The Laws of Medicine

Siddhartha Mukherjee opens with an anecdote from his time as a medical student in Boston. He watched a senior surgeon, Dr. Castle, operate on a woman with a tumor in her lower intestine. When a substitute resident encountered unexpected bleeding, Castle asked what the resident would do if no one knew whether the patient had a bleeding disorder, then swiftly closed the vessels himself. The resident adapted and completed the surgery. Afterward, Castle offered a key insight: "It's easy to make perfect decisions with perfect information. Medicine asks you to make perfect decisions with imperfect information" (4). This tension between knowledge and uncertainty becomes the book's central concern.

Mukherjee recalls that his medical school curriculum, beginning in 1995, equipped him with facts but not with tools for navigating incomplete or ambiguous data. During his residency beginning in 2000, he read Lewis Thomas's essay collection The Youngest Science repeatedly. Thomas, a physician, scientist, and author, describes his internship at Boston City Hospital in the 1930s, when medical interventions had almost no measurable impact on illness. The ineffectiveness of pre-1930s treatments gave rise to "therapeutic nihilism," championed by figures like the physician William Osler at Johns Hopkins, who chose to observe and categorize diseases rather than treat them. This restraint proved productive: By the 1930s, careful observation had built foundational models of disease enabling doctors to understand conditions like heart failure and diabetes. Mukherjee contrasts the crude 1937 treatment for heart failure with the sophisticated arsenal available by the 1990s, then asks whether medicine qualifies as a true science governed by laws. He spent much of his residency searching for such laws, envisioning them as navigational rules for young doctors.

Mukherjee's first law states that a strong intuition is much more powerful than a weak test. He illustrates this with Mr. Carlton, a 56-year-old patient who presented with severe unexplained weight loss and fatigue. Four weeks of exhaustive testing yielded no diagnosis. Then one evening, Mukherjee saw Carlton conversing in the hospital lobby with a man previously admitted for a heroin-related skin infection. The incongruity triggered a realization: Carlton was likely using heroin. This inference explains the earlier difficulty drawing blood from scarred veins and raises the prior probability, the estimated likelihood of a diagnosis before testing, that Carlton had HIV. An HIV test confirmed this, and further testing established AIDS as the diagnosis.

Mukherjee reframes diagnostic challenges as probability games: Clinicians assign an initial probability that symptoms stem from a particular condition, then gather evidence to adjust that probability until it warrants a confirmatory test. Every test has false-positive and false-negative rates, so results can only be interpreted meaningfully in context. This insight traces back to Thomas Bayes, an 18th-century English clergyman whose most important work on probability was discovered posthumously. Bayes addressed the inverse of standard probability: inferring unknown causes from observed outcomes. Mukherjee applies Bayesian reasoning to contemporary controversies, including prostate-specific antigen (PSA) blood testing for prostate cancer, BRCA1 gene testing for inherited breast and ovarian cancer risk, and hypothetical Ebola airport screening. In each case, without adequate prior information about the tested population, even sophisticated tests produce misleading results.

Mukherjee's second law states that "normals" teach us rules, while "outliers" teach us laws. He opens with Tycho Brahe, the 16th-century Danish astronomer whose measurements supported a model of the cosmos that worked for every planet except Mars. Brahe assigned the problem of Mars's retrograde motion, its apparent backward movement across the sky, to his young assistant Johannes Kepler. Kepler treated the outlier as the most important data point and eventually discovered that all planetary orbits are ellipses, not circles. The exception became the key to formulating Kepler's Laws of planetary motion.

Mukherjee extends this principle to medicine through the history of autism research. In 1908, psychiatrists classified withdrawn children prone to repetitive behaviors as having a variant of schizophrenia. By the 1960s, a theory emerged that emotionally cold parents caused the condition, hardening into the "refrigerator mom" hypothesis. However, outlying data did not fit: The condition ran in families and showed a concordance rate, the frequency with which both twins share a condition, of 50 to 80 percent in identical twins. By 2012, genome-sequencing studies confirmed genetic rather than parental causes. Mukherjee argues that current medical models are incomplete hybrids that create the illusion of understanding until an outlier exposes their flaws. He introduces David Solit, a cancer scientist who in 2009 began studying "exceptional responders," rare patients who responded dramatically to otherwise failing drugs. In one trial, a 73-year-old woman's tumors nearly disappeared after treatment with a drug called everolimus that had failed all other participants. Solit sequenced her tumor's genes and identified mutations in TSC1 as a likely driver, transforming a single anomaly into a portal for gene-guided investigation. Drawing on Karl Popper's criterion of falsifiability, Mukherjee argues that outliers are the data that test and potentially overthrow existing medical models.

Mukherjee's third law states that for every perfect medical experiment, there is a perfect human bias. He discovered this during his oncology fellowship beginning in 2003, when he inherited patients on a trial of a drug described as a molecular cousin of Gleevec, a revolutionary cancer medication. The response rate among his patients appeared spectacular, but overall trial results revealed only a 15 percent response rate. Graduating fellows had transferred only drug-responding patients to incoming fellows and moved nonresponders to attending physicians, creating an unintentional but sharp distortion.

Mukherjee traces hope as a source of bias through the history of radical mastectomy. By the early 1900s, the surgeon William Halsted developed the procedure, removing the breast, underlying muscles, and deep lymph nodes to prevent relapse. His hypothesis was logically coherent but incorrect: In most cases, cancer had already spread before surgery. No randomized trial was launched until 1980. Led by the surgeon Bernie Fisher, the trial struggled to enroll patients because American surgeons were reluctant to question a procedure they trusted. The results showed no benefit over conservative surgery with radiation, and an estimated 100,000 to 500,000 women underwent the radical procedure unnecessarily between 1900 and 1985. Mukherjee also explores subtler biases in randomized controlled trials, noting that enrollees represent a self-selected subset whose results may not generalize. The best clinicians, he suggests, are bias hunters who understand when data applies to their patients and when it does not.

In closing, Mukherjee argues that uncertainty remains central to medicine despite technological progress. Lewis Thomas had predicted that high-precision instruments would map all bodily functions, leaving little uncertainty. Reality has diverged: Medicine has taken on vastly more complex projects, requiring deeper engagement with priors, outliers, and biases. Mukherjee illustrates with a frontier case: a six-year-old girl in Philadelphia with relapsed leukemia whose immune cells were genetically modified to attack her cancer. The treatment triggered a near-fatal inflammatory response driven by interleukin-6, an immune signaling molecule. Doctors administered a serendipitously available drug that blocked interleukin-6, and the girl's organs returned to normal function. The case involves all three laws: identifying which lab values drove the crisis, determining whether her response was normal or exceptional, and evaluating a therapy with no comparable precedent. Mukherjee acknowledges that his laws are personal, shaped by his own training. He suspects any future laws of medicine will also concern the nature of information and uncertainty. He characterizes medicine as a discipline still learning to reconcile pure knowledge with real knowledge, calling it the most human of the sciences.