I didn’t want you all to have to wait all weekend for the truth: Walter Willet didn’t really say, “I’ve never met a statistical person I didn’t like,” but he is sort of the Will Rogers of nutrition.
Everybody likes him, me included. Like Will Rogers was about politics, Willett is a staunch nutrition middle-of-the-roader who thinks fat it not so bad after all, but hey now, let’s not go any kind of crazy here, because saturated fat will still kill you in a New York minute probably maybe.
I spent a lot of time with him earlier this year—okay, really just his book, but his book is so sweet and personal that I felt just like I was sitting at the master’s feet—which were clad in my imagination in the most sensible of shoes—as he unfolded for me the saga of nutritional epidemiology.
What I’m about to say is said with all due respect to the man himself (he’s basically created a whole freekin’ discipline for goodness sake). This is simply my reading of a particular text located within a particular context, i.e. this is what happens when they let English majors into science programs.
There are many reasons why nutritional epidemiology may not be up to the task of giving us a sound basis for nutrition policy. But why take my word for it? If you want to understand the heart of nutritional epidemiology—the driving force behind our bold 40-year march in the misguided direction of one-size-fits-all dietary recommendations—you must read Walter Willett’s Nutritional Epidemiology. It is a book I love more every time I read it, and I say this in all sincerity.
While I suppose it was written as a sort of textbook, and it is certainly used as one, it doesn’t really read like a textbook. It is part apology and part defense, and is much more about “why” than “how.” And the “why?” that it tries to answer to is “Why apply the techniques of epidemiology to nutrition and chronic disease?”
In this regard, it is a touching masterpiece. Walter Willett, MD, DrPH is a professor at the Harvard School of Public Health and at Harvard Medical School. He is considered by many to be the father of nutritional epidemiology. To stretch the analogy, you can think of nutritional epidemiology as his child. Reading the book this way, it almost moves me to tears (again, not joking*), for I find this book to be a father’s sweet and sad paean to a beautiful prince full of promise, who has grown into a spoiled, churlish, and lazy adult, unfit to rule the kingdom, but with too much of the dreams of many poured into him to banish altogether. And the dreams of the father are the most poignant of all.
Apparently, to Willett’s eternal dismay, the whole field got started off on the wrong foot by focusing on dietary cholesterol (as a cause) and serum cholesterol (as an outcome), associations—as we now know—that turned out to be weak, inconsistent, nonexistent, or even the inverse of what was expected (pp. 5-6, 417-418) . We now know that sub-fractions of serum cholesterol affect heart disease risk differently (LDL-C vs HDL-C, for instance) and that different foods affect different aspects of serum cholesterol differently, making the relationship to overall heart disease risk even more obscure, which seems to be par for the course in this field, as Willett readily admits.
Here, according to Willett, is what we don’t know and can’t do in nutritional epidemiology:
- We don’t know any given individual’s true intake. It can only be estimated with greater or lesser degrees of error. (p. 65)
- We don’t know any given individual’s true status for a nutrient. Ditto above. (p. 174)
- We don’t know the true nutrient content of any given food that a person might eat. Double ditto. (pp. 23-24)
- We don’t know what factors/nutrients in a food may operate together to prevent/cause disease. Similarly, we don’t how foods commonly found together in dietary patterns may operate together to prevent/cause disease. (pp. 15, 21-22, 327-328)
- We have a really hard time separating calorie intake from nutrient intake (Ch. 11). Ditto nutrients and food patterns, food patterns and lifestyle patterns, etc. (pp. 10, 15, 22)
- We can’t separate metabolic consequences of food intake patterns from the food itself, i.e. what we are looking at in any given data set is really metabolism of food, not food. (p. 15)
- We don’t know what really causes the chronic diseases we study in nutrition epidemiology (p. 12); age, genetics, education, income, and lifestyle factors may influence, modify, or be more important than any dietary factor in the origins of these diseases (pp. 10, 15).
- We can’t distinguish between causal and coincidental associations. Furthermore, weak associations could be causal; strong associations can be coincidental (p. 12).
- Associations we do find are likely to be weak; we will often find no associations at all. Even if we do find statistically significant associations between nutrients and disease, they may be clinically or practically irrelevant and should not necessarily be used to make public health recommendations. (pp. 12-14, 21).
But wait! Willett cries. Don’t give up! This book is also a defense of those shortcomings—although one blinkered by what I must assume is Willett’s love for the field. I am always a little touched and frustrated by the section on why we find so many instances of lack of association between an ostensible nutritional cause and a disease outcome in nutrition epidemiology. Willett meticulously lists the possible reasons one by one as to why we may not be able to “observe a statistically significant association when such an association truly exists” (pp. 12-14). At no time does he venture to offer up the possibility that perhaps—and how would we know one way or the other?—no such association does truly exist.
A new edition of the book is coming out; this should make the old edition cheap in comparison. I won’t read the new edition because I’m afraid it would ruin my romance with the old edition, which is the one I recommend to you.
If you think Gary Taubes is “a poisonous pea in an ideological pod” (as I’ve heard him called), read this book (especially Ch 17 on “Diet and Coronary Heart Disease”). On the other hand, if you think population studies investigating nutrition and chronic disease are basically a gigantic undifferentiated crock of malarkey, read this book. Why? Because there are no clear answers and no real heroes. If you want to know the strengths and weakness of nutritional epidemiology, best to hear them outlined in excruciating and loving detail by Willett himself.
You don’t have to read it cover to cover. Skip around. You’ll learn in passing some methodology behind the folly of trying to forge links between specific nutrients in food to long-term chronic diseases that have multiple and complex origins (just the sections on how we collect information about what we think people are eating are eye-opening in that regard—Ch. 4-8). But I think (I hope) you’ll also hear the voice of a father wise enough to know that children are—must be—brought into this world on grand faith, one that hopes that they will make the world a better place than before, and that his child—nutritional epidemiology—is no different. Willett believes in this child and the book is a statement of that faith.
Please draw your own conclusions, here’s mine: Faith is not science.
Any parent out there knows this: you seem at first to have a child of your own, but you end up sending an adult out into the world who is no longer yours and never really was. The mistakes, limitations, failures, shortcomings belong only to that grown child, not to the parent. But still. It may be hard to acknowledge the fact that your precious one is no better than the other kids and probably won’t save the world. Sometimes, when I’m reading this book—when I’m supposedly studying for an exam—I am caught unawares by the sighs of disappointment, the rally of excuses, and finally the prickly justifications: The prince must be allowed to rule; the king knows he’s a weak little louse, but he’s all we’ve got.
I know—and any of us who are students of literature know—that this is the king’s tragic flaw. The prince can’t save the kingdom; the empire must crumble. But here is the king, holding brick and mortar together through sheer force of will, somehow acknowledging and somehow—at the same time—unaware, that this particular castle was built on sand in the first place. In this book, I hear Willett’s love for a hopelessly flawed field, a touching declaration of blind optimism, and I love this book, and I deeply respect the man himself, for showing that to me.
Note: I don’t expect anybody but dweeby English majors to get the title of this post, but for dweeby wanna-bees, see T. S. Eliot’s “The Love Song of J. Alfred Prufrock.” It just makes my heart sing with joy that Willett refers to his diet of preference as the “prudent” diet.
Stay tuned for N of 1 Nutrition: Part 4, when you’ll hear Dr. Roger J. Williams say:
“Nutrition is for real people. Statistical humans are of little interest.”
*Admittedly, it could be eye strain. I am OLD.
Page numbers and chapters refer to the following edition:
Willett, W. Nutrition Epidemiology, 2nd ed. New York: Oxford University Press, 1998.