Kaplan-Meier Survival Curve
A step function plot showing the estimated probability of survival over time — used in clinical trials and epidemiology to compare treatment groups and visualize time-to-event data.
// 01 — The chart
What it looks like
A Kaplan-Meier curve comparing overall survival between Drug A (green) and Placebo (red). Each step down represents an observed event. Tick marks indicate censored observations.
// 02 — Definition
What is a Kaplan-Meier curve?
A Kaplan-Meier curve (also called a survival curve or KM plot) is a non-parametric statistic used to estimate the survival function from time-to-event data. It produces a distinctive staircase-shaped line that starts at 1.0 (100% survival) and steps downward each time an event — such as death, disease recurrence, or equipment failure — is observed.
What makes the Kaplan-Meier estimator so powerful is its ability to handle censored data. In clinical trials, not every patient reaches the endpoint: some drop out, some are lost to follow-up, and some are still alive when the study ends. Censored observations are marked on the curve with small tick marks, indicating that the patient was still event-free when last observed.
Each step in the curve represents a recalculation of the survival probability. The height of the step is proportional to the number of events occurring at that time point relative to the number of subjects still at risk. This makes the curve a true probability estimate, not just a raw count.
Origin: The Kaplan-Meier estimator was introduced by Edward L. Kaplan and Paul Meier in their 1958 paper published in the Journal of the American Statistical Association. It remains one of the most cited papers in statistics, with over 60,000 citations, and is the standard method for survival analysis in biomedical research.
// 03 — Anatomy
Parts of a Kaplan-Meier curve
// 04 — Usage
When to use it — and when not to
- Analyzing time-to-event data in clinical trials (overall survival, progression-free survival)
- Comparing survival between two or more treatment groups side by side
- Your dataset contains censored observations that must be accounted for
- Presenting results to a medical or biostatistical audience familiar with the format
- You want to estimate median survival time visually from the 0.5 probability line
- Showing the full trajectory of event occurrence over the study period
- Your data is not time-to-event — KM curves are specifically designed for survival analysis
- You need to adjust for confounding variables — use Cox regression instead
- Your audience is non-technical and unfamiliar with probability curves
- You have very few events or subjects, making the curve unreliable
- You want to show cumulative incidence of competing risks — use a competing risks model
- The proportional hazards assumption is severely violated between groups
// 05 — Reading guide
How to read a Kaplan-Meier curve
Follow these steps whenever you encounter a Kaplan-Meier curve in a clinical paper or report.
Identify what event is being measured
The Y-axis shows survival probability, but “survival” can mean different things: overall survival (death), progression-free survival (disease worsening), or even mechanical failure. Read the title and axis labels to understand the specific endpoint.
Note the time scale on the X-axis
Check whether time is measured in days, months, or years. Also check the total follow-up period — the curve becomes less reliable at later time points when fewer patients remain at risk.
Compare the separation between curves
When comparing treatment groups, look at how far apart the curves are. Wider separation means a larger survival difference. If curves cross, it may indicate the treatment effect changes over time.
Find the median survival time
Draw a horizontal line from 0.5 on the Y-axis to the curve, then drop down to the X-axis. This gives you the time at which 50% of subjects have experienced the event. Compare medians across groups to quantify the treatment benefit.
Check the number at risk and p-value
Look for the “number at risk” table below the chart — it shows how many subjects remain at each time point. Also check the log-rank test p-value, which tells you whether the difference between curves is statistically significant.
// 06 — Common mistakes
Mistakes to watch out for
Ignoring the number at risk
The tail of a Kaplan-Meier curve often represents only a handful of patients. A dramatic drop at month 30 might involve just 2–3 subjects, making it statistically unreliable. Always check the at-risk table and consider truncating the curve where the sample becomes too small.
Omitting censoring marks
Censoring tick marks are essential for readers to assess data quality. Without them, it’s impossible to tell whether a long plateau means no events occurred or everyone was censored. Always display censoring indicators on the curve.
Over-interpreting crossing curves
When two Kaplan-Meier curves cross, it usually means the proportional hazards assumption is violated. The log-rank test may not be valid in this case, and a different statistical approach (like restricted mean survival time) may be more appropriate.
Confusing survival probability with percentage alive
The Y-axis is an estimate derived from the Kaplan-Meier formula, not a simple count of who is alive. It accounts for censoring and time-varying risk sets. Treating it as a raw percentage leads to misinterpretation, especially when censoring is heavy.
Not showing confidence intervals
A Kaplan-Meier curve without confidence bands can be misleading, especially with small sample sizes. The pointwise 95% confidence interval widens over time as fewer subjects remain, and displaying it helps readers judge uncertainty.
// 07 — Real-world examples
Where you’ll see Kaplan-Meier curves used
Oncology: Comparing immunotherapy vs chemotherapy
Landmark trials like KEYNOTE-024 use Kaplan-Meier curves to show overall survival and progression-free survival, demonstrating that pembrolizumab significantly extends survival compared to platinum-based chemotherapy in non-small-cell lung cancer.
Clinical TrialsCardiology: Heart failure device trials
Device trials such as PARADIGM-HF use Kaplan-Meier curves to compare the time to cardiovascular death or heart failure hospitalization between sacubitril/valsartan and enalapril, showing the separation of curves over years of follow-up.
Medical DevicesEngineering: Equipment reliability analysis
Manufacturing companies use Kaplan-Meier curves to estimate the survival function of components — such as ball bearings or turbine blades — where some units are removed from service before failure (censored), making traditional failure-rate calculations inaccurate.
Reliability Engineering// 08 — At a glance
Quick reference
// 09 — Variations
Types of survival curves
The classic Kaplan-Meier estimator has several important variations used in different analytical contexts.
KM with confidence bands
Adds shaded 95% confidence intervals around the survival estimate to show statistical uncertainty, which widens as sample size decreases.
Cumulative incidence curve
The complement of the survival curve (1 – S(t)), showing the probability of having experienced the event by time t. Steps upward instead of downward.
Multi-group comparison
Overlays survival curves for multiple treatment arms or subgroups, each in a different color, with a log-rank test comparing them.
KM with number-at-risk table
Includes a table below the X-axis showing how many subjects remain at risk at each time point, adding crucial context for interpretation.
// 10 — FAQs
Frequently asked questions
What is a kaplan-meier survival curve?+
A Kaplan-Meier curve (also called a survival curve or KM plot) is a non-parametric statistic used to estimate the survival function from time-to-event data. It produces a distinctive staircase-shaped line that starts at 1.0 (100% survival) and steps downward each time an event — such as death, disease recurrence, or equipment failure — is observed.
When should you use a kaplan-meier survival curve?+
Use a kaplan-meier survival curve when analyzing time-to-event data in clinical trials (overall survival, progression-free survival). It also works well when comparing survival between two or more treatment groups side by side, and when your dataset contains censored observations that must be accounted for.
When should you avoid a kaplan-meier survival curve?+
Avoid a kaplan-meier survival curve when your data is not time-to-event — KM curves are specifically designed for survival analysis. It is also a poor fit when you need to adjust for confounding variables — use Cox regression instead, or when your audience is non-technical and unfamiliar with probability curves.
Is a kaplan-meier survival curve suitable for dashboards?+
Yes — a kaplan-meier survival curve can work well in dashboards as long as the panel is large enough for readers to perceive the encoded values, has a clear title, and includes the legend or axis labels needed to interpret it.
What category of chart is a kaplan-meier survival curve?+
Kaplan-Meier Survival Curve belongs to the Scientific family of charts. Charts in that family are designed to answer the same kind of question, so they often work as alternatives when one doesn't quite fit your data.
How do you read a kaplan-meier survival curve?+
Start with the axis labels and legend, then look at the overall shape before zooming into individual marks. Compare prominent features against the rest of the data, and verify any conclusion against the underlying numbers when precision matters.