Bean counting
What does someone who has been reading too much about nutrition eat?
After my talk at ML 4 Healthcare, I got a few questions about what I eat. I suppose this is a natural question for someone speaking about vitamins, diet, and nutrition. Maybe in a future blog, I’ll try to explain more about what led me to these diet-centric topics in the first place, but I can’t deny that at least part of it is that I recently started thinking more about what I eat.
I started my current diet after Coach Patrick told me I needed to eat more protein for better recovery. The party line (bro science) recommendation for strength fitness enthusiasts is to eat 1g of protein for every 1lb of body weight daily. I weighed 185 pounds then, so I had to figure out how to eat 185 grams of protein per day. I’m sure this recommendation is overkill, and it’s just an easy-to-describe, heuristic target. The literature on how much protein you need to grow and maintain muscle is, as you might expect, a total mess. But whatever, it’s a constraint in the diet problem and is as sensible as any other USDA allowance. Let’s just roll with it.
I told Laurent Lessard about Patrick’s recommendation, and Laurent told me to get the myfitnesspal app. The app has an immense database of every food out there. Stigler computed a subsistence diet using 77 foods. Myfitnesspal has over 14 million. When I started using this app, it would let you scan a barcode and automatically populate your daily log for you. They have sadly since started charging for this feature.
I started logging my food to see if I was hitting my protein goals. But to my dismay, I also started to realize foods had a ton of hidden calories. We’d do pizza nights on Mondays, and my hippie pizza from Whole Foods had 1500 calories. My breakfast was cereal and granola. Healthy! But granola would have 600 calories a cup. I added almonds to my smoothies because they were delicious and filling. Californians love almonds. They are the healthy nut. But they have 800 calories per cup.
With the aid of the app, I changed some of my regular meals around, found some reasonable substitutes, and ended up with a calorie deficit of around 750 calories. I had no idea that kind of thing was possible. I told Lauren (my lovely wife) how shocked I was, and she bemusedly quipped, “Welcome to being a woman.” Yeah, I get it.
After a couple of months of meal logging, I didn’t need the app anymore. I got a feel for what worked for me to get a ballpark of what I needed. And I’ve stuck with this “diet” for about four years. There are definitely things I avoid. I don’t eat pizza and pasta except on special occasions. I don’t tend to order sandwiches when out. I have a rough idea of how much I should eat. When I’m not traveling, I’ll prepare most of my meals, maybe having restaurant food or take-out four times a week. I don’t eat the “free lunches” that come with the university life.
I’m not a keto or Atkins devotee, but I do eat a rather high-protein diet. Protein is around ⅓ of my calories. A typical breakfast for me will be three eggs, a slice of toast, and some nonfat yogurt with fresh fruit and berries. I eat chicken rice bowls a lot. But I like them! I sous vide the chicken breast and have a dozen chili crisps in the fridge. There’s a reason places like Sweetgreen and Cava are popular.
I learned many things through this dietary shift. First, eating a recommended “performance diet” doesn’t have to be that crazy. It required thought and discipline, but I love to cook and have enjoyed the new creative challenge of keeping it interesting.
Second, I personally found high-protein meals that are less calorically dense are more satiating. If I eat a half pound of chicken, a cup of rice, and a serving of broccoli (150g), I feel more full than if I eat a personal pizza and a salad. But the pizza meal has over three times the calories. That’s wild! The signaling paths in our bodies are wildly complex.
I now appreciate the purist chicken, broccoli, and rice diet some athletes commit to. Boneless, skinless chicken breasts are more or less equivalent to protein powder. Rice is almost exclusively carbohydrates. Broccoli provides other essential nutrients, doesn’t have a lot of calories, and adds some chew to your food. Eating the same three ingredients at every meal certainly made diet calculations simple. But I also can see how obsessively doing this forever would lead to an eating disorder.
Third, and this will undoubtedly make me sound like a cuckoo fitness influencer, but US food marketing interests dispose us to unhealthy diets. I don’t know what we can make of nutrition labels, as the USDA gives companies a lot of lee-way with what they list there. I totally get how you can be misled by food advertising. “Healthy” foods with “net carbs” advertised on their packaging can pack a ton of calories. Food labels also have a ton of leeway and regulations let them be off by 20%. And even unprocessed foods will have rather varied nutritional content. We don’t get a very accurate count of what we’re consuming unless we stick with the exact same provider of chicken, frozen broccoli, and rice and periodically measure with a food calorimeter.
I don’t think anyone needs high precision obsession to get to a healthy diet. But I don’t have clean recommendations for how to fix things either. I have found a diet that works for me, and, in the process, come to appreciate how trying to find recommendations that work for everyone is impossible.
There are probably some recommendations that would "work" (in a strictly nutritional sense) for mostly everyone: "eat a substantial and varied amount of fruits and vegetables" will rarely be bad advice. In my view, the challenge of dietary recommendations (beyond the limited scientific understanding of many health effects and individual variability) appears when zooming into more specific advice and a more general concept of "working", i.e., how to make dietary changes that "stick", according to your socioeconomic circumstances and preferences. I follow a similar philosophy as you regarding meal choices (eyeballing nutrient content¹), and I think it works for me as a "numbers person", but would that work for anyone? Is there even a viable alternative?
Some governments in Western Europe are pushing Nutriscore (a grade based on quantization of a linear formula on macros and salt), and supposedly there is some evidence for it, but it just feels wrong to me. Even if you don't consider the strategic / adversarial implications and the different dimensions of nutrition (overall calorie intake, sodium content, proteins, micronutrients, fiber...), "nutritional adequacy" is a property of a complete diet, not a property of any individual product. If we put the optimization-tinted glasses on, all those terms are closer to constraints than additive terms in the objective function, so the actual marginal value of a given step depends on the current point, but let's try to avoid reasoning about hedonic Lagrange multipliers. I feel Nutriscore is neither a sufficient replacement for actual nutritional knowledge nor a necessary one at all, even after tweaking out all the nonsense². If you search for consumer advice in the context of Nutriscore, you are presented with a number of rules ("you should only compare foods/beverages within the same family", "good grades do not mean eat unbounded amounts", "prefer fresh produce") that undermine its pretense of simplicity (arguably the main selling point over general nutritional advice).
Coming back to the question of generality: assuming we have clearly defined nutritional goals, there's the question of how easy would it be for a specific person to swap some meals in or out depending on tastes, financial and "temporal" budgets, cultural contexts³, convenience, satiety... Here, I feel that there's a lot of interesting stuff that could be said, though advice along those lines will probably have to trascend the purely nutritional aspects. Ultimately, I think effective advice likely has to combine some basic principles (which macronutrient volumes and micronutrients are important for specific health/performance outcomes, and which foods contain them) with more practical suggestions that highlight potential options for adjusting your dietary habits.
¹ To be honest, I wonder if you can get any significant benefit from doing Excel-tracking at all! It's easy to find "nutritional content per 100g" tables, but how big are the errors bars and variability of those numbers? How accurately can you judge weights when not cooking at home? Are your actual macro targets exactly what some linear rules of thumb tell you? Surely a systematic approach is a big improvement over nothing, but it would surprise me if it was not in the zone of diminishing returns on precision. Some people would probably find themselves more comfortable with putting stuff on an app or spreadsheet, though.
² Early implementations featured stuff like "D-grade olive oil and Iberian ham", which invited some turmoil here in Spain (especially when compared to A/B-grade cookies or sodas), though apparently they kept adjusting categories and coefficients to mitigate counterproductive advice. At that point, however, what's the point of adding formulas and grades if they are adjusted until the results are the desired ones (i.e., the "expert consensus") that you already received?
³ One example that comes to mind: pulses/legumes. Quite useful from a nutritional PoV, as they are provide fiber, micronutrients, satiety and proteins (with all the aminoacid composition caveats that are not an issue for most people because of grains) without major drawbacks, but their popularity varies markedly around the world. Beans and rice is foundational in Latin America, and bean/chickpea/lentil stews are a staple in traditional Spanish food (as well as other Mediterranean countries). People coming from those gastronomic traditions (or people more curious about "world food") will probably have an easier time incorporating servings of those kinds of dishes if you have to, say, replace "pasta as a starter" with options beyond a bland salad.
So much to say on this but I cannot understand/buy/make sense of the 1g/body weight of protein entirely. As with everything fitness, it is overkill and completely at odds with reality. There are even crazier apps to manage macros and the whole obsession with it is senseless (of course, sour grapes for this slightly protein-intolerant fitness nerd) but it is also very interesting. More later as one preps for first day of class here in the valley of the sun.