What this covers
Peter Attia is interviewed by Patrick about how to reason when the evidence you would want does not exist. He contrasts Medicine 2.0, whose randomized controlled trials suit quick, lethal problems like heart attack, HIV, and infection, with longevity, which fits none of those criteria. His alternative is triangulation from three imperfect sources: observation of centenarians, non-human experiments, and thirty years of molecular findings. He also argues that "first, do no harm" is already routinely violated, so the real task is weighing risks and benefits probabilistically.
The range is wide. On evolution, he says it guides exercise, food, and sleep but not mindfulness or pharmacology, and he claims both paleo and vegan arguments are incomplete. The APOE-4 variant, which protects against parasitic brain infection but raises Alzheimer's risk, illustrates how selection favored reproduction over late life. He covers mTOR, autophagy, senescent cells, and parabiosis, and he is cautious about human translation. He asserts 7 to 9 hours of sleep is needed, and that genetics can trump lifestyle, since centenarians on average smoke more and exercise less. The sharpest skepticism goes to the microbiome: fecal transplant is approved only for recurrent C. difficile, and he calls the microbiome-analysis industry largely unfounded.
The practical thread is about targets and feedback. About 80% of deaths after age 40 come from four causes, and he walks through where car crashes kill. He stresses that the final decade often runs at half of functional capacity. Behavior change, he argues, needs fast feedback, as the two-second speech-delay experiment shows, and most longevity levers do not supply it.
Attia argues that longevity cannot be settled by clinical trials, so people should use evolutionary context, probabilistic risk-weighing, and real-time feedback to pursue healthspan, not lifespan alone.
- Longevity fails the criteria that make randomized trials work: the outcomes are slow, multifactorial, and not randomizable, so decisions must be probabilistic and empirical.
- Evolution is a useful guide for movement, sleep, and fasting, but not for mindfulness, pharmacology, or modern chronic diseases.
- Behavior changes stick when feedback is fast and gamified, and most longevity levers lack that feedback.
Simple daily habits capture most health benefit, so minimalist care without detailed biomarkers may be practical.
- Movement, diet, rest, mindfulness, and relationships are simple habits that capture roughly 85% of the health benefit, so a minimalist model without detailed biomarkers may be the practical answer.
“I would add relationships, would be the fifth... They're from my point of view, 85% of what you talked about today”