ILLUSTRATIVE EXAMPLE
Registry Study for a
Transplant Outcomes Question
Cumulative incidence of graft loss by donor type in a multicenter transplant registry, with death as a competing event, cause-specific and subdistribution models and registry data checks, reported to STROBE and RECORD.
This example does not describe a real client or project. Every engagement is confidential, during the project and after it ends.
RESEARCH QUESTION
What the Registry Study Set Out to Ask
This registry study example follows a transplant research group with access to a multicenter registry of kidney transplant recipients. The group asks whether recipients of living-donor kidneys have a different risk of graft loss over ten years than recipients of deceased-donor kidneys. Recipients can die with a functioning graft, which prevents graft loss from ever being observed, so death is a competing event and not simply a reason to stop following a patient.
Recipients are not assigned to a donor type, and donor type is tied to recipient age, health, waiting time and transplant era, so the question is framed as an association in registry data. The plan fixes the cohort, the event definitions and the models before any outcome is examined, and it asks two different questions: how many recipients lose a graft by a given time, and whether donor type changes the rate of graft loss.
STUDY DESIGN AND DATA
A Multicenter Registry with About 8,000 Recipients
A retrospective cohort drawn from a registry of about 8,000 adult kidney transplant recipients from about 25 centers, with up to ten years of follow-up. Each record holds donor type, recipient age, cause of kidney disease, diabetes, waiting time, transplant year, center and the dates of graft loss, death and last follow-up. A registry holds what centers report, so the plan first audits completeness by center and era, how graft loss is defined, and how missing donor characteristics arise, before any model is fitted.
- Design: retrospective registry cohort with time to graft loss and death as a competing event
- Outcome: graft loss, with death with a functioning graft as the competing event and last follow-up as censoring
- Software requested: SAS, with programs and logs
METHOD AND ASSUMPTIONS
Why Two Models, and What Must Hold
Treating death as censoring estimates graft survival in a world where recipients cannot die, and one minus the Kaplan-Meier estimate overstates the cumulative incidence of graft loss. The plan estimates the cumulative incidence of graft loss and of death by donor type, reads the two curves together, and fits two models, each tied to its question: a cause-specific Cox model for the rate of graft loss among recipients still at risk, and a Fine-Gray model for the effect of donor type on the cumulative incidence of graft loss.
The two models can point in different directions, because a donor type that lowers the rate of graft loss can also change who survives long enough to lose a graft, and the report explains both and does not choose the more favorable one. Proportional hazards are checked for the cause-specific model, and transplant era and center enter as strata or through clustered standard errors, depending on the number of events per center. The number of covariates is limited by the number of graft losses, covariates are chosen on clinical grounds and not from the outcome, and missing donor characteristics are handled by multiple imputation with the complete-case result alongside.
OUTPUT AND INTERPRETATION
How the Results Are Read
The output has a flow diagram from the registry to the analysis cohort, tables of data completeness by center and era, cumulative incidence curves for graft loss and death by donor type with 95% confidence intervals, and the cause-specific and subdistribution hazard ratios with their 95% confidence intervals. The cumulative incidence at five and ten years is given as an absolute risk for each donor type, so the size of the difference is readable beside the hazard ratios.
Sensitivity analyses cover the exclusion of the earliest era, a different covariate set, the complete-case analysis and a center-level frailty term. The interpretation is worded as an association between donor type and graft loss in this registry, adjusted for the recorded covariates, and it says that factors the registry does not record, such as recipient frailty, can still explain part of the difference. The findings apply to the registry's centers and eras.
REPORTING
What the Researcher Received
A written report structured to STROBE with its RECORD extension for studies using routinely collected health data, with statistical wording for the methods, matched to the output, for the team to check and incorporate. The research question, the clinical interpretation and the conclusions remain with the research team.