Divan A. Burger proposes a new bi-phasic non-linear regression model for characterizing the early bactericidal activity of tuberculosis treatments. The model analyzes colony forming unit count data from patient sputum during the first days or weeks of treatment. Statistical inference about treatment efficacy is based on a Bayesian non-linear mixed effects regression model fitted jointly to all trial patient data.
Use Cases
- Modeling CFU count decline over time based on the bi-phasic regression framework described
- Comparing treatment efficacy based on posterior predictive distributions of slope parameters
- Judging whether CFU decline is mono-linear or bi-linear using the Bayesian NLME model
Strengths
- The model incorporates linear and bi-linear regression as special cases, providing flexibility
- The Bayesian NLME approach allows for joint analysis of all patient data from a trial
Limitations
- Column-level documentation is absent; field semantics must be inferred after download
- Row count is unknown, which may limit suitability assessment
Provenance
- Source
- Divan A. Burger
- Collection Method
- Likely contains data from clinical trials assessing early bactericidal activity of TB drugs.