Treated survival time at the same survival proportion

Multiplies the time by which a given proportion of the reference group has had the event by the time ratio to give the corresponding time in the treated group. Applied at the median it gives the treated median; applied at every percentile it traces the whole treated curve.

Signature

t_T = TR * t_C
Inputs
InputsDefinitionUnit
TRTime ratio of treatment to reference from the accelerated failure time modelratio, no unit
t_CTime by which the reference group reaches a stated survival proportion, for example the mediantime, for example months
Output
t_TTime by which the treated group reaches the same survival proportion as the reference group at t_Ctime, same unit as t_C

Function

Accelerated failure time function

Models the logarithm of the time to an event as a linear function of covariates plus a scaled error term whose distribution defines the model, such as Weibull, log-logistic or log-normal. A covariate multiplies time itself, so the survival curve of the treated group is the reference curve stretched or shrunk along the time axis: S(t|x = 1) equals S_0(t / TR), where TR is the time ratio.

Try this function

Implementations

  • Excel

    Treated percentile time from named cells

    Excel multiplies the reference time at the chosen percentile by the time ratio.

    =TimeRatio*ReferenceTime

Assumptions

  • Time ratio holds at the chosen percentile

    The model's constant time ratio implies the same multiplication at every percentile, so S_T(t_T) equals S_C(t_C). Mean survival is multiplied by the same factor; a restricted mean over a fixed horizon is not.

Worked examples

  • Median survival with a time ratio of 1.5

    A reference median of 12 months and a time ratio of 1.5 give a treated median of 18 months, as in the article. Equivalently, S(18 | x = 1) equals S_0(12), which is 0.5.

    TR = 1.5; t_C = 12; t_T = 18

Common errors

  • Scaling the restricted mean or life-years within a horizon

    Life-years accrued within a fixed horizon are an area under the curve truncated at that horizon, so they do not scale by the time ratio. Multiplying modelled life-years by 1.5 overstates the gain.

  • Multiplying survival probabilities by the time ratio

    The time ratio acts on the time axis, not on the probability axis. Multiplying a survival probability of 0.60 by 1.5 gives 0.90, which has no meaning under the model.

Sources

  • Stretching the survival curve along the time axis

    Bradburn MJ, Clark TG, Love SB, Altman DG. Survival analysis part II: multivariate data analysis, an introduction to concepts and methods. British Journal of Cancer. 2003;89(3):431-436. Section on accelerated failure time models (the proportion event-free in one group at t_1 equals that in the other at a constant multiple of t_1).

    View source →

Canonical Identity

Stable URI · Machine-readable · Resolvable · CC BY 4.0