← Back to Tutorials

How to Digitize Kaplan-Meier Curves for Meta-Analysis

⚡ Quick Answer

Yes. Import the Kaplan-Meier figure into Engauge Digitizer, set X to time and Y to survival probability (0–1), trace each treatment arm as its own curve, and export time–probability pairs as CSV. Those pairs feed directly into IPD reconstruction (Guyot et al. 2012) for your meta-analysis.

Why Digitize Kaplan-Meier Curves?

In oncology and cardiology, trial results are published as Kaplan-Meier survival plots — but the underlying patient data stays with the authors. For a meta-analysis or systematic review, researchers routinely recover the numbers by digitizing the published curves, then reconstruct individual patient data (IPD) using the well-validated Guyot et al. (2012) algorithm (implemented in R packages like IPDfromKM). Digitizing is step one of that pipeline — and it is the step this guide covers.

Step-by-Step: Digitizing a KM Curve

  1. Import the figure: Use the highest-resolution version available — the journal PDF, not a thumbnail. Drag it into Engauge Digitizer.
  2. Set the axes: In the Axis Wizard: X axis = time (days, months, or years — match the label), Y axis = survival probability from 0 to 1 (if the axis shows percentages, use 0–100 and convert later). Place axis points on clearly labeled grid lines.
  3. Add one curve per arm: Open the Curves List (View > Curves) and click Add New Curve for each treatment arm — name them “Treatment” and “Control” so the CSV columns stay organized.
  4. Trace the steps: KM curves are staircases: use the Point Match tool or trace manually, clicking at every step — especially where the curve drops. Ignore the censoring tick marks; they mark dropouts, not survival values.
  5. Export: File > Export to CSV. You now have time–probability pairs per arm, ready for IPD reconstruction.

💡 Tip: Copy the number-at-risk table printed under most KM plots into your spreadsheet too — IPD reconstruction algorithms use it to calibrate the estimates. It is typed by hand, not digitized.

From CSV to IPD: The Next Steps

Digitizing gives you the curve coordinates. Reconstruction turns them into patient-level data:

  1. Prepare inputs: Your digitized time–probability pairs plus the number-at-risk counts at each time point.
  2. Run the algorithm: Use the Guyot et al. (2012) method — the R package IPDfromKM (with its Shiny app) is the most accessible implementation.
  3. Validate: Plot the reconstructed curve against your digitized points; they should overlap closely. Report the digitizing software, version, and method in your paper’s methods section.

⚠️ Note: Reconstructed IPD is an estimate for research synthesis — excellent for meta-analysis, but not a substitute for the original trial data in every context. Always report that the data were reconstructed from published curves.

🎓 From extraction to meta-analysis

Digitizing curves is one step of a systematic review. The Google Data Analytics certificate on Coursera teaches spreadsheets, statistics foundations, and data cleaning — the exact skills behind a publishable meta-analysis.

Disclosure: this is an affiliate link; we may earn a commission at no extra cost to you.

Related Guides

How to Digitize Bar Charts with Error Bars

Recover means and uncertainty from bar plots for meta-analysis.

Is Digitized Data Accurate Enough to Publish?

What validation studies say about digitizing precision.

Frequently Asked Questions

Can I digitize a Kaplan-Meier curve with Engauge Digitizer?

Yes. Import the figure, set the X axis to time and the Y axis to survival probability (0 to 1), then trace each survival curve with the automatic tools or manually. Export the time–probability pairs as CSV for further analysis.

How do I handle the censoring tick marks on a KM curve?

Ignore them while tracing — censor marks show where patients left the study, but they are not part of the survival step function. Trace the stepped line itself; the ticks are visual noise for digitizing purposes.

What is IPD reconstruction from a Kaplan-Meier curve?

IPD (individual patient data) reconstruction estimates patient-level survival times from a published KM curve, usually with the number-at-risk table. The standard method is the Guyot et al. (2012) algorithm, implemented in R packages such as IPDfromKM. It lets researchers run survival analyses without access to the original trial data.

How many points should I digitize along the curve?

Capture every step of the staircase, especially around drops — those carry the information. A good rule: digitize at least 50–100 points per curve, denser where the curve falls steeply. More points give smoother, more faithful IPD reconstruction.

Can I digitize multiple treatment arms from one figure?

Yes. Add a separate curve for each arm in the Curves List (View > Curves > Add New Curve), name them (e.g. “Treatment” and “Control”), and trace each one. They export as separate columns in one CSV file.

Ready to try it yourself?

Download Engauge Digitizer free and follow along with this guide.

Download Engauge Digitizer