Gary Taubes is probably the most influential health journalist of the past 10 years. He wrote two articles about the problems with low-fat diets - this one for Science magazine, and this one for the New York Times Magazine - that did more than any other reporting to shift the nutrition paradigm. I think it's safe and accurate to say that if Taubes hadn't made it okay for journalists to write about the evidence supporting low-carb diets, my career would've taken a different course.  

Taubes has a new book out this month, called Good Calories, Bad Calories. It's 600+ pages, and, from what I was told by a journalist who knows the author, that's the lean-and-mean version. At one point, the manuscript was several hundred pages longer.

Yesterday the New York Times Magazine published another major article by Taubes, called "Do We Really Know What Makes Us Healthy?" I'm not sure if it's an excerpt from the book, but it sure is worth the time it takes to read it.

We don't need Taubes or any other journalist to tell us that this year's health scare will be refuted next year, or five or 10 or 20 years from now, while the health-promoting food or medicine or lifestyle choice touted today will be declared useless or even deadly down the road. That's obvious.

But we do need someone with a wider lens to show us how and why this happens, and how to understand the shift in thinking when it occurs.

The first step toward understanding these reversals, Taubes explains, is knowing the difference between the two main types of research that doctors depend on when they give advice to their patients.

On big-picture health issues, the first type of finding that gets attention is usually from epidemiological studies -- a large group of people, followed over many years. If a subgroup has lower rates of a particular disease, the researchers dive into the data to figure out what makes them different from the others. They generally find something -- they eat more of one thing and less of another, they take a particular medicine, they do or don't drink or smoke, they get more exercise, or they tip the waiters better when they eat meals in restaurants.

Those behaviors or food choices or medications are correlated with the positive result, and that's what gets reported. If it's a major finding -- vitamin X is correlated with lower risk of cancer Y -- it quickly becomes part of the healthy-living canon.

The next type of study, a clinical or experimental study, uses real people to test the correlation. Some will get vitamin X, some will get a placebo, and nobody will know who's been given which until the results are in.

These double-blind, placebo-controlled studies often refute the correlations discovered in the epidemiological research. Sometimes, the interventions make the condition worse or more likely to occur in certain populations. And the media reports that, generally leaning toward extreme interpretations of the data: "Vitamin X doesn't prevent cancer Y. It actually causes it!"

So why do the people in the epidemiological research have less cancer Y? Taubes explains:

Conversely, disease and bad health tend to cluster among people who meet the opposite descriptions:

Now I wonder how we should interpret diet and weight-loss studies with this new perspective. Is it possible that a certain type of study participant will do better with any type of diet? And a certain type of subject won't see much in the way of results, no matter if he's given low-fat, low-carb, or anything in between?

And I really wonder about exercise studies. I was talking with a friend the other day about fitness publishing, and the types of recommendations we make. There's almost no such thing as a "typical" result with any type of workout or training system. Our bias is always toward the outliers -- the people who get the most extreme results. We use those outliers to inspire readers, without noting the difference between possibility ("You too could get results like this!") and probability (only one guy out of two dozen got the results we're touting).

For example, high-intensity training -- HIT -- is sold to the public almost entirely on the results of a handful of outliers. The first outlier was a bodybuilder named Casey Viator. In the legendary Colorado Experiment, in 1973, Viator was said to have gained 63 pounds of muscle in 28 days while losing 18 pounds of fat, for a net gain of 45 pounds of body weight. The legend says he did this without steroids, using an extreme type of HIT training on Nautilus machines, and taking a full day off after every workout.

If you read the fine print, you see that Viator had already been a genetically blessed bodybuilder. He'd won Mr. America in 1971 when he was 19, making him the youngest champion ever. Before the Colorado Experiment, he'd lost 40 pounds because of an infection following an industrial accident. So most of the muscle he added was actually tissue he had only recently lost and was regaining.

It's also worth noting that Viator peaked as a bodybuilder in 1971 -- he never won anything else until 1980, when he won three minor contests. Jones trained Viator for his successful run in his late teens, which certainly shows that HIT worked well for Viator. But we can't infer that Viator wouldn't have been just as successful with more traditional bodybuilding training. Since he couldn't possibly have done both types of training simultaneously, we'll never know.

What we do know is that HIT-type training -- usually defined as doing one set of each exercise to failure, two or three times a week -- has never looked particularly good when compared to multi-set training in which no sets are taken to failure. ("Failure" in this context is the point at which your muscles are so exhausted they can't move the weight another inch.)

Taubes' article isn't about HIT, of course. But it is a good reminder that studies rarely tell the entire story about any particular type of diet or lifestyle choice. My own addition to the conversation -- that you should always be skeptical about case studies of selected outliers -- is too trivial a point to bother arguing.

I only bring it up because anecdotes about outliers seem to drive the entire health and fitness industry. Which, of course, isn't trivial at all to people like me.

Monday blog meat

• Curiously, the L.A. Times has a major feature today on the same subject covered by Taubes -- why health research is so often contradictory. I don't know if it's worth your time to read both, but it's worth clicking the L.A. Times link just to read the opening paragraph.
• In other news, 60 percent of adults in the U.K. would rather die than work out. Meanwhile, in the U.S., 60 percent of adults would rather lie to researchers than admit they'd choose death over exercise.