Meta-analyses of Total and Individual Fructose-containing Sugars and Incident Cardiometabolic Disease

Background: Fructose has become a focus of intense concern regarding its links to the obesity epidemic and increasing burden of cardiometabolic disease. There have been dozens of editorials, commentaries, and letters in the scientific literature and numerous pieces in the lay and social media calling for efforts to restrict its intake and even regulate it like tobacco or alcohol. Uncontrolled ecological analyses which have linked increasing fructose intake with increasing obesity, diabetes, and hypertension rates and animal models of fructose induced metabolic syndrome and hypertension, which overfeed fructose at levels of exposure far beyond actual population levels of intake, have been used to underpin this debate. Evidence from well-adjusted prospective cohort studies also suggest a positive association between the consumption of sugar-sweetened beverages and increased energy consumption and weight gain. But not all meta-analyses of the available prospective cohort studies have supported this conclusion for SSBs, and no meta-analyses have investigated the effect of total fructose-containing sugars which also include grain and fruit sources on incident overweight/obesity, diabetes, metabolic syndrome, hypertension, gout, and cardiovascular disease. Despite the limitations in extrapolating from the available observational data and their inconsistency with data from controlled trials in humans (the highest level of evidence used in evidence based medicine) which do not show any adverse metabolic effects under isocaloric feeding conditions, the heart and diabetes associations have taken a risk reduction approach to added fructose-containing sugars, setting highly restrictive upper thresholds for their intake to achieve and maintain healthy body-weights and avoid adverse lipid effects.

Objective: To improve the evidence on which recommendations and public health policy are based, we will conduct a series of systematic reviews and meta-analyses of the role of fructose-containing sugars in the development of cardiometabolic disease in prospective cohort studies. A total of 5 analyses are proposed: (1)overweight/obesity, (2) diabetes/metabolic syndrome, (3) hypertension, (4) gout, and (5) coronary heart disease (CHD).

Design: The planning and conduct of the proposed meta-analyses will follow the Cochrane handbook for systematic reviews of interventions. The reporting will follow the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines.

Data sources. MEDLINE, EMBASE, CINAHL and The Cochrane Central Register of Controlled Trials (Clinical Trials; CENTRAL) will be searched using appropriate search terms, supplemented by manual, hand searches of bibliographies.

Study selection: We will include prospective cohort studies investigating the relation of fructose-containing (fructose, sucrose, and HFCS) sugars to incident overweight/obesity, diabetes, metabolic syndrome, hypertension, gout, and cardiovascular disease.

Data extraction. Two investigators will independently extract information about study design, sample size, subject characteristics, fructose form, fructose exposure levels, duration/person-years of follow-up, background diet profile, adjustments of models. Risk ratios for clinical outcomes will be extracted or derived from clinical event data across quantiles of exposure. Risk of bias will be assessed using the Cochrane Risk of Bias tool.

Outcomes: Each of the 5 proposed analyses will assess a different cardiometabolic disease outcome: (1) overweight/obesity, (2) diabetes/metabolic syndrome, (3) hypertension, (4) gout, and (5) CHD.

Data synthesis. The natural log-transformed relative risks of clinical outcomes comparing the highest exposure level to the reference group from each cohort will be pooled using the generic inverse variance method with random effects models. Heterogeneity will be assessed by Cochrane's Q and quantified by I2. Sensitivity analyses and a priori subgroup analyses will be undertaken to explore sources of heterogeneity including the effect of underlying disease status, sex, sugar type (fructose, sucrose, HFCS), follow-up (<10-years, >=10-years), level of adjustment of models, and Cochrane risk of bias on the effect of fructose. Significant unexplained heterogeneity will be investigated by additional post hoc subgroup analyses. Meta-regression analyses will assess the significance of subgroups analyses. Dose-response analyses will be undertaken using random-effects generalized least squares trend estimation models (GLST), appropriate for weighted regression of summarized dose-response data with dependent components(i.e. the reference exposure level). If insufficient evidence of a linear relationship is found, then we will do spline curve modeling (the MKSPLINE procedure) to characterize segments of the dose response curve where a linear approximation best describes the data. Publication bias will be assessed by the inspection of funnel plots and using Begg's and Egger's tests.

Knowledge translation plan: The results will be disseminated through interactive presentations at local, national, and international scientific meetings and publication in high impact factor journals. Target audiences will include the public health and scientific communities with interest in nutrition, diabetes, obesity, and cardiovascular disease. Feedback will be incorporated and used to improve the public health message and key areas for future research will be defined. Applicant/Co-applicant Decision Makers will network among opinion leaders to increase awareness and participate directly as committee members in the development of future guidelines.

Preliminary findings: To address the uncertainties in the evidence, we conducted a series of Canadian Institutes of Health Research (CIHR) funded systematic reviews and meta-analyses of controlled feeding trials of the effect of fructose on cardiometabolic risk (ClinicalTrials.gov registration number: NCT01363791). We found that fructose in isocaloric substitution for other sources of carbohydrate (isocaloric trials) does not increase body weight, lipids, blood pressure, uric acid, or insulin and even improves glycemic control. There was, however, a signal for harm under certain conditions. High doses of fructose increased triglycerides in isocaloric trials, and fructose providing excess energy at extreme doses relative to control diets (hypercaloric trials) also increased body weight, triglycerides, and uric acid. The implications of these findings for "real world" dietary advice, however, were complicated by several factors. First, fructose is not commonly consumed in isolation as a sweetener. Sucrose and HFCS are the primary fructose-containing sweeteners in the U.S. diet. Second, the level of fructose exposure in the available trials was well above population levels of intake, exceeding the 95th-percentile for U.S. intake in most of the isocaloric trials and in all of the hypercaloric trials, in which the excess energy brought by fructose was an important source of confounding. Finally, the available trials investigated effects on biomarkers of disease and not clinically meaningful events. The proposed systematic review and meta-analyses of prospective cohort studies will address these limitations directly by investigating the relation of self-reported, "real world" intakes of all fructose-containing sugars (fructose, sucrose, and HFCS) to the development overweight/obesity, diabetes/metabolic syndrome, hypertension, gout, and cardiovascular disease.

Significance: The proposed project will aid in knowledge translation related to the effects of dietary fructose on overweight/obesity, diabetes/metabolic syndrome, hypertension, gout, and cardiovascular disease, strengthening the evidence-base for recommendations and improving health outcomes through informing consumers and guiding future research.