Omega-3, omega-6, and total dietary polyunsaturated fat for prevention and treatment of type 2 diabetes mellitus: systematic review and meta-analysis of randomised controlled trials

Abstract

Objective: To assess effects of increasing omega-3, omega-6, and total polyunsaturated fatty acids (PUFA) on diabetes diagnosis and glucose metabolism.

Design: Systematic review and meta-analyses.

Data sources: Medline, Embase, Cochrane CENTRAL, WHO International Clinical Trials Registry Platform, Clinicaltrials.gov, and trials in relevant systematic reviews.

Eligibility criteria: Randomised controlled trials of at least 24 weeks' duration assessing effects of increasing α-linolenic acid, long chain omega-3, omega-6, or total PUFA, which collected data on diabetes diagnoses, fasting glucose or insulin, glycated haemoglobin (HbA_1c), and/or homoeostatic model assessment for insulin resistance (HOMA-IR).

Data synthesis: Statistical analysis included random effects meta-analyses using relative risk and mean difference, and sensitivity analyses. Funnel plots were examined and subgrouping assessed effects of intervention type, replacement, baseline risk of diabetes and use of antidiabetes drugs, trial duration, and dose. Risk of bias was assessed with the Cochrane tool and quality of evidence with GRADE.

Results: 83 randomised controlled trials (mainly assessing effects of supplementary long chain omega-3) were included; 10 were at low summary risk of bias. Long chain omega-3 had little or no effect on likelihood of diagnosis of diabetes (relative risk 1.00, 95% confidence interval 0.85 to 1.17; 58 643 participants, 3.7% developed diabetes) or measures of glucose metabolism (HbA_1c mean difference -0.02%, 95% confidence interval -0.07% to 0.04%; plasma glucose 0.04, 0.02 to 0.07, mmol/L; fasting insulin 1.02, -4.34 to 6.37, pmol/L; HOMA-IR 0.06, -0.21 to 0.33). A suggestion of negative outcomes was observed when dose of supplemental long chain omega-3 was above 4.4 g/d. Effects of α-linolenic acid, omega-6, and total PUFA on diagnosis of diabetes were unclear (as the evidence was of very low quality), but little or no effect on measures of glucose metabolism was seen, except that increasing α-linolenic acid may increase fasting insulin (by about 7%). No evidence was found that the omega-3/omega-6 ratio is important for diabetes or glucose metabolism.

Conclusions: This is the most extensive systematic review of trials to date to assess effects of polyunsaturated fats on newly diagnosed diabetes and glucose metabolism, including previously unpublished data following contact with authors. Evidence suggests that increasing omega-3, omega-6, or total PUFA has little or no effect on prevention and treatment of type 2 diabetes mellitus.

Systematic review registration: PROSPERO CRD42017064110.

Fig 1

Effect of long chain omega-3, α-linolenic acid, omega-6, and total polyunsaturated fatty acids (PUFA), on new diagnosis of diabetes. *Verified: see author response. †Reported as rise in blood sugar. ‡Diabetes excluded at baseline. Risk of bias key: A=random sequence generation (selection bias); B=Allocation concealment (selection bias); C=blinding of participants and personnel (performance bias); D=blinding of outcome assessment (detection bias); E=incomplete outcome data (attrition bias); F=selective reporting (reporting bias); G=attention; H=compliance; I=other bias

Fig 2

Effect of long chain omega-3, α-linolenic acid, omega-6, and total polyunsaturated fatty acids (PUFA) on glycated haemoglobin (%). *Units unclear. †Median change from baseline (zero in both arms). ‡In participants with impaired glucose metabolism at baseline. §In participants normoglycaemic at baseline. ¶Geometric means. **Change in medians. ††Medians only provided. ‡‡Median change. §§Barcelona hospital cohort at 5 years, Casas 2016. Risk of bias key: A=random sequence generation (selection bias); B=Allocation concealment (selection bias); C=blinding of participants and personnel (performance bias); D=blinding of outcome assessment (detection bias); E=incomplete outcome data (attrition bias); F=selective reporting (reporting bias); G=attention; H=compliance; I=other bias

Fig 3

Effect of long chain omega-3, α-linolenic acid, omega-6, and total polyunsaturated fatty acids (PUFA) on homoeostatic model assessment for insulin resistance (HOMA-IR) score. *Median change from baseline. †Geometric means. ‡Medians. §Reus subcohort, 2 year data. Risk of bias key: A=random sequence generation (selection bias); B=Allocation concealment (selection bias); C=blinding of participants and personnel (performance bias); D=blinding of outcome assessment (detection bias); E=incomplete outcome data (attrition bias); F=selective reporting (reporting bias); G=attention; H=compliance; I=other bias

Fig 4

Effect of long chain omega-3, α-linolenic acid, omega-6, and total polyunsaturated fatty acids (PUFA) on fasting serum insulin (pmol/L). *Reported SDs seemed to be SEs, so converted. †Geometric means. ‡Change in median. §Change, no measure of variance supplied. ¶Reus subcohort, 2 year data. Risk of bias key: A=random sequence generation (selection bias); B=Allocation concealment (selection bias); C=blinding of participants and personnel (performance bias); D=blinding of outcome assessment (detection bias); E=incomplete outcome data (attrition bias); F=selective reporting (reporting bias); G=attention; H=compliance; I=other bias

Fig 5

Effect of long chain omega-3, α-linolenic acid, omega-6, and total polyunsaturated fatty acids (PUFA) on fasting serum or plasma glucose (mmol/L). *Numbers verified. †Reported SDs seemed to be SEs, so converted. ‡Median change. §In participants with impaired glucose metabolism at baseline. ¶In participants normoglycaemic at baseline. **Geometric means. ††Change in median. ‡‡Change (no SDs supplied). §§Median change; data reported as mean and interquartile range. ¶¶Data verified. ***Barcelona hospital cohort at 5 years, Casas 2016. Risk of bias key: A=random sequence generation (selection bias); B=Allocation concealment (selection bias); C=blinding of participants and personnel (performance bias); D=blinding of outcome assessment (detection bias); E=incomplete outcome data (attrition bias); F=selective reporting (reporting bias); G=attention; H=compliance; I=other bias