Avoidable flaws in observational analyses: an application to statins and cancer

Area of Science:

• Health Informatics
• Epidemiology
• Biostatistics

Background:

• Real-world data (RWD) availability fuels debate on its role in medical treatment benefit-risk assessment.
• Observational studies often show discrepancies with randomized trials, e.g., statins and cancer risk.
• Flaws in observational studies may be mitigated by emulating target trials.

Purpose of the Study:

• To assess the role of emulating a target trial in reducing bias in real-world data analyses.
• To compare estimates from a target trial emulation with traditional analytic approaches for statins and cancer.

Main Methods:

• Emulation of a target trial using electronic health records from 733,804 UK adults over 10 years.
• Comparison of cancer risk estimates between target trial emulation and previously applied analytic methods.
• Analysis of intention-to-treat and per-protocol survival differences.

Main Results:

• Target trial emulation yielded cancer-free survival differences of -0.5% (ITT) and -0.3% (per-protocol).
• Previous analytic approaches produced estimates suggesting strong protective effects, differing from target trial emulation.
• Significant bias was observed in traditional analytic approaches compared to the target trial emulation.

Conclusions:

• Explicitly emulating a target trial is crucial for reducing bias in observational analyses of medical treatments.
• Target trial emulation offers a more reliable method for assessing treatment effects using real-world data.
• This methodology enhances the validity of benefit-risk assessments in healthcare.