Real-World Pitfalls of Analyzing Real-World Data: A Cautionary Note and Path Forward

Area of Science:

• Oncology Research
• Biostatistics
• Real-World Data Analysis

Background:

• Real-world data (RWD) are increasingly utilized in oncology research for clinical trend and patient outcome insights.
• However, RWD present unique challenges and potential pitfalls in survival analyses.
• Common issues include handling missing data, classification errors, and time-related biases.

Purpose of the Study:

• To demonstrate common pitfalls in survival analyses using real-world data (RWD).
• To highlight issues such as incorrect surrogate markers for missing data, ignoring early deaths, and guarantee-time bias.
• To emphasize the need for appropriate analytical methods when using RWD in oncology.

Main Methods:

• Utilized the American Society of Clinical Oncology (ASCO) CancerLinQ Discovery (CLQD) multiple myeloma (MM) data set.
• Compared overall survival (OS) by analyzing scenarios with known vs. presumed diagnosis dates, inclusion vs. exclusion of early deaths (0 months), and matched vs. unmatched cohorts for guarantee-time bias.
• Analyses were performed using STATA Version 17.0.

Main Results:

• Using presumed diagnosis dates as surrogates for missing data led to significantly different median OS (107 vs. 40 months).
• Excluding early deaths (within 1 month) in secondary AML (sAML) exaggerated median OS (46 vs. 39 months).
• Accounting for guarantee-time bias in MM patients with second primary malignancies (SPMs) substantially altered OS results (HR changed from 0.73 to 1.30).

Conclusions:

• Awareness of analytical pitfalls is essential for maximizing the benefits of RWD in oncology.
• Incorrect handling of missing data, early deaths, and time-related biases can lead to erroneous survival estimates.
• Employing robust statistical methods is critical for reliable survival analyses with RWD.