Real-World Actionability Analysis of Comprehensive Genomic Profiling Versus Single/Small-Gene Panels Authors: Paul Spin, Bela Bapat, Chad Moretz, Robert Dumanois, Adrienne M Gilligan, Mostafa Shokoohi, Emily Borgundvaag, John Fox, Carlo Bifulco, David Bartlett Published: 2026 ABSTRACT AbstractIntroductionComprehensive genomic profiling (CGP) enables identification of patients eligible for targeted treatments, making it essential in the management of advanced cancer. This retrospective real-world study compared actionable mutations in CGP-tested patients with advanced/metastatic solid tumors to those who received single-gene/small-panel (SP) tests.MethodsPatients aged > 18 years with advanced/metastatic solid tumors (including non-small cell lung cancer (NSCLC), colorectal cancer, prostate cancer, breast cancer, or melanoma) with a CGP or SP test reported between 1 January 2018, and 31 December 2022, were included. OncoKB-derived actionability was compared between the two cohorts. Inverse probability of treatment weighting (IPTW) was used to adjust for baseline characteristics. A weighted generalized linear model with log link was used to report actionability ratio (AR) and 95% confidence intervals (CI).ResultsAmong 406 patients (CGP-tested cohort: 202, SP-tested cohort: 204), approximately half were diagnosed with NSCLC. After adjusting for baseline characteristics, CGP testing detected more actionable alterations than SP: OncoKB level 1 (50.0% versus 31.4%; AR 1.76 [95% CI: 1.24, 2.51]; p = 0.002), OncoKB level 2 (22.8% versus 10.3%; AR 2.53 [95% CI: 1.17, 5.48]; p = 0.018), and OncoKB level 1 or level 2 or level R1 (55.9% versus 41.2%; AR 1.5 [95% CI: 1.11, 2.01]; p = 0.008).ConclusionsCGP testing identified more actionable genetic alterations compared with SP testing methods for patients with advanced/metastatic NSCLC, colorectal cancer, prostate cancer, breast cancer, or melanoma. Expanding reimbursement and coverage for CGP testing as well as expanding CGP use can facilitate equitable treatment access for patients with advanced/metastatic cancer.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1007/s40487-026-00453-8.Keywords: Actionable mutations, Advanced tumors, Comprehensive genomic profiling, Single-gene testing, Small-panel testing Key Summary Points Why carry out this study? Comprehensive genomic profiling (CGP) can identify a wider range of clinically actionable biomarkers than single-gene or small-panel (SP) testing, but there is limited real-world evidence comparing these approaches across multiple advanced or metastatic solid tumors.Patients with advanced or metastatic NSCLC, colorectal cancer, prostate cancer, breast cancer, or melanoma may benefit from broader biomarker testing to help guide treatment decisions.This study examined whether CGP identifies more clinically actionable biomarkers than SP testing in routine clinical practice. What was learned from the study? CGP identified more actionable alterations than SP testing, including OncoKB level 1 biomarkers (50.0% versus 31.4%), level 2 biomarkers (22.8% versus 10.3%), and level 1, 2, or R1 biomarkers overall (55.9% versus 41.2%; weighted actionability ratio 1.50; p = 0.008).These findings suggest that CGP may improve the identification of biomarker-informed treatment opportunities for patients with advanced or metastatic cancer, although further research is needed to evaluate associated clinical outcomes and cost-effectiveness. Introduction Around 200 genomic biomarkers have been identified among various cancer types [1, 2], making genetic testing essential for prognosis and targeted treatments, particularly in advanced cancers [1, 3–7]. Clinical genomic testing often involves single-gene or hotspot testing to detect point mutations and certain insertions or deletions in a single or small-panel of genes. While these tests may be accurate and can be routinely performed in hospitals, they do not assess a large number of genes simultaneously and may miss clinically relevant alterations or rearrangements [8–11]. Comprehensive genomic profiling (CGP) uses next-generation sequencing (NGS) to simultaneously detect multiple types of genomic alterations across a broader number of genes, including pan-tumor biomarkers such as microsatellite instability (MSI) and tumor mutational burden (TMB) [11–14]. CGP testing helps to characterize mutation patterns across related genes in established cancer pathways, thereby informing treatment decisions, prognosis, and monitoring [1, 8, 13, 15]. Despite the depth of genomic information provided by CGP, its use has been limited by higher costs, lack of perceived clinical utility among payers, and inconsistent insurance coverage [16–21]. The reluctance to pay for and use CGP may also reflect a perceived absence of evidence on the economic value of CGP compared with established tests [18, 19, 21, 22]. To date, there is limited published evidence comparing CGP with other standard targeted assays used in clinical practice. To address this gap, this retrospective cohort study compared the identification of actionable mutations in CGP-tested patients with advanced or metastatic solid tumors (non-small cell lung cancer (NSCLC), colorectal cancer, prostate cancer, breast cancer, or melanoma) to those who received SP testing using real-world data. Methods Study Design and Patient Population A retrospective cohort was derived using data from integrated electronic medical record (EMR) database, covering more than 160 million patients from more than 2000 physician offices, ambulatory centers, and more than 500 hospitals across 30 health systems in the USA. The study period was defined as the first diagnosis of advanced or metastatic cancer available in the medical notes to 31 March 2023. The index period used for patient selection was from 1 January 2018, to 31 December 2022. Patients whose first CGP test occurred during the index period were eligible for inclusion in the study, with their index date (time zero of follow-up) defined as the date of their CGP test order date. The follow-up period was defined from CGP test order date to first systemic anticancer pharmaceutical treatment initiation (Fig. 1). Eligible patients were diagnosed with advanced or metastatic solid tumors including NSCLC, colorectal cancer, prostate cancer, breast cancer, or melanoma, based on International Classification of Diseases, Tenth Revision (ICD-10) codes and clinical notes. Patients were excluded if they had less than 3 months of follow-up from the first test order date, defined as no encounters at the same site for 3 or more months postindex. Adult patients who received molecular testing during the index period based on current procedural terminology (CPT) codes (Supplementary Tables S1 and S2) and confirmed through review of the medical notes were categorized into CGP-tested and SP-tested cohorts. Confirmation of CGP testing was determined via curation protocol that required (a) a CPT code for a CGP test; (b) a named genetic marker and test result; (c) a report of NTRK/TMB. Diagram of the study period, follow-up period, and index period. The study period was set from the first advanced or metastatic cancer diagnosis (as far back as available in the medical notes) to 31 March 2023. The index period, from 1 January 2018, to 31 December 2022, refers to patient eligibility, with the first CGP test (index date) occurring during this period. Inclusion and exclusion criteria were applied during the index period. Follow-up was set from the CGP test order date to first systemic anticancer pharmaceutical treatment initiation. CGP comprehensive genomic profiling The small-panel (SP)-tested cohort was drawn from over 19,000 patients. To facilitate manual chart review while ensuring a sufficient pool of control patients would be available for analysis, 5:1 propensity score matching (PSM) [23] using nearest neighbor matching (NNM) without replacement was used to sample SP-tested patients for subsequent chart review and screening for inclusion in the cohort. The cohorts were matched on age at advanced/metastatic diagnosis, geographic region, year of advanced/metastatic disease, and sex. Controls identified via PSM NNM were subsequently screened in five batches, starting with the best available match, to confirm their clinical diagnosis and use of a SP test. The 1:1 matched cohort formed the “analysis set” used for all subsequent data analysis. Study Outcomes The primary outcome of this study was the OncoKB-derived proxy for treatment actionability [24]. OncoKB is a precision oncology knowledge base that provides evidence-based clinical actionability of cancer gene alterations [24]. Of specific interest were OncoKB levels 1, 2, and R1, which indicate an Food and Drug Administration (FDA)-recognized biomarker predictive of response to an FDA-approved drug, a standard of care biomarker recommended by the NCCN or other professional guidelines predictive of response to an FDA-approved drug, and a standard of care biomarker predictive of resistance to an FDA-approved drug, respectively [2, 24, 25]. OncoKB actionability was defined as the gene, alteration, and cancer type either fully (i.e. gene positivity with alteration noted) or partially matching (i.e. gene positivity with one or more alteration not confirmed in medical notes; matching is assumed) OncoKB recommendations. Pan‑tumor biomarkers such as MSI and TMB were captured when explicitly reported in the medical record; however, incomplete documentation limited systematic incorporation of these biomarkers into the primary OncoKB-based endpoint. PD-L1 expression was assessed separately [26]. As a result, clinical actionability was categorized on the basis of biomarker detection as: (a) OncoKB level 1; (b) OncoKB level 2; (c) OncoKB level R1; (d) OncoKB level 1, 2 or R1; and (e) PD-L1, which was reported separately and not included in the definition of level 1, level 2, and level R1. Covariates The covariates adjusted for included age at the time of test, sex, geographic region, payer type, cancer type, number of metastasis locations, Charlson Comorbidity Index (CCI) score excluding malignancy [27], American Joint Committee on Cancer (AJCC) staging, Eastern Cooperative Oncology Group performance status (ECOG PS), prior number of lines of treatment (LOT), and prior treatment (chemotherapy, targeted therapy, immunotherapy, other). Statistical Analyses Covariates with less than 50% missing data in the matched dataset were imputed using multiple imputation by chained equations (MICE) with the predictive mean matching (PMM) algorithm assuming data were missing at random [28, 29]. A total of 30 imputations were performed on covariates listed in Table 1. To balance covariates between the CGP-tested and SP-tested cohorts, inverse probability of treatment weighting (IPTW) was applied using the average treatment effect (ATE) estimate [30]. Weights were calculated as the inverse probability of being assigned to CGP, conditioning on the observed covariates. To avoid extreme weights, the weights were stabilized by multiplying the IPTW weights by the marginal probability of being assigned to CGP. Weights were then averaged across the 30 imputed datasets to generate a single weight per patient in comparative analysis. Absolute standardized mean differences (ASMD) before and after weighting were reported, with ASMD < 0.10 indicating good balance and < 0.20 indicating moderate balance. A generalized linear model with log link was applied to compare actionability between cohorts. Both unweighted and weighted actionability ratio (i.e., risk ratio) and their 95% confidence intervals (CI) were reported. All analyses were performed using R version 4.3.3. Table 1Overview of baseline characteristics for CGP-tested patients versus SP-tested patients before and after adjustment with IPTWBefore IPTWaAfter IPTWCGP-tested cohort(n = 202)SP-tested cohort(n = 204)ASMDbCGP-tested cohort(ESS = 82)SP-tested cohort(ESS = 122)ASMDbAge at index date, years, mean (SD)65.5 (10.2)60.2 (12)0.47762.6 (10.8)62 (11.3)0.050Sex, (%)0.0040.023Female58.9%59.3%56.9%59.2%Male41.1%40.7%43.1%40.8%Region, (%)North Central1.6%10.5%0.3836.7%6.5%0.009North East10.2%14.1%0.12012%12.4%0.013South74.3%25.1%1.14450%47.4%0.06West13.9%50.3%−0.85531.3%33.6%0.056Payer type, (%)Commercial15.1%33.2%0.43627.5%26.7%0.021Medicaid9.1%10.8%0.0578.5%11%0.083Medicare55%37.2%0.36743%41.2%0.038Self-pay20.8%18.8%0.05120.9%21.1%0.006Cancer type, (%)Breast20.8%20.1%0.00721.8%23.1%0.013Colorectal21.8%21.1%0.00723.2%20.3%0.028NSCLC48.5%50%0.01546.5%49.3%0.028Melanoma3%4.4%0.0142.5%3.3%0.008Prostate5.9%4.4%0.0156%4%0.02Number of metastases, mean (SD)2.3 (1.3)2.2 (1.5)0.0222.3 (1.3)2.2 (1.5)0.069CCI (without malignancies), mean (SD)2.9 (2.2)1.3 (1.6)0.8512 (2.1)1.8 (1.9)0.145AJCCI7.7%5.8%0.0887.2%7.6%0.017II14.8%14.2%0.01712.9%13.3%0.013III29.4%28.2%0.03232.3%32%0.007IV48.1%51.8%0.08047.6%47.1%0.011ECOG PS0–182.8%83.3%0.01786%84.7%0.0432+ 17.2%16.7%0.01714%15.3%0.043Prior number of LOT, mean (SD)0.7 (0.9)0.4 (0.6)0.4110.5 (0.7)0.5 (0.6)0.086Prior treatmentNaive48.5%65.2%0.16757%57.5%0.005Experienced51.5%34.8%0.41143%42.5%0.086AJCC American Joint Committee on Cancer, ASMD absolute standardized mean difference, CCI Charlson Comorbidity Index, CGP comprehensive genomic profiling, ECOG PS Eastern Cooperative Oncology Group performance score, ESS effective sample size, IPTW inverse probability treatment weighting, LOT line of therapy, n sample size, NSCL non-small cell lung cancer, SD standard deviation, SP single-gene or small-panelaData were imputed before IPTWbASMDs < 0.10 indicate well-balance distribution between study groups, and ASMD between 0.11 to < 0.20 indicate moderate balance Overview of baseline characteristics for CGP-tested patients versus SP-tested patients before and after adjustment with IPTW AJCC American Joint Committee on Cancer, ASMD absolute standardized mean difference, CCI Charlson Comorbidity Index, CGP comprehensive genomic profiling, ECOG PS Eastern Cooperative Oncology Group performance score, ESS effective sample size, IPTW inverse probability treatment weighting, LOT line of therapy, n sample size, NSCL non-small cell lung cancer, SD standard deviation, SP single-gene or small-panel aData were imputed before IPTW bASMDs < 0.10 indicate well-balance distribution between study groups, and ASMD between 0.11 to < 0.20 indicate moderate balance Ethical Approval All data used in the study have been certified as statistically deidentified by an independent third-party such that, this study does not require review and approval by ethics committees. Informed consent and consent for publication were not applicable as this was a retrospective study without active patient recruitment, and all retrieved clinical data were deidentified. The study was conducted in accordance with the Declaration of Helsinki of 1964, and its later amendments. Results Patient Identification and Sample Size A total of 215 CGP-tested patients were identified. After conducting the 5:1 PSM NNM, 1075 SP-tested patients were initially identified as possible controls, subject to further screening based on medical notes. If a patient in the SP-tested cohort from the first matched set was excluded during screening, another patient with the same tumor type was selected from the next matched set. The final cohort for comparative analysis between the two cohorts consisted of 406 patients, with 202 patients in the CGP-tested cohort and 204 patients in the SP-tested cohort. Baseline Variables Nearly half of the patients in both groups were diagnosed with NSCLC (n = 98 [48.5%] in CGP-tested versus n = 102 [50%] in SP-tested), followed by colorectal cancer (n = 44 [21.8%] versus n = 43 [21.1%]), breast cancer (n = 42 [20.8%] versus n = 41 [20.1%]), prostate cancer (n = 12 [5.9%] versus n = 9 [4.4%]), and melanoma (n = 6 [3%] versus n = 9 [4.4%]). Baseline characteristics before and after IPTW adjustment are presented in Table 1. After adjustment by IPTW, a well-balanced distribution was achieved between the CGP-tested and SP-tested cohorts for age (weighted mean: 62.6 versus 62.0 years; ASMD: 0.050), sex (weighted %: 56.9% versus 59.2% female sex, ASMD: 0.023), geographical region (e.g., weighted % in Southern region 50% versus 47.4%; ASMD: 0.060), payer type (e.g., weighted % with Medicare: 43% versus 41.2%; ASMD = 0.006), tumor type (e.g., weighted % with lung cancer: 46.5% versus 49.3%; ASMD = 0.028); number of metastases (weighted mean: 2.3 versus 2.2; ASMD: 0.069); AJCC stages (e.g., weighted % at stage IV: 47.6% versus 47.1%; ASMD: 0.011); ECOG PS (e.g., weighted % with ECOG PS 0 or 1: 86% versus 84.7%; ASMD: 0.043) and prior number of LOT (weighted mean: 0.5 versus 0.5; ASMD = 0.086), and previous treatment (e.g., weighted % treatment-naive: 57% versus 57.5%; ASMD: 0.005), while the malignancy-free CCI had a moderate balance (weighted mean: 2 versus 1.8; ASMD: 0.145). Actionability Patients in the CGP-tested cohort had a significantly higher percentage of actionable findings compared with patients in the SP-tested cohort (Table 2), with 113 (55.9%) of the CGP-tested cohort having a positive biomarker result that was OncoKB level 1 or level 2 or level R1 eligible compared with 84 (41.2%) in the SP-tested cohort (weighted actionability ratio [wAR] 1.5 [95% CI: 1.11, 2.01]; p = 0.008). Half (n = 101 [50.0%]) of patients in the CGP cohort had OncoKB level 1 eligibility compared with one-third (n = 64 [31.4%]) in the SP-tested cohort (wAR 1.76 [95% CI: 1.24, 2.51]; p = 0.002). Over twice as many patients in the CGP-tested cohort reported OncoKB level 2 eligibility compared with the SP-tested cohort (n = 46 [22.8%] versus n = 21 [10.3%], respectively; wAR: 2.53 [95% CI: 1.17, 5.48]; p = 0.018). There was no difference in the proportion of positive test results detecting level R1 eligibility (n = 26 [12.9%] versus n = 26 [12.7%]; wAR: 1.2 [95% CI: 0.53, 2.74]; p = 0.663). There were more positive tests detecting PD-L1 biomarker in the CGP-tested cohort (n = 86 [42.6%]) compared with the SP-tested cohort (n = 59 [28.9%]), though this difference was not statistically significant (wAR 1.41 [95% CI: 0.95, 2.09]; p = 0.093).The prevalence (%) of actionable biomarkers varied between cohorts and across cancer types (Fig. 2). Among patients with an actionable result (OncoKB level 1, 2, or R1), the distribution of biomarkers was broadly similar between CGP-tested and SP-tested cohorts for frequently observed genes such as KRAS (n = 30 [26.5%] versus n = 24 [28.6%]), BRAF (n = 17 [15.0%] versus n = 12 [14.3%]), and EGFR (n = 25 [22.1%] versus n = 29 [34.5%]). However, CGP testing identified additional actionable alterations across a wider range of biomarkers, including ERBB2, PIK3CA, and BRCA2, which were less frequently observed in the SP-tested cohort. Tumor-specific analyses demonstrated similar patterns across cancer types, with CGP identifying a broader spectrum of actionable alterations beyond those most commonly observed with SP testing. (Supplementary Tables S3–S8). Table 2Analysis of actionable biomarkers and recommended treatment in level 1, level 2, R1, and PD-L1 in CGP-tested and SP-tested cohortsActionable alterationsCGP-tested cohort(N = 202)SP-tested cohort(N = 204)Unweighted estimatesIPTW estimatesn (%)n (%)Actionability ratio (95% CI)pActionability ratio (95% CI)pOncoKB level 1101 (50.0%)64 (31.4%)1.59 (1.25, 2.04) < 0.0011.76 (1.24, 2.51)0.002OncoKB level 246 (22.8%)21 (10.3%)2.21 (1.37, 3.57)0.0012.53 (1.17, 5.48)0.018OncoKB R126 (12.9%)26 (12.7%)1.01 (0.61, 1.68)0.9701.2 (0.53, 2.74)0.663OncoKB levels 1 or 2 or R1113 (55.9%)84 (41.2%)1.36 (1.11, 1.67)0.0031.5 (1.11, 2.01)0.008PD-L186 (42.6%)59 (28.9%)1.47 (1.13, 1.93)0.0051.41 (0.95, 2.09)0.093OncoKB level 1 is defined as an FDA-recognized biomarker predictive of response to an FDA-approved drug. OncoKB level 2 is defined as a standard of care biomarker recommended by the NCCN or other professional guidelines predictive of response to an FDA-approved drug. OncoKB level R1 is defined as a standard of care biomarker predictive of resistance to an FDA-approved drug. PD-L1 is reported separately in these analyses and not captured in OncoKB level 1, OncoKB level 2, OncoKB level R1, or OncoKB levels 1 or 2 or R1CGP comprehensive genomic profiling, CI confidence intervals, IPTW inverse probability of treatment weighting, N sample size, n number with event; p -value, PD-L1 programmed death-ligand 1, SP single-gene or small-panel Analysis of actionable biomarkers and recommended treatment in level 1, level 2, R1, and PD-L1 in CGP-tested and SP-tested cohorts OncoKB level 1 is defined as an FDA-recognized biomarker predictive of response to an FDA-approved drug. OncoKB level 2 is defined as a standard of care biomarker recommended by the NCCN or other professional guidelines predictive of response to an FDA-approved drug. OncoKB level R1 is defined as a standard of care biomarker predictive of resistance to an FDA-approved drug. PD-L1 is reported separately in these analyses and not captured in OncoKB level 1, OncoKB level 2, OncoKB level R1, or OncoKB levels 1 or 2 or R1 CGP comprehensive genomic profiling, CI confidence intervals, IPTW inverse probability of treatment weighting, N sample size, n number with event; p -value, PD-L1 programmed death-ligand 1, SP single-gene or small-panel Frequency map of the number of cases per tumor type with actional mutations in OncoKB level 1 and 2 categories in both CGP-tested and SP-tested cohorts. CGP comprehensive genomic profiling, N sample size, SP single-gene or small-panel Discussion High-quality evidence comparing CGP with other standard targeted assays traditionally used in clinical practice is limited. The present retrospective study aimed to help fill this gap by comparing the actionability of identified mutations and subsequent recommended treatment between cancer patients (lung, breast, colorectal, melanoma, prostate) tested with CGP or SP testing. CGP-testing was found to be associated with the identification of more actionable genetic results across a number of advanced cancer types compared with patients who underwent SP testing. These findings help to further support the potential adoption and coverage of CGP in routine clinical practice beyond NSCLC. Other published literature in both clinical and real-world settings have assessed the impact of CGP testing versus SP testing with findings generally consistent with the present study. In contrast, however, the majority focused on only patients with primary NSCLC. In one prospective single-center study with CGP testing of 134 patients, additional actionable alterations beyond those identified previously in SP tests were found in 31% of patients and could directly impact treatment decisions and clinical trial eligibility [31]. A similar study reported that, among 561 patients with NSCLC who had received negative SP test results, 46% received positive CGP results for actionable biomarkers missed on single-gene testing. In fact, initial SP testing was found to use up much of the tumor tissue sample, resulting in the inability to conduct CGP at all or higher sequencing failures and turnaround times [32]. Similar to the current study, Bapat et al. in 2022 found that CGP testing was associated with a significantly higher proportion of patients with > 1 actionable biomarker compared to SP (34% versus 15%; p < 0.001), in patients with advanced/metastatic NSCLC treated in a United States (US) community health setting. The proportion of patients receiving approved targeted therapy or immunotherapy (9% versus 3%; p < 0.001), and being eligible for clinical trials (56% versus 4%; p < 0.001) was also significantly higher with CGP than SP testing [33]. Wallenta and colleagues reported similar results in that CGP identified actionable biomarkers at more than twice the rate of single panel testing (32% versus 14%; p < 0.001), with a higher chance of receiving matched treatment in any line. Furthermore, patients who received matched therapies had improved survival compared with those who did not [26]. A major distinction of the present study compared with previously published literature is the inclusion of patients with breast cancer, colorectal cancer, prostate cancer, and melanoma. The reported findings not only align with previous results in NSCLC as summarized above [26, 31–33], but also demonstrates that the benefits of CGP in terms of biomarker actionability extend to other cancer types as well. A more granular evaluation of actionable biomarker distributions provides further context for these findings. Across all tumor types, commonly observed actionable alterations—including KRAS, EGFR, and BRAF—were identified at similar frequencies between CGP- and SP-tested cohorts. However, CGP testing identified a broader range of actionable alterations across additional biomarker classes, including ERBB2, PIK3CA, and DNA damage repair genes such as BRCA2. Tumor-specific analyses demonstrated similar patterns, with CGP consistently capturing actionable alterations beyond those most frequently observed in SP testing. The current coverage for CGP in the USA is limited, with coding practices not fully refined [34, 35]. Previous studies have reported low rates of reimbursement for CGP- or NGS-based testing, despite the inclusion of associated billing codes [35]. Variable coverage may also reinforce existing racial and socio-economic inequities in access to genetic testing [36–38]. The findings of more actionable genetic results and the receipt of a matched recommended treatment with CGP testing in the present study, in combination with other published analyses indicating minimal cost and budget impact, suggest that CGP testing warrants consideration for expanded coverage. Extending coverage of CGP and the removal of barriers such as prior authorization, could expand access to CGP testing to more patients with advanced or metastatic cancer, potentially contributing to significant added diagnostic, treatment, and potential economic value. Further studies are needed to quantify the economic value and clinical impact of expanded CGP access. Community oncologists who deal with many cancer types on a daily basis may struggle to stay current with which small-gene panel to order for which cancers. They may also be limited by insurance coverage determinations to define panel choices. Often, the concern for delay in processing over insurance denials can lead to a smaller panel being used inappropriately or suboptimal treatment pathways relative to CGP-guided targeted treatments. Routine use of CGP for metastatic cancer could potentially reduce these testing inefficiencies. Also, existing small-gene panels can evolve into bigger gene panels. For example, ThyroSeq expanded from a 7-gene panel in 2011 to a 56 gene panel in 2014, and a 112 gene panel in 2017 [39]. Using CGP will avoid wasted time and effort in validating and marketing each new panel individualized for each cancer type as new data emerges on the significance of specific mutations. Patients often drive testing as they seek eligibility for clinical trials. This leads to unnecessary duplication where a small-gene panel has to be updated to CGP for a comprehensive assessment of clinical trial eligibility. In an older study examining the impact of CGP, the authors identified 61.3% of 333 patients having an off-label therapy available, and 77.9% were potentially eligible for a clinical trial based on CGP results [40]. The timeliness of CGP is important as well, where a delay in receiving results can impact the feasibility of using the results. Tredan et al. performed a randomized controlled trial of CGP versus a limited panel of 87 genes demonstrating an increase of 14.8% in molecular-based recommended therapies [41]. However, out of 192 patients with molecular-based recommended therapies, only 43 received directed therapy. The authors emphasize the importance of “tight timeframes” for these patients who have failed multiple treatments and are deteriorating. Early, routine use of CGP in patients diagnosed with metastatic cancer would avoid last minute tests at a critical time in decision-making where survival time is limited. A major strength of this analysis is the use of real-world data which typically allows for the inclusion of a broader patient population than captured in clinical trials or a single academic institution [42, 43]. Thus, our findings may be more representative of the target population in real-world clinical practice. Additionally, the present analysis used clinically rich EMR data and employed data curation methodologies to obtain a real-world sample. The use of EMR data has not previously been used widely in this space; however, the consistency with published literature helps to validate the current findings as well as the novel data and methods employed [26, 31–33]. Furthermore, prior to this study, limited research had been published comparing CGP to SP testing across advanced cancer types, with the present study being the first published evidence for some. This study has important limitations inherent in the use of routinely collected EMR data. Because this was a retrospective study based on routinely documented results in unstructured EMR notes, detailed analytical characteristics of the underlying molecular assays (e.g., gene content, genomic footprint, depth of coverage, and bioinformatic pipelines) were not consistently available across patients and could not be standardized. Accordingly, CGP and SP testing were defined operationally using CPT codes and manual chart abstraction of reported biomarkers and results, reflecting how EMR systems selectively capture genomic test data. In addition, actionability frameworks differ across clinical guidelines [44]. Because the present analysis relied on unstructured EMR documentation, we used a single framework (OncoKB) to enable consistent abstraction and comparison across cohorts [25]. Future research using richer structured molecular data should examine concordance across multiple frameworks. Notably, a high degree of missing data (more than 50%) was present for baseline variables (race, ethnicity, pathology). These variables were excluded from population adjustment, which limits our ability to assess whether findings differ across demographic subgroups. Furthermore, heterogeneity and gaps within the EMR data lead to challenges in identifying eligible patients, resulting in fewer-than-intended patients being included in the final analyses. The small sample has reduced statistical power and precision and could limit the generalizability of findings, particularly for less-represented tumor types. Additionally, as the data for these analyses span from 2018 to 2022, they may not fully reflect current testing patterns. Given the small sample size, it was not feasible to obtain reliable temporal trends. Additional analyses with higher sample size, perhaps by obtaining data from multiple sources or databases, should be conducted in the future to further expand on these findings. EMR information is designed to support the delivery of health care and billing rather than research purposes, which can create challenges. Importantly, the dataset lacked clinical outcomes (e.g., progression-free survival, overall survival) and cost data, preventing assessment of whether increased actionable biomarker detection translates to improved patient outcomes or represents good economic value. As such, EMR-derived data may contain gaps and coding errors. Despite this, the value of EMR data for research has been well established and successfully used in a wide range of investigations, studies, and disease areas [45, 46]. The present study also employed a comprehensive and iterative process to address potential informational biases, such as confirming the identification of CGP testing with CPT codes as well as a review of the medical notes for validation. It is also important to acknowledge that SP testing may remain clinically appropriate in certain scenarios. For example, when clinical presentation strongly suggests a specific actionable mutation with high pretest probability (such as EGFR mutations in never-smoking East-Asian patients with lung adenocarcinoma [47]), rapid SP testing may provide timely results to guide immediate treatment decisions. Additionally, SP testing may be preferred when tissue quantity is limited and a specific, high-probability target needs to be evaluated first. The decision between CGP and SP testing should consider clinical context, urgency of treatment initiation, tissue availability, and the probability of identifying actionable alterations beyond commonly targeted mutations. Our findings suggest that CGP offers advantages in detecting a broader range of actionable biomarkers, but testing strategy should be individualized based on clinical circumstances. Conclusions CGP is the next step employing NGS technology in the management of advanced and metastatic cancers, shifting beyond the limitations of SP testing to a more comprehensive and precise approach. In the current study, CGP testing was found to identify more actionable genetic results than SP testing methods. This suggests that clinicians can connect more patients with targeted therapies that can help improve overall survival. These results suggest that expanded reimbursement could help to achieve optimal and personalized treatments where clinically indicated, though further studies are needed to incorporate clinical outcomes and weigh these potential benefits against the relative cost of CGP testing. Supplementary Information Below is the link to the electronic supplementary material. Supplementary file1 (PDF 444 KB) Acknowledgements The authors would like to thank Naleen Raj Bhandari, Alexandra Hall, and Mary Nesline for their support in this project. Medical Writing, Editorial, and Other Assistance Editorial assistance in the preparation of this article was provided by Shiva Shahabi of EVERSANA and was funded by the Conafay Group. Author Contributions Conceptualization: John Fox, David Bartlett, Robert Dumanois, Carlo Bifulco, Paul Spin, Mostafa Shokoohi, Bela Bapat, and Chad Moretz; Methodology: John Fox, David Bartlett, Robert Dumanois, Carlo Bifulco, Paul Spin, Mostafa Shokoohi, Bela Bapat, and Chad Moretz; Formal analysis and investigation: John Fox, Robert Dumanois, Adrienne M. Gilligan, Emily Borgundvaag, Paul Spin, Mostafa Shokoohi, Bela Bapat, and Chad Moretz; Writing—original draft preparation: John Fox, David Bartlett, Adrienne M. Gilligan, Carlo Bifulco, Emily Borgundvaag, Paul Spin, Mostafa Shokoohi, Bela Bapat, and Chad Moretz; Writing—review and editing: John Fox, David Bartlett, Robert Dumanois, Adrienne M. Gilligan, Carlo Bifulco, Paul Spin, Mostafa Shokoohi, Bela Bapat, and Chad Moretz; Funding acquisition: Robert Dumanois; Resources: John Fox, David Bartlett, and Bela Bapat; Supervision: Paul Spin, Mostafa Shokoohi, Bela Bapat, and Chad Moretz. All authors read and approved the final manuscript. Funding This study was funded by the Conafay Group. The Rapid Service Fee was funded by the Conafay Group. Data Availability The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. Declarations Conflict of Interest John Fox is a full-time employee at Illumina, Inc., manufacturer of sequencers for both large and small panels discussed in this manuscript. Bela Bapat is a full-time employee and stock owner at Illumina, Inc., and declares that Illumina, Inc. has no financial interest in or financial conflict with the subject matter discussed in the manuscript. Carlo Bifulco reports stock ownership and serving on scientific advisory board of PrimeVax, BioAI, and Pictor Labs, serving on scientific advisory board of Lunaphore, consulting or advisory relationships (compensated) at Sanofi, Agilent, Roche, Incendia, Bristol Myers Squibb (BMS), and receiving institutional research support from Illumina, Inc. These disclosures are unrelated to the preparation of the current manuscript, with the exception of the institutional research support from Illumina Inc., and no entity had any influence over the research content or interpretation. Robert Dumanois is an employee and stock owner of Thermo Fisher Scientific and declares that Thermo Fisher Scientific does not have any financial interest in or financial conflict with the subject matter discussed in the manuscript. Adrienne M. Gilligan is an employee and stock owner of Eli Lilly and Company and declares that Eli Lilly and Company does not have any financial interest in or financial conflict with the subject matter discussed in the manuscript. Paul Spin, Mostafa Shokoohi, and Emily Borgundvaag are employees of EVERSANA, which consults for biomedical and pharmaceutical companies, and declare that EVERSANA has no financial interest in or financial conflict with the subject matter discussed in the manuscript. Chad Moretz and David Bartlett declare that they have no competing interests. Ethical Approval All data used in the study have been certified as statistically de-identified by an independent third-party such that, this study does not require review and approval by ethics committees. Informed consent and consent for publication were not applicable as this was a retrospective study without active patient recruitment, and all retrieved clinical data were deidentified. The study was conducted in accordance with the Helsinki Declaration of 1964, and its later amendments. References Associated Data Supplementary Materials Supplementary file1 (PDF 444 KB) Data Availability Statement The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.