Cranmer et al. comparison of partitioned survival analysis and state-transition multi-state modelling (oncology case study)
This study compares two methods used to estimate the cost-effectiveness of cancer treatments: one splits patient survival into distinct time periods (partitioned survival), and the other models how patients move between health states over time (multi-state model). The results show that the choice of method can greatly affect cost-effectiveness conclusions.
At a glance
Use when
Comparing long-term cost-effectiveness of oncology treatments,When individual patient data on survival and progression are available,When structural uncertainty needs to be assessed
Avoid when
When only short-term outcomes are of interest,When transition data are too sparse for reliable MSM estimation,When model transparency is critical and complexity must be minimized
Inputs
Overall survival (OS) and progression-free survival (PFS) data, cost inputs (treatment, adverse events, hospitalizations), utility values stratified by progression status, parametric survival model forms
Outputs
Incremental cost-effectiveness ratios (ICERs), discounted life years, quality-adjusted life years (QALYs), transition probabilities (for MSM)
How it works
The study compares a partitioned-survival analysis (PartSA) and a semi-Markov multi-state model (MSM) in an oncology cost-effectiveness context. PartSA used parametric survival models (via 'flexsurv') fitted to overall survival (OS) and progression-free survival (PFS) data. MSM (using 'mstate') modeled transitions between 'progression-free', 'post-progression', and 'death' states. Both models included treatment costs, adverse events, hospitalizations, and progression-stratified utilities, with outcomes discounted at 3.5% over 15 years. ICERs were £342,474 (PartSA) and £411,574 (MSM), with MSM showing greater variability in scenario analyses.
- HTA domains
- Clinical Effectiveness, Costs & Economic Evaluation
- Assumptions
- Survival curves can be extrapolated using parametric models; permitted transitions in MSM reflect clinical reality; utility and cost patterns are consistent across time and states; discounting at 3.5% applies to costs and outcomes
- Strengths
- Enables long-term extrapolation from limited trial data; MSM captures more complex disease pathways; PartSA is simpler and more transparent; both allow incorporation of real-world evidence
- Limitations
- Structural uncertainty is rarely explored; results highly sensitive to model choice and parametric assumptions; MSM may produce unstable ICERs with very small QALY gains; both rely on accurate extrapolation beyond observed data
- Also known as
- PartSA, MSM, Partitioned Survival Analysis, Multi-State Model, Semi-Markov Model
Questions this answers
References & sources
Similar by meaning
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.