Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

Optimizing comparative effectiveness evidence from transfusion medicine trials: informing clinical practice and study designs

Portela, Gerard T. (2024) Optimizing comparative effectiveness evidence from transfusion medicine trials: informing clinical practice and study designs. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

[img] PDF
Restricted to University of Pittsburgh users only until 16 May 2026.

Download (2MB) | Request a Copy


Blood transfusions are frequently used interventions to prevent life-threatening anemia but can cause complications and cost U.S. hospitals almost $3 billion per year. There is a debate regarding optimal transfusion strategies for patients with acute myocardial infarction (MI) and for high-mortality risk patients with sickle cell disease (SCD). Applications of contemporary statistical methods may enhance recommendations for transfusion practices in these populations. The objectives of this dissertation were to produce evidence for transfusion recommendations to treat anemia in patients with acute MI and to determine a trial endpoint that attains sufficient statistical power for discriminating effectiveness of automated red cell exchange from standard of care in patients with SCD. In aim 1, we emulated a target trial of four transfusion strategies with hemoglobin thresholds between 7g/dL to 10g/dL in patients with acute MI and anemia using data from the Myocardial Ischemia and Transfusion (MINT) trial. The risk of 30-day death or MI increased progressively as hemoglobin transfusion thresholds decreased; a threshold of <9-10g/dL may be preferred for this population. In aim 2, we created individualized treatment rules (ITRs) to determine the optimal transfusion strategy for patients with acute MI and anemia. We did not identify treatment effect modifiers for 30-day death or MI, nor for 30-day death. An interpretable ITR for 30-day death, MI, revascularization, readmission, or heart failure led to reduced risk of this outcome compared to assigning all MINT participants a liberal or a restrictive strategy. In aim 3, we simulated a 150-patient randomized trial, Sickle Cell Disease and Cardiovascular Risk - Red Cell Exchange trial, to determine scenarios in which a count, time-to-event, or prioritized rank endpoint achieved sufficient statistical power to discriminate the effect of automated red cell exchange from standard of care. In most scenarios, a count endpoint analyzed with a negative binomial regression achieved higher power than the other endpoint analyses. Our results highlight the importance of informed endpoint selection in small treatment trials for populations with rare diseases. Together, our findings contribute valuable information for formulating transfusion practice guidelines and for designing studies to improve public health for clinical populations with anemia.


Social Networking:
Share |


Item Type: University of Pittsburgh ETD
Status: Unpublished
CreatorsEmailPitt UsernameORCID
Portela, Gerard T.gtp14@pitt.edugtp140000-0002-1871-6117
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairBrooks, Maria M.mbrooks@pitt.edumbrooks
Committee MemberBertolet, Marniemhb12@pitt.edumhb12
Committee MemberSwanson, Sonja A.sas766@pitt.edusas766
Committee MemberNovelli, Enrico M.noveex@upmc.eduemn3
Date: 16 May 2024
Date Type: Publication
Defense Date: 2 April 2024
Approval Date: 16 May 2024
Submission Date: 11 April 2024
Access Restriction: 2 year -- Restrict access to University of Pittsburgh for a period of 2 years.
Number of Pages: 187
Institution: University of Pittsburgh
Schools and Programs: School of Public Health > Epidemiology
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: clinical trials, transfusion medicine, causal inference, target trial emulation, intention-to-treat, machine learning
Date Deposited: 16 May 2024 17:35
Last Modified: 16 May 2024 17:35


Monthly Views for the past 3 years

Plum Analytics

Actions (login required)

View Item View Item