The Preventable Causes of Death in the United StatesComparative Risk Assessment of Dietary, Lifestyle, and Metabolic Risk Factors

Death certificates tell us what killed someone — heart disease, cancer, or stroke — but they don't reveal why those diseases occurred. That gap is crucial for public health policy. If we want to prevent deaths, we need to identify which modifiable behaviors and conditions are actually driving them. This is the question Danaei and colleagues aimed to answer in a landmark comparative risk assessment of preventable mortality in the United States. Their ambition was specific: to take twelve modifiable dietary, lifestyle, and metabolic risk factors, apply one consistent framework, and estimate how many of the deaths in 2005 each factor was responsible for. The twelve risk factors included high blood pressure, high blood glucose, high LDL cholesterol, overweight and obesity, physical inactivity, tobacco smoking, alcohol use, high dietary salt, high trans fatty acids, and low intakes of omega-3 fatty acids, polyunsaturated fatty acids, and fruits and vegetables. One study, twelve factors, one coherent method. The method they used is called comparative risk assessment, and the central concept is the population-attributable fraction — essentially, the share of deaths that would be prevented if everyone's exposure to a given risk shifted from where it currently is to the lowest-risk level the evidence supports. That lowest-risk level is called the theoretical minimum risk exposure distribution. For tobacco, that means zero exposure. For blood pressure, it is a systolic reading of 115 millimeters of mercury. For body mass index, it is 21. For dietary salt, it means essentially eliminating excess sodium. These are not arbitrary targets — they come from the lowest mortality levels observed across epidemiological studies and low-exposure populations. To determine the relative risks linking each exposure to specific diseases, Danaei and colleagues drew from published systematic reviews and meta-analyses, or conducted new ones when necessary. Crucially, they adjusted for regression dilution bias — a statistical artifact where imprecise single measurements make a risk factor appear weaker than it is — specifically for blood pressure, LDL cholesterol, and blood glucose. The result is relative risks that are larger and more accurate. They then propagated all the uncertainty through five hundred Monte Carlo simulations, drawing from the distributions of both exposures and relative risks, multiplying that by National Center for Health Statistics disease-specific death counts, and reporting ninety-five percent confidence intervals from the full range of simulated results. Now, let’s look at the findings. Tobacco smoking caused an estimated four hundred sixty-seven thousand deaths in 2005 — about one in five adult deaths in the United States. High blood pressure caused three hundred ninety-five thousand, roughly one in six. These two risk factors stand apart from everything else in scale. Tobacco's toll was spread across cancers, cardiovascular disease, and respiratory disease. It caused one hundred ninety thousand cancer deaths alone — thirty-three percent of all cancer deaths that year — driven by relative risks as high as twenty-one for lung cancer in men and twelve point five in women. High blood pressure, meanwhile, was the dominant force in cardiovascular mortality, responsible for forty-five percent of all cardiovascular deaths. It was also the leading cause of death specifically in women, accounting for two hundred thirty-one thousand female deaths, while smoking remained the top killer in men. These aren't correlations dressed up as causation. The relative risks were derived from large cohort studies adjusted for confounders, and for blood pressure and LDL cholesterol, they were further validated by randomized trial data. The numbers represent the best available estimates of what would be prevented if these exposures were eliminated. Below that top tier sits a cluster of risks that each cause mortality in the hundreds of thousands but rarely appear on any death certificate. Overweight and obesity caused an estimated two hundred sixteen thousand deaths. Physical inactivity caused one hundred ninety-one thousand. Together, those two come close to the tobacco number — a fact worth pondering. Then come the dietary risks. High dietary salt was responsible for one hundred two thousand deaths. Low omega-3 fatty acid intake — meaning not enough seafood — caused eighty-four thousand. High trans fatty acid consumption caused eighty-two thousand. High LDL cholesterol added another one hundred thirteen thousand. Low fruit and vegetable intake contributed fifty-eight thousand. These numbers are striking mainly because none of those causes appear when someone dies of a heart attack. The death certificate states ischemic heart disease. The upstream cause — years of excess salt raising blood pressure or a diet chronically low in omega-3s — is invisible. Age patterns add another layer. About seventy percent of deaths attributable to physical inactivity and sixty-eight percent attributable to high salt occurred after age seventy. But for LDL cholesterol, overweight and obesity, trans fats, and low omega-3 intake, forty percent or more of attributable deaths struck people under seventy. These are not just conditions of old age. A significant fraction of this mortality impacts people in the middle decades of life. Alcohol deserves its own treatment because it is the one risk factor in this analysis where the data show both real harms and real benefits in the same framework. The protective effect on the cardiovascular system is genuine — Danaei and colleagues estimate that alcohol averted approximately twenty-six thousand deaths from ischemic heart disease, ischemic stroke, and diabetes. That's a meaningful number. But it does not survive the full accounting. Alcohol caused roughly ninety thousand deaths from injuries, violence, liver disease, cancers, hemorrhagic stroke, and alcohol-use disorders, leading to a net attributable death count of about sixty-four thousand. Seventy percent of those deaths were in men. Additionally, twenty-nine percent of the chronic-disease burden fell on heavy drinkers — men consuming more than sixty grams of alcohol per day and women consuming more than forty — a group that showed no cardiovascular benefit whatsoever. In light drinkers, the protective effects on heart disease and diabetes did outweigh some harms for certain chronic outcomes. However, the population-level calculus is clear: net harm, not net benefit. The alcohol analysis also illustrates the hardest methodological challenge in this type of work — the overlap problem. Risk factors don't operate in isolation. A person who is obese may also be sedentary, have high blood pressure, and eat too much salt. Attributing a death to any one of those factors involves real uncertainty because they partially cause each other and share disease outcomes. Danaei and colleagues addressed this by using relative risks adjusted for potential confounders and by running sensitivity analyses that incorporated correlations between risk factors. Incorporating those correlations changed estimated attributable deaths by between three and thirty-one percent depending on the risk. That's not trivial. The authors are explicit: the numbers for individual risk factors overlap and cannot simply be added together. If you tried to sum the twelve figures, you'd be double-counting deaths. So, what should we do with these numbers? The study was designed explicitly to inform health policy and resource allocation — to answer the question of where interventions would save the most lives. The answer from the data is clear: tobacco and high blood pressure are the dominant targets. Together, they account for more preventable deaths than any other combination of factors in this analysis. Effective interventions for both exist and are well understood. However, the findings carry important limits worth considering. The estimates assume current exposure distributions and project against a theoretical ideal, so they do not predict a timeline for benefit — hazardous exposures accumulate slowly, and mortality declines lag behind behavioral changes. Several relative risks, particularly for self-reported behaviors like physical activity, fruit and vegetable intake, and omega-3 consumption, were not corrected for measurement error, which may mean some estimates are conservative. Moreover, because many dietary and metabolic risks operate partly through the same intermediate pathways — for instance, obesity raising blood pressure, dietary patterns affecting LDL and glucose — there is real mediation between factors that this analysis can only partially disentangle. Ultimately, the paper shows that the leading causes of preventable death in the United States are not mysteries. They are quantifiable, modifiable, and the tools to address most of them are largely known. The challenge isn't knowledge; it's will. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.