MolDx has a non-coverage LCD for esophageal biomarkers to aid in detection of Barrett's esophagus and its progression. On September 4, 2025, they held a multi-jurisdictional advisory panel on the topic (aka Contractor Advisory Committee, CAC).
At this point, all materials are posted: The agenda (trivial), the panel questions (#17), and the bibliography (#23), as were available for the meeting. And they've added a YouTube archive video of the meeting, and, an official 29-page transcript.
Find the materials all here:
https://med.noridianmedicare.com/web/jeb/policies/lcd/cac#upcomingmeetings
###
AI Corner
###
I asked Chat GPT to review the meeting transcript. Note that I fed some of the materials into Chat GPR for its "insights" back in September, also, before the full materials were available online. For example, then, I asked it to compare 2021 CAC questions with the newest 2025 CAC questions.
###
I'm printing the Chat GPT reading as it came out, but I think it's a little too conservative.
Therefore, further below, I'm including a prompt asking Chat GPT to look at fact patterns and interpretations that are ultimately more favorable. This created a second AI dialog.
The two dialogs would be of most use to a reader who knows the field well and can use the two Chat GPT outputs to judge, for himself/herself, which AI viewpoint is more convincing.
###
SUMMARY OF AI OUTPUT
Executive Summary
MolDx currently maintains a non-coverage LCD for molecular biomarker tests intended to aid in detection and risk stratification of Barrett’s esophagus (BE) and its progression to esophageal adenocarcinoma. On September 4, 2025, MolDx convened a multi-jurisdictional Contractor Advisory Committee (CAC) to reconsider the topic. All materials are now publicly available, including the 17 formal panel questions, a 23-item bibliography, a two-hour archived video, and a 29-page official transcript.
Below, I present two AI-assisted analyses of the meeting.
The first is a more conservative reading, concluding that MolDx is unlikely to shift from non-coverage in the near term absent stronger evidence of clinical utility (CU), particularly in Medicare populations. That interpretation emphasizes MolDx’s longstanding AV/CV/CU framework and its caution regarding generalizability, dysplasia detection sensitivity, and durable management change.
However, I also asked the AI to reanalyze the transcript from a more favorable interpretive lens—remaining scientifically grounded but recognizing that policy signals often appear in tone, structure, and line of questioning before they appear in formal LCD language. That counterpoint suggests Dr. Bien-Willner’s questioning may reflect constructive policy drafting rather than skepticism, with a plausible ≥50% likelihood of a limited favorable LCD proposal within six months—particularly one focused on high-risk, triage-oriented, rule-out use cases.
Readers familiar with MolDx and the BE biomarker literature may find value in reviewing both analyses side by side and drawing their own conclusions about where this reconsideration is likely headed.
###
MORE CONSERVATIVE (AND FIRST) AI READING
###
Below is a structured, journalist-style summary of the CAC meeting, followed by a policy forecast grounded in the tone of the discussion — especially comments attributed to Dr. Bien Willner (MolDx Medical Director) — and aligned with the existing non-coverage LCD.
Key documents referenced:
CAC Reconsideration Questions
Bibliography submitted to the CAC
Noridian/MolDx CAC Transcript (29 pp., ~2 hrs)
Part I — Journalistic Summary of the CAC Meeting
The Framing: Not “Does It Work?” but “Is It Ready?”
The CAC was structured around a very specific tension:
Can non-endoscopic or minimally invasive molecular tests for Barrett’s Esophagus (BE) meaningfully improve detection and management in Medicare patients — and is the evidence strong enough to change coverage?
The tone of the meeting was analytical rather than adversarial. The panelists largely agreed on three background facts:
Endoscopy (EGD) is imperfect — under-sampling, patchy dysplasia, variable adherence to Seattle protocol.
BE surveillance is inconsistently applied, and many EAC cases occur without prior BE diagnosis.
Non-endoscopic approaches are clinically appealing, especially in high-risk GERD populations.
But the real debate centered on whether currently available molecular tests have demonstrated:
Analytical validity (AV)
Clinical validity (CV)
Clinical utility (CU)
— in a Medicare-relevant population.
That distinction dominated the discussion.
Major Themes from the Panel
1. Current EGD Use is Selective and Imperfect
In response to Questions 1–4 :
Panelists emphasized:
Not all GERD patients are referred for EGD.
Surveillance intervals are variably followed.
There are access, compliance, and no-show issues.
EGD is invasive, costly, and subject to sampling error.
However, several gastroenterologists pushed back on the idea that EGD is failing broadly. The issue is not that EGD is useless — it is that it is not scalable as a population screening tool.
That nuance matters.
MolDx appeared focused not on dissatisfaction with EGD — but on whether molecular tests demonstrably improve outcomes beyond it.
2. “Rule-Out” vs “Rule-In” Framing Was Central
Questions 5–8 drove a key conceptual distinction:
A rule-out test must have extremely high NPV and sensitivity.
A rule-in test must have strong PPV and management consequences.
Panelists were generally more comfortable with the “rule-out” model.
However, MolDx repeatedly steered discussion toward:
What is the acceptable false negative rate?
What harm occurs if dysplasia is missed?
Is surveillance actually reduced in practice?
There was skepticism that published studies sufficiently demonstrate:
Durable rule-out performance over time.
Clear reduction in downstream EGDs.
Improved stage-shift in EAC detection.
The BEST trials (UK) and methylation biomarker studies (Moinova, Iyer, Shaheen, etc.) were acknowledged as important — but repeatedly described as:
Enriched cohorts
UK health-system specific
Not directly translatable to US Medicare practice
That last point was emphasized more than once.
3. Clinical Utility Remains the Fracture Line
Several panelists acknowledged:
Strong NPV in some studies
Good patient tolerability
Growing US-based prospective data
However, MolDx questioning repeatedly returned to:
Has anyone demonstrated reduction in mortality?
Has anyone shown fewer unnecessary EGDs in real-world Medicare?
Is there prospective evidence of management change sustained over time?
Dr. Bien Willner’s comments (as reflected in the transcript ) were measured but pointed. His posture appeared to be:
Molecular tests are biologically plausible.
The literature is evolving.
But MolDx requires evidence of clinical utility, not just accuracy.
He appeared particularly focused on:
Long-segment vs short-segment disease performance.
Dysplasia detection sensitivity.
Operator variability in sample retrieval.
Generalizability to Medicare-age populations (>65).
That is consistent with MolDx’s AV/CV/CU framework already articulated in the LCD.
4. Medicare Population Applicability Was a Repeated Concern
Multiple times the discussion returned to:
How many study participants were Medicare age?
Were comorbidities represented?
Does the risk-benefit calculus change in older adults?
MolDx historically requires population congruence. The transcript suggests they are not willing to extrapolate from predominantly younger screening cohorts.
5. No Clear Consensus for Immediate Coverage
Crucially:
The panel did not coalesce around a strong statement that “current evidence supports coverage.”
Instead, the sentiment was:
Promising.
Potentially valuable.
Evidence improving.
Not yet definitive.
That tone matters.
Part II — What is MolDx Likely to Do?
We now overlay the meeting tone with the existing non-coverage LCD (L39256) .
The current LCD states:
“There are currently no existing tests that have demonstrated AV, CV, and clinical utility (CU) to fulfill the necessary criteria.”
The policy explicitly requires:
Demonstrated analytical validity
Demonstrated clinical validity
Demonstrated clinical utility
Technical Assessment completion
The CAC transcript suggests MolDx still believes the CU bar has not been met.
Let’s examine possible outcomes.
Scenario 1: Do Nothing (Most Likely)
Probability: High
Rationale:
The CAC did not produce overwhelming clinical endorsement.
No new mortality or management-change trials were presented.
Evidence remains accuracy-focused, not outcome-focused.
Dr. Willner’s tone suggests caution, not imminent reversal.
MolDx may simply:
Maintain non-coverage.
State that evidence remains insufficient for CU.
Encourage further prospective Medicare-relevant data.
This would be fully consistent with the current LCD structure.
Scenario 2: Revise the Non-Coverage LCD (Moderately Likely)
Probability: Moderate
Possible revisions:
Clarify evidentiary thresholds.
Specify required endpoints.
Define acceptable rule-out sensitivity benchmarks.
Clarify Medicare-age cohort expectations.
MolDx sometimes uses CAC meetings to sharpen policy language rather than reverse position.
This would signal:
“We are not covering yet — but here is what you must show.”
If this occurs, expect:
More explicit CU criteria.
Possibly narrowed language around enriched populations.
Possibly explicit discussion of US vs UK generalizability.
Scenario 3: Issue Favorable Coverage (Unlikely in Near Term)
Probability: Low (short term)
Why unlikely:
No strong consensus statement from CAC.
No new randomized US Medicare-based outcome trials.
Persistent concerns about dysplasia detection sensitivity.
No demonstrated mortality reduction.
MolDx rarely pivots to favorable LCD absent:
Strong prospective multicenter US data.
Clear management change evidence.
Clear reduction of unnecessary EGDs.
Evidence of stage shift in EAC.
None of these appeared convincingly demonstrated in the transcript.
Strategic Reading of Dr. Willner’s Posture
Dr. Willner’s questions suggest MolDx is thinking in this framework:
Is this replacing EGD? If yes, the bar is extremely high.
Is this triaging EGD? Then show reduction in unnecessary EGDs.
Is this improving dysplasia detection? Then show improved outcomes.
Is this Medicare relevant? Show data in older populations.
He did not dismiss the technology. But he did not signal readiness to move.
The subtext was:
“Show me real-world management change and durable outcomes.”
Bottom Line
The CAC meeting did not read like a prelude to a favorable LCD.
It read like:
An evidence inventory exercise.
A pressure test of clinical utility.
A signal that MolDx still views CU as unmet.
Most likely path:
Maintain non-coverage, possibly refine evidentiary expectations.
What would change the outcome?
Large US prospective trials in Medicare-age patients.
Demonstration of reduced EGD utilization.
Clear stage-shift or dysplasia detection benefit.
Economic modeling specific to US Medicare.
Strong US society endorsement incorporating these tests into guidelines.
Until then, MolDx is unlikely to move. [End of first dialog]
While staying honest to the fact patterns and scientifically reasonable, provide a counterpoint analysis from an objective expert who believes Dr Bien Willner's tone and questions were, for him, relatively favorable and a favorable LCD proposal is at least 50% in the next six months.
Remember that your analysis must be cogent and fact-based and written as if you were preparing for somewhat skeptical readers (who are more conservative about LCD change).
###
###
SECOND AND MORE FAVORABLE AI READING
Here’s a fact-pattern-honest counterpoint—written as if by an objective expert who expects skeptical readers, but who nonetheless reads Dr. Gabriel Bien-Willner’s tone and line of questioning as relatively favorable and thinks a favorable LCD proposal is ≥50% likely within ~6 months.
Why an “optimistic read” is scientifically reasonable
A conservative reader will correctly note: a CAC transcript is not a coverage decision, and MolDx’s current LCD posture is non-coverage (i.e., “no tests have yet demonstrated AV/CV/CU”).
That said, the way a payer medical director structures and moderates a CAC can be an early signal of direction. In this transcript, Dr. Bien-Willner’s approach looks less like “building a denial record” and more like “building a coverage record with guardrails.”
1) The meeting design itself looks like reconsideration with a pathway, not a dead-end
MolDx didn’t convene a generic, adversarial debate.
Instead, the meeting is tightly organized around triage logic: screening context, compliance failures, what “reliable” would mean, and what threshold performance is needed for rule-out vs rule-in testing—the exact architecture you’d use to draft a limited favorable LCD.
If MolDx were simply reaffirming non-coverage, they could have leaned almost entirely on “insufficient CU” and stopped. Instead, they pushed panelists to articulate (on record) how management would change given certain performance characteristics—especially NPV-driven “stop” rules (“if negative, you’re done”).
That is the kind of record you create when you may later justify coverage conditioned on performance + use-case.
2) Dr. Bien-Willner repeatedly steers toward actionable decision rules (a coverage hallmark)
A skeptic may interpret his questioning as “picking holes.” But listen to the form: he asks panelists to state operational consequences of test results, not just general enthusiasm.
Example: when a panelist initially gives a broad answer, Bien-Willner pushes for the precise management impact (“how would that information… change management?”), and once the respondent states the practical outcome (“negative → no endoscopy”), he accepts it and moves forward.
That “pin-down-the-workflow” style is typical when drafting coverage criteria such as:
Who is eligible (risk-factor enriched symptomatic population)
What the test is used for (rule-out triage vs rule-in enrichment)
What happens next (EGD referral yes/no)
What harms are tolerable (false positives vs false negatives)
Those are precisely the questions in the formal reconsideration list.
3) He emphasizes “put it on the record”—that’s not neutral; it’s record-building
He explicitly notes that some questions seem self-evident, but he wants them “stated on the record” (e.g., pathology required for BE diagnosis).
A denial record can do this too, but what he chooses to put on the record matters: he highlights the binary nature of diagnosis, the barriers to EGD, and the potential clinical logic of a high-NPV triage pathway. These are coverage-supportive facts if MolDx later proposes a carefully limited policy.
4) He allows, and does not undercut, repeated panel consensus on the “NPV/triage” value proposition
Across multiple experts, the repeated “headline claim” is: with an extremely high NPV, clinicians would not proceed to EGD for BE workup, reserving endoscopy for positives or other indications.
Importantly, Bien-Willner does not argue against that logic. He probes for thresholds (“how high is high enough?”), which is exactly what you need to operationalize coverage.
A conservative reader might say, “they still need CU.” True. But the panel is offering the CU mechanism in plain terms: reduced procedural burden and enriched endoscopy yield, while maintaining patient safety via high NPV. MolDx may be signaling willingness to accept “decision-analytic utility” + pragmatic outcomes (reduced endoscopies, improved yield) as a practical CU surrogate—if supported by data.
5) The bibliography choice subtly supports a “near-term pivot” narrative
The bibliography includes multiple recent (2024–2025) clinical validation/utility-adjacent studies and real-world evaluations (including prospective validation and real-world NHS evaluation items, as well as newer US-context publications).
An objective but optimistic read is: MolDx is assembling a dossier of freshening evidence sufficient to justify at least a draft favorable LCD with restrictions.
How a favorable LCD could happen without MolDx “changing its personality”
A skeptical reader’s strongest objection is: “MolDx rarely covers without clear CU.” That’s fair, and it’s exactly why the most plausible favorable outcome is not broad coverage, but narrow, conditional coverage that converts the CU debate into checkable conditions.
A likely “favorable but controlled” proposal (consistent with the questions posed ) would look like:
Indication-limited: high-risk symptomatic GERD patients meeting guideline-style risk enrichment (not general population).
Use-case-limited: explicitly triage/rule-out (and/or “EGD enrichment”), not “replace EGD” and not broad surveillance claims.
Performance-anchored: required minimum sensitivity/NPV in intended-use population, with special emphasis on not missing HGD/EAC (a theme the panel itself emphasizes).
Operational requirements: defined follow-up pathways for positives, and defined re-testing intervals or triggers for negatives.
Data development hook: registry, outcomes tracking, or coverage-with-data expectations (even if informal), to convert “we need CU” into “we will watch CU.”
This would allow MolDx to remain logically consistent with its current LCD language (AV/CV/CU expectations) while still moving from “no” to “yes, but only here—and only like this.”
Why ≥50% in six months is defensible (even to conservative readers)
This is not “they sounded excited.” It’s a timing-and-process argument:
They are already in reconsideration mode (new key questions + updated bibliography + CAC).
The moderator’s questioning is policy-drafting oriented, not merely cross-examining.
The panel repeatedly supplies the specific CU mechanism MolDx needs (triage to reduce EGDs, improve yield, address access/compliance barriers).
MolDx can issue a favorable proposal that is narrow—which is operationally easier than reversing to broad coverage.
On that basis, an objective expert can reasonably say: a draft favorable LCD proposal (even if constrained and later modified) is at least as likely as “no change,” because the meeting seems designed to generate the exact record needed for a constrained yes.
What would falsify this optimistic interpretation?
To stay honest for skeptical readers, three “tells” would push the probability back down:
If MolDx reiterates that no published evidence demonstrates management change in Medicare-age cohorts, and insists on hard outcomes (stage shift, mortality), the path slows. (The current LCD posture already points that way.)
If MolDx frames operator variability/sample failure as unmanageable in real-world practice, that undermines the triage narrative (because false negatives become politically unacceptable).
If MolDx treats UK/NHS-linked evidence as non-transferable without US Medicare replication, they may prefer “revise non-coverage” over “propose coverage.”
Bottom line counterpoint
A conservative reader is right that MolDx still cares about CU and Medicare relevance. But the structure and tone of Dr. Bien-Willner’s questioning in this CAC—pinning down workflows, thresholds, and decision consequences, and repeatedly eliciting the “negative → stop” management model—can reasonably be read as constructively favorable.
That pattern supports a plausible near-term outcome: not a sweeping reversal, but a limited favorable LCD proposal aimed at triage/rule-out use in a defined high-risk population, potentially paired with data expectations.
###
