Introduction: The transplanted kidney has a certain lifetime and it depends on recipient and donor related, and many other clinical factors. Considering graft failure risk, patients are followed over years to monitor their kidney function during the post-transplantation period. This longitudinal observational period provides longitudinal and survival outcomes which are related and allow to model graft failure mechanism.
Method: The research questions that motivated this study arose from a clinical study that included approximately 80 longitudinal variables on 1002 patients who underwent kidney transplantation at Bursa Acıbadem Hospital (Turkey) between December 2011 and January 2021. In this context, the relationship between kidney survival and temporal changes in creatinine and hematocrit, which are kidney function biomarkers, was examined with multivariate joint models.
Results: In the multivariate joint model determined, the risk of kidney loss increased as the age at transplantation was high (hazard ratio 1.02) and the donor type was cadaver (hazard ratio 1.08), but only the age at transplantation was found to be statistically significant (p<0.001). No effect of gender and total number of missmatches was observed (p>0.05). When the effect of longitudinal variables on survival was examined, an increase in creatinine on the logarithmic scale increased the risk, while an increase in hematocrit reduced the risk (hazard ratios 4.52 and 0.84, respectively; p<0.001).
Conclusion: The relationship between longitudinal data of kidney function biomarkers and kidney survival has been demonstrated. It is beneficial to incorporate all available information in this type of multivariate model. It is expected that dynamic predictions will be integrated into current systems in the future, with these models utilizing all data from a patient's medical history. This makes multivariate joint modeling a valuable tool in the era of personalized medicine as it provides doctors with a deeper understanding of disease progression and the ability to select the most suitable treatment at a specific follow-up time.