Kaplan-Meier Curve

The Kaplan-Meier (KM) curve is the standard survival analysis visualization. A KM curve without an at-risk table is incomplete by modern reporting standards —readers cannot evaluate whether the curves at late time points are based on adequate numbers. Licklider generates the at-risk table, censoring marks, and hazard ratio automatically from the event-time data.

STEP 1 —When to Use

Use a KM curve when: the outcome is time-to-event (survival, time to relapse, time to first event); the data include censoring (patients lost to follow-up or event-free at study end); you want to compare survival distributions between two or more groups (log-rank test); you are reporting OS, PFS, EFS, or any time-to-event endpoint in clinical research.

At-risk table is mandatory: the number of subjects at risk at each time point must appear below the KM curves. This allows readers to evaluate the reliability of the curve at late time points —where small n makes the curve unreliable.

[Image placeholder: KM curve with 2 groups. Step-function curves with censoring tick marks (vertical small tick on the curve at censored time points). At-risk table below: two rows (one per group) showing n at risk at each labeled time point. Median survival with 95% CI for each group. Log-rank p-value and HR shown as annotation.]

Complete KM figure: step curves + censoring marks + at-risk table + median survival + log-rank p + HR.

STEP 2 —Journal Requirement

At-risk table mandatory: n at risk at each time point for each group (CONSORT, JAMA, Lancet, NEJM all require this).
Censoring marks: tick marks at each censored time point on the survival curve.
Log-rank p-value and hazard ratio (HR) with 95% CI must be reported.
Median survival with 95% CI per group must be stated.
For curves that do not reach 50% (median not reached): "median not reached" must be stated.
If more than 2 groups: correction for multiple log-rank comparisons is required.

STEP 3 —Licklider Implementation

Input

Time column (numeric, in declared units: days, months, years)
Event column (binary: 1 = event occurred, 0 = censored)
Group variable (for multi-group comparison)
Confidence interval method: Greenwood (default), log transformation, log-log

Output

KM step-function curves per group
95% CI shaded band per curve (Greenwood formula)
Censoring tick marks at each censored time point
At-risk table below the figure (n at risk at user-specified or auto-spaced time points)
Median survival + 95% CI per group (annotated on figure or in legend)
Log-rank test p-value and HR with 95% CI (from Cox proportional hazards model)
Restricted mean survival time (RMST) optional alternative to median

[Image placeholder: Licklider KM curve. Two groups (n=45 and n=43). Step curves with 95% CI bands, censoring ticks on curves. At-risk table below with n at risk at 0, 6, 12, 18, 24, 30 months. HR = 0.62 (95% CI [0.39, 0.98]) and log-rank p = 0.041 shown in top-right corner. Median survival: Group A = 18.4 months (95% CI [14.2, 22.6]), Group B = 24.8 months (95% CI [19.3, NR]).]

Licklider KM curve: CI bands, censoring ticks, at-risk table, HR with 95% CI, log-rank p, and median survival per group.

STEP 4 —Draft Output (Draft / Needs review)

Kaplan-Meier survival curves for Group A (n = 45) and Group B (n = 43).
Shaded regions show 95% confidence intervals (Greenwood formula).
Tick marks indicate censored observations. Numbers at risk at each
time point are shown in the at-risk table below the figure.
Median survival: Group A = 18.4 months (95% CI [14.2, 22.6]);
Group B = 24.8 months (95% CI [19.3, not reached]).
Log-rank test: chi²(1) = 4.2, p = 0.041. Hazard ratio (Group A vs. Group B)
= 0.62 (95% CI [0.39, 0.98]), Cox proportional hazards model.