A Clinical Pharmacogenetic Model to Predict the Efficacy of Methotrexate Monotherapy in Recent-Onset Rheumatoid Arthritis

Transcription

1 ARTHRITIS & RHEUMATISM Vol. 56, No. 6, June 2007, pp DOI /art , American College of Rheumatology A Clinical Pharmacogenetic Model to Predict the Efficacy of Methotrexate Monotherapy in Recent-Onset Rheumatoid Arthritis Judith A. M. Wessels, 1 Sjoerd M. van der Kooij, 1 Saskia le Cessie, 1 Wietske Kievit, 2 Pilar Barerra, 2 Cornelia F. Allaart, 1 Tom W. J. Huizinga, 1 and Henk-Jan Guchelaar, 1 for the Pharmacogenetics Collaborative Research Group ISRCTN no Supported by the Foundations Nuts Ohra, Amsterdam, and De Drie Lichten, Leiden, The Netherlands, AutoCure, and the European Union s Sixth Framework Programme for Research (FP6). The BeSt study is supported by the Dutch College for Health Insurance, with additional funding provided by Centocor and Schering-Plough. 1 Judith A. M. Wessels, PharmD, Sjoerd M. van der Kooij, MD, Saskia le Cessie, PhD, Cornelia F. Allaart, MD, Tom W. J. Huizinga, MD, PhD, Henk-Jan Guchelaar, PharmD, PhD: Leiden University Medical Center, Leiden, The Netherlands; 2 Wietske Kievit, MSc, Pilar Barerra, PhD: University Medical Centre Nijmegen, Nijmegen, The Netherlands. Members of the Pharmacogenetics Collaborative Research Group, in addition to the authors of this article, are as follows: R. J. H. M. van der Straaten, PhD (Leiden, The Netherlands), R. de Jonge, PhD (Rotterdam, The Netherlands), Y. P. M. Goekoop- Ruiterman, MD (Leiden, The Netherlands), J. K. de Vries-Bouwstra, MD (Amsterdam, The Netherlands), D. van Schaardenburg, MD (Amsterdam, The Netherlands), M.-L. Westedt, MD (The Hague, The Netherlands), M. de Bois, MD (The Hague, The Netherlands), P. L. van Riel, MD, PhD (Nijmegen, The Netherlands), M. J. Coenen, PhD (Nijmegen, The Netherlands). Dr. Allaart has received speaking fees (less than $10,000) from Schering-Plough. Professors Huizinga and Guchelaar hold a patent for the methotrexate-responsive predictive model. Address correspondence and reprint requests to Henk-Jan Guchelaar, PharmD, PhD, Leiden University Medical Center, Department of Clinical Pharmacy and Toxicology, PO Box 9600, 2300 RC Leiden, The Netherlands. H.J.Guchelaar@lumc.nl. Submitted for publication December 8, 2006; accepted in revised form February 20, Objective. To develop a clinical pharmacogenetic model to predict the efficacy of methotrexate (MTX) in rheumatoid arthritis (RA). Methods. Two hundred five patients with newly diagnosed RA and active disease were treated with MTX (initiated at a dosage of 7.5 mg/week and increased to 15 mg/week after 4 weeks) and folic acid (1 mg/day). If the Disease Activity Score (DAS) was >2.4 at 3 months, the dosage of MTX was increased up to 25 mg/week. Twenty-four baseline variables possibly influencing disease state and drug response were selected. In addition, 17 polymorphisms in 13 genes related to the MTX mechanism of action, purine and pyrimidine synthesis, were determined. Factors were compared between responders (defined as patients with a DAS <2.4 at 6 months) and nonresponders. In case of differences, a stepwise selection procedure identified the predictors for response. A clinical score was designed by simplifying regression coefficients of the independent variables. Cutoff levels were chosen based on the clinical score, and positive and negative response rates were calculated. An evaluation of the model was performed in a second group of patients. Results. The model for MTX efficacy consisted of sex, rheumatoid factor and smoking status, the DAS, and 4 polymorphisms in the AMPD1, ATIC, ITPA, and MTHFD1 genes. This prediction model was transformed into a scoring system ranging from 0 to Scores of <3.5 had a true positive response rate of 95%. Scores of >6 had a true negative response rate of 86%. Sixty percent of the patients were categorized as either responders or nonresponders, whereas 32% of the patients were categorized using a nongenetic model. Evaluation of the model in 38 additional patients with RA supported the results. Conclusion. This study established a model for predicting the efficacy of MTX in patients with RA. This pharmacogenetic model may lead to better-tailored initial treatment decisions in patients with RA. Disease activity in rheumatoid arthritis (RA) leads to progressive joint and cartilage destruction (1). Patients with active RA generally experience declining functionality and disability within 2 years of disease onset (2). Recent clinical studies demonstrated that 1765

2 1766 WESSELS ET AL early therapeutic interventions with combination strategies, with or without anti tumor necrosis factor (anti- TNF) agents, are superior to monotherapy with diseasemodifying antirheumatic drugs (DMARDs) and considerably improve the prognosis (3 5). Although combination strategies are more efficacious, such an intensive approach probably is not necessary for all patients with newly diagnosed RA (6,7). In addition, combination treatment with anti-tnf agents or corticosteroids possibly introduces other risks, such as serious infection or osteoporosis (8,9). As a result, it is important to determine whether patients have a high probability of response to DMARD monotherapy, in order to preclude or prescribe combination treatment. Methotrexate (MTX) is the most widely used DMARD in clinical practice, although the factors determining the efficacy of MTX are largely unknown. Previously, the influences of demographic, clinical, immunologic, and genetic factors on the state of disease in patients with RA have been studied (10 18). Specifically, polymorphisms in genes coding for methylenetetrahydrofolate reductase (MTHFR), adenosine monophosphate deaminase (AMPD1), aminoimidazole carboxamide ribonucleotide transformylase (ATIC), serine hydroxymethyltransferase 1 (SHMT1), and inosine triphosphate pyrophosphatase (ITPA) demonstrate association with the MTX response (19 22). Nongenetic factors such as low disease activity at baseline, male sex, treatment with nonsteroidal antiinflammatory drugs (NSAIDs), and lower levels of creatinine clearance are also related to the efficacy of MTX (23). However, these associations have not been transformed into clinical decision-making tools to guide MTX treatment in patients. Therefore, the aim of this study was to develop a clinical pharmacogenetic model to predict the efficacy of MTX monotherapy in patients with recent-onset RA. PATIENTS AND METHODS RA patient population. The 205 patients enrolled in this study comprised a subcohort of the 508 patients participating in the BeSt (Behandelstrategieën voor Reumatoide Artritis [Treatment strategies for rheumatoid arthritis]) study (5). The BeSt study was a randomized, multicenter clinical study comparing the clinical efficacy of 4 different treatment strategies in patients with recent-onset RA: sequential monotherapy starting with MTX (n 126), step-up from MTX to combination therapy with MTX and sulfasalazine (n 121), initial combination therapy with MTX, sulfasalazine, and high-dose (with tapering) prednisone (n 133), or initial biologic therapy with infliximab plus MTX (n 128). Inclusion criteria were a diagnosis of early RA, as defined by the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) 1987 revised criteria (24), age 18 years, and symptom duration of 2 years. Patients were also required to have active RA, with 6 of 66 swollen joints, 6 of 68 tender joints, and either an erythrocyte sedimentation rate (ESR) 28 mm/hour or a global health score of 20 mm on a mm visual analog scale (VAS), where 0 best and 100 worst. Exclusion criteria were previous treatment with DMARDs other than antimalarials and concomitant treatment with an experimental drug. The local ethics committee at each participating hospital approved the study protocol. All patients gave informed consent before enrollment into the study. Study design and treatment. Patients who were allocated to initial monotherapy with MTX and for whom DNA samples were available (n 205) were included in the current analysis. The primary goal of therapy in the BeSt study was clinical response as defined by a Disease Activity Score (DAS) (25) in 44 joints of 2.4. The DAS is a validated composite outcome measure consisting of the Ritchie Articular Index (RAI) (26), the swollen joint count (SJC), general well-being as indicated by the patient on a VAS, and the ESR. A research nurse who was blinded to the allocated treatment group assessed the DAS every 3 months. All patients included in this analysis began treatment with a regimen of oral MTX at a dosage of 7.5 mg weekly, with the dosage increasing to 15 mg weekly after 4 weeks, in combination with folic acid (1 mg daily). In the event of insufficient clinical response (DAS 2.4) at the 3-month followup visit, the MTX dosage was increased stepwise, from 5 mg every 2 weeks to 25 mg weekly. In case of adverse reactions, MTX was continued at the lowest tolerated dosage. In case of intolerance, MTX could also be given parenterally. If MTX was not tolerated at all, the patient was treated according to the next treatment step. Concomitant treatment with NSAIDs and intraarticular injections of corticosteroids was allowed for all treatment groups. Systemic corticosteroid therapy was not allowed. Intraarticular corticosteroids were more often used in MTX nonresponders (34 of 99 [34%]) than in MTX responders (13 of 87 [15%]) (P 0.002). For the current analysis, clinical data from the first 6 months of followup were used to represent MTX treatment only. MTX response evaluation. Responders were defined as patients who were receiving MTX and had a DAS of 2.4 at 6 months (good clinical response). Nonresponders were defined as patients who were receiving MTX at the 6-month followup and had a DAS of 2.4. Of the 205 patients enrolled in the study, 19 were unavailable for efficacy analyses (2 patients moved; 1 patient refused to continue receiving MTX after a short period, without having an adverse drug event; 5 patients did not have their DAS assessed; 1 patient began receiving sulfasalazine before the evaluation; and 10 patients had discontinued MTX treatment permanently after experiencing adverse drug events). Consequently, 186 patients remained eligible for the MTX efficacy evaluation at 6 months. Patients who experienced adverse drug events but continued to be treated with MTX at 6 months were included in the analysis. Selection of demographic, clinical, and immunologic factors. Baseline variables possibly influencing the patient s disease state and MTX response were selected based on the literature (10 18,20,23,27,28). The following factors were iden-

3 MODEL TO PREDICT EFFICACY OF MTX IN RA 1767 tified: sex; rheumatoid factor (RF) status; age; duration of joint symptoms; alcohol consumption; smoking; body mass index; menopausal status; hormone supplementation; scores for physician s assessment of disease activity, pain, patient s assessment of disease activity, patient s assessment of global health, and morning stiffness on a point VAS; Health Assessment Questionnaire (HAQ) score (29); ESR; C-reactive protein level; DAS; SJC; RAI; kidney function (defined as creatinine clearance); anti cyclic citrullinated peptide (anti- CCP) status; NSAID use; and the existence of comorbidity based on drug use (other than RA disease related drugs). In some patients, the anti-ccp assay was not performed at the time of inclusion in the BeSt study. Because anti-ccp status is unlikely to change with treatment (28), the anti-ccp status after beginning treatment was also used. Selection of single-nucleotide polymorphisms (SNPs). Seventeen SNPs in 13 candidate genes related to the MTX mechanism of action, purine and pyrimidine synthesis (30 32), were selected, taking into consideration the following criteria (33,34): validated SNP, SNP causes nonsynonymous amino acid change, indications for clinical relevance from previous publications, and a preferred minimal genotype frequency of 10% (20,21,35 48). DNA was isolated from peripheral white blood cells by the standard manual salting-out method. As a quality control, positive controls (Control DNA CEPH ; Applied Biosystems, Foster City, CA) and negative controls (water) were used. In addition, 5 10% of samples were genotyped in duplicate, and no inconsistencies were observed. Genotyping techniques, success rates, and genotype frequencies for 10 of the 17 SNPs in this population, together with their association with MTX response, were previously reported (19,21). These SNPs were in genes coding for adenosine monophosphate deaminase (AMPD1), aminoimidazole carboxamide ribonucleotide transformylase (ATIC), inosine triphosphate pyrophosphatase (ITPA), methionine synthase (MTR), methionine synthase reductase (MTRR), dihydrofolate reductase (DHFR), methylenetetrahydrofolate reductase (MTHFR), and the reduced folate carrier (RFC). The following SNPs were analyzed: MTHFR 677C T (rs ), MTHFR 1298A C (rs ), DHFR 473G A (DNA alignment, rs ), DHFR 35289A G (DNA alignment, rs ), RFC 80G A (rs ), MTRR 66A G (rs ), MTR 2756A G (rs ), AMPD1 34C T (rs ), ITPA 94A C (rs ), and ATIC 347C G (rs ). Seven SNPs were de novo genotyped for this analysis in our population. These SNPs were in genes coding for methylenetetrahydrofolate dehydrogenase (MTHFD1), serine hydroxymethyltransferase (SHMT1), folylpolyglutamate synthase (FPGS), gamma glutamyl hydrolase (GGH), and thymidylate synthetase (TYMS). The SNPs were MTHFD1 1958G A (rs ), SHMT1 1420C T (rs ), TYMS 28-bp tandem repeat polymorphism in the promoter region, FPGS 114G A (rs ), FPGS 1994A G (rs10106), GGH 452C T (rs ), and GGH 16T C (rs ). Real-time polymerase chain reaction (PCR) using TaqMan technique for MTHFD1 and SHMT1 were used in genotyping. Both assays were performed according to protocols provided by the manufacturer (Applied Biosystems). FPGS and GGH SNPs were genotyped using the pyrosequencer method and protocols (Pyrosequencing, Uppsala, Sweden). Double (2R2R) or triple (3R3R) 28-bp tandem repeats in the promoter region of the TYMS gene were visualized on agarose gel directly after the PCR. Genotype distributions were as follows: for MTHFD1 1958G A, 29% GG, 50% GA, 22% AA; for SHMT1 1420C T, 54% CC, 42% CT, 4% TT; for FPGS 114G A, 49% GG, 45% GA, 6% AA; for FPGS 1994A G, 31% AA, 48% GA, 21% GG; for GGH 452C T, 83% CC, 17% CT; for GGH 16T C, 53% TT, 39% CT, 7% CC; and for TYMS 28-bp repeats, 31% 3R3R, 48% 2R3R, 21% 2R2R, 0.4% 2R6R. The success rates for these assays were as follows: for MTHFD1 1958G A, 99%; for SHMT1 1420C T, 99.5%; for TYMS 28-bp repeat, 99%; for FPGS 114G A, 91.2%; for FPGS 1994A G, 97.1%; for GGH 452C T, 98.5%; for GGH 16T C, 94.1%. Genotype frequencies for all 17 SNPs were in Hardy- Weinberg equilibrium. The mean overall success rate was 97.6%. Our population comprised 93.2% Caucasians (n 191), 2.4% Asians (n 5), 1.0% Africans (n 2), and 3.4% other (n 3 Hindustani, n 3 Surinamese, n 1 Israeli). RA patient group for replication. Thirty-eight patients were recruited to replicate the genetic model. The validation cohort data were collected after development of the predictive model, but the investigators were blinded to either the clinical parameters or the genotyping results for the replication group. Fourteen patients were available from the BeSt study; they were treated with initial MTX monotherapy, but DNA samples were obtained after the clinical pharmacogenetic model was developed. Twenty-four patients were available from the RA cohort of the rheumatology department at the University Medical Center Nijmegen (UMCN), consisting of 553 patients (5,49). Patients were eligible for the replication cohort if they fulfilled the ACR 1987 criteria for RA, started with MTX monotherapy, had not been treated previously with DMARDs other than antimalarial agents, and had disease duration of less than 2 years. In addition, clinical data comprising the prediction model and DNA samples had to be available. In the UMCN cohort, 352 patients were recorded as having received MTX at any time point. Of these 352 patients, only 36 received MTX monotherapy as their primary DMARD for more than 6 months. Twenty-four of these 36 patients were used for validation of the clinical model; the other 12 patients were excluded because no DNA was available, poor quality DNA was available, or prednisone was prescribed as additional therapy in doses 10 mg. Statistical analysis. Variables between responders and nonresponders were compared using Student s t-test, the Mann-Whitney U test, or the chi-square test, depending on the tested variable. Candidate variables with a P value 0.1 were selected for the multiple regression analysis. Except for anti- CCP status, missing baseline values were not replaced, because the data were 98.6% complete. Missing anti-ccp data were imputed, because the data were 87% complete (n 178 patients). The unknown anti-ccp status per patient was replaced by the expected fractional probability of a patient being anti-ccp positive or anti-ccp negative, given the RF status. Differences in genotype distribution for response were tested by 2 2 cross-tabulations for carriers versus noncarri-

4 1768 WESSELS ET AL Table 1. Comparison of baseline variables between responders and nonresponders* Baseline variable Responders (n 87) Nonresponders (n 99) P Age, years Female sex, % Alcohol consumption, % Smoking, % Body mass index, kg/m Disease Activity Score Swollen joint count, median (IQR) 11 (8 18) 14 (11 20) Ritchie Articular Index, median (IQR) 11 (7 15) 16 (13 20) Duration of symptoms, median (IQR) weeks 24 (13 54) 25 (15 42) VAS physician s assessment of disease activity score VAS pain score VAS patient s assessment of disease activity score VAS patient s assessment of global health score VAS morning stiffness score Health Assessment Questionnaire score ESR, median (IQR) mm/hour 34 (18 50) 40 (26 65) C-reactive protein, median (IQR) mg/liter 20 (9 42) 24 (14 58) Rheumatoid factor positive, % Anti-CCP positive, % Creatinine clearance, ml/minute Nonsteroidal antiinflammatory drug use, % Comorbidity, % Postmenopausal women, % Hormone supplementation, % * Except where indicated otherwise, values are the mean SD. Some values for anti cyclic citrullinated peptide (anti-ccp) were not determined at baseline. Comorbidity was defined as the patient s use of drugs other than those to treat rheumatoid arthritis. IQR interquartile range; VAS visual analog scale (0 100 mm); ESR erythrocyte sedimentation rate. ers, with analysis by 2-sided chi-square test. For continuous variables such as the DAS, their contribution to the MTX response was studied as both continuous and categorical variables, in quartiles. In addition, the interaction between RF, smoking, and anti-ccp status was explored (13,50). The interaction between age ( 50 years, years, and 60 years) and sex was also studied (17). In the backward selection procedure, the most significant independent variables were identified using P 0.10 as the removal criterion. This was done for both genetic and clinical candidate variables (the pharmacogenetic model) and only the clinical variables (the nongenetic model). The predicted probability for MTX efficacy in a logistic regression model is related to the covariates in the equation A B1 x1 B2 x2 B3 x3... Bk xk (where A the estimated constant term, x a particular value for an explanatory variable, and B the regression coefficient). The exponential of the regression coefficient, e B, is an estimate of the adjusted odds ratio. The estimated probability for response was calculated for each individual patient with a set of variable values. A receiver operating characteristic (ROC) curve was derived to evaluate the discriminative performance of the model. Cross-validation was performed to control for overfitting (51). Weighted scores for the simplified model were assigned by rounding the regression coefficients in the final model to the nearest number ending in.5 or.0. Negative regression coefficients were inverted to obtain only positive weighted scores. Categories within a variable were grouped if regression coefficients led to identical scores. The calculated scores per individual were compared with the observed responses to MTX. Higher calculated scores reflect a higher probability of nonresponse to MTX. Several clinical score cutoff levels that represent (approximately) 0.80 or 0.20 probability of response were chosen to classify patients as nonresponders, intermediate responders, or responders. The true positive response rate and the true negative response rate were calculated. All statistical analyses were performed using SPSS 11.5 software (SPSS, Chicago, IL). RESULTS Univariate analysis of baseline variables in relation to the efficacy of MTX monotherapy. Table 1 shows the characteristics of responders and nonresponders at baseline. After 6 months of treatment with MTX, 87 patients (47%) were responders. Among these responders, 43% were receiving MTX at a dosage of 15 mg weekly, and 57% were receiving MTX at a dosage of 25 mg weekly. Sixteen variables were selected for the multivariate analysis, based on a P value of 0.1. Prior to multivariate analysis, the interaction between RF, smoking, and anti-ccp status was studied. In this context, patients were categorized into each possible combination of RF, smoking, and anti-ccp status. Data

5 MODEL TO PREDICT EFFICACY OF MTX IN RA 1769 Table 2. Comparison of de novo typed single-nucleotide polymorphisms between responders and nonresponders* Responders (n 87) Nonresponders (n 99) P MTHFD1 1958G A GG versus A allele carriers G allele carriers versus AA SHMT1 1420C T CC versus T allele carriers C allele carriers versus TT TYMS 28-bp repeat 3R/3R versus 2R repeat carriers R repeat carriers versus 2R/2R FPGS 114G A AA versus G allele carriers A allele carriers versus GG FPGS 1994A G AA versus G allele carriers A allele carriers versus GG GGH 452C T CC versus T allele carriers C allele carriers versus TT GGH 16T C TT versus C allele carriers T allele carriers versus CC * Values are the percentage. 2R2R double 28-bp tandem repeat; 3R3R triple 28-bp tandem repeat. showed a difference in MTX response for the combined variable RF and smoking status (P 0.088) but no difference (P 0.1) for combinations including anti- CCP status. Accordingly, RF-positive and RF-negative patients were divided into 2 additional groups according to smoking status. No significant interaction between age and sex was observed. However, sex and menopausal status were combined into a new variable with 3 categories based on a biologic rational: men, premenopausal women, and postmenopausal women. Therefore, a total of 15 nongenetic variables were selected for the multivariate analysis. Table 2 displays the comparison of wild-type and/or mutant allele carriers for the de novo genotyped SNPs between responders and nonresponders. Only MTHFD1 1958G A, which compared G allelic carriers with homozygous mutant AA genotypes, showed a possible trend toward a difference between responders and nonresponders. In addition, 3 of the 10 previously genotyped SNPs were associated with a good clinical response to MTX at 6 months. These SNPs were AMPD1 34C T, ITPA 94C A, and ATIC 347C G (21). Specifically, AMPD1 34T allele carriers and patients with the ITPA CC genotype and the ATIC 347CC genotype were more likely to achieve good clinical response. Thus, 4 SNPs were selected for the subsequent multivariate analysis. Multivariate analysis of baseline variables in relation to the efficacy of MTX monotherapy. The independent predicting variables resulting from the stepwise selection procedure were sex, RF status, smoking status, DAS at baseline, SJC, the HAQ score, and 4 polymorphisms in AMPD1, ATIC, ITPA, and MTHFD1. Because the SJC is a composite measure of the DAS, there is a large correlation between these 2 variables. Adding the DAS at baseline and the SJC to the model yielded, due to colinearity, coefficients that were difficult to interpret. Addition of the HAQ score also yielded coefficients that were difficult to interpret. This is likely attributable to strong correlations between the HAQ score and the DAS, as described previously (52). These coefficients did not allow the HAQ score and the SJC to be entered simultaneously with the DAS at baseline in the model. Therefore, the definite model to obtain a simple pharmacogenetic score for MTX monotherapy efficacy consisted of the independent variables sex, RF status, smoking status, DAS at baseline, and 4 polymorphisms in the AMPD1, ATIC, ITPA, and MTHFD1 genes. All factors were significantly (P 0.05) associated with the response to MTX at 6 months, and the model had an explained variance (Cox and Snell correlation coefficient) of 35%.

6 1770 WESSELS ET AL Table 3. Regression coefficients of the logistic regression model and the assigned scores for predicting efficacy of methotrexate monotherapy at 6 months* Variable Score B OR (95% CI) Female sex Premenopausal ( ) Postmenopausal ( ) Male sex 0 DAS at baseline and 5.1 2nd quartile ( ) 3rd quartile ( ) ( ) RF-negative nonsmoker 0 RF-negative smoker ( ) RF-positive nonsmoker ( ) RF-positive smoker ( ) MTHFD1 1958AA genotype ( ) AMPD1 34CC genotype ( ) ITPA 94A allele carrier ( ) ATIC 347G allele carrier ( ) Other genotype 0 *B regression coefficient in the definite model; OR odds ratio; 95% CI 95% confidence interval; DAS Disease Activity Score; RF rheumatoid factor. Higher scores represent a higher probability of nonresponse to methotrexate. Validity and predictive value of the clinical pharmacogenetic model for MTX monotherapy efficacy. In our population, the predicted probability of a response to MTX varied from to (0 responder, 1 nonresponder). The probability of response was converted into a simplified clinical score. The regression coefficients of the logistic regression model and the assigned points per variable for the simplified prediction are shown in Table 3. The scores in our population ranged from 0 to 9.5, with a lower score reflecting a higher probability of response to MTX. The cutoff values were set at 3.5 points for responders and 6 points for nonresponders. The score of 3.5 had a true positive rate of 95% (95% confidence interval [95% CI] 82 99%). The true positive rate reflects the proportion of patients with a high probability of experiencing MTX efficacy who were true responders. A score of 6 had a true negative response rate of 86% (95% CI 76 93%). This reflects the proportion of patients with a low probability of experiencing MTX efficacy who were true nonresponders. The numbers of predicted and observed patients per response category are shown in Table 4. With the clinical pharmacogenetic model, 110 patients (60%) were classified as responders or nonresponders. As shown in Figure 1, a ROC curve of the pharmacogenetic model was prepared. The discriminative ability of the model (area under the curve [AUC]) was 85% (95% CI 80 91%). Cross-validation of the definite model resulted in a ROC curve for the prediction of MTX response of 79% (95% CI 73 86%). To improve the usefulness of predicting the efficacy of MTX monotherapy for the intermediateresponse group, an additional parameter was introduced for patients with a score 3.5 but 6 points (n 74). We evaluated whether a decrease in the DAS of 1.2, 3 Table 4. Observed and predicted numbers of responders and nonresponders to MTX therapy at 6 months, using the pharmacogenetic and nongenetic models* Predicted response Observed response Nonresponder Intermediate Responder Pharmacogenetic model Nonresponder Responder Nongenetic model Nonresponder Responder * Nonresponders and responders were defined as patients having a Disease Activity Score (DAS) of 2.4 or 2.4, respectively, after receiving methotrexate (MTX) for 6 months. In the pharmacogenetic model, nonresponders were defined as having a prediction-derived score of 6, intermediate responders had a score of 3.5 and 6, and responders had a score of 3.5. Values for 2 patients are missing because of incomplete genotyping. In the nongenetic model, independent variables included sex, DAS at baseline, and rheumatoid factor status in combination with smoking status. In both models, cutoff levels were chosen based on the clinical score, which represents the probabilities of a response to MTX of (approximately) 0.80 and 0.20.

7 MODEL TO PREDICT EFFICACY OF MTX IN RA 1771 Figure 1. Receiver operating characteristic curve for predicting the response to methotrexate, including clinical and pharmacogenetic factors. Factors incorporated in the pharmacogenetic model are sex, Disease Activity Score (DAS) at baseline, rheumatoid factor (RF) status, smoking status, and genotypes for ATIC, AMPD1, ITPA, and MTHFD1. Factors incorporated in the nongenetic model are the DAS at baseline, sex, RF status, and smoking status. months after initiating MTX therapy, could improve classification of this group. Selecting patients with a predicted intermediate probability of response to MTX showed that if patients achieved a decrease of 1.2 in the DAS at 3 months, the likelihood of achieving good clinical response at 6 months was 78% (31 of 40 patients; 95% CI 62 89%). The likelihood of being a nonresponder was 76% (26 of 34 patients; 95% CI 59 89%) if patients had not achieved a decrease in the DAS of 1.2 at 3 months. Predictive value of nongenetic clinical baseline variables in relation to MTX monotherapy efficacy. To clarify whether pharmacogenetic testing adds an adequate amount of information for predicting the MTX treatment response in all patients (e.g., for patients with favorable clinical phenotypes such as nonsmoking status, RF negativity, and a DAS of 3.8), we developed a second predictive model excluding genetic variables. Using the identical selection procedure, a nongenetic model consisting of the independent predicting variables sex, DAS at baseline, and RF status in combination with smoking status was generated. To compare the discriminative ability of this nongenetic model with the pharmacogenetic model, a ROC curve was prepared (Figure 1). The comparison of ROC curves showed that the sensitivity and specificity of the model are higher when genetic information is included. The discriminative ability (AUC) of the nongenetic model was 79% (95% CI 72 85%). However, this model categorized only 60 of the patients (32%) as responders or nonresponders (Table 4). The explained variance (Cox and Snell correlation coefficient) was 25%. Replication of the clinical pharmacogenetic model for MTX monotherapy efficacy. The response at 6 months after starting MTX treatment in the group of RA patients (n 38) in the replication model was 45% (n 17), with an average MTX dosage of 19 mg weekly (range 5 25 mg weekly). Seventy-one percent of the patients in this cohort were women, and 62% were RF positive. No significant differences between the replication cohort and the BeSt population were observed for age, sex, RF status, and MTHFD, ATIC, AMPD, and ITPA genotype frequencies. The DAS at baseline in the replication cohort was 3.8, which was significantly lower than the DAS at baseline in the BeSt population (Table 1). The calculated simple scores in the replication group of RA patients ranged from 1 to 9. The true positive response rate for this cohort was 70% (7 of 10 patients;

8 1772 WESSELS ET AL 95% CI 35 93%), and the true negative response rate was 72% (13 of 18 patients; 95% CI 47 90%). In addition, 28 patients (68%) were categorized as responders and nonresponders, whereas 10 patients (32%) were categorized as intermediate responders. DISCUSSION This study demonstrates that the response to MTX therapy in patients with recent-onset RA can be predicted. In this study, we used a clinical pharmacogenetic prediction model including 8 genetic and nongenetic factors to categorize patients into 3 groups: nonresponders with a low probability of response, patients with an intermediate probability of response, and responders with a high probability of response to MTX monotherapy. This model was converted into a simple scoring system ranging from 0 to A score of 3.5 had a true positive response rate of 95%. A score of 6 had a true negative response rate of 86%. The scoring system categorized 60% of the patients as responders or nonresponders. The prediction for patients in the intermediateresponse group was further improved by introducing an additional criterion: a decrease in the DAS of 1.2 after 3 months of therapy. The positive response rate for the intermediate-response group was 78% if a decrease in the DAS of 1.2 was achieved at 3 months, whereas for patients with a change in the DAS of 1.2, the negative response rate was 75%. Prior to the application of this clinical pharmacogenetic model, several aspects must be considered. Although our current study amplifies results of prior studies by including other data, this system to identify MTX-sensitive patients must be prospectively validated in other populations, such as patients with newly diagnosed RA or patients previously treated with other DMARDs (e.g., patients with persistent RA). Additionally, positive and negative response rates depend on the response percentages in the study population. In our study, the true positive response rate with strict criteria was high. The good clinical response rate in our overall population was 47% after 6 months of treatment. However, response rates may differ in outpatient clinic settings, where treatment strategies may be less stringent. The validation of our model in a modest replication cohort supports the concept of predicting MTX monotherapy efficacy with a clinical pharmacogenetic model. The positive and negative response rates in this cohort were lower but were still 70%. However, replication in a larger prospective cohort is required. Moreover, refinement and further improvement of the prediction model are warranted. Additional research may also focus on the contribution of other factors such as inflammatory cytokine levels, because variability in drug response may also reflect changes in the complex biologic milieu in which drugs interact with their target molecules (53). Our analysis was performed in a primarily Caucasian population, excluding patients who discontinued treatment with MTX due to adverse drug events (n 10). Patients with other ethnicities may display disparate allelic frequencies (54), which may influence the derived weighted scores for response prediction. Besides, it must be recognized that our model predicts MTX efficacy based on baseline variables only, excluding patients who discontinued treatment due to toxicity. However, an additional analysis including patients who discontinued MTX due to adverse drug events as nonresponders in the model showed that the factors predicting response to MTX remained similar (data not shown). In that model, negative and positive response rates were 84% and 87%, respectively. Our model was developed to predict good clinical response as defined by a DAS 2.4 after 6 months of MTX treatment. These criteria differ slightly from the original European League Against Rheumatism criteria, because DAS improvement from baseline is not incorporated, but are in concordance with the BeSt study design (5,55). The DAS has proven to be a prognostic marker for the prevention of joint and cartilage destruction (1,15). Although our model has not shown its value to predict prolonged and sustained suppression of disease activity, other studies have already demonstrated that early suppression of disease activity is associated with better long-term outcomes (2,15). Apart from clinical factors, genetic variants are ideal markers for response, given their absolute and persistent presence at disease onset. Yet, the precise interaction between MTX and enzymes in our model is unclear, even though MTX or its polyglutamated form directly inhibits ATIC and AMPD1 and may act through the release of endogenous antiinflammatory adenosine (32). In addition, SNPs in AMPD1, ATIC, and ITPA have been proven to alter their enzyme functionality and may influence the cellular pyrimidine and purine analog balance (35,36,42,48). More importantly, our data showed no significant differences in disease activity measures at baseline or in MTX dosage at 6 months between patients with and those without favorable genotypes (data not shown). This suggests that the selected

9 MODEL TO PREDICT EFFICACY OF MTX IN RA 1773 genes are favorable only if patients are treated with MTX, and that the achievement of a good clinical response is not influenced by differences in disease activity at the time of diagnosis. However, more research is required to elucidate the MTX mechanism of action, the precise role of enzymes, and the functionality of SNPs in RA. It must be realized that other studies have shown an association between SHMT 1420C T, MTHFR 677C T, and MTHFR 1298A C with MTX efficacy (19,22). Due to our multiple stepwise backward statistical approach, only those factors contributing most to MTX efficacy at 6 months of treatment were selected. However, other molecular profiling studies have shown that several different gene profiles may reflect a common set of phenotypes (53,56). This means that depending on how the model was developed, the components of the model may differ, while a set of components still predict MTX response for the same subgroups of patients with RA. Interestingly, our analysis included sex as a factor for MTX response. Results of other studies suggest that women have higher disease activity than men due to the influence of estrogens on the immune system (11,57). Our data show a trend toward a difference between premenopausal and postmenopausal women and men (Table 3). However, other factors such as altered enzyme activity levels due to differences in gene expression or different pharmacokinetics between men and women are also plausible explanations (58,59). Finally, pharmacogenetic tests are expected to be routinely available in the near future. Nonetheless, adapting the DAS with strict cutoff values and pharmacogenetic testing in daily clinical practice may be a challenge. For this reason, models with and those without pharmacogenetic factors were compared. We conclude from this comparison that pharmacogenetic testing adds a sufficient amount of information by approximately doubling the number of patients categorized as responders or nonresponders, with increased sensitivity and specificity (Table 4). Our derived simple scoring system may assist clinicians in initiating optimal treatment at the time of disease onset by identifying patients who have a high probability of responding to MTX monotherapy. Finally, therapy recommendations based on the clinical pharmacogenetic model are suggested to assist initial treatment decisions and to enhance the clinical usefulness of the model. Table 5 displays suggestions for each category of response. Once the probability of response to MTX monotherapy is low, combination therapies are probably Table 5. Recommendations for clinical application of the simple score to select patients eligible for methotrexate monotherapy Category Score 6 Score 6 but 3.5 Score 3.5 Clinical consequence Low probability of response to methotrexate monotherapy. Consider a different diseasemodifying antirheumatic drug or a combination strategy. Intermediate probability of response to methotrexate monotherapy. Evaluate after 3 months of treatment. A. If a decrease in Disease Activity Score of 1.2, the probability of response increases with dosage escalation to 25 mg/week. B. If a decrease in the Disease Activity Score of 1.2, consider a different diseasemodifying antirheumatic drug or a combination strategy. High probability of response to methotrexate monotherapy. Dosage escalation to 25 mg/ week if necessary. the best alternative. However, it is beyond the scope of this study to conclude whether these combinations should include MTX. In spite of several restrictions, clinical use of the prediction model has the potential to be effective in improving the health status of individual patients with RA. Once prospectively validated, this model may prove that clinical and pharmacogenetic determinants for drug response can be transposed into a simple score that categorizes patients into response groups and makes tailored therapy for patients with RA possible. In conclusion, our data show that the efficacy of MTX monotherapy is predictable in patients with recent-onset RA. This pharmacogenetic model, in conjunction with clinical factors, may lead to more-informed and better-tailored initial treatment decisions in patients with RA. ACKNOWLEDGMENTS We thank the following members of the BeSt study group for their contributions to this project (all locations are in The Netherlands): W. M. de Beus (Medical Center Haaglanden, The Hague), M. H. W. de Bois (Medical Center Haaglanden, The Hague), F. C. Breedveld (Leiden University Medical Center), G. Collée (Medical Center Haaglanden, The Hague), B. A. C. Dijkmans (VU Medical Center, Amsterdam), J. A. P. M. Ewals (Haga Hospital, The Hague), A. H. Gerards (Vlietland Hospital, Schiedam), B. A. M. Grillet (De Honte Hospital, Terneuzen), J. H. L. M. van Groenendael (Fransiscus Hospital, Roosendaal), K. H. Han (MCR Zuider, Rotterdam), J. M. W. Hazes (Erasmus Medical Center, Rotterdam), H. M. J. Hulsmans (Haga Hospital, The Hague), J. M. de Jonge-Bok (Groene Hart Hospital, Gouda), P. J. S. M. Kerstens (Jan van Breemen Institute, Amsterdam), M. V. van

10 1774 WESSELS ET AL Krugten (Hospital Walcheren, Vlissingen), H. van der Leeden (retired), W. F. Lems (Slotervaart Hospital, Amsterdam), M. F. van Lieshout-Zuidema (Spaarne Hospital, Haarlem), A. Linssen (Kennemer Gasthuis, Haarlem), P. A. H. M. van der Lubbe (Vlietland Hospital, Schiedam), C. Mallée (Kennemer Gasthuis, Haarlem), H. K. Markusse (deceased), A. J. Peeters (Reinier de Graaf Groep, Delft), H. K. Ronday (Haga Hospital, The Hague), D. van Schaardenburg (Jan van Breemen Institute, Amsterdam), P. E. H. Seys (Lievensberg Hospital, Bergen op Zoom), R. M. van Soesbergen (retired), P. B. J. de Sonnaville (Oosterschelde Hospital, Goes), I. Speyer (Bronovo Hospital, The Hague), J. Ph. Terwiel (Spaarne Hospital, Haarlem), A. E. Voskuy (VU Medical Center, Amsterdam), M. L. Westedt (Bronovo Hospital, The Hague), S. ten Wolde (Kennemer Gasthuis, Haarlem), and M. H. de Jager (Albert Schweitzer Hospital, Dordrecht). AUTHOR CONTRIBUTIONS Drs. Barerra, Guchelaar, and Huizinga had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study design. Wessels, van der Kooij, Allaart, Huizinga, Guchelaar. Acquisition of data. Wessels, van der Kooij, Kievit, Barrera, Allaart. Analysis and interpretation of data. Wessels, van der Kooij, le Cessie, Barrera, Allaart, Huizinga, Guchelaar. Manuscript preparation. Wessels, van der Kooij, Barrera, Allaart, Huizinga, Guchelaar. Statistical analysis. Wessels, van der Kooij, le Cessie, Huizinga. Collection of DNA. Barrera. ROLE OF THE STUDY SPONSOR The rheumatologists participating in the Foundation for Applied Rheumatology Research were responsible for the study design and data collection in the BeSt study. The authors are responsible for the current subcohort data analysis, including genotyping, interpretation of data, manuscript preparation, and the decision to publish. Centocor and Schering-Plough did not participate in any of these activities. REFERENCES 1. Welsing PM, Landewe RB, van Riel PL, Boers M, van Gestel AM, van der Linden S, et al. The relationship between disease activity and radiologic progression in patients with rheumatoid arthritis: a longitudinal analysis. Arthritis Rheum 2004;50: Mottonen T, Hannonen P, Korpela M, Nissila M, Kautiainen H, Ilonen J, et al, for the FIN-RACo trial group. Delay to institution of therapy and induction of remission using single-drug or combination disease-modifying antirheumatic drug therapy in early rheumatoid arthritis. Arthritis Rheum 2002;46: Mottonen T, Hannonen P, Leirisalo-Repo M, Nissila M, Kautiainen H, Korpela M, et al, for the FIN-RACo trial group. Comparison of combination therapy with single-drug therapy in early rheumatoid arthritis: a randomised trial. Lancet 1999;353: Landewe RB, Boers M, Verhoeven AC, Westhovens R, van de Laar MA, Markusse HM, et al. COBRA combination therapy in patients with early rheumatoid arthritis: long-term structural benefits of a brief intervention. Arthritis Rheum 2002;46: Goekoop-Ruiterman YP, de Vries-Bouwstra JK, Allaart CF, van Zeben D, Kerstens PJ, Hazes JM, et al. Clinical and radiographic outcomes of four different treatment strategies in patients with early rheumatoid arthritis (the BeSt study): a randomized, controlled trial. Arthritis Rheum 2005;52: Matteson EL, Weyand CM, Fulbright JW, Christianson TJ, Mc- Clelland RL, Goronzy JJ. How aggressive should initial therapy for rheumatoid arthritis be? Factors associated with response to non-aggressive DMARD treatment and perspective from a 2-yr open label trial. Rheumatology (Oxford) 2004;43: Scott DL. Evidence for early disease-modifying drugs in rheumatoid arthritis. Arthritis Res Ther 2004;6: Bongartz T, Sutton AJ, Sweeting MJ, Buchan I, Matteson EL, Montori V. Anti-TNF antibody therapy in rheumatoid arthritis and the risk of serious infections and malignancies: systematic review and meta-analysis of rare harmful effects in randomized controlled trials. JAMA 2006;295: Van Everdingen AA, Jacobs JW, Siewertsz van Reesema DR, Bijlsma JW. Low-dose prednisone therapy for patients with early active rheumatoid arthritis: clinical efficacy, disease-modifying properties, and side effects: a randomized, double-blind, placebocontrolled clinical trial. Ann Intern Med 2002;136: Green M, Marzo-Ortega H, McGonagle D, Wakefield R, Proudman S, Conaghan P, et al. Persistence of mild, early inflammatory arthritis: the importance of disease duration, rheumatoid factor, and the shared epitope. Arthritis Rheum 1999;42: Tengstrand B, Ahlmen M, Hafstrom I. The influence of sex on rheumatoid arthritis: a prospective study of onset and outcome after 2 years. J Rheumatol 2004;31: Morel J, Combe B. How to predict prognosis in early rheumatoid arthritis. Best Pract Res Clin Rheumatol 2005;19: Rantapaa-Dahlqvist S, de Jong BA, Berglin E, Hallmans G, Wadell G, Stenlund H, et al. Antibodies against cyclic citrullinated peptide and IgA rheumatoid factor predict the development of rheumatoid arthritis. Arthritis Rheum 2003;48: Criswell LA, Lum RF, Turner KN, Woehl B, Zhu Y, Wang J, et al. The influence of genetic variation in the HLA DRB1 and LTA TNF regions on the response to treatment of early rheumatoid arthritis with methotrexate or etanercept. Arthritis Rheum 2004;50: Gossec L, Dougados M, Goupille P, Cantagrel A, Sibilia J, Meyer O, et al. Prognostic factors for remission in early rheumatoid arthritis: a multiparameter prospective study. Ann Rheum Dis 2004;63: Anderson JJ, Wells G, Verhoeven AC, Felson DT. Factors predicting response to treatment in rheumatoid arthritis: the importance of disease duration. Arthritis Rheum 2000;43: Symmons DP. Environmental factors and the outcome of rheumatoid arthritis. Best Pract Res Clin Rheumatol 2003;17: Symmons DP. Epidemiology of rheumatoid arthritis: determinants of onset, persistence and outcome. Best Pract Res Clin Rheumatol 2002;16: Wessels JA, de Vries-Bouwstra JK, Heijmans BT, Slagboom PE, Goekoop-Ruiterman YP, Allaart CF, et al. Efficacy and toxicity of methotrexate in early rheumatoid arthritis are associated with single-nucleotide polymorphisms in genes coding for folate pathway enzymes. Arthritis Rheum 2006;54: Dervieux T, Furst D, Lein DO, Capps R, Smith K, Walsh M, et al. Polyglutamation of methotrexate with common polymorphisms in reduced folate carrier, aminoimidazole carboxamide ribonucleotide transformylase, and thymidylate synthase are associated with methotrexate effects in rheumatoid arthritis. Arthritis Rheum 2004;50: Wessels JA, Kooloos WM, de Jonge R, de Vries-Bouwstra JK, Allaart CF, Linssen A, et al. Relationship between genetic variants in the adenosine pathway and outcome of methotrexate treatment in patients with recent-onset rheumatoid arthritis. Arthritis Rheum 2006;54: Dervieux T, Greenstein N, Kremer J. Pharmacogenomic and metabolic biomarkers in the folate pathway and their association

11 MODEL TO PREDICT EFFICACY OF MTX IN RA 1775 with methotrexate effects during dosage escalation in rheumatoid arthritis. Arthritis Rheum 2006;54: Hoekstra M, van Ede AE, Haagsma CJ, van de Laar MA, Huizinga TW, Kruijsen MW, et al. Factors associated with toxicity, final dose, and efficacy of methotrexate in patients with rheumatoid arthritis. Ann Rheum Dis 2003;62: Arnett FC, Edworthy SM, Bloch DA, McShane DJ, Fries JF, Cooper NS, et al. The American Rheumatism Association 1987 revised criteria for the classification of rheumatoid arthritis. Arthritis Rheum 1988;31: Van der Heijde DM, van t Hof MA, van Riel PL, Theunisse LM, Lubberts EW, van Leeuwen MA, et al. Judging disease activity in clinical practice in rheumatoid arthritis: first step in the development of a disease activity score. Ann Rheum Dis 1990;49: Ritchie DM, Boyle JA, McInnes JM, Jasani MK, Dalakos TG, Grieveson P, et al. Clinical studies with an articular index for the assessment of joint tenderness in patients with rheumatoid arthritis. QJM 1968;37: Felson DT, Anderson JJ, Boers M, Bombardier C, Chernoff M, Fried B, et al. The American College of Rheumatology preliminary core set of disease activity measures for rheumatoid arthritis clinical trials. Arthritis Rheum 1993;36: Ronnelid J, Wick MC, Lampa J, Lindblad S, Nordmark B, Klareskog L, et al. Longitudinal analysis of citrullinated protein/peptide antibodies (anti-cp) during 5 year follow up in early rheumatoid arthritis: anti-cp status predicts worse disease activity and greater radiological progression. Ann Rheum Dis 2005;64: Fries JF, Spitz P, Kraines RG, Holman HR. Measurement of patient outcome in arthritis. Arthritis Rheum 1980;23: Ulrich CM, Robien K, Sparks R. Pharmacogenetics and folate metabolism: a promising direction. Pharmacogenomics 2002;3: Krajinovic M, Moghrabi A. Pharmacogenetics of methotrexate. Pharmacogenomics 2004;5: Cronstein BN. Low-dose methotrexate: a mainstay in the treatment of rheumatoid arthritis. Pharmacol Rev 2005;57: Huizinga TW, Pisetsky DS, Kimberly RP. Associations, populations, and the truth: recommendations for genetic association studies in Arthritis & Rheumatism. Arthritis Rheum 2004;50: Hattersley AT, McCarthy MI. What makes a good genetic association study? Lancet 2005;366: Marie S, Heron B, Bitoun P, Timmerman T, Van den Berghe G, Vincent MF. AICA-ribosiduria: a novel, neurologically devastating inborn error of purine biosynthesis caused by mutation of ATIC. Am J Hum Genet 2004;74: Cao H, Hegele RA. DNA polymorphisms in ITPA including basis of inosine triphosphatase deficiency. J Hum Genet 2002;47: Marinaki AM, Ansari A, Duley JA, Arenas M, Sumi S, Lewis CM, et al. Adverse drug reactions to azathioprine therapy are associated with polymorphism in the gene encoding inosine triphosphate pyrophosphatase (ITPase). Pharmacogenetics 2004;14: Weisman MH, Furst DE, Park GS, Kremer JM, Smith KM, Wallace DJ, et al. Risk genotypes in folate-dependent enzymes and their association with methotrexate-related side effects in rheumatoid arthritis. Arthritis Rheum 2006;54: Kohlmeier M, da Costa KA, Fischer LM, Zeisel SH. Genetic variation of folate-mediated one-carbon transfer pathway predicts susceptibility to choline deficiency in humans. Proc Natl Acad Sci U S A 2005;102: Skibola CF, Smith MT, Hubbard A, Shane B, Roberts AC, Law GR, et al. Polymorphisms in the thymidylate synthase and serine hydroxymethyltransferase genes and risk of adult acute lymphocytic leukemia. Blood 2002;99: Chave KJ, Ryan TJ, Chmura SE, Galivan J. Identification of single nucleotide polymorphisms in the human -glutamyl hydrolase gene and characterization of promoter polymorphisms. Gene 2003;319: Kalsi KK, Yuen AH, Rybakowska IM, Johnson PH, Slominska E, Birks EJ, et al. Decreased cardiac activity of AMP deaminase in subjects with the AMPD1 mutation: a potential mechanism of protection in heart failure. Cardiovasc Res 2003;59: Gaughan DJ, Kluijtmans LA, Barbaux S, McMaster D, Young IS, Yarnell JW, et al. The methionine synthase reductase (MTRR) A66G polymorphism is a novel genetic determinant of plasma homocysteine concentrations. Atherosclerosis 2001;157: Dervieux T, Kremer J, Lein DO, Capps R, Barham R, Meyer G, et al. Contribution of common polymorphisms in reduced folate carrier and -glutamylhydrolase to methotrexate polyglutamate levels in patients with rheumatoid arthritis. Pharmacogenetics 2004;14: Bosco P, Gueant-Rodriguez RM, Anello G, Barone C, Namour F, Caraci F, et al. Methionine synthase (MTR) 2756 (A 3 G) polymorphism, double heterozygosity methionine synthase 2756 AG/ methionine synthase reductase (MTRR) 66 AG, and elevated homocysteinemia are three risk factors for having a child with Down syndrome. Am J Med Genet A 2003;121: Stranzl T, Wolf J, Leeb BF, Smolen JS, Pirker R, Filipits M. Expression of folylpolyglutamyl synthetase predicts poor response to methotrexate therapy in patients with rheumatoid arthritis. Clin Exp Rheumatol 2003;21: Cheng Q, Wu B, Kager L, Panetta JC, Zheng J, Pui CH, et al. A substrate specific functional polymorphism of human -glutamyl hydrolase alters catalytic activity and methotrexate polyglutamate accumulation in acute lymphoblastic leukaemia cells. Pharmacogenetics 2004;14: Morisaki T, Gross M, Morisaki H, Pongratz D, Zollner N, Holmes EW. Molecular basis of AMP deaminase deficiency in skeletal muscle. Proc Natl Acad Sci U S A 1992;89: Welsing PM, van Riel PL. The Nijmegen inception cohort of early rheumatoid arthritis. J Rheumatol Suppl 2004;69: Klareskog L, Stolt P, Lundberg K, Kallberg H, Bengtsson C, Grunewald J, et al. A new model for an etiology of rheumatoid arthritis: smoking may trigger HLA DR (shared epitope) restricted immune reactions to autoantigens modified by citrullination. Arthritis Rheum 2006;54: Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med 1996;15: Drossaers-Bakker KW, de Buck M, van Zeben D, Zwinderman AH, Breedveld FC, Hazes JM. Long-term course and outcome of functional capacity in rheumatoid arthritis: the effect of disease activity and radiologic damage over time. Arthritis Rheum 1999; 42: Roden DM, Altman RB, Benowitz NL, Flockhart DA, Giacomini KM, Johnson JA, et al. Pharmacogenomics: challenges and opportunities. Ann Intern Med 2006;145: Hughes LB, Beasley TM, Patel H, Tiwari HK, Morgan SL, Baggott JE, et al. Racial/ethnic differences in allele frequencies of single nucleotide polymorphisms in the methylenetetrahydrofolate reductase gene and their influence on response to methotrexate in rheumatoid arthritis. Ann Rheum Dis 2006;65: Van Riel PL, van Gestel AM, van de Putte LB. Development and validation of response criteria in rheumatoid arthritis: steps towards an international consensus on prognostic markers. Br J Rheumatol 1996;35 Suppl 2: Massague J. Sorting out breast-cancer gene signatures. N Engl J Med 2007;356: Forslind K, Hafstrom I, Ahlmen M, Svensson B. Sex: a major predictor of remission in early rheumatoid arthritis? Ann Rheum Dis 2007;66: Jawaheer D, Lum RF, Gregersen PK, Criswell LA. Influence of male sex on disease phenotype in familial rheumatoid arthritis. Arthritis Rheum 2006;54: Harris RZ, Benet LZ, Schwartz JB. Gender effects in pharmacokinetics and pharmacodynamics. Drugs 1995;50: