|
|
|
|
|
The Burden
of Caution |
|
Author: Bruce Lanphear, MD, MPH Editor's Note: his article was originally published on May 19, 2026 by Plagues, Pollution & Poverty on Substack. History’s deadliest exposures were once considered acceptable risks. There is a peculiar dialect in epidemiology. Once you notice it, you can’t unhear it. “These findings suggest that exposure to X may possibly increase the risk of Y.” Suggest. May. Possibly. Risk. By the time the sentence reaches its period, the meaning has nearly evaporated. Epidemiologists are not the only offenders. Scientists everywhere hedge, and some caution is necessary. The world is full of people selling certainty—wellness gurus promising miracle cures, tech investors convinced AI will solve human loneliness, biotech companies declaring breakthroughs before the experiments are barely finished. Skepticism has its place. Science moves forward because its claims are tested, questioned, and sometimes undone. But epidemiology has drifted into a different problem. We have become so afraid of overstating conclusions that we often understate them beyond recognition. And that matters because epidemiologists do not merely generate evidence. We also shape how evidence is understood, debated, delayed, regulated, or ignored. The Long Education in Caution The training begins early. Students are taught—correctly—that association does not prove causation. Confounding lurks everywhere. Bias exists in every dataset. Replication matters. One study proves little. Young researchers quickly learn that reviewers rarely criticize excessive restraint. They criticize boldness. A paper that concludes, “The evidence indicates that lead exposure causes cardiovascular disease,” will attract more scrutiny than one that says, “The findings suggest lead exposure may be associated with cardiovascular outcomes.” So scientists adapt. We layer conclusions with qualifiers—suggest; may; could; possibly; consistent with—partly out of caution and partly because bold conclusions attract scrutiny. Over time, the meaning can become so diluted that the conclusion loses much of its force. Meanwhile, the public and many of our colleagues read this language literally. If we keep saying “may,” people reasonably conclude: “Ah. They’re not really sure.” Industry lawyers understand this perfectly. They have spent decades weaponizing scientific caution against science itself. The Tobacco Playbook The pattern is familiar. Early evidence linking smoking to lung cancer was observational, imperfect, and fiercely contested. Tobacco companies amplified every uncertainty while scientists responded with caution. Yet the evidence accumulated: dose-response relationships, biologic plausibility, characteristic DNA mutations, and consistent findings across populations. Eventually the causal link became undeniable. The same cycle has repeated with asbestos, air pollution, endocrine disrupting chemicals, and climate change: evidence strengthens while scientific language remains tentative long after the broader conclusion is already clear. When Caution Obscures Meaning Environmental epidemiology is especially vulnerable to this problem because randomized trials are often impossible or unethical. We cannot randomly assign pregnant women to be dosed with pesticides or children to lead. Instead, we rely on observational studies, mechanistic evidence, animal experiments, biomonitoring, and triangulation across multiple lines of evidence. That is not a flaw. It is how population science usually works. As epidemiologist Miguel Hernán has argued, treating “causal” as a dirty word can itself distort science. The goal of much observational research is causal inference, even when randomized trials are impossible or unethical. No one demanded randomized trials to conclude that cholera spread through contaminated water or that asbestos causes mesothelioma. Public health has often advanced by assembling converging evidence under imperfect conditions. We do not require randomized trials of parachutes before concluding they prevent death when people jump from airplanes. Yet epidemiologists sometimes communicate as though uncertainty itself were the central finding. A curious thing happens when every conclusion is wrapped in layers of qualification: people begin to wonder whether scientists know anything at all. And in fairness, part of this problem belongs to us. We often complain that policymakers ignore evidence while failing to acknowledge how our own writing can contribute to delay. Scientific caution is important. But our conclusions can become so restrained and qualified that they obscure the weight of the evidence—and delay has consequences. The Difference Between Humility and Evasion Humility is essential in science. Nature routinely humiliates people who think they have solved everything. But humility is not the same as evasiveness. A humble scientist says: “This is what the evidence currently indicates, and here are its limitations.” An evasive scientist says: “The findings suggest there may possibly be a potential association.” One clarifies uncertainty. The other obscures meaning. Sir Austin Bradford Hill understood this tension decades ago when he wrote: “All scientific work is incomplete… liable to be upset or modified by advancing knowledge.” But Hill immediately added something equally important: “That does not confer upon us a freedom to ignore the knowledge we already have.” The Seduction of Absolute Proof Part of the problem is that society increasingly expects science to deliver certainty when science mostly delivers probability. People want a smoking gun. Epidemiology often offers something messier: patterns, gradients, convergence, consistency. But that is—or should be—enough. When dozens of studies point in the same direction, when animal studies align with human evidence, when biologic mechanisms make sense, and when exposure reduction improves outcomes, demanding “absolute proof” becomes less a scientific standard than a strategy for permanent delay. Industries understand this well. They do not need to disprove epidemiology. They merely need to preserve uncertainty long enough to continue business as usual. Regulators face a related tension. Acting too early risks criticism for overreach; acting too late rarely carries the same immediate penalty. The result is a system that often treats uncertainty as a reason to delay action rather than a reason for caution about continued exposure. Scientists, often unintentionally, can reinforce this dynamic by speaking as though every conclusion were forever provisional. Climate science offers a revealing parallel. For years, scientists described climate change in highly cautious language—projected risks, possible impacts, uncertainty ranges—even as the underlying evidence grew increasingly strong. Over time, many climate scientists began speaking more directly, not because they had abandoned rigor, but because the evidence had become overwhelming and because excessive restraint was itself contributing to public misunderstanding. Environmental epidemiology now faces a similar challenge. Saying What We Mean None of this means epidemiologists should become crusaders or abandon rigor. The world has enough people making grand claims with fragile evidence. But we can communicate more clearly without sacrificing honesty. Sometimes “the evidence indicates” is more accurate than “the findings suggest.” Sometimes “consistent with a causal relationship” is more truthful than “may possibly increase risk.” And sometimes, after decades of evidence, we should simply say: this exposure causes harm. Not because science is infallible, but because endless hesitation carries its own risks. Epidemiologists help shape not only scientific understanding, but political will, regulatory action, and public perception. We should choose our words carefully. But we should also occasionally choose them bravely. ■ |
|
|
