A note on promotion time cure models with a new biological consideration

TL;DR

The paper addresses the need to model survival when intra-tumor heterogeneity drives progression. It introduces the generalized promotion time cure model (GPTCM), which partitions tumor cells into $L$ subtypes with cluster-specific promotion times and proportions $p_l$, yielding the population survival $S_{pop}(t)=e^{- heta F(t)}$ where $F(t)=\sum_{l=1}^L p_l F_l(t)$ and covariates enter through $\theta=\exp(\xi_0+X\xi)$ and $F_l(t)$. The authors derive the formulation, connect it to last-activation and reliability analyses, discuss identifiability and cluster-importance measures, and validate the approach via simulation showing feasible ML estimation and accurate recovery of parameters with larger samples. The GPTCM provides a flexible framework to integrate multi-scale data (clinical, cellular, and molecular) for improved prediction and potential identification of cell-type–specific prognostic drivers, with implications for personalized cancer therapy and reliability engineering contexts where multi-subsystem heterogeneity is relevant.

Abstract

We introduce a generalized promotion time cure model motivated by a new biological consideration. The new approach is flexible to model heterogeneous survival data, in particular for addressing intra-sample heterogeneity. We also indicate that the new approach is suited to model a series or parallel system consisting of multiple subsystems in reliability analysis.

Figures (2)

• Figure 1: Illustration of series system, parallel system, parallel-series system and series-parallel system.
• Figure 2: Survival and hazard plots of the GPTCM with two clusters. Weibull distributions with cluster-specific scale parameters and common shape parameter are as an example: $S_l(t)=e^{-(t/\lambda_l)^\kappa}$, $\lambda_l = \frac{\mu_l}{\Gamma(1+1/\kappa)}$, $l\in \{1,2\}$, $(\log\mu_1,\log\mu_2)=(-0.1, 1)$, $(p_1,p_2)=(0.3, 0.7)$, $\theta=2$, and the Weibull's shape parameter $\kappa=0.5$ (solid line), $\kappa=1$ (dashed line) or $\kappa=3$ (dotted line).