213 ‒ Liquid biopsies and cancer detection | Max Diehn, M.D. Ph.D.

KEYPOINTS

• 01

  Liquid biopsies began with a clinical frustration.

  Diehn, a Stanford radiation oncologist, treats many early-stage lung cancers that haven't spread; targeted high-dose radiation cures most, but about 20–25% recur. Imaging can't tell after treatment who still harbours microscopic disease, and by the time a tumour is large on a scan — at 10×10×10 cm the tumour burden itself is fatal — options are worse. He wanted a blood test that could see cancer earlier.

• 02

  Older blood markers are limited by normal-tissue production.

  PSA, CEA and CA 19-9 have long been used, but normal cells make them too — normal prostate cells make prostate-specific antigen — so they lack specificity for screening.

• 03

  Sensitivity and specificity must be read together, with prevalence.

  Sensitivity is how many people with cancer test positive; a mammogram might be about 85% sensitive, missing 15 of 100 cancers. Calling everyone positive yields 100% sensitivity but useless specificity. Attia stresses that in a low-prevalence population, positive predictive value can be modest even for a test that sounds great.

• 04

  Low-dose CT is a proven lung-cancer screen.

  Early chest X-ray studies showed no benefit, but the landmark National Lung Screening Trial randomized high-risk smokers to low-dose CT or chest X-ray and found about a 20% relative reduction in lung-cancer deaths. With a single-digit absolute benefit (say 5%), the number needed to screen is around 20. False positives were a concern in the original protocol; modern protocols reduce them. Radiation is low relative to the regulatory 50 mSv annual limit.

• 05

  The field moved from circulating tumour cells to cell-free DNA.

  Early-2000s work focused on intact tumour cells in blood, filtered using markers like CD45 (a pan-white-cell marker). Attention shifted to cell-free DNA: double-stranded fragments of about 170 bases. Healthy people have roughly 1–5 ng/mL in plasma, so 20 mL of blood yields about 10 mL of plasma and 10–50 ng of DNA.

• 06

  Tumour DNA is a tiny fraction of the signal in early cancer.

  In early-stage patients circulating tumour DNA is often well below 0.01% — sometimes 0.001% — of cell-free DNA. About 99.9% of fragments map to normal germline sequence.

• 07

  Tracking many mutations at once multiplies sensitivity.

  If a patient's tumour has been sequenced, looking for 10 of its mutations instead of one raises sensitivity about tenfold, since you don't need to see all of them. Prior methods bottomed out near 0.01% (one in 10,000); Diehn's lab tackled this from around 2014, and newer methods go further.

• 08

  Filtering out white-blood-cell mutations is essential.

  About 80–90% of cell-free DNA comes from white blood cells. With age, leukocytes acquire mutations through clonal hematopoiesis, which then appear in plasma. Diehn's team sequences both cell-free and leukocyte DNA to subtract them. Cancer-derived fragments are also a few bases shorter on average.

• 09

  Minimal residual disease detection is the near-term win.

  After treatment, ultra-sensitive tests can detect microscopic residual disease with a very high positive predictive value: detectable ctDNA means the cancer will very likely return. This is easier than screening because the tumour's mutations are already known.

• 10

  Pan-cancer screening is far harder.

  Without a known tumour, tests like Grail's look at methylation patterns of cell-free DNA. Specificity is above 99%, but presented stage I lung-cancer sensitivity is about 5% or less — and stage I is exactly where blood tests matter, since larger tumours show on CT.

• 11

  Diehn's target for a breakthrough.

  He argues 50–75% sensitivity for true stage I cancers while keeping specificity above 99% would be a big step. Dropping specificity to 80% inflates sensitivity by luck — a trade some groups make.

• 12

  Cell-free RNA could add tissue-of-origin information.

  Diehn says cell-free RNA exists in plasma and is a much more nascent field than cell-free DNA. Because RNA reflects which genes are active, it could in principle reveal what tissue a signal comes from and add information beyond mutations or methylation. His long-term vision is a single blood draw that detects cancer and points to where it is.