Treatment Effect Heterogeneity

Concept

---

Theoretically, Treatment Effect Heterogeneity describes variation in the magnitude or direction of a treatment effect across individuals or predefined population subgroups. It is founded on causal inference, statistical interaction and potential outcomes theory. The concept exists because interventions rarely produce identical effects for all patients owing to differences in demographics, disease severity, genetics, comorbidities or other effect modifiers. Recognising treatment effect heterogeneity is essential for personalised medicine, precision health and efficient resource allocation.

Mathematically, Treatment Effect Heterogeneity is represented by allowing treatment effects to vary as a function of observed or unobserved characteristics. The recognised mathematical framework incorporates interaction terms within regression models, hierarchical models or causal inference methods to estimate subgroup-specific treatment effects while accounting for uncertainty. The presence of statistically significant interaction indicates that the treatment effect differs across levels of the modifying variable.

In practice, Treatment Effect Heterogeneity is evaluated using subgroup analyses, interaction tests, hierarchical modelling, meta-regression, individual participant data meta-analysis or machine learning approaches. In health economics it is incorporated into cost-effectiveness models to estimate subgroup-specific costs, health outcomes and incremental cost-effectiveness ratios, thereby informing reimbursement decisions and targeted treatment recommendations.

Purpose

---

Used to identify, estimate and quantify differences in treatment effectiveness across patient groups to support personalised healthcare, subgroup-specific economic evaluation and evidence-based resource allocation.

Mathematical Formulae

Primary Formula

Y = ?? + ??T + ??X + ??(T ? X) + �

where ?? represents treatment effect heterogeneity.

Supporting Formulae

Conditional average treatment effect:

CATE(x) = E[Y(1) ? Y(0) � X = x]

Average treatment effect:

ATE = E[Y(1) ? Y(0)]

Null hypothesis for interaction:

H?: ?? = 0

Related Mathematical Methods

• Interaction Analysis
• Subgroup Analysis
• Generalised Linear Models
• Hierarchical Models
• Meta-Regression
• Individual Participant Data Meta-analysis
• Causal Inference
• Bayesian Hierarchical Modelling

---

Example

A randomised trial evaluates a new diabetes therapy. The overall treatment effect is an increase of 0.6 quality-adjusted life-years (QALYs). An interaction analysis shows patients younger than 65 years gain 0.9 QALYs, whereas patients aged 65 years or older gain 0.3 QALYs. The estimated interaction coefficient is statistically significant (?? = 0.6, p = 0.01), demonstrating treatment effect heterogeneity. Separate cost-effectiveness analyses are therefore undertaken for each age subgroup.

---

Excel Implementation

Function	Example Formula	Health Economics Application
LINEST	=LINEST(Y2:Y201,A2:C201,TRUE,TRUE)	Estimate regression coefficients including treatment-by-subgroup interaction
IF	=IF(D2>=65,"Older","Younger")	Define subgroup membership
AVERAGEIFS	=AVERAGEIFS(E:E,B:B,"Treatment",D:D,"<65")	Estimate subgroup-specific treatment outcomes
T.TEST	=T.TEST(E2:E51,F2:F51,2,2)	Compare treatment effects between subgroups

---

VBA (Optional)

Automate subgroup-specific regression analyses and generate treatment effect heterogeneity summaries across multiple patient characteristics.

---

Sources

• Kent DM, Rothwell PM, Ioannidis JPA, Altman DG, Hayward RA. Assessing and Reporting Heterogeneity in Treatment Effects in Clinical Trials.
• Hern�n MA, Robins JM. Causal Inference: What If.
• Harrell FE. Regression Modeling Strategies.
• NICE. Health Technology Evaluations: The Manual.
• Drummond MF, Sculpher MJ, Claxton K, Stoddart GL, Torrance GW. Methods for the Economic Evaluation of Health Care Programmes.
• ISPOR Good Practice Reports.