Joint modeling of longitudinal and time to event data with application to tuberculosis research

Abstract

Due to tuberculosis (TB) being one of the top ten diseases in Africa with the highest mortality rate, a crucial objective is to find the appropriate medication to cure patients and prevent people from contracting the disease. Since this statistic is not improving sufficiently, it is evident that there is a need for new anti-TB drugs. One of the main challenges in developing new and effective drugs for the treatment of TB is to identify the combinations of effective drugs when subsequent testing of patients in pivotal clinical trials are performed. During the early weeks of the treatment of TB, trials of the early bactericidal activity assess the decline in colony-forming unit (CFU) count of Mycobacterium TB in the sputum of patients containing smear-microscopy-positive pulmonary TB. A previously published dataset containing CFU counts of treated patients over 56 days is used to perform joint modeling of the nonlinear data over time and the patients’ sputum culture conversion (i.e., the time-to-event outcome). It is clear from the results obtained that there is an association between the longitudinal and time-to-event outcomes.