Influence of Ascorbic Acid Supplementation on Type 2 Diabetes Mellitus in Observational and Randomized Controlled Trials; a Systematic Review with Meta-analysis

Purpose.There are controversial data regarding the beneficial effects of ascorbic acid (AA) supplementation in type 2 diabetes mellitus (T2DM). In this systematic review, we aimed to criticize the current relevant data from both observational and randomized controlled trials (RCTs). Methods. All observational and RCTs conducted to assess anti-hyperglycemic effects of AA in diabetics, published before January 2013, were included. To obtain all related studies Google Scholar, PubMed, Scopus, IranMedex, and Magiran web databases were searched. Exclusion criteria were animal studies, and studies conducted in Type 1 DM, children or pregnant women. Main outcome measures were fasting blood sugar (FBS), and glycated hemoglobin (HbA1c). According to degree of heterogeneity, fixed or random effect models were employed. Meta-analyses were done using Stats Direct software, version 3.0.97. The quality of included articles and publication bias were also assessed. Results. We selected 38 articles; 26 observational studies and 12 RCTs. Due to severe methodological heterogeneity in all observational studies and some of RCTs, we could pool data from only 5 RCTs in a meta-analysis. Single intake of AA versus placebo showed a significant effect on FBS; with the standardized mean difference (SMD):-20.59, 95% confidence intervals (95% CI):-40.77 to-0.4 (p= 0.04), but non-significant effect on HbA1c; SMD:-0.46, 95% CI:-1.75 to 0.84 (p= 0.4). Effect of other antioxidants with/without AA supplementation on FBSwere nonsignificant; SMD:-4.26 (p= 0.8), and SMD:-12.04 (p= 0.3), respectively. Also, their effect on HbA1c was non-significant; SMD: 0.53 (p= 0.11), and SMD: 0.28 (p= 0.34), respectively. Conclusions. Our study supports the positive effect of AA in reduction of FBS in diabetics, however, due to insufficient evidence ragarding long term safety of AA supplementation and limited number of RCTs, the long term use of this vitamin for its anti-diabetic properties cannot be strongly recommended.


INTRODUCTION
Diabetes mellitus (DM), as one of the most important worldwide health problems, shows an increasing prevalence.Currently, there are approximately 381 million diabetic patients, a figure that expects to rise to 592 million by 2035 (1).Various studies have established the key role of oxidative stress in the pathophysiology of diabetes and its complications (2)(3)(4).Oxidative stress reflects an imbalance between the formation of reactive oxygen species (ROS) and body's antioxidant defense system (2)(3)(4).
It has been shown that chronic diseases such as diabetes can diminish the antioxidative status of the body and increase the oxidative load (5).Under diabetic conditions, ROS are produced mainly through the glycation reaction.Oxidative stress can in turn promote glycation of hemoglobin (6) and impair the ability of β-cells of the pancreas for insulin secretion (7)._________________________________________ On the other hand, several epidemiological studies have shown that individuals with low concentration of antioxidants are at increased risk of diabetes complications (8,9).
Although the human body has its own antioxidant defense systems, this defense mechanism can be reinforced by the application of external source of antioxidants.Enzymatic or nonenzymatic antioxidants such as vitamins may have a role in oxidative stress (3,(10)(11)(12).The main source of majority of these antioxidants is the consumed food.Fruits, vegetables, and grains are among the richest sources of dietary antioxidants (5).Vitamins C and E are the well-known dietary antioxidants that may have beneficial effects against oxidative stress in diabetes.Several epidemiological studies have shown that individuals with low concentration of antioxidants are at increased risk of diabetes complications (8,9).Recently, Xu et al (13) published the effect of vitamin E (VE) supplementation on diabetes improvement.Along with their work, we aimed to critically and systematically, assess the effect of ascorbic acid (AA) in diabetes.
Vitamin C or AA is a hydrophilic antioxidant that depends on the employed dosage could have either prooxidant -or antioxidant effects (14).At low concentration, AA shows pro-oxidant function and helps in ROS formation, whilst its antioxidant function is found at higher concentrations (15,16).This vitamin as an essential micronutrient is acquired primarily through the consumption of fruits, and vegetables (17).However, AA is also readily available as an out of the counter drug that is usually consumed by healthy people.Data showed that 12.4% of the US adults take this vitamin as a dietary supplement (18).
While, high intake of AA might have a toxic effect (19), excess amounts of AA can be excreted through urine because of its water-soluble characteristic.Normally, consumption of doses up to 2000 mg/day is safe for the general population (20).The beneficial effect of AA consumption in diabetes is controversial.Some data support the idea that due to impairment of insulin secretion and ascorbate cycle in DM, AA is necessary to optimize the insulin secretory function of the islet cells (21).Another important function of AA is its ability to regenerate VE and some other antioxidants (22).
Overall, it has been hypothesized that the antioxidant effects of AA may improve the glycemic status of the DM, though not enough ABBRIVIATIONS 95% CI, 95% confidence intervals AA, ascorbic acid ADA, American Diabetes Association DBP, diastolic blood pressure FBS, fasting blood sugar HbA1c, glycated hemoglobin HDL-C, high density lipoprotein cholesterol IFG, impaired fasting glucose IGT, impaired glucose tolerance LDL-C, low-density lipoprotein cholesterol NGT, normal glucose tolerance OGTT, oral glucose tolerance test PL, placebo PRISMA, preferred reporting items for systematic reviews and meta-analyses RR, relative risk RTC, randomized controlled trials SBP, systolic blood pressure SMD, standardized mean difference STROBE, strengthening the reporting of observational studies in epidemiology T2DM, type 2 diabetes mellitus TC, total cholesterol TG, triglycerides VE, vitamin E WHO, World Health Organization evidences exist in the literature to strongly support this idea.The present meta-analyses systematic review is a novel work, because we focused on antihyperglycemic effect of AA according to data, separately extracted from both observational and randomized controlled trials.Specifically, our main outcome measures were the assessment of association between AA and FBS, HbA1c or incidence of diabetes.

Search strategy
All relevant available observational studies, including cohort, case-control or cross-sectional studies as well as randomized controlled trials (RCTs) conducted to assess anti-hyperglycemic effect of AA in human and published before January 2013, were included.To obtain all related studies Google Scholar, PubMed, Scopus, IranMedex, and Magiran web databases were systematically searched.The used search terms were "antioxidant", "diabetes", "vitamin C", "vit.C", "ascorbic acid", limited in human.In order to obtain the relevant information, we sent at least 3 e-mails to corresponding authors, whenever the data was incomplete.In the next step, the title and then the abstract of papers were examined.The duplicated articles were excluded, and then potentially eligible studies were retrieved for perusal in full text.

Study selection
All observational studies or RCTs that met the following criteria were included: 1) observational studies or RCTs involving T2DM patients; 2) fasting blood sugar (FBS), glycosylated hemoglobin (HbA1c), incidence of diabetes was estimated at baseline and at the end of a single intake of AA or its mixture with other antioxidants; 3) using oral glucose tolerance test (OGTT) or other accepted criteria for defining diabetes.
Studies that followed the above criteria, but conducted primarily in the healthy population, children, pregnant women, or patients with type 1 DM were excluded.Other exclusion criteria were defined as animal studies, in vitro studies, review articles, letters to the editor and thesis.No language restriction was set.

Data extraction and quality assessment
The following data were extracted: design of study, participants characteristics (age, sex), sample size in each groups (treatment or placebo), number of diabetic patients, number of participants received or not received AA, frequency and dosage of each supplementation (vitamin, placebo, or the mixture of vitamin with other antioxidants), concomitant therapy, duration of intervention, follow up, serum levels of AA, mean levels of FBS and HbA1c, case ascertainment, and endpoints.Primary outcome measures included the net changes in serum levels of FBS and HbA1c or incidence of diabetes after AA supplementation.The secondary outcomes were changes in serum concentrations of total cholesterol (TC), triglycerides (TG), low-density lipoprotein cholesterol (LDL-C), high density lipoprotein cholesterol (HDL-C), and insulin as well as diastolic (DBP) and systolic blood pressure (SBP).The methodological quality of each included observational/ cross-sectional studies, and RCTs was assessed using the STROBE (strengthening the reporting of observational studies in epidemiology) and Jadad scoring system (23,24), respectively.Two authors independently assessed the quality of all studies.Five selected items from the recommended checklist of STROBE (24) was used for quality assessment.The items included: a) 1 point for confirmed DM diagnosis, according to the American Diabetes Association (ADA) or other accepted criteria; b) 1 point for providing the eligibility criteria; c) 1 point for presenting the key elements of study design; d) 1 point if dietary intake was estimated, using a valid questionnaire or tool, to measure the intake of AA and/or other antioxidants' nutrient; and e) 1 point for describing the characteristics of study participants.Studies that did not fulfill more than two criteria (≤ 3 points) were classified as low quality.
In Jadad scale, for each part addressed in the study, one point was considered, with possible scores ranged between 0 and 5 (randomized, double-blinding, description of withdrawals and dropouts, generation of random numbers and allocation concealment) (24).RCTs with score <3 were considered as low quality.

Data synthesis and analysis
As mentioned previously, our main outcome measures were the assessment of association between AA and FBS, HbA1c or incidence of diabetes.According to World Health Organization (WHO) definition, diabetes mellitus is described as having a FBS ≥ 126 mg/dl or 2-h OGTT with 75 g glucose ≥ 200 mg/dl (25).Those with 2-h plasma glucose of 140 to 199 mg/dl, or FBS 100 to 125 mg/dl are classified as impaired glucose tolerance (IGT) or impaired fasting glucose (IFG), respectively (25).Subjects with FBS< 110 mg/dl are defined as normal glucose tolerance (NGT) (26).HbA1c, an index of average glucose control over the last 3 months, is considered as another marker of hyperglycemia when its value was greater than or equal to 6.5% (25).
Data from selected studies were extracted in the form of 2×2 tables by study characteristics.Included studies were weighted by effect size and then pooled.Data were analyzed using StatsDirect software version 3.0.97.Relative Risk (RR) and 95% confidence intervals (95% CI) were calculated using Mantel-Haenszel, Rothman-Boice (for fixed effects) and DerSimonian-Laird (for random effects) methods.The Cochran Q test was used to test heterogeneity.If the Q-statistic for heterogeneity was significant at the level of 0.1 or few included studies, random effects model was employed (27).In other cases the fixed effects model was used (28).Standardized effect size and 95% CI were calculated using Mulrow-Oxman (for fixed effects) and DerSimonian-Laird (for random effects) methods.The degree of heterogeneity was quantified using I 2 statistic which is an estimate of the total variation across studies due to heterogeneity (29).Egger and Begg-Mazumdar tests were used to evaluate publication bias indicators in funnel plot (30).A p value ≤0.05 was considered as statistically significant.

RESULTS
This study is reported according to PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) guideline (31).

Primary prevention of T2DM
Non-significant effect on DM risk, significant reduction in DM risk in without history high cholesterol, nonsignificant 2 or 3-way interactions among the agents for DM risk

Primary prevention of DM
Significant inverse association between baseline AA level and FBS after multiple adjusted, nonsignificant inverse association between AA intake and FBS
Due to using the 24 hours recall food intake tool, the net amount of AA intake was unclear.Moreover, because of the methodological heterogeneity, performing the meta-analysis to assess the effect of AA intake on glycemic improvement was not possible.The quality score of the included observational studies varied from 3 to 5. Most of the studies classified as high quality with scores ≥3.Within 11,471 participants in 14 observational studies (32,33,34,37,40,41,44,45,48,50,53,54,56,57) conducted to assess correlation between FBS and AA intake, 1,148 subjects were diabetics and 610 case were IGT (Impaired glucose tolerance) or IFG (impaired fasting glucose).According to these studies, DM or susceptibility to it presented in 15.32% of participants.There was a significant inverse association between manifestation of either DM or IGT/IFG and AA intake in 1,250 cases (10.90% of total participants).Interestingly, in the majority of cases (1,173 cases, 10.22% of total participants), sole intake of AA was used.

Association between HbA1c and AA intake
Overall, 11 observational studies assessed the association between HbA1c levels and AA intake (32,34,38,39,42,47, 49, 51, 52, 55, 56).Among 44,900 participants that were enrolled in these studies 3,466 and 632 subjects became diabetics or IGT/ IFG, respectively.In the other word, only in 9.13% of total participants in these studies, diabetes or susceptibility to diabetes development was found.A significant inverse association was found between HbA1c levels and antioxidants intake in 3,783 of the cases (8.42% of total participants), of which 1,686 cases (3.75% of total participants) used AA alone.

Association between DM risk and AA intake
From 51,857 participants in 5 eligible observational studies (35,39,43,46,50) which assessed the association between DM risk and AA intake, only 2,365 (4.56%) diabetics were defined.Between onset of diabetes and antioxidants' intake, a significant inverse association was found in 735 cases (1.42% of total participants).In all the participants, sole intake of AA was considered.

AA supplementation and diabetes in RCTs
Among selected studies that shown in Tables-2a, 2b, 3a, 3b, 4a and 4b, only two studies (59, 60) conducted just in males or females, and the rest was conducted in both genders.The types of RCTs among the 7 eligible studies were double-blind (59-61, 63, 67-69), in 3 studies (64, 65, 66) non-blind, 1 open-label (58), and 1 parallel (62).A number of the participants varied from 54 to 6,574 with treatment duration from 4wk to 9 years.In 7 studies (59, 60, 63-67) AA intake was compared to nutrient antioxidants with/without AA that later included vitamins E, β-carotene, zinc, selenium, copper, magnesium, and eicosapentaenoic acid.AA dosage varied from 120 mg up to 2 g/day.In most of these studies, the net change of FBS and HbA1c after consumption of AA was reported and shown a significant decrease in FBS or HbA1c levels after consumption (59, 61-65, 67).

Qualitative analysis
The quality of these 38 eligible studies was checked by STROBE or Jadad scale according to the design of the study (Table -1a and Table-6).

Meta-analysis
Main outcome measures included net changes in FBS, HbA1c or incidence of DM by single or mixture consumption of AA with other antioxidants.Most of the trials were too heterogeneous to perform a meta-analysis except 5 RCTs (58, 59, 61, 63, 65) that details of the findings are described below.

Effect of AA in comparison to placebo on FBS levels in diabetic patients
The summary for the standardized effect size of mean differences of FBS in diabetic patients "∆FBS" for AA therapy, in five trials, compared to placebo, in four studies (58, 59,61, 65), was -20.59 with 95% CI: -40.77 to -0.4 (p= 0.04, Figure 2-a).
The Cochrane Q test for heterogeneity indicated that the studies are not heterogeneous (p= 0.69) and could be combined, thus the fixed effects for individual and summary of effect size for standardized mean was applied.In the evaluation of publication bias, Egger regression test on the normalized effect vs. precision for all included studies of "∆FBS" among AA vs. placebo therapy in diabetic patients was 0.84 (95% CI: -2.22 to 3.91, p= 0.45) and Begg-Mazumdar Kendall's test on the standardized effect vs. variance indicated tau= 0.4, p= 0.48 (Figure 3-a).

Effect of antioxidants in comparison to placebo therapy on FBS level in diabetic patients
The summary for standardized effect size of mean differences of FBS in diabetic patients "∆FBS" for antioxidants therapy, in three included trials, compared to placebo, in two studies (59, 65), was -4.26 with 95% CI: -36.85 to 28.32 that was greater than null (p= 0.8, Figure 2-b).The Cochrane Q test for heterogeneity indicated that the studies are not heterogeneous (p= 0.94) and therefore could be combined.However, because of low number of included studies, the random effects for individual and summary of effect size for standardized mean was applied.Publication bias for "∆FBS" in diabetic patients among antioxidants vs. placebo therapy could not be evaluated because of too few strata.

Effect of AA plus antioxidants in comparison to placebo therapy on FBS levels in diabetic patients
The summary for standardized effect size of mean differences of FBS in diabetic patients "∆FBS" for AA plus antioxidants therapy, in four included trials, compared to placebo, in three studies (59, 63, 65), was -12.04 with 95% CI: -37.34 to 13.26 that was greater than null (p= 0.3, Figure 2-c).The Cochrane Q test for heterogeneity indicated that the studies are not heterogeneous (p= 0.52) and could be combined, thus the fixed effects for individual and summary of effect size for standardized mean was applied.In the evaluation of publication bias, Egger regression test on normalized effect vs. precision for all included studies was -1.5 (95% CI: -6.46 to 3.45, p= 0.32) and Begg-Mazumdar Kendall's test on standardized effect vs. variance indicated tau= -0.33, p= 0.33 (Figure 3-b).

Effect of AA in comparison to placebo therapy on HbA1c in diabetic patients
The summary for the standardized effect size of mean differences of HbA1c in diabetic patients "∆HbA1c" for AA therapy, in five included trials, compared to placebo, in four studies (58,59, 61, 65), was -0.46 with 95% CI: -1.75 to 0.84 that was greater than null (p= 0.4, Figure 4-a).The Cochrane Q test for heterogeneity indicated that the studies are heterogeneous (p < 0.0001) and could not be combined, thus the random effects for individual and summary of effect size for standardized mean was applied.In the evaluation of publication bias, Egger regression of normalized effect vs. precision for all included studies of "∆HbA1c" among AA vs. placebo therapy in diabetic patients was 11.52 (95% CI: 5.5 to 17.54, p= 0.01) and Begg-Mazumdar Kendall's test on the standardized effect vs. variance indicated tau= 0.6, p= 0.48 (Figure 5-a).

Effect of antioxidants in comparison to placebo therapy on HbA1c in diabetic patients
The summary for the standardized effect size of mean differences of HbA1c in diabetic patients "∆HbA1c" for antioxidants therapy, in three included trials, compared to placebo, in two studies (59, 65), was 0.53 with 95% CI: -0.11 to 1.17 that was greater than null (p= 0.11, Figure 4-b).The Cochrane Q test for heterogeneity indicated that the studies are not heterogeneous (p= 0.9) and could be combined, but because of the low number of included studies the random effects for individual and summary of effect size for standardized mean was applied.Publication bias for included studies for "∆HbA1c" in diabetic patients among antioxidants vs. placebo therapy could not be evaluated because of too few strata.

Effect of AA plus antioxidants in comparison to placebo therapy on HbA1c in diabetic patients
The summary for standardized effect size of mean differences of HbA1c in diabetic patients "∆HbA1c" for AA plus antioxidants therapy, in four included trials, compared to placebo, in three studies (59, 63, 65), was 0.28, ( 95% CI: -0.3 to 0.85 greater than null, p= 0.34, Figure 4-c).The Cochrane Q test for heterogeneity indicated that the studies are not heterogeneous (p= 0.63) and could be combined, thus the fixed effects for individual and summary of effect size for standardized mean was applied.In evaluation of publication bias, Egger regression on normalized effect vs. precision for all included studies of "∆HbA1c" among AA plus antioxidants vs. placebo therapy in diabetic patients was -1.87 (95% CI: -5.64 to 1.89, p= 0.17

Effect of AA in comparison to placebo therapy on TC in diabetic patients
The summary for standardized effect size of mean differences of TC in diabetic patients "∆TC" for AA therapy, in five included trials compared to placebo, in four studies (58, 59, 61, 65), was= -15.16 with 95% CI: -28.57 to -1.75 (p= 0.03, Figure 6-a).The Cochrane Q test for heterogeneity indicated that the studies are not heterogeneous (p= 0.71) and could be combined, thus the fixed effects for individual and summary of effect size for standardized mean was applied.In the evaluation of publication bias, Egger regression of normalized effect vs. precision for all included studies of "∆TC" among AA vs. placebo therapy in diabetic patients was 1.24 (95% CI: -1.65 to 4.13, p= 0.27) and Begg-Mazumdar Kendall's test on the standardized effect vs. variance indicated tau= 0.4, p= 0.48 (Figure 7 a-for the outcome of "∆TC" in studies with AA comparing to placebo therapy, b-for the outcome of "∆Tg" in studies with AA comparing to placebo therapy, c-for the outcome of "∆LDL-C" in studies with AA comparing to placebo therapy.

Effect of antioxidants in comparison to placebo therapy on TC level in diabetic patients
The summary for the standardized effect size of mean differences of TC in diabetic patients "∆TC" for antioxidants therapy, in three included trials compared to placebo in two studies (59, 65), was -12.66 with 95% CI: -53.04 to 27.73 that was greater than null (p= 0.54, Figure 6-b).The Cochrane Q test for heterogeneity indicated that the studies are not heterogeneous (p= 0.09) and could be combined however, because of the low number of included studies, the random effects for individual and summary of effect size for standardized mean was applied.Publication bias for included studies for "∆TC" in diabetic patients among antioxidants vs. placebo therapy could not be evaluated because of too few strata.

Effect of AA plus antioxidants in comparison to placebo therapy on TC level in diabetic patients
The summary for the standardized effect size of mean differences of TC in diabetic patients "∆TC" for AA plus antioxidants therapy, in three included trials compared to placebo in two studies (59, 65) was -14.45 with 95% CI: -52.84 to 23.95 that was greater than null (p= 0.46, Figure 6-c).The Cochrane Q test for heterogeneity indicated that the studies are not heterogeneous (p= 0.15) and could be combined however, because of few low number of included studies the random effects for individual and summary of effect size for standardized mean was applied.Publication bias for included studies for "∆TC" in diabetic patients among AA plus antioxidants vs. placebo therapy could not be evaluated because of too few strata.

Effect of AA in comparison to placebo therapy in Tg in diabetic patients
The summary for the standardized effect size of mean differences of Tg in diabetic patients "∆Tg" for AA therapy, in five included trials compared to placebo in four studies (58, 59, 61, 65), was= -21.93 with 95% CI: -48.55 to 4.69 that was greater than null (p= 0.11, Figure 8-a).The Cochrane Q test for heterogeneity indicated that the studies are not heterogeneous (p= 0.89) and could be combined, thus the fixed effects for individual and summary of effect size for standardized mean was applied.In evaluation of publication bias, Egger regression on normalized effect vs. precision for all included studies of "∆Tg" among AA vs. placebo therapy in diabetic patients was -0.01 (95% CI: -1.7 to 1.67, p= 0.98) and Begg-Mazumdar Kendall's test on the standardized effect vs. variance indicated tau= 0, p= 0.82 (Figure 7

Effect of antioxidants in comparison to placebo therapy on Tg level in diabetic patients
The summary for standardized effect size of mean differences of Tg in diabetic patients "∆Tg" for antioxidants therapy, in three included trials compared to placebo in two studies (59, 65), was -6.56 with 95% CI: -60.31 to 47.19 that was greater than null (p= 0.8, Figure 8-b).The Cochrane Q test for heterogeneity indicated that the studies are not heterogeneous (p= 0.89) and could be combined, but because of low number of included studies the random effects for individual and summary of effect size for standardized mean was applied.Publication bias for included studies for "∆Tg" in diabetic patients among antioxidants vs. placebo therapy could not be evaluated because of too few strata.

Effect of AA plus antioxidants in comparison to placebo therapy on Tg level in diabetic patients
The summary for standardized effect size of mean differences of Tg in diabetic patients "∆Tg" for AA plus antioxidants therapy, in three included trials compared to placebo in two studies (59, 65), was -7.18 with 95% CI: -63.34 to 48.98 that was greater than null (p= 0.8, Figure 8-c).The Cochrane Q test for heterogeneity indicated that the studies are not heterogeneous (p= 0.8) and could be combined, but because of low number of included studies the random effects for individual and summary of effect size for standardized mean was applied.Publication bias for included studies for "∆Tg" in diabetic patients among AA plus antioxidants vs. placebo therapy could not be evaluated because of too few strata.

Effect of AA in comparison to placebo therapy in LDL-C in diabetic patients
The summary for the standardized effect size of mean differences of LDL-C in diabetic patients "∆LDL-C" for AA therapy, in five included trials compared to placebo in four studies (58, 59, 63, 65), was -12.59 with 95% CI: -22.34 to -2.84 (p= 0.01, Figure 9-a).The Cochrane Q test for heterogeneity indicated that the studies are not heterogeneous (p= 0.07) and could be combined, thus the fixed effects for individual and summary of effect size for standardized mean was applied.In the evaluation of publication bias, Egger regression of normalized effect vs. precision for all included studies of "∆LDL-C" among AA vs. placebo therapy in diabetic patients was 0.

Effect of antioxidants in comparison to placebo therapy on LDL-C level in diabetic patients
The summary for the standardized effect size of mean differences of LDL-C in diabetic patients "∆LDL-C" for antioxidants therapy in three included trials compared to placebo in two studies (59, 65), was 22.38 with 95% CI: -0.51 to 45.26, greater than null (p= 0.06, Figure 9-b).The Cochrane Q test for heterogeneity indicated that the studies are not heterogeneous (p= 0.32) and could be combined, but because of the low number of included studies, the random effects for individual and summary of effect size for standardized mean was applied.Publication bias for included studies for "∆LDL-C" in diabetic patients among antioxidants vs. placebo therapy could not be evaluated because of too few strata.

Effect of AA plus antioxidants in comparison to placebo therapy on LDL-C level in diabetic patients
The summary for the standardized effect size of mean differences of LDL-C in diabetic patients "∆LDL-C" for AA plus antioxidants therapy, in three included trials compared to placebo in two studies (59, 65), was 13.59 with 95% CI: -9.14 to 36.32, greater than null (p= 0.2, Figure 9-c).The Cochrane Q test for heterogeneity indicated that the studies are not heterogeneous (p= 0.82) and could be combined, but because of low number of included studies the random effects for individual and summary of effect size for standardized mean was applied.Publication bias for included studies for "∆LDL-C" in diabetic patients among AA plus antioxidants vs. placebo therapy could not be evaluated because of too few strata.

Secondary outcome measures
Due to few available data for HDL-C, insulin, SBP, and DBP in eligible RCTs, heterogeneity assessment and pooling data were impossible.However, detail of these data are shown in tables 4(a, b) and 5(a, b).

DISCUSSION
Overall, observational studies have shown an inverse association between AA status or selfreported intake of AA with/without antioxidants and development of T2DM that was significant in 20.74% of the total participants.However, due to methodological heterogeneity, meta-analysis of the reported data was impossible.The meta-analysis of 5 eligible RCTs involving 385 subjects revealed that, even though a significant decrease in FBS levels following AA consumption in diabetic subjects might be seen, the changes were not significant when it comes to measuring the HbA1c levels after intervention with AA and/or other antioxidants.Although reduction in FBS was observed in two trials (59, 61), the greatest reduction was found in a the trials where AA was administered for at least 3 months and with a minimum dose of 1,250 mg per day (61).However, the meta-analysis of 2 RCTs that compared the effect of other antioxidants vs. placebo and also meta-analysis of 3 other RCTs that evaluated the mixed mode consumption of antioxidants and AA vs. placebo on FBS were not significant.Many data have established the key role of oxidative stress in the glycation of hemoglobin, peroxidation of cell membrane lipids, and finally tissue damage (3,70).A further evidence for the biological plausibility of these findings has been provided by recent studies in which the effects of AA in glucose metabolism have been assessed.AA has various functions against the oxidative process.This vitamin can scavenge ROS, inhibit the launch of chain reactions that lead to protein glycation, and protect against lipid peroxidation (71)(72)(73).Ascorbyl radical and dehydroascorbic acid are oxidation products of ascorbic acid that can be reduced back to AA by glutathione.In the meantime, the AA can support recycling back of VE and glutathione from their oxidized forms (71,74).
A variety of epidemiological and observational studies have been conducted to assess the effects of AA on oxidative stress conditions in diabetics and have reported conflicting results.Observational prospective cohorts have indicated that low levels of serum AA are associated with a reduced risk of diabetes (52,54,75).A large, 12-year populationbased study involving 21,831 participants in the European Prospective Investigation of Cancer-Norfolk Prospective Study identified 735 incident cases of diabetes and revealed a significantly lower DM risk by elevation of serum AA (odds ratio=0.38,CI 95%: 0.28-0.52)(39).In contrast, some RCTs have demonstrated no association between AA supplementation and the risk of T2DM.One randomized, open label, double-blind intervention trial reported no improvement in blood pressure, FBS, HbA1c, TG and HDL-C, after intervention with 500 mg/day of AA for 3 months in a group of 30 T2DM subjects, compared to baseline and also placebo group (58).The sole significant change was observed in the present study, was a remarkable improvement in TC and LDL-C levels.Some researchers suggested an improvement in glycemic control after AA supplementation (68,69).For instance, study of Delvarianzadeh et al. (61) showed a link between AA intake and HbA1c whereas Shoff et al. (55) reported non-significant difference in mean HbA1c among the highest vs. lowest quintiles of AA supplementation in 2,141 subjects.Accordingly, our meta-analysis showed no significant association between HbA1c and AA intake vs. placebo.
Several possible reasons may account for the observed controversial results in observational studies and RCTs.Under different physiological conditions, AA can have preoxidant or antioxidant effects (76).However, required serum concentration and doses that are needed to induce the oxidative stress are different from those that are required for induction of other effects.This feature of AA is not usually considered when epidemiological studies are performed (76).Normal or high physiological level of AA (60-100 µmol/l) can attenuate the oxidative damage (77-79), while its prooxidant function that occurs in the presence of some metals such as copper and iron, can promote the oxidative damage (80).A level of 200 mg AA that is usually obtained from vitamin C-rich foods produces an average serum concentration of 90 µmol (81).In normal conditions, physiological availability of AA is low, which mostly is due to the instability of this vitamin, poor intestinal absorption, and easy excretion (75).High levels of glucose in the blood can induce intracellular AA deficiency, which is caused by competition of glucose with this vitamin for tissue uptake (82)(83)(84).Moreover, bioavailability of AA also depends on amounts of transporting proteins and their binding affinity (85) which is impaired in chronic conditions such as diabetes.It is known that cellular uptake of AA is orchestrated by blood levels of both glucose and insulin (86,87).Therefore, the presence of hyperglycemia in diabetic subjects could increase the urinary loss of this vitamin and subsequently results in lower levels of AA in diabetics (88).Taking it as a whole, it is clear that diabetic subjects require higher doses of AA than recommended dietary allowance (RDA).Among healthy men and women, the daily RDA of AA is 90 and 75 mg, respectively.In hyperglycemic subjects, this measure should be increased by 35 mg/day (89).AA intake in all RCTs included in our meta-analysis was at least 200 mg daily from 8 weeks to 3 months.Two RCTs that investigated AA supplementation of 200 and 1,250 mg daily, showed a significant effect on FBS versus placebo over 8-12 weeks (59, 61).However, by daily supplementation of 200 (63) or 500 mg daily (58) for 12 weeks, or 500 mg daily for 4-8 weeks (65) the FBS was not improved versus placebo.After pooling data, effect size was increased and showed the significant beneficial effect of AA intake on FBS versus placebo.Due to few RCTs, we could not perform subgroup analysis to determine type of AA supplemet, effective dosage and treatment duration.
Although, observed benefits in some of our observational studies were related to coadministration of AA and antioxidants (34,41,42,47,52), in majority of studies, single AA was taken (32,36,37,39,40,45,49,53,54).It is thought that the benefits reported in the epidemiological studies in which AA and/or antioxidants are coadministrated, might be related to intake of higher amounts of fruits and vegetables, as a complex mixture of micronutrients or synergistic interaction between natural antioxidants (56).It is expected that co-administration of two or more vitamins and antioxidants is more effective than single supplementation ( 90), and selecting the kind of antioxidants to combine together has a crucial importance.The benefits of mixture consumption of a hydrophilic (AA) with a hydrophobic (VE) antioxidant have been reported ( 91) as this benefit was shown in our included studies (34,41,42,47,52).However, as we mentioned previously, observed benefitial effects in our RCTs and most of observational studies were related to sole intake of AA.Steinberg et al ( 92) suggested that since many of pathological changes seen in diabetes have been developed few years before clinical presentation, it may take more than 5 years for antioxidant therapy to reverse the pathological changes.It is clear that beneficial effect of AA supplementation after a few weeks cannot be documented based on HbA1c measurement, because this measure reflects a mean glucose level over the last three months.We have to mention that total daily dosages of AA in included studies had a wide variation ( 93).In some included studies in our systematic review, small dose of AA and/or other antioxidants was employed while in others, high dose of antioxidants was administered.It should be noted that large intake of AA does not necessarily guaranty the full absorption.In fact, plasma AA concentration is tightly controlled by three mechanisms: intestinal absorption, tissue transport, and renal reabsorption.In addition, in response to sudden high oral intake of AA, exess AA is largely excreted in the urine (94)(95)(96).AA is generally considered safe in normal individuals, but in special conditions such as renal stones, hyperoxaluria, dialysis, renal failure or kidney transplantation, administration of high amounts of AA could be harmful due to oxalate formation (97).One of the diabetes complication is diabetic nephropathy that happens few years after onset of diabetes.Therefore, high dose of AA therapy should be avoided in these conditions.AA administration is also contraindicated in patients with systemic iron overload due to increased iron absorbtion, and in angioplastic patients due to increased risk of cardiovascular diseases (95,97).Respectively, Lee et al (95) found that high dose of AA could increase the rate of cardiovascular complications in a 15 year prospective study.
Analysis of our secondary outcomes revealed a significant improvement in TC, and LDL-C levels in the AA group compared to placebo or antioxidants treated subjects, a result that was consistent with the findings of Ginter et al (98).As it is shown in human and animal studies, long term AA intake leads to elevation of ascorbate concentration in the liver, which subsequently could result in an enhanced rate of cholesterol transformation to bile acids (99,100).
We could not see the beneficial effects of other antioxidants; eicosapentaenoic acid (59), or VE (63, 65) with/without AA on FBS, this finding was similar to results obtained by previous metaanalysis in diabetics (13,101).
Our study supports the role of AA in reduction of FBS in diabetics, however, due to lack of evidences on long term safety of the vitamin supplementation and insufficient available RCTs involved in our meta-analysis, we cannot strongly recommend the long term use of this vitamin for its anti-diabetic properties.Although American Diabetes Associationhas recommended 8-10 daily servings of fruits and vegetables as a source of AA for diabetic patients (102), we think, according to our findings regarding the positive effect of AA on FBS, serving the only natural source of AA is not enough and AA supplements should be considered in diabetics.
Our study has some strengths and of course some limitations.Firstly, this meta-analysis, for the first time, assessed the effects of AA supplementation on plasma concentrations of FBS and HbA1c.Moreover, the trials included in this meta-analysis were all RCTs, which allow reliable inferences about causality.Among our limitations, we should mention that we needed to include small trials with limited subjects and varied dosage of AA.This variation limited our ability to performe subgroup analysis and definie dosage and duration of AA recommendations in T2DM.Our second limitation was considerable trial heterogeneity.Our third limitation was the wide variation in quality of RCTs included in this meta-analysis.Of 12 trials, only 4 trials had score equal to 4 (high-quality studies) and the others were catergorized as low quality studies.Moreover only for few studies, we were able to pool the data and perform the metaanalysis.These conditions could affect the confidence of this meta-analysis.Publication bias could be a potential limitation in this study.However, we tried to explore the possibility of this bias by using funnel plot and Egger's test and found that publication bias did not have significant effect on the results of AA and/or other antioxidants supplementation on FBS and lipid profiles.Finally, except in study performed by Lee et al. (95), no study has yet assessed the long-term safety and efficacy of AA intake on other tissues and organs.We concluded that our systematic review of observational studies and meta-analysis of RCTs identified a significant correlation between AA and improvements in FBS level in diabetics.However, yet large-scale randomized trials are needed to investigate the effect of AA supplementation on FBS and HbA1c.Taken together, it should be appropriate to suggest that diabetic subjects without contraindication of AA intake might benefit more from taking a combination of antioxidants, from either natural sources and/ or fortified foods, and AA supplementation.

Figure 2 .Figure 3 .
Figure 2. Individual and pooled relative risk for the outcome of "changing in fasting blood sugar" in the studies in diabetic patients; a-AA comparing to placebo therapy, b-antioxidants comparing to placebo therapy, c-AA plus antioxidants comparing to placebo therapy

Figure 4 .
Figure 4. Individual and pooled relative risk for the outcome of "∆HbA1c" in diabetic patients; a-AA comparing to placebo therapy, b-antioxidants comparing to placebo therapy, c-AA plus antioxidants comparing to placebo therapy.

Figure 5 .
Figure 5. Publication bias indicators for the outcome of "∆HbA1c" in diabetic patients; a-AA comparing to placebo therapy, b-AA plus antioxidants comparing to placebo therapy.

Figure 6 .
Figure 6.Individual and pooled relative risk for the outcome of "∆TC" in diabetic patients; a-AA comparing to placebo therapy, b-antioxidants comparing to placebo therapy, c-AA plus antioxidants comparing to placebo therapy.

Figure 7 .
Figure 7. Publication bias indicators in diabetic patients;a-for the outcome of "∆TC" in studies with AA comparing to placebo therapy, b-for the outcome of "∆Tg" in studies with AA comparing to placebo therapy, c-for the outcome of "∆LDL-C" in studies with AA comparing to placebo therapy.

cFigure 8 .
Figure 8. Individual and pooled relative risk for the outcome of "∆Tg" in diabetic patients; a-AA comparing to placebo therapy, b-antioxidants comparing to placebo therapy, c-AA plus antioxidants comparing to placebo therapy.

Figure 9 .
Figure 9. Individual and pooled relative risk for the outcome of "∆LDL-C" in diabetic patients; a-AA comparing to placebo therapy, b-antioxidants comparing to placebo therapy, c-AA plus antioxidants comparing to placebo therapy.

Table 1 -
a. Observational studies of ascorbic acid intake and type 2 diabetes mellitus

Table 2 -a. Randomized
controlled trials of ascorbic acid intake and type 2 diabetes mellitus

Table 2 -
b. Response to treatment in eligible RCTs of ascorbic acid intake and type 2 diabetes mellitus

Table 3 -
a. Secondary outcomes in eligible RCTs of ascorbic acid intake and type 2 diabetes mellitus Ref.Mean±SD baseline TC (mg/dl) in study groups and groups size (n) Mean±SD baseline TG (mg/dl) in study groups and groups size (n) Mean±SD baseline LDL-C (mg/dl) in study groups and groups

Table 3 -
b. Response to secondary outcomes in eligible RCTs of ascorbic acid intake and type 2 diabetes mellitus

Table 4 -
a. Secondary outcomes in eligible RCTs of ascorbic acid intake and type 2 diabetes mellitus RefMean±SD baseline HDL-C (

Table 4 -
b. Response to secondary outcomes in eligible RCTs of ascorbic acid intake and type 2 diabetes mellitus

Table 5 -
a. Secondary outcomes in eligible RCTs of ascorbic acid intake and type 2 diabetes mellitus

Table 5 -
b. Response to secondary outcomes in eligible RCTs of ascorbic acid intake and type 2 diabetes mellitus

Table 6 .
Quality assessment of RCTs included in the meta-analysis Y: yes; U: unclear.

4 a
Effect size meta-analysis plot [fixed effects]

a
Effect size meta-analysis plot [random effects]

a
Effect size meta-analysis plot [fixed effects] Nutrition Survey of people aged 65 years and over.DiabetMed, 2004; 21:677-684.43.Montonen J, Knekt P, Järvinen R, Reunanen A. Dietary antioxidant intake and risk of type 2 diabetes.Diabetes Care, 2004; 27:362-366.44.Merzouk S, Hichami A, Madani S, et al.Antioxidant status and levels of different vitamins determined by high performance liquid chromatography in diabetic subjects with multiple complications.Gen Physiol Biophys, 2003; 22:15-27.45.Birlouez-Aragon I, Delcourt C, Tessier F, Papoz L; POLA Study Group.Pathologies Oculaires Liées à l'Age.Associations of age, smoking habits and diabetes with plasma vitamin C of elderly of the POLA study.Int J Vitam Nutr Res, 2001; 71:53-59.46.Ford ES.Vitamin supplement use and diabetes mellitus incidence among adults in the United States.Am J Epidemiol, 2001; 153:892-897.47.Boeing H, Weisgerber UM, Jeckel A, Rose HJ, Kroke A. Association between glycated hemoglobin and diet and other lifestyle factors in a nondiabetic population: cross-sectional evaluation of data from the Potsdam cohort of the European Prospective Investigation into Cancer and Nutrition Study.Am J Clin Nutr, 2000; 71:1115-1122.48.Firoozrai M, Soloukizadeh N, Danesh Doust L, Ghafari M. Levels of ascorbic acid (vitamin C) in plasma and mononuclear leukocytes of patients with type II diabetes mellitus.J Iran University of Medical Sciences, 2000; 21: 207-212.49.Sargeant LA, Wareham NJ, Bingham S, et al.Vitamin C and hyperglycemia in the European Prospective Investigation into Cancer-Norfolk (EPIC-Norfolk) study: a population-based study.Diabetes Care, 2000; 23:726-732.50.Will JC, Ford ES, Bowman BA.Serum vitamin C concentrations and diabetes: findings from the Third National Health and Nutrition Examination Survey, 1988-1994.Am J Clin Nutr, 1999; 70:49-52.51.Maxwell SRJ, Thomason H, Sandler D, et al.Poor glycemic control is associated with reduced serum free radical scavenging (antioxidant) activity in noninsulin-dependent diabetes mellitus.Ann Clin Biochem, 1997; 34: 638-644.52.Sundaram RK, Bhaskar A, Vijayalingam S, et al.Antioxidant status and lipid peroxidation in type II diabetes mellitus with and without complications.Clin Sci (Lond), 1996; 90:255-260.53.Feskens EJ, Virtanen SM, Räsänen L, et al.Dietary factors determining diabetes and impaired glucose tolerance.A 20-year follow-up of the Finnish and Dutch cohorts of the Seven Countries Study.Diabetes Care, 1995; 18:1104-1112.54.Sinclair AJ, Taylor PB, Lunec J, Girling AJ, Barnett AH.Low plasma ascorbate levels in patients with type 2 diabetes mellitus consuming adequate dietary vitamin C. Diabet Med, 1994; 11:893-898.55.Shoff SM, Mares-Perlman JA, Cruickshanks KJ, et al.Glycosylated hemoglobin concentrations and vitamin E, vitamin C, and beta-carotene intake in diabetic and nondiabetic older adults.Am J Clin Nutr,1993; 58:412-416.56.Lysy J, Zimmerman J. Ascorbic acid status in diabetes mellitus.Nutr Res, 1992; 12: 713-720.57.Institute of Medicine (U.S.), Panel on Dietary Antioxidants and Related Compounds, Dietary reference intakes for vitamin C, vitamin E, selenium, and carotenoids: a report of the Panel on Dietary Antioxidants and Related Compounds, Subcommittees on Upper Reference Levels of Nutrients and of Interpretation and Use of Dietary Reference Intakes, and the Standing Committee on the Scientific Evaluation of Dietary Reference Intakes, Food and Nutrition Board, Institute of Medicine, Washington D.C., National Academy Press., 2000.58.Bhatt J, Thomas S, MJ N. Effect of oral supplementation of vitamin C on glycemic control and lipid profile in patients with type 2 diabetes mellitus.Int J Pharm Pharm Sci, 2012; 4: 524-527.59.Shakouri Mahmoudabadi MM, Djalali M, Djazayery SA, et al.Effects of eicosapentaenoic acid and vitamin C on glycemic indices, blood pressure, and serum lipids in type 2 diabetic Iranian males.J Res Med Sci, 2011; 16: 361-367.60.Song Y, Cook NR, Denburgh MV, Manson JE.Effects of vitamins C and E and β-carotene on the risk of type 2 diabetes in women at high risk of cardiovascular disease: a randomized controlled trial.Am J Clin Nutr, 2009; 90: 429-437.61.Delvarianzadeh M, Abbasian M, Norouzi P. Effect of vit C supplement on fasting blood glucose and plasma lipid level in type II diabetic patients.Full texts in Persian.Knowledge and Health, 2008; 3: 39-43.62. Afkhami-Ardekani M and Shojaoddiny-Ardekani A. Effect of vitamin C on blood glucose, serum lipids & serum insulin in type 2 diabetes patients.Indian J Med Res, 2007; 126: 471-474.63.Farvid MS, Siasi F, Jalali M. The impact of vitamin C and E, magnesium and Zinc on glycemic control and insulin resistance in type II diabetic patients.Tehran University Medical Journal, 2007; 64: 67-75.64.Czernichow S, Couthouis A, Bertrais S, et al.Antioxidant supplementation does not affect fasting plasma glucose in the supplementation with antioxidant vitamins and minerals (SU.VI.MAX) study in France: association with dietary intake and plasma concentrations.Am J Clin Nutr, 2006; 84: 395-399.65.Farvid M, Saadat N, Valai N, Aminpour A. Comparison of the supplementary effects of vitamins E, C and their combination on blood