NEPHROLOGY - ORIGINAL PAPER

Association between TNF-a 2308G/A polymorphismand diabetic nephropathy risk: a meta-analysis

Yuliang Zhao • Jiqiao Yang • Ling Zhang •

Zheng Li • Yingying Yang • Yi Tang •

Ping Fu

Received: 14 March 2013 / Accepted: 6 June 2013

� Springer Science+Business Media Dordrecht 2013

Abstract

Purpose TNF-a -308G/A polymorphism has been

implicated in the susceptibility of diabetic nephropa-

thy, but studies have reported inconclusive results.

The present study investigated the relationship

between -308G/A polymorphism in the TNF-a gene

and diabetic nephropathy risk by meta-analysis.

Methods Data from PubMed, Embase, Ovid, Coch-

rane Library, China National Knowledge Infrastruc-

ture, Wanfang, VIP and China Biology Medicine disc

databases were evaluated and analyzed. Statistical

analysis was performed using RevMan 4.2 and Stata

10.0 software.

Results A total of 1,277 diabetic nephropathy cases

and 1,740 controls in eight case-controlled studies

were identified for data analysis. The results suggested

that A allele carriers (GA ? AA) may not have an

altered risk of diabetic nephropathy when compared

with homozygote GG carriers with boarder-line sta-

tistical significance (OR = 0.84, 95 % CI = 0.71–

1.00, p = 0.05 for GA ? AA vs. GG). However, in

Asian subgroup analysis, the A allele variant was

associated with a decreased diabetic nephropathy risk

(OR = 0.69, 95 % CI = 0.51–0.94, p = 0.02 for

GA ? AA vs. GG).

Conclusions Meta-analysis suggests that the A allele

of TNF-a -308G/A polymorphism might be protec-

tive against diabetic nephropathy with ethnic selec-

tivity. Future studies are needed to validate these

findings.

Keywords TNF-a -308G/A polymorphism �Gene �Diabetic nephropathy � Meta-analysis

Introduction

Diabetic nephropathy is a progressive kidney disease

caused by long-term diabetes mellitus and commonly

causes chronic kidney failure and end-stage renal

disease [1, 2]. Diabetic nephropathy is the result of

interactions between hereditary and acquired factors.

The etiology of diabetic nephropathy is not yet fully

understood. While diabetes mellitus appears to be the

primary disease, it has been reported that factors such

as inflammatory pathways, cytokines and other bio-

logical processes may be important in the

Zhao Yuliang and Yang Jiqiao contributed equally to this

work.

Y. Zhao � L. Zhang � Y. Yang � Y. Tang � P. Fu (&)

Division of Nephrology, West China Hospital, Sichuan

University, No. 37, Guoxue Alley, Chengdu 610041,

Sichuan Province, China

e-mail: fupinghx@163.com

J. Yang

West China Medical School, Sichuan University,

Chengdu, China

Z. Li

West China School of Stomatology, Sichuan University,

Chengdu, China

123

Int Urol Nephrol

DOI 10.1007/s11255-013-0490-3

pathogenesis of diabetic nephropathy [3, 4]. A number

of genes have been suggested as diabetic nephropathy

candidate genes, including transforming growth fac-

tor-b1 [5], carnosine dipeptidase 1 [6], ERBB receptor

feedback inhibitor 1 [7] and tumor necrosis factor-

alpha (TNF-a) [4]. Of them, TNF-a is one of the most

widely studied genes.

TNF-a is a pro-inflammatory cytokine involved in

the regulation of immune cells and plays an important

role in the pathogenesis of many types of diseases,

including diabetes mellitus and diabetic nephropathy

[8, 9]. Patients with diabetic nephropathy can have

abnormal levels of TNF-a in the serum and can,

therefore, differ from the non-diabetic nephropathy

populations with regard to inflammation and other

pathophysiologic activities. The TNF-a gene is a

member of the TNF superfamily located on chromo-

some 6q21 within the class III region of the major

histocompatibility complex. Several polymorphisms

have been identified, such as -308G/A (rs1800629),

-857C/T (rs179972) and -1031T/C (rs1799964). Of

them, -308G/A polymorphism has been most exten-

sively studied. -308G/A polymorphism in the TNF-agene might alter the properties and levels of the

molecule and might, therefore, contribute to the

etiology of diabetic nephropathy. It has been reported

that -308G/A polymorphism is associated with the

onset of obesity [10], and insulin resistance in diabetes

mellitus might also be related [11]. Such findings

suggest the possible roles that -308G/A polymor-

phism might play in the pathogenesis of diabetic

nephropathy. Associations between -308G/A poly-

morphism and the risk of diabetic nephropathy have

been widely studied; however, results were inconclu-

sive. Therefore, the present meta-analysis was per-

formed for the purpose of overcoming the limitations

of individual studies. This is, to our knowledge, the

first meta-analysis examining associations between

TNF-a -308G/A polymorphism and diabetic

nephropathy risk.

Materials and methods

Study identification and selection

A systematic search of the literature was performed to

investigate the associations between -308G/A poly-

morphism in the TNF-a gene and diabetic nephropathy

risk. PubMed, Embase, Ovid, Cochrane Library, the

China National Knowledge Infrastructure, VIP, Wan

Fang and the China Biology Medicine disc databases

were used (the last search was performed on March 10,

2013). The search terms were ‘‘diabetic nephropathy’’ in

combination with ‘‘TNF-a’’ or ‘‘tumor necrosis factor-

alpha’’ and in combination with ‘‘polymorphism’’ or

‘‘variant’’ or ‘‘mutation’’. There were no limitations to

language. Inclusion criteria were the following: (a) stud-

ies evaluating the association between -308G/A poly-

morphism in the TNF-a gene and diabetic nephropathy

risk; (b) case-controlled study design with diabetic

nephropathy participants and non-diabetic nephropathy

participants; and (c) studies with sufficient data (geno-

type distributions of cases and controls) available to

calculate an odds ratio (OR) with a 95 % confidence

interval (95 % CI). Exclusion criteria were the follow-

ing: (a) studies based on family or sibling pairs;

(b) studies with genotype frequencies or numbers not

reported; and (c) case reports, reviews or conference

abstracts. If more than one case-controlled study was

published by the same author(s) using the same case

series or an overlapping case series, the most suitable

studies with the largest number of cases or latest

publication dates were selected.

Data extraction

Two reviewers (ZYL and YJQ) independently

extracted data and reached a consensus on all items.

The following information was extracted from each

study: author, publication year, country of origin,

ethnicity, sample size, type of diabetes mellitus and

genotype number in cases and controls.

Statistical analysis

The strength of the association between -308G/A

polymorphism in the TNF-a gene and diabetic

nephropathy risk was measured by OR and 95 % CI.

The statistical significance of summary OR was

determined using a Z-test. The genetic models used

for the data analysis for the polymorphism were as

follows: (1) dominant model: GA ? AA versus GG;

(2) recessive model: AA versus GA ? GG; and (3)

other genetic models: AA versus GG and A versus G.

Heterogeneity was assessed by a v2-based Q statistic,

and a P value of \0.10 was considered statistically

significant. Pooled OR was analyzed by a fixed-effects

Int Urol Nephrol

123

model (the Mantel–Haenszel method) or a random-

effects model (the DerSimonian and Laird method)

according to heterogeneity. If the p value was[0.10,

the pooled OR was calculated using the fixed-effects

model; otherwise, the random-effects model was used

[12]. To analyze the ethnic-specific effects, subgroup

analysis was performed by ethnicity for the dominant

genetic model.

Publication bias was analyzed by visual inspection

of asymmetry in funnel plots, and Egger’s tests were

also carried out for statistical assessment. Sensitivity

analysis was conducted by sequentially deleting a

single study each time in an attempt to identify any

potential influence of an individual data set on the

pooled OR. All statistical tests were performed using

RevMan 4.2 and Stata 10.0 software.

Results

Study selection and characteristics

A total of 108 results were identified after an initial

search of the selected databases (Fig. 1). After reading

titles and abstracts, 64 potential studies in the correct

article types which were relevant to the association

between polymorphisms in the TNF-a gene and

diabetes were included for full-text review. After

reading full texts, 33 studies were excluded for not

being relevant to diabetic nephropathy and 20 were

excluded for not being relevant to -308G/A poly-

morphism in the TNF-a gene. Eleven studies remained

for data extraction. At this step, three studies were

excluded for not reporting usable data. A total of eight

case-controlled studies were identified [4, 13–19].

Diabetic nephropathy was diagnosed by renal

biopsy in the Babel et al. [19] study and assessed

from medical records as an established diagnosis in the

Krayenbuehl et al. [17] study. A diagnostic criterion

for diabetic nephropathy of serum creatinine (Scr)

C265 lmol/L and urinary albumin excretion rate

[ 200 mg/L was applied in the Prasad et al. [14]

study, while Wang et al. [13] included diabetic

patients with Scr C150 lmol/L and urinary albumin/

creatinine ratio C25 mg/mmol. In the other four

studies, the diagnostic criteria for diabetic nephropa-

thy were not addressed [4, 15, 16, 18]. The character-

istics, genotype and allele distributions from each

case-controlled study are listed in Table 1.

Quantitative data synthesis

Total analysis

A total of 1,277 cases and 1,740 controls were included

in meta-analysis of the relationship between -308G/A

polymorphism and the risk of diabetic nephropathy.

The heterogeneity of GA ? AA versus GG for all eight

case-controlled studies was analyzed. The value of v2

was 9.95 with 6 degrees of freedom, and p = 0.13 in a

fixed-effects model (Fig. 2). The I2 value is another

index for heterogeneity. The I2 value was 39.7 %,

suggesting absence of heterogeneity. The fixed-effects

model was, therefore, chosen to synthesize data.

Overall, OR was 0.84 (95 % CI = 0.71–1.00), and

the test for overall effect Z value was 1.92 (p = 0.05)

for GA ? AA versus GG model. Three other genetic

models were also used, and the summary of results of

genetic comparisons is listed in Table 2.

Subgroup analysis

Four Caucasian [15, 17–19] and four Asian [4, 13, 14,

16] case-controlled studies were included in the

subgroup analysis by ethnicity (Fig. 3). A positive

result for decreased diabetic nephropathy risk was

observed in the Asian studies (OR = 0.69, 95 %

CI = 0.51–0.94, p = 0.02 for GA ? AA vs. GG). No

statistically significant association was found in the

Caucasian studies (OR = 0.92, 95 %CI = 0.75–1.15,

P = 0.47 for GA ? AA vs. GG). Results from ethnic-

specific subgroup analysis are listed in Table 2.

Publication bias

Publication bias was assessed using Begg’s funnel

plots and Egger’s tests. The shape of the funnel plots

seemed symmetrical for the GA ? AA versus GG

comparative genetic model, suggesting an absence of

publication bias (Fig. 4). An Egger’s test was per-

formed to provide statistical evidence of funnel plot

symmetry, with the result indicating a lack of publi-

cation bias for meta-analysis (t = -0.76, p = 0.480).

Sensitivity analysis

To assess the stability of the results from meta-

analysis, we performed a one-study-removed sensi-

tivity analysis for the dominant genetic model. The

Int Urol Nephrol

123

result was statistically positive (p = 0.02), with the

study of Buraczynsk K. et al. [18] excluded (Table 3).

Statistically insignificant results (p [ 0.05) were

observed after sequentially excluding each of the

other studies (Table 3).

Discussion

Individual susceptibility to diabetic nephropathy var-

ies with environmental exposure and genetic

background [20]. The -308G/A polymorphism in

the TNF-a gene has been reported to be associated

with the risk of diabetic nephropathy [4, 18], but the

results remain inconclusive. The present study used

meta-analysis on 1,277 diabetic nephropathy cases

and 1,740 controls from eight case-controlled studies

to investigate the association between -308G/A

polymorphism in the TNF-a gene and diabetic

nephropathy risk. The results indicated that individ-

uals who carry the A allele (GA ? AA) may not have

an altered risk of diabetic nephropathy compared with

Fig. 1 Flow diagram of the studies included and excluded. CNKI the China National Knowledge Infrastructure, CBM the China

Biology Medicine disc

Table 1 Characteristics of publications included and distribution of genotypes and alleles among diabetic nephropathy patients and

controls

Authors Patients/

controls

DM

type

Country Ethnicity Years DN patients Controls A allele (%)

GG GA AA GG GA AA DN Control

Prasad et al. [14] 196/224 2 India Asian 2007 178 16 2 195 27 2 5.1 6.9

Lee et al. [4] 122/125 2 Korea Asian 2005 116 6 0 108 17 0 2.5 6.8

Lindholm et al. [15] 427/780 1&2 Sweden Caucasian 2008 254 152 21 443 292 45 22.7 24.5

Babel et al. [19] 44/113a 2 Germany Caucasian 2006 34 7 3 76 33 4 14.8 18.1

Buraczynsk et al. [18] 37/115a 1&2 Poland Caucasian 2004 22 13 2 86 24 5 23 14.8

Kung et al. [16] 24/23 2 China Asian 2010 0 24 0 0 23 0 50 50

Wang et al. [13] 388/323 2 China Asian 2008 326 62 261 62 NA NA

Krayenbuehlet al. [17] 39/37 2 Switzerland Caucasian 2006 30 9 28 9 NA NA

DM diabetes mellitus, DN diabetic nephropathy, NA not applicablea Studies with healthy controls

Int Urol Nephrol

123

homozygote GG carriers in the general population.

The -308G/A polymorphism might contribute to

decreased risk of diabetic nephropathy in Asians.

The P value of the result in general analysis was

0.05. This critical value requires accurate evaluation

and should be interpreted with caution. We had

previously identified an increased type 2 diabetes risk

for TNF-a -308A allele carriers in an earlier meta-

analysis. These outcomes indicate the complexity of

the pathogenesis of diabetes mellitus and its related

complications. Correlations between genes, ethnicity,

environment and other factors may potentially influ-

ence such diseases.

All possible methodological issues in the meta-

analysis were thoroughly investigated. No heteroge-

neity was found during the analyses of any genetic

model. There was no apparent publication bias,

although there was the potential that some data may

have been omitted, such as that from conference

abstracts. Sensitivity analysis indicated that data from

a single publication may have a significant influence

on the overall result [18]. Generally, the stability and

accuracy of the meta-analysis were good.

Of the eight original case-controlled studies

included, control populations were diabetes patients

without diabetic nephropathy in two studies [18, 19]

and disease-free patients in the other six studies [4,

13–17] (annotated in Table 1). This may lead to

potential bias and could be a reason for the results from

sensitivity analysis. The study by Kung W. J. et al. [16]

reported data with both diabetes mellitus and healthy

controls. We extracted data from both sets of case-

controlled designs, and similar results were obtained.

This meta-analysis had some limitations. First, as

the case-controlled studies used involved Caucasian

and Asian populations, the results may only be

applicable to those two ethnic groups. Further studies

are required to investigate the association in other

populations. Second, since diabetic nephropathy is

more frequent in men [21], gender-specific subgroup

Fig. 2 Meta-analysis for the association between diabetic nephropathy risk and TNF-a -308G/A polymorphism (GA ? AA vs. GG):

total analysis

Table 2 Comparison results of the total and subgroup analyses of the -308G/A polymorphism in different genetic models

Genetic model Ethnicity Studies Participants OR (95 % CI) Z p I2, % PHet Effect model

Overall 8 3,017 0.84 (0.71, 1.00) 1.92 0.05 39.7 0.13 Fixed

GA ? AA versus GG Caucasian 4 1,592 0.92 (0.75, 1.15) 0.72 0.47 40.1 0.17 Fixed

Asian 4 1,425 0.69 (0.51, 0.94) 2.33 0.02 29.2 0.24 Fixed

AA versus GA ? GG Overall 4 1,936 0.95 (0.60, 1.52) 0.21 0.84 0 0.75 Fixed

AA versus GG Overall 4 1,372 0.92 (0.58, 1.48) 0.33 0.74 0 0.76 Fixed

A versus G Overall 6 2,230 0.89 (0.75, 1.05) 1.39 0.17 41.1 0.13 Fixed

PHet: p value for heterogeneity

The bold values mean that their association is significant

Int Urol Nephrol

123

analysis should have been carried out if the original

data were found to be sufficient. Third, the diagnostic

criteria for diabetic nephropathy varied and were not

precisely described in some original articles. This

inconsistency may be a cause of potential bias. Fourth,

the majority of studies were based on type 2 diabetes

mellitus, whereas type 1 diabetes was only seen in two

studies. Therefore, we did not carry out type 1

diabetes-based subgroup analysis. Fifth, precise num-

bers of GA heterozygote or AA homozygote carriers

were inaccessible in two articles. Thus, related data

were not included in analyses with the corresponding

genetic models (AA vs. GA ? GG, AA vs. GG and A

vs. G). This may also be a cause of potential bias.

Sixth, only those studies found in the selected

databases were included for data analysis. Other

relevant published or unpublished studies with null

results might have been omitted.

Despite these limitations, bias has been minimized

as much as possible throughout this meta-analysis

through the methods employed for study identifica-

tion, data selection and statistical analysis, and for the

control of publication bias and sensitivity.

Fig. 3 Meta-analysis for the association between diabetic nephropathy risk and TNF-a -308G/A polymorphism (GA ? AA vs. GG):

subgroup analysis by ethnicity

Fig. 4 Begg’s funnel plot for publication bias in a selection of

studies (GA ? AA vs. GG)

Table 3 Sensitivity analysis by omitting each study in fixed-

effects model

Study omitted OR 95 % CI p

Prasad et al. [14] 0.84 0.61–1.17 0.30

Lee et al. [4] 0.87 0.70–1.10 0.25

Lindholm et al. [15] 0.79 0.53–1.17 0.24

Babel et al. [19] 0.86 0.71–1.03 0.09

Buraczynsk et al. [18] 0.81 0.67–0.97 0.02

Kung et al. [16] 0.82 0.62–1.09 0.17

Wang et al. [13] 0.82 0.56–1.19 0.29

Krayenbuehl et al. [17] 0.81 0.59–1.11 0.18

Int Urol Nephrol

123

In conclusion, this study is the first meta-analysis to

have assessed the association between TNF-a -308G/

A polymorphism and diabetic nephropathy risk. The

results suggest that the TNF-a -308A variant may be

associated with a decreased diabetic nephropathy risk

in Asians but not in Caucasians. Larger-scale case-

controlled studies are required to confirm these

findings.

Conflict of interest The authors declare that they have no

conflict of interest.

References

1. Grace BS, Clayton P, McDonald SP (2012) Increases in

renal replacement therapy in Australia and New Zealand:

understanding trends in diabetic nephropathy. Nephrology

(Carlton) 17(1):76–84

2. Zhuo L, Zou GM, Li WG, Lu JH, Ren WW (2013) Preva-

lence of diabetic nephropathy complicating non-diabetic

renal disease among Chinese patients with type 2 diabetes

mellitus. Eur J Med Res 18(1):4

3. Baban B, Liu JY, Mozaffari MS (2012) Endoplasmic

reticulum stress response and inflammatory cytokines in

type 2 diabetic nephropathy: role of indoleamine 2,

3-dioxygenase and programmed death-1. Exp Mol Pathol.

doi:10.1016/j.yexmp.2012.1011.1004

4. Lee SH, Lee TW, Ihm CG, Kim MJ, Woo JT, Chung JH

(2005) Genetics of diabetic nephropathy in type 2 DM:

candidate gene analysis for the pathogenic role of inflam-

mation. Nephrology (Carlton) 10:S32–S36

5. El-Sherbini SM, Shahen SM, Mosaad YM, Abdelgawad

MS, Talaat RM (2013) Gene polymorphism of transforming

growth factor-b1 in Egyptian patients with type 2 diabetes

and diabetic nephropathy. Acta Biochim Biophys Sin

45(4):330–338

6. Kurashige M, Imamura M, Araki S, Suzuki D, Babazono T,

Uzu T, Umezono T, Toyoda M, Kawai K, Imanishi M

(2013) The influence of a single nucleotide polymorphism

within CNDP1 on susceptibility to diabetic nephropathy in

Japanese women with type 2 diabetes. PLoS ONE 8(1):

e54064

7. Lee IS, Lee JH, Kim HJ, Lee JM, Lee SK, Kim HS, Lee JM,

Park KS, Ku BJ (2013) Novel ERBB receptor feedback

inhibitor 1 (ERRFI1) ?808 T/G polymorphism confers

protective effect on diabetic nephropathy in a Korean pop-

ulation. Dis Markers

8. Kaynar K, Ulusoy S, Ovali E, Vanizor B, Dikmen T, Gul S

(2005) TGF-beta and TNF-alpha producing effects of lo-

sartan and amlodipine on human mononuclear cell culture.

Nephrology (Carlton) 10(5):478–482

9. Mekinian A, Tamouza R, Pavy S, Gestermann N, Ittah M,

Mariette X, Miceli-Richard C (2011) Functional study of

TNF-a promoter polymorphisms: literature review and

meta-analysis. Eur Cytokine Netw 22(2):88–102

10. Chang WT, Wang YC, Chen CC, Zhang SK, Liu CH, Chang

FH, Hsu LS (2012) The -308G/A of Tumor Necrosis Factor

(TNF)-a and 825C/T of Guanidine Nucleotide Binding

Protein 3 (GNB3) are Associated with the Onset of Acute

Myocardial Infarction and Obesity in Taiwan. Int J Mol Sci

13(2):1846–1857

11. Guzman-Flores JM, Escalante M, Sanchez-Corona J, Gar-

cıa-Zapien AG, Cruz-Quevedo EG, Munoz-Valle JF,

Moran-Moguel MC, Saldana-Cruz AM, Flores-Martınez SE

(2013) Association analysis between -308G/A and -238G/

A TNF-alpha gene promoter polymorphisms and insulin

resistance in Mexican women with gestational diabetes

mellitus. J Investig Med 61(2):265–269

12. Yu ZY, Chen LS, Zhang LC, Zhou TB (2012) Meta-analysis

of the relationship between ACE I/D gene polymorphism

and end-stage renal disease in patients with diabetic

nephropathy. Nephrology (Carlton) 17(5):480–487

13. Wang Y, Wu CX, Tang JY, Chen ZE (2008) Relationship

between tumor necrosis factor-a G-308A gene polymor-

phism and risk of nephropathy in obese Chinese type 2

diabetic patients. Biao Ji Mian Yi Fen Xi Yu Lin Chuang

(Chinese) 4(15):71–75

14. Prasad P, Tiwari AK, Kumar KM, Ammini AC, Gupta A,

Gupta R, Thelma BK (2007) Association of TGFb1, TNFa,

CCR2 and CCR5 gene polymorphisms in type-2 diabetes

and renal insufficiency among Asian Indians. BMC Med

Genet 8(1):20

15. Lindholm E, Bakhtadze E, Cilio C, Agardh E, Groop L,

Agardh CD (2008) Association between LTA, TNF and

AGER polymorphisms and late diabetic complications.

PLoS ONE 3(6):e2546

16. Kung WJ, Lin CC, Liu SH, Chaung HC (2010) Association

of interleukin-10 polymorphisms with cytokines in type 2

diabetic nephropathy. Diabetes Technol Ther 12(10):

809–813

17. Krayenbuehl PA, Wiesli P, Schmid M, Schmid C, Ehses JA,

Hersberger M, Vetter W, Schulthess G (2007) TNF-alpha

-308G [A polymorphism modulates cytokine serum con-

centrations and macrovascular complications in diabetic

patients on aspirin. Exp Clin Endocrinol Diabetes 115(5):

322–326

18. Buraczynska K, Koziol-Montewka M, Majdan M, Tokarz

A, Ksiazek A (2004) Genetic determination of TNF and

myeloperoxidase production in dialyzed patients with dia-

betic nephropathy. Ren Fail 26(6):633–639

19. Babel N, Gabdrakhmanova L, Hammer MH, Schoenemann

C, Skrypnikov V, Poliak N, Volk HD, Reinke P (2006)

Predictive value of cytokine gene polymorphisms for the

development of end-stage renal disease. J Nephrol 19(6):

802–807

20. Okada H, Fukui M, Tanaka M, Matsumoto S, Mineoka Y,

Nakanishi N, Asano M, Yamazaki M, Hasegawa G, Na-

kamura N (2013) Visit-to-Visit blood pressure variability is

a novel risk factor for the development and progression of

diabetic nephropathy in patients with Type 2 diabetes.

Diabetes care

21. Mollsten A, Vionnet N, Forsblom C, Parkkonen M, Tarnow

L, Hadjadj S, Marre M, Parving HH, Groop PH (2011) A

polymorphism in the angiotensin II type 1 receptor gene has

different effects on the risk of diabetic nephropathy in men

and women. Mol Genet Metab 103(1):66–70

Int Urol Nephrol

123