Estimating Parameters in Models for Hepatitis C Virus Dynamics
Author Information
Author(s): Thiébaut Rodolphe, Guedj Jérémie, Jacqmin-Gadda Hélène, Chêne Geneviève, Trimoulet Pascale, Neau Didier, Commenges Daniel
Primary Institution: Bordeaux 2 University, Bordeaux, France
Hypothesis
Can a method be developed to accurately estimate parameters of non-linear mixed models while accounting for undetectable viral load values?
Conclusion
The proposed method provides unbiased estimates when the assumed distribution is correct and is easy to implement with standard software.
Supporting Evidence
- Taking into account left-censoring reduced the relative bias on fixed effects to less than 2%.
- Estimates of the death rate of infected cells differed significantly when left-censoring was considered.
- The method is easy to implement using standard software like NLMIXED in SAS.
- Confidence intervals were larger when accounting for left-censoring, indicating increased variability.
Takeaway
This study shows that using a better method to handle undetectable virus levels can help doctors get more accurate information about how well treatments are working.
Methodology
The study used a full likelihood approach to estimate parameters while distinguishing between observed and left-censored measurements.
Potential Biases
Potential bias exists if the assumed distribution of viral load is incorrect.
Limitations
The method relies on the assumption of a log-normal distribution of viral load measures.
Participant Demographics
Participants were 17 HIV co-infected patients undergoing treatment for HCV.
Statistical Information
P-Value
0.001
Confidence Interval
95% CI: 0.14; 0.26
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website