By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
Peter Attia MD213 ‒ Liquid biopsies and cancer detection | Max Diehn, M.D. Ph.D.
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
CHAPTERS
- 0:02 – 10:55
Training with Pat Brown and DNA microarrays
Peter Attia and Max Diehn reconnect from their Stanford MD-PhD days. Diehn explains choosing Pat Brown's lab for DNA microarray technology and describes his T cell and RNA-related dissertation work, including the challenge of keeping RNA intact.
- •Diehn chose Pat Brown's lab because microarrays could measure expression of nearly all genes at once
- •His dissertation covered immunology and cancer biology, including a catalog of genes turned on in activated T cells
- •RNA is chemically unstable, so samples have to be stabilized, for example by working in a cold room
- •The technique measured only protein-coding transcripts chosen ahead of time, not non-coding RNAs
- 10:55 – 14:08
Choosing oncology and radiation oncology
Diehn explains that his father's lymphoma diagnosis during college set him on an oncology path. He then describes weighing surgical, medical and radiation oncology and why radiation oncology won out.
- •His father's lymphoma convinced him to be on the doctor side and to improve treatments
- •He considered surgical oncology but thought it would be harder to balance with research
- •A rotation prompted by his wife's research year introduced him to radiation oncology
- •He liked the extra time with patients, the technology, and the lack of molecular lab work in the field
- 14:08 – 18:50
Residency, research track, and Stanford's culture
Diehn describes combining residency with about two years of postdoc research through the Holman Pathway. He also explains why Stanford's radiation oncology department, founded by Henry Kaplan, valued laboratory research.
- •Residency was an internship plus four years of radiation oncology, with part of that time traded for research
- •The research track gave him a preview of life as a faculty member, with a half day to a day of clinic per week
- •Henry Kaplan's work curing Hodgkin's disease put the department on the map and brought early hype
- •Existing lab-based faculty and mentors made the department a good fit
- 18:50 – 26:47
The clinical frustration behind liquid biopsies
Diehn says his liquid biopsy work started from a clinical frustration: after treating early-stage lung cancer he could not tell who would recur. He and Attia discuss the roughly billion-cell limit of CT imaging and the blind spot for micrometastatic disease.
- •Diehn follows clinical needs and the data rather than a preset research area
- •About 20-25% of early-stage lung cancer patients recur, and follow-up was reactive scanning
- •CT can barely see tumors around 8 mm to 1 cm, roughly a billion cells
- •Micrometastatic deposits across the body could add up to a detectable blood signal
- 26:47 – 34:50
Protein biomarkers, sensitivity and specificity
Attia asks about PSA, CEA and CA 19-9, and Diehn explains that protein biomarkers lack specificity because normal cells make them too. The two then define sensitivity and specificity and use a thought experiment to show the trade-off.
- •Protein biomarkers are made by normal cells as well as cancer cells, which limits specificity
- •Sensitivity is the true positive rate, and specificity is the equivalent for people without disease
- •Attia relays a thought experiment: telling 1,000 women they have breast cancer gives 100% sensitivity but a useless test
- •Reporting sensitivity without specificity is not meaningful
- 34:50 – 53:22
Lung cancer burden, risk factors, and low-dose CT
They review lung cancer as the leading cause of cancer death, with falling incidence and mortality, and discuss its risk factors beyond smoking. They then cover the National Lung Screening Trial and the benefits and limits of low-dose CT.
- •Lung cancer is the top cancer killer, though trends are improving through less smoking, better treatment and screening
- •Non-smoking risks include pollution, radon and genetics, and adenocarcinoma is most common in non-smokers
- •Pack-years and secondhand smoke are hard to quantify, and secondhand smoke does not qualify for screening
- •The NLST showed about a 20% relative reduction in lung cancer deaths with low-dose CT versus chest X-ray
- •Imaging decisions weigh benefit against hard-to-quantify radiation risk
- 53:22 – 1:06:43
Starting the lab work and reproducibility problems
Diehn describes funding early blood collection from his startup funds and finding that protein biomarkers did not hold up in his patients. He and Attia discuss why failed replications often go unpublished and how that biases the literature.
- •Blood was collected before and after treatment, funded by startup money because grant agencies want proof first
- •Protein markers like CEA and CA 19-9 could not be reproduced as reported
- •Proving a result does not replicate requires large studies that cost time and money
- •Selection bias favors positive findings, so reproducing positive results matters
- 1:06:43 – 1:12:26
Circulating tumor cells and why Diehn moved on
Diehn explains why circulating tumor cells proved impractical: they are rare, require immediate processing, and also appear in healthy people. In early-stage lung cancer most patients had zero. Attia notes the approach might still help guide adjuvant therapy.
- •Usually 10-20 mL of blood is drawn, and in stage 1-2 lung cancer the CTC count was most often zero
- •Samples must be processed within hours, which prevents building frozen biobanks
- •Healthy people can have epithelial-looking cells with normal genomes, which hurts specificity
- •High CTC levels predict recurrence but would miss most patients who recur
- 1:12:26 – 1:24:44
Cell-free DNA biology
Diehn explains cell-free DNA: short, histone-protected fragments of about 170 bases in plasma at very low concentrations. The conversation covers where it comes from, how it is quantified, and a tangent on Attia's blood.
- •Prenatal diagnostics work on fetal DNA in maternal blood inspired the move to cancer
- •Healthy people have about 1-5 ng/mL of cell-free DNA in plasma, and the fragments are about 165-170 bases
- •Histones protect the fragments, while DNases in blood clear DNA quickly
- •The release mechanism is poorly understood and is probably multifactorial, including apoptosis and necrosis
- •Quantification is usually done with a simple fluorescent dye method
- 1:24:44 – 1:34:59
Detecting circulating tumor DNA by its mutations
Diehn explains that cancer-specific mutations distinguish tumor DNA from normal cell-free DNA. Tumor DNA may be only 1% of the total in advanced disease and far less in early disease. Next-generation sequencing with multiple patient-specific mutations overcame the sensitivity problem.
- •Mutations are specific to cancer cells and track the cause of the disease
- •Tumor DNA is about 1% of cell-free DNA in advanced lung cancer and often below 0.01% in early disease
- •Sequencing the patient's tumor first identifies mutations to look for in the blood
- •Tracking dozens of mutations in parallel raises sensitivity versus tracking one
- •Truncal or clonal mutations present in every tumor cell are most useful
- 1:34:59 – 1:42:11
Cell-free RNA, ctDNA terminology, and methylation
Diehn describes cell-free RNA as an early-stage field that could reveal gene expression, such as PD-L1 or tumor type. He distinguishes cell-free DNA from ctDNA, then explains methylation signatures for tissue of origin and how their sensitivity compares to mutation-based methods.
- •Cell-free RNA could show which genes are on or off, which DNA mutations cannot
- •Cell-free DNA is the total, and circulating tumor DNA is the tumor-derived fraction
- •Methylation patterns indicate tissue of origin and are more stereotyped than mutations
- •Mutation-based methods are far more sensitive, with Diehn's third-generation method reaching one part in a million versus about 0.1% for methylation
- 1:42:11 – 1:51:27
Screening test performance, predictive value, and the need for trials
Attia and Diehn work through why a test with 50% sensitivity and 99%+ specificity is still weak for screening low-risk people. Diehn explains that stage I lung cancer is heterogeneous and argues that randomized trials showing cancer-specific survival benefit are needed.
- •Reported sensitivity across stages can be inflated by including many stage IV patients, and stage I lung sensitivity for current tests is about 5% or less
- •At 1% pretest probability, 50% sensitivity and 99.5% specificity give a positive predictive value around 40%
- •Stage I includes patients with and without hidden micrometastases, and tests may preferentially catch the former
- •Diehn says commercial efforts are not planning randomized trials like the low-dose CT study, and he thinks they are needed
- 1:51:27 – 1:59:15
Approved liquid biopsy tests and the regulatory landscape
Diehn explains what the FDA-approved tests like Guardant are designed for: identifying actionable mutations in advanced cancer. He also clarifies the difference between FDA approval and CLIA-based laboratory-developed tests such as GRAIL's.
- •Guardant and Foundation tests find mutations in metastatic cancer to guide targeted therapy, the easiest liquid biopsy problem
- •They are not designed for early detection or for checking cure after treatment
- •CellSearch, a CTC assay, was approved first but is little used since it gives no actionable information
- •CLIA-certified labs can offer tests without FDA review of each test, which is how GRAIL and others reach patients
- 1:59:15 – 2:10:49
Minimal residual disease and guiding adjuvant therapy
Attia describes the sci-fi goal of blood-based early detection. Diehn explains that the nearer step is minimal residual disease testing, now in trials at Stanford, and discusses which cancers it suits, current colorectal and breast progress, and avoiding overtreatment.
- •ctDNA after treatment has a very high positive predictive value for recurrence
- •A Stanford trial gives immunotherapy to early-stage lung patients who test positive for ctDNA
- •The approach fits cancers where only a minority recur, not pancreatic cancer
- •Colorectal cancer is furthest along, while breast and prostate are harder due to low ctDNA levels
- •Repeat testing could avoid overtreating patients who are already cured
- 2:10:49 – 2:26:07
Toward multi-modal cancer screening and its limits
Diehn describes Lung CLIP, a mutation-based machine learning screening method, and the prospect of combining mutations, methylation and fragment features. They discuss realistic sensitivity targets, practical uses of blood tests, funding growth, and Diehn's call for rigorous clinical proof.
- •Lung CLIP sequences plasma and white blood cells to subtract clonal hematopoiesis mutations, then uses machine learning on features like fragment length
- •The model reached about 20% stage I sensitivity at about 98% specificity
- •Attia estimates a test might need about 80% sensitivity and 99.5% specificity at 1% pretest probability
- •Diehn is hopeful for 50-75% but says it is unknown whether it is achievable
- •Near-term uses include cancers without screening and improving access, but not replacing better existing tests
- •Liquid biopsy research funding and interest have grown dramatically since 2012
