Simulating the multidisciplinary case review workflow: NGS results, evidence-level annotation, and consensus precision-oncology recommendations
A molecular tumor board (MTB) begins where standard-of-care decision trees end. Before any genomic report is opened, the case coordinator (typically an oncology nurse navigator or research coordinator) assembles a structured clinical summary: diagnosis and staging, pathology, prior therapy lines with response and toxicity, performance status, comorbidities, and the specific clinical question driving referral. This framing determines which genomic findings will matter.
Precision oncology now generates more actionable data per patient than any single oncologist can reliably interpret alone. A single comprehensive genomic profiling (CGP) panel can return 50–500 gene alterations; separating true drivers from passenger mutations, germline artifacts, and sequencing noise requires combined expertise in oncology, molecular pathology, clinical genetics, and biostatistics.
MTBs formalize this interpretation into a structured, reproducible process. Unlike an informal curbside consult, an MTB convenes a standing multidisciplinary panel on a fixed schedule, documents its reasoning, and issues a written recommendation that becomes part of the medical record. The case presentation format itself is standardized (see Stage 5) so that recommendations are consistent, auditable, and defensible for insurance appeals and tumor registries.
MTBs are now mandated or strongly recommended by several accreditation bodies, including the Commission on Cancer (CoC), for centers seeking comprehensive cancer program status.
A well-formed case presentation opens with a structured clinical dossier, typically 1–2 slides or a single-page synopsis:
• Demographics and performance status: age, ECOG/Karnofsky score — directly bears on trial eligibility and treatment intensity tolerance • Diagnosis and staging: histology, grade, AJCC stage at diagnosis and at referral, sites of metastatic disease • Pathology: original biopsy findings, IHC markers (e.g., PD-L1 TPS/CPS, HER2, ER/PR), any prior molecular testing • Treatment timeline: every line of therapy in sequence — surgery, chemotherapy regimens with cycle counts, radiation fields and doses, prior targeted therapy or immunotherapy — annotated with best response (RECIST) and reason for discontinuation (progression vs. toxicity vs. patient choice) • Current clinical status: symptoms, labs, imaging findings driving the referral • The specific question for the board: "Is there an actionable target?" "Does this patient qualify for trial X?" "Should we test germline BRCA given this somatic finding?"
Treatment-line sequencing matters enormously for interpretation: a KRAS G12C mutation detected after three lines of chemotherapy is interpreted differently than the same mutation found at initial diagnosis, because resistance mechanisms and clonal evolution reshape the mutational landscape under therapeutic pressure.
Studies of case presentation quality consistently show that incomplete treatment histories are the single largest driver of MTB recommendation delay — panels frequently must table a case and request additional records before a genomic finding can be safely acted upon.
Comprehensive genomic profiling panels (FoundationOne CDx, MSK-IMPACT, Tempus xT, Guardant360) sequence hundreds of cancer-associated genes at deep coverage (500–1000×) to detect single-nucleotide variants, small indels, copy-number alterations, gene fusions, and genomic signatures like tumor mutational burden (TMB) and microsatellite instability (MSI). The molecular pathologist's presentation of this report is the evidentiary core of the MTB session.
1. Sample QC: tumor content must exceed 20% (macrodissection performed if needed); DNA/RNA extracted from FFPE tissue or fresh-frozen material.
2. Library preparation and hybrid capture: targeted bait sets pull down exons of the panel gene set (hybridization capture is preferred over amplicon methods for larger panels because it tolerates degraded FFPE DNA and captures structural variants).
3. Sequencing: paired-end sequencing on Illumina NovaSeq platforms to 500–1000× unique depth — deep coverage is essential because tumor purity, subclonality, and formalin-induced artifacts all reduce effective variant allele frequency (VAF).
4. Variant calling: somatic callers (MuTect2, Strelka2, VarDict) compare tumor reads against a matched normal (blood) or an unmatched population database to filter germline polymorphisms. Structural variant callers detect fusions; copy-number algorithms infer amplifications/deletions from read-depth ratios.
5. Annotation and filtering: raw calls are annotated against reference databases (COSMIC, ClinVar, gnomAD) and filtered for sequencing artifacts, strand bias, and low VAF noise (typical reporting threshold: VAF ≥5%).
6. Signature calculation: TMB (mutations/Mb) and MSI status are computed panel-wide and reported alongside individual variants — both are themselves biomarkers for immunotherapy response.
The presenting pathologist typically walks the board through variants in a fixed order of biological weight:
• Hotspot SNVs/indels in known oncogenes (EGFR, KRAS, BRAF): high-confidence driver mutations at recurrently mutated codons • Gene fusions (ALK, ROS1, RET, NTRK): often missed by DNA-only panels unless RNA-based fusion calling is included — a key reason modern panels sequence RNA in parallel • Copy-number amplifications (ERBB2/HER2, MET, MYC): distinguished from polysomy by copy-number thresholds (typically ≥6 copies reportable) • Tumor suppressor loss-of-function (TP53, PTEN, CDKN2A): usually not independently druggable but shape prognosis and resistance biology • Genomic signatures: TMB-high (≥10 mut/Mb) qualifies for pembrolizumab regardless of histology (tissue-agnostic FDA approval, 2020); MSI-High similarly qualifies for checkpoint inhibition
Variant Allele Frequency (VAF) contextualizes clonality: a driver mutation at 45% VAF in a diploid, 80%-pure tumor is likely clonal (present in nearly every cancer cell — a strong therapeutic target); a variant at 3% VAF may represent a minor subclone, sequencing noise, or clonal hematopoiesis of indeterminate potential (CHIP) if detected in blood-based liquid biopsy.
CHIP is a critical confounder in liquid biopsy interpretation: age-related clonal expansion of blood stem cells can produce circulating TP53, DNMT3A, or ASXL1 mutations that did not originate from the tumor at all — misattributing these to the cancer is a well-documented MTB pitfall.
A raw variant call is clinically meaningless without curated interpretation. Three complementary knowledgebases anchor evidence-based annotation: OncoKB (MSK) assigns FDA/NCCN-aligned actionability levels, CIViC (Clinical Interpretation of Variants in Cancer, Washington University) is a fully open, community-curated evidence resource, and COSMIC (Catalogue Of Somatic Mutations In Cancer, Sanger Institute) provides the largest reference catalogue of somatic variant frequency across tumor types.
OncoKB (Chakravarty et al., JCO Precision Oncology 2017) is the most widely used clinical evidence framework in US MTBs and is FDA-recognized as a source of clinical evidence for companion diagnostics. Its therapeutic levels:
• Level 1: FDA-recognized biomarker predictive of response to an FDA-approved drug in that specific tumor type (e.g., EGFR L858R + osimertinib in NSCLC) • Level 2: Standard-of-care biomarker per professional guidelines (NCCN) predictive of response, potentially in a different tumor type than FDA label • Level 3A/3B: Compelling clinical evidence from trials supports the biomarker as predictive, but not yet standard of care (3A = same tumor type, 3B = different tumor type) • Level 4: Compelling biological evidence supports the biomarker as predictive, but clinical evidence is insufficient • Resistance levels (R1/R2): biomarker predicts resistance to an otherwise indicated therapy
This ladder directly maps to the "actionable variants" metric MTBs track: only Level 1–2 findings typically support an immediate, guideline-concordant prescription; Level 3–4 findings route toward clinical trial screening rather than standard prescribing.
CIViC (civicdb.org) differs from OncoKB in being fully open-access and crowd-curated by an international community of clinicians and researchers, with every evidence item traceable to a specific publication and assigned an independent evidence level (A–E) and evidence direction (supports/does not support). Its transparency and API-first design make it popular for building institutional variant-annotation pipelines and for cases involving rarer tumor types or emerging biomarkers not yet formally reviewed by OncoKB curators.
COSMIC (cancer.sanger.ac.uk) is not primarily a therapeutic-actionability database but a mutation frequency and functional catalogue: it answers "how often has this exact variant been seen across cancer genomes, and in which tumor types?" This is essential for distinguishing a recurrent, biologically validated hotspot from a private, potentially artifactual variant unique to one patient. COSMIC's Cancer Gene Census further classifies genes as tier 1 (strong causal role, extensively documented) or tier 2 (emerging/less-validated evidence).
MTBs typically cross-reference all three: OncoKB/CIViC for "what should we do about this variant," COSMIC for "how confident are we this variant is real and recurrent." Discordance between databases (e.g., CIViC citing emerging trial evidence OncoKB has not yet incorporated) is itself a common discussion point in the multidisciplinary session.
A variant absent from COSMIC entirely is not automatically an artifact — it may be a genuinely novel finding — but it raises the bar for orthogonal confirmation (repeat testing, IHC/FISH validation) before the board treats it as actionable.
The defining feature of an MTB, distinguishing it from a single physician reading a genomic report, is structured multidisciplinary deliberation. A functioning panel typically includes medical oncology, molecular pathology, clinical/cancer genetics, and a research or trials coordinator, often joined by radiation oncology, surgical oncology, pharmacy, bioinformatics, and — increasingly — patient advocates, depending on institutional model.
• Medical oncologist(s): bring the whole-patient view — treatment history, performance status, comorbidities, and practical prescribing constraints (drug access, toxicity profile, patient preference). Often the primary "owner" of the clinical question.
• Molecular pathologist: presents and technically validates the genomic report — sequencing quality metrics, VAF interpretation, distinguishing true positives from artifacts, and flagging results needing orthogonal confirmation (FISH, IHC, repeat biopsy).
• Clinical/cancer geneticist: evaluates whether a detected somatic variant has a plausible germline origin (e.g., BRCA1/2, Lynch syndrome genes) warranting germline testing and family counseling — a decision with implications extending beyond the patient to relatives.
• Research/trials coordinator: maintains real-time knowledge of open clinical trials, matches variants to trial eligibility criteria (including basket and umbrella trials such as NCI-MATCH and TAPUR), and manages the logistics of trial referral.
• Additional participants as needed: radiation oncology (local control options), surgical oncology (resectability of oligometastatic disease), clinical pharmacy (drug interactions, dosing), bioinformaticians (pipeline and variant-calling questions), and increasingly a patient advocate or palliative care representative to keep goals-of-care central to the discussion.
Most MTBs follow a semi-standardized discussion flow once the case and molecular findings are presented:
1. Clarifying questions: panel members confirm clinical details (prior therapy responses, current performance status, patient goals) before addressing the molecular findings.
2. Variant-by-variant review: typically starting with the highest evidence-level findings — is this variant clonal, is it a known resistance mechanism to a prior therapy, does it match an approved or trial therapy in this specific tumor type or only by extrapolation from another cancer.
3. Cross-specialty flags: pathology may flag an unusual VAF pattern; genetics may flag a variant warranting germline workup; the trials coordinator cross-checks live trial eligibility (age, prior lines, organ function, geographic feasibility).
4. Weighing competing options: when multiple actionable findings compete (e.g., a Level 2 off-label option vs. a Level 3 trial with a novel agent), the panel debates sequencing — which to try first, and what would trigger switching.
5. Practical constraints: drug access/insurance coverage, trial site distance, patient fitness for a demanding regimen — oncology and pharmacy typically raise these before a recommendation is finalized.
This structure is deliberately similar to a case conference or morbidity/mortality review: documented, time-boxed per case, and designed so that the reasoning — not just the conclusion — is captured for the medical record.
A 2020 survey of NCI-designated cancer centers found that virtually all reported multidisciplinary attendance as the single greatest determinant of MTB recommendation quality — panels missing genetics or trials representation were significantly more likely to issue incomplete recommendations requiring follow-up.
The tangible output of an MTB is a written, consensus-based recommendation returned to the treating oncologist within days of the session — not a binding order, but a documented, evidence-graded opinion the treating team weighs alongside clinical judgment and patient preference. A substantial and growing literature has quantified just how often this process changes what actually happens to the patient.
Documented MTB output typically follows a standardized template regardless of institution, designed for both the treating physician and for retrospective audit/registry purposes:
• Case summary: diagnosis, stage, treatment history (as presented in Stage 1) • Molecular findings table: gene, alteration, VAF/copy number, evidence level (OncoKB/CIViC), and source database citations • Panel discussion summary: key points raised, dissenting opinions if any, and the clinical reasoning connecting findings to the recommendation • Recommendation: specific — a named drug and line of therapy, a named clinical trial (with NCT number) and eligibility caveats, a germline referral, or "no actionable finding at this time, recommend re-review at progression" • Level of consensus: whether the recommendation was unanimous or reflects a majority view with noted alternative opinions • Follow-up plan: who is responsible for closing the loop with the patient and treating team, and the timeline for re-review if disease progresses
This structured format is what allows MTB recommendations to be tracked longitudinally — did the patient receive the recommended therapy, and if not, why — a feedback loop increasingly used to refine institutional MTB practice.
Multiple retrospective cohort studies across academic cancer centers converge on a consistent finding: MTB review changes the treatment plan in roughly 20–40% of discussed cases relative to what would have happened under standard single-oncologist interpretation of the same genomic report. Reported changes include: initiation of a targeted therapy not otherwise considered, referral to a specific clinical trial, addition of germline genetic counseling, or de-escalation (recognizing that an apparent target is not truly actionable, e.g., a subclonal or VAF-implausible finding).
Studies specifically measuring clinical trial enrollment after MTB discussion report roughly 10–15% of discussed patients subsequently enroll in a matched trial — a meaningful improvement over baseline trial enrollment rates in oncology, which hover around 3–5% of all cancer patients nationally.
Survival benefit data are harder to establish given selection bias (patients discussed at MTB are not randomly selected), but several single-center and multi-center studies (e.g., using the WINTHER and I-PREDICT trial frameworks) report improved progression-free survival in patients who received an MTB-matched therapy compared to those who received unmatched or standard-of-care therapy, supporting the biological plausibility that better-matched targeted treatment improves outcomes, not merely that MTB-reviewed patients differ systematically from non-reviewed patients.
The most consistent number across the outcomes literature — repeated in single-center series from MSK, MD Anderson, and multi-institutional consortia alike — is that molecular tumor board review changes the documented treatment plan in roughly one in three to one in five cases, making the MTB one of the highest-yield interventions in the precision oncology workflow.