On Monday, 13 November, Giresun University Department of Statistics will continue our seminars with the participation of Dr. Dinçer Göksülük. The seminar will be held online at 12:00 via Google meet.
You can access the current seminar programme of the semester from here and you can also follow the seminars on our department's YouTube channel.

Kind regards,
Giresun University
Department of Statistics
-------------------
Seminar Title 
Joint Modelling of Survival and Longitudinal Data in Peritoneal Dialysis Patients: Individualised Dynamic Risk Estimates

Seminar Summary 
Objective: Peritoneal dialysis (PD) treatment is frequently used in patients with late renal failure and the survival status of PD patients continues to be studied in the field of nephrology. In order to evaluate the course of the disease, biomarkers related to renal function (serum albumin, creatinine, etc.) and various biochemical parameters (peritonitis rate, gromelurel filtration rate, etc.) are measured repeatedly during the follow-up period. When the studies in the literature are examined, it is seen that the mean values of these repeated measurements taken at baseline and/or during the follow-up period are mostly taken into consideration in survival analyses.  Although this approach is not wrong, it ignores the change of repeated measurements over time and may lead to biased results. This study aims to reveal the relationship between various biomarkers and mortality in PD patients by taking into account the repeated measurement structure. In addition, it is aimed to obtain dynamic survival estimates at the patient level and to dynamically monitor the course of the disease during follow-up.

Method: In this study, data of 511 patients who started peritoneal dialysis (PD) treatment in Erciyes University Nephrology Department between 1995 and 2007 were retrospectively evaluated [1]. Patients were followed up from the beginning of PD until death, kidney transplantation, transition to haemodialysis, withdrawal from follow-up, or termination of PD treatment. The mortality response variable was death from all causes observed during PD. The effect of the change in biomarkers over time on survival was evaluated using a joint modelling approach [2, 3]. Cox regression method was used for the survival model and mixed-effects linear model was used for longitudinal data. Serum albumin values were considered as repeated measurements in the study.

Results: According to the results obtained, mean serum albumin values were not associated with mortality. However, serum albumin values measured repeatedly had a significant and inverse relationship with mortality (HR: 2.43, 95% CI: 1.48 - 4.16). Peritonitis rate, haemodialysis history and total number of comorbidities were positively correlated with mortality and mortality rates increased 1.74, 3.21 and 1.41 times, respectively. Survival status showed a significant variability among patients. When personalised dynamic risk estimates were considered, it was observed that the overall survival estimates obtained from the Cox regression model were not representative of all patients, and the individual survival estimates obtained from the mixed-effects model differed significantly from the overall estimates. 

Conclusion: When the relationship between serum albumin values and survival is evaluated with appropriate statistical modelling techniques over repeated measurements, more accurate and unbiased predictions are obtained. Dynamic risk estimates obtained on an individual basis play an important role in predicting the course of the disease and in the follow-up of patients. In addition, as new data are obtained during follow-up, risk estimates can be dynamically updated and personalised decisions can be made.

Time
Mon 13 Nov 2023 ⋅ 12pm - 1pm (Turkey Time)