Are we assuming too much with our statistical assumptions? Lessons learned from the ALTTO trial.

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Oncology Research

Background:

• Randomized clinical trials (RCTs) with time-to-event endpoints rely on statistical assumptions.
• These assumptions include event rates, proportionality of treatment effects, and event differences over time and between subgroups.

Purpose of the Study:

• To highlight how routinely applied statistical assumptions can impact RCT result reporting.
• To use the Adjuvant Lapatinib and/or Trastuzumab Treatment Optimization (ALTTO) trial experience to illustrate these impacts.

Main Methods:

• Analysis of data from the ALTTO RCT (NCT00490139), which enrolled 8381 patients with HER2-positive early breast cancer.
• Evaluation of the impact of statistical assumptions on reporting of time-to-event endpoints.
• Comparison of results from Cox models and accelerated failure time models.

Main Results:

• Early stopping for efficacy in RCTs can be misleading; futility stopping rules are important for patient safety.
• Pre-specifying subgroup analyses and focusing on relevant endpoints are crucial for capturing clinically significant differences.
• Non-proportional hazards can significantly affect conclusions, and the Cox model may be misleading if not carefully considered.
• The assumption that more events always increase statistical power needs re-evaluation, considering changes in hazard rates over time.
• Strict type 1 error control in multi-arm trials can limit the ability to answer multiple research questions.

Conclusions:

• Statistical assumptions in RCTs require careful consideration to ensure accurate and reliable reporting of results.
• Methodological rigor in trial design and analysis, including subgroup analysis and handling of non-proportional hazards, is essential.
• Future RCTs should critically evaluate standard assumptions and consider alternative analytical approaches to better interpret time-to-event data.