Performance and Clinical Benefit of Comprehensive Genomic Profiling of GI Tumors in a Comprehensive Cancer Center Authors: Giulia Maddalena, Valentina Angerilli, Federico Nichetti, Gianmarco Ricagno, Jessica Gasparello, Elena Mattiuzzo, Maria Caterina De Grandis, Elena Carcea, Carlotta Ceccon, Anna Roma, Eleonora Perissinotto, Sara Sperotto, Giacomo Di Paolo, Marta Sbaraglia, Marco Maruzzo, Francesca Bergamo, Angelo Paolo Dei Tos, Matteo Fassan, Sara Lonardi Published: 2026 ABSTRACT PURPOSEComprehensive genomic profiling (CGP) is increasingly adopted in the management of patients affected by GI cancers. However, the applicability, performance, and clinical utility of CGP in the real-world setting are still undefined.METHODSWe retrospectively evaluated CGP performance and clinical benefit in consecutive patients with GI tumor at the Veneto Institute of Oncology-IRCCS, Padua. We assessed CGP success and its advantage over routine diagnostics in detecting actionable molecular targets. A custom list of gain alterations was defined, including targets not classified as ESMO Scale for Clinical Actionability of molecular Targets tier IA at the time of analysis.RESULTSOf the 1,450 samples, 140 (9.7%) were inadequate for CGP. Failure was mainly due to low quantity of extracted tumor DNA (P = .002). Of the 1,265 metastatic patients, GAs were detected in 355 (28.1%) cases, of whom 55 (15.5%) were treated with targeted therapy. Survival did not differ between patients with no GAs and those with GAs who were not treated accordingly (median overall survival 12.0 v 10.9 months), whereas it was significantly longer for those receiving treatment for actionable GAs (26.4 months). This advantage was confirmed in an exploratory, inverse probability of treatment-weighted analysis (adjusted hazard ratio, 0.72 [95% CI, 0.58 to 0.89]; P = .002). Among 283 (83.7%) untreated patients with complete follow-up, 41.6% did not receive targeted therapy due to lack of clinical trial or failure to meet inclusion criteria.CONCLUSIONAppropriate specimen selection and early molecular assessment at the time of advanced disease diagnosis are essential to detect clinically actionable molecular alterations and therapeutic opportunities in patients with GI cancers. Our results support CGP in comprehensive cancer centers. INTRODUCTION Next-generation sequencing (NGS) of tumor samples is a crucial tool in the shift from a one-size-fits-all paradigm to a tailored therapeutic decision making. In 2018, the ESMO Precision Medicine Working Group defined the ESMO Scale for Clinical Actionability of molecular Targets (ESCAT)1 and drafted the first guidelines to endorse the use of multigene, panel-based NGS in different malignancies to guide treatment with matching drugs.2 Overtime, since the sensitivity of detecting alterations is directly related to sequence coverage, comprehensive genomic profiling (CGP, approximately 300-500 genes) panels have been progressively preferred over limited (approximately 10-50 genes) ones.2-5 From these, clinicians may identify novel actionable molecular targets to increase treatment opportunities with tailored anticancer drugs within clinical trials or with off-label access programs, with the final goal to pursue precision oncology.6 CONTEXT Key Objective How does real-world comprehensive genomic profiling (CGP) perform and affect clinical benefit in patients with GI cancer? This study provides a unique, large-scale retrospective analysis of CGP's practical success rates and treatment barriers. Knowledge Generated Targeting genomic alterations identified through CGP provided a survival benefit for patients with GI cancer compared with standard treatment (adjusted hazard ratio 0.72, P = .002). However, nearly 10% of patient samples failed quality checks due to low DNA quantity, and over 40% of patients with actionable alterations could not receive targeted therapy due to inclusion criteria failures or a lack of available clinical trials. Relevance To maximize the clinical benefit of CGP, tissue specimen has to be carefully selected and tests should be offered at the diagnosis of advanced disease. Additionally, our work stresses the clinical need to expand trial accessibility and develop broader expanded-access programs. Key Objective Key Objective How does real-world comprehensive genomic profiling (CGP) perform and affect clinical benefit in patients with GI cancer? This study provides a unique, large-scale retrospective analysis of CGP's practical success rates and treatment barriers. How does real-world comprehensive genomic profiling (CGP) perform and affect clinical benefit in patients with GI cancer? This study provides a unique, large-scale retrospective analysis of CGP's practical success rates and treatment barriers. Knowledge Generated Knowledge Generated Targeting genomic alterations identified through CGP provided a survival benefit for patients with GI cancer compared with standard treatment (adjusted hazard ratio 0.72, P = .002). However, nearly 10% of patient samples failed quality checks due to low DNA quantity, and over 40% of patients with actionable alterations could not receive targeted therapy due to inclusion criteria failures or a lack of available clinical trials. Targeting genomic alterations identified through CGP provided a survival benefit for patients with GI cancer compared with standard treatment (adjusted hazard ratio 0.72, P = .002). However, nearly 10% of patient samples failed quality checks due to low DNA quantity, and over 40% of patients with actionable alterations could not receive targeted therapy due to inclusion criteria failures or a lack of available clinical trials. Relevance Relevance To maximize the clinical benefit of CGP, tissue specimen has to be carefully selected and tests should be offered at the diagnosis of advanced disease. Additionally, our work stresses the clinical need to expand trial accessibility and develop broader expanded-access programs. To maximize the clinical benefit of CGP, tissue specimen has to be carefully selected and tests should be offered at the diagnosis of advanced disease. Additionally, our work stresses the clinical need to expand trial accessibility and develop broader expanded-access programs. Nonetheless, data on applicability, performance, and impact of CGP beyond standard clinical practice in the real-world setting are limited. About the technique, a number of caveats might negatively affect the actual benefit of the analysis: for instance, inappropriate selection of tumor biospecimens, along with preanalytic and analytic issues, can significantly impair the accuracy of the molecular profiling.7,8 However, estimates of the sensitivity and specificity of CGP techniques were derived mainly from experimental settings, and data on the reproducibility of the performance in real-world practice are very limited. With regard to the clinical implications, many targeted drugs are under investigation in basket or umbrella trials, and a number of compounds or combinations have obtained accelerated approval based on the impressive evidence of efficacy, even without survival outcome results.9,10 Moreover, although the FDA recognized the need for a different regulatory attitude toward benefit-risk analyses within the agnostic paradigm, the Molecular Tumor Board (MTB) has now become the key next step to grant an efficient interpretation and application of CGP results in clinical practice.11,12 Yet, the downstream impact of CGP-based targeted agents in terms of survival outcomes and toxicity profiles remains inconclusive, despite an increasing body of literature for various cancer types.13-15 Within these considerations, precision medicine has gained momentum in the field of GI cancers. In this evolving scenario, CGP has been implemented—and its results exploited—heterogeneously across GI tumor entities, countries, and cancer centers, often relying on molecular prescreening platforms offered as part of clinical trials.16 Here, we evaluated the performance of CGP, its output, and related clinical relevance in a large cohort of patients with GI cancer profiled beyond standard NGS panels within a referral comprehensive cancer center. METHODS This retrospective observational study included consecutive patients with advanced GI cancers treated at the Veneto Institute of Oncology (Italy) between December 2019 and January 2022. Patients with an expected survival longer than 3 months and available tumor tissue were eligible and underwent CGP, using the FoundationOne CDx panel, mainly within clinical trials or after MTB evaluation. Quality assessment considered factors such as DNA quantity and quality, tumor cell content, and tissue adequacy. Cases with insufficient quality were reviewed by pathologists to identify preanalytical issues. Clinical, molecular, and demographic data were collected and updated through February 2024. The study evaluated the added value (gain) of CGP by identifying predefined actionable alterations not routinely assessed. Patients with such alterations could receive targeted therapies based on clinical evidence, mutation characteristics, drug availability, and individual clinical factors. When multiple alterations were present, treatment selection prioritized expected benefit and feasibility. Primary objectives were to assess CGP feasibility and its clinical utility in identifying actionable targets. Secondary objectives included evaluating sample limitations and exploratory outcomes. The Wilcoxon-Mann-Whitney test, Fisher's exact test, or Pearson's chi-square test was used to define bivariate associations of continuous and categorical variables. To account for patient heterogeneity, the study used inverse probability of treatment weighting (IPTW) and propensity score modeling. This statistical approach allowed for a fair comparison of survival outcomes between patients who received targeted therapy and those who did not, adjusting for variables such as age, tumor type, and prior treatments. The study was conducted in accordance with the Declaration of Helsinki, and all patients expressed and signed study-specific informed consent forms. Patients’ treatment was conducted under the auspices of an institutionally validated program, and reviewed and approved by the IRB to ensure compliance with all ethical and safety standards. Detailed information on the study methods is provided in the Supplementary Material. RESULTS Inappropriate Selection/Management of the Biospecimen The study flow chart is reported in Figure 1. In our case study, CGP was proposed to a total of 1,450 patients with GI tumors, of whom 1,310 (90.3%) had samples adequate for sequencing; 946 (65.2%) of 1,450 samples were derived from resection of primary or metastatic lesions, with the remaining 504 (34.8%) derived from tumor biopsies. Study flow chart. CGP, comprehensive genomic profiling. Among the 140 (9.7%) cases evaluated as inadequate and thus failing the analysis, the major causes reported were low tumor nuclei content (47.1% of cases), low DNA quality (26.4%), and low DNA quantity (14.3%). In terms of sample adequacy, CGP analysis failure was proportionally higher in tumor biopsy specimens than in resection samples (P < .001; Data Supplement, Table S2). Reasons for inadequate analysis were similar for biopsy and resection samples, except for low tumor quantity (P = .002), more frequently reported in biopsy samples, as expected (Data Supplement, Table S3). For each specific tumor entity, inadequacy was reported with a different rate between types of samples (Data Supplement, Table S4): of note, inadequate CRC samples were more frequently obtained from surgical resection than biopsy (46.8% v 20.5%), highlighting how quantity is not the only issue for tissue analysis. Furthermore, inadequacy of biopsy samples was more frequent in hepato-biliary-pancreatic tumors (53.8% v 25.8% for surgical resections). We next performed an internal histopathological reevaluation of the specimens that were centrally deemed inadequate: 62 (44.3%) of 140 inadequate cases were samples with low cellularity (<10% of neoplastic cells within the section), nine (6.4%) had extensive necrosis (>80% of the neoplastic component), 10 (7.1%) had fixation artifacts (either hypo- or hyperfixation artifacts), one (0.7%) had freeze artifacts due to intraoperative examination of the specimen, 39 (27.9%) had scarce material (ie, low tissue volume), and 19 (13.6%) had no detectable microscopic alteration. Of the 62 samples with low cellularity, the low neoplastic cell content was due to high mucin content (ie, mucinous neoplasms) in four cases (2.9%). Examples of potential pitfalls in CGP analysis of formalin-fixed, paraffin-embedded (FFPE) biopsy and surgical specimens are shown in Figure 2. Potential pitfalls in NGS analysis of FFPE samples. Representative images of potential pitfalls in CGP analysis of FFPE bioptic and surgical specimens. (A,B) Small biopsy samples composed of hematic material enclosing few cancer cells; (C) adenocarcinoma glands with extensive necrosis; (D) mucinous adenocarcinoma characterized by low cellularity and mucinous acellular component; (E) cautery artifacts; (F) freeze artifacts due to intraoperative examination; (G) evidence of chemoembolization and necrosis; (H) hyperfixation artifacts (ie, clefts between epithelium and surrounding connective tissue); (I) hypofixation artifacts (ie, fading nuclei due to autolysis). CGP, comprehensive genomic profiling; FFPE, formalin-fixed, paraffin-embedded; IPTW, inverse probability of treatment weighting; NGS, next-generation sequencing. Adequate and Low-Quality Report Analysis Among 1,310 adequate specimens, 879 (67.1%) were primary tumors and 431 (32.9%) were metastatic sites. Six hundred twenty-three (70.9%) of the 879 primary tumor tissue samples were resection samples and 256 (29.1%) were biopsies; of the 431 metastasis samples, 260 (60.3%) were resected tissue and 171 (39.6%) were biopsies. The prevalence of sampling type was different in primary and metastatic tissue samples (P = .002), as expected. A total of 252 samples (19.2% of 1,310 adequate specimens), despite having an informative report, were classified as low-quality, 168 (19.0%) of 427 resection specimens and 84 (19.7%) of 883 biopsy samples, without statistically significant differences between the two groups. The major cause reported was insufficient DNA quality (Data Supplement, Table S5). According to the post hoc, internal revision, of the 252 low-quality samples, 72 (28.6%) were samples with low tumor cellularity (<20% tumor cells), which was attributable to high mucin content (ie, mucinous neoplasms) in 21 (8.3%) of them. Other reasons for low-quality report were scarce tumor tissue (<500 tumor cells) in 27 (10.7%), extensive necrosis in 27 (10.7%), fixation artifacts (either hypo- or hyperfixation artifacts) in 19 (7.5%), and freeze artifacts from intraoperative examination in eight (3.2%), whereas 99 (39.3%) had no detectable microscopic pitfalls potentially responsible for the low quality. Tumor Mutational Burden and MSI Gain Rate Overall, microsatellite instability (MSI) and tumor mutational burden (TMB) evaluation failed in 245 (18.5%) and 225 (17.2%) cases, of which 162 (66.1%) and 153 (68.0%) were associated with low-quality reports. Among evaluable cases, 131 (12.4%) were classified as MSI-H, whereas 200 (18.4%) had high TMB (>10 muts/Mb). A detailed summary of MSI status and TMB values is reported in the Data Supplement (Table S6 and Fig S1). Focusing on discordant cases, two (0.2%) patients had MSI-H tumors with low TMB, whereas 66 (7.1%) MSS cases had high TMB; in CRC, this subgroup was 6.4%. Among MSS cases, a significantly higher proportion of upper GI and rare cancers had high TMB (13.3% and 14.3%, respectively), whereas hepato-biliary-pancreatic cases were confirmed to be tumors with low mutational burden (2.6% of cases had high TMB). Of note, median TMB values of MSS patients with >10 muts/Mb (MSS high TMB group) were significantly lower than those of patients with MSI-H (11.4 muts/Mb v 45.4 muts/Mb, P < .001). Seven cases were found harboring POLE/D1 pathogenic mutations.17,18 Of these, three patients had MSI-H tumors; therefore, it is possible that mutations in mismatch repair–related genes were passenger events secondary to the POLE/D1 alteration; one patient had MSS, high TMB CRC and one had high TMB CRC with nonevaluable MSI status (Data Supplement, Fig S1), whereas two cases had nonevaluable MSI or TMB. Molecular Profiling, Therapeutic Opportunities, and Outcomes The study cohort (N = 1,310) counted 1,265 (96.6%) patients with advanced (stage IV) GI tumors; patients with a localized (stage I-III) disease (n = 45) were excluded from further analysis because of the different treatment setting; of note, 44.4% (n = 20) of them carried a gain alteration. A druggable variant was identified thanks to the CGP analysis in 355 (28.1%) cases of advanced disease patients, representing a 24.5% gain in the overall study population. Specific GAs for each group are presented in Figure 3A and in the Data Supplement (Table S7). Gain in information was higher for CRC and upper GI cancers than hepato-biliary-pancreatic and rare subgroups as displayed in Table 1; among patients with identified GAs, prevalence was 51.0% in CRC, 24.5% in hepato-biliary-pancreatic group, 18.6% in upper GI tumors, and 5.9% in rare diseases. (A) Frequency of gain alterations according to tumor site of origin. (B) Overall survival according to the presence of gain alterations and subsequent implementation of targeted therapy in patients with advanced-stage disease and complete follow-up. (C) Kaplan-Meier curves for OS after IPTW adjustment comparing patients with GAs treated versus not treated with targeted therapies. CRC, colorectal cancer; HR, hazard ratio; IPTW, inverse probability of treatment weighting; MSI, microsatellite instability; NA, not applicable; OS, overall survival. TABLE 1.Characteristics of Patients Affected by Metastatic GI Cancers With a Report of CGP (n = 1,310)CharacteristicOverall (N = 1,265)No Gain Alterations (n = 910)Gain Alteration Detected (n = 355)Gain Alterations, Not Treated (n = 299)aGain Alterations, Treated (n = 56)a P Sex, female, No. (%)543 (42.9)399 (43.8)144 (40.6)124 (41.5)20 (35.7).414Age (median, IQR)64.2 (55.1, 71.8)64.6 (55.8, 71.9)61.0 (51.8, 70.0)62.5 (53.5, 72.7)64.1 (54.9, 70.0).443Tumor site, No. (%).008 CRC601 (47.5)420 (46.2)181 (51.0)155 (51.8)26 (46.4) Hepato-biliary-pancreatic375 (29.6)288 (31.6)87 (24.5)75 (25.1)12 (21.4) Rare cancersb95 (7.5)74 (8.1)21 (5.9)13 (4.3)8 (14.3) Upper GI194 (15.3)128 (14.1)66 (18.6)56 (18.7)10 (17.9)Abbreviations: CGP, comprehensive genomic profiling; CRC, colorectal cancer. a One treated patient received a single administration of tailored treatment and was then lost to follow-up; therefore, the patient has been excluded from further evaluations. b Rare tumors include neuroendocrine tumor, appendix, small intestine and anal neoplasia. Characteristics of Patients Affected by Metastatic GI Cancers With a Report of CGP (n = 1,310) Abbreviations: CGP, comprehensive genomic profiling; CRC, colorectal cancer. One treated patient received a single administration of tailored treatment and was then lost to follow-up; therefore, the patient has been excluded from further evaluations. Rare tumors include neuroendocrine tumor, appendix, small intestine and anal neoplasia. Of note, 56 (15.8%) and six (1.7%) patients had two and three concomitant GAs identified, respectively. The most frequently identified GAs were PIK3CA mutations (n = 147, 32.8%), followed by high TMB in the presence of MSS/pMMR status (MSS high TMB, n = 65, 14.5%). In the Data Supplement (Fig S2), we reported four images of FFPE cases with particular molecular alteration discovered thanks to CGP. Clinical gain has been evaluated in 263 of 355 patients harboring a potentially actionable alteration, after excluding 17 patients with GAs who were lost to follow-up (of which one was treated with a tailored agent) and 75 patients with insufficient performance status to be considered for further treatment. The clinical characteristics of the 263 patients with advanced disease harboring GAs, evaluated at the time of CGP request, are reported overall and according to the subsequent implementation of targeted therapies (Treated v Not Treated) in Table 2. Among the 55 treated cases, the most frequently actioned GAs were high TMB in the presence of MSS/pMMR status (MSS high TMB; n = 12, 21.8%), MSI-H status in non-CRC (n = 10, 18.2%), and ERBB2 amplifications (n = 11, 20.0%). In terms of tumor type, tailored treatments have been delivered mostly in CRC (45.5% of treated population), but by proportion, rare and upper GI cancers were the most enriched by the gain in opportunity (50.0% and 23.3%, respectively). Tailored agents were implemented within clinical trials (38.2%) and by institutional out-of-pocket (ie, the treatment was supplied by IOV) off-label prescriptions (36.4%) in most cases, whereas a nominal access program was available for the remaining 25.5% of patients. Conversely, among nontreated cases, the most frequently reported reason for treatment miss was the lack of available clinical trials or off-label/nominal use programs (n = 106, 37.5%), especially for patients affected by hepato-biliary-pancreatic cancers; nonpermissive performance status was a limitation in 26.5% (n = 75) cases (Table 3). At the time of writing, considering the total number of patients with advanced disease (n = 1,265), an additional treatment opportunity has been provided to 4.5% of the population, and for the remaining 89 (7.0%) patients without evidence of progressive disease during ongoing treatment or surveillance, this may represent a future option. TABLE 2.Baseline Characteristics, as Evaluated at the Time of CGP Request, in Metastatic Patients With Identified GAs, Complete Follow-Up, and Grouped According to the Subsequent Implementation of Targeted TherapiesCharacteristicOverall, (N = 263)Not Treated, (n = 208)Treated, (n = 55) P Sex, female, No. (%)102 (38.8)83 (39.9)19 (34.5).468Age (median, IQR)62.4 (53.1, 70.7)61.6 (52.4, 70.8)64.1 (55.4, 70.3).311Tumor site, No. (%).040 CRC139 (52.9)114 (54.8)25 (45.5) Hepato-biliary-pancreatic65 (24.7)53 (25.5)12 (21.8) Rare cancersa16 (6.1)8 (3.8)8 (14.5) Upper GI43 (16.3)33 (15.9)10 (18.2)Number of treatment lines before CGP request, No. (%).166 051 (19.4)41 (19.7)10 (18.2) 187 (33.1)74 (35.6)13 (23.6) 2+125 (47.5)93 (44.7)32 (58.2)ECOG PS at CGP request, No. (%).258 0140 (53.2)107 (51.4)33 (60.0) 1+123 (46.8)101 (48.6)22 (40.0)Gain alteration, No. (%)b AKT mutation1 (0.4)1 (0.5)— BRAF V600E mutation2 (0.8)1 (0.5)1 (1.8) BRAF non-V600E mutation9 (3.4)7 (3.4)2 (3.6) BRCA1/2 mutation15 (5.7)12 (5.8)3 (5.5) EGFR amplification5 (1.9)5 (2.4)— ERBB2 amplification20 (7.6)9 (4.3)11 (20.0) ERBB2 mutation13 (4.9)12 (5.8)1 (1.8) FGFR2 fusion/rearrangement7 (2.7)3 (1.4)4 (7.3) IDH1 mutation16 (6.1)15 (7.2)1 (1.8) KRAS G12C mutation6 (2.3)6 (2.9)— MET amplification9 (3.4)9 (4.3)— MSI-H15 (5.7)5 (2.4)10 (18.2) MSS high TMB36 (13.7)24 (11.5)12 (21.8) NRAS mutation6 (2.3)6 (2.9)— NTRK1/2/3 fusion/rearrangement4 (1.5)3 (1.4)1 (1.8) PIK3CA mutation89 (33.8)82 (39.4)7 (12.7) POLE/POLD1 mutation5 (1.9)4 (1.9)1 (1.8) RET fusion/rearrangement2 (0.8)1 (0.5)1 (1.8) ROS1 fusion/rearrangement3 (1.1)3 (1.4)—ESCAT tier, No. (%) IB16 (6.1)15 (7.2)1 (1.8) IC21 (8.0)9 (4.3)12 (21.8) IIB31 (11.8)16 (7.7)15 (27.3) IIIA173 (65.8)150 (72.1)23 (41.8) IIIB22 (8.4)18 (8.7)4 (7.3)Type of targeted therapy, No. (%) Anti-HER2 mAb or ADC——12 (21.8) ICI——23 (41.8) PARP-i——3 (5.5) TKI or small molecule inhibitor——17 (30.9)NOTE. From the overall population with GAs detected (N = 355), a total of 17 cases were later excluded from the analysis due to loss of follow-up, 16 patients from the nontreated group and one from the treated group: 75 patients were excluded from the comparison because of a poor performance status.Abbreviations: ADC, adenocarcinoma; CGP, comprehensive genomic profiling; ECOG PS, Eastern cooperative oncology group performance status; ESCAT, ESMO Scale for Clinical Actionability of molecular Targets; GA, gain alteration; HER2, human epidermal growth factor receptor 2; ICI, immune checkpoint inhibitors; mAb, monoclonal antibody; MSI, microsatellite instability; PARP-i, poly (ADP-ribose) polymerase inhibitors; TKI, tyrosine kinase inhibitors. a Rare tumors include neuroendocrine tumor, appendix, small intestine, and anal neoplasia. b When more than one GA was identified, only the priority alteration for each patient is reported here. Baseline Characteristics, as Evaluated at the Time of CGP Request, in Metastatic Patients With Identified GAs, Complete Follow-Up, and Grouped According to the Subsequent Implementation of Targeted Therapies NOTE. From the overall population with GAs detected (N = 355), a total of 17 cases were later excluded from the analysis due to loss of follow-up, 16 patients from the nontreated group and one from the treated group: 75 patients were excluded from the comparison because of a poor performance status. Abbreviations: ADC, adenocarcinoma; CGP, comprehensive genomic profiling; ECOG PS, Eastern cooperative oncology group performance status; ESCAT, ESMO Scale for Clinical Actionability of molecular Targets; GA, gain alteration; HER2, human epidermal growth factor receptor 2; ICI, immune checkpoint inhibitors; mAb, monoclonal antibody; MSI, microsatellite instability; PARP-i, poly (ADP-ribose) polymerase inhibitors; TKI, tyrosine kinase inhibitors. Rare tumors include neuroendocrine tumor, appendix, small intestine, and anal neoplasia. When more than one GA was identified, only the priority alteration for each patient is reported here. TABLE 3.Modalities of Targeted Treatment Implementation (top) and Reason for Targeted Therapy Exclusion (bottom) Among Patients With Identified GAs at CGPModality of Targeted TreatmentOverall, N = 55 (%)CRC, n = 25 (%)Hepato-Biliary-Pancreatic, n = 12 (%)Rare Cancers, n = 8 (%)Upper GI, n = 10 (%) Clinical trial21 (38.2)13 (52.0)5 (41.7)3 (37.5)— Nominal use14 (25.5)6 (24.0)4 (33.3)1 (12.5)3 (30.0) Off-label20 (36.4)6 (24.0)3 (25.0)4 (50.0)7 (70.0)Reason for Targeted Therapy ExclusionOverall, N = 283 (%)CRC, n = 140 (%)Hepato-Biliary-Pancreatic, n = 75(%)Rare Cancers, n = 13 (%)Upper GI, n = 55 (%) No clinical trial or off-label/nominal use program available106 (37.5)50 (35.7)38 (50.7)4 (30.8)14 (25.5) Ongoing treatment or in follow-up89 (31.4)58 (41.4)13 (17.3)4 (30.8)14 (25.5) Poor performance status75 (26.5)26 (18.6)22 (29.3)5 (38.5)22 (40.0) Screening failure for trial inclusion/exclusion criteria13 (4.6)6 (4.3)2 (2.7)—5 (9.1)Abbreviations: CGP, comprehensive genomic profiling; CRC, colorectal cancer; GAs, gain alterations. Modalities of Targeted Treatment Implementation (top) and Reason for Targeted Therapy Exclusion (bottom) Among Patients With Identified GAs at CGP Abbreviations: CGP, comprehensive genomic profiling; CRC, colorectal cancer; GAs, gain alterations. After a median follow-up of 22.2 months (IQR, 11.9-37.7), although the survival from metastatic disease is similar in treated versus nontreated patients (HR, 0.84 [95% CI, 0.58-1.22]; P = .352), the observation accounting for time of GCP testing showed significantly longer median overall survival (OS) for patients with GAs who received tailored treatment (HR, 0.56 [95% CI, 0.39-0.80]; P = .001), whereas similar survival was observed between patients with GA who did not receive targeted therapies and those without GAs (HR, 1.04 [95% CI, 0.88-1.23]; P = .612), as shown in Figure 3B. An IPTW-adjusted analysis was performed, aimed at assessing the impact of targeted therapies in an unbiased modality. Patients unable to receive targeted therapy due to poor clinical condition were excluded to avoid biasing results. Before adjustment, significant differences in patients' age, tumor entity, and number of prior treatment lines were detected; as shown in the Data Supplement (Fig S3 and Table S8), IPTW resulted in successful correction of d values. After IPTW, the median OS for patients who received CGP-guided therapy was significantly longer compared with nontreated cases (adjusted HR, 0.72 [95% CI, 0.58-0.89]; P = .002, Fig 3C). Among patients who received targeted therapy, the majority (38.2%) were treated with immune checkpoint inhibitors due to the presence of MSI-H, POLE/POLD1 mutations, and/or MSS high TMB, followed by small molecule inhibitors against genomic fusions/rearrangements (eg, NTRK1/2/3 or FGFR2) or point mutations in genes of interest (eg, BRAF or PIK3CA, Table 2). The median PFS was 15.6 (95% CI, 8.5-25.3) months, whereas the ORR and DCR were 40.0% (with two complete responses and 20 partial responses) and 74.5%, respectively. DISCUSSION Our work presents real-world findings from a large cohort of patients with GI cancers, highlighting the challenges and benefits of CGP in clinical practice, including tissue selection and implementation of targeted therapies. In our study, CGP succeeded in 90.3% of cases, consistent with prior studies19 and comparable with WGS/WES success rates of up to 88%.20 Tumor tissue adequacy remains a significant hurdle for CGP, as FFPE samples often yield suboptimal DNA due to over- or underfixation, leading to degradation, crosslinking, and genomic alterations. Moreover, DNA extraction methods and prior pathological processing can also affect sample quality, especially for small biopsies.6,8,21,22 Recent trials often mandate a new pretreatment biopsy.23 However, a new specimen could not always be the best tissue sample for pathological evaluations, and a new biopsy is not always accepted by patients. This is also exemplified by the testing failure in the WIDE study, in which basal biopsy was requested per protocol.24 These results underscore the need for pathologists to carefully select and annotate the most representative tumor tissue (ie, excluding non-neoplastic tissue, tumor necrosis, and areas with excessive inflammation), when multiple blocks are available.25 Additionally, the integration of liquid biopsy has emerged as a valuable tool for achieving a noninvasive molecular characterization of tumors, also enabling dynamic monitoring of tumor evolution.26 On the basis of our results, clinicians could gain information on nonroutine tumor molecular alterations in 28.1% of patients, with a gain of therapeutic opportunities in 24.5% and an overall new potential treatment for 15.2% of patients. This confirms that implementation of comprehensive molecular assessment in clinical routine significantly outperforms targeted NGS panels, as previously reported.11,12 We also highlighted that 37.5% of patients with actionable alterations did not receive targeted therapy due to a lack of clinical trials or off-label options, similarly to previous reports.27 In this context, maintaining multibasket trials is essential to expand access to innovative therapies, as shown by the ROME study, where patients with advanced tumors were profiled and randomly assigned to standard care or matched therapy.23,28 The study demonstrated a significant improvement in ORR and PFS with targeted therapy compared with SoC, whereas the lack of an OS difference was likely influenced by the high crossover rate (52% of patients) and by the heterogeneity of the treated population. Notably, the authors highlighted several practical hurdles, including limited profiling success, drug availability, and rapid patient deterioration, which precluded further postrandomization therapy. Despite these challenges, ROME provides an ideal trial framework that now needs to be adapted to the complexities of real-world practice. In this regard, our experience supports and extends the ROME findings, contextualizing them within the daily clinical management of GI malignancies. Poor performance status played a crucial role (26.5%) in limiting the implementation of targeted therapies in our cohort as well; therefore, together with the attrition rate and the decreased benefit of anticancer therapies in later treatment lines, our work encourages to take the opportunity to offer CGP early in the treatment timeline to open therapeutic windows and optimize sequencing strategies, also potentially reducing costs associated with repeated biopsies and testing.29. An exploratory prognostic analysis revealed a statistically and clinically significant outcome advantage for patients in whom a GA was identified and targeted treatment was implemented, whereas there was no difference between the remaining patients with or without identified GAs. These data need to be contextualized given the heterogeneity of tumor entity, molecular alterations, and clinical history of the included patients, but they are consistent with the findings from the ROME trial. The prognostic impact of molecular alterations in different tumor entities and stages is well established in literature.30-32 However, for codon-specific GAs considered, prognostic data are limited and not definitive.33,34 To this point, sample size of specific molecular subgroups did not give us the power to investigate survival according to a molecular matching, representing a limitation of our study. However, after IPTW (matching clinical variables) and exclusion of patients with insufficient clinical conditions for targeted treatment implementation, a significant survival benefit from targeted therapies was confirmed. Regarding survival from the diagnosis of metastatic disease, the lack of observed outcome differences should be interpreted in the context of cohort heterogeneity and the potential dilution of treatment effect across the broader treatment trajectory. Furthermore, the overall outcome may also have been influenced from multimodality strategies that emerged over the last decade. Importantly, our study cannot demonstrate a definitive advantage of CGP compared with selective profiling or smaller genomic panels, as it was not designed as a randomized controlled trial to directly compare these approaches. However, through the use of CGP, we were able to identify a broad spectrum of actionable alterations that, both at the time of the study and in most cases till today, are not targeted as standard practice in GI cancers.3 Since tailored therapies have been administered predominantly (over 70%) within clinical trials or through institutional, out-of-pocket off-label prescriptions, our findings support the integration of CGP into the routine clinical practice of comprehensive cancer centers, where study protocols are typically implemented at earlier stages. Across different institutional settings, the added value of CGP compared with more limited, tumor-specific gene panels remains debated and may be influenced by financial considerations, which are largely country-dependent. To date, data suggest that CGP, although associated with higher upfront costs than small selective panels, may be more cost-efficient in identifying patients eligible for personalized treatment.15,35 When considering the health care pathway, which includes services before tumor profiling, the profiling itself, and the subsequent MTB discussion, most costs arise before CGP. In this context, the cost difference between small-panel and comprehensive NGS testing corresponds to a mean impact of 2%-3% on the overall diagnostic pathway, in a scenario where sequencing costs continue to decline.35 Beyond direct clinical costs, the financial and logistical burden of care is exacerbated by the fact that nearly half of patients seek to use private medical evaluations. Furthermore, the prevalence of independent, digital health–seeking behaviors underscores the necessity for more proactive, data-driven precision medicine frameworks.36 Overall, our data confirm that CGP might enhance the identification of actionable molecular alterations and therapeutic opportunities in up to 25% of patients with advanced GI cancers. Specimen selection and early molecular evaluation at the time of advanced disease diagnosis are crucial. Notably, the earlier the evaluation, the greater the chance for patients with preserved clinical conditions to be considered for tailored agents and clinical trials.