Abstract
Background
Inter-individual variation in response to anti-TNFα therapy may be explained by genetic variability in disease pathogenesis or mechanism of action. Recent genome wide association studies (GWAS) in IBD have increased our understanding of the genetic susceptibility to IBD.
Aim
Test associations of known IBD susceptibility loci and novel “pharmacogenetic” GWAS identified loci with primary non-response to anti-TNFα in pediatric IBD patients and develop a predictive model of primary non-response.
Methods
Primary non response was defined using the HBI for CD and partial Mayo score for UC. Genotyping was performed using the Illumina Infinium platform. Chi square analysis tested associations of phenotype and genotype with primary non-response. Genetic associations were identified by testing known IBD susceptibility loci and by performing a GWAS for primary non-response. Step-wise multiple logistic regression was performed to build predictive models.
Results
Non-response occurred in 22 of 94 subjects. Six known susceptibility loci were associated with primary non-response (p < 0.05). Only the 21q22.2/BRWDI loci remained significant in the predictive model. The most predictive model included 3 novel “pharmacogenetic” GWAS loci, the previously identified BRWD1, pANCA and a UC diagnosis (R2 =0.82 and AUC = 0.98%). The relative risk of non-response increased 15 fold when number of risk factors increased from 0–2 to ≥ 3.
Conclusion
The combination of phenotype and genotype is most predictive of primary non response to anti-TNFα in pediatric IBD. Defining predictors of response to anti-TNFα may allow the identification of patients who will not benefit from this class of therapy.
Introduction
Natural history observations in both early and later onset inflammatory bowel disease (IBD) have prompted the increasing use of anti-TNFα therapy for IBD patients (1,2). Both the REACH (3) pediatric study and the ACCENT 1 (4) study demonstrate that infliximab is effective for the induction and maintenance response and remission in a subset of CD patients. However, the clinical trial data for all anti-TNFα therapies among adult CD patients report that 40% of patients do not respond to the induction phase (primary non-responder) and that approximately 40% of those patients who do enter the maintenance phase of the trial lose response over time (4,5,6). The pediatric REACH trial reported that close to 90% of children responded to induction, suggesting a more robust acute response to anti-TNFα therapy in children as compared to adults with CD (1,2). This primary response outcome did not, however, require children to have weaned corticosteroids to meet response criteria. This would be a more clinically robust outcome definition given that the importance of steroid sparing in the induction and maintenance phase of these therapies. Moreover, approximately 40% of children, like their adult counterparts, who entered the maintenance phase lost response and were no longer in remission and off steroids at 12 months. Primary non-response appears to be less of an issue in children than adults based on the REACH trial, but more studies are needed to assess the true incidence of primary non-response in children in a non clinical trial setting. The adult UC trials (ACT 1 and ACT 2) reported similar response rates among adult UC patients receiving infliximab as the CD trials (7). Infliximab is being used off label in children with UC and the official clinical trial for indication is currently underway. These data do suggest that children respond better to induction therapy as compared to adult CD patients, but there are many differences in the patient population and outcome measures making a comparison across trials difficult and hard to interpret. One variable that may be an important predictor of response is duration of disease at initiation of therapy, which has been suggested based on post hoc analysis in both the certolizumab and adalimumab trials conducted in the adult CD population (5,6).
However inter-individual variability in therapeutic response may be better explained by genetic variability as it relates to disease pathogenesis and mechanism of action of this class of therapies. To date, limited candidate gene association studies with anti-TNFα have been reported (8–11). Other than NOD2 (12) and IBD5 (13), IBD susceptibility genes identified via genome wide linkage approach or Genome Wide Association Studies (GWAS) (14–18) have not been evaluated as predictors of response to anti-TNFα therapies. It would be logical as an initial step to test the known IBD susceptibility genes, under the hypothesis that anti-TNF therapy interrupts the pathophysiologic pathway these genes represent. Thus far, NOD2 was not found to be associated with therapeutic response to infliximab in these limited studies. It is conceivable that disease susceptibility genes do not influence the ultimate response to therapeutic targets given the multifactorial influences on disease and the relatively unknown functionality of these susceptibility genes. However, the GWAS approach, which identifies portions of the genome that contain genetic variants associated with specific phenotypes, may identity novel variants that contribute to therapeutic outcome i.e. discovery of genetic loci that are potentially responsible for the mechanism of altered drug response, in this case to anti-TNFα. There may also be important non genetic factors that influence or modify primary response to anti-TNFα. Among the serologic immune responses, pANCA (11,19) has been shown to be negatively associated with primary response in both CD and UC patients. PANCA is present in both CD and UC and defines a specific colitis phenotype suggesting a degree of overlap in the underlying pathway biology of these two disease subtypes (20). The results of recent genetic studies have also suggested this overlap (21). Other than duration of disease at initiation of therapy, there are likely important clinical and demographic variables that also influence therapeutic outcomes. To date, however, it remains unknown whether these are independent of genetic variability. Thus, the aims of this study were to test associations of known IBD susceptibility loci as well as novel loci identified by our pharmacogenetic GWAS with anti-TNFα response in pediatric IBD patients, and to develop a predictive model of anti-TNFα primary non-response using clinical phenotype, serologic and genetic variables.
Methods
Patient Population
A total of 94 pediatric CD and UC patients (age at diagnosis < 21 years) followed at Cedars-Sinai Medical Center (CSMC) by one clinician (MD) were enrolled in this study. All subjects must have received at least 2 doses of (weeks 0 and 2) of infliximab to be eligible. Infliximab was chosen as the first line anti-TNFα used in both CD and UC for children. This study was approved by the IRB.
Phenotyping
All data was collected by chart review and stored in a secured database. For the purpose of this study phenotype was defined as all variables that were not genetic.
Clinical Phenotype
These included demographic and clinical variables: age, gender, IBD subtype (CD vs. UC), disease duration, age at diagnosis, age at initiation of infliximab, immunomodulator history, steroid history, Harvey Bradshaw Index (HBI) activity scores, Partial Mayo scores and reason for infliximab discontinuation.
Immune Phenotype
Serum was collected on all patients and analyzed at CSMC. Serum immune responses: anti-saccharomyces cereviciae antibodies (ASCA IgG and IgA), perinuclear anti-nuclear cytoplasmic antibody (pANCA), anti-flagellin (anti-CBir1), anti-outer membrane porin C (anti-OmpC) and anti-Pseudomonas fluorescens-associated sequence I2 (anti-I2) were analyzed blinded to therapeutic responsiveness by ELISA as previously described (22).
Genotype
Genotyping was performed at Children's Hospital of Philadelphia (CHOP) using the Illumina Human550 platform (n=70) and the Medical Genetics Institute at CSMC using the Illumina Human610 platform for CD samples (n=17) and HumanCNV370 platform for UC samples (n=11) (23). First, genotype data were tested for association between previously reported IBD susceptibility SNPs and anti-TNFα response. Table 1 illustrates the 28 SNPs included in this part of the analysis and references the GWAS that first reported these associations with disease (14–18). Twenty-one (21) SNPs from the Barrett CD meta-analysis GWAS (14), 5 SNPS from the various UC GWAS (15–17) and 2 SNPs from the Kugathasan pediatric IBD GWAS (18) were analyzed. Second, the genome wide data were tested for association with anti-TNFα response (“pharmacogenetic” GWAS, see below).
Table 1.
Known Genetic Susceptibility Loci
| Chromosome/Loci of interest | GWAS Reference 1 = Adult CD (14) 2 = Adult UC (15–17) 3=Pediatric IBD (18) |
Chromosome/Loci of interest | GWAS Reference 1= Adult CD (14) 2 = Adult UC (15–17) 3=Pediatric IBD (18) |
|---|---|---|---|
| 1p13 / PTPN22 | 1 | 6q27 / CCR6 | 1 |
| 1p36 | 2 | 8q24.13 / TRIB1 | 1 |
| 1q23 / ITLN1 | 1 | 9p24 / JAK2 | 1 |
| 1q24 | 1 | 9q32 / TNFSF15 | 1 |
| 1q21.2 / ECM1 | 2 | 10q21 / ZNF365 | 1 |
| 1q32.1 / IL10 | 2 | 10q24 / NKX2–3 | 1 |
| 1p31 / IL23R | 1 | 12q15 | 2 |
| 2q37 / ATG16L1 | 1 | 13q14 / C13orf31 | 1 |
| 3p21 / MST1 | 1 | 16q12 / NOD2 | 1 |
| 5p13 / PTGER4 | 1 | 17q12 / CCL2 | 1 |
| 5q31 / IBD5 | 1 | 17q21/ORMDL3/STAT3 | 1 |
| 5q33 / IL12b | 1 | 20q13/ TNFRSR6B | 3 |
| 6p21/HLA-DQA1/TNFα, TNFβ | 2 | 21q22 / COSLG | 1 |
| 6p22 / CDKAL1 | 1 | 21q22.2 / BRWD1 | 3 |
Outcomes and Definitions
The primary outcome of this study was to identify genetic loci associated with primary non-response defined for the purpose of this study as:
CD: Failure to decrease HBI ≥ 3 points (24) or increase from baseline at week 10 or 4 weeks after their last infusion if they did not receive the 3rd induction dose
UC: Failure to decrease ≥ 2 points or increase from baseline in the sum of stool frequency and rectal bleeding subscores of the partial Mayo score (7) and no improvement in the physician's global assessment subscore at week 10 or 4 weeks after their last infusion if they did not receive the 3rd induction dose
Statistical Analysis
Univariate Analysis
Association between clinical and demographic variables and primary non-response
The Chi-square test was used to check the association of primary non-response with the following categorical variables: CD vs. UC, male vs. female, serum immune response positivity, percentage of immunomodulator use at the start of infliximab and primary non-response. The Student t test was used for associations of continuous variables; age of diagnosis, disease duration and duration of immunomodulator use at the start of infliximab.
Association between known IBD susceptibility loci and primary non-response
The Chi-square test was applied to test the association between each SNP (Table 1) and primary non-response. A dominant model based on the presence of the rare allele was assumed. Relative risk was calculated by comparing the risk of non-response in the patients with a specific genotype versus those without the genotype.
Pharmacogenetic genome wide association study (GWAS)
Principal components (PC) analysis (using Eigenstrat) was conducted to examine population stratification (25). All the subjects formed one cluster with no significant outliers. We did not need to correct for population stratification during the association analysis, as the first ten PC evaluated were not significantly associated with primary non-response.
For the purpose of quality control, SNPs with a minor allele frequency (MAF) <0.01, genotype failure rate >0.10, HWE P value <0.001 were excluded from the analysis. Allelic association between an individual SNP and primary non-response was carried out by chi-square test in PLINK (26). The first 10 SNPs with the most significant results were then retained for modeling. Following quality control, 301,742 SNPs were available in all data sets for analysis.
Multivariate Analysis
Predictive models of primary non-response
Models to predict non-response were built using step-wise multiple logistic regression, combining a) IBD susceptibility SNPs, b) the top 10 hits from the pharmacogenetic GWAS analysis, c) serology status and d) IBD subtypes. Since rs5975493 and rs7059861 are in high linkage disequilibrium, only rs7059861 was kept in the model. Exact logistic regression was used if the estimate from a regular regression model was not available. The significance level for a variable to enter and stay in the model was 10%. The likelihood based pseudo-R squared from the logistic regression model was used to measure the strength of association as well as the proportion of variance of the outcome accounted by the model's independent variables (27). Models were built at five different levels:
demographic variables, serology status and IBD subtype only
known IBD susceptibility SNPs only
model I and II combined
pharmacogenetic GWAS SNPs (dominant model assumed) only
model III and IV combined (final model). (known IBD susceptibility SNPs, pharmacogenetic GWAS top hits, serologic status and clinical variables)
Clinical Utility Measures
The area under the Receiver Operating Characteristic (ROC) curve (AUC) was used as a measure of the predictive performance of the final model. The risk score was calculated based on the final model by assigning each risk phenotype or genotype as 1 point. Sensitivity [# true positives /(#of true positives + #of false negatives)], specificity [#of true negatives/(#of true negatives + #of false positives)], accuracy [(sensitivity + specificity)/2] and positive likelihood ratio test (sensitivity/(1-specificity)) for primary non response were also calculated for the final model (28). All statistical analysis was conducted by SAS software v9.1 (SAS Institute; Cary, NC.).
Results
Patient population and phenotype associations
Of the 94 patients evaluated, 22 patients (23 %) met the criteria of primary non-response. Table 2 illustrates the key demographic data for both responders and non responders. A diagnosis of UC (p < 0.0001) and pANCA positivity (p=0.0001) were associated with primary non-response. Gender, mean age at diagnosis, disease duration at initiation of infliximab, percentage of immunomodulator use and duration of use at start of infliximab did not differ between the two groups.
Table 2.
Phenotype Associations with Therapeutic Outcomes to Anti-TNFα
| CLINICAL VARIABLE | NON-RESPONSE | RESPONSE | P Value |
|---|---|---|---|
| UC vs. CD | 14:8 | 6:66 | <0.0001 |
| Mean age at diagnosis (years) | 10.1 | 10.5 | 0.71 |
| Gender M:F | 8:14 | 40:32 | 0.12 |
| Mean Disease duration at start of IFX (months) | 26 | 25 | 0.96 |
| IMM use at start of IFX (%) | 79% | 92.8% | 0.1 |
| Mean Duration of IMM at start of IFX (months) | 12.7 | 15.3 | 0.57 |
| pANCA + | 76.2% | 29% | 0.0001 |
| ASCA+ | 0% | 46.9% | 0.0003 |
Genetic Associations: Univariate Analysis
Known IBD Susceptibility Loci
Of the 28 previously identified genetic loci (Table 1), 6 were found to be significantly associated with primary non-response. Figure 1 illustrates the frequency of primary non-response for the different genotypes of these 6 SNPs. Four of the 6 SNPs are from the CD meta-analysis (14), 1 from the UC GWAS (15–17) and 1 from pediatric IBD GWAS (18). For this analysis, the dominant model of the rare allele was assumed. The common allelic variant was associated with non-response in 4 of the 6 SNPs. Table 3 compares the reported IBD risk allele with the allele found to be associated with non-response in this study for all 6 SNPs. In only 2 of the SNPs was the disease risk allele the same as that found to be associated with non response to anti-TNFα.
Figure 1.
Genotype associations of known IBD susceptibility loci with primary non-response. The dominant rare allele model (i.e. presence of the rare allele) was assumed for the statistical analyses). The relative risk (RR) and frequency of non-response are shown for the genotypes for each locus that met p value significance in univariate analysis.
Table 3.
Allelic Variants
Pharmacogenetic GWAS
Table 4 lists the results of the chi square analyses for the pharmacogenetic GWAS. Only those SNPs with a p value < 10−4 are listed.
Table 4.
Pharmacogenetic GWAS: p < 0.0001
| Chromosome | SNP | Position* | Gene db129*, | Other loci**, | OR | P |
|---|---|---|---|---|---|---|
| 13 | rs1155848 | 78786477 | RBM26 | 35.73 | 5.35E-07 | |
| 5 | rs1592749 | 159904599 | 5.949 | 1.09E-06 | ||
| X | rs765132 | 133150270 | 347475 |402425 |644403 |
35 | 1.09E-06 | |
| 6 | rs4707930 | 72463040 | 6.603 | 5.55E-06 | ||
| 10 | rs7905482 | 81853944 | PLAC9 |389988 642506 |642521 |642538 |727879 |
4.875 | 1.08E-05 | |
| 20 | rs6100556 | 57701043 | PHACTR3 | PHACTR3 | 4.951 | 1.23E-05 |
| X | rs7059861 | 133191565 | 347475 |402425 |
8.333 | 1.62E-05 | |
| X | rs5975453 | 133212999 | 347475 402425 |
8.333 | 1.62E-05 | |
| 10 | rs4077511 | 5568172 | CALML3 CALML5 100132159 |
8.214 | 1.74E-05 | |
| 2 | rs975664 | 75222305 | TACR1 | TACR1 | 0.1667 | 1.75E-05 |
| 3 | rs4855535 | 69017124 | FAM19A4 | FAM19A4 | 5.619 | 1.77E-05 |
| 17 | rs4796606 | 36917613 | KRT13 |KRT15 |KRT32 |KRT35 |KRT38 |KRT37 |KRT36 |
11.73 | 1.98E-05 | |
| 8 | rs2943177 | 88121513 | CNBD1 | CNBD1 | 6.412 | 2.06E-05 |
| 8 | rs11991611 | 139676572 | COL22A1 | FAM135B COL22A1 |
15.29 | 2.12E-05 |
| 10 | rs3740543 | 129140065 | DOCK1 | DOCK1 | NPS | 4.932 | 2.18E-05 |
| 21 | rs2825673 | 19888587 | 100128057 | 6.438 | 2.28E-05 | |
| 1 | rs7521532 | 70351206 | LRRC7 | SFRS11 |LRRC7 |
7.143 | 2.54E-05 |
| 3 | rs4605505 | 141323576 | CLSTN2 | CLSTN2 | 7.143 | 2.54E-05 |
| 8 | rs7003556 | 5213929 | 7.143 | 2.54E-05 | ||
| 14 | rs1243519 | 94370110 | GSC | 6 | 3.06E-05 | |
| 5 | rs2044111 | 22897160 | CDH12 | 4.444 | 3.54E-05 | |
| 6 | rs2103867 | 47374655 | TNFRSF21 | TNFRSF21 | 12.07 | 3.6E-05 |
| 7 | rs17168564 | 14977349 | 8.286 | 3.68E-05 | ||
| 6 | rs10485363 | 13245751 | PHACTR1 | PHACTR1 | 5.182 | 3.87E-05 |
| 5 | rs7726515 | 129736249 | 5.187 | 4E-05 | ||
| 11 | rs835780 | 44816947 | TSPAN18 | 5.187 | 4E-05 | |
| 11 | rs835791 | 44823513 | TSPAN18 | 5.187 | 4E-05 | |
| 1 | rs3795727 | 154856074 | HAPLN2 | GPATCH4 |HAPLN2 BCAN IQGAP3 646129 |
4.5 | 4.1E-05 |
| 6 | rs6906890 | 13069206 | PHACTR1 | PHACTR1 | 5.899 | 4.31E-05 |
| 11 | rs7124825 | 44812803 | TSPAN18 | 0.1263 | 4.31E-05 | |
| 19 | rs302827 | 61102034 | NLRP13 | NLRP13 NLRP8 NLRP1 |
4.545 | 4.45E-05 |
| 12 | rs2723829 | 11819364 | ETV6 | ETV6 | 4.333 | 4.47E-05 |
| 2 | rs1372256 | 141070866 | LRP1B | LRP1B | 4.626 | 4.65E-05 |
| 4 | rs13138970 | 155494510 | DCHS2 | DCHS2 | 5.297 | 4.75E-05 |
| 20 | rs1205434 | 36302930 | KIAA1755 | BPI | TGM2 KIAA1755 |
0.1526 | 4.75E-05 |
| 13 | rs9556658 | 96476252 | OXGR1 | 6.111 | 4.92E-05 | |
| 2 | rs7588326 | 75202554 | TACR1 | TACR1 | 0.1826 | 5.07E-05 |
| 2 | rs3771823 | 75205456 | TACR1 | TACR1 | 0.1826 | 5.07E-05 |
| 20 | rs1555901 | 20762132 | 0.1826 | 5.07E-05 | ||
| X | rs4465121 | 133249071 | PHF6 | 7 | 5.26E-05 | |
| 6 | rs12527937 | 16652564 | ATXN1 | ATXN1 | 5.5 | 5.74E-05 |
| 7 | rs10269232 | 10442565 | 100128638 | 6.368 | 6.33E-05 | |
| 6 | rs3757105 | 73950398 | KCNQ5 | KCNQ5 | 24.2 | 6.51E-05 |
| 18 | rs1667216 | 27383929 | DSG2 | TTR | 4.952 | 6.75E-05 | |
| 12 | rs278917 | 39666755 | CNTN1 | CNTN1 | 4.172 | 6.93E-05 |
| 6 | rs9404502 | 104605814 | 100129694 | 5.409 | 7.01E-05 | |
| X | rs5977968 | 133151511 | 347475 402425 644403 |
7.333 | 7.03E-05 | |
| 1 | rs12567958 | 154861280 | HAPLN2 | GPATCH4 HAPLN2 BCAN IQGAP3 646129 |
4.167 | 7.03E-05 |
| 8 | rs1880473 | 88079222 | CNBD1 | CNBD1 | 5.085 | 7.09E-05 |
| 4 | rs7689941 | 182246739 | hCG_2025798 | 728081 | 4.244 | 7.3E-05 |
| 17 | rs12937472 | 15194290 | PMP22 TEKT3 |
8.509 | 7.5E-05 | |
| 5 | rs4301261 | 159902033 | 5.433 | 7.59E-05 | ||
| X | rs1264379 | 132893432 | GPC3 | GPC3 | 6.092 | 7.59E-05 |
| X | rs6529954 | 4194349 | 0.1781 | 7.69E-05 | ||
| 22 | rs3088103 | 25250048 | TPST2 | CRYBA4 |TPST2 TFIP11 HPS4 644380 653715 729905 100128401 HMGB1L10 |
4.308 | 7.77E-05 |
| 6 | rs4711716 | 42375201 | TRERF1 | TRERF1 387535 653802 |
8.727 | 7.82E-05 |
| 7 | rs10464448 | 141433796 | MGAM | MGAM | 0.134 | 8.32E-05 |
| X | rs12559781 | 4225952 | 7.973 | 8.49E-05 | ||
| 7 | rs2540678 | 36227249 | EEPD1 | EEPD1 | 4.6 | 8.54E-05 |
| 4 | rs7659755 | 182246126 | hCG_2025798 | 728081 | 4.461 | 8.62E-05 |
| 13 | rs770389 | 50043157 | LOC730194 | 730194 | 5.1 | 8.96E-05 |
| 21 | rs2825699 | 19934102 | 5.1 | 8.96E-05 | ||
| 12 | rs7309734 | 39564210 | CNTN1 | CNTN1 | 4.171 | 9.7E-05 |
| 2 | rs11903032 | 3840992 | 5.469 | 9.78E-05 | ||
| 8 | rs10808755 | 68679235 | CPA6 | CPA6 | 0.08701 | 9.87E-05 |
Defined using dbSNP Build 129
Loci include genes that overlap each SNP within 100kb of 5' end and 10kb of 3' end of each gene. Numbers refer to dbGene (NCBI).
Multivariate Analysis
Predictive models of non-response
Logistic multiple regression was employed to develop models of primary non-response. Five different models were developed. Model 1 examined the significance of pANCA and IBD subtype (UC vs. CD). Both pANCA (OR 5.4; p= 0.01) and the diagnosis of UC (OR 15.0; p = 0.0001) remained significant in model I, with an R squared (R2) of 0.48. Model II (R2 = 0.30) examined the 6 SNPs from the univariate analysis. Four (4) of the 6 known IBD susceptibility SNPs remained significant; rs2188962 (5q31) (OR 3.3; p =0.04), rs6908425 (6p22/CDKAL1) (OR 3.2; p = 0.04), rs2836878 (21q22/BRWD1) (OR 3.3; p = 0.05) and rs2395185 (6p21/HLA-DAQ1) (OR 4.6; p =0.01). Table 5 shows the results of Model III which included serology, IBD subtype and known susceptibility SNPs (combining the variables of models I and II). Only 3 SNPs survived the model when combined with these other independent variables. Model IV analyzed the top 10 SNPs from our pharmacogenetic GWAS and only 4 SNPs remained significant; rs975664 (TACR1) (OR 17.6, p = 0.0006), rs4855535 (FAM19A4)(OR 8.8, p = 0.006), rs4796606 (KRT32 KRT35 KRT36 KRT13)(OR 13.4, p= 0.01) and rs765132 (OR 30.1, p= 0.03). The R squared for this particular model was 0.67. The results of our final model (V) are shown in Table 6. Model V (R2 = 0.82, including all variables) examined the associations of pANCA, diagnosis of UC, the 6 known susceptibility SNPs and the top 10 SNPs from the pharmacogenetic GWAS. UC, pANCA, 3 SNPs from the pharmacogenetic GWAS, and rs2836878 (21q22/BRWD1), a known susceptibility SNP originally reported in the pediatric IBD GWAS (18), remained significant.
Table 5.
Model III: known susceptibility SNPs and phenotype
| Model | Variable | P value | OR | R squared |
|---|---|---|---|---|
| III: | UC vs. CD | 0.0008 | 14.7 | 0.59 |
| pANCA pos | 0.004 | 15.3 | ||
| rs6908425 6p22.3 CDKA1 |
0.05 | 4.6 | ||
| rs2836878 21q22.2 BRWD1 |
0.02 | 9.8 | ||
| rs2395185 6p21 HLA-DQA1 |
0.047 | 5.4 |
Table 6.
Model V: Final Model
| Model | Variable | P value | OR | R squared |
|---|---|---|---|---|
| V: | Diagnosis: UC vs. CD | 0.008 | 28.9 | 0.82 |
| pANCA pos vs neg | 0.03 | 15.4 | ||
| rs975664 2p12 / TACR1 |
0.01 | 26.5 | ||
|
rs4855535 3p14 / FAM19A4 |
0.02 | 10.8 | ||
| rs6100556 20q13 / PHACTR3 |
0.02 | 13.8 | ||
| rs2836878 21q22 / BRWD1 |
0.07 | 8.0 |
Clinical Utility Measures
The potential clinical utility of the final model (V) was calculated. Table 7 lists the AUC, sensitivity, specificity, accuracy and positive likelihood ratio of non response in a patient who had at least 3 of the 6 risk factors of non response based on model V. In addition, the negative likelihood ratio, i.e. the likelihood that a patient will not be a non-responder if at least 3 risk factors are absent, was calculated at 0.06.
Table 7.
Clinical Utility Measures
| Model | Variable | AUC | Sensitivity | Specificity | Accuracy | Positive Likelihood Ratio |
|---|---|---|---|---|---|---|
| V: | Diagnosis | 0.98 | 0.95 | 0.88 | 0.92 | 8 |
| pANCA | ||||||
| rs975664 2p12 TACR1 | ||||||
| rs4855535 3p14 FAM19A4 | ||||||
| rs6100556 20q13 PHACTR3 | ||||||
| rs2836878 21q22 BRWD1 |
The relative risk of non-response was calculated based on the number of risk factors (model V) carried by an individual patient. Both the frequency of non-response and the relative risk increase with increasing number of risk factors (p < 0.0001) (Figure 2).
Figure 2.
Relative Risk (RR) of non-response based on number of risk factors derived from Model V (the most general model). Risk of non response was compared between patients with 2 or less risk factors as compared to 3–4 markers and then as compared to at least 5 of the 6 risk factors for non-response.
Discussion
Anti-TNFα is an important and effective class of therapies for the management of both adult and pediatric IBD patients. Clinical experience suggests inter-individual variation in efficacy, both induction and maintenance, and in the occurrence of side effects. There are likely multiple host factors that influence these variations such as disease and immune phenotype as well as genetic background. The ability to predict which patient would have a lower likelihood of response before treatment is initiated in order to minimize exposure to potentially ineffective therapies may be an important consideration in our IBD patients. In the current era of risk; benefit balance; this concept may be very timely.
In this study we tested the associations of known and novel genetic loci with primary response outcome and developed a predictive model of primary non response using clinical phenotype, serologic and genetic variables. Six of the 28 known susceptibility loci tested were found to be associated with primary non response in the univariate analysis. The relative risk of primary non response ranged from 2.1–2.9. When tested in the multivariate analysis, only 4 of these loci remained significant. However when combined with the novel pharmacogenetic GWAS loci, only 1 loci remained significant. This one locus was initially reported in the pediatric IBD GWAS reported by Kugathasan et al (18). The functional significance of this locus remains unknown. Of the 10 novel pharmacogenetic GWAS loci tested in our predictive model, 3 remained significant in the final model. TACR1 is a receptor for substance P a known pro-inflammatory molecule. PHACTR3 (phosphatase and actin regulator 3) is associated with the nuclear scaffold in proliferating cells. While there is little known about FAM19A4 it is thought to be structurally related to MIP1α and function as a chemokine. The pharmacogenetic GWAS identified top loci did substantially improve the strength of the prediction of non-response compared to known susceptibility loci. In addition, a diagnosis of UC and pANCA positivity was independently associated with primary non-response. Table 8 compares the r-squared and AUC values for all 5 models. The combination of genotype, phenotype and serotype was the best predictive model of non response to anti-TNFα with an r-squared of 0.82 and an AUC of 0.98, and substantially better than the models that included only known IBD SNPs. (models II or III).
Table 8.
| MODEL | MODEL DESCRIPTOR | R SQUARED | AUC |
|---|---|---|---|
| 1 | Diagnosis and pANCA | 0.48 | 0.9 |
| II | Known Susceptibility SNPs only | 0.3 | 0.8 |
| III | Diagnosis, pANCA and known susceptibility SNPs | 0.59 | 0.93 |
| IV | Pharmacogenetic GWAS SNPS only | 0.67 | 0.94 |
| V | Diagnosis, pANCA, known susceptibility SNPs AND pharmacogenetic GWAS SNPs | 0.82 | 0.98 |
Our findings suggest that the majority of the known IBD susceptibility loci do not appear to greatly modify or influence primary response outcomes to anti-TNFα in pediatric IBD patients. This raises the possibility that the majority of genes that are associated with risk of disease may not influence the immune pathways that should be targeted to control or modify disease activity. The results of previously reported candidate gene association studies with anti-TNFα response have not been translated into the clinic and the functional significance of the genes tested remain unknown (8–13). The functionality of the cytokines and/or receptor renders them of interest as it relates to therapeutic outcome. With the GWAS approach, however, no a priori assumptions need to be set, as there is not a prior focus on a particular protein or target or enzyme as it relates to drug response. This hypothesis generating approach allows us to identify genetic variants that are associated with response and non-response and thus potentially identify pathways that are responsible and may well be apparent from a functional and mechanistic perspective. In this study we have analyzed the associations with primary non-response only, as we believed this was of the greatest clinical relevance. In future studies, the genetic, immunogenic and phenotypic influences on loss of response also need to be explored in order to determine those patients who may quickly lose their initial response.
Despite our significant findings, this study was limited by its relatively small sample size and lack of replication cohort, hence it is essentially a pilot study. It should be noted that previous important genetic discoveries were made using similarly small discovery cohorts for GWAS including the original publication of TNFSF15 (n=94) (29) as well as the genetic associations with statin induced myopathy (85 cases, 90 controls) (30) and with Kawasaki disease (119 cases, 135 controls) (31). In this study we included all pediatric IBD subjects receiving anti-TNFα therapy. Currently clinicians treat both CD and UC with anti-TNF therapy and the clinical trial data suggest similar primary non response and steroid free remission outcomes for both disease subtypes. Moreover there is genetic and serologic evidence that there is pathway biology overlap within the spectrum of CD and UC phenotype (20,21). In this study we did analyze UC, CD and shared susceptibility loci and the pharmacogenetic GWAS would identify loci that are independent of disease phenotype.
Defining predictors of response to anti-TNFα will assist clinicians in choosing the appropriate therapy for the appropriate IBD patient, with the goal of maximizing efficacy and minimizing toxicity. As research progresses in defining the characteristics of patients who require biologics, of equal importance will be the research as proposed herein to individualize therapy based on who will or will not respond to different classes of IBD therapeutic interventions. The development of adverse events to anti-TNF therapies such as lymphoma, and sepsis naturally induce caution in clinicians who would like the ability to appropriately select patients who are most likely to respond to these therapies.. It is important that the findings from this study be replicated in an independent cohort and also tested in adult IBD patients. We are conducting a large prospective study with predefined outcomes in an independent pediatric cohort, which, in addition to primary non-response, will also examine loss of response and sustained response as an outcome. If replicated, the novel pharmacogenetic information gained from this pilot study has the potential to improve the management of patients in the clinic with more targeted use of the existing anti-TNFα agents, Such approaches may change the way we conduct large scale clinical trials, such that only patients with a higher probability of response to specific therapies will be enrolled in an attempt to minimize exposure to ineffective therapies and thus protect patients from treatment related serious and potentially fatal adverse events. The ultimate goal of this approach is the right therapy for the right patient.
Acknowledgments
This study was supported in part by NIH/NIDDK grant P01-DK046763; Cedars-Sinai Medical Center Inflammatory Bowel/ Immunobiology Institute Research Funds; the Feintech Family Chair in IBD (SRT); the Cedars-Sinai Board of Governors' Chair in Medical Genetics (JIR); Abe and Claire Levine Chair in Pediatric IBD (MD), NCRR grant M01-RR00425 (GCRC) and DERC grant DK063491.
Footnotes
Disclosures: Marla C Dubinsky receives grant support from Centocor and UCB pharma
References
- 1.Beaugerie L, Seksik P, Nion-Larmurier I, Gendre JP, Cosnes J. Predictors of Crohn's disease. Gastroenterology. 2006;130:650–656. doi: 10.1053/j.gastro.2005.12.019. [DOI] [PubMed] [Google Scholar]
- 2.Van Limbergen J, Russell RK, Drummond HE, Aldhous MC, Round NK, Nimmo ER, Smith L, Gillett PM, McGrogan P, Weaver LT, Bisset WM, Mahdi G, Arnott ID, Satsangi J, Wilson DC. Definition of phenotypic characteristics of childhood-onset inflammatory bowel disease. Gastroenterology. 2008;135:1114–22. doi: 10.1053/j.gastro.2008.06.081. [DOI] [PubMed] [Google Scholar]
- 3.Hyams J, Crandall W, Kugathasan S, Griffiths A, Olson A, Johanns J, Liu G, Travers S, Heuschkel R, Markowitz J, Cohen S, Winter H, Veereman-Wauters G, Ferry G, Baldassano R, REACH Study Group Induction and maintenance infliximab therapy for the treatment of moderate-to-severe Crohn's disease in children. Gastroenterology. 2007;132:863–73. doi: 10.1053/j.gastro.2006.12.003. [DOI] [PubMed] [Google Scholar]
- 4.Hanauer SB, Feagan BG, Lichtenstein GR, Mayer LF, Schreiber S, Colombel JF, Rachmilewitz D, Wolf DC, Olson A, Bao W, Rutgeerts P, ACCENT I Study Group Maintenance infliximab for Crohn's disease: the ACCENT I randomized trial. Lancet. 2002;359:1541–1549. doi: 10.1016/S0140-6736(02)08512-4. [DOI] [PubMed] [Google Scholar]
- 5.Colombel JF, Sandborn WJ, Rutgeerts P, Enns R, Hanauer SB, Panaccione R, Schreiber S, Byczkowski D, Li J, Kent JD, Pollack PF. Adalimumab for maintenance of clinical response and remission in patients with Crohn's disease: the CHARM trial. Gastroenterology. 2007;132:52–65. doi: 10.1053/j.gastro.2006.11.041. [DOI] [PubMed] [Google Scholar]
- 6.Schreiber S, Khaliq-Kareemi M, Lawrance IC, Thomsen OO, Hanauer SB, McColm J, Bloomfield R, Sandborn WJ, PRECISE 2 Study Investigators Maintenance therapy with certolizumab pegol for Crohn's disease. N Engl J Med. 2007;357:239–250. doi: 10.1056/NEJMoa062897. [DOI] [PubMed] [Google Scholar]
- 7.Rutgeerts P, Sandborn WJ, Feagan BG, Reinisch W, Olson A, Johanns J, Travers S, Rachmilewitz D, Hanauer SB, Lichtenstein GR, de Villiers WJ, Present D, Sands BE, Colombel JF. Infliximab for induction and maintenance therapy for ulcerative colitis. New England Journal of Medicine. 2005;353:2462–76. doi: 10.1056/NEJMoa050516. [DOI] [PubMed] [Google Scholar]
- 8.Pierik M, Vermeire S, Steen KV, Joossens S, Claessens G, Vlietinck R, Rutgeerts P. Tumour necrosis factor-a receptor 1 and 2 polymorphisms in inflammatory bowel disease and their association with response to infliximab. Aliment. Pharmacol. Ther. 2004;20:303–310. doi: 10.1111/j.1365-2036.2004.01946.x. [DOI] [PubMed] [Google Scholar]
- 9.Mascheretti S, Hampe J, Kuhbacher T, Herfarth H, Krawczak M, Folsch UR, Schreiber S. Pharmacogenetic investigation of the TNF/TNF-receptor system in patients with chronic active Crohn's disease treated with infliximab. Pharmacogenomics. 2002;J. 2:127–136. doi: 10.1038/sj.tpj.6500091. [DOI] [PubMed] [Google Scholar]
- 10.Louis E, Vermeire S, Rutgeerts P, De Vos M, Van Gossum A, Pescatore P, Fiasse R, Pelckmans P, Reynaert H, D'Haens G, Malaise M, Belaiche J. A positive response to infliximab in Crohn disease: association with a higher systemic inflammation before treatment but not with_308 TNF gene polymorphism. Scand. J. Gastroenterol. 2002;37:818–824. [PubMed] [Google Scholar]
- 11.Taylor KD, Plevy SE, Yang H, Landers CJ, Barry MJ, Rotter JI, Targan SR. LTa 1–1-1–1 haplotype is associated with negative response in Crohn's disease ANCA pattern and LTA haplotype relationship to clinical responses to anti-TNF antibody treatment in Crohn's disease. Gastroenterology. 2001;120:1347–1355. doi: 10.1053/gast.2001.23966. [DOI] [PubMed] [Google Scholar]
- 12.Vermeire S, Louis E, Rutgeerts P, De Vos M, Van Gossum A, Belaiche J, Pescatore P, Fiasse R, Pelckmans P, Vlietinck R, Merlin F, Zouali H, Thomas G, Colombel JF, Hugot JP. NOD2/CARD15 does not influence response to infliximabin Crohn's disease. Gastroenterology. 2002;123:106–111. doi: 10.1053/gast.2002.34172. [DOI] [PubMed] [Google Scholar]
- 13.Urcelay E, Mendoza JL, Martinez A, Fernandez L, Taxonera C, Diaz-Rubio M, de la Concha EG. IBD 5 (5q31) TT is associated with negative response in Crohn's disease. IBD5 polymorphisms in inflammatory bowel disease: association with response to infliximab. World J. Gastroenterol. 2005;11:1187–1192. doi: 10.3748/wjg.v11.i8.1187. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Barrett JC, Hansoul S, Nicolae DL, Cho JH, Duerr RH, Rioux JD, Brant SR, Silverberg MS, Taylor KD, Barmada MM, Bitton A, Dassopoulos T, Datta LW, Green T, Griffiths AM, Kistner EO, Murtha MT, Regueiro MD, Rotter JI, Schumm LP, Steinhart AH, Targan SR, Xavier RJ, Libioulle C, Sandor C, Lathrop M, Belaiche J, Dewit O, Gut I, Heath S, Laukens D, Mni M, Rutgeerts P, Van Gossum A, Zelenika D, Franchimont D, Hugot JP, de Vos M, Vermeire S, Louis E, Belgian-French IBD Consortium. Wellcome Trust Case Control Consortium. Cardon LR, Anderson CA, Drummond H, Nimmo E, Ahmad T, Prescott NJ, Onnie CM, Fisher SA, Marchini J, Ghori J, Bumpstead S, Gwilliam R, Tremelling M, Deloukas P, Mansfield J, Jewell D, Satsangi J, Mathew CG, Parkes M, Georges M, Daly MJ. Genome-wide association defines more than 30 distinct susceptibility loci for Crohn's disease. Nat Genet. 2008;40:955–62. doi: 10.1038/NG.175. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Franke A, Balschun T, Karlsen TH, Sventoraityte J, Nikolaus S, Mayr G, Domingues FS, Albrecht M, Nothnagel M, Ellinghaus D, Sina C, Onnie CM, Weersma RK, Stokkers PC, Wijmenga C, Maria Gazouli M, Strachan D, McArdle WL, Vermeire S, Rutgeerts P, Rosenstiel P, Krawczak M, Vatn MH, the IBSEN study group. Mathew CG. Schreiber Sequence variants in IL10, ARPC2 and multiple other loci contribute to ulcerative colitis susceptibility. Nat Genetics. 2008;40:710–712. doi: 10.1038/ng.221. [DOI] [PubMed] [Google Scholar]
- 16.Fisher SA, Tremelling M, Anderson CA, Gwilliam R, Bumpstead S, Prescott NJ, Nimmo ER, Massey D, Berzuini C, Johnson C, Barrett JC, Cummings FR, Drummond H, Lees CW, Onnie CM, Hanson CE, Blaszczyk K, Inouye M, Ewels P, Ravindrarajah R, Keniry A, Hunt S, Carter M, Watkins N, Ouwehand W, Lewis CM, Cardon L, Welcome Trust Case Control Consortium. Lobo A, Forbes A, Sanderson J, Jewell DP, Mansfield JC, Deloukas P, Mathew CG, Parkes M, Satsangi J. Genetic determinants of ulcerative colitis include the ECM1 locus and five loci implicated in Crohn's disease. Nat. Genet. 2008;40:710–712. doi: 10.1038/ng.145. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Silverberg MS, Cho JH, Rioux JD, McGovern DP, Wu J, Annese V, Achkar JP, Goyette P, Scott R, Xu W, Barmada MM, Klei L, Daly MJ, Abraham C, Bayless TM, Bossa F, Griffiths AM, Ippoliti AF, Lahaie RG, Latiano A, Pare P, Proctor DD, Regueiro MD, Steinhart AH, Targan SR, Schumm LP, Kistner EO, Lee AT, Gregersen PK, Rotter JI, Brant SR, Taylor KD, Roeder K, Duerr RH. Ulcerative colitis-risk loci on chromosomes 1p36 and 12q15 found by genome-wide association study. Nat Genetics. 2009;41(2):216–20. doi: 10.1038/ng.275. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Kugathasan S, Baldassano RN, Bradfield JP, Sleiman PM, Imielinski M, Guthery SL, Cucchiara S, Kim CE, Frackelton EC, Annaiah K, Glessner JT, Santa E, Willson T, Eckert AW, Bonkowski E, Shaner JL, Smith RM, Otieno FG, Peterson N, Abrams DJ, Chiavacci RM, Grundmeier R, Mamula P, Tomer G, Piccoli DA, Monos DS, Annese V, Denson LA, Grant SF, Hakonarson H. Loci on 20q13 and 21q22 are associated with pediatric-onset inflammatory bowel disease. Nat Genetics. 40:1211–1215. doi: 10.1038/ng.203. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Ferrante M, Vermeire S, Katsanos KH, Noman M, Van Assche G, Schnitzler F, Arijs I, De Hertogh G, Hoffman I, Geboes JK, Rutgeerts P. Predictors of early response to infliximab in patients with ulcerative colitis. Inflammatory Bowel Diseases. 2007;13:123–8. doi: 10.1002/ibd.20054. [DOI] [PubMed] [Google Scholar]
- 20.Vasiliauskas EA, Plevy SE, Landers CJ, Binder SW, Ferguson DM, Yang H, Rotter JI, Vidrich A, Targan SR. Perinuclear antineutrophil cytoplasmic antibodies in patients with Crohn's disease define a clinical subgroup. Gastroenterology. 1996;110:1810–1819. doi: 10.1053/gast.1996.v110.pm8964407. [DOI] [PubMed] [Google Scholar]
- 21.Anderson CA, Massey DC, Barrett JC, Prescott NJ, Tremelling M, Fisher SA, Gwilliam R, Jacob J, Nimmo ER, Drummond H, Lees CW, Onnie CM, Hanson C, Blaszczyk K, Ravindrarajah R, Hunt S, Varma D, Hammond N, Lewis G, Attlesey H, Watkins N, Ouwehand W, Strachan D, McArdle W, Lewis CM, Wellcome Trust Case Control Consortium. Lobo A, Sanderson J, Jewell DP, Deloukas P, Mansfield JC, Mathew CG, Satsangi J, Parkes M. Investigation of Crohn's disease risk loci in ulcerative colitis further defines their molecular relationship. Gastroenterology. 2009;136:523–9. doi: 10.1053/j.gastro.2008.10.032. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Dubinsky MC, Kugathasan S, Mei L, Picornell Y, Nebel J, Wrobel I, Quiros A, Silber G, Wahbeh G, Katzir L, Vasiliauskas E, Bahar R, Otley A, Mack D, Evans J, Rosh J, Oliva Hemker M, Leleiko L, Crandall W, Langton C, Landers C, Taylor KD, Targan SR, Rotter JI, Markowitz J, Hyams J, the Western Regional Pediatric IBD Research Alliance. Pediatric IBD Collaborative Research Group. the Wisconsin Pediatric IBD Alliance Increased immune reactivity predicts aggressive complicating Crohn's disease in children. Clinical Gastroenterology and Hepatology. 2008;6:1105–11. doi: 10.1016/j.cgh.2008.04.032. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Gunderson KL, Steemers FJ, Lee G, Mendoza LG, Chee MS. A genome-wide scalable SNP genotyping assay using microarray technology. Nat Genet. 2005;37:549–54. doi: 10.1038/ng1547. [DOI] [PubMed] [Google Scholar]
- 24.Harvey RF, Bradshaw JM. A simple Index of Crohn's disease activity. Lancet. 1980;1:514. doi: 10.1016/s0140-6736(80)92767-1. [DOI] [PubMed] [Google Scholar]
- 25.Price AL, Patterson NJ, Plenge RM, Weinblatt ME, Shadick NA, Reich D. Principal components analysis corrects for stratification in genome-wide association studies. Nat Genet. 2006 Aug;38(8):904–9. doi: 10.1038/ng1847. [DOI] [PubMed] [Google Scholar]
- 26.Purcell S, Neale B, Todd-Brown K, et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet. 2007;81:559–75. doi: 10.1086/519795. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Stokes ME, Davis CS, Koch GG. Categorical data analysis using the SAS system. 2nd edition SAS press; Cary,NC,USA: 2000. [Google Scholar]
- 28.Armitage P, Berry G, Matthews JNS. Statistical Methods in Medical Research. Fourth Edition Wiley-Blackwell; Malden, MA: 2002. [Google Scholar]
- 29.Yamazaki K, McGovern D, Ragoussis J, Paolucci M, Butler H, Jewell D, Cardon L, Takazoe M, Tanaka T, Ichimori T, Saito S, Sekine A, Iida A, Takahashi A, Tsunoda T, Lathrop M, Nakamura Y. Single nucleotide polymorphisms in TNFSF15 confer susceptibility to Crohn's disease. Human Molecular Genetics. 2005;14:3499–506. doi: 10.1093/hmg/ddi379. [DOI] [PubMed] [Google Scholar]
- 30.SEARCH Collaborative Group. Link E, Parish S, Armitage J, Bowman L, Heath S, Matsuda F, Gut I, Lathrop M, Collins R. SLCO1B1 variants and statin-induced myopathy--a genomewide study. New England Journal of Medicine. 2008;359:789–99. doi: 10.1056/NEJMoa0801936. [DOI] [PubMed] [Google Scholar]
- 31.Burgner D, Davila S, Breunis WB, Ng SB, Li Y, Bonnard C, Ling L, Wright VJ, Thalamuthu A, Odam M, Shimizu C, Burns JC, Levin M, Kuijpers TW, Hibberd ML, International Kawasaki Disease Genetics Consortium A genome-wide association study identifies novel and functionally related susceptibility Loci for Kawasaki disease. PLoS Genetics. 2009;5:e1000319. doi: 10.1371/journal.pgen.1000319. [DOI] [PMC free article] [PubMed] [Google Scholar]