# Confounding by indication

Source: https://onco.cc/terms/confounding-by-indication/  
OnCo record `confounding-by-indication` (Term). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

People often receive different treatments because their health or cancer differs. Those starting differences can distort a comparison of outcomes between treatments.

## Summary

A primary observational breast cancer study examined how treatment selection could make chemotherapy appear associated with recurrence and how adjustment methods changed the estimate. Measured-covariate adjustment did not eliminate the study's concern about residual confounding. A treatment association in routine-care data must be interpreted with the reasons treatment was selected. Propensity scores and large sample size do not automatically reproduce random assignment.

## Fields

- Kind: Term
- Last checked: 2026-11-11
- Also known as: treatment-selection confounding; indication bias

## Notes

- Confounding by indication arises when a reason for choosing treatment also relates to the outcome, such as giving more intensive therapy to people with more severe disease. Baseline fitness, access and prior care can also influence both treatment selection and outcome. A treatment-outcome association alone does not separate those pathways from the effect of treatment. The causal role and timing of each variable require the actual study context.
- Fictional six-week symptom example: each treatment group contains 100 people. With A, 60 of 80 people in better baseline health improve and 5 of 20 in poorer health improve. With B, the corresponding counts are 15 of 20 and 20 of 80. Within each health category both treatments therefore have the same rates, 75 and 25 per cent, yet the pooled rates are 65 and 35 per cent because the baseline mixtures differ. This arithmetic illustrates a confounding explanation, not a recipe showing that matching one factor fixes every real comparison. Other differences, imperfect measurements, missing follow-up and clocks still matter.

## Sources

- Treatment-selection confounding in an observational breast cancer study (2009): https://doi.org/10.1016/j.jclinepi.2009.03.001
- Specifying a target trial prevents immortal time bias and other self-inflicted injuries in observational analyses: https://doi.org/10.1016/j.jclinepi.2016.04.014
- The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.: https://doi.org/10.1371/journal.pmed.0040296

## Connected records

- terms: [Evidence from measurement to patient benefit](https://onco.cc/terms/measurement-to-benefit-evidence/), [Real-world evidence](https://onco.cc/terms/real-world-evidence/), [Target-trial emulation](https://onco.cc/terms/target-trial-emulation/), [Transporting trial evidence to another population](https://onco.cc/terms/trial-population-transportability/)

---
JSON: https://onco.cc/api/v1/entities/confounding-by-indication.json