Authors: Yating Zhao, Ruixia Zhu, Danni Wang, Xu Liu
Categories: Research Articles, Research Article
Source: Annals of Clinical and Translational Neurology
Doi: 10.1002/acn3.50892
Diabetic neuropathy (DN) is one of the most common complications of diabetes that occurs in more than 67% of individuals with diabetes. Genetic polymorphisms may play an important role in DN development. However, until now, the association between genetic polymorphisms and DN risk has remained unknown. We performed a systematic review, meta‐analysis, and trial sequential analysis (TSA) of the association between all genetic polymorphisms and DN risk.
Relevant published studies examining the relationship between all genetic polymorphisms and DN were obtained based on a designed search strategy up to 28 February 2019. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated to assess overall pooled effects of genetic models as well as in subgroup analyses. Sensitive analysis and publication bias were applied to evaluate the reliability of the study. Moreover, TSA was conducted to estimate the robustness of the results.
We conducted a systematic review of a total of 1256 articles, and then 106 publications reporting on 136 polymorphisms of 76 genes were extracted. We performed 107 meta‐analyses on 36 studies involving 12,221 subjects to derive pooled effect estimates for eight polymorphisms. We identified that ACE I>D, MTHFR 1298A/C, GPx‐1 rs1050450, and CAT ‐262C/T were associated with DN, while MTHFR C677T, GSTM1, GSTT1, and IL‐10 ‐1082G/A were not. Sensitivity analysis, funnel plot, and Egger’s test displayed robust results. Furthermore, the results of TSA indicated sufficient sample size in studies of ACE, GPx‐1, GSTM1, and IL‐10 polymorphisms.
Our study assessed the association between ACE I>D, MTHFR C677T, MTHFR 1298A/C, GPx‐1 rs1050450, CAT ‐262C/T, GSTM1, GSTT1, and IL‐10 ‐1082G/A polymorphisms and DN risk. We hope that the data in our research study are used to study DN genetics.
As a global public threat, diabetes mellitus (DM) is a life‐long disease that involves multiple organs and systems, and the morbidity of diabetes among adults could rise to 552 million by 2030.1, 2 As the most common complication of diabetes, diabetic neuropathy (DN) including diabetic autonomic neuropathy and somatic sensorimotor neuropathy has a prevalence of 8% in newly diagnosed diabetic patients and over 50% in patients with a long course of disease.3, 4 DN may produce a series of clinical manifestations including numbness, tingling, pain, and/or weakness which considerably decrease the quality of life in patients.5 Currently, the risk factors and pathogenesis of DN have drawn increasing attention.
Many factors are known to be associated with DN susceptibility, including smoking, obesity, poor glycemic control, and duration of diabetes, but there are still some potential factors leading to the occurrence of DN, such as genetic variants.6, 7 In 1997, Vague P et al. first found an association between the ATP1 A1 gene polymorphism and DN risk.8 Since then, an increasing number of studies have been carried out to investigate the association between various genetic polymorphisms and DN susceptibility, such as ACE I/D, MTHFR C677T and GSTM1.9, 10 For example, in 2012, Jurado et al.11 reported that the ID genotype of the ACE I/D polymorphism had a protective effect on the development of DN. However, others drew a completely different conclusion in that the ID genotype may lead to an increased DN risk.2, 12 Similarly, a significant association between the MTHFR gene C677T mutation and DN was observed by Yigit in 2013,13 which could not be replicated by Russo in 2016.14
Till now, the findings of individual studies were not always consistent, and no systematic review covered all genetic polymorphisms has been reported. To fill this gap in medical literature worldwide, we performed the first systematic review and meta‐analysis involving all the available evidence in the field of genetic variants and DN susceptibility.
A comprehensive literature search was performed in the PubMed and Embase databases up to 28 February 2019, using the following “diabetic neuropathy/diabetic polyneuropathy/diabetic peripheral neuropathy/DPN/cardiovascular autonomic neuropathy/CAN” and “polymorphism/variant/genotype/allele/SNP/mutation”. As a complement, we also checked the reference list of the meta‐analyses and review articles on genetic association for DN, in case the references they used had been missed in original search.
Studies were included if they met the following (1) case–control studies; (2) for the association between any genetic polymorphism and DN susceptibility; (3) sufficient allele and genotype data to calculate the odds ratios (ORs) with 95% confidence intervals (CIs); (4) studies published in English. If two papers included the same dataset, but one included additional data not found in the other paper, only the later was included. Any genetic polymorphism with three or more published studies was included in our meta‐analysis.
By using a standardized form, two investigators independently extracted the following the name of the first author, publication year, region, ethnicity, sample size, allele and genotype frequencies, genotyping methods, age‐ and gender‐matched status, type of diabetes, type of neuropathy, Newcastle‐Ottawa Quality Assessment Scale (NOS) score, and *P *value for Hardy–Weinberg equilibrium (HWE) in the control group. The quality of studies was evaluated using the NOS and scores >5 were considered to be of high quality, otherwise, they were thought to be with low quality.
We used Stata 12.0 software to conduct the meta‐analysis for each genetic polymorphism to determine the pooled ORs and 95% CIs. We calculated the pooled results under all five genetic models (allelic, recessive, dominant, homozygous, and heterozygous model). Heterogeneity was measured by the I ^2^ statistic, and I ^2^ > 50% was considered significant heterogeneity. The random‐effects model was used if significant heterogeneity existed or else the fixed‐effects model was adopted. Subgroup analyses were performed based on ethnicity, genotyping methods, age‐ and gender‐matched status, HWE status of controls, quality of studies, source of control, type of diabetes, and type of neuropathy. The sensitivity analyses were conducted by sequentially omitting each study to detect the stability of pooled results and source of heterogeneity. Publication bias was explored using visual inspection of the funnel plot and Egger’s test. P < 0.05 was considered to be statistically significant.
Meta‐analysis may lead to a false‐positive or negative conclusion.15 Hence, we used trial sequential analysis (TSA) to reduce these statistical errors.16 TSA is a novel statistical analysis method that uses a combination of techniques that provides required information size (RIS), a threshold of statistically significant effect, for evaluating whether sufficient evidence is included and whether a result is reliable or not, in meta‐analysis. Additionally, a threshold of futility could be tested by TSA to find a conclusion of no effect before reaching the information size by using TSA software (version 0.9.5.10 beta) (Copenhagen Trial Unit, Centre for Clinical Intervention Research, Rigshospitalet, Copenhagen, Denmark). We computed the RIS based on an alpha risk of 5%, a beta risk of 20%, a relative risk reduction of 20% and a two‐sided boundary type. For those analyses that the Z‐curve reached the RIS line or monitoring the boundary line or futility area, it indicates that enough samples are included in the studies, and their results are credible. Otherwise, the amount of information is not large enough, and more evidence is needed.
In total, 1256 articles were retrieved according to our search strategy. First, we excluded 1032 articles by duplicate screening as well as title and abstract reviewing. Second, after full‐text reviewing, 118 studies containing 60 letters, reference abstracts and reviews; 38 studies not relevant to DN; 12 studies not focused on DN susceptibility; and eight studies not written in English were excluded. Third, 106 eligible articles were selected in our systematic review, and the relationship between all 136 genetic polymorphisms and DN susceptibility was extracted and listed in Table S1. Finally, for any polymorphism with three or more published studies and sufficient genotype data to extract, we keep it into our meta‐analysis. A total of 36 studies were involved in the meta‐analysis, and the entire process of study selection is shown in Figure 1.
Figure 1 Flow diagram of the study selection process.
Thirty‐six studies with 4515 cases and 7706 controls were included in the meta‐analysis according to the inclusion and exclusion criteria.13, 14, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39 The general characteristics of the studies are summarized in Table 1. Among the 36 studies, 6 were related to ACE I/D, 8 to MTHFR C677T, 3 to MTHFR 1298A/C, 5 to GPx‐1 rs1050450, 3 to CAT ‐262C/T, 4 to GSTM1 and GSTT1 and 3 to IL‐10 ‐1082G/A. In these studies, 26 studies were performed in the Caucasian population and 10 remaining studies were performed in the Asian population. The genotyping methods included polymerase chain reaction‐restriction fragment length polymorphism, TaqMan, polymerase chain reaction‐sequence specific primers, and amplification refractory mutation system‐polymerase chain reaction. For the quality of studies, all of them except four14, 18, 21, 37 scored more than 5 in NOS. In addition, for the HWE of controls, most of the articles met HWE equilibrium, while 10 studies failed.24, 25, 26, 30, 31, 32, 37, 39
The ACE I>D polymorphism was investigated in six studies along with DN (884 cases, 1570 controls).17, 18, 19, 20, 21, 22 A significant association was uncovered between the ACE I>D genetic polymorphism and DN risk under allelic and homozygous models (D vs. I: OR = 1.23, 95% CI = 1.08–1.39; DD vs. II: OR = 1.50, 95% CI = 1.15–1.95) (Fig. 2). Furthermore, stratified analyses based on ethnicity, quality of studies, matched status, type of diabetes and type of neuropathy were conducted for allele, recessive, and dominant models, with results presented in Table 2. Finally, increased susceptibility was found in the recessive model in the high‐quality study group as well as in the age‐ and gender‐matched group. We subsequently performed sensitivity analyses to explore the influence of an individual study on the pooled results, and our results did not change when omitting each study in the allelic and homozygous models (Figure S1).
Figure 2 Forests for ACE I>D polymorphism and DN risk. (A) allele model (D vs. I); (B) homozygous model (DD vs. II). DN, diabetic neuropathy.
Totally, there were 8 studies13, 14, 20, 23, 24, 25, 26, 27 (813 cases, 1544 controls) associated with MTHFR C677T and DN involved in the meta‐analysis. Five of eight studies were performed in the Caucasian population, and the other three studies were performed in the Asian population. The pooled results of the five genetic models did not show any significant difference (Fig. 3). Further subgroup analyses were conducted, and no significant result was observed (Table 3).
Figure 3 Forests for MTHFR C677T polymorphism and DN risk. (A) allele model (T vs. C); (B) recessive model (TT vs. TC + CC); (C) dominant model (TC + TT vs. CC); (D) homozygous model (TT vs. CC); (E) heterozygous model (TC vs. CC). DN, diabetic neuropathy.
We included three studies (223 cases, 650 controls) published on the relationship between the MTHFR 1298A/C polymorphism and DN in this meta‐analysis.23, 24, 25 Two of them were performed in the Caucasian population and the other one in the Asian population. Using the AA genotype as the reference, two genetic models revealed a significant association between the MTHFR 1298A/C polymorphism and DN (CC + AC vs. AA: OR = 1.44, 95% CI = 1.03–2.01; AC vs. AA: OR = 1.51, 95% CI = 1.07–2.11; Fig. 4). In addition, the stratified analyses according to ethnicity suggested that MTHFR 1298C/T was correlated with DN in the Caucasian population (CC + AC vs. AA: OR = 1.57, 95% CI = 1.02–2.41).
Figure 4 Forests for MTHFR 1298A/C polymorphism and DN risk. (A) dominant model (CC + AC vs. AA); (B) heterozygous model (AC vs. AA). DN, diabetic neuropathy.
Five studies31, 32, 33 (784 cases, 1957 controls) were combined to analyze the association between the GPx‐1 rs1050450 polymorphism and DN. Three of five studies were performed in the Caucasian population, and the other two studies were conducted in the Asian population. The pooled OR values of four models revealed a significant association between GPx‐1 rs1050450 and DN risk (T vs. C: OR = 1.43, 95% CI = 1.26–1.64; TT + CT vs. CC: OR = 1.74, 95% CI = 1.46–2.08; TT vs. CC: OR = 1.58, 95% CI = 1.17–2.12; CT vs. CC: OR = 1.78, 95% CI = 1.48–2.14; Fig. 5). Stratification accounting for the type of diabetes revealed increased DN risk in the T2DM group (Table 4). Additionally, a similar relationship was detected under allelic and dominant models in the group with Caucasian ethnicity, sensorimotor neuropathy and controls in HWE (Table 4). In addition, each single study was omitted sequentially, without obvious alteration of overall statistical significance in sensitivity analysis (Figure S1).
Figure 5 Forests for GPx‐1 rs1050450 polymorphism and DN risk. (A) allele model (T vs. C); (B) dominant model (TT + CT vs. CC); (C) homozygous model (TT vs. CC); (D) heterozygous model (CT vs. CC). DN, diabetic neuropathy.
The analysis of the CAT ‐262C/T polymorphism associated with DN included 3 studies (350 cases, 465 controls), which were all performed in the Caucasian population.28, 29, 30 Using the CC genotype as a reference, we found a protective effect of the CAT ‐262C/T polymorphism against the susceptibility of DN (T vs. C: OR = 0.71, 95% CI = 0.57–0.87; TT vs. CT + CC: OR = 0.53, 95% CI = 0.36–0.77; TT vs. CC: OR = 0.54, 95% CI = 0.35–0.82; Fig. 6). When stratified by the type of diabetes, a decreased risk was identified in the T1DM group (T vs. C: OR = 0.68, 95% CI = 0.53–0.86; TT vs. CT + CC: OR = 0.51, 95% CI = 0.35–0.76), but not T2DM group.
Figure 6 Forests for CAT‐262C/T polymorphism and DN risk. (A) allele model (T vs. C); (B) recessive model (TT vs. CT + CC); (C) homozygous model (TT vs. CC). DN, diabetic neuropathy.
The meta‐analysis including four studies30, 34, 35, 36 (516 cases, 573 controls) about GSTM1 null/present polymorphism and DN reflected no significant difference (OR = 1.21, 95% CI = 0.94–1.56, Fig. 7). Concerning GSTT1 null/present polymorphism, four studies30, 34, 35, 36 (500 cases, 589 controls) were enrolled in the meta‐analysis. The pooled results also failed to show any significant difference (OR = 0.96, 95% CI = 0.30–3.04, Fig. 8). The sensitivity analysis showed no significance after excluding any of the studies (Figure S1).
Figure 7 Forest for GSTM1 null/present polymorphism and DN risk. DN, diabetic neuropathy.
Figure 8 Forest for GSTT1 null/present polymorphism and DN risk. DN, diabetic neuropathy.
In the meta‐analysis of the IL‐10 ‐1082G/A polymorphism and DN, three studies37, 38, 39 were involved (444 cases, 358 controls). The pooled results showed no significance between IL‐10 ‐1082G/A and DN (Fig. 9).
Figure 9 Forests for IL‐10 ‐1082G/A polymorphism and DN risk. (A) allele model (G vs. A); (B) recessive model (AA vs. AG + GG); (C) dominant model (AA + AG vs. GG); (D) homozygous model (AA vs. GG); (E) heterozygous model (AG vs. GG). DN, diabetic neuropathy.
In addition to the genetic polymorphisms discussed above, we also found that some other polymorphisms had statistical significance on DN risk in 33 individual studies, such as CACNA 1A rs2248069, CYBA rs4673, FTO rs17817449, IL2RA rs706778, SCN10A rs7375036, CTLA‐4 rs5742909, GNB3 C825T, and NOS3 Glu298Asp.8, 20, 28, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68 Due to the small number of relevant studies or insufficient data for genotype frequency, these studies could not be enrolled in our meta‐analysis. Therefore, we performed a systematic review of these polymorphisms and listed them in Table 5, with the purpose of providing clues in future searches for genetic risk factors of DN.
Funnel plot and Egger’s test were employed to appraise the publication bias among all eight studies. By visual detection of funnel plots, six genetic variants including ACE I>D, MTHFR C677T, GPx‐1 rs1050450, CAT ‐262C/T, GSTM1 null/present and GSTT1 null/present, showed symmetric shapes, which demonstrated that no publication bias existed and was further confirmed by Egger’s test. In contrast with these variants, we detected mild publication bias in MTHFR 1298A/C and IL‐10 polymorphisms. As for MTHFR 1298A/C, marginal bias could be found in the allelic model (P = 0.025). In the recessive genetic model of IL‐10, a statistically significant difference could be found by Egger’s test (P = 0.023). The visual inspection of the funnel plot and P value of Egger’s test of all included studies are summarized in Figure S2 and Table 6, respectively.
Among the eight studies mentioned above, three studies performed on the ACE I>D polymorphism, GPx‐1 rs1050450 polymorphism, and IL‐10 ‐1082G/A polymorphism concluded that a sufficient number of samples were used in the analyses, and conclusive results could be obtained. Specifically, in the study of the ACE I>D polymorphism, the Z‐curve of the allelic and homozygous model crossed either the TSA monitoring boundary or RIS line, confirming that the ACE I>D polymorphism was associated with increased DN risk. For the GPx‐1 rs1050450 polymorphism, in the allelic, dominant and heterozygous models, we detected that the Z‐curve exceeded the RIS line, which revealed enough evidence for significant results. With regard to the IL‐10 ‐1082G/A polymorphism, as the Z‐curve entered the futility area in the allelic and dominant models, we came to a confirmed conclusion that the IL‐10 polymorphism had no relationship with DN susceptibility. However, the TSA results of the other five genetic variants did not show adequate information involved in the meta‐analysis. More relevant studies are necessary to prove our findings in the future. The TSA results for all the included studies are shown in Figure S3.
As we all know, the systematic review and meta‐analysis approach used in this study is the most comprehensive method to detect genetic risk factors in most human diseases.69 To date, there is no complete systematic review and meta‐analysis reporting the potential association between all genetic polymorphisms and DN risk. Using widely accepted genetic models and subgroup analyses based on ethnicity, HWE status, quality of studies and so on, we performed this comprehensive systematic review that provided empirical support for exploring the relationship between relevant genetic polymorphisms, such as ACE I/D, MTHFR C677T, MTHFR 1298 A/C, GPx‐1 rs1050450, CAT ‐262C/T, GSTM1, GSTT1, IL‐10 ‐1082G/A, and DN susceptibility.
ACE is a key component of the renin–angiotensin system that converts angiotensin (Ang) I to Ang II. Ang II impacts endothelial damage and microcirculatory dysfunction.70 Therefore, insufficient blood supply to peripheral nerves due to microcirculatory dysfunction is considered a possible pathological mechanism of DN.71 As the starting factor affecting Ang II level, ACE activity is influenced by the presence of an insertion (I) or deletion (D) of a 287‐base pair fragment in intron 16 of the ACE gene resulting in a common variant, with the D allele being associated with higher ACE activity.72 This allele has been previously observed to probably associate with microvascular complications of diabetes.73, 74, 75 In this study, we statistically confirmed that the ACE I/D polymorphism was significantly associated with DN risk. The D allele had a 1.23‐fold risk for DN compared with the I allele, and a 50% increased risk of DN was identified in DN patients with the DD genotype compared with the II genotype.
MTHFR is a key regulatory enzyme in homocysteine metabolism that converts homocysteine back to methionine via the re‐methylation pathway.76 Therefore, deficiency of MTHFR increases the odds for hyperhomocysteinemia.77 Meantime, it was reported that homocysteine levels and the prevalence of hyperhomocysteinemia were strongly associated with DN.78 Mutations of the MTHFR gene have been defined, and C677T and A1298C variants are the two of the most explored.77 Both are functional polymorphisms that lead to decreased enzymatic activity, resulting in elevated homocysteine levels.77 The association between MTHFR gene polymorphisms and the susceptibility of DN has been investigated in several studies but with inconsistent results. Therefore, we performed this meta‐analysis involving all the available evidence of these two genetic variants and DN risk. In our study, only the MTHFR 1298A/C polymorphism showed a significant association with DN in the pooled results, while no significant difference was found in the analysis of MTHFR C677T. In vitro studies showed that hyperhomocysteinemia affected nervous function either by direct cytotoxicity or by oxidative damage.79, 80 Oxidative stress is associated with the development of apoptosis in neurons and supporting glial cells and could be the unifying mechanism that leads to nervous system damage in diabetes.81, 82
GPx‐1 is a gene that encodes an antioxidant enzyme. Its main role is protecting cells against oxidative damage by reducing hydrogen peroxide and organic peroxidases to H2O2 with reduced glutathione.83 As one of the GPx‐1 polymorphisms, rs1050450, which reduces the activity of this enzyme, may cause an adverse effect on the vascular system and microvascular complications of diabetes.84, 85 The present study aimed to evaluate the association of the rs1050450 polymorphism in the GPx‐1 gene with DN. For our pooled results, we detected that GPx‐1 rs1050450 showed a significant difference in the risk for DN. In the subgroup analysis, we found a similar result in the Caucasian population, as well as in the T2DM and sensorimotor neuropathy groups. The exact mechanism of the observed effect of GPx‐1 gene polymorphism on susceptibility to DN is unknown. We speculate that changing the capacity of the antioxidant enzyme by the rs1050450 polymorphism may lead to increased oxidative damage which was found to be an important pathophysiological mechanism involved in DN.
CAT is a widespread enzyme that can catalyze the decomposition of H2O2 to water and molecular oxygen, which inactivate free oxygen radicals and peroxides in the process of oxidative stress existing in DN.86 Therefore, CAT plays an important role in the pathogenesis of DN. From the current meta‐analysis of CAT ‐262C/T and DN risk, our findings suggested that the T allele showed a protective effect on DN development, with nearly 29% and 47% decreased susceptibility in the allelic and recessive genetic models, respectively. Additionally, all three studies involved in this meta‐analysis are performed in the Caucasian population. Thus, there may be a low risk for DN in T allele carriers of Caucasians. However, no related study was conducted in an Asian population. The role of CAT ‐262C/T in DN requires further studies for non‐Caucasian populations.
Glutathione S‐transferases (GSTs) are a family of antioxidant enzymes that play important antioxidant roles in the elimination of reactive oxygen species.87 GSTM1 and GSTT1 genes are polymorphic in humans, and the null genotypes are accompanied by a lack of enzyme activity.88 The GSTM1 and GSTT1 polymorphisms have been reported as risk factors for DN in the past but without consistent results. According to our pooled data, none of these two genetic polymorphisms showed a significant difference in the risk for DN. However, due to the limited number of further studies and the inadequate number of included samples indicated in TSA, confirming the association between either of the two genetic polymorphisms and DN is difficult. Future studies with larger sample sizes are required.
Limitation also existed in our study. First, several genes have just been investigated in small cohorts and in only Caucasian populations such as GSTT1, GSTM1, and CAT ‐262C/T. Second, we confined the enrolled studies to publications in English. Third, obvious heterogeneity could be detected among some meta‐analyses, such as MTHFR C677T and GSTM1 null/present which influences the credibility of our results. Therefore, we performed subgroup and sensitivity analyses to explore the source of heterogeneity, which was often from different study designs, measurement errors and ethnic diversity. Unfortunately, heterogeneity was not eliminated by these methods, which indicated that all factors mentioned before should be considered together. Fourth, mild publication bias was detected in MTHFR 1298A/C and IL‐10 polymorphisms, and TSA showed inadequate information involved in the analyses for MTHFR, CAT and GST genes. Thus, the comprehensive analyses should be interpreted with caution. Finally, we did not analyze the gene‐gene and gene‐environment interactions in our current meta‐analysis due to insufficient information.
In conclusion, we demonstrated that ACE I/D, MTHFR 1298A/C, GPx‐1 rs1050450, and CAT‐262C/T were associated with DN susceptibility but MTHFR C677T, GSTM1, GSTT1, and IL‐10 −1082G/A were not. More studies performed in different ethnicities with larger sample sizes are required to confirm our findings in the near future.
The authors declare no financial or other conflicts of interests.
This work was supported by the National Natural Science Foundation of China (grant no. 81400950, 81501006).
This work was funded by National Natural Science Foundation of China grants 81400950 and 81501006.