A flexible parametric accelerated failure time model and the extension to time-dependent acceleration factors

Principles of artificial intelligence in radiooncology

PURPOSE

In the rapidly expanding field of artificial intelligence (AI) there is a wealth of literature detailing the myriad applications of AI, particularly in the realm of deep learning. However, a review that elucidates the technical principles of deep learning as relevant to radiation oncology in an easily understandable manner is still notably lacking. This paper aims to fill this gap by providing a comprehensive guide to the principles of deep learning that is specifically tailored toward radiation oncology.

METHODS

In light of the extensive variety of AI methodologies, this review selectively concentrates on the specific domain of deep learning. It emphasizes the principal categories of deep learning models and delineates the methodologies for training these models effectively.

RESULTS

This review initially delineates the distinctions between AI and deep learning as well as between supervised and unsupervised learning. Subsequently, it elucidates the fundamental principles of major deep learning models, encompassing multilayer perceptrons (MLPs), convolutional neural networks (CNNs), recurrent neural networks (RNNs), transformers, generative adversarial networks (GANs), diffusion-based generative models, and reinforcement learning. For each category, it presents representative networks alongside their specific applications in radiation oncology. Moreover, the review outlines critical factors essential for training deep learning models, such as data preprocessing, loss functions, optimizers, and other pivotal training parameters including learning rate and batch size.

CONCLUSION

This review provides a comprehensive overview of deep learning principles tailored toward radiation oncology. It aims to enhance the understanding of AI-based research and software applications, thereby bridging the gap between complex technological concepts and clinical practice in radiation oncology.

Mortality risk associated to arsenic exposure after a major disaster. Results from the Manfredonia occupational cohort study 1976-2021

BACKGROUND

On September 1976, due to the explosion of an ammonia-washing column at the petrochemical plant in Manfredonia (Italy), 39 tonnes of arsenic were released into the atmosphere, contaminating the plants and the neighbourhoods close to it. The aim of this study is to present the results of a 45-year follow up of exposed workers with a focus on residential exposure.

METHODS

We contacted Italian General Registries Offices and updated the vital status of persons involved in the clean-up activities following the disaster. The outcome of interest was the overall and cause-specific mortality. An accelerated failure time (AFT) approach was used when appropriate to model the risk of mortality.

RESULTS

1772 workers contributing 67,743 person years were considered in the analysis. For overall mortality, results of the age-adjusted AFT model show an accelerator factor of 0.89 (95%CI 0.80-0.99) among contract workers, which means a shortening of survival in comparison to the reference category (plastic area workers). When accounting for latency greater than 20 years, higher mortality rates for lung cancer were observed among workers residing in Manfredonia.

DISCUSSION

An increased risk of mortality among workers who were more exposed to arsenic during the clean-up activities has been observed. In fact, a loss of 5 years of life among more exposed workers was calculated. Furthermore, the mortality rates of residents in Manfredonia were higher than those of workers residing elsewhere.

INFLA score: a novel inflammatory marker for assessing cardiometabolic disease risk in obese individuals

BACKGROUND

The low-grade inflammation score (INFLA-score) is a composite index that assesses chronic inflammatory status using multiple inflammatory markers. However, its correlation with cardiometabolic diseases (CMDs) in obese populations remains unclear.

METHODS

We conducted a prospective cohort study involving 79,160 participants with obesity (BMI ≥ 30 kg/m2) from the UK Biobank. The INFLA-score was calculated based on high-sensitivity C-reactive protein, leukocyte count, platelet count and granulocyte/lymphocyte ratio. We employed Kaplan-Meier survival curves, multivariable Cox regression, restricted cubic splines and accelerated time-to-failure models to analyse the association between the INFLA-score and CMDs risk, including coronary heart disease (CAD), stroke and type 2 diabetes mellitus (T2DM).

RESULTS

Over a median follow-up of 161.41 months, we recorded 14,903 CMDs events, comprising 7184 CAD cases, 1914 strokes and 7924 T2DM cases. Cox regression analysis revealed that each unit increase in the INFLA-score corresponded to a 1.5%, 1.1%, 1.2% and 2.4% increase CMDs risk (HR: 1.015, 95% CI 1.013-1.018), CAD risk (HR: 1.011, 95% CI 1.007-1.015), stroke risk (HR: 1.012, 95% CI 1.004-1.020) and T2DM risk (HR: 1.024, 95% CI 1.020-1.028), respectively. Restricted cubic spline analysis indicated a non-linear relationship between cumulative INFLA-score and CMDs risk (P = 0.044). Subgroup analysis revealed interactions between sex, age, history of lipid-lowering drug use, and INFLA-score regarding CMDs risk. Sensitivity analysis corroborated the main findings.

CONCLUSION

Our findings strongly support the close association between INFLA-score and CMDs risk, particularly notable in women, those aged < 55, and individuals with a history of lipid-lowering drug use. These findings offer new insights into the role of inflammation in obesity-related CMDs, suggesting potential applications for prevention and identification of high-risk populations.

Decompressive craniectomy plus best medical treatment versus best medical treatment alone for spontaneous severe deep supratentorial intracerebral haemorrhage: a randomised controlled clinical trial

BACKGROUND

It is unknown whether decompressive craniectomy improves clinical outcome for people with spontaneous severe deep intracerebral haemorrhage. The SWITCH trial aimed to assess whether decompressive craniectomy plus best medical treatment in these patients improves outcome at 6 months compared to best medical treatment alone.

METHODS

In this multicentre, randomised, open-label, assessor-blinded trial conducted in 42 stroke centres in Austria, Belgium, Finland, France, Germany, the Netherlands, Spain, Sweden, and Switzerland, adults (18-75 years) with a severe intracerebral haemorrhage involving the basal ganglia or thalamus were randomly assigned to receive either decompressive craniectomy plus best medical treatment or best medical treatment alone. The primary outcome was a score of 5-6 on the modified Rankin Scale (mRS) at 180 days, analysed in the intention-to-treat population. This trial is registered with ClincalTrials.gov, NCT02258919, and is completed.

FINDINGS

SWITCH had to be stopped early due to lack of funding. Between Oct 6, 2014, and April 4, 2023, 201 individuals were randomly assigned and 197 gave delayed informed consent (96 decompressive craniectomy plus best medical treatment, 101 best medical treatment). 63 (32%) were women and 134 (68%) men, the median age was 61 years (IQR 51-68), and the median haematoma volume 57 mL (IQR 44-74). 42 (44%) of 95 participants assigned to decompressive craniectomy plus best medical treatment and 55 (58%) assigned to best medical treatment alone had an mRS of 5-6 at 180 days (adjusted risk ratio [aRR] 0·77, 95% CI 0·59 to 1·01, adjusted risk difference [aRD] -13%, 95% CI -26 to 0, p=0·057). In the per-protocol analysis, 36 (47%) of 77 participants in the decompressive craniectomy plus best medical treatment group and 44 (60%) of 73 in the best medical treatment alone group had an mRS of 5-6 (aRR 0·76, 95% CI 0·58 to 1·00, aRD -15%, 95% CI -28 to 0). Severe adverse events occurred in 42 (41%) of 103 participants receiving decompressive craniectomy plus best medical treatment and 41 (44%) of 94 receiving best medical treatment.

INTERPRETATION

SWITCH provides weak evidence that decompressive craniectomy plus best medical treatment might be superior to best medical treatment alone in people with severe deep intracerebral haemorrhage. The results do not apply to intracerebral haemorrhage in other locations, and survival is associated with severe disability in both groups.

FUNDING

Swiss National Science Foundation, Swiss Heart Foundation, Inselspital Stiftung, and Boehringer Ingelheim.

The AccelerAge framework: a new statistical approach to predict biological age based on time-to-event data

Aging is a multifaceted and intricate physiological process characterized by a gradual decline in functional capacity, leading to increased susceptibility to diseases and mortality. While chronological age serves as a strong risk factor for age-related health conditions, considerable heterogeneity exists in the aging trajectories of individuals, suggesting that biological age may provide a more nuanced understanding of the aging process. However, the concept of biological age lacks a clear operationalization, leading to the development of various biological age predictors without a solid statistical foundation. This paper addresses these limitations by proposing a comprehensive operationalization of biological age, introducing the "AccelerAge" framework for predicting biological age, and introducing previously underutilized evaluation measures for assessing the performance of biological age predictors. The AccelerAge framework, based on Accelerated Failure Time (AFT) models, directly models the effect of candidate predictors of aging on an individual's survival time, aligning with the prevalent metaphor of aging as a clock. We compare predictors based on the AccelerAge framework to a predictor based on the GrimAge predictor, which is considered one of the best-performing biological age predictors, using simulated data as well as data from the UK Biobank and the Leiden Longevity Study. Our approach seeks to establish a robust statistical foundation for biological age clocks, enabling a more accurate and interpretable assessment of an individual's aging status.

Characterizing quantile-varying covariate effects under the accelerated failure time model

An important task in survival analysis is choosing a structure for the relationship between covariates of interest and the time-to-event outcome. For example, the accelerated failure time (AFT) model structures each covariate effect as a constant multiplicative shift in the outcome distribution across all survival quantiles. Though parsimonious, this structure cannot detect or capture effects that differ across quantiles of the distribution, a limitation that is analogous to only permitting proportional hazards in the Cox model. To address this, we propose a general framework for quantile-varying multiplicative effects under the AFT model. Specifically, we embed flexible regression structures within the AFT model and derive a novel formula for interpretable effects on the quantile scale. A regression standardization scheme based on the g-formula is proposed to enable the estimation of both covariate-conditional and marginal effects for an exposure of interest. We implement a user-friendly Bayesian approach for the estimation and quantification of uncertainty while accounting for left truncation and complex censoring. We emphasize the intuitive interpretation of this model through numerical and graphical tools and illustrate its performance through simulation and application to a study of Alzheimer's disease and dementia.

Effect of glucagon-like peptide-1 receptor agonists and sodium-glucose cotransporter-2 inhibitors on time to outcome in type 2 diabetes cardiorenal outcome trials

INTRODUCTION

In randomized controlled trials (RCTs), treatment effects are commonly reported as hazard ratio, a measure often misinterpreted as a relative risk reduction. The acceleration factor (AF) indicates the extent to which a treatment increases/decreases the time before the occurrence of an outcome and gives useful insights in the interpretation of trials' results.

METHODS

Using individual time-to-event data reconstructed from Kaplan-Meier plots, we estimated AFs for the primary outcomes (POs) and all-cause mortality in glucagon-like peptide-1 receptor agonists (GLP1-RAs) or sodium-glucose cotransporter-2 inhibitors (SGLT2-is) cardiorenal outcome trials in subjects with type 2 diabetes.

RESULTS

AFs were estimated from 28 Kaplan-Meier plots of 19 RCTs. Compared to placebo, most GLP1-RAs increased the time before the onset of POs (from 9 % to 59 %) and all-cause mortality (from 8 to 13 %). Similarly, SGLT2-is increased time before the onset of POs (from 19 % to 87 %) and all-cause mortality (from 13 % to 42 %).

CONCLUSIONS

The AFs provide a complementary and easier-to-interpret measure of treatment effect that could be useful to improve the shared decision-making.

Time-varying effects are common in genetic control of gestational duration

Preterm birth is a major burden to neonatal health worldwide, determined in part by genetics. Recently, studies discovered several genes associated with this trait or its continuous equivalent-gestational duration. However, their effect timing, and thus clinical importance, is still unclear. Here, we use genotyping data of 31 000 births from the Norwegian Mother, Father and Child cohort (MoBa) to investigate different models of the genetic pregnancy 'clock'. We conduct genome-wide association studies using gestational duration or preterm birth, replicating known maternal associations and finding one new fetal variant. We illustrate how the interpretation of these results is complicated by the loss of power when dichotomizing. Using flexible survival models, we resolve this complexity and find that many of the known loci have time-varying effects, often stronger early in pregnancy. The overall polygenic control of birth timing appears to be shared in the term and preterm, but not very preterm, periods and exploratory results suggest involvement of the major histocompatibility complex genes in the latter. These findings show that the known gestational duration loci are clinically relevant and should help design further experimental studies.

Time-varying effects are common in genetic control of gestational duration

Preterm birth is a major burden to neonatal health worldwide, determined in part by genetics. Recently, studies discovered several genes associated with this trait or its continuous equivalent - gestational duration. However, their effect timing, and thus clinical importance, is still unclear. Here, we use genotyping data of 31,000 births from the Norwegian Mother, Father and Child cohort (MoBa) to investigate different models of the genetic pregnancy "clock". We conduct genome-wide association studies using gestational duration or preterm birth, replicating known maternal associations and finding one new foetal variant. We illustrate how the interpretation of these results is complicated by the loss of power when dichotomizing. Using flexible survival models, we resolve this complexity and find that many of the known loci have time-varying effects, often stronger early in pregnancy. The overall polygenic control of birth timing appears to be shared in the term and preterm, but not very preterm periods, and exploratory results suggest involvement of the major histocompatibility complex genes in the latter. These findings show that the known gestational duration loci are clinically relevant, and should help design further experimental studies.

Concurrent dietary intake to nitrate, thiocyanate, and perchlorate is negatively associated with hypertension in adults in the USA

We aimed to comprehensively evaluate the association of urinary nitrate, thiocyanate, and perchlorate metabolites with hypertension among a nationally representative sample of the US adult population. This cross-sectional study investigated data from 15,717 adults aged more than 20 years obtained from the National Health and Nutritional Examination Survey (NHANES) for the years 2005-2016. In the survey, urinary levels of nitrate, thiocyanate, and perchlorate were measured using ion chromatography combined with electrospray tandem mass spectrometry. Blood pressure was calculated as the mean of three measurements. Hypertension was defined as (a) systolic BP ≥130 and/or diastolic BP ≥80 mmHg and/or (b) self-report. Multivariate logistic regression and weighted quantile sum (WQS) regression models were applied to estimate the association between exposure to multiple inorganic anions and hypertension. Restricted cubic spline (RCS) regressions were fitted to discern the potential relationship between the anion exposure and hypertension. These innovation methods used to support our results. Overall, 7533 (49.95%) people with and 7638 (50.35%) without hypertension were included in this study. In the multivariable-adjusted logistic regression models, urinary nitrate (P < 0.001) and perchlorate (P < 0.001) were independently negatively associated with increased occurrence of hypertension, while urinary thiocyanate was insignificantly associated with hypertension (P = 0.664). The WQS regression index showed that, in combination, the three inorganic anions mixture were negatively correlated with hypertension (adjusted OR 0.89; 95% CI 0.83-0.95, P < 0.001). Urinary nitrate was the most heavily weighted component in the hypertension model (weight = 0.784). RCS regression demonstrated that nitrate (nonlinearity P = 0.205) and perchlorate (nonlinearity P = 0.701) were linearly associated with decreased occurrence of hypertension. Concurrent exposure to nitrate, thiocyanate, and perchlorate is associated with a decreased risk of hypertension, with the greatest influence coming from nitrate probably; urinary specific thiocyanate alone had an insignificant association with hypertension.