AUTHORS
Joyce Tiong, MD, Ang Li, MD, MS2, Dang Nhu Hai Pham, PhD3, Danielle Guffey4, Juan Ortiz5, Tianshi Wu, MD6, Zihan Yang, MS7, Barbara Lam Jr, MD8, Ming Lim, MBBCh, MS9, Shruti Chaturvedi Jr, MBBS, MS10, Allyson Pishko I, MD11, Omid Jafari, Senior Software Engineering and Programming Associate12, Senthil Sukumar, MD,MSc13
AFFILIATIONS
1 Baylor College of Medicine, Houston, Texas, United States
2 Baylor College of Medicine, Houston, Texas, United States
3 Baylor College of Medicine, Houston, Texas, United States
4 Baylor College of Medicine, Houston, Texas, United States
5 Baylor College of Medicine, Houston, Texas, United States
6 Baylor College of Medicine, Houston, Texas, United States
7 University of Texas Health Science Center at Houston, Houston, Texas, United States
8 University of Washington, Seattle, Washington, United States
9 University of Utah, Salt Lake City, Utah, United States
10 Johns Hopkins University, Baltimore, Maryland, United States
11 University of Pennsylvania, Philadelphia, Pennsylvania, United States
12 Baylor College of Medicine, Houston, Texas, United States
13 Baylor College of Medicine, Houston, Texas, United States
ABSTRACT
Introduction
Real-world comparative effectiveness studies in immune thrombotic thrombocytopenic purpura (iTTP) are constrained by disease rarity and misclassification when ICD-10 codes are used alone. We evaluated early caplacizumab (CAP) and rituximab (RTX), each with therapeutic plasma exchange (TPE), in a nationwide cohort of iTTP patients.
Methods
Study data came from Epic Cosmos, a collaborative dataset of Epic health systems including more than 300 million patients from all 50 states, D.C., and non-U.S. sites (cosmos.epic.com). We identified U.S. hospitalizations from 2019-2025 using a validated iTTP phenotype incorporating severe thrombocytopenia at presentation, moderate to high-risk PLASMIC score, ADMATS13 <10%, iTTP-directed therapy, and ICD-10 codes. The analytic cohort was restricted to patients who received TPE by hospital day 2 and survived through day 4.
We conducted two separate day 4 landmark comparative effectiveness analyses. The CAP analysis compared CAP receipt on or before day 4 with no CAP by day 4. The RTX analysis used the corresponding definition for RTX. The day 4 landmark fixed exposure classification and marked the start of mortality follow-up. Patients who initiated the corresponding drug after day 4 remained in the no-early-drug comparator because excluding them based on future treatment would introduce selection bias. Patients could contribute to both analyses, and concomitant use of the other study drug was allowed. Separate propensity-score (PS) models with overlap weighting balanced demographics, hospital characteristics, disease severity, laboratory values, comorbidities, corticosteroids, and concomitant early CAP or RTX. Outcomes after day 4 were all-cause mortality through last follow-up, achievement of platelet-count recovery (clinical response), time to platelet-count recovery, and in-hospital clinically relevant bleeding (CRB). PS-weighted Cox models estimated mortality hazard ratios (HRs); weighted outcome models estimated risk ratios or ratios of mean times (RRs). Sensitivity analyses excluded comparator patients who later initiated the corresponding drug.
Results
The analytic cohort included 1,596 patients from 192 hospitals. The median mortality follow-up was 837 days. Overall deaths by 30, 90, and 365 days were 44 (2.8%), 59 (3.7%), and 75 (4.7%), respectively; 106 (6.6%) died during all available follow-up.
In the CAP analysis, 327 patients received CAP by day 4 (median day 1 [IQR 0.5-2]) and 1,269 did not; among the latter, 191 initiated CAP after day 4 (median day 9 [IQR 6-12]). Early CAP was associated with more patients achieving clinical response (98.1% vs 93.0%; RR 1.06 [95% CI 1.03-1.08]) and a shorter time to response among responders (mean 3.6 vs 5.0 days; RR 0.72 [0.66-0.79]), but not significantly lower mortality (HR 0.74 [0.43-1.25]). CRB was more frequent with early CAP (9.0% vs 3.7%; RR 2.42 [1.50-3.92]). In the sensitivity analysis excluding patients who initiated CAP after day 4, the mortality estimate was similar (HR 0.68 [0.40-1.17]).
In the RTX analysis, 391 patients received RTX by day 4 (median day 2 [IQR 1-3]) and 1,205 did not; among the latter, 568 initiated RTX after day 4 (median day 7 [IQR 6-9]). Early RTX was associated with more patients achieving clinical response (96.8% vs 93.8%; RR 1.03 [1.01-1.06]) and lower mortality (HR 0.53 [0.29-0.97]). Time to response (mean 4.4 vs 4.6 days; RR 0.94 [0.85-1.05]) and CRB (5.4% vs 4.6%; RR 1.18 [0.71-1.97]) did not differ significantly. In the sensitivity analysis excluding patients who initiated RTX after day 4, the mortality association remained significant (HR 0.41 [0.21-0.79]).
Conclusion
Across 192 U.S. hospitals, early CAP and RTX were associated with distinct benefits: CAP with a shorter time to clinical response despite more bleeding, and RTX with lower long-term mortality. The day-4 landmark approach, inclusion of delayed initiators in the primary comparators, and concordant sensitivity analyses address key biases introduced by treatment timing in real-world data. These findings extend controlled-trial and international observational evidence to heterogeneous U.S. practice and support timely integration of modern TTP therapies. Residual confounding and inability to distinguish initial from recurrent TTP episodes limit causal interpretation.
DISCLOSURES
Joyce Tiong, MD: No relevant conflicts of interest to declare.