Do Fructose-Containing Sugars Lead to Adverse Health Consequences? Results of Recent Systematic Reviews and Meta-analyses^,,

Abstract

Sugars have replaced fat as the dominant public health nutrition concern. A fructose-centric view of cardiometabolic disease has emerged whereby fructose-containing sugars are thought to have deleterious effects on body weight, fasting and postprandial blood lipids, glycemia, blood pressure, uric acid, and markers of nonalcoholic fatty liver disease. Long-term prospective cohort studies have not supported these associations when assessing the relation between total fructose-containing sugars at any amount of intake and incident cardiometabolic disease. Conversely, a consistent signal for harm has been reported for sugary beverages when comparing the highest with the lowest intakes. These associations, however, do not hold at moderate intakes, which are more reflective of real-world intakes, are subject to important collinearity effects, and have small risk estimates with modest population-attributable risk fractions. Higher-level evidence from controlled feeding trials shows that fructose-containing sugars in either liquid or solid form have adverse cardiometabolic effects only when they supplement diets with excess calories compared with the same diets without the excess calories. In the absence of harm when fructose-containing sugars are exchanged for other sources of carbohydrate under energy-matched conditions, excess calories appear to be the dominant consideration. Like with the earlier fat story, it is difficult to separate the contribution of fructose-containing sugars from that of other sources of excess calories in the epidemic of obesity and cardiometabolic disease. Attention needs to remain focused on reducing the overconsumption of all caloric foods associated with obesity and cardiometabolic disease, including sugary beverages and foods, and promoting greater physical activity.

Introduction

Dietary fat has recently been replaced by sugars as the dominant concern in public health nutrition. The original debate between fat and sugar first occurred in the 1960s and 1970s between Professor Ancel Keys, who published the landmark Seven Countries Study (1) and Professor John Yudkin, who published the popular book entitled Pure White and Deadly (2). It was the publication of the Seven Countries Study (1) that shifted attention to the role of fat in chronic disease and shaped low-fat dietary advice for the next 40 y. Over the past decade the focus has shifted back to sugars and, in particular, added fructose-containing sugars and their main proxy, sugary beverages (3). Although experimental models offer plausible biochemical mechanisms to support this fructose-centric view of cardiometabolic disease, whereby the fructose moiety of sugars acts as an unregulated substrate for de novo lipogenesis, increases uric acid concentrations and impairs satiety signals (through insulin, leptin, and ghrelin) (4), the mechanisms by which fructose-containing sugars affect obesity and cardiometabolic disease may relate more to a combination of hedonic pathways, whereby sugary foods are overconsumed because of their high palatability or to incomplete compensation of liquid calories from sugary beverages (5, 6). The evidence that any of these mechanisms operate at “real world” levels of exposure in free-living people, however, requires careful examination. We assessed the highest level of evidence from prospective cohort studies and controlled dietary trials to help answer the question of whether fructose-containing sugars are uniquely contributing to the epidemics of obesity and cardiometabolic disease.

Sugars and Cardiometabolic Risk in Prospective Cohorts

Prospective cohort studies currently provide the strongest evidence of a link between fructose-containing sugars and incident cardiometabolic disease. Drawing inference from these studies is complicated by the format in which fructose-containing sugars are consumed. These studies failed to show a consistent relation of total fructose-containing sugars in the diet with BMI (7) or incident cardiometabolic disease, including diabetes (8), hypertension (9), and coronary heart disease (CHD) (10). Gout is the one exception in which a relation has been seen (11). In contrast, a consistent relation for harm was reported when assessing the association between sugary beverages and cardiometabolic diseases. A WHO-commissioned systematic review and meta-analysis of 38 prospective cohort studies showed evidence of a significant association between sugary beverages and the risk of overweight/obesity (7). Another systematic review and meta-analysis of 11 studies showed a similar association with diabetes and metabolic syndrome (12). Individual cohort studies also showed a significant relation of sugary beverages with higher BMI (13), hypertension (14), gout (11), CHD (15), and stroke (16). Although larger effect estimates were seen where energy-unadjusted models were preferred, these associations remained significant even after adjustment for energy. The one exception has been for BMI in children and adolescents (17, 18).

It is unclear, however, why a significant association is found for sugary beverages but not total fructose-containing sugars. One explanation may relate to the food matrixes in which fructose-containing sugars are found. Fructose-containing sugars that are found in fluid form may be less sufficiently compensated for energy at subsequent meals, thus leading to increased overall energy intake and weight gain (5, 6). Although this hypothesis is physiologically possible, it remains unsupported. A recent systematic review of acute feeding trials found a similar lack of energy compensation among liquid, semisolid, and solid preloads after 120 min (5). Furthermore, the few chronic feeding trials that attempted to test whether this mechanism translates into meaningful weight gain over the longer term also found no clinically meaningful differences between sugary beverages and snack foods (19, 20). Therefore, it remains unclear whether sugary beverages will lead to decreased energy compensation more than any other highly palatable snack foods and greater weight gain over the longer term under free-living conditions. Another explanation for the differences may relate to the contribution of fructose-containing sugar intake from nutrient-dense fruit and vegetables as well as whole-grain products. Both of these important sources of fructose-containing sugars were associated with weight loss and improved metabolic outcomes in large prospective cohort studies (21–23) and randomized dietary trials (24, 25) and, as such, may balance any harm associated with fructose-containing sugars.

Sugars Compared with Other Major Risk Factors

What is the significance of these associations between sugary beverages and cardiometabolic diseases to public health? Strong parallels are being drawn between fructose-containing sugars and tobacco, with the implication that sugary beverages (as the main proxy for added sugars) are to obesity and cardiometabolic disease as tobacco smoke is to lung and cardiovascular disease (4). Is there justification for this view? Pooled estimates from meta-analyses for the RR of developing overweight/obesity, diabetes, and metabolic syndrome are <1.3 when comparing the highest with the lowest levels of exposure for sugary beverages in energy-adjusted models and even smaller in energy-unadjusted models (12). These risk estimates are small in comparison with other established risk factors for chronic diseases and do not remain significant at moderate levels of exposure near the 50th percentile of intake in the United States. For example, the RR for diabetes comparing the highest with the lowest level of adiposity (as BMI, waist circumference, or waist-to-hip ratio) is as high as 12 in men and 39 in women, with the association being significant at all levels (26). If we invoke the popular analogy with cigarette smoking and lung cancer, then the RR estimates are even higher. A pooled analyses of 5 prospective cohort studies [including many of the same cohorts that looked at the relation of sugar-sweetened beverage (SSB) exposure with incident diabetes] showed that the RR for lung cancer at the highest level of exposure (>40 cigarettes/d) was 43 in men and 60 in women (27). The associations for lung cancer are also significant at all levels of exposure, such that the RRs are >10-fold higher for lung cancer at the lowest level of exposure [where ∼19% of the population have this level of exposure or greater (28)] than for sugary beverages at the highest level of exposure for any cardiometabolic outcome (where a similar proportion of people, ∼20%, have this level of exposure among the same pooled cohorts). The question becomes whether the small effect sizes for the associations between SSBs and incident cardiometabolic diseases are magnified at population levels of exposure. One way to examine this question is to look at population-attributable risk modeling. In a pooled analysis that assessed the population risk of the burden of disease attributable to 67 risk factors, smoking was ranked second, accounting for 6.3% of the disease burden in the world, whereas sugary beverages were ranked 32nd, accounting for <1.1% of disease burden in the world (29). Compared with smoking, both the risk estimates and population-attributable risk fractions for sugary beverages are small. Whether these small associations seen for sugary beverages will translate into meaningful public health consequences in the context of other risk factors that cluster with SSB intake remains to be seen.

Sugars Compared with Other Dietary Factors

The small RRs associated with sugary beverages suggest the involvement of multiple factors in chronic disease pathways. Numerous dietary factors other than sugary beverages have been associated with cardiometabolic risk. Diets high in salt and trans FAs and low in PUFAs and in fruit and vegetables are associated with higher population-attributable risk of diabetes and/or ischemic heart disease–related mortality (29, 30). A pooled analysis of 3 Harvard cohorts not adjusted for energy also showed that an increase in 1 serving of French fries (+1.52 kg), potato chips (+0.77 kg), unprocessed meat (+0.43 kg), processed meat (+0.42 kg), trans fat (+0.29 kg), or boiled, baked, or mashed potatoes (+0.26 kg) resulted in greater or similar weight gain than did intake of sugary beverages (+0.45 kg) for every 4 y of follow-up (Figure 1) (22). The contribution to the diet of these other foods is not trivial: the servings per day provided by unprocessed meats (0.6), potatoes (0.4), processed meats (0.2–0.4), and potato chips (0.1–0.2) were similar to those provided by sugary beverages (0.2–0.3) in the Harvard cohorts. Similarly, the RR estimates for the relation of sugary beverages with cardiometabolic outcomes were shown to be smaller than or overlap with those for smoking, lack of exercise, and other established dietary risk factors such as the intake of high dietary salt, low n–3 FAs, or high trans FAs (30). Taken together, these comparatively small effect sizes seen only at the extremes of intake suggest that sugary beverages are not uniquely contributing to cardiometabolic risk.

FIGURE 1.

Significant weight changes for an increase in the consumption of different food items in the HPFS, the NHS, and the NHS II for every 4 y of follow-up as reported in reference 22. Total follow-up was 20 y in NHS I and HPFS and 12 y in NHS II. Increased consumption is based on servings per day for all items except for trans fat (% of total energy) and fried foods (servings/wk). Values are pooled mean changes with 95% CIs adjusted for age, baseline BMI at the start of each 4-y interval, sleep duration, and changes in physical activity, smoking, alcohol use, television watching, and each additional food item. HPFS, Health Professionals Follow-Up Study; NHS, Nurses’ Health Study.

Sugars as a Marker of an Unhealthy Lifestyle Pattern

Food and lifestyle choices do not exist in a vacuum. If a food is added or taken away, it has downstream effects on other related food and lifestyle choices. An unintended consequence of reducing added sugars may be to increase other highly palatable energy-dense foods. There is an historical precedent for this concern in the United States. Low-fat dietary advice and a flood of low-fat foods resulted in a shift in the partitioning of macronutrients but did not result in a decrease in calories, with a resulting increase in the prevalence of overweight/obesity through the 1980s and 1990s (31). The same scenario appears to be playing out now for sugars. Coinciding with the emergence of added sugars as a dominant public health concern over the past decade, the intake of added sugars has decreased by almost 20%, whereas overall energy intake has remained high, because of reciprocal increases in other carbohydrates, protein, and fat intakes (32). Why, how, and to what extent this compensation takes place for sugars is unclear. There are important collinearity effects exerted by other dietary and lifestyle factors that are determinants of the quantity and quality of compensation. It becomes a real challenge to disentangle these complicated interactions among different foods and the dietary and lifestyle patterns they comprise under free-living conditions. For example, high consumers of sugary beverages tend to consume more calories, exercise less, smoke more, and have a poorer dietary pattern (14, 15), all of which can be difficult to measure and adjust for completely in observational studies. Dietary pattern analyses, which take advantage of this collinearity, showed that a Western dietary pattern (characterized by high intakes of processed meat, red meat, refined grains, French fries, sweets and desserts, and sugary beverages) is associated with cardiometabolic disease outcomes including weight gain and increased risk of diabetes, CHD, and mortality from CHD in energy-adjusted models (33–35). When comparing the highest with the lowest amount of intake of a Western dietary pattern, the effect sizes for weight gain (+7.03 kg over 8 y) (33), diabetes (RR: 2.56–2.93) (34), and CHD (RR: 1.46) (35) are larger than those reported previously for sugary beverages alone (17, 18), and these associations remained unchanged even after adjustment for sugary beverages [in the 2 analyses in which this adjustment was performed (33, 34)]. Taken together, high intake of sugary beverages may be a marker of an overall unhealthy lifestyle characterized by high energy intake, inactivity, and a poor dietary pattern, in which the components in aggregate contribute to a high risk of obesity and cardiometabolic disease.

Evidence from Controlled Trials of Sugars and Sugary Beverages

To address the possible confounders identified in prospective cohort studies, controlled dietary trials provide the greatest protection against bias. A series of previously published controlled feeding trials that assessed the effect of sugars on cardiometabolic risk were identified. These published studies can be categorized into 3 main types of trials: 1) substitution trials, in which added fructose–containing sugars are exchanged for other macronutrient sources under energy-matched conditions; 2) addition trials, in which fructose-containing sugars are added to the diets compared with the same diets alone without the excess sugars; and 3) subtraction trials, in which the energy from added fructose-containing sugars is displaced, either using water and/or noncaloric beverages, from background diets compared with the same background diets still containing the added fructose–containing sugars. These 3 study designs ask different questions, allowing one to separate the effects of the fructose-containing sugars from those of the calories they provide.

Evidence from systematic reviews and meta-analyses of substitution trials, in which added fructose-containing sugars were provided in isocaloric exchange for other carbohydrates, failed to show a signal for harm. A WHO-commissioned systematic review and meta-analysis assessing the effect of all added fructose-containing sugars on measures of adiposity in randomized trials showed that added fructose-containing sugars compared with other sources of carbohydrate under energy-matched conditions did not affect body weight (7). Another systematic review of the available trial evidence found a similar lack of effect of added sucrose in isocaloric exchange with starch at amounts up to 25% of energy on blood lipids, glucose, and insulin control (36). However, whether there is an effect when sucrose supplies >25% of calories is unclear because there are few trials that assessed the effects above this level. There is also insufficient evidence to draw reliable conclusions on the effect of substituting fructose-containing for other macronutrients such as fat and protein. In this regard, one trial was conducted to compare the substitution of milk with sugary beverages in children in a 4-mo trial (37) and in adults in a 6-mo trial (38). The findings from the trials showed that there were no body weight or body fat differences; however, there was an increase in ectopic liver and visceral fat in the adults at high intakes of SSBs (1 L/d providing 106 g sugars/d) in isocaloric comparison with milk (38). These preliminary data suggest a potential signal for harm where sugars in liquid form is substituted for milk at very high doses that are well above normal amounts of intake, where the average population intake of added sugars from sugary beverages is 30.2 g/d (32).

A different story has emerged for addition trials in which added fructose-containing sugars supplement diets with excess calories compared with the same diets alone without the excess calories. These trials showed a consistent signal for harm across different trial conditions. The same WHO-commissioned systematic review and meta-analysis mentioned previously (7) and 2 updated systematic reviews and meta-analyses (37, 39) showed that the supplementation of diets with excess calories (+150–530 kcal of extra energy) from sugary beverages resulted in significant weight gain over 3 wk to 24 mo in controlled feeding trials. The weight gain achieved was proportional to the degree of calorie supplementation, with partial compensation for the extra calories provided, such that the weight gain achieved was less than that which would be predicted by the extra calories (16). A similar increasing effect has been seen with TGs when energy from sugary beverages supplement background diets (38) or where added fructose-containing sugars provide extra calories in ad libitum substitutions with starch (36). However, in the absence of a comparator matched for calories in these trials, one cannot conclude that the weight gain would have been different for any other source of excess calories.

Subtraction trials, in which calories from added sugars is reduced in diets compared with the same diets still containing the added sugars, have failed to deliver the expected benefits. Three recent systematic reviews and meta-analyses all found a lack of weight gain or weight loss in subtraction trials in which calories from sugary beverages were displaced by using water and/or noncaloric beverages (7, 37, 40). Sensitivity analyses, however, did show improvements when analyses were restricted to adult participants (7) and overweight/obese participants (39) but not children (7, 40). Compensation under ad libitum conditions (whereby a decrease in energy from one food will tend to be freely compensated by replacement with energy from another food) may be one reason for the lack of overall benefit across groups. It may also explain why strategies to reduce calories from sugary beverages do not appear to differ from those of general weight-loss advice. The Choose Healthy Options Consciously Everyday (CHOICE) trial showed similar weight loss for interventions that reduced calories from high-fructose corn syrup–sweetened beverages using water or noncaloric sweetened (“diet”) beverages compared with an intervention providing general weight-loss information at 6 mo (41).

Evidence from Controlled Trials of Dietary Fructose

In the absence of an effect on cardiometabolic risk of fructose-containing sugars beyond their calories, it is unclear whether there is an effect that is particular to the fructose moiety of sugars to which all cardiometabolic harm has been ascribed. The question remains: do fructose’s unique set of biochemical, metabolic, and endocrine responses translate into meaningful increases in cardiometabolic risk? To help address this question, we previously performed a series of Canadian Institutes of Health Research–funded systematic reviews and meta-analyses of >50 controlled dietary trials in >1000 participants (clinicaltrials.gov identifier: NCT01363791). Two main types of trials were identified in these published meta-analyses, which allowed for the effect of fructose to be isolated from that of energy (59). These included the following: 1) substitution trials, in which fructose was exchanged for other carbohydrate sources under calorie-matched conditions (Figure 2A), and 2) addition trials, in which fructose was added to background diets providing excess calories compared with the same background diets without the excess calories (Figure 2B).

FIGURE 2.

Forest plots of summary estimates from recent meta-analyses of the effect of different fructose interventions on cardiometabolic endpoints in controlled dietary trials. The meta-analyses were grouped broadly on the basis of the interventions in question. (A) Fructose substitution trials (isocaloric comparison), in which fructose was exchanged for other carbohydrate sources under calorie-matched conditions, and (B) fructose addition trials (hypercaloric comparison), in which fructose-supplemented control diets with excess calories at high doses were compared with the same control diets alone without the excess calories. Summary estimates (diamonds) were derived from pooled trial-level data. To allow the summary estimates for each endpoint to be displayed on the same axis, MDs were transformed to SMDs. Pseudo 95% CIs for each transformed SMD were derived directly from the original MD and 95% CI. *Significant interstudy heterogeneity as assessed by the Cochran Q statistic and quantified by the I² statistic, P < 0.10. ALT, alanine aminotransferase; DBP, diastolic blood pressure; FBG, fasting blood glucose; FBI, fasting blood insulin; GBP, glycated blood protein; HDL-C, HDL cholesterol; IHCL, intrahepatocellular lipid; LDL-C, LDL cholesterol; MAP, mean arterial pressure; MD, mean difference; NAFLD, nonalcoholic fatty liver disease; SBP, systolic blood pressure; SMD, standardized mean difference; TC, total cholesterol.

Pooled analyses of the substitution trials showed that fructose in isocaloric exchange for other carbohydrates (even under positive energy balance or in fluid form) does not adversely affect cardiometabolic risk factors and even benefits glycemic control and blood pressure (42–45) (Figure 2A). Nevertheless, these same pooled analyses of addition trials showed that diets supplemented with fructose providing excess calories (+18–97% of energy) at extreme doses (+104–250 g/d), which are well above the 95th percentile for intake in the population (46), compared with the same diets alone (without the excess calories) increased body weight and uric acid concentrations (42–45) as well as fasting TGs, postprandial TGs, and nonalcoholic fatty liver disease markers (intrahepatocellular lipid, alanine aminotransferase), glucose, and insulin (45, 47–49) (Figure 2B). Formal tests of interaction suggested that these adverse cardiometabolic effects seen in the addition trials were more attributable to the excess calories than the fructose. A subset of 5 of the isocaloric trials included in our systematic reviews and meta-analyses (50–55) used excess energy diets (positive energy balance) in both the fructose and comparator (glucose) arms, so permitted the effect of fructose to be isolated from that of energy under matched, yet excess energy, feeding conditions. If we restrict our meta-analyses to these trials, there was no evidence of harm and even a possible blood pressure benefit of diets supplemented with fructose providing excess energy as long as the comparison with glucose-supplemented diets was matched for the excess energy. Individual trials, however, did show an increase in fasting TGs (53) and fasting blood glucose (51). Three earlier meta-analyses also suggested possible high-dose thresholds for harm for the effect of fructose on fasting lipids [>60 g/d (49) and >100 g/d (56–58)] and postprandial [>50 g/d (56)] TGs under energy-matched conditions, thresholds that could not be replicated in updated analyses on fasting lipids (49, 60) and postprandial TGs (48). Further trials are needed to assess whether ad libitum replacement of fructose-containing sugars with other sources of calories leads to similar, greater, or smaller increases in calorie intake, weight gain, and cardiometabolic risk.

Conclusions

Like with the earlier fat story, it is difficult to separate the contribution of fructose-containing sugars from that of other factors in the epidemic of obesity and cardiometabolic disease. Larger, longer, and higher-quality randomized trials are needed to directly investigate the role of fructose-containing sugars on disease outcome. However, because of their modest population-attributable risk fractions and lack of demonstrated harm compared with other sources of excess calories in the diet likely to replace them, public health interventions that solely target sugars are unlikely to be the solution. Sugary beverages and foods are one of many pathways to overconsumption. Other highly palatable foods such as savory snack foods, refined grains, potato products, and processed meats are also associated with overconsumption, leading to weight gain and cardiometabolic complications. Dietary patterns that bring these factors together have the greatest influence on weight gain and cardiometabolic risk and represent the best opportunity for successful interventions. Attention needs to remain focused on reducing overconsumption of all caloric foods associated with obesity and cardiometabolic disease, including sugary beverages and foods, and promoting greater physical activity.