Accounting for Misclassification of Binary Outcomes in External Control Arm Studies for Unanchored Indirect...

Area of Science:

• Biostatistics
• Clinical Trial Design
• Health Economics

Background:

• Single-arm control trials are increasingly used for treatment evaluation but face methodological limitations.
• Regulatory agencies express concerns, yet these trials are sometimes required.
• Accurate indirect treatment comparisons are crucial, especially with external control arms from real-world data.

Purpose of the Study:

• To quantify bias from ignoring binary outcome misclassification in unanchored indirect comparisons.
• To propose a likelihood-based method, the outcome-corrected model, to address this bias.

Main Methods:

• Simulations were used to assess bias and coverage probabilities when misclassification is ignored.
• A novel outcome-corrected model was developed and evaluated.
• The methodology was applied to real-world hepatocellular carcinoma trial data.

Main Results:

• Ignoring outcome misclassification led to significant bias and poor coverage probabilities in simulations.
• The outcome-corrected model demonstrated reduced bias and improved confidence interval coverage.
• The model also showed improvements in root mean square error across various scenarios.

Conclusions:

• Addressing outcome misclassification is vital for reliable indirect treatment comparisons.
• The proposed outcome-corrected model enhances the accuracy and reliability of unanchored indirect comparisons.
• This method offers a practical solution for utilizing real-world data in treatment evaluations.