October 2012
Tian Hu, Katherine T. Mills, Lu Yao, Kathryn Demanelis, Mohamed Eloustaz, William S. Yancy, Jr, Tanika N. Kelly, Jiang He, Lydia A. Bazzano
American Journal of Epidemiology, Volume 176, Issue suppl_7, 1 October 2012, Pages S44–S54, https://doi.org/10.1093/aje/kws264

 

Abstract

The effects of low-carbohydrate diets (≤45% of energy from carbohydrates) versus low-fat diets (≤30% of energy from fat) on metabolic risk factors were compared in a meta-analysis of randomized controlled trials. Twenty-three trials from multiple countries with a total of 2,788 participants met the predetermined eligibility criteria (from January 1, 1966 to June 20, 2011) and were included in the analyses.

Data abstraction was conducted in duplicate by independent investigators. Both low-carbohydrate and low-fat diets lowered weight and improved metabolic risk factors. Compared with participants on low-fat diets, persons on low-carbohydrate diets experienced a slightly but statistically significantly lower reduction in total cholesterol (2.7 mg/dL; 95% confidence interval: 0.8, 4.6), and low density lipoprotein cholesterol (3.7 mg/dL; 95% confidence interval: 1.0, 6.4), but a greater increase in high density lipoprotein cholesterol (3.3 mg/dL; 95% confidence interval: 1.9, 4.7) and a greater decrease in triglycerides (−14.0 mg/dL; 95% confidence interval: −19.4, −8.7).

Reductions in body weight, waist circumference and other metabolic risk factors were not significantly different between the 2 diets. These findings suggest that low-carbohydrate diets are at least as effective as low-fat diets at reducing weight and improving metabolic risk factors. Low-carbohydrate diets could be recommended to obese persons with abnormal metabolic risk factors for the purpose of weight loss. Studies demonstrating long-term effects of low-carbohydrate diets on cardiovascular events were warranted.

There were an estimated 937 million overweight and 396 million obese people worldwide in 2005 (1). Moreover, it was estimated that 68.0% of American adults were either overweight or obese in 2009 (2). Overweight and obesity are important risk factors for diabetes, cardiovascular disease, cancer, and premature death. The high prevalence of obesity has become a serious public health challenge.

The dietary recommendations for weight loss from the American Heart Association and the National Institutes of Health emphasize the importance of low-fat, high-carbohydrate diets (3, 4). However, low-carbohydrate diets have recently become very popular for weight loss (5–7). Because low-carbohydrate diets may include significant amounts of fat and cholesterol, which have been associated with elevated low density lipoprotein (LDL) cholesterol levels, there is concern about their adverse effects on metabolic risk factors.

Some previous studies (6, 8–12), but not others (13–15), showed significant changes in metabolic risk factors associated with low-carbohydrate diets. Many clinical trials of low-carbohydrate diets have small sample sizes and insufficient statistical power to detect small changes in metabolic risk factors that may have public health importance.

A previous meta-analysis of clinical trials comparing low-carbohydrate and low-fat diets reported differences in metabolic risk factors between the 2 diets (16). However, that study included only 5 trials, with a total of 447 participants (16). Since then, several larger trials of longer duration have been published (6, 11, 13, 14, 17). Given this recent additional evidence, a meta-analysis of randomized controlled trials comparing the effects of low-carbohydrate diets with those of low-fat diets on metabolic risk factors is timely and important to public health.

In the present meta-analysis, we aimed to estimate the long-term (6 or more months) effect of low-carbohydrate diets compared with those of low-fat diets on body weight, waist circumference, and metabolic risk factors, including systolic and diastolic blood pressure, total cholesterol, LDL cholesterol, high density lipoprotein (HDL) cholesterol, ratio of total to HDL cholesterol, triglycerides, fasting blood glucose, and serum insulin.

In addition, we explored the possible underlying reasons (i.e., study duration, diabetic status, age, gender, and levels of carbohydrate intake in diets) for the observed heterogeneity of effect sizes.

MATERIALS AND METHODS

Data sources and searches

We used the MEDLINE online database (from January 1, 1966 to June 20, 2011), EMBASE, Web of Science, and the Cochrane Database of Systematic Reviews to identify randomized controlled trials that compared the low-carbohydrate diet with the low-fat diet. The following key words or medical subject headings on MEDLINE were used: (“low-carbohydrate diet” or “low sugar diet” or “carbohydrate restriction” or “diet, carbohydrate-restricted”) AND (“body mass index” or “BMI” or “waist circumference” or “obesity” or “diabetes” or “blood glucose” or “hypertension” or “HDL” or “LDL” or “triglycerides” or “cholesterol” or “lipids” or “dyslipidemias” or “blood pressure” or “diabetes mellitus” or “heart diseases” or “cardiovascular diseases”). The search was restricted to include studies classified as randomized controlled trials, and no language restriction was applied. In addition, manual searches of the references from selected original research and review articles were conducted.

 

Study selection

To be included in this meta-analysis, the studies had to be randomized controlled trials conducted in adults that compared a low-carbohydrate diet (≤45% of energy from carbohydrates) with a low-fat diet (≤30% of energy from fat) (18, 19), had an intervention duration of 6 months or more, and reported metabolic risk factors as outcomes.

Studies were excluded if treatment allocation was not random, the study included participants less than 18 years of age, there was no difference in carbohydrate or fat intakes between diets, there were differences other than macronutrient and energy intake between the 2 diets, metabolic risk factors were not reported by treatment status, the variance of outcomes could not be calculated, and/or the duration of intervention was less than 6 months (3).

When the results of a study were published more than once, only the most recent or most complete article was included in the analysis. Two investigators independently reviewed all potentially relevant publications and made decisions on inclusion. Where these decisions conflicted, additional investigators (co-authors) were involved to discuss discrepancies until mutual agreement was reached.

 

Data extraction and quality assessment

Two investigators independently abstracted the data in duplicate using a standardized data collection form. The following data were collected: article title, primary author's name, year and source of publication, country of origin, study design (parallel, cross-over, or factorial trial), blinding (open, single, or double), dietary composition, loss to follow-up, intention-to-treat analysis, characteristics of the study population (sample size, age, sex, prior disease status, and baseline levels of metabolic risk factors), and the net changes in metabolic risk factors with measures of variance, overall and by prespecified subgroups. Disagreement was resolved by consensus with additional investigators.

 

Data synthesis and analysis

Mean net change was calculated by subtracting mean change (from baseline to end of trial) in the low-fat-diet group from mean change in the low-carbohydrate-diet group for each metabolic risk factor. Pooled mean net changes and their 95% confidence intervals in metabolic risk factors were calculated using DerSimonian and Laird random-effects models (20).

The presence of heterogeneity was assessed with the Q test and the extent of heterogeneity was quantified with the I-squared index. To assess publication bias, we constructed a funnel plot for each outcome. Begg's rank correlation test was used to examine the asymmetry of the funnel plot, and Egger's weighted linear regression test was used to examine the association between the mean effect estimate and its variance. Where these tests or funnel plots indicated possible publication bias, we used the trim-and-fill method to determine whether publication bias could have accounted for the results we observed.

Additionally, sensitivity analyses were conducted by excluding each study in turn, by removing studies with a completion rate less than 70%, by removing studies with fewer than 20 participants per group, and by removing studies among postsurgery patients or those with severe diseases, such as cancer, to evaluate their relative influence on the pooled estimates.

Finally, subgroup analyses including diabetic versus nondiabetic samples, very-low-carbohydrate (≤60 g carbohydrates per day) versus moderate-carbohydrate (>60 g carbohydrates per day) diets, longer (≥12 months intervention) versus shorter (<12 months) study duration, predominantly male (at least 50% men) versus predominantly female (at least 50% women) samples, and older (mean age ≥50 years) versus younger (mean age <50) samples were conducted to examine differences in all study outcomes between the 2 diets.

The Bonferroni and false discovery rate methods were used to correct for multiple comparisons (21). All analyses were conducted using Stata, version 10 (StataCorp LP, College Station, Texas).

 

RESULTS

The flow of studies in our meta-analysis is depicted in Figure 1. From 785 potentially relevant references, 503 records remained after duplicates between databases were eliminated, and 406 articles were eliminated after reviewing titles and abstracts. A total of 97 full-text articles were reviewed for eligibility. Of those, 23 studies met all of the eligibility criteria and were included in the meta-analysis (5, 6, 8–14, 17, 22–34). These studies included data from 2,788 trial participants (1,392 on low-carbohydrate diets and 1,396 on low-fat diets).

The characteristics of these 23 randomized controlled trials are presented in Table 1. All trials were parallel except for 1 trial that had a factorial design (13). Trial participants were usually not blinded to their assignment because of the nature of the intervention—most interventions provided dietary instruction, leaving food buying and/or preparation to the participants.

Study duration ranged from 6 to 24 months, and 16 studies had an intervention duration of 12 months or longer (6, 8, 9, 11–14, 17, 22, 23, 25, 26, 28–31). Most trials were conducted among obese or overweight patients without cardiovascular diseases or diabetes mellitus. However, 4 studies were conducted in patients with diabetes (23, 26, 28, 29), 1 was conducted in patients with prediabetes (31), and 1 was conducted in patients with coronary heart disease (23).

The goal dietary nutritional composition varied across the studies, with carbohydrate consumption ranging from 4% to 45% (weighted mean, 23%) of energy intake in the low-carbohydrate group and fat ranging from 10% to 30% (weighted mean, 26%) of energy intake in the low-fat group. Self-reported mean energy intake weighted by trial sample sizes was similar for both diets at approximately 2,000 kcal.