SLNB · Melanoma outcomes research
SLNB: survival analysis for melanoma outcomes
Sentinel lymph node biopsy (SLNB) is an invasive surgical procedure used for staging in melanoma. Because it carries risks and costs, we study its association with outcomes using large melanoma datasets. This page is an illustrative walkthrough of the time-to-event methods used for such analyses. The data shown here are simulated for demonstration purposes. Results from real-world cohorts are to be reported separately in peer-reviewed publications.
01 The challenge
Not all patients finish a study at the same time
In time-to-event studies, some patients experience the event of interest while others are still event-free when follow-up ends. This incomplete information is called censoring, and it makes standard comparisons unreliable.
Censored patients still carry information: we know they were event-free up to their last contact. Survival analysis is built to use that partial information correctly.
Simple summaries like “average time to event” are not appropriate here because they discard partial follow-up and can give biased estimates of event-free probability.
For readability, the panel below shows 40 illustrative patients; the real cohorts I work with include several thousand patients.
02 The survival curve
Estimating event-free probability over follow-up
The Kaplan-Meier curve tracks the probability of remaining event-free as time passes. Each step down marks an observed event. Censoring is shown as tick marks and does not make the curve drop. Flat stretches indicate intervals with no observed events.
When comparing two groups (here: SLNB vs no SLNB), the gap between curves suggests a difference in event rates. The hazard ratio summarises the relative event rate between groups over time: below 1 corresponds to lower hazard in the SLNB group; above 1 corresponds to higher hazard (worse than the comparator in this illustrative example).
03 Adjusted comparisons
Accounting for differences between groups
In real-world cohorts, patients who undergo SLNB can differ from those who do not (for example in age, sex, and tumour characteristics). To reduce bias from measured differences, we use propensity score methods (e.g., weighting, matching, or stratifying patients by their estimated propensity to receive SLNB).