SIDEBAR to July 28 blog
Yes. The interesting thing about the Valar Labs comment is that it is not attacking CMS’s new SaMS framework at all. It embraces the framework and argues that CMS applied it inconsistently to Valar’s two bladder tests.
What is Valar’s pricing problem?
Valar has three new H&E/AI oncology services:
| Valar test | 2027 CPT | CMS proposed APC | Proposed payment | Valar asks |
|---|---|---|---|---|
| Vesta Bladder Risk Stratify | 1063T / X568T | 1542 | $350.50 | APC 1509, $750.50 |
| Vesta Bladder BCGPredict | 1064T / X569T | 1542 | $350.50 | APC 1509, $750.50 |
| Vitara Pancreas ChemoPredict | 1097T / X614T | 1509 | $750.50 | Keep $750.50 |
So the practical request is quite simple: more than double payment for each of the two bladder tests, from $350.50 to $750.50. The pancreas test is already where Valar wants it.
Valar thinks CMS has somehow treated the two bladder codes as related/similar procedures and effectively subjected them to a roughly 50% reduction. The letter says this appears to reflect a misunderstanding because the two tests are separately ordered, produce separate reports, and answer different clinical questions: one is prognostic for recurrence/progression and surgical decisions; the other predicts response to BCG.
What rules does Valar say should apply?
Valar's argument is essentially four layers.
First, O1 separate payment. Valar strongly supports CMS's proposed new status indicator O1, with payment specifications like status indicator S: SaMS gets a separate APC payment, rather than being packaged. That is particularly important because these tests are performed after the hospital encounter and their results affect later oncology management.
Second, New Technology APCs are the right interim home. Valar accepts CMS's premise that ordinary clinical APCs don't fit algorithm-driven services very well. New Technology APCs are therefore a reasonable bridge while CMS collects claims/cost information and develops a permanent SaMS methodology. That tracks CMS's own stated transitional approach.
Third, comparability should drive the APC assignment. Valar doesn't have an established CLFS price for these brand-new Category III codes. So it says CMS should look to similar H&E AI oncology services, particularly 0220U, 0376U, 0414U, 0418U, 0512U and 0513U—all of which CMS proposes to put in APC 1509 at $750.50. It gives especially strong weight to new Category III codes 1106T/X623T and 1107T/X624T, whose descriptors are almost twins of Valar's bladder codes except for tumor site and clinical endpoint; CMS also put those at $750.50.
Fourth, don't apply a multiple-procedure-type discount to the two bladder tests. This is a particularly good argument under CMS's own proposed architecture. CMS defines O1 as “separate APC payment.” Valar's point is that 1063T and 1064T are independently useful tests rather than two components of one procedure. CMS itself proposes O1 as the separately payable SaMS indicator.
There is also an older packaging argument: Valar cites CMS's prior rationale for separately paying some cancer algorithm tests because they are relatively disconnected from the encounter where the specimen was obtained and inform subsequent treatment.
And yes—the tables are unusually useful
Pages 5–7 are probably the most reusable part of the comment. Valar essentially builds a mini-reference table for H&E computational pathology reimbursement.
For each service it supplies:
code/test/company → specimen → workflow/resources → algorithm methodology → proposed APC/SI/payment.
It covers six existing PLA H&E-AI services:
0220U PreciseDx Breast; 0376U ArteraAI Prostate; 0414U LungOI; 0418U PreciseDx Breast Biopsy; 0512U Tempus p-MSI; and 0513U Tempus p-Prostate.
Then it adds Valar's three tests and the two very similar new Category III breast/prostate codes 1106T and 1107T. That's an 11-row comparison table, and it is much richer than CMS's table because Valar has supplied the operational details—accessioning, QC, pathologist review, WSI digitization, image transfer/storage, GPU inference, report generation, etc.
For digital-pathology work, that table is quite valuable independent of Valar's lobbying position.
Does Valar call out CMS's strange Table 62?
Surprisingly, essentially no.
CMS's actual Table 62 contains exactly 10 codes, not eleven. CMS says it selected them according to a purportedly simple rule: if the descriptor contained no laboratory method and only algorithmic analysis, CMS classified it as a SaMS laboratory analysis.
But the table itself is rather obviously troublesome. Among the ten are:
0511U, whose descriptor actually says “tumor cell culture in 3d microenvironment”—rather hard to characterize as having no laboratory method.
0208U, which expressly describes mRNA gene-expression analysis of 108 genes from an FNA specimen.
81416, which is sequence analysis of a comparator exome—again, rather conspicuously a sequencing code rather than merely an algorithm running on previously generated data.
That seems internally inconsistent with CMS's own sentence immediately preceding the table: no laboratory methods in the descriptor; only algorithmic analysis. CMS also inexplicably puts 81416 in APC 1575, $10,001–$15,000, while the genuine H&E-AI codes cluster at APC 1509.
Valar does not take CMS to task for any of that. Nor does it point out the rule text's erroneous reference to 42 CFR 493.21 in defining the underlying CLIA concept. Instead, Valar very sensibly cherry-picks the part of Table 62 that helps its case—the H&E AI comparators—and leaves the rest alone.
Its only mild comment about Table 62 is that Valar assumes its own codes weren't in the published Table 62 because the final Category III CPT numbers weren't yet available when CMS prepared the proposed rule.
So I would characterize the comment this way:
Valar accepts CMS's SaMS theory, accepts O1, accepts New Technology APCs, and even uses CMS's Table 62 as precedent. Its dispute is narrowly about horizontal equity: if substantially similar H&E-AI oncology tests get APC 1509/$750.50, Valar's two bladder tests should too. It does not challenge the much more fundamental oddities in CMS's selection of the ten “lab SaMS” codes.
And in the process, Valar has inadvertently given everybody else a very nice little comparative inventory of the emerging H&E-AI CPT/PLA universe and its proposed Medicare pricing.
##
##
Valar's argument is logical and well-crafted. Yet, it may represent the very problem CMS alluded to in the original OPPS pages about CLFS and digital. CMS may worry we image a $10 H&E slide once, and every ten seconds run a new algorithm in memory - $750, $1500, $2250, $3000, etc.
##
Yes. Valar is a very well-written illustration of the exact economic problem CMS says it is trying to solve.
CMS says rather candidly that OPPS was built to pay for things whose costs are largely material resources, whereas SaMS may derive its value from proprietary algorithms and “scalable, non-material costs.” CMS then adds that the cost-based system may provide limited transparency and “may not effectively constrain pricing.” It separately flags subscription, license, and per-use or “per-click” fees as raising program-integrity concerns.
Now put Valar into that framework.
A patient already has an FFPE block. You make an H&E slide; perhaps the slide costs tens of dollars. You digitize it once. Thereafter, the same WSI data file can potentially support:
Risk Stratify → $750.50
BCGPredict → another $750.50
conceivably some future biomarker/prediction algorithm → another $750.50
and so forth.
The marginal physical resource cost of the second and third analyses may be dramatically lower than the first. I would not assert that they literally take only seconds without Valar-specific evidence—there may be QC, data handling, pathologist oversight, computing, report production, etc.—but economically the important point survives: the cost of running another trained algorithm against an already-created digital image is unlikely to resemble the resource structure of repeating a $750 physical medical procedure.
And Valar's particular request almost highlights the issue. It says its two bladder tests are genuinely different clinical services, which is perfectly plausible: one predicts recurrence/progression and one predicts BCG response. Therefore, Valar says, pay each one separately at $750.50.
That is a sound coding/clinical-distinctness argument, but it does not answer the CMS policymaker's payment-economics question:
If two different answers are generated from the same slide and substantially the same digital infrastructure, should Medicare pay 2 × $750 merely because there are two separately coded outputs?
Indeed, CMS seems to have anticipated almost exactly this problem. Its proposed O1 status indicator would initially work like status indicator S—separate payment without a multiple-procedure reduction. But CMS specifically asks whether SaMS should instead get a status indicator functioning like T, so that multiple-procedure discounting would apply, expressly mentioning program-integrity concerns.
That makes Valar's situation a nearly textbook example:
Valar's position:
“These are two separate clinically meaningful tests, therefore don't discount either one.”
CMS's problem:
“Fine—but if both are software analyses of the same previously created image, paying the full standalone price every time may wildly overpay the incremental resource cost and create an incentive to proliferate separately coded algorithms.”
There is an even deeper problem. A 50% multiple-procedure reduction is itself borrowed from a world of procedures and may be economically arbitrary for software. If the first algorithm costs $750 and the incremental computational/reporting cost of the second is, hypothetically, $30 or $100, then paying the second one $375 still doesn't solve the problem. Conversely, some software may entail significant licensing, human review, cloud infrastructure, validation and ongoing development costs that make a pure marginal-compute-cost price too low.
So CMS really needs something more sophisticated than either:
$750 + $750 + $750, or
$750 + $375 + $375.
The long-term answer could be some combination of a base payment plus incremental-analysis payment, episode bundling, multiple-SaMS packaging, capped payment per underlying dataset/image, or a valuation methodology that recognizes both fixed intellectual-property costs and very low marginal production costs.
And this is why I think the Valar comment is unusually instructive. Valar may be entirely right under the payment rules CMS has proposed, while simultaneously demonstrating why CMS believes those rules cannot be the permanent answer. The company's excellent tables actually make the policy problem more visible: once H&E becomes a reusable digital substrate, an unlimited number of separately coded algorithms can potentially sit on top of it.
That, much more than whether this bladder algorithm deserves $350 or $750, is the elephant in CMS's SaMS discussion.
