Abstract

Background. Cytokine gene polymorphisms have been associated with poor outcomes after renal transplantation such as chronic allograft nephropathy (CAN), graft rejection (GR) and graft failure (GF), but the effects of these polymorphisms are still controversial. We therefore conducted a systematic review, with individual patient data (IPD) where possible, to determine the association between cytokine polymorphisms (TGF-β1, TNF-α and IL-10) and outcomes after renal transplantation.

Methods. Five investigators were willing to participate and provided IPD. The outcomes of interest were GF, GR and CAN. Subjects with at least one of these were classified as having poor outcomes. Heterogeneity of gene effects was assessed. Multiple logistic regression was applied to assess gene effects, adjusting for clinical variables such as HLA matching and age.

Results. One-thousand and eighty-seven subjects were included in the IPD meta-analysis. Pooled results showed no evidence of heterogeneity and indicated that the strongest variables determining poor outcomes are HLA mismatching (OR = 1.6–1.8 for ≥3 HLA-A, -B, -DR mismatches compared with those with <3 mismatches) and age (OR = 1.2–1.4 for age 45 years or more). Incremental information on risk of a poor outcome is provided by the TGF-β1c10 polymorphism (OR = 1.5, P = 0.034, 95% CI: 1.0–2.2 for TC genotype compared to TT genotype). Haplotypes of TGF-β1 at c10 and c25 were inferred and the C-C haplotype was a marker of a poor outcome (OR = 1.3, P = 0.177, 95% CI: 1.0–2.3). Three polymorphisms of the IL-10 gene at −1082, −819, −592 are in strong linkage disequilibrium with each other (correlation coefficients: 0.6–1) and inferred haplotypes between these three loci show some association, with ACC increasing the risk of poor events com- pared to GCC (OR = 1.3, P = 0.044, 95% CI: 0.9–1.6).

Conclusion. Pooled results to date suggest possible association between both the TGF-β1 c10 polymorphism and a 3-SNP-haplotype of IL-10 and poor outcomes in renal transplantation, but this needs to be confirmed in larger studies.

Introduction

Cytokine gene polymorphisms have been associated with adverse outcomes of renal transplantation such as delayed graft function (DGF), chronic allograft nephropathy (CAN) and graft rejection (GR). These polymorphisms include variants in the genes coding for IL-10 at −1082, −819 and −592, IL-6 at −174 [ 1–10 ], transforming growth factor-beta1 (TGF-β1) at codons 10 and 25 [ 1–6 , 8,9 , 11 ], tumour necrosis factor-α (TNF-α) at position −308 [ 1–6 , 8–10 ] and interferon gamma (IFN-λ) in intron 1 at +874 [ 2 , 9 , 12 ].

The IL-6 and IL-10 genes are located at chromosome 7p21 and 1q31-q32 (MIM # 147620 and 124092, respectively). The SNPs (single nucleotide polymorphisms) studied to date are at positions −174 (G→C) for IL-6, −1082 (G→A), −819 (C→T) and −592 (C→A) for IL-10; the latter three are all in the promoter regions and are thought to influence IL-10 production , which may affect the risk of GR. In contrast, TNF-α located at 6p21.3 (MIM # 191160), TGF-β1 located at 2p13 (MIM # 190170 ) and IFN-γ located at 12q14 (MIM # 147570) mediate immune response and may have pro- or anti-inflammatory effects depending on the context. The SNPs are G→A substitution at position −308 for TNF-α, T→C at codon 10 and G→C at codon 25 for TGF-β1, and T→A at +874 for IFN-γ.

The effects of these SNPs on graft outcomes are still controversial, partly due to lack of power in the original studies and partly due to linkage disequilibrium between the multiple adjacent SNPs. However, we hypothesized that gene polymorphisms associated with increased production of pro-inflammatory and decreased production of anti-inflammatory cytokines may have deleterious effects on the transplant outcome. We therefore conducted a systematic review, with individual patient data (IPD) where possible, with the following primary aims:

  • First, to estimate the direction, magnitude and genetic mode of effect of these cytokine gene polymorphisms on poor outcomes of transplantation by looking at each SNP separately, adjusting for clinical variables such as HLA matching, cold ischaemic time, age and sex.

  • Second, to estimate the combined SNP effects by looking at haplotypes, again adjusting for clinical variables given above.

Materials and methods

Search strategy

MEDLINE (from January 1966 to January 2007) was searched using the PubMed search engine. The search strategies were (interleukin OR ‘IL-2’ OR ‘IL-15’ OR ‘IL-4’ OR ‘IL-6’ OR ‘IL-10’ OR ‘IL-13’) AND (gene OR polymorphism) AND (‘renal transplant * ’) AND (rejection OR ‘graft failure’); (‘Transforming growth factor beta * ’ OR TGF beta) AND (gene OR polymorphism) AND (‘renal transplant * ’) AND (rejection OR ‘graft failure’); (‘tumour necrosis factor alpha’ OR TNF alpha) AND (gene OR polymorphism) AND (‘renal transplant * ’) AND (rejection OR ‘graft failure’); (‘interferon-gamma’ OR INF) AND (gene OR polymorphism) AND (‘renal transplant * ’) AND (rejection OR ‘graft failure’).

Inclusion criteria

Using the above search strategies, 48 studies were identified. Abstracts and full papers if needed were reviewed; only IL-10: −1082, −819, −592; TNF-α-308, TGF-β1: codon 10 and 25 and IFN-γ: +874 polymorphisms had at least two to three studies assessing association between polymorphisms and transplant outcomes and so these were selected for pooling. Any human population-based association studies, regardless of sample size, were included if they met the following criteria:

  • Association studies between cytokine gene polymorphisms (i.e. IL-10: −1082, −819, −592; TNF-α-308, TGF-β1: codon 10 and 25, and IFN-γ: +874) and clinical outcomes of renal transplantation, which included GF, CAN and acute or chronic GR.

  • Having a stable graft function group as a comparator.

The reference lists of the retrieved articles were also reviewed to identify publications on the same topic. Where there were multiple publications from the same study group, the most complete and recent results were used.

Outcome of interest

The outcomes of interest were GF, acute or chronic GR and CAN. GF was defined as return to dialysis. GR was defined as a >10–20% rise in serum creatinine from baseline within 3 months after transplant (acute GR) or later (chronic GR) with confirmation by biopsy; although many practitioners use reversibility with immunosuppressives as part of the definition of rejection, this information was not consistently available and had to be dropped from the definition. CAN was defined as chronic deterioration of GF (e.g. percent decrease in glomerular filtration rate >5%/year), with/without confirmation by biopsy. However, this was differentiated from chronic GR by biopsy. Although pathologic mechanisms of GF and chronic rejections may be quite similar but different from acute rejection and CAN, the limited numbers of participants and studies, and the lack of overlap in polymorphisms studies, meant there was insufficient power to analyse each outcome separately. We therefore combined them by classifying cases as those who had at least one of these outcomes.

Data collection and management

Corresponding authors of included studies were contacted with requests to provide IPD, consisting of genetic polymorphisms, demographic and clinical variables known to be related to renal transplant outcomes, i.e. age of recipient, cold ischaemia time, gender, source of transplant and HLA compatibility. Data cleaning and checking were performed separately for each centre. Any unclear coding or outlier was clarified by contact with the authors.

Statistical analysis

Hardy–Weinberg equilibrium (HWE) was assessed in the control group for each study using the exact test [ 13–15 ]. Only studies that observed HWE were included in analysis of gene effects. Logistic regression analysis was applied by fitting gene into a logit model. Study variable was also included in this model since included studies had differences in studied subjects, design, measurements, etc., and these might confound the gene effects. Random- and fixed-effect models were applied for fitting the study effect; if the random-effect model explained more of the variation in the outcome, this model was applied; otherwise the fixed-effect model was used. In addition, transplant variables such as age, gender, cold ischaemic time (in hours) and number of HLA mismatches were also included in the model. Age and HLA mismatches were categorized according to mean and median. The likelihood-ratio test was applied to select the most parsimonious model. All analyses were performed using STATA version 9.2 [ 16 ] except for inferring haplotype, which was performed using SimHap beta 2.1 [ 17 ]. A P -value <0.05 was considered statistically significant, except heterogeneity test in which p < 0.10 was used.

Results

As displayed in Figure 1 , 48 studies were identified, of which 13 were considered eligible [ 1 , 12 , 18 ]. The number of studies looking at each cytokine polymorphisms were as follows:

Fig. 1

Flowchart indicating studies included in this review.

Fig. 1

Flowchart indicating studies included in this review.

  • TGF-β1: codon 10 T/C = 8 studies, codon 25 G/C = 8 studies;

  • IL-10: −1082 G/A = 11 studies, −819 C/T = 5 studies, −592 C/A = 5 studies;

  • TNF-α: −308 G/A = 11 studies, −1032 T/C = 1 study, −859 G/A = 1 study, −865 C/A = 1 study;

  • IL-6: −174 C/G = 5 studies;

  • IL-2: −330 T/G = 1 study;

  • IL-1-R1: −970 C/T = 1 study.

Except for one whose full paper was in Chinese and whose contact details could not be located [ 8 ], all 13 authors were e-mailed an invitation to collaborate. Although 10 initially agreed to collaborate, only 5 authors [ 1 , 5 , 7 , 11 , 18 ] were able to provide IPD. Characteristics of theses studies are described in Table 1 : mean age ranged from 36 to 48 years, males constituted between 50 and 67% of the groups and number of HLA mismatches ranged from 1 to 4. Only four studies provided data for the type of donor and cold ischaemic time; percentage of cadaveric donors ranged from 0 (i.e. all were living donor) to 93 and cold ischaemic time ranged between 9 and 20 h. Each study had GF, GR or CAN as the outcome and these three outcomes were combined, resulting in 478 (44.0%) cases (i.e. poor clinical outcomes) and 609 (56.0%) controls (i.e. those with stable graft function).

Table 1

General characteristics of included studies

Authors Year n Ethnicity Mean age Mean age Percentage HLA-A, -B, -DR Cold ischaemic Outcome 
       of cadaveric  mismatches  time  
       donors  (median)  (h, median)  
Morgun [7] 2003 63 Caucasian 38 51 – AGR 
McDenial [ 5 ]  2002 76 African American 39 57 93 18 CAN, GF, CGR 
Dmitrienko [ 1 ]  2005 100 Caucasian 44 57 60 GF, AGR 
Inigo [ 11 ]  2003 715 Caucasian 48 67 – 20 CAN, GF, AGR 
Chow [ 18 ]  2005 133 Asian 36 50 63 GF, AGR 
Authors Year n Ethnicity Mean age Mean age Percentage HLA-A, -B, -DR Cold ischaemic Outcome 
       of cadaveric  mismatches  time  
       donors  (median)  (h, median)  
Morgun [7] 2003 63 Caucasian 38 51 – AGR 
McDenial [ 5 ]  2002 76 African American 39 57 93 18 CAN, GF, CGR 
Dmitrienko [ 1 ]  2005 100 Caucasian 44 57 60 GF, AGR 
Inigo [ 11 ]  2003 715 Caucasian 48 67 – 20 CAN, GF, AGR 
Chow [ 18 ]  2005 133 Asian 36 50 63 GF, AGR 

AGR, acute graft rejection; CAN, chronic allograft nephropathy; GF, graft failure; CGR, chronic graft rejection.

Among these five IPD studies, all studies genotyped TGF-β1 10 T/C [ 1 , 5 , 7 , 11 , 18 ], four studies genotyped TGF-β1 25 G/C [ 1 , 5 , 7 , 11 ], four studies genotyped TNF-α-308 A/G [ 1 , 5 , 7 , 11 ], three studies genotyped IL-10-1082 G/A [ 1 , 5 , 7 ], two studies genotyped IL-10-819 C/T and IL-10-592 [ 1 , 5 , 7 ] C/A and two studies genotyped INF-γ+874 T/A, IL-6-174 G/C and IL-2-330 T/G, cytokine polymorphisms [ 5 , 7 ]. Only four polymorphisms (i.e. TGF-β1 10 T/C, TGF-β1 25 G/C, TNF-α-308 A/G, IL-10-1082 G/A) were common and thus had enough IPD to assess gene effects; the remainder were rarer and thus were insufficient to pool data. The response rates (i.e. willing to collaborate and provided data) for these corresponding four polymorphisms were 5/8 (62.5%), 4/8 (50.0%), 4/11 (36.4%) and 3/11 (27.3%) studies, respectively.

Genotype frequencies and transplant factors between cases and controls are described in Table 2 . The pooled prevalence of alleles T, G, G and G for TGF-β1 10 T/C, TGF-β1 25 G/C, TNF-α-308 A/G and IL-10-1082 G/A were 52.3% (95% CI: 47.9–56.8%), 93.7% (95% CI: 90.6–96.8%), 79.1% (95% CI: 75.7–82.6%) and 41.7% (95% CI: 34.2–49.2%), respectively, with no heterogeneity except for TGF-β1 25 G/C.

Table 2

Distribution of genotypes and transplant variables between groups of outcome of renal transplantation: univariate analyses

 Outcome  ORs 
Cytokine polymorphisms Poor number (%) Good number (%) P -value   OR * 95% CI  Heterogeneity chi-square ( P -value)  
TGF-β1-10 T/C       
   TT 87 (26.1) 98 (30.1)    
   TC 177 (53.0) 151 (46.5) 0.101 1.4 0.9–2.0 3.6 (0.457) 
   CC 70 (20.9) 76 (23.4) 0.889 1.0 0.7–1.6 2.2 (0.695) 
TGF-β1 25 G/C       
   GG 198 (84.6) 184 (89.7)    
   GC 31 (13.3) 18 (8.8) 0.297 1.4 0.8–2.6 3.0 (0.381) 
   CC 5 (2.1) 3 (1.5) 0.719 1.3 0.4–5.5 1.7 (0.625) 
TNF-α-308 G/A       
   GG 248 (70.9) 325 (74.4)    
   GA 83 (23.7) 95 (21.7) 0.399 1.2 0.8–1.6 0.7 (0.885) 
   AA 19 (5.4) 17 (3.9) 0.370 1.4 0.7–2.7 5.9 (0.120) 
IL-10-1082 G/A       
   GG 22 (18.0) 21 (18.0)    
   GA 46 (37.7) 48 (41.0) 0.917 1.0 0.5–2.0 1.1 (0.578) 
   AA 54 (44.3) 48 (41.0) 0.370 1.3 0.6–2.6 1.1 (0.580) 
Transplant factors Poor number (%) Good number (%) P -value  OR (95% CI)   
Age       
   ≥45 250 (53.1) 286 (48.0) 0.039  1.3 (1.0–1.7)  
   <45 221 (46.9) 310 (52.0)     
Gender       
   Male 304 (64.3) 366 (60.9) 0.176  1.2 (0.9–1.5)  
   Female 169 (35.7) 235 (39.1)     
No. of HLA mismatches       
   ≥3 347 (75.1) 385 (66.6) 0.002  1.7 (1.2–2.3)  
   <3 115 (24.9) 193 (33.4)     
Cold ischaemic time (h)       
   ≥18 224 (46.9) 297 (48.8) 0.930  0.9 (0.8–1.3)  
   <18 254 (53.1) 312 (51.2)     
 Outcome  ORs 
Cytokine polymorphisms Poor number (%) Good number (%) P -value   OR * 95% CI  Heterogeneity chi-square ( P -value)  
TGF-β1-10 T/C       
   TT 87 (26.1) 98 (30.1)    
   TC 177 (53.0) 151 (46.5) 0.101 1.4 0.9–2.0 3.6 (0.457) 
   CC 70 (20.9) 76 (23.4) 0.889 1.0 0.7–1.6 2.2 (0.695) 
TGF-β1 25 G/C       
   GG 198 (84.6) 184 (89.7)    
   GC 31 (13.3) 18 (8.8) 0.297 1.4 0.8–2.6 3.0 (0.381) 
   CC 5 (2.1) 3 (1.5) 0.719 1.3 0.4–5.5 1.7 (0.625) 
TNF-α-308 G/A       
   GG 248 (70.9) 325 (74.4)    
   GA 83 (23.7) 95 (21.7) 0.399 1.2 0.8–1.6 0.7 (0.885) 
   AA 19 (5.4) 17 (3.9) 0.370 1.4 0.7–2.7 5.9 (0.120) 
IL-10-1082 G/A       
   GG 22 (18.0) 21 (18.0)    
   GA 46 (37.7) 48 (41.0) 0.917 1.0 0.5–2.0 1.1 (0.578) 
   AA 54 (44.3) 48 (41.0) 0.370 1.3 0.6–2.6 1.1 (0.580) 
Transplant factors Poor number (%) Good number (%) P -value  OR (95% CI)   
Age       
   ≥45 250 (53.1) 286 (48.0) 0.039  1.3 (1.0–1.7)  
   <45 221 (46.9) 310 (52.0)     
Gender       
   Male 304 (64.3) 366 (60.9) 0.176  1.2 (0.9–1.5)  
   Female 169 (35.7) 235 (39.1)     
No. of HLA mismatches       
   ≥3 347 (75.1) 385 (66.6) 0.002  1.7 (1.2–2.3)  
   <3 115 (24.9) 193 (33.4)     
Cold ischaemic time (h)       
   ≥18 224 (46.9) 297 (48.8) 0.930  0.9 (0.8–1.3)  
   <18 254 (53.1) 312 (51.2)     

* Adjusting study effect.

Genotype frequencies for the four polymorphisms are described in Table 2 . Before pooling gene effects across studies, heterogeneity was assessed. There was no evidence of heterogeneity for all three polymorphisms (see Table 2 ). The fixed-effect logistic model was thus applied to assess gene effects. After adjusting for the study effect, the impact of these three polymorphisms was not significant whereas two transplant factors, i.e. age ≥45 years and number of HLA-A, -B, -DR mismatches ≥3, were significantly associated with a poor outcome. These two variables were therefore included in the same model as each polymorphism, as described in Table 3 . We found that after adjusting for age and number of HLA mismatches, TGF-β1 10 T/C genotype was significantly associated with a poor outcome; those having the TC genotype were ∼1.5 times (95% CI OR: 1.0–2.2) more likely to develop a poor outcome than those with the TT genotype. In addition, having three or more HLA-A, -B, -DR mismatches increased the risk of having a poor outcome by 80% (OR = 1.8, 95% CI: 1.2–2.6) compared with those with less than three mismatches. In addition, recipients aged 45 years or older were 1.4 times (95% CI of OR: 1.0–2.0) more likely to have poor outcomes than recipients aged <45 years. Although associations with TGF-β1 25 G/C, TNF-α-308 A/G, and IL-10-1082 G/A were not significant, there was a trend towards association. For instance, those with the GC genotype for TGF-β1 25 G/C had ∼30% (OR = 1.3, 95% CI of OR: 0.7–2.3) higher risk of poor outcomes compared with the GG genotype; the risks were ∼20% (OR = 1.2, 95% CI of OR: 0.8–1.7) and 30% (OR = 1.3, 95% CI of OR: 0.6–2.7) higher for GA and AA, respectively, as compared to the GG genotype for TNF-α-308 A/G; for IL-10-1082 G/A, the risk of having poor outcome was ∼30% (OR = 1.3, 95% CI: 0.6–2.9) higher in AA genotype compared with GG genotype but GA and GG had a similar risk (OR = 1.0, 95% CI: 0.5–2.2).

Table 3

Assessing genetic effects on renal transplantation adjusting for transplant factors using individual data: multivariate analysis

Factors OR P 95% CI 
TGF-β1 10 T/C    
   TT   
   TC 1.51 0.034 1.03–2.22 
   CC 1.02 0.938 0.64–1.61 
No. of HLA mismatches    
   ≥3 1.78 0.002 1.24–2.59 
   <3   
Age    
   ≥45 1.42 0.042 1.01–1.99 
   <45   
    
TGF-β1 25 G/C    
   GG   
   GC 1.27 0.458 0.67–2.39 
   CC 1.02 0.976 0.24–4.43 
No. of HLA mismatches    
   ≥3 2.01 0.003 1.26–3.19 
   <3   
Age    
   ≥45 1.19 0.377 0.80–1.80 
   <45   
    
TNF-α-308 G/A    
   GG   
   GA 1.18 0.353 0.83–1.68 
   AA 1.33 0.449 0.64–2.71 
No. of HLA mismatches    
   ≥3 1.57 0.023 1.06–2.32 
   <3   
Age    
   ≥45 1.34 0.061 0.99–1.81 
   <45   
IL-10-1082 G/A    
   GG   
   GA 1.0 0.989 0.46–2.20 
   AA 1.30 0.521 0.59–2.85 
No. of HLA mismatches    
   ≥3 1.96 0.035 1.05–3.70 
   <3   
Age    
   ≥45 0.89 0.676 0.51–1.55 
   <45   
Factors OR P 95% CI 
TGF-β1 10 T/C    
   TT   
   TC 1.51 0.034 1.03–2.22 
   CC 1.02 0.938 0.64–1.61 
No. of HLA mismatches    
   ≥3 1.78 0.002 1.24–2.59 
   <3   
Age    
   ≥45 1.42 0.042 1.01–1.99 
   <45   
    
TGF-β1 25 G/C    
   GG   
   GC 1.27 0.458 0.67–2.39 
   CC 1.02 0.976 0.24–4.43 
No. of HLA mismatches    
   ≥3 2.01 0.003 1.26–3.19 
   <3   
Age    
   ≥45 1.19 0.377 0.80–1.80 
   <45   
    
TNF-α-308 G/A    
   GG   
   GA 1.18 0.353 0.83–1.68 
   AA 1.33 0.449 0.64–2.71 
No. of HLA mismatches    
   ≥3 1.57 0.023 1.06–2.32 
   <3   
Age    
   ≥45 1.34 0.061 0.99–1.81 
   <45   
IL-10-1082 G/A    
   GG   
   GA 1.0 0.989 0.46–2.20 
   AA 1.30 0.521 0.59–2.85 
No. of HLA mismatches    
   ≥3 1.96 0.035 1.05–3.70 
   <3   
Age    
   ≥45 0.89 0.676 0.51–1.55 
   <45   

We further inferred haplotypes for TGF-β at positions 10 and 25 using the EM algorithm based on the four studies [ 1 , 5 , 7 , 11 , 18 ] that had data for both polymorphisms. These two SNPs were moderately linked ( r = 0.3) and three of the possible four haplotypes were identified (Table 4 ). We found that the C-C haplotype had a 30% (OR = 1.3; 95% CI = 1.0–2.3) higher risk of a poor outcome than the T-G haplotype, whereas the C-G haplotype had a similar risk as the T-G haplotype (OR = 0.9, 95% CI = 0.7–1.4).

Table 4.

Estimation of haplotype effects on the outcome of renal transplantation

TGFβ1 (codons 10 and 25)  Case n = 468 (%)   Control n = 408 (%)   OR a (95% CI)  
   T-G 262 (55.9) 233 (57.1) 
   C-G 165 (35.3) 151 (37.0) 0.9 (0.7–1.4) 
   C-C 41 (8.8) 24 (5.9) 1.30 (1.0–2.3) 
IL-10 (-1082-819-592)  Case n = 354 (%)   Control n = 342 (%)   OR a (95% CI)  
   G-C-C 163 (47.7) 164 (46.3) 
   A-C-C 82 (24.0) 93 (26.3) 1.3 (0.9–1.6) 
   A-T-A 97 (28.3) 97 (27.4) 1.1 (0.8–1.4) 
TGFβ1 (codons 10 and 25)  Case n = 468 (%)   Control n = 408 (%)   OR a (95% CI)  
   T-G 262 (55.9) 233 (57.1) 
   C-G 165 (35.3) 151 (37.0) 0.9 (0.7–1.4) 
   C-C 41 (8.8) 24 (5.9) 1.30 (1.0–2.3) 
IL-10 (-1082-819-592)  Case n = 354 (%)   Control n = 342 (%)   OR a (95% CI)  
   G-C-C 163 (47.7) 164 (46.3) 
   A-C-C 82 (24.0) 93 (26.3) 1.3 (0.9–1.6) 
   A-T-A 97 (28.3) 97 (27.4) 1.1 (0.8–1.4) 

a Adjusting study effect.

Since only three studies [ 1 , 5 , 7 ] had IPD for the IL-10-1082 polymorphism, combing these IPD with three other studies that had only summary data [ 2 , 4 , 6 ] (data were extractable from published papers) might increase the power for detecting association. The summary data were thus expanded to simulate individual data and combined with the IPD data. The pooled prevalence of the G allele was 45.0% (95% CI: 38.3–51.6). Two pooled odds ratios were estimated, OR 1 (AA versus GG) and OR 2 (GA versus GG), as described in Table 5 . Both of these odds ratios were homogenous (OR 1 chi-square = 2.6, df = 5, P = 0.766; OR 2 chi-square = 6.6, df = 5, P = 0.255). The pooled OR 1 and OR 2 using fixed-effect models were 1.25 (95% CI: 0.80–1.88) and 1.10 (95% CI: 0.90–1.41), respectively, i.e. recipients with AA and GA genotypes had ∼25 and 10% higher risk of having poor outcomes than recipients with GG genotypes. Although these risks were not statistically significant, there was a trend towards association. We therefore assessed the haplotype effect of three polymorphisms in the same gene, i.e. IL-10: −1082 (G/A), −819 (C/T) and −592 (C/A). Only two studies had individual data for all three polymorphisms and these were combined with one study where summary data of haplotype were extractable. The three SNPs were moderately to highly linked, i.e. the r coefficients were 0.6, 0.6 and 1 for position 1082 versus −819, −1082 versus −592 and −819 versus −592, respectively. Haplotypes were inferred using the EM algorithm and only three haplotypes were identified (Table 4). We found that having the A-C-C haplotype carried a borderline significant risk of having a poor outcome (OR = 1.26, 95% CI: 0.94–1.57) compared with the G-C-C haplotype, whereas having the A-T-A haplotype carried only 10% (95% CI: 0.82–1.37) higher risk of a poor outcome than the G-C-C haplotype.

Table 5

Distribution of IL-10-1082 polymorphism genotypes in poor and good outcome groups.

 Poor outcome Good outcome  OR 1  OR 2 
IL-10-1082 n GG GA AA n GG GA AA HWE (95% CI) (95% CI) 
Morgun [ 7 ]  20 10 43 14 22 0.091 1.6 (0.3, 9.1) 2 (0.3, 12.1) 
Melk [ 6 ]  75 42 25 30 12 11 0.454 1.9 (0.6, 6.9) 3.1 (0.9, 10.2) 
McDenial [ 5 ]  43 16 24 33 16 13 1.000 2.5 (0.5, 12.7) 1.3 (0.3, 6.9) 
Hutchings [ 2 ]  41 29 60 26 26 0.777 0.6 (0.2, 2.6) 2.2 (0.6, 8.3) 
Marshall [ 4 ]  114 43 44 27 95 33 45 17 0.834 1.2 (0.6, 2.6) 0.8 (0.4, 1.4) 
Dmitrienko [ 1 ]  59 17 22 20 41 10 18 13 0.531 1.0 (0.3, 2.6) 0.7 (0.3, 1.9) 
Pooled OR (95% CI)          1.3 (0.8, 1.9) 1.1 (0.9, 1.4) 
 Poor outcome Good outcome  OR 1  OR 2 
IL-10-1082 n GG GA AA n GG GA AA HWE (95% CI) (95% CI) 
Morgun [ 7 ]  20 10 43 14 22 0.091 1.6 (0.3, 9.1) 2 (0.3, 12.1) 
Melk [ 6 ]  75 42 25 30 12 11 0.454 1.9 (0.6, 6.9) 3.1 (0.9, 10.2) 
McDenial [ 5 ]  43 16 24 33 16 13 1.000 2.5 (0.5, 12.7) 1.3 (0.3, 6.9) 
Hutchings [ 2 ]  41 29 60 26 26 0.777 0.6 (0.2, 2.6) 2.2 (0.6, 8.3) 
Marshall [ 4 ]  114 43 44 27 95 33 45 17 0.834 1.2 (0.6, 2.6) 0.8 (0.4, 1.4) 
Dmitrienko [ 1 ]  59 17 22 20 41 10 18 13 0.531 1.0 (0.3, 2.6) 0.7 (0.3, 1.9) 
Pooled OR (95% CI)          1.3 (0.8, 1.9) 1.1 (0.9, 1.4) 

Discussion

We investigated the effect of a number of cytokine gene polymorphisms on renal transplant outcome. We obtained individual level data from five previous studies; the addition of three other studies with partially reconstructed IPD from summary data raised the total number of cases and controls to ∼300 and 350, respectively, depending on the polymorphism. Despite this number, we were unable to demonstrate any statistically significant effects based on single-locus analysis, due either to no substantial genetic effect or to insufficient power to detect more modest genetic effects. For example, detecting an odds ratio of 1.3 with an allele frequency of roughly 50% (e.g. TGF-β1 10 T/C) would require over 450 case–control pairs if an additive model was operating but over 1100 pairs if a recessive or dominant model was operating. At minor allele frequencies of under 10% (e.g. TGF-β1 25 G/C), over 1100 pairs are needed to detect this effect size if an additive model is operating, and over 20 000 for a recessive model.

Gains in power can be made using linked, i.e. haplotype, data [ 19 ] and this was pursued. We found some indication of an effect of TGF-β1 c10 and c25 haplotypes and to a lesser extent for the haplotype of the three IL-10 polymorphisms, although these effects appear modest. These genetic effects are in addition to the other co-variables in the model, i.e. HLA matching and age. We can get some idea of the relative importance of the genetic variable in comparison to the other variables by looking at the change in pseudo R2 in the model with and without the variable of interest. In the multivariate model incorporating HLA matching, age and TGF-β1 10 T/C, the delta pseudo R2 is 39%, 12% and 16%, respectively; this indicates that the polymorphism contributes approximately half the predictive power that HLA matching does, and more than age. Hence, despite a very modest effect size, this polymorphism contributes a reasonable amount in predicting poor events compared to other accepted variables.

Among the five IPD studies, the majority of subjects were Caucasians (three studies), whereas only one study had African Americans, and one study had Asian subjects. The pooled prevalences of alleles T, G and G for TGF-β1 10 T/C, TGF-β1 25 G/C and TNF-α-308 A/G in Caucasian studies were 55.2% (95% CI: 49.7–60.6%), 94.5% (95% CI: 91.3–97.8%) and 79.5% (95% CI: 75.9–83.1%), respectively. These corresponding prevalences of alleles were 50% (95% CI: 35.6–64.4%), 81.6% (95% CI: 69.3–93.9%) and 79.1% (95% CI: 75.7–82.6%) in African American subjects. Only one TGF-β1 10 T/C polymorphism was studied in Asians and the prevalence of T allele was 45% (95% CI: 35.8–54.3%). These prevalences were very similar and thus gene effects should not be much different across ethnicities.

These results come with some caveats. It is difficult to reconcile what is known about the biology of these cytokines with the direction of the effect seen; this however may reflect a lack of complete understanding about the mechanisms and serve as a warning about overly simple assumptions about the effect of polymorphisms.

Cytokines participate in extremely complex cascades and can have opposing effects. For example, although the IL-10-1082 polymorphism leads to increased IL-10 levels both in vivo and in vitro [ 20,21 ], the results of increased IL-10 are difficult to predict. On one hand IL-10 downregulates MHC class II molecules, ICAM expression and inhibits the synthesis of proinflammatory cytokines (INFc, TNF-α, IL-1, IL-2, IL-6 and IL-8), all of which have anti-inflammatory effects. On the other hand, IL-10 promotes B cell activation and maintains B cell viability by inhibiting apoptosis, both of which would have pro-inflammatory effects. The −1082 polymorphism at IL-10 is located in the promoter region, with the wild-type allele G leading to greater promoter activity and higher IL-10 expression levels than the A allele. Although McWarle et al . reported some conflicting results, perhaps dependent on the population used [ 22 ], in our study the variant allele is paradoxically associated with an increase in poor outcomes, similarly to liver transplant patients [ 23 ].

TGF-β1 is produced by a variety of cell types, including cells of the immune system. TGF-β1 has also been shown to be a potent immunomodulatory molecule, with both immunosuppressive and pro-inflammatory abilities. TGF-β1 can, under certain conditions, stimulate proliferation and differentiation of T cells [ 24,25 ], but most commonly has been shown to inhibit lymphocyte proliferation, downregulate receptor and cytokine expression and inhibit generation of cytotoxic T lymphocytes [ 26,27 ]. Over-expression of TGF-β1 was reported to correlate with CAN [ 28 ], the rate of decline in renal function and cyclosporine toxicity in human allografts [ 29,30 ]. The TGF-β1 10 T/C polymorphism is in a region of the gene that codes for an α-helix that directs transport of the TGF-β1 protein through the cell; the leucine to proline amino acid substitution changes the charge and disrupts the helix, leading to lower levels of TGF-β1 production. The arginine to proline change due to the polymorphism at TGF-β1 25 G/C interferes with enzymatic cleavage and also lowers the levels of TGF-β1. These lower levels of TGF-β1 are associated with poorer outcomes, which is contrary to biological expectation and to the previous work with kidney and other transplant patients, e.g. heart, lung [ 23 ].

It is also worth noting that these results are driven mainly by case–control studies; these studies are subject to bias and it is interesting that a cohort study of these polymorphisms does not confirm these findings [ 31 ]. This may be explained by a lack of power with prospective cohorts, or bias, in that those with adverse events are recruited as cases in case–control studies and although they may have a whole genetic profile associated with that adverse event, only a few polymorphisms are genotyped and all of the increased risk is assigned to those (known) polymorphisms.

Advantages and drawbacks of meta-analysis of IPD and summary data have been reported and discussed previously [ 32,33 ]. The IPD meta-analysis method is more flexible, particularly in exploring causes of heterogeneity [ 34 ] and in assessing whether potential causes of heterogeneity affect individual- or study-level variables or both [ 35,36 ]. For instance, for those five studies where IPD data were available, the effect of HLA-A, -B, -DR mismatches was assessed using both summary data and IPD for pooling odds ratio of TC versus CC for TGF-β1 10 T/C. For summary data, there was no significant association between the HLA mismatch and odds ratio using the meta-regression approach (coefficient = −0.01, SE = 0.02, P = 0.587) whereas the association was present for IPD data using logistic regression. If the number of studies was not taken into account, this could be interpreted as indicating that the HLA mismatch did not affect poor outcomes at the study level but did affect poor outcomes at the individual level.

In the present study, we could adjust for only a few clinical factors, i.e. HLA matching, cold ischaemic time, age and sex. Other factors (e.g. immunosuppressive drugs and dosage, viral hepatitis infection, duration of dialysis, etc.) that were previously associated with poor outcomes after renal transplantation were not available in the datasets obtained. Since the IPD meta-analysis is a retrospective collaboration, it is difficult to get clinical variables that have been assessed and measured using similar methods across all studies; standardization is best done as a prospective collaboration. Another limitation was that there was insufficient information to separate the outcomes more discretely, i.e. separating CAN from CR or acute rejection from chronic rejection. For example, in post hoc analyses, only three studies had sufficient data to pool for acute rejection and the estimates obtained were well within the confidence intervals for the combined outcomes. For instance, for the TGF-β1 10 T/C polymorphism, the gene effect of TC versus TT on acute rejection was 1.45 (95% CI: 0.81–2.59) compared to the effect on the combined outcome of 1.51 (95% CI: 1.03–2.22). It is possible that genetic associations might be clearer if the outcome could be more clearly specified. Although the pathologic mechanisms may be different between outcomes, the lack of heterogeneity suggests that this may not be a major influence in the current results.

In summary, this IPD meta-analysis has demonstrated that a 2-SNP-haplotype of TGF-β1 and a 3-SNP-haplotype of IL-10 show a possible association with poor outcomes in renal transplantation, but this needs to be confirmed in larger studies.

Conflict of interest statement . None declared.

References

1
Dmitrienko
S
Hoar
DI
Balshaw
R
, et al.  . 
Immune response gene polymorphisms in renal transplant recipients
Transplantation
 , 
2005
, vol. 
80
 (pg. 
1773
-
1782
)
2
Hutchings
A
Guay-Woodford
L
Thomas
JM
, et al.  . 
Association of cytokine single nucleotide polymorphisms with B7 costimulatory molecules in kidney allograft recipients
Pediatr Transplant
 , 
2002
, vol. 
6
 (pg. 
69
-
77
)
3
Loucaidou
M
Stitchbury
J
Lee
J
, et al.  . 
Cytokine polymorphisms do not influence acute rejection in renal transplantation under tacrolimus-based immunosuppression
Transplant Proc
 , 
2005
, vol. 
37
 (pg. 
1760
-
1761
)
4
Marshall
SE
McLaren
AJ
Haldar
NA
, et al.  . 
The impact of recipient cytokine genotype on acute rejection after renal transplantation
Transplantation
 , 
2000
, vol. 
70
 (pg. 
1485
-
1491
)
5
McDaniel
DO
Barber
WH
Nguyan
C
, et al.  . 
Combined analysis of cytokine genotype polymorphism and the level of expression with allograft function in African-American renal transplant patients
Transpl Immunol
 , 
2003
, vol. 
11
 (pg. 
107
-
119
)
6
Melk
A
Henne
T
Kollmar
T
, et al.  . 
Cytokine single nucleotide polymorphisms and intrarenal gene expression in chronic allograft nephropathy in children
Kidney Int
 , 
2003
, vol. 
64
 (pg. 
314
-
320
)
7
Morgun
A
Shulzhenko
N
Rampim
GF
, et al.  . 
Interleukin-2 gene polymorphism is associated with renal but not cardiac transplant outcome
Transplant Proc
 , 
2003
, vol. 
35
 (pg. 
1344
-
1345
)
8
Tian
Y
Ma
W
Zhang
Y
Influence of cytokine gene polymorphism on renal transplantation
Zhonghua Wai Ke Za Zhi
 , 
2002
, vol. 
40
 (pg. 
256
-
258
)
9
Tinckam
K
Rush
D
Hutchinson
I
, et al.  . 
The relative importance of cytokine gene polymorphisms in the development of early and late acute rejection and six-month renal allograft pathology
Transplantation
 , 
2005
, vol. 
79
 (pg. 
836
-
841
)
10
Weimer
R
Mytilineos
J
Feustel
A
, et al.  . 
Mycophenolate mofetil-based immunosuppression and cytokine genotypes: effects on monokine secretion and antigen presentation in long-term renal transplant recipients
Transplantation
 , 
2003
, vol. 
75
 (pg. 
2090
-
2099
)
11
Inigo
P
Lario
S
Campistol
JM
, et al.  . 
Role of transforming growth factor beta-1 gene polymorphisms in the development of chronic allograft nephropathy in renal transplant recipients
Nefrologia
 , 
2003
, vol. 
23
 (pg. 
312
-
320
)
12
Asderakis
A
Sankaran
D
Dyer
P
, et al.  . 
Association of polymorphisms in the human interferon-gamma and interleukin-10 gene with acute and chronic kidney transplant outcome: the cytokine effect on transplantation
Transplantation
 , 
2001
, vol. 
71
 (pg. 
674
-
677
)
13
Hernandez
JL
Weir
BS
A disequilibrium coefficient approach to Hardy-Weinberg testing
Biometrics
 , 
1989
, vol. 
45
 (pg. 
53
-
70
)
14
Chen
JJ
Thomson
G
The variance for the disequilibrium coefficient in the individual Hardy-Weinberg test
Biometrics
 , 
1999
, vol. 
55
 (pg. 
1269
-
1272
)
15
Nam
J
Simple test for the Hardy-Weinberg law for HLA data with no observed double blanks
Biometrics
 , 
1995
, vol. 
51
 (pg. 
354
-
357
)
16
StataCorp
Stata Statistical Software: Release 9.0.
 , 
2005
Collage Station, TX
Stata Corporation
17
McCaskie
PA
Carter
KW
Palmer
LJ
SimHap: a comprehensive modelling framework and a simulation-based approach to haplotypic analysis of population-based data
2004
 
18
Chow
KM
Szeto
CC
Poon
P
, et al.  . 
Transforming growth factor-beta1 gene polymorphism in renal transplant recipients
Ren Fail
 , 
2005
, vol. 
27
 (pg. 
671
-
675
)
19
Thakkinstian
A
D’Este
C
Attia
J
Haplotype analysis of VDR gene polymorphisms: a meta-analysis
Osteoporos Int
 , 
2004
, vol. 
15
 (pg. 
729
-
734
)
20
Tagore
A
Gonsalkorale
WM
Pravica
V
, et al.  . 
Interleukin-10 (IL-10) genotypes in inflammatory bowel disease
Tissue Antigens
 , 
1999
, vol. 
54
 (pg. 
386
-
390
)
21
Turner
DM
Williams
DM
Sankaran
D
, et al.  . 
An investigation of polymorphism in the interleukin-10 gene promoter
Eur J Immunogenet
 , 
1997
, vol. 
24
 (pg. 
1
-
8
)
22
Warle
MC
Farhan
A
Metselaar
HJ
, et al.  . 
Are cytokine gene polymorphisms related to in vitro cytokine production profiles?
Liver Transpl
 , 
2003
, vol. 
9
 (pg. 
170
-
181
)
23
Hutchinson
IV
The role of transforming growth factor-beta in transplant rejection
Transplant Proc
 , 
1999
, vol. 
31
 (pg. 
9S
-
13S
)
24
Cerwenka
A
Bevec
D
Majdic
O
, et al.  . 
TGF-beta 1 is a potent inducer of human effector T cells
J Immunol
 , 
1994
, vol. 
153
 (pg. 
4367
-
4377
)
25
Swain
SL
Huston
G
Tonkonogy
S
, et al.  . 
Transforming growth factor-beta and IL-4 cause helper T cell precursors to develop into distinct effector helper cells that differ in lymphokine secretion pattern and cell surface phenotype
J Immunol
 , 
1991
, vol. 
147
 (pg. 
2991
-
3000
)
26
Fox
FE
Ford
HC
Douglas
R
, et al.  . 
Evidence that TGF-beta can inhibit human T-lymphocyte proliferation through paracrine and autocrine mechanisms
Cell Immunol
 , 
1993
, vol. 
150
 (pg. 
45
-
58
)
27
Morris
DR
Kuepfer
CA
Ellingsworth
LR
, et al.  . 
Transforming growth factor-beta blocks proliferation but not early mitogenic signaling events in T-lymphocytes
Exp Cell Res
 , 
1989
, vol. 
185
 (pg. 
529
-
534
)
28
Suthanthiran
M
Human renal allograft rejection: molecular characterization
Nephrol Dial Transplant
 , 
1998
, vol. 
13
 
Suppl 1
(pg. 
21
-
24
)
29
Cuhaci
B
Kumar
MS
Bloom
RD
, et al.  . 
Transforming growth factor-beta levels in human allograft chronic fibrosis correlate with rate of decline in renal function
Transplantation
 , 
1999
, vol. 
68
 (pg. 
785
-
790
)
30
Pankewycz
OG
Miao
L
Isaacs
R
, et al.  . 
Increased renal tubular expression of transforming growth factor beta in human allografts correlates with cyclosporine toxicity
Kidney Int
 , 
1996
, vol. 
50
 (pg. 
1634
-
1640
)
31
Mytilineos
J
Laux
G
Opelz
G
Relevance of IL10, TGFbeta1, TNFalpha, and IL4Ralpha gene polymorphisms in kidney transplantation: a collaborative transplant study report
Am J Transplant
 , 
2004
, vol. 
4
 (pg. 
1684
-
1690
)
32
Oxman
AD
Clarke
MJ
Stewart
LA
From science to practice. Meta-analyses using individual patient data are needed
JAMA
 , 
1995
, vol. 
274
 (pg. 
845
-
846
)
33
Stewart
LA
Parmar
MK
Meta-analysis of the literature or of individual patient data: is there a difference?
Lancet
 , 
1993
, vol. 
341
 (pg. 
418
-
422
)
34
Smith
CT
Williamson
PR
Marson
AG
An overview of methods and empirical comparison of aggregate data and individual patient data results for investigating heterogeneity in meta-analysis of time-to-event outcomes
J Eval Clin Pract
 , 
2005
, vol. 
11
 (pg. 
468
-
478
)
35
Simmonds
MC
Higgins
JP
Covariate heterogeneity in meta-analysis: criteria for deciding between meta-regression and individual patient data
Stat Med
 , 
2007
, vol. 
26
 (pg. 
2982
-
2999
)
36
Thompson
SG
Higgins
JP
How should meta-regression analyses be undertaken and interpreted?
Stat Med
 , 
2002
, vol. 
21
 (pg. 
1559
-
1573
)

Comments

0 Comments