Input From the Clinical Expert Consulted by CADTH

A substantial proportion of patients lose their response to therapy over time or do not achieve a minimal response with their first therapy. In addition, some treatments have more adverse effects than others. There is a need for medications with new mechanisms of action or a different safety profile to offer alternate treatment options for these patients who have an inadequate response or intolerance to therapy. According to the clinical expert, guselkumab may be used as first- or second-line biologic therapy. It may be a preferred first-line treatment for patients with moderate psoriasis in addition to musculoskeletal disease. Guselkumab may be used in combination with methotrexate or leflunomide; however, there is no evidence to support its use in combination with other biologics.

Response to therapy is based on a reduction in the number of inflamed joints, improvement in the skin, and improvements in patient-reported outcomes (i.e., based on assessments of pain, function, and fatigue). According to the clinical expert, a major improvement would constitute at least a 50% improvement, but may also include achievement of minimal disease activity (MDA) and remission as measured by specific instruments. An initial response may be expected within 3 months, with more significant improvement by 6 months after initiating therapy. The expert indicated that if a patient shows no change within 3 to 6 months, they would be considered a nonresponder and may be switched to another medication. The expert stated that ideally, guselkumab would be prescribed by specialists who are familiar with the drug and its uses, or at least in consultation with a dermatologist or rheumatologist.

Clinician Group Input

CADTH received 1 clinician group input submission from the Canadian Rheumatologist Psoriatic Arthritis Interest Group, based on responses from 6 clinicians who practise in academic and community settings. This clinician input largely agreed with the input received from the clinician consulted by CADTH. No major contrary views among the views provided by the clinical experts consulted by CADTH for this review were presented.

Description of Studies

Three double-blind, placebo-controlled, randomized controlled trials (RCTs) met the inclusion criteria for the systematic review. The pivotal trials (DISCOVER-1 and DISCOVER-2) included patients with active PsA who had experienced an inadequate response to cDMARDs, apremilast, and/or NSAIDs. The DISCOVER-1 study (N = 381) enrolled a mixed population that included patients with no prior biologic treatment experience; however, up to 30% of patients had previously received 1 or 2 TNF alpha inhibitors. In the DISCOVER-2 study, all enrolled patients were biologic-naive (N = 741). The COSMOS study enrolled patients with active PsA who were intolerant to or had experienced an inadequate response to 1 or 2 TNF alpha inhibitors (N = 285). The trials were conducted mainly in Europe, with some sites in Asia, the US, Australia, Israel, and Canada (DISCOVER-1 only).

Patients were randomized to receive placebo or guselkumab 100 mg SC at week 0, week 4, and every 8 weeks thereafter for 24 weeks. The DISCOVER trials included a third treatment group (guselkumab 100 mg every 4 weeks). However, the latter dosage is not consistent with the Health Canada–recommended dose; therefore, results for this treatment group are not included in this report. The total trial durations were 52 weeks (DISCOVER-1), 100 weeks (DISCOVER-2), and 48 weeks (COSMOS), with placebo patients switching to guselkumab starting at week 24. During the trials, patients could continue receiving methotrexate, leflunomide, hydroxychloroquine, sulfasalazine, oral corticosteroids, or NSAIDs if the doses were stable and did not exceed the protocol-specified maximum doses. Early escape therapy consisting of cDMARDs, corticosteroids, or NSAIDs — or a switch to guselkumab (COSMOS) — was available at week 16 for patients who had an improvement of less than 5% in tender and swollen joint counts.

The primary outcome in all trials was the proportion of patients who achieved an American College of Rheumatology (ACR) 20% improvement (ACR 20) at week 24. The ACR 20 was defined as a greater than or equal to 20% improvement from baseline in both swollen joint count (66 joints) and tender joint count (68 joints), and a greater than or equal to 20% improvement from baseline in 3 of the 5 assessments: patient’s assessment of pain, patient’s global assessment of disease activity, physician’s global assessment of disease activity, the Health Assessment Questionnaire Disability Index (HAQ-DI), and C-reactive protein (CRP). Other outcomes of interest included the change from baseline in the HAQ-DI and Short Form (36) Health Survey (SF-36) Physical Component Summary (PCS) and the impact on plaque psoriasis (as measured by the Investigator’s Global Assessment of Psoriasis [IGA] or Psoriasis Area and Severity Index [PASI] response).

The mean age of patients enrolled ranged from 44.9 years (standard deviation [SD] = 11.9 years) to 49.1 years (SD = 12.3 years) across treatment groups in the 3 trials. The proportion of female patients ranged from 46% to 55%, and most patients were White (89% to 98%; not reported for the COSMOS study). The mean number of swollen joints ranged from 9.0 (SD = 5.7) to 12.3 (SD = 6.9), and the mean number of tender joints ranged from 18.2 (SD = 10.7) to 21.6 (SD = 13.1). Approximately 2-thirds of patients had psoriatic involvement affecting at least 3% of their body surface area (BSA). Two-thirds of patients reported enthesitis, while approximately 40% had dactylitis at baseline. The majority of patients (54% to 63%) were receiving methotrexate at baseline, with a lower percentage (0% to 7%) receiving other permitted cDMARDs. In the DISCOVER trials, 14% to 20% of patients were receiving oral corticosteroids at baseline versus 4% to 5% of patients in the COSMOS study.

Efficacy Results

In the DISCOVER-1 study, 52.0% of patients in the group receiving guselkumab every 8 weeks achieved ACR 20 at 24 weeks, compared with 22.2% of patients in the placebo group. The absolute difference was 29.8% (95% confidence interval [CI], 18.6% to 41.1%; P < 0.001), favouring guselkumab versus placebo (Table 2). The proportion of patients who achieved at least an American College of Rheumatology 50% improvement (ACR 50) was 29.9% versus 8.7% (absolute difference = 21.4%; 95% CI, 12.1% to 30.7%) for guselkumab every 8 weeks versus placebo; the proportion of patients who achieved an American College of Rheumatology 70% improvement (ACR 70) was 11.8% versus 5.6% (absolute difference = 6.4%; 95% CI, −0.3% to 13.1%) for guselkumab every 8 weeks versus placebo. However, ACR 50 and ACR 70 were not controlled for multiple testing and should be interpreted considering the inflated risk of type I error rate.

Among patients who were biologic-naive (DISCOVER-2), 64.1% and 32.9% achieved ACR 20 at 24 weeks in the group receiving guselkumab every 8 weeks and the placebo group, respectively, with an absolute difference of 31.2% (95% CI, 22.9% to 39.5%; P < 0.001) (Table 2). The proportion of patients who achieved ACR 50 was 31.5% versus 14.2% (absolute difference = 17.2%; 95% CI, 10.0% to 24.4%), and the proportion of patients who achieved ACR 70 was 18.5% versus 4.1% (absolute difference = 14.5%; 95% CI, 9.1% to 19.9%). ACR 50 and ACR 70 were not controlled for multiple testing (i.e., the type I error rate has not been controlled).

For biologic-experienced patients who were enrolled in the COSMOS study, 44.4% and 19.8% achieved ACR 20 at week 24 in the group receiving guselkumab every 8 weeks and the placebo group, respectively. The absolute difference between groups favoured guselkumab: 24.6% (95% CI, 14.1% to 35.2%; P < 0.001) (Table 2). The difference also favoured guselkumab every 8 weeks versus placebo for the proportion who achieved ACR 50 (19.6% versus 5.2%; absolute difference = 14.3%; 95% CI, 7.2% to 21.4%; P < 0.001). ACR 70 was achieved by 7.9% versus 1.0% of patients in the group receiving guselkumab every 8 weeks versus the placebo group, with an absolute difference of 6.8% (95% CI, 2.6% to 11.1%). ACR 70 was not controlled for multiple testing.

In the DISCOVER trials, the odds ratios of ACR 20 response were generally consistent across subgroups based on prior TNF alpha inhibitor use, use of non-biologic DMARDs, oral corticosteroids, or NSAIDs at baseline; however, the trials may not have been powered to detect subgroup differences. The COSMOS study did not report data for any subgroups of interest to this review.

Disability was assessed using the HAQ-DI, a patient-reported, 20-question instrument that assesses the degree of difficulty a person has in accomplishing tasks in 8 functional areas (dressing, arising, eating, walking, hygiene, reaching, gripping, and activities of daily living). The overall score is the average of 8 domains ranging from 0 (no disability) to 3 (completely disabled). The change from baseline to week 24 in the HAQ-DI favoured guselkumab every 8 weeks versus placebo in all trials (Table 2). The least squares (LS) mean differences versus placebo reported were −0.25 (95% CI, −0.36 to −0.13; P < 0.001) in the DISCOVER-1 study; −0.24 (95% CI, −0.32 to −0.15; P < 0.001) in the DISCOVER-2 study; and −0.17 (95% CI, −0.28 to −0.06; P = 0.003) in the COSMOS study. Across the trials, the between-group and within-group differences did not exceed the 0.35 minimal important difference (MID) cited by the sponsor,⁶ with the exception of the change from baseline within the group receiving guselkumab every 8 weeks in the DISCOVER-2 study.

The change from baseline to week 24 in the SF-36 PCS favoured guselkumab every 8 weeks versus placebo in all 3 studies (Table 2). The LS mean differences were 4.1 (95% CI, 2.4 to 5.9; P < 0.001) in DISCOVER-1, 4.0 (95% CI, 2.7 to 5.2; P = 0.011) in DISCOVER-2, and 3.9 (95% CI, 2.5 to 5.4; P < 0.001) in the COSMOS study. The Clinical Study Report defined at least a 5-point increase as clinically meaningful, but a MID of 3.74 points has also been reported in the literature.⁷ No statistically significant differences were detected between guselkumab every 8 weeks and placebo in the change from baseline to week 24 in the SF-36 Mental Component Summary (MCS).

In all trials, psoriasis skin disease outcome measures were analyzed in the subgroup of patients who had psoriasis affecting greater than or equal to 3% BSA and an IGA score of greater than or equal to 2 at baseline (55% to 74% of patients per treatment group). Psoriasis severity was assessed using composite physician-reported assessments: IGA and PASI response. For the IGA, the severity of a patient’s psoriasis is scored as cleared (0), minimal (1), mild (2), moderate (3), or severe (4). IGA response was defined as a score of 0 or 1, and at least a 2-point decrease from baseline. The PASI evaluates the extent and severity of psoriasis and is scored from 0 to 72 points, with a PASI score greater than 10 considered to represent severe disease. Patients with 90% improvement in their PASI score (PASI 90) or 100% improvement in their PASI score (PASI 100) would meet PASI 90 or PASI 100 response criteria, respectively.

The proportion of patients who achieved an IGA response at week 24 was higher in the groups receiving guselkumab every 8 weeks than in the placebo groups in the DISCOVER-1 study (57.3% versus 15.4%; absolute difference = 42.0% [95% CI, 28.9% to 55.1%; P < 0.001]) and DISCOVER-2 study (70.5% versus 19.1%; absolute difference = 50.9% [95% CI, 42.2% to 59.7%; P < 0.001]) (Table 2). In the COSMOS study, 48.1% versus 9.4% of patients in the guselkumab-every-8-weeks versus placebo group achieved in IGA response, with an absolute difference of 38.8% (95% CI, 27.3% to 50.4%); however, the P value has not been adjusted to control for multiple testing and should be interpreted with caution because of the potential for inflated type I error rate. PASI 100 response at week 24 was a secondary outcome in the COSMOS study. In the guselkumab group, 30.8% of patients achieved a PASI 100 response compared with 3.8% of patients in the placebo group (absolute difference = 27.4% [95% CI, 17.9% to 36.8%; P < 0.001). In the DISCOVER trials, the proportion of patients who achieved a PASI 100 response was nominally higher for the guselkumab versus placebo groups; however, these outcomes were not controlled for multiple testing and should be interpreted considering the inflated risk of type I error rate.

For patients with enthesitis or dactylitis at baseline, the results of the DISCOVER-2 and COSMOS studies suggest an improvement in enthesitis or dactylitis end points, with guselkumab every 8 weeks relative to placebo, but no statistically significant difference was detected between groups in DISCOVER-1. Based on the preplanned pooled analysis of data from DISCOVER-1 and DISCOVER-2, 49.6% and 29.4% of patients in the guselkumab-every-8-weeks and placebo groups had resolution of enthesitis at week 24, with a between-group difference of 20.1% (95% CI, 11.8% to 28.5%; P = 0.03). The proportion of patients whose dactylitis resolved at week 24 was 59.4% versus 42.2% in the guselkumab-every-8-weeks versus placebo groups (between-group difference = 18.0%; (95% CI, 7.4% to 28.6%; P = 0.03). None of the trials detected a statistically significant difference between guselkumab every 8 weeks and placebo in the proportion of patients who reported a clinically important improvement in axial disease based on the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI). It is noteworthy that these outcomes were tested in subgroups of patients that may not have been balanced with respect to baseline demographic and disease characteristics between treatment groups, due to the lack of stratification at randomization.

Symptoms of fatigue were assessed using the Functional Assessment of Chronic Illness Therapy – Fatigue Scale (FACIT-Fatigue). The scores ranged from 0 to 52, with lower scores reflecting more severe fatigue. Estimates of the MID ranged from 3.1 points to 4 points.^8,9 The DISCOVER-1 study reported an LS mean difference of 3.4 points (95% CI, 1.4 to 5.4); DISCOVER-2 reported an LS mean difference of 4.0 (95% CI, 2.5 to 5.5); and COSMOS reported an LS mean difference of 3.6 (95% CI, 1.7 to 5.4) for guselkumab every 8 weeks versus placebo. This outcome was not controlled for multiple testing (i.e., the type I error rate has not been controlled).

Radiographic progression was a major secondary outcome in the DISCOVER-2 study. Progression was assessed using the modified van der Heijde-Sharp (vdH-S) score, which ranges from 0 (best) to 528 (worst) and is the sum of the joint erosion score and the joint space narrowing score.¹⁰ At 24 weeks, the study failed to detect a statistically significant difference between guselkumab every 8 weeks and placebo in the change from baseline in vdH-S score. However, the duration of the trial may have been insufficient to detect a difference.

Harms Results

The frequency of adverse events was generally similar between groups in all trials, with 42% to 54% of patients in the groups receiving guselkumab every 8 weeks and 41% to 60% of patients in the placebo groups reporting 1 or more adverse events during the 24-week treatment period (Table 3). Nasopharyngitis (4% to 13%), upper respiratory tract infection (2% to 5%), and increased alanine aminotransferase (2% to 6%) were the most common adverse events in the guselkumab groups, with a comparable frequency of these events reported in the placebo groups. Generally, the frequency of infections was similar in the guselkumab groups (16% to 26%) and placebo groups (18% to 25%) across trials. Few serious infections were reported (0% to 0.5% in the guselkumab groups).

The frequency of serious adverse events ranged from 1% to 4% in the groups receiving guselkumab every 8 weeks and 3% to 4% in the placebo groups. No specific serious adverse events were reported in more than 1 patient per treatment group. Adverse events that resulted in treatment discontinuation were generally low and similar between groups (1% to 3%). In the DISCOVER-1 study, 1 patient in the placebo group died due to cardiac failure. No other deaths were reported in the first 24 weeks of the trials.

Table 2

Summary of Key Efficacy Results From Pivotal and Protocol Selected Studies.

Table 3

Summary of Key Safety Results From Pivotal and Protocol Selected Studies.

Critical Appraisal

The risk of bias related to randomization and treatment allocation concealment was rated as low for all studies. In general, patient characteristics and co-interventions appeared to be balanced between groups at baseline. The trials were double blind and took steps to maintain blinding of patients and investigators. Joint assessments were conducted by an independent rater who was not otherwise involved in the trial. Therefore, the risk of bias in the measurement of the outcomes was low for all trials. The frequency of withdrawal in all trials was low and similar between groups; therefore, there is a low risk of bias due to missing outcome data. The full analysis set (FAS), which excluded only 1 randomized patient in DISCOVER-2, was used for all efficacy outcomes; therefore, the analyses were appropriate for estimating the effect of assignment to the intervention.

In all trials, the primary and other dichotomous end points were analyzed using a Cochran-Mantel-Haenszel (CMH) test that was stratified by randomization stratification factors, with missing data imputed as nonresponders. The DISCOVER trials used an adjusted analysis of covariance (ANCOVA) model, and the COSMOS study used an unadjusted mixed-effects model for repeated measures (MMRM) to analyze continuous outcomes. Missing data were imputed under the missing-at-random assumption. This assumption may not hold true, but was not thought to be a major source of bias. Efficacy analyses were based on the composite estimand, with any patients meeting treatment failure criteria considered nonresponders for binary end points, or as no change from baseline for continuous measures. Treatment failure criteria included early study withdrawal or discontinuation of the study drug, or initiation of new treatments for PsA. This estimand, which considers any treatment failure end point to be an unfavourable outcome, may be considered a more conservative estimate of treatment effects. Of note, the COSMOS study incorrectly assigned 20 patients to early escape despite these patients not having met the escape criteria. Although the sponsor conducted sensitivity analyses to explore the impact of this error, these analyses cannot fully address the potential bias. The type I error rate was controlled for the primary and selected secondary outcomes in all studies. However, several outcomes of interest to this review were not controlled for multiplicity; these data should be interpreted with caution, given the potential for inflated type I error rate. Randomization was not stratified by the presence of psoriasis, enthesitis, dactylitis, or axial disease; thus, interpretation of the results for these outcomes should consider the possibility of imbalances in baseline demographic and disease characteristics between treatment groups in these subpopulations. The primary outcome was ACR 20 response; however, according to the clinical expert, this represents the minimum level of improvement that may be relevant to patients. In practice, the goal of therapy is to achieve higher levels of response.

Although the trials were 48 weeks to 100 weeks in duration, the comparative period was limited to 24 weeks for this chronic condition. For outcomes such as radiologic changes, the duration of treatment may have been insufficient to detect the impact of guselkumab. Moreover, none of the trials included an active control group; thus, direct evidence comparing guselkumab to other DMARDs available in Canada is not available.

With regards to external validity, the clinical expert did not identify any substantial limits to generalizability based on the patient population enrolled. The guselkumab dosing regime used in the trials (i.e., once every 8 weeks) was consistent with the Health Canada–recommended dose, and the expert stated that concomitant utilization of cDMARDs was similar to what may be expected in clinical practice. However, the expert also noted that the use of oral corticosteroids in the DISCOVER trials was higher than would be expected in Canada (14% to 20%). The use of a placebo comparator as an add-on to cDMARDs and NSAIDs is not consistent with Canadian practice for patients who have demonstrated an inadequate response to cDMARDs or bDMARDs. The trials excluded patients who had previously been treated with biologics other than TNF inhibitors; thus, the efficacy in patients with intolerance or inadequate response to other biologics, such as Janus kinase (JAK) or other IL inhibitors, is not known.

Description of Studies

The sponsor conducted a network meta-analysis (NMA) of RCTs that assessed the comparative efficacy and safety of guselkumab and 13 other bDMARDs for the short-term treatment of acute PsA. The indirect treatment comparison (ITC) was based on a systematic literature review, and 34 RCTs provided data to inform the Bayesian NMA. Analyses were conducted for the overall PsA population, with subgroup analyses restricted to patients who were biologic-naive or biologic-experienced. Treatment durations were 12 weeks to 24 weeks.

One other NMA was identified by CADTH through a literature search. The NMA by McInnes et al. (2022)¹⁵ evaluated the efficacy and safety of licensed and unlicensed bDMARDs for patients with active PsA. A total of 46 RCTs, which were identified through a systematic review, were included in the Bayesian analyses. The MNA included 19 biologics with outcomes assessed at 12 weeks to 26 weeks.

Efficacy Results

For the overall PsA population in the sponsor-submitted NMA, the results suggest that patients who received guselkumab every 8 weeks ||| || |||| |||||| || ||||||| || ||| || |||||||| than those who received apremilast or abatacept, but were |||| |||||| || respond than infliximab, golimumab, and secukinumab 300 mg (34 RCTs; ordinal baseline risk-adjusted random-effects model). Comparisons between guselkumab and other biologics had 95% credible intervals (CrIs) that |||||||||| ||| |||| |||||. The NMA results for PASI 90 response |||||||| |||||||||| |||||| ||||||||||| ||| ||| ||| ||||| |||||||||| ||||||| |||||||||||| ||| ||||||||||| ||||||||| ||| ||||||||||. Other comparisons with biologics reported 95% CrIs that |||||||| ||| ||||. For the change in HAQ-DI, SF-36 PCS or MCS, or vdH-S scores, |||| || ||| ||||||||||| |||||||| guselkumab over other biologics, with analyses reporting 95% CrIs that |||||||||| ||| ||||. Across all networks, many comparisons showed imprecise 95% CrIs that included the possibility of appreciable benefit and/or worse outcomes, and || |||||||||| ||||||| |||||||||| ||| ||||| |||||||||, which limits the ability to draw conclusions from these data.

The results in the biologic-naive population were generally |||||||||| |||| ||| |||||||| || ||| ||||||| ||||||||||, although the networks were smaller (17 studies to 30 studies), not all comparators were available for all outcomes, and with |||| || ||| |||||||| ||||||| ||||||| |||||||||. The NMAs in biologic-experienced patients included fewer trials (8 studies to 16 studies) and the reported results were imprecise, limiting the ability to draw conclusions.

The results for ACR 20 and PASI 90 in the NMA by McInnes et al. (2022)¹⁵ were largely consistent with the findings of the sponsor-submitted NMA.

Harms Results

In the overall population, the analysis of serious adverse events in the sponsor-submitted NMA |||||| ||| ||| |||| |||||||||| ||| |||| ||| |||| |||||||||||| |||| |||||||| ||||| |||| ||| |||||||||| |||||| |||||||||| || |||||||||||| ||| || |||| ||| |||||||| |||||||| ||||||||||||| |||||| ||||||| ||||||. For all other comparisons, the 95% CrI was |||||||||| limiting the ability to draw a conclusion. In most cases, ||| ||||||| || ||| ||| || ||||||| |||||| |||||| ||||||| ||||||||.

||| ||| ||| ||||||| |||||| || ||||||| ||||||| |||||| || |||||||||||||| ||| ||||||| ||| |||||||||||||||||||| |||||||| || ||||||| |||||| ||| ||| |||| |||||||||| ||| ||||| |||| |||| ||||||||||| ||||||| |||| |||| |||||||| ||| ||||||| || |||| | ||||||||||. No comparative safety data for guselkumab were reported in the NMA by McInnes et al. (2022).¹⁵

Critical Appraisal

Although the sponsor-submitted ITC was based on a systematic review, 46 RCTs were excluded from the NMA, and the criteria for selecting trials or outcomes for analysis were not stated. Heterogeneity in patient and study characteristics was identified, and it is unclear if the transitivity assumption has been met. The authors of the NMA attempted to address potential variability in effect modifiers by using a baseline risk-adjusted model, but it is unclear if these effect modifiers have the same level of effect on the active arms. Given that it is unclear to what extent placebo response is an adequate proxy for specific characteristics or effect modifiers, uncertainty remains in these analyses. Subgroup analyses based on prior treatment exposure were conducted to create more homogenous patient populations, but some of these analyses included data from a limited number of trials, and often showed substantial uncertainty, with wide CrIs. There were no subgroup or sensitivity analyses conducted to explore the potential impact of differences in the timing of outcome assessment, duration of disease, background therapies, or year of study.

ACR and PASI percentage improvements were analyzed using an ordinal model, which assumed that the relative treatment effects were the same for each response level. Thus, although data were reported separately for each response level, the inferences for each comparison are the same across the ACR 20, ACR 50, and ACR 70 levels or PASI levels. It is not clear if this assumption of the model holds true (i.e., if relative treatment effects are consistent across response levels), given that data were pooled for different time points. No sensitivity analyses were run to examine the impacts of this assumption.

In the sponsor-submitted NMA, there was limited ability to assess the consistency between direct and indirect evidence because there were few closed loops (i.e., there were only 4 head-to-head studies), and the statistical tests for inconsistency are generally underpowered. Further, most of the contributing trials were judged to be at high or unclear risk of bias in at least 1 domain.

Issues with heterogeneity in patient and study characteristics, lack of ability to assess consistency, and potential bias in the included RCTs were also identified as limitations for the NMA by McInnes et al. (2022).¹⁵

The indirect evidence was limited to short-term efficacy and safety; thus, longer-term comparative effects are uncertain.

Description of Studies

Efficacy and safety data were available for the uncontrolled extension phase of DISCOVER-1 (52 weeks), DISCOVER-2 (100 weeks), and COSMOS (48 weeks) trials. Descriptive results for patients who received guselkumab 100 mg every 8 weeks are summarized in this section, including results for patients from the placebo group in the COSMOS study who crossed over to guselkumab.

Efficacy Results

The extension phase data suggest that treatment effects may be maintained in patients who remain on guselkumab therapy every 8 weeks for 48 weeks to 100 weeks. In DISCOVER-1, 76 patients out of 112 patients (68%), and in COSMOS, 120 patients out of 172 patients (70%), achieved ACR 20 at week 48 or week 52. In the DISCOVER-2 study, 85 patients out of 234 patients (79%) achieved ACR 20 at week 52, and 183 patients out of 223 patients (82%) achieved ACR 20 at week 100. PASI 100 response was reported by 36 patients out of 75 patients (48%) in DISCOVER-1 (week 52), by 94 patients out of 169 patients (57%) in DISCOVER-2 (100 weeks), and by 80 patients out of 121 patients (66%) in COSMOS (48 weeks).

Harms Results

During the extension period, 31% to 72% of patients reported 1 or more adverse event; 3% to 9% reported a serious adverse event; and 1% to 3% stopped treatment due to adverse events. No deaths were reported. Infections were reported in 43% of patients in DISCOVER-1 and in 29% and 38% of patients at week 52 and week 100 of the DISCOVER-2 study. Over the 48-week treatment period of the COSMOS study, 22% of patients who received guselkumab every 8 weeks experienced an infection.

Critical Appraisal

Limitations of the extension study include selection bias and lack of a control group. Data were available only as descriptive statistics; and because there were no comparator groups, the interpretation of the results is limited. The outcomes were based on observed case data, with no imputation for missing data, and reflect treatment effects in patients who continue on therapy. As such, the results may overestimate the response in the broader population, given that patients who drop out are more likely to have unfavourable outcomes or poor tolerance to therapy.

Unmet Needs

The goals of therapy are to control the symptoms of PsA (i.e., pain stiffness, swelling, fatigue), reduce impacts on patients’ quality of life and function, and prevent the progression of joint damage. A substantial proportion of patients do not achieve a minimal response with their first therapy or lose their response over time. In addition, some treatments have more adverse effects than others. Thus, there is a need for medications with new mechanisms of action or a different safety profile to offer alternate treatment options. There also needs to be a choice for patients in terms of preference with regards to the frequency of injections, injection versus infusion, and oral administration versus injection.

Place in Therapy

According to the clinical expert, guselkumab may be used as first- or second-line biologic therapy. It may be a preferred first-line treatment for patients with moderate psoriasis in addition to musculoskeletal disease. Guselkumab may be used in combination with methotrexate or leflunomide. There is no evidence to support its use in combination with other biologics. The expert stated that it would not be appropriate to recommend that patients try other biologics before receiving guselkumab.

Patient Population

There are currently no biomarkers that may be used to identify the patients who will be most suited for a particular drug. Clinicians make these decisions based on patient history, physical examination, imaging, and discussion.

There is currently no diagnostic test for PsA. Rheumatologists can usually diagnose PsA after evaluating the patient and may seek confirmation of the psoriasis diagnosis from a dermatologist. The expert stated that 30% to 80% of psoriasis patients had a missed PsA diagnosis.

Assessing Response to Treatment

Response to therapy is based on a reduction in the number of inflamed joints, improvement in the skin, and patient-reported outcomes. According to the clinical expert, a major improvement would be considered an improvement of at least 50%, but may also include achievement of MDA and remission as measured by specific instruments. Assessment of response requires the physician to perform a physical examination, including a joint count and skin evaluation, and also includes patient-reported outcomes (i.e., assessment of pain, function, fatigue). Less frequently, structural damage may be assessed using X-rays.

An initial response may be expected within 3 months, with more significant improvement by 6 months after initiating therapy. The expert indicated that if a patient shows no change within 3 months to 6 months, they would be considered a nonresponder and may be switched to another medication.

Discontinuing Treatment

Patients may be switched to another medication if no improvement is observed within 3 months to 6 months or if the patient experiences intolerable or serious adverse effects. If a patient has responded in some but not all manifestations, another medication may be added to address the area that has not responded.

Prescribing Conditions

Ideally, guselkumab would be prescribed by specialists who are familiar with the drug and its uses, or at least in consultation with a dermatologist or rheumatologist. The administration of guselkumab may take place in any setting, including self-administration by the patient in their home.

Inclusion and Exclusion Criteria

All 3 included studies enrolled adults greater than or equal to 18 years of age with active PsA diagnosed at least 6 months prior who met the Classification Criteria for Psoriatic Arthritis (Table 7). In the DISCOVER-1 and COSMOS studies, patients were required to have least 3 tender and swollen joints — and in DISCOVER-2, at least 5 tender and swollen joints — to qualify. Key inclusion criteria differences were as follows:

• In the DISCOVER studies, patients had to have documented evidence of inadequate response or intolerance to standard PsA therapies, including non-biologic DMARDs (≥ 3 months), apremilast (≥ 4 months), and/or NSAID therapy (≥ 4 weeks). In DISCOVER-1, up to 30% of patients enrolled in had previously received 1 or 2 TNF alpha inhibitors. DISCOVER-2 enrolled patients with no prior exposure to bDMARDs.
• COSMOS enrolled patients with an inadequate response or intolerance to 1 or 2 TNF alpha inhibitors.

In all trials, the patients enrolled were required to have at least 1 PsA subset: distal interphalangeal joint involvement, polyarticular arthritis with absence of rheumatoid nodules, arthritis mutilans, asymmetric peripheral arthritis, or spondylitis with peripheral arthritis. Patients were also required to have active plaque psoriasis with 1 or more lesions at least 2 cm in diameter, nail changes of psoriasis, or documented history of plaque psoriasis.

Key exclusion criteria were the presence of other inflammatory diseases, a nonplaque form of psoriasis or drug-induced psoriasis, significant risk of suicide, recent serious infection, or herpes zoster, tuberculosis, HIV, hepatitis B, or hepatitis C. Those with prior treatment with JAK inhibitors or recent exposure to a prohibited medication without an adequate washout period were also excluded.

Baseline Characteristics

The demographics of patients were generally similar within and across trials (Table 8). The mean age of patients enrolled ranged from 44.9 years (SD = 11.9 years) to 49.1 years (SD = 12.3 years) across treatment groups. The proportion of female patients ranged from 46% to 55%, and most patients were White (89% to 98% in the DISCOVER studies; not reported for the COSMOS study). The mean number of swollen joints ranged from 9.0 (SD = 5.7) to 12.3 (SD = 6.9), and the mean number of tender joints ranged from 18.2 (SD = 10.7) to 21.6 (SD = 13.1). Approximately 2-thirds of patients had psoriatic involvement affecting at least 3% of their BSA. Two-thirds of patients reported enthesitis, while approximately 40% had dactylitis at baseline. Some potential imbalances were noted in the COSMOS study in the proportion of female patients (55% versus 46%) and psoriasis at baseline (70% versus 55%), and in the DISCOVER-2 study in the proportion with enthesitis at baseline (64% versus 73%) in the groups receiving guselkumab every 8 weeks versus placebo, respectively.

Across the studies, 89% to 95% of patients per treatment group had received a prior non-biologic DMARD, of which methotrexate was the most commonly received drug (80% to 92%). Prior use of any immunosuppressive or apremilast was less common, and was reported by 2% to 6% of patients. In the DISCOVER-1 and DISCOVER-2 studies, 39% to 48% of patients had previously received systemic corticosteroids, while 87% to 92% had received NSAIDs (data not reported for COSMOS).

In the DISCOVER-1 study, 41 patients (32%) and 39 patients (31%) in the guselkumab and placebo groups, respectively, had received prior TNF alpha inhibitors, including 7 and 4 patients, respectively, who had received 2 prior drugs from this class. The reasons for stopping TNF alpha inhibitor therapy in the guselkumab and placebo groups, respectively, were financial (42% and 41%), inadequate response (37% and 31%), adverse events (2% and 8%), contraindications (5% and 3%), or unspecified (22% and 23%). In the COSMOS studies, all patients had received TNF alpha inhibitors. Among them, 88% had received 1 prior drug and 12% had received 2 prior TNF inhibitors. As per the inclusion criteria, no patients in DISCOVER-2 had prior exposure to TNF inhibitors.

Table 8

Summary of Baseline Characteristics.

Clinical Response in PsA Symptoms

In all studies, a trained, independent joint assessor who was not otherwise involved in the conduct of the trial evaluated each of 68 joints for tenderness and each of 66 joints for swelling (hips were excluded for swelling). This assessor also conducted enthesitis and dactylitis assessments.

ACR 20, ACR 50, and ACR 70 are measurements of improvement in multiple disease assessment criteria. The ACR 20 is defined as a greater than or equal to 20% improvement from baseline in both swollen joint count (66 joints) and tender joint count (68 joints), and a greater than or equal to 20% improvement from baseline in 3 of the following 5 assessments: the patient’s assessment of pain (visual analogue scale [VAS]), the patient’s global assessment of disease activity (arthritis, VAS), the physician’s global assessment of disease activity (VAS), the HAQ-DI, and CRP. ACR 50 and ACR 70 response criteria require changes from baseline of 50% and 70%, respectively. The ACR 20 is generally accepted as the MID, indicating a response to treatment, while the ACR 50 and ACR 70 more likely reflect truly important change for the long-term management of arthropathy.^43,44

The MDA is a composite measure that encompasses different aspects of PsA disease.⁴⁵ To meet the MDA criteria, patients had to achieve at least 5 out of the following 7 items: tender joint count less than or equal to 1, swollen joint count less than or equal to 1, PASI score less than or equal to 1, patient’s assessment of pain less than or equal to 15 (VAS), patient’s global assessment of disease activity less than or equal to 20 (arthritis and psoriasis, VAS), HAQ-DI score less than or equal to 0.5, and tender entheseal points less than or equal to 1.⁴⁵

The Disease Activity Score 28 (DAS 28) using CRP (DAS 28 CRP) is composite measure of disease activity developed for patients with rheumatoid arthritis. It combines tender joints (28 joints) and swollen joints (28 joints) of the arms, shoulders, and knees with the CRP and the patient’s global assessment of disease activity.⁴⁶ The range of the DAS 28 is 0 to 9.4, with higher scores indicating more active disease.⁴⁷ No evidence was found to support the validity, reliability, or responsiveness in patients with PsA. According to the clinical expert consulted by CADTH, the relevance of this outcome measure in patients with PsA has been questioned because the DAS 28 does not include an assessment of inflammation in the feet, which is more common in patients with PsA than in those with rheumatoid arthritis. Assessment of 66 and 68 swollen and tender joints over the 28-joint count in patients with PsA has been endorsed by GRAPPA and the Outcome Measures in Rheumatology working group.⁴⁸

Functioning or Disability

The HAQ-DI is a patient-reported, 20-question instrument that assesses the degree of difficulty a patient has in accomplishing tasks in 8 functional areas (dressing, arising, eating, walking, hygiene, reaching, gripping, and activities of daily living). Each functional area is scored from 0 (indicating no difficulty) to 3 (indicating inability to perform a task in that area). The overall score is the average of 8 domains, ranging from 0 (no disability) to 3 (completely disabled). The Clinical Study Report (CSR) states that in PsA, a decrease in score of 0.35 has been determined to indicate a clinically meaningful improvement.⁶ Other sources have reported MIDs of 0.131 and 0.16 in patients with PsA.^49,50

Health-Related Quality of Life

The SF-36 is a 36-item general health status instrument that consists of 8 health domains: physical functioning, role physical, bodily pain, general health, vitality, social functioning, role emotional, and mental health. The SF-36 also provides 2 component summaries, the PCS and the MCS, which are derived by aggregating the 8 domains according to a scoring algorithm. All domain and component scores are based on a scale of 0 to 100, with higher scores indicating higher health status. Scores were standardized with a mean of 50 points and an SD of 10 points in the general US population. The CSR defined a change of greater than or equal to 5 points in the PCS and MCS as a clinically meaningful improvement; however, this threshold was not specific to patients with PsA.^11,51 MIDs of 3.74 points for the PCS and 1.77 points for the MCS have been reported for patients with PsA.^7,52

Psoriasis-Specific Measures

Psoriasis-specific end points were evaluated in the subgroup of patients with psoriatic involvement affecting at least 3% of their BSA and an IGA score of 2 (mild) or higher at baseline.

The IGA is a composite score of physician assessment of the overall severity of the patient’s psoriatic lesions. The static version of the IGA was used in all included studies, and is a measurement of disease severity at a given time point. The investigator assessed the overall severity of induration, erythema, and scaling of lesions based on a 5-point scale scored as 0 (no evidence), 1 (minimal), 2 (mild), 3 (moderate), and 4 (severe). The IGA score was the average of the induration, erythema, and scaling scores. The patient’s psoriasis was assessed as cleared (0), minimal (1), mild (2), moderate (3), or severe (4).⁵³ IGA response was defined as a score of 0 or 1 and at least a 2-point decrease from baseline.

The PASI is a widely used instrument in psoriasis trials that grades the extent and severity of psoriatic lesions. It combines an assessment of the BSA affected in 4 anatomic regions (head, trunk, arms, and legs) and the severity of desquamation, erythema, and plaque induration or infiltration (thickness) in each region. Scores range from 0 points to 72 points, with a PASI score greater than 10 representing more severe disease.⁵⁴ A PASI 75, PASI 90, or PASI 100 response is defined, respectively, as a greater than or equal to 75%, greater than or equal to 90%, or greater than or equal to 100% improvement in PASI score from baseline. No MID for the PASI has been estimated among patients with PsA.

Other Musculoskeletal Outcomes

Dactylitis was assessed in both hands and feet using a scoring system from 0 to 3 for each digit (where 0 = no dactylitis, 1 = mild dactylitis, 2 = moderate dactylitis, and 3 = severe dactylitis). The results were summed to produce a final score ranging from 0 to 60. Enthesitis was assessed using the Leeds Enthesitis Index (LEI), which counts the number of painful entheses among the left and right lateral epicondyle humerus, left and right medial femoral condyle, and left and right Achilles tendon insertion. The LEI index ranges from 0 to 6.⁵⁵ No MID was identified for the LEI and dactylitis score. In all trials, only patients with dactylitis or enthesitis at baseline were included in the analysis of these end points.

The BASDAI is a patient-reported assessment for ankylosing spondylitis. It consists of 6 questions relating to the 5 major symptoms of ankylosing spondylitis (fatigue, spinal pain, peripheral joint pain, localized tenderness, and morning stiffness). Each question is scored using a 10-unit VAS to indicate the degree of symptoms over the past week. The total score ranges from 0 to 10, with higher scores indicate greater disease severity.^56,57 According to the CSR, a 50% decrease in the score is clinically meaningful.^11,58 The MID in patients with PsA is unclear. In the clinical trials, only patients with spondylitis with peripheral arthritis as the primary arthritic presentation of PsA completed the BASDAI.

Symptoms

FACIT-Fatigue consists of 13 questions that assess a patient’s level of fatigue and tiredness over the last 7 days. Each question was graded on a 5-point scale (where 0 = not at all, 1 = a little bit, 2 = somewhat, 3 = quite a bit, and 4 = very much). The score ranges from 0 to 52, with lower scores reflecting more severe fatigue.⁸ The CSR states that in rheumatology, a change of 4 points is considered meaningful and has been used in the PsA population.⁸ An MID of 3.1 points has also been reported for patients with PsA.⁹

Radiologic End Points

For the assessment of radiographic changes, single radiographs of the hands and feet were evaluated by central independent readers and scored using the modified vdH-S score. The modified vdH-S score, which ranges from 0 (best) to 528 (worst), is the sum of the joint erosion score and the joint space narrowing score.¹⁰ In addition to reporting the change from baseline in the modified vdH-S score, the DISCOVER-2 study also reported the number of patients with radiographic progression based on the smallest detectable change. The smallest detectable change is defined as the amount of change from baseline for which any smaller change cannot be reliably distinguished from random error in the measurement. The MID in patients with PsA is not known.

Safety

Treatment-emergent adverse events included any adverse events that occurred after the start of study drug administration and any adverse events that were present at baseline but worsened in severity after the start of study drug administration. Serious adverse events included any event that resulted in death, was life-threatening, required hospitalization or prolongation of existing hospitalization, resulted in persistent or significant disability, was a congenital anomaly, was the suspected transmission of an infectious drug through a medicinal product, or was medically important (e.g., required intervention to prevent 1 of the other outcomes listed).

Primary Analysis Methods

In all of the trials, the primary analysis of efficacy outcomes was based on the composite estimand, which assessed the treatment effects based on the outcome variable as well as events that met the treatment failure criteria. Patients who met any of the treatment failure criteria up to week 24 were considered nonresponders for the binary (responder) outcomes, or were considered as no change from baseline for the continuous outcomes. Treatment failure was defined as 1 of the following: discontinued study drug injections due to any reason; terminated study participation due to any reason; initiated or increased the dose of non-biologic DMARDs (methotrexate, sulfasalazine, hydroxychloroquine, leflunomide) or oral corticosteroids over baseline for PsA; initiated protocol-prohibited medications or therapies for PsA; or met early escape criteria (COSMOS only). An alternate estimand, the treatment policy strategy, was the primary estimand for radiologic outcomes in DISCOVER-2 (refer to the Sensitivity and Subgroup Analyses section for details).

The statistical methods used in the included studies have been summarized in Table 10. The primary outcome (i.e., ACR 20 at week 24) and other binary outcomes were tested using a CMH test (or Fisher’s exact test, if CMH was not appropriate) stratified by baseline factors specific to each trial. In DISCOVER-1, the stratification factors were the use of a non-biologic DMARD at baseline (yes, no) and prior exposure to TNF alpha inhibitors (yes, no). DISCOVER-2 patients were stratified by baseline use of non-biologic DMARDs (yes, no) and most recent CRP value (< 2.0 mg/dL versus ≥ 2.0 mg/dL). In the COSMOS study, the stratification factors were non-biologic DMARD use at baseline (yes, no) and number of prior TNF alpha inhibitors (1 or 2). The differences in response rates and 95% CI between the guselkumab and placebo groups were calculated based on the Wald statistic. Patients who met any of the treatment failure criteria were considered nonresponders at week 24, regardless of their ACR 20 response status. In addition, any patients with missing data were imputed as not achieving the response.

In the DISCOVER trials, major secondary continuous outcomes were analyzed using an ANCOVA model based on multiple imputation data, with missing data imputed under the missing-at-random assumption (Table 10). The model included covariates for baseline score and each studies’ stratification factors. Other continuous outcomes (e.g., FACIT-Fatigue, BASDAI) were analyzed using an MMRM with no imputation for missing data in the DISCOVER trials. In the COSMOS study, continuous outcomes were analyzed using an MMRM with no imputation for missing data. Any patients who met the treatment failure criteria were imputed as no change (i.e., no improvement) from baseline in the model.

Sensitivity and Subgroup Analyses

Several sensitivity analyses were planned for the primary outcome in the DISCOVER-1 and DISCOVER-2 studies. These included the following: a tipping point analysis where all combinations of missing data, imputed as responders and nonresponders, were analyzed; a treatment policy analysis, which was based on observed ACR 20 response data for all patients regardless of whether treatment failure criteria were met before week 24 and using multiple imputation methods for missing data (missing-at-random assumption); a tipping point analysis based on multiple imputation methods for missing data for the treatment policy estimand; an alternate composite analysis where patients who stopped treatment for reasons other than lack of efficacy were not considered as nonresponders; and a per-protocol analysis that was based on the last non-missing data before stopping therapy or meeting treatment failure criteria in patients who did not have any major protocol violations.

In the DISCOVER trials, subgroup analyses based on treatment history and concomitant DMARD use were of interest to this review. These preplanned analyses for the ACR 20 end point were based on a logistic regression model and reported as odds ratios and 95% CIs. P values for the treatment by subgroup interaction terms were reported.

In the COSMOS study, a supplementary analysis was conducted based on the treatment policy estimand, which evaluated treatment effects using all observed data, regardless of the intercurrent events that met the treatment failure criteria, with imputation for missing data. In this trial, this estimand ignored the week 16 crossover of placebo patients who met the early escape criteria and received guselkumab. The sponsor stated that this analysis biases the results against guselkumab because it does not consider any benefit that patients in the placebo group may have received from switching to active treatment. A second supplementary analysis was conducted where discontinuation of the study drug due to reasons other than lack of efficacy was not considered as treatment failure.

A total of 20 patients in the COSMOS study (8 patients [8%] in the placebo group and 12 patients [6%] in the guselkumab group) were incorrectly assigned to early escape. Patients in the placebo group were incorrectly switched to guselkumab, and there was no change to the study drug received by the patients who had been incorrectly assigned to early escape in the guselkumab group. Two sensitivity analyses were conducted for these patients. The first ignored the early escape criterion, which meant that the 12 patients on guselkumab were not considered to have experienced treatment failure (i.e., no change in therapy); however, the 8 placebo patients were analyzed as a treatment failure because they had received a prohibited medication. The second analysis set the week 16 to week 24 end point values as missing. These were imputed using multiple imputation methods for binary outcomes or MMRM for continuous end points.

The preplanned subgroup analyses based on patient demographics were not of interest to this review and have not been summarized in this report.

Statistical Power

The DISCOVER-1 study was estimated to have greater than 90% power at a 2-sided alpha level of 0.05 for the primary outcome of ACR 20 response. This estimate was based on enrolling a total of 360 patients and assuming a 40% ACR 20 response rate in the guselkumab group and a 20% ACR 20 response in the placebo group at week 24.

For the DISCOVER-2 study, assuming ACR 20 response rates of 45% and 20% in the guselkumab and placebo groups, respectively, a sample size of 228 patients per group would have 99% power to detect a significant treatment effect at a 2-sided significance level of 0.05. With a sample size of 228 patients per group, the trial would have 90% power to detect a difference in the mean change from baseline in the vdH-S score, assuming a mean change of 0.9 in the placebo group and 0.3 in the guselkumab group, with an SD of 2.5. No references were listed to support the assumptions used in the calculations.

The COSMOS study was planned to have 90% power to detect a difference in the primary outcome (i.e., ACR 20) at a 2-sided significance level of 0.05 based on a sample size of 163 patients in the guselkumab group and 82 patients in the placebo group, assuming a response rate of 41% versus 20% for guselkumab and placebo, respectively. The estimated ACR 20 response rates were based on data from an ustekinumab trial in patients previously treated with TNF alpha inhibitors.⁵⁹

Table 10

Statistical Analysis of Efficacy End Points.

Control of Type I Error Rate

In the DISCOVER trials, the overall type I error was controlled at a significance level of less than or equal to 0.05 for the primary outcome (i.e., ACR 20) and select major secondary outcomes. Due to regulatory requirements, there were 2 pre-specified multiplicity adjustment procedures used: 1 for the US and 1 for countries except the US (global). In both procedures, the primary outcome was tested first for the group receiving guselkumab every 4 weeks versus the placebo group, and then for the group receiving guselkumab every 8 weeks versus the placebo group. Testing of select major secondary outcomes proceeded if the primary outcome was statistically significant. In the DISCOVER-1 study, the global testing procedure included 8 major secondary outcomes that were tested for each guselkumab dose group independently in the order specified in Table 11. In the US testing procedure, the 2 dosage groups were tested in parallel, as shown in Figure 5. The overall significance level of less than or equal to 0.05 was controlled for the 24-week IGA response and the HAQ-DI and SF-36 PCS outcomes. Other secondary outcomes were not controlled for type I error rate. The testing procedures stated that ACR 20 at week 16, ACR 50 at week 16, ACR 50 at week 24, ACR 70 at week 24, and change from baseline DAS 28 at week 24 were tested (with a nominal P value) only if the primary outcome for the corresponding guselkumab dosage group was statistically significant.

In the DISCOVER-2 study, the US-based testing procedure controlled the type I error at an overall significance level of less than or equal to 0.05 for the primary outcome (i.e., ACR 20) and for 3 major secondary end points that were not highly correlated with each other (IGA response, HAQ-DI, and SF-36 PCS at week 24). Figure 6 shows the testing procedure for these 4 outcomes in the 2 dosage groups. The global testing procedure controlled the type I error rate for 12 outcomes, as shown in Figure 7, and included a pooled analysis of the resolution of enthesitis or dactylitis based on data from DISCOVER-1 and DISCOVER-2. Five other outcomes highly correlated with ACR 20 were tested (with nominal P values) only if statistical significance was achieved for ACR 20 response at week 24 for the corresponding guselkumab dose group. These outcomes included ACR 20 and ACR 50 response at week 16, ACR 50 and ACR 70 response at week 24, and change from baseline in DAS 28 at week 24. Other secondary outcomes were not formally tested, and nominal P values were reported. Of note: the Health Canada reviewer’s report stated that the US analyses were deemed the preferred method.⁴⁰

For the COSMOS study, the type I error rate was controlled for the primary outcome and 4 major secondary outcomes. These were tested in a fixed sequence (week 24 HAQ-DI, ACR 50, SF-36 PCS, PASI 100 response). Statistical testing was performed sequentially only if the previous outcome reported a P value of less than 0.05. The P values for the other outcomes reported were not controlled for multiple testing, and the sponsor stated that these were descriptive and not to be used to determine statistical significance.

Table 11

Hierarchical Sequence for Testing Secondary Outcomes.

Figure 5

US-Specific Multiplicity Testing Procedure — DISCOVER-1 — Redacted.

Figure 6

US-Specific Multiplicity Testing Procedure — DISCOVER-2 — Redacted.

Figure 7

Global Multiplicity Testing Procedure — DISCOVER-2 — Redacted.

Analysis Populations

In all studies, most efficacy outcomes were analyzed based on the FAS, which included all randomized patients who received at least 1 dose of the study drug according to their randomized group. Psoriasis-specific outcomes (i.e., IGA or PASI response) were analyzed in specific FAS subpopulations that had greater than or equal to 3% BSA psoriatic involvement and an IGA score of greater than or equal to 2 at baseline. Patients included in the enthesitis analyses had to have at least 1 tender enthesis among the 6 sites included in the LEI at baseline. The dactylitis outcomes were analyzed in patients with dactylitis at baseline, and the BASDAI was analyzed in the subset of patients with spondylitis and peripheral joint PsA as their primary presentation. DISCOVER-2 also included a preplanned, pooled analysis of enthesitis and dactylitis outcomes that were based on relevant subgroup data from the DISCOVER-1 and DISCOVER-2 trials.

The safety population included all randomized patients who received at least 1 dose of the study drug and were analyzed according to the actual treatment received.

Clinical Response in PsA Symptoms

In the DISCOVER-1 study, which enrolled both biologic-naive and biologic-experienced patients, 52.0% of patients in the group receiving guselkumab every 8 weeks achieved ACR 20 at 24 weeks compared with 22.2% of patients in the placebo group. The absolute difference was 29.8% (95% CI, 18.6% to 41.1%; P < 0.001) (Table 14) favouring guselkumab versus placebo. At week 24, 29.9% of patients versus 8.7% of patients in the guselkumab versus placebo groups, respectively, achieved ACR 50 (absolute difference = 21.4%; 95% CI, 12.1% to 30.7%; P < 0.001), and 11.8% versus 5.6% achieved ACR 70 (6.4%; 95% CI, −0.3% to 13.1%; P = 0.086). ACR 50 and ACR 70 outcomes were controlled for multiple testing based on the global testing procedure rather than the US-specific testing procedure. The US-specific procedure was preferred by Health Canada.

Among patients who were biologic-naive (in DISCOVER-2), 64.1% and 32.9% of patients achieved ACR 20 in the group receiving guselkumab every 8 weeks and the placebo group, respectively, with an absolute difference of 31.2% (95% CI, 22.9% to 39.5%; P < 0.001) (Table 14). The difference between guselkumab and placebo in the proportion of patients who achieved ACR 50 was 17.2% (95% CI, 10.0% to 24.4%), and the proportion of patients who achieved ACR 70 was 14.5% (95% CI, 9.1% to 19.9%). However, neither outcome was statistically tested according to the global testing procedure due to failure of a prior outcome in the hierarchy. ACR 50 and ACR 70 were not controlled for multiple testing in the US-specific procedure.

Among biologic-experienced patients in the COSMOS study, 44.4% and 19.8% achieved ACR 20 at week 24 in the guselkumab and placebo groups, respectively. The absolute difference between groups (24.6%) favoured guselkumab (95% CI, 14.1% to 35.2%; P < 0.001) (Table 14). The difference for guselkumab versus placebo for the proportion who achieved ACR 50 was 19.6% versus 5.2%; the absolute difference was 14.3% (95% CI, 7.2% to 21.4%; P < 0.001). ACR 70 response was achieved by 7.9% versus 1.0% of patients in the guselkumab versus placebo groups, with an absolute difference of 6.8% (95% CI, 2.6% to 11.1%). ACR 70 was not controlled for multiple testing and should be interpreted considering the inflated risk of type I error rate.

In the DISCOVER-1 and 2 studies, the results of the sensitivity analyses for ACR 20 were generally consistent with the results of the primary analysis. In the COSMOS study, the analysis that was based on observed case data regardless of meeting treatment failure criteria reported an absolute difference of 10.9% (95% CI, −1.3% to 23.2%), which the investigators stated was due to the doubled ACR 20 response rate in the placebo group, noting that almost half of patients in the placebo group received early escape therapies starting at week 16. Other sensitivity analyses (per-protocol population and analyses adjusting for errors in early escape assignment) showed results that were consistent with the primary analysis.

The proportion of patients who achieved MDA at week 24 was higher in the group receiving guselkumab every 8 weeks versus the placebo group in all 3 studies. However, this outcome was not controlled for multiple testing and should be interpreted with caution because of the potential for inflated type I error rate. The absolute differences between guselkumab and placebo were 11.9% (95% CI, 2.9% to 20.9%) for DISCOVER-1, 18.9% (95% CI, 12.8% to 25.0%) for DISCOVER-2, and 11.7% (95% CI, 5.6% to 17.7%) for the COSMOS study (Table 14).

Subgroup analyses of interest to this review for the DISCOVER studies are shown in figures 11 to 14 in Appendix 3. The odds ratios for ACR 20 response were generally consistent across subgroups based on prior use of TNF alpha inhibitors, non-biologic DMARDs, oral corticosteroids, or NSAIDs at baseline. The COSMOS study did not report data for any subgroups of interest.

Table 14

ACR and MDA Response at Week 24.

Additional data on the ACR 20 and ACR 50 response rates at week 16 are summarized in Appendix 3, Table 42. These findings were generally consistent with the ACR 20 results at week 24; however, the ability to draw statistical inferences from the data may be limited, whether because no statistical testing was performed (DISCOVER-2), there was no control for multiple testing (COSMOS), or control of the type I error rate was based on global rather than US-specific testing procedures (DISCOVER-1). For ACR 50, the DISCOVER-1 study failed to detect a statistically significant difference between guselkumab every 8 weeks and placebo at week 16 (based on global testing procedure). No statistical testing was performed in DISCOVER-2 (due to failure of a prior outcome). The response rates were nominally higher for the guselkumab versus placebo groups in the COSMOS study, but there was no control of the type I error rate.

Data on the changes from baseline in DAS 28 CRP scores are summarized in Appendix 3, Table 43. The point estimates for the LS mean differences ranged from −0.61 to −0.73, suggesting an improvement with guselkumab every 8 weeks versus placebo in all trials; however, in the COSMOS study, this outcome was not controlled for multiple testing, and in the DISCOVER studies, it was controlled for multiple testing based on the global testing procedure rather than the US-specific procedure.

Function or Disability

Changes from baseline to week 24 in the HAQ-DI favoured guselkumab every 8 weeks versus placebo in all trials. The LS mean differences reported were −0.25 (95% CI, −0.36 to −0.13; P < 0.001) in the DISCOVER-1 study, −0.24 (95% CI, −0.32 to −0.15; P < 0.001) in the DISCOVER-2 study, and −0.17 (95% CI, −0.28 to −0.06; P = 0.003) in the COSMOS study. Across the trials, only the group receiving guselkumab every 8 weeks in the DISCOVER-2 study achieved a change from baseline that exceeded the 0.35 MID cited in the study’s protocol.^6,7,51

The proportion of HAQ-DI responders was reported as another secondary outcome in the DISCOVER trials (i.e., no control for type I error rate). Response was defined achieving a point improvement of greater than or equal to a 0.35 in the HAQ-DI score from baseline (among the subset of patients who had a HAQ-DI score ≥ 0.35 points at baseline). At 24 weeks in the DISCOVER-1 study, 50.9% of patients (57 of 112) in the guselkumab group and 29.1% of patients (32 of 110) in the placebo group achieved HAQ-DI responses (absolute difference = 21.8%; 95% CI, 9.3% to 34.2%). In DISCOVER-2, 50.0% of patients (114 of 228) and 31.4% of patients (74 of 236) in the guselkumab versus placebo groups, respectively, achieved HAQ-DI responses (absolute difference = 18.7%; 95% CI, 10.0% to 27.3%).

Table 15

Change From Baseline in HAQ-DI to Week 24.

Health-Related Quality of Life

The change from baseline in the SF-36 PCS was a secondary outcome in all studies. Most treatment groups reported an improvement at week 24, with a difference that favoured guselkumab every 8 weeks versus placebo in all studies (Table 16). The LS mean difference was 4.1 (95% CI, 2.4 to 5.9; P < 0.001) in the DISCOVER-1 study, 4.0 (95% CI, 2.7 to 5.2; P = 0.011) in the DISCOVER-2 study, and 3.9 (95% CI, 2.5 to 5.4; P < 0.001) in the COSMOS study. The CSR defined an increase of at least 5 points as clinically meaningful, but a MID of 3.74 has also been reported in the literature.^7,51

All trials failed to detect a statistically significant difference between guselkumab every 8 weeks and placebo in the change from baseline to week 24 in the SF-36 MCS (Table 16).

Table 16

Change From Baseline in the SF-36 PCS and MCS to Week 24.

Psoriasis Skin Disease

In all trials, psoriasis skin disease outcome measures were analyzed in the subgroup of patients who had psoriasis affecting greater than or equal to 3% of their BSA and an IGA score of greater than or equal to 2 at baseline (55% to 74% of patients per treatment group). Randomization was not stratified by the presence of skin disease at baseline.

The proportion of patients who achieved an IGA response at week 24 was higher in the group receiving guselkumab every 8 weeks than in the placebo group in the DISCOVER-1 study (57.3% versus 15.4%; absolute difference = 42.0% [95% CI, 28.9% to 55.1%; P < 0.001]) and the DISCOVER-2 study (70.5% versus 19.1%; absolute difference = 50.9% [95% CI, 42.2% to 59.7%; P < 0.001]). In the COSMOS study, 48.1% versus 9.4% of patients achieved an IGA response in the guselkumab and placebo groups; the absolute difference was 38.8% (95% CI, 27.3% to 50.4%). However, the P value was not adjusted to control for multiple testing; it should be interpreted with caution because of the potential for inflated type I error rate.

PASI 100 response at week 24 was a secondary outcome in the COSMOS study. In the group receiving guselkumab every 8 weeks, 30.8% of patients achieved a PASI 100 response compared with 3.8% of patients in the placebo group (absolute difference = 27.4%; 95% CI, 17.9% to 36.8%; P < 0.001]). The PASI 90 and PASI 75 results in the COSMOS study were nominally higher in the guselkumab versus placebo groups, but the type I error rate was not controlled for these outcomes. The PASI 100, PASI 90, and PASI 75 response rates in the DISCOVER trials were nominally higher for guselkumab versus placebo, but these outcomes were not part of the statistical testing procedure and should be interpreted with caution, considering the potential for inflated type I error rate (Table 17).

Table 17

IGA and PASI Response at Week 24.

Other Musculoskeletal Disease

Outcomes related to enthesitis and dactylitis were analyzed for patients reporting these symptoms at baseline. Approximately 2-thirds of patients reported enthesitis at baseline (57% to 72%), whereas 35% to 45% of patients reported dactylitis across treatment groups in the 3 trials. Randomization was not stratified by the presence of enthesitis or dactylitis at baseline. These outcomes were not controlled for multiple testing; thus, P values of less than 0.05 should be interpreted with caution, considering the potential for inflated type I error rate. The results for these outcomes in each trial are shown in Table 18.

The DISCOVER-2 study planned a prior pooled analysis of the resolution of enthesitis and dactylitis end points using combined data from the DISCOVER-1 and DISCOVER-2 trials. Based on the pooled analysis, 49.6% and 29.4% of patients in the group receiving guselkumab every 8 weeks and the placebo group had resolution of enthesitis at week 24, with a between-group difference of 20.1% (95% CI, 11.8% to 28.5%; P = 0.03 [based on the US-specific testing procedure]). The proportion of patients whose dactylitis resolved at week 24 was 59.4% versus 42.2% in the guselkumab versus placebo group; the between-group difference was 18.0% (95% CI, 7.4% to 28.6%; P = 0.03 [based on the US-specific testing procedure]). In accordance with the global testing procedure, these outcomes were not formally tested due to the failure of a prior outcome in the testing hierarchy.

Table 18

Enthesitis and Dactylitis Outcomes at 24 Weeks.

The change in the BASDAI score was analyzed for the subgroup of patients with spondylitis and peripheral joint PsA as their primary presentation at baseline (18% to 37% of patients per treatment group across the studies). Randomization was not stratified by the presence of axial disease at baseline. The sponsor stated that a 50% improvement in the BASDAI represents a clinically important change.⁵⁸

Among biologic-naive patients (DISCOVER-2), 39% and 22% in the guselkumab and placebo groups, respectively, achieved at least a 50% improvement in the BASDAI score at week 24 (absolute difference = 15%) (95% CI, 0.4% to 30%; not controlled for type I error rate) (Table 19). The proportion of patients with an improvement of greater than or equal to 50% was the same in both treatment groups (19%) among patients who were biologic-experienced in the COSMOS study. In the mixed population enrolled in the DISCOVER-1 study, 42% and 13% of patients in the guselkumab and placebo groups achieved at least a 50% improvement, respectively, with an absolute difference of 24% (95% CI, −1% to 49%).

Table 19

Improvement in BASDAI Score at Week 24.

PsA Symptoms

FACIT-Fatigue scores range from 0 to 52, with higher scores indicating less fatigue. This outcome was not controlled for multiple testing; thus, P values of less than 0.05 should be interpreted with caution, considering the potential for inflated type I error rate. MIDs of 3.1 to 4 points have been reported in the literature.^8,9

The FACIT-Fatigue scores suggest improvement in fatigue, with point estimates for the within-group change from baseline to week 24 that exceeded the MID for all groups receiving guselkumab every 8 weeks and for the placebo group in the DISCOVER-2 study (Table 20). The DISCOVER-1 study reported an LS mean difference of 3.4 points (95% CI, 1.4 to 5.4); the DISCOVER-2 study reported an LS mean difference of 4.0 (95% CI, 2.5 to 5.5); and the COSMOS study reported an LS mean difference of 3.6 (95% CI, 1.7 to 5.4) for guselkumab every 8 weeks versus placebo. In all studies, the between-group difference exceeded the MID of 3.1, while only the DISCOVER-2 study met the 4-point threshold.

Radiologic Outcomes

The change from baseline to week 24 in the modified vdH-S score was a major secondary outcome in the DISCOVER-2 study. At 24 weeks, the LS mean difference observed for guselkumab versus placebo was −0.43 (95% CI, −0.90 to 0.03) (P = 0.068) (Table 21).

The proportions of patients with no progression (defined as a change from baseline that was less than or equal to the smallest detectable change for the modified vdH-S total score) were 89.0% in the group receiving guselkumab every 8 weeks and 84.0% in the placebo group (absolute difference = 5.0%; 95% CI, –1.0% to 10.9%).

Table 20

Change From Baseline in FACIT-Fatigue Score to Week 24.

Table 21

Change From Baseline in Modified vdH-S of PsA to 24 Weeks.

Adverse Events

The frequency of adverse events was generally similar between groups, with 42% to 54% of patients in the groups receiving guselkumab every 8 weeks and 41% to 60% of patients in the placebo groups reporting 1 or more adverse events during the 24-week treatment period (Table 22). Nasopharyngitis (4% to 13%), upper respiratory tract infection (2% to 5%), and increased alanine aminotransferase (2% to 6%) were the most commonly reported adverse events in the guselkumab groups. The corresponding proportions of patients experiencing these adverse events in the placebo groups were 4% to 6% for nasopharyngitis, 3% to 7% for upper respiratory tract infections, and 3% to 6% for increased alanine aminotransferase.

In the COSMOS study, adverse event data were reported separately for 45 patients who switched from placebo to guselkumab every 8 weeks before week 24. Six patients (13%) experienced an adverse event, including 1 patient with a serious adverse event (a buttock injury) and 2 patients with infections (4%).

Serious Adverse Events

The frequency of serious adverse events ranged from 1% to 4% in the groups receiving guselkumab every 8 weeks and from 3% to 4% in the placebo groups (Table 22). No specific serious adverse events were reported in more than 1 patient per treatment group (Table 23). In the DISCOVER-1 study, only 1 of the serious adverse events (plasma cell myeloma in the guselkumab group) resulted in treatment discontinuation. One patient in the guselkumab group of the COSMOS study experienced increased alanine aminotransferase that was categorized as a serious adverse event. This patient was subsequently diagnosed with autoimmune hepatitis and the study drug was stopped.

Treatment Discontinuation Due to Adverse Events

Adverse events that resulted in treatment discontinuation were generally low (1% to 3%) and similar between groups (Table 22). Other than 2 patients in the placebo group of the DISCOVER-1 study, who stopped treatment due to psoriasis, all other events were reported in 1 patient (Table 23).

Mortality

In the DISCOVER-1 study, 1 patient in the placebo group died due to cardiac failure. No other deaths were reported in the guselkumab groups or placebo groups during the first 24 weeks of the 3 studies.

Notable Harms

Generally, the frequency of infections was similar in the groups receiving guselkumab every 8 weeks (16% to 26%) and the placebo groups (18% to 25%) across trials (Table 22). In the DISCOVER-1 study, no patients in the guselkumab group reported a serious infection. Two patients in the placebo group experienced a serious infection, which included a serious upper respiratory tract infection and an abscessed limb. In the DISCOVER-2 study, 1 patient in each group reported a serious infection, including pyrexia in the guselkumab group and post-procedural fistula in the placebo group. In the COSMOS study, 1 patient in the guselkumab group reported a serious community-acquired pneumonia that required hospitalization. No serious infections were reported among patients who received placebo. No cases of tuberculosis or opportunistic infections were reported in the first 24 weeks of all 3 trials.

The other notable harms identified in the systematic review protocol were infrequently reported (i.e., injection-site reactions, anaphylactic adverse events, and liver-related adverse events).

Table 22

Summary of Harms Up to 24 Weeks (Safety Set).

Table 23

Description of SAEs or Adverse Events Leading to Treatment Discontinuation Up to 24 Weeks (Safety Set).

Internal Validity

The risk of bias related to randomization and treatment allocation concealment was rated as low for all studies, and, in general, the patient characteristics and co-interventions appeared to be balanced between groups at baseline. However, some differences were noted in the proportion of female patients and patients with psoriasis at baseline in the COSMOS study and in the proportion of patients with enthesitis at baseline in the DISCOVER-2 study. No issues were identified in the methods used to randomize patients and conceal allocation. The trials were double blind and took steps to maintain blinding of patients and investigators. In the DISCOVER studies, where the frequency of dosing varied across treatment groups (i.e., every 4 weeks or every 8 weeks), patients in the group receiving the study drug every 8 weeks also received placebo injections at alternate visits to match the every-4-weeks dosing schedule of the other guselkumab group. In all studies, an independent joint assessor not otherwise involved in the conduct of the trial evaluated patients’ joints for tenderness and swelling and the presence of enthesitis and dactylitis. Central raters assessed radiographs in DISCOVER-2. Therefore, the risk of bias in the measurement of the outcomes was low for all trials. In general, the type and frequency of adverse events were similar across groups, so there was low risk that adverse events led to substantial unblinding.

In all trials, the primary and other dichotomous end points were analyzed using a CMH test that was stratified by randomization stratification factors, with missing data imputed as nonresponders. The DISCOVER trials used an adjusted ANCOVA model, while the COSMOS study used an unadjusted MMRM model to analyze continuous outcomes. Missing data were imputed under the missing-at-random assumption, which may not hold true; however, this was not thought to be a major source of bias, given the low frequency of withdrawals and the use of the composite estimand to conduct the analyses. The efficacy analyses were based on the FAS (randomized and treated), which is not a true intention-to-treat population; however, only 1 patient in the DISCOVER-2 study was excluded. As a result, the analyses were appropriate for estimating the effect of assignment to the intervention.

All efficacy analyses were based on the composite estimand, where any patients who met the treatment failure criteria were considered nonresponders for binary end points, or to have experienced no change from baseline for continuous measures (e.g., HAQ-DI). Based on the composite estimand, patients who discontinued the study or study drug, initiated or increased the dose of cDMARDs or corticosteroids, initiated prohibited medication, or, in the COSMOS study, met early escape criteria, were considered treatment failures. This estimand, which considers any treatment failure end points to be unfavourable outcomes, may be considered a more conservative estimate of treatment effects, and was the estimand preferred by Health Canada. Of note, the COSMOS study incorrectly assigned 20 patients to early escape even though they had not met the escape criteria. As a result, 8 patients in the placebo group (8.3%) were switched to guselkumab treatment at week 16, and 12 patients (6.3%) in the guselkumab group were analyzed as nonresponders in the primary analysis. Although the sponsor conducted sensitivity analyses to explore the impact of this error, these analyses cannot fully address the potential bias.

The type I error rate was controlled for the primary and selected secondary outcomes in all studies. Two different testing strategies were used in the DISCOVER trials to meet US and other regulators’ requirements. The Health Canada reviewer’s report indicated that the US-based strategy was preferred.⁴⁰ Several outcomes of interest to this review were tested, and nominal P values reported (e.g., MDA, ACR 70, PASI response, resolution of enthesitis and dactylitis, FACIT-Fatigue); however, any results with a P of less than 0.05 should be interpreted with caution, considering the potentially inflated type I error rate. Although the subgroup analyses were pre-specified, there is no evidence that the studies were powered to detect subgroup differences, and the 95% CIs for some subgroups were imprecise.

Of note, skin-related end points — as well as enthesitis, dactylitis, and axial disease outcome measures — were tested in subgroups of patients affected with these conditions at baseline. Randomization was not stratified for the presence of these conditions; thus, the interpretation of the results for these outcomes should consider the possibility of imbalances in baseline demographic and disease characteristics between treatment groups in these subpopulations. According to the expert consulted by CADTH, the DAS 28 is not relevant for PsA because it assesses only 28 joints and excludes the lower extremities (which may be important in PsA).⁴⁸ Moreover, CADTH found limited data on the validity — and no data on the reliability and responsiveness — of the DAS 28 in patients with PsA. Thus, although this end point was a major secondary outcome in the DISCOVER trials, it should be interpreted with caution. Based on CADTH’s review, there was also limited evidence to support the validity or reliability of the BASDAI and the enthesitis and dactylitis scales used in the clinical trials. In addition, the MID for the change in vdH-S scores was unknown.

Although the trials were 48 weeks to 100 weeks in duration, the comparative period was limited to 24 weeks for this chronic condition. For outcomes such as radiologic changes, this duration of treatment may have been insufficient to detect the impact of guselkumab every 8 weeks. Moreover, none of the trials included an active control group (other than the group receiving guselkumab every 4 weeks, which is not a Health Canada–approved dosage); thus, direct evidence comparing guselkumab every 8 weeks to other biologics available in Canada is not available.

The primary outcome was ACR 20 response, but according to the clinical expert, this represents the minimum level of improvement that may be relevant to patients. In practice, the goal of therapy is to achieve higher levels of response, such as MDA or low disease activity as measured by the DAPSA. The COSMOS study included ACR 50 in its statistical testing procedures to control the type I error rate, but according to the US-based procedure in the DISCOVER trials, ACR 50 and ACR 70 were not controlled for multiple testing. The MDA end point was not controlled for multiple testing in any of the trials, which limits the ability to draw statistical inference from these data.

External Validity

The trials enrolled adults who, on average, were in their mid to late forties and predominantly White. There were a roughly equal number of male and female patients enrolled. Although patients with only 3 or 5 tender or swollen joints could enrol, the mean number of tender joints was about 20 and the mean number of swollen joints was about 11. The mean PASI score was generally less than 10, which the expert stated was consistent with scores seen in rheumatology clinics, given that most patients had received DMARDs (which may help to control psoriasis). Approximately 8% of patients in the trials had not previously received a cDMARD before being enrolled. Of the patients screened, 13%, 36%, and 39% were excluded. However, based on reporting in the CSRs, there was limited information about the characteristics of these patients and the reasons for their screening failures. Most of the trials were conducted in Europe, with some sites in Asia, the US, Australia, and Israel. Only the DISCOVER-1 study included sites in Canada, which enrolled 15 patients (4%). Based on the inclusion and exclusion criteria and the characteristics of patients, the clinical expert did not identify any substantial limits to the generalizability of the results of the trials.

The guselkumab-every-8-weeks dosing regime used in the trials was consistent with the Health Canada–recommended dose, and the clinical expert stated that concomitant utilization of cDMARDs was similar to what may be expected in practice. However, the clinical expert also noted that the use of oral corticosteroids in the DISCOVER trials (14% to 20%) was higher than would be expected in Canada, where chronic corticosteroid use is generally avoided. The trials excluded patients who had previously been treated with biologics other than TNF inhibitors; thus, the efficacy in patients with intolerance or inadequate response to other biologics is not known.

The trials assessed several of the outcomes that patients report as important. However, not all were part of the statistical testing strategy; as a result, the ability to draw conclusions from these data may be limited. The 24-week duration of the comparative part of the trials may be considered short, given that PsA is a chronic condition. Moreover, the comparator used in the trials (i.e., placebo plus cDMARD) may not reflect current practice, given that most patients who show an inadequate response cDMARDs or bDMARDs DMARDs would be offered biologic therapy.

Objectives

The objective of the sponsor-submitted ITC was to assess the relative efficacy and safety of guselkumab compared with other biologics after 12 weeks to 24 weeks of therapy in patients with active PsA.

Study Selection Methods

The ITC was based on a systematic literature review of RCTs that evaluated biologic and non-biologic DMARDs in adults with active PsA. The review was based on a protocol that was drafted a priori and published to PROSPERO in April 2020. The systematic review was first conducted in 2018, was updated in 2020 and 2021, and was published in 2021.^62,68 Table 24 outlines the patients, interventions, comparators, outcomes, and study designs included in the current version of the sponsor’s systematic review. The review included phase II and III RCTs in adults with active PsA who were treated with 1 or more DMARD or placebo for at least 12 weeks. The review included a broad scope of treatments for PsA; however, the authors made no statement with regards to doses of treatments that would be accepted in the systematic review. No limits were placed on the outcomes of interest to the review.

The systematic review was based on a peer-reviewed literature search of the Embase, MEDLINE, and Cochrane Central databases and a search of ClinicalTrials.gov. Studies were selected based on the criteria listed in Table 24, with screening conducted by 2 reviewers independently. Data were extracted by 1 reviewer and verified by a second. Trials were assessed for study quality by a single reviewer using the National Institute for Health and Care Excellence clinical effectiveness quality assessment checklist. Although not explicitly stated, it appears that no trials were excluded based on study quality.

Additional criteria were applied to determine which studies were included in the quantitative analysis; however, no rationale was provided for these criteria, and it is unclear if these were determined a priori. Based on the description of the excluded studies, the ITC was limited to phase III RCTs of a biologic (including small-molecule inhibitors apremilast, tofacitinib, and upadacitinib) at dosages approved by the FDA or European Medicines Agency. It is not clear how outcomes were selected for analysis in the ITC.

Table 24

Study Selection Criteria and Methods for Sponsor-Submitted ITC.

ITC Analysis Methods

The authors conducted an NMA using Bayesian methods (Table 25). The authors stated that a random-effects model was selected a priori as the primary analysis method due to the known clinical heterogeneity in PsA trials. Unadjusted fixed-effects models would also be run, but would be accepted only if there were compelling reasons, such as the between-trial heterogeneity prior not being updated by the data, leading to conclusions (based on overlap of the 95% CrIs) that are contrary to primary trial results. Further, a baseline risk-adjusted model (adjusting for placebo response, described by the authors as a proxy for measured and unmeasured patient characteristics) was planned to adjust for known heterogeneity in patient and study characteristics, as has been done in other NMAs for PsA.⁶⁹ Model fit was assessed based on 4 factors: estimated regression coefficient (beta) and a 95% CrI that excludes the null for a meta-regression covariate; lower between-study SD compared with the unadjusted model; lower deviance information criterion (by ≥ 5 points) compared with the unadjusted model; and posterior mean of the residual deviance approximately equal to the number of treatment arms. The first 2 factors were given more weight when determining if the baseline risk-adjusted model fit best.

An inconsistency model was used to identify loops with inconsistency, although the authors stated that this method had low power to detect a difference. Of note, the network had few closed loops; only 5 studies had an active comparator. The authors stated that their interpretation of the reliability of estimates was based on their quality assessment of the transitivity assumption and evaluation of the estimated between-trial heterogeneity.

The NMA included trials with outcomes assessed at 12 weeks to 24 weeks. Although not explicitly stated, it appears that each dosage regimen was a separate node. Outcomes assessed include ACR response, PASI response, change from baseline in HAQ-DI, Psoriatic Arthritis Response Criteria response, SF-36 PCS and MCS, change from baseline in vdH-S, adverse events, and serious adverse events. No rationale was provided for the selection of these outcome measures.

The primary analysis included a mixed population of biologic-naive and biologic-experienced patients. Subgroup analyses were run, restricted to patients who were biologic-naive and biologic-experienced. For the overall and biologic-experienced populations, sensitivity analyses were run to test the impact of the early escape errors in the COSMOS study. The primary analysis used the main findings from COSMOS, with a secondary analysis that accounted for errors in early escape.

Table 25

ITC Analysis Methods.

Summary of Included Studies

The systematic review included 169 citations of 80 RCTs. Of these, 46 RCTs were excluded from the ITC because the trials included a treatment or dose that was not approved by the European Medicines Agency or FDA (17 studies), included only 1 relevant treatment (1 study), did not include a biologic (21 studies), or were a phase II study (8 studies). A total of 34 RCTs published between 2004 and 2021 were included in the ITCs.

Fourteen bDMARDs were included in the NMA: abatacept, adalimumab, apremilast; certolizumab pegol, etanercept, golimumab, guselkumab, infliximab, ixekizumab, risankizumab, secukinumab, tofacitinib, upadacitinib, and ustekinumab. Most studies included a placebo group (32 trials). Of the 16 studies that compared active treatments, most included 2 dosing regimens of the same drug. Four trials compared 2 different active treatments. The sample size of the studies ranged from 100 patients to 1,281 patients.

The authors of the ITC acknowledged that there was heterogeneity across studies and provided data tables describing the patient and study characteristics of the included trials. The enrolment criteria of all studies specified adults (≥ 18 years) with active PsA; however, the definition of active PsA varied across trials. Most trials required patients to have at least 3 tender or swollen joints and a history or presence of plaque psoriasis at enrolment. The mean age of patients enrolled ranged from 45.7 years to 53.6 years. In terms of gender and race, 41.9% to 61% of patients enrolled were male, and 76.5% to 99.7% were White (not reported in 2 studies). The mean duration of PsA ranged from 3.1 years to 13.4 years across the studies. At baseline, the mean number of swollen and tender joints ranged from 6.1 to 18.3 and 14 to 27.3, respectively. The proportion of patients with psoriasis affecting more than 3% of their BSA ranged from 42% to 100% (not available in 6 studies). At baseline, the mean PASI score ranged from 5.9 to 12.2 (not reported in 9 studies).

Eighteen trials involved patients who were biologic-naive, 4 trials involved patients who were biologic-experienced, and 12 trials involved a mixed population. In the mixed-population studies, 14% to 61% of patients had prior experience with a biologic treatment. Where reported, 21% to 100% of patients had previously received cDMARDs (not reported in 16 studies). In most studies, patients were receiving methotrexate at baseline (38% to 83% of those enrolled). No information was provided about other co-interventions received during the trials.

The primary time point in 9 trials was 12 weeks. Ten trials had time points of 14 weeks to 16 weeks, and 15 trials lasted 24 weeks. In general, the trials of TNF inhibitors, JAK inhibitors, and apremilast were 16 weeks or shorter, whereas the studies of abatacept and the IL-17, IL-12 to 23, and IL-23 inhibitors were 24 weeks in duration. The exceptions were adalimumab and secukinumab — for which there were data from both shorter- and longer-duration trials — and etanercept, for which there were data from only a 24-week trial. Twelve of the 16 24-week studies (75%) had an early escape route available before the primary efficacy end point, which is a potential source of confounding. There were no details provided on how patients who received escape therapy were handled in each study.

To explore clinical heterogeneity within and across the studies, bar plots of the distribution of placebo responses and scatterplots of the relationships between placebo response and the treatment effect from each study were examined by the authors of the ITC (data not available). The authors stated that the placebo response varied across trials and tended to be lower for stricter ACR categories. Across the trials, the ACR 20 response rate in the placebo groups ranged from 9% to 33% (median = 20%; N = 32). The authors also noted that the trials evaluating guselkumab tended to have a higher-than-average placebo response (i.e., ACR 20 = 20% to 33%). The authors stated that a strong, negative association between treatment effect and placebo response was evident for all ACR response levels, with studies with larger placebo responses appearing to have smaller treatment effects. The authors of the ITC stated that failure to adjust for baseline risk could result in biased estimates of comparative efficacy. A summary of the CADTH assessment of clinical heterogeneity is provided in Table 26.

The authors of the ITC rated most of the included studies as having a low overall risk of bias; however, 53% were rated as having an unclear risk of bias, and 18% were rated as having a high risk of bias in at least 1 domain. Eight of 34 studies (24%) had unclear risk of bias related to randomization and/or allocation concealment. Blinding of patients, care providers, and/or outcome assessors was unclear in 6 studies (18%); and study participants were unblinded in 1 study (3%). The authors of the ITC assessed that the groups were balanced at the start of all studies, and that only 1 study had an unclear risk of bias related to imbalances in dropouts. Seventeen studies (50%) had an unclear risk of bias and 1 study (3%) had a high risk of bias related to the methods used to account for missing data. Three studies (9%) had a high risk of reporting bias.

Table 26

Assessment of Homogeneity for Sponsor-Submitted ITC.

Results

Adverse Events

The NMA of adverse events was conducted using a baseline risk-adjusted, random-effects model for the overall population (32 studies) and biologic-naive populations (17 studies). ||| ||||||| ||||||| |||| |||||||||| ||| |||| | ||||| |||| || |||||||||| ||||||| |||||| |||| ||||||||||| |||||||||| || ||||||||||| ||| ||||| ||||||||||| |||||| ||| |||| |||| |||||||| ||| ||||| ||| || |||| ||||| |||| |||||||||| || ||| |||||||||||||| ||||||||||| ||| ||||||||||| ||||||| |||||||||| ||| ||||||||| |||||||| ||| ||| |||| |||||||| ||| |||| ||| |||| |||||||||| ||| ||| || ||| |||||||||||||||||||| |||||||||| || |||||||| |||||||||| ||||| |||||| |||||| ||||||||| |||| ||||||||||| ||| |||| | ||||| |||| || ||||||| |||||| |||| |||||||||| ||| ||| |||||||||||| ||||| ||||||||||| |||||||| ||| |||| |||| |||||||| ||| ||||| |||| |||| ||||||| ||||||||| |||||||.

Figure 8

Evidence Network for ACR Response in the Overall PsA Population.

Table 27

Key NMA Results for the Overall PsA Population — Guselkumab Every 8 Weeks Versus Biologics.

Figure 9

Evidence Network for ACR Response in Biologic-Naive PsA Population.

Table 28

Key NMA Results for the Biologic-Naive Population — Guselkumab Every 8 Weeks Versus Biologics.

Figure 10

Evidence Network for ACR Response in Biologic-Experienced PsA Population.

Table 29

Key NMA Results for the Biologic-Experienced Population — Guselkumab Every 8 Weeks Versus Biologics.

Description of Studies

The extension period for DISCOVER-1 was from week 24 to week 52, with a safety follow-up phase of 8 weeks after week 52. For DISCOVER-2, the active treatment phase was from week 24 to week 100, with follow-up for efficacy at week 52 and week 100 and a safety follow-up phase of 12 weeks after week 100. For the COSMOS study, the extension period was from week 24 to week 48, with a safety follow-up phase of 8 weeks.

In all 3 trials, during the extension period, groups of patients who had been randomized to receive guselkumab 100 mg every 8 weeks or every 4 weeks continued on the same regimen, whereas those randomized to placebo switched to guselkumab 100 mg every 8 weeks or every 4 weeks. Data from the guselkumab every-4-weeks groups in the DISCOVER-1 and DISCOVER-2 studies (including patients who crossed over from placebo to guselkumab every 4 weeks at 24 weeks) have not been summarized in this section because this dose is not consistent with the Health Canada–recommended dose and dose adjustment, as mentioned earlier in the report. Results for the patients in the placebo group of the COSMOS trial who crossed over to guselkumab 100 mg every 8 weeks at 24 weeks have been provided as supporting information (hereafter referred to as the placebo crossover group).

Statistical Analysis

For the extension periods in the DISCOVER-1, DISCOVER-2, and COSMOS studies, no hypothesis testing of treatment differences was conducted. Descriptive statistics were used to summarize most efficacy data. Analyses of end points were conducted according to randomized treatment groups and based on observed data.

Results

Patient Disposition

Table 30 describes the patient disposition for the extension period. For DISCOVER-1 (week 52), 13 patients (10%) discontinued from the study early, whereas 13 patients (5%) and 27 patients (11%) discontinued the study early in DISCOVER-2 (week 52) and DISCOVER-2 (week 100), respectively. Moreover, in DISCOVER-1 (week 52), DISCOVER-2 (week 52), and DISCOVER-2 (week 100), 11 patients (9%), 14 patients (6%) and 25 patients (10%) discontinued the study drug before end of the study, respectively. The rates of discontinuation of the study drug were higher for DISCOVER-1 and for DISCOVER-2 during the active treatment period compared to the randomized phase.

In the COSMOS study, the rates of discontinuation of the study drug were 12% in the guselkumab-every-8-weeks group and 14% in the placebo group before week 44. However, the rates for early discontinuation of the study drugs were lower during the randomized phase, with 15 patients (8%) and 10 patients (10%) for the guselkumab-every-8-weeks and placebo groups, respectively.

Table 30

Patient Disposition for the Extension Period.

Clinical Response in PsA Symptoms

Table 31 describes ACR and MDA responses for the active treatment period. In the DISCOVER-1 study, 67.9%, 43.4%, and 28.9% of patients achieved an ACR 20, ACR 50, and ACR 70 responses, respectively, at week 52. In the DISCOVER-2 study, 79.1%, 51.3%, and 29.5% of patients achieved ACR 20, ACR 50, and ACR 70 responses, respectively, at week 52. At week 100, the corresponding values were 82.1%, 60.7%, and 39.3%.

For COSMOS, 69.8%, 47.1%, and 27.9% of patients achieved ACR 20, ACR 50, and ACR 70 responses, respectively, at week 48.

Although differences were not tested statistically, in all cases, the rate of response appeared to increase numerically with longer follow-ups. The proportions of patients who maintained ACR 20, ACR 50, and ACR 70 responses at week 52 were 88.5%, 83.8%, and 80.0%, respectively, in DISCOVER-1, and were 91.7%, 87.2%, and 82.6%, respectively, in DISCOVER-2. The proportions of patients who maintained ACR 20, ACR 50, and ACR 70 responses at week 100 were 90.4%, 81.9%, and 80.9% in DISCOVER-2. The proportions of patients who maintained ACR 20, ACR 50, and ACR 70 responses at week 48 were 85.6%, 85.4%, and 81.3% in COSMOS.

The proportions of patients with an MDA response at week 24 were 23.6% in DISCOVER-1, 26.5% in DISCOVER-2, and 17.4% in COSMOS. At 52 weeks, the proportions of patients with MDA responses were 33.9% in DISCOVER-1 and 32.9% in DISCOVER-2. At 100 weeks, 44.6% of patients had an MDA response in DISCOVER-2. At 48 weeks, 32.5% of patients had an MDA response in COSMOS. Although the differences were not tested statistically, in all cases, the rate of response appeared to increase numerically with longer follow-up.

For the COSMOS study, the proportions of patients with MDA response were 32.5% in the guselkumab-every-8-weeks group and 29.8% in the placebo group at week 48. The proportions of patients who maintained MDA response at week 52 were 75.9% and 82.5% in DISCOVER-1 and DISCOVER-2, respectively. The proportions of patients who maintained MDA response at week 100 were 78.3% (based on patients achieving MDA response at week 24) and 85.3% (based on patients achieving MDA response at week 52) for DISCOVER-2.

Table 31

ACR and MDA Response for the Extension Period (FAS).

Function or disability

Table 32 describes the changes from baseline in HAQ-DI scores for the active treatment period. The mean changes in HAQ-DI at week 24 were –0.34 (SD = 0.57), –0.40 (SD = 0.54), and –0.29 (SD = 0.55) for DISCOVER–1, DISCOVER–2, and COSMOS, respectively. The mean changes from baseline at 52 weeks were –0.44 (SD = 0.56) and –0.48 (SD = 0.56) for DISCOVER–1 and DISCOVER–2, respectively. The mean changes from baseline at week 48 and week 100 were –0.47 (SD = 0.60) and –0.59 (SD = 0.58) for COSMOS and DISCOVER-2, respectively.

The proportions of patients who achieved a clinically meaningful HAQ-DI response (defined as a score greater than or equal to 0.35 improvement from baseline) were 53.2% at week 24 and 57.4% at week 52 for DISCOVER-1 and 52.5% at week 24, 60.9% at week 52, and 70.7% at week 100 for DISCOVER-2. For COSMOS, the proportions of patients achieving an improvement of greater than or equal to 0.35 from baseline were 47.4% at week 24 and 62.3% at week 48 for the guselkumab-every-8-weeks group, and 39.3% at week 24 and 48.7% at week 48 for the placebo crossover group.

The proportions of patients who maintained an improvement of greater than or equal to 0.35 from baseline in HAQ-DI score at week 52 were 84.9% and 92.0% in DISCOVER-1 and DISCOVER-2, respectively. The proportions of patients who maintained an improvement of greater than or equal to 0.35 from baseline in HAQ-DI score at week 100 were 88.7% (based on patients with HAQ-DI response at week 24) and 90.4% (based on patients with HAQ-DI response at week 52) in DISCOVER-2.

Table 32

Change From Baseline in HAQ-DI Scores for the Extension Period (FAS).

Health-Related Quality of Life

The mean changes from baseline in SF-36 PCS and MCS scores from week 24 through the long-term follow-up period are provided in Table 33. The mean changes from baseline in the SF-36 PCS for week 24 were 6.51 (SD = 7.71), 7.84 (SD = 8.03), and 5.83 (SD = 7.17) for DISCOVER-1, DISCOVER-2, and COSMOS, respectively. The mean changes from baseline at 52 weeks were 7.28 (SD = 8.07) and 9.51 (SD = 8.32) for DISCOVER-1 and DISCOVER-2, respectively. The mean changes from baseline at week 48 and week 100 were 8.44 (SD = 7.27) and 11.28 (SD = 9.28) for COSMOS and DISCOVER-2, respectively.

The proportions of patients who achieved a clinically meaningful SF-36 PCS score (≥ 5-point improvement from baseline) were 53.7% at week 24 and 53.5% at week 52 for DISCOVER-1; for DISCOVER-2, they were 63.4% at week 24, 67.1% at week 52, and 70.1% at week 100. For COSMOS, the proportions of patients achieving a greater than or equal to 5-point improvement from baseline in SF-36 PCS were 46.0% at week 24 and 55.6% at week 48 for the group receiving guselkumab every 8 weeks, and 22.9% and 26.0% at week 24 and week 48, respectively, for the placebo crossover group.

The proportions of patients who maintained a greater than or equal to 5-point improvement in SF-36 PCS scores at week 52 were 74.1% and 81.8% in DISCOVER-1 and DISCOVER-2, respectively. The proportions of patients who maintained a greater than or equal to 5-point improvement from baseline in SF-36 PCS scores at week 100 were 82.0% (based on SF-36 responders at week 24) and 85.4% (based on responders at week 52).

The mean changes from baseline in SF-36 MCS for week 24 were 3.03 (SD = 10.62), 4.45 (SD = 9.96), and 2.59 (SD = 10.18) for DISCOVER-1, DISCOVER-2, and COSMOS, respectively. The mean changes from baseline at 52 weeks were 5.14 (SD = 9.17) and 4.47 (SD = 9.78) for DISCOVER-1 and DISCOVER-2, respectively. The mean changes from baseline at week 48 and week 100 were 4.98 (SD = 9.94) and 3.64 (SD = 10.95) for COSMOS and DISCOVER-2, respectively. The change from baseline in the SF-36 MCS was either maintained or increased numerically (i.e., was not tested statistically) at the various time points evaluated in the 3 studies.

The proportions of patients who achieved a clinically meaningful (≥ 5-point improvement from baseline) SF-36 MCS score were 39.0% at week 24 and 46.5% at week 52 for DISCOVER-1, and 40.3% at week 24, 44.9% at week 52, and 46.4% at week 100 for DISCOVER-2. For COSMOS, the proportions of patients achieving a greater than or equal to 5-point improvement from baseline in SF-36 MCS were 36.6% at week 24 and 42.7% at week 48 for the group receiving guselkumab every 8 weeks, and 30.1% and 45.9%, respectively, at week 24 and week 48 for the placebo crossover group.

The proportions of patients who maintained a greater than or equal to 5-point improvement in SF-36 MCS scores at week 52 were 82.2% and 82.3% in DISCOVER-1 and DISCOVER-2, respectively. The proportions of patients who maintained a greater than or equal to 5-point improvement from baseline in SF-36 MCS scores at week 100 were 79.1% (based on SF-36 responders at week 24) and 77.2% (based on responders at week 52).

Table 33

Change From Baseline in SF-36 PCS and MCS for the Extension Period (FAS).

Psoriasis Skin Disease

Table 34 shows patients’ IGA and PASI responses for the active treatment period. In all trials, psoriasis skin disease outcome measures were analyzed in the subgroup of patients who had psoriasis affecting greater than or equal to 3% of their BSA and an IGA score of greater than or equal to 2 at baseline for the randomized phase. Although the difference was not tested statistically, the proportions of patients who achieved an IGA response at week 52 were numerically higher than at week 24 in the DISCOVER-1 and DISCOVER-2 studies (58.0% versus 69.3%, and 72.1% versus 77.1%, respectively). For the COSMOS study, the proportions of patients who achieved an IGA response at week 48 were 81.0% for the group receiving guselkumab every 8 weeks and 78.4% for the placebo crossover group.

The proportions of patients who maintained their IGA responses at week 52 were 81.8% and 85.4% in the DISCOVER-1 and DISCOVER-2 studies; at week 100 in the DISCOVER-2 study, these proportions were 88.2% and 89.1% at week 24 and week 52, respectively.

The proportions of patients who achieved PASI 100, PASI 90, and PASI 75 responses at week 24 were 25.9%, 50.6%, and 76.5% in DISCOVER-1; 46.5%, 70.3%, and 80.8% in DISCOVER-2; and 41.1%, 65.1%, and 76.0% for COSMOS, respectively. For DISCOVER-1 and DISCOVER-2, the proportions of patients who achieved PASI 100, PASI 90, and PASI 75 responses at week 52 were 48%, 66.7%, and 80.0%; and 54.7%, 77.1%, and 88.8%, respectively. For DISCOVER-2, the proportions at week 100 were 57.3%, 75.0%, and 87.8% for PASI 100, PASI 90, and PASI 75, respectively. For COSMOS, the proportions achieving PASI 100, PASI 90, and PASI 75 responses at week 48 were 66.1%, 84.3%, and 93.4%, respectively, for the group receiving guselkumab every 8 weeks, and 49.0%, 66.7%, and 88.2%, respectively, for the placebo crossover group.

The proportions of patients who maintained PASI 100, PASI 90, and PASI 75 responses at week 52 were 80.0%, 84.6%, and 86.2% in DISCOVER-1; and 75.9%, 88.3%, and 97.1% in DISCOVER-2, respectively. For DISCOVER-2, the proportions of patients who maintained PASI 100, PASI 90, and PASI 75 responses at week 100 were 76.3%, 86.2%, and 93.9% at week 24 and 86.7%, 88.2%, and 94.5% at week 52.

Table 34

IGA and PASI Response for the Extension Period (FAS).

Other Musculoskeletal Disease

Table 35 presents enthesitis and dactylitis outcomes for the extension period. The proportions of patients with resolution of enthesitis and dactylitis at week 52 were 56.3% and 79.5% for DISCOVER-1; these were 65.5% and 81.9% for DISCOVER-2, respectively. The proportions of patients with resolution of enthesitis and dactylitis at week 100 were 77.5% and 91.1% for DISCOVER-2, respectively. For COSMOS, the proportions of patients with resolution of enthesitis and dactylitis at week 48 were 67.5% and 80.0%, respectively.

The mean reductions from baseline in enthesitis scores (based on the LEI) and dactylitis scores (based on observed data) at week 52 were –1.8 (SD = 1.66) and –7.8 (SD = 10.55) for DISCOVER-1, and –1.9 (SD = 1.65) and –7.3 (SD = 9.74) for DISCOVER-2, respectively; at week 100, these scores were –2.1 (SD = 1.65) and –7.9 (SD = 10.12) for DISCOVER-2, respectively. For COSMOS, the mean reductions from baseline in enthesitis scores and dactylitis scores were –2.0 (SD = 1.92) and –6.1 (SD = 6.41) for the group receiving guselkumab every 8 weeks, and –1.8 (SD = 1.70) and –7.3 (SD = 7.04) for the placebo crossover group, respectively.

The proportions of patients who maintained resolution of enthesitis (based on the LEI) and dactylitis at week 52 were 74.1% and 96.4% in DISCOVER-1, and 88.4% and 92.2% in DISCOVER-2, respectively. For DISCOVER-2, the proportions of patients who maintained resolution of enthesitis (based on the LEI) and dactylitis at week 100 were 94.0% and 95.7%, respectively, at week 24, and 96.8% and 96.4%, respectively, at week 52.

Table 35

Enthesitis and Dactylitis Outcomes for the Extension Period (FAS).

Table 36 shows the improvement in BASDAI scores for the active treatment period. The changes in BASDAI scores were analyzed for the subgroup of patients who had spondylitis and peripheral joint PsA with BASDAI scores greater than 0 as their primary presentation at baseline. The proportions of patients with a greater than or equal to 50% improvement at week 24 were |||||| ||||| ||| ||||| in DISCOVER-1, DISCOVER-2, and COSMOS, respectively. The proportions at week 52 were ||||| ||| ||||| in DISCOVER-1 and DISCOVER-2; the proportions at week 100 were ||||| in DISCOVER-2; and at week 48, the proportions were ||||| ||| ||||| for the group receiving guselkumab every 8 weeks and the placebo crossover group in the COSMOS study, respectively.

Table 36

Improvement in BASDAI Scores for the Extension Period (FAS).

PsA Symptoms

Table 37 shows the changes in FACIT-Fatigue scores for the active treatment period. The mean improvements from baseline in FACIT-Fatigue scores at week 24 were 5.86 (SD = 10.39), 8.03 (SD = 9.89), and 5.9 (SD = 9.56) in DISCOVER-1, DISCOVER-2, and COSMOS, respectively. The mean improvements from baseline in FACIT-Fatigue scores at week 52 were 7.48 (SD = 9.63) and 8.93 (SD = 9.47) in DISCOVER-1 and DISCOVER-2; at week 100, these were ||||| ||| | |||||| in DISCOVER-2; and at week 48, these were ||| ||| | |||||| ||| ||| ||| | |||||| for the group receiving guselkumab every 8 weeks and the placebo crossover group in the COSMOS study, respectively.

Table 37

Change From Baseline in FACIT-Fatigue Score for the Extension Period (FAS).

Radiologic Outcomes

Table 38 shows the modified vdH-S scores for PsA for the DISCOVER-2 study at week 52 and week 100, based on the smallest detectable change.

The proportion of patients with no progression (defined as a change from baseline that was less than or equal to the smallest detectable change for the modified vdH-S total score) was ||||| from week 24 to week 52, and ||||| from week 52 to week 100 during the extension period. These results were consistent with the results for the group receiving guselkumab every 8 weeks ||||||| during the randomized phase.

Table 38

Modified vdH-S Scores of PsA (Based on the Smallest Detectable Change) for the Extension Period (FAS).

Harms

The summary of harms is presented in Table 39. More than half of patients had at least 1 adverse event through the end of the studies in DISCOVER-1 (69%) and DISCOVER-2 (63% after 52 weeks and 72% after 100 weeks). In the COSMOS study, the proportions of patients with at least 1 adverse event (cumulative) in the group receiving guselkumab every 8 weeks and the placebo crossover group were 50% and 46%, respectively. The most common events were nasopharyngitis, upper respiratory tract infection, bronchitis, enthesopathy, arthralgia, increase of alanine aminotransferase and aspartate aminotransferase, neutropenia, hypertension, and headache. No deaths were reported during the active treatment period. The rates of patients with at least 1 adverse or serious adverse event increased numerically through the end of study versus week 24. No deaths were reported through the end of the study. There were few serious adverse events in all trials through the end of the extension period.

The rate of infections was highest (43%) in the DISCOVER-1 study (week 52), following 38% and 29% in the DISCOVER-2 study (week 100 and week 52, respectively). In the COSMOS study, the rates of infection (cumulative) were 22% and 14% for the group receiving guselkumab every 8 weeks and the placebo crossover group, respectively. All other notable harms occurred only rarely during the long-term follow-up period. Some may have increased numerically, but the frequency was still quite low.

Table 39

Summary of Harms (Safety Set) for the Extension Period.

Critical Appraisal

ClinicalTrials.gov

Produced by the US National Library of Medicine. Targeted search used to capture registered clinical trials.

• [Search – Studies with results | tremfya OR guselkumab OR cnto-1959 OR cnto1959 | Psoriatic Arthritis]

WHO ICTRP

International Clinical Trials Registry Platform, produced by the WHO. Targeted search used to capture registered clinical trials.

• [Search terms – (tremfya OR guselkumab OR cnto-1959 OR cnto1959) AND psoriatic arthritis]

Health Canada’s Clinical Trials Database

Produced by Health Canada. Targeted search used to capture registered clinical trials.

• [Search terms – tremfya OR guselkumab]

EU Clinical Trials Register

European Union Clinical Trials Register, produced by the European Union. Targeted search used to capture registered clinical trials.

• [Search terms – guselkumab AND psoriatic arthritis]

Assessment of Validity, Reliability, and Responsiveness

The ACR tender and swollen joint counts have been deemed as appropriate for disease activity in patients with rheumatoid arthritis by the ACR expert committee and shown to be a valid outcome measure in RCTs in terms of content and construct validity, inter-rater reliability, and responiiveness.^43,44 The construct validity was also assessed using the known-group approach in a study of 164 patients with PsA and peripheral arthritis from 2 randomized placebo-controlled trials of TNF inhibitors (IMPACT trial, N = 104; etanercept trial, N = 60).⁴³ The number of patients who achieved ACR 20, 50, 70 in both active drug groups was significantly higher than the placebo groups.⁴³ The ACR criteria has demonstrated to have poor to substantial inter-observer reliability in PsA. Gladman and colleagues⁹⁰ conducted study of 10 patients with PsA who represented a broad range of joint inflammation, joint damage, and spinal involvement. The study was based on a Latin Square design. Each patient was examined by each of 10 rheumatologists who were members of the Spondyloarthritis Research Consortium of Canada. Generally, the study reported that the inter-observer reliability was poor to substantial (intra-class correlation coefficients [ICC] ranged from 0.10 to 0.80). However, right grip strength, number of damaged joints, left lateral flexion, and cervical flexion demonstrated almost perfect agreement (ICC ranged from 0.81 to 0.92).⁹⁰

The responsiveness of the ACR criteria was in the same study that assessed construct validity.⁴³ The study compared responsiveness and discriminative capacity of the Psoriatic Arthritis Response Criteria, ACR, European League Against Rheumatism (EULAR) response criteria, the DAS, and core-set measures. The ACR improvement criteria with 20%, 50% and 70% (ACR 20, 50, 70) cut-off points, EULAR response criteria, and the Psoriatic Arthritis Response Criteria were used as criteria for improvement. Standardized response mean (SRM) and effect size were calculated as statistics for responsiveness. In both active drug groups, the SRM ranged from –1.98 to –0.61 and effect size ranged from –2.84 to –0.57 while the SRM ranged from –0.40 to 0.02 and effect size ranged from –0.37 to 0.15 in the placebo groups.⁴³

Minimal Important Difference

The MID for ACR has not been estimated in patients with PsA, however, the ACR 20 is generally accepted as the MID indicating a response to treatment, while the ACR 50 and 70 more likely reflect truly important change for the long-term management of arthropathy.^43,44,91

Assessment of Validity, Reliability, and Responsiveness

No studies were found that validated BASDAI in PsA patients. In previous research, the BASDAI has been shown to have good validity, test-retest reliability, and responsiveness in patients with ankylosing spondylitis.⁹³ Haywood and colleagues conducted study of patients with a confirmed diagnosis of ankylosing spondylitis (N = 349) in England and Scotland assessing the validity, reliability, and responsiveness of the BASDAI.⁹³ Convergent validity of the BASDAI was assessed by testing a priori hypotheses about the strength of correlation with the Ankylosing Spondylitis Quality of Life Questionnaire (ASQoL). A high level of correlation (Pearson’s correlation coefficient r > 0.70) was hypothesized between the BASDAI and ASQoL. A strong correlation was observed between the BASDAI and ASQoL (r = 0.79).⁹³ The construct validity was assessed using the known-group approach. Patients who were able to work had significantly higher BASDAI scores compared with those unable to work due to ill health.⁹³ The BASDAI demonstrated strong internal consistency (Cronbach alpha = 0.87) and 2-week test-retest reliability (ICC 0.87, 95% 0.83 to 0.91).⁹³ The BASDAI were compared for responsiveness to change over the 6-month period by calculating the modified SRM were calculated for patients reporting an improvement or deterioration on health transition (general or ankylosing spondylitis-specific). The BASDAI produced high levels of responsiveness for groups of patients whose ankylosing spondylitis-specific (modified SRM ranged from –0.47 to 0.60) or general health (modified SRM ranged from –1.02 to 0.75) had improved or deteriorated according to transition question responses.⁹³

MID

The MID for the BASDAI has been determined based on data from 1 RCT (n = 205) as a change of −1.96 on the 10-point BASDAI scale in patients with ankylosing spondylitis.⁹⁴ The MID for BASDAI has not been estimated in patients with PsA.

Assessment of Validity, Reliability, and Responsiveness

No studies were found that validated DAS 28 in PsA patients. The validity and responsiveness of the DAS 28 was assessed in a study of 391 patients with active rheumatoid arthritis from 2 RCTs of abatacept (ATTAIN trial, N = 258; AIM trial, N = 133).⁴⁷ The k coefficients with quadratic weights were calculated as statistic for criterion validity. There was general agreement in classifying patients as none, moderate and good EULAR responders using DAS 28 CRP and ESR, with a k of 0.80 (95% CI 0.76 to 0.83) indicating good agreement. Healy and colleagues conducted a study of 28 patients with PsA underwent clinical assessment over a period of 6 months. Convergent validity was assessed using Spearman’s correlation was used to determine the relationship between the entheseal indices and disease activity as measured by the swollen and tender joint counts, patient and physician global VAS, CRP level, HAQ, and LEI. Strong to moderate correlations were observed between DAS 28 and the previously mentioned measures (r ranged from 0.344 to 0.651).⁹⁵

Responsiveness was assessed for the DAS 28 CRP using EULAR response criteria as anchors, the treatment difference was –18.83; the percentage improvement was –14.42%, the SRM was –0.31, and the relative efficiency was 1.93.⁴⁷

MID

No evidence of MID for DAS 28 was identified in patients with PsA. A study of 105 patients with early rheumatoid arthritis from a RCT assessed the thresholds for DAS 28 using an absolute level of disease activity, the study reported a value of ≤ 3.2 defined as the threshold for a low disease activity state and < 2.6 as the threshold for remission.⁸⁸

Assessment of Validity, Reliability, and Responsiveness

The FACIT-Fatigue demonstrated good content validity. Cella et al. (2005) completed a qualitative study with 12 patients; 2 (17%) had mild, 8 (67%) had moderate, and 2 (17%) had severe PsA disease activity; 7 (58%) attributed fatigue to PsA, and 7 (58%) rated fatigue as important or extremely important. Most patients considered the FACIT-Fatigue items relevant to their PsA experience and understood the item content and response options well.^8,96 A Toronto PsA cohort study of 135 patients with PsA assessed the validity and reliability of FACIT-Fatigue.⁹⁷ The study reported high levels of internal consistency (Cronbach alpha = 0.96) and 1-week test-retest reliability (ICC = 0.95).⁹⁷ A strong correlation was observed between the FACIT-Fatigue and modified Fatigue Severity (r = −0.79, 95% CI −0.85 to −0.72) indicating good convergent validity.⁹⁷ Construct validity was assessed using the known-group approach, FACIT-Fatigue scores were lower in patients with overwhelming fatigue and fibromyalgia than in those without (P < 0.001).⁹⁷ FACIT-Fatigue demonstrated good internal consistency (Cronbach coefficient alpha ≥ 0.90) and test-retest reliability (ICC = 0.95).⁹⁶ Apart from the Health Transition Item (which has a recall period of 1 year), correlations between FACIT-Fatigue and SF-36 domains generally exceeded 0.60 (all were > 0.50; P < 0.0001). There was a strong correlation with SF-36 Vitality (ρ > 0.80).⁹⁶

MID

A suggested MID for the FACIT-Fatigue in patients with rheumatoid arthritis is between 3 and 4 points using the Multidimensional Assessment of Fatigue and SF-36 Vitality as anchors.⁸ This MID was found in a sample of 271 patients (77% female, 81% White) with a median age of 56 years (range = 28 years to 84 years), a median tender joint count of 26 (range = 9 to 68), and a median swollen joint count of 15 (range = 2 to 43).⁸ Using disease activity measured by Patient’s Global Assessment of Psoriasis and Arthritis and SF-36 Vitality domain as anchors, a minimal clinically important difference for the FACIT-Fatigue total score was estimated as 3.1 points in patients with PsA.⁹⁶

Assessment of Validity, Reliability, and Responsiveness

The HAQ-DI, PCS score of the 12-Item Short Form Survey (PCS 12) and the Psoriatic Arthritis Impact of Disease instrument functional capacity score (PsAID-FC) were assessed for construct validity. The 3 patient-reported outcome measures correlated strongly with each other (ρ > 0.7); and moderately to strongly with patient global assessments for arthritis (ρ = 0.61 to 0.78), pain (ρ = 0.61 to 0.77); moderately with tender joint count (P = 0.39 to 0.51) and DAPSA (ρ = 0.55 to 0.72); weakly with swollen joint count (ρ = 0.19 to 0.32); and very weakly with patient global assessment for skin (ρ = 0.24 to 0.36).⁵⁰ Leung et al. (2021) analyzed data available from a longitudinal study in 14 countries of consecutive adults with definite PsA of at least 2 years in duration.⁵⁰ A total of 414 patients (52% male) were analyzed. They reported an internal consistency measured by Cronbach alpha of 0.92 for HAQ-DI. Ceiling effects were noted in a third of patients.⁹⁹ A study collected data from 2 studies conducted in UK and Singapore, respectively. A total of 69 patients with PsA (UK study, N = 31; Singapore study, N = 38) were assessed for test-retest reliability.¹⁰⁰ The study reported almost perfect test-retest reliability in the UK study (ICC 0.90, 95% CI 0.79 to 0.95) in a 1-week interval and Singapore study (ICC 0.94, 95% CI 0.89 to 0.97) in a 2-week interval.¹⁰⁰ All 3 patient-reported outcome measures for physical function were more sensitive for worsening than improvement. Moderate effect sizes (Cohen d) were seen in all 3 patient-reported outcome measures in measurement of worsening (HAQ-DI 0.46,95% CI 0.12 to 0.78; PCS12 − 0.57, 95% CI −0.92 to −0.24; PsAID-FC 0.51, 95% CI 0.16 to 0.87). The SRM for worsening for HAQ-DI, PCS12, and PsAID-FC were 0.37(95% CI 0.10 to 0.61), −0.45(95% CI 0. −0.74 to −0.19), and 0.38(95% CI 0.12 to 0.66) respectively.⁵⁰

MID

Mease et al. (2011)⁶ have estimated that the MID for the HAQ-DI in PsA patients using anchor-based methods is 0.35 (unlike 0.22 for rheumatoid arthritis), while the MID has been estimated to be 0.131 in PsA patients using an anchor-based approach (equal bidirectional magnitudes for improvement and worsening) by Kwok and Pope (2010).⁴⁹ Discrepancies in the MID estimates may partly be explained by differences in the HAQ-DI score of the patients studied at baseline.¹²³ In the study by Mease et al. (2011),⁶ patients had a mean HAQ-DI score at baseline of 1.16, corresponding to moderate functional impairment. In contrast, patients in the study by Kwok and Pope (2010) had less functional impairment at baseline, with a mean HAQ-DI score of 0.732.⁵⁰ Blackmore and colleagues (1995) have shown the HAQ-DI adequately captures clinically important changes in functional status and pain in patients with PsA.¹²⁴ However, the HAQ-DI focuses on physical disability and may not adequately capture disability in patients with predominantly skin disease.¹²⁴ Modified versions of the HAQ to include spinal domains or skin disease assessment have not proven to be significantly better in assessment of health status in PsA than the original HAQ-DI.^124,125 Leung et al. (2020), reported the MID for improvement and worsening for HAQ-DI were –0.16 (SD: 0.87) for improvement, and 0.30 (SD: 0.81) for worsening using patient-rated “improved” or “worse” disease status as anchors in patients with PsA.⁹⁹

Validity

The most recent study assessing the validity of the IGA evaluated data from 4 phase III clinical studies of tofacitinib in patients with psoriasis (N = 3,641).¹⁰¹ Construct validity was assessed using a known-group approach, measuring the relationship between IGA and PASI through a repeated measures model. Using the PASI score as a categorical anchor, the differences in the IGA scores between the ‘clear’ group (PASI score of 0) and the other groups (PASI score > 0) were statistically significant, and increased as psoriasis became more severe (i.e., with larger PASI scores), indicating that the IGA could discriminate between different degrees of disease severity.¹⁰¹

Convergent validity was assessed in the aforementioned study by comparing the IGA with 3 additional outcome measures: the PASI, patient global assessment, and Dermatology Life Quality Index (DLQI).¹⁰¹ Pearson correlation coefficients between IGA and the 3 scales ranged from 0.4 to 0.79 at primary assessment time points in the 4 clinical studies. Results showed a strong correlation between the IGA and the PASI (r ranged from 0.77 to 0.79) even though the IGA does not consider the amount of BSA affected by psoriasis. A moderately strong correlation was found between the IGA and DLQI (0.44 to 0.57). None of the correlations were large (> 0.8), indicating that the IGA considers some different information than the PASI and DLQI. Baseline correlations were smaller likely due to the limited range of responses from relatively homogeneous patients. An earlier study by Cappelleri et al. (2013)¹⁰² conducted a psychometric validation of the IGA using data from a phase II study of tofacitinib among 197 patients. Results found a relatively high correlation between the IGA and Patient Global Assessment as well as the IGA and PASI with correlation coefficient r > 0.5, except for at baseline. The study also examined correlations between the IGA and the Ocular Comfort Index (OCI) and the pain/discomfort assessment with all correlation coefficient r < 0.2, indicating relatively high divergent validity. These findings were consistent with several other studies examining the 6-point IGA.^109-111

Reliability

A recent systematic review noted that the 5-point IGA appears to be based on a formative model where a change in the severity of the disease does not necessarily mean a change in all 3 components of erythema, induration, and scaling; therefore, an assessment of internal consistency would not be applicable.¹²⁷ However, Callis Duffin et al. (2019)¹⁰¹ assessed the internal consistency reliability, demonstrating that the scoring items (erythema, induration, and scaling) were highly consistent with each other (Cronbach coefficient alpha ≥ 0.90) at the primary assessment points in all 4 trials. The internal consistency reliability was less convincing (Cronbach coefficient alpha 0.50 to 0.63) for the values observed at baseline, likely a result of the specific inclusion criteria of the trials.¹⁰¹ This study also evaluated the consistency of IGA measurements between screening and baseline visits, when no change in terms of disease severity was expected. The ICC for the pooled data was 0.70, suggesting an acceptable test-retest reliability over a stable period.

Cappellleri et al. (2013)¹⁰² also examined the test-retest reliability of IGA scores at baseline and week 2 among patients with little to no change in their PASI scores, resulting in a high ICC of 0.80. The study also examined internal consistency reliability of the IGA, with Cronbach coefficient alpha > 0.80 at all time points from week 2 onwards, indicating high correlations after baseline. Langley et al. (2013)¹¹¹ conducted an analysis of variance to examine the variability in the IGA and PASI. The intra-rater variation in the 6-point IGA was lower than the PASI with SDs of 0.2 and 2.5, respectively.¹¹¹ The systematic review by Puzenat et al. also reported low intra-observer variability but moderate inter-observer variability for the IGA.¹²⁸

Responsiveness

No evidence regarding the responsiveness of the IGA was identified from the literature at this time.

The validity, reliability, and responsiveness of the IGA instrument have not been evaluated in patients with PsA.

MID

The MID for IGA has not been estimated in patients with PsA. It is generally accepted that a clinically meaningful score in the IGA is a score of 0 (“clear”) or 1 (“almost clear” or “minimal”).⁵³ Furthermore, some trials of patients with moderate to severe psoriasis define efficacy as a 2-point reduction in the total IGA score.¹⁰³ Both Cappellleri et al. (2013)¹⁰² and Callis Duffin et al. (2019)¹⁰¹ assessed the clinically important difference for the IGA. Both studies used the patient global assessment score as a continuous anchor and determined a clinically important difference score of 0.52 (95% CI, 0.42 to 0.56) and 0.55 (95% CI, 0.546 to 0.563) in each study, respectively. It should be noted that the clinically important difference does not necessarily imply a minimum clinically important difference and although it is not possible to measure less than 1 category difference for an individual on the IGA scale, the clinically important difference was deemed appropriate to determine a group difference.¹⁰²

Assessment of Validity, Reliability, and Responsiveness

Healy and colleagues conducted a study of 28 patients with PsA underwent clinical assessment over a period of 6 months. Convergent validity was assessed using Spearman’s correlation was used to determine the relationship between the entheseal indices and disease activity as measured by the swollen and tender joint counts, patient and physician global VAS, CRP level, HAQ, and DAS 28. Moderate to weak correlations were observed between LEI and the previously mentioned measures (r ranged from –0.157 to 0.462).⁵⁵ The LEI index showed a large effect size of 0.82 at 6 months and significant response to change indicating adequate responsiveness.⁵⁵ No evidence was identified to support the reliability of the LEI for patients with PsA.

MID

The MID for LEI has not been estimated in patients with PsA.

Assessment of Validity, Reliability, and Responsiveness

Construct validity was assessed using the know-group approach in an observational PsA cohort study of 344 patients, it was found that patients who achieved sustained MDA (sustained MDA was defined as achieving MDA on consecutive visits for a minimum duration of 12 months) had a reduction in joint damage progression, where 69% of patients who achieved sustained MDA showed no progression of joint damage, compared with 51% in the control group, in addition the mean change in damaged joint counts was 0.931 in the sustained MDA group and 2.245 in the controls (P < 0.001).^104,105 In addition, Queiro et al. (2017) reported the relationship between MDA and presence of radiographic erosions in the hands and feet in a cross-sectional study. Patients in MDA were less likely to have evidence of hand erosions compared with those who were not (P < 0.05); however, there were no significant differences among patients when evaluating presence of erosions in the feet.^106,129 Coates et al.(2019) conducted a review of 20 studies evaluated the relationship between MDA and disease activity reported by the patient (measured as a patient-reported overall indicator of disease activity). The κ coefficient between MDA and patients’ rating of whether they were in a minimal disease state was 0.30 indicating weak convergent validity.¹⁰⁶ Coates and colleagues also reported moderate agreement (κ ranged from 0.73 to 0.75) with 3 alternative definitions of treatment responses: PASDAS, Composite Psoriatic Disease Activity Index-4 and Composite Psoriatic Disease Activity Index-3 to assess convergent validity in the same study assessed the construct validity of MDA.¹³⁰ Agreement was also strong for the MDA skin domain (MDA-joints) (κ = 0.86) indicating adequate internal consistency.¹⁰⁶

Threshold for MDA

The MDA-5, defined as meeting 5 of 7 of the cut-offs for the previously mentioned MDA domains, is generally accepted as the threshold for achieving MDA.^105,107

Validity

Simpson et al. (2015)¹⁰⁸ studied data from a phase III clinical trial (N = 445) in patients with plaque psoriasis to validate 3 systems of physician-scoring psoriasis severity, which included the PASI, static IGA, and Lattice System IGA measures. Construct validity of PASI was assessed by evaluating the correlation between the PASI score and the DLQI score, a skin-related HRQoL measure in grading psoriasis severity. The PASI correlated moderately with both the DLQI overall score as well as a single item of DLQI related to psoriasis symptoms (0.36 ≤ r ≤ 0.54), demonstrating that psoriasis severity is correlated with the DLQI score. The same study also investigated the content validity of the 3 measures by assessing the relative impact of the individual components of the measures on HRQoL using multiple linear regression analysis; BSA was most consistently associated with DLQI scores, followed by plaque induration and erythema. The scaling score was found to be minimally and inconsistently associated with DLQI scores which may be in part due to the static measurement of scaling which does not encompass the flaking of the skin over time which can be very distressing to patients.¹⁰⁸ The authors therefore concluded that weighing erythema, induration and scaling equally would not accurately capture the varying degrees to which these factors affect the patient’s rating of quality of life. Lastly, the construct and content validity of the PASI were found to be stronger during active treatment compared to pre-therapy.¹⁰⁸

Another study of 10 trained dermatologists evaluating 9 adult patients with plaque-type psoriasis assessed the correlations of PASI with other commonly used instruments in psoriasis, including the BSA and the IGA.¹⁰⁸ The authors reported a strong correlation with both measures (Pearson correlation coefficient > 0.78 and > 0.61, respectively).¹¹⁰ Similarly, Berth-Jones et al. (2016) (14 trained dermatologists, 16 patients with chronic plaque psoriasis) reported a strong correlation between PASI and the Lattice System IGA (Spearman’s rank correlation, r = 0.92), and a moderate correlation with the IGA (r = 0.73).¹⁰⁹ Berth-Jones et al. (2016) also found that the PASI and IGA were in good agreement for the clearance state (kappa = 0.64) but poor agreement for the severe state (kappa = 0.18).¹⁰⁹

Reliability

The reliability of the PASI measure has been assessed in several studies.^109-112 Bożek et al. (2017) reported the ICCs for all components of the PASI to be > 0.75, indicating very good intra-rater reliability, except for scaling (ICC = 0.72) in 9 patients with plaque-type psoriasis. The highest ICC was observed for the area score (ICC = 0.97). The coefficient of variation for the PASI was 36.9 overall, indicating moderate inter-rater reliability. The highest variability was observed for the head and neck (coefficients of variation = 117.8) and the lowest variability was for the area score (coefficients of variation = 26.8).^109-112 Langley et al. (2004) (17 physicians, 25 patients with psoriasis) reported similar results, with higher variability observed in the PASI scores derived by inexperienced physicians compared with experienced investigators (σ = 3.2 versus 1.2).¹¹¹ Berth-Jones et al. (2016) found excellent intra-rater and inter-rater reliability for the PASI score (ICCs > 0.81) in.¹⁰⁹ Fink et al. (2018) validated the methodology of ‘image-based’ versus commonly used ‘live’ PASI measurements in a pilot study, followed by validating in an observational cohort study.¹¹² They investigated the precision and reproducibility of automated, computer- guided PASI measurements in comparison with 3 trained physicians. PASI scores of 120 patients affected by plaque psoriasis of various severities were prospectively evaluated by 3 formally trained physicians by means of total body images. Each observer independently performed 2 rounds of image-based PASI calculations in all patients at 2 different time points.¹¹² Overall, 720 image-based PASI scores were calculated with a mean PASI of 8.8 (range 0.7 to 34.8). An inter-rater variability with an ICC of 0.895 and mean absolute difference of 3.3 PASI points were observed. Intra-rater variability showed a mean ICC of 0.877 and a mean absolute difference of 2.2 points.¹¹²

Responsiveness

The PASI score was found to have a moderate sensitivity to change in psoriasis patients (ICC ranged from 0.5 to 0.8).¹¹³ In a review by Spuls et al. (2010)¹³⁶ the authors commented on the responsiveness of PASI being weak when less than 10% of the BSA is affected given that the PASI score would be entirely dependent on the plaque severity scores, and therefore may underestimate the general degree of improvement.

Minimal Important Difference

The MID for PASI has not been estimated in patients with PsA. A systematic review by Mattei et al. (2013)¹¹⁴ including 13 RCTs evaluating biologics in psoriasis, reported that a ≥ 75% reduction in the PASI score translates to clinically-significant HRQoL improvement in patients assessed using the DLQI. This is based on the several studies that have demonstrated that a reduction in PASI scores can predict a reduction in DLQI scores, particularly when the patients were achieving a PASI 75 or higher (PASI 75 versus PASI 50 to 75 versus mean difference of 3.24).¹¹⁴

Assessment of Validity, Reliability, and Responsiveness

No evidence of validity, reliability, and responsiveness is identified for the dactylitis score in patients with PsA.

MID

The MID for the dactylitis score has not been estimated in patients with PsA.

Assessment of Validity, Reliability, and Responsiveness

Validity and reliability were assessed in a study of 168 patients with PsA.⁵² Construct validity was assessed using the known-group approach, patients with PsA were grouped under severe disease based on HAQ > 1.0, BASDAI > 50 and DAS 28 > 5.1. All 8 scales and summary scores in the severe groups were significantly worse than the less severe groups.⁵² Inter-scale coefficients (Cronbach alpha) were calculated to assess internal reliability, substantial to moderate levels of agreement were observed for most of scales of SF-36 (Cronbach alpha ranged from 0.33 to 0.70).⁵² Leung and colleagues (2021) analyzed data of 31 patients with PsA in a UK observational study. The study reported that SF-36 physical function domain demonstrated almost perfect test-retest reliability in patients with PsA (ICC 0.96, 95% CI 0.92 to 0.98).¹⁰⁰ The responsiveness has been assessed for the SF-36 in RCTs in patients with PsA. The median effect sizes were 0.77 (95% CI 0.60 to 0.93) and 0.23 (95% CI 0.09 to 0.36) for intervention and control groups, respectively.^139-141

Minimal Important Difference

The MID for either the PCS or MCS of the SF-36 is typically between 2.5 and 5 points.⁵² Leung and colleagues (2010) reported MIDs of 3.74 and 1.77 for the PCS and MCS subsections, respectively, in PsA patients treated with anti-TNF alpha drugs using an anchor-based approach.^7,52

Assessment of Validity, Reliability, and Responsiveness

In a study of 105 patients with PsA (mean age = 50.2, SD = 12.1), the vdH-S method showed strong convergent validity, when correlated with Simplified Psoriatic Arthritis Radiographic Score (SPARS) (r  = 0.926, P  < 0.0001).¹¹⁷

In 1 study, hand and feet radiographs from 50 patients with PsA were scored at 2 time points by 2 assessors for the modified vdH-S score. The radiographs of 10 patients were scored by both the readers using all 4 techniques in random order to assess inter-rater reliability and then scored 1 month later to estimate intra-rater reliability using ICCs. Sensitivity to change was estimated using a SRM and smallest detectable change. The inter- and intra-rater reliability were almost perfect agreement: ICC = 0.95 (95% CI 0.83 to 0.99) at baseline and 0.99 (95% CI 0.96 to 1.00) at follow-up (mean: 26 months, SD: 9.6 months) for inter-rater reliability; Rater 1 ICC = 0.97 (95% CI 0.90 to 0.99) and Rater 2 0.99 (95% CI 0.98 to 0.99) for intra-rater reliability. The modified vdH-S score has the ability to detect change at 1.2%. The sensitivity to change of the methods using the SRM demonstrated the modified vdH-S as having the greatest ability to detect change at a level of 0.79. The feasibility of the modified vdH-S method was estimated based on the mean time taken to score each film. The modified vdH-S method took 14.4 minutes to score.¹¹⁷

Minimal Important Difference

No reported MID was found for PsA patients.