- Published
- Updated
A single hazard ratio may not adequately describe the entire follow-up experience. If effects change over time, assess both assumption suitability and clinical meaning. Successful generation of model output is not sufficient justification for treating one estimate as universally applicable across time.
Define the question and assumptions
Review event patterns together with the intended treatment-effect question. Diagnostics, visual inspection and clinical knowledge can complement one another. Alternative summaries may target different quantities; do not select them only because results appear more favourable. Explain primary and supportive methods prospectively with appropriate expertise, including what each method can reasonably describe.
Hypothetical study example
In a hypothetical trial, groups are similar early and diverge later. One ratio does not clearly describe which aspect of that experience is summarised. The team presents time patterns and uncertainty instead of interpreting the ratio as an identical benefit at every time without discussing the assumption.
Design review checks
- Review events and numbers at risk throughout the available follow-up period.
- Explain assumption checks and their connection to the clinical interpretation.
- Distinguish the quantities targeted by alternative summary methods.
Record the decision and its limits
Accepted reporting separates the main method, assumption assessment and additional analyses. Where few people remain late in follow-up, identify uncertainty in graphical patterns. Do not present post-result method searching as a confirmatory plan. A hazard ratio is not directly interchangeable with a risk ratio; make the time horizon and remaining limitations understandable.
Frequently asked questions
- Does one hazard ratio represent every time point?
- Not necessarily. Time-varying patterns can limit that interpretation.
- Does an alternative method answer the same question?
- Not always; define its target quantity and interpretation explicitly.