Combining randomized trial data to estimate heterogeneous treatment effects

Authors  Hatfield, Laura; Kramer, Daniel;  Normand, Sharon-Lise
Year Published 2015
Link to white paper


Heart failure arises, progresses, and responds to therapy differently in different people. Yet clinical trials often lack power to estimate treatment effects for subgroups, or enforce eligibility criteria that exclude some patients entirely. Combining information across trials increases power for subgroup estimates and expands generalizibility. However, naively pooling patient-level data sacrifices the benefits of randomization, and pooling study-level estimates must consider trial heterogeneity.We develop and illustrate approaches for combining information across trials to estimate effects in men and women with heart failure who are treated with implantable cardioverter-defibrilliator (ICD) alone or in combination with cardiac resynchronization therapy (CRT-D). We consider individual- and trial-level factors that may confound or mediate subgroup treatment effects. For example, ischemic disease is more common in men; could this explain why women appear to benefit more from CRT-D than men?Our Bayesian models estimate sex-specific treatment effects across trials, accounting for uncertainty, confounding, and mediation. We find that with a very small number of heterogeneous studies, hierarchical modeling offers few benefits over conventional effect pooling,producing wider credible intervals but little shrinkage. We also find little evidence for residual confounding within subgroups, but some evidence of interactions between left bundle branch blockage and ischemic etiology in the sex-specific treatment effects, suggesting further study.


LAH and SLN are supported by contract DHHS/FDA-223201110172C and grant 1U01FD004493-01 from the Center for Devices and Radiological Health, US Food and Drug Administration.DBK is supported by a Paul B. Beeson Career Development Award (NIA K23AG045963).

A Population-Based Analysis of Robotic-Assisted Mitral Valve Repair

Journal The Annals of Thoracic Surgery
Authors Paul, Subroto; Isaacs, Abby; Jalbert, Jessica; Osakwe, Nonso; Salemi, Arash; Girardi, Leonard; Sedrakyan, Art
Year Published 2015
Link to publication


BACKGROUND: Robotic-assisted mitral valve repair is becoming more frequently performed in cardiac surgery. However, little is known about its utilization and safety at a national level.

METHODS: Patients undergoing mitral valve repair in the United States from 2008 to 2012 were identified in the National Inpatient Sample. Inhospital mortality, complications, length of stay, and cost for patients undergoing robotic-assisted mitral valve repair were compared with patients undergoing nonrobotic procedures.

RESULTS: We identified 50,408 isolated mitral valve repair surgeries, of which 3,145 were done with robotic assistance. In a propensity score matched analysis of 631 pairs of patients, we found no difference between patients undergoing robotic-assisted and nonrobotic-assisted mitral valve repair with respect to inhospital mortality, complications, or composite outcomes in unadjusted or multivariable analyses. Robotic-assisted mitral valve repair surgery was associated with a shorter median length of stay (4 versus 6 days, p < 0.001), and there was no difference in median total costs between the two procedures.

CONCLUSIONS: In our analysis of a large national database with its inherent limitations, robotic-assisted mitral valve repair was found to be safe, with an acceptable morbidity and mortality profile.

National trends in utilization and in-hospital outcomes of mechanical versus bioprosthetic aortic valve replacements

Journal The Journal of Thoracic and Cardiovascular Surgery
Authors Isaacs, Abby; Shuhaiber, Jeffrey; Salemi, Arash; Isom, O. Wayne; Sedrakyan, Art
Year Published 2015
Link to publication


OBJECTIVE: Substantial controversy surrounds the choice between a mechanical versus bioprosthetic prosthesis for aortic valve replacement (AVR), based on age. This study aims to investigate national trends and in-hospital outcomes of the 2 prosthesis choices.

METHODS: All patients aged >18 years in the National Inpatient Sample who received an AVR between 1998 and 2011 were considered. Valve-type use was examined by patient, procedural, and hospital characteristics, after which we matched patients based on their propensity score for receiving a bioprosthetic valve and compared their in-hospital outcomes.<

RESULTS: Bioprosthetic valves comprised 53.3% of 767,375 implanted valves, an increase in use from 37.7% in the period 1998 to 2001 to 63.6% in the period 2007 to 2011. The median age was 74 years for patients receiving bioprosthetic valves, and 67 years for those receiving mechanical valves. Use of bioprosthetic valves increased across all age groups, most markedly in patients age 55 to 64 years. Compared with patients receiving mechanical valves, these patients had a higher incidence of renal disease (8.0% vs 4.2%), coronary artery disease (58.5% vs 50.5%), concomitant coronary artery bypass grafting (46.7% vs 41.9%), and having surgery in a high-volume (>250 cases per year) center (31.3% vs 18.5%). Patients receiving bioprosthetic valves had a higher occurrence of in-hospital complications (55.9% vs 48.6%), but lower in-hospital mortality (4.4% vs 4.9%) than patients receiving mechanical valves. This difference was confirmed in propensity-matched analyses (complications: 52.7% vs 51.5%; mortality: 4.3% vs 5.2%).

CONCLUSIONS: Use of bioprosthetic valves in AVR increased dramatically from 1998 to 2011, particularly in patients age 55 to 64 years. Prosthesis selection varied significantly by facility, with low-volume facilities favoring mechanical valves. Aortic valve replacement with a bioprosthetic valve, compared with a mechanical valve, was associated with lower in-hospital mortality.

Population-Based Estimates of the Prevalence of Uterine Sarcoma Among Patients With Leiomyomata Undergoing Surgical Treatment

Journal JAMA Surgery
Authors Mao, Jialin ; Pfeifer,Samantha; Zheng, Xi Emily; Schlegel, Peter; Sedrakyan, Art
Year Published 2015
Link to publication


Uterine leiomyomata are one of the most common gynecologic problems among women in the United States, with an annual diagnosis range from 2.0 to 12.8 per 1000 reproductive-age women.1 Intervention is a standard management for symptomatic patients, and various procedures include open and laparoscopic hysterectomy, myolysis, uterine artery embolization, and magnetic resonance–guided focused ultrasonographic surgery.

The practice of electric morcellation has been used by gynecologic surgeons during laparoscopic and robotic-assisted hysterectomies and myomectomies as a less invasive alternative to open surgery. In April 2014, the US Food and Drug Administration (FDA) stated that they discouraged the use of this technique over concern that morcellation may spread unsuspected sarcoma tissue. Based on the literature, the FDA reported that 1 in 352 women have unsuspected uterine sarcoma while undergoing surgery for presumed benign leiomyoma. A recent study using an all-payer database found that 1 in 368 women who underwent morcellation had uterine cancer. However, the estimates in this study were limited by the selective participation of hospitals and by the lack of pathologic confirmation. The literature estimates used by the FDA are prone to referral and reporting bias. We sought to determine the population-based estimates of the prevalence of uterine sarcoma, as well as the risks of major complications following open surgery.

Unique device identifiers for coronary stent postmarket surveillance and research: a report from the Food and Drug Administration Medical Device Epidemiology Network Unique Device Identifier demonstration.

Journal The American Heart Journal
Authors Tcheng, James; Crowley, Jay; Tomes, Madris; Reed, Terrie; Dudas, Joseph; Thompson, Kweli; Garratt, Kirk; Drozda, Joseph Jr.;  on behalf of the MDEpiNet UDI Demonstration Expert Workgroup
Year Published 2014
Link to publication

BACKGROUND: Although electronic product identification in the consumer sector is ubiquitous, unique identification of medical devices is just being implemented in 2014. To evaluate unique device identifiers (UDIs) in health care, the US Food and Drug Administration (FDA) funded the Medical Device Epidemiology Network initiative, including a demonstration of the implementation of coronary stent UDI data in the information systems of a multihospital system (Mercy Health). This report describes the first phase of the demonstration.

METHODS: An expert panel of interventional cardiologists nominated by the American College of Cardiology and the Society for Cardiovascular Angiography and Interventions was convened with representatives of industry, health system members of the Healthcare Transformation Group, the American College of Cardiology National Cardiovascular Data Registry, and FDA to articulate concepts needed to best use UDI-associated data. The expert panel identified 3: (1) use cases for UDI-associated data (eg, research), (2) a supplemental data set of clinically relevant attributes (eg, stent dimensions), and (3) governance and administrative principles for the authoritative management of these data.

RESULTS: Eighteen use cases were identified, encompassing clinical care, supply chain management, consumer information, research, regulatory, and surveillance domains. In addition to the attributes of the FDA Global Unique Device Identification Database, 9 additional coronary stent-specific attributes were required to address use case requirements. Recommendations regarding governance were elucidated as foundational principles for UDI-associated data management.

CONCLUSIONS: This process for identifying requisite extensions to support the effective use of UDI-associated data should be generalizable. Implementation of a UDI system for medical devices must anticipate both global and device-specific information.