Mike Beversluis

Thursday, July 30, 2009

Let's make a linear curve fit through one data point...

Megan McArdle posted the other day about the obesity "crisis", and continued on today with somewhat strident language:

3. We don't have any good way to make people shorter, but we do know how to make them lose weight.

Actually, this is rubbish: we don't know how to lose weight. Some of the things Paul Campos is saying about obesity are controversial, but this isn't. Every single study which has attempted to make overweight people get thin without very risky surgery has failed completely and utterly. Fewer than 1% of patients ever keep the weight off.

Now, being fat is not the end of the world, as I was gradually discovering more and more over the last few years. Fat may be a bit of an exaggeration, but "fleshy" is probably about right. In the euphemism of so many online dating profiles, I was "average". That's all fine and good, but I disagree that it's a proven fact that you cannot loose weight, since that is what I have done, fairly slowly, this year. Yes, it's unlikely that dieters on average will succeed (I've seen 95% elsewhere), and yes people yo-yo diet, but I have sought to avoid this through data analysis. Let me illustrate: Here is a graph of my measured weight over the last few months:

The black curve is the raw measurement and the red is a 10-day exponentially-weighted moving average. The month of February was a baseline measurement. After that, I ate an average of 1650 calories per day, which resulted in losing about 1.5 lb per week for the last five months. This is a moderate pace, and fits within the FDA's guidelines for weight loss rates. The total amount may seem large, but at 6'5", the 30 lbs here represents about 3 inches around my waist. About a month after starting the diet, I also started a mild exercise program which totals about 100 minutes per week. This has been enough for slight fitness improvement, but as you can estimate, it didn't really affect my weight loss. That's expected, since that much exercise works out to about 600 calories per week, which is small compared to the average 5200 calorie loss. My main interest was to not lose muscle during this dieting process, and so far, that's held true. Actually, I've gained muscle, so this data underestimates my fat loss.

I have been followed the method outlined in the Hacker's Diet, which simply involves counting calories. I've been doing this online with the rather nice site FitDay, which allows you to estimate your nutritional intake in a few minutes. I used Hacker's Diet online to record my weight each morning because I liked the built-in smoothing, which has been critical for a continued sense of improvement. There is noise in the daily measurements, and it can be difficult to appreciate the steady 0.2 lb trend when things fluctuate +/-2lb. This idea is outlined in that online text, but having done it this long, I will tell you it makes a huge motivational difference to see that the trend is still moving down.

This allowed me to stop using appetite, or more accurately, stop using the feeling of fullness, to signal when to start or stop eating. A ~700 calorie deficit is roughly equivalent to skipping a meal, so basically I sharply downsized my lunch, and cut out eating out, snacks and beer. I have not been 100% strict about this, more like 75%. I usually follow it on the work week, and then ease up a little on the weekends. The day's activities during work make it easier to eat less. These excursions are not binges, but are usually about what I ate during my baseline.

What I've found is that this time, while not exactly fun, hasn't been hard either. I don't have the strong hunger pain that McArdle describes. Contrary to most dieting advice, I have found it's easier to fast for part of the day than it is to eat a little throughout the day - I'm about as hungry at lunch if I haven't eaten yet as I am if I had breakfast. In fact, eating breakfast makes me hungrier at lunch. Once the topic is broached, though, I have also tried to fill up on bulky, low-calorie food like fruits and vegetables. Since I have a sweet tooth, I've used pinapple and non-fat yogurts to substitute for candy and ice-cream. I've switched to leaner cuts of meat, and like frozen shrimp as an easy low-fat protein.

So far my metabolism has been fairly steady, but her point about minumum sustainable weights might have some validity, so I'm curious if this will change now that I am back to my old college weight. I wasn't extra-lean then, so I'm going to try and continue for another few months and see what happens. According to her, my metabolism should drop down to zero, and I should be starving hungry all of the time, dreaming of cheeseburgers (which hasn't happened yet either)(because I have had some during this period) but I suspect that mixing up my daily intake has kept my body from going into a stong starvation mode. I think part of the results from her cited studies was due to crash dieting. Most weight gain is gradual, and I think if you keep the weight loss gradual, you are less likely to invoke these strong metabolic responses. To make a comparison, it's like suggesting that no-one can become a jogger unless they can immediately run a competitive marathon. Baby steps, Bob, baby steps. But perhaps it's also affected by total bodyfat and not just a daily calory deficit. We shall see.

This is where data analysis is very helpful, since I can measure small long-term changes. I'd esitmate that my resolution is on the order of 0.1lb per month, so that you can easily pull out a gradual shift. The downside of this is that I will need to keep recording my weight daily, and most likely keep recording my food intake too. But to tell the truth, doing so online has been very easy. The upshot is that since getting here has been surprisingly easy, I'm very confident that I will be able to maintain this lower weight.

So, I think it's very possible for most people to succeed dieting, relatively painlessly, by using data processing. As silly or trite as that may sound, tools do make a difference. I would guess that some people would be much hungrier than I when losing 1.5 lb per week, but that suggests a lower rate of weight loss should be tolerable. I did find it much harder to lose 1.6 lbs than 1.4 lbs per week, so this should be tailored to the individual's comfort to ensure long-term discipline. Unless something is physically wrong, which is probably <5% of the population, they should be comfortable, if not thrilled, with a zero-weight gain diet, so there should exist a tolerable deficit point, even if it was only 0.2 lb per week.

Now, having said all that - McArdle is absolutely right about both the obnoxious moral tenor of the conversation, and with the surprising (to me) lack of health problems for overweight people. There is little affect of even rather high-obesity on life expetancy - it starts to really ramp up over 35-40, so I'd have to weigh over 300 lbs. As I now stand, I have traversed an entire BMI range, and my long-term health expectancy is dominated by other factors. I'm still happy I did it, and I think people who want to for asthetic reasons, without need for expensive diet plans or health clubs, by using measurements to replace their appetites in an age of overabundant calories.



Post a Comment

<< Home