This is a prospective cohort study approved by the University of Louisville Institutional Review Board. All patients signed a written informed consent. Consecutive patients with a diagnosis of CD were identified from a large prospectively maintained genetic database, from a large University digestive disease practice, encompassing the period 1/1998 to 4/2016. Inclusion criteria were CD patients who had received anti-TNF therapy, and whose medical records were available, with information about receipt of anti-TNF therapy, its duration, efficacy, and cessation where applicable. Included patients received appropriate drug doses and had a follow-up of at least 12 mo following treatment initiation[16,17]. Patients were excluded if anti-TNF treatment was stopped due to side-effects, local and/or systemic allergy, or if it was impossible to distinguish from the medical records whether the drug worked.

Additional data collected from the medical records included gender, race, socioeconomic status (patient’s

zip code of residence was used to obtain median household income based on United States census data from the American Community Survey 2014 - 5 year estimates)[18], surgical history, and clinical state of the disease according to the Montreal classification for CD, including age at diagnosis, location, disease behavior, and the presence or absence of perianal disease[19].

The main outcome measure (following exposure to the drug) was response to therapy. Participants were grouped as ever-responders if they had initial response to anti-TNF treatment (even if this was later lost due to antibody formation) or non-responders in accordance with the treating physician decision. The patient genotypes (see below) were assessed as the predictors of outcome. Possible confounders and effect modifiers included age, gender, race, and socioeconomic status disease, as well as disease characteristics (such as Montreal criteria).

Peripheral blood was collected by venipuncture (after written informed consent) in EDTA-vacutainers (BD, Franklin Lakes, NJ, United States) and stored at 4 °C until further use.

Genomic DNA was extracted from blood samples using the illustra GenomiPhi V2 DNA Amplification Kit (GE Healthcare, Pittsburgh, PA, United States) using the manufacturer's protocol[20]. Briefly, the blood was initially diluted with PBS buffer. Blood was then lysed: 1 μL of diluted blood was lysed with 1 μL of cell lysis solution (400 mmol/L KOH, 10 mmol/L EDTA, 100 mmol/L DTT), followed by the addition of 1 μL of neutralization buffer (400 mmol/L HCl, 600 mmol/L Tris-HCl, pH 7.5). Whole genome amplification was then performed: 17 μL of master mix [7 μL sample buffer, 9 μL reaction buffer, and 1 μL enzyme mix from the illustra GenomiPhi V2 DNA Amplification Kit (GE Healthcare, Pittsburgh, PA, United States)] was added to each sample for a total reaction volume of 20 μL. Amplification was performed according to the following program: 30 °C for two hours, followed by 65 °C for 10 min, then cooled to 4 °C. Following whole genome amplification, DNA concentration was determined using NanoDrop^® 2000 spectrophotometry. The samples were diluted and stored at -20 °C until analysis.

SNPs selection: A PubMed literature search was conducted using the keywords “tumor necrosis factor-alpha”, “anti-TNF”, “infliximab”, “adalimumab”, “polymorphism”, “Crohn’s disease”, “response”, “biomarker” using Boolean operators (AND), (OR), (NOT). Results were narrowed down to original studies investigating SNPs including frequency of alleles and genotypes for different groups. We included SNPs that had demonstrated association with CD or anti-TNF treatment, those that had biological relevance, and those that had an expected minor allele frequency ≥ 5%. Both new genetic associations and previously described efforts were investigated. SNPs were excluded if they had been extensively investigated and if there was no prognostic value for the combination of CD and anti-TNF treatment response. As a result of this search, the following seven SNPs within 5 genes were selected for study and assessed in each patient’s DNA sample: ATG16L1 (rs10210302, T300A rs2241880), Fas ligand (-843 rs763110), IBD5 (rs2522057), FCGR 3A (rs396991), and TNF (-308 rs1800629, -238 rs361525).

SNP assessment was performed using TaqMan^® predesigned genotyping assays (Life Technologies^®, Carlsbad CA)[21]. The TaqMan^® genotyping assays were diluted to a 20× working stock solution with 1× TE buffer (10 mmol/L Tris-HCl, 1 mmol/L EDTA, pH 8.0, in DNase-free, sterile-filtered water) and stored at -20 °C, as recommended by the manufacturer.

MicroAmpR Fast Optical 96-well reaction plates (Applied Biosystems, Foster City, CA, United States) were used. Six microliters of master mix was used for each assay (5.5 μL of TaqMan^® Universal Master Mix II, no UNG [Applied Biosystems™] together with 0.5 μL of 20 × working assay [TaqMan^® predesigned Genotyping Assays]). Five microliters of DNA (4.5 ng/μL) was added to the plate. PCR reactions were performed using a Step-One Plus^® RT-PCR System (Life Technologies^®, Carlsbad, CA, United States) and the following program: 95 °C for 10 min, followed by 40 cycles of 95 °C for 15 s, and then 60 °C for 1 min. Analysis was performed using Step-One Plus^® software v2.1 (applied Biosystems, Foster City, CA, United States). Each genotype was independently assigned by two investigators. In cases of disagreement, assignment was reached by consensus. All laboratory work and genotyping was done at the Price Institute of Surgical Research, Louisville Kentucky, United States.

Descriptive and analytical statistics were performed using SAS version 9.4 statistical software[22]. Genotype frequencies, demographic, and disease characteristics were compared using a χ² test (or Fisher’s exact test for 2 × 2 tables). Socioeconomic status was calculated according to the national percentile of the patient’s median household income divided into quartiles (0-25, 26-50, 51-75 and 76-100) and compared using a χ² test. Comparison of continuous variables was performed using a two-sample t-test or ANOVA. In order to explore for the presence of bias in the cohort, a group of contemporary subjects who did not receive anti-TNF treatment were compared with patients included in this study. Following this, characteristics between ever-responders and non-responders in the study group were then compared[23].

Hardy-Weinberg equilibrium was determined for each SNP (Table 1). Univariable logistic regression was modeled for the probability of anti-TNF treatment failure for each covariate. Multivariable logistic regression models were used for separate SNPs and covariates exhibiting a trend towards a significant difference (P < 0.15)[24]. Final models included OR and 95%CI. A P-value of < 0.05 was considered statistically significant.

The statistical methods of this study were reviewed by Pan J and Rai SN.

Table 1 shows a flow diagram of patient selection; 121 patients were selected for study. Of these, 21 (17.4%) patients were primary non-responders to anti-TNF treatment and 100 (82.6%) patients were ever-responders to anti-TNF treatment. A quarter of these initial ever-responders (25/100) lost response at a later time and were termed secondary non-responders. The patient population was predominantly Caucasian (92.6%), with a higher proportion of women (58.7%) (Table 2). With regards to clinical parameters, 90/121 (74.4%) patients were diagnosed between the ages of 17 and 40 years of age (Montreal A2). Most CD patients, 74/121 (61.2%), had combined ileocolonic disease (Montreal L3), whereas 15/121 (12.4%) had isolated ileal disease (Montreal L1), and 32/121 (26.4%) had only colonic disease (Montreal L2). Only 3/121 (2.5%) patients had upper GI (Montreal L4) involvement, all of whom were responders. Montreal L4 disease was analyzed separately from L1-3, due to the fact that, according to the Montreal classification, it is not mutually exclusive and can be added to any of the other locations when concomitant upper GI disease is present[19]. The population was fairly evenly distributed with respect to disease behavior with 36/121 (30%) patients having non-stricturing, non-penetrating disease (Montreal B1), 45/121 (37%) patients having stricturing disease (Montreal B2), and 40/121 (33%) patients having penetrating disease (Montreal B3). In addition, 37 of 121 (31%) patients had perianal disease (Montreal P designation). Table 2 shows the clinical and demographic data of the participants, as well as these data for the non-responder and ever-responder groups. None of the clinical or demographic characteristics were significantly different between these 2 groups.

When comparing the characteristics of the patients who received anti-TNF treatment and included in the study (n = 121) with those who did not receive anti-TNF treatment (n = 152) in order to ascertain the presence of bias, no difference was found in 3 of the 4 variables examined: gender (P = 0.27), race (P = 0.95), or socioeconomic status (P = 0.23). The patients included in the study who received anti-TNF treatment were, however, younger (41.6 years old, 95%CI: 39.2-44.0) than those that did not receive anti-TNF treatment (49.4 years old, 95%CI: 47.1-51.7) (P < 0.001).

We assessed 7 different SNPs associated with 5 different genes and observed less than 5% technical failure rate in all assays. Table 3 shows the SNPs tested as well as their genotype and allele distribution. Comparison of genotypes between ever responders and non-responders (Table 4) identified a significant difference in the Fas ligand SNP rs763110 genotypes (P = 0.042). Patients with a CC genotype (as compared to those with a TC or TT genotype) were more likely to be non-responders to anti-TNF treatment, (P = 0.016; OR = 0.31, 95%CI: 0.11-0.83). Genotypes of another SNP, such as -308 (rs1800629), within the TNF gene demonstrated a trend towards correlation with response to anti-TNF treatment (P = 0.088) when grouping genotypes AA and GA compared to genotype GG (P = 0.093, OR = 2.29, 95%CI: 0.85-6.17).

The vast majority of participants were Caucasian. Only 9 patients were African American, all of whom were ever-responders. Analyzing the Caucasians separately as a sensitivity analysis achieved similar results for the grouped Fas ligand SNP (P = 0.029) and for the grouped -308 TNF gene SNP (P = 0.049). No significant difference was observed for the remaining SNPs studied: ATG16L1 (rs10210302, T300A rs2241880), IBD5 (rs2522057), FCGR 3A (rs396991), and TNF (-238 rs361525).

Results of the univariable comparisons are shown in Table 5. In univariable analyses, the Fas ligand SNP (rs763110) demonstrated a difference between ever-responders and non-responders with borderline significance (P = 0.058) and significance when grouping TC and TT genotypes together (P = 0.020). The comparison of -308 SNP (rs1800629) genotypes between ever-responders and non-responders (P = 0.130) became more different when grouping AA and GA genotypes together (P = 0.099). Univariate variables with P < 0.15 were included in the multivariable analysis for the comparison between anti-TNF treatment ever-responders and non-responders. Both Montreal disease behavior (P = 0.125) and perianal disease classification (P = 0.086) were included in the multivariable analysis.

Logistic multivariable regression models were developed for the Fas ligand (rs763110) SNP, the TNF gene -308A/G (rs1800629) SNP, and their combination. The multivariable logistic regression models included genotype data for each of these two SNPs (with genotypes grouped as described above), the Montreal disease behavior classification, and the Montreal perianal disease classification (Table 6). The Fas ligand SNP (rs763110) CC genotype was predictive of non-response, as compared to the TC and TT genotypes (P = 0.009, OR = 4.30, 95%CI: 1.45-12.80). In the -308 TNF gene (rs1800629) SNP multivariable model, the AA and GA genotypes were significantly predictive of non-response as compared to the GG genotype (P = 0.049, OR = 2.88, 95%CI: 1.01-8.22). Patients with the combination of the Fas ligand (rs763110) CC genotype and presence of the TNF -308 A allele (genotypes AA or GA as opposed to GG) had nearly five-fold higher odds of being non-responders (P = 0.015, OR = 4.76, 95%CI: 1.35-16.77). This occurred in 16 (13%) of our patients. Montreal disease behavior and the presence of perianal disease were not found to be predictive in any of the multivariable models.

Anti-tumor necrosis factor (TNF) agents will not be effective in a subset of patients with Crohn’s disease (CD). Predicting the subset of patients who do not respond to anti-TNF treatment is important clinically and economically.

A possible conduit to predict response to anti-TNF therapy could be through genetic testing. The authors chose to examine a series of single nucleotide polymorphisms (SNPs) within genes that have been linked either with CD and/or with anti-TNF treatment response in order to determine whether these could aid in predicting response to anti-TNF treatment in CD patients.

Two SNPs Fas ligand and TNF gene -308 were associated with response to anti-TNF treatment.

Genotyping these SNPs from DNA obtained from peripheral blood may help define which CD patients are likely to benefit from anti-TNF therapy and permit efficient and cost-effective treatment by avoiding expensive therapy that is likely to fail and permitting selection of other treatments more likely to succeed.

Single nucleotide polymorphisms are a type of genetic variation in which a change is found in a single nucleotide at a specific position in the genome.

The authors investigate genetic factors that might help define which CD patients are likely to benefit from anti-TNF therapy. This is an interesting paper.