FormalPara Key Summary Points 

Why carry out this study?

While biologic therapies for psoriatic arthritis (PsA), guselkumab and ustekinumab, met their primary endpoints versus placebo in their respective trials, understanding the relative efficacy between these two treatments in the long-term is unknown.

Indirect treatment comparisons can help inform clinical decision-making when head-to-head evidence is lacking, as in the PsA disease space.

What was learned from the study?

This study allows for a comparison of guselkumab and ustekinumab in both biologic-naïve and biologic-experienced patients.

A multivariable logistic regression analysis was performed using individual patient-level data (IPD) to estimate adjusted comparisons of efficacy between guselkumab and ustekinumab while accounting for between-patient differences.

Leveraging IPD from phase 3 studies in the absence of head-to-head data demonstrated that both guselkumab doses were generally more effective versus ustekinumab for joint and skin outcomes up to 52 weeks, consistently in both biologic-naïve and biologic-experienced groups.

Introduction

Treatment for psoriatic arthritis (PsA), a chronic inflammatory condition, has evolved from traditional disease-modifying antirheumatic drugs (DMARDs) to targeted synthetic and biologic DMARDs [1, 2]. Specifically, biologic therapies that inhibit steps in the interleukin (IL) pathways, including IL-23 and IL-12/23 inhibitors, have emerged as effective PsA treatments for improving joint and skin outcomes [3]. Guselkumab is a fully human immunoglobulin G1 (IgG1) lambda monoclonal antibody that binds selectively to the p19 subunit of IL-23 [4], while ustekinumab, another fully human monoclonal antibody, binds to the p40 subunit shared by IL-12 and IL-23 [5]. Both treatments demonstrated superior efficacy in their respective trials [6,7,8,9]; however, head-to-head trials are lacking.

Indirect treatment comparison (ITC) methods assess comparative effectiveness and help inform treatment selection in the absence of head-to-head randomized controlled trials (RCTs) [10]. Network meta-analyses (NMA) compare multiple therapies directly and indirectly through a common anchor, such as placebo [11, 12]. A 2022 NMA of PsA trials demonstrated that guselkumab every 8 weeks (Q8W) and every 4 weeks (Q4W) had better skin efficacy (i.e., Psoriasis Area and Severity Index [PASI] 75, 90, and 100), than most other targeted therapies, including ustekinumab 45 mg [13]. Guselkumab Q4W was better than ustekinumab 90 mg [13]. For American College of Rheumatology (ACR) 20, 50, and 70, only guselkumab Q4W was better than ustekinumab 45 mg, while both guselkumab doses were comparable to most other agents, including ustekinumab 90 mg [13]. However, an NMA in separate biologic-naïve and biologic-experienced populations found that guselkumab Q8W was comparable to ustekinumab 45 mg and ustekinumab 90 mg for ACR 20 and PASI 75 at week 24 [14].

Notably, there are methodological challenges with conducting NMAs assessing long-term comparative efficacy. A key reason for the lack of long-term NMAs is that trial designs typically include a planned cross-over to active treatment after an initial placebo-controlled period [15]. This makes anchored analyses invalid and instead, methodologically appropriate analyses require a comparison of continuously treated active patients which is achieved by leveraging individual patient-level data (IPD) from a treatment’s RCT. Thus, the current objective was to use IPD to indirectly compare guselkumab and ustekinumab on joint and skin efficacy in biologic-naïve and biologic-experienced patients with PsA up to week 52.

Methods

Data Sources

Included studies were of similar trial design: DISCOVER-1, DISCOVER-2, PSUMMIT 1, and PSUMMIT 2. Study overviews and treatment doses are presented in the Supplementary Material.

Data Analysis

A flow diagram outlining the analytical procedures used is presented in Supplementary Figure S1. The primary comparative analyses between guselkumab 100 mg Q8W and 100 mg Q4W versus ustekinumab (45 and 90 mg) were performed for biologic-naïve (n = 1170) and biologic-experienced (n = 197) populations separately. Patient-level data for biologic-naïve patients were pooled for all active treatment arms from DISCOVER-1 and DISCOVER-2 and from PSUMMIT 1 and PSUMMIT 2, and IPD for biologic-experienced patients were from DISCOVER-1 and PSUMMIT-2. Pooled data included baseline characteristics and ACR 20 and PASI 90 outcomes. Notably, data for ustekinumab 45 mg and ustekinumab 90 mg from PSUMMIT 1 and PSUMMIT 2 were combined into ustekinumab 45/90 mg given that the weight-based dosing for ustekinumab is based on the specific pharmacokinetics of the molecule (e.g., drug absorption, distribution, metabolism). Two scenario analyses were also conducted to explore the separate ustekinumab doses at weeks 16, 24, and 52 versus both guselkumab doses. Scenario 1 examined only ustekinumab 45 mg as it is the recommended dose by the Food and Drug Administration (FDA) and the European Medicines Agency (EMA) for patients without concomitant moderate-to-severe psoriasis. Scenario 2 examined the recommended dose of ustekinumab 45 mg for those with a body weight ≤ 100 kg and ustekinumab 90 mg only for patients with a body weight > 100 kg, as per the FDA and EMA label. Guselkumab data for the Q8W and Q4W doses were not combined because the dose options are not related to pharmacokinetics; instead, the dose is prescribed dependent on whether or not a patient is at high risk for joint damage, according to clinical judgement.

The primary outcomes of interest, ACR 20 and PASI 90 response rates, were analyzed through 52 weeks of follow-up, using multivariable logistic regression adjusting for differences in patient characteristics between both guselkumab and ustekinumab treatment cohorts. The model was estimated with pre-specified covariates selected based on clinical input from a trained and practicing rheumatologist. Observed prognostic factors at baseline in the trial included: gender, age, body weight, PsA duration, number of swollen and tender joints, Health Assessment Questionnaire-Disability Index (HAQ-DI), C-reactive protein (CRP), dactylitis status, enthesitis status, PASI level, body surface area (BSA) affected by psoriasis, and methotrexate (MTX) use. The adjusted odds ratios (OR) for the baseline characteristics for biologic-naïve and biologic-experienced populations are presented in Supplementary Figures S2–S5.

Results are expressed as ORs with 95% confidence intervals (CI) and were considered statistically significant if the 95% CI for ORs excluded the null value of 1. Odds ratios resulting from the multivariable logistic regression model were translated into predicted response rates for ustekinumab, simulating the same patient population as the one enrolled in the guselkumab trials. Non-responder imputation was used for missing response data in the analyses. Analyses were performed using SAS software (version 9.4, SAS Institute Inc., Cary, NC, USA).

Sensitivity Analysis—Propensity Score-Based Modeling Approach

A propensity score-based modeling approach was used to conduct a sensitivity analysis for ACR 20 and PASI 90 responses at week 52; results are presented in Supplementary Figures S6 to S19. The propensity score is the conditional probability of being assigned to the guselkumab Q8W treatment arm given a set of covariates. Inverse probability weighting (IPW) aims to create a pseudo-population where the guselkumab Q8W arm and the ustekinumab arm being compared are homogeneous. The resulting weighted logistic regression model is where the single observations are up-weighted or down-weighted according to their propensity score. Analyses used different weighting schemes based on propensity scores, including the average treatment effect in the treated population (i.e., guselkumab Q8W) (ATT),Footnote 1 standardized average treatment effect (sATT), and IPW/average treatment effect (ATE).Footnote 2

Ethics Approval

The primary studies (i.e., DISCOVER-1, DISCOVER-2, PSUMMIT 1, and PSUMMIT 2) were conducted in accordance with the Declaration of Helsinki and the International Committee on Harmonization of Good Clinical Practice Guidelines. Protocols were reviewed and approved by each site’s governing institutional review board or ethics committee. All participants provided written informed consent.

Results

Patient Population

Tables 1 and 2 present the baseline characteristics for biologic-naïve and biologic-experienced patients on guselkumab Q8W and Q4W and ustekinumab 45/90 mg.

Table 1 Baseline characteristics for biologic-naïve patient population
Full size table
Table 2 Baseline characteristics for biologic-experienced patient population
Full size table

Significant differences across treatment groups (Table 1) were observed between both biologic-naïve treatment cohorts for number of swollen joints > 15 (19.4 vs. 29.1%), number of tender joints > 15 (51 vs. 66.9%), HAQ-DI score > 2 (4.5 vs. 13.6%), CRP ≥ 2 (24.1 vs. 93.8%), enthesitis status (60.9 vs. 70.3% without enthesitis), and BSA affected by psoriasis (19.6 vs. 25.5% with < 3% affected). Significant differences across treatment groups (Table 2) were observed between both biologic-experienced treatment cohorts for gender (38.1 vs. 58.5% male), number of swollen joints > 15 (13.2 vs. 47.5%), number of tender joints > 15 (47.4 vs. 83.1%), HAQ-DI > 2 (4.9 vs. 27.1%), CRP ≥ 2 (26.3 vs. 98.3%), enthesitis status (39.5 vs. 21.2% without enthesitis), and use of MTX at baseline (43.2 vs. 68.3% using MTX). Other characteristics examined were comparable. The imbalances between trials and prognostic values for all commonly available baseline characteristics are presented in Supplementary Table S1 and S2.

Primary Analysis

ACR 20 Response

Biologic-Naïve Patients

Figure 1 presents the observed and predicted ACR 20 responses based on the multivariable regression model from the pooled DISCOVER and PSUMMIT trials. The predicted ACR 20 response for ustekinumab relative to guselkumab Q8W accounts for differences between study populations and is generally aligned with the observed rates from the PSUMMIT trials. Both guselkumab doses showed significantly higher ACR 20 responses (Q8W = 54.9%; OR = 1.97; 95% CI 1.37, 2.84; Q4W = 56.5%; OR = 2.04; 95% CI 1.40, 2.96) versus ustekinumab (38.3%) from week 16 onwards. At week 52, both dose regimens of guselkumab had a significantly higher observed probability of achieving an ACR 20 response (Q8W = 70.7%; OR = 1.88; 95% CI 1.28, 2.76; Q4W = 71.7%; OR = 1.92; 95% CI 1.29, 2.86) versus ustekinumab (54.1%). The predicted probability for ustekinumab versus guselkumab Q8W at week 52 was 56.3% and the corresponding risk difference was 14.5% (95% CI 9.9, 19.1).

Fig. 1
figure 1

Observed ACR 20 response for guselkumab (DISCOVER) and adjusted ustekinumab (PSUMMIT) in bio-naïve patients. Adjusted estimated odds ratios using multivariate logistic regression model and risk differences for ACR 20 in bio-naïve patients up to week 52. The yellow line indicates the predicted ACR 20 response for ustekinumab relative to guselkumab Q8W. Sample sizes: GUS Q8: n = 335; GUS Q4; n = 336; UST; n = 499. ACR American College of Rheumatology, CI confidence interval, GUS guselkumab, Q4W every 4 weeks, Q8W every 8 weeks, UST ustekinumab

Full size image

Supplementary Figure S2 presents the results from the multivariable regression model for evaluating the prognostic strength of covariates used for ACR 20 at week 52. ACR 20 response rates were generally consistent across covariates. However, the following covariates were associated with a significantly lower likelihood of an ACR 20 response at week 52: females, individuals with a body weight > 90 kg, number of swollen joints between 10 and 15, and HAQ-DI scores > 2.

Biologic-Experienced Patients

Figure 2 presents the observed and predicted ACR 20 responses based on the multivariable regression model from DISCOVER-1 and PSUMMIT 2. Both doses of guselkumab showed significantly higher ACR 20 responses (Q8W = 46.3%; OR = 2.70; 95% CI 1.12, 6.53; Q4W = 63.2%; OR = 5.31; 95% CI 2.15, 13.10) versus ustekinumab (24.6%) at week 16 and from week 24 onwards (Q8W = 56.1%; OR = 2.57; 95% CI 1.11, 5.93; Q4W = 57.9%; OR = 2.63; 95% CI 1.12, 6.17) versus ustekinumab (35.6%). Only guselkumab Q4W had a significantly higher ACR 20 response (63.2%; OR = 3.39; 95% CI 1.41, 8.13) versus ustekinumab (33.9%) at week 20. At week 52, both dose regimens of guselkumab had a significantly higher observed probability of achieving an ACR 20 response (Q8W = 58.5%; OR = 2.72; 95% CI 1.17, 6.31; Q4W = 68.4%; OR = 4.77; 95% CI 1.95, 11.63) versus ustekinumab (35.6%). However, the CIs were generally large versus the estimated difference due to the smaller sample size of biologic-experienced patients. The predicted probability for ustekinumab versus guselkumab Q8W at week 52 was 34.2% and the corresponding risk difference was 24.3% (95% CI 10.4, 38.2).

Fig. 2
figure 2

Observed ACR 20 response for guselkumab (DISCOVER) and adjusted ustekinumab (PSUMMIT) in bio-experienced patients. Adjusted estimated odds ratios using multivariate logistic regression model and risk differences for ACR 20 in bio-experienced patients up to week 52. The yellow line indicates the predicted ACR 20 response for ustekinumab relative to guselkumab Q8W. Sample sizes: GUS Q8: n = 41; GUS Q4; n = 38; UST; n = 118. ACR American College of Rheumatology, CI confidence interval, GUS guselkumab, Q4W every 4 weeks, Q8W every 8 weeks, UST ustekinumab

Full size image

Supplementary Figure S3 presents the results from the multivariable regression model for evaluating the prognostic strength of covariates used for ACR 20 at week 52. ACR 20 response rates were generally consistent across covariates. However, a body weight > 90 kg was associated with a significantly lower likelihood of an ACR 20 response at week 52.

PASI 90 Response

Biologic-Naïve Patients

Figure 3 presents the observed and predicted PASI 90 responses based on the multivariable regression model from the pooled DISCOVER and PSUMMIT trials. In patients with BSA ≥ 3% at baseline, both doses of guselkumab showed significantly higher PASI 90 responses (Q8W = 51.5%; OR = 2.33; 95% CI 1.52, 3.56; Q4W = 54.8%; OR = 2.57; 95% CI 1.67, 3.97) versus ustekinumab (32.3%) from week 16 onwards. At week 52, both dose regimens of guselkumab had a significantly higher observed probability of achieving an ACR 20 response (Q8W = 70.5%; OR = 2.59; 95% CI 1.68, 3.99; Q4W = 75.9%; OR = 3.19; 95% CI 2.03, 5.00) versus ustekinumab (47.6%). The predicted probability for ustekinumab versus guselkumab Q8W at week 52 was 48.0% and the corresponding risk difference was 22.5% (95% CI 17.3, 27.7).

Fig. 3
figure 3

Observed PASI 90 response for guselkumab (DISCOVER) and adjusted ustekinumab (PSUMMIT) in bio-naïve patients. Adjusted estimated odds ratios using multivariate logistic regression model and risk differences for PASI 90 in bio-naïve patients up to week 52. The yellow line indicates the predicted PASI 90 response for ustekinumab relative to guselkumab Q8W. Sample sizes: GUS Q8: n = 264; GUS Q4; n = 270; UST; n = 372. CI confidence interval, GUS guselkumab, PASI Psoriasis Area and Severity Index, Q4W every 4 weeks, Q8W every 8 weeks, UST ustekinumab

Full size image

Supplementary Figure S4 presents the results from the multivariable regression model for evaluating the prognostic strength of covariates used for PASI 90 at week 52. PASI 90 response rates were consistent across all covariates.

Biologic-Experienced Patients

Figure 4 presents the observed and predicted PASI 90 responses based on the multivariable regression model from DISCOVER-1 and PSUMMIT 2. In patients with BSA ≥ 3% at baseline, both doses of guselkumab showed significantly higher PASI 90 responses (Q8W = 40.0%; OR = 6.31; 95% CI 1.95, 20.39; Q4W = 35.5%; OR = 5.16; 95% CI 1.60, 16.69) versus ustekinumab (12.9%) at week 16. Only guselkumab Q4W had significantly higher PASI 90 responses (54.8%; OR = 3.71; 95% CI 1.36, 10.16) versus ustekinumab (30.6%) at week 24. At week 52, both dose regimens of guselkumab had a significantly higher observed probability of achieving an ACR 20 response (Q8W = 50.0%; OR = 3.96; 95% CI 1.39,11.27; Q4W = 74.2%; OR = 13.10; 95% CI 4.18, 41.04) versus ustekinumab (25.9%). The predicted probability for ustekinumab versus guselkumab Q8W at week 52 was 20.2% and the corresponding risk difference was 29.8% (95% CI 12.7, 47.0).

Fig. 4
figure 4

Observed PASI 90 response for guselkumab (DISCOVER) and adjusted ustekinumab (PSUMMIT) in bio-experienced patients. Adjusted estimated odds ratios using multivariate logistic regression model and risk differences for PASI 90 in bio-experienced patients up to week 52. The yellow line indicates the predicted PASI 90 response for ustekinumab relative to guselkumab Q8W. Sample sizes: GUS Q8: n = 30; GUS Q4; n = 31; UST; n = 85. CI confidence interval, GUS guselkumab, PASI Psoriasis Area and Severity Index, Q4W every 4 weeks, Q8W every 8 weeks, UST ustekinumab

Full size image

Supplementary Figure S5 presents the results from the multivariable regression model for evaluating the prognostic strength of covariates used for PASI 90 at week 52. PASI 90 response rates were consistent across all covariates.

Sensitivity Analysis—Propensity Score-Based Modeling Approach

ACR 20 and PASI 90 Responses

To maximize comparability between guselkumab Q8W versus ustekinumab, IPW was used to balance baseline characteristics that were dissimilar between groups. Unlike regression modelling, IPW reduces the problem of adjusting for many covariates to one of weighting on the propensity score.

Propensity scores were derived with a logistic regression model using a binary treatment variable as the dependent variable and selected baseline covariates as explanatory variables. The estimated propensity scores were then used to derive weights for each patient using estimand-specific weighting formulas (ATE, ATT, and sATT).

Biologic-Naïve Patients

The estimated OR for baseline variables for the propensity score model from which weights were derived in the biologic-naïve population are presented in Supplementary Figure S6. Significant imbalances between treatment groups were observed for age ≥ 65 years, number of tender joints from 10 to 15, HAQ-DI score from 1–2 and > 2, CRP levels from 1 to < 2 and ≥ 2, and BSA with 10–< 20% affected by psoriasis.

Like the primary analysis, sensitivity analyses showed significantly higher ACR 20 for guselkumab Q8W versus ustekinumab. The results based on the propensity score model were consistent with the primary analysis at week 52 (OR = 1.88; 95% CI 1.28, 2.76). In each respective ATE, ATT, and sATT analysis, similar results were also shown in the multivariate weighted logistic regression model that adjusted for gender, age, weight, PsA duration, number of swollen joints, number of tender joints, HAQ-DI, CRP value at baseline, dactylitis status, enthesitis status, baseline PASI, BSA affected by psoriasis, and use of MTX at baseline (ATE: OR = 1.84; 95% CI 1.37, 2.46; ATT: OR = 2.20; 95% CI 1.60, 3.02; sATT: OR = 2.20; 95% CI = 1.64, 2.95) and the stepwise model that adjusted for a select combination of those covariates (ATE: OR = 1.64; 95% CI 1.23, 2.20; ATT: OR = 1.67; 95% CI 1.21, 2.31; sATT: OR = 1.67; 95% CI 1.24, 2.24). The estimated adjusted OR for baseline characteristics at week 52 for ACR 20 in the biologic-naive group using a weighted logistic regression model and ATE, ATT, and sATT weights are presented in Supplementary Figures S7 to S9.

Like the primary analysis, sensitivity analyses showed significantly higher PASI 90 for guselkumab Q8W versus ustekinumab. The results based on the propensity score model were consistent with the primary analysis at week 52 (OR = 2.59; 95% CI 1.68, 3.99). In the ATE analysis, similar results were also shown in the multivariate model that adjusted for gender, age, weight, PsA duration, number of swollen joints, number of tender joints, HAQ-DI, CRP value at baseline, dactylitis status, enthesitis status, baseline PASI, BSA affected by psoriasis, and use of MTX at baseline (ATE: OR = 2.01; 95% CI 1.49, 2.70; and the stepwise model that adjusted for a select combination of those covariates (ATE: OR = 1.88; 95% CI 1.41, 2.53). However, the multivariate model that adjusted for gender, age, weight, PsA duration, number of swollen joints, number of tender joints, HAQ-DI, CRP value at baseline, dactylitis status, enthesitis status, baseline PASI, BSA affected by psoriasis, and use of MTX at baseline only showed numerically higher PASI 90 responses for guselkumab Q8W in the ATT and sATT analyses (ATT: OR = 1.23; 95% CI 0.89, 1.69; sATT: OR = 1.23; 95% CI 0.92, 1.65) and numerically higher PASI 90 responses for guselkumab Q8W versus ustekinumab were also seen for the stepwise weighted model that adjusted for a select combination of those covariates (ATT: OR = 1.09; 95% CI 0.79, 1.51; sATT: OR = 1.09; 95% CI 0.81, 1.46). The estimated adjusted OR for baseline characteristics at week 52 for PASI 90 in the biologic-naive group using a weighted logistic regression model and ATE, ATT, and sATT weights are presented in Supplementary Figure S10 to S12.

Biologic-Experienced Patients

The estimated OR for baseline variables for the propensity score model from which weights were derived in the biologic-experienced population are presented in Supplementary Figure S13. Significant imbalances between treatment groups were for patients who were female, number of swollen joints > 15, and no use of MTX at baseline.

Like the primary analysis, sensitivity analyses showed significantly higher ACR 20 for guselkumab Q8W versus ustekinumab. The results based on the propensity score model were consistent with the primary analysis at week 52 (OR = 2.72; 95% CI 1.17, 6.31). In each respective ATE and sATT analysis, similar results were also shown in the multivariate weighted logistic regression model that adjusted for gender, weight, number of swollen joints, HAQ-DI, dactylitis status, enthesitis status, and use of MTX at baseline (ATE: OR = 5.17; 95% CI 2.35, 11.38; sATT: OR = 2.13; 95% CI 1.04, 4.40). However, once the multivariate weighted logistic regression model adjusted for gender, weight, number of swollen joints, HAQ-DI, dactylitis status, enthesitis status, and use of MTX at baseline in the ATT analysis, the results were only numerically higher for guselkumab Q8W (OR = 2.13; 95% CI 0.88, 5.16). The estimated OR for baseline characteristics at week 52 for ACR 20 in the biologic-experienced group using a weighted logistic regression model and ATE, ATT, and sATT weights are presented in Supplementary Figures S14 to S16.

Like the primary analysis, sensitivity analyses showed significantly higher PASI 90 for guselkumab Q8W versus ustekinumab. The results based on the propensity score model were consistent compared to the primary analysis at week 52 (OR = 3.96; 95% CI 1.39,11.27). In each respective ATE, ATT, and sATT analysis, similar results were also shown in the multivariate model that adjusted for gender, weight, number of swollen joints, HAQ-DI, dactylitis status, enthesitis status, and use of MTX at baseline (ATE: OR = 7.50; 95% CI 3.45, 16.32; ATT: OR = 5.98; 95% CI 2.12, 16.85; sATT: OR = 5.98; 95% CI 2.73, 13.12). The estimated OR for baseline characteristics at week 52 for PASI 90 in the biologic-experienced group using a weighted logistic regression model and ATE, ATT, and sATT weights are presented in Supplementary Figures S17 to S19.

Scenario Analyses

ACR 20 and PASI 90

Two different scenario analyses were explored. Scenario 1 examined only ustekinumab 45 mg as it is the recommended dose by the FDA and the EMA for patients with PsA without concomitant moderate-to-severe psoriasis. Scenario 2 examined the recommended dose of ustekinumab 45 mg for those with a body weight ≤ 100 kg and ustekinumab 90 mg only for patients with a body weight > 100 kg, as per the FDA and EMA label.

Biologic-Naïve Patients

Both scenario analyses demonstrated results consistent with the primary analysis (Supplementary Table S3). Compared to the 45 mg dose of ustekinumab (Scenario 1) and to ustekinumab 45 mg (body weight ≤ 100 kg) or ustekinumab 90 mg (body weight > 100 kg) (Scenario 2), both doses of guselkumab showed significantly higher ACR 20 and PASI 90 responses at weeks 16, 24, and 52.

Biologic-Experienced Patients

Both scenario analyses demonstrated results consistent with the primary analysis (Supplementary Table S4). Compared to the 45 mg dose of ustekinumab (Scenario 1) and to ustekinumab 45 mg (body weight ≤ 100 kg) or ustekinumab 90 mg (body weight > 100 kg) (Scenario 2), both doses of guselkumab showed significantly higher ACR 20 responses at weeks 16 and 52. For PASI 90, both doses of guselkumab showed significantly higher PASI 90 responses at weeks 16, 24, and 52, except for guselkumab Q8W in Scenario 1 at week 24.

Discussion

The lack of direct comparisons and treatment variety complicates PsA therapy decisions for clinicians and payers [10, 16, 17]. Using ITC methods can help assess comparative effectiveness between treatments to aid clinical decision-making [10]. Indirect comparisons using IPD from the DISCOVER and PSUMMIT trials demonstrate guselkumab’s and ustekinumab’s joint and skin efficacy in biologic-naïve and biologic-experienced populations [6,7,8,9]. From week 16 through 52, both guselkumab doses had significantly higher ACR 20 and PASI 90 versus ustekinumab, except for ACR 20 for guselkumab Q8W versus ustekinumab at week 20 (OR = 1.74, 95% CI 0.75, 4.06) and PASI 90 at week 24 (OR = 2.26, 95% CI 0.83, 6.18) in the biologic-experienced population.

Similarly, a 2022 NMA by Mease et al. revealed superior efficacy for guselkumab Q4W over ustekinumab 45 mg for ACR 20 and PASI 90, while both guselkumab doses were comparable to ustekinumab 90 mg for ACR 20; only guselkumab Q4W showed superior efficacy for PASI 90 [13]. Notably, the NMA was conducted in a mixed biologic-naïve and biologic-experienced population during the induction period, included COSMOS guselkumab data, and implemented a multinomial model for ACR and PASI data [13]. However, there were some differences between the current analysis and an NMA comparing PASI 90 responses in a biologic-naïve and biologic-experienced population [14], which may primarily be the result of unstable/uninformative estimates provided by the NMA. As anchored analyses, NMAs preserve randomization and allow unbiased estimates of relative treatment effects if patient populations do not differ regarding effect modifying variables [12, 18]. While NMAs require fewer assumptions than unanchored comparisons, this comes at the cost of increased variance; the consequence is that treatment effects are unstable in terms of direction and magnitude, a situation illustrated by large CIs. In disease areas where important confounders are either known or accounted for via inclusion/exclusion criteria, an IPD-adjusted analysis of the active arms as conducted in the present study may be more reliable.

Additional support for the efficacy of guselkumab relative to ustekinumab is provided by a frequentist NMA by Lu et al. during the induction period. Hierarchical clustering identified distinct treatment clusters for ACR 20 and PASI 75. Guselkumab was in a high-ranking cluster for both efficacy parameters [19]. Conversely, ustekinumab 90 mg ranked high for skin efficacy, but moderate for joint efficacy; ustekinumab 45 mg ranked moderately for both [19]. In contrast to the current results, guselkumab was comparable to both ustekinumab doses for ACR 20 in the biologic-naïve patient population [19]. However, the Lu et al. NMA used only phase 2 guselkumab data, and did not consider variation in placebo response across trials, introducing heterogeneity [20]. Lastly, the current findings are aligned with a previous study comparing guselkumab and ustekinumab for maintenance therapy in psoriasis [21]. IPD-based comparisons revealed significantly higher PASI 90 for guselkumab versus ustekinumab at week 16 and week 40.

Differential joint and skin efficacy between guselkumab and ustekinumab may stem from their distinct sites of action. While both target the IL-23/T-helper cell 17 (Th17) pathway, guselkumab selectively binds to IL-23’s p19 subunit [4] whereas ustekinumab binds to the shared p40 subunit of IL-12 and IL-23 [5]. Interleukin-23 drives PsA disease activity and increases production of downstream pro-inflammatory cytokines like IL-17 [22]. A 2021 study by Sweet et al. examined the protein expression of potential inflammatory biomarkers in patients with PsA treated with guselkumab or ustekinumab [23]. At week 4, guselkumab and ustekinumab induce a significant reduction in IL-17A and IL-17F concentrations; however, only guselkumab induced a further reduction in these biomarkers at week 24 [23]. Although the mechanism for the sustained reduction in IL-17s is unclear, the stronger suppression with guselkumab versus ustekinumab aligns with the enhanced joint and skin efficacy observed in the current study.

This analysis has several strengths. First, it employs multivariable regression analysis using IPD, allowing for adjustment of cross-trial differences between DISCOVER and PSUMMIT trial patients. The model was estimated with pre-specified covariates, minimizing bias with input from a clinical expert. Additionally, the credibility of the methodology is reinforced by consistent results between the propensity-score based sensitivity analysis and covariate approach. Second, compared to a naïve ITC, a methodologically rigorous form of ITC was used by leveraging IPD from pivotal guselkumab and ustekinumab trials [24]. Lastly, guselkumab’s efficacy remained consistent and robust across primary, scenario, and sensitivity analyses.

Limitations should also be acknowledged. The validity of relative treatment effects from unanchored ITCs are dependent on the difficult to reach assumption that all observable and unobservable clinically relevant prognostic factors are balanced [25]. Multivariable regression models in the primary analysis further rely on the assumption that covariate effects are the same for both treatments (no effect modifiers), and all analyses rely on no important unmeasured effect modifiers if results are intended to be transported to new populations. It is possible that covariates not considered in the analysis could have impacted the outcomes of interest and therefore residual bias may accompany the estimated relative treatment effects. Relatedly, the RCTs included in the IPD analyses were conducted in different settings and countries, and during different time periods, potentially introducing confounding factors and biases. However, to minimize bias, several clinically relevant patient characteristics were considered in the analysis. Lastly, omission of COSMOS trial data was due to distinct definitions of biologic-experienced patients versus DISCOVER-1. In COSMOS, patients had an inadequate response to tumor necrosis factor (TNF) inhibitors [26]; in DISCOVER-1, patients must have had previously received TNF inhibitors to be considered biologic-experienced [6].

Conclusions

In conclusion, leveraging IPD from phase 3 studies demonstrated that both guselkumab doses were generally more effective versus ustekinumab for joint and skin outcomes up to 52 weeks, consistently in both biologic-naïve and biologic-experienced groups. Future studies are encouraged to assess long-term comparative efficacy beyond 52 weeks.