JAMA Network Open publishes a must-read paper on comprehensive genomic profiling (CGP) by Chow et al. The paper headlines "a difference in Medicare Advantage patients" - less profiling - but the difference is tiny (25% vs 26%!). But there are LARGE differences by geography, and LARGE differences by cancer type.
Find it here: https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2849451
How was CGP defined?
First, they use the term "BGP" - broad genomic profiling - to mean 10 genes or more (including by old-school stack coding of individual oncogenes, think EGFR+KRAS+etc.)
They count codes like 81445, 81455, PLA codes for CGP, and also count 81479, especially critical in MolDx states. See Supplement, Table 2. TAT wasn't an issue. The time span was 2 months before surgery and 6 months after.
The method picks up CGP as 81479 under Moldx; but (I believe) misses CGP performed as part of an inpatient DRG payment and not coded separately.
How was metastatic defined?
I'm not sure. They write, "We restricted the sample to patients with metastatic cancer based on the presence of a secondary malignant neoplasm ICD-10 code (C77, C78, or C79) in the 2 months before through 6 months after their initial cancer diagnosis. Patients with a solid tumor cancer diagnosis code in the year prior to their index date were excluded, as we aimed to focus on patients with incident disease."
It's ridiculous we have a zillion ICD-10 codes and codes for "bitten by a duck on the left hand" but not clear coding for "metastatic cancer." Insane.
What was the Medicare Advantage difference?
Statistically valid, but very tiny. 254,720 patients. About half Medicare Advantage, median age 74, "metastatic cancer." 25.3% had CGP/BGP, of which 25.8% FFS, 24.6% MA insurance.
What was the impact of geography?
Big, but very patchy. Here is a heat map. Range of CGP/BGP about 15%-35%.
Large; about 40% in metastatic lung cancer, about 10% in prostate and kidney cancer, and in between those find 15% in breast cancer, 25-30% in colon cancer.
###
AI Corner
A Must-Read Article on CGP in Medicare Metastatic Cancer
JAMA Network Open has published a must-read paper by Chow et al. on the use of broad genomic profiling — what many of us would call comprehensive genomic profiling, or CGP — in Medicare beneficiaries with metastatic cancer.
The article headline, in effect, is about Medicare Advantage versus traditional fee-for-service Medicare. That is a real finding. It is statistically significant. But it is also worth saying plainly: the absolute difference is tiny. In this cohort, 25.8% of fee-for-service beneficiaries received broad genomic profiling, compared with 24.6% of Medicare Advantage beneficiaries. That is a 1.2 percentage point difference. With more than 250,000 patients, even tiny differences become statistically real. But in policy terms, the much larger story is elsewhere.
The Biggest Finding: CGP Remains Underused
The big story is that most Medicare beneficiaries with metastatic cancer did not receive broad genomic profiling at all. Overall, only about one-quarter of patients received BGP during the study window. Even among cancer types where broad genomic profiling was explicitly recommended, fewer than half of patients received it. For the precision oncology field, that should be the flashing red light.
After years of guideline development, FDA approvals, NCD 90.2, MolDx policies, commercial test growth, and endless conference panels on precision oncology, only about one in four older Medicare patients with newly diagnosed metastatic cancer received broad genomic profiling in this study.
The Medicare Advantage Difference: Real, But Small
The Medicare Advantage finding should not be dismissed, but it should be kept in proportion. The authors found FFS patients were more likely to receive BGP, with an adjusted odds ratio of 1.08 overall. The gap widened over time, and it was larger in cancers with equivocal guideline recommendations than in cancers with explicit recommendations.
That pattern is interesting. It may suggest that Medicare Advantage plans are more restrictive at the margins, especially where clinical consensus is less forceful. But the MA-versus-FFS contrast is not the dominant effect in the paper. It is a modest signal sitting inside a much larger national underuse problem.
For policy readers, the take-home message is not “Medicare Advantage blocks CGP.” The better summary is: CGP remains underused in metastatic cancer; use varies enormously by geography and tumor type; and Medicare Advantage may add a small additional access headwind, especially in less clearly guideline-mandated settings.
Geography: The Much Larger Variation
The second big story is geography. Chow et al. found striking variation across hospital referral regions, with adjusted BGP use ranging from about 14% to 36%. That is not a subtle payer effect. That is a map of uneven adoption.
In one region, a patient with metastatic cancer may have roughly a one-in-seven chance of receiving broad profiling; in another, more than one-in-three. The authors even note striking differences between nearby regions, such as San Angelo and Odessa, Texas. That kind of variation suggests local practice culture, access to oncology networks, molecular tumor boards, ordering workflows, lab relationships, and regional reimbursement habits may matter as much as — or more than — the broad category of MA versus FFS.
Cancer Type: Lung Leads, Others Lag
The third major story is cancer type. Broad genomic profiling was used far more often in lung cancer than in some other metastatic cancers. Lung cancer was around the 40% range, while prostate and kidney cancers were much lower, roughly around 10%. Breast cancer was also surprisingly low, around the mid-teens, while colorectal cancer was closer to the 25% to 30% range.
These differences are not random. They reflect the historical depth of biomarker-driven therapy in lung cancer, varying guideline strength, oncologist expectations, and perhaps the availability of obvious targeted-treatment pathways.
How Chow et al. Defined Broad Genomic Profiling
The paper uses the term BGP, for broad genomic profiling, rather than CGP. Their working definition is sequencing of 10 or more genes on a single day. That is an important operational definition. It means the authors are not just counting 500-gene panels. They are also trying to capture broader molecular testing through older billing patterns, including “stacked” individual gene codes.
The supplement is especially useful here. The authors count familiar genomic sequencing codes such as 81455 and 81456, intermediate panel code 81445, many PLA codes for named commercial tests, and unlisted molecular pathology code 81479. The inclusion of 81479 is especially important in MolDx states, where major CGP tests may appear under an unlisted code rather than a neat, named, national CPT code.
Claims Data: Powerful, But Imperfect
Claims-based CGP studies always have blind spots. If a genomic test is performed during an inpatient stay and absorbed into the DRG rather than separately billed, it may not be visible in the same way. Conversely, an algorithm that counts 81479 or stacked codes has to make assumptions about what the code represented.
Chow et al. address this thoughtfully, but the limitation is inherent in claims research. Claims are not laboratory information systems. They are billing exhaust.
How “Metastatic Cancer” Was Defined
The metastatic cancer definition is also worth pausing over. The authors define metastatic disease using secondary malignant neoplasm ICD-10 codes — C77, C78, or C79 — during the period from 2 months before through 6 months after the initial cancer diagnosis. That is a reasonable claims-based approach, but it also highlights an absurdity of our coding system.
We have ICD-10 codes for astonishingly granular events, including famously silly external-cause examples, but in real-world oncology data, “metastatic cancer” still has to be reconstructed clumsily and indirectly.
For a health system that wants to measure quality in precision oncology, that is a serious infrastructure problem.
Policy Takeaway: This Is an Implementation Gap
This article is valuable because it gives precision oncology stakeholders a national claims-based picture of where we are. The picture is not reassuring.
Better coverage policy is part of the answer, but probably not the whole answer. The geographic findings point to implementation gaps: ordering systems, oncologist awareness, tissue pathways, reflex testing, payer prior authorization, lab contracting, and institutional habits. The cancer-type differences point to guideline clarity and therapeutic actionability.
The claims-code issues point to the continuing difficulty of measuring modern molecular diagnostics in a payment system still built around older coding concepts.
Bottom Line
Chow et al. provide much more than a Medicare Advantage paper. They provide a snapshot of precision oncology’s incomplete diffusion into real-world Medicare cancer care. The headline may be MA versus FFS. The real story is that access to broad genomic profiling remains inconsistent, underdeveloped, and highly dependent on where the patient lives and what cancer label appears on the claim.
