ENBIS-17 in Naples

9 – 14 September 2017; Naples (Italy) Abstract submission: 21 November 2016 – 10 May 2017

Predictive Risk Factors of Default from Tuberculosis Treatment Using Mixture Cure Model

12 September 2017, 10:30 – 10:50


Submitted by
Anicet Tchibozo
Anicet Tchibozo (KEYRUS BIOPHARMA)
Background: In 2015, 10.4 million people fell ill with tuberculosis (TB), causing 1.8 million deaths. We assessed the risk of default from TB treatment in Benin.

Methods: TB cohort was extracted from Benin TB registry. TB case and default were defined using WHO guidelines. Predictive risk factors were assessed with Cox regression, integrating cure models. SAS 9.2 was used for statistical analyses. Multiple imputation was used for missing values. FREQ procedure was used for tabulating default across predictive factors. LIFETST was used for default survival function, PHREG procedure for default hazard and NLMIXED procedure through the PSPMCM SAS® macro respectively.

Results: 5.22% (64/1226) of defaults emerged of which 3% occurred within the first two months with a Conditional Probability of Default (CPD) = 0.025 ± 0.004). When further Adjusting, HIV+ patients were roughly 4 times as higher at risk of defaulting compared to HIV- patients (hazard ratio [HR] = 3.82 [2.28; 6.41], P < 0.0001), patients with a past TB history where less at risk of defaulting compared to new TB patients (HR = 0.13 [0.03; 0.53], P = 0.0045) and senior TB patients where 3.5 times as higher at risk of defaulting compared to middle-age TB patients (HR = 3.49 [1.25; 9.74], P = 0.0170). This was confirmed by the log-logistic cure model, as HIV/AIDS and Age increased default probability whereas TB history reduced default.

Conclusion: this study provides evidences that HIV/AID, TB history and Age were the major predictive factors of defaulting from anti-TB treatment in Benin.
View paper

Return to programme