Clinical Care/Education/Nutrition BRIEF REPORT Effect of Cinnamon on Glucose Control and Lipid Parameters 1,2 3 WILLIAM L. BAKER, PHARMD, BCPS JEFFREY KLUGER, MD “cinnamomum aromaticum” in combina- 2 1,2 GABRIELA GUTIERREZ-WILLIAMS, PHARMD CRAIG I. COLEMAN, PHARMD tion with “diabetes mellitus.” Results were 1,2 C. MICHAEL WHITE, PHARMD, FCP, FCCP limited to clinical trials in humans. A manual search of retrieved articles was also performed. Three investigators inde- OBJECTIVE — To perform a meta-analysis of randomized controlled trials of cinnamon to pendently reviewed potentially relevant better characterize its impact on glucose and plasma lipids. articles and abstracted necessary data. The mean change in each study end RESEARCH DESIGN AND METHODS — A systematic literature search through July point from baseline was treated as a con- 2007 was conducted to identify randomized placebo-controlled trials of cinnamon that reported tinuous variable, and the weighted mean data on A1C, fasting blood glucose (FBG), or lipid parameters. The mean change in each study difference was calculated as the difference end point from baseline was treated as a continuous variable, and the weighted mean difference was calculated as the difference between the mean value in the treatment and control groups. A between the mean value in the treatment random-effects model was used. and control groups. Advanced statistical methods were used to impute change RESULTS — Five prospective randomized controlled trials (n ϭ 282) were identified. Upon scores as suggested by Follman and col- meta-analysis, the use of cinnamon did not significantly alter A1C, FBG, or lipid parameters. leagues (12,13). We conducted subgroup Subgroup and sensitivity analyses did not significantly change the results. and sensitivity analyses to assess whether diabetes type had an effect on our results. CONCLUSIONS — Cinnamon does not appear to improve A1C, FBG, or lipid parameters in A random-effects model was used to cal- patients with type 1 or type 2 diabetes. culate weighted mean difference and 95% CIs. Statistical heterogeneity was ad- Diabetes Care 31:41–43, 2008 dressed using the I2 statistic. Visual in- spection of funnel plots was used to assess innamon contains biologically ac- analysis of randomized controlled trials of for publication bias. The funnel plot is a tive substances that have demon- cinnamon to better characterize its impact pictorial representation of each study C strated insulin-mimetic properties. on glucose and plasma lipids. plotted by its effect size on the horizontal In vitro (1,2) and in vivo (3,4) studies axis and variance on the vertical axis. If have shown that cinnamon enhances glu- RESEARCH DESIGN AND the plot represents an inverted symmetri- cose uptake by activating insulin receptor METHODS cal funnel, it is said that publication bias is kinase activity, autophosphorylation of To be included in this meta-analysis, trials unlikely. Statistics were performed using the insulin receptor, and glycogen syn- had to be randomized placebo-controlled StatsDirect, version 2.5.8 (StatsDirect, thase activity. Other recent studies have trials of cinnamon and report data on Cheshire, England). demonstrated the ability of cinnamon to A1C, fasting blood glucose (FBG), or lipid reduce lipid levels in fructose-fed rats, po- parameters. tentially via inhibiting hepatic 3-hydroxy- Using the above-mentioned inclusion RESULTS 3-methylglutaryl CoA reductase activity criteria, we conducted a systematic litera- The initial search yielded 24 potential lit- (5,6). ture search of MEDLINE, CINAHL, Web erature citations. Of those, 14 citations Several clinical trials (7–11) have in- of Science, and the Cochrane Library were human studies, and only 6 were vestigated the impact of cinnamon on glu- from the earliest possible date through clinical trials. Furthermore, one citation cose and plasma lipid concentrations in July 2007. We used the following medical was excluded from the analysis because it was not a trial of cinnamon. Thus, a total patients with diabetes but yielded con- subject headings and keywords: “cinna- ϭ flicting results and had modest sample mon,” “cinnamomum,” “cinnamomum cas- of five clinical trials (n 282 subjects, sizes. Therefore, we performed a meta- sia,” “cinnamomum zeylanicum,” and follow-up range 5.7–16.0 weeks) were in- ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● cluded in the meta-analysis (7–11). All of 1 the studies used cinnamomum cassia, and From the Department of Pharmacy Practice, School of Pharmacy, University of Connecticut, Storrs, doses ranged from 1 to 6 g. Four studies Connecticut; the 2Department of Drug Information, Hartford Hospital, Hartford, Connecticut; and the 3Department of Cardiology, Hartford Hospital, Hartford, Connecticut. provided powder-filled capsules (7– Address correspondence and reprint requests to Craig I. Coleman, Pharm D, Assistant Professor of 9,11), while one provided aqueous-filled Pharmacy Practice, University of Connecticut, Hartford Hospital, 80 Seymour St., Hartford, CT 06102-5037. capsules (10). Four of the studies dosed E-mail: [email protected]. cinnamon during meals (8–11). The Received for publication 29 August 2007 and accepted in revised form 28 September 2007. Published ahead of print at http://care.diabetesjournals.org on 1 October 2007. DOI: 10.2337/dc07- study by Khan et al. (9) examined three 1711. different doses of cinnamon, and the re- Abbreviations: FBG, fasting blood glucose. sults were combined in this meta-analysis A table elsewhere in this issue shows conventional and Syste`me International (SI) units and conversion because no dose-response relationship factors for many substances. was found with cinnamon between 1 and © 2008 by the American Diabetes Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby 6 g (9). Studies were in type 2 diabetic marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. subjects (8–11) or adolescents with type DIABETES CARE, VOLUME 31, NUMBER 1, JANUARY 2008 41 Cinnamon and diabetes Table 1—Results of meta-analysis of randomized controlled trials evaluating cinnamon Base-case Type 1 diabetes only Type 2 diabetes only Weighted mean difference Weighted mean difference Weighted mean difference (95% CI) n (95% CI) n (95% CI) n A1C 0.07 (Ϫ0.11 to 0.26) 4 (204) 0.30 (Ϫ0.01 to 0.70) 1 (57) 0.01 (Ϫ0.20 to 0.22) 3 (147) FBG Ϫ17.15 (Ϫ47.58 to 13.27) 4 (207) NA — Ϫ17.15 (Ϫ47.58 to 13.27) 4 (207) Total cholesterol Ϫ9.63 (Ϫ35.94 to 16.67) 4 (207) NA — Ϫ9.63 (Ϫ35.94 to 16.67) 4 (207) Triglycerides Ϫ28.44 (Ϫ61.81 to 4.94) 4 (207) NA — Ϫ28.44 (Ϫ61.81 to 4.94) 4 (207) HDL cholesterol 1.58 (Ϫ0.74 to 3.89) 3 (147) NA — 1.58 (Ϫ0.74 to 3.89) 3 (147) LDL cholesterol Ϫ4.71 (Ϫ18.12 to 8.71) 4 (207) NA — Ϫ4.71 (Ϫ18.12 to 8.71) 4 (207) All results are reported in mg/dl as weighted mean difference (95% CI) using a random-effects model. n ϭ number of studies (number of subjects). 1 diabetes (7). Studies were conducted in dence in cinnamon’s impact on long- mon regulation of insulin signaling. Horm the U.S. (7,8), Europe (10,11), and Paki- term glycemic control. Res 50:177–182, 1998 stan (9). Patient withdrawals were appro- There are some additional limitations 2. Jarvill–Taylor KJ, Anderson RA, Graves priately reported in all studies. to this meta-analysis that should be DJ: A hydroxychalcone derivative from Upon meta-analysis, the use of cinna- noted. First, we identified only a small cinnamon functions as a mimetic for in- sulin in 3T3–L1 adipocytes. J Am Coll mon did not significantly alter A1C, FBG, number of eligible studies. Thus, our Nutr 20:327–336, 2001 or lipid parameters (Table 1). No statisti- meta-analysis may be underpowered to 3. Qin B, Nagasaki M, Ren M, Bajotto G, Os- cal heterogeneity was observed for the detect statistically significant differences hida Y, Sato Y: Cinnamon extract (tradi- 2 ϭ A1C or HDL analyses (I 0%). Each of in many of the end points. Post hoc sam- tional herb) potentiates in vivo insulin- the other analyses displayed a high degree ple size calculations suggest that if the dif- regulated glucose utilization via enhanced of statistical heterogeneity (I2 Ͼ 79.6% ferences were due to a real effect rather insulin signaling in rats. Diabetes Res Clin for all). Visual inspection of funnel plots than chance, then 1,166–6,853 patients Pract 62:139–148, 2003 (not shown) could not rule out publica- would be needed. Even if the beneficial 4. Cao H, Polansky MM, Anderson RA: Cin- tion bias for any analysis. changes observed in some of end points namon extract and polyphenols affect the After conducting subgroup and sen- were found to be statistically significant, expression of tristetraprolin, insulin re- sitivity analyses, the exclusion of non- their clinical significance could still be de- ceptor, and glucose transporter 4 in blinded trials (9), or evaluating type 1 (7) bated. We cannot determine the reason mouse 3T3–L1 adipolytes. Arch Biochem Biophys 459:214–222, 2007 and type 2 diabetes (8–11) separately did for differences between the Khan study 5. Kannappan S, Jayaraman T, Rajasekar R, not significantly change our meta- and the others. Ethnicity or cultural di- Ravichandran MK, Anuradha CV: Cinna- analysis’ results. Little to no statistical het- etary differences, dose, lack of verification mon bark extract improves glucose me- erogeneity was observed for any of these of double blinding, or chance resulting tabolism and lipid profile in the fructose- subsequent analyses. from small sample size in the Khan study fed rat. Singapore Med J 47:858–863, could explain the disparate findings.