Claude Opus worked from a transcript of the CLFS ALM day, the agenda of 47 speakers or organizations, and the Excel spreadsheet of the circa-110 codes under discussion.
# # # #
The frame of the day
What CMS convened on June 10 was not a pricing decision but a price-nomination hearing. Under §1834A(f)(3) of the Social Security Act, CMS has to take public comment before setting a payment basis for new and substantially revised CDLT codes, and this five-hour webinar — the CY2027 Annual Laboratory Meeting — is that statutory forum.
The opening remarks set the boundaries tightly: every presenter is arguing for one of exactly two outcomes for each code, crosswalk (assign the payment of an existing comparable code) or gap-fill (kick the code to the MACs, who derive a carrier-specific amount over roughly nine months, post it ~April 30, and accept reconsideration). No dollar figures are debated; no policy is entertained. The meeting feeds a separate machine — the CDLT FACA advisory panel on July 14–15, then preliminary determinations in early October, a 30-day comment window, finalization, and a second meeting/panel cycle on September 15–16, with reconsideration rights layered on top.
A few contextual flags from the introduction matter for any expert read. CMS reminded everyone that legislation CAA 2026 §6226 (enacted Feb 3, 2026) revised the PAMA data-reporting requirements and continued to cushion the private-payer-rate phase-in (no more than a 15% annual cut through 2029) — so the rate environment around all of this is still being held artificially flat. They also disclosed, somewhat unusually for an opening statement, that a handful of recently established CPT codes are not on the agenda despite incoming letters and questions, and "will not be discussed." That is a live procedural dispute parked in plain sight. [BQ: It involves digital pathology codes.]
The code list itself frames the day as much as the agenda does. Of ~110 codes, the overwhelming majority are proprietary (PLA 0xxxU or xxxxU) rather than Category I, and the substantive center of gravity is unmistakable: oncology ctDNA/MRD assays, rare-disease whole-genome and exome sequencing (often proband-plus-comparator trios), transplant rejection signatures, and — the single most conspicuous cluster — a wave of DNA-methylation "positive/negative risk" tests spanning psychiatry, neurology, hepatology, cardiology, and substance use.
How speakers typically presented
The format is a near-ritual. A presenter — almost always a VP of Reimbursement, Market Access, or a reimbursement consultant rather than the bench scientist — walks a slide deck through four beats: the clinical indication, the analytic workflow (extraction → library/assay → algorithm → report), the resources required, and then the recommendation with a comparator table. The argument is structured almost entirely as analytic equivalence to a chosen anchor code: same analyte type, same platform, same number of targets, same specimen, same reportable output, therefore crosswalk to code X. Abbott's TBI panel (X300U) crosswalking to 0570U is the canonical clean case — genuinely the same two-analyte immunoassay on a different platform.
Two presentation strategies recur and are worth flagging because they bear directly on valuation:
The first is anchor selection toward the high end — companies pick the most generous defensible comparator. The second, more aggressive, is additive ("stacked") crosswalking: rather than map to a single code, presenters build a price by summing codes. Variantyx recommended trios priced as a base genome code "plus two comparators"; ClearNote built X282U as 0410U + 86301; ACLA constructed its Alport recommendation as a level-8 plus a level-9 sequencing code, notably applying a 0.5 multiplier "to represent the efficiencies" of a single panel — almost the only downward adjustment volunteered all day.
The professional societies (ACLA, AMP, ASM) presented differently from the labs. They tended to recommend gap-fill where novelty is real (the reconsidered carbapenem-resistance code 87183 is the shared example), to coordinate openly ("we are in alignment with ASM/ACLA"), and to flag where AMA descriptor conventions strip methodology out of the code (e.g., GFAP chemiluminescence not appearing in the descriptor). They functioned as the day's only systematic counterweight to anchor-shopping.
The panel's questions: how often, and why
This is where an observer learns the most. The CDLT advisory panel — the same body that meets formally in July — is present and listening, but the questioning is carried by a tiny rotating cast: Joe Lennerz (MGH), a "Dr. Bridges" (CMS), Chris Chong, and Mark Rumpler. Across 47 presentations, only a minority drew any question at all — a rough quarter to a third — and the overwhelming default transition is "any questions from the panel? … okay, thank you," straight to the next speaker. Lennerz and Bridges were the most active and by far the most technical.
The striking thing is the content of the questions: almost without exception they probe the resource inputs that determine whether the proposed crosswalk is honest. Rumpler asked Agendia about the computational burden of 465 housekeeping genes and asked another presenter about batch size. Bridges, who had clearly pre-reviewed Biodesix's 0634U, asked for the validated limit of detection per variant, the limit of blank, crosstalk in the 11-plex design, and concordance against an FDA-approved companion diagnostic on shared samples — and the presenter could only promise to "follow up." Lennerz's exchange with TruDiagnostic is the cleanest illustration of the whole day's tension: he established that the "risk score" for twelve chronic-disease tests is actually a binary output, that sign-out is a pipeline QC check with lab-director review and no molecular pathologist, and that cancer indications (including HCC) are being run on whole blood, not tissue — all while the company seeks to crosswalk to 0020M, a 30,000-locus tumor-tissue methylation classifier.
So the questions, when they come, are doing one job: testing whether "resource equivalence" survives contact with the actual analytic and professional facts. They are not about clinical utility, evidence of benefit, or price level — those are out of scope by design.
What is unusual about the day
Several things should catch an expert's eye:
The methylation-risk-score flood is genuinely anomalous. A single company (TruDiagnostic) brought twelve codes asking to anchor all of them to one oncology tissue classifier on a pure "same array, wet-lab cost dominates" theory; Quantigen brought a parallel set (tobacco, alcohol, lung-cancer hazard ratios from methylation). This is a coordinated attempt to establish a high epigenetic-screening price tier by equating chronic-disease binary screens with cancer diagnostics.
The advocacy asymmetry is structural, not incidental. The applicant supplies the anchor; the panel is mostly silent; ~70% of codes pass with no interrogation at all. Combined with the 10-minute one-way format and no requirement to disclose actual costs, the operative default is "accept the proposed crosswalk unless someone on the panel happens to challenge it."
The presence of professional reimbursement intermediaries fronting multiple clients is notable — McDermott+ presented for Abbott, for the Point of Care Testing Association, and for the Coalition for 21st Century Medicine in the same day. That is normal in this world but worth naming: the "public" comment is substantially a managed-advocacy exercise.
Finally, the "missing codes" disclosure and the fact that one presenter (One Lambda's X253U) explicitly invoked a prior panel vote ("nine to one to crosswalk") as precedent both reveal how path-dependent and quasi-adjudicative this nominally informational meeting actually is.
How the day shapes valuation — rational, fair, or skewed
The honest answer is that the format pushes simultaneously toward both poles.
It promotes rational, data-based valuation in real ways. Everything is on the record and the code list is published in advance. Panel members pre-review specific codes (Bridges: "I reviewed this code") and ask exactly the questions that separate a fair crosswalk from a borrowed one — LOD, professional interpretation, batch economics, concordance. The societies inject an independent voice for gap-fill where novelty is genuine. And the architecture gives multiple deliberative bites: this meeting, then the FACA panel, then comment, then reconsideration. The resource-equivalence framing, at its best, is a disciplined cost-based logic that resists paying for marketing claims.
But the same architecture skews in predictable directions. Because applicants nominate their own anchors and the dominant move is crosswalk rather than gap-fill (crosswalk references outran gap-fill roughly four-to-one), the process systematically imports the price of yesterday's most generous comparable into today's test — and additive stacking lets a presenter manufacture a price larger than any existing single code. The "resource equivalence" standard, taken literally, deliberately excludes clinical value, evidentiary rigor, and indication from the pricing conversation, which is precisely how a whole-blood binary screen can argue for parity with a tissue tumor classifier. And because only a minority of codes are questioned, the quality control is a spotlight, not a floor: a code gets rigor if Lennerz or Bridges happens to pull the thread, and a pass if they don't.
The net effect is a process whose rationality is real but conditional — conditional on a handful of expert panelists having the bandwidth and pre-reading to contest self-interested anchors across 47 presentations in five hours, and conditional on the July panel and the comment cycles correcting what June lets through. For the codes that get interrogated, the day works as designed; for the long tail that does not, valuation is effectively set by the applicant's slide deck.''
###
### SIDEBAR ONE
###
Crosswalk vs gap-fill tally
Classifying each of the 47 agenda slots by what the presenter clearly advocated (noting that several multi-code presenters split their recommendations):
- Clearly favored crosswalk (every code presented was a crosswalk request): ~34. This is the bulk of the day — essentially all the single-product labs (Abbott, CNSide, Agendia, Biodesix, Mayo, Guardant, Veracyte, TruDiagnostic's full set of 12, Variantyx, Natera, Castle, BillionToOne, Quantigen, GeneDx, and so on).
- Clearly favored gap-fill (every code was a gap-fill request): 4. Virant (0604U, 0605U), Pairidex (0644U, 0645U), ID-Fish (0580U, 0615U), and IGeneX (0636U–0639U). Notably, three of these four are rare-disease / tick-borne immunoblot or personalized-MRD assays whose sponsors argued no comparable code exists.
- Mixed — recommended gap-fill for some codes and crosswalk for others: ~8–9. This is where almost all the gap-fill advocacy lives, and it is concentrated in the multi-code society and coalition presentations: AMP, ASM, ACLA, the Coalition for 21st Century Medicine (9 crosswalks + 1 gap-fill), Thermo Fisher (0648U crosswalk / X269U gap-fill), POCTA, Genomic Health (X278U gap-fill / X294U crosswalk), Labcorp Oncology PGDx (gap-fill for the 0646U baseline, crosswalk for the rest), and AdvaMedDx (predominantly crosswalk, with one ambiguously worded code).
The headline is the asymmetry: crosswalk advocacy outnumbered clean gap-fill advocacy by roughly 34 to 4, and even the mixed presentations leaned crosswalk. Gap-fill was pushed mainly by the professional societies (where it reflects a principled "no true comparator exists / let the MACs cost it" stance) and by a handful of genuinely novel assays — almost never by a single-product lab that had a usable anchor available.
# # #
# # #
SIDEBAR TWO
# # #
The four presentations that drew the most questions
Counting distinct questions put to each presenter, three incidents stand clearly above the rest, and the fourth slot is a tie among several two-question exchanges, of which Castle Biosciences is the most substantive. The four:
| Incident | Code(s) | # Q's | Questioner(s) | What was probed | Presenter's response |
|---|---|---|---|---|---|
| Thermo Fisher (Dumanois) | 0648U | ~5 | Bridges (CMS) alone | Content completeness: which guideline biomarkers are covered by tumor type, MSI/TMB reporting, additional CDx indications + on-label vs RUO claims, RNA-fusion sensitivity/DV200/failure rate, concordance vs larger panels | Deflected nearly everything to written follow-up |
| Biodesix (Pascal) | 0634U | 4 | Bridges (CMS) alone | Analytic validation: LOD per variant, limit of blank, crosstalk in the 11-plex, concordance vs an FDA ESR1 companion Dx, justification of rarer variants | Partial answers (3–4 copy LOD, 100% concordance vs Guardant360), exact data promised later |
| TruDiagnostic (Mallon) | 0616U–0627U | 4 | Rumpler + Lennerz | Crosswalk justification: do loci differ by indication or is it one array; what the "score" actually outputs; molecular-pathologist sign-out vs pipeline QC; confirm whole-blood (not tissue) even for the cancer indications | Answered directly — and the answers undercut the equivalence claim |
| Castle Biosciences (Goldberg) | 0635U | 2 | Rumpler + a second panelist | Crosswalk magnitude: 487 genes vs a 50–100-gene anchor — is a 1:1 crosswalk appropriate; does the larger panel consume more material/sequencing/reagents | Held firm, refused to concede, repeated "gene count reflects analytic content, not resource burden" |
(The fourth position is genuinely a tie: CNSide 0640U, Precision Medicine 0628U, Pairidex 0644/0645U and Theranostix X258U each also drew two questions. Castle is the most useful to profile because it crystallizes the same dispute TruDiagnostic raised.)
Differences among the four incidents
The cleanest axis of difference is what kind of doubt the questions expressed. Thermo and Biodesix were both interrogated by Bridges, who had pre-reviewed the codes, and her questions were about the test itself — Thermo's were content/coverage questions (is this panel under-covering guideline biomarkers, is it missing MSI/TMB), and she even volunteered that the crosswalk looked reasonable; Biodesix's were pure assay-performance questions (LOD, limit of blank, crosstalk, concordance). Those two incidents exposed data gaps — facts that simply weren't in the room and had to be sent later.
TruDiagnostic and Castle were different in kind: the questions weren't fishing for missing data but pressing a conceptual dispute about the crosswalk anchor. Both presenters were arguing that a large, algorithm-driven test should be priced at a smaller-content comparator because the wet-lab work is equivalent, and in both cases a panelist pushed on exactly that move — Lennerz establishing that TruDiagnostic's "score" is a bare binary with pipeline QC and no pathologist sign-out (while still seeking the price of a 30,000-locus tumor-tissue classifier), and Rumpler pushing Castle on whether 487 genes really equate to a 50–100-gene anchor. The presenters' postures diverged sharply: TruDiagnostic conceded the distinguishing facts; Castle refused to concede and repeated its framing.
Secondary differences: Thermo and Biodesix were single-code, single-questioner (Bridges) episodes; TruDiagnostic and Castle were multi-panelist, and TruDiagnostic uniquely carried twelve codes onto one anchor.
Commonalities
All four were crosswalk requests — not one gap-fill among them — so every one of these exchanges was, functionally, a stress-test of a crosswalk justification. All four were probed by the same three-person technical core (Bridges, Rumpler, Lennerz) and by no one else; there were no peer-lab or public questions in any of them. Every question stayed strictly on analytic and resource inputs — assay performance, sequencing depth, professional review, loci counts — and none touched clinical utility or the dollar level, consistent with Lennerz's explicit instruction elsewhere in the meeting that utility and cost paradigms "are not factors we can take into account." And in all four, the presenter could not fully close the question in the room; each ended with a promise to respond in writing — which maps onto the chair's closing note that several presenters were asked to submit follow-up answers by June 17.
