Please activate JavaScript!
Please install Adobe Flash Player, click here for download

ISCB2014_abstract_book

ISCB 2014 Vienna, Austria • Abstracts - Poster Presentations 117Wednesday, 27th August 2014 • 11:00-11:30 Monday25thAugustTuesday26thAugustThursday28thAugustAuthorIndexPostersWednesday27thAugustSunday24thAugust P3.1.157 On the estimation of survival of HIV/AIDS patients on anti-retroviral therapy: an application to interval censored data PK Swain1 , G Grover1 1 University of Delhi, Delhi, India   The main objective of this paper is to estimate the survival of HIV/AIDS patients who are undergoing Antiretroviral Therapy treatment in an ART centre, Delhi, India. Non Parametric Maximum Likelihood Estimation NPMLE (E-M) for interval censoring and KM survival plot for left, right and mid-point imputation have been used to estimate the survival of these patients. It has been observed that that the mid-point imputed survival plot has a very similar and consistent pattern as obtained by NPMLE (E-M) method. Considering these mid-point imputed values as right censored data, Cox PH model and Accelerated Failure time Model (AFTM) have been applied to study the effects of prognostic factors like age, sex, mode of transmission, baseline CD4 cell count, hemoglobin, baseline weight and smoking habits on the survival of the patients. The Akaike Information Criterion (AIC) has been employed to compare the efficiency of the mod- els and Cox-Snell residual to test proportionality assumption.   P3.1.160 The relevance of joint modelling of longitudinal and competing risks data in the analysis of a peritoneal dialysis program L Teixeira1 , I Sousa2 , A Rodrigues3,4 , D Mendonça1,5 1 ICBAS-UP, Porto, Portugal, 2 Department of Mathematics and Applications, UMinho, Guimarães, Portugal, 3 CHP-HGSA, Porto, Portugal, 4 UMIB/ICBAS-UP, Porto, Portugal, 5 ISPUP, Porto, Portugal   In many clinical studies such as on peritoneal dialysis program, the pres- ence of a longitudinal outcome repeatedly registered along the follow-up time and the occurrence of a specific event is common. The many well- established models proposed to analyse longitudinal and time-to-event outcomes separately are not suitable to analyse data when the longitu- dinal and survival outcomes are associated. Then, a joint modelling ap- proach is required. In the last years, joint modelling of longitudinal and survival data has re- ceived much attention and an increase in the use in clinical studies was verified. Although, some joint models were adapted in order to allow for competing endpoints, this methodology has not been widely dissemi- nated. The present study has as main objectives to compare different joint mod- elling approaches of longitudinal and survival data in a competing risks setting and to illustrate their relevance in the analysis of a peritoneal di- alysis program. With these models it was possible to evaluate the associa- tion between a longitudinal clinical parameter (such as albumin) and the events of interest (death, transfer to haemodialysis and renal transplanta- tion), besides the identification of predictors of each of these outcomes. Results obtained with this methodology, which could not have been ob- tained with standard survival models, produced new information about peritoneal dialysis and contributed for a better knowledge and manage- ment of peritoneal dialysis program. P3.1.162 An illness-death model of chronic kidney disease progression A Thakkinstian1 , P Vejakama1 , A Ingsathit1 1 Faculty of Medicine Ramathibodi Hospital, Mahidol University, Bangkok, Thailand Objectives: To estimate probability of ESRD, death, and death after ESRD (ESRD-death) in CKD patients using illness-death model. Methods: Retrospective CKD cohort data were retrieved from one prov- ince (20 districts) in Thailand years 1997-2011. Illness-death models con- sisted of 3 transitions: death (transition 1), ESRD (transition 2), and ESRD- death (transition 3). A flexible-parametric survival with restricted-cubic spline was applied, and probability of each state was estimated. Results: Among 32106, 30634 CKDs I-IV were initial state, 2573/30634 (8.4%) subjects developed ESRD, but 55 subjects were lost since ESRD diagnosis, 6175/30634 (20.2%) subjects died without ESRD, and 1472/32106(4.6%) subjects initially enrolled with ESRD. Of 3990 ESRDs, 2457 (61.6%) subjects died. Probability of ESRD at 2-years, 5-years, and 10-years were 6.33 (95% CI: 6.13%, 6.55%), 5.30 (5.07%, 5.53%), and 2.47% (2.27%, 2.66%), respectively. These corresponding probabilities were re- spectively 4.98% (4.78%, 5.18%), 16.50 (16.12%, 16.89%), 36.37% (35.57%, 37.17%) for death; and 2.58% (2.43%, 2.73%), 7.27 (7.00, 7.54), and 13.04 (12.62%, 13.45%) for ESRD-death. Finally, probabilities of survival with ESRD-free were 86.03 (85.70%, 86.38%), 70.85% (70.37%, 71.33%), and 48.12% (47.31%, 48.93%). Risk of death was higher in diabetes than non-diabetes with hazard ratios of 1.21 (1.15, 1.28), 1.83 (1.15, 1.28), 1.65 (1.47, 1.86) for death, ESRD, and ESRD- death, respectively. In addition, ESRD-diabetes was about 1.37 (1.20, 1.56) times significantly higher risk of death than non-ESRD-diabetes. Conclusions: This study provided progression of CKD in Thai setting. Probabilities of ESRD, ESRD-death, and death with ESRD-free were esti- mated. Diabetes was higher risk for both ESRD and death than non-dia- betes.   P3.1.170 An illness-death model of HIV infection U Udomsubpayakul1 , P Subhaluksuksakorn1 , P Chancharas2 , B-O Thepthien3 , S Sriwanichakorn2 1 Ramathibodi Hospital, Mahidol University, Bangkok, Thailand, 2 Ministry of Public Health, Nonthaburi, Thailand, 3 ASEAN Institute for Health Development, Mahidol University, Nakhon Pathom, Thailand Objectives: To estimate probability of lost to follow-up (lossFU), death, and death after lossFU in HIV infected patients using illness-death model. Methods: Data were retrieved from ThaiHIV-registry of 21 provinces, the National Health Security Office (NHSO) years 2008 to 2012. Illness-death models were constructed with 3 transitions: death (transition 1), lossFU (transition 2), and lossFU-death (transition 3). State’s probability was es- timated using a restricted-cubic spline regression. Prognostic factors (i.e., sex, age, opportunistic infection(OI), anti-retroviral treatment (ART), health-coverage, and hospital-change) were then assessed. Results: Among 8692 HIV patients entered to the initial state, 2453 (28.2%) patients were lossFU, 6,239 adhered with clinics but 769 died. Of 2453, 783 patients died after lossFU. Probability of lossFU at 2-years and 5-years were 17.48% (95% CI: 16.72%, 18.23%) and 21.77 (20.64%, 22.91%), respectively. Corresponding probabilities were respectively 8.15% (7.58%, 8.72%) and 10.93 (10.17%, 11.70%) for death; 8.96% (8.36%, 9.56%) and 12.32 (11.25%, 13.02%) for lossFU-death. Age ≥ 55 years and males were 1.28 (95%CI: 1.124, 1.47) and 1.21 (1.14, 1.29) times higher risk than age <55 years and females. ART, social, and government health-coverage were respectively 85% (84%, 86%), 17%(8%, 25%) and 41% (28%, 50%) lower risk than non-ART and universal scheme. Conversely, OI and hospital- change were 55% (40%, 71%) and 26% (17%, 36%) higher risk than non-OI and non-hospital-change.

Pages Overview