PTPN22 Gene Polymorphism (C1858T) Is Associated with Susceptibility to Type 1 Diabetes: A Meta-Analysis of 19,495 Cases and 25,341 Controls


Corresponding authors: Li-Min Lun, MD, PhD and Chao Xuan, PhD, The Affiliated Hospital of Medical College, Qingdao University, No. 59, Haier Road, Qingdao, 266101, China. Tel: +86–532-82913083; Fax: +86–532-82913083; E-mail: and or


The protein tyrosine phosphatase N22 (PTPN22) gene C1858T polymorphism has been reported to be associated with susceptibility to type 1 diabetes (T1D) in relatively small sample sizes. This study aimed at investigating the pooled association by carrying out a meta-analysis on the published studies. The Medline, EBSCO, and BIOSIS databases were searched to identify eligible studies published in English before June 2012. The association was assessed by odds ratio (OR) with 95% confidence intervals (CI). The presence of heterogeneity and publication bias was explored by using meta-regression analysis and Begg's test, respectively. A total of 28 studies were involved in this meta-analysis. Across all populations, significant associations were found between the PTPN22 C1858T polymorphism and susceptibility to T1D under genotypic (TT vs. CC [OR = 3.656, 95% CI: 3.139–4.257], CT vs. CC [OR = 1.968, 95% CI: 1.683–2.300]), recessive (OR = 3.147, 95% CI: 2.704–3.663), and dominant models (OR = 1.957, 95% CI: 1.817–2.108). In ethnicity- and sex-stratified analyses, similar associations were found among Caucasians and within Caucasian male and female strata. The meta-analysis results suggest that the PTPN22 C1858T polymorphism was associated with susceptibility to T1D among the Caucasian population, and males who carried the -1858T allele were more susceptible to T1D than females.


Type 1 diabetes (T1D) is a common chronic disease, which is characterised by selective autoimmune destruction of the pancreatic β-cells and subsequent dependence on exogenous insulin. T1D is usually diagnosed in children and young adults (Notkins & Lernmark, 2001). Epidemiologic studies reveal a significant environmental contribution to the pathogenesis of T1D (Knip et al., 2005; Borchers et al., 2010). Familial aggregation and twin studies indicate the presence of genetic factors for susceptibility to this condition (Barnett et al., 1981; Pociot & McDermott, 2002). The disease is highly heritable and the heritability is estimated to be 72–88%. Siblings of patients with T1D are at 15 times greater risk to get this disease than the general population and concordance in monozygotic twins is as much as 50% (Grant & Hakonarson, 2009).

The importance of genetic factors in susceptibility to T1D is also supported by recent data from linkage disequilibrium mapping studies and genome-wide association studies (GWAS) (Hakonarson et al., 2007; Hakonarson et al., 2008; Grant et al., 2009). As expected, several regions potentially associated with susceptibility to T1D were identified, such as HLA (6p21.32), INS (11p15.5), IFIH1 (2q24.2), CTLA4 (2p33.2), and PTPN22 (1p13.2) (Cooper et al., 2008; Hakonarson & Grant, 2009).

The protein tyrosine phosphatase N22 (PTPN22) gene maps to chromosome 1p13.3–p13.1 and encodes a lymphoid-specific phosphatase (LYP) known to be important in negative control of T-cell activation and in T-cell development (Cohen et al., 1999). In humans, the nonsynonymous change C1858T of the PTPN22 gene (rs2476601) results in an amino acid substitution at codon 620 from arginine (Arg) to tryptophan (Trp). This substitution is located in the proline-rich sequence of LYP and likely affects its catalytic activity compared with the wild type (Vang et al., 2005). Experimental evidence suggests that the LYP 620Trp variant causes a gain of physiological function of the wild-type phosphatase; that is, 620Trp has more negative regulatory activity than the more common wild-type 620Arg allele (Rieck et al., 2007; Vang et al., 2007).

Since Bottini and coworkers (Bottini et al., 2004) first noted an association between PTPN22 Gene Polymorphism (C1858T) and susceptibility to T1D, other studies have been undertaken to replicate this work. A number of previous studies in which meta-analyses have been performed, have reported significant results (Lee et al., 2007; Peng et al., 2012; Tang et al., 2012). However, the previous meta-analyses were limited in having relatively small sample sizes, and lacking sex-stratified analyses and secondary statistical analyses. Therefore, we re-evaluated the evidence for the association between the PTPN22 gene polymorphism (C1858T) and susceptibility to T1D in a noticeably enlarged sample size by using the present updated meta-analysis. In addition, sex-stratified analyses were also performed.

Materials and Methods

Search Strategy

All studies reporting the association between the PTPN22 polymorphism and susceptibility to T1D published in English before June 2012 were identified by comprehensive computer-based searches of Medline, EBSCO, and BIOSIS. The following keywords were used for searching: (‘‘diabetes’’ or ‘‘T1D’’) and (“polymorphism*” or “variant*”) and (‘‘protein tyrosine phosphatase nonreceptor’’ OR “PTPN22” or “lymphoid protein tyrosine phosphatase” or “LYP”). The most complete and recent results were used when there were multiple publications from the same study group. The references of reviews and retrieved articles were also searched simultaneously to find additional eligible studies.

Inclusion Criteria

Two investigators reviewed all identified studies independently to determine whether an individual study was eligible for inclusion. The selection criteria for studies to be considered for this meta-analysis were as follows: (1) PTPN22 polymorphisms in T1D (including GWAS data); (2) case-control or case-cohort study; (3) proper T1D diagnosis criteria (WHO criteria: blood glucose level, proneness to ketosis, and absolute insulin dependency); (4) original data; and (5) not animal studies. The study would be excluded if the information could not be obtained.

Data Extraction

Two investigators extracted the data independently, and the result was reviewed by a third investigator. The following information was extracted from a study: first author, year of publication, study population (country, ethnicity), the number of patients and controls for a study, and genotype information. If any data essential to analysis were not available from a study, best efforts were made to contact the authors to fill in the missing data.

Statistical Analysis

Allele frequencies at the PTPN22 (C1858T) polymorphisms from each respective study were determined by the allele counting method (Xuan et al., 2011a). Genotype distributions of controls were used to estimate the frequencies of the putative risk allele (-1858T) using the inverse variance method (Xuan et al., 2011b; Xuan et al., 2012). The deviation from Hardy-Weinberg equilibrium (HWE) for distribution of the allele frequencies was analysed by Fisher's exact test in control groups. We examined the contrast of the TT versus CC, TT versus CT and also examined the recessive genetic model (TT vs. CC + CT) and the dominant genetic model (CT + TT vs. CC). The associations between PTPN22 (C1858T) polymorphisms and T1D susceptibility were estimated by OR and its 95% CI. The significance of the pooled OR was determined by the Z-test; P < 0.05 was considered statistically significant. In order to evaluate the ethnicity- and gender-specific effects, stratified analyses were performed.

Heterogeneity across the eligible studies was tested using the Q-test, and it was considered statistically significant when P < 0.1 (Higgins & Thompson, 2002; Bowden et al., 2011).

Heterogeneity was also quantified with the I2 metric (I2 = (Qdf)/Q × 100%). The I2 < 25%, no heterogeneity; I2 = 25–50%, moderate heterogeneity; I2 = 50–75%, large heterogeneity, I2 > 75%, extreme heterogeneity). When the effects were assumed to be homogenous (P > 0.1, I2 < 50%), the fixed-effects model was used; otherwise, the random-effects model was more appropriate. If heterogeneity was detected, meta-regression analysis was used to try to find the source of the heterogeneity. Sensitivity analysis was performed to evaluate the stability of the crude results, which were pooled with the random-effects model.

Cumulative meta-analysis was performed for the PTPN22 polymorphism to evaluate the trend of pooled OR for the genetic contrast. It also provided a framework for updating a genetic effect from all studies as well as a measure of how much the genetic effect changes as evidence accumulates. The cumulative meta-analysis studies were chronologically ordered by publication year, and the pooled OR was obtained at the end of each year.

Begg's test was used to measure publication bias, which was shown as a funnel plot (Begg & Mazumdar, 1994). The P < 0.05 was considered representative of statistically significant publication bias. All analyses were performed using STATA software, version 10.0 (Stata Corporation, College Station, TX, USA) and R statistical software, version 2.15.2 (


Studies Included in the Meta-Analysis

A total of 136 abstracts were retrieved through the Medline, EBSCO and BIOSIS databases that met the inclusion criteria. Two reviewers then selected the relevant studies independently. Forty-four relevant studies that described the association between the PTPN22 C1858T polymorphism and T1D were identified. However, after reading the full articles and contacting the authors, we excluded three meta-analysis studies (Lee et al., 2007; Peng et al., 2012; Tang et al., 2012), one study that had only case data (Maier et al., 2006), one duplicate study (Smyth, Cooper et al., 2008), four family-based research studies (Ladner et al., 2005; Qu et al., 2005; Onengut-Gumuscu et al., 2006; Steck, Baschal et al., 2009), and seven studies (Aarnisalo et al., 2008; Bjornvold et al., 2008; Zoledziewska et al., 2008; Lempainen et al., 2009; Steck, Zhang et al., 2009; Maziarz et al., 2010; Plagnol et al., 2011) in which information could not be obtained after authors were contacted. Finally, 28 studies (30 cohorts) (Bottini et al., 2004; Smyth et al., 2004; Gomez et al., 2005; Kahles et al., 2005; Zheng & She, 2005; Zhernakova et al., 2005; Fedetz et al., 2006; Hermann et al., 2006; Steck et al., 2006; Chelala et al., 2007; Cinek et al., 2007; Nielsen et al., 2007; Santiago et al., 2007; Baniasadi & Das, 2008; Cervin et al., 2008; Douroudis et al., 2008; Petrone et al., 2008; Saccucci et al., 2008; Smyth, Plagnol et al., 2008; Dultz et al., 2009; Korolija et al., 2009; Chagastelles et al., 2010; Fichna et al., 2010; Klinker et al., 2010; Kordonouri et al., 2010; Stene et al., 2010; Zhebrun et al., 2011; Kisand & Uibo, 2012) that met the inclusion criteria, consisting of 19,495 cases and 25,341 controls, were considered in the meta-analysis. Figure 1 shows the process of study selection and exclusion, with specification of reasons. In this stratified analysis, except for one South Asian study (Baniasadi & Das, 2008), 27 different studies (29 cohorts), consisting of 19,366 cases and 25,232 controls with specific focus on the Caucasian population, were included in ethnic groups. In another stratified analysis, seven studies (Kahles et al., 2005; Nielsen et al., 2007; Santiago et al., 2007; Cervin et al., 2008; Saccucci et al., 2008; Smyth, Plagnol et al., 2008; Klinker et al., 2010) on the Caucasian population were included in gender-specific groups. The main characteristics of the included studies are listed in Table 1.

Table 1. The detailed characteristics of all eligible studies
    Sample size-1858 T allele (%)  
StudyYearCountry (ethnicity)GenderPatientsControlsPatientsControlsHWE PCharacteristics
  1. C, Caucasians; A, South Asians; T1D, type 1 diabetes; HWE, Hardy-Weinberg equilibrium; NA, not available.

  2. *Data were, respectively, provided by authors of Nunzio Bottini, Nigel Ovington, Jasmina Kravic, and Matthew W. Klinker (see Acknowledgements).

  3. †A GWAS study.

  4. #Deviated from HWE.

Bottini et al.2004USA (C)All29439519.0511.650.63The same ethnic background.
  Italy (C) 17421440.892.101.00Mean age of patients was 7.4 ± 0.3 years, and 60.2% were males. This control sample consisted of healthy individuals from the same Italian population.
Smyth et al.2004UK (C)All1573171817.0410.451.00The case subjects <16 years of age, have a mean age at onset of T1D at 7.5 ± 4 years. The regional distribution of case- and population-based control subjects is matched.
Zhernakova et al.2005Netherlands (C)All33452817.968.711.00The age at onset of the T1D patients was on average 8.7 years (range 1–17).
Zheng & She2005USA (C)All396117814.528.571.00The sporadic case and control subjects were obtained from north central Florida.
Gomez et al.2005Spain (C)All1103087.734.381.00There were 56 girls and 54 boys patients; the mean age was 8.8 ± 6.3 years. Controls were matched to patient groups by gender, ethnicity, and socioeconomic status, the mean age was 47 ± 9 years, and 90% were women.
Kahles et al.2005Germany (C)All22023919.3211.300.75The mean age at onset of patients was 10 years (range 1–44 years). The average ages TID patients and controls were 27.9 ± 12.9 (range 6–69) and 49.4 ± 16.6 (range 23–83) and the male: female was 1:1.1 and 1:0.8.
Steck et al.2006USA (C)All69051516.169.030.79The mean age at onset of patients was 11.2 years (range: 0.3–54).
Hermann et al.2006Finland (C)All54653823.8113.940.21The mean age at onset of patients was 8.2 ± 4.1 years. The control group comprised healthy infants (51.2% boys).
Fedetz et al.2006Ukraine (C)All29624221.1114.050.19The age range was from 16 to 65 (36.0 ± 11.6) for patients and from 18 to 60 years (35.4 ± 11.5) for the controls. The male/female ratio was 1.1 in patients and 2.0 in the control group.
Cinek et al.2007Czech (C)All37240020.8310.250.59The patients were 188 males, 184 females who developed T1D <15 years (range 4.1–10.9). Hospital-based health controls.
  Azeri (C)All1602712.810.371.00The patients were 81 males, 79 females who developed T1D <19 years (6–12). The controls were 79 males and 192 females, median age 21 years (range 19–21).
Nielsen et al.2007Denmark (C)All25335416.019.180.06The mean age at onset of patients was 7.3 years (range 0.6–16.6). The controls came from unrelated healthy blood donors.
Chelala et al.2007France (C)All88544215.889.620.27NA
Santiago et al.2007Spain (C)All31655410.926.680.73The patients were 157 males, 159 females. The age at onset for patients ranged from 1 to 55 years. The age of controls ranging from 18–60 years.
Baniasadi & Das2008India (SA)All1291093.492.751.00The mean age at onset of the disease was 15.4 ± 6.6 years, and sex (male: female) ratio was 78:51. Age, sex and ethnicity matched normal controls.
Saccucci et al.*2008Italy (C)All2982506.713.20.22The Rome cases had a mean age at sample collection of 15.6 ± 0.37 years; 46.3% were males and 53.7% were females. The age of controls was 36.8 ± 0.64 years.
Douroudis et al.2008Estonia (C)All17023025.0013.910.41The patients were 86 female, 84male; the mean age was 29.53 ± 17.86 years. The younger population-based controls (43.3 ± 11.2 years); the older hospital-based controls (49.8 ± 17.9 years).
Smyth et al.*2008bUK (C)All89841093017.299.560.66The mean age at onset of the disease was 7.5 years (range 0.5–16). The controls were obtained from the British 1958 Birth Cohort (
Petrone et al.2008Italy (C)All55854510.304.680.33The patients were 296 males, 262 females, who developed T1D was 14.9 ± 7.8 years. The controls were 278 males and 267 females, the mean age was 30 ± 5 years.
Cervin et al.*2008Sweden (C)All145367418.1810.080.38The mean age at onset of the disease was 17.9 ± 9.2 years, and the mean age of controls was 70.1 ± 2.9 years.
Korolija et al.2009Croatia (C)All10219328.9211.660.48The mean age at onset of the disease was 23.5 years (range 17–29.5). Median age for controls was 70 years (range 58–78).
Dultz et al.2009Germany (C)All7010010.718.000.11Patients and controls resided in the same area and were unrelated, excluding genetic admixture between the groups.
Stene et al.2010Norway (C)All33998516.2210.610.18The cases of T1D diagnosed <15 years. Age, sex and ethnicity matched normal controls.
Chagastelles et al.2010Brazil (C)All21124114.695.191.00The mean age of patients was 15.13 ± 5.72 years, and the mean age of controls was 43.50 ± 7.95 years.
Klinker et al.*2010Finland (C)All579153521.1615.150.77The mean age at diagnosis of the disease was 26.00 ± 7.16 years, and the mean age of controls was 62.9 ± 7.5 years.
Kordonouri et al.2010Germany (C)All24320915.2310.531.00The mean age at diagnosis of the disease was 8.6 years (range 0.1-17.8).
Fichna et al.2010Poland (C)All21523618.6011.650.53The mean age of disease onset was 8.3 ± 4.3 years, and the mean age of controls was 59.6 ± 11.5 years.
Zhebrun et al.2011Russia (C)All15020020.6717.500.05The diagnosis was established < 20 years.
Kisand & Uibo2012Estonia (C)All15422925.6513.9713.97The mean age of patients was 22.0 ± 14.3 years, and the mean age of controls was 45.9 ± 14.5 years.
Figure 1.

Flow chart of selection of studies and specific reasons for exclusion from the meta-analysis.

Pooled Prevalence of PTPN22–1858T in the Controls

There was slight heterogeneity among the 29 cohorts of the Caucasian population studies (χ2 = 38.68 df = 28, P = 0.086; I2 = 27.6%). The pooled PTPN22–1858T allele frequency using the random-effects model was 9.80% (range: 3.20–17.50%). The pooled PTPN22–1858T allele frequency among the male and female Caucasian population was 6.99% (range: 3.47–14.22%) and 8.95% (range: 0.37–15.86%).

The pooled PTPN22–1858T allele frequency was 2.75% in the South Asia population. Genotype distributions in the controls of all studies were in agreement with HWE, except for a study in the female subgroup (Saccucci et al., 2008).

Meta-Analysis Results

We investigated the association between the PTPN22 C1858T polymorphism and susceptibility to T1D for each study. Overall, when all the eligible studies were pooled with fixed- or random-effects models, significant associations were observed in all genetic models for TT versus CC (OR = 3.656, 95% CI: 3.139–4.257; P = 1.83E-62), CT versus CC (OR = 1.968, 95% CI: 1.683–2.300; P = 1.90E-17), recessive model (OR = 3.147, 95% CI: 2.704–3.663; P = 1.26E-49; Fig. 2), and dominant model (OR = 1.957, 95% CI: 1.817–2.108; P = 2.94E-70). The Z-test indicated that the pooled ORs were statistically significant.

Figure 2.

Pooled OR (recessive model) and 95% CI of individual studies and pooled data for the association between polymorphism C1858T and type 1 diabetes (T1D) in the overall population.

In the stratified analysis by ethnicity, significant associations were found in the Caucasian population when all studies were pooled with a fixed-effects model for TT versus CC (OR = 3.659, 95% CI: 3.141–4.261; P = 1.83E-62), CT versus CC (OR = 1.967, 95% CI: 1.682–2.299; P = 2.25E-17), recessive model (OR = 3.149, 95% CI: 2.705–3.665; P = 1.46E-49) or the random-effects model for dominant inheritance model (OR = 1.962, 95% CI: 1.821–2.113; P = 2.46E-70). The Z-test indicated that the pooled ORs were statistically significant. In addition, the same associations were found in the male and female Caucasian populations. The main results of the meta-analysis are shown in Table 2.

Table 2. Main results of the pooled ORs in meta-analysis
     Test of
  Sample sizeTest of heterogeneityTest of associationbias
SubgroupmodelPatientsControlsQPI2 (%)OR95% CIZPzP
  1. *Pooled with random-effects model.

OverallTT vs. CC19,49525,34120.070.8910.03.6563.139–4.25716.681.83E-620.860.392
 TT vs. CT  22.260.8090.01.9681.683–2.3008.501.90E-170.390.695
 Recessive model  19.980.8930.03.1472.704–3.66314.811.26E-490.930.354
 Dominant model  45.840.02436.71.957*1.817–2.10817.722.94E-700.610.544
CaucasianTT vs. CC19,36625,23220.020.8640.03.6593.141–4.26116.681.83E-621.110.268
 TT vs. CT  22.240.7700.01.9671.682–2.2998.482.25E-170.470.639
 Recessive model  19.970.8660.03.1492.705–3.66514.801.46E-491.180.237
 Dominant model  44.830.02337.51.962*1.821–2.11317.732.46E-700.990.320
Male CaucasianTT vs. CC582875313.930.6860.04.3183.228–5.7759.866.20E-230.600.548
 TT vs. CT  4.200.6490.02.3941.777–3.2265.749.47E-090.001.000
 Recessive model  3.820.7010.03.7722.823–5.0408.982.71E-190.300.764
 Dominant model  5.970.4260.01.9601.806–2.12716.121.85E-580.600.548
Female CaucasianTT vs. CC549675868.800.18631.83.5372.704–4.6259.232.71E-200.300.764
 TT vs. CT  9.940.32613.61.8591.412–2.4484.429.87E-060.001.000
 Recessive model  7.970.24024.83.0352.323–3.9658.144.44E-160.300.764
 Dominant model  10.970.08945.51.983*1.655–2.3777.421.17E-130.600.548

Meta-Regression for Heterogeneity and Sensitivity Analysis

We found moderate heterogeneity among the Caucasian population when the OR was pooled in the dominant model. Meta-regression analysis was used to try to find the source of the heterogeneity in general variables. However, the year of publication (P = 0.449), age of onset in T1D patients (young or not; P = 0.487), genotype methods (PCR-RFLP, Taq-man, Sequencing, Gene-chip, or Mass spectrometry; P = 0.543), and source of controls (population- or hospital-based; P = 0.880) were not the source of heterogeneity. The heterogeneity may come from the statistical method or unknown variables, and the random-effects method for pooled OR was appropriate under this condition.

We conducted sensitivity analysis to evaluate the stability of the crude results, which were pooled with the random-effects model among the Caucasian population. When any single study was deleted, the corresponding pooled ORs were not substantially altered (Fig. 3), suggesting that the results of this meta-analysis were stable. In addition, we removed the study by Saccucci and coworkers (Saccucci et al., 2008) in the stratified analysis due to the genotype distribution in the control groups of the study slightly deviating from HWE. We found that the corresponding pooled ORs were not substantially altered in the female Caucasian population (data not shown).

Figure 3.

Sensitivity analysis for Caucasian population in dominant genetic model. This figure shows the influence of individual studies on the summary OR. The middle vertical axis indicates the overall OR and the two vertical axes indicate its 95% CI. Every hollow round indicates the pooled OR when the left study is omitted in this meta-analysis. The two ends of every broken line represent the 95% CI.

Cumulative Meta-Analysis

As information accumulated, the trend of association in estimated risk effect was shown in the cumulative meta-analysis for the recessive model (Fig. 4) in the association between PTPN22 C1858T polymorphism and susceptibility to T1D.

Figure 4.

Cumulative meta-analysis: the pooled OR with corresponding 95% CI for recessive model in the overall population at the end of each study; information steps are shown.

Publication Bias

Begg's test and a funnel plot were performed to assess the publication bias of the literature. The results suggested no evidence of publication bias (Table 2, Fig. 5).

Figure 5.

Funnel plot of C1858T polymorphism and susceptibility to T1D (recessive model) in the overall population (z = 0.93, P = 0.354).


The incidence of T1D is rapidly increasing in multiple regions of the world with a trend toward earlier onset. The risk of complex diseases such as T1D is generally thought to be influenced by multiple genetic and environmental factors. Epidemiological, clinical, and molecular studies have provided strong evidence for the role of genetics in determining susceptibility to T1D (Villano et al., 2009). The most important genes associated with T1D in the Caucasian population are the HLA class II genes (primarily HLA-DRB1, HLA-DQA1, and HLA-DQB1 genes), which encode the highly polymorphic antigen-presenting proteins (Hinks et al., 2005). This can explain about 50% of the familial clustering of T1D and has been confirmed subsequently in many other populations (Vyse & Todd, 1996).

Since GWAS technology has been used for the study of complex disease genetics, T1D has been at the forefront of a rapidly moving field (Grant & Hakonarson., 2009; Craddock et al., 2010). GWAS studies have identified more than 50 loci that are associated with T1D, and this has led to our appreciation of the integral role of pathways within immune systems in the pathogenesis of T1D (Aminkeng et al., 2010).

The potential association of the PTPN22 gene with susceptibility to T1D was recently identified outside the HLA region in the pre-GWAS era. The PTPN22 C1858T variant leads to the replacement of Arg by Trp in position 620 of the LYP protein (Vang et al., 2005). The 620Arg is a critical residue in LYP that binds the SH3 domain of Csk (a signalling factor), and the 620Trp protein fails to bind Csk. Both LYP and Csk are crucial gatekeepers of T-cell antigen receptor (TCR) signalling (Cloutier & Veillette, 1996). Studies have demonstrated that the threshold for TCR signalling is altered by the SNP (Begovich et al., 2004). Biochemical studies on primary human T1D patient T cells showed that the two alleles of LYP indeed behave differently in T-cell signalling (Bottini et al., 2006).

Many studies have reported the association between the PTPN22 C1858T polymorphism and susceptibility to T1D in relatively small sample sizes. In order to evaluate the association in a larger population, Lee and coworkers first performed a meta-analysis in 2006 (data updated to November 2005) and reported a significant result (Lee et al., 2007). However, only six comparisons and 7263 participants were include in their study. In the last 6 years, many case-control studies were reperformed to investigate the association, including a GWAS study. Recently, Peng and associates performed a meta-analysis to evaluate the association and reported significant results in European and American populations (17,770 Participants, data updated to September 2011) (Peng et al., 2012). Tang and colleagues re-evaluated this association with similar results (34,237 participants, data updated to October 2011) (Tang et al., 2012). However, the two previous meta-analyses have limitations in the sex-stratified analyses and secondary statistical analysis. Therefore, it is essential for us to perform a new updated meta-analysis to evaluate this association. In this study, we enlarged the sample size to 44,836 participants (19,495 cases and 25,341 controls), and performed sensitivity analysis, cumulative meta-analysis, and Begg's test to evaluate the stability of the results. Moreover, we performed pooled analyses stratifying on sex and ethnicity. Importantly, four authors provided unpublished data (3382 cases and 6339 controls) for this meta-analysis, which were not included in their original papers.

In our results, the frequency of the putative risk allele -1858T was 9.80% (95% CI: 9.28–10.36%) in the Caucasian population, and 2.75% (95% CI: 2.21–3.29%) in the South

Asia population. The prevalence of the -1858T allele varied from 6.99% to 8.95% in different gender groups, where the lowest was for male Caucasian and the highest was for female Caucasian. The meta-analysis results showed that association exists between the PTPN22 C1858T polymorphism and susceptibility to T1D for all genetic models among the overall and Caucasian populations (Table 2). Furthermore, the results indicated that the same associations were found among males and females in the stratified analysis, and males who carried the -1858T allele were more susceptible to T1D than were females. The results of cumulative meta-analysis revealed the temporal trend of pooled results. To assess publication bias, Begg's tests were performed. The results of these tests suggested that publication bias, which might influence the results of this meta-analysis, could be ruled out.

Some limitations of this meta-analysis should be discussed. First, this meta-analysis only focused on papers published in the English language and those which were reported in other languages might bias the present results. Second, the significance of heterogeneity among studies was observed when we pooled ORs in the dominant genetic model. We tried to find the source of the heterogeneity by using the method of meta-regression. However, the relationships between heterogeneity and some general variables were not detected. Although we used a random-effects model to pool the ORs, sensitivity analysis indicated that the results were stable.

Despite these limitations, our results support the PTPN22–1858T allele as a susceptibility factor for T1D among the Caucasian population, and males who carried the -1858T allele were more susceptible to T1D than females. The effect of the variants on the expression levels and the possible functional role of the variants in T1D should be addressed in further studies.


We thank Nunzio Bottini (Department of Internal Medicine, Division of Clinical Immunology, University of Rome, Rome, Italy and Institute for Genetic Medicine, Keck School of Medicine, University of Southern California, Los Angeles, CA, USA), Nigel Ovington (the Juvenile Diabetes Research Foundation/We​llcome Trust Diabetes and Inflammation Laboratory, University of Cambridge, UK), Jasmina Kravic (Department of Clinical Sciences-Diabetes & Endocrinology, Clinical Research Center, Skåne University Hospital, Malmö, Sweden), Matthew W. Klinker (Max McGee National Research Center for Juvenile Diabetes, Department of Pediatrics, Medical College of Wisconsin, Milwaukee, WI, USA) for providing data in their study. The work described in this paper was fully supported by grants from the National Natural Science Foundation of China (No. 81170148), International S & T Cooperation Program of China (No. 2009DFB30560) and the National Basic Research Program of China (No. 2010CB529500), and Tianjin Municipal Science and Technology Commission 09ZCZDSF04200 and 10JCYBJC26400, Tianjin Municipal Research Grant for Applied Basic and Frontier Technology, China, Binhai Key Platform for Creative Research Program (2012-BH110004), Tianjin Binhai New Area Health Bureau (2011BHKZ001 & 2012BWKZ008) and (2011BHKY002 & 2012BWKY024).