Systematic Review of Herbs and Dietary Supplements for Glycemic Control in Diabetes

We searched MEDLINE, OLDMEDLINE, CAM-PubMed, HealthSTAR, and the Cochrane Library Database from 1960 to March 2002 using the MeSH terms CAM, alternative therapies, hypoglycemic plants, and individual herb and supplement names from popular sources, each crossed with the term diabetes mellitus. In addition, we contacted experts in the field to identify studies, and we also hand-searched references of key articles. We did not include supplements made from animal components. Fish oil supplementation, for example, has been examined in prior meta-analyses (46,47). We also did not include soluble fiber supplements, which overlap considerably with dietary fiber treatment and already play a role in conventional diabetes nutrition advice (48–51).

We limited studies to those published in the English language and restricted our search to herbs and supplements for glycemic control and symptoms of hyperglycemia. We excluded trials that primarily examined diabetic complications such as neuropathy, nephropathy, or retinopathy. We included studies in subjects with impaired glucose tolerance or those specifically at risk for diabetes (e.g., older, sedentary, obese individuals with a family history of diabetes). As supporting evidence, we also examined studies of glycemic control in healthy volunteers. To assess quality of RCTs, we employed the Jadad scale, a previously validated instrument that assesses trials based on appropriate randomization, blinding, and description of study withdrawals or dropouts (52,53). Quality of evidence for specific herbs or supplements was assessed using the U.S. Preventive Services Task Force criteria (54) (online appendix A; http://care.diabetesjournals.org) and the American Diabetes Association evidence grading system for clinical practice recommendations (55) (online appendix B). Clinical guidelines were based on the criteria developed by Weiger et al. (56) (online appendix C). These criteria place individual CAM therapies along a continuum that encompasses “recommend” (high-quality evidence supports both efficacy and safety), “accept/consider recommending” (evidence supports both efficacy and safety), “accept” (evidence regarding efficacy is inconclusive but supports safety), and “discourage” (evidence indicates either inefficacy or serious risk).

A total of 108 clinical studies were found examining 25 single herbs, 11 combination herb formulas, and 9 vitamin/mineral supplements as potential therapy for diabetes. Of these, 58 were controlled clinical trials in patients with diabetes or impaired glucose tolerance (42 randomized, 16 nonrandomized). Only four of the controlled trials included patients with type 1 diabetes (57–60). In addition, there were 12 trials examining glycemic parameters in healthy individuals. The remaining studies were 36 uncontrolled prospective cohort trials and 2 case reports.

We present the available evidence for 26 of the substances with either one or more controlled clinical trials in patients with diabetes or impaired glucose tolerance. Methodological details and results of these trials are summarized in Tables 1–3 (Single herbs, Multiple herb combinations, Vitamin and Mineral Supplements). These tables include supplement name, reference, study design, sample population, intervention, control, outcomes, direction of evidence, Jadad score, and reported adverse effects. Other studies, including some RCTs that examined glycemic parameters in healthy individuals, are not listed in the tables but are presented as supporting evidence when applicable.

The most common outcomes measures encountered in these trials included the following parameters of glycemic control: fasting and postprandial plasma glucose, response to glucose tolerance tests (GTTs), insulin and C-peptide levels, protein glycosylation (long-lived intracellular glycated hemoglobin and shorter-lived plasma fructosamine), and clinical insulin requirements. Many of the vitamin studies also examined measures of insulin sensitivity, hepatic glucose production, glucose oxidation, and glucose uptake. Oftentimes, only a few of the above measures were studied in any particular trial. A significant positive change in at least one of these important parameters was required to categorize the trial as positive.

Table 1 presents the controlled clinical trials of single herbs for glycemic control in patients with diabetes. Of the single herbs studied, the higher-quality RCTs (with Jadad scores of 3 or greater) are available for Coccinia indica, ginseng species, Bauhinia forficata, and Myrcia uniflora. One RCT for Allium sativum is also of adequate quality but was conducted in nondiabetic individuals. Other herbs, Allium cepa, Ocimum sanctum, Ficus carica, Silibum marianum, Opuntia streptacantha, and Trigonella foenum, have been studied in poorer-quality RCTs. Gymnema sylvestre and Momordica charantia have been studied in only nonrandomized controlled trials.

Coccinia indica (ivy gourd) is a creeping plant that grows wildly in many parts of the India subcontinent, and is used to treat “sugar urine” (madhumeha) in Ayurveda, a traditional East Indian healing system. The mechanism of action of Coccinia indica is not well understood, but the herb appears to have insulin-mimetic properties (61–63).

The one RCT of this herb (n = 32), conducted in India, reported significant changes in glycemic control following 6 weeks’ use of powder from locally obtained crushed dried leaves in poorly controlled or otherwise untreated patients with type 2 diabetes (64). Another three-arm controlled clinical trial (n = 70) compared 12 weeks’ use of dried herb pellets made from fresh leaves with no treatment and oral hypoglycemic agents (chlopropamide) in patients with type 2 diabetes (61). The magnitude of change seen with the herb was similar to that with a conventional drug. Two other open-label prospective trials offer supporting evidence of a hypoglycemic effect (62,63). No adverse events were reported in these trials. The preliminary evidence suggests that the potential role for Coccinia indica in diabetes warrants further study. (U.S. Preventive Services Task Force Level I, American Diabetes Association Guidelines Level A)

Several different plant species are often referred to as ginseng. These include Chinese or Korean ginseng (Panax ginseng), Siberian ginseng (Eleutherococcus senticosus), American ginseng (P. quiquefolius), and Japanese ginseng (P. japonicus). Panax species (from the root panacea) are often touted for their “cure-all” adaptogenic properties, immune-stimulant effects, and their ability to increase stamina, concentration, longevity, and overall well-being (37). Preparations use the herb’s root; some sources report greater efficacy with roots that are greater than 3 years old. Principal components are believed to be the triterpenoid saponin glycosides (ginsenosides or panaxosides). Hypoglycemic effects have been shown in streptozotocin rat models (65). Reported mechanisms of action include decreased rate of carbohydrate absorption into the portal hepatic circulation, increased glucose transport and uptake mediated by nitric oxide, increased glycogen storage, and modulation of insulin secretion (39).

Most clinical trials we found utilized American ginseng, with many examining the herb’s short-term effects on patients with type 2 diabetes after a standard oral GTT (66,67). Two longer-term trials administered American ginseng for 8 weeks (n = 36 and n = 24); both reported decreases in fasting blood glucose and HbA_1c (68,69). Only one case of insomnia was reported in these trials. Three other short-term metabolic trials in healthy volunteers also found decreases in postprandial glucose (66,70,71). All but one of the clinical trials we examined were from the same investigator group. The available evidence for American ginseng in diabetes suggests a possible hypoglycemic effect; however, the trials are small and longer-term studies are needed. (Level I, A)

Allium sativum (garlic), a member of the lily family, is most commonly used worldwide for flavorful cooking. Much of the clinical literature on garlic has focused on its potential antioxidant activity and microcirculatory effects (e.g., allicin and ajoene for use in hypertension and hyperlipidemia). Few studies have examined its effects on insulin and glucose handling, although some attention has been given to allyl propyl disulfide, a volatile oil, and S-allyl-cysteine sulfoxide, a sulfur containing amino acid (39,72). Experiments in animal models with alloxan-induced diabetes have shown moderate reductions in blood glucose; no effect is seen in pancreatectomized animals (72,72). Allium cepum (onion) also contains allyl propyl disulphide and has similar purported hypoglycemic properties. Reported mechanisms of allium species include increased secretion or slowed degradation of insulin, increased glutathione peroxidase activity, and improved liver glycogen storage (39,41).

The highest quality RCT of Allium sativum in humans was actually designed to examine thrombocyte aggregation in nondiabetic individuals (n = 60). However, the investigators found significant decreases in fasting serum glucose (74). The only available trial of garlic in patients with type 2 diabetes (n = 33) did not find consistent glucose or insulin responses after 1 month of supplementation (75). The only clinical trial available for Allium cepa is a small RCT of allyl propyl disulphide extract capsules from onion in nondiabetic volunteers (n = 6); investigators showed an acute decrease in fasting blood glucose and increase in insulin, supporting an insulin-mediated effect (76). No adverse events were reported in these trials. The limited data provide conflicting evidence for allium species in glycemic control. (Level I, C)

Ocimum sanctum (holy basil) is another commonly used herb in Ayurveda (related species include Ocimum album and Ocimum basilicum). Studies in animal models suggest hypoglycemic effects (77), although the mechanism of action remains unknown. Postulated effects include enhanced β-cell function and insulin secretion. The one available controlled clinical trial of Ocimum sanctum (n = 40) showed positive effects on both fasting and postprandial glucose in patients with type 2 diabetes using a local preparation of fresh leaf powder mixed in water for 4 weeks (78). No adverse effects were reported. Further information is needed before the efficacy of Ocimum sanctum in diabetes can be fully assessed. (Level III, C)

Trigonella foenum graecum (fenugreek) is a legume extensively cultivated in India, North Africa, and the Mediterranean. It is a common condiment used in Indian cooking and commonly used herb in Ayurveda. Defatted seeds of fenugreek, which are rich in fiber, saponins, and protein, have been described in early Greek and Latin pharmacopoeias for hyperglycemia. Although the seed portion is often mentioned, other parts of the herb have also been investigated. Purported mechanisms include delay of gastric emptying, slowing carbohydrate absorption, and inhibition of glucose transport from the fiber content, as well as increased erythrocyte insulin receptors and modulation of peripheral glucose utilization. Many studies in alloxan-rat models have shown modulated exocrine pancreatic secretion (79).

There are several trials available for fenugreek in type 2 diabetes; however, most are noncontrolled (158). Of the available RCTs, they are generally poorer-quality studies with small numbers (n = 5–15) and from a single investigator group. Nonetheless, these trials, including a single trial in type 1 diabetes, have reported improved glycemic control using seed powder incorporated into unleavened bread (59,80). Another trial in healthy volunteers (n = 38) incorporated several short-term randomized crossover experiments administering different preparations of fenugreek before oral GTT. In these series of trials, whole raw seeds, extracted seed powder, gum isolate of seeds, and cooked whole seeds seemed to decrease postprandial glucose levels, whereas degummed seeds and cooked leaves did not (79). Other open-label prospective cohort studies have followed patients on fenugreek for up to 6 months with reported benefits in glycemic control (79,81–84). No adverse effects were reported in these trials. There is some preliminary evidence for the efficacy of fenugreek that suggests further studies may be warranted. (Level II-2, C)

Indigenous to rainforests and tropical areas of South America, Bauhinia forficata has been used in traditional treatment of diabetes in that area. In Brazilian herbal medicine, Bauhinia species have been referred to as “vegetable insulin.” Another commonly used South American herb is Myrcia uniflora. As part of a national effort to identify potential plant species useful in glucose control, two small crossover studies (n = 16 and n = 18) by one investigator administered each of these herbs as tea infusions to separate groups of patients three times daily for 8 weeks. No significant differences in glucose or HbA_1c were detected between study herb infusion and a placebo tea using Imperata brasiliensis. No adverse effects were reported (85). This limited preliminary evidence does not support the hypoglycemic effect of these two plant species. (Level I, American Diabetes Association level not applicable if no studies show benefit)

Ficus carica (fig leaf) is a popular plant used for patients with diabetes in Spain and other areas in Southwestern Europe. Its active component is unknown. Several studies in animal models with diabetes have shown both short- and long-term hypoglycemic effects, although human trials are lacking. Potential hypolipdemic effects in diabetic rats have also been shown (86–88). Its mechanism for glucose effect is unknown; however, some studies suggest facilitation of glucose uptake peripherally. The one available clinical trial is a small crossover study of fig leaf tea for 4 weeks in patients with type 1 diabetes (n = 10); investigators showed a decrease in postprandial glucose and insulin requirements, but no change in fasting glucose when compared with the control commercial tea (60). No effect was seen in C-peptide levels, thereby supporting a non-insulin-mediated effect. No adverse effects were reported. Clearly, more information is needed before the efficacy of Ficus carica can be properly assessed. (Level III, C)

Opuntia streptacantha (nopal) or the prickly pear cactus can be found in arid regions throughout the Western hemisphere, including the southwestern U.S., and is commonly used for glucose control by those of Mexican descent. It has a high-soluble fiber and pectin content, which may affect intestinal glucose uptake, partially accounting for its hypoglycemic actions (65). Animal models have reported decreases in postprandial glucose and HbA_1c with synergistic effects with insulin. Studies in pancreatectomized animals report that hypoglycemic activity is not dependent on the presence of insulin (89). Most human studies of nopal have been published in Spanish and, thus, are not included in this review. We found only two controlled short-term metabolic trials (n = 14 and n = 32) published in the English language, both by the same investigator (90,91). These reported improvements in patients with type 2 diabetes with decreased fasting glucose and decreased insulin levels, suggesting enhanced insulin sensitivity. No side effects were reported in these trials. The limited data suggests a possible hypoglycemic effect of nopal; however, longer-term clinical trials are needed. (Level III, C)

Silibum marianum (milk thistle), a member of the aster family, has been primarily studied for its purported effects on alcoholic and viral hepatitis, rather than for glycemic control. However, silymarin is rich in flavonoids, potent antioxidants, and some have postulated a potential benefit for those who have insulin resistance secondary to hepatic damage (39). Mechanisms are based on the herb’s antioxidant activity and effects on hepatocyte stabilization with decreased glutathione oxidation, as well as on restoration of normal malondialdehyde concentrations.

The one available clinical trial examined cirrhotic patients with type 2 diabetes (n = 60) using a commercially available preparation (“Legalon” 600 mg/day; IBI Lorenzini, Milan, Italy) for 12 months, with significant improvements in glycemic control when compared with no treatment (92). No adverse effects were reported. Further information and higher quality clinical trials are needed to further investigate milk thistle in glycemic control. (Level III, C)

Gymnema sylvestre is another commonly used herb in Ayurveda. The plant is a woody climber that grows in tropical forests of central and southern India. According to common folklore, chewing the leaves causes a loss of sweet taste, hence the popular Hindi name of the plant “gurmar,” meaning “destroyer of sugar.” Early animal studies reported blood glucose-lowering effects in animals with residual pancreatic function, but no effect in total pancreatectomized animals. Studies of an ethanol leaf extract, GS4, in diabetic rat and rabbit models have reported regeneration of islets of Langerhans, decreases in blood glucose, and increases of serum insulin (58). Mechanism of action is unknown; postulated theories include an increase in glucose uptake and utilization, increase in insulin release through cell permeability, increase in β-cell number, and stimulation of β-cell function (39,93).

Two nonrandomized controlled clinical trials are available, both from the same investigator group. Groups of patients with type 1 diabetes (n = 64) and type 2 diabetes (n = 47) showed improved glycemic control with chronic adjunctive use of GS4 extract compared with those who received conventional treatment alone (58,94). The evidence for the beneficial effect of Gymnema sylvestre in diabetes is suggestive, although inconclusive given the limited data. (Level II-1, C)

Momordica charantia is a vegetable indigenous to tropical areas, including India, Asia, South America, and Africa, also known as balsam pear, karela (karolla), and bitter melon. Reported preparations of the herb range from injectable extracts to fruit juice to fried melon bits (39,95–97). Active components are thought to be charantin, vicine, and polypeptide-p (an unidentified insulin-like protein similar to bovine insulin). Theoretical mechanisms include increased insulin secretion, tissue glucose uptake, liver muscle glycogen synthesis, glucose oxidation, and decreased hepatic gluconeogenesis. Studies in alloxan-induced diabetic rabbits have suggested hypoglycemic effects (98).

Two controlled short-term metabolic trials in patients with type 2 diabetes (n = 18 and n = 9) have reported acute effects on blood glucose with Momordica charantia fruit juice, as well as subcutaneous vegetable insulin extract (95,99). Two other small, uncontrolled open-label trials also reported positive effects on glycemic control after longer-term use (7–11 weeks) (96,97). No adverse effects were reported in these trials. Some, albeit limited, data suggest a potential effect of Momordica charantia in diabetes. However, further information in RCTs is needed. (Level III, C)

Aloe vera is the most well-known species of aloe, a desert plant resembling the cactus in the Liliaceae family. It is popularly used to treat burns and promote wound healing. The dried sap of the Aloe vera is a traditional remedy for diabetes in the Arabian peninsula (33), although aloe gel is preferred over the sap as the latter contains the laxative anthraquinone (100). Aloe gel, obtained from the inner portion of the leaves, contains glucomannan, a hydrosoluble fiber which may in part account for its hypoglycemic effects (39). Reports in animal models have been inconsistent (100–103). Two nonrandomized clinical trials (n = 40 and n = 76) are available from the same investigator group that reported improved fasting blood glucose with 6 weeks of juice made from aloe gel (100,104). Case reports of five type 2 diabetic individuals reported decreases in fasting blood glucose as well as HbA_1c (101). No adverse effects were reported in these trials. The preliminary data suggest a potential effect of Aloe vera in glycemic control; however, further information is needed. (Level II-1, C)

Other herbs that have been studied solely in uncontrolled trials include berberine (105), Cinnamomym tamala (106), curry (107), Eugenia jambolana (108), gingko (109), Phyllanthus amarus (110), Pterocarpus marsupium (111), Solanum torvum (112), and Vinca rosea (113).

Table 2 presents the controlled clinical trials of multiple herb combinations for glycemic control in patients with diabetes.

TCM encompasses a system of healing that has origins over 2,000 years old. It emphasizes the importance of a balanced and harmonious flow of “qi,” or “life force,” and employs diverse modalities such as acupuncture, massage, qigong, and an individualized approach to herbal medicine (20). We found few trials of TCM in the English language; most have been published in Chinese and were unavailable for this review.

One controlled clinical trial of a multiple herb combination examined a specific formulation containing Coptis chinensis, Astragalus membranaceus, and Lonicera japonica. Among a host of other plants used in TCM for the treatment of diabetes, these plants were selected for study by the Chinese Academy of Medical Science based on experiential reports of efficacy and safety. Mechanisms of action are not well reported, but may include decreasing digestive carbohydrate absorption. This formula is not thought to influence action of insulin. Using a 2×2 factorial design (n = 216) with TCM verum pill or placebo and glibenclamide verum pill or placebo, investigators reported that the two treatments together were more efficacious than either alone (114). Of 216 patients, there was one report of diarrhea and one report of dry mouth. Also, one case of hypoglycemia occurred in the combined treatment group.

A much smaller trial (n = 12) of lower quality examined another TCM preparation, Xiaoke tea. Little is written about this formulation in English literature. It appears not to affect insulin concentrations and was ineffective in rats that lack endogenous insulin. The trial did not report details about the constituents of the treatment tea, and investigators reported no difference in glycemic parameters as compared with an “ordinary” tea infusion (115). Another controlled clinical trial (n = 148) examined a formulation called Semen Persical Decoction for Purgation with Addition (SPDPA), a combination of eight different herbs and reported decreases in fasting blood glucose not significantly different from changes seen with glyburide (116). No adverse effects were reported with this formulation. The available studies suggest that some TCM formulations, but not others, may have beneficial effects. However, the data are certainly limited and no formula has been studied in more than one trial. (Level I, C)

Native American medicine refers to the healing practices from the people indigenous to North America; the approach combines awareness of mind, body, and spirit and ritualistic observances with practices of herbalism. One clinical trial (n = 40) specifically examined an herbal tea preparation containing Populus tremuloides (trembling aspen) and Heracleum lanatum (cow parsnip) prescribed by an Alexis band Sioux healer (117). Investigators reported no glycemic benefit over a control tea containing mint and fennel seed. Little is known scientifically about this formula, and it has not been studied elsewhere. The limited evidence for this Native American herb preparation does not support its use in glycemic control. (Level I, American Diabetes Association level not applicable if studies show no benefit)

Tibetan medicine is a traditional system of healing that has influences from China, Persia, India, and Greece, incorporating concepts from Ayurveda as well as psychological, philosophical, and spiritual aspects of Buddhism. Herbalism, especially from the Himalayas, plays an important role. Although of poorer quality, one large RCT (n = 200) was available that examined individualized Tibetan herb prescription based on age, sex, personality, pulse, and urine characteristics in traditional diagnosis (118). Individual plant species and postulated mechanisms were not reported. At 6 months, the study suffered a large number of dropouts (44%); however, investigators analyzed data by intention-to-treat, and improvements were nevertheless reported in fasting plasma glucose, postprandial glucose, and HbA_1c values. No adverse effects were noted. These limited data are inconclusive regarding use of individual Tibetan herb prescriptions in type 2 diabetes. (Level II-2, C)

We identified six other specific combination herb formulations that have been studied in patients with diabetes, three from Ayurveda (D-400, MA-471, and Ayush-82) (33,119–121) and three from Siddha (Chendooram, Sandanapodi, and Kadal Azhinjil) (122–125). None have been examined in RCTs—only open-label prospective cohort studies or case reports.

Table 3 presents the controlled clinical trials of vitamin/mineral supplements for glycemic control in patients with diabetes. Of the studies examining vitamin and mineral supplements for glycemic control, the higher-quality RCTs (with Jadad scores of 3 or greater) are available for chromium, magnesium, vitamin E, and l-carnitine (126–137). Vanadium has been studied in only nonrandomized controlled trials (138–140).

Chromium (Cr3), a trace element in its trivalent form, is required for the maintenance of normal glucose metabolism. Experimentally, chromium deficiency is associated with impaired glucose tolerance, which can be improved with supplementation (35). Most individuals with diabetes, however, are not chromium deficient. In addition to glucose control, the supplement has been studied for its effects on weight control, lipids, and bone density. Its action is linked with glucose tolerance factor (GTF), and has been shown to increase the number of insulin receptors, to enhance receptor binding, and to potentiate insulin action. Some suggest that chromium picolinate is the preferred form because it is utilized more efficiently (141).

Of the eight RCTs examining chromium in those with diabetes or impaired glucose tolerance, preparations differ and the results are mixed. Among the larger trials, one using organic chromium in brewer’s yeast (n = 78) and another using chromium chloride (n = 180) reported decreases in fasting and postprandial glucose (127,128). However, another trial by Anderson (n = 110) utilizing chromium pidolate did not find changes in glycemic control (142). One large noncontrolled open-label trial of chromium picolinate followed 833 type 2 diabetic patients in China for up to 10 months. Investigators reported a decrease in fasting and postprandial glucose and a decrease in fatigue, excessive thirst, and frequent urination (143). These studies all reported no adverse effects. A recent meta-analysis by Althuis et al. (144) that included 15 RCTs (only 4 included diabetic individuals) reported that chromium had no effect on glucose or insulin concentrations in nondiabetic subjects; however, the data among patients with diabetes were inconclusive. Althuis et al. also suggested that more trials should be performed in North America, as the generalizabiltiy of trials conducted in China is unknown given regional differences in diet and nutritional status. (Level I, C)

Hypomagnesemia is common in patients with diabetes, especially those with glycosuria, ketoacidosis, and excess urinary magnesium losses. Deficiency of magnesium can potentially cause states of insulin resistance. Studies have examined magnesium’s potential role in the evolution of such complications as neuropathy, retinopathy, thrombosis, and hypertension. However, its role in glycemic control is unknown. Magnesium is a cofactor in various enzyme pathways involved in glucose oxidation, and it modulates glucose transport across cell membranes. It may increase insulin secretion and/or improve insulin sensitivity and peripheral glucose uptake. It has been shown to have no effect on hepatic glucose output and nonoxidative glucose disposal (35,40). Because it is an intracellular cation, it is difficult to measure accurately, and total body stores are seldom measured.

Of the seven RCTs examining magnesium supplementation for glycemic control in diabetes, only two small lower-quality trials from one investigator group (n = 8 and n = 9) reported a decrease in fasting plasma glucose and increase in postprandial insulin (145,146). Of the three highest-quality trials (Jadad score of 3), magnesium did not change blood glucose or HbA_1c (130–132). One trial (n = 128) did find a decrease in serum fructosamine, a short-term marker of glycemic control. Another study (n = 40) reported one subject with an exanthem and one who had transient gastrointestinal pain with magnesium supplementation. (Interestingly, the trial by Eriksson and Kohvakka [132] contained a study arm that administered vitamin C supplements, which unlike magnesium, did show improvements in glycemic control. To our knowledge, this is the only report of vitamin C for glucose control.) The available data for magnesium are mixed, and thus the evidence for efficacy in diabetes is inconclusive. (Level I, C)

Diabetes produces a state of increased free radical activity. The purported effects of vitamin E on glucose control relate to the vitamin’s potent lipophilic antioxidant activity, with possible influences on protein glycation, lipid oxidation, and insulin sensitivity and secretion. Through unknown mechanisms, it may also affect nonoxidative glucose metabolism (35,40).

Of the controlled trials that examined vitamin E for glucose control, the direction of the evidence for patients with type 2 diabetes is positive in four of six, with doses ranging from 100 to 1,600 mg/day for 2–4 months’ supplementation. The largest of these trials (n = 53), however, was a double-blind placebo-controlled crossover trial that found no change in serum glucose, fructosamine, or HbA_1c (136). One clinical trial examined patients with type 1 diabetes (n = 35) and reported decreases in protein glycosylation after 3 months of low-dose 100 IU/day vitamin E (57). Thus far, the available evidence for vitamin E in glycemic control is mixed and inconclusive. (Level I, C)

Several in vitro studies have helped to elucidate l-carnitine’s role in metabolism, suggesting that it acts as a modulator of fuel substrate utilization in cells, influencing free fatty acid and glucose oxidation. Few have examined it clinically in patients with diabetes. Three small controlled short-term metabolic trials examined the acute effects in type 2 diabetes (n = 18, n = 15, and n = 9), showing that intravenous carnitine (or its derivative acetyl-l-carnitine) administration can possibly effect insulin sensitivity and enhance glucose uptake and storage (137,147,148). There are no longer-term clinical studies of l-carnitine for glucose control and no studies of orally administered preparations. Thus, the available data are limited, and no conclusions can be made regarding its possible use in diabetes management. (Level I, A)

Vanadium has been described as either a nonessential nutrient or a nutrient that is required only in minute quantities, as no physiological role of the trace element has yet to be found (35,149). Human deficiency has not been documented. There are no accurate assays in clinical settings, and there is no recommended daily allowance. Vanadium exists in several valence forms, with vanadyl (+5) sulfate and sodium metavanadate (+4) being the most common supplement forms. Its mechanism of action in glycemic control is thought to be primarily insulin-mimetic with upregulation of insulin receptors. In animal models, it has been shown to facilitate glucose uptake and metabolism and to enhance insulin sensitivity. Clinically, it may enhance glucose oxidation and glycogen synthesis, and it may modulate hepatic glucose output (35). Three very small controlled clinical trials (n = 6–8) have reported decreased fasting blood glucose (138–140); two of these trials also reported significant changes in HbA_1c and insulin sensitivity (138,139). Two noncontrolled open-label studies, also with small sample sizes, nonetheless offer supporting evidence (150,151). Goldfine et al. (151) included type 1 diabetic patients (n = 5) who decreased their insulin requirements after 2 weeks of treatment. Gastrointestinal discomfort, including diarrhea, nausea, and flatulence, was reported by a large proportion of patients in all the vanadium trials. Organically chelated compounds, however, are thought to cause less gastrointestinal irritation than vanadium salts (149). The evidence for efficacy of vanadium in glucose control is suggestive, but as yet no RCTs are available. (Level II-1, C).

Also known as thioctic acid, a disulfide compound synthesized in the liver, α-lipoic acid is a potent lipophilic antioxidant. It is a cofactor in many multienzyme complexes and may also play a role in glucose oxidation (152). Experimental in vitro data have shown possible effects in enhancing glucose uptake in muscle and preventing glucose-induced protein modifications. One multiple-dosage controlled trial is available in patients with type 2 diabetes (n = 74), and it reported positive effects on glucose uptake and insulin sensitivity with 600–1,800 mg/day α-lipoic acid for 4 weeks; however, the trial showed no changes in fasting blood glucose (153). Another noncontrolled trial offers supportive evidence for a change in insulin sensitivity (152). The available data are limited and suggest that further elucidation of α-lipoic acids actions is needed. (Level II-3, C)

A total of 108 human trials of herbs and vitamin/mineral supplements for glycemic control were obtained. Most trials examined supplements as an adjunct to conventional treatment with diet and/or medication. Of the available trials, 58 were controlled (42 RCTs) and conducted specifically in individuals with diabetes or impaired glucose tolerance. Among these controlled trials, statistically significant treatment effects were reported in 88% (23 of 26) of those examining single herbs, 60% (3 of 5) of those examining combination herbs, and 67% (18 of 27) of those examining vitamin and mineral supplements. However, many trials were of poor quality. More than half of the RCTs (24 of 42, 57%) scored 2 or less on the Jadad scale. (No RCT achieved a score of 5.) Thirteen trials had sample sizes of 10 or fewer patients. In addition, there were generally few trials per supplement, making it difficult to draw definitive conclusions regarding efficacy. Nevertheless, no major safety concerns were reported in these trials. Few mild adverse effects, mainly gastrointestinal irritation, were reported for ginseng, Native American herb tea, TCM pill, magnesium, and vanadium (see Tables). For the following supplements, >50% of controlled clinical trials (at least two trials) suggested efficacy: Coccinia indica, Trigonella foenum, American ginseng, nopal, Gymnema sylvestre, Aloe vera, Momordica charantia, chromium, and vanadium. Of these, the best evidence is available for Coccinia indica and American ginseng. Supplements that appear effective but have only been studied in nonrandomized trials include Gymnema sylvestre, Aloe vera, and vanadium. Supplements that appear to be effective in short-term metabolic trials include Momordica, nopal, and l-carnitine.

In assessing the quality of the evidence, we employed the American Diabetes Association criteria for clinical guidelines (55). The evidence for the majority of supplements earned a C level rating, mostly for supportive evidence from RCTs with methodological flaws or uncontrolled studies, or conflicting evidence with weight supporting the recommendation (online appendix B). Those supplements that earned an A rating include Coccinia indica, American Ginseng, and l-carnitine, with supportive evidence from at least one adequate RCT. However, according to the criteria described by Weiger et al. (56), no herb or supplement has sufficient evidence to actively recommend or discourage its use among patients with diabetes. That is, evidence regarding efficacy is inconclusive or not rigorous enough to meet the outlined requirements of efficacy, yet the herb or supplement appears to be generally safe. Physicians should thus keep an open mind in advising patients who might already be using these supplements.

The American Diabetes Association and the American Dietetic Association do not have specific recommendations for the use of herb or vitamin/mineral supplements in people with diabetes. Broad recommendations for the general public are that healthy people at low risk for nutritional deficiencies meet their requirements with natural food sources. Those at increased risk for deficiencies, such as the elderly, strict vegetarians, those following very low-calorie diets, and other special populations, may benefit from multivitamin supplements (35).

Despite the lack of formal recommendations, the American Diabetes Association has acknowledged patient interest and use of CAM supplements for diabetes. In A Step-by-Step Approach to Complementary Therapies and Guidelines for Using Vitamin, Mineral, and Herbal Supplements (154,155), safety is the main theme. Practical information for patients on choosing supplements is outlined (e.g., looking for products with recognized symbols of quality: USP, NF, TruLabel, ConsumerLabs, etc.; looking for products with an expiration date; avoiding foreign products unless quality is known; and avoiding companies that make sensational claims or have misleading labels, etc). The American Diabetes Association also warns against combining supplements and prescription drugs without the physician’s knowledge and against stopping prescribed medication without the physician’s knowledge. They advise discontinuing supplements before medical procedures (e.g., surgeries or anesthesia) and in the event of an adverse effect.

Although the trials contained in this review reported very few adverse effects, other sources mentioned potential or theoretical effects for six supplements. Theoretical cross-allergenicity was mentioned with silymarin as a member of the aster family (daisy) and Trigonella as a member of the leguminosae family (peanuts), although no actual cases have been reported. The most important potential drug-herb interaction was that of garlic or Trigonella with warfarin, as both herbs may have limited anticoagulant properties. Momordica may increase risk of potassium depletion, so caution might be taken with those on laxatives or diuretics. Ginseng used in conjunction with monoamine oxidase inhibitors, phenelzine, or stimulants may cause an enhanced euphoric effect. Other adverse effects have been reported with Panax ginseng (Asian) (e.g., hypertension, hypotension, mastalgia, vaginal bleed, and insomnia), although the literature on diabetes has largely involved Panax quiquefolius (American). Rare topical reactions have been reported with nopal, garlic, and α-lipoic acid. Of note, one case of hypoglycemic coma has been reported with overdosage of Momordica charantia (36,37,39, www.naturaldatabase.com).

Currently, there is not yet sufficient evaluation of herbs, vitamins, and mineral supplements for glucose control in diabetes. Aside from relatively poor study methodological quality, this area of supplement research has been fraught with several complications.

First, the multiple constituent nature of botanical products has made standardization a challenging task. Proponents of herbal remedies caution that in standardizing to one constituent, resulting extracts may have lost a proportion of benefit as compared with the whole plant (156). Precise considerations of purity, chemical composition, and potency of derivatives may be grossly influenced by the age of the plant (especially of roots), the source location, the season of harvest, the method of drying and crude preparation, etc. In the literature we examined, several herb studies used “homemade” or otherwise unspecified preparations. Although individual companies have begun to standardize supplements, there is a general lack of consistency across the market. With vitamin and mineral supplements, these issues are less relevant.

In addition, the development of proper supplement regulation and safety codes has been slow. Currently, all dietary supplements (including herbal products) are regulated under the Dietary Supplement Health and Education Act of 1994 (DSHEA), which specifically differentiates supplements from drugs. Consequently, DSHEA does not require the extensive premarket approval that the Food and Drug Adminstration requires for a prescription drug, and although it calls for “good manufacturing practices [GMP],” the burden of proof that a supplement is unsafe lies with the government, leaving manufacturers to operate unchecked. This has contributed to skepticism among clinicians, and makes it especially difficult for physicians to responsibly recommend supplements to patients. In the absence of external regulation, the industry has taken steps to police itself. For example, the National Nutritional Foods Association (NNFA), representing about one-third to one-half of retailers and manufacturers of natural products in the U.S., has encouraged the adoption of strict, self-imposed GMP standards, as well as initiatives such as the TruLabel program (in which products are subjected to random laboratory testing by independent third-party auditors to verify contents) (42).

Research of vitamin and mineral supplements has also been hindered by a lack of accurate and meaningful assays that detect functional micronutrient deficiencies. In the case of chromium, for example, it is postulated that supplementation of targeted individuals might be more beneficial. Some speculate that positive results seen in large studies in diabetic patients in China may be due to the population’s relative chromium deficiency. However, without reliable assays, these theories have remained difficult to test (144).

Finally, the existing literature in this area includes a considerable amount of study population heterogeneity. Future research may need to more precisely define targeted diabetic populations with regard to disease classification, severity, optimal adjunctive interventions, and perhaps nutrient deficiencies. It will also be important to further elucidate mechanisms of action so that applicability to type 1 or type 2 diabetes can be clarified.

As interest in the potential benefit of herbs and supplements for diabetes grows, it will become increasingly important to monitor the progress of the clinical literature and to communicate these findings to patients. Based on this review, there is insufficient evidence to actively recommend or discourage use of any particular supplement, although most appeared to be generally safe. Preliminary evidence of several herbs and supplements suggest that further RCTs may be warranted. The seven most promising supplements include Coccinia indica, American ginseng, Momordica charantia, nopal, l-carnitine, Gymnema sylvestre, Aloe vera, and vanadium. Until more definitive studies help to clarify our questions, clinicians should remain cautious, yet open-minded, regarding adjunctive use of these supplements. They should be guided not only by sound clinical judgement, but also by patients’ preferences, needs, and values. As we further our understanding of herbs and dietary supplements, we might begin to develop a framework for a medical system capable of incorporating those complementary therapies proven to be beneficial.

Since our review of this topic, the report of a large multicenter trial (n = 3,654), which examined the effects of vitamin E with and without ramipril in high-risk patients with diabetes, has been published. Although this study was primarily concerned with cardiovascular events and mortality, it does report that there were no differences in change of HbA_1c between groups (157).

The authors thank Dr. Alan Moses and Karen Chalmers for their thoughtful review of the manuscript.