The Cause or Merely a Coincidence?

Students from dozens of countries and every continent filled seats in the lecture hall’s curved rows. Wide steps narrowed with each lower level in the amphitheater of learning – one of many at Harvard School of Public Health. Individual microphones at each seat and a pair of triple-decker chalk boards upfront gave every student a voice and our professor a podium of authority. His unparalleled reputation had filled nearly every seat. Our lesson for that day: how to distinguish cause from coincidence.

“In 1981, the New England Journal of Medicine published a study linking coffee consumption with pancreatic cancer,” our professor said. “The study was publicized in the New York Times but later proven wrong when other experiments could not replicate its findings.”

The hall was silent. Disproving findings in top journals is rare.

“What was the experiment’s mistake?” he asked. “How was cause confused with coincidence?” The flasks and mugs atop nearly every desk of the lecture hall caffeinating our minds demonstrated the question’s relevance.

“The bedrock principle of science is repeat experimentation,” his voice echoed.

When studies first reveal the possibility of new knowledge, they must be scrutinized through repetition. This means that scientific understanding evolves – changes – with time. New assertions either get overturned or reinforced depending on results when others repeat the experiment.

The professor explained, “These authors failed to adequately consider a hidden confounder – tobacco. Their mistake is our lesson topic today.”

In medicine, confounding can hide true causes of an illness or true impacts of a treatment. A confounder is anything that affects whatever is being studied and any second factor thought to affect it. Confounding is often subtle and can lead to conclusions being later proven false by repeat experimentation.

To explain confounding in every-day terms, consider wildfires. We know that the true relationship between wildfires and firetrucks is protective – of course – firetrucks extinguish fires. Yet, if we were unfamiliar observers flying high overhead and unaware of this relationship, we might observe a plume of smoke, see wildfire together with many emergency vehicles, and incorrectly connect the dots. All we would know is that fire and firetrucks are commonly found together. We might assume firetrucks cause fire. A confounder would be anything affecting both the fire and firetrucks that is the actual true underlying cause – in this example … firefighters. Only after watching firefighters get inside firetrucks to extinguish pre-existing fires, could we learn that, really, firefighters were increasing the number of firetrucks in order to decrease the number of fires. The firetrucks don’t do anything on their own.

Similarly, confounding led to the misattribution of coffee as a cause of pancreatic cancer. The flawed study interviewed 367 people with pancreatic cancer and 643 controls, looking for correlation to smoking, alcohol, and coffee. Unfamiliar with the true relationships, its authors observed that controls more frequently abstained from coffee compared to people with cancer (14% vs. 5%). Yet, controls also more often avoided smoking (36% vs. 28%). So, pancreatic cancer was less common for both non-coffee drinkers and non-smokers. They didn’t know why. Maybe both coffee and tobacco contributed to cancer? Maybe only tobacco contributed but smoking was more common in coffee drinkers? Maybe the reverse was true and only coffee contributed? Maybe neither did and another confounder was hiding unseen? Because the authors only categorized smoking broadly and did not quantitatively account for each patient’s aggregated lifetime smoking history, it was later determined that heavier smokers were more likely to get pancreatic cancer and also more likely to drink coffee. Repetition of the experiment clarified smoking as the cause and coffee as a mere coincidence.

As Nevada County residents navigate claims of ways to cure cancer, the standard-of-care medical options of surgery, radiation, chemotherapy, (and newly) immunotherapy are tried and true and vetted by hundreds of repeat experiments. Claims of easier or all-natural cures might hold promise, but nearly none have passed the test of repeat experimentation. If they had, they would be the standard of care.

Dr. Hess is the Medical Director of Radiation Oncology at Sierra Nevada Memorial Hospital. Dr. Hess’ views are his own and do not reflect official positions of CommonSpirit Health, Dignity Health, or Sierra Nevada Memorial Hospital. Some aspects of this article are fictionalized history but based on a true story. All names are fictitious to protect confidentiality.