Dataset Information

BITES: balanced individual treatment effect for survival data.

ABSTRACT:

Motivation

Estimating the effects of interventions on patient outcome is one of the key aspects of personalized medicine. Their inference is often challenged by the fact that the training data comprises only the outcome for the administered treatment, and not for alternative treatments (the so-called counterfactual outcomes). Several methods were suggested for this scenario based on observational data, i.e. data where the intervention was not applied randomly, for both continuous and binary outcome variables. However, patient outcome is often recorded in terms of time-to-event data, comprising right-censored event times if an event does not occur within the observation period. Albeit their enormous importance, time-to-event data are rarely used for treatment optimization. We suggest an approach named BITES (Balanced Individual Treatment Effect for Survival data), which combines a treatment-specific semi-parametric Cox loss with a treatment-balanced deep neural network; i.e. we regularize differences between treated and non-treated patients using Integral Probability Metrics (IPM).

Results

We show in simulation studies that this approach outperforms the state of the art. Furthermore, we demonstrate in an application to a cohort of breast cancer patients that hormone treatment can be optimized based on six routine parameters. We successfully validated this finding in an independent cohort.

Availability and implementation

We provide BITES as an easy-to-use python implementation including scheduled hyper-parameter optimization (https://github.com/sschrod/BITES). The data underlying this article are available in the CRAN repository at https://rdrr.io/cran/survival/man/gbsg.html and https://rdrr.io/cran/survival/man/rotterdam.html.

Supplementary information

Supplementary data are available at Bioinformatics online.

SUBMITTER: Schrod S

PROVIDER: S-EPMC9235492 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BackgroundPerforming well-powered randomised controlled trials (RCTs) of new treatments for rare diseases is often infeasible. However, with the increasing availability of historical data, incorporating existing information into trials with small sample sizes is appealing in order to increase the power. Bayesian approaches enable one to incorporate historical data into a trial's analysis through a prior distribution.MethodsMotivated by a RCT intended to evaluate the impact on event-free survival of mifamurtide in patients with osteosarcoma, we performed a simulation study to evaluate the impact on trial operating characteristics of incorporating historical individual control data and aggregate treatment effect estimates. We used power priors derived from historical individual control data for baseline parameters of Weibull and piecewise exponential models, while we used a mixture prior to summarise aggregate information obtained on the relative treatment effect. The impact of prior-data conflicts, both with respect to the parameters and survival models, was evaluated for a set of pre-specified weights assigned to the historical information in the prior distributions.ResultsThe operating characteristics varied according to the weights assigned to each source of historical information, the variance of the informative and vague component of the mixture prior and the level of commensurability between the historical and new data. When historical and new controls follow different survival distributions, we did not observe any advantage of choosing a piecewise exponential model compared to a Weibull model for the new trial analysis. However, we think that it remains appealing given the uncertainty that will often surround the shape of the survival distribution of the new data.ConclusionIn the setting of Sarcome-13 trial, and other similar studies in rare diseases, the gains in power and accuracy made possible by incorporating different types of historical information commensurate with the new trial data have to be balanced against the risk of biased estimates and a possible loss in power if data are not commensurate. The weights allocated to the historical data have to be carefully chosen based on this trade-off. Further simulation studies investigating methods for incorporating historical data are required to generalise the findings.

Project description:BackgroundRisk of the outcome is a mathematical determinant of the absolute treatment benefit of an intervention, yet this can vary substantially within a trial population, complicating the interpretation of trial results.MethodsWe developed risk models using Cox or logistic regression on a set of large publicly available randomized controlled trials (RCTs). We evaluated risk heterogeneity using the extreme quartile risk ratio (EQRR, the ratio of outcome rates in the lowest risk quartile to that in the highest) and skewness using the median to mean risk ratio (MMRR, the ratio of risk in the median risk patient to the average). We also examined heterogeneity of treatment effects (HTE) across risk strata.ResultsWe describe 39 analyses using data from 32 large trials, with event rates across studies ranging from 3% to 63% (median?=?15%, 25th-75th percentile?=?9-29%). C-statistics of risk models ranged from 0.59 to 0.89 (median?=?0.70, 25th-75th percentile?=?0.65-0.71). The EQRR ranged from 1.8 to 50.7 (median?=?4.3, 25th-75th percentile?=?3.0-6.1). The MMRR ranged from 0.4 to 1.0 (median?=?0.86, 25th-75th percentile?=?0.80-0.92). EQRRs were predictably higher and MMRRs predictably lower as the c-statistic increased or the overall outcome incidence decreased. Among 18 comparisons with a significant overall treatment effect, there was a significant interaction between treatment and baseline risk on the proportional scale in only one. The difference in the absolute risk reduction between extreme risk quartiles ranged from -3.2 to 28.3% (median?=?5.1%; 25th-75th percentile?=?0.3-10.9).ConclusionsThere is typically substantial variation in outcome risk in clinical trials, commonly leading to clinically significant differences in absolute treatment effects Most patients have outcome risks lower than the trial average reflected in the summary result. Risk-stratified trial analyses are feasible and may be clinically informative, particularly when the outcome is predictable and uncommon.

Project description:ObjectiveTo examine risk of malignancy and death in patients with kidney transplant who receive the immunosuppressive drug sirolimus.DesignSystematic review and meta-analysis of individual patient data.Data sourcesMedline, Embase, and the Cochrane Central Register of Controlled Trials from inception to March 2013.EligibilityRandomized controlled trials comparing immunosuppressive regimens with and without sirolimus in recipients of kidney or combined pancreatic and renal transplant for which the author was willing to provide individual patient level data. Two reviewers independently screened titles/abstracts and full text reports of potentially eligible trials to identify studies for inclusion. All eligible trials reported data on malignancy or survival.ResultsThe search yielded 2365 unique citations. Patient level data were available from 5876 patients from 21 randomized trials. Sirolimus was associated with a 40% reduction in the risk of malignancy (adjusted hazard ratio 0.60, 95% confidence interval 0.39 to 0.93) and a 56% reduction in the risk of non-melanoma skin cancer (0.44, 0.30 to 0.63) compared with controls. The most pronounced effect was seen in patients who converted to sirolimus from an established immunosuppressive regimen, resulting in a reduction in risk of malignancy (0.34, 0.28 to 0.41), non-melanoma skin cancer (0.32, 0.24 to 0.42), and other cancers (0.52, 0.38 to 0.69). Sirolimus was associated with an increased risk of death (1.43, 1.21 to 1.71) compared with controls.ConclusionsSirolimus was associated with a reduction in the risk of malignancy and non-melanoma skin cancer in transplant recipients. The benefit was most pronounced in patients who converted from an established immunosuppressive regimen to sirolimus. Given the risk of mortality, however, the use of this drug does not seem warranted for most patients with kidney transplant. Further research is needed to determine if different populations, such as those at high risk of cancer, might benefit from sirolimus.

Project description:We recently developed a new method random-intercept accelerated failure time model with Bayesian additive regression trees (riAFT-BART) to draw causal inferences about population treatment effect on patient survival from clustered and censored survival data while accounting for the multilevel data structure. The practical utility of this method goes beyond the estimation of population average treatment effect. In this work, we exposit how riAFT-BART can be used to solve two important statistical questions with clustered survival data: estimating the treatment effect heterogeneity and variable selection. Leveraging the likelihood-based machine learning, we describe a way in which we can draw posterior samples of the individual survival treatment effect from riAFT-BART model runs, and use the drawn posterior samples to perform an exploratory treatment effect heterogeneity analysis to identify subpopulations who may experience differential treatment effects than population average effects. There is sparse literature on methods for variable selection among clustered and censored survival data, particularly ones using flexible modeling techniques. We propose a permutation-based approach using the predictor's variable inclusion proportion supplied by the riAFT-BART model for variable selection. To address the missing data issue frequently encountered in health databases, we propose a strategy to combine bootstrap imputation and riAFT-BART for variable selection among incomplete clustered survival data. We conduct an expansive simulation study to examine the practical operating characteristics of our proposed methods, and provide empirical evidence that our proposed methods perform better than several existing methods across a wide range of data scenarios. Finally, we demonstrate the methods via a case study of predictors for in-hospital mortality among severe COVID-19 patients and estimating the heterogeneous treatment effects of three COVID-specific medications. The methods developed in this work are readily available in the R${\textsf {R}}$ package riAFTBART$\textsf {riAFTBART}$ .