For a while, it looked as if technology’s ability to track, record, and analyze personal data was going to revolutionize weight loss. By providing data, fitness devices and calorie counting apps gave us control. Knowing how many steps we’d taken (or, conversely, how many M&Ms we’d eaten) allowed us to calculate just how far we were from shedding the next pound.

In theory, at least. But what if the data is wrong?

In a new study published in the Journal of Personalized Medicine, Stanford researchers tested the accuracy of seven popular fitness trackers, including the Apple Watch, Fitbit, Microsoft Band, and PulseOn. Across the board, the devices weren’t very good at measuring the real number of calories burned. Fitbit, the most accurate device, had a median error rate of 27% — in other words, if your Fitbit indicates you’ve burned 100 calories, on average, that number is 27 calories off.

Not great. But far better than PulseOn, which had a median error rate of 93%. (Fortune attempted to reach out to PulseOn for comment, but its site was down.)

In general, across devices, counts were “all over the place,” says Euan Ashley, one of the study’s authors and an associate professor of medicine at Stanford University. “Sometimes it was too high, sometimes it was too low.”

For anyone who plans meals around the number of calories they’ve purportedly burned, these inconsistencies are a problem.

But rather than curse the devices for misleading numbers, Ashely recommends reevaluating the decision to count calories in the first place. This extends to calorie-tracking apps. While broadly useful as accountability tools, “we’re not very good at estimating calories in food, either.” Instead, he recommends a more analog approach: make a conscious effort to move more, and subscribe to a diet rich in vegetables, grains, and other whole, healthy foods.

Calories’ relationship to weight loss is itself in flux. For decades, it was believed to be simple: a calorie deficit of 3,500 calories translated into the loss of one pound. But as our understanding of metabolism — particularly how the body fights back against weight loss — grows, the neat simplicity of this equation is being questioned.

The Stanford study adds another layer to our shifting (and increasingly complex) understanding of weight loss. Whereas once it was simple (burn more than you consume), recent research suggests exercise often leads to weight gain, fitness trackers can inhibit weight loss, and a “healthy” diet changes based on genetics.

Even when the numbers are accurate, weight loss is more complicated than counting calories.

DIGITAL HEALTH

Can an app help treat depression? A teaching hospital in New York City is piloting a smartphone-based program that connects patients with care managers to treat anxiety and depression. While primary care physicians remain in control of patient health, managers provide supplementary behavioral-based services, including mental health assessments and appointment information, via the app. In a recent trial, participants using the platform showed higher rates engagement and improved symptoms of depression and anxiety. (MobiHealthNews)

INDICATIONS

Researchers are developing a drug modeled on two protective genetic mutations. Studies have shown that people carrying specific variants in a gene tied to triglyceride metabolism, a type of blood fat, have a significantly lower risk of developing heart disease. Now, the quest is underway to develop heart treatments that mimic the genetic mutations. (New York Times)

Cannabis can reduce seizures in children with epilepsy. In a double-blind, placebo-controlled study, cannabidiol, a component of marijuana that doesn’t contain psychoactive properties, was found to significantly decrease the number of seizures in children with a serious form of epilepsy. In the study, participants ingested a liquid containing the cannabis component. Manufactured by GW Pharmaceuticals, the drug has yet to receive FDA-approval. (The Verge)

THE BIG PICTURE

CBO report: the House healthcare bill would leave 23 million more uninsured. The nonpartisan budget office found that under the AHCA, 14 million more Americans would be uninsured by next year, a number that would grow to 23 million within a decade. In addition, those with pre-existing conditions would face “extremely high premiums” in states that get waivers. And while premiums would eventually go down for younger and healthier people, sick and older Americans would pay significantly more for coverage. (New York Times)

REQUIRED READING

Exclusive: This Data Crunching Startup Just Made a Big Acquisition, by Barb Darrow

Cyber Giant Tanium Raises $100 Million, CEO Defiant About Company Culture, by Jeff John Roberts

This Tech Cycling Company Is Now Worth $1 Billion, by Leena Rao

Automation, Robots, and Job Losses Could Make Universal Income a Reality, by Barb Darrow

The iPhone Decade, by Aaron Pressman and Nicolas Rapp