0:00 – Intro & welcome
0:34 – Disclosures & background on the Keto-CTA study
1:01 – Study design & lean mass hyper-responder criteria
1:45 – CT scans, imaging methods & data acquisition
2:10 – Semi-quantitative vs quantitative analysis explained
3:06 – Timeline of analyses & initial expectations
3:29 – First Cleerly results & early concerns
4:11 – Shared findings: baseline plaque & lack of LDL/ApoB association
5:21 – Paper publication & critics’ reactions
6:07 – Prespecified methodology: why QAngio matters
6:59 – Social media debates & misinterpretations
7:33 – The “limited contrast” argument emerges
8:33 – Breaking down claims about “all high LDL”
9:42 – Visualizing the actual LDL spread
11:01 – LDL variability: 49 mg/dL to 591 mg/dL
12:30 – Problems with the “limited exposure” critique
13:08 – Revisiting the lipid hypothesis: dose-dependent effects
14:07 – Are these the “healthiest” participants? Context & caveats
15:14 – Participant risk factors: CAC scores, A1C, glucose, hs-CRP
16:33 – Why Dave reduced public debates for a time
17:01 – Issues discovered in Cleerly’s dataset
18:13 – Introducing HeartFlow: a blinded independent analysis
19:05 – Why a fourth analysis was critical
20:02 – Record-breaking multi-analysis study design
20:34 – Preliminary results warning
21:04 – Non-calcified plaque volume % change: conflicting findings
22:00 – Absolute vs relative change: why it matters
22:28 – Total plaque volume & percent atheroma volume
23:17 – How HeartFlow & QAngio compare to reference studies
24:40 – Top-level observations & agreements
25:07 – Plaque regression: the star of the show
26:14 – Direct verification of regression by Dr. Budoff
27:00 – Red flags in Cleerly’s data: no negative values
28:33 – The “noise floor” problem explained
30:28 – Why multiple analyses were necessary
31:15 – Every imaging study shows regression — except Cleerly
32:02 – Statistical implausibility of Cleerly’s results
32:37 – Final conclusions on regression evidence
33:29 – Does regression imply safety for lean mass hyper-responders?
34:11 – No association of LDL/ApoB with plaque presentation or progression
35:02 – What’s next: study extension & Triad study
35:38 – Reflecting on six years since Keto-CTA’s conception
36:32 – Acknowledgments & closing remarks

Summary

The speaker, an engineer and founder of the Citizen Science Foundation, presented a detailed overview of the ketoCTA study, which investigates the effects of a ketogenic diet on cholesterol and coronary plaque development. Initiated in 2015, the study followed 100 individuals who exhibited increased cholesterol levels after adopting a ketogenic diet, a phenomenon the speaker termed “lean mass hyper-responder” (LMHR). These participants were selected based on specific cholesterol criteria (LDL ≥190, HDL ≥60, triglycerides ≤80) and underwent coronary CT angiography (CTA) scans at baseline and after one year (totaling 200 scans).

The study employed multiple methods of analyzing the scans: semi-quantitative expert reviews, and various quantitative analyses using AI and algorithms, including platforms named Clearly, Qangio (the prespecified methodology), and HeartFlow. Early results revealed notable discrepancies between the different analyses, particularly with the Clearly platform, which showed no instances of plaque regression—a finding at odds with both the other analyses and established scientific knowledge that some plaque regression is expected even in healthy populations.

The speaker emphasized the broad variation in LDL cholesterol levels among study participants, ranging from very low to extremely high (including cases of homozygous familial hypercholesterolemia), countering criticisms that there was insufficient LDL variability (“limited contrast”) to assess LDL’s impact on plaque progression. Despite the large spread and high LDL levels, none of the analyses demonstrated an association between LDL cholesterol or ApoB levels and the progression or presence of coronary plaque in this cohort, challenging aspects of the lipid hypothesis within this specific group.

The speaker also addressed social media critiques and the challenges faced during the study, including delays and disagreements over data analysis and interpretation. Due to concerns over the Clearly analysis, a blinded quality control review was requested but not completed, leading the team to commission an additional independent analysis via HeartFlow. The HeartFlow and Qangio analyses showed close agreement and revealed instances of plaque regression in some participants, which was independently verified by expert readers.

The talk concluded with a balanced message: while some participants showed plaque regression despite high LDL levels, this does not imply that being a lean mass hyper-responder guarantees low cardiovascular risk. The cohort was generally healthier than average but included individuals with various risk factors and pre-existing conditions. The study’s findings are preliminary, with ongoing efforts to analyze further data and extend the study. The speaker expressed gratitude to the research team, contributors, and participants and highlighted the significance of this milestone in the journey that began nearly six years ago.

Highlights

• 🩺 Study of 100 keto diet adopters with elevated cholesterol tracked over one year using coronary CT angiography.
• 🧬 Multiple independent methods analyzed the same 200 scans, revealing discrepancies between analysis platforms.
• 🔍 No association found between LDL cholesterol/ApoB levels and plaque progression in this cohort.
• 📊 Study population demonstrated an exceptionally wide range of LDL cholesterol, including extreme familial hypercholesterolemia.
• 🔄 Evidence of plaque regression verified by multiple analyses and expert readers, contradicting one major analysis (Clearly).
• 🛑 Challenges with data quality and analysis led to commissioning additional independent review (HeartFlow).
• ⚠️ Regression does not guarantee low cardiovascular risk; cohort includes individuals with varying health statuses.

Key Insights

• 🧪 Multi-modal imaging analysis reveals complexity and variability in plaque progression assessment:
  The ketoCTA study uniquely employed four independent analyses on the same set of 200 coronary CT scans, including semi-quantitative expert reads and three AI-driven quantitative platforms (Clearly, Qangio, HeartFlow). This multi-layered approach highlighted how different methodologies can produce divergent results, particularly in measuring plaque regression or progression. It underscores the importance of using multiple analytical tools and blinded quality controls to validate imaging studies, especially in novel cohorts.

• 📉 Plaque regression occurs even among individuals with high LDL cholesterol:
  Contrary to the dominant lipid hypothesis narrative, multiple analyses in this study documented regression of coronary plaque in a subset of participants, including those with LDL levels comparable to familial hypercholesterolemia. This finding challenges the assumption that high LDL universally predicts plaque progression and cardiovascular risk, suggesting that other factors may modulate atherosclerosis in lean mass hyper-responders or ketogenic dieters.

• 📈 No correlation between LDL cholesterol/ApoB and plaque progression in this cohort:
  Despite the large variability of LDL levels (ranging from under 50 to nearly 600 mg/dL), the study found virtually no association between LDL or ApoB and either the baseline plaque burden or the change in plaque volume over one year. This lack of correlation, consistent across multiple analyses, suggests that LDL cholesterol may not be the sole or primary driver of atherosclerosis progression in this specific population, prompting a re-examination of the lipid hypothesis in diverse metabolic contexts.

• ⚖️ Wide LDL cholesterol variability refutes the “limited contrast” critique:
  Critics argued that the study’s participants all had uniformly high LDL, limiting the ability to detect dose-response relationships. However, the data showed a broad LDL distribution with a standard deviation of 84 mg/dL and a range from 49 to 591 mg/dL—arguably the widest spread in any prospective imaging study to date. This diversity provides a strong test of the LDL-plaque relationship, reinforcing the study’s conclusions.

• 🔍 Importance of absolute versus relative changes in plaque measurements:
  The speaker pointed out a critical methodological issue: relying solely on relative percent changes in plaque volume can be misleading. Absolute changes provide a clearer picture, better aligned across different analytical platforms. This distinction is crucial for interpreting imaging results and for understanding the biological significance of plaque dynamics in clinical research.

• 🛑 Data quality and transparency are vital, especially amid scientific controversy:
  The Clearly platform’s refusal to conduct a blinded quality control review after concerns were raised highlights challenges in data transparency and scientific rigor. The study’s response—to commission an additional independent analysis with HeartFlow—demonstrates a commitment to robust science and the value of independent validation, particularly in contentious areas like keto and lipid research.

• ⚠️ Plaque regression does not equate to cardiovascular safety:
  The study emphasizes that even with documented plaque regression, lean mass hyper-responders are not guaranteed protection from atherosclerotic cardiovascular disease. The cohort was healthier than average but included individuals with diabetes, obesity, hypertension, and high coronary artery calcium scores. This nuanced view cautions against oversimplified conclusions and underscores the need for ongoing research into risk stratification and mechanisms beyond cholesterol levels.

Conclusion

This presentation offers a comprehensive and transparent look at the ketoCTA study’s progress, challenges, and preliminary findings related to cholesterol, ketogenic diets, and coronary artery plaque. It highlights the complexity of interpreting advanced imaging data, the need for rigorous multi-method validation, and the evolving understanding of lipid biology in cardiovascular risk. The findings challenge conventional lipid dogma in this unique population, while also acknowledging limitations and the necessity for further research.