Poor early response to methotrexate portends inadequate long-term outcomes in patients with moderate-to-severe psoriasis: Evidence from 2 phase 3 clinical trials

ORIGINAL

ARTICLE

Poor early response to methotrexate portends inadequate long-term outcomes in patients with moderate-to-severe psoriasis: Evidence from 2 phase 3 clinical trials Kenneth B. Gordon, MD,a Keith A. Betts, PhD,b Murali Sundaram, PhD,c James E. Signorovitch, PhD,b Junlong Li, PhD,b Meng Xie, BS,b Eric Q. Wu, PhD,b and Martin M. Okun, MD, PhDc,d Milwaukee, Wisconsin; Boston, Massachusetts; North Chicago, Illinois; and Fort Atkinson, Wisconsin Background: Most methotrexate-treated psoriasis patients do not achieve a long-term PASI75 (75% reduction from baseline Psoriasis Area and Severity Index score) response. Indications of nonresponse can be apparent after only 4 weeks of treatment. Objective: To develop a prediction rule to identify patients unlikely to respond adequately to methotrexate. Methods: Patient-level data from CHAMPION (NCT00235820, N = 110) was used to construct a prediction model for week 16 PASI75 by using patient baseline characteristics and week 4 PASI25. A prediction rule was determined on the basis of the sensitivity and specificity and validated in terms of week 16 PASI75 response in an independent validation sample from trial M10-255 (NCT00679731, N = 163). Results: PASI25 achievement at week 4 (odds ratio = 8.917) was highly predictive of response with methotrexate at week 16. Patients with a predicted response probability \30% were recommended to discontinue methotrexate. The rates of week 16 PASI75 response were 65.8% and 21.1% (P \ .001) for patients recommended to continue and discontinue methotrexate, respectively. Limitations: The CHAMPION trial excluded patients previously treated with biologics, and the M10-255 trial had no restrictions. Conclusion: A prediction rule was developed and validated to identify patients unlikely to respond adequately to methotrexate. The rule indicates that 4 weeks of methotrexate might be sufficient to predict long-term response with limited safety risk. ( J Am Acad Dermatol http://dx.doi.org/10.1016/j.jaad.2017.08.017.) Key words: discontinuation; methotrexate; moderate-to-severe psoriasis; Psoriasis Area and Severity Index (PASI); prediction; response.

From the Medical College of Wisconsin, Milwaukeea; Analysis Group Inc, Bostonb; AbbVie Inc, North Chicagoc; and Fort HealthCare, Fort Atkinson.d Funding sources: Supported by AbbVie. AbbVie participated in the interpretation of data, review, and approval of the manuscript. Conflicts of interest: Dr Gordon receives research support from AbbVie, Amgen, Jannsen, Eli Lilly, Boerhinger Ingelheim, Dermira, and Celgene and serves as consultant for AbbVie, Amgen, Boerhinger Ingelheim, Jannsen, Eli Lilly, Celgene, Pfizer, Sun Pharmaceuticals, and Novartis. Drs Betts, Signorovitch, Li, and Wu and Ms Xie are employed by Analysis Group Inc, which received payment from AbbVie Inc for participation in this research. Dr Sundaram is a former employee of AbbVie and

may own AbbVie stock or stock options. Dr Okun is a former employee of AbbVie and serves as a consultant for AbbVie, Gilead Science, and Crescendo Biosciences. Accepted for publication August 8, 2017. Correspondence to: Keith A. Betts, PhD, 111 Huntington Ave, 14th floor, Boston, Massachusetts 02199. E-mail: keith.betts@ analysisgroup.com. Published online October 6, 2017. 0190-9622/$36.00 Ó 2017 by the American Academy of Dermatology, Inc. http://dx.doi.org/10.1016/j.jaad.2017.08.017

1

J AM ACAD DERMATOL

2 Gordon et al

n 2017

Methotrexate is one of the most commonly utidetermining whether methotrexate would be benelized systemic treatments for psoriasis, having been ficial for them. successfully used for [50 years in the clinical setting.1,2 Randomized controlled trials have demonMETHODS strated that methotrexate has lower efficacy than Study population cyclosporine but can be used continuously for Data from the CHAMPION randomized controlled years.3-7 Recommendations for the dosing, schedtrial was used to develop the model, and data from uling, monitoring, and folic the M10-255 randomized acid supplementation apcontrolled trial was used to CAPSULE SUMMARY proaches vary. Generally, evaluate the model. Subjects methotrexate is given as a randomized to methotrexate Most methotrexate-treated psoriasis single weekly oral dose. in CHAMPION were assigned patients fail to achieve a PASI75 (75% Often low doses are given the following dosing rereduction from baseline Psoriasis Area initially, and then adjusted gimen: 7.5 mg on weeks 0-1, Severity Index score) response. based on the individual pa10 mg on weeks 2-3, and A parsimonious scoring algorithm to tient’s disease control and 15 mg on weeks 4-7, with predict a patient’s probability of side effect profile.8 dose escalation to 20 mg on response to methotrexate was Limitations of methoweeks 8-11 and 25 mg on developed and validated. trexate treatment for psoriaweeks 12-15 if PASI50 (50% sis include slow onset of reduction from baseline PASI) This decision rule allows for early action, modest efficacy, and response was not achieved at identification of patients unlikely to toxicity. Rare side effects weeks 8 or 12. Subjects ranrespond adequately to methotrexate. include hepatotoxicity, hemdomized to methotrexate in atotoxicity, and allergic M10-255 were assigned the pneumonitis, and common side effects include following dosing regimen: 5 mg on week 0, 10 mg nausea, malaise, diarrhea, and headaches.9 on week 1, and 15 mg on weeks 2-9, with dose Identifying the characteristics of patients unlikely to escalation to 20 mg on weeks 10-15 and 25 mg on respond to methotrexate would allow clinicians to weeks 16-24 if PASI75 (75% reduction from baseline limit patient exposure to this medication, thereby PASI) response was not achieved or Physician’s avoiding the risk for toxicity and unnecessary delays Global Assessment (PGA) score was worse than in initiating more effective treatment. The benefitminimal on weeks 10 or 16. This study analyzed the risk balance of methotrexate would be enhanced intent-to-treat (ITT) population for the methotrexate because the patients who continue methotrexate arm in each trial. This study was a retrospective would have an enhanced probability of response. analysis of deidentified clinical trial data, and did The CHAMPION clinical trial (NCT00235820) not require institutional review board approval. was a phase 3, randomized, double-blind trial that compared the treatments adalimumab, oral methoModel development and validation trexate, and placebo for subjects with moderate-toThe primary outcome was PASI75 response at severe psoriasis.10 The M10-255 clinical trial week 16. A patient’s PASI score is a measure of body (NCT00679731) was a phase 3, randomized, surface area involvement and psoriasis severity14; double-blind clinical trial that compared methoPASI75 is defined as $75% improvement in PASI trexate and briakinumab.11 By using the methoscore from baseline. trexate treatment arms in these 2 controlled clinical A prediction model for PASI75 response with trials, we developed and evaluated a prediction methotrexate at week 16 was developed by using model for a patient’s probability of response to the patients randomized to methotrexate in methotrexate on the basis of patient characteristics CHAMPION. The prediction model included baseand early clinical improvements in Psoriasis Area line patient and disease characteristics, as well as and Severity Index (PASI).12,13 Even though initial achievement of PASI25 (25% reduction from baseline methotrexate doses were low and increased gradPASI) at week 4 (an early indication of patient ually in these trials, outcomes at week 4 proved to response to methotrexate). Model selection was be a powerful predictor of long-term treatment performed via best subsets regression and clinical success. In addition, the risk of serious hepatotoxic input. The best prediction model was chosen on the or hematologic adverse events was low at the basis of the area under the curve (AUC) and the methotrexate doses used up to week 4, so patients Hosmer-Lemeshow test15 in M10-255. The full set of experienced little risk of acute toxicity while considered baseline characteristics is shown in d

d

d

J AM ACAD DERMATOL

Gordon et al 3

VOLUME jj, NUMBER j

Abbreviations used: AUC: DLQI: ITT: PASI: PASI25: PASI75: PGA: PROs: PtGA: ROC: VAS:

area under the curve Dermatology Life Quality Index intent-to-treat Psoriasis Area and Severity Index 25% reduction from baseline Psoriasis Area Severity Index score 75% reduction from baseline Psoriasis Area Severity Index score Physician’s Global Assessment patient-reported outcomes Patient’s Global Assessment receiver operating characteristic visual analog scale

Table I. Prediction models containing different numbers of covariates were considered to examine the tradeoff between complexity and predictive ability in possible models. Prediction rule and subgroup analysis By using the prediction model built using the CHAMPION trial data, the receiver operating characteristic (ROC) curve was plotted to show the sensitivity (true positive rate) and 1 e specificity (false positive rate) at each probability cutoff value in the M10-255 trial. A true positive was defined as a patient with predicted probability of treatment response above the cutoff value who was actually observed to respond to methotrexate at week 16. A false positive was defined as a patient with predicted probability of treatment response above the cutoff value who was not observed to respond to methotrexate at week 16. The optimal cutoff based on sensitivity and specificity was chosen as the point closest to the upper left corner of the ROC plot (the upper left corner corresponds to true positive rate of 100% and false positive rate of 0%). A prediction rule was then determined. Patients were predicted to respond and recommended to continue if their predicted probability of response to methotrexate at week 16 was equal to or above the optimal cutoff, and patients were predicted to not respond and recommended to discontinue if their predicted probability of response to methotrexate at week 16 was below the optimal cutoff. Study measures The following clinical efficacy outcomes and patient-reported outcomes (PROs) at week 16 and week 24 were compared between patients who were recommended to continue and patients recommended to discontinue methotrexate: observed PASI75, Dermatology Life Quality Index (DLQI), visual analog scale (VAS) for plaque psoriasis pain, VAS

Table I. Baseline characteristics and efficacy outcomes of the methotrexate arms in CHAMPION and M10-255 Category

Baseline demographics Age, years, mean 6 SD Female sex, n (%) Weight, kg, mean 6 SD Psoriasis duration, years, mean 6 SD Psoriatic arthritis, n (%) Psoriatic arthritis duration, years, mean 6 SD Prior therapy Biologic, n (%) Nonbiologic, n (%) Week 4 clinical characteristics PASI score, mean 6 SD Physician Global Assessment Very severe, n (%) Severe, n (%) Moderate, n (%) Mild, n (%) Minimal, n (%) Clear, n (%) Outcomes at week 4 PASI75 achievement, n (%) PASI25 achievement, n (%) PGA clearance, n (%) Outcomes at week 16 PASI75 achievement, n (%)

CHAMPION

M10-255

(N = 110)

(N = 163)

41.6 6 12.0 37 (33.6) 83.1 6 17.5 19.0 6 10.2

43.1 6 12.9 52 (31.9) 82.0 6 18.6 19.1 6 11.3

19 (17.3) 8.9 6 11.0

28 (17.2) 7.2 6 5.1

11 (10.0) 48 (43.6)

34 (20.9) 89 (54.6)

15.2 6 8.1

13.5 6 6.9

5 28 58 15 3 1

(4.5) (25.5) (52.7) (13.6) (2.7) (0.9)

2 44 87 25 3 2

(1.2) (27.0) (53.4) (15.3) (1.8) (1.2)

3 (2.7)

6 (3.7)

45 (40.9)

79 (48.5)

4 (3.6)

5 (3.1)

39 (35.5)

63 (38.7)

PASI, Psoriasis Area Severity Index; PASI25, 25% reduction from baseline Psoriasis Area Severity Index score; PASI75, 75% reduction from baseline Psoriasis Area Severity Index score; PGA, Physician’s Global Assessment; SD, standard deviation.

for psoriatic arthritis pain, psoriasis-related pruritus, and Patient’s Global Assessment (PtGA). Continuous variables were compared by using 2-sample t tests, and categorical variables were compared by using chi-squared tests or Fisher’s exact tests. In addition, grade 3 elevations (based on Common Toxicity Criteria version 3.0) in selected hepatic and hematologic parameters were evaluated through the first 4 weeks of methotrexate treatment.

RESULTS Patient characteristics in CHAMPION and M10255 Patients were similar at baseline across the 2 trials (Table I) with the exception that M10-255 had greater

J AM ACAD DERMATOL

4 Gordon et al

n 2017

Table II. Prediction model for treatment response to methotrexate at week 16 (scored by PASI75 achievement) Full model Parameter

Model training parameters Intercept Baseline demographics Age in years Female sex Weight in kg Psoriasis duration in years Prior use of biologic therapy Week 4 clinical characteristics PASI25 achievement Model validation statistics (out of sample)

Estimate

0.205

Simple model Odds ratio

Estimate

Odds ratio

-

0.390

-

0.018 0.531 e0.021 e0.052 e1.647

1.018 1.700 0.980 0.949 0.193

0.002

1.002

e0.024

0.977

e1.323

0.266

2.363

10.627

2.188

8.917

Estimate

P value

Estimate

P value

0.818 8.332

.402

0.80 10.91

.21

C-statistic/area under the curve Hosmer-Lemeshow

The odds ratio is a measure of the association between a model covariate and the outcome. An odds ratio [1 indicates that a 1-unit increase in the covariate increases the odds of treatment response to methotrexate. PASI, Psoriasis Area Severity Index; PASI25, 25% reduction from baseline Psoriasis Area Severity Index score; PASI75, 75% reduction from baseline Psoriasis Area Severity Index score.

proportions of patients with previous biologic therapy and previous nonbiologic therapy. Through the first 4 weeks of methotrexate treatment, 4 of 273 (1.5%) patients experienced grade 3 elevations in liver function parameters, 2 of 273 experienced grade 3 elevations in hematologic parameters, and no patients fulfilled the criteria for Hy’s law. Prediction models The following model was determined to be predictive of week 16 patient response, defined as PASI75 achievement (hereafter referred to as the full model): logit ðprobability of week 16 PASI75 response with methotrexateÞ ¼ 0:205 þ ð2:363 3 week 4 PASI25 achievementÞ ð0:021 3 weightÞ ð0:052 3 psoriasis durationÞþ ð0:531 3 femaleÞþð0:018 3 ageÞ ð1:647 3 prior use of biologic therapyÞ Early clinical improvement measured by PASI25 achievement at week 4 was highly predictive of response with methotrexate at week 16 (P \ .001). The model demonstrated excellent discrimination (AUC = 0.82) and calibration (Hosmer-Lemeshow test, P = .40) when validated with the M10-255 trial (Table II). A second model was developed using the same method with the additional constraint that there had to be #4 covariates. The following simpler model

could be more suitable for use by physicians (hereafter referred to as the simple model): logit ðprobability of week 16 PASI75 response with methotrexateÞ ¼ 0:390þ ð2:188 3 week 4 PASI25 achievementÞ ð0:024 3 weightÞþð0:002 3 ageÞ ð1:323 3 prior use of biologic therapyÞ Again, PASI25 achievement at week 4 was highly predictive of response with methotrexate at week 16 (P \ .001). The model demonstrated excellent discrimination (AUC = 0.80) and calibration (Hosmer-Lemeshow test, P = .21) when validated with the M10-255 trial (Table II). PASI25 was the most predictive covariate in the full and simple models and, therefore, had potential to be a strong predictor on its own. Therefore, week 4 PASI25 achievement alone was used as a rule to predict week 16 PASI75. Those who achieved PASI25 at week 4 were predicted to achieve PASI75 at week 16 and were thus recommended to continue, and others were predicted to be nonresponders and were recommended to discontinue. Evaluation of prediction rules The predicted probability of response to methotrexate at week 16 was calculated for all patients in the M10-255 trial using the prediction models (including both full and simple models developed by using the CHAMPION data). Patients with a predicted probability of response greater than the

J AM ACAD DERMATOL

VOLUME jj, NUMBER j

Fig 1. Out-of-sample ROC curve of prediction model for treatment response to methotrexate (MTX). The ROC curve corresponds to the full model: week 16 treatment response to MTX = intercept 1 week 4 PASI25 achievement 1 weight 1 psoriasis duration 1 female sex 1 age 1 prior use of biologic therapy. The colors indicate the probability threshold used for the binary prediction. For example, a threshold of 0.5 indicates that binary predictions were made by classifying patients scoring [0.5 as treatment responders and otherwise as treatment nonresponders. The ROC curve plots the sensitivity and 1  specificity for all possible thresholds. The prediction model for treatment response to MTX was developed using the best subset regression model, ie, choosing a subset of potential covariates (baseline and week 4 clinical characteristics). The goodness of the fit of the models were assessed on the basis of the area under the curve (AUC) and Hosmer-Lemeshow tests. The outcome for the prediction model was treatment response to MTX, defined as an achievement of a $75% reduction in PASI score from baseline at week 16. The model was created with the intent-to-treat (ITT) population from the CHAMPION trial (N = 110 MTX-treated patients) for whom the baseline covariates and the outcome data were available. The validation sample ROC curve was evaluated with the ITT population from the M10-255 trial (N = 163 MTX-treated patients). The blue point in the validation sample ROC curve shows the optimal combination of sensitivity and specificity for prediction in the full model, as measured by the minimum distance to the upper left corner of the graph. The corresponding probability threshold used for binary prediction is 0.30. The purple point in the validation sample ROC curve shows the sensitivity and specificity of prediction using the 0.30 threshold in the simple model: week 16 treatment response to MTX = intercept 1 week 4 PASI25 achievement 1 weight 1 age 1 prior use of biologic therapy. The red point in the validation sample ROC curve shows the sensitivity and specificity of the decision rule in which week 4 PASI25 achievers are predicted to be week 16 responders, and all others are predicted to be nonresponders. PASI25, 25% reduction from baseline Psoriasis Area and Severity Index score; ROC, receiver operating characteristic.

Gordon et al 5

optimal cutoff (30%; as determined by balancing true positive and true negative rates from the ROC analysis, Fig 1) were recommended to continue methotrexate therapy. When using the full model, 77 patients were recommended to continue and 86 recommended to discontinue methotrexate. Predictions using the simple model and PASI25 only were generally consistent with the full model. When using the simple model, 73 patients were recommended to continue and 90 recommended to discontinue methotrexate. The kappa statistics for concordance16 between the full and simple model predictions was 0.852 (P \.001). When applying the prediction rule of PASI25 achievement, 79 patients were recommended to continue and the rest (n = 84) recommended to discontinue methotrexate. Plotting the sensitivity and 1  specificity for each the full model, simple model, and PASI25 alone shows that the full model resulted in the best prediction model, followed by the simple model, and then PASI25 alone (Fig 1). Thus, a more complex model resulted in incremental improvements in prediction (full model: sensitivity 0.791, specificity 0.750). However, the performance of all 3 models was similar, suggesting that the simple model (sensitivity 0.716, specificity 0.740) and PASI25 (sensitivity 0.731, specificity 0.688) could be a good parsimonious option. Week 4 PASI25 in particular can be used as an easy-to-use and model-free prediction rule that provides most of the predictive performance of the multivariate models. Patient baseline characteristics by subgroup Patients recommended to continue had significantly lower baseline disease activity measured by PASI, PtGA (category of uncontrolled disease), and PGA (category of severe disease); had less psoriasisrelated pruritus; and were less likely to have had prior systemic therapy (Table III). The group of patients recommended to continue versus the group recommended to discontinue had a significantly greater proportion of women and had a lower baseline weight. Long-term efficacy outcomes and PROs by subgroup The proportion of patients who achieved PASI75 among those recommended to continue methotrexate (68.8% at week 16) was significantly greater than the proportion of patients who achieved PASI75 among those recommended to discontinue (16.3% at week 16) (P \ .001) at both week 16 and 24 (Table IV). The PASI75 response rate among those recommended to continue methotrexate was 30.1 percentage points higher than the week 16 PASI75

J AM ACAD DERMATOL

6 Gordon et al

n 2017

Table III. Subgroup analysis results for select baseline characteristics in the methotrexate arm of M10-255 stratified by week 4 prediction of week 16 PASI75 achievement, based on the optimal prediction rule Recommended to continue

Recommended to discontinue

(N = 77)

(N = 86)

P value

35 (45.5) 74.7 6 16.0 4 (5.2)

17 (19.8) 88.6 6 18.4 30 (34.9)

.0004 \.0001 \.0001

10.1 6 6.4 6 (7.8)

16.5 6 5.8 0 (0)

\.0001 .0100

Characteristics

Baseline demographics Female sex, n (%) Weight, kg, mean 6 SD Prior biologic therapy, n (%) Week 4 clinical characteristics PASI score, mean 6 SD PASI75 achievement, n (%) PGA Very severe, n (%) Severe, n (%) Moderate, n (%) Mild, n (%) Minimal, n (%) Clear, n (%) PGA clearance, n (%) PROs Psoriasis-related pruritus at week 0, mean 6 SD PtGA at week 4 Uncontrolled disease, n (%) Limited disease control, n (%) Good disease control, n (%) Complete disease control, n (%)

1 7 41 23 3 2 5

(1.3) (9.1) (53.2) (29.9) (3.9) (2.6) (6.5)

(1.2) (43.0) (53.5) (2.3) (0) (0) (0)

1.0000 \.0001 .9754 \.0001 .1032 .2216 .0219

3.4 6 2.3

4.4 6 2.7

.0083

5 35 33 3

28 29 26 3

(6.6) (46.1) (43.4) (3.9)

1 37 46 2 0 0 0

(32.6) (33.7) (30.2) (3.5)

\.0001 .1091 .0817 1.0000

Among those recommended to continue, there was a missing observation for psoriasis-related pruritus and PtGA. PASI, Psoriasis Area Severity Index; PASI75, 75% reduction from baseline Psoriasis Area Severity Index score; PROs, patient-reported outcomes; PGA, Physician’s Global Assessment; PtGA, Patient’s Global Assessment; SD, standard deviation.

response rate for all subjects randomized to methotrexate in M10-255. Patients recommended to continue generally had better PROs at week 16 and week 24. In particular, differences in DLQI were significant at both time points. Patients recommended to discontinue performed worse in both clinical measures and PROs and continued to perform worse in the longer term (at week 24). Therefore, it would likely have been more beneficial for these patients if they had been discontinued from methotrexate at week 4.

DISCUSSION In this post-hoc analysis of clinical trial data, a model consisting of age, sex, weight, psoriasis duration, prior use of biologic therapy, and week 4 PASI25 achievement was determined to be predictive of week 16 patient response (PASI75 achievement). On the basis of the prediction model, patients with a \30% probability of response to methotrexate at week 16 should consider alternative therapies. As a simpler rule, patients who do not achieve PASI25 at week 4 should consider an alternative treatment approach. This result is similar to the findings of a post-hoc analysis of ixekizumab (an interleukin 17

monoclonal antibody tested in moderate-to-severe psoriasis), which found early clinical improvement in disease symptoms at weeks 4 and 6 to be predictive of later achievement of PASI75.17 The ixekizumab study recommended assessing early response using PASI40 at week 4. Applying this threshold to the current analysis of methotrexate would increase the true positive fraction (the percentage of patients recommended to continue who actually respond), but such a threshold may be too high for methotrexate and lead to some patients (who would have responded to methotrexate) prematurely switching. Regardless of the threshold, the findings of the current study and the ixekizumab study underscore the value of early clinical monitoring in establishing treatment futility efficiently. In addition, in a retrospective study, an association between serum calcium level and the efficacy of methotrexate for the treatment of severe plaque psoriasis was identfied.18 To verify this association in the current study, a sensitivity analysis was conducted to include serum calcium level in the full model. The serum calcium level was found to be neither be significantly associated with PASI75 response nor change the effect of other predictors.

J AM ACAD DERMATOL

Gordon et al 7

VOLUME jj, NUMBER j

Table IV. Subgroup analysis of PROs and long-term outcomes in the methotrexate arm of M10-255 stratified by week 4 prediction of week 16 PASI75 achievement, based on the optimal prediction rule Recommended to continue

Recommended to discontinue

(N = 77)

(N = 86)

P value

53 (68.8) 3.5 6 4.0 9.4 6 14.5 17.8 6 18.1

14 (16.3) 6.7 6 6.2 16.5 6 22.6 26.0 6 24.6

\.0001 .0001 .0171 .3285

Outcome

PASI75 at week 16 and PROs at week 12 PASI75 achievement, n (%) DLQI, mean 6 SD VAS for plaque psoriasis pain, mean 6 SD VAS for psoriatic arthritis pain, mean 6 SD PtGA Uncontrolled disease, n (%) Limited disease control, n (%) Good disease control, n (%) Complete disease control, n (%) PASI75 and PROs at week 24 PASI75 achievement, n (%) DLQI, mean 6 SD VAS for plaque psoriasis pain, mean 6 SD VAS for psoriatic arthritis pain, mean 6 SD PtGA Uncontrolled disease, n (%) Limited disease control, n (%) Good disease control, n (%) Complete disease control, n (%)

2 21 46 8

(2.6) (27.3) (59.7) (10.4)

49 (63.6) 3.2 6 4.6 11.1 6 16.7 19.3 6 16.9 4 19 39 15

(5.2) (24.7) (50.6) (19.5)

19 33 30 4

(22.1) (38.4) (34.9) (4.7)

21 (24.4) 5.7 6 5.5 16.5 6 22.8 24.1 6 26.9 18 33 26 9

(20.9) (38.4) (30.2) (10.5)

.0002 .1328 .0015 .1613 \.0001 .0030 .0828 .5816 .0033 .0611 .0079 .1049

DLQI, Dermatology Life Quality Index; PASI, Psoriasis Area Severity Index; PASI25, 25% reduction from baseline Psoriasis Area Severity Index score; PASI75, 75% reduction from baseline Psoriasis Area Severity Index score; PROs, patient-reported outcomes; PtGA, Patient’s Global Assessment; SD, standard deviation; VAS, visual analog scale.

The current study has some limitations. Although the populations of the CHAMPION and M10-255 trials were similar, the CHAMPION trial excluded patients with previous exposure to systemic antitumor necrosis factor therapies, including thalidomide, infliximab, etanercept, and adalimumab. The M10-255 trial had no such restriction. The prediction model was developed in a separate independent trial (CHAMPION) from the one the prediction rule was evaluated in (M10-255). The current study used the dosing schedules of methotrexate presented in the CHAMPION and M10-255 trials to predict the week 16 PASI75 response. Although a dosing schedule of low starting doses and gradual dose increments is widely used in the prescription of methotrexate, other dosing schedules may be implemented. Further research could assess the accuracy of the prediction rule in alternative clinical trials and in patients prescribed different methotrexate dosing schedules. Across both studies, the incidence of methotrexate-treated patients experiencing serious (Common Toxicity Criteria grade 3) abnormalities of liver function or hematologic parameters was 2.2%. The incidence was similar for subjects in the 2 studies (though the dosing regimens differed slightly). Only subjects who received at least 15 mg methotrexate

weekly experienced these abnormalities. According to our model, dermatologists could reliably assess if patients are ultimately unlikely to be PASI75 responders by evaluating their response at week 4, by which time those subjects adhering to the CHAMPION dosing regimen (7.5 mg at weeks 0 and 1, 10 mg at weeks 2 and 3) would not yet have up-titrated past a methotrexate dosage of 10 mg per week. These results suggest that a 4-week trial of methotrexate, which had a low risk for acute hepatotoxicity or hematotoxicity, might suffice to portend whether patients are likely to have a satisfactory longer-term response. However, this inference is based on a relatively small number of study subjects, and may not be generalizable to patients with multiple or severe medical comorbidities who might not have qualified for clinical trial enrollment. In conclusion, a parsimonious scoring algorithm to predict a patient’s probability of response to methotrexate and assist with treatment decisionmaking was developed and validated on the basis of patient characteristics and early clinical improvements. Patients predicted to not reach PASI75 and recommended to discontinue methotrexate at week 4 were likely to have suboptimal clinical response and PROs on average and might benefit from early switching in the longer term. In combination with

J AM ACAD DERMATOL

8 Gordon et al

their clinical knowledge, physicians can use this algorithm to make educated treatment decisions regarding patients unlikely to respond to methotrexate. REFERENCES 1. Edmundson WF, Guy WB. Treatment of psoriasis with folic acid antagonists. AMA Arch Derm. 1958;78(2):200-203. 2. Smith KC. Systemic therapy of psoriasis using methotrexate. Skin Ther Lett. 2000;6(3):1-2; 5. 3. Dogra S, Krishna V, Kanwar AJ. Efficacy and safety of systemic methotrexate in two fixed doses of 10 mg or 25 mg orally once weekly in adult patients with severe plaque-type psoriasis: a prospective, randomized, double-blind, dose-ranging study. Clin Exp Dermatol. 2012;37(7):729-734. 4. Fallah Arani S, Neumann H, Hop WC, Thio HB. Fumarates vs. methotrexate in moderate to severe chronic plaque psoriasis: a multicentre prospective randomized controlled clinical trial. Br J Dermatol. 2011;164(4):855-861. 5. Flytstrom I, Stenberg B, Svensson A, Bergbrant IM. Methotrexate vs. ciclosporin in psoriasis: effectiveness, quality of life and safety. A randomized controlled trial. Br J Dermatol. 2008; 158(1):116-121. 6. Heydendael VM, Spuls PI, Opmeer BC, et al. Methotrexate versus cyclosporine in moderate-to-severe chronic plaque psoriasis. N Engl J Med. 2003;349(7):658-665. 7. Hsu S, Papp KA, Lebwohl MG, et al. Consensus guidelines for the management of plaque psoriasis. Arch Dermatol. 2012; 148(1):95-102. 8. Pathirana D, Ormerod AD, Saiag P, et al. European S3-guidelines on the systemic treatment of psoriasis vulgaris. J Eur Acad Dermatol Venereol. 2009;23(Suppl 2):1-70. 9. Barker J, Horn EJ, Lebwohl M, et al. Assessment and management of methotrexate hepatotoxicity in psoriasis patients:

n 2017

10.

11.

12.

13.

14. 15. 16. 17.

18.

report from a consensus conference to evaluate current practice and identify key questions toward optimizing methotrexate use in the clinic. J Eur Acad Dermatol Venereol. 2011; 25(7):758-764. Saurat JH, Stingl G, Dubertret L, et al. Efficacy and safety results from the randomized controlled comparative study of adalimumab vs. methotrexate vs. placebo in patients with psoriasis (CHAMPION). Br J Dermatol. 2008;158(3): 558-566. Reich K, Langley RG, Papp KA, et al. A 52-week trial comparing briakinumab with methotrexate in patients with psoriasis. N Engl J Med. 2011;365(17):1586-1596. Erder MH, Signorovitch JE, Setyawan J, Yang H, Parikh K, Betts KA, Xie J, Hodgkins P, Wu EQ. Identifying patient subgroups who benefit most from a treatment: using administrative claims data to uncover treatment heterogeneity. J Med Econ. 2012;15(6):1078-1087. Setyawan J, Yang H, Cheng D, Cai X, Signorovitch J, Xie J, Erder M H. Developing a risk score to guide individualized treatment selection in attention deficit/hyperactivity disorder. Value Health. 2015;18(6):824-831. Fredriksson T, Pettersson U. Severe psoriasiseoral therapy with a new retinoid. Dermatologica. 1978;157(4):238-244. Hosmer DW Jr, Lemeshow S, Sturdivant RX. Applied logistic regression. John Wiley & Sons; 2013:398. Cohen J. A coefficient of agreement for nominal scales. Educ Psychol Meas. 1960;20:37-46. Zhu B, Edson-Heredia E, Cameron GS, et al. Early clinical response as a predictor of subsequent response to ixekizumab treatment: results from a phase II study of patients with moderate-to-severe plaque psoriasis. Br J Dermatol. 2013; 169(6):1337-1341. Zhai Z, Chen L, Yang H, et al. Can pretreatment serum calcium level predict the efficacy of methotrexate in the treatment of severe plaque psoriasis? J Am Acad Dermatol. 2015;73(6): 991-997.e993.