How to digitize a Kaplan-Meier curve
Rebuilding patient-level data from a published survival curve needs two things: the curve's coordinates and the numbers at risk. Both are read here in one pass, exactly when the paper is a PDF, and the patient-level data can be rebuilt from them for you.
What is read
- Each arm's curve, as the point where every new level begins.
- The number-at-risk table, row by row, with the times its columns stand under.
- Censoring marks are left out: they are not steps.
Steps
- Upload the paper's PDF, or a sharp picture of the figure with its table.
- Check the curve drawn back on the figure, arm by arm.
- Download Excel: the curve and the table are on the same sheet.
Limits
From a PDF drawing the steps are exact. From a picture, steps closer together than a pixel are merged in the picture itself, so no tool can separate them.
We have measured this on survival curves we drew ourselves, not yet on figures from journals. Check the result against the medians the paper reports.
Questions
Does it rebuild the patient-level data?
Yes, with the Review pass, for a curve read from a PDF with its number-at-risk table. The Excel file gets a sheet with one row for each patient: a time, and whether it ended in the event or in censoring. On 300 survival charts drawn from patients we simulated and read back from the PDF, the rebuilt data had the right number of patients in all 644 groups and exactly the right number of events in 627 of them (within one in 639). The hazard ratio between two groups came out within 2% of the true one on 247 of 251 charts, and within 5% on 250.
Is reading the curve itself free?
Yes. The steps of each curve and the number-at-risk table come out free from a PDF, without limit. The pass adds the rebuilt patients.