How do scientists find cancer risk factors?
Mostly by watching large groups of people over time. Researchers compare those who develop cancer with those who do not, looking at how the two groups differ in behavior or exposure.
The National Cancer Institute is careful about what that shows. Such studies, on their own, cannot prove that a behavior or substance causes cancer.
Two ways a finding can mislead
NCI names both. The result could be chance. Or the true risk factor could be something other than the one under suspicion.
There is a third problem the research summaries dwell on: measurement. NCI's health professional summary notes that observational studies rely on self-reported dietary assessments, which are prone to substantial measurement error. That error does not cancel out.
Findings like these often reach the news before they are settled, and NCI says this can lead to wrong ideas about how cancer starts and spreads.
What moves a link toward a cause
Two things, and NCI names them together.
The first is repetition. When many studies all point to a similar association, chance becomes a poorer explanation.
The second is mechanism. When a possible biological route exists that could explain how the factor causes cancer, scientists can be more confident about the relationship. Repetition plus mechanism is the standard NCI describes.
The strongest evidence comes from a randomized trial, where people are assigned by chance to change an exposure or not. Those are rare, because nobody can randomly assign a person to smoke for 30 years.
When trials contradict the observations
This is the part that gets skipped, and it is worth sitting with.
Population studies pointed to high-fiber diets to prevent colon growths, and the advice followed. Then a 2017 Cochrane review looked at randomized trials of supplemental fiber. It found a lack of evidence that more dietary fiber reduces the return of adenomatous polyps over a 2- to 8-year period. Adenomatous polyps are precancerous growths in the colon, and the review covered people with a history of them.
Red meat went the same way. Reviews of cohort studies and randomized trials found small or uncertain effects on overall cancer rates and deaths. The overall quality of that evidence was graded low to very low. A separate review found no clear link between total, animal, or plant protein intake and cancer deaths.
Meanwhile some findings hold up hard. NCI's summary says smoking is causally associated with 13 types of cancer, and that the associations are often much stronger than those seen for obesity.
The pattern that convinces epidemiologists
Newer cohort work on medical imaging shows what strong evidence looks like.
In a group of 10.9 million Australians, records documented CT scans in people aged 0 to 19. They were then followed into young adulthood. Compared with those who had no CT scan, those who had at least one were more likely to be diagnosed with cancer. The relative risk was 1.24, with a 95% confidence interval of 1.20 to 1.29. Average follow-up in the scanned group was 9.5 years.
The detail that matters most is the shape of the result. A statistically significant dose-response relationship was observed, with risk rising with each additional scan. More exposure, more risk, is among the most convincing signals in this field.
Check the date on the page
NCI's patient page on risk factors carries a last-updated date of December 23, 2015. The health professional Cancer Prevention Overview, which discusses the same evidence in far more detail, was updated on February 12, 2025.
So a useful question for any headline is a simple one: was this an observational study or a randomized trial, and has anyone repeated it? Our overview of cancer risk factors covers the best-supported ones, and alcohol and cancer risk shows what a well-replicated link looks like.
Sources
Want the full picture? Read our complete explanation: Understanding Cancer Risk Factors
Know someone who needs this?
Plenty of people are looking for something like this and do not know where to start. If this would help a friend or someone you love, send it on — we have written an opening line so you do not have to stare at an empty message. You can change every word of it.
Your message is written and sent in your own email or messaging app — we never see who you send it to, and nothing is added to any list.