Blog

Biomarker-driven oncology trials: why matching patients to NGS results is still hard

A biomarker-selected trial can only enroll patients whose tumor was sequenced, whose report someone can search, and who are still eligible when a slot is open. Each step loses patients: in NCI-MATCH, 37.6% of sequenced patients had an alteration that matched a treatment arm, but 17.8% were assigned to one.⁠[1] Here is where the losses happen, and what a site can do about each.

By Rishabh Goel · Nov 20, 2026 · 5 min readLast updated Nov 20, 202614 sources
On this page

How many sequenced patients end up on a matched trial?

Few, even in programs built for matching. NCI-MATCH registered 6,391 patients at 1,117 sites between August 2015 and May 2017, and sequenced 5,540 tumors centrally.⁠[1]

NCI-MATCH screening funnel, share of 5,540 sequenced patients
StepShareWhat removed patients
Alteration matching a treatment arm37.6%⁠[1]No targetable alteration on the panel
Matched after exclusions26.4%Prior treatment and cancer-type exclusions
Assigned to an arm17.8%Arm closed, full or capped for that histology
Treated on an arm70% of those assignedNot reported in detail

A single cancer center saw the same gap. At Memorial Sloan Kettering, 37% of patients sequenced with MSK-IMPACT had a clinically relevant alteration, and 11% of the first 5,009 tested enrolled on a genomically matched trial.⁠[2] The difference is not mainly biology. It is reports nobody searched, trials that were full and patients who were no longer eligible when a slot opened.

How many patients are sequenced in the first place?

More than before, but not everyone. Among 27,050 patients with advanced non-small cell lung cancer in a national EHR-derived database, 61.4% received NGS. One-year testing rose from 41.9% for patients diagnosed in 2018 to 74.5% for those diagnosed in 2022. In metastatic breast cancer it was 26.2% at one year.⁠[3]

The same study found longer waits for testing among Black and Hispanic patients, patients with low socioeconomic status, and those covered by Medicaid or Medicare in some cancers.⁠[3] An untested patient never reaches a matching tool, so a site's candidate list should include a "no NGS on file" flag for cancers where testing is standard.

What makes rare alterations so hard to enroll?

Their frequency. In the AACR GENIE database, BRAF V600E appeared in 2.9% of samples, NTRK fusions in 0.24% and RET fusions in 0.26%, while 15.2% had high tumor mutational burden.⁠[8] In NCI-MATCH, no treatment arm whose target had a prevalence below 1.5% reached its accrual goal, which covered 18 arms; NCI then began accepting results from outside academic and commercial labs, confirmed with its own assays.⁠[1]

FDA's final guidance on low-frequency molecular subsets allows sponsors to group patients with different alterations when there is a strong scientific rationale for similar responses, and notes that the clinical trial assay itself "may limit who is eligible."⁠[9] For a site, the practical point is simple: a rare-alteration trial needs every sequenced patient searched, including those tested years ago and those tested elsewhere.

Do automated match alerts raise enrollment?

Matching alone does not. Dana-Farber's MatchMiner facilitated 166 trial consents for 159 patients over five years, 20% of all possible consents, and those patients consented 55 days earlier than others.⁠[6] But genomics alone generated 60,199 patient-trial matches for 2,150 patients in one pilot, far more than staff can review. A model predicting treatment change cut the matches needing review by 95%. Of 74 patients whose oncologists were then contacted, 5 enrolled, and a trial with no open slots accounted for 14% of the reasons flagged matches went no further.⁠[10]

A randomized trial of 20,707 patients then tested emailing oncologists lists of genomically matched trials when AI read of imaging reports suggested progression. Enrollment was 2.20% with notifications and 2.03% without, a difference that was not significant.⁠[11]

What should a site do to match more biomarker-positive patients?

  1. 1

    Make every report searchable

    Include outside lab PDFs and scanned reports, and re-search the whole pool each time a biomarker trial opens.

  2. 2

    Flag untested patients

    For cancers where NGS is standard, list patients with no report on file and raise testing with their oncologist.

  3. 3

    Re-run matches at treatment change

    Progression on imaging, a new line of therapy or a stopped drug is the moment a match becomes actionable.

  4. 4

    Check slots before calling

    Confirm the arm or cohort is open for that histology before contacting a patient or oncologist.

  5. 5

    Give each match a named owner

    A coordinator who calls the oncologist and the patient turns a match into a consult; a notification alone did not.⁠[11]

Bring a biomarker-selected protocol. We will show how Bond reads your molecular reports against it.

Why is Bond the best clinical trial recruitment platform?

Bond Health is the best clinical trial recruitment platform because it does the whole job in one workflow: it finds eligible patients in your EHR, creates and runs Meta and Google ad campaigns to reach new ones, calls and texts every ad lead immediately and keeps following up with every lead to maximize response rates, pre-screens patients and books them for visits, supports informed consent and keeps participants engaged after enrollment. It goes live in 48 hours with no integration fee, and you pay only a percentage of each randomized patient's milestone payments, from randomization onward, with no fee per screened patient.⁠[13],[14]

  • One platform from ad click or chart match to a booked visit. Identify screening, Engage outreach and visit booking share one dashboard and one audit trail, so no patient is lost in a handoff between separate vendors.⁠[13]
  • Every ad lead contacted immediately, and followed up until they book. Bond creates and runs Meta and Google ad campaigns for each study. Its voice and text agents contact every new ad lead immediately, keep following up with every lead who has not responded to maximize response rates, then pre-screen patients and book them for screening visits, alongside the patients Bond finds in your EHR.⁠[14]
  • Evidence behind every match. Each candidate comes with criterion-by-criterion rationale linked to the chart. Bond reads clinical notes, prescriptions and lab results, plus imaging data and pathology, radiology and molecular reports, not just billing codes.⁠[13],[14]
  • Every major EHR and CTMS, with EHR integration in 48 hours. Bond connects to Epic, Oracle Health (Cerner), MEDITECH, athenahealth, eClinicalWorks, NextGen, Veradigm, Greenway Prime Suite, OncoEMR and the other major EHRs, and works with the CTMS systems sites use, including CRIO, Advarra OnCore, Advarra Clinical Conductor, RealTime and Veeva SiteVault. Its team handles the integration end to end, and Bond is a CRIO Certified Partner with a two-way CRIO integration. See integrations.⁠[13],[14]
  • You pay only when a patient is randomized. There is no fee per screened patient, no integration fee and no charge for creating and running your ad campaigns: Bond's only fee is a percentage of the milestone payments for each randomized patient, from randomization through later study visits. See pricing.⁠[14]

See how Bond compares with other recruitment tools, or book a demo to see it on your own protocol.

Frequently asked questions

Can a site use a patient's existing NGS report to pre-screen for a trial?
Generally yes, under the site's HIPAA and IRB rules for reviewing records, because results from routine care may be used for eligibility without first obtaining consent. A research-only test needs consent first.⁠[12]
What share of sequenced patients should a site expect to enroll on a matched trial?
Published programs report low figures: 17.8% of sequenced NCI-MATCH patients were assigned to an arm, and 11% of the first 5,009 patients sequenced at Memorial Sloan Kettering enrolled on a genomically matched trial.⁠[1],[2]
Do protocols accept local NGS results?
Many do, but check each one. NCI-MATCH moved to accepting outside lab results for rare targets and confirmed them with its own assays.⁠[1]

Sources

  1. 1.
    Molecular Landscape and Actionable Alterations in a Genomically Guided Cancer Clinical Trial: National Cancer Institute Molecular Analysis for Therapy Choice (NCI-MATCH) · Journal of Clinical Oncology (Flaherty KT et al.), via PubMed Central, 2020
    Read October 2026 via NCBI full text. Quotes: "NCI-MATCH opened on August 12, 2015; 1,117 sites registered 6,391 patients until… moreRead October 2026 via NCBI full text. Quotes: "NCI-MATCH opened on August 12, 2015; 1,117 sites registered 6,391 patients until registration for centralized molecular screening closed on May 22, 2017"; "Of the 5,954 samples submitted, 5,540 (93.0%) were sequenced successfully."; "Molecular alterations for assignment to an NCI-MATCH subprotocol were present in 37.6% of patients"; "When molecular, prior treatment, and specific cancer exclusions were accounted for, the match rate was 26.4%. Lack of subprotocol availability ... led to a treatment assignment rate of 17.8% (n = 985 of 5,540"; "Seventy percent of assigned patients received treatment on a subprotocol."; "No subprotocol whose targeted alteration had a prevalence of < 1.5% reached the accrual goal in this phase of the trial (18 subprotocols)."; "we have now used external academic and commercial sequencing platforms that are being used in routine practice, to continue accrual in NCI-MATCH, and will confirm the outside NGS result with the NCI-MATCH assays."
  2. 2.
    Mutational landscape of metastatic cancer revealed from prospective clinical sequencing of 10,000 patients · Nature Medicine (Zehir A et al., Memorial Sloan Kettering), via PubMed Central, 2017
    Quote: "We observed that 37% of patients harbored a clinically relevant alteration, and 11% of the first 5,009 patients to receive… moreQuote: "We observed that 37% of patients harbored a clinically relevant alteration, and 11% of the first 5,009 patients to receive MSK-IMPACT testing were subsequently enrolled on a genomically matched clinical trial."
  3. 3.
    Trends and Disparities in the Use of Next-Generation Sequencing in Patients With Cancer in the United States · JAMA Network Open (Hage Chehade C et al.), via PubMed Central, 2026
    Retrospective cohort, 280 US cancer clinics, EHR-derived database, diagnoses 2018 to 2022; 27,050 patients with advanced NSCLC.… moreRetrospective cohort, 280 US cancer clinics, EHR-derived database, diagnoses 2018 to 2022; 27,050 patients with advanced NSCLC. Quotes: "A total of 16 599 (61.4%) received NGS after aNSCLC diagnosis."; "For patients diagnosed in 2018, cumulative incidence was 41.9% (95% CI, 40.7%-43.1%) at 1 year"; "For those diagnosed in 2022, cumulative incidence was 74.5% (95% CI, 72.8%-76.0%) at 1 year"; "The cumulative incidence of NGS was 26.2% (95% CI, 25.4%-27.0%) at 1 year after mBC diagnosis"; "Patients with low socioeconomic status, Black or Hispanic patients, and those covered by Medicaid or Medicare experienced significantly longer time to testing in some of these cancers". The PubMed abstract describes the cohort as patients "who underwent NGS", which conflicts with the full text; the full-text wording is used here.
  4. 4.
    Real-world integration of genomic data into the electronic health record: the PennChart Genomics Initiative · Genetics in Medicine (Lau-Min KS et al., Penn Medicine), via PubMed Central, 2021
    Quotes: "Currently, most genetic results are reported in unstructured PDF documents."; "it is preferable to store them in… moreQuotes: "Currently, most genetic results are reported in unstructured PDF documents."; "it is preferable to store them in discrete, computable format to enable electronic searching, clinical decision support (CDS), and secondary use". Mostly germline genetics.
  5. 5.
    Next-Generation Sequencing and the Clinical Oncology Workflow: Data Challenges, Proposed Solutions, and a Call to Action · JCO Precision Oncology (Conway JR, Warner JL, Rubinstein WS, Miller RS; Harvard, Vanderbilt and ASCO), 2019
    Abstract read via PubMed. Quotes: "the electronic health records of today seem ill suited for managing genomic data"; "along with… moreAbstract read via PubMed. Quotes: "the electronic health records of today seem ill suited for managing genomic data"; "along with a commitment on the part of sequencing laboratories to consistently provide structured genomic data for clinical use." Narrative review.
  6. 6.
    MatchMiner: an open-source platform for cancer precision medicine · npj Precision Oncology (Klein H et al., Dana-Farber Cancer Institute), via PubMed Central, 2022
    Built and evaluated by the same center; March 2016 to March 2021. Quotes: "MatchMiner facilitated 166 trial consents (MatchMiner… moreBuilt and evaluated by the same center; March 2016 to March 2021. Quotes: "MatchMiner facilitated 166 trial consents (MatchMiner consents, MMC) for 159 patients"; "We found MMC consented to trials 55 days (22%) earlier than non-MMC."; "MatchMiner facilitated 20% (166 out of 847) of all possible MMC."; "Genomic and clinical features from the external report are inputted manually and all available trials are searched."
  7. 7.
    Clinical Use of Precision Oncology Decision Support · JCO Precision Oncology (Johnson A et al., MD Anderson Cancer Center), 2017
    Abstract read via PubMed. Quotes: "Trial enrolment was significantly higher for patients with actionable/potentially actionable… moreAbstract read via PubMed. Quotes: "Trial enrolment was significantly higher for patients with actionable/potentially actionable alterations (92/333, 27.6%) than those with unknown (16/136, 11.8%) and non-actionable (2/66, 3%) alterations"; "Over half of alterations annotated were of unknown significance or non-actionable."
  8. 8.
    Tumor-Agnostic Precision Medicine from the AACR GENIE Database: Clinical Implications · Clinical Cancer Research (Gouda MA et al.), via PubMed Central, 2023
    Quote from the results: "4,912 samples harbored BRAF V600E mutations (2.9%), 410 harbored NTRK fusions (0.24% of all samples and… moreQuote from the results: "4,912 samples harbored BRAF V600E mutations (2.9%), 410 harbored NTRK fusions (0.24% of all samples and 1.6% of samples profiled for structural variants), and 396 harbored RET fusions (0.26% of all samples and 1.5% of samples profiled for structural variants). In addition, 25,527 samples (15.2%) were classified as having high TMB." The abstract swaps the NTRK and RET percentages; the results text is used here.
  9. 9.
    Developing Targeted Therapies in Low-Frequency Molecular Subsets of a Disease: Guidance for Industry · US Food and Drug Administration, 2018
    Final guidance, October 2018. Quotes from the PDF: "the FDA will accept grouping patients with different molecular alterations if… moreFinal guidance, October 2018. Quotes from the PDF: "the FDA will accept grouping patients with different molecular alterations if it is reasonable to expect that the grouped patients will have similar pharmacological responses based on a strong scientific rationale"; "The FDA recognizes that the clinical trial assay method may limit who is eligible for clinical trials."
  10. 10.
    Identifying Oncology Clinical Trial Candidates Using Artificial Intelligence Predictions of Treatment Change: A Pilot Implementation Study · JCO Precision Oncology (Kehl KL et al., Dana-Farber Cancer Institute), via PubMed Central, 2024
    Single-center pilot, no control group. Quotes: "60,199 patient-trial matches were generated for 2,150 patients on the basis of… moreSingle-center pilot, no control group. Quotes: "60,199 patient-trial matches were generated for 2,150 patients on the basis of genomics alone. Of these, 3,168 patient-trial matches (5%) corresponding to 525 patients were flagged for ONN review by our model, representing a 95% reduction"; "the trial had no slots (14%)"; "Of 74 patients whose oncologists were contacted, 10 (14%) had a consult regarding a trial and five (7%) enrolled."
  11. 11.
    Clinical Trial Notifications Triggered by Artificial Intelligence-Detected Cancer Progression: A Randomized Trial · JAMA Network Open (Mazor T et al., Dana-Farber Cancer Institute), via PubMed Central, 2025
    Quotes: "Of 20 707 patients randomized ..., 13 802 were randomized to the intervention arm and 6905 to the control arm. The… moreQuotes: "Of 20 707 patients randomized ..., 13 802 were randomized to the intervention arm and 6905 to the control arm. The intervention had no significant impact on the trial enrollment rate (intervention, 2.20% [95% CI, 1.97%-2.46%]; control, 2.03% [95% CI, 1.72%-2.39%] ... P = .41)"; "The intervention consisted of emailing lists of genomically matched trials to physicians treating patients who became trial ready"; notifications "for patients with tumor progression based on AI interpretation of imaging reports did not increase therapeutic trial enrollment."
  12. 12.
    Screening Tests Prior to Study Enrollment: Guidance for Institutional Review Boards and Clinical Investigators · US Food and Drug Administration, 1998
    Information sheet, January 1998, final. Quotes: "informed consent must be obtained prior to initiation of any clinical procedures… moreInformation sheet, January 1998, final. Quotes: "informed consent must be obtained prior to initiation of any clinical procedures that are performed solely for the purpose of determining eligibility for research"; "Procedures that are to be performed as part of the practice of medicine and which would be done whether or not study entry was contemplated, such as for diagnosis or treatment of a disease or medical condition, may be performed and the results subsequently used for determining study eligibility without first obtaining consent."
  13. 13.
  14. 14.
    Bond Health product information · Bond Health, 2026Capabilities, pricing and compliance status described by Bond Health, October 2026.

Related pages

Ready to get started?

Enroll patients faster.

See how Bond can accelerate enrollment for your clinical trials with a personalized demo.

Book a demo

Up to 3×

faster enrollment

90%+

matching accuracy

48 hrs

to EHR integration