How to Read a Cancer Survival Statistic

A cancer survival statistic tells you what fraction of people diagnosed with a cancer are still alive after a set period, usually five years. It does not tell you how many were cured, how many were helped by treatment, or whether catching a cancer earlier changed anyone's fate. Five-year relative survival and net survival are careful attempts to strip out deaths from unrelated causes, yet even these refined measures can rise while the number of people dying stays exactly the same. That gap between what a survival number says and what people assume it means is the most important thing to understand before you trust one.

What the number actually counts

Start with the plain version. Five-year survival is the share of patients alive five years after diagnosis. If 100 people are diagnosed and 70 are alive at year five, five-year survival is 70 percent. Simple, but incomplete, because some of those deaths and survivals have nothing to do with the cancer.

Two adjustments try to fix that. Relative survival compares the observed survival of a cancer group to the expected survival of a similar group from the general population matched on age, sex, and calendar year. If a cancer group survives at 70 percent while a matched cancer-free group would have survived at 90 percent, relative survival is roughly 78 percent, the ratio of observed to expected. Net survival asks a sharper hypothetical: what survival would be if the cancer were the only possible cause of death. A 2022 scoping review in the journal Cancers describes the Pohar Perme estimator, proposed in 2012, as the method that became the reference standard for net survival, precisely because it removes background mortality in a way that lets populations with different underlying death rates be compared fairly.

These are genuine improvements. They let you compare a young population against an older one, or one country against another, without a difference in ordinary mortality masquerading as a difference in cancer outcomes. But they share one blind spot: they measure time from diagnosis, and diagnosis is exactly the moment that screening moves.

Lead-time bias: the clock starts earlier, the ending is the same

Imagine a cancer that will kill someone at age 70 no matter what. If it is found because of symptoms at 67, that person has a three-year survival and counts as a death within five years. Now imagine the same cancer, in the same person, found by a screening test at 60. The person still dies at 70. But now they survived ten years from diagnosis and pass the five-year mark as a success.

The National Cancer Institute uses almost exactly this example and draws a blunt conclusion: in its version, the person "does not live even a second longer." Nothing about the disease changed. Only the starting line moved. This is lead-time bias, and it is inherent in any comparison based on survival. Earlier diagnosis mechanically inflates survival statistics whether or not earlier diagnosis helps.

Length-time bias and overdiagnosis: screening finds the slow ones

The second distortion is subtler. Cancers grow at wildly different speeds. Fast, aggressive tumors spend little time in a detectable but symptomless state, so a periodic screening test is unlikely to catch them in that window. Slow, indolent tumors linger in that state for years, so screening catches them preferentially. The pool of screen-detected cancers is therefore enriched for the least dangerous ones, which have good survival regardless of treatment. That is length-time bias.

Its extreme form is overdiagnosis: finding a cancer that would never have caused symptoms or death in a person's lifetime. Every overdiagnosed case is, by definition, a five-year survivor, so each one pushes the survival statistic up while helping no one. The National Cancer Institute cites estimates that around 19 percent of screen-detected breast cancers, and somewhere between 20 and 50 percent of screen-detected prostate cancers, are overdiagnosed. Add enough of these harmless cases to the denominator and survival can climb, in the institute's illustration, from 40 percent to 80 percent while the exact same number of people die.

What survival cannot tell you, and what can

Put the two biases together and you get the core warning. A rise in five-year survival can reflect real therapeutic progress, or it can reflect nothing but an earlier clock and a fuller pool of harmless cases. Survival alone cannot distinguish these. The measure that can is mortality: deaths from the cancer per unit of population over time, across everyone, screened and unscreened alike. Mortality does not care when the clock started, and overdiagnosed cases do not lower it. This is why the National Cancer Institute treats a reduction in cancer deaths in a randomized trial, not a rise in survival, as the reliable way to judge whether screening reduces deaths.

The size of any benefit deserves the same scrutiny. The National Lung Screening Trial found that low-dose CT screening produced a 20.3 percent relative reduction in lung-cancer mortality compared with chest radiography. That relative figure sounds dramatic until you see the rates behind it: roughly 247 lung-cancer deaths per 100,000 person-years in the CT group versus about 309 in the radiography group. Both numbers describe the same result. Relative reductions look large; the underlying rates tell you how much the absolute risk actually moved.

None of this means survival statistics are worthless or that screening does not work. A 2014 analysis in PLoS ONE by Maruvka, Tang, and Michor argues that the opposite trap also exists: dismissing every survival gain as an artifact can hide real progress, because rising incidence and changing detection can obscure genuine treatment improvements unless you account for them. That paper concluded that increases in five-year survival, once normalized against incidence, largely reflected real advances in cancer care rather than lead-time bias alone. Their point and the institute's point are two halves of the same lesson. A survival number is a starting question, not an answer. Before you accept one, ask which measure it is, whether the comparison groups were screened differently, and whether anyone has shown a matching drop in mortality.

This article is educational and is not medical advice; decisions about screening should be made with a qualified clinician who knows your history.