Beyond the Established Sites: Reading the Wider Cancer Signals

A landmark pooled analysis asked whether leisure-time activity was associated with incidence across 26 cancer types. It found inverse associations for 13, but a large study does not make every result equally causal, precise, or useful for an individual. The wider map needs careful reading.

What a 26-site study can and cannot do

Moore and colleagues combined data from 12 prospective cohorts in the United States and Europe. At baseline, participants reported leisure-time physical activity; researchers followed them for cancer incidence. The pooled sample included 1.44 million adults, with a median age of 59, and 186,932 new cancers over a median 11 years. The large sample made it possible to examine more uncommon cancer sites than a single cohort could handle.

The study compared activity at the 90th percentile with activity at the 10th percentile within each cohort. Because those percentiles were cohort-specific, “high activity” did not mean one shared dose such as a particular number of minutes per week. A percentile contrast asks whether people toward the active end of each study had different incidence from those toward the less-active end; it does not define an exercise prescription.

Thirteen cancer types had statistically inverse associations, two had higher incidence estimates, and 11 did not show a statistically clear association in that analysis. “Not statistically clear” does not prove no effect; the study may lack precision for a smaller association. Likewise, a statistically significant result is not automatically causal. Each finding has to be weighed against evidence from other cohorts, exposure measurement, plausible confounding, and independent reviews.

Signals across sites

Among the inverse associations were esophageal adenocarcinoma, liver, lung, kidney, gastric cardia, endometrial, myeloid leukemia, myeloma, colon, head and neck, rectal, bladder, and breast cancer. Some estimates were larger than others, but size alone does not tell us which findings are most credible. The WCRF/AICR and US Physical Activity Guidelines Advisory Committee have separately graded evidence across sites, and their evidence categories are more informative than treating all 13 associations as equivalent.

In the Moore analysis, malignant melanoma and prostate cancer had small positive associations with higher activity. The melanoma comparison was HR 1.27 (95% CI, 1.16–1.40); prostate was HR 1.05 (95% CI, 1.03–1.08). These associations should not be read as evidence that exercise causes either cancer. Outdoor activity may correlate with ultraviolet exposure, and prostate cancer detection can vary with healthcare use. The cohort design cannot fully isolate those possibilities.

The overall total-cancer HR of 0.93 is a relative hazard comparison, not a seven-percentage-point difference in absolute risk. It also averages across cancer types that differ in biology, baseline incidence, detection, and established causes. An overall association can be useful for population research but cannot replace site-specific prevention evidence.

Evidence grades are not a popularity contest

WCRF/AICR graded physical activity as convincing evidence for lower colon-cancer risk and probable evidence for lower postmenopausal breast and endometrial risk. It also reported probable evidence for vigorous activity and breast cancer before and after menopause. For several other sites, evidence remained limited or suggestive. Its grades reflect more than whether one pooled study returned a p-value: they consider consistency, study design, dose-response, precision, and biological plausibility.

The ACS and the 2018 Physical Activity Guidelines Advisory Committee discuss a broader set of sites with strong or moderate evidence, including kidney, bladder, gastric cardia, and esophageal adenocarcinoma. Review groups can differ in scope and grading frameworks. That is not a reason to select whichever label sounds strongest; readers should note the review, its year, and the specific cancer site.

In a 26-outcome analysis, some positive results are expected by chance even if all true effects are null. Moore et al. estimated a false-discovery rate of 7% across the tested sites. That statistical check lowers concern that chance alone accounts for every inverse finding, but it does not remove confounding or validate each site as a causal effect. Replication and independent evidence grading still matter.

Incidence is not mortality

This analysis studied cancer incidence: new diagnoses during follow-up. It did not establish that activity reduced deaths from each of the same 26 cancers. Incidence may change because of disease occurrence, screening, diagnosis, or detection intensity. Mortality depends on incidence plus stage, treatment, tumor biology, and survival. Those outcomes must not be swapped in summaries or charts.

There is separate research on activity after a cancer diagnosis and on survival among people living with cancer. That evidence asks different questions and involves different populations. A prediagnosis activity association cannot be presented as proof that exercise treats cancer, prevents recurrence, or substitutes for care. Treatment and survivorship decisions are clinician territory.

It is also possible for a behavior to be associated with lower incidence of one cancer and show no association—or a higher association—for another site. Cancer is a group of diseases, not one outcome. A statement such as “exercise prevents cancer” erases the site-specific variation that the evidence actually shows.

Why not every signal deserves equal emphasis

First, the 26-site analysis asked many questions at once. Multiple comparisons raise the chance that some associations will appear statistically persuasive even when the underlying relationship is weak. Moore et al. estimated a 7% false-discovery rate across tested sites, which is a helpful correction but not a certificate of causality for each result. The number should be read as part of the study's analysis, not as a promise that exactly one or two findings are false.

Second, confounding differs by site. Smoking is strongly related to lung cancer and may also relate to activity patterns; Moore et al. reported that smoking status modified the lung association. Outdoor exercise can increase sun exposure, which complicates melanoma findings. Prostate testing and diagnosis may differ according to healthcare use. Adjustment and subgroup checks can help, but residual differences may remain.

Third, strength of evidence depends on replication and how much supporting evidence exists beyond a single pooled dataset. The WCRF/AICR review graded colon evidence as convincing, and postmenopausal breast and endometrial evidence as probable. For other sites, its assessments were more limited or suggestive. That broader synthesis is a better guide to certainty than a simple list of statistically significant outcomes.

Finally, cancer sites are not interchangeable. A relative risk for liver cancer does not tell us the likely absolute benefit in a population with a different baseline rate. An association for colon cancer cannot be transferred to rectal cancer, and an estimate for overall breast cancer does not automatically apply to every molecular subtype. Keep each result tied to its exact site and population.

How to use the wider map

For practical decisions, prioritize broad activity guidance and avoid chasing individual site percentages. Adults can use the ACS target of 150–300 minutes of moderate activity or 75–150 minutes of vigorous activity weekly, or an equivalent combination. These recommendations support several health outcomes; they are not a cancer-specific dose formula. Start where you are, build gradually, and adapt the activity to your health and preferences.

Risk is shaped by multiple factors, including age, inherited susceptibility, tobacco, alcohol, body fatness, infections, environmental exposures, and chance. The relative contribution differs by cancer site. Exercise may contribute to lower risk for some cancers, but it does not make a person immune and it cannot erase other known risk factors.

Screening is a separate prevention and early-detection tool. Follow current screening recommendations that apply to your age and risk profile. New symptoms deserve assessment even in someone who exercises regularly. This page is not a screening guide and does not reproduce the site's separate screening toolkit.

Questions, answered briefly

• 🧮 Does 13 of 26 mean half of cancers are preventable through exercise? No. It means 13 sites had inverse observational associations in one analysis. It says nothing about the share of cases that activity would prevent if changed.
• 🔬 Should melanoma or prostate findings make people avoid exercise? No. The higher associations were observational and may reflect residual factors such as sun exposure or detection. They are not evidence to avoid activity.
• 📉 Is HR 0.93 a seven-point decrease in risk? No. It is a relative hazard comparison at cohort-specific activity percentiles. Absolute risk depends on the population and follow-up period.
• 🛡️ Does activity replace screening? No. Incidence associations and early detection are different; follow screening guidance and seek care for symptoms.

Related Topics