Antioxidant agents for delaying diabetic kidney disease progression: A systematic review and meta-analysis

Background Oxidative stress is a key player in the genesis and worsening of diabetic kidney disease (DKD). We aimed at collecting all available information on possible benefits of chronic antioxidant supplementations on DKD progression. Study design Systematic review and meta-analysis. Population Adults with DKD (either secondary to type 1 or 2 diabetes mellitus) Search strategy and sources Cochrane CENTRAL, Ovid-MEDLINE and PubMed were searched for randomized controlled trials (RCTs) or quasi-RCTs without language or follow-up restriction. Intervention Any antioxidant supplementation (including but not limited to vitamin A, vitamin C, vitamin E, selenium, zinc, methionine or ubiquinone) alone or in combination. Outcomes Primary outcome was progression to end-stage kidney disease (ESKD). Secondary outcomes were change in albuminuria, proteinuria, serum creatinine and renal function. Results From 13519 potentially relevant citations retrieved, 15 articles referring to 14 full studies (4345 participants) met the inclusion criteria. Antioxidant treatment significantly decreased albuminuria as compared to control (8 studies, 327 participants; SMD: -0.47; 95% CI -0.78, -0.16) but had apparently no tangible effects on renal function (GFR) (3 studies, 85 participants; MD -0.12 ml/min/1.73m2; 95% CI -0.06, 0.01). Evidence of benefits on the other outcomes of interest was inconclusive or lacking. Limitations Small sample size and limited number of studies. Scarce information available on hard endpoints (ESKD). High heterogeneity among studies with respect to DKD severity, type and duration of antioxidant therapy. Conclusions In DKD patients, antioxidants may improve early renal damage. Future studies targeting hard endpoints and with longer follow-up and larger sample size are needed to confirm the usefulness of these agents for retarding DKD progression.

Introduction Diabetes Mellitus (DM) remains one of the most challenging global epidemics of the twentyfirst century. More than 350 million people worldwide are estimated to be affected by this metabolic disorder [1]. Diabetic kidney disease (DKD) currently ranks as the first cause of endstage kidney disease (ESKD), accounting for approximately 50% of cases in the developed world [2]. As many as 50% of individuals with longtime DM usually develop some degree of renal damage during their lifetime [3]. Progressive impairment in renal function is associated with an increased risk of cardiovascular events and hospitalizations, particularly in ESKD patients needing chronic renal replacement therapy by dialysis or kidney transplantation. Current strategies available for slowing-down DKD progression largely failed to achieve stable results in the long term. Alternative or additive approaches for maximizing reno-protection are thus eagerly advocated [4]. It is nowadays well recognized that oxidative stress plays a major role in the genesis and worsening of DKD [5]. A persistent state of hyperglycemia and the increase in advanced glycation end products (AGEs) elicit the generation of reactive oxygen species (ROS) which, in turn, enhance chronic inflammation and glomerular and tubular hypertrophy, eventually impairing overall renal function. Sparse evidence has now accrued indicating that antioxidant supplements may bring significant benefits to DKD patients, including the reduction of urinary albumin and total protein excretion and the normalization of glomerular filtration rate [6]. This raises the question as to whether such supplements should be systematically recommended for improving reno-protection in diabetic patients, particularly with early signs of renal damage.
We therefore aimed at performing a systematic review and meta-analysis of randomized clinical trials to investigate whether chronic antioxidant supplementations may represent a potential tool for slowing down disease progression in patients with diabetic kidney disease.

Data source and search strategy
Ovid-MEDLINE, PubMed and CENTRAL databases were searched for articles without time or language restriction up to November 15, 2016 using high sensitive search strategies (S1 Table). References from pertinent studies and eminent full-reviews were screened for additional articles. The search was designed and performed by three Authors (DB, GB and RB).

Study selection and data extraction
We included any randomized controlled trial (RCT) and quasi-RCT (trials in which allocation to treatment was made by alternation, use of alternate medical records, date of birth or other expected methods) testing the effects of any antioxidant supplementation (including but not limited to vitamin A, vitamin C, vitamin E, selenium, zinc, methionine or ubiquinone, either alone or in combination to other treatments) on renal endpoints in patients with diabetic kidney disease (DKD).
Studies were considered regardless of dosage or duration of administration of antioxidants and type of comparator. For cross-over studies the first period was considered.
DKD was defined as evidence of renal damage (chronic kidney disease, CKD) related to diabetic disease. CKD, in turn, was defined according to the National Kidney Foundation-Kidney Disease Outcomes Quality Initiative (NKF KDOQI) guidelines by the presence of a reduced glomerular filtration rate (GFR) <90 mL/min/1.73 m 2 or the persistence of hyperfiltration and/or urinary abnormalities such as pathological albuminuria, proteinuria or hematuria in subjects with GFR!90 mL/min/1.73 m 2 [9].
The primary endpoint of interest was progression to end-stage kidney disease (ESKD), defined as need for chronic renal replacement therapy, kidney transplantation or doubling of serum creatinine from baseline values. Secondary outcomes were change in albuminuria, proteinuria, serum creatinine and renal function (creatinine clearance/eGFR). Information on any adverse event was also collected, when reported.
Studies were excluded if: 1) dealing with diabetic patients with no evidence of kidney disease or dealing with CKD patients without diabetes; 2) focusing on DKD patients on chronic or acute renal replacement therapy (e.g. hemodialysis or peritoneal dialysis); 3) testing the effects of synthetic antioxidants (e.g. bardoxolone methyl) or antioxidant mixtures which exact composition was not defined; 4) not providing short or long-term data on the outcomes of interest.
Studies where at least part of the population satisfied the above criteria were included in the review.
Titles and abstracts were screened independently by two authors (VC, GG) who discarded studies that were not pertinent to the topic. Case reports, reviews, editorials and studies performed on children (age<18) were excluded from qualitative analyses but screened for potential additional references. Two Authors (VC, GG) independently assessed the retrieved abstracts and the full text of these studies to determine eligibility according to the inclusion/ exclusion criteria.
A third reviewer (GD) solved possible discrepancies on study judgments. Data extraction and analysis were carried out by two reviewers (VC, GG) and independently verified by another (GD).
reported narratively. The effects of treatment on continuous variables were assessed as mean difference (MD) or standardized mean difference (SMD), as appropriate.
Data were pooled using the random-effects model. To ensure robustness of the model and susceptibility to outliers pooled data were also analyzed with the fixed-effects model. Data expressed as median and range were converted to mean and SD by applying the Hozo formula [10]. Heterogeneity was assessed by the Chi 2 test on N-1 degrees of freedom, with an alpha of 0.05 considered for statistical significance and the Cochrane-I 2 [11]. I 2 values of 25%, 50% and 75% were considered to correspond to low, medium and high levels of heterogeneity, respectively. Possible sources of heterogeneity were explored by sensitivity analysis.

Risk of bias assessment
Likelihood of bias in the single RCTs was evaluated according to the checklist developed by the Cochrane Renal Group that considers the presence of potential selection bias (random sequence generation and allocation concealment), performance bias (blinding of investigators and participants), detection bias (blinding of outcome assessors), attrition bias (incomplete outcome data), reporting bias (selective reporting) and possible other sources of bias.

Summary of findings and quality of the evidence
A "Summary of findings" table summarizing pooled evidence for the main outcomes was constructed according to the GRADE method [12]. The five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) were taken into account to assess the quality of a body of evidence for the main pre-specified outcomes. All decisions to downgrade or upgrade the quality of studies were justified using footnotes and comments were made, when appropriate, to help readers' understanding of the review.

Search results
Fig 1 shows the flow diagram of the study selection process. Thirteen thousand five hundred and nineteen potentially relevant citations were initially found. By screening titles and abstracts, a total of 13491 references were excluded for various reasons (search overlap, study population or intervention not pertinent, review articles or case reports). Amongst the 28 studies selected for full text examination, 13 studies were excluded because not RCTs (n = 6) or because the study population (n = 6) or the intervention (n = 1) did not fulfil the review criteria. A total of 15 articles referring to14 full studies (4345 participants) were reviewed in detail. Eight studies [13][14][15][16][17][18][19][20] (327 participants) providing suitable numerical data on the outcomes of interest contributed to pooled meta-analyses. Main characteristics of the included studies are described in Table 1.
Ten RCTs were double blind [13-17, 19, 20, 22, 24, 25] and two studies were open label [23,26]. In two studies, blinding was unclear [18,21]. Attrition bias was low in all but one study [21] in which the overall drop-out rate was as high as 29%. Reporting bias was low in two studies [20,24] and unclear in the remainder. Risk of funding bias was potentially high in one study [24]. No further sources of bias were identified.

Effects of antioxidant supplements on primary outcomes
In Lonn [24] et al., the number of patients experiencing needing to start chronic dialysis due to DKD progression was not different between the intervention and the control group (p = 0.96).

Effects of antioxidant supplements on secondary outcomes
Urinary albumin excretion (UAE). In a pooled analysis of eight studies (327 patients) [13][14][15][16][17][18][19][20], antioxidant treatment produced a significant decrease in UAE levels as compared to control (SMD: -0.47; 95% CI -0.78, -0.16, Fig 2a). The analysis had mild level of heterogeneity (Chi 2 = 19.23, df = 9 (P = 0.02); I 2 = 53%) which was totally nullified in a sensitive analysis not including the only study with an open label design [18]. No publication bias was observed by visual inspection of the funnel plot and Egger's regression test (Fig 3). The quality of the body of evidence for this outcome (GRADE) resulted high after being upgraded for strong magnitude of effects and downgraded for study limitations (very short follow-up of all the included studies) ( Table 3).
These findings were in line with results obtained by four single studies [21,22,25,26], not suitable to be included in the meta-analysis. Conversely, two other studies [23,24] did not report significant changes in UAE after antioxidant therapy as compared with placebo.
In subgroup analyses performed according to the type of antioxidant administered, benefits on UAE were confirmed in trials testing Vitamin E (4 studies/5 separate arms-194 participants;      2c). Lonn et al [24] reported also information on the incidence of new cases of microalbuminuria which was statistically not different between the intervention-(35.3%) and the control group (37.5%) (p = 0.14).
Proteinuria. Farvid et al. [15] did not report significant changes in proteinuria after combined treatment with zinc + magnesium or Vitamin C + Vitamin E.
Serum creatinine. In one trial [14], no significant difference was observed in end of treatment serum creatinine values between the intervention and the control group (p = 0.55).
Renal function. In data pooled from 3 RCTs [13,16,17] (85 patients), antioxidant supplements had no tangible effects on GFR as compared with control (MD -0.12 ml/min/1.73m 2 ; 95% CI -0.06, 0.01; Fig 4). Heterogeneity in this analysis was absent (Chi 2 = 0.68, p = 0.71; I 2 = 0%). The quality of the body of evidence (GRADE) for this outcome resulted very low after being downgraded for study limitations (very short follow-up and small study populations) and evidence of imprecision (wide confidence intervals) ( Table 3). In one single study not suitable to be included in the meta-analysis [16], there was no significant difference in the end of treatment clearance of creatinine between individuals randomized to Vitamin E or placebo (p = 0.34).
Adverse events. Six RCTs did not report significant adverse events in both the active and control groups [13,14,16,19,20,24]. In Parham et al. [17] two individuals apiece for each study group had episodes of epigastric pain. In Farvid et al. [15] two individuals in the active group withdrew because of side effects (not further specified) in the 1 st week of study. The remaining studies [18, 21-23, 25, 26], did not look systematically at side events.

Discussion
This systematic review and meta-analysis aimed at assessing potential benefits of antioxidant agents on kidney disease progression in diabetic individuals with altered renal function. Unfortunately, despite diverse randomized controlled trials matching the review criteria were retrieved, the question as to whether antioxidant supplements should be systematically recommended for retarding DKD remains unclarified for various reasons.
First, scarce or no information was available on hard renal outcomes-that is kidney disease progression to end-stage as defined as the need for chronic dialysis or kidney transplantation. Conversely, the majority of studies were powered, both in terms of sample size and follow-up length, only to explore surrogate endpoints such as the absolute change in urinary albumin excretion or renal function. Second, only a minority of the included trials provided outcome data in a format suitable to be incorporated in meta-analyses while most information was confined to single studies or provided in a descriptive way only. This limitation hampered the robustness and generalizability of findings and prevented performing more complex, stratified statistical analyses according to various study characteristics, as originally pre-planned.
Third, despite heterogeneity in pooled analyses was unexpectedly low to moderate, there was high variability among trials with respect to population characteristics, study duration, severity of renal impairment and, above all, type and dose of antioxidant administered. No less important, most RCTs were of low methodological quality and risk of bias was unclear for the majority of the items analyzed, hence limiting the overall quality of evidence available.
In a meta-analysis of eight studies (371 individuals), antioxidants administration led to a significant reduction in albuminuria as compared to control. Such a finding was consistent with those obtained by four single RCTs [21,22,25,26] not suitable to be included in the meta-analysis and echoed previous observations reported by a wealth of evidence based on uncontrolled or non-randomized studies [28][29][30][31][32]. Pathological urinary albumin excretion (UAE) is one of the most robust and earliest signs of diabetes-induced kidney injury. This abnormality is originally consequent to a damaged glomerular filtration barrier resulting in an increased permeability to plasma proteins. Interestingly, all pathological changes underlying UAE worsening, including abnormal renal hemodynamics, glomerular basement membrane thickening, mesangial expansion, extracellular matrix accumulation and glomerulosclerosis, are largely driven by diabetes-induced ROS, setting the rationale for treatments targeting enhanced oxidative stress [33].
Of note, in this meta-analysis the observed reduction in albuminuria was as high as -0.33 [SMD, 95% CI -0.61, -0.04], there was a quite strong magnitude of effects and there was no evidence of publication bias. Taken all together, these observations might be of high importance in terms of potential clinical impact of antioxidant treatments in DKD patients. Although enthusiastic, however, these findings must necessary be toned down in the wake of the (still) exploratory nature, the low number and the above-mentioned limits of the included trials, being rather more useful for setting the stage for confirmatory studies. New evidence appears also necessary for clarifying whether the observed benefits may be attributable to the whole class of antioxidants rather than to single compounds. For instance, with respect to UAE, this latter possibility would be suggested by sensitive analyses showing a persistence of benefits in trials testing vitamin E and an apparent lack of effects in those testing vitamin C.
As aforementioned, another important aim of this review was also to ascertain possible direct benefits of antioxidant agents on renal function. In a previous Cochrane review focusing on chronic kidney disease of various etiology, antioxidant therapy significantly reduced incidence of ESKD, serum creatinine levels and improved creatinine clearance, although findings relied on few trials of suboptimal quality and a very small number of events [34].
In the diabetic setting, preliminary evidence exists indicating that antioxidant supplementation may, in some cases, normalize GFR or creatinine clearance, particularly in individuals with hyperfiltration [35][36][37].
In our review, such benefits were apparently not confirmed in a pooled analysis of three studies (85 patients) and in two other single studies [14,16] not reporting significant variations in GFR or serum creatinine after antioxidant therapy.
As persistent changes in renal function are notoriously achieved by long-term treatments, this lack of effect may be influenced by the very short therapy duration and the low number of patients with significantly impaired renal function at baseline, leaving the issue of direct benefits of antioxidant therapy on renal function still open.
Our review has strengths and limitations that deserve mentioning. Strengths are a pre-published protocol and a literature search, study selection, data extraction, cumulative analyses, bias and quality assessment that have all been performed following current best methodological standards for systematic reviews.
As alluded to before, the key limitation of this review mostly relies on the robustness of information available from the majority of the included studies. All the reviewed studies but one were single-center and long-term data on clinically relevant outcomes, such as need for dialysis or kidney transplantation and change in renal function, were basically lacking.
Alternative strategies including the adoption of measured GFR or other valid, non-surrogate estimates of renal function/damage might be a way for future trials for overcoming the need for large sample sizes and long follow-up times [38].
Furthermore, the short follow up period may prevent establishing durability of the observed effects (e.g. on UAE) and the quite high heterogeneity among study cohorts may limit the generalizability and applicability of findings to the whole DKD population.
In conclusion, despite cumulative findings point at some benefits of antioxidant therapy (particularly vitamin E) on early signs of renal damage, there is still no robust evidence supporting a widespread use of any of these agents as an alternative (or additive) therapy for retarding diabetic kidney disease progression. Future studies targeting hard rather than surrogate endpoints and with longer follow-up and larger sample size are advocated.