181 - Viewing cancer through an evolutionary lens: a radically different approach to treatment

KEYPOINTS

• 01

  Cancer is an evolutionary problem, and treatment selects for resistance.

  Gatenby, a radiologist who trained in physics at Princeton, argues that oncology lacks the first principles that let physics make sense of messy data—as Newton did for planetary motion. His organizing principle is Darwinian: a tumor is a population of competing cells, and every therapy is a selective force that will eventually favor resistant variants unless it is managed strategically.

• 02

  Models don't need to predict ten years out to be useful.

  Attia compared cancer modeling to economic models that can't forecast GDP a decade ahead, and Gatenby agreed that cancer sits toward the chaotic end of the spectrum (Attia rated it an eight or nine out of ten). The goal is not long-range prediction but answering 'what should I do next?' with models that are recalibrated to each patient's data as treatment proceeds.

• 03

  The pesticide lesson: maximum dose clears the field for resistant cells.

  The diamondback moth has been hit with every pesticide developed and by the 1980s some populations resisted all of them. In Gatenby's illustration, a heavy spray kills the sensitive 80% and leaves the resistant 20% a whole field without competitors. Resistance mechanisms carry a fitness cost, so in a lower-pressure environment sensitive moths can out-compete resistant ones—potentially shrinking the resistant share over time.

• 04

  Adaptive therapy uses sensitive cells to suppress resistant ones.

  In metastatic prostate cancer, his team gave abiraterone until PSA fell to 50% of its baseline, then stopped, letting drug-sensitive cells regrow and compete with resistant ones before restarting. The idea is to keep a controllable tumor burden rather than eradicate sensitive cells and release resistant ones from competition.

• 05

  The pilot trial roughly doubled time to progression.

  Compared with patients on the standard maximum tolerated dose until progression, median time to progression went from about 14 months (consistent with the literature) to about 30 months with adaptive therapy, a 16-month difference—with less total drug. Some patients were still cycling about five years later. It was a small, non-randomized pilot, and models had predicted control would be lost somewhere between 2 and 20 cycles depending on the unknown starting fraction of resistant cells.

• 06

  The key surprise was a sevenfold fitness advantage.

  The trial data implied that sensitive cells were about seven times fitter than resistant cells off-drug, versus the two- to threefold the team had assumed. Instead of plateauing, resistant populations declined, and modeling suggested three successive well-timed cycles could drive resistant cells toward extinction. They think competition for space and substrate, scarce in poorly perfused tumors, drives the differential.

• 07

  Small populations are vulnerable—hit them with a sequence, not a single strike.

  Small populations can go extinct from minor shifts in birth and death rates (stochasticity), and tumors may show an Allee effect, where cells benefit from numbers, for example to build blood vessels. Gatenby favors an 'extinction' strategy borrowed from ecology: after a first strike shrinks the tumor, follow-up therapies attack the remnant before resistance emerges, as in pediatric leukemia regimens.

• 08

  Probability math explains why one drug rarely cures large tumors.

  A 100-gram tumor holds on the order of 100 billion cells, so some pre-existing resistant cells are near-certain. Knocking a population down to millions and then adding a second and third drug changes the odds dramatically. Attia's analogy: the US could survive a 90% population loss in theory, but a nation of 33 million under ongoing attack would be a very different, vulnerable entity.

• 09

  Immunotherapy may work best on minimal disease.

  Response to anti-PD-1 drugs is best in people with relevant mutations; first-generation interleukin-2 had perhaps 10–20% response in melanoma, but always in full-blown metastatic disease. Gatenby suggests immune approaches might perform very differently in patients with no visible disease, when the cancer population is small.

• 10

  Most cells that spread die; metastasis is about the soil.

  When cancer cells are injected into mice, nearly all die at distant sites. Around 30–40% of women undergoing mastectomy for apparently localized breast cancer may have cancer cells in their bone marrow, yet not all relapse. He and Attia discussed few cancers that kill without spreading, such as glioblastoma.