Effects of Vitamin C Supplementation on Glycemic Control and Cardiovascular Risk Factors in People With Type 2 Diabetes: A GRADE-Assessed Systematic Review and Meta-analysis of Randomized Controlled Trials

Type 2 diabetes remains a serious driver of chronic diseases, including cardiovascular disease. Improving glycemic control is important for managing type 2 diabetes; however, improving blood pressure (BP), lipid levels, and insulin sensitivity are also important targets to reduce the risk of cardiovascular disease in type 2 diabetes (1). Excess reactive oxygen species and oxidative stress are among the pathophysiological factors underlying impaired glucose metabolism and vascular complications of type 2 diabetes (2,3). Emerging evidence from predominantly short-term and small randomized controlled trials (RCTs) has suggested that antioxidant therapy may be effective in improving glycemic control and cardiovascular risk factors in people with type 2 diabetes (4–7).

Vitamin C is a water-soluble antioxidant that has been investigated therapeutically in people with type 2 diabetes. Prior systematic reviews of RCTs have focused on the effects of vitamin C supplementation on glycemic control (8), lipids (9), BP (10), and endothelial function (11), although these were not specific to people with type 2 diabetes. Other meta-analyses investigated effects of vitamins in people with type 2 diabetes (12–14); however, supplements were not exclusive to vitamin C only. While these prior reviews included many relevant studies, the substantial increase in published studies over recent years warrants an updated and more focused evaluation of vitamin C supplementation on cardiometabolic risk factors in people with type 2 diabetes. Furthermore, limited information on evidence quality and evidence certainty also warrants further evaluation to ascertain potential clinical translatability of vitamin C supplementation. The aim of this review, therefore, is 1) to investigate the efficacy of oral vitamin C supplementation in improving glycemic control, blood lipids, BP, and oxidative stress in people with type 2 diabetes and 2) to assess evidence certainty on the basis of the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) approach.

A systematic review and meta-analysis of RCTs (PROSPERO registry CRD42019140113) was undertaken, targeting effects of oral vitamin C supplementation on primary outcomes of glycemic control (HbA_1c, fasting glucose), blood lipids (triglycerides, total cholesterol, HDL cholesterol, LDL cholesterol), and BP (systolic, diastolic). Secondary outcomes were postprandial glucose, fasting insulin, HOMA of insulin resistance (HOMA-IR), clamp insulin sensitivity, and oxidative stress markers. Databases searched were Cochrane Library, Scopus, Embase, and MEDLINE-PubMed. Clinical trial registries (ClinicalTrials.gov, ANZCTR, EU Clinical Trial Register, and ISRCTN) were also searched for additional completed studies. Searches were not limited to English-language records. Database and registry searches were conducted three times from 5 September 2019 to 8 September 2020. Specific search strategies used are outlined in Supplementary Table 1.

Studies included were RCTs involving participants with type 2 diabetes. Exclusion criteria included comparisons that used intravenous (or nonoral) vitamin C administration, involved participants with type 1 diabetes, involved participants who were pregnant or <18 years old, lacked a control group, included other supplements besides vitamin C, and were <2 weeks in duration.

The selection of the studies was performed by two reviewers (S.A.M. and G.D.W.) independently using Covidence systematic review software (Veritas Health Innovation, Melbourne, Victoria, Australia). Studies were screened on the basis of their titles and abstracts. Studies that could not be ruled out had their full texts evaluated, and their eligibility was determined. Any discrepancies during selection of studies were resolved by reviewer consensus, although if a consensus could not be reached, a third reviewer (M.A.K.) adjudicated the decision.

The following data were extracted from selected RCTs by two reviewers (S.A.M. and G.D.W.) independently: author, publication year, study design, number of subjects per arm, losses per arm, sex, mean age, mean BMI, mean duration of diabetes, duration of supplementation, dose/dosage regimen, baseline vitamin C concentration, baseline HbA_1c, diabetes and other medications, and outcome data. Quantitative and qualitative reports on adverse effects observed were also recorded.

Any discrepancies during data extraction were resolved by reviewer consensus, although if a consensus could not be reached, a third reviewer (M.A.K.) adjudicated the decision. Authors of studies were contacted in instances where full-text articles could not be obtained or if data were unclear in the articles. Two full-text studies could not be obtained (15,16), and accurate BP values could not be established for two studies (17,18). The latter studies were subsequently excluded from BP analyses.

The Cochrane risk-of-bias tool (19) was used to evaluate bias in studies, including domains of random sequence generation, allocation concealment, blinding of participants and researchers, blinding of outcome assessment, incomplete outcome data, and selective reporting. Other biases, including compliance with treatment and diet and lifestyle confounding, were also assessed. Two reviewers (S.A.M. and G.D.W.) evaluated biases independently, with any discrepancies adjudicated by consensus.

Between-group differences in (within-group) pre-post changes were determined for outcomes. For parallel design trials, mean pre-post differences and SDs were used or calculated from the pretreatment and posttreatment data. If not provided, the pre-post change SD was calculated assuming a correlation coefficient of 0.7. This value was established on the basis of correlation coefficient values determined using data from two studies in our review (5,20). We also undertook sensitivity analyses of all outcomes using a correlation coefficient value of 0.5 for which no changes in statistical outcomes occurred compared with use of a 0.7 correlation coefficient (Supplementary Table 2). One study (21) did not provide variance data, and for that study, SD values consistent with average vitamin C and control group data across all other included studies were imputed.

Crossover trials were regarded similarly to parallel trials, with separate vitamin C and control arms. One crossover study (22) did not provide pretreatment values, and therefore, for that study, only posttreatment data were used.

Three studies (21–23) included multiple vitamin C groups with different doses. For these studies, all vitamin C groups were combined into a single group, with collective means and SDs determined according to recommended methods (24). Other studies (23,25–28) did not include posttreatment data for some outcomes. These studies were not included in the main analyses; however, they were included in sensitivity analyses, where pretreatment data were carried forward as posttreatment data. Significant results of main analyses were not affected by inclusion of these studies (Supplementary Table 2).

Pooled estimates and 95% CIs of effect sizes were calculated using random-effects modeling with DerSimonian-Laird methods in Review Manager software (29). Heterogeneity between studies was assessed using Cochran Q and I² statistics. Ninety-five percent prediction intervals (95% PIs) were additionally calculated to highlight study heterogeneity. A 95% PI estimates where the possible effects of vitamin C supplementation are to be expected for 95% of similar studies that might be conducted in the future (30). Mean difference (MD) effect sizes were used for all outcomes except clamp insulin sensitivity and oxidative stress outcomes, for which standardized MDs were used to account for significant variations in measurements/methods used. Data generated from Review Manager software were used to generate forest plots with Stata IC 16.1 software (StataCorp, College Station, TX). P < 0.05 was used to establish statistical significance for all statistical tests used.

Subgroup and meta-regression analyses were conducted on outcome measures that contained at least 10 studies. Prespecified subgroup analyses were based on BMI (nonobese <30 kg/m² vs. obese ≥30 kg/m²), plasma vitamin C concentration (hypovitaminosis C <23 μmol/L vs. ≥23 μmol/L), vitamin C dose used (lower dose <1,000 mg/day vs. high dose ≥1,000 mg/day), and treatment duration (shorter term <12 weeks vs. longer term ≥12 weeks). The rationale for these analyses was to assess potential sources of heterogeneity and to identify possible subgroups of differing vitamin C efficacy. Post hoc subgroup analyses investigating the potential impact of the control comparator group (active control vs. placebo) and overall study risk of bias (four or more of seven low-risk domains of bias vs. fewer than four of seven low-risk domains of bias) on outcomes were also undertaken. Heterogeneity as a result of potential modifying effects of various factors were further explored in (post hoc) meta-regression analyses using Stata software. Baseline biological factors of HbA_1c, BMI, age, duration of diabetes, and vitamin C concentration may plausibly affect the relative efficacy of treatment as explored previously (8–10), while trial-related factors, such as vitamin C dose, treatment duration, and study participant number, might further explain heterogeneity and efficacy of outcomes (8–10). Sensitivity analyses were conducted to evaluate only low-risk-of-bias studies on the basis of individual Cochrane risk-of-bias domains (Supplementary Table 2).

GRADEpro GDT software (31) was used to assess certainty of evidence for primary outcomes on the basis of areas of study design, risk of bias, inconsistency, indirectness, imprecision, and other considerations, such as publication bias, effect size, and potential confounding. Grades of evidence and their explanations include the following:

• High certainty: We are very confident that the true effect lies close to that of the estimate of the effect.

• Moderate certainty: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.

• Low certainty: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect.

• Very low certainty: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect (32).

We regarded a threshold of I² > 50% to denote inconsistency. For imprecision, we rated outcomes down if 95% CIs overlapped with nonclinically important effects. Minimal thresholds of clinically important changes considered were HbA_1c ≥0.5% (33), fasting glucose ≥1 mmol/L (34), total cholesterol ≥1 mmol/L (35), LDL cholesterol ≥0.3 mmol/L (36), HDL cholesterol ≥0.03 mmol/L (37), triglycerides ≥1 mmol/L, and systolic BP (SBP) and diastolic BP (DBP) >2 mmHg (38). We rated evidence certainty of outcomes down for risk of bias if sensitivity analyses investigating only low-risk studies on the basis of individual Cochrane risk-of-bias domains yielded inconsistent statistical outcomes across domains. We also rated evidence certainty of outcomes down for indirectness because our primary outcomes are surrogate rather than patient-important outcomes (39).

Small study effects were assessed using funnel plots and the Egger regression test when there were at least 10 studies, including at least 1 medium/large sample study (40).

Of 2,318 studies identified for screening, 68 full-text articles were reviewed in depth (Supplementary Fig. 1). Of these studies, 28 were deemed eligible for inclusion in the review (Table 1). Clinical trial registry searches yielded no additional studies to those found published in the searched databases. The main reasons for study exclusion were the wrong study design, repeat or redundant data, and the wrong participant population. Only eight studies (5,6,20,41–45) were rated as low risk in at least four of seven Cochrane risk-of-bias domains (Supplementary Fig. 2).

Vitamin C supplementation decreased HbA_1c to a statistically and clinically significant degree (MD −0.54% [95% CI −0.90, −0.17%]; P = 0.004 [95% PI −2.02, 0.94%]; n = 1,133 in 16 studies) (Fig. 1A), whereas fasting glucose was decreased to a statistically significant but not a clinically significant degree (MD −0.74 [95% CI −1.17, −0.31] mmol/L; P = 0.0007 [95% PI −2.44, 0.96 mmol/L]; n = 1,305 in 19 studies) (Fig. 1B) compared with control. Vitamin C also decreased postprandial glucose to a statistically significant degree (MD −0.95 [95% CI −1.83, −0.06] mmol/L; P = 0.04 [95% PI −4.67, 2.77 mmol/L]; n = 235 in four studies) (Supplementary Fig. 3) but had no significant effects on fasting insulin, HOMA-IR, or clamp insulin sensitivity (Supplementary Fig. 3) compared with control.

Vitamin C supplementation decreased SBP (MD −6.27 [95% CI −9.60, −2.96] mmHg; P = 0.0002 [95% PI −15.36, 2.82 mmHg]; n = 466 in eight studies) (Fig. 1C) and DBP (MD −3.77 [95% CI −6.13, −1.42] mmHg; P = 0.002 [95% PI −10.65, 3.11 mmHg]; n = 466 in eight studies) (Fig. 1D) to a statistically and clinically significant degree compared with control. A study excluded from BP analyses reported changes in BP as a proportion of participants above or below the threshold of 130/80 mmHg (18). Although raw data were not provided, that study showed a decrease in the proportion of participants above 130/80 mmHg from 88.8 to 42.1% after 12 months of vitamin C supplementation compared with a change from 92.1 to 83.5% after placebo supplementation.

Vitamin C supplementation decreased triglycerides (MD −0.20 [95% CI −0.36, −0.04] mmol/L; P = 0.01 [95% PI −0.79, 0.39 mmol/L]; n = 1,065 in 17 studies) (Fig. 2A) and total cholesterol (MD −0.27 [95% CI −0.43, −0.10] mmol/L; P = 0.001 [95% PI −0.89, 0.35 mmol/L]; n = 1,014 in 17 studies) (Fig. 2B) to a statistically significant but not clinically significant degree compared with control. Vitamin C had no statistically or clinically significant effects on LDL cholesterol (MD −0.23 [95% CI −0.48, 0.03] mmol/L; P = 0.08 [95% PI −1.25, 0.79 mmol/L]; n = 988 in 16 studies) (Fig. 2C), although it increased HDL cholesterol to a clinically important but not statistically significant degree (MD 0.06 [95% CI 0.00, 0.13] mmol/L; P = 0.06 [95% PI −0.17, 0.29 mmol/L]; n = 1,022 in 17 studies) (Fig. 2D) compared with control.

Three markers of lipid oxidation (malondialdehyde [MDA], F₂-isoprostanes, and LDL oxidation) were measured across more than one study and were included in meta-analyses. Vitamin C supplementation decreased plasma MDA to a statistically significant degree (standardized MD −1.25 [95% CI −1.88, −0.62]; P = 0.0001 [95% PI −3.44, 0.94]; n = 220 in five studies) but had no significant effects on F₂-isoprostanes or LDL oxidation compared with control (Supplementary Fig. 3). A study that measured susceptibility of LDL to oxidation that was not included in the analysis (28) reported no effect of vitamin C supplementation on this outcome. Another study found significantly lower plasma superoxide with vitamin C supplementation compared with placebo after 4 months (6). Also, skeletal muscle 2,7-dichlorodihydrofluorescein oxidation was found to decrease during insulin-stimulated conditions with vitamin C supplementation (42), although no change in whole-blood or skeletal muscle glutathione/glutathione disulfide ratio was found in that study.

Study heterogeneity was significant (I² > 50%) for all glycemic control and lipid outcomes, DBP, and MDA, although not for other outcomes. Subgroup and meta-regression analyses of potentially modifying factors revealed that the most prominent effects on heterogeneity and efficacy were due to study sample size, study duration, and baseline HbA_1c (Supplementary Tables 3 and 4). Residual heterogeneity decreased below I² = 50% for HbA_1c, triglycerides, total cholesterol, and HDL cholesterol when the modifying factor was sample size in meta-regression analyses (Supplementary Table 4). With each increase of one participant per study, vitamin C significantly improved HbA_1c (−0.009%), fasting glucose (−0.007 mmol/L), triglycerides (−0.003 mmol/L), total cholesterol (−0.004 mmol/L), LDL cholesterol (−0.004 mmol/L), and HDL cholesterol (0.002 mmol/L). With each increase of 1 day in study duration, vitamin C significantly improved HbA_1c (−0.006%), fasting glucose (−0.005 mmol/L), triglycerides (−0.003 mmol/L), total cholesterol (−0.003 mmol/L), LDL cholesterol (−0.004 mmol/L), and HDL cholesterol (0.002 mmol/L). Meta-regression also revealed some significant effects when the modifying factors were baseline HbA_1c and daily vitamin C dose. For every 1% increase in baseline HbA_1c, vitamin C significantly decreased HbA_1c (−0.47%), fasting glucose (−0.57 mmol/L), triglycerides (−0.18 mmol/L), total cholesterol (−0.32 mmol/L), and LDL cholesterol (−0.55 mmol/L). For every 100 mg/day increase in vitamin C dose, vitamin C increased fasting glucose (0.09 mmol/L) and LDL cholesterol (0.07 mmol/L). Baseline age, BMI, diabetes duration, and vitamin C concentration had no significant modifying effect on heterogeneity or efficacy with vitamin C supplementation in subgroup or meta-regression analyses for any outcome except HDL cholesterol, which decreased with increasing baseline age (−0.01 mmol/L per 1-year increase) and increasing baseline BMI (−0.04 mmol/L per 1 kg/m² BMI increase). The number of low-risk-of-bias domains in studies were not found to significantly modify the efficacy and heterogeneity findings in either subgroup or meta-regression analyses.

Sensitivity analyses that were based on individual Cochrane risk-of-bias domains found mostly consistent significant effects favoring vitamin C when considering only low-risk-of-bias studies for DBP, MDA, HbA_1c, and SBP (Supplementary Table 2). Other measures either did not significantly favor vitamin C across most bias domains (LDL cholesterol, HDL cholesterol, F₂-isoprostanes, LDL oxidation, clamp insulin sensitivity, HOMA-IR) or produced statistically mixed results across individual bias domains (fasting glucose, fasting insulin, triglycerides, total cholesterol) when considering only low-risk-of-bias studies (Supplementary Table 2). There were no statistically significant small study effects found that were suggestive of publication bias (40) (Supplementary Fig. 4).

Table 2 presents a summary of findings with GRADE evidence for primary outcomes, and Supplementary Table 5 shows GRADE evidence profiles for these outcomes. SBP was found to have a moderate certainty of decreasing statistically and clinically with vitamin C supplementation. HbA_1c and DBP were found to have a very low certainty of decreasing statistically and clinically with vitamin C supplementation. Fasting glucose, triglycerides, total cholesterol, LDL cholesterol, and HDL cholesterol were evaluated as having very low evidence certainty, whether changes with vitamin C were statistically significant or not.

Adverse effects reporting was included in 13 studies (5,6,18,20,21,41,42,46–51) (Supplementary Table 6). Five studies (6,20,41,46,49) reported no adverse effects of supplementation without specifying actual outcomes assessed. One study (42) reported temporary gastrointestinal upset in one participant with vitamin C, while another study (5) mentioned a feeling of depression in one participant with placebo supplementation. Another study (21) reported no significant difference between vitamin C and control groups for symptoms including nausea, vomiting, diarrhea, giddiness, headache, oral mucosal erosion, and fatigue, although the number of affected individuals was not specified. Foroghi et al. (48) noted that 15% of participants in the vitamin C treatment group complained of physical discomfort, with no reports in the placebo group. Ragheb et al. (51) found significantly better quality-of-life survey scores following vitamin C supplementation in relation to role limitation to physical health and emotional problems compared with control. The most comprehensive reporting of adverse effects was documented in the study of Gillani et al. (18) in which significantly fewer adverse events were reported with vitamin C supplementation (n = 23, including two hypoglycemic episodes, three hyperglycemic episodes, and five wrong timing of medication intake) compared with placebo (n = 89, including 30 hypoglycemic episodes, 45 hyperglycemic episodes, and 11 wrong timing of medication). Finally, renal and hepatic function tests were conducted in several studies (5,6,18,41,42,50), which found no significant differences in outcomes between vitamin C and control groups.

The major findings of this review are that vitamin C supplementation significantly lowers HbA_1c, fasting and postprandial glucose, triglycerides, total cholesterol, SBP and DBP, and MDA in people with type 2 diabetes. However, because of study biases, heterogeneity between studies, indirectness of outcomes, and imprecision of findings, evidence certainty is mostly very low across these measures. The 95% PI included zero for most significant outcomes. This means that while on average vitamin C might have beneficial effects, it might not always be beneficial in an individual (i.e., clinical, study) setting. A notable limitation of studies included was that they were predominantly short term (<6 months) with a relatively small participant number (n < 100). Interestingly, the only study that was >6 months and with >250 participants yielded the largest improvements in glycemic control, lipids, and BP (18). Therefore, this review highlights that vitamin C supplementation may be beneficial for improving glycemic control and reducing risk factors associated with cardiovascular disease in type 2 diabetes. However, further investigations using larger sample sizes and longer supplementation periods are required to confirm these potential cardiometabolic benefits.

The mean magnitude of change in HbA_1c across studies is consistent with a clinical improvement (−0.54%). In contrast, a prior systematic review found no significant effect of vitamin C supplementation on HbA_1c in a subgroup analysis of people with type 2 diabetes (−0.15%) (8). We also found a larger effect size for fasting glucose than did that prior review (−0.74 vs. −0.44 mmol/L) (8). Notably, our review contained 10 additional studies (n = 812) for these outcomes that were published since that prior review. For glycemic control outcomes, subgroup analyses revealed greater effects of vitamin C when study duration was ≥12 weeks. Meta-regression analyses found that factors of increasing baseline HbA_1c, increasing study duration, and increasing sample size were potential modifying factors in terms of treatment efficacy. Thus, vitamin C might have greater effects on glycemic control when taken long term and for patients with higher HbA_1c levels.

Potential antihyperglycemic mechanisms of vitamin C action have not been well established but might plausibly include antioxidant effects of vitamin C that promote improved insulin sensitivity (5,6). However, direct evidence is scant and weak to support this effect in people with type 2 diabetes, with no significant effects and very low evidence certainty found for effects of vitamin C supplementation on clamp insulin sensitivity and HOMA-IR in our review.

The mean decreases in SBP (−6.27 mmHg) and DBP (−3.77 mmHg) were consistent with clinically meaningful improvements. These findings lend support to a prior systematic review (10) that found improvements in both SBP (−4.71 mmHg) and DBP (−4.07 mmHg) with vitamin C in a subgroup analysis in people with type 2 diabetes. However, that review included studies with concomitant intake of other antioxidant compounds along with vitamin C, thus limiting conclusions about vitamin C specifically. Given the moderate level of evidence certainty for SBP findings, it is possible that one of the strongest beneficial effects of vitamin C supplementation in people with type 2 diabetes is a reduction in BP.

Potential hypotensive mechanisms of vitamin C might relate to its potential to enhance nitric oxide (NO) synthesis and bioavailability through its antioxidant actions. Vitamin C is believed to scavenge superoxide (52) and, therefore, may decrease NO reactivity with superoxide and limit formation of the potential vasculature-damaging reactive species peroxynitrite. Vitamin C has also been shown to preserve concentrations of the endothelial NO synthase cofactor tetrahydrobiopterin, in turn maintaining NO production through endothelial NO synthase (53).

The significant improvements for triglycerides (−0.2 mmol/L) and total cholesterol (−0.27 mmol/L) observed are arguably not of clinical significance. These findings partially support subgroup analyses in people with type 2 diabetes in a previous systematic review that similarly reported a statistically significant decrease in triglycerides (−0.15 mmol/L) with vitamin C supplementation (9). Despite small effect sizes for lipid outcomes overall, effects were larger for most lipid outcomes in subgroup analyses when study duration was ≥12 weeks and vitamin C dose was <1,000 mg/day. Meta-regression analyses found that factors of increasing baseline HbA_1c, increasing study duration, and increasing sample size were potential modifying factors in terms of treatment efficacy on lipid outcomes. Findings of increased efficacy with lower vitamin C doses implies an upper dose limit of effectiveness of vitamin C. However, it should be noted that there were only four studies in the review that used doses >1,000 mg/day and only two studies that used a dose of <500 mg/day. Thus, beneficial lipid-related effects may be most evident when modest doses are taken long term and for patients with higher HbA_1c.

We found significant effects of vitamin C supplementation on plasma MDA. A prior systematic review found significant effects of vitamin supplementation on MDA in people with type 2 diabetes (12). However, that review was not specific to vitamin C. Evidence for effects of vitamin C as a single-compound supplement on MDA is limited to a small number of studies and has significant between-study heterogeneity. Studies have shown vitamin C to be an effective antioxidant at reducing lipid peroxidation in human plasma (54,55). However, given current low-quality evidence in people with type 2 diabetes, further studies are required to evaluate effects of vitamin C on lipid peroxidation and other oxidative stress markers in type 2 diabetes.

A limitation of this review was a failure to extract relevant data from four studies (15–18) that may have been included in the analyses of lipids and BP. However, it is unlikely that inclusion of these studies would have decreased the effect sizes observed because abstract (15) and results (18) data from these studies have suggested significant improvements with vitamin C supplementation. Also, while we reported few notable adverse effects in studies, our review did not involve a comprehensive search of adverse effects. More focused reviews of adverse effects of vitamin C supplementation in people with type 2 diabetes are required.

Studies included in the review did not include patient-important outcomes, such as diabetic complications, cardiovascular disease incidence, or mortality. The lack of evidence on these outcomes in RCTs with vitamin C supplementation in people with type 2 diabetes is a gap in the literature that requires further investigation.

While we did include exploratory subgroup and meta-regression analyses of outcomes on the basis of plasma vitamin C status, most studies included in the review did not measure participants’ vitamin C concentrations. Therefore, it remains unclear whether baseline vitamin C status might affect the outcomes explored in our review. Finally, most studies included in the review involved the concomitant use of diabetes treatments, and some studies allowed BP- and lipid-lowering medications. Therefore, we cannot draw any clear conclusions about the efficacy of vitamin C as a stand-alone therapeutic agent for diabetes management but more so as an add-on therapy to existing diabetic therapies.

Vitamin C supplements are relatively inexpensive and widely available. Thus, vitamin C might be a potentially cost-effective treatment for people with type 2 diabetes. However, there is currently a lack of investigation of the cost-effectiveness of this approach. Thus, further investigation is required to evaluate the potential cost-effectiveness of vitamin C supplementation for the management of people with type 2 diabetes.

While our study only investigated effects of supplemental vitamin C on outcomes, it is possible that regular high dietary intakes of vitamin C could yield similar outcomes for people with type 2 diabetes, although there is a paucity of well-controlled studies in this area. Furthermore, foods rich in vitamin C may also be rich in additional nutrients, like vitamins, minerals, phytonutrients, and dietary fiber, that may have additional beneficial effects for people with diabetes. Future research should address effects of vitamin C supplements versus other antioxidant compounds or antioxidant-rich diets to establish whether the effects are specific to vitamin C or are more general with regard to vitamin C as an antioxidant.

Our findings, which are based largely on short-term studies with low evidence quality, suggest that vitamin C supplementation may be potentially effective for improving glycemic control and BP in people with type 2 diabetes. These effects may be greatest in people with higher HbA_1c and in those who regularly supplement with vitamin C. However, given limitations of studies, vitamin C supplementation cannot currently be recommended for type 2 diabetes management, with further investigations using larger sample sizes and longer supplementation periods that are powered to stratify effects on the basis of baseline glycemic control required to confirm beneficial effects of vitamin C supplementation.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. S.A.M. and G.D.W. conceived the study. S.A.M. and G.D.W. extracted and reviewed the data independently, and M.A.K. resolved any discrepancies. S.A.M. and G.D.W. performed the statistical analyses. S.A.M., G.D.W., and M.A.K. interpreted the data. S.A.M. wrote the manuscript with input from G.D.W. and M.A.K.