“It Isn’t Always Fun.” – Failure Breeds Success
No one who cannot rejoice in the discovery of his own mistakes deserves to be called a scholar.” Donald Foster admitting that his work to establish Shakespeare as the author of an obscure poem was wrong.
How one responds to failures and recovers from mistakes are signatures of a life, a career, and of character. Of course, it is easier to analyze where others fail and what they do about it than to see one’s own mishappenings. Here I consider some gaps in the thinking and some outright gaffs of Ancel Keys and what he did constructively about them.
Total fat-quality of fat
Keys held out too long that only total dietary fat, not the quality of fat, was important in causing elevated blood cholesterol and fatty arteries. He had focused only on total fat in the tentatives of his early cultural comparisons in Italy and Japan. When others demonstrated otherwise, he was stung. At the time he had no experimental data of his own. A mistake.
But then he set about to do the best studies ever done on the issue, using cross-over controlled trials of diet fatty acid substitutions for carbohydrates, keeping calories constant. In one of the first meta-analyses in medical science, he combined data from many of these experiments to derive the Keys Equation, which for almost 40 years has been the best formulation for predicting diet fat effects on blood cholesterol level.
Good and Bad Cholesterol
In a second example, Keys held out too long that only beta-lipoprotein in the serum was related to heart attack risk, based on his early ecologic findings in which beta and not alpha showed strong cultural differences parallel to disease rates, for example, between Japanese on the mainland and Japanese in California. Not having individual predictive data on alpha, or HDL, from the Minnesota and Seven Countries Studies, he fell victim, for the second time, of the “ecological fallacy,” which states that group data cannot be automatically applied to the individual. And so he joined with the common wisdom, the ill-fated consensus in the Cooperative Lipoprotein Study.1 There the majority rejected John Gofman’s prescience on the matter, derived from his ultracentrifugal partitioning of the blood lipoproteins into high and low density particles. And for too long thereafter, Keys also pooh-poohed the Framingham Study’s clincher in longitudinal data of individual “protection” from CHD by HDL level, at least as it applied in a U.S. culture having high average LDL levels and coronary rates.
It was then too late to obtain lipid fractions from the early Seven Countries surveys, which were based on dried filter paper samples of serum. Keys dropped out of the polemic on lipid fractions and exploited maximally his data in the landmark monograph, The Seven Countries Study. A Multivariate Analysis of Coronary Heart Disease Risk. The result was a greater understanding of the population causes of coronary disease as contrasted to the individual, which led in turn to the important idea of sick and well populations and to the population strategy of heart attack prevention.
Miss one boat, catch another.
Math and Arithmetic
Keys was among the first medical scientists to apply mathematics and statistics, using regressions from the time of his earliest researches. He was also the first to envision and mount a long-term cohort study on characteristics of coronary disease risk in men in the Minnesota Business and Professional Men’s Study. Uncharacteristically, however, he failed in the 1940s to do the math on sample size. The Framingham Study, with a more adequate sample, had risk data available in six years that Keys had to wait 21 years for in the businessmen. He was clearly bested in this crucial early issue of cohort study design.
Keys reacted by studying risk factor heterogeneity in whole cultures contrasting in diet, where he eventually obtained the classical evidence that individual risk traits were universally valid predictors across populations but that dietary saturated fatty acids and blood cholesterol level were necessary causal factors for both population and individual risk.
The Body of Evidence
Elsewhere I have recounted the detailed story of Keys’s mental lapse when he bluntly advanced his hypothesis about the central causal importance of diet and heart attacks before skeptical peers at the earliest WHO conference on atherosclerosis in the mid-50s. Failing to summarize the whole body of evidence, cutting directly to his conclusion, and then being abruptly challenged, Keys cited a single piece of evidence, which the experts were easily able to diminish. Stung, I suspect, he went forth to organize the Seven Countries Study, achieving about as definitive evidence as could have been assembled in its day to establish the major influence of diet and blood lipids on both population rates and individual risk of heart attack.
Keys early recognized the relationship among diet, physical activity, and overweight and challenged medical and actuarial claims of its centrality to the risk of premature cardiovascular disease and death in many populations. Stung again, I suspect, by having little obvious influence on this widely prevailing thought, he has continued well into his 90s to study those relationships and to challenge the common wisdom. He established that almost any relationship can be demonstrated between body mass and coronary disease or all-causes death, depending presumably on cultural differences in arriving at the obese state. In population-based samples of rural and working men followed for 30 years, those in the center of the distribution did better than those at the extremes. Weight gain in middle years foretold greater longevity than stable or decreasing weight in those years, even after eliminating early deaths. This, in turn, was not wholly explained by stopping smoking as cause of the weight gain.
These are just a few examples from just one career in which important lessons can be learned about how to deal with mistakes and failures.
Been knocked down?