Effects of free sugars on blood pressure and lipids: a systematic review and meta-analysis of nutritional isoenergetic intervention trials 1 3

Transcription

1 Effects of free sugars on blood pressure and lipids: a systematic review and meta-analysis of nutritional isoenergetic intervention trials 1 3 Elena Fattore, 4 * Francesca Botta, 4 Carlo Agostoni, 6 and Cristina Bosetti 5 Department of 4 Environmental Health Science and 5 Epidemiology, IRCCS- Istituto di Ricerche Farmacologiche Mario Negri, Milan, Italy; and 6 Clinical Sciences and Community Health- DISCCO, Università degli Studi di Milano, Intermediate Pediatric Care Unit, Fondazione IRCCS Cà Granda, Ospedale Maggiore Policlinico, Milan, Italy ABSTRACT Background: Sugar has been suggested as a central risk factor in the development of noncommunicable diseases. Objective: We assessed the evidence of the effects of free sugars compared with complex carbohydrates on selected cardiovascular disease risk factors. Design: We conducted a systematic review and meta-analysis of intervention trials to compare diets that provide a given amount of energy from free sugars with a control diet that provides the same amount of energy from complex carbohydrates. The primary outcomes were: blood pressure, total cholesterol, low density lipoprotein (LDL) cholesterol, high density lipoprotein (HDL) cholesterol, triacylglycerols, apolipo A-I and B, or very low density lipoprotein cholesterol. Body weight was also recorded but was not a primary outcome of the studies. Results: In all, 28 studies involving 510 volunteers were included. When free sugars were substituted for complex carbohydrates, no significant increases were detected in systolic or diastolic blood pressure, and no heterogeneity was observed. There were significant increases in HDL cholesterol, LDL cholesterol, and triacylglycerols, although for LDL cholesterol and triacylglycerols there was significant heterogeneity between studies and evidence of publication bias. After adjustment for missing studies, these increases lost significance. Subgroup analyses showed that diets providing the largest total energy intake and energy exchange enhanced the effect of free sugars on total and LDL cholesterol and triacylglycerols. The increase of triacylglycerols was no longer significant when studies with the highest risk of bias were excluded or when only randomized trials were considered. Free sugars had no effect on body weight. Conclusions: In short- or moderate-term isoenergetic intervention trials, the substitution of free sugars for complex carbohydrates had no effect on blood pressure or body weight and an unclear effect on blood lipid profile. Further independent trials are required to assess whether the reduction of free sugars improves cardiovascular disease risk factors. This review was registered at as CRD Am J Clin Nutr 2017;105: Keywords: blood pressure, cholesterol, triacylglycerols, body weight, free sugars, complex carbohydrates, intervention trials, meta-analysis INTRODUCTION Sugar was suggested as early as the 1970s as being involved in the etiology of some diseases (1), but only within the last 5 y has it entered the arena, together with tobacco and alcohol, as a central risk factor in the development of noncommunicable diseases (2). In addition to its role in dental caries, for which there is moderate evidence of a direct relation with sugar intake (3), increasing attention has been paid to how dietary sugars affect obesity, type 2 diabetes, and cardiometabolic and kidney diseases (4 6). The role of sugar in weight gain was strongly suggested by the temporal relation between the overconsumption of high-fructose corn syrup an industrial caloric sweetener used in place of sucrose and the obesity epidemic in the United States (7). In addition, meta-analyses of observational and intervention studies have suggested an effect of dietary sugars on blood pressure and lipid biomarkers of cardiovascular diseases (6). In 2015, the WHO issued recommendations for a reduction of daily sugar intake to,10% of the total energy intake for adults and children, noting that a further reduction to,5% would provide additional health benefits (8). Fructose in particular has been considered responsible for the pathologic conditions mentioned previously. Although fructose is a carbohydrate like glucose, it has a completely different metabolic profile because it does not stimulate insulin secretion, provides a rapid substrate for lipid synthesis, and may increase uric acid concentrations and stimulate endogenous inflammatory glycated products (9, 10). Despite ongoing research on the topic, some uncertainties still persist. For example, it is not clear whether the detrimental effects are caused by sugar intake or an excess of calories. Controlled isocaloric trials in which sugars were isoenergetically replaced 1 Supported by Soremartec Italia. 2 The funders had no role in the study design, implementation, data analysis, or interpretation. 3 Supplemental Figures 1 20 and Supplemental Tables 1 and 2 are available from the Online Supporting Material link in the online posting of the article and from the same link in the online table of contents at nutrition.org. *To whom correspondence should be addressed. elena.fattore@ marionegri.it. Received May 30, Accepted for publication October 17, First published online December 21, 2016; doi: /ajcn Am J Clin Nutr 2017;105: Printed in USA. Ó 2017 American Society for Nutrition

2 CARDIOVASCULAR EFFECTS OF FREE SUGARS 43 with other carbohydrates failed to show any effect on body weight (11, 12), and a meta-analysis that showed small but significant effects on biomarkers of cardiovascular disease risk was marred by unsolved heterogeneity (6). To further disentangle the role of sugar intake on important cardiovascular disease risk markers, we conducted a systematic review and meta-analysis of nutritionally controlled intervention trials that assessed the effect of the isoenergetic substitution of sugars (fructose, sucrose, and glucose) for complex carbohydrates on blood pressure, selected lipid biomarkers of cardiovascular disease risk (total cholesterol, LDL cholesterol, HDL cholesterol, triacylglycerols, apolipo A-I and B, and VLDL cholesterol), and body weight. To avoid misunderstandings or inconsistencies regarding the definition of sugars, we adoptedtheterminologyandclassificationusedbythewho, which defines free sugars as monosaccharides (such as glucose and fructose) and disaccharides (such as sucrose or table sugar) added to foods and drinks by manufacturers, cooks, or consumers and sugars naturally present in honey, syrups, fruit juices, and fruit juice concentrates (12). METHODS Search strategy and selection of studies This systematic review and meta-analysis (CRD ) was conducted following Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (13). We identified relevant articles published up to 22 October 2015 through searches of the literature in PubMed/MEDLINE, EMBASE, and the Cochrane Library. The following search strategy strings were used: (sucrose or fructose or sugar* or simple carbohydrate* ) and (cholesterol* or lipoprotein* or apolipoprotein* or LDL* or HDL* or systolic or triglyceride* or triacylglycerol* or diastolic or blood pressure ) and (trial or clinical trial or comparative study or randomized controlled trial ) for PubMed/Medline; sucrose or fructose or sugar or simple carbohydrate and (cholesterol or lipoprotein or apolipoprotein or LDL or HDL or systolic or triglyceride or triacylglycerol or diastolic or blood pressure ) and (trial or clinical trial or comparative study or randomized controlled trial ) for EMBASE; and (sucrose or fructose or sugar* or simple carbohydrate* ) and (cholesterol* or lipoprotein* or apolipoprotein* or LDL* or HDL* or systolic or triglyceride* or triacylglycerol* or diastolic or blood pressure ) for the Cochrane Library. We retrieved and assessed potentially relevant articles and checked the reference lists of all papers of interest to identify additional relevant publications. Studies were considered eligible if they 1) included original data from controlled dietary intervention trials that compared a diet containing a given amount of energy provided by free sugars (glucose, fructose, or sucrose) with a control diet for which the same amount of energy was provided by complex carbohydrates; 2) included interventions that lasted $2 wk; 3) provided mean values after the intervention and control diets for $1 of the parameters of interest (i.e., systolic or diastolic blood pressure, total cholesterol, LDL cholesterol, HDL cholesterol, triacylglycerols, apolipo A-I and B, or VLDL cholesterol) and provided corresponding measures of dispersion or sufficient data to derive them; 4) were conducted in humans; and 5) were published in English. We excluded trials for which there were no isocaloric substitutions of free sugars for other forms of carbohydrates and those for which the intervention and control diets were not isocaloric (i.e., energy intake was not strictly controlled). Abstracts and full-text articles were screened for inclusion by 2 independent reviewers, and disagreements were resolved by discussion with a third reviewer. Data collection and quality assessment For each selected study, we extracted information on authors, publication year, country, baseline characteristics of the study population (sex, age, BMI or body weight, and health status), number of subjects involved, study design ( or parallel), intervention and control diets (including the composition of carbohydrates in terms of free or complex sugars, types of free sugars, total energy intake, percentage of energy from sugars, and percentage of energy exchanged by the specific test sugar), duration of intervention, and body weights after the intervention and control diets. In addition, for each parameter of interest, we extracted the means after the intervention and control diets with corresponding SEs, SDs, or 95% CIs when available; we also extracted the mean differences between the 2 groups with the corresponding SEs, SDs, 95% CIs, P values, or t values. We assigned quality scores to the trials based on the risk of bias with the use of the Cochrane Collaboration tool (14). Statistical analysis We compared a diet containing a given amount of energy provided by free sugars with diets for which the same amount of energy was provided by complex carbohydrates. In a few studies that considered different types of free sugars (i.e., sucrose, fructose, or glucose), these sugars were combined into a single group (15, 16); in other studies (17 20), data from different intervention groups with similar diets were combined. For the analyses, total cholesterol, LDL cholesterol, HDL cholesterol, triacylglycerols, and VLDL cholesterol were converted into milligrams per deciliter; apolipo were converted into milligrams per liter. For each study and parameter of interest, we computed the mean difference between the intervention and control diets and the corresponding SE if not already available in the original study; in this case, for studies, we adjusted the SE of the mean difference to take into account the nonindependence of the comparison groups with the use of P or t values for this difference (21). If a P value was reported as nonsignificant, we imputed a value of 0.5. In a few studies (15, 16, 18, 22) that did not report any P or t values, this adjustment was not made; similarly, adjustments were not possible when the mean difference was zero. We pooled study-specific estimates to calculate the weighted mean difference (WMD) of the parameters of interest between the comparison diets with the use of both fixed- and randomeffects models (23, 24). Only the results from the latter, however, are presented to take into account the heterogeneity of the effect estimates (this being more conservative). We assessed heterogeneity between studies with the use of the chisquare test and defined significant heterogeneity as P, 0.10 (23) and quantified the heterogeneity with the use of the I 2 statistic (25), which represents the percentage of the total

3 44 FATTORE ET AL. variation across studies attributable to heterogeneity rather than chance. Subgroup analyses and meta-regression models were conducted on a priori defined covariates such as age, sex, health status, body weight, study area, study design, duration of the intervention, total energy from diet, percentage of energy exchanged by the test sugar, and source of funding. A sensitivity analysis was conducted by removing 1 study at a time and removing studies at a higher risk of bias. For each parameter, we provided forest plots, in which a square was plotted for each study, whose projection on the underlying scale corresponds to the study-specific mean difference. The area of the square is proportional to the inverse of the variance of the mean difference and thus gives a measure of the amount of statistical information available. A diamond was used to plot the summary WMD and the corresponding 95% CI. Publication bias was evaluated by visually inspecting funnel plots and quantified by the Egger s and Begg s tests (26, 27). For selected parameters, we used the trim-and-fill method to correct for funnel plot asymmetry arising from publication bias (23, 24). STATA version 13 (StataCorp LP) was used for all statistical analyses. RESULTS Literature flow We identified 2670 records from the original literature search: 976 from PubMed, 828 from Embase, and 866 from the Cochrane Library. After excluding duplicate records, we screened 1675 citations. Of these, 180 were considered of interest based on their title and abstract, and their full texts were retrieved for detailed evaluation. Eight additional studies were identified from the reference list of the retrieved papers. Sixty papers were considered potentially eligible for the meta-analysis. After a closer evaluation, 32 studies were additionally excluded for different reasons (Supplemental Table 1), thus leaving 28 studies for this review and meta-analysis (Figure 1). Trial characteristics The studies involved 510 volunteers (51% women) aged y. Twenty-eight studies compared free sugars with starch (15 20, 22, 28 47) and 1 with maltodextrin (48); 22 had a design (15 20, 22, 28 32, 34 36, 38 42, 45, 47), and 6 were FIGURE 1 Flowchart of the study selection for the systematic review and meta-analysis.

4 CARDIOVASCULAR EFFECTS OF FREE SUGARS 45 TABLE 1 Main characteristics of the studies included in the review and meta-analysis 1 First author, year (ref) Country Study design No. and sex of subjects age, y Health status BW, kg Intervention diet Control diet Study duration Groen, 1966 (28) Antar, 1970 (29) Birchwood, 1970 (30) Little, 1970 (17) Mann, 1972 (31) Grande, 1974 (18) Reiser, 1979 (32) Israel Crossover 5 men; 10 women Canada Crossover 13 men; 2 women Canada Crossover 10 men; 1 woman Canada Crossover 2 men; 2 women 40.2 Healthy Sucrose: 66.4% CHOs, 14.6%, 18.9% fats 51 Hyperlipoproteinemic (type II V) with CVD complications 42.4 Hyperlipoproteinemic (type II V) with CAD complications 34.8 Hyperlipoproteinemic (type II) with CVD complications Canada Crossover 3 men 54.7 Hyperlipoproteinemic (type III and IV) with CVD complications Canada Crossover 2 men 38.5 Hyperlipoproteinemic South Africa Randomized (experiment I) (experiment II) United States Crossover 10 men; 9 women (type V) with CVD complications 9 men Normolipidemic subjects admitted to hospital for nonmetabolic conditions 71.1 Sucrose: 50% CHOs (40% sucrose), 15%, 35% fats, 2% fibers 66.5 Sucrose: 50% CHOs (41% sucrose), 15%, 35% fats, 2% fibers Sucrose: 25% CHOs (20% sucrose), 65% fats, 10%, 1.2% fibers Sucrose: 20% CHOs (14% sucrose), 70% fats, 10%, 1.6% fibers Sucrose: 70% CHOs (4.5% starch and 54% sucrose), 10% fats, 20%, 1.6% fibers Sucrose: 55% CHOs (23% sucrose), 30% fats, 15% 12 men Healthy Sucrose: 46.1% CHOs (22.6% sucrose), 36.9 fats, 17% ; fruit: 45.8% CHOs (21.8% fructose), 37.2 fats, 17% 11 men Healthy Sucrose: 48.1% CHOs (18.8% sucrose), 34.7% fats, 17.2% Healthy Sucrose: 43% CHOs (30% sucrose), 42% fats, 15% Bread: 62.1% CHOs, 18.4%, 19.3% fats Starch: 50% CHOs (40% starch), 15%, 35% fats, 2% fibers Starch: 50% CHOs (40% starch and 1% sucrose), 15%, 35% fats, 1.5% fibers Starch: 25% CHOs (20% starch and 1% sucrose), 65% fats, 10%, 0.8% fibers Starch: 20% CHOs (14% starch and 1% sucrose), 70% fats, 10%, 0.8% fibers Starch: 70% CHOs (56.5% starch and 3.5% sucrose), 10% fats, 20%, 2.8% fibers Potato and rice: 55% CHOs (23% potato and rice), 30% fats, 15% Wheat flour: 46.3% CHOs (6% sucrose and 34% starch), 36.7 fats, 17% Wheat flour: 47.8% CHOs (2.7% sucrose and 33.2% starch), 4.4% wheat, 34.9% fats, 17.3% ; peas: 47% CHOs (2.7% sucrose and 34.1% starch), 5.9% peas, 35.4% fats, 17.6% ; garbanzos: 46.9% CHOs (2.8% sucrose and 34.1% starch), 5.6% garbanzos, 35.4% fats, 17.6% Starch: 43% CHOs (30% starch), 42% fats, 15% 5wk 4-wk intervention; 1 2-wk washout 2-wk intervention; 1-wk washout 2-wk intervention; 1-wk washout 2-wk intervention; 1-wk washout 2-wk intervention; 1-wk washout 2wk 2wk 2wk 6-wk intervention; 4-wk washout (Continued)

5 46 FATTORE ET AL. TABLE 1 (Continued ) First author, year (ref) Country Study design No. and sex of subjects age, y Health status BW, kg Intervention diet Control diet Study duration Reiser, 1981 (22) Israel, 1983 (15) Peterson, 1986 (19) Abraira, 1988 (33) Cooper, 1988 (34) Grigoresco, 1988 (35) Osei, 1989 (36) Porta, 1989 (37) Reiser, 1989 (20) United Kingdom United Kingdom Randomized Randomized parallel parallel Australia Randomized France Randomized Italy Randomized parallel 12 men; 12 women 12 men; 12 women 10 men; 2 women 7 men; 4 women Men: 38.6; women: 35.1 CHOs-sensitive Men: 77.8; women: 67.4 Sucrose 18%: 44% CHOs (18% sucrose), 42% fats, 15% ; sucrose 33%: 44% CHOs (33% sucrose), 42% fats, 15% 36.8 Hyperinsulinemic 72.8 Sucrose 33%: 44% CHOs (33% sucrose), 11% starch, 42% fats, 14% ; sucrose 18%: 44% CHOs (18%, sucrose), 26% starch, 42% fats, 14% Type 1 diabetes Type 2 diabetes Sucrose: 54% CHOs (18% simple sugars), 27% fats, 19% 9 men 61.4 Type 2 diabetes Sucrose: 50% CHOs (36% sucrose), 35% fats, 15% 8 men; 1 woman 6 men; 11 women 5 men; 3 women 5 men; 8 women 4 men; 4 women United States Crossover 10 men 11 men 61.4 Type 2 diabetes Sucrose: 50% CHOs (36% sucrose), 35% fats, 15% Type 2 diabetes Sucrose: 52% CHOs (28 g sucrose), 33% fats, 15% Type 2 diabetes Fructose: 50% CHOs (30 g fructose), 30% fats, 20% Type 2 diabetes Fructose: 50% CHOs (17% fructose), 35% fats, 15% 60 Type 2 diabetes Sucrose: 1200 kcal (10% sucrose), 1500 kcal (10% sucrose), 1850 kcal (10% sucrose) Hyperinsulinemic Nonhyperinsulinemic Fructose: 51% CHOs (20% fructose), 36% fats, 13% Sucrose 5%: 44% CHOs (5% sucrose and 39% starch), 42% fats, 15% Sucrose 5%: 44% CHOs (5% sucrose and 39% starch), 42% fats, 14% Control: 53% CHOs (10% simple sugars and 43% starch), 27% fats, 20% Starch: 50% complex CHOs, 35% fats, 15% Starch: 50% complex CHOs, 35% fats, 15% Starch: 52% CHOs (28 g starch), 33% fats, 15% Control starch: CHOs (30 g starch), 30% fats, 20% Complex CHOs: 50% CHOs (mostly complex), 35% fats, 15% Starch: 1200 kcal (10% starch), 1500 kcal (10% starch), 1850 kcal (10% starch) Starch: 51% CHOs (20% starch), 36% fats, 13% 6wk 6wk 6wk 4wk 4wk 6wk 8wk 24-wk intervention; 4-wk washout 24 wk 5wk Santacroce, 1990 (38) Bantle, 1992 (39) Italy Randomized 6 men; 6 women 7 men; 11 women 27 Type 1 diabetes Sucrose: 52.5% CHOs (30 g sucrose), 31% fats, 16% 45 Type 1 and 2 diabetes Fructose: 55% CHOs (17% fructose), 30% fats, 15% Control: 52.6% CHOs (mostly complex), 31% fats, 17% Starch: 55% CHOs (,3% fructose), 30% fats, 15% 8wk 4wk (Continued)

6 CARDIOVASCULAR EFFECTS OF FREE SUGARS 47 TABLE 1 (Continued ) First author, year (ref) Country Study design No. and sex of subjects age, y Health status BW, kg Intervention diet Control diet Study duration Swanson, 1992 (40) Bantle, 1993 (41) Koivisto, 1993 (42) Piatti, 1993 (43) Malerbi, 1996 (16) Surwit, 1997 (44) Black, 2006 (45) Vaisman, 2006 (48) Finland Double-blind randomized 7 men; 7 women 4men 8 women 4 men; 6 women Healthy Fructose: 265g CHOs (132 g starch and 100 g fructose), 71 g fats, 79 g Type 2 diabetes Sucrose: 55% CHOs (19% fructose), 30% fats, 15% Type 2 diabetes Fructose: 50% CHOs (20% fructose), 30% fats, 20% Italy Parallel 11 women 40 Obese Simple sugar: 15% complex Italy Parallel 1 man; 6 women Brazil Crossover 7 men; 9 women CHOs, 45% simple CHOs, 20% fats, 20% 40 Obese Simple sugar: 15% complex CHOs, 45% simple CHOs, 20% fats, 20% Type 2 diabetes High sucrose: 55% CHOs (19% sucrose. 35% polysaccharides, and 1% fructose), 35% fats, 15% High fructose: 55% CHOs (1% sucrose, 34% polysaccharides, and 20% fructose), 35% fats, 15% United States Parallel 20 women Overweight or obese High sucrose: 73.3% CHOs (43% sucrose), 10.8% fats, 18.7% United States Parallel 22 women Overweight or obese High sucrose: 73.3% CHOs (43% sucrose), 10.8% fats, 18.7% Ireland Randomized Israel Double-blind randomized parallel Israel Double-blind randomized parallel 14 men Healthy High sucrose: 55% CHOs (25% sucrose and 17% starch), 30 35% fats, 10 15% 5 men; 8 women 7 men; 5 women Starch: 262 g (201 g starch and 14 g fructose), 71 g fats, 79 g, Starch: 55% CHOs (52% bread and potatoes), 30% fats, 15% Complex CHOs: 50% complex CHOs, 30% fats, 20% Starch: 45% complex CHOs, 15% simple CHOs, 20% fats, 20% Starch: 45% complex CHOs, 15% simple CHOs, 20% fats, 20% High starch: 50% CHOs (4% sucrose, 45% polysaccharides, 1% fructose), 35% fats, 15% Low sucrose (starch): 70.9% CHOs (4% sucrose and 67% starch), 10.8% fats, 18.7% Low sucrose (starch): 70.9% CHOs (4% sucrose and 67% starch), 10.8% fats, 18.7% Low sucrose (starch): 55% CHOs (10% sucrose and 34% starch), 30 35% fats, 10 15% Type 2 diabetes Fructose: 7.5 g, 3 times/d Control: 7.5 g maltodextrin, 3 times/d Type 2 diabetes Fructose: 7.5 g, 3 times/d Control: 7.5 g maltodextrin, 3 times/d 4wk 4wk 4-wk intervention; 4-wk washout 3wk 3wk 4-wk intervention; 2-wk washout 6wk 6wk 6-wk intervention; 4-wk washout 4wk 4wk (Continued)

7 48 FATTORE ET AL. TABLE 1 (Continued ) BW, kg Intervention diet Control diet Study duration age, y Health status No. and sex of subjects Study design First author, year (ref) Country 6-wk intervention; 4-wk washout Sucrose 5%: 54.8% CHOs (37.6% starch and 5.2% sucrose), 32.9% fats, 12.3% Obese and overweight Sucrose 15%: 55% CHOs (24.7% starch and 14.9% sucrose), 32.8% fats, 12.1% 9 men; 4 women Randomized United Kingdom Lewis, 2013 (47) 8wk Starch: 45.3% CHOs (6.2% Fructose: 43.8% CHOs (32% fructose), 19.8 fats, 33.8% fructose), 17.8% fats, 37.7% 24 women Obese women with PCOS Norway Randomized parallel Johnson, 2015 (49) 8wk Starch: 45.3% CHOs (6.2% Fructose: 43.8% CHOs (32% fructose), 19.8 fats, 33.8% fructose), 17.8% fats, 37.7% 27 women Obese women with PCOS Norway Randomized parallel 1 BW, body weight; CAD, coronary artery disease; CHO, carbohydrate; CVD, cardiovascular disease; PCOS, polycystic ovary syndrome; ref, reference. 2 6 SE. 3 Range. 4 6 SD. parallel studies (33, 37, 43, 44, 48, 49). Studies measuring blood pressure lasted 6 wk, except for 1 study (42) that lasted 4 wk, whereas the duration of the intervention ranged from 4 (16, 33, 39 42) to 24 wk (36) for LDL cholesterol, 3 (43) to 24 wk (36) for HDL cholesterol, and 2 (17, 18, 29, 30, 50) to 24 wk (36, 37) for triacylglycerols. Fifteen studies (15, 18, 19, 22, 31, 34 36, 38 42, 45, 47) and 4 parallel studies (13, 33, 37, 48, 49) were randomized. Eleven studies were conducted in the United States (15, 18, 20, 22, 32, 33, 36, 39 41, 44); the remainder were conducted in various countries around the world. Two studies were conducted on women only (44, 49), 4 on men only (18, 20, 31, 45), and the others on both. Three studies were conducted on young people (mean age: #30 y) (18, 38, 49, 51), 15 on people aged y (15, 17, 20, 22, 28, 30 32, 35, 39, 40, 43 45, 47), and 10 on people.50 y (16, 17, 19, 29, 33, 34, 36, 37, 41, 42, 48). Six studies included healthy volunteers (18, 20, 28, 31, 32, 40), 10 were conducted on type 2 diabetic subjects (16, 19, 33 35, 37, 41, 42, 48, 52), 2 studies were conducted on type 1 diabetic subjects (19, 38), and 1 study included both type 1 and 2 diabetic subjects (39). Three studies were conducted on carbohydrate-sensitive people (15, 20, 22), 3 on people suffering from hyperlipoproteinemia or hyperlipidemia with cardiovascular complications (17, 29, 30), and 4 on obese subjects (43, 44, 47, 49). Energy provided by the diet ranged from 800 (43) to 3240 (20) kcal/d, and the energy exchanged by the intervention sugars and carbohydrates ranged from w5% (48) to 50% (17) of the total energy provided by the diet. Fifteen studies (16, 17, 28 31, 33 35, 39 41, 44, 45, 47) were funded by contributions from private companies and 5 by national and public research institutions (18, 36, 37, 42, 49). For 8 studies (15, 19, 20, 22, 32, 38, 43, 48), the financial source was not specified. The main characteristics of the studies considered are provided in Table 1. Cardiovascular disease risk factors Isoenergetic substitutions of free sugars for complex carbohydrates had no significant effects on systolic (WMD: 0.61 mm Hg; 95% CI: 21.64, 2.86 mm Hg) and diastolic blood pressure (WMD: 1.36 mm Hg; 95% CI: 20.45, 3.18 mm Hg) (Figure 2A, B). No heterogeneity was detected between studies. There was a significant increase in LDL (WMD: 7.13 mg/dl; 95% CI: 3.15, mg/dl) and HDL cholesterol (WMD: 1.32 mg/dl; 95% CI: 0.59, 2.05 mg/dl) (Figure 3A, B) when free sugars were substituted for complex carbohydrates. Significant heterogeneity was found for LDL (I 2 = 64.4%; P = ) but not HDL cholesterol. Serum triacylglycerols (Figure 4) showed a significant increase when free sugars were substituted for complex carbohydrates (WMD: 8.27 mg/dl; 95% CI: 3.04, mg/dl), with significant heterogeneity between studies (I 2 = 59.3%; P = ). Similarly, total cholesterol showed a significant increase (WMD: 9.26 mg/dl; 95% CI: 4.86, mg/dl), with heterogeneity between studies (I 2 = 70.6%; P = ). No significant differences were found for apolipo A-I and B, but only a few studies reported results for these biomarkers (Supplemental Figures 1 and 2) and VLDL cholesterol (Supplemental Figure 3). There was no effect on body weight (WMD: kg; 95% CI: 21.83, 1.35 kg) or heterogeneity

8 CARDIOVASCULAR EFFECTS OF FREE SUGARS 49 FIGURE 2 WMDs and corresponding 95% CIs in systolic (A) and diastolic blood pressure (B) (mm Hg) after isoenergetic exchange between free sugars and complex carbohydrates in dietary intervention trials. WMDs were calculated from random-effect models. A square was plotted for each study, the center projection of which on the underlying scale that corresponds to the study-specific WMD and the area is proportional to the inverse of the variance of the log WMD. The whiskers represent the 95% CI. A diamond was used to plot the overall WMD, the center of which represents the WMD and its extremes the corresponding 95% CI. ID, identifier; WMD, weighted mean difference. between studies when free sugars were substituted for complex carbohydrates (Figure 5). Subgroup analyses showed some differences across strata of percentages of energy exchanged, total energy intake, and body weight for total cholesterol, LDL cholesterol, and triacylglycerols, with somewhat larger increases when the percentage of energy exchanged with free sugars was.25% (Figure 6A), when total energy intake was.2500 kcal (Figure 6B), and for subjects who were normoweight (Figure 6C). The residual I 2 after accounting for the percentage of energy exchanged were 71.5%, 54.0%, and 59.6% for total cholesterol, LDL cholesterol, and triacylglycerols, respectively; thus, the percentage of energy exchanged explained only a small proportion of between-studies heterogeneity for these lipid biomarkers. When accounting for total energy intake, corresponding values were 51.7%, 43.6%, and 54.6%, and when accounting for body weight they were 59.5%, 60.6%, and 46.5%. Thus, a substantial part of heterogeneity (.50%) was explained only for total cholesterol by total energy diet. LDL cholesterol increases were larger in studies conducted in the United States (WMD: 9.95 mg/dl; 95% CI: 6.83, mg/dl) than in those from other countries (WMD: 3.13 mg/dl; 95% CI: 23.41, 9.67 mg/dl) and in subjects aged y (WMD: 9.84 mg/dl; 95% CI: 6.30, mg/dl) than in older volunteers (WMD: 2.74 mg/dl; 95% CI: 22.46, 7.93 mg/dl). Triacylglycerol

9 50 FATTORE ET AL. FIGURE 3 WMDs and corresponding 95% CIs in LDL (A) and HDL cholesterol (B) (mg/dl) after isoenergetic exchanges between free sugars and complex carbohydrates in dietary intervention trials. WMDs were calculated from random-effect models. A square was plotted for each study, the center projection of which on the underlying scale corresponds to the study-specific WMD and the area is proportional to the inverse of the variance of the log WMD. The whiskers represent the 95% CI. A diamond was used to plot the overall WMD, the center of which represents the WMD and its extremes the corresponding 95% CI. ID, identifier; WMD, weighted mean difference.

10 CARDIOVASCULAR EFFECTS OF FREE SUGARS 51 FIGURE 4 WMDs and corresponding 95% CIs in serum triacylglycerols (mg/dl) after isoenergetic exchange between free sugars and complex carbohydrates in dietary intervention trials. WMDs were calculated from random-effect models. A square was plotted for each study, the center projection of which on the underlying scale corresponds to the study-specific WMD and the area is proportional to the inverse of the variance of the log WMD. The whiskers represent the 95% CI. A diamond was used to plot the overall WMD, the center of which represents the WMD and its extremes the corresponding 95% CI. ID, identifier; WMD, weighted mean difference. increases were also larger in studies in the United States (WMD: mg/dl; 95% CI: 6.04, mg/dl) than in other countries (WMD: 1.30 mg/dl; 95% CI: 23.82, 6.41 mg/dl) and in studies (WMD: 10.8 mg/dl; 95% CI: 5.07, 16.3 mg/dl) than in parallel studies (WMD: mg/dl; 95% CI: 222.5, 5.64 mg/dl). A subgroup analysis for triacylglycerols also showed that in all subgroups for which heterogeneity was no longer significant, the effect on serum triacylglycerols lost significance. Finally, nonsignificant effect modification was observed for the funding source or duration of the intervention, except for increases in LDL cholesterol and triacylglycerols in studies with shorter (,6 wk) than longer ($6 wk) intervention duration (data not shown). Meta-regression models showed that after adjusting for all covariates, only the effects of the percentage of energy exchanged and total dietary energy on total cholesterol were still significant (data not shown). The meta-analysis of fructose compared with complex carbohydrates (Supplemental Figures 4 10) and sucrose compared with complex carbohydrates (Supplemental Figures 11 17) showed that fructose and sucrose had similar effects on blood pressure and other cardiovascular biomarkers; only VLDL cholesterol was significantly higher when fructose was substituted for complex sugars, although only in 2 studies (Supplemental Figure 8). Sensitivity analysis and risk of bias Sensitivity analyses in which 1 study at a time was removed did not substantially change the effect estimates for any of the cardiovascular markers analyzed. Results of quality assessment by the Cochrane Collaboration tool identified 13 studies (15, 17, 19, 20, 22, 28 30, 32, 34, 37, 38, 43) with a high risk of bias (Supplemental Table 2). A sensitivity analysis when these studies were excluded showed that the effect for triacylglycerols was no longer significant (WMD: 0.07 mg/dl; 95% CI: 21.47, 1.61 mg/dl), and there was no heterogeneity between studies (Figure 7A). Similarly, when no randomized studies were excluded, the effect for triacylglycerols (WMD: 2.97 mg/dl; 95% CI: 20.84, 6.77 mg/dl) and heterogeneity between studies (I 2 = 23.6%; P = 0.17) lost significance (Figure 7B). Publication bias was detected for triacylglycerols (Supplemental Figures 18 and 19) and LDL cholesterol (Supplemental Figure 20). For triacylglycerols, the trim-and-fill method estimated, respectively, 8 and 6 missing studies among those that compared free sugars with complex carbohydrates and sucrose with complex carbohydrates, and for LDL cholesterol 5 missing studies among those that compared sucrose with complex carbohydrates. After adjusting for asymmetry, the increases in triacylglycerols caused by free sugars (WMD: 1.57 mg/dl; 95% CI: 24.18, 7.33 mg/dl) and

11 52 FATTORE ET AL. FIGURE 5 WMDs and corresponding 95% CIs in body weight (kg) after isoenergetic exchange between free sugars and complex carbohydrates in dietary intervention trials. WMDs were calculated from random-effect models. A square was plotted for each study, the center projection of which on the underlying scale corresponds to the study-specific WMD and the area is proportional to the inverse of the variance of the log WMD. The whiskers represent the 95% CI. A diamond was used to plot the overall WMD, whose center represents the WMD and its extremes the corresponding 95% CI. ID, identifier; WMD, weighted mean difference. sucrose (WMD: 2.83 mg/dl; 95% CI: 23.21, 8.87 mg/dl) lost significance, as did the increase in LDL cholesterol (WMD: 1.95 mg/dl; 95% CI: 22.77, 6.68 mg/dl). DISCUSSION This systematic review and meta-analysis shows that in nutritional intervention trials in which free sugars provided the same amount of energy as complex carbohydrates, there were no significant changes in systolic and diastolic blood pressure. Analyses of fructose and sucrose separately found no meaningful differences between the 2 sugars. These results are consistent with those of a recent systematic review and meta-analysis of prospective cohort studies by Jayalath et al. (53), who found no association between fructose intake and hypertension. However, they did not confirm a previous meta-analysis that showed a significant reduction in diastolic blood pressure when fructose was exchanged for other carbohydrates in isocaloric trials (54). This might be explained by the low statistical power in our results because the meta-analysis included few studies that compared fructose with complex carbohydrates because of different inclusion criteria. In other studies, both in humans (55 57) and animals (58), fructose has been reported to increase blood pressure. However, these increases occurred at relatively high doses, i.e., w40 60% of dietary total energy, whereas the doses in the intervention diets included in our meta-analysis were generally,20%. Serum cholesterol showed significant increases when simple sugars were substituted for complex carbohydrates. However, although the effect of simple sugars on HDL is clear, the evidence of an effect on LDL cholesterol is less clear because residual heterogeneity persists despite comprehensive subgroup analyses. In addition, there was evidence of publication bias when sucrose was substituted for complex carbohydrates. For serum triacylglycerols, the effect of free sugars is not convincing for several reasons: a significant effect was found but with significant heterogeneity, the adjusted estimate to correct for publication bias was not significant, and the results that excluded studies with a higher risk of bias or that included only randomized controlled trials, which provide the highest level of evidence, showed no significant effect or heterogeneity. The strongest effect modification on blood lipid biomarkers was for the percentages of energy exchanged in the intervention and for total dietary energy intake, showing that the effect of free sugars on these biomarkers is affected both by dose and overall caloric intake. However, these 2 covariates explain only a small part of the heterogeneity between studies for triacylglycerols, whereas total dietary energy explained a substantial part of the heterogeneity for total and LDL cholesterol. The effects of other covariates such as country, body weight, and age were probably driven by total energy intake because studies in the United States, those that include normoweight persons, and those that include people aged y provided diets higher in energy than those from outside the United States that involved overweight, obese, and even elderly people. The effect of the study design for triacylglycerols also probably

12 CARDIOVASCULAR EFFECTS OF FREE SUGARS 53 FIGURE 6 Results of the stratified analysis according to percentage of energy exchanged (A), total energy intake (B), and body weight (C) showing the WMDs and corresponding 95% CIs for total cholesterol, LDL cholesterol, and triacylglycerols after isoenergetic exchange between free sugars and complex carbohydrates in dietary intervention trials. WMDs were calculated from random-effect models. A square was plotted for the overall WMD within each subgroup, the center projection of which on the underlying scale corresponds to the overall WMD and the area is proportional to the inverse of the variance of the log WMD. The whiskers represent the 95% CI. Arrows were used in case of extremely wide 95% CI. WMD, weighted mean difference. resulted from total energy intake because almost all parallel studies (showing no increase in serum triacylglycerols) provided diets with,2000 kcal/d. We found no major differences in the effect of individual free sugars on other outcomes considered, except for an increase of VLDL cholesterol when fructose was substituted for complex

13 54 FATTORE ET AL. FIGURE 7 Results of the sensitivity analysis that excluded studies with a higher risk of bias (A) and included only randomized intervention trials (B) showing the WMDs and corresponding 95% CIs for total cholesterol, LDL cholesterol, HDL cholesterol, and triacylglycerols after isoenergetic exchange between free sugars and complex carbohydrates in dietary intervention trials. WMDs were calculated from random-effect models. A square was plotted for the overall WMD within each subgroup, the center projection of which on the underlying scale corresponds to the overall WMD and the area is proportional to the inverse of the variance of the log WMD. The whiskers represent the 95% CI. Arrows were used in case of extremely wide 95% CI. WMD, weighted mean difference. carbohydrates. However, this result should be interpreted with caution because it was based on only 2 studies (20, 42). Thus, further investigation is needed. Results on body weight confirmed that in nutritional intervention trials in which free sugars provided the same amount of energy as complex carbohydrates no changes occurred, with consistent study estimates. The absence of heterogeneity between studies indicates that there is no effect of any covariate on body weight. This confirms reports from other systematic (11, 12) and critical reviews (59, 60) of intervention trials in which the caloric intake was the main determinant of weight gain and is consistent with observational studies that did not show any effect of free sugars on body weight after adjusting for energy intake (61). This adjustment, also in relation to age, sex, and physical activity, is a critical step lacking in most studies that evaluate the effects of single isolated nutrients. This meta-analysis has some limitations and strengths. The limitations include those inherent to the original studies, most of which were of small sample size, of short duration, and supported by private companies and therefore potentially affected by conflicts of interest. However, the subgroup analysis for the duration of the intervention and source of funding did not indicate that these covariates had any real role in underestimating the effects. Another limitation of this meta-analysis is the lack of a validated quality scoring system for studies. The tools of the Cochrane Collaboration developed for clinical trials are not completely suitable for nutritional trials, which are mostly conducted on freeliving volunteers following dietary advice. This could have weakened the results of the quality assessment. A strength of this meta-analysis is the stringent eligibility criteria that excluded studies without a control group (i.e., sequential studies), studies in which the isocaloric balance was obtained from other macronutrients (e.g., fats) in addition to complex carbohydrates, and studies in which the isoenergetic balance was not stringent. As a consequence, only a limited number of studies were included in the final meta-analysis for some outcomes (e.g., apolipo and VLDL cholesterol), so we may not have detected any effect because of the low statistical power. Overall, this meta-analysis indicates that free sugars when substituted for complex carbohydrates in isocaloric intervention trials increase both HDL and LDL cholesterol, although the estimate for LDL cholesterol was affected by heterogeneity. Even

14 CARDIOVASCULAR EFFECTS OF FREE SUGARS 55 if an effect of free sugars on triacylglycerols cannot be excluded, the evidence of an increase is not convincing. The evidence of publication bias further weakens the effect estimates for LDL cholesterol and triacylglycerols and raises the key question of how unpublished negative results impair the scientific evidence and thus drive opinions that are not proven. Finally, there is no evidence that free sugars, when isoenergetically substituted for complex carbohydrates, increase blood pressure, and there is clear evidence that they do not increase body weight. In conclusion, this meta-analysis does not establish whether the isoenergetic substitution of free sugars with complex carbohydrates improves cardiovascular disease risk factors. Rather, it highlights the limitations of the individual studies, their wide variability, and the heterogeneity of the treatments, all of which make it difficult to draw firm conclusions. Therefore, additional independent randomized intervention trials of adequate sample size and duration involving a wider and up-to-date set of blood lipid biomarkers (62, 63) are needed. The traditional biomarkers, i.e., total, LDL, and HDL cholesterol, do not seem to predict cardiovascular events and mortality (64, 65). Finally, this metaanalysis also highlights the importance of overall dietary caloric intake as a basic determinant of cardiovascular disease risk factors. Thus, dietary recommendations should consider the effects of single macronutrients only after accounting for total dietary energy. We thank JD Baggott for editorial assistance. The authors responsibilities were as follows EF, CA, and CB: designed the study; EF and CB: conducted the library search and wrote the manuscript; EF and FB: extracted and controlled the data; FB and CB: analyzed the data; EF: had primary responsibility for the final content; and all authors: read and approved the final manuscript. EF, CA, and CB have received honoraries for seminars, congress presentations, and didactic activities from Soremartec Italia, which might have an interest in this work. The remaining author reported no conflicts of interest related to the study. REFERENCES 1. Yudkin J. Sugar and disease. Nature 1972;239: Lustig RH, Schmidt LA, Brindis CD. Public health: the toxic truth about sugar. Nature 2012;482: Freeman R. Moderate evidence support a relationship between sugar intake and dental caries. Evid Based Dent 2014;15: Johnson RJ, Segal MS, Sautin Y, Nakagawa T, Feig DI, Kang DH, Gersch MS, Benner S, Sanchez-Lozada LG. Potential role of sugar (fructose) in the epidemic of hypertension, obesity and the metabolic syndrome, diabetes, kidney disease, and cardiovascular disease. Am J Clin Nutr 2007;86: Ludwig DS, Peterson KE, Gortmaker SL. Relation between consumption of sugar-sweetened drinks and childhood obesity: a prospective, observational analysis. Lancet 2001;357: Te Morenga LA, Howatson AJ, Jones RM, Mann J. Dietary sugars and cardiometabolic risk: systematic review and meta-analyses of randomized controlled trials of the effects on blood pressure and lipids. Am J Clin Nutr 2014;100: Bray GA, Nielsen SJ, Popkin BM. Consumption of high-fructose corn syrup in beverages may play a role in the epidemic of obesity. Am J Clin Nutr 2004;79: WHO. Sugars intake for adults and children. Geneva (Switzerland): WHO; Houston M, Minich DM. Fructose-containing sugars do not raise blood pressure or uric acid at normal levels of human consumption. J Clin Hypertens (Greenwich) 2015;17: Malik VS, Hu FB. Fructose and cardiometabolic health: what the evidence from sugar-sweetened beverages tells us. J Am Coll Cardiol 2015;66: Sievenpiper JL, de Souza RJ, Mirrahimi A, Yu ME, Carleton AJ, Beyene J, Chiavaroli L, Di Buono M, Jenkins AL, Leiter LA, et al. Effect of fructose on body weight in controlled feeding trials: a systematic review and meta-analysis. Ann Intern Med 2012;156: Te Morenga L, Mallard S, Mann J. Dietary sugars and body weight: systematic review and meta-analyses of randomised controlled trials and cohort studies. BMJ 2012;346:e Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gotzsche PC, Ioannidis JP, Clarke M, Devereaux PJ, Kleijnen J, Moher D. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration. PLoS Med 2009;6:e Higgins JP, Altman DG, Gotzsche PC, Juni P, Moher D, Oxman AD, Savovic J, Schulz KF, Weeks L, Sterne JA, et al. The Cochrane Collaboration s tool for assessing risk of bias in randomised trials. BMJ 2011;343:d Israel KD, Michaelis OE 4th, Reiser S, Keeney M. Serum uric acid, inorganic phosphorus, and glutamic-oxalacetic transaminase and blood pressure in carbohydrate-sensitive adults consuming three different levels of sucrose. Ann Nutr Metab 1983;27: Malerbi DA, Paiva ES, Duarte AL, Wajchenberg BL. Metabolic effects of dietary sucrose and fructose in type II diabetic subjects. Diabetes Care 1996;19: Little JA, Birchwood BL, Simmons DA, Antar MA, Kallos A, Buckley GC, Csima A. Interrelationship between the kinds of dietary carbohydrate and fat in hyperlipoproteinemic patients. 1. Sucrose and starch with polyunsaturated fat. Atherosclerosis 1970;11: Grande F, Anderson JT, Keys A. Sucrose and various carbohydratecontaining foods and serum lipids in man. Am J Clin Nutr 1974;27: Peterson DB, Lambert J, Gerring S, Darling P, Carter RD, Jelfs R, Mann JI. Sucrose in the diet of diabetic patients just another carbohydrate? Diabetologia 1986;29: Reiser S, Powell AS, Scholfield DJ, Panda P, Ellwood KC, Canary JJ. Blood lipids, lipo, apo, and uric acid in men fed diets containing fructose or high-amylose cornstarch. Am J Clin Nutr 1989;49: Elbourne DR, Altman DG, Higgins JP, Curtin F, Worthington HV, Vail A. Meta-analyses involving cross-over trials: methodological issues. Int J Epidemiol 2002;31: Reiser S, Bickard MC, Hallfrisch J. Blood lipids and their distribution in lipo in hyperinsulinemic subjects fed three different levels of sucrose. J Nutr 1981;111: Greenland S. Quantitative methods in the review of epidemiologic literature. Epidemiol Rev 1987;9: Higgins JPT, Green SB. Cochrane handbook for systematic reviews of interventions: version London: The Cochrane Collaboration; Higgins JP, Thompson SG. Quantifying heterogeneity in a metaanalysis. Stat Med 2002;21: Egger M, Davey Smith G, Schneider M, Minder C. Bias in metaanalysis detected by a simple, graphical test. BMJ 1997;315: Thornton A, Lee P. Publication bias in meta-analysis: its causes and consequences. J Clin Epidemiol 2000;53: Groen JJ, Balogh M, Yaron E, Cohen AM. Effect of interchanging bread and sucrose as main source of carbohydrate in a low fat diet on the serum cholesterol levels of healthy volunteer subjects. Am J Clin Nutr 1966;19: Antar MA, Little JA, Lucas C, Buckley GC, Csima A. Interrelationship between the kinds of dietary carbohydrate and fat in hyperlipoproteinemic patients. 3. Synergistic effect of sucrose and animal fat on serum lipids. Atherosclerosis 1970;11: Birchwood BL, Little JA, Antar MA, Lucas C, Buckley GC, Csima A, Kallos A. Interrelationship between the kinds of dietary carbohydrate and fat in hyperlipoproteinemic patients. 2. Sucrose and starch with mixed saturated and polyunsaturated fats. Atherosclerosis 1970;11: Mann JI, Truswell AS. Effects of isocaloric exchange of dietary sucrose and starch on fasting serum lipids, postprandial insulin secretion and alimentary lipaemia in human subjects. Br J Nutr 1972;27: Reiser S, Hallfrisch J. Michaelis OE 4th, Lazar FL, Martin RE, Prather ES. Isocaloric exchange of dietary starch and sucrose in humans. I. Effects on levels of fasting blood lipids. Am J Clin Nutr 1979;32: Abraira C, Derler J. Large variations of sucrose in constant carbohydrate diets in type II diabetes. Am J Med 1988;84: Cooper PL, Wahlqvist ML, Simpson RW. Sucrose versus saccharin as an added sweetener in non-insulin-dependent diabetes: short- and medium-term metabolic effects. Diabet Med 1988;5: Grigoresco C, Rizkalla SW, Halfon P, Bornet F, Fontvieille AM, Bros M, Dauchy F, Tchobroutsky G, Slama G. Lack of detectable deleterious effects on metabolic control of daily fructose ingestion for 2 mo in NIDDM patients. Diabetes Care 1988;11: