Low-fat Diets for Long-term Weight Loss What Do Decades of Randomized Trials Conclude? HSPH Nutrition Department Seminar Series October 5, 2015 Deirdre Tobias, ScD Instructor of Medicine Harvard Medical School & Brigham and Women's Hospital Division of Preventive Medicine
Disclosures None
IN OUT WEIGHT LOSS Macronutrients - Fat = 9 kcal/g - Protein = 4 kcal/g - Carbohydrates = 4 kcal/g Hypothesis generating? - Calories in vs. calories out - Reduce total calories weight loss - Reducing total fat the most efficient way to reduce total caloric intake - Weight loss (and overall dietary) recommendations centered around low fat diets (<30% calories from fat) - 1960s diet heart hypothesis, groundwork against fat already laid
1992 USDA Food Pyramid: Fats & Oils: Use sparingly Bread, Cereal, Rice & Pasta: 6-11 servings/day
FAIL
Hundreds of short-term trials: Weight change Palatability Satiety Energy expenditure Adherence measures intermediates
Previous Meta-Analyses Hooper, et al (BMJ, 2012) N=33 RCTs Favors low fat WMD=-1.6kg (95% CI=-2.0, -1.2); I 2 =75% Includes trials of 6-12 months Excluded weight loss trials Over-samples trials among those w/ illnesses or risk factors (e.g., breast cancer, cardiovascular disease) N=3 healthy populations N=8 with similar attention to intervention groups
Previous Meta-Analyses Johnston, et al (JAMA, 2014) N=48 RCTs Favors low fat WMD=-1.6kg (95% CI=-2.0, -1.2); I 2 =75% Includes trials of 3+ months duration Overweight/obese populations with weight loss goal Included named diets (Atkins, Jenny Craig, etc)
Aim: to summarize the large body of RCT literature on low fat vs. other dietary interventions on long-term weight change
Methods Systematic literature review Inclusion criteria Randomized clinical trials 1 year of follow-up Trials often observe max weight loss ~6 months, with regain Adult populations Report change in body weight & variance by group Compare low fat vs. other dietary interventions Non-dietary intervention components were allowed if similar between groups (e.g., exercise recommended for both) Exclusions: pregnant populations, dietary supplements or meal replacement beverages, etc
Methods Meta-analysis Outcome: change in body weight DerSimonian and Laird random effects model to compute mean change difference between low fat vs. comparator diet Heterogeneity: conducted X 2 tests and I 2 values (>50% ~ moderate) Data extraction (independently by 2 investigators) Population characteristics Intervention details Outcome measure and variance (weight change, adherence measures) Intention-to-treat results were extracted when available Stratification by intervention and population characteristics Publication bias (Begg s and Egger s tests and funnel plot)
Results: Study attrition diagram Citations identified and screened: 3517 Initial acceptance on the basis of title or abstract: 333 Rejected after review of full manuscript: 280 Reasons for rejection: 17 Non-randomized 44 <12 month follow-up 52 No low-fat diet comparison of interest 13 Multicomponent intervention(s) 38 No weight outcome of interest 5 Not English language 5 meeting abstract, commentary, review article 2 unable to locate original article 12 data not extractable/no variance measure 92 duplicate study population Accepted for inclusion in meta-analysis: 53
Results: Characteristics of RCTs Trial characteristics N=53 RCTs N=37 from North America N=27 were 1 year in duration N=20 among participants with chronic disease (e.g., cancer, CVD) N=35 weight loss trials, N=10 no weight goal, N=5 maintain Diverse dietary interventions Low fat: <10% to 30% of calories from fat Calories from fat was not the entire focus of all interventions Wide range of comparator diets (e.g., usual diet, low carb, Mediterranean pattern) Intensity varied from pamphlets at baseline to feeding trials Caloric restriction not a component of all weight loss trials
Weight Loss Trials N=33 Vs. Low-Carbohydrate (n=18) Vs. Other Higher Fat (n=19) Vs. Usual Diet (n=8) 53 RCTs Low-Fat vs. Other Non-Weight Loss Trials N=18 Vs. Low-Carbohydrate (n=0) Vs. Other Higher Fat (n=7) Vs. Usual Diet (n=11) Weight Maintenance Trials N=6 Vs. Low-Carbohydrate (n=0) Vs. Other Higher Fat (n=3) Vs. Usual Diet (n=3)
Results: Weighted mean difference in weight (kg) for low fat vs. control dietary interventions N=53 RCTs included (n=68,128 adults) Overall mean weight loss = 2.7 kg (SD = 2.8) Weight Loss Goal: WMD = -0.41 (-1.29, 0.48) I 2 =84% p-het<0.0001 No Weight Goal: WMD = -1.54 (-2.32, -0.76)* I 2 =85%, p-het<0.0001 Maintenance: WMD = -0.70 (-0.88, -0.53) I 2 =0%, p-het=0.9
vs. usual diet vs. other vs. low-carb Forest Plot: Weight Loss Trials Random effects pooled WMD (kg) for low fat vs. control dietary interventions from RCTs reporting at least 1 year of follow-up 1.15 (0.52, 1.79) 0.36 (-0.66, 1.37) -5.41 (-7.29, -3.54) Favors low fat Favors control
vs. usual diet vs. other Forest Plot: No Weight Loss Random effects pooled WMD (kg) for low fat vs. control dietary interventions from RCTs reporting at least 1 year of follow-up 0.26 (-0.39, 0.91) -2.22 (-3.00, -1.45) Favors low fat Favors control
vs. usual diet vs. other Forest Plot: Maintenance Random effects pooled WMD (kg) for low fat vs. control dietary interventions from RCTs reporting at least 1 year of follow-up -0.95 (-2.00, -0.10) -0.70 (-0.88, -0.52) Favors low fat Favors control
vs. Usual Diet E.g., WHI n=48,835 postmenopausal women Primary endpoints: breast and colorectal cancer Low-fat intervention: 20% fat, fruits and vegetables 5/d, grains 6/d Small group sessions, individual meetings, tailored diet Maintain body weight Control: no contact with study dieticians Howard, et al (JAMA, 2006)
vs. Usual Diet Mean follow-up = 7.5 years >70% FFQ response Fat intake LF: 38.8% 29.8%* C: 38.3% 38.1 Fiber LF: +2.2 g* C: -0.2 g Frt/Veg LF: +1.4 s/d C: +0.2 s/d Physical Activity LF: +1.1 METs/wk C: +0.9 METs/wk Howard, et al (JAMA, 2006)
vs. Usual Diet Mean follow-up = 7.5 years Body weight LF: 0.8 kg* C: 0.1 kg Waist circumference LF: +1.6 cm* C: +1.9 cm Waist-hip-ratio: LF: +0.02 C: +0.02 Howard, et al (JAMA, 2006)
vs. Usual Diet Address a different research question Cannot isolate the independent effect of dietary fat Numerous other components Confounding by intervention intensity/attention from study staff
Weight Loss Trials - with Similar Intensity
Weight Loss Trials - with Similar Intensity
Weight Loss Trials - with Similar Intensity
Weight Loss Trials - with Similar Intensity
Weight Loss Trials - with Similar Intensity
No Weight Loss - with Similar Intensity
No Weight Loss - with Similar Intensity
No Weight Loss - with Similar Intensity
No Weight Loss - with Similar Intensity
No Weight Loss - with Similar Intensity
Strengths and Limitations Meta-analyses are limited to the strengths & limitations of individual RCTs Results were robust to sensitivity analyses by study quality measures (sample size, ITT, Jadad Score) Stratification by important trial characteristics indicated significant sources of heterogeneity Relative intervention intensity of control group Magnitude of difference in fat intake (% calories and TGs) Moderate unexplained heterogeneity
Conclusions Long-term effect of low-fat diets on body weight depends on the intensity of intervention in the comparison group. When compared to dietary interventions of similar intensity, evidence from RCTs does not support low-fat diets over other dietary interventions.
Conclusions and Potential Implications Modest (at best) long-term weight change achieved on any of the diets evaluated. Additional research is needed to identify optimal diets for weight control, which may extend beyond a focus on macronutrient composition (e.g. dietary patterns). Additional health-related intermediate endpoints and outcomes need to be considered (blood pressure, diabetes, etc).
Back where we all started? Suisse Report (2015)
Overall AHEI 2010 NHANES adults aged 20-85 years improved diet quality Wang D, et al (JAMA Int Med 2014)
Recommend foods vs. fats
Recommend foods vs. fats
Thank you! Thank you to co-authors: Mu Chen JoAnn E. Manson David S. Ludwig Walter Willett MD Frank B. Hu Supported by NIDDK K01 #DK103720, American Diabetes Association Mentor-Based Postdoc #7-12-MN-34, and BNORC U54CA155626