Why Earlier Detection Can Flatter Survival Figures

Survival statistics are widely used to argue for screening, and they contain a structural distortion. Finding a disease earlier lengthens survival as it is measured, whether or not the person lives any longer.

Survival is measured from diagnosis

Five-year survival counts the proportion of people alive five years after diagnosis, which makes the diagnosis date the starting line.

Screening moves that starting line earlier, since the whole purpose is to find disease before it produces symptoms that would have prompted testing.

If the date of death is unchanged, moving the start earlier automatically increases the interval between the two points. The number rises without anything about the disease changing.

This is called lead time

The extra interval created purely by earlier detection is lead time, and it is added to every screen-detected case regardless of outcome.

A person who would have been diagnosed at symptom onset and lived a further three years now appears to survive six after a screening diagnosis three years earlier.

The apparent improvement is an artefact of measurement. It is real in the sense that the person knows about the disease for longer, which is not the same as living longer.

Slow-growing disease is over-represented

Screening catches disease during the window before symptoms appear, and slow-growing disease spends far longer in that window than aggressive disease does.

Aggressive cases are more likely to develop and present between screening rounds, so they turn up as symptomatic diagnoses rather than screen-detected ones.

The screen-detected group is therefore enriched with slower disease that would have had better outcomes anyway, which flatters the figures a second time.

Overdiagnosis adds a third distortion

Some detected disease would never have caused symptoms at all during a person's life, and those cases are counted as diagnoses that were survived.

Adding cases that were never going to cause death to the denominator and the survivor count raises the survival rate mechanically.

This is why survival can improve markedly after a screening programme starts while the number of deaths from the disease remains unchanged.

What figure avoids the problem

Mortality rates in a whole population, counted per head rather than per diagnosis, are not affected by when a diagnosis was made.

Randomised trials comparing invited and uninvited groups measure the same thing directly, which is why programme decisions rest on them rather than on survival statistics.

None of this means screening does not work. It means survival percentages are the wrong instrument for judging whether it does, and they are the figure most often quoted.