Cure models
For therapies where a subgroup of patients appears to achieve long-term remission, extrapolating a single parametric curve to a lifetime horizon may not capture a plateau in survival. EasyHTA’s Partitioned survival with cure model template (PSM-cure) addresses this by building a composite curve in three phases: the observed KM data, a parametric extrapolation, and general-population mortality adjusted by a standardised mortality ratio.
The cure fraction
The three phases
The curve follows the trial data over the observed period, a parametric extrapolation after that, and general-population mortality once the cohort is treated as cured. You control the two points where it changes over:
- The switch point, where the curve stops following the observed data and starts following your selected parametric distribution.
- The cure point, where it stops following that distribution and follows population mortality instead, adjusted by the standardised mortality ratio (SMR).
The switch point cannot fall after the cure point, and neither can fall after the time horizon. Overall survival and progression-free survival are fitted and extrapolated separately, so each endpoint, on each arm, has its own switch point and cure point.
The composite curve itself, and the survival it implies at the cure point, are visible on the curve-fitting screen as soon as a switch point, cure point, starting age and SMR are entered, alongside the parametric fit. The Excel export carries the same curve in its own sheet.
The app checks the join at both points and flags anything that looks off: a visible gap between the observed and parametric curves at the switch point, and a parametric hazard that has fallen below SMR-adjusted population mortality by the cure point, which would mean the model already implies better-than-general-population mortality before the cure assumption even applies.
Population mortality and the SMR
After the cure point, survival follows age-specific mortality from a national life table, multiplied by the SMR. An SMR of 1.0 asserts that surviving patients face exactly general-population mortality. Values above 1 carry a persistent excess risk relative to the general population.
Set the cohort’s mean age at model entry so the life table is read from the right age, and choose the life table for your reference country. The table used is recorded with the analysis, so a re-run reproduces the same result. The cohort can optionally be split by sex.
The optional Background mortality floor prevents the extrapolated phase from declining more slowly than SMR-adjusted general-population mortality, which can otherwise happen when a well-fitting distribution has a very flat tail. It is off by default, and turning it on is worth recording in your assumptions log.
Testing the cure assumption
On the PSM-cure template, the one-way DSA includes the cure point, the switch point, and the SMR alongside costs, utilities, and the discount rate. Each is swept across a range around your base case, moving both arms and both endpoints together, and the tornado notes when the ICER doesn’t respond in a straight line across that range. The PSA samples costs, utilities, and the fitted survival-curve parameters, but holds the cure point and the SMR fixed at their base-case values and states this in its output. The distribution and the time horizon stay fixed in both: test either by re-running the model with the alternative choice.
The Sensitivity tab also has a Cure point and SMR grid, which moves the two together and shows the incremental net monetary benefit of every combination at your willingness-to-pay threshold.
When to use it
A cure assumption needs external support. Before using this template, consider whether there is a plateau in the observed KM curve that persists well beyond the point where events cluster, whether long-term registry or follow-up evidence supports sustained remission in this indication, and whether the cure point you have chosen falls where that evidence is strongest. NICE DSU TSD21 discusses cure and other survival models that TSD14 does not cover, and is the appropriate reference when a standard parametric model is not adequate.
If none of that support exists, Partitioned survival, the base template with no cure point and no SMR (though it can still apply the optional Background mortality cap), is the more defensible starting point. See Parametric distributions for the fitting step that precedes this one, and Quickstart for the end-to-end workflow.
References
Latimer NR (2011, last updated 2013). NICE DSU Technical Support Document 14: Undertaking survival analysis for economic evaluations alongside clinical trials, extrapolation with patient-level data. Sheffield: NICE DSU. sheffield.ac.uk/nice-dsu
Rutherford MJ, Lambert PC, Sweeting MJ, Pennington B, Crowther MJ, Abrams KR, Latimer NR (2020). NICE DSU Technical Support Document 21: Flexible methods for survival analysis. Sheffield: NICE DSU. sheffield.ac.uk/nice-dsu
LinkedIn