Randomised controlled trials (RCTs) have observed contrasting results on the effects of vitamin C on circulating biomarkers of glycaemic and insulin regulation. We conducted a systematic review and meta-analysis of RCTs testing the effect of vitamin C administration on glucose, HbA1c and insulin concentrations. Four databases (PubMed, Embase, Scopus and Cochrane Library) were used to retrieve RCTs published from inception until April 2016 and testing the effects of vitamin C in adult participants. The screening of 2008 articles yielded 22 eligible studies (937 participants). Overall, vitamin C did not modify glucose, HbA1c and insulin concentrations. However, subgroup analyses showed that vitamin C significantly reduced glucose concentrations (−0.44 mmol/l, 95% CI: −0.81, −0.07, P=0.01) in patients with type 2 diabetes and in interventions with a duration greater than 30 days (−0.53%, 95% CI: −0.79, −0.10, P=0.02). Vitamin C administration had greater effects on fasting (−13.63 pmol/l, 95% CI: −22.73, −4.54, P<0.01) compared to postprandial insulin concentration. Meta-regression analyses showed that age was a modifier of the effect of vitamin C on insulin concentration. Furthermore, the effect size was associated with baseline BMI and plasma glucose levels, and with the duration of the intervention. In conclusion, greater reduction in glucose concentrations observed in patients with diabetes, older individuals and with more prolonged supplementation. Personalised interventions with vitamin C may represent a feasible future strategy to enhance benefits and efficacy of interventions. Nevertheless, results need to be interpreted cautiously due to limitations in the primary studies analysed.
Diabetes is an important predictor of cardiovascular morbidity and mortality;1 six people die from diabetes’ complications every minute.2 These detrimental effects on human health are primarily linked to a higher level of oxidative stress and subsequent lowering of antioxidant capacity, which could be linked to the pathogenesis and later complications of type 2 diabetes.2 Oxidative stress is linked to an increased systemic inflammation, damage of the endothelium, impairment of insulin secretion and interference with glucose disposal.3 Vitamin C is an essential nutrient with antioxidant properties.4 Ascorbic acid (the reduced form of vitamin C) scavenges physiologically relevant reactive oxygen species and reactive nitrogen species.5 Moreover, vitamin C is involved in the regeneration of circulatory antioxidant molecules including glutathione, α-tocopherol, urate and β-carotene.5
Previous epidemiological analyses have showed that a higher antioxidant capacity was linked to reduced insulin resistance and better glucose control.6, 7 Moreover, these findings were corroborated by the findings of lower blood concentrations of vitamins C in patients with diabetes than healthy controls.8, 9 A lower risk of diabetes in individuals with higher plasma concentrations of ascorbic acid (odd ratio: 0.38; 95% CI: 0.28, 0.52) was reported in a 12-year cohort study.10 In addition, women at high risk of cardiovascular diseases received greater benefits from vitamin C supplementation (500 mg/day) in terms of diabetes risk (risk ratio: 0.89; 95% CI: 0.78, 1.02) during a 9.2 year (median) follow-up period.11
However, randomised controlled trials (RCTs) testing whether vitamin C had significant effects on biomarkers of glucose control (glucose, HbA1c concentrations) and insulin sensitivity have reported mixed findings.12, 13, 14, 15 Therefore, we conducted a systematic review of the literature on the topic by performing a meta-analysis of RCTs investigating whether vitamin C administration would exert favourable effects on glucose, HbA1c and insulin concentrations. An additional aim was to evaluate the influence of other potential factors which could modify the effects of supplemental vitamin C. These included the dose, duration and route of vitamin C administration but also the individual characteristics of the participant such as age, body mass index [BMI], health status, baseline blood glucose and vitamin C concentrations).
Four databases (PubMed, Embase, Scopus, and Cochrane Library) were searched to identify eligible studies published until April 2016. Reference lists of each eligible article were manually examined to find relevant reviews and published papers that were not included in the first set of results. These specific keywords were entered: vitamin C, ascorbic acid, hyperglycaemia, glucose blood level, glycaemia, haemoglobin, HbA1c, haemoglobin A1c, insulin, insulin resistance, insulin sensitivity, advanced glycosylation end products, glycated protein, glycoproteins and randomised controlled trials.
Criteria for inclusion were: 1) randomised controlled trial (design of the study and blinding were not taken into account); 2) adult participants aged 18 years independent of health status or phenotypic characteristics; 3) single administration of vitamin C; 4) report changes in blood glucose concentration, insulin, insulin sensitivity and HbA1c; 5) written in the English-language. In addition, we excluded trials employing glucose clamp techniques as circulating glucose levels were maintained at a pre-determined, narrow concentration by adjusting the glucose infusion rates as a measure or peripheral insulin sensitivity. Titles and abstracts were assessed by two independent investigators (ADW, NDW) to check eligibility for inclusion. A concordant decision between the two investigators meant that articles were either assessed in the full-text phase or removed from the main analysis. However, if disagreement between the two investigators was present, the article was selected for a more detailed assessment of the inclusion criteria. If needed, the two investigators reviewed together the article and agreed on a final inclusion/exclusion outcome.
Data extraction and quality assessment
Two independent investigators (AWA, ADW) performed the data extraction. The information extracted included: authors, year of publication, study design, sample size, health status, age, gender, dose and route of administration, study duration, baseline BMI, glucose and insulin concentrations and blood vitamin C levels at the beginning and at the end of the trial. Any discrepancy in the data extraction was discussed and an agreement between the two investigators was reached. We attempted to retrieve raw data from the graphs if results were not clearly reported in any other section of the manuscript.
We utilised the modified Jadad score to evaluate the risk of bias. The score rates the study quality by taking into account randomisation, blinding and dropout.18 A score of 0 to 2 indicates high risk while a score of 3 to 5 indicates a low risk of bias.
The net differences in glucose, insulin and HbA1c concentrations between the intervention and control groups at the end of each RCT were considered as the primary outcomes of the meta-analysis. A random effect model was used to analyse the data and to minimise the influence of heterogeneity on the study results. Forest plots were used to present graphically the results for glucose and insulin and HbA1c outcomes. STATA 12 (StataCorp. 2011. College Station, TX, USA) software was used to perform the meta-analysis. Inverse variance weighting was used to calculate the weighted mean differences (WMDs) and 95% confidence intervals. For crossover trials, we used a separate mean and SD for the intervention and control groups. This approach was chosen to have more conservative estimate of the effect size for these studies, which may reduce the power and the magnitude of the effect size.19
Heterogeneity was addressed by conducting stratified analyses of the studies according to participant characteristics [age, health status, blood glucose and plasma vitamin C concentrations (median)]. Additional stratifications were conducted to evaluate the influence of the dose, duration, design and quality of the trials. We restricted the interpretation of stratified analyses to subgroups with more than 5 studies.16
Meta-regression was used to assess whether age, BMI, vitamin C dose, study duration and baseline blood glucose and vitamin C concentrations modified the effect of vitamin C on glucose, insulin and HbA1c outcomes.
Egger’s regression test and funnel plots were used to evaluate the risk for publication bias.20 Calculation of Cochrane Q statistics evaluated the degree of heterogeneity across trials, which was considered as significant if P<0.1. Heterogeneity was classified according to the I2 test in the following categories:<25%=low risk, 25–75%=moderate risk, >75% =high risk.21
The different stages involved in the identification of the RCTs are presented in Figure 1. After removal of duplicates, the initial search of the literature databases yielded 2008 articles. Manual searching of relevant material produced an additional three studies. Title and abstract screening identified 180 candidate publications which were retrieved in full-text for further evaluation. Twenty-two studies were included in the systematic review.
The 22 studies included a total of 937 participants (median =28; range 8–170 participants for each study) and an overall median age of 49 (range 22–60) years. Sixteen trials had a parallel design and six were crossover trials. However, we should also clarify that two independent subgroups were identified in two articles, and therefore the final analyses were based on 24 trials (Table 1). Of these trials, 23 described the effect of vitamin C on glucose,12, 13, 14, 15, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38 9 on insulin23, 25, 27, 31, 33, 34, 35, 36, 37 and 10 on HbA1c concentrations.15, 22, 23, 24, 26, 28, 30, 32, 39 Nineteen trials collected fasting blood samples12, 13, 15, 22, 23, 24, 25, 26, 27, 28, 29, 30, 32, 34, 35, 37, 38 and four trials collected postprandial blood.14, 31, 33, 36
In addition, 13 trials comprised individuals with type 2 diabetes, eight trials recruited healthy individuals, two recruited those with type 1 diabetes and one study investigated coronary artery diseases patients (Table 1). Interventions lasted between 1 and 120 days (median duration: 30 days) and the doses administered varied between 72 and 6000 mg per day (median: 1000 mg).
The median Jadad score was 3 (range: 2–5). Bias risk was low for 17 studies, whereas a high risk of bias (score<3) was found in five studies (Table 1). The randomisation protocol was reported in six studies.15, 23, 25, 26, 28, 39 The duration of the wash-out period in the cross-over studies ranged between 1 to 7 weeks.
Vitamin C and glucose concentration
Overall, glucose concentration were not modified by vitamin C administration (−0.09 mmol/l, 95% CI: −0.34, 0.18, P=0.50); and studies were significantly heterogeneous (I2=55.6; P<0.01) (Figure 2). On the other hand, stratified analyses revealed a decrease in glucose concentration (−0.44 mmol/l, 95% CI: −0.81, −0.07, P=0.02) in diabetic patients (Table 2). Furthermore, vitamin C supplementation for >30 days induced a greater reduction of glucose levels (−0.53 mmol/l, 95% CI: −0.97, −0.10, P=0.02). Nevertheless, further analyses showed that participant characteristics (age, baseline vitamin C status) or study characteristics (design, dose, route and the quality of the included studies) had a minimal effect on the effect size. Meta-regression analyses revealed greater benefits of vitamin C administration on glucose concentrations in those with higher baseline BMI (β: −0.05, CI: −0.10, −0.01, P=0.03) and blood glucose concentration (β: −0.17, CI: −0.30, −0.04, P=0.01); in addition, effects were greater in studies with longer duration (β: −0.01, CI: −0.02, −0.001, P=0.04) (Figure 3). Baseline insulin concentration did not modify the effect of supplemental vitamin C on glucose concentrations (Supplementary Figure 1).
Vitamin C and insulin concentration
Supplemental vitamin C did not reduce plasma insulin concentration (−5.92 pmol/l, 95% CI: −14.29, 2.46, P=0.16) and studies were characterised by low heterogeneity (I2=27; P=0.20) (Figure 4a). Stratified analyses (fasting vs postprandial) showed a significant beneficial effect on fasting (−11.63 pmol/l, 95% CI: −22.73, −4.55, P<0.01) but not on postprandial insulin concentrations (4.46 pmol/l, 95% CI: −5.90, 14.82, P=0.81) (Table 2). Meta-regression analyses showed a greater insulin response to vitamin C administration in older (β: −0.70, CI: −1.29, −0.11, P=0.03) and in overweight and obese (β: −3.31, CI: −6.72, 0.10, P=0.05) participants. There was a tendency for a significant greater effect in participants with higher baseline plasma glucose concentration (β: −3.65, CI: −7.46, 0.15, P=0.06) (Figure 5) whereas a similar trend was not found for baseline insulin concentration (Supplementary Figure 2).
Vitamin C and HbA1c concentration
HbA1c concentrations were not modified by vitamin C administration (−0.02%, 95% CI: −0.19, 0.15, P=0.83) and generally responses were similar across studies (I2=0; P=0.60) (Figure 4b). Moreover, stratified and meta-regression analyses did not identify any confounding factor modifying the association between vitamin C and HbA1c concentration (Table 2 and Supplementary Figure 3).
Publication bias was not present for glucose (Supplementary Figure 4), insulin (Supplementary Figure 5) and HbA1c outcomes (Supplementary Figure 6). Moreover, the Egger’s regression test (β: −0.09, P=0.85; β: −0.30, P=0.82; β: −0.07, P=0.91 for glucose, insulin and HbA1c, respectively) showed a non-significant publication bias.
Vitamin C supplementation was not associated with beneficial effects on glucose concentrations. However, stratified analyses revealed a beneficial effect of supplemental vitamin C in patients with diabetes. Moreover, important information was also obtained from the meta-regression analyses, which showed a significant role of age, BMI, study duration, baseline concentrations of glucose and vitamin C in modifying the effects of vitamin C on measures of glycaemic control.
Overproduction of free radicals and/or low antioxidant capacity is recognised pathogenetic steps in the impairment of insulin signalling and development of diabetic hyperglycaemia. Previous experimental studies demonstrated that an increased production of free radicals in muscular tissue may impair insulin signalling and glucose uptake.40 The resultant insulin resistance was attributed to the production of stress-activated p38 mitogen-activated protein kinase (p38 MAPK) secondary to the oxidative insult.41 Moreover, insulin-secreting β cells are damaged by free radicals because of their low antioxidant capacity.42
Vitamin C has reducing properties and its physiological functions are based on its ability to reduce oxidised species43 and also involved in hydroxylation reactions, nucleic acid and histone demethylation, and synthesis of carnitine and catecholamines.44 The observational data from NHANES III showed that a decrease in vitamin C blood concentrations was associated a significant increase in risk for diabetes, but the effect was mostly explained by differences in dietary vitamin C intake in the population.9 Diabetes is also associated with a reduced renal reuptake of vitamin C.45 Patients suffered from diabetes-induced microalbuminuria have an increased urinary excretion of vitamin C and consequent reduced plasma levels of vitamin C.46, 47
The mechanistic basis linking diabetes to vitamin C deficiency were conceptualised in the ‘latent scurvy’ hypothesis, which proposed that a reduced transport of vitamin C could be linked to lower intracellular vitamin C levels. The hypothesis stems primarily from the structural similarity between glucose and ascorbic acid, and, possibly, a reduced re-conversion of intracellular DHA into ascorbate as a consequence of an increased oxidative stress in hyperglycaemic conditions.48, 49 Furthermore, the excessive systemic free-radical production associated with diabetes may further decline the antioxidant capacity of vitamin C, which may cause further vitamin C depletion.48 However, the biological plausibility of these alleged mechanisms is still elusive, which remain to be identified and tested in future studies.
Previously published meta-analyses of antioxidant vitamins supplementation in patients with diabetes demonstrated contradictory results. For instance, Akbar et al.50 reported no significant effects of antioxidant vitamins supplementation on glucose or insulin concentrations but a significant decrease in HbA1c. However, in the latter analysis, only three studies administered vitamin C alone.50 Similarly, another meta-analysis tested whether the administration of vitamin C may modify insulin sensitivity and found no significant improvement in homoeostasis model assessment (HOMA) index.51 In contrast with the above two meta-analyses, the data synthesis from five randomised controlled trials of vitamin C supplementation revealed significant reduction in fasting blood glucose, but with no significant effect on HbA1c.52 In contrast with the above studies, our results were based on the inclusion of a greater number of high-quality studies. In addition, our detailed subgroup and meta-regression analyses of these studies helped to explain the heterogeneity in response in glycaemic control after vitamin C administration.
This study reported greater benefits of supplemental vitamin C in diabetic patients and if individuals had higher baseline plasma glucose concentration and higher BMI. However, these are likely to be interdependent, since higher baseline glucose concentration and higher BMI are closely associated with diabetes. These positive results of supplemental vitamin C in patients with type 2 diabetes may indicate an opportunity for personalisation of such interventions in future clinical trials.
Previous investigation revealed a decrease in the production of reactive oxygen species in high BMI smokers but not normal BMI smokers.53 Similarly, vitamin C administration improved vascular health in disease-free individuals and with a low habitual consumption of fruit and vegetables.54 Furthermore, multivitamin supplementation for 6 years significantly improved cerebrovascular disease mortality in a population with lower micronutrient status.55 Taken together, these results may suggest that the level of vitamin concentrations at baseline may represent a critical factor in predicting the cardiovascular and metabolic responses to nutritional interventions tackling oxidative stress.56
The utility of any meta-analysis depends on the characteristics and qualities of the included studies. Two major limitations of our analysis are the small sample size and the modest quality of the trials. Because small and low-quality studies tend to produce inflated effect sizes,16 we interpret our results cautiously and advocate for studies with more robust study design to would take into account these methodological limitations.
In conclusion, the effects of supplemental vitamin C on glycaemic control were significantly modified by study design and participants’ characteristics. Stratified analyses revealed a significant improvement of glucose concentration in patients with diabetes, older participants and in studies with longer duration. However, suitably designed personalised nutritional interventions are needed to evaluate the efficacy of vitamin C for the protection of vascular health.
AWA is funded by the Republic of Iraq. PROSPERO Database registration: CRD42015023344.
The authors’ were responsible for the following tasks: AWA drafted the manuscript; AWA, JL, MS and JCM conceived the idea for the study and developed the search strategy; AWA, NDW and ADW conducted the search and summarised the data. All authors contributed to the data analysis, verification, writing and revising the manuscript.
About this article
Supplementary Information accompanies this paper on European Journal of Clinical Nutrition website (http://www.nature.com/ejcn)