HEADER: CMS RELEASES PAMA CODES WITH "NO DATA"
CMS’s PAMA “no data” list included roughly 90 conventional CPT codes, seven M-codes, and about 300 PLA codes—fully 60% of all active PLA codes.
The PLA pattern is not simply “new codes lack data”: code age was a weak predictor. Instead, commercial traction differed sharply by test type. Infectious-disease and transplant assays were far more likely to have data, while whole-genome, red-cell genetics, and therapeutic-drug tests were less likely. Oncology was surprisingly average overall.
Read about CMS plans for a "no data" meeting Sept 15-16, 2026: Plans here.
###
This summer, CMS called for labs nationwide to submit pricing data for claims in 1H2025, which CMS can use to reset a new fee schedule for labs for CY 2027, '28, '29.
CMS has released a spreadsheet of over 400 codes for which NO pricing data was submitted.
Find it here:
https://www.cms.gov/files/document/pama-test-codes-no-pp-data.pdf
CMS will seek public comment and hold a public meeting about how the 400-odd codes should be priced for 2027.
CMS posted the data as a PDF File. I exported that into an XLS and stored it open access in the cloud. note that the Google Sheets has 2 tabs, one with all codes with no data, the second just PLA codes with no data. Google Sheets is easily downloaded back into XLS.
###
AI ASSESSES the PAMA Data
I asked Chat GPT to compare the full list of Spring 2025 PLA codes, with the new list of Spring 2025 PLA codes with no commercial pricing data. Result; Chat GPT is a nerd! Seriously, Chat GPT provides a number of observational and exploratory analyses, you'll have to decide which ones might matter to you.
###
CMS’s newly released PAMA files allow an unusually simple experiment: compare proprietary laboratory analysis (PLA) codes for which CMS received no applicable private-payer information for 1H2025 against the remaining PLA codes active on the Clinical Laboratory Fee Schedule in 2Q2025.
The result is more interesting than a simple story of “new codes have no data.”
First, the headline number is surprisingly large
After cleaning the two CMS spreadsheets, there were 495 unique active PLA U-codes in the 2Q2025 CLFS file. Of these:
304 codes — 61.4% — were on CMS’s “no data” list.
191 codes — 38.6% — were not on the list and therefore are treated here as having PAMA data.
The cleaning matters. The CMS no-data file contained eight non-PLA G, P, and Q codes, which were excluded. The CLFS file contained duplicate identical entries for 0240U and 0241U, which were counted only once. All 304 PLA codes on the no-data list successfully matched to their longer CLFS descriptors.
Thus, nearly two-thirds of active PLA codes apparently generated no reportable PAMA private-payer data for the period.
That fact alone is striking.
Code age helps — but much less than expected
PLA numbers are issued roughly sequentially, so the numeric code provides a useful, although imperfect, proxy for age.
One might expect a straightforward pattern: older PLA codes have had years to obtain payer coverage, establish billing pathways, and generate private-payer claims, while newly issued codes have not.
The data do not show such a clean progression.
| PLA code-number cohort | Active codes | No-data codes | Percent no data |
|---|---|---|---|
| 0001–0150 | 114 | 57 | 50.0% |
| 0151–0300 | 145 | 114 | 78.6% |
| 0301–0450 | 136 | 73 | 53.7% |
| 0451–0551 | 100 | 60 | 60.0% |
The very oldest cohort does perform better: only half of codes 0001U–0150U lacked data. But the relationship is emphatically not monotonic. The 0151U–0300U generation performed dramatically worse than both older and newer cohorts.
At finer resolution, an extraordinary 43 of 47 codes from 0151U through 0200U — 91.5% — had no data. Some of this reflects a large block of highly specialized red-cell and blood-group genotyping codes, but those codes do not explain the entire effect.
The median code number was actually 270.5 among no-data codes versus 320 among codes with data. In other words, simple code age is a surprisingly poor predictor.
Test type tells a much more interesting story
Using the informative CLFS long descriptors, several recognizable test families can be compared. These classifications are descriptive and sometimes overlap, but the contrasts are large.
| Descriptor feature | Codes | No data | Percent no data |
|---|---|---|---|
| All PLA codes | 495 | 304 | 61% |
| Infectious disease | 51 | 18 | 35% |
| Transplant-related | 13 | 2 | 15% |
| AI/image-analysis related | 16 | 7 | 44% |
| Cell-free DNA/ctDNA | 26 | 13 | 50% |
| NGS-related | 50 | 27 | 54% |
| Pharmacogenomics/drug metabolism | 29 | 17 | 59% |
| Oncology overall | 147 | 92 | 63% |
| Whole-genome/exome related | 20 | 17 | 85% |
| Red-cell/blood-group genetics | 34 | 31 | 91% |
| Drug testing/therapeutic monitoring | 16 | 15 | 94% |
Several findings stand out.
Infectious-disease PLA codes were much more likely to produce PAMA data. Only 35% lacked data, compared with 61% of PLA codes overall. This is consistent with infectious-disease tests entering relatively conventional laboratory workflows with large numbers of commercial patients.
Transplant testing was even more striking. Only 2 of 13 transplant-related PLA codes lacked data. Although the sample is small, this is almost the mirror image of the overall PLA universe.
At the opposite extreme, 31 of 34 red-cell/blood-group genetic codes had no data, as did 15 of 16 codes related to prescription-drug testing or therapeutic drug monitoring.
Whole-genome/exome testing also stood out: 17 of 20 codes had no PAMA data.
PGx itself was not unusually disadvantaged
Pharmacogenomics provides an instructive counterexample.
A broad group of 29 drug-metabolism/PGx codes produced a 59% no-data rate — essentially the same as the 61% baseline for PLA codes overall.
Thus, a simple conclusion that “commercial payers do not pay PGx” is not supported by this comparison.
There were important differences inside PGx. Several highly specific older CYP2D6 component codes lacked data, while a number of later multigene pharmacogenomic panels did generate PAMA data.
The apparent market behavior seems to depend more on the particular test and billing model than on the label “pharmacogenomics.”
Oncology was remarkably average — until individual cancers were examined
Oncology is the largest identifiable group, with 147 codes. Overall, 63% lacked PAMA data, almost identical to the 61% rate for the complete PLA population.
But oncology was anything but homogeneous.
Among the larger recognizable subgroups:
Prostate: 9 of 18 no data — 50%
Breast: 7 of 12 — 58%
Lung: 5 of 9 — 56%
Colorectal: 13 of 14 — 93%
Hematolymphoid: 4 of 5 — 80%
Bladder: 5 of 5 — 100%
The smaller groups need cautious interpretation, but colorectal is especially notable. The no-data colorectal codes include assays spanning urine metabolites, microRNA, protein algorithms, tissue AI, methylation, cfDNA, NGS, and conventional KRAS/NRAS testing.
This suggests that even a very mainstream cancer indication does not guarantee meaningful private-payer reporting for a particular proprietary test.
Nor did “cutting-edge technology” automatically predict no data
Another surprising finding is that several technologies commonly thought of as newer or more exotic were not especially enriched among the no-data codes.
Only 50% of the 26 codes whose descriptors referenced cell-free DNA or ctDNA lacked data. NGS-related codes were at 54%. A small group involving AI or image analysis was at 44%.
By contrast, whole-genome/exome codes were at 85%.
This distinction is important. The sequencing technology itself does not appear to determine whether a PLA code generates PAMA data. The clinical application and commercial pathway appear to matter much more.
Price was also a weak discriminator
The CLFS payment amount did not divide the two populations particularly well.
The median CLFS rate was about $451 for no-data codes versus $598 for codes with data. But the distributions overlapped enormously.
Even among PLA tests priced at $2,000 or more, 45 of 84 — 54% — still had no PAMA data.
Higher-priced tests were somewhat more likely to have data, but there was no obvious price threshold at which private-payer reporting suddenly appeared.
What does “no data” really measure?
The analysis suggests that PAMA no-data status should not be interpreted simply as “private payers do not cover this test.”
At the code level, CMS is observing whether applicable laboratories reported applicable private-payer information for that HCPCS code during the collection period. Absence of data can therefore reflect several different commercial circumstances: extremely low test volume, Medicare-heavy utilization, limited private-payer coverage, use by laboratories outside the applicable-laboratory reporting universe, or a test that has simply failed to gain routine billing traction.
Conversely, the appearance of PAMA data establishes that some applicable private-payer transactions occurred; it does not establish broad national coverage or strong utilization.
Still, the binary signal is remarkably informative.
The strongest pattern in these CMS files is not that new PLA codes fail while old PLA codes succeed. Nor is it that oncology succeeds while PGx fails, or that expensive sequencing tests succeed while inexpensive tests fail.
Instead, the PAMA data seem to reveal something closer to commercial embedding: whether a proprietary assay has found its way into ordinary private-payer laboratory transactions at applicable laboratories.
Some relatively mundane infectious-disease and transplant tests clearly have. Many rare-disease, blood-group, therapeutic-drug, and whole-genome tests clearly have not.
And in 1H2025, that latter category still represented more than 60% of the active PLA code universe.
COMPARE TO CMS 2024 UTILIZATION
We compared the "zero" utilization codes in PAMA data to actual CMS utilization of each code in 2024.
About 30 codes had CMS utilization (between 11 and 2700) in 2024. Most were regular Cat I, not PLA.
Another 80 codes had CMS utilization, but suppressed because 10 units or less.
The remaining about 300 codes had no CMS utilization, either.
 |
| click to enlarge |