DNA testing won’t guide dieters to the weight-loss regimen most likely to work for them, scientists reported on Tuesday. Despite some earlier studies claiming that genetic variants predict whether someone has a better chance of shedding pounds on a low-carbohydrate or a low-fat diet, and despite a growing industry premised on that notion, the most rigorous study so far found no difference in weight loss between overweight people on diets that “matched” their genotype and those on diets that didn’t.

The findings make it less likely that genetics might explain why only some people manage to lose weight on a low-carb diet like Atkins and why others succeed with a low-fat one (even though the vast majority of dieters don’t keep off whatever pounds they lose). Unlike cancer treatments, diets can’t be matched to genotype, the new study shows.

The results underline “how, for most people, knowing genetic risk information doesn’t have a big impact,” said Timothy Caulfield, of the University of Alberta, a critic of quackery. “We know weight loss is tough and sustained weight loss is even tougher. Genetics are relevant … [but] it seems highly unlikely that providing genetic risk information is going to be the magical formula that is going to fix this complex problem.”

The study, called DIETFITS and published in the Journal of the American Medical Association by researchers at Stanford University Medical School, randomly assigned 609 overweight adults, aged 18 to 50, to either a healthy low-fat or healthy low-carb diet. The volunteers got 22 hourlong classes with dietitians on healthy low-fat diets (eat less oils, fatty meats, full-fat dairy, and nuts) or low-carb ones (reduce cereals, grains, rice, starchy vegetables, and legumes), as well as on the dangers of eating mindlessly. Both groups were instructed to eat lots of vegetables and very few foods with added sugars, trans fats, or refined flour.

People were remarkably compliant about following their assigned low-fat or low-carb diet. Halfway through the study, fat consumption was 50 and 87 grams per day, respectively, while carb consumption was 211 and 113 grams per day, a pattern that held for the full 12 months. There was virtually no difference in weight loss between the two groups after 12 months: 11.7 pounds in the low-fat group and 13.2 in the low-carb one, a difference that was not statistically significant or meaningful in real life.

The researchers then analyzed weight loss among people whose DNA “matched” or clashed with their assigned diet. That was based on which variants of three genes—called PPARG, ADRB2, and FABP2, which are involved in processes such as fat and carbohydrate metabolism—they had. Earlier research suggested that these variants could predict who would successfully lose weight on which kind of diet.

Of the 244 people with the low-fat genotype, 130 happened to land in the low-fat diet group, meaning they were on the “right” diet for their DNA. That was the case for 97 of the 180 people with the low-carb genotype. The others were mismatched.

“There was no significant difference in weight change among participants matched vs mismatched to their diet assignment,” the researchers wrote. There was also no DNA/diet interaction for waist circumference, body mass index, or body fat percentage.

“I had this whole rationale for why these three [DNA variants] would have an effect,” said Stanford’s Christopher Gardner, co-author of the $8 million study. He previously led a smaller study, in 2010, finding that overweight women whose genotype matched their diet lost 13 pounds in a year while those who were mismatched lost just over 4 pounds. “But let’s cut to the chase: We didn’t replicate that study, we didn’t even come close. This didn’t work.”

One of the companies that sells DNA tests that promise to guide customers to the most-likely-to-succeed diet, Pathway Genomics, did not respond to requests for comment. A spokeswoman for another leading DNA/diet company, Habit, said it agrees that DNA alone “isn’t enough to develop personalized dietary recommendations” and that the company therefore factors in blood biomarkers and other information “when making personalized dietary recommendations.” Craig Pickering of DNAFit said that genes besides the three in the Stanford study can contribute to weight loss as well as fat loss and gain, as a small 2007 study found, and that it is “putting the finishing touches” on a study showing that “subjects on a genetically matched diet lost more weight” than did those on a one-size-fits-all low-carb diet.

Gardner’s own 2010 study, of 140 overweight women divided into various DNA and diet types, reached the same conclusion. Those on genetically matched diets seemed to do better. But that apparently reflected extreme outliers. For instance, a few women whose DNA matched their diet and were fanatic about calorie-counts and exercise shed pounds, making the matched group look great. Similarly, a few women whose DNA did not “match” went through a divorce or other upheaval, ate for emotional comfort, gained weight, and made the mismatched group look terrible—a reminder that so many emotional, economic, metabolic, social, and other forces affect someone’s chance of losing weight that the effect of genes gets lost in the noise.

“It just looked so cool back then,” said Gardner. “We were so excited and thought this [matching DNA profile to weight-loss diet] would work. It’s humbling, and just underlines the importance of replication” of tantalizing preliminary findings with larger, more rigorous studies.

Republished with permission from STAT. This article originally appeared on February 20, 2017