01.Association07.Attributable fraction, number needed to treat, and relative excess risk due to interaction - sporedata/researchdesigneR GitHub Wiki

1. Common condition general public

Attributable fraction, the number needed to treat, and relative excess risk due to interaction are all measures for causal inference. They do not reach the level of granularity of confounding control achieved by propensity score matching, the difference in difference, and other methods, but provide different types of information that might be useful.

The number needed to treat (NNT) is the average number of patients who need to be treated to prevent one additional bad outcome. It is defined as the inverse of the absolute risk reduction (defined as the difference between the incidence in the exposed vs the non-exposed).

Although popularized by the Evidence-Based Medicine movement as a clinically relevant type of effect size, NNT often has very wide confidence intervals, making its interpretation often meaningless -- see Misleading Statistics: The Problems Surrounding Number Needed to Treat and Number Needed to Harm

The attributable fraction is the proportional reduction in a condition among a given population (or mortality) that would occur if one were to reduce risk factor levels to an alternative scenario. Last, the relative excess risk due to interaction is a deviation from additivity of effects on a relative-risk scale.

A form of inference that can use probability to quantify the evidence about an unknown quantity such as the absolute risk is the Bayesian inference. Bayesian analysis can find that there is, for example, a 60% probability that the absolute risk reduction is 3% or more. For a particular observed set of results from a clinical trial, these Bayesian probabilities can differ between analyses that use different prior distributions [2].

• Attributable fraction - Causal analysis
• Number needed to treat - Cost-effectiveness analysis
• Relative excess risk - Models for outcomes measured a single time and Models with grouped outcomes

1. Books

   • Causal Inference in Statistics - A Primer

2. Articles

   • Estimation of causal effect measures with the R-package stdReg
   • Estimating the Relative Excess Risk Due to Interaction: A Bayesian Approach
   • How to calculate the Number Needed to Treat (NNT)
   • The numbers needed to treat and harm (NNT, NNH) statistics: what they tell us and what they do not.
   • Number needed to treat

• Attributable fraction - estimates the portion of outcomes that could be avoided if the exposure were eliminated, and as a result, can provide meaningful information to the healthcare system.

• Number needed to treat - is used as an epidemiological measure of the minimum number of people needed to treat in order to avoid worse outcomes - e.g., populace vaccination ratio.

• Relative excess risk due to interaction - is a suitable measure to compare outcomes based only on the treatment or exposure, not considering the background risk.

• stdReg [1].

[1] Sjölander A. Regression standardization with the R package stdReg. European journal of epidemiology. 2016 Jun 1;31(6):563-74.
[2] Yarnell CJ, Abrams D, Baldwin MR, Brodie D, Fan E, Ferguson ND, ... & Goligher EC. Clinical trials in critical care: can a Bayesian approach enhance clinical and scientific decision making?. The Lancet Respiratory Medicine. 2021 Fev;9(2):207-216.