Decision Curve Analysis for Multiple Treatments
This method helps doctors and patients choose the best treatment when there are several options by showing which strategy—like using a personalized prediction model or treating everyone the same way—leads to better outcomes across different patient preferences.
At a glance
Use when
Evaluating prediction models for treatment selection in settings with multiple therapeutic options; supporting personalized medicine decisions using synthesized trial evidence; comparing clinical utility of different treatment strategies in HTA or guideline development
Avoid when
When only two treatment options exist and standard DCA suffices; when no prediction model for treatment benefit is available; when NMA evidence is sparse or inconsistent; when threshold values cannot be reliably defined
Inputs
Prediction model for treatment benefit, network meta-analysis results, treatment-specific threshold values (minimum risk difference acceptable for treatment), patient-level or population-level data for model application
Outputs
Net benefit curves comparing personalized and one-size-fits-all treatment strategies across threshold values, graphical display of optimal decision strategies, identification of conditions under which the model adds clinical value
How it works
This extension of decision curve analysis integrates evidence from network meta-analysis (NMA) to evaluate personalized treatment selection models when multiple therapies are available. It calculates the net benefit of strategies—such as 'treat all', 'treat none', or 'treat based on a prediction model'—across a range of treatment threshold values (i.e., the minimum acceptable risk reduction for a given treatment). The net benefit is plotted against threshold combinations to identify the optimal decision strategy. The method was demonstrated using an NMA-based prediction model for relapsing-remitting multiple sclerosis with four treatment options.
- Project
- HTx
- Funding
- Horizon 2020
- Project status
- Completed 2024
- HTA domains
- Clinical Effectiveness
- Technology
- Non-specific
- Assumptions
- Treatment thresholds reflect patient preferences or clinical trade-offs; the prediction model accurately estimates individual treatment effects; NMA provides valid relative effect estimates; linearity of net benefit over threshold ranges
- Strengths
- Incorporates multiple treatment options and evidence from network meta-analysis; allows comparison of personalized versus population-level strategies; accounts for clinical consequences and patient preferences through threshold values; provides visual and quantitative decision support
- Limitations
- Results depend on the quality of the prediction model and NMA; requires definition of clinically meaningful threshold values for each treatment; may not consistently favor personalized models across all thresholds; limited generalizability if model performance varies across subgroups
- Also known as
- Extended Decision Curve Analysis, DCA for Multiple Treatments, NMA-based Decision Curve Analysis
Questions this answers
- › How can we compare different treatments when there are more than two options?
- › Is a personalized treatment model better than giving the same treatment to all patients?
- › How do patient preferences affect which treatment is best?
- › Can we use results from multiple clinical trials to guide individual treatment choices?
- › When does a prediction model actually improve clinical decisions?
- › What happens if small differences in treatment benefit matter in practice?
Related methods
Similar by meaning
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.