Skip to contentScholarGate
LibraryBookshelfDeskReview StudioAssistant
Sign in
On this page
IntuitionHow it worksWhen to use itStrengths & limitationsCommon pitfallsApplicationsFrequently asked🔒 Read the full methodSourcesRelated methods
Cite this pageSpotted an issue on this page? Report or suggest a fix →
Home›Epidemiology›Matched Survival Analysis — Matched Cohort Survival Analysis
Process / pipelineClinical / epidemiology

Matched Survival Analysis — Matched Cohort Survival Analysis

Matched Cohort Survival Analysis · Also known as: matched time-to-event analysis, propensity-matched survival analysis, matched Kaplan-Meier analysis, paired survival analysis

Matched survival analysis combines a matching design — typically propensity score matching or exact matching on key covariates — with time-to-event methods such as Kaplan-Meier estimation and the Cox proportional hazards model. By pairing treated and control subjects who are similar on observed confounders before estimating survival curves or hazard ratios, the approach reduces confounding bias in non-randomised studies and produces more credible comparisons of event-free survival between exposure groups.

ScholarGate
  1. Process / pipeline
  2. v1
  3. 2 Sources
  4. PUBLISHED
Cite this page →
Tools & resources
Download slides
Learn & explore

Read the full method

Members only

Sign in with a free account to read this section.

Sign in

Method map

The neighbourhood of related methods — select a node to explore.

Matched Survival Analysis
Cox proportional hazardsKaplan-Meier EstimatorLog-Rank TestPropensity Score Matching

When to use it

Use matched survival analysis when you have observational time-to-event data and want to compare survival between a treated and a control group while controlling for measured confounders — and when randomisation was not feasible. It is well-suited to clinical registry data, pharmacoepidemiology, and comparative effectiveness studies where sample sizes are sufficient to achieve good balance after matching (typically at least 50–100 matched pairs). Do not use it when the number of measured confounders greatly exceeds the available sample, when there is substantial unmeasured confounding (in which case instrumental variable methods or sensitivity analyses for hidden bias should be added), when the proportional hazards assumption is badly violated without a remedy, or when the matching process discards so many subjects that the matched sample no longer represents the target population of interest.

Strengths & limitations

Strengths
  • Reduces confounding by observed covariates without requiring a fully parametric outcome model, making causal comparisons in observational data more credible.
  • Produces intuitive, directly interpretable outputs: Kaplan-Meier curves visually display cumulative survival for matched groups.
  • The stratified log-rank test and robust Cox model correctly account for the within-pair correlation introduced by matching.
  • Flexible matching algorithms (propensity score, exact, Mahalanobis) allow the approach to be tailored to different data structures and clinical questions.
  • Widely accepted in clinical and epidemiological journals, making results easy to communicate to medical audiences.
Limitations
  • Controls only for measured covariates; unmeasured or unknown confounders remain a threat to validity and should be addressed with sensitivity analyses (e.g., E-value, Rosenbaum bounds).
  • Matching typically discards unmatched subjects, potentially reducing statistical power and limiting generalisability to the matched population rather than the full cohort.
  • The Cox model's proportional hazards assumption must be checked; violations require time-varying coefficients, stratified models, or alternative estimators (e.g., restricted mean survival time).
  • With many covariates and a small sample, propensity score estimation can be unstable, leading to poor balance even after matching.

Frequently asked

Why use the stratified log-rank test rather than the ordinary log-rank test on matched data?

Matching creates pairs (or sets) of subjects who are intentionally similar. These pairs are not independent observations — they share covariate values by design. The ordinary log-rank test assumes independence and underestimates variance when applied to matched data, inflating the Type I error rate. The stratified log-rank test conditions on the matched pairs by treating each pair as a stratum, correctly accounting for the within-pair correlation.

How do I handle ties in survival times within matched pairs?

Ties are handled by the standard tie-breaking options in the Kaplan-Meier estimator and Cox model (Breslow, Efron, or exact methods). Matching itself does not introduce additional complexity around ties; the choice of tie method should follow the same guidance as in unmatched survival analysis, with Efron's approximation being a common default in most statistical software.

What if my Cox model violates the proportional hazards assumption after matching?

First confirm the violation with Schoenfeld residual plots and the global proportional hazards test. If violated, consider adding time-by-covariate interaction terms, using a stratified Cox model, reporting restricted mean survival time (RMST) differences as an alternative summary measure, or using flexible parametric models. The violation should be reported and the chosen remedy justified.

Can I use inverse probability of treatment weighting (IPTW) instead of matching for survival outcomes?

Yes. IPTW is a closely related propensity-score method that retains all subjects (no discarding) by reweighting observations by the inverse of their probability of receiving their actual treatment. For survival outcomes, IPTW can be combined with the Kaplan-Meier estimator (producing marginal structural Kaplan-Meier curves) or with a weighted Cox model. Matching and IPTW address the same confounding problem; the choice depends on the study's estimand (matched population vs. full population) and the sample size available.

How large a sample do I need for matched survival analysis?

There is no fixed universal rule, but as a practical guide you need enough subjects to (1) achieve good covariate balance after matching — typically at least 50–100 matched pairs — and (2) observe a sufficient number of events for reliable survival estimation and Cox model fitting. As a rough heuristic, at least 10 events per covariate in the Cox model is recommended. Very small matched samples (fewer than 30 pairs) yield wide confidence intervals and unstable Kaplan-Meier estimates.

Sources

  1. Austin, P. C. (2014). Graphical assessments of the balance of propensity score matched samples: A SAS macro. Journal of Statistical Software, 58(7), 1-29. Also see Austin, P. C. (2017). A tutorial on multilevel survival analysis: Methods, models and applications. International Statistical Review, 85(2), 185-203. link ↗
  2. Collett, D. (2015). Modelling Survival Data in Medical Research (3rd ed.). CRC Press. ISBN: 9781439856789

How to cite this page

ScholarGate. (2026, June 3). Matched Cohort Survival Analysis. ScholarGate. https://scholargate.app/en/epidemiology/matched-survival-analysis

Related methods

Cox proportional hazardsKaplan-Meier EstimatorLog-Rank TestPropensity Score Matching

Which method?

Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.

  • Cox proportional hazardsEpidemiology↔ compare
  • Kaplan-Meier EstimatorStatistics↔ compare
  • Log-Rank TestSurvival↔ compare
  • Propensity Score MatchingResearch Statistics↔ compare
Compare side by side →

Similar methods

Matched Kaplan-Meier AnalysisMatched Cox Proportional HazardsMatched Cohort StudyRisk-adjusted survival analysisRetrospective survival analysisRisk-adjusted Kaplan-Meier analysisSurvival AnalysisKaplan-Meier Analysis

Related reference concepts

Study Matching and StratificationSurvival Analysis and Time-to-Event MethodsCox Regression ModelsProportional Hazards AssumptionKaplan-Meier Survival CurvesCensoring and Follow-Up Data

Spotted an issue on this page? Report or suggest a fix →

ScholarGate — Matched Survival Analysis (Matched Cohort Survival Analysis). Retrieved 2026-07-21 from https://scholargate.app/en/epidemiology/matched-survival-analysis · Dataset: https://doi.org/10.5281/zenodo.20539026
Quick facts
Originator
Building on Kaplan & Meier (1958) and Cox (1972); matching framework formalised in observational study design literature (Rosenbaum & Rubin, 1983)
Year
1983 (propensity-score matching); applied to survival outcomes throughout 1990s–2000s
Type
Observational study analytic method
DataType
Time-to-event data with censoring; matched pairs or matched sets from cohort or registry data
Subfamily
Clinical / epidemiology
Related methods
Cox proportional hazardsKaplan-Meier EstimatorLog-Rank TestPropensity Score Matching
ScholarGate

A content-first reference library for research methods — what each one is, how it works, and where it comes from.

Open data (CC-BY)

Explore

  • Library
  • Search the library…
  • Browse by field
  • Fields
  • Journey
  • Compare
  • Which method?

Reference

  • Subjects
  • Atlas
  • Glossary
  • Methodology
  • Philosophy

Your tools

  • Bookshelf
  • Desk
  • Chat

Company

  • About
  • Pricing
  • Contact
  • Suggest a method

Entries are compiled from published sources for reference. Verifying the accuracy and suitability of any information for your own use remains your responsibility.

© 2026 ScholarGate · A research-method reference library
  • Privacy
  • Cookies
  • Terms
  • Delete account