Supplementary Appendix

Transcription

1 Supplementary Appendix This appendix has been provided by the authors to give readers additional information about their work. Supplement to: Gladman D, Rigby W, Azevedo VF, et al. Tofacitinib for psoriatic arthritis in patients with an inadequate response to TNF inhibitors. N Engl J Med 2017;377: DOI: /NEJMoa

2 Tofacitinib in Psoriatic Arthritis with Inadequate Response to TNF Inhibitors D. Gladman, 1 W. Rigby, 2 V.F. Azevedo, 3 F. Behrens, 4 R. Blanco, 5 A. Kaszuba, 6 E. Kudlacz, 7 C. Wang, 7 S. Menon, 7 T. Hendrikx, 8 and K.S. Kanik 7 1 University of Toronto, Toronto Western Hospital, Toronto, Canada; 2 Geisel School of Medicine at Dartmouth, Dartmouth Hitchcock Medical Center, Lebanon, NH, USA; 3 Universidade Federal do Paraná, Curitiba, Brazil; 4 Goethe University and Fraunhofer IME Project Group Translational Medicine and Pharmacology TMP, Frankfurt, Germany; 5 Hospital Universitario Marques de Valdecilla, IDIVAL, Avda Valdecilla, Santander, Spain; 6 Specjalistyczne Gabinety Lekarskie DERMED, Lodz, Poland; 7 Pfizer Inc, Groton, CT, USA; 8 Pfizer Inc, Collegeville, PA, USA Supplementary information Inclusion and exclusion criteria Page 4 Protocol amendments after study commencement Page 6 Secondary efficacy endpoints Page 7 Adjudication of safety events Page 8 Sample size calculations Page 9 Hierarchical procedures for declaring significance Page 10 Supportive analyses for primary endpoints Page 12 Figure S1. American College of Rheumatology 20% response rate and change Page 14 from baseline in Health Assessment Questionnaire-Disability Index at month 3. Figure S2. American College of Rheumatology 20% response rate and change Page 16 from baseline in Health Assessment Questionnaire-Disability Index at month 3 by number of prior TNFi inadequate responses 1

3 Figure S3. Change from baseline in American College of Rheumatology Page 17 response component scores throughout the study to month 6 Figure S4. American College of Rheumatology 50% and 70% response rates Page 19 throughout the study to month 6 Figure S5. Psoriatic Arthritis Response Criteria and Psoriasis Area and Page 21 Severity Index response rates throughout the study to month 6 Figure S6. Change from baseline in Leeds Enthesitis Index, SpondyloArthritis Page 23 Research Consortium of Canada enthesitis index, and Dactylitis Severity Score throughout the study to month 6 Figure S7. Change from baseline in Functional Assessment of Chronic Illness Page 25 Therapy-Fatigue and Short Form-36 Version 2 Physical Functioning domain score throughout the study to month 6 Figure S8. Change from baseline in EuroQoL 5-Dimensions domain scores Page 26 throughout the study to month 6 Table S1. Patient enrollment by region, country, and study site for sites that Page 28 enrolled at least five patients Table S2. Difference from placebo (95% CI) at month 3 for selected efficacy Page 30 endpoints (full analysis set) Table S3. Additional baseline demographics and characteristics (safety Page 33 analysis set) Table S4. Additional efficacy endpoints and patient-reported outcomes at Page 35 month 3 and month 6 (full analysis set) Table S5. Most common adverse events (all causality); reported in 5% of Page 38 patients in any treatment group (safety analysis set) 2

4 Table S6. Laboratory parameter values at month 3 and month 6 (safety Page 39 analysis set) References Page 41 3

5 Inclusion and exclusion criteria Eligible patients were aged 18 years ( 20 years in Taiwan); had a diagnosis of psoriatic arthritis for 6 months; fulfilled ClASification criteria for Psoriatic Arthritis (CASPAR) at screening; 1 had active plaque psoriasis (diagnosed or confirmed by a dermatologist or a sponsor-approved rheumatologist) at screening and active arthritis ( 3 tender/painful joints and 3 swollen joints) at screening and baseline; and had an inadequate response to 1 tumor necrosis factor inhibitor (TNFi; lack of efficacy and/or treatment-related adverse event determined by or reported to the physician and recorded on the case report form). Patient exclusion criteria included current non-plaque forms of psoriasis (except nail psoriasis); current or recent history of severe, progressive, or uncontrolled renal, hepatic, hematologic, gastrointestinal, metabolic, endocrine, pulmonary, cardiovascular, or neurologic disease; evidence of active, latent, or inadequately controlled Mycobacterium tuberculosis infection; blood dyscrasias within 3 months of first study drug dose (including confirmed hemoglobin <10g/dL, white blood cell count <3.0 x 10 9 /L, absolute neutrophil count 1.5 x 10 9 /L, absolute lymphocyte count <1.0 x 10 9 /L, platelet count <100 x 10 9 /L); aspartate aminotransferase or alanine aminotransferase >1.5 x upper limit of normal at screening; estimated creatinine clearance <40 ml/min; history of any autoimmune rheumatic disease other than psoriatic arthritis; history of lymphoproliferative disorder; history of recurrent herpes zoster, disseminated herpes zoster, or disseminated herpes simplex; history of active infection requiring hospitalization or parenteral antimicrobial therapy within 6 months prior to first study drug dose; current or history of malignancy (except adequately treated or excised non-metastatic basal or squamous cell cancer of the skin or cervical carcinoma in situ); and prior treatment with a non-b-cell-specific lymphocyte-depleting agent. Permitted conventional synthetic disease-modifying antirheumatic drugs (DMARDs): 4

6 Methotrexate: maximum dose 20 mg/week. Patients must have received methotrexate for at least 4 months prior to, and been on a stable dose for at least 4 weeks prior to, first dose of study drug. Sulfasalazine: maximum dose 3 g/day. Patients must have received sulfasalazine for at least 2 months prior to, and been on a stable dose for at least 4 weeks prior to, first dose of study drug. Leflunomide: maximum dose 20 mg/day. Patients must have received leflunomide for at least 4 months prior to, and been on a stable dose for at least 4 weeks prior to, first dose of study drug. Other conventional synthetic DMARDs were assessed on a case-by-case basis by the Study Investigator and Sponsor. Whilst an inadequate response to prior TNFi was required, patients had to have discontinued TNFi before the first dose of study drug. Washout periods were as follows: Etanercept, 4 weeks; Adalimumab, 10 weeks; Infliximab, 8 weeks; Golimumab, 10 weeks; Certolizumab, 10 weeks. 5

7 Protocol amendments after study commencement An amendment was made to the protocol on December 13, 2013, and included the addition of minimum treatment duration and washout information for certolizumab per marketing approval for psoriatic arthritis, in addition to standardization of the washout period for other biologic agents. The exclusion criteria were updated to include localized infection and patients at risk of gastrointestinal perforation. In addition, the use of sexual abstinence as a contraceptive method only when consistent with the patient s preferred and usual lifestyle was added. A second protocol amendment was made on October 03, 2014, to clarify the use of prohibited medications and rescue medication use prior to study visits. The final protocol amendment was made on February 01, 2015, and included an additional contraceptive requirement for women of childbearing potential in Canada based upon the Health Canada Guidance document, Considerations for Inclusion of Women in Clinical Trials and Analysis of Sex Differences. The washout period for prohibited medications was also clarified, following a regulatory request. Information was added in the Adjudicated Safety Events and Safety Event Review Committee sections, i.e. inclusion of gastrointestinal perforation review committee. 6

8 Secondary efficacy endpoints PASI75 is defined as 75% improvement in Psoriasis Area and Severity Index score from baseline, and is measured over 3 domains (erythema, induration, scaling), each scored 0 4, with 4 indicating more severe symptoms. PASI examines 4 body regions (head and neck; hands and arms; chest, abdomen and back [trunk]; buttocks, thighs, and legs), each scored 0 6 for the area affected. Psoriatic Arthritis Response Criteria is assessed over 4 domains: tender/painful joint count (68 joints), swollen joint count (66 joints), Physician s Global Assessment of Arthritis (VAS), and Patient s Global Assessment of Arthritis (VAS). Response is based on improvement in two of the criteria, one of which must be the number of tender/painful or swollen joints, without worsening of any criteria. Dactylitis Severity Score (DSS) is measured by scoring the tenderness of each digit on a scale 0 3, giving a maximum score of 60. A score of 3 indicates extreme tenderness. Enthesitis, as measured by the SpondyloArthritis Research Consortium of Canada score is assessed across 16 sites, and the Leeds Enthesitis Index (LEI) is assessed across 6 sites. For both, higher scores indicate a greater number of sites with enthesitis. Functional Assessment of Chronic Illness Therapy-Fatigue (FACIT-F) comprises 13 items to assess fatigue. The score ranges from 0 to 52 with higher scores indicating less fatigue. The Short Form-36 Version 2 (SF-36v2) is a 36-item generic health status measure across 8 general health domains that can be combined into mental or physical component summary scores. EuroQol 5- Dimension Health State Profile is a five-domain questionnaire. Each domain is scored from 0 3, with higher score showing greater impairment. 7

9 Adjudication of safety events External, independent adjudication committees made up of experts in each field were established to standardize the assessment of specific safety events. These included cardiovascular endpoints, opportunistic infections, malignancies, gastrointestinal, and hepatic events. Members of each adjudication committee were blinded to treatment assignment, to enable unbiased assessments. Each committee was provided with an adjudication package containing information on each potential event to determine whether these events met a consistent set of clinical criteria for classification. Potential events were initially reviewed by two blinded assessors. A third assessment was made in the case of disagreement between the first two assessors. Cases of herpes zoster were adjudicated as opportunistic infections if they were disseminated (nature of dissemination was designated), multidermatomal (non-adjacent or >2 adjacent dermatomes), or had two adjacent dermatomes. Adjudicated cardiovascular events of myocardial infarction, cerebrovascular event (stroke), or cardiovascular death (coronary, cerebrovascular, cardiac) were considered major adverse cardiovascular events. 8

10 Sample size calculations Sample size calculations for the primary endpoints were based on randomization in a 2:2:1:1 ratio to tofacitinib 5 mg twice daily (BID) (planned N=130), tofacitinib 10 mg BID (N=130), placebo advancing to 5 mg BID (N=65), or placebo advancing to 10 mg BID (N=65), where the two placebo treatment sequences were combined. For American College of Rheumatology 20% improvement (ACR20) response rate, assuming a placebo response rate of 15%, a sample size of 130 patients per treatment group was expected to provide 84% power for a difference between tofacitinib and placebo of at least 15% and 97% power for a difference of at least 20%. For Health Assessment Questionnaire-Disability Index, a sample size of 130 patients per treatment group was expected to provide 98% power for a difference between two means of at least 0.3, assuming standard deviation of

11 Hierarchical procedures for declaring significance Three families of hierarchical testing procedures were used. The first family included the primary endpoints as described in the main body of the manuscript and a set of key secondary endpoints at month 3. The Type 1 error was controlled globally within this family. Upon achieving significance for the primary endpoints (4 tests), a hierarchical procedure for declaring significance for key secondary endpoints was based on testing tofacitinib 10 mg BID versus placebo, followed by tofacitinib 5 mg BID versus placebo for each endpoint in the following order: PASI75, change from baseline (Δ) in LEI, ΔDSS, ΔSF-36v2 Physical Functioning domain, ΔFACIT-F (3 endpoints in order of testing priority: FACIT-F total score, FACIT-F experience domain score, and FACIT-F impact domain score). ACR response family of endpoints at month 3 ACR50 and ACR70 can be viewed as extensions of ACR20 and all belong to the ACR family of endpoints; therefore, a hierarchical approach to testing the ACR20, ACR50, and ACR70 response rates at month 3 was used for each endpoint within each dose and doses within each endpoint. Tofacitinib 10 mg BID at a given endpoint could only achieve significance if it was significant at the prior endpoint; tofacitinib 5 mg BID at a given endpoint could only achieve significance if it was significant at the prior endpoint and if tofacitinib 10 mg BID was significant at the same endpoint. Though this testing scheme did not protect the Type I error for the family of all possible comparisons, it provided Type I error protection for testing the family of ACR endpoints within each dose and doses within each ACR endpoint. Establishing onset of efficacy with ACR20 (ACR20 response time course) To be more rigorous in establishing the onset of efficacy, a hierarchical approach was used 10

12 for the ACR20 response rate from 3 months to earlier time points (order of testing: month 3, 2, 1, and week 2). Tofacitinib 10 mg BID at a given time point could only achieve significance if it was also significant at the prior time point. Tofacitinib 5 mg BID could only achieve significance if it was significant at the prior time point and tofacitinib 10 mg BID was significant at the same time point. Although this testing scheme did not protect the Type I error for the family of all possible comparisons, it provided Type I error protection for testing the family of ACR20 time points within each dose and doses within each time point. 11

13 Supportive analyses for primary endpoints ACR20 response rates As a supportive analysis to the primary analysis for ACR20 response, a generalized marginal linear model for ACR20 from week 2 through month 3 was used. This model included fixed effects for visit (discrete: week 2, month 1, month 2, month 3), treatment (tofacitinib 5 mg BID, tofacitinib 10 mg BID, placebo), and time by treatment interaction; the dependent variable was logit of the probability of response, without explicit imputation for missing values (though implicitly imputed by using the repeated measures model). A common first-order autoregressive (AR[1]) variance-covariance matrix for all treatment groups was used to model the variability among observations within a patient. This analysis assumed patients remained on therapy after discontinuation from the study. From this model, the estimated ACR20 response rates at month 3 were 52.6%, 50.1%, and 26.2% for tofacitinib 5 and 10 mg BID and placebo, respectively. The 95% confidence interval for the odds ratio in ACR20 response between each tofacitinib dose and placebo excluded 1. The results are supportive of the primary conclusion. Change from baseline in HAQ-DI As a supportive analysis to the primary analysis for the change from baseline in HAQ-DI, a missing not at random multiple imputation approach was used. 2 The specific multiple imputation implemented applied the jump to reference imputation approach to the active treatment groups but missing at random imputation approach to the combined placebo group. The joint distribution of a patient s data across the time points up to and including month 3 was assumed to be multivariate normal. Each treatment group was assumed to have different sets of means. For dropouts randomized to an active treatment group, applying jump to 12

14 reference, the means of the multivariate normal distributions across time for each dropout were equal to those estimated for that active group prior to dropping out, and equal to those estimated for the placebo group post-dropout. For each imputation, the parameters of the multivariate normal distributions were sampled from the posterior distributions using Markov chain Monte Carlo methods. The distribution of each patient s imputed values, conditional on his or her past values, was hence multivariate normal and a function of the patient s past values and the sampled parameters. The effect of past values on the imputed values depended on the strength of the correlation and the residuals (past values minus their expected values). For active group dropouts, the residuals were computed using their respective active group means at the time points prior to dropout. As a result, all the dropouts in the active groups would rapidly lose the estimated average treatment effect attributed to their respective active treatment after dropping out. While measures at time points prior to month 3 were used, once multiple complete datasets (1000 datasets) were generated through imputation, an ANCOVA model was used to analyze just the data through month 3 for treatment comparison between each tofacitinib dose and placebo. Inference was performed according to the Rubin s rule 3 for combining the multiple estimates and standard errors. From this analysis, the least squares mean changes from baseline in HAQ-DI at month 3 were -0.40, -0.36, and for tofacitinib 5 and 10 mg BID, and placebo, respectively. The 95% confidence interval for the difference in least squares means at month 3 between each tofacitinib dose and placebo excluded 0. The results were supportive of the primary conclusion. This analysis assumed patients remained receiving a conventional synthetic DMARD (but not tofacitinib) after discontinuation from the study. 13

15 Figure S1. American College of Rheumatology 20% response rate and change from baseline in Health Assessment Questionnaire-Disability Index at month 3. Shown are response rates for ACR20 and change from baseline in HAQ-DI at month 3 (Panel A and B, respectively), for patients who received tofacitinib 5 mg BID, tofacitinib 10 mg BID, or placebo. All data are shown for the full analysis set. Missing ACR20 data were imputed as non-response to treatment (patients with non-missing ACR20 response at month 3: placebo, n=116; tofacitinib 5 mg BID, n=124; tofacitinib 10 mg BID, n=120). No imputation was applied for missing HAQ-DI data (patients with non-missing data at month 3: placebo, n=117; tofacitinib 5 mg BID, n=124; tofacitinib 10 mg BID, n=120; number of patients included in 14

16 repeated measures analysis, respectively, n=131, n=129, and n=132). Asterisks represent the comparison with placebo, with * indicating nominal P 0.05, ** indicating nominal P 0.01, and *** indicating nominal P 0.001; achieved statistical significance at P 0.05 vs. placebo for global Type 1 error control per the pre-specified stepdown testing procedure. Δ, change from baseline; ACR20, 20% improvement in American College of Rheumatology criteria; BID, twice daily; HAQ-DI, Health Assessment Questionnaire-Disability Index; LS, least squares; SE, standard error. 15

17 Figure S2. American College of Rheumatology 20% response rate and change from baseline in Health Assessment Questionnaire-Disability Index at month 3 by number of prior TNFi inadequate responses. Shown are response rates for ACR20 (A) and change from baseline HAQ-DI (B) at month 3 for patients who received tofacitinib 5 mg BID, tofacitinib 10 mg BID, or placebo by number of inadequate responses to prior TNFi. All data shown are for the full analysis set. Missing ACR20 data were imputed as non-response to treatment. No imputation was applied for missing HAQ-DI data. Δ, change from baseline; ACR, American College of Rheumatology; BID, twice daily; HAQ-DI, Health Assessment Questionnaire-Disability Index; IR, inadequate response; LS, least squares; TNFi, tumor necrosis factor inhibitor; SE, standard error. 16

18 Figure S3. Change from baseline in American College of Rheumatology response component scores throughout the study to month 6. 17

19 Shown are the LS mean change from baseline in swollen joint count (A), tender/painful joint count (B), PtGA of arthritis (VAS, mm) (C), PGA of arthritis (VAS, mm) (D), CRP (mg/l) (E), and PAAP (VAS, mm) (F) throughout the study to month 6 for patients who received tofacitinib 5 mg BID, tofacitinib 10 mg BID, or placebo advancing to tofacitinib 5 mg or 10 mg BID at month 3 (placebo treatment sequences combined for visits at or prior to month 3). All data shown are for the full analysis set. No imputation was applied for missing data. The dashed line at month 3 indicates the end of the placebo-controlled period. Δ, change from baseline; BID, twice daily; hscrp, high sensitivity C-reactive protein; LS, least squares; PAAP, Patient s Assessment of Arthritis Pain; PGA, Physician s Assessment of Arthritis; PtGA, Patient s Assessment of Arthritis; SE, standard error; VAS, visual analog scale. 18

20 Figure S4. American College of Rheumatology 50% and 70% response rates throughout the study to month 6. Shown are response rates for ACR50 (A) and ACR70 (B) throughout the study to month 6, for patients who received tofacitinib 5 mg BID, tofacitinib 10 mg BID, or placebo advancing to tofacitinib 5 mg or 10 mg BID at month 3 (placebo treatment sequences combined for visits at or prior to month 3). All data are shown for the full analysis set. Missing values were imputed as non-reponse to treatment. The dashed line at month 3 indicates the end of the placebo-controlled period. Asterisks represent the comparison with placebo, with ** indicating 19

21 nominal P 0.01; achieved statistical significance at P 0.05 per the pre-specified step-down testing procedure for Type 1 error control within the ACR family responses. ACR50/70, 50% or 70% improvement in American College of Rheumatology domains; BID, twice daily; SE, standard error. 20

22 Figure S5. Psoriatic Arthritis Response Criteria and Psoriasis Area and Severity Index response rates throughout the study to month 6. Shown are response rates for PsARC (A) and PASI75 (B) throughout the study to month 6 for patients who received tofacitinib 5 mg BID, tofacitinib 10 mg BID, or placebo advancing to tofacitinib 5 mg or 10 mg BID at month 3 (placebo treatment sequences combined for visits at or prior to month 3). All data shown are for the full analysis set. Missing data were imputed as non-response to treatment. The dashed line at month 3 indicates the 21

23 end of the placebo-controlled period. Asterisks represent the comparison with placebo, with *** indicating nominal P 0.001; achieved statistical significance at P 0.05 per the pre-specified step-down testing procedure for global Type 1 error control. BID, twice daily; PASI75, 75% improvement in Psoriasis Area and Severity Index; PsARC, Psoriatic Arthritis Response Criteria; SE, standard error. 22

24 Figure S6. Change from baseline in Leeds Enthesitis Index, SpondyloArthritis Research Consortium of Canada enthesitis index, and Dactylitis Severity Score throughout the study to month 6. 23

25 Shown are the LS mean change from baseline in LEI (A), SPARCC enthesitis index (B), and DSS (C) throughout the study to month 6, for patients with a baseline score greater than 0 who received tofacitinib 5 mg BID, tofacitinib 10 mg BID, or placebo advancing to tofacitinib 5 mg or 10 mg BID at month 3 (placebo treatment sequences combined for visits at or prior to month 3). No imputation was applied for missing data. The dashed line at month 3 indicates the end of the placebo-controlled period. Δ, change from baseline; BID, twice daily; DSS, Dactylitis Severity Score; LEI, Leeds Enthesitis Index; LS, least squares; SE, standard error; SPARCC, SpondyloArthritis Research Consortium of Canada. 24

26 Figure S7. Change from baseline in Functional Assessment of Chronic Illness Therapy- Fatigue and Short Form-36 Version 2 Physical Functioning domain score throughout the study to month 6. Shown are change from baseline FACIT-F total score (A) and SF-36v2 Physical Functioning domain score (B) throughout the study to month 6 for patients who received tofacitinib 5 mg BID, tofacitinib 10 mg BID, or placebo advancing to tofacitinib 5 mg or 10 mg BID at month 3 (placebo treatment sequences combined for visits at or prior to month 3). All data shown are for the full analysis set. No imputation was applied for missing data. An increase in domain score represents an improvement in that domain. The dashed line at month 3 indicates the end of the placebo-controlled period. Δ, change from baseline; BID, twice daily; FACIT-F, Functional Assessment of Chronic Illness Therapy- Fatigue; LS, least squares; SE, standard error; SF-36v2, Short Form-36 Version 2, acute. 25

27 Figure S8. Change from baseline in EuroQoL 5-Dimensions domain scores throughout the study to month 6. 26

28 Shown are change from baseline EQ-5D mobility domain score (A), self-care domain score (B), usual activities domain score (C), pain/discomfort domain score (D), and anxiety/depression domain score (E) throughout the study to month 6 for patients who received tofacitinib 5 mg BID, tofacitinib 10 mg BID, or placebo advancing to tofacitinib 5 mg or 10 mg BID at month 3 (placebo treatment sequences combined for visits at or prior to month 3). All data shown are for the full analysis set. No imputation was applied for missing data. The dashed line at month 3 indicates the end of the placebo-controlled period. A reduction in domain score represents an improvement in that domain. Δ, change from baseline; BID, twice daily; EQ-5D, EuroQol 5-Dimension Health State Profile; LS, least squares; SE, standard error. 27

29 Table S1. Patient enrollment by region, country, and study site for sites that enrolled at least five patients. Region Country Study site Number of patients randomized United United States States/Canada (N=118) (N=118) Western Europe/Australia (N=122) Australia (N=16) Belgium (N=17) France (N=2) Germany (N=40) Spain (N=23) United Kingdom (N=24) Czech Republic (N=2) 28

30 Eastern Europe/Russian Federation (N=89) Poland (N=54) Russian Federation (N=31) Slovakia (N=2) Rest of World (N=66) Brazil (N=11) Mexico (N=46) Taiwan (N=9) Total centers screened 546 patients. 98 centers enrolled 395 patients into the study. N, number of patients enrolled. 29

31 Table S2. Difference from placebo (95% CI) at month 3 for selected efficacy endpoints (full analysis set). Endpoint Difference vs. placebo (95% CI) Tofacitinib 5 mg BID (N=131) Tofacitinib 10 mg BID (N=132) ACR20 response rate (%) 26.0 (14.7, 37.2) 23.3 (12.1, 34.5) LS mean change from baseline in HAQ-DI -0.3 (-0.4, -0.1) -0.2 (-0.3, -0.1) PASI75 response rate (%) * 7.3 (-4.3, 18.9) 29.3 (16.2, 42.3) LS mean change from baseline in LEI LS mean change from baseline in DSS LS mean change from baseline in SF-36v2 Physical Functioning domain LS mean change from baseline in FACIT-F -0.9 (-1.4, -0.3) -0.8 (-1.3, -0.3) -3.3 (-5.4, -1.2) -3.5 (-5.7, -1.4) 3.3 (1.3, 5.3) 2.4 (0.4, 4.4) 3.9 (1.6, 6.2) 2.8 (0.5, 5.1) ACR50 response rate (%) 15.3 (5.4, 25.2) 13.5 (3.8, 23.3) ACR70 response rate (%) 6.9 (-1.3, 15.1) 4.5 (-3.4, 12.4) LS mean change from baseline in swollen joint count LS mean change from baseline in tender joint count LS mean change from baseline in PtGA (mm) LS mean change from baseline in PGA (mm) LS mean change from baseline in hscrp (mg/l) -4.9 (-6.5, -3.2) -4.0 (-5.7, -2.3) -5.4 (-8.1, -2.7) -5.2 (-7.9, -2.5) (-20.7, -8.3) (-19.0, -6.5) (-16.7, -6.1) (-18.4, -7.7) -6.5 (-11.5, -1.4) -6.9 (-12.0, -1.9) 30

32 LS mean change from baseline in PAAP (mm) Patients achieving decrease from baseline 0.35 in HAQ-DI (%) Patients achieving absence of enthesitis (by LEI) (%), Patients achieving absence of dactylitis, (%), LS mean change from baseline SPARCC enthesitis index (-20.0, -7.9) (-19.2, -7.1) 22.4 (10.2, 34.6) 13.1 (1.2, 25.0) 18.3 (4.8, 31.7) 10.8 (-1.6, 23.3) 22.9 (6.5, 39.4) 22.2 (5.7, 38.7) -1.2 (-2.2, -0.3) -1.5 (-2.5, -0.6) Patients achieving MDA (%) 8.4 (-1.0, 17.8) 6.7 (-2.5, 15.9) Patients achieving PsARC (%) 29.8 (18.3, 41.2) 19.5 (7.9, 31.0) LS mean change from baseline in DAS28-3(CRP) -0.8 (-1.0, -0.5) -0.6 (-0.9, -0.3) LS mean change from baseline in EQ-5D domain scores Mobility -0.1 (-0.2, -0.01) -0.1 (-0.2, 0.01) Self-care -0.1 (-0.2, 0.00) -0.1 (-0.2, 0.01) Usual activities -0.1 (-0.2, 0.03) -0.1 (-0.2, 0.1) Pain/discomfort -0.2 (-0.3, -0.1) -0.2 (-0.3, -0.1) Anxiety/depression -0.1 (-0.2, 0.1) -0.1 (-0.2, 0.04) * In patients with baseline BSA 3% and baseline PASI >0. In patients with baseline score >0. Among patients with baseline score Post hoc analyses. Missing values for ACR20, ACR50, ACR70, PASI75, decrease from baseline 0.35 in HAQ-DI, enthesitis absence, dactylitis absence, MDA, and PsARC, were considered as 31

33 non-response. LS means and 95% CI for the continuous endpoints were calculated based on a mixed model for repeated measures without imputation for missing values. Enthesitis absence is defined as a resolution of enthesitis (LEI), i.e. absence of enthesitis in all of the 6 assessed sites; dactylitis absence is defined as having a resolution of dactylitis (DSS), i.e. absence of dactylitis in all of the 20 assessed digits. ACR20/50/70, 20%, 50%, or 70% improvement in American College of Rheumatology domains; BID, twice daily; BSA, body surface area; CI, confidence interval; DAS28-3(CRP), Disease Activity Score in 28 joints with C-reactive protein; DSS, Dactylitis Severity Score; EQ-5D, EuroQol 5-Dimension Health State Profile; FACIT-F, Functional Assessment of Chronic Illness Therapy-Fatigue; HAQ-DI, Health Assessment Questionnaire-Disability Index; hscrp, high-sensitivity C-reactive protein; LEI, Leeds Enthesitis Index; LS, least squares; MDA, Minimal Disease Activity; PAAP, Patient s Assessment of Arthritis Pain; PASI, Psoriasis Area and Severity Index; PASI75, 75% improvement in PASI; PGA, Physician s Global Assessment of arthritis; PsARC, Psoriatic Arthritis Response Criteria; PtGA, Patient s Global Assessment of arthritis; SF-36v2, Short Form-36 Version 2; SPARCC, SpondyloArthritis Research Consortium of Canada. 32

34 Table S3. Additional baseline demographics and characteristics (safety analysis set*). Placebo (N=131) Tofacitinib 5 mg BID (N=131) Tofacitinib 10 mg BID (N=132) Patient demographics Mean weight, kg (SD) 82.1 (16.2) 87.9 (22.9) 87.1 (19.7) Mean BMI, kg/m 2 (SD) 29.5 (5.5) 30.5 (7.1) 31.0 (6.7) Baseline disease characteristics Presence of enthesitis, SPARCC >0, n (%) Mean SPARCC enthesitis index score (SD) Mean PGA of arthritis, VAS, mm (SD) Mean PtGA of arthritis, VAS, mm (SD) 100 (76.3) 96 (73.3) 108 (81.8) 5.4 (3.5) 5.8 (4.1) 6.9 (4.6) 53.7 (21.2) 53.5 (20.9) 55.8 (20.8) 55.8 (23.7) 57.4 (22.9) 58.5 (22.3) Mean PAAP, VAS, mm (SD) 54.9 (25.3) 56.4 (24.1) 59.5 (22.3) Mean DAS28-3(CRP), (SD) 4.4 (1.0) 4.5 (1.0) 4.7 (1.2) Median hscrp, mg/l (range) 4.4 ( ) 5.7 ( ) 4.9 ( ) Prior TNFi use, n (%) Adalimumab 67 (51.1) 67 (51.1) 56 (42.4) Certolizumab 6 (4.6) 13 (9.9) 19 (14.4) Etanercept 65 (49.6) 69 (52.7) 66 (50.0) Golimumab 26 (19.8) 26 (19.8) 27 (20.5) Infliximab 35 (26.7) 45 (34.4) 45 (34.1) * All patients who received 1 dose of study medication. Among patients with baseline score >0. 33

35 Nominal P values were determined using chi-squared tests for categorical variables and Kruskal-Wallis tests for continuous variables. Nominally significant differences among treatment groups were reported for prior use of certolizumab (nominal P=0.0262); all other differences were not significant. BID, twice daily; BMI, body mass index; DAS28-3(CRP), Disease Activity Score in 28 joints with C-reactive protein; hscrp, high sensitivity C-reactive protein; PAAP, Patient s Assessment of Arthritis Pain; PGA, Physician s Global Assessment of arthritis; PtGA, Patient s Global Assessment of arthritis; SD, standard deviation; SPARCC, SpondyloArthritis Research Consortium of Canada; TNFi, tumor necrosis factor inhibitor; VAS, visual analog scale. 34

36 Table S4. Additional efficacy endpoints and patient-reported outcomes at month 3 and month 6 (full analysis set*). Month 3 (Active treatment vs. combined placebo) Month 6 Placebo Tofacitinib Tofacitinib Placebo Placebo Tofacitinib Tofacitinib 5 mg BID 10 mg BID tofacitinib tofacitinib 5 mg BID 10 mg BID 5 mg BID 10 mg BID (N=131) (N=131) (N=132) (N=66) (N=65) (N=131) (N=132) Patients achieving decrease from baseline 0.35 in HAQ-DI, n/n1 (%) Patients achieving absence of enthesitis (by LEI),, n/n1 (%) Patients achieving absence of dactylitis,, n/n1 (%) 32/116 (27.6) 58/116 (50.0) 50/123 (40.7) 20/93 (21.5) 33/83 (39.8) 32/99 (32.3) 15/47 (31.9) 19/46 (41.3) 37/83 (44.6) 43/99 (43.4) 18/63 (28.6) 34/66 (51.5) 33/65 (50.8) 16/30 (53.3) 15/33 (45.5) 38/66 (57.6) 36/65 (55.4) LS mean change from baseline -1.3 (0.3) -2.5 (0.3) -2.8 (0.3) -2.6 (0.5) -2.4 (0.5) -2.6 (0.4) -3.1 (0.4) in SPARCC enthesitis index [87] [92] [96] [40] [43] [91] [93] (SE) [N1] 35

37 Patients achieving minimal disease activity, n (%) Patients achieving Psoriatic Arthritis Response Criteria, n (%) 19 (14.5) 30 (22.9) 28 (21.2) 12 (18.2) 19 (29.2) 31 (23.7) 31 (23.5) 38 (29.0) 77 (58.8) 64 (48.5) 34 (51.5) 36 (55.4) 77 (58.8) 68 (51.5) LS mean change from baseline -0.6 (0.1) -1.4 (0.1) -1.2 (0.1) -1.5 (0.1) -1.5 (0.1) -1.6 (0.1) -1.4 (0.1) in DAS28-3(CRP), (SE) [N1] [117] [123] [118] [56] [55] [123] [113] LS mean change from baseline in EQ-5D domain scores (SE) [N1] Mobility (0.04) (0.04) (0.04) (0.06) (0.06) (0.04) (0.04) [117] [124] [120] [56] [55] [122] [112] Self-care (0.04) (0.04) (0.04) (0.06) (0.06) (0.04) (0.04) [117] [122] [120] [56] [55] [120] [112] Usual activities (0.04) (0.04) (0.04) (0.07) (0.07) -0.3 (0.05) (0.05) [117] [124] [120] [56] [55] [122] [112] Pain/discomfort (0.04) (0.04) (0.04) (0.07) (0.07) (0.05) (0.05) 36

38 [117] [124] [120] [56] [55] [121] [112] Anxiety/depression (0.05) [117] (0.04) [124] (0.05) [120] (0.07) [56] (0.07) [55] (0.05) [122] (0.05) [112] Enthesitis absence is defined as a resolution of enthesitis (LEI), i.e. absence of enthesitis in all of the 6 assessed sites; dactylitis absence is defined as having a resolution of dactylitis (DSS), i.e. absence of dactylitis in all of the 20 assessed digits. Missing values for patients achieving decrease from baseline 0.35 in HAQ-DI, enthesitis absence, dactylitis absence, MDA, and PsARC were handled by non-responder imputation. LS means were calculated based on a mixed model for repeated measures without imputation for missing values. * All randomized patients who received 1 dose of study medication (N=394). Among patients with baseline score In patients with baseline score >0. Post hoc analyses. BID, twice daily; DAS28-3(CRP), Disease Activity Score in 28 joints with C-reactive protein; EQ-5D, EuroQol 5-Dimension Health State Profile; HAQ-DI, Health Assessment Questionnaire-Disability Index; LS, least squares; MDA, Minimal Disease Activity; N, number of patients in treatment group; n, number of patients with response; N1, number of patients with data available; PsARC, Psoriatic Arthritis Response Criteria; SE, standard error; SPARCC, SpondyloArthritis Research Consortium of Canada. 37

39 Table S5. Most common adverse events (all causality); reported in 5% of patients in any treatment group (safety analysis set). Baseline to month 3 Baseline to month 6 Placebo Tofacitinib 5 mg BID Tofacitinib 10 mg BID Placebo 5 mg BID Placebo 10 mg BID Tofacitinib 5 mg BID Tofacitinib 10 mg BID Patients with adverse event, n (%) (N=131) (N=131) (N=132) (N=66) (N=65) (N=131) (N=132) Upper respiratory tract infection 6 (4.6) 10 (7.6) 6 (4.5) 4 (6.1) 7 (10.8) 12 (9.2) 7 (5.3) Nasopharyngitis 3 (2.3) 10 (7.6) 7 (5.3) 4 (6.1) 1 (1.5) 14 (10.7) 12 (9.1) Headache 7 (5.3) 5 (3.8) 10 (7.6) 3 (4.5) 4 (6.2) 10 (7.6) 12 (9.1) Nausea 7 (5.3) 4 (3.1) 3 (2.3) 5 (7.6) 4 (6.2) 5 (3.8) 7 (5.3) Diarrhea 1 (0.8) 6 (4.6) 5 (3.8) 2 (3.0) 2 (3.1) 10 (7.6) 8 (6.1) Hypertension 2 (1.5) 4 (3.1) 4 (3.0) 2 (3.0) 2 (3.1) 8 (6.1) 5 (3.8) Sinusitis 3 (2.3) 3 (2.3) 3 (2.3) 4 (6.1) 2 (3.1) 3 (2.3) 5 (3.8) Dizziness 1 (0.8) 6 (4.6) 1 (0.8) 0 1 (1.5) 7 (5.3) 1 (0.8) Bronchitis 0 3 (2.3) 4 (3.0) 1 (1.5) 2 (3.1) 3 (2.3) 7 (5.3) BID, twice daily. 38

40 Table S6. Laboratory parameter values at month 3 and month 6 (safety analysis set*). Up to month 3 Up to month 6 Placebo Tofacitinib 5 mg BID Tofacitinib 10 mg BID Placebo tofacitinib 5 mg BID Placebo tofacitinib 10 mg BID Tofacitinib 5 mg BID Tofacitinib 10 mg BID (N=131) (N=131) (N=132) (N=66) (N=65) (N=131) (N=132) Mean change from baseline ALC x10 3 cells/mm 3 (SD) Mean change from baseline ANC x10 3 cells/mm 3 (SD) Mean change from baseline hemoglobin, g/dl (SD) Mean change from baseline platelets x10 3 cells/mm 3 (SD) Mean % change from baseline LDL-c, mg/dl (SD) Mean % change from baseline HDL-c, mg/dl (SD) Mean % change from baseline triglycerides, mg/dl (SD) 0.1 (0.4) -0.1 (0.6) 0.1 (0.5) 0.0 (0.5) 0.2 (0.4) -0.2 (0.6) 0.1 (0.5) -0.1 (1.8) -0.3 (1.8) -0.7 (1.9) -0.3 (2.0) -0.9 (2.0) -0.1 (2.1) -0.8 (2.0) -0.2 (0.8) -0.1 (0.9) -0.3 (0.9) -0.2 (1.1) -0.1 (1.0) -0.2 (1.0) -0.3 (1.0) -2.7 (57.6) (52.2) (66.3) (60.4) (53.1) (63.5) (71.4) 5.0 (21.2) 7.9 (18.6) 12.9 (26.4) 9.0 (21.4) 17.7 (29.2) 7.6 (17.6) 11.6 (25.2) -1.6 (17.0) 9.1 (18.0) 15.1 (18.8) 4.2 (16.1) 14.4 (20.6) 9.6 (20.6) 11.8 (21.2) 7.1 (45.4) 6.2 (37.6) 13.4 (56.0) 10.7 (44.0) 2.9 (53.1) 5.8 (39.0) 24.2 (59.0) 39

41 Mean change from baseline creatinine, mg/dl (SD) Mean change from baseline creatine kinase, U/L (SD) (0.1) 0.02 (0.1) 0.04 (0.1) 0.01 (0.1) 0.02 (0.1) 0.02 (0.1) 0.03 (0.1) -8.8 (140.1) 51.3 (90.3) 89.8 (173.2) 38.6 (39.3) 21.6 (205.3) 56.7 (147.8) 93.6 (139.5) AST increase, n (%) >1 x ULN 8 (6.2) 24 (18.5) 29 (22.1) 13 (20.0) 8 (12.3) 37 (28.5) 41 (31.1) 2 x ULN 1 (0.8) 4 (3.1) 2 (1.5) 0 1 (1.5) 5 (3.8) 3 (2.3) 3 x ULN 0 1 (0.8) 1 (0.8) (0.8) 1 (0.8) ALT increase, n (%) >1 x ULN 21 (16.2) 29 (22.3) 34 (26.0) 19 (29.2) 16 (24.6) 38 (29.2) 47 (35.6) 2 x ULN 3 (2.3) 7 (5.4) 4 (3.1) 3 (4.6) 3 (4.6) 10 (7.7) 6 (4.5) 3 x ULN 0 2 (1.5) 1 (0.8) (1.5) 2 (1.5) * All patients who received 1 dose of study medication. ALC, absolute lymphocyte count; ALT, alanine aminotransferase; ANC, absolute neutrophil count; AST, aspartate aminotransferase; BID, twice daily; HDL-c, high-density lipoprotein cholesterol; LDL-c, low-density lipoprotein cholesterol; n, number of patients with event; SD, standard deviation; ULN, upper limit of normal. 40

42 References 1. Taylor W, Gladman D, Helliwell P, Marchesoni A, Mease P, Mielants H. Classification criteria for psoriatic arthritis: development of new criteria from a large international study. Arthritis Rheum 2006;54: Carpenter JR, Roger JH, Kenward MG. Analysis of longitudinal trials with protocol deviation: a framework for relevant, accessible assumptions, and inference via multiple imputation. J Biopharm Stat 2013;23: Rubin DB. Multiple imputation for nonresponse in surveys. New York,