Page 44 - ISAKOS 2021 Newsletter Volume 1
P. 44

Interpreting Registry Data and Its Effect on a Surgeon’s Decision to Change Implants
Christopher Vertullo, MBBS, PhD, FRACS(Orth), FAOrthA
Knee Research Australia & Menzies Health Institute, Griffith University, AUSTRALIA
With >30 countries having national joint replacement registries, registry data are increasingly prevalent. However, interpretating the data related to prosthesis and technique choice can be problematic. Decades of TKR design innovation have resulted in a large number of variations in prostheses and surgical techniques. Recently, we defined surgeon preference with respect to those variations as “choosing a particular technique or implant to maximise surgeon-assessed attribute utility (i.e. value) within the parameters of accepted standards and regulation” while recognizing that the drivers for surgeons to change their preferences remain uncertain. In an era of patient-centric care and increasing arthroplasty costs, these drivers of preference deserved further examination.
What Constitutes Practice-Changing Evidence?
There is a lack of a definition for what constitutes surgeon- assessed attribute utility, and moreover, what constitutes universally accepted practice-changing evidence. The advantages of different clinical, observational, and experimental trial outcome data types are debatable, but none is without disadvantage and potential bias. Case series, particularly by designer-surgeons, suffer confounders such as performance bias and selection bias, whereas randomized controlled trials are typically underpowered to show small but clinically important differences and can lack generalizability. The validity of proxies for TKR survivorship such as radiostereometric analysis remains uncertain as they only examine prosthesis fixation stability; similarly, the clinical applicability of kinetic and kinematic analyses as proxies for patient function is also problematic.
Registry data are pragmatic and comprehensive; however, they are observational in nature, and, as a result, can also be affected by confounders. Nonetheless, such data can be used to perform unique studies that no other technique can accomplish. Registry confounders include patient-related factors, surgeon-related factors, surgical technique, and prosthesis design.
Interestingly, in the Australian Orthopaedic Association National Joint Replacement Registry (AOANJRR), prosthesis selection is the dominant factor when a surgeon’s revision risk is above the 99.7% confidence limits when examined in funnel plots compared with the mean.
Novel methods can be utilized to control for the confounders inherent in observational registry data. Matched-prosthesis analysis is a method to pragmatically compare two prostheses while controlling for variables such as fixation type, polyethylene type, and patellar resurfacing status. A recent matched-prosthesis analysis in which oxidized zirconium was compared with cobalt-chrome as a bearing surface demonstrated that the more expensive zirconium offered no advantage1. Another method, instrumental variable (IV) analysis, allows pseudo-randomization of registry data. Surgeon preference for a particular style or design of prosthesis is an accepted instrumental variable as it satisfies three key assumptions: (1) it is associated with the exposure under study, (2) it affects the outcome only through the exposure (exclusion restriction), and (3) it is independent of confounders. Recently, IV analysis was used to examine how the risk of revision is affected by surgeon preference for posterior-stabilized TKR, selective patellar resurfacing, and hybrid fixation2. One other advantage of using surgeon preference in an IV analysis, besides allowing for pseudo- randomization of observational data, is to potentially improve surgeons’ cognitive synthesis by directly linking their preferences to subsequent revision risk.
Other methods of controlling for confounders are also utilized, such as cohort stratification and risk adjustment, matched-patient cohorts, and propensity score matching. Recent registry studies using these methods have demonstrated a higher risk of infection in association with posterior-stabilized TKR.
Recognizing that not all confounders can be eliminated from purely observational data, some registries have commenced registry-nested randomized trials. The CRISTAL trial, a cluster randomized, crossover, non-inferiority trial of aspirin compared with low-molecular-weight heparin for prophylaxis against venous thromboembolism in patients undergoing hip or knee arthroplasty, is one such study that is currently underway.
Catalysts for Change
Surgeons can decide to change their prosthesis in response to a variety of catalysts. However, the tipping point for preference change by an individual surgeon is complex and multifaceted and is altered by factors such as surgeon barriers, institutional barriers, surgeon cognitive biases, heuristics, data source, and data ownership.

   42   43   44   45   46