The Case of the Fitbit-Defying Metabolism

Calorie Monitoring as Crude Method 

In the past three weeks, I’ve burned 6300 more calories than I’ve consumed. I know this because I have an app on my phone (Lose It!) that keeps track of the calories I eat. This app is integrated with my Fitbit Flex wristband, which keeps track of the calories I burn. The math says that I should lose a pound for every 3500 calories burned. The bathroom scale, however, says I have gained two pounds.

I’m frustrated, but not at all surprised that this is not adding up (or subtracting) like it should. As the journalist Gary Taubes says, this notion of calories in vs. calories out would work if calories in the body obeyed simple rules of physics (if a calorie of protein was the same as a calorie of fat as a calorie of carbohydrate), but they do not. Nor does the human body behave like a machine, simply burning fuel as it’s added. While Taubes focuses on uncontroversial, general facts about carbohydrates and insulin, there are still other factors that affect how individual bodies strive for equilibrium. In a New York Times piece called “The Fat Trap,” Tara Parker Pope explores many of the complexities of human weight gain—complexities from person to person, and complexities within one body over time. There are a multitude of genes and a plethora of hormones, some only recently discovered, that play roles in hunger and metabolism, not to mention the changes in the body’s metabolic efficiency that take place in the body both when it begins losing weight, and once it has lost weight.

Of course, there is room for error in my calorie monitoring system. For one, the calorie-counting app on my phone relies on me manually logging the precise amounts of the foods I eat. That could account for discrepancies between what I actually consume and what the records say. But I carry my phone with me so as never to forget an entry, and I strive for accuracy. If I have a Greek yogurt, I scan the barcode to get the exact amount of calories (it also records carbs, proteins, and fat: while I know my body treats calories from each differently, I have ignored these particulars in favor of testing the simpler calories-in calories-out model). The portion is controlled—one container, exactly. When my garden is dormant and I eat more processed foods like this, I can more easily control for the numbers. (When the garden is thriving, I’m eating healthier anyway, but that’s another story.) Still, at a family function the other day, I had a bowl of homemade chili and a choice: I could have asked for the recipe and programmed it into my phone, or I could have, as I did, searched the database for a chili that looked close and settled for an approximation.

And yet, while my “calories in” is measured by a manual log, and “calories out” is measured by a tiny three-dimensional accelerometer on my wrist, it’s hard to tell which is the greater source of error. Both programs calculate (estimate) my Basal Metabolic Rate—BMR, my rest metabolism—using averages for someone of my age, gender, height, and weight. Researching the assumptions behind the equations used to obtain these estimates has been like falling down a rabbit hole: of five online BMR calculators, three gave one value and two gave another. Both equations (the Mifflin-St Jeor and the Harris-Benedict) use height, weight, age, and gender. I don’t know which is better—more accurate—but I did find a discrepancy between my calories-in counter and my calories-out wristband.

Lose It! uses the Mifflin equation, an estimate of my daily activity level, and my weight loss goal to determine a daily caloric budget. They assume a sedentary lifestyle, but users can correct for this on a given day by logging physical activity or, if they happen to have integrated a device like mine, earning a “Fitbit adjustment.” On days when the Fitbit Flex says I have burned more than Lose It! estimates, I get the difference subtracted from my net amount of calories in, a non-zero Fitbit adjustment. However, the adjustment only goes in one direction: if Fitbit says I burned fewer calories than Lose It!, calories are not added to my net intake—I simply have a Fitbit adjustment of 0 for the day. I recently discovered on a couple 0 adjustment days that Fitbit thought I was burning less than Lose It! predicted. As I said, I don’t know which estimate of my rest metabolism is correct, but to get Lose It! and Fitbit on the same page, I erred on the side of caution and changed the setting on the program that predicted I burn more.

The next step in the evolution of fitness data technology is a calories-burned monitor that learns from me: if I could teach the Fitbit Flex by giving it weight data (hey Fitbit, my weight hasn’t changed despite your predictions), it could adjust its algorithm for my body’s rest metabolism by incorporating that data over time.

Trying to sort this out has been maddening. None of it changes the simple fact that I have burned far more calories than I consumed by the safest estimates. I stay under budget with food. I put in thirty minutes on the treadmill despite my sense that this has little to do with the number on the scale. As Taubes put it, “despite half a century of efforts to prove otherwise, scientists still can’t say that exercise will help keep off the pounds”—a claim so unsettling it’s hard to believe, so I remain agnostic. Last week, I spent an afternoon digging a trench for an electrical line instead of using the treadmill. At least that activity has already yielded a tangible result.