Methods

How medical evidence is actually built, and why it takes so long

The gap between a promising laboratory result and an accepted treatment is not bureaucracy. It is the process by which a claim earns the right to be believed.

9 minute readMedicine

A finding in a dish is a hypothesis about a living system. A finding in a mouse is a hypothesis about a human. Each step forward is a test of whether the previous result survives contact with a more complicated situation, and most results do not survive.

Why the earliest results mislead

Cultured cells are grown in conditions chosen for convenience. Constant temperature, abundant nutrients, no immune system, no circulation, no neighbouring tissue. Those conditions make experiments possible and they remove nearly everything that makes a body a body.

A compound that behaves well in that setting has demonstrated one thing only, that it can do something under ideal conditions. It has not shown that it reaches the right tissue in an organism, that it survives the liver, that it does nothing unwanted elsewhere, or that the effect persists beyond the length of the experiment.

The staged structure

Human trials are usually described in phases. Early phases involve small numbers of people and are concerned mainly with safety and dose. Later phases enlarge the group and shift the question toward whether the treatment works, measured against either an existing standard or an inactive comparison.

The comparison is the part that carries the weight. Without it there is no way to separate the effect of the treatment from the natural course of the condition, from the effect of being observed and cared for, or from the tendency of people to enter a study when their symptoms are at their worst and to improve afterwards regardless.

Randomisation and blinding

Assigning participants at random distributes unknown differences roughly evenly across groups, including differences nobody thought to measure. Blinding, where neither the participant nor the assessor knows the assignment, addresses the fact that expectation influences both how people report and how observers record.

These are not procedural formalities. They are the mechanisms that make the eventual number mean something. A result produced without them is not a weaker version of the same finding. It is a different and much less informative kind of statement.

Reading announcements carefully

Useful questions when a result is announced: how many people, compared against what, measured how, and over what period. An improvement in a laboratory marker is not the same as an improvement in how long or how well someone lives, and the two are frequently reported in the same sentence.

The slowness is often criticised, sometimes fairly. But the process exists because the alternative was tried for most of recorded history, and the results of confident practice without controlled comparison are documented at length in the history of medicine.