Key Takeaways
- In the BRAID trial, abbreviated MRI and contrast-enhanced mammography each found roughly three times as many invasive cancers as automated breast ultrasound. GE HealthCare was one of three funders.
- USPSTF said the evidence was insufficient in 2024. The American College of Physicians went further in 2026 and now advises against supplemental ultrasound and MRI in average-risk women.
- CISNET has modeled supplemental screening in dense breasts twice since 2015, in 2024 and again in 2026. Both efforts modeled MRI, and even MRI came back conditional. Ultrasound still has no modern economic analysis.
- The USPSTF "I" statement has a direct financial consequence. No A or B grade means no federal requirement to cover supplemental screening without cost sharing.
- State coverage mandates do not reach self-funded employer plans, which is where 67 percent of covered workers sit. For most of the commercial market, this is a voluntary benefit design decision.
- ABUS has a real case built on access and safety. The launch made a different one.
GE HealthCare launched a new automated breast ultrasound platform this month. Fourteen months earlier, a randomized trial the company helped pay for ranked that modality last among supplemental options for dense breasts. The evidence is not missing. It showed up, and the commercial plan did not account for it.
The trial that should have set the strategy
On August 3, GE HealthCare launched Invenia ABUS Prime and ABUS StreamVue for supplemental screening in women with dense breasts. The announcement leads with speed. Reading time down 33 percent. Scan speed up 40 percent. Sensitivity up to 93 percent. It frames the market with two statistics, that 40 percent of women have dense breasts and that those women face four to six times higher risk. It says nothing about how anyone gets paid.
Hold onto both of those market statistics. I come back to them later, because they are not quite right, and the way they are wrong tells you something.
Fourteen months earlier, The Lancet published interim results from BRAID. It is the only randomized head-to-head comparison of supplemental screening options in dense breasts, run across ten UK screening sites with 9,361 women aged 50 to 70 who had dense breasts and a clean mammogram. Each woman got abbreviated MRI, automated breast ultrasound, contrast-enhanced mammography, or standard care.
| Supplemental modality | Cancers per 1,000 | Invasive per 1,000 |
|---|---|---|
| Contrast-enhanced mammography | 19.2 | 15.7 |
| Abbreviated MRI | 17.4 | 15.0 |
| Automated breast ultrasound | 4.2 | 4.2 |
MRI and contrast-enhanced mammography each found roughly three times as many invasive cancers as ABUS. The cancers they found were half the size. The gap between abbreviated MRI and ABUS was statistically significant.
The funders were Cancer Research UK, GE HealthCare, and Bayer Healthcare.
So a company helped pay for the trial where its modality finished last, then launched a new product in that modality leading with scan speed. The fair reading is that two teams made two decisions on two timelines and never compared notes. That is the pattern worth looking at, because it is common and it is expensive.
BRAID has real limits and they should be said plainly. These are interim results from one round of screening. The endpoint is cancer detection, not survival, and the authors are clear that the trial does not measure overdiagnosis. The ultrasound arm ran on older hardware between 2019 and 2024, so it does not test today's platform or AI-assisted reading. Those are fair points. They are also the things a company says after publication instead of the evidence it builds before.
The guidelines moved the same direction
Coverage policy does not run on trials. It runs on the guideline documents that medical policy committees actually cite, and those have been consistent.
In 2024 the US Preventive Services Task Force concluded that the evidence is insufficient to weigh the benefits and harms of supplemental screening with ultrasound or MRI in women with dense breasts and a negative mammogram. That is an "I" statement, and it applies no matter how dense the breast. USPSTF reached it the same way this piece is arguing. The supplemental screening studies reported detection, not health outcomes.
Then the American College of Physicians published Version 2 of its screening guidance, released online April 17, 2026 and carried in the June print issue. ACP advises against supplemental MRI or ultrasound in average-risk women and suggests clinicians consider supplemental tomosynthesis instead, weighing benefit, harm, radiation, availability, patient preference, and cost.
A major professional society moved from "we do not know" to "do not do this." On either date, it landed before the August launch.
The comparative literature lines up. A 2025 systematic review found that against mammography alone, MRI picked up 18.92 additional cancers per 1,000 screens. ABUS found 2.30, handheld ultrasound 2.57, and tomosynthesis 1.69. Contrast-enhanced mammography looked comparable to MRI on limited data. The same review found the economic modeling across the field to be all over the map. NCCN treats whole-breast ultrasound as the fallback when contrast-enhanced mammography or molecular breast imaging is not available.
The economics were never settled, and attention went elsewhere
Supplemental ultrasound in dense breasts has two cost-effectiveness studies, and they are not close.
| Sprague et al., 2015 | Grady et al., 2025 | |
|---|---|---|
| Journal | Annals of Internal Medicine | Cancer Causes & Control |
| Funding | National Cancer Institute | None external |
| Method | Three validated CISNET models | Single-clinic Monte Carlo model |
| Result | $325,000 per QALY | $7,071 per QALY |
| Harms | 354 false-positive biopsy recommendations per 1,000 women | Included in model |
| Verdict | Big cost, small benefit | Cost-effective |
A medical policy committee sees a 46-fold spread on the one question it exists to answer. It also sees two studies that do not carry equal weight. Sprague ran three independently validated models on federal funding. Grady ran a model at a single breast clinic that owns and operates the technology. That second study is transparent and honestly done. Committees still discount work produced by people who sell the thing being studied. That is process, not prejudice, and every market access team should plan around it.
The obvious pushback is that Sprague is from 2015 and predates today's machines. Here is the part that makes it worse rather than better. CISNET has refreshed its dense-breast modeling twice since then, and both times it modeled MRI.
Stout and colleagues published a three-model analysis in JAMA Internal Medicine in 2024 on biennial tomosynthesis in women aged 50 to 74, with and without supplemental MRI by density. Adding MRI for extremely dense breasts moved deaths averted from 7.4 to 7.6 per 1,000 and pushed false-positive biopsy recommendations from 151 to 180. Widening it to heterogeneously dense breasts as well got deaths averted to 8.0 and false-positive biopsy recommendations to 343.
Then in 2026, Tosteson and colleagues published a full health and economic analysis in Annals of Internal Medicine, NCI funded, from a federal payer perspective. It opens by noting that federally mandated density notification is what motivates the question in the first place, which is the same thread running through this piece. Adding MRI for extremely dense breasts averted 0.1 to 0.8 additional breast cancer deaths per 1,000 and produced 22 to 186 additional false-positive biopsy recommendations. The authors concluded that supplemental MRI could be cost-effective for women with extremely dense breasts and at least twice average risk, but only if MRI costs and false-positive biopsy rates come down.
Two federally funded modeling efforts, two years apart, ran the full benefit, harm, and cost analysis on supplemental screening in dense breasts. Both aimed at MRI. Even MRI, the modality the guidelines favor, came back conditional. Ultrasound kept the 2015 number and never got the analysis at all.
That is not neglect. Academic modeling follows guideline relevance and federal funding priorities, and ultrasound has neither. Which means the only party with a reason to fund ultrasound economics is the industry selling it, and the industry has not.
Radiologists are not supplying the operational case either
When coverage is not there, the fallback pitch is efficiency. Faster reads, more capacity, less burnout. That pitch now has data against it.
Elmi and colleagues surveyed 215 breast radiologists through the Society of Breast Imaging, published in Clinical Imaging on August 11, 2026.
| Expected benefit | Non-users who expect it | Users who report it |
|---|---|---|
| Fewer recalls | 59.3% | 34.7% |
| Fewer biopsies | 36.4% | 9.1% |
| Less burnout | 56.0% | 29.4% |
Cost was the top barrier to adoption at 53.3 percent. Users described AI as a second opinion rather than something they act on directly.
Now set that against a 33 percent reading time reduction in a launch deck. Both numbers can be true. A controlled measurement and a deployed result are different things, and only the second one shows up in an operating budget.
Who actually has to pay, and who does not
This is where the coverage story usually gets told wrong, and the mistake is expensive.
If you come from the device side, a short detour is worth it, because the structure below is not obvious and it decides who your customer actually is.
There are two ways an employer buys health coverage, and they sit under different rulebooks.
| Fully insured plan | Self-funded plan | |
|---|---|---|
| Who bears the risk | The carrier, paid a premium | The employer, paying claims from its own funds |
| Is it an insurance product | Yes, a regulated policy | No, the carrier only administers and rents the network |
| Who regulates it | The state | Federal law, under ERISA |
| Does a state coverage mandate reach it | Yes | No, ERISA preempts the mandate |
| Who owns the coverage decision | Set by the policy and state law | The employer, free to carve services in or out |
The catch is that a member cannot tell which one they hold. The card carries a national carrier's logo either way. Same logo, same network, same claims system, completely different rulebook. And this is settled ground, not a loophole. State coverage mandates are insurance regulation, and ERISA keeps them off self-funded plans by design.
Lines of business do not share decisions. A national carrier runs commercial, Medicare Advantage, and Medicaid as separate businesses with separate policy, separate economics, and separate approvers. A win in one does not travel to the others. Its self-funded book is a fourth thing entirely, where the carrier administers the plan but the employer owns the coverage decision and can carve services in or out at will.
Screening age matters here. Breast screening runs roughly 40 to 74. Most of that volume sits in employer coverage before 65 and moves to Medicare after. A reimbursement strategy built around Medicare is aimed at the smaller half of the population.
With that in place, here is what actually applies.
Notification is federal and it obligates nobody. The FDA's updated Mammography Quality Standards Act rule took effect September 10, 2024. Every facility in the country now has to tell women their breast density using standard language. Telling a woman she has dense breasts creates demand. It does not create a payment obligation for anyone.
The ACA hook is missing, and USPSTF is why. Under the ACA, plans have to cover preventive services that USPSTF grades A or B with no cost sharing to the patient. Screening mammography has a B and gets that treatment. Supplemental ultrasound and MRI got an "I," so they do not. There is no federal requirement to cover them at all, let alone at zero cost share. The guideline grade and the payment rule are the same decision.
State mandates are real but they reach less than half the market. Thirty-nine states and DC require some coverage of supplemental breast imaging, though they vary on which modalities count and whether the patient still owes a copay. Those laws are insurance regulation, which means ERISA keeps them off self-funded employer plans. Sixty-seven percent of covered workers are in self-funded plans, and at employers with 200 or more workers it is 80 percent. Employer-sponsored insurance covers 154 million people under 65, which is the same working-age population that gets screened.
Put it together and the picture is not the one in most commercial decks.
| Segment | What requires coverage |
|---|---|
| Fully insured commercial, mandate state | State law, with modality and cost-sharing limits |
| Fully insured commercial, non-mandate state | Nothing |
| Self-funded employer plan, any state | Nothing. ERISA preempts the state mandate and the USPSTF "I" removes the ACA hook |
| Medicare | No national coverage determination for supplemental screening |
For most of the commercial market, covering supplemental breast screening is a voluntary plan design choice. A benefits leader makes it, not a medical director, and they make it against a USPSTF "I" statement and now an ACP recommendation against.
Here is why that should change how a commercial team works. Winning a carrier's medical policy committee feels like winning the account. It is not. It wins the fully insured book, which in a large-employer market is one customer in five. The other four are self-funded employers who never saw that decision, are not bound by it, and are making their own call with a benefits consultant. Nobody from a medtech commercial team is in that conversation, which means the largest single block of the addressable market is being left to decide on its own with no information from the people who built the product.
There is one more wrinkle worth knowing. Employers on high-deductible plans paired with HSAs have to be careful about waiving cost sharing here. If dense breast tissue gets read as a condition rather than a preventive finding, zero cost share could put the plan's tax-advantaged status at risk. Benefits counsel raises this, and it slows adoption further.
The AI itself has no way to get paid
Automated breast ultrasound bills under CPT 76641 and 76642. Those are established Category I codes. The AI inside the system rides along in that same technical fee and earns nothing extra. The same coding wall shows up in AI-guided POCUS, where a validated workflow still has no clean path to a code.
Eric Rubin, ACR's CPT advisor to the AMA, explained why. CPT codes describe distinct procedures. A radiologist already gets paid for spotting the findings these tools spot, so the algorithm is not a new service. Improving something that already has a code does not earn a code.
Category III does not rescue this. Codes 0689T and 0690T cover quantitative ultrasound tissue characterization across body sites and are not breast screening codes. X579T, released July 1, 2026 and effective January 1, 2027, belongs to QT Imaging for 3D ultrasound tomography. A Category III code lets you report a service. It does not make anyone pay for it, and commercial plans usually call these services investigational.
One structural tension sits underneath all of this. The 510(k) pathway rewards you for arguing your device is basically the same as something already on the market. A Category I CPT code requires you to argue it is meaningfully different. Companies optimize for whichever gate comes first, and clearance always comes first.
Back to those two market statistics
Radiologists grade breast density on a four-point scale. Almost entirely fatty. Scattered areas of fibroglandular density. Heterogeneously dense. Extremely dense. The word "dense" covers the top two categories, and the bottom category and the top category are each roughly 10 percent of women.
Now take GE's two framing numbers.
"40 percent of women have dense breasts." It is closer to half of women who get screened, once you combine heterogeneously dense and extremely dense. This one understates.
"Four to six times higher risk." That figure is real, and it compares extremely dense breasts against almost entirely fatty breasts. It is the two ends of the scale, each about a tenth of the population. Most women who get told they have dense breasts are in the heterogeneously dense category, and NCCN puts their odds ratio at 1.3 against scattered density. Extremely dense comes in at 1.8. Pooled estimates comparing all dense against all non-dense run about 1.8 to 2.1.
So a woman reading her density notification and hearing "four to six times" is being handed a number that describes a comparison she is probably not in.
Neither correction weakens the clinical case for screening the right women. Both are the kind of thing a clinical reviewer circles when a coverage request crosses their desk, and getting them right costs nothing. When your market-sizing statistic overstates by a factor of three and your prevalence statistic understates, someone on the other side of the table will eventually notice, and they will discount everything else in the packet.
Why this keeps happening
I have argued across four parts of the Capital Sales series that medtech's real constraint is the business model, not the technology. This is that argument playing out live.
Part two called the evidence problem exactly. Clinical data that is inadequate, delayed, or institution-specific. The best economic result in this category comes from a single clinic that operates the device. The randomized evidence arrived after the roadmap was locked, funded in part by the company it landed on.
The mechanism is simple. Capital equipment revenue books at install. Whether recalls drop, whether capacity improves, whether cost per cancer found goes down, all of that shows up on the hospital's income statement. None of it shows up on the manufacturer's. A company that captures nothing from the outcome has no reason to spend money proving the outcome, and no reason to let that evidence drive the roadmap.
Regulatory evidence gets built on schedule because FDA requires it. Pathway and economic evidence is optional, so it arrives late, and then it arrives as news.
Underfunding health economics is not carelessness. It is what the model rewards.
The argument that was sitting right there
ABUS has a genuine case. It is just not a detection case.
BRAID recorded zero adverse events in the ultrasound arm. The contrast-enhanced mammography arm produced 24 iodinated contrast reactions, one severe, plus three extravasations. ABUS needs no contrast, no IV, no gadolinium, and none of the infrastructure an MRI suite demands. It runs in community and rural sites where abbreviated MRI and CEM are not options at all. It costs far less per exam.
That is an access and safety argument. It is a solution claim rather than a device claim, and it happens to be true.
It also cannot be made with the evidence that exists, because it needs a different question. Does supplemental ultrasound beat no supplemental screening, in a place with no MRI, at a price the plan will absorb? Nobody has run that study. Different comparator, different endpoints, different economic model.
The stronger argument was available and the launch made the weaker one. That is what a product mindset does. It optimizes the thing the company sells instead of the path the patient travels.
What a market access plan here has to include
The economic model tells you which endpoints your clinical study has to be powered for, and that call comes years before anyone builds a launch plan.
That means standard of care in the covered population. In this category it now means abbreviated MRI and contrast-enhanced mammography.
Recall rate, biopsy rate, interval cancers, time to diagnosis. Those move cost-effectiveness results. Detection rate alone does not.
Work from a site that owns the equipment gets discounted no matter how good it is.
Cost-effectiveness tells a plan whether something is worth buying. Budget impact tells them whether they can afford it this year, and that second question decides more coverage policies than the first.
Two-thirds of the commercial market is self-funded, which means the decision maker is a benefits leader and their consultant. Almost nobody in imaging AI is in that room.
Per-study pricing, utilization-linked terms, performance guarantees. Any of them makes an economic claim credible in a way a white paper never will.
CMS has proposed Software as a Medical Service, a payment category for algorithmic software that produces its own diagnostic output, with a new status indicator that pays separately. Comments close August 31 under CMS-1850-P. The near-term dollars for commercial breast screening are small. The evidence expectations being written right now are not.
Erik's final hot take
The gray space in this market is not missing evidence. USPSTF ruled. ACP went further. A randomized trial answered the comparative question. CISNET refreshed its models and aimed them at MRI. The evidence showed up. It was just answering a different question than the one the commercial team was asking.
The white space is that supplemental ultrasound still has an unmade case. Access and safety, in the places where the better modalities do not exist, for the two-thirds of the market where nobody is required to cover anything. That is a fundable study and a reachable buyer, and neither is being worked.
Whoever runs that study will set how this category gets judged. Everyone else will keep explaining why installed base is not turning into utilization.
Frequently asked questions
What did the BRAID trial find about automated breast ultrasound?
BRAID randomized 9,361 women with dense breasts and negative mammograms across ten UK screening sites. Interim results from the first round showed cancer detection of 19.2 per 1,000 for contrast-enhanced mammography, 17.4 for abbreviated MRI, and 4.2 for automated breast ultrasound. MRI and contrast-enhanced mammography found roughly three times as many invasive cancers, and those cancers were half the size.
Does USPSTF recommend supplemental ultrasound for dense breasts?
No. The 2024 statement concluded the evidence is insufficient to weigh benefits and harms of supplemental ultrasound or MRI in women with dense breasts and a negative mammogram. That is an "I" statement, which means neither a recommendation for nor against.
What does the American College of Physicians say?
ACP's Version 2 guidance, released online in April 2026 and printed in the June 2026 issue, advises against supplemental MRI or ultrasound in asymptomatic average-risk women and suggests considering supplemental tomosynthesis instead.
Does the ACA require plans to cover supplemental breast screening?
No. The ACA requires no-cost-sharing coverage for preventive services USPSTF grades A or B. Screening mammography carries a B. Supplemental ultrasound and MRI carry an "I," so no federal coverage requirement applies to them.
Do state dense breast coverage laws apply to my employer plan?
Often not. Thirty-nine states and DC mandate some coverage of supplemental breast imaging, but those are insurance laws and ERISA preempts them for self-funded employer plans. Sixty-seven percent of covered workers are in self-funded plans, rising to 80 percent at employers with 200 or more workers.
Why can imaging AI rarely get its own CPT code?
CPT codes describe distinct procedures. When an algorithm improves detection of findings a radiologist is already paid to identify, the code set does not see a new service. ACR's CPT advisor to the AMA has cited this as the reason many cleared AI tools will never be separately paid.
Is supplemental ultrasound screening cost-effective?
It is contested. Sprague and colleagues, using three validated CISNET models on NCI funding, estimated $325,000 per QALY. Grady and colleagues, using a single-center model, estimated $7,071. No independent analysis has reconciled them. CISNET has since modeled supplemental screening in dense breasts twice, in 2024 and 2026, and both efforts modeled MRI rather than ultrasound.
How much does breast density really raise risk?
The four to six fold figure compares extremely dense breasts to entirely fatty breasts. NCCN cites odds ratios closer to 1.3 for heterogeneously dense and 1.8 for extremely dense against scattered density, with pooled estimates for dense against non-dense around 1.8 to 2.1.
Sources
Clinical and economic citations retrieved from PubMed.
- Gilbert FJ, Payne NR, Allajbeu I, et al. Comparison of supplemental breast cancer imaging techniques, interim results from the BRAID randomised controlled trial. Lancet. 2025;405(10493):1935-1944. doi.org/10.1016/S0140-6736(25)00582-3
- US Preventive Services Task Force, Nicholson WK, Silverstein M, et al. Screening for Breast Cancer: US Preventive Services Task Force Recommendation Statement. JAMA. 2024;331(22):1918-1930. doi.org/10.1001/jama.2024.5534
- Qaseem A, Harrod CS, Balk EM, et al. Screening for Breast Cancer in Asymptomatic, Average-Risk Adult Females: A Guidance Statement From the American College of Physicians (Version 2). Annals of Internal Medicine. 2026;179(6):842-856. doi.org/10.7326/ANNALS-25-05116
- Stout NK, Miglioretti DL, Su YR, et al. Breast Cancer Screening Using Mammography, Digital Breast Tomosynthesis, and Magnetic Resonance Imaging by Breast Density. JAMA Internal Medicine. 2024;184(10):1222-1231. doi.org/10.1001/jamainternmed.2024.4224
- Tosteson ANA, Stout NK, Su YR, et al. Outcomes of Density-Targeted Supplemental Breast Magnetic Resonance Imaging Screening by Breast Cancer Risk: Long-Term Health and Economic Considerations. Annals of Internal Medicine. 2026;179(4):486-496. doi.org/10.7326/ANNALS-25-00792
- Sprague BL, Stout NK, Schechter C, et al. Benefits, harms, and cost-effectiveness of supplemental ultrasonography screening for women with dense breasts. Annals of Internal Medicine. 2015;162(3):157-166. doi.org/10.7326/M14-0692
- Duggan SN, Azharuddin M, Hernández R, et al. Supplemental imaging modalities for breast cancer screening in women with dense breasts: A systematic review with economic considerations. Breast. 2025;85:104668. doi.org/10.1016/j.breast.2025.104668
- KFF Employer Health Benefits Survey, 2025 Annual Survey Summary of Findings.
- FDA, Final Rule to Amend the Mammography Quality Standards Act, effective September 10, 2024.
- DenseBreast-info, state insurance coverage law map.
- NCCN Breast Cancer Screening and Diagnosis, updated March 5, 2026.
- Elmi A, et al. Clinical Imaging, August 11, 2026, reported at AuntMinnie.
- GE HealthCare press release, August 3, 2026.
This piece is analysis and commentary based on public sources as of the date noted, and nothing in it constitutes legal, clinical, or financial advice. It references GE HealthCare's public product announcement and publicly funded, peer-reviewed clinical and economic literature as market examples, and its inclusion of any named party implies no criticism beyond the analysis stated. Interpretive conclusions are the author's opinion, offered for analytical discussion. All views reflect independent professional judgment and do not represent the views or positions of any current or former employer or affiliated organization.