Accelerated Failure Time Research Articles

Rank-based sequential feature selection for high-dimensional accelerated failure time models with main and interaction effects

High-dimensional accelerated failure time (AFT) models are commonly used regression models in survival analysis. Feature selection problem in high-dimensional AFT models is addressed, considering scenarios involving solely main effects or encompassing both main and interaction effects. A rank-based sequential feature selection (RankSFS) method is proposed, the selection consistency is established and illustrated by comparing it with existing methods through extensive numerical simulations. The results show that RankSFS achieves a higher Positive Discovery Rate (PDR) and lower False Discovery Rate (FDR). Additionally, RankSFS is applied to analyze the data on Breast Cancer Relapse. With a remarkable short computational time, RankSFS successfully identifies two crucial genes.

Association between blood pressure and new onset of chronic kidney disease in non-diabetic Japanese adults: A population-based longitudinal study from 1998 to 2023

Background and aimsHypertension is a risk factor for developing chronic kidney disease (CKD). Studies of adult participants in the USA reported that hypertension increased the risk of developing CKD even in the non-diabetic population. However, studies in non-diabetic populations are limited and additional studies in other races are required. This study aimed to examine the relationship between hypertension and the development of CKD in non-diabetic Asian adults. Methods and resultsIn this longitudinal study, non-diabetic Japanese adults who took annual checkups from 1998 to 2023 were included. CKD was defined as <60 mL/min/1.73 m2, and hypertension was classified into four levels according to the guidelines of the American College of Cardiology/American Heart Association. The Weibull accelerated failure time model was selected because the proportional hazards assumption was violated. Of the 7363 (men: 40.3%) people in the final cohort, 2498 (men: 40.1%) developed CKD after a mean follow-up of 7.99 years. Elevated blood pressure (BP) and hypertension stage 2 had a 9% (95% confidence interval [CI]: 1%–16%) and 11% (95% CI: 5%–17%) shorter survival time to CKD onset, respectively, than normal BP. Hypertension stage 1 also had a shorter survival to CKD onset by point estimate, but all 95% CIs crossed 1 in all models. ConclusionsIn a relatively healthy Asian population without diabetes, controlling BP to an appropriate range reduces the risk of developing CKD.

Impact of countdown timer on drivers’ anticipation at the onset of yellow at signalized intersections

ABSTRACT Countdown timers at traffic signals display the time remaining for a phase to terminate, influencing driver behaviour. Studies on their effects at the onset of yellow have generally focused on drivers’ stop/go decisions and dilemma zone, without explicitly addressing how they help drivers anticipate the signal change and make ‘informed decisions’. Therefore, this study first analysed the impact of countdown timer on drivers’ anticipation by comparing brake initiation, deceleration, and speed in the absence and presence of countdown timer. Next, driver behaviour was studied using ‘Accelerated Failure Time’ models. A driving simulator experiment was designed for data collection, with demographic and additional driving behaviour data gathered through a questionnaire. When countdown timer was present, drivers anticipated the onset of yellow, resulting in significant variations in driver behaviour. These variations, in turn, influenced the duration of braking and the time taken to cross the stop line from the onset of yellow.

CureAuxSP: An R package for estimating mixture cure models with auxiliary survival probabilities

Background and Objective:There is a rising interest in exploiting aggregate information from external medical studies to enhance the statistical analysis of a modestly sized internal dataset. Currently available software packages for analyzing survival data with a cure fraction ignore the potentially available auxiliary information. This paper aims at filling this gap by developing a new R package CureAuxSP that can include subgroup survival probabilities extracted outside into an interested internal survival dataset. Methods:The newly developed R package CureAuxSP provides an efficient approach for information synthesis under the mixture cure models, including Cox proportional hazards mixture cure model and the accelerated failure time mixture cure model as special cases. It focuses on synthesizing subgroup survival probabilities at multiple time points and the underlying method development lies in the control variate technique. Evaluation of homogeneity assumption based on a test statistic can be automatically carried out by our package and if heterogeneity does exist, the original outputs can be further refined adaptively. Results:The R package CureAuxSP provides a main function SMC.AxuSP() that helps us adaptively incorporate external subgroup survival probabilities into the analysis of an internal survival data. We also provide another function Print.SMC.AuxSP() for printing the results with a better presentation. Detailed usages are described, and implementations are illustrated with numerical examples, including a simulated dataset with a well-designed data generating process and a real breast cancer dataset. Substantial efficiency gain can be observed by our results. Conclusions:Our R package CureAuxSP can make the wide applications of utilizing auxiliary information possible. It is anticipated that the performance of mixture cure models can be improved for the survival data with a cure fraction, especially for those with small sample sizes.

Modelling Time to Recovery from Multidrug Resistant Tuberculosis in Southern Ethiopia

Introduction: Multidrug resistant tuberculosis (MDR-TB) is a global heath threat, resistant to key anti-TB drugs. It is ranked among the top 10 causes of death worldwide. Therefore, the current study investigated time to recovery from MDR-TB in southern Ethiopia. Data, Materials and Methods: Restrospective data from selected hospitals in SNNPR (January 2016 to December 2021) were analyzed. A cluster sample of 301 MDR-TB patients (131 NEMMCH, 121 BH, 49 AGH) was considered. Results: Among the 301 cases, 116 (38.5%) were censored. While 185 (61.5%) were recovered. Parametric shared frailty models were employed to account unobserved heterogeneity among the Hospitals and patients and AFT models were employed. the median recovery time of MDR-TB is 22 months. The clustering effect of frailty model was hospitals. Weibull-gamma shared frailty model was appropriate for this data. Conclusion: The final model showed that males have higher recovery rates than females. Extra pulmonary MDR-TB and Urban residency correleted with longer recovery times. The recovery rate increases with increasing baseline weight, education level, and occupation. But, the recovery rate decreases with smoking, co-morbidities, previous drug history, history of TB, and alcohol use Recommendation: All concerned bodies should be cognizant on the risk factors of MDR-TB in SNNP region By providing on early case detection and appropriate treatment of drug-susceptible MDR-TB, since it is essential to shorten the recovery time of MDR-TB patients in line with WHO guidelines

The impact of maternal smoking during pregnancy and the age of smoking initiation on incident dementia: A prospective cohort study.

The impact of early-life tobacco exposure on dementia has remained unknown. Using the UK Biobank, the associations of maternal smoking during pregnancy (MSDP) and age of smoking initiation (ASI) with the onset time of all-cause dementia were estimated with accelerated failure time models. The effects of MSDP and ASI on brain structure and their genetic correlation to Alzheimer's disease (AD) were analyzed. A Mendelian randomization (MR) analysis was conducted. The time ratios for smokers starting in childhood, adolescence, and adulthood (vs never smokers) were 0.87 (0.76 to 0.99), 0.92 (0.88 to 0.96), and 0.95 (0.89 to 1.01). MSDP and smoking in adolescence altered many brain regions, including the hippocampus. In genetic analysis, MSDP was genetically and causally linked to AD, and a younger ASI was genetically correlated to a higher AD risk. Early-life smoking accelerated dementia onset and was genetically correlated to AD. MSDP demonstrated genetic and causal linkage to AD risks. Unlike the commonly used Cox proportional hazards model, this article uses a parametric survival analysis method - the accelerated failure model - to explore the relationship between exposure to onset time. It can be used as an alternative method when the proportional hazards assumption is not met. Genetic analyses including genetic correlation study and MR analysis and brain structure analyses were conducted to support our findings and explore the potential mechanisms. The study reveals the relationship between different smoking initiation periods and the onset time of dementia and shows that earlier smoking exposure has a more significant impact on dementia. It emphasizes the importance of preventing early smoking. In the future, more research focusing on the relationship between early exposure and dementia is called for to provide more detailed prevention measures for dementia that cover all age groups.

A general model-checking procedure for semiparametric accelerated failure time models

A general model-checking procedure for semiparametric accelerated failure time models

Accelerated failure time frailty model for modeling multiple systems subject to minimal repair

AbstractThis article presents accelerated failure time models with and without frailty for modeling multiple systems subject to minimal repair. The study considers the conventional accelerated failure time model, the accelerated failure time model with Gamma frailty, and proposes the accelerated failure time model with weighted Lindley frailty, which has attractive properties such as a closed‐form Laplace transform. The proposed model extends the accelerated failure time model with the intensity function of a power law process. It retains the direct physical interpretation of the original accelerated failure time model, in which the role of covariates is to accelerate or decelerate the time to each repair. This framework includes parametric approaches to model fitting, which we consider for estimating the vector of regression parameters under this model and the parameter in the baseline intensity functions. The methodology is illustrated with a simulation study and a toy example to demonstrate the applicability of these models in the industrial context.

Social and spatial disparities in individuals’ mobility response time to COVID-19: A big data analysis incorporating changepoint detection and accelerated failure time models

Although an increasing number of studies have investigated the lag time between the outbreak of COVID-19 and behavior change, few have accurately measured response times to the epidemic at the individual scale as well as their social and spatial heterogeneities. Using a large-scale, long time series dataset of individual-level mobile phone trajectories from Shenzhen, China, we compared six changepoint detection (CPD) algorithms in terms of their performance in detecting true changepoints (CPs) in time series data of individuals’ daily travel distances. We found that the kernel-based CPD method outperformed other algorithms. We thus adopted this method to calculate Shenzhen residents’ mobility response times to the outbreak of COVID-19 and further used an accelerated failure time (AFT) model to explore factors affecting response times. The results suggest that the average and median mobility response times to the outbreak in Shenzhen were 4.64 days and 4 days, respectively. Males and the elderly responded more slowly to the outbreak, while responses were faster among residents in neighborhoods with a higher percentage of highly educated, married, or employed individuals; with better regional accessibility to the city center, railway stations, or the airport; and with higher residential density. These findings can assist policymakers in determining the policy timeline and re-assessing the effectiveness and equity impact of mobility restriction policies and designing more responsive policies varying by social groups and built environment features, helping build socially-resilient neighborhoods in the post-COVID era.

Urban Unemployment Duration Analysis in Post-Reform China

We link China’s accession to the WTO with the unemployment duration of urban jobseekers. Using data from the Rural Urban Migration in China (RUMiC) of a survey conducted in early 2008, we construct an inflow sample of those who started a period of joblessness between January 2002 and December 2007 to estimate the unemployment duration for local urban jobseekers. Deploying the Cox Proportional Hazards and Accelerated Failure Time models, we find reduced trade policy uncertainty significantly shortens unemployment duration, with hazard-rate varying across internal in-migrants and urban natives, due to variations in job accessibility and reservation wages.

Comparative Analysis of Women's Breast Cancer Survival Time at Three Selected Government Referral Hospitals in Ethiopia's Amhara Region Using Parametric Shared Frailty Models.

One in five people will eventually develop cancer, and one in eleven women will lose their lives to the disease. The main aim of this study is to determinants of survival time of women with breast cancer using appropriate Frailty models. A study involving 632 Ethiopian women with breast cancer was conducted between 2018 and 2020, utilizing medical records from Felege-Hiwot Referral Hospital, the University of Gondar, and Dessie Referral Hospital. To compare survival, the Kaplan-Meier plot (s) and Log rank test were employed; to assess mean survival, one-way analysis of variance and the t test were utilized. The factors influencing women's survival times from breast cancer were identified using the parametric shared frailty model and the accelerated failure time model. The median time to die for breast cancer patients treated at FHRH, UoGCSH, and DRH was 14.91 months, 11.14 months, and 12.32 months, respectively. The parametric model of shared frailty fit those who were statistically significant in univariate analysis. The results showed that survival of women with breast cancer was significantly influenced by age, tumor size, comorbidity, nodal status, stage, histologic grade, and type of primary treatment initiated. When comparing mean survival times between hospitals, the results showed a significant difference; patients who were treated in FHRH live significantly longer than patients treated in UoGCSH and DRH, whereas patients treated in UoGCSH have comparatively lower survival. Women with stage IV and comorbidities have 22.4% and 27.1% shorter expected survival, respectively. This finding suggests that improving the availability and accessibility of radiation therapy and surgery, eliminating disparities between hospitals, raising awareness of early signs and symptoms of breast cancer and encouraging women to seek clinical help, and highlighting women with comorbidities at diagnosis are important ways to increase survival time.

Modeling the impact of driving aggression on lane change performance Measures: Steering compensatory Behavior, lane change execution duration and crash probability

The current study aimed to investigate the influence of driving aggression on lane change behavior. Drivers belonging to three levels of aggressive driving styles (aggressive, moderately aggressive, and non-aggressive) participated in lane change experiments, and 232 lane change maneuvers were recorded. Steering behavior during lane change maneuvers was investigated using three steering performance measures, including the standard deviations of steering angle, steering reversal rates, and maximum steering swerve. Three separate generalized estimating equation models were developed to analyze the effects of aggressive driving behavior on the standard deviation of steering angle, steering reversal rates, and maximum steering swerve. A Weibull accelerated failure time model with clustered heterogeneity was developed to analyze the effect of aggressive driving behavior on lane change execution duration. Additionally, to model the crash probability during lane change, a random effects logistic regression model was developed. The findings indicate that aggressive and moderately aggressive drivers completed lane changes 59% and 38% faster, respectively, compared to non-aggressive drivers. This implies a propensity for aggressive and moderately aggressive drivers to execute lane changes more swiftly than non-aggressive drivers. Aggressive drivers increased their standard deviation of steering angle, steering reversal rates, and maximum steering swerves by 180%, 60%, and 62%, respectively, compared to non-aggressive drivers. This implies that aggressive drivers engaged in extreme steering behavior as a compensatory measure for their reckless driving in order to avoid a collision during lane change maneuvering. However, despite their vigorous compensatory steering behavior, the crash rate (number of crashes observed) increased by 5.10 and 2.83 times for aggressive and moderately aggressive drivers, respectively, compared to non-aggressive drivers. The findings have implications for developing driving assistance systems and lane change simulation models.

Leveraging machine learning: Covariate-adjusted Bayesian adaptive randomization and subgroup discovery in multi-arm survival trials

Clinical trials evaluate the safety and efficacy of treatments for specific diseases. Ensuring these studies are well-powered is crucial for identifying superior treatments. With the rise of personalized medicine, treatment efficacy may vary based on biomarker profiles. However, researchers often lack prior knowledge about which biomarkers are linked to varied treatment effects. Fixed or response-adaptive designs may not sufficiently account for heterogeneous patient characteristics, such as genetic diversity, potentially reducing the chance of selecting the optimal treatment for individuals. Recent advances in Bayesian nonparametric modeling pave the way for innovative trial designs that not only maintain robust power but also offer the flexibility to identify subgroups deriving greater benefits from specific treatments. Building on this inspiration, we introduce a Bayesian adaptive design for multi-arm trials focusing on time-to-event endpoints. We introduce a covariate-adjusted response adaptive randomization, updating treatment allocation probabilities grounded on causal effect estimates using a random intercept accelerated failure time BART model. After the trial concludes, we suggest employing a multi-response decision tree to pinpoint subgroups with varying treatment impacts. The performance of our design is then assessed via comprehensive simulations.

Disparities in Mortality by Sexual Orientation in a Large, Prospective Cohort of Female Nurses

Extensive evidence documents health disparities for lesbian, gay, and bisexual (LGB) women, including worse physical, mental, and behavioral health than heterosexual women. These factors have been linked to premature mortality, yet few studies have investigated premature mortality disparities among LGB women and whether they differ by lesbian or bisexual identity. To examine differences in mortality by sexual orientation. This prospective cohort study examined differences in time to mortality across sexual orientation, adjusting for birth cohort. Participants were female nurses born between 1945 and 1964, initially recruited in the US in 1989 for the Nurses' Health Study II, and followed up through April 2022. Sexual orientation (lesbian, bisexual, or heterosexual) assessed in 1995. Time to all-cause mortality from assessment of exposure analyzed using accelerated failure time models. Among 116 149 eligible participants, 90 833 (78%) had valid sexual orientation data. Of these 90 833 participants, 89 821 (98.9%) identified as heterosexual, 694 (0.8%) identified as lesbian, and 318 (0.4%) identified as bisexual. Of the 4227 deaths reported, the majority were among heterosexual participants (n = 4146; cumulative mortality of 4.6%), followed by lesbian participants (n = 49; cumulative mortality of 7.0%) and bisexual participants (n = 32; cumulative mortality of 10.1%). Compared with heterosexual participants, LGB participants had earlier mortality (adjusted acceleration factor, 0.74 [95% CI, 0.64-0.84]). These differences were greatest among bisexual participants (adjusted acceleration factor, 0.63 [95% CI, 0.51-0.78]) followed by lesbian participants (adjusted acceleration factor, 0.80 [95% CI, 0.68-0.95]). In an otherwise largely homogeneous sample of female nurses, participants identifying as lesbian or bisexual had markedly earlier mortality during the study period compared with heterosexual women. These differences in mortality timing highlight the urgency of addressing modifiable risks and upstream social forces that propagate and perpetuate disparities.

Evaluation of the European Society of Cardiology Risk Assessment Score in Incident Systemic Sclerosis-Associated Pulmonary Arterial Hypertension.

Patients with pulmonary arterial hypertension (PAH) may be stratified as low, intermediate, or high risk of 1-year mortality. In 2022, the European Society of Cardiology (ESC) updated and simplified its risk stratification tool, based on three variables: World Health Organization functional class, serum N-terminal pro-brain type natriuretic peptide and six-minute walk distance, applied at follow-up visits, intended to guide therapy over time. We applied the 2022 ESC risk assessment tool at baseline and follow-up (within 2 years) to a multinational incident cohort of systemic sclerosis-associated PAH (SSc-PAH). Kaplan-Meier curves, Cox hazards regression, and accelerated failure time models were used to evaluate survival by risk score. At baseline (n = 260), the majority of SSc-PAH (72.2%) were graded as intermediate risk of death according to the 2022 tool. At follow-up, according to 2022 tool, half (55.5%) of the cohort were classified as low or intermediate-low risk. The 2022 risk model at follow-up was able to differentiate survival between risk strata. All three individual parameters (World Health Organization functional class, N-terminal pro-brain type natriuretic peptide, six-minute walk distance) were significantly associated with mortality at baseline and/or follow-up. The 2022 ESC risk assessment strategy applied at baseline and follow-up predicts survival in SSc-PAH. Treatment decisions for SSc-PAH should include risk assessments, aiming to achieve low-risk status according to the 2022 ESC guidelines.

Time to initiation of antenatal care and its predictors among pregnant women in Ethiopia: a multilevel mixed-effects acceleration failure time model

ObjectiveTo assess the time to initiation of antenatal care (ANC) and its predictors among pregnant women in Ethiopia.DesignRetrospective follow-up study using secondary data from the 2019 Ethiopian Mini-Demographic and Health...

The impact of positive and negative psychological affect and overconfidence from major family events on new venture survival

This paper investigates how family events interacting with entrepreneurs’ psychological affect and overconfidence impact new venture viability. We use panel data from the Australian Household, Income and Labor Dynamics survey, focusing on family event-induced psychological affect entrepreneurs experience as a predictor of new venture survival. Our accelerated failure time model shows that although negative family events interact with entrepreneur overconfidence to spur cautious behaviour, positive events interacting with overconfidence have the biggest impact (negative) on new ventures. The study enhances our understanding of the embeddedness of family in the entrepreneurial process and challenges past research by revealing how positive family events can have a greater negative impact on new venture survival than negative ones.

Bayesian extension of the Weibull AFT shared frailty model with generalized family of distributions for enhanced survival analysis using censored data

ABSTRACT In survival analysis, the Accelerated Failure Time (AFT) shared frailty model is a widely used framework for analyzing time-to-event data while accounting for unobserved heterogeneity among individuals. This paper extends the traditional Weibull AFT shared frailty model using half logistic-G family of distributions (Type I, Type II and Type II exponentiated) through Bayesian methods. This approach offers flexibility in capturing covariate influence and handling heavy-tailed frailty distributions. Bayesian inference with MCMC provides parameter estimates and credible intervals. Simulation studies show improved model predictive performance compared to existing models, and real-world applications demonstrate its practical utility. In summary, our Bayesian Weibull AFT shared frailty model with Type I, Type II and Type II exponentiated half logistic-G family distributions enhances time-to-event data analysis, making it a versatile tool for survival analysis in various fields using STAN in R.

Longitudinal Patterns of Multimorbidity in Gulf War Era Veterans With and Without Gulf War Illness.

Objectives: To examine whether severe Gulf War illness (SGWI) case status was associated with longitudinal multimorbidity patterns. Methods: Participants were users of the Veteran Health Administration Health Care System drawn from the Gulf War Era Cohort and Biorepository (n = 840). Longitudinal measures of multimorbidity were constructed using (1) electronic health records (Charlson Comorbidity Index; Elixhauser; and Veterans Affairs Frailty Index) from 10/1/1999 to 6/30/2023 and (2) self-reported medical conditions (Deficit Accumulation Index) since the war until the survey date. Accelerated failure time models examined SGWI case status as a predictor of time until threshold level of multimorbidity was reached, adjusted for age and sociodemographic and military characteristics. Results: Models, adjusted for covariates, revealed that (1) relative to the SWGI- group, the SGWI+ group was associated with an accelerated time for reaching each threshold and (2) the relationship between SGWI and each threshold was not moderated by age. Discussion: Findings suggest that veterans with SGWI experienced accelerated aging.

Segmentation of patients with small cell lung cancer into responders and non-responders using the optimal cross-validation technique

BackgroundThe timing of treating cancer patients is an essential factor in the efficacy of treatment. So, patients who will not respond to current therapy should receive a different treatment as early as possible. Machine learning models can be built to classify responders and nonresponders. Such classification models predict the probability of a patient being a responder. Most methods use a probability threshold of 0.5 to convert the probabilities into binary group membership. However, the cutoff of 0.5 is not always the optimal choice.MethodsIn this study, we propose a novel data-driven approach to select a better cutoff value based on the optimal cross-validation technique. To illustrate our novel method, we applied it to three clinical trial datasets of small-cell lung cancer patients. We used two different datasets to build a scoring system to segment patients. Then the models were applied to segment patients into the test data.ResultsWe found that, in test data, the predicted responders and non-responders had significantly different long-term survival outcomes. Our proposed novel method segments patients better than the standard approach using a cutoff of 0.5. Comparing clinical outcomes of responders versus non-responders, our novel method had a p-value of 0.009 with a hazard ratio of 0.668 for grouping patients using the Cox proportion hazard model and a p-value of 0.011 using the accelerated failure time model which approved a significant difference between responders and non-responders. In contrast, the standard approach had a p-value of 0.194 with a hazard ratio of 0.823 using the Cox proportion hazard model and a p-value of 0.240 using the accelerated failure time model indicating the responders and non-responders do not differ significantly in survival.ConclusionIn summary, our novel prediction method can successfully segment new patients into responders and non-responders. Clinicians can use our prediction to decide if a patient should receive a different treatment or stay with the current treatment.