Effect of Low-Fat vs Low-Carbohydrate Diet on 12-Month Weight Loss in Overweight Adults and the Association With Genotype Pattern or Insulin Secretion: The DIETFITS Randomized Clinical Trial
Christopher D. Gardner, John F. Trepanowski, Liana C. Del Gobbo, Michelle E. Hauser, Joseph Rigdon, John P. A. Ioannidis, Manisha Desai, Abby C. King - JAMA, 2018
In this 12-month randomized trial, there was no significant difference in weight loss between a healthy low-fat and a healthy low-carbohydrate diet (HLF -5.3 kg vs HLC -6.0 kg; between-group difference 0.7 kg, CI crossing zero). Neither a predefined genotype pattern nor baseline insulin secretion predicted which diet worked better. An inconvenient result for low-carb advocates is that LDL cholesterol rose in the low-carbohydrate group, with 12-month LDL changes significantly favouring the low-fat diet.
Key findings
- No significant difference in 12-month weight loss: HLF -5.3 kg vs HLC -6.0 kg (mean between-group difference 0.7 kg; 95% CI -0.2 to 1.6 kg, crossing zero).
- No significant diet-genotype pattern interaction (P = 0.20) and no significant diet-insulin secretion (INS-30) interaction (P = 0.47) for weight loss.
- Both diets improved lipids and lowered blood pressure, insulin, and glucose, except that LDL cholesterol increased in the healthy low-carbohydrate group.
- 12-month LDL cholesterol changes significantly favoured the healthy low-fat diet.
Effect measures
- Risk Difference: 0.7 kg (HLC minus HLF; CI crosses zero)95% CI -0.2 to 1.6 kg
- Other: HLF -5.3 kg vs HLC -6.0 kg
- Other: P = 0.20 (not significant)
- Other: P = 0.47 (not significant)
Why this evidence tier (High)
- Risk of bias:
- A large, well-conducted randomized trial with a prespecified primary outcome and prespecified genotype/insulin hypotheses; unblinded by the nature of a diet intervention, with 79% completion.
- Precision:
- Adequately powered (609 participants); the weight-difference interval is narrow and crosses zero, supporting a genuine null rather than an underpowered one.
- Directness:
- Directly compares two real-world dietary patterns on weight and on lipid and metabolic markers over a full year.
- Consistency:
- The null weight difference is consistent with other head-to-head diet trials; the LDL rise on low-carbohydrate aligns with other low-carbohydrate trials.
- Funding / COI:
- Supported by NIH (NIDDK and NHLBI) grants, the Stanford Clinical and Translational Science Award, and the Nutrition Science Initiative (NuSI), a low-carbohydrate-leaning funder co-founded by journalist Gary Taubes. Notably the NuSI funding did NOT bias the result toward low-carbohydrate: the trial found no weight-loss advantage and LDL rose on low-carbohydrate. Authors reported no personal conflicts of interest.
High certainty for the null: low-fat and low-carbohydrate produced similar weight loss, genotype and insulin secretion did not predict response, and LDL rose on low-carbohydrate despite a low-carb-leaning funder.
- Population:
- Adults aged 18-50 without diabetes, BMI 28-40 (overweight/obese), 57% women; randomized to healthy low-fat (HLF, n = 305) vs healthy low-carbohydrate (HLC, n = 304) diet for 12 months; 481 (79%) completed.
- Conflicts of interest:
- Authors reported no personal conflicts of interest (ICMJE forms). Funding included the Nutrition Science Initiative, a low-carbohydrate-leaning funder co-founded by Gary Taubes; that funding source did not bias the result toward low-carbohydrate.
- Funding:
- NIH (NIDDK grant 1R01DK091831; NHLBI grants 1K12GM088033 and T32HL007034), the Nutrition Science Initiative, and the Stanford Clinical and Translational Science Award.
Limitations
- Two minor JAMA corrections were issued (April 3, 2018: lipid unit conversions in Table 3; April 24, 2018: clarification of the funding description); neither affects the cited findings.
- Unblinded diet intervention with self-reported intake; 79% completion at 12 months.
- Measured weight and risk factors, not clinical cardiovascular events.
- Partly funded by the Nutrition Science Initiative, a low-carbohydrate-leaning funder, though the result did not favour low-carbohydrate.