Pivotal Studies and Protocol-Selected Studies
Description of Studies
Five pivotal multinational double-blind randomized controlled trials (RCTs) met the criteria for this systematic review: SELECT-COMPARE, SELECT-MONOTHERAPY, SELECT-NEXT, SELECT-BEYOND, and SELECT-EARLY. All five studies enrolled adults with adult-onset RA, and in all but the EARLY trial (where patients were methotrexate naive), patients’ symptoms had been inadequately controlled with a DMARD. In COMPARE (N = 1,629) patients had to be inadequately controlled on methotrexate, in NEXT (N = 661) patients had to be inadequately controlled on any csDMARD, and in BEYOND (N = 499) patients had to be inadequately controlled on a biologic DMARD (bDMARD). The five studies reflected patients with RA who had received various prior treatments and had switched to different regimens. In EARLY (N = 947), patients could not have been considered as methotrexate inadequate responders and could not have been on any DMARD beside methotrexate for longer than three weeks; the patients were randomized to receive upadacitinib 15 mg once daily, upadacitinib 30 mg once daily, or methotrexate (7.5 mg or 10 mg) once weekly. In MONOTHERAPY, patients had had a csDMARD experience and patients were randomized to upadacitinib 15 mg once daily, upadacitinib 30 mg once daily, or methotrexate (to continue on a prior stable dose) with no other added therapy in the background. In COMPARE, patients had been considered as csDMARD inadequate responders but not bDMARD inadequate responders and moved into the trial with a background therapy of methotrexate; they were randomized (2:1:2) into upadacitinib 15 mg once daily, adalimumab 40 mg injection every other week, or placebo. In NEXT, patients had been considered csDMARD inadequate responders but not bDMARD inadequate responders and moved into the study with csDMARD background therapy; they were randomized into upadacitinib 15 mg once daily, upadacitinib 30 mg once daily, or placebo. In BEYOND, patients had been considered bDMARD inadequate responders and moved into the study with csDMARD background therapy; they were randomized into upadacitinib 15 mg once daily, upadacitinib 30 mg once daily, or placebo. The primary outcome in the COMPARE, NEXT, and BEYOND studies was the proportion of patients achieving an American College of Rheumatology (ACR) 20 (20% improvement in ACR criteria) response at 12 weeks. The primary outcome in MONOTHERAPY was achieving ACR20 at 14 weeks, and the primary outcome in EARLY was achieving 50% improvement in ACR criteria (ACR50) at 12 weeks. Key secondary outcomes that were accounted for type I errorincluded HRQoL on the Health Assessment Questionnaire–Disability Index (HAQ-DI), the Disease Activity Score-28 and C-reactive protein (DAS28-CRP), and the modified total Sharp score (mTSS). The primary outcome was reported at 12 weeks in all studies except MONOTHERAPY, in which it was reported at week 14. Each of these studies has an ongoing long-term extension study.
Efficacy Results
Efficacy results are summarized in . The primary outcome in EARLY of ACR50 at week 12 showed a response rate of 52.1% (95% confidence interval [CI], 46.6 to 57.5) in the upadacitinib and 28.3% (95% CI, 23.4 to 33.3) in the methotrexate arms, with a rate difference of 23.7 (95% CI, 16.3 to 31.1). The primary outcome in MONOTHERAPY of ACR20 at week 14 showed a response rate of 67.7% (95% CI, 61.5 to 74.0) in the upadacitinib and 41.2% (95% CI, 34.6 to 47.8) in the methotrexate arms, with a rate difference of 26.5 (95% CI, 17.5 to 35.6). The primary outcome in COMPARE of ACR20 at week 12 showed a response of 70.5% (95% CI, 67.0 to 74.0) in the upadacitinib, 63.0% (95% CI, 57.8 to 68.2) in the adalimumab, and 36.4% (95% CI, 32.7 to 40.1) in the placebo arms, with a rate difference of 7.5 (95% CI, 1.2 to 13.8) versus adalimumab and 34.1 (95% CI, 29.0 to 39.2) versus placebo. The primary outcome in NEXT of ACR20 at week 12 was 63.8% (95% CI, 57.5 to 70.1) in the upadacitinib and 35.7% (95% CI, 29.4 to 42.1) in the placebo groups, with a rate difference of 28.1 (95% CI, 19.1 to 37.0). In BEYOND, the primary outcome of ACR20 at week 12 was 64.6% (95% CI, 57.3 to 72.0) in the upadacitinib and 28.4% (95% CI, 21.6 to 35.2) in the placebo groups, with a rate difference of 36.2 (95% CI, 26.2 to 46.2).
Summary of Key Results From Pivotal and Protocol-Selected Studies.
The HAQ-DI was a secondary outcome in all studies. Results of upadacitinib 15 mg versus the placebo and methotrexate groups showed a statistically significant magnitude of difference greater than the estimated minimal important difference of 0.22 in all the studies at week 12 (week 14 in MONOTHERAPY). The DAS28-CRP was a secondary outcome in all studies. Results of upadacitinib 15 mg versus the placebo and methotrexate groups showed a statistically significant mean difference in the results in all the studies. The mTSS was reported in COMPARE (week 24) and EARLY (week 26) as a secondary outcome, with a responder analysis where a responder was defined as having no change in mTSS. In both studies, the mean difference was statistically significantly in favour of upadacitinib and, similarly, the response rate in upadacitinib-treated patients was statistically significantly higher than placebo (COMPARE, rate difference 7.5 [95% CI, 3.0 to 12.1]) and than methotrexate (EARLY, rate difference 9.8 [95% CI, 3.5 to 16.2]). Other secondary outcomes included the proportion of patients with LDA, the Short Form (36) Health Survey (SF-36), and morning stiffness. Overall, these results are consistent in showing the benefit of upadacitinib 15 mg once daily over the placebo and methotrexate control arms.
One of the comparisons provided in the studies was that of upadacitinib versus adalimumab in the COMPARE study. The sponsor’s initial outcome for this comparison was to achieve a noninferiority on ACR50 at week 12; the result showed upadacitinib to be statistically superior to adalimumab (rate difference 16.1 [95% CI, 9.9 to 22.3]). In addition, upadacitinib was superior to adalimumab in the HAQ-DI measure, but the treatment difference did not exceed the identified minimal important difference of 0.22 points (least squares mean difference −0.11 [95% CI, −0.184 to −0.036]). Other outcomes beyond ACR50, HAQ-DI, and patients’ assessments of pain, in comparing upadacitinib 15 mg to adalimumab, were outside the statistical testing hierarchy. However, upadacitinib showed better results than adalimumab in all examined outcomes except in mTSS.
Harms Results
In COMPARE, 64.2% of upadacitinib patients, 60.2% of adalimumab patients, and 53.2% of placebo patients experienced an adverse event. In MONOTHERAPY, the percentages were 47.5% in upadacitinib and 47.2% in placebo groups. In NEXT, the percentages were 56.6% in upadacitinib and 48.9% in placebo groups. In BEYOND, the percentages were 55.5% in upadacitinib and 56.2% in placebo groups. In EARLY, the percentages were 64.0% in upadacitinib and 65.3% in placebo groups. Respiratory tract infections were the most common adverse events in all the included studies. Serious adverse generally occurred in less than 5% of patients across the studies. In COMPARE, 3.7% of upadacitinib patients, 4.3% of adalimumab patients, and 2.9% of placebo patients experienced a serious adverse event. In MONOTHERAPY, the percentages were 5.1% in upadacitinib and 2.8% in methotrexate groups. In NEXT, the percentages were 4.1% in upadacitinib and 2.3% in placebo groups. In BEYOND, the percentages were 4.9% in upadacitinib and 0% in placebo groups. In EARLY, the percentages were 4.7% in upadacitinib and 4.1% in methotrexate groups. No single serious adverse event was most common across the five included studies.
According to the available data, there was a numerically higher incidence of herpes zoster infection in the upadacitinib treatment groups when contrasted to non-upadacitinib treatment groups. Over the course of the efficacy and extension phases of the studies, malignancies were reported in 1.1% of all patients who started and stayed on upadacitinib and 1.2% of patients who switched over to upadacitinib for the extension phase. Overall, notable harms identified for this review did not show explicit imbalance between groups, with the exception of a numerically higher proportion of neutropenia in COMPARE, BEYOND, and EARLY. Also, there was no explicit imbalance in the number of thromboembolic events between upadacitinib-treated patients and other groups.
Critical Appraisal
The main limitations of the pivotal studies include an imbalance in the discontinuation rate between groups in the EARLY study (14.6% withdrawals in methotrexate and 8.5% in upadacitinib) as well as a disproportional discontinuation rate between adalimumab (8.3%) and upadacitinib (4.8%) in the COMPARE study. It is not clear if these imbalances in discontinuation could have led to any sort of bias in the outcomes. In addition, the lack of direct comparison against other existing JAK inhibitors reduces the ability to determine the benefit of upadacitinib versus other existing JAK inhibitors (e.g., baricitinib). Also, several outcomes that were identified in our protocol and reported in the studies fell outside the statistical testing hierarchy and thus need to be interpreted with consideration of type I error.
As well, the EARLY study included treatment-naive patients, which does not align with the Health Canada indication for the treatment of adults with moderately to severely active RA who have had an inadequate response or intolerance to methotrexate. Additionally, although the trials’ inclusion and exclusion criteria as well as the baseline demographic characteristics of enrolled patients are generally in line with other RA clinical trials, according to the clinical expert, the enrolled patients in these trials do not represent the majority of patients present in clinical practice. Specifically, the high ratio of female, white, and rheumatoid factor–positive patients may not be representative of Canadian patients. However, there is no clear evidence that these factors will lead to variation in the response to the treatment.
Indirect Comparisons
Description of Studies
Two indirect treatment comparisons (ITCs) are discussed in this review: one submitted by the sponsor and one identified in the literature search conducted by CADTH. Both used a Bayesian network meta-analysis (NMA) approach. The sponsor-submitted ITC is a systematic review of upadacitinib and existing csDMARD, bDMARD, and JAK inhibitors. The published ITC identified by CADTH, by Song et al., only compared upadacitinib to tofacitinib.
Efficacy Results
In the csDMARD-experienced population, the authors of the sponsor’s ITC reported that upadacitinib achieved the highest numerical probability of achieving ACR20, 50, and 70 (70% improvement in the ACR criteria) at both the 12- and 24-week end points when contrasted with the probability of other bDMARD and csDMARD interventions. Wide credible intervals (CrIs) were reported across all calculations. However, no comparative result versus bDMARD or JAK inhibitors was provided to assess the magnitude of potential difference in treatment response between upadacitinib and other bDMARD or JAK inhibitors. In the bDMARD-experienced population, the results for upadacitinib were available for 12 weeks only, showing CrIs that are wider than those reported in the csDMARD-experienced patients and suggesting a lack of statistical robustness in the data.
Song et al. reported that upadacitinib had a higher odds ratio (OR) of achieving the efficacy outcome than did tofacitinib. However, the CrI was wide and included the null (OR 1.52 [95% CrI, 0.64 to 3.26]). A recent draft publication by the Institute for Clinical and Economic Review examining upadacitinib reported conclusions similar to the sponsor’s submitted ITC, but the report had insufficient details and outcome results to be included in this review.
Harms Results
The sponsor’s submitted ITC did not report a safety outcomes analysis. Song et al. provided an analysis of serious adverse events, where the results suggest that the OR of serious adverse events is lower in upadacitinib than in other comparators with a CrI that includes the null.
Critical Appraisal
Several limitations increase the uncertainty in the results provided in the ITC discussed in this review. The sponsor’s ITC did not provide an indirect comparison result versus comparators beyond csDMARD. Also, CrIs were wide across the reported outcomes, suggesting considerable statistical heterogeneity in the included studies. In addition, the sponsor’s ITC did not provide the results of inconsistency modelling but reported that it was conducted and that no inconsistency was observed. The ITC by Song et al. does not provide sufficient information regarding the included studies’ characteristics, the baseline demographics of enrolled patients, or the methods of combining different routes of administration and different definitions of inadequate responders and potential outcomes. An informed judgment of potential clinical or methodological heterogeneity cannot be made in the absence of this information.
Other Relevant Evidence
Description of Studies
Each of the five pivotal studies consisted of two periods, with the first ranging from 12 weeks to 48 weeks in duration and the second ranging from 192 weeks to five years. At the time of this review, data up to 48 weeks were available for SELECT-COMPARE, SELECT-MONOTHERAPY, and SELECT-EARLY. SELECT-NEXT and SELECT-BEYOND included data up until week 60. The five included studies were double-blind RCTs, followed by an open-label extension (SELECT-COMPARE and SELECT-EARLY) or a blinded long-term extension (SELECT-MONOTHERAPY, SELECT-NEXT, and SELECT-BEYOND).
Efficacy Results
The proportion of patients meeting the ACR20, 50, and 70 response criteria at week 48 was 64.7% (95% CI, 61.0 to 68.3), 49.5% (95% CI, 45.6 to 53.3), and 36.1% (95% CI, 32.4 to 39.8), respectively, in SELECT-COMPARE. In the SELECT-EARLY study, at week 48, 91.9% (95% CI, 88.5 to 95.3) of patients met the ACR20 criteria, 79.3% (95% CI, 74.2 to 84.3) met the ACR50 criteria, and 63.3% (95% CI, 57.2 to 69.3) met the ACR70 criteria.
The three remaining studies used the as-observed dataset to describe their results and reported that 87.2% (95% CI, 82.2 to 92.2) met the ACR20 criteria at week 48 (SELECT-MONOTHERAPY) and 76.7% (95% CI, 69.5 to 83.9) and 85.0% (95% CI, 79.6 to 90.3) met it at week 60 (SELECT-NEXT and SELECT-BEYOND). Further, 69.5% (95% CI, 62.7 to 76.4) of patients met the ACR50 criteria at week 48 in SELECT-MONOTHERAPY, and 52.2% (95% CI, 43.8 to 60.7) and 72.6% (95% CI, 65.9 to 79.4) did at week 60 in SELECT-BEYOND and SELECT-NEXT, respectively. The ACR70 response rate at week 48 was 45.5% (95% CI, 38.1 to 52.8) in SELECT-MONOTHERAPY; at week 60, it was between 33.3% (95% CI, 25.4 to 41.3) and 51.5% (95% CI, 44.0 to 59.0) in SELECT-BEYOND and SELECT-NEXT.
Harms Results
The pooled harms data included data from patients treated with upadacitinib 15 mg for up to one year in any of the five included studies. In total, 2,630 patients were included. Of the 2,630 patients, ▬ reported experiencing an adverse event, ▬ experienced a serious adverse event, ▬ stopped treatment owing to adverse events▬ ▬ ▬ ▬ ▬
Critical Appraisal
The results of the long-term efficacy and safety outcomes are limited by a lack of comparator in SELECT-NEXT, SELECT-BEYOND, and SELECT-MONOTHERAPY. In addition, only descriptive statistics were provided and any statistical testing that was performed was not included in the statistical hierarchy; thus, there is a risk of type I error. Missing data were only accounted for in SELECT-COMPARE, which used a nonresponder imputation method to label missing data as nonresponders. Consequently, the efficacy of upadacitinib in patients in SELECT-COMPARE was lower than in the other four studies, which used nonimputed, observed data. It is uncertain whether this is due to the bias introduced by classifying missing data as nonresponders or by an overestimation of the results caused by not imputing for missing data.
