What's the Expected Shape of a Weight Loss Curve?

We "discovered" low-carb diets about 6-7 months ago by reading Gary Taubes' recent book, Good Calories, Bad Calories (highly recommended, if admittedly somewhat controversial). Deciding that some of our own struggles with weight control even (or especially) in face of a 30-40-mile per week exercise program might be due to excessive carb consumption, we embarked on our own dietary revisions and experiments.

But regardless of the weight-loss method adopted, no one seems to show any accurate data on weight loss vs. time. You can find plenty of anecdotal data: "I lost 50 lbs. (10, 20, 100, whatever) in x weeks (months, days)." You can find before and after photos. You can find data correlated with lipid profiles, blood pressure, cardiac incidents, diabetes, even death. You can even find guidelines that suggest that you should plan to lose, say 1–2 lbs. per week, but is that right? Does it depend on how much you need to lose? Should the weight loss be linear over some period of time? Obviously, it can't be linear indefinitely, since your weight must necessarily level off at some point. The next simplest model might be an exponential approach to some target weight value. That model might arguably be correct if you make an incremental change to your diet/lifestyle that causes you to have a new natural weight setpoint. If that new setpoint is a bunch of pounds different from where you are now, what is the expected time constant for approaching the new set point? Or is that model too simple too? And for that matter, how much noise should you expect to see in the data? What "normal" day-to-day fluctuations should one expect to see? None of these questions seems to be answered in the literature, even though they are clearly easy to measure and clearly of great psychological importance to anyone trying to lose weight.

As scientists, we, of course, have been keeping records! The data are still somewhat limited and anecdotal. Observations so far:

(1) Low carb diets really work for us as means of reversing persistent long-term weight gain. We'll talk about our particular diets in another posting, but so far, we've seen losses of 10% body weight without much effort or hunger. It's a permanent diet change, but once you get used to it, it's painless, healthy and easy to continue.

(2) Over the time we have recorded so far (about 7 months), we still can't clearly distinguish between a linear fit and an exponential fit to the data. But all that really proves so far is that, if the exponential model is valid, we are seeing a time constant of somewhere in the range of 10-14 months. (We have each lost about 20 lbs so far and expect to be able to lose another 20 or so.)

(3) There are normal daily fluctuations of ± 1.5–2 lbs (remember that if you drink a pint, it's a pound added, and there are all sorts of things that can easily cause variations in fluid retention of a pound or two).

(4) There may be ~monthly cycles that may or may not match menstrual cycles that account for a couple of pounds of cyclic variations.

(5) There appear to be effects (usually temporary) associated with heavy exercise that go beyond simple fluid replacement issues. Again, these can account for temporary weight increases of a pound or two.

(6) There are also clearly seasonal and environmental effects due, at least in part, to the body's response to average temperature. This is another fluid retention effect, I think.

(7) There is often a perceived "plateau" effect where one seems to be stuck at certain levels for up to a few weeks at a time and then suddenly lose a few pounds. These are often reported anecdotally, especially when measurements are made infrequently (weekly or less often) but, in our case (daily measurements), we can't really prove that they're not just artifacts of the scatter in the data. We certainly see clusters of data points that are above and below the long-term trend lines, but we rarely see points that are appreciably outside a 1–2 lb. tolerance band about the longterm trend lines.