importance: Advanced stage non-small cell lung cancer (NSCLC) patients with no driver mutations are typically treated with immune checkpoint inhibitor (ICI)-based therapy, either in the form of monotherapy or concurrently with chemotherapy, while treatment modality selection is based mainly on programmed death ligand 1 (PD-L1) expression levels in the tumor. However, PD-L1 assays are only moderately predictive of therapeutic benefit. Objective: To develop a novel decision-making tool for physicians treating NSCLC patients on whether to administer immune checkpoint inhibitor (ICI) therapy alone or in combination with chemotherapy. Design, setting, and participants: This multicenter observational study includes patients from an ongoing clinical trial (PROPHETIC; NCT04056247). Patients were recruited from 13 different centers (total n=425; 58 patients were excluded) from June 2016 and June 2021. Plasma samples were obtained prior to treatment initiation, and deep proteomic profiling was conducted. PROphet® computational model for predicting clinical benefit (CB) probability at 12 months was developed based on the plasma proteomic profile. The model performance was validated in a blinded manner. Following validation, training and prediction was performed over the entire cohort using cross-validation methodology. The patients were divided into four groups based on their PD-L1 expression level combined with their CB probability, and the survival outcome was examined for each group. The data were analyzed from July to October 2022. Main outcome and measures: Clinical benefit from ICI-based treatment, overall survival (OS) and progression-free survival (PFS). Results: The model displayed strong predictive capability with an AUC of 0.78 (p-value = 5.00e-05), outperforming a PD-L1-based predictive model (AUC = 0.62; p-value 2.76e-01), and exhibited a significant difference in OS and PFS between patients with low and high CB probabilities. When combining CB probability with PD-L1 expression levels, four patient subgroups were identified; (i) patients with PD-L1 ≥ 50% and a negative PROphet result who significantly benefit from ICI-chemotherapy combination therapy compared to ICI monotherapy; (ii) patients with PD-L1 ≥ 50% and

Phase III Study Comparing Cisplatin Plus Gemcitabine With Cisplatin Plus Pemetrexed in Chemotherapy-Naive Patients With Advanced-Stage Non–Small-Cell Lung Cancer

Abstract: PURPOSE Cisplatin plus gemcitabine is a standard regimen for first-line treatment of advanced non–small-cell lung cancer (NSCLC). Phase II studies of pemetrexed plus platinum compounds have also shown activity in this setting. PATIENTS AND METHODS This noninferiority, phase III, randomized study compared the overall survival between treatment arms using a fixed margin method (hazard ratio [HR] < 1.176) in 1,725 chemotherapy-naive patients with stage IIIB or IV NSCLC and an Eastern Cooperative Oncology Group performance status of 0 to 1. Patients received cisplatin 75 mg/m2 on day 1 and gemcitabine 1,250 mg/m2 on days 1 and 8 (n = 863) or cisplatin 75 mg/m2 and pemetrexed 500 mg/m2 on day 1 (n = 862) every 3 weeks for up to six cycles. RESULTS Overall survival for cisplatin/pemetrexed was noninferior to cisplatin/gemcitabine (median survival, 10.3 v 10.3 months, respectively; HR = 0.94; 95% CI, 0.84 to 1.05). Overall survival was statistically superior for cisplatin/pemetrexed versus cisplatin/gemcitabine in patients with adenocarcinoma (n = 847; 12.6 v 10.9 months, respectively) and large-cell carcinoma histology (n = 153; 10.4 v 6.7 months, respectively). In contrast, in patients with squamous cell histology, there was a significant improvement in survival with cisplatin/gemcitabine versus cisplatin/pemetrexed (n = 473; 10.8 v 9.4 months, respectively). For cisplatin/pemetrexed, rates of grade 3 or 4 neutropenia, anemia, and thrombocytopenia (P ≤ .001); febrile neutropenia (P = .002); and alopecia (P < .001) were significantly lower, whereas grade 3 or 4 nausea (P = .004) was more common. CONCLUSION In advanced NSCLC, cisplatin/pemetrexed provides similar efficacy with better tolerability and more convenient administration than cisplatin/gemcitabine. This is the first prospective phase III study in NSCLC to show survival differences based on histologic type.

A decision-making tool for first-line treatment in advanced non–small-cell lung cancer based on plasma proteome biomarkers.

Abstract: 9122 Background: Initial treatment selection for advanced NSCLC patients without driver mutations is mainly based on evaluating PD-L1 expression levels in the tumor tissue. However, PD-L1 assays are only moderately predictive. In addition, the guidelines for the PD-L1≥50% subpopulation are not definitive, enabling the usage of immune checkpoint inhibitors (ICI) either as a monotherapy or combined with chemotherapy. Here we aim to provide a decision-making tool for the first-line treatment of advanced NSCLC patients based on plasma derived biomarkers obtained before treatment initiation. Methods: Pre-treatment plasma samples were collected from 545 NSCLC patients receiving ICI-based therapy. Clinical benefit was evaluated based on progression-free survival (PFS) at 12 months as a threshold. Deep plasma proteomic profiling was performed using aptamer technology. Based on the proteomic profiles, a computational model, termed PROphet, was developed and evaluated in a blinded manner on the validation subset (n = 272). The model output, clinical benefit probability, was used to stratify the patients into two groups, PROphet-positive or -negative. Results: The model displayed good performance with a high correlation between the predicted clinical benefit and the observed clinical benefit rate (R2= 0.97), outperforming a PD-L1-based prediction model (R2= 0.35). Patients classified as PROphet-positive achieved significantly longer overall survival (OS) than PROphet-negative patients, with a median OS of 25.9 versus 10.8 months (hazard ratio, HR = 0.51; 95% confidence interval, CI = 0.37-0.70; p-value < 0.001). Next, we examined the clinical utility of combining the PROphet output with PD-L1 levels by comparing different treatment modalities. Focusing on PD-L1≥50%, patients with PROphet-positive results displayed similar OS and PFS when receiving ICI monotherapy or ICI-chemotherapy (OS HR = 0.77; CI = 0.42-1.43; p-value = 0.4096), suggesting that these patients may consider monotherapy, thus avoiding chemotherapy-related toxicity. Conversely, PD-L1≥50% patients with a PROphet-negative result displayed a significantly longer OS and PFS when receiving ICI-chemotherapy in comparison to ICI monotherapy, with a median OS that was not reached versus 11.10 months in the combination therapy and monotherapy groups, respectively (HR = 0.29; CI = 0.14-0.59; p-value < 0.001). These patients should consider combination therapy, avoiding a treatment that is suboptimal for them. Results for PD-L1 1-49% and < 1% provide additional clinical information for each subgroup. Conclusions: Plasma proteomic profiling can provide biomarkers that accurately predict the benefit of ICI-based treatment. The model output complements tissue PD-L1 expression levels as a tool to assist therapeutic decisions.