Neuenschwander M, Ballon A, Weber KS, Norat T, Aune D, Schwingshackl L, Schlesinger S. Role of diet in type 2 diabetes incidence: umbrella review of meta-analyses of prospective observational studies. BMJ 2019;365:l2368.
OBJECTIVE To summarise the evidence of associations between dietary factors and incidence of type 2 diabetes and to evaluate the strength and validity of these associations.
DESIGN Umbrella review of systematic reviews with metaanalyses of prospective observational studies.
DATA SOURCES PubMed, Web of Science, and Embase, searched up to August 2018.
ELIGIBILITY CRITERIA Systematic reviews with meta-analyses reporting summary risk estimates for the associations between incidence of type 2 diabetes and dietary behaviours or diet quality indices, food groups, foods, beverages, alcoholic beverages, macronutrients, and micronutrients.
RESULTS 53 publications were included, with 153 adjusted summary hazard ratios on dietary behaviours or diet quality indices (n=12), food groups and foods (n=56), beverages (n=10), alcoholic beverages (n=12), macronutrients (n=32), and micronutrients (n=31), regarding incidence of type 2 diabetes. Methodological quality was high for 75% (n=115) of meta-analyses, moderate for 23% (n=35), and low for 2% (n=3).
Quality of evidence was rated high for an inverse association for type 2 diabetes incidence with increased intake of whole grains (for an increment of 30 g/day, adjusted summary hazard ratio 0.87 (95% confidence interval 0.82 to 0.93)) and cereal fibre (for an increment of 10 g/day, 0.75 (0.65 to 0.86)), as well as for moderate intake of total alcohol (for an intake of 12-24 g/day v no consumption, 0.75 (0.67 to 0.83)). Quality of evidence was also high for the association for increased incidence of type 2 diabetes with higher intake of red meat (for an increment of 100 g/day, 1.17 (1.08 to 1.26)), processed meat (for an increment of 50 g/day, 1.37 (1.22 to 1.54)), bacon (per two slices/day, 2.07 (1.40 to 3.05)), and sugar sweetened beverages (for an increase of one serving/ day, 1.26 (1.11 to 1.43)).
CONCLUSIONS Overall, the association between dietary factors and type 2 diabetes has been extensively studied, but few of the associations were graded as high quality of evidence. Further factors are likely to be important in type 2 diabetes prevention; thus, more well-conducted research, with more detailed assessment of diet, is needed.
Of all lifestyle factors that relate to health and disease, probably the most difficult ones to evaluate are dietary factors, including foods, nutrients, and beverages. People eat foods that contain a very large number of nutrients, so the effects of any single nutrient cannot be ascertained just from knowing the consumption of a single food. Further, diet consists of many types of food in combination, which also complicate an understanding of health effects of individual factors. Finally, even for “alcohol,” people obtain it in many different beverages, most of which contains substances other than alcohol.
On top of the inherent difficulties in assessing dietary effects on health, the amount of a substance being consumed is very difficult to assess, and the pattern of eating and drinking provide additional complexity. Given all of these difficulties, it is quite amazing that we have high quality evidence for health effects of any substance consumed.
The present review paper, based on 53 meta-analyses from prospective cohort observational studies, provides a good example of the complexity of studying the effects of dietary factors and the risk of a very common disease, diabetes mellitus. Data on very few foods or nutrients are adequate to clearly determine their effects. Among those that have been identified as reducing the risk of diabetes, high-quality evidence was found only for whole grains, cereal fiber, and alcohol. As for dietary factors that have strong evidence of increasing the risk of diabetes, the authors of this paper identify only red meat, processed meat, bacon, and sugar sweetened beverages.
Comments by specific Forum members: Reviewer Ellison noted: “This is an unusual paper, a massive review of meta-analyses (‘umbrella review’) based on prospective studies relating many aspects of the diet (including beverages) for their relation to the development of diabetes. It is interesting that analyses of only a few dietary factors are judged as presenting high quality of evidence, but included as potential ways to prevent diabetes are whole grains, cereal fibre and moderate alcohol intake. (Factors with high quality evidence that increase the risk of diabetes are red meat, processed meat, bacon, and sugar sweetened beverages.) While the differences according to type of beverage are sometimes not statistically significant, it is interesting that in all comparisons, consumers of wine showed greater protection against diabetes that consumers of spirits.”
Forum members considered that there are a number of strengths to this analysis. It is based on previous meta-analyses of 53 prospective studies which summarized 277 high-quality primary studies. While there have been several previous such “umbrella” analyses that were based on foods and nutrients, they did not include the effects of alcohol consumption. In the present paper the authors evaluated both linear and non-linear models, and were able to adjust for most common potential confounders of effect. Among the weaknesses are that most of the primary studies did not include alcohol as an exposure, and changes in consumption of factors were not evaluated. Further, the authors were not able to adjust for the pattern of drinking in their alcohol analyses. Because the results are essentially the same as for previous umbrella studies, some Forum reviewers were not impressed with the new information being presented. Others, however, thought that the data presented were very useful in evaluating dietary exposures related to the development of diabetes.
Reviewer Skovenborg noted: “In the soup of (too much low quality) evidence you will often taste a spicy mix of hidebound positions and policy considerations. One example: the 2018 WHO draft guidelines on dietary saturated fatty acids and trans fatty acids recommend reducing total intake of saturated fat and replacing it with polyunsaturated and mono-unsaturated fatty acids. The recommendations failed to take into account considerable evidence that the health effects of saturated fat vary depending on the specific fatty acid and on the specific food source (Astrup et al).
“The field of nutritional epidemiology also fails to take the variation between people into consideration. One example: ‘In summary, substantial heterogeneities in BMI, both over time and across different population groups, the fact that the bulk of the variation in BMI is between individuals (and not between populations), and our inability to explain much of this inter-individual variation in BMI together suggest that an exclusive focus on population-wide preventive strategies is unlikely to make a noticeable difference in reversing the obesity epidemic’ (Kim et al). Last but not least alcohol intake, the most popular ‘prügelknabe’ among political epidemiologists as a nutrient is often not measured by the same standard as other nutrients. An example: I suppose that the link between sugary drinks and breast cancer (Chazelas et al found that a 100 ml increase in sugary drink consumption was associated with a 22% increased risk of breast cancer) will not get the same attention as alcohol and breast cancer.
Skovenborg concluded: “I am not able to give an expert opinion of Neuenschwander et al’s meta-analysis, however, to me this analysis looks to be as good as these meta-analyses get, and for once it is comforting to see that the authors have applied the same yardstick to all the dietary factors, including alcohol as a part of diet.”
Does the present analysis expand our knowledge of diet and diabetes? Forum member Puddey stated: “From the perspective of alcohol consumption and risk of Type 2 diabetes mellitus I am uncertain what this umbrella review tells us above and beyond what has already been well-canvassed in the 2 previous meta-analyses quoted by the authors (Li et al; Huang et al). Those studies also concluded that light to moderate alcohol consumption decreases the risk of Type 2 diabetes mellitus, and that this evidence is of high quality and that the decrease is more prominent with consumption of wine. In addition, this umbrella review does not enable consideration of issues raised in previous meta-analyses as to how the alcohol – diabetes relationship is modified by gender, adiposity, smoking and family history of diabetes, all important issues with respect to shaping any public health advice on alcohol use in the prevention of type 2 diabetes mellitus. Mild to moderate alcohol intake, and especially the intake of wine, have been consistently associated with healthier dietary patterns than those seen in never drinkers or heavier drinkers, and this umbrella review does not enable us to tease out whether any effects of alcohol intake or type of alcoholic beverage are independent from dietary influence.
“Another umbrella review of 86 meta-analyses and systematic reviews of risk factors for type 2 diabetes incidence was published by Bellou et al earlier this year and reported almost identical results for alcohol, with a decrease in risk when moderate drinkers were compared to never drinkers. The evidence was graded as highly suggestive rather than convincing (the highest grade). In that study the conclusions with respect to dietary factors were also very similar with six showing convincing or highly suggestive relationships with Type 2 diabetes mellitus, an increase in relative risk seen with processed meat and sugar-sweetened beverages and a decrease with whole grain products, a healthy dietary pattern and dietary heme iron (although in each case the demonstrated effect sizes were considered modest).”
Reviewer Ellison added: “We realize that in all meta-analyses, a large number of assumptions are required, and further assumptions are needed in umbrella studies. Do these always lead to a better estimate of ‘the truth’? I like to also consider results of individual, large-scale prospective studies in which the participants are similar to each other (as in the large Harvard studies of nurses and health professionals). Because of less variation in education, lifestyle, and other socio-economic factors in such studies, there tend to be fewer differences among subjects in many ‘confounding variables’ which may simplify the analyses when evaluating an exposure. An example is the very early report from the Nurses’ Health Study of alcohol and diabetes by Stampfer et al, in 1988, in which there was a 20% or greater reduction in risk associated with moderate drinking, even when obesity was taken into consideration. Subsequent papers from these studies have helped clarify the relation of many other lifestyle factors to the risk of diabetes (e.g., Hu et al 2001; Salmeron et al; Hu et al 2003; Schulze et al; Ardisson Korat et al).”
References from Forum critique
Ardisson Korat AV, Willett WC, Hu FB. Diet, lifestyle, and genetic risk factors for type 2 diabetes: a review from the Nurses’ Health Study, Nurses’ Health Study 2, and Health Professionals’ Follow-up Study. Curr Nutr Rep 2014;3:345–354.
Astrup A, Bertram HCS, Bonjour J-P, et al. WHO draft guidelines on dietary saturated and trans fatty acids: time for a new approach? BMJ 2019; 366 doi: https://doi.org/10.1136/bmj.l4137
Bellou V, Belbasis L, Tzoulaki I, Evangelou E. Risk factors for type 2 diabetes mellitus: An exposure-wide umbrella review of meta-analyses. PLoS One. 2018;13:e0194127. doi: 10.1371/journal.pone.0194127. eCollection 2018
Chazelas E, Srour B, Desmetz E, et al. Sugary drink consumption and risk of cancer: results from NutriNet-Santé prospective cohort exposure-wide umbrella review of meta-analyses. PLoS One 2018;13:e0194127.
Hu FB, Manson JE, Stampfer MJ, Colditz G, Liu S, Solomon CG, et al. Diet, lifestyle, and the risk of type 2 diabetes mellitus in women. N Engl J Med 2001;345:790-797.
Hu FB, Li TY, Colditz GA, Willett WC, Manson JE . Television watching and other sedentary behaviors in relation to risk of obesity and type 2 diabetes mellitus in women. JAMA 2003;289:1785-1791.
Huang J, Wang X, Zhang Y. Specific types of alcoholic beverage consumption and risk of type 2 diabetes: A systematic review and meta-analysis. J Diabetes Investig 2017;8:56-68.
Kim, R, Lee DH, Subramanian SV. Understanding the obesity epidemic. BMJ 2019;366 doi: https://doi.org/10.1136/bmj.l4409
Li XH, Yu FF, Zhou YH, He J. Association between alcohol consumption and the risk of incident type 2 diabetes: a systematic review and dose-response meta-analysis. Am J Clin Nutr 2016;103:818-829.
Salmeron J, Hu FB, Manson JE, Stampfer MJ, Colditz GA, Rimm EB, et al. Dietary fat intake and risk of type 2 diabetes in women. Am J Clin Nutr 2001;73:1019-1026.
Schulze MB, Manson JE, Ludwig DS, Colditz GA, Stampfer MJ, Willett WC, Hu FB. Sugar-sweetened beverages, weight gain, and incidence of type 2 diabetes in young and middle-aged women. JAM. 2004;292:927-934.
Stampfer MJ, Colditz GA, Willett WC, Manson JE, Arky RA, Hennekens CH, Speizer FE. A prospective study of moderate alcohol drinking and risk of diabetes in women. Am J Epidemiol 1988;128:549-558.
Of all lifestyle factors that relate to health and disease, probably the most difficult ones to evaluate are dietary factors, including foods, nutrients, and beverages. The present paper consists of a meta-analysis (an “umbrella” analysis) based on 53 previous meta-analyses from prospective cohort observational studies. This analysis includes 153 adjusted summary hazard ratios on dietary behaviors or diet quality indices, food groups and foods, beverages, alcoholic beverages, macronutrients, and micronutrients related to the incidence of type 2 diabetes. Data from previous meta-analyses on very few foods or nutrients are adequate to clearly determine their effects, but among those that have been identified as reducing the risk of diabetes, high-quality evidence was found only for whole grains, cereal fiber, and alcohol. As for dietary factors that have strong evidence of increasing the risk of diabetes, the authors of this paper identify only red meat, processed meat, bacon, and sugar-sweetened beverages.
Forum members considered this to be a well-done, appropriate analysis in a paper that summarizes well the current scientific literature. Some members were concerned about the number of assumptions that scientists must make when doing a simple meta-analysis, and especially when attempting an umbrella one; they consider it key to also evaluate single, large prospective studies where adjustments for many confounding socio-economic factors are less likely to be important.
Overall, the results of this paper strongly support most of the findings of previous meta-analyses and umbrella analyses and indicate only a few dietary factors for which “high-quality” evidence exists for an association with the risk of developing diabetes. The clearest evidence from the analyses in this paper, which was based only on prospective cohort studies, is that cereal grains and moderate alcohol lower the risk, while meats and sugar-sweetened beverages increase the risk of diabetes.
Given the complexity of evaluating dietary factors for their relation to disease outcomes, it will be useful for future studies to attempt to measure a multitude of foods, beverages, nutrients, and eating patterns, preferably over time, to evaluate further the association with disease. For now, current evidence suggests that cereals, grains, and moderate alcohol (that decrease risk) and meat and sugar-sweetened beverages (that increase risk) have the most consistent data for a relation to subsequent diabetes mellitus.
Comments on this critique by the International Scientific Forum on Alcohol Research have been provided by the following members:
Giovanni de Gaetano, MD, PhD, Department of Epidemiology and Prevention, IRCCS Istituto Neurologico Mediterraneo NEUROMED, Pozzilli, Italy
R. Curtis Ellison, MD, Professor of Medicine, Section of Preventive Medicine & Epidemiology, Boston University School of Medicine, Boston, MA, USA
Ramon Estruch, MD, PhD, Hospital Clinic, IDIBAPS, Associate Professor of Medicine, University of Barcelona, Spain
Harvey Finkel, MD, Hematology/Oncology, Retired (Formerly, Clinical Professor of Medicine, Boston University Medical Center, Boston, MA, USA)
Tedd Goldfinger, DO, FACC, Desert Cardiology of Tucson Heart Center, University of Arizona School of Medicine, Tucson, AZ, USA
Fulvio Mattivi, MSc, CAFE – Center Agriculture Food Environment, University of Trento, via E. Mach 1, San Michele all’Adige, Italy
Linda McEvoy, PhD, Department of Radiology, University of California at San Diego (UCSD), La Jolla, CA, USA
Ian Puddey, MD, Emeritus Professor, Faculty of Health & Medical Sciences, University of Western Australia, Nedlands, Australia
Erik Skovenborg, MD, specialized in family medicine, member of the Scandinavian Medical Alcohol Board, Aarhus, Denmark
Creina Stockley, PhD, MSc Clinical Pharmacology, MBA; Adjunct Senior Lecturer at the University of Adelaide, Australia
Arne Svilaas, MD, PhD, general practice and lipidology, Oslo University Hospital, Oslo, Norway
Pierre-Louis Teissedre, PhD, Faculty of Oenology–ISVV, University Victor Segalen Bordeaux 2, Bordeaux, France
Fulvio Ursini, MD, Dept. of Biological Chemistry, University of Padova, Padova, Italy
David Van Velden, MD, Dept. of Pathology, Stellenbosch University, Stellenbosch, South Africa