News today of Anthem covering blood-based Alzheimer tests (here, here, policy here) reminded me of the tortured history of CMS and amyloid PET scans, plus the fact that no MACs have yet brought out any Alzheimer test policies (neither for serum tau nor PET).
I asked Chat GPT to research and write up the history of CMS coverage of Amyloid PET. (Prompt at bottom).
# # #
Amyloid PET in Medicare:
From Extreme Control to Near–Auto-Payment
Research found one variance from your request. IDEAS produced a conspicuously split result. It was strongly positive on a comparatively soft physician-behavior endpoint and negative on its harder health-care utilization endpoint. Much of the favorable publicity blurred that distinction.
Medicare's coverage of amyloid PET imaging has undergone a remarkable reversal over the past decade.
In 2013, CMS adopted one of its most restrictive diagnostic imaging policies, limiting coverage almost exclusively to Coverage with Evidence Development studies because evidence that amyloid PET improved patient outcomes was lacking. The resulting IDEAS study showed that PET scans frequently changed physicians' diagnoses and management plans but failed to demonstrate reductions in hospitalizations or emergency department visits, despite promotional messaging that often emphasized the management findings. A second study, New IDEAS, largely confirmed the management-change results in a more diverse population without establishing improved clinical outcomes.
By 2023, however, the emergence of anti-amyloid therapies fundamentally changed the policy landscape. CMS withdrew its National Coverage Determination entirely, leaving coverage decisions to Medicare Administrative Contractors.
Ironically, this shifted Medicare from one of its most tightly controlled diagnostic policies to a largely permissive system with few published contractor coverage policies and substantial reliance on routine claims processing.
1. The original CMS decision was unusually restrictive
The commercial amyloid PET tracer era began with FDA approval of florbetapir in 2012, followed by additional tracers in 2013 and 2014.
In September 2013, CMS concluded that the evidence did not adequately show that amyloid PET improved meaningful health outcomes, or that management guided by amyloid PET improved such outcomes. Medicare therefore covered essentially one amyloid PET scan per beneficiary only through Coverage with Evidence Development, principally in approved studies addressing difficult differential diagnoses or clinical-trial selection. [1]
That decision was stringent, but it was intellectually coherent.
At the time:
Amyloid positivity did not by itself establish that amyloid was causing a patient’s cognitive impairment.
A meaningful proportion of cognitively normal older adults had amyloid deposition.
There was no disease-modifying treatment whose use depended upon amyloid confirmation.
A more certain pathological classification did not necessarily produce better treatment or better outcomes.
CMS was effectively saying:
“An informative scan is not enough. Demonstrate that obtaining the information benefits the patient.”
2. IDEAS was designed around two very different claims
The original IDEAS study was led by Gil Rabinovici at the University of California, San Francisco, with the Alzheimer’s Association and the American College of Radiology among the principal participating organizations.
Its first major analysis included 11,409 Medicare beneficiaries with mild cognitive impairment or dementia of uncertain cause.
Aim 1: Did the scan change what physicians planned to do?
Yes—dramatically.
Management changed within approximately 90 days in about 60% of patients with mild cognitive impairment and 64% of patients with dementia. Diagnoses also changed to or from Alzheimer disease in roughly one-third of patients. [2]
But this endpoint was much less impressive than it sounded.
First, it measured change, not improvement.
Second, the changes included prescribing or discontinuing symptomatic medications, counseling, referrals, and other actions whose net benefit was not established.
Third, the study was nonrandomized and had no concurrent control group for this endpoint.
Fourth, patients were enrolled only when a dementia specialist considered the cause of impairment uncertain and believed that the PET result was likely to alter diagnosis or management.
In other words, the population was partly selected precisely because the physician expected to change something after seeing the scan.
Finding frequent management changes was important, but it was not surprising. It also did not establish that those changes were correct or beneficial.
Aim 2: Did the scan reduce hospitalizations or emergency-department use?
No. This is the result many observers remember less clearly.
The prespecified target was at least a 10% reduction in 12-month hospitalizations or emergency-department visits compared with matched Medicare controls.
Hospitalization occurred in approximately 24.0% of IDEAS participants versus 25.1% of controls. This amounted to a nonsignificant relative reduction of approximately 4.5%, well below the prespecified target.
Emergency-department use was essentially identical, at approximately 44.8% in both groups.
The investigators concluded that amyloid PET was not associated with a significant reduction in either hospitalization or emergency-department use. [3]
It is therefore fair to say:
IDEAS succeeded on its process endpoint but failed on its principal downstream health-care utilization endpoint.
Calling the entire study “negative” overlooks the management-change result.
Calling it a demonstration of improved patient care overlooks the utilization result and confuses changed management with better outcomes.
3. The favorable publicity went beyond what the evidence established
The 2019 IDEAS and UCSF publicity described the first phase as demonstrating “clinical impact.” It stated that amyloid PET could improve diagnostic accuracy and lead to better medical management. [4]
The release was not literally false. It clearly described these findings as first-phase results and acknowledged that the hospitalization and emergency-department analysis was still pending.
Nevertheless, the rhetoric traveled farther than the evidence.
“Management changed” became “management improved.”
“Diagnostic classification changed” became “diagnosis improved.”
A large observational percentage became evidence of “clinical utility.”
The unresolved question of patient outcomes was deferred to the future and received far less rhetorical emphasis.
This is a familiar maneuver in diagnostic-test advocacy: emphasize the largest proximal effect and permit readers to infer an unproven distal benefit.
The current IDEAS website remains strongly affirmative, describing the project as major evidence for the clinical utility of amyloid PET while also acknowledging that the hospitalization and emergency-department endpoint was not met. [5]
The publicity can therefore reasonably be characterized as selectively accurate but highly promotional.
4. New IDEAS largely repeated the management-change exercise
New IDEAS was created to enroll a more ethnoracially and clinically diverse Medicare population, addressing one of the major limitations of the original study.
Its principal published report included approximately 5,700 participants and again found management changes in about 59% of patients, exceeding its prespecified benchmark of 30%. [6]
That is useful evidence that amyloid PET information affects specialist behavior in a broader population.
It is not randomized evidence that those changes improve:
cognition;
function;
quality of life;
caregiver burden;
safety;
hospitalization;
emergency-department use; or
survival.
The major New IDEAS result therefore remains another management-change study rather than a demonstration of improved patient outcomes.
5. The 2023 CMS reversal was driven primarily by the treatment environment
On October 13, 2023, CMS removed National Coverage Determination 220.6.20 entirely.
CMS ended Coverage with Evidence Development, eliminated the once-in-a-lifetime national restriction, and transferred reasonable-and-necessary determinations to the Medicare Administrative Contractors under section 1862(a)(1)(A). [7]
The decisive change was not that IDEAS had demonstrated fewer hospitalizations.
In fact, the negative hospitalization and emergency-department analysis appeared online on October 9, 2023, four days before CMS issued its final decision. [3]
What had changed was the therapeutic environment.
With anti-amyloid monoclonal antibodies entering clinical practice, amyloid confirmation had become necessary to:
identify patients who actually had amyloid pathology;
avoid exposing amyloid-negative patients to ineffective treatment and risks such as amyloid-related imaging abnormalities;
determine eligibility for treatment; and
potentially assess plaque clearance or inform decisions about discontinuing treatment.
CMS explicitly recognized that the old Coverage with Evidence Development questions had been formulated before amyloid-based treatment selection became clinically important. It also concluded that the one-scan lifetime limit was poorly suited to the emerging treatment environment. [7]
CMS did not really conclude:
“IDEAS proved that amyloid PET improves health outcomes.”
CMS effectively concluded:
“Amyloid PET has become a necessary enabling test for a new category of covered treatment, so the 2013 national research-only restriction is no longer workable.”
That is a different argument, and a substantially more defensible one.
6. The absence of MAC policies creates near–auto-payment, but not a formal guarantee
As of July 2026, there does not appear to be a current amyloid-specific Local Coverage Determination in the public Medicare Coverage Database.
The Medicare Administrative Contractors appear not to have translated CMS’s delegation into explicit, prospective coverage criteria.
Of course, the absence of an LCD does not legally mean that every claim is automatically covered.
The underlying section 1862(a)(1)(A) reasonable-and-necessary requirement remains in effect. In the absence of an NCD or LCD, a contractor can still:
allow a clean claim through ordinary automated processing;
apply coding or diagnosis edits;
request records through medical review;
deny an individual claim as insufficiently documented; or
conduct post-payment review.
The most accurate description is therefore:
De facto permissive payment under individual-claim discretion, rather than a formal entitlement to automatic payment.
Still, there is considerable practical overlap between having no published policy and routine auto-payment.
Medicare fee-for-service claims data identified 908 amyloid PET scans during October through December 2023 and 2,951 during January through March 2024. Scans appeared in all 12 MAC regions. [8]
Substantial numbers of claims were therefore plainly being paid after the NCD disappeared.
Utilization varied enormously among MAC jurisdictions, from approximately 31 to 613 scans per 100,000 beneficiaries with mild cognitive impairment or dementia during the first quarter of 2024. [8]
Some of that variation presumably reflected scanner availability, referral patterns, specialist concentration, and the locations of anti-amyloid treatment centers.
Nevertheless, it undercuts CMS’s prediction that evidence-based MAC discretion would produce reasonably consistent national coverage.
Medicare Advantage is a separate matter. Medicare Advantage plans may employ prior authorization and more explicit medical-necessity criteria even when traditional Medicare has no LCD.
7. From explicit overcontrol to silent permissiveness
The description of the policy history as a movement from one extreme to the opposite is essentially correct.
In 2013, CMS imposed a highly visible national presumption of noncoverage, a one-scan lifetime limit, research protocols, and centralized control. CMS demanded evidence at the far end of the diagnostic-utility chain: improved patient outcomes.
In 2023, CMS did not replace that policy with a carefully bounded national coverage rule. It simply removed the rule and delegated the question to the contractors.
The MACs then largely declined to publish local rules, leaving coverage to ordinary claims processing and comparatively invisible case-by-case discretion.
The resulting regulatory cycle looks like this:
A national evidentiary standard so demanding that routine access was nearly impossible.
A large Coverage with Evidence Development study showing that the information changed physician behavior but did not improve the selected utilization outcomes.
Promotional presentation of the physician-behavior result as evidence of clinical improvement.
A new treatment environment that made amyloid confirmation operationally necessary.
Withdrawal of national controls without transparent replacement criteria.
The central policy lesson is that diagnostic utility exists at several distinct levels:
Does the test measure amyloid accurately?
Does it clarify the diagnosis?
Does it change management?
Does it improve treatment selection?
Does it improve patient outcomes?
IDEAS provided strong evidence for the middle steps, but not the final one.
The subsequent emergence of anti-amyloid drugs made the middle steps much more consequential. CMS therefore relaxed coverage for a legitimate new reason.
But instead of acknowledging that revised logic through a modernized national coverage policy, CMS shifted the issue into decentralized claims administration.
That was administratively expedient, but intellectually untidy and procedurally opaque.
It produced access without a public rule—probably preferable to continuing the obsolete 2013 prohibition, but not a particularly admirable endpoint for Medicare coverage policymaking.
The policy history therefore seems to have moved from one extreme to the other: from highly centralized, tightly controlled coverage through research studies to a largely permissive system with no clear national rule and few, if any, explicit local ones. How should this sequence of events be understood and interpreted?
# # #
# # #
PROMPT USED:
When amyloid PET tracers were introduced around 2013, CMS determined that Medicare generally would not cover them except through certain controlled research studies. That decision led to the IDEAS study, which was led by investigators at UCSF. # # As I recall, IDEAS produced a mixed—and in an important respect negative—result. Amyloid PET frequently changed physicians’ diagnoses or management plans, but the study did not achieve its harder outcomes, such as reducing hospitalizations or emergency-department visits. Nevertheless, UCSF and the study’s sponsors presented the findings as favorably as possible in their public statements. At the time, I found some of that messaging distasteful and potentially misleading because it seemed to blur the distinction between changing management and improving patient outcomes. # # There was later a follow-up study, New IDEAS, although I know less about its findings. # # Then, in 2023, CMS eliminated the national coverage determination for amyloid PET entirely and left coverage to the discretion of the Medicare Administrative Contractors. Yet no contractor appears to have issued a specific local coverage policy. In practical terms, that suggests that many claims may now pass through routine claims processing in the absence of any published restriction, subject of course to ordinary medical review and post-payment audit authority.
References
Centers for Medicare & Medicaid Services. “Decision Memo for Beta Amyloid Positron Emission Tomography in Dementia and Neurodegenerative Disease.” September 27, 2013.
https://www.cms.gov/medicare-coverage-database/view/ncacal-decision-memo.aspx?NCAId=265&proposed=N
Rabinovici GD, Gatsonis C, Apgar C, et al. “Association of Amyloid Positron Emission Tomography With Subsequent Change in Clinical Management Among Medicare Beneficiaries With Mild Cognitive Impairment or Dementia.” JAMA. 2019;321(13):1286-1294.
https://jamanetwork.com/journals/jama/fullarticle/2729371
Rabinovici GD, Gatsonis C, Apgar C, et al. “Amyloid PET and Subsequent Health Care Use Among Medicare Beneficiaries With Mild Cognitive Impairment or Dementia.” JAMA Neurology. Published online October 9, 2023.
https://jamanetwork.com/journals/jamaneurology/fullarticle/2809940
IDEAS Study. “JAMA Publishes IDEAS Study Results.”
https://www.ideas-study.org/News/JAMA-IDEAS-Study-Results
IDEAS Study. “Original IDEAS Study.”
https://www.ideas-study.org/Original-Study
New IDEAS Research Group. New IDEAS study report on changes in clinical management following amyloid PET. Alzheimer’s & Dementia. 2025.
https://alz-journals.onlinelibrary.wiley.com/doi/10.1002/alz.70504
Centers for Medicare & Medicaid Services. “Decision Memo for Beta Amyloid Positron Emission Tomography in Dementia and Neurodegenerative Disease.” October 13, 2023.
https://www.cms.gov/medicare-coverage-database/view/ncacal-decision-memo.aspx?ncaid=308&proposed=N
JAMA Neurology. Analysis of Medicare amyloid PET utilization following removal of the national coverage determination.
https://jamanetwork.com/journals/jamaneurology/fullarticle/2827462