SAT0433 TRIAL SIMULATION TO INFORM ENROLLMENT CRITERIA AND OUTCOME MEASURES FOR PRAGMATIC TRIALS IN PsA

Abstract

Background:Randomized controlled trials (RCTs) in psoriatic arthritis (PsA) have traditionally enrolled a homogenous subgroup of patients with more polyarticular disease, and the outcome measure used in PsA RCTs (ACR20) may not be ideal to measure differences between two active therapies nor capture change in patients with lower joint counts.Objectives:We conducted a simulation study to determine how changing the inclusion criteria and the primary outcome measure would impact the outcome of a future RCT.Methods:We used the Tight Control of PsA (TICOPA)1trial to inform simulation of two hypothetical head-to head trials comparing MTX to TNFi with 100 patients per arm. Within TICOPA, we identified MTX and TNFi new users; the visit at drug initiation became the hypothetical trial baseline visit, and the follow up visit was 12 weeks later. These data informed prediction models to simulate enrolled patients. We utilized propensity score-adjusted outcome models to account for potential confounding by indication. Trial 1, modeled after the SEAM-PsA trial,2used typical enrollment criteria (≥3 tender joint count (TJC) and ≥3 swollen joint count (SJC))2; Trial 2 required ≥1 TJC/SJC.1For each trial, five binary outcomes were simulated: ACR20, Disease Activity in PsA (DAPSA), clinical DAPSA (cDAPSA), Routine Assessment of Patient Index Data (RAPID3), and PsA Disease Activity Score (PASDAS), where low disease activity was the cutoff for continuous measures. Each hypothetical trial was simulated 1000 times, and the distribution of estimated effects was summarized using standard summary statistics and graphs.Results:Among 188 patients in TICOPA, 179 patients initiated MTX, and 43 patients initiated TNFi within the first 36 weeks. Among these, 107 MTX initiators and 15 TNFi initiators had ≥3 TJC and ≥3 SJC at drug initiation. Baseline characteristics of those in the “severe” (≥3 TJC and ≥3 SJC) and not severe (not meeting ≥3 TJC and ≥3 SJC) are shown in Table 1. Among “severe” patients, the mean probability of achieving ACR20 across simulations was approximately 0.27 in both arms and the observed relative risk (RR) TNFi vs MTX severe cohort across simulations was 1.0, IQR 0.84-1.17 (the RR in the SEAM trial at 24 wks was 1.20, 95%CI:1.05-1.35). In the “full cohort”, the median RR was 1.0, IQR 0.81-1.04. Trials using PASDAS, cDAPSA, and RAPID3 were more likely to differentiate between TNFi and MTX in the severe cohort (figure) but in the full cohort the results favored MTX.Table 1.Observed characteristics at drug initiationSevere (n=148)Not Severe (n=75)MTX (n=127)TNFi (n=21)SMDMTX (n=52)TNFi (n=23)SMDTICOPA Arm (no. (%))Standard Care57 (45%)4 (19%)0.5828 (54%)3 (13%)0.96Intensive Management70 (55%)17 (81%)24 (46%)20 (87%)Female (no. (%))65 (51%)11 (52%)0.0222 (42%)12 (52%)0.20TJC (mean (SD))17.8 (15.3)19.1 (17.3)0.083.5 (4.6)16.4 (19.1)0.93SJC (mean (SD))9.2 (7.4)10.2 (12.1)0.102.4 (3.4)2.2 (2.4)0.05Severe = ≥3 tender and ≥3 swollen jointsNot-severe = <3 tender or <3 swollen joints*Pseudo-baseline characteristics were at the time of drug initiation. In cases where the patient started a TNFi between visits, these were the values at the previous visit.Abbreviations: SMD = standardized mean difference, TJC=tender joint count, SJC=swollen joint countConclusion:Including patients with lower joint counts in an RCT reduced the ability to detect change with therapy. Additionally, among the outcome measures used to detect a difference between two active therapies, PASDAS, cDAPSA, and RAPID3 outperformed ACR20.