Treatment-switching adjustment methods for survival (RPSFTM / IPE / IPCW) — NICE DSU TSD 16/24
These methods help correct survival estimates in clinical trials when some patients switch from the control to the treatment group, which can bias results. They use statistical techniques to account for this crossover and give a clearer picture of a treatment’s true effect.
At a glance
Use when
There is substantial crossover from control to treatment in a randomized trial and the intention-to-treat analysis may underestimate treatment efficacy.
Avoid when
Switching is minimal or random; individual patient data are unavailable; key assumptions (e.g., positivity, ignorability of switching) are clearly violated.
Inputs
Individual patient data from randomized trials, including time-to-event outcomes, treatment assignment, switching times, and covariates for weighting or adjustment.
Outputs
Adjusted estimates of treatment effect on survival (e.g., hazard ratios, survival curves, restricted mean survival time) that account for treatment switching.
How it works
A suite of methods including Rank Preserving Structural Failure Time Models (RPSFTM), Iterative Parameter Estimation (IPE), and Inverse Probability of Censoring Weights (IPCW) to adjust overall survival estimates in the presence of treatment switching in randomized controlled trials. RPSFTM uses counterfactual survival assumptions, IPE iteratively estimates treatment effects under structural models, and IPCW weights patients by the inverse probability of not switching to maintain randomization integrity. These are detailed and evaluated in NICE DSU Technical Support Documents 16 and 24.
- HTA domains
- Clinical Effectiveness, Costs & Economic Evaluation, Patient and Social Aspects
- Assumptions
- RPSFTM assumes no unmeasured confounding for switching and a common treatment effect across subgroups; IPCW assumes correct specification of the censoring model and positivity; all methods assume that the timing and pattern of switching are adequately captured.
- Strengths
- Preserves randomization integrity; allows for unbiased estimation of treatment effects under plausible assumptions; supported by simulation studies and real-world applications; recommended by health technology assessment agencies like NICE.
- Limitations
- Reliant on strong assumptions (e.g., no unmeasured confounding, correct model specification); requires individual patient data; sensitivity to model choice and parameter assumptions; complex to implement and interpret.
- Also known as
- RPSFTM, IPE, IPCW, NICE DSU TSD 16, NICE DSU TSD 24, Treatment switching adjustment methods
Questions this answers
- › What is the true effect of a treatment when patients in the control group switch to the treatment during the trial?
- › How can survival estimates be adjusted for crossover in randomized trials?
- › Which method is most appropriate for adjusting overall survival in the presence of treatment switching?
- › How does treatment switching bias the estimation of long-term treatment effects?
- › Can we reconstruct counterfactual survival outcomes in the absence of switching?
- › How can we preserve the benefits of randomization when crossover is common?
References & sources
Similar by meaning
Beta record. Based on the original catalogue summary; primary-source enrichment pending.