Why CT Angiogram Scans are Highly Questionable (Simon Hill)

#1

I have posted several times my skepticism regarding the utility of CT angiogram scans as they are conducted today.

From what I can see, these scans are extremely unreliable, differing as much as 600% in computing total plaque burden between platforms and they use methodologies which are very questionable insofar as they are unable to reliably measure key variables (such as size of vessels, vessel walls etc.). There is a tremendous amount of noise.

Furthermore, the AI analysis is quite flawed and lacking in nuance in comparison with qualified human analysis which uses a wider evidence base.

These are among the reasons why I never had a CT angio scan done - I would love to get one, but only once the methodology changes and there has been some validated marker matrix established.

For now, I have simply gotten an ordinary CAC scan. It tells me the calcified plaque burden in my heart arteries. Obviously, it’s quite limited - it tells me nothing about the plaque in my whole body vasculature (also a limitation of CT angio scans) or more importantly soft plaque which can be more dangerous. Here a CIMT might be useful to complement the CAC scan.

I suppose there may still be utility to CT angios - if you use the same platform (for example Cleerly), the same machine and the same settings and protocol taken at two points in time, years apart, to show if there is any progression of plaque. That’s it. But to me - that’s too little. Therefore, for now, I’m staying away from CT angio until the technology and methodology improves.

The experience of Simon Hill, who had his arteries scanned simultaneously by three different platforms - one of which is a research platform not even available commercially at this point - is illustrative of the quagmire current CT angio technology is, as can be seen in the yt video below.

I am not trying to tell anyone whether they should or should not go in for a CT angio. I am just saying that you should be aware of the limitations of the current technology and analysis and that you cannot draw far going conclusions about your plaque burden using these tools as they are configured today.

To save time, start with about minute 25:00 or so, as the time spent before is just setting it all up.

I Thought I Reversed My Artery Plaque by 48%. Here’s What Really Happened | Simon Hill | EP#439 (via The Proof with Simon Hill)

1 Like
#2

I. Executive Summary

This clinical presentation by Simon Hill deconstructs the biostatistical and pathophysiological realities behind his self-tracked serial coronary computed tomography angiography (CCTA) scans. Following public attention regarding a reported “48% coronary plaque regression” over a 16.5-month interval, Hill presents a rigorous audit of the underlying raw imaging data, highlighting the severe measurement artifacts, algorithm-dependent variance, and clinical limitations inherent to contemporary artificial intelligence (AI)-driven quantitative coronary computed tomography (AI-QCT).

Between December 2024 (age 38) and April 2026 (age 39), serial scans conducted at the Lundquist Institute under Dr. Matthew Budoff were analyzed using three distinct commercial software engines: QAngio, Cleerly, and HeartFlow. Despite identical acquisition parameters, total plaque volume varied six-fold at follow-up (QAngio: 11.5 mm³; HeartFlow: 60 mm³; Cleerly: 74 mm³). Longitudinal trajectory analysis similarly diverged: QAngio showed near-absolute stability (+1.1 mm³), Cleerly suggested a ~20% progression, and HeartFlow calculated a 48% regression. Hill acknowledges that the 48% regression figure is an artifact of proprietary lumen-vessel wall edge-detection algorithms and short scan intervals. In very low absolute plaque burdens (~1–6 grains of rice spread across 60 cm of coronary vasculature), minor variations in luminal segmentation thresholds generate disproportionately large percentage swings.

Pathophysiologically, the episode reinforces the concept of cumulative lifetime exposure to apolipoprotein B (ApoB)-containing lipoproteins—the “area under the curve” (AUC)—explaining why high-fiber plant-based dietary patterns and low-risk lifestyle metrics cannot fully erase vascular damage accumulated over three decades of omnivorous dietary intake (LDL-C≈125–130 mg/dL) alongside polygenic susceptibility.

Armed with expert radiological re-review (favoring expert-overseen QAngio metrics indicating plaque stabilization with soft-to-fibrous/calcified remodeling), Hill transitioned from pure lifestyle monotherapy (ApoB 70–80 mg/dL) to aggressive pharmacological lipid lowering with the PCSK9 inhibitor evolocumab (Repatha) targeting ApoB≤40 mg/dL(LDL-C<55 mg/dL). This decision aligns with the landmark GLAGOV trial and the 2026 ACC/AHA Dyslipidemia Guidelines, which emphasize early intervention, lifetime cumulative ApoB burden reduction, and plaque stabilization over premature reliance on AI volumetric short-interval follow-ups.

II. Insight Bullets

  • Serial coronary computed tomography angiography (CCTA) allows direct volumetric quantification and phenotypic classification (calcified, fibrous, low-attenuation soft plaque) of subclinical atherosclerosis.
  • An initial reported 48% coronary plaque volume reduction over 16.5 months was confirmed to be a platform-specific measurement discrepancy rather than true biological clearance.
  • The host processed identical DICOM raw datasets through three leading quantitative platforms: QAngio, Cleerly, and HeartFlow.
  • At follow-up, absolute plaque volume measurements differed six-fold across platforms: QAngio registered 11.5 mm³, HeartFlow registered 60 mm³, and Cleerly registered 74 mm³.
  • Directional trajectory diverged between platforms: QAngio reported stability (+1.1 mm³), Cleerly reported a ~20% increase, and HeartFlow reported a 48% decrease.
  • At minimal plaque burdens (11.5–74 mm³ spread across ~60 cm of coronary vasculature), marginal differences in vessel edge-detection algorithms generate massive relative percentage swings.
  • Atherosclerosis development is fundamentally driven by the lifetime cumulative exposure to ApoB-containing lipoproteins (“area under the curve”), rather than acute point-in-time serum concentrations.
  • Plant-based dietary patterns adopted at age 30 lowered the host’s LDL-C from ~125–130 mg/dL into the 70s, but could not retroactively eliminate previous arterial remodeling.
  • The large-scale European prospective cohort study (REACT cohort across Denmark and Spain) demonstrated that over one-third of asymptomatic men in their 30s harbor peripheral or coronary atherosclerosis, rising to two-thirds in their 40s.
  • In the REACT study cohort, asymptomatic men in their 30s who had detectable coronary plaque displayed a median plaque volume of 42.6 mm³ on QAngio, placing the host’s 11.5 mm³ burden in the lower quartile of affected individuals.
  • Population percentile rankings remain highly platform-dependent; the host’s plaque volume registered at the 55th percentile on Cleerly (Miami Heart reference) but jumped to the 80th percentile on HeartFlow’s proprietary registry.
  • In the Miami Heart Study (JACC: Cardiovascular Imaging), investigators warned of AI-driven quantitative CT overestimating soft plaque relative to human readers.
  • Leading cardiac imaging expert Dr. Matthew Budoff identified semi-automated platforms with trained human-in-the-loop oversight (e.g., QAngio) as clinically superior and less prone to automated edge-detection artifacts than fully automated AI engines.
  • The host’s underlying plaque phenotype demonstrated functional stabilization over 16.5 months: soft, lipid-rich plaque decreased by ~40% with a compensatory increase in fibrous and dense calcified tissue.
  • In the landmark GLAGOV randomized controlled trial, evolocumab-induced LDL-C lowering to a median of 36.6 mg/dL drove significant atheroma regression on intravascular ultrasound (IVUS) without a discernible lower efficacy threshold.
  • The 2026 ACC/AHA Dyslipidemia Guideline re-established aggressive numerical targets, recommending an LDL-C<70 mg/dL (or <55 mg/dL in high-risk categories) when coronary artery calcium (CAC) exceeds the 75th percentile for age and sex.
  • A CAC score of 4 Agatston units at age 38 exceeds the 75th percentile due to the near-zero background incidence of coronary calcification in young adult males.
  • The host rejected daily oral statins and initially trialed lerodalcibep before establishing maintenance therapy on the monoclonal antibody PCSK9 inhibitor evolocumab (Repatha) to achieve ApoB≤40 mg/dL.
  • Adjunctive non-lipid therapy includes 2,400 mg/day of aged garlic extract, citing randomized clinical trials demonstrating attenuation of low-attenuation plaque progression.
  • Quantitative coronary serial plaque progression assessment requires an absolute minimum scanning interval of 2 to 5 years under identical scanner geometries to exceed measurement noise limits.
  • Rapid plaque progression is clinically defined as an increase in percent atheroma volume (PAV) of ≥1.0% per year; all three platforms confirmed the host remained well below this risk benchmark.

III. Adversarial Claims & Evidence Table

Claim from Video Speaker’s Evidence Scientific Reality (Current Data) Evidence Grade (A-E) Verdict
Serial CCTA demonstrated a true 48% coronary plaque regression in 16.5 months. HeartFlow automated AI volumetric analysis reporting total plaque volume reduction. Refuted by parallel multi-platform analysis and radiologic consensus. QAngio showed stability (+1.1 mm³) while Cleerly suggested a ~20% increase. In low-volume disease (11–74 mm³), minor segmentation boundary shifts cause massive percentage artifacts. True short-interval regression of this magnitude without intense lipid-lowering pharmacotherapy is biologically implausible (Choi et al., 2021). Level B Unsupported(Measurement artifact; speaker conceded this in the episode)
Fully automated AI-QCT algorithms systematically overestimate coronary soft plaque burden. The Miami Heart Study (co-authored by Dr. Matthew Budoff in JACC: Cardiovasc Imaging). Confirmed. Multiple clinical validations indicate that fully automated AI segmentation tools frequently mistake perivascular adipose tissue, partial-volume artifacts, and motion blur for non-calcified low-attenuation plaque, identifying pathology in vessels adjudicated as normal by expert human over-readers (Budoff et al., 2024; van Rosendael et al., 2023). Level A Strong Support
Atherosclerosis is governed by cumulative lifetime ApoB exposure (AUC), not current levels. Epidemiological studies, Mendelian randomization, and lifetime risk trajectories. Robustly validated. Lifetime Mendelian randomization models establish that low lifelong circulating LDL-C/ApoB reduces ASCVD risk three-fold more per unit reduction than late-life statin initiation, proving that vascular retention of ApoB-containing lipoproteins is cumulative and non-reversible by diet alone (Ference et al., 2017). Level A Strong Support
Plaque regression scales linearly with LDL-C down to 20 mg/dL with no lower threshold. Post-hoc analysis of the IVUS-based GLAGOV randomized clinical trial using evolocumab. Supported in secondary prevention cohorts. GLAGOV(Nicholls et al., JAMA 2016) demonstrated significant percent atheroma volume (PAV) regression (-0.95%) at achieved LDL-C 36.6 mg/dL, with regression observed down to ~20 mg/dL. However, translating this to primary prevention in young asymptomatic individuals with minimal plaque burden remains an extrapolation. Level B Plausible(Established in secondary CAD; extrapolated in primary prevention)
Aged garlic extract (2,400 mg/day) halts progression of low-attenuation plaque. Previous CCTA trials evaluating aged garlic extract (AGE) in coronary disease cohorts. Supported by small-to-moderate Phase II RCTs. Multiple trials by Budoff and colleagues demonstrated that 2,400 mg daily of Kyolic aged garlic extract significantly reduced low-attenuation plaque progression and slowed CAC progression compared to placebo over 12 months (Zeb et al., 2018; Matsumoto et al., 2016). Phase III cardiovascular outcome trials remain lacking. Level B Plausible(Surrogate imaging endpoints positive; lacks hard MACE outcomes)
Serial CCTA for plaque progression tracking requires a minimum 2–5 year interval. Clinical guidance and statement from the American College of Cardiology (ACC). Fully supported by radiological standards. Measurement variance and biological remodeling velocity dictate that serial CCTA imaging at intervals <2 years captures predominantly machine noise, contrast variation, and heart-rate gating discrepancies rather than true clinical progression (Hecht et al., SCCT Guidelines 2023). Level A Strong Support
A plant-based diet combined with resistance/cardio training stabilizes plaque. Personal serial QAngio results showing stability and phenotypic transition from soft to fibrous tissue. Plausible biological mechanism. While intensive lifestyle modifications (Ornish, Esselstyn) demonstrated angiographic stability or modest regression, modern prospective trials confirm that lifestyle rarely achieves the extreme ApoB targets (<40 mg/dL) required to reliably induce volumetric regression across large populations without lipid-lowering pharmacotherapy (Leenders et al., 2023). Level C Plausible
1 Like