Safety and Efficacy of B-Cell Depletion with Rituximab for the Treatment of Systemic Sclerosis-associated Pulmonary Arterial Hypertension: A Multicenter, Double-Blind, Randomized, Placebo-controlled Trial

Abstract

Rationale: Systemic sclerosis (SSc)-pulmonary arterial hypertension (PAH) is one of the most prevalent and deadly forms of PAH. B cells may contribute to SSc pathogenesis. Objectives: We investigated the safety and efficacy of B-cell depletion for SSc-PAH. Methods: In an NIH-sponsored, multicenter, double-blinded, randomized, placebo-controlled, proof-of-concept trial, 57 patients with SSc-PAH on stable-dose standard medical therapy received two infusions of 1,000 mg rituximab or placebo administered 2 weeks apart. The primary outcome measure was the change in 6-minute-walk distance (6MWD) at 24 weeks. Secondary endpoints included safety and invasive hemodynamics. We applied a machine learning approach to predict drug responsiveness. Measurements and Main Results: We randomized 57 subjects from 2010 to 2018. In the primary analysis, using data through Week 24, the adjusted mean change in 6MWD at 24 weeks favored the treatment arm but did not reach statistical significance (23.6 ± 11.1 m vs. 0.5 ± 9.7 m; P = 0.12). Although a negative study, when data through Week 48 were also considered, the estimated change in 6MWD at Week 24 was 25.5 ± 8.8 m for rituximab and 0.4 ± 7.4 m for placebo (P = 0.03). Rituximab treatment appeared to be safe and well tolerated. Low levels of RF (rheumatoid factor), IL-12, and IL-17 were sensitive and specific as favorable predictors of a rituximab response as measured by an improved 6MWD (receiver operating characteristic area under the curve, 0.88-0.95). Conclusions: B-cell depletion therapy is a potentially effective and safe adjuvant treatment for SSc-PAH. Future studies in these patients can confirm whether the identified biomarkers predict rituximab responsiveness. Clinical trial registered with www.clinicaltrails.gov (NCT01086540).

Keywords: pulmonary hypertension; systemic sclerosis; treatment.

Figures

Figure 1.

CONSORT diagram. *Two subjects who withdrew early from the study also discontinued study treatment early. Three subjects (two rituximab and one placebo) were excluded from the mITT population as they were ineligible. The reasons included 1) abnormal neutrophil count <1.5 × 109/L at screening (1.3 and 1.0 × 109/L at screening and baseline, respectively), 2) not on stable dose of pulmonary arterial hypertension medication for 4 weeks before randomization (stopped bosentan 5 d before randomization), and 3) not on stable dose of prednisone for 4 weeks before randomization (prednisone 40 mg/d on Days −8 to −4 and tapered over the first 100 d). The number at the “Week 24 Visit” box indicates the number of participants in the randomized population who completed the Week 24 visit. mITT = modified intent-to-treat; PAH = pulmonary arterial hypertension.

Figure 2.

Model estimated mean (±SE) changes in 6-minute-walk distance (6MWD) from baseline are shown in the modified intent-to-treat (mITT) cohort. (A) The primary endpoint model, including 6MWD data up to Week 24, demonstrates a least-squares mean difference in the distances at Week 24 of 23.1 m (P = 0.12). (B) Secondary analysis of primary endpoint model, including all 6MWD data to Week 48, shows an estimated mean change walk distance from baseline to 24 weeks of 25.5 ± 8.79 m for rituximab and 0.4 ± 7.43 m for placebo (P = 0.03), a finding that diminished by Week 48. The numbers indicate the number of mITT participants with available 6MWD data at each time point and are shown at the bottom of the figure.

Figure 3.

Cumulative distribution plot of change in 6-minute-walk distance (6MWD) from baseline to Week 24. Although 48% of the rituximab-treated subjects demonstrated ⩾33 m in 6MWD, only 15% of placebo-treated subjects achieved this threshold. Similarly, a higher proportion of rituximab subjects (38% vs. 15%) achieved an improvement of ⩾50 m. In this figure, the short-dashed vertical gray line represents the ⩾33 m threshold, and the long-dashed vertical gray line represents the ⩾50 m threshold.

Figure 4.

Kaplan-Meier curves of (A) time to clinical worsening and (B) time to addition or modification of pulmonary arterial hypertension–specific therapies across treatment arms. Subjects were censored at their last study visit if an event did not occur or if they terminated early from the study. Although overall rates of deterioration did not differ substantially across rituximab and placebo arms, the majority of worsening in the placebo arm appears to be just before or after the 24 weeks’ time. Shaded areas represent confidence intervals. PAH = pulmonary arterial hypertension.

Figure 5.

Relationship between B-cell depletion and clinical response in the (A) rituximab and (B) placebo treatment arms. Change in 6MWD (solid line) and raw B-cell count (dashed line) from baseline to Week 48. 6MWD = 6-minute-walk distance.

Figure 6.

Machine learning discovery of baseline predictors of rituximab clinical response. (A) Receiver operating characteristic curves show performance of the extreme gradient boosted decision trees (XGBOOST) algorithm for predicting rituximab response (6-minute-walk distance [6MWD] improvement ⩾33 m). Cross-validated performance (across dataset subsampling iterations) is displayed for two XGBOOST models, including one model that included all measured input variables (red curve; area under the curve [AUC] = 0.84) and another model that was trained with recursive feature elimination and selected RF (rheumatoid factor), IL-12, and IL-17 as the most informative predictors (blue curve; AUC = 0.95). The XGBOOST algorithm outperformed three alternative machine learning algorithms (Table E1). These alternative algorithms selected a few additional candidate predictors of rituximab response (see Figure E4), although RF, IL-12, and IL-17 were consistently identified as the top biomarkers. Baseline plasma concentrations of (B) RF, (C) IL-12, and (D) IL-17 are compared between clinical responders (green) and nonresponders (black) stratified across treatment arms. Rituximab-treated patients with a 6MWD increase ⩾33 m at Study Week 24 had lower baseline concentrations of RF (P < 0.001), IL-12 (P = 0.002), and IL-17 (P = 0.009). Figures are box plots superimposed on shaded violin plots, which show the density of data points across the distribution. MFI = mean fluorescence intensity.