Ginkgo biloba Extract for Patients with Early Diabetic Nephropathy: A Systematic Review

Objectives. To evaluate the effectiveness and safety of a Ginkgo biloba extract for patients with early diabetic nephropathy. Methods. Randomised controlled trials (RCTs) conducted on adults with early diabetic nephropathy which used Gingko biloba extract were included. The major databases were searched, and manufacturers of Gingko biloba products were contacted for information on any published or unpublished studies. Two authors independently extracted the data from the included studies. Data analysis was conducted using Review Manager 5.0 software. Results. Sixteen RCTs were included. Ginkgo biloba extract decreased the urinary albumin excretion rate (UAER), fasting blood glucose (FBG), serum creatinine (SCR), and blood urea nitrogen (BUN). The extract also improved hemorheology. The methodological quality in the included studies was low. The explicit generation of the allocation sequence was described in only 6 trials. None of the included trials were confirmed to use blinding. Three studies had observed adverse events. One study using angiotensin-converting enzyme inhibitor (ACEi) reported mild cough in both groups. No serious adverse effects were reported. Conclusions. Gingko biloba extract is a valuable drug which has prospect in treating early diabetic nephropathy, especially with high UAER baseline level. The safety for early diabetic nephropathy is uncertain. Long-term, double-blinded RCTs with large sample sizes are still needed to provide stronger evidence.


Introduction
Diabetic nephropathy is one of the most serious complications in patients with diabetes. It is now the major leading cause of end-stage renal failure throughout the world [1].
e natural history of diabetic nephropathy includes �ve stages, starting with initial hyperfunction and hypertrophy at diagnosis, followed by increased glomerular �ltration rate (GFR) with normal albumin excretion, incipient diabetic nephropathy (characterised by microalbuminuria), then overt clinical nephropathy leading to progressive renal failure, followed by end-stage renal disease (ESRD) with uremia [2]. roughout the process, the relentless decline in renal function that starts at stage 4 (overt clinical nephropathy) with a mean GFR fall rate of approximately 1 mL/min 2 is a prominent feature in patients with diabetic nephropathy. When diabetic nephropathy develops into ESRD, renal replacement therapy (RRT) is required for survival. is has a major societal impact because of the enormous �nancial burden on patients and governments. Moreover, the survival of patients with diabetes undergoing dialysis is much worse than that of nondiabetic patients [3,4]. e UK Prospective Diabetes Study (UKPDS 64) [5] showed that the annual death rate of patients with elevated plasma creatinine or who were undergoing RRT was 19 microalbuminuria was 3.0%, and their condition did not progress to RRT even aer a year. erefore, early therapeutic intervention in patients with diabetic nephropathy is essential. e use of an ACEi or an angiotensin II receptor blocker (ARB) is now a component of standard therapy for patients with diabetic nephropathy, along with the control of glucose, lipids, and blood pressure. However, many patients continue to show progressive kidney damage. One study [6] that included a US population showed that among patients with diabetic nephropathy, the effect of ACEi/ARB in slowing progression to ESRD was not optimal. ere is still the need to develop new therapies to improve diabetic nephropathy treatment.
At present, many botanical medicines are applied as complementary therapy for diabetic nephropathy. Gingko biloba extract is one of the few Chinese herbal preparations that is recognised by the international medical community [7]. It is derived from the leaves of Gingko biloba, the only surviving species of the Ginkgoaceae family of tree. Now it is one of the most popular over-the-counter herbal dietary supplements in the world. A standardised Gingko biloba extract contains 22%-27% �avone glycosides and 5%-7% terpenes [8]. Studies have shown that patients with type 2 diabetes might bene�t from ingesting Gingko biloba extract to improve platelet function, alter platelet-vessel wall interactions [9], and reduce malondialdehyde levels in platelets [10].
Reduced GFR and increased albuminuria are independent risk factors for diabetic nephropathy [11]. In China, Gingko biloba extract has been used widely as a supplement to improve albuminuria and kidney function during the early stage (characterised by microalbuminuria) of diabetic nephropathy. Although the mechanisms underlying the pathogenesis of diabetic nephropathy are not completely understood, oxidative stress and cytokines are important factors of disease progression. Many in vitro and in vivo experiments have indicated that Gingko biloba extract can reduce relative total superoxide dismutase activity [12,13] aer adjusting for the expression of cytokines [14] in patients with diabetic nephropathy.
Many clinical trials have been conducted to assess the effectiveness and safety of Gingko biloba extract on early diabetic nephropathy, particularly in China. However, there is still no compelling evidence of the effectiveness and safety of Gingko biloba extract for early diabetic nephropathy, and a paper remains to be conducted. e aim of this paper was to summarise the evidence on the effectiveness and safety of Gingko biloba extract for early diabetic nephropathy and to provide the best available evidence for clinical practice and further research on diabetic nephropathy treatment.

Criteria for Considering Studies for this Review.
We included all RCTs conducted on adults (≥18 years) with early diabetic nephropathy, according to Mogensen stage III [2] (incipient urinary albumin excretion rate (UAER) of 20-200 g/min), and that used Gingko biloba extract, irrespective of blinding, publication status, or language.
Ginkgo biloba extract was added to the intervention group. All groups in the randomised trial received the same conventional treatment including glucose, lipid, and blood pressure control. e extract combined with ACEi/ARB was compared to ACEi/ARB alone. All of the included studies used standardised Gingko biloba extract containing 24% �avone glycosides and 6% terpenes, irrespective of dosage or form.
e primary outcome measurement was the incipient urinary albumin excretion rate (UAER) at the end of the study. Secondary outcomes were SCR, BUN, FBG, postprandial blood glucose (PBG), hemorrheology indices, and adverse reactions. We checked the reference lists of all acquired articles and called authors to ask for unpublished studies. In addition, we contacted manufacturers of Gingko biloba products to ask for information about any published and unpublished studies. We did not apply any language restrictions.

Data Collection.
Two authors (L. Zhang and W. Mao) independently assessed the title or abstract of each record to select potential eligible studies. Full articles were retrieved for further assessment if they were graded as included or unclear. en they independently assessed the full articles to decide Evidence-Based Complementary and Alternative Medicine 5 which ones were to be included. Differences were resolved by a third author ( Z. H. Wen). Two authors (L. Zhang and Z. Z. Liu) independently extracted the data using a self-developed data-extraction form, which included the following data.
(3) Participants: disease or condition, diagnostic criteria, inclusion and exclusion criteria, total number and number in comparison groups, and baseline characteristics.
(4) Interventions: treatment duration, the name and form of medication, the composition or ingredients, manufacture and quality control, and dose and administration.
(6) Followup: length of follow-up, any results of followup, reason and number of dropouts and withdrawals, and method of analysis.

Quality Assessment.
Two authors (L. Zhang and W. Mao) independently assessed the selected trials for methodological quality using the Cochrane Collaboration tool for assessing risk of bias [31]. Trials were assessed with respect to sequence generation, allocation concealment, blinding, incomplete outcome data, selective outcome reporting, and other sources of bias. We called the authors of all of the selected trials to con�rm the information of the above six domains, whether they were described in the articles or not. We resolved discrepancies by discussion. Sometimes we consulted authors Z. H. Wen and �. F. �uo to make the �nal decision.

Measures of Treatment Effect.
To measure the effects of treatments, we considered the primary and secondary outcomes detailed above. For continuous data, weighted mean differences between groups and its 95% con�dence intervals were calculated when the same measurement scale was used [32].

Missing Data.
Not all of the trials provided difference of means and its standard deviation (SD) before and aer treatment in both groups. e SD was calculated using the following formula [33]: We assumed the value of the correlation coefficient (Corr) to be 0.5 between before and aer treatment.

Assessment of Heterogeneity.
We used a chi-square test to test heterogeneity and set the signi�cance level at 0.1, in view of its low power. We also used the 2 statistic to quantify heterogeneity. e 2 statistic describes the percentage of variability in effect estimates that is due to heterogeneity rather than sampling error. A value >50% may indicate substantial heterogeneity.

Assessment of Publication Bias.
We investigated publication biases of studies that included more than �ve trials [34] using the funnel plot.
2.9. Data Analysis. Data analysis was conducted using Review Manager 5.0 soware, and STATA 11.0 SE. A metaanalysis was conducted on the primary and secondary outcomes mentioned above to summarise and to compare the efficacy of treatment with Gingko biloba extract to that of control intervention. We pooled the data using a �xed-effects model, and a randomised-effect model was used if heterogeneity was signi�cant. e role of different baseline data was tested by metaregression analysis. In addition, we performed subgroup analyses according to different interventions, if a sufficient number of RCTs was found.

Included Studies.
A total of 16 RCTs, all published trials that were conducted in China, were included. For details of studies selection and included studies see Figure 1 and Table  2.We contacted all authors of trials who did not report a randomised method. However, six authors could describe the process clearly. Five trials were published between 2000 and 2005 [21,22,24,26,27], and the remaining 11 trials were published between 2006 and 2010. e 16 included trials involved 1099 participants. Five of these studies compared ACEi/ARB alone to Gingko biloba extract combined with ACEi/ARB in patients with early diabetic nephropathy [15-17, 19, 28]. e others compared conventional treatment to the effect of Gingko biloba extract added to conventional treatment. e number of participants included in each trial varied between 45 and 112. e majority of trials included about 60 participants, and two trials included more than 100 patients [16,24]. e mean age of the participants in the treatment groups varied between 36.7 and 66.2 years, whereas that in the control group was between 37.1 to 68.2 years. e remaining six trials had a mean age of about 50 years [17,19,23,24,28,30]. All trials included male and female participants.  [25,29], WHO 1999 for eight trials [15,16,18,19,21,24,26,28], WHO 1998 for one trial [22], and WHO 1997 for one trial [27]. Two trials [16-20, 23, 25, 28-30]  3.1.6. Excluded Studies. Initially, 105 trials were prepared to be included in our study. However, aer we read the articles and called the authors, 89 trials were excluded due to the following reasons. Twenty-eight trials did not conform to the literature inclusion criteria aer reading the entire article and examining the data. Outcome measure data were missing or unclear in eight articles. Two articles had been published in different journals with the same data. ree trials were retrospective studies, and one trial used improper statistical methods. Finally, 48 trials were excluded aer we called the authors and con�rmed that randomisation was not used. (Table  4). All 16 trials were randomised trials; however, explicit generation of the allocation sequence was described in 6 trials or (by their authors during telephone calls) [16,18,20,22,23,28] in which a random number generator was used. Four trials explicitly described adequate allocation concealment. In these trials, the allocation sequence was concealed by an opaque envelope [18,23] or administered by a third party who was not involved in the study [16,22]. e authors of two other trials [16,22] did not describe allocation concealment clearly. �one of the included trials were con�rmed to use blinding. is may have been due to the difficulty in preparing of placebo, which is similar to Ginkgo biloba extract, with the same appearance and feeling during injection.

Assessing Risk of Bias in the Included Studies
Aer we telephoned authors, a bias of incomplete outcome data was reported in 4 trials [ For Group 1, the UAER data were available in 10 trials with a total sample size of 661 participants. At the end of the study, UAER decreased with an overall effect size of 34.64 g/min (95% CI, from 28.90 to 40.37, ) (Figure 2(a)), in favour of the Gingko biloba extract group using a random-effects model. e test for heterogeneity in the total group indicated an 2 value of 90%. is signi�cant Favours Ginkgo Test for subgroup differences: 2 = 0.07, df = 1 ( = 0.79), 2 = 0% Heterogeneity: 2 = 12, df = 9 ( = 0.21); 2 = 25% Test for overall effect:     heterogeneity could be explained by the different UAER baseline levels of the patients enrolled in these trials. We then divided group 1 into two subgroups based on baseline UAER levels. Subgroup 1.1.1 (Figure 2(a)) had a UAER baseline of >150 g/min and a higher effect size of 74.52 g/min (from 63.89 to 85. 15, ) with no signi�cant heterogeneity in the analysis. Subgroup 1.1.2 (Figure 2(a)) had a UAER baseline 100 g/min and lower effect size of 23.88 g/min (from 22.58 to 25.18, ) with low heterogeneity in the analysis.
e role of different baseline UAER levels was tested by metaregression using STATA 11.0 SE. e results showed a signi�cant in�uence of the covariables (Table 1).
For Group 2, three trials including 376 participants reported urea levels, which decreased with an overall effect size of 27.95 g/min (from 22.06 to 33.84, ) (Figure 2(b)).

Blood Glucose.
Data for FBG were presented in 11 trials with a total of 762 participants (Figure 3(a)). FBG decreased with an overall effect size of 0.44 mmol/L (0.31 to 0.57, 0.00001) with a �xed-effects model in favour of the Gingko biloba extract group. PBG was presented in four trials with a total of 284 participants (Figure 3(b)). PBG decreased with an overall effect size of −0.16 mmol/L (from −0.72 to 0.40, 0.58), which was not statistically signi�cant compared to the control group.

Kidney Function
. SCR data were presented in 10 trials with a total of 728 participants (Figure 4(a)). SCR decreased with an overall effect size of 3.27 mol/L (from 1.17 to 5.37, 0.00 ) with a �xed-effects model in favour of the Gingko biloba group. But no statistically signi�cant effect in SCR was observed in subgroup 3.1.1 ( 0. ) (Figure 4(a)). Seven trials including 526 participants were analysed. BUN decreased by 0.68 mmol/L compared to the control group (from 0.36 to 1.01, 0.00001; Figure 4(b)). But in the subgroup analysis, BUN increased by 0.02 mmol/L, which was not statistically signi�cant compared to the control group (from 0.62 to 1.40, 0.0 ; Figure 4(b)).

Hemorrheology.
In three trials with 220 participants, high shear viscosity and low shear viscosity were presented. e former decreased with an overall effect size of 1.17 mPa⋅s (from 0.08 to 2.25, 0.00 ; Figure 5(a)), and the latter decreased with an overall effect size of 1.68 mPa⋅s (from 0.83 to 2.53, ; Figure 5(b)). Plasma viscosity was presented in three trials, with a total of 180 participants. Plasma viscosity decreased by 0.30 mPa⋅s compared to controls (from 0.10 to 0.49, ; Figure 5(c)).

Adverse Effects.
No study reported serious adverse effects such as bleeding [8]. Mild cough was reported in only one trial [19], in which the intervention was combined with ACEi. 2 trials [22,23] reported no adverse effects in patients during observation. e other 13 trials did not report whether adverse effects were observed.

Publication
Bias. e funnel plots of FBG and BUN ( Figure 6) appear asymmetric, suggesting evidence of publication bias. No publication bias was observed form the funnel plots of UAER and SCR.

Summary of Findings.
e principal �nding of this paper is that Gingko biloba extract may be bene�cial for early diabetic nephropathy by decreasing the UAER, lowering FPG, and improving kidney function and the hemorrheology outcome measurements. Microalbuminuria is associated with an increased likelihood of progression of generic chronic kidney disease (CKD) to more advanced stages or even to ESRD [35][36][37][38], and increased albumin exertion, even below the lower limit of conventional microalbuminuria, is associated with an increased likelihood for both all cause and CV-related mortality in patients with CKD [39]. Hence, the level of microalbuminuria has more of an impact on renal endpoints (now de�ned as the need to start RRT) than the level of GFR per se in CKD stages from 1 to 3 [40,41].
ACEi and ARB reduce urinary albumin excretion and slow the progression of CKD [42]. However, new therapies to improve the prognosis of diabetic nephropathy are still needed. Microalbuminuria is a sign of systemic endothelial dysfunction, which may impair tubular epithelial albumin reabsorption. Extracts of Gingko biloba have been used to protect the vasculature of patients with diabetes [43,44].
Our paper showed that Gingko biloba extract decreased UAER level during early diabetic nephropathy. In addition, the baseline UAER level may in�uence the effect of Gingko biloba according to metaregression analysis. UAER with a high baseline level >150 g/min was decreased by 74.52 g/min; however, UAER with a low level baseline of <100 g/min was decreased by 23.88 g/min. e effect of Gingko biloba may be more remarkable with a high level of pretreatment. However, more studies are needed to con�rm the relationship between baseline UAER levels and the effect of Gingko biloba.
Abnormal hemodynamic indexes, such as increase in whole blood viscosity and plasma viscosity, are risk factors for development of diabetic nephropathy [45][46][47]. Our paper showed that Gingko biloba extract decreased blood viscosity and plasma viscosity in patients with early diabetic nephropathy. One study [48] conducted in Taipei also con-�rmed the effect of Gingko biloba extract on hemorheological indices in patients with diabetic retinopathy which are also microvascular complication of diabetes. Molecular biological mechanism of hemorheological disturbance in diabetes is not very clear. It may be associated with lipid peroxidation stress or reduced antioxidant vitamin E content of cell membranes [48]. Although experiments have demonstrated that Gingko biloba has the effect of antioxidation [49,50], further studies about the effect and mechanism of Gingko biloba on hemorrheology are still needed. In all trials included in our paper, kidney function was reported as SCR and BUN. Well-controlled UAER and improved haemodynamics play important roles in kidney protection. However, the effects of extract on SCR and BUN found in our subgroup analysis were not statistically signi�cant, although the pooled effect si�e of the entire group had statistical signi�cance. It was possibly due to the short course (from 14 to 60 days) of treatment in the included trials.
Ginkgo biloba extract may in�uence BG control by decreasing the effect of oral glucose-lowering drugs [51]. However, a series of clinical randomised studies has shown that Gingko biloba extract decreases BG by increasing pancreatic -cell function [52,53] without producing insulin resistance in patients with type 2 diabetes with pancreatic exhaustion [54]. In our analysis, Gingko biloba had effect on lowering FBG, but no obvious in�uence on �BG.

Safety Assessment.
As some studies reported, Gingko biloba extract could increase the risk of bleeding [8], such as intracranial hemorrhage or hyphema [55,56]. ese reports were mainly case reports, with dose varying from 100 mg to 240 mg daily and durations from 4 weeks to 52 weeks. And Gingko biloba extract in these reports was oen used with anticoagulants or aer surgery. In this paper, no one reported bleeding with lower dose (between 35 mg to 70 mg daily) and shorter duration (between 2 weeks to 9 weeks). But most studies included in this paper did not report whether they made observation on adverse events. e safety of Gingko biloba extract for early diabetic nephropathy still needs further research to estimate if there are any side effect.
In addition, several clinical studies observed herb drug interactions between Ginkgo biloba extract and conventional medicinals used in early diabetic nephropathy, such as tolbutamide [57,58], talinolol [59,60], metformin [61], nifedipine [62], and atorvastatin [63]. It was reported that combined use of Ginkgo biloba extract with tolbutamide [57] or talinolol [59] might affect their efficacy at higher doses (between 320 mg daily and 360 mg daily, resp.). In this review, no study reported herb drug interaction caused by Ginkgo biloba extract. May be it is due to the lower doses (from 35 mg to 70 mg daily) of Ginkgo biloba extract used in all the included trials. However, there is still absence of evidence about the in�uence of Ginkgo biloba extract on the efficacy of conventional medicinals. Further investigation of herb drug interaction with conventional treatment should be undertaken.

Overall Completeness and Applicability of Evidence.
Our �ndings are generalisable to the ma�ority of patients with early diabetic nephropathy being treated with Gingko biloba extract added to conventional medicine or ACEi/ARB. It is possible that the effect of Gingko biloba extract on decreasing microalbuminuria will be stronger at the high baseline level. However, additional RCTs are needed to con�rm this.

Quality of the Evidence.
Although 16 trials were included in our paper, randomisation methods were elaborated in 6 trials. We con�rmed that none of the trials were blinded, which could have affected the results. Some doctors might pay more attention to patients in a treatment group than those in a control group. In addition, some of the studies were conducted several years ago, and the authors may have forgotten the details of some incomplete outcomes, which could have resulted in selective reporting.

Potential Biases in the Paper Process.
Variation among trials was apparent in terms of sample size, treatment course, dosage, and forms of Gingko biloba extract used. Extract was used in combination with ACEi/ARB in several studies. However, the effects of Gingko biloba extract alone, Gingko biloba extract combined with ACEi, and Gingko biloba extract combined with ARB could not be clari�ed on some outcome measurements due to the small number of studies.
Conventional therapy including control of glucose, lipid, and BP is necessary as a general intervention according to the guidelines. However, this conventional therapy was not be well reported in the included studies, and we were unable to ad�ust for potential in�uences in our analyses.
e statistical heterogeneity among trials in this study was apparently substantial. Heterogeneity in UAER, FBG, and BUN mainly originates from subgroups allocated according to different baseline values or interventions. Heterogeneity in hemorrheology may be explained by differences in baseline characteristics among participants.
Although a series of comprehensive search strategies were conducted, the results of our paper were based on published studies. e funnel plots of FBG and BUN appear asymmetric, suggesting evidence of publication bias in the literature. But poor methodological quality in smaller studies and heterogeneity due to different interventions should also be considered as possible reasons for asymmetry.

Conclusion
Gingko biloba extract is a valuable drug which has prospect in treating early diabetic nephropathy, as it may decrease UAER and FBG and improve hemorrheology. Patients with a higher baseline UAER level may gain a better effect from Gingko biloba extract. e safety of Gingko biloba extract for early diabetic nephropathy is still uncertain. Herb drug interactions between Ginkgo biloba extract and conventional medicinals need further investigation to indicate. Nevertheless, the methodological quality in included studies was not perfect, and there were potential biases due to different interventions and short courses of treatment that may have affected the results. Long-term, double-blinded RCTs with large sample sizes are still needed to provide stronger evidence. #5 #4 and #3 Quick limits "humans, " limit "MJ, " advanced limits "RCT, " "CT" #6 "ginkgo biloba"/exp OR "egb 761"/exp OR "tavonin" OR "tebonin"/exp OR "rokan"/exp OR "tanakan"/exp OR "li 1370"/exp OR "egb" OR "bilobalid" OR "kaveri"/exp OR ginkgo$ OR gingko$ OR ginko$ OR gingko OR gincosan #7 #4 and #6 Quick limits "humans, " limit "MJ, " advanced limits "RCT, " "CT" #8 #5 OR #7 Disease Search #9 # 3 OR #6 Subject: diabetic nephropathy Subheading: Disease management Drug therapy Prevention Rehabilitation erapy Quick limits "humans" Advanced limits "RCT, " "CT".

�on��ct of �nterest
ere are no con�ict of interests for any of the authors of this study.