Home > DARE Reviews > Fructose consumption and consequences...

PubMed Health. A service of the National Library of Medicine, National Institutes of Health.

Database of Abstracts of Reviews of Effects (DARE): Quality-assessed Reviews [Internet]. York (UK): Centre for Reviews and Dissemination (UK); 1995-.

Database of Abstracts of Reviews of Effects (DARE): Quality-assessed Reviews [Internet].

Fructose consumption and consequences for glycation, plasma triacylglycerol, and body weight: meta-analyses and meta-regression models of intervention studies

G Livesey and R Taylor.

Review published: 2008.

CRD summary

This review concluded that oral fructose can have beneficial effects on glycated haemoglobin and adverse effects on fasting and postprandial triacylglycerols, but these were dose dependent. Reductions of fructose intake may exchange one risk (dyslipidaemia) by another (dysglycaemia). The authors' conclusions reflect the limited evidence presented. Large studies of long duration in diabetics and hyperlipidaemics were not found.

Authors' objectives

To evaluate the effects of dietary fructose on glycated haemoglobin (HbA1c), plasma triacylglycerol and body weight, with emphasis on treatment dose.


MEDLINE via PubMed and Cochrane Central Register of Controlled Trials (CENTRAL) were searched from 1966 to June 2006; search terms were reported. Only English-language journals were included.

Study selection

Randomised and non-randomised controlled trials (analysed separately and in combination) that compared orally administered diets with fructose replacing other substrates on HbA1c, plasma triacylglycerol and body weight were eligible for inclusion. Studies were included if they could be assigned to one of three administration methods (as defined by the review) and specified dose and duration of treatment. Eligible studies were of healthy individuals or individuals with impaired fasting glucose or glucose tolerance, type 2 diabetes, hypertriacylglycerolaemia or other forms of hyperlipidaemia, or elevated risk of coronary heart disease. Studies that used high-fructose corn syrup were excluded.

Included studies varied by participant age, health status and body mass index/body weight. Treatment durations ranged from: 2 to 26 weeks, mostly ≤13 weeks, for HbA1c; 0.14 to 78 weeks, mainly ≤4 weeks, for fasting triacylglycerols; 0.05 to 6 weeks, mostly ≤1 week, for post-prandial triacylglycerols; and 0.57 to 26 weeks, mostly ≤10 weeks, for body weight. Some studies had type 1 diabetes or mixed type 1 and 2 diabetes. Some women were menopausal. Fructose doses ranged from 17.5g to 350g per day. Controls received either substitute available carbohydrate (glucose, sucrose, dextromaltose, starch or maltodextrin) or no substitute (subjects ate more diet). One study of fasting triacylglycerol control and intervention patients received moderate amounts of supplementary oils.

Only one reviewer [Author information] performed the study selection.

Assessment of study quality

Study quality was assessed using the Jadad scale (criteria on randomisation, blinding and drop-outs). Each study received a score from 0 to a maximum 3.

Two reviewers independently extracted information for the validity assessment before any disagreements were resolved through joint discussion. Jadad scores were machine calculated subsequently [Author information].

Data extraction

Differences in outcomes between diets with and without fructose were extracted or calculated, either as differences between treatment arms in the change from baseline to follow-up or the end of study measurement. Standard deviations and errors were extracted, calculated using exact t, p-values or 95% confidence intervals (CIs). The authors stated that standard error of differences (paired within subject) where not reported in original studies were imputed dependent on treatment mean, duration of treatment and reported values from other studies. This was rather than dropping studies and incurring a related bias [Author information].

Two reviewers independently extracted data. Any disagreements were resolved through joint discussion.

Methods of synthesis

Pooled weighted mean difference (WMD) and 95% confidence intervals (CIs) were calculated using fixed-effect meta-analysis or (if I2 was not 0%) random-effects meta-analysis. Significance of combined means was assessed using the z-test. Heterogeneity was assessed using the Q and I2 statistics and subgroup analysis. Heterogeneity was further investigated using a range of meta-regression techniques, fitted by restricted maximum likelihood. Combined study trend was assessed with the Knapp and Hartung t-test, and the between-study variance with the likelihood-ratio test. Correlation between dietary inputs was assessed using an F-ratio test to identify potential causes of concern. Results from the meta-regressions were reported as trends (betas) with 95% CIs. Departure of subgroups of different health or body weight status, sex, age, type of meal from the general trend was assessed my meta-analysis of residual differences and application of z-test. Publication bias was investigated using both funnel plots and trim-and-fill analyses. Pseudo 95% CIs were estimated using the z-scores.

Repeated measures were treated according to available data. Data from the last time point was used to maintain independence of observations when few of the studies reported repeats. When >20% of studies reported repeats, the data were pooled across repeats and located at the average time point. When most studies reported repeats, but the number of studies was limited, and the outcome was expected to vary by duration of treatment, intermediate time points were retained in the analysis to facilitate fitting of meta-regression models that included time as a predictor variable (i.e. epsilon correction for repeated measures was assumed =1 [Author information]). Sensitivity of trends to different methods of handing repeated measures was undertaken.

Results of the review

A total of 112 studies (42 reports) were included in the review: More than half (54%) of the studies were of cross-over design. A high proportion (35%) of these did not report randomisation. Most studies scored 1 or 2 on the 0 to 3 Jadad scale. Attrition was less than 20% in all studies. There were only two double blinded studies and one single blinded study.

HbA1c (eight studies comprising six cross-over and two parallel studies, totalling 106 participants, provided just 14 point estimates): Overall, the addition of fructose to the diet reduced HbA1c concentration regardless of the substitute used (glucose or starch). There was no data comparing fructose and sucrose. Fructose dose, duration of fructose treatment, and the severity of dysglycaemia interacted as a single determinant of the treatment effect size. Treatment effect was greatest in individuals with poor baseline glycaemic control (high HbA1c). Funnel plot and trim-and-fill analyses revealed no significant publication bias.

Fasting plasma triacylglycerol (FPTG) (60 studies, comprising 33 cross-over, 20 sequence, four sandwich and three parallel studies, totalling 589 participants): Overall, RCTs with 100g/day or less fructose had no significant effect on FPTG. Fructose intake up to 350g/day increased FPTG. Non-significant residual differences from a general trend with dose were reported for RCTs and studies in which randomisation was not reported (RCTs: ResD -0.09, 95% CI -0.39 to 0.21. Non-RCTs: ResD 0.06, 95% CI -0.14 to 0.28). There was evidence of statistical heterogeneity for RCTs (I2=0.93) and non-RCTs (I2=0.91). FPTG increased as fructose dose increased, but the size of the effect decreased with duration of treatment. Subgroups of males, females, healthy persons, type 1 diabetics, type 2 diabetics, hyperlipidaemics, normal weight, overweight and obese persons showed no significant departure from the general trend for the dose response. Funnel plots and trim-and-fill analyses revealed no significant publication bias.

Postprandial triacylglycerol (PPTG) (25 studies, comprising 17 cross-over, five sequence, two parallel and one sandwich studies, totalling 234 participants): There was a significant reduction in PPTG in studies that monitored PPTG for five hours or less (-0.02 mmol/L, 95% CI -0.03 to -0.01; 13 studies). There was evidence of significant heterogeneity (Χ2=0.001). In studies that monitored PPTG between six and 24 hours, there was no statistically significant evidence of a dose dependency effect (12 studies). Further studies are needed.

Body weight (19 studies, comprising 12 cross-over, three parallel, two sandwich and two sequence studies, totalling 246 participants): Fructose intakes of 100g/day or less did not have a significant effect on body weight. There was no evidence of statistical heterogeneity (I2=0%). There was no evidence of publication bias using funnel plots and trim-and-fill analyses. Data on the effect of fructose intakes greater than 100g/day on body weight were limited.

The sensitivity of combined effects to different ways of handing repeat measures was investigated. Heterogeneity was variably studied as appropriate across outcomes in terms of choice of control substitute, background diet, health state, baseline weight/BMI, age and gender of the participants, adaptation, mode of fructose ingestion, and study quality. The results were reported in the review.

Authors' conclusions

Fructose intakes of 90g or less per day led to significant improvements (lowering) in levels of HbA1c. There were no significant changes for FPTG or body weight in adults for fructose intakes of 100g or less per day. Elevation of FPTG by fructose was significant and adverse only above 100g/day or higher dose. Significant effects on PPTG were evident when fructose intakes were greater than 50g/day, but dose-dependent effects were not significantly greater than for control carbohydrates (glucose or starch). Reductions of fructose intake may exchange one risk (dyslipidaemia) by another (dysglycaemia).

CRD commentary

This review had clearly stated inclusion and exclusion criteria in terms of participants, study design, interventions, comparators, and outcomes. Appropriate efforts were made to minimise reviewer bias and error during data extraction and study assessment. In searching only two relevant databases, some studies may have been missed. Study selection by only one reviewer may have overlooked relevant studies. Only English-language studies were included, so language bias is not ruled out. Despite these omissions, funnel plots and trim-and-fill analyses showed no significant evidence of publication bias. Statistical analyses, with perhaps the exception of the use of repeated measures as independent observations for glycated protein, appear appropriate. Overall, the authors conclusions reflect the limited evidence presented. The authors themselves state that further original studies would improve on reliability of quantitative aspects provided standard errors on treatment differences are published rather than have meta-analyst impute them, and that original authors reach for higher scores for study quality.

Implications of the review for practice and research

Practice: The findings of this review may be of interest to health professionals in the context of using these health markers as indicators of disease progression and drug efficacy.

Research: The longer-term effects of high fructose intake of greater than 50g/day on quality of life in subjects with increased dysglycaemia or dyslipidaemia need further evaluation.


Dansico Sweeteners (Redhill, United Kingdom).

Bibliographic details

Livesey G, Taylor R. Fructose consumption and consequences for glycation, plasma triacylglycerol, and body weight: meta-analyses and meta-regression models of intervention studies. American Journal of Clinical Nutrition 2008; 88(5): 1419-1437. [PubMed: 18996880]

Indexing Status

Subject indexing assigned by NLM


Adult; Biological Markers /blood; Blood Glucose /metabolism; Body Weight /drug effects; Clinical Trials as Topic; Diabetes Mellitus /metabolism; Dose-Response Relationship, Drug; Female; Fructose /administration & dosage /metabolism; Hemoglobin A, Glycosylated /analysis; Humans; Hyperlipidemias /metabolism; Male; Middle Aged; Regression Analysis; Risk Factors; Sweetening Agents /administration & dosage /metabolism; Triglycerides /blood



Database entry date


Record Status

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.

CRD has determined that this article meets the DARE scientific quality criteria for a systematic review.

Copyright © 2014 University of York.

PMID: 18996880

PubMed Health Blog...

read all...

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...