Innovative estimation of survival using log-normal survival modelling on ACCENT database

Impact of Blood-Count-Derived Inflammatory Markers in Psoriatic Disease Progression

Psoriasis is a chronic immune-mediated disease, linked to local and systemic inflammation and predisposing patients to a higher risk of associated comorbidities. Cytokine levels are not widely available for disease progression monitoring due to high costs. Validated low-cost and reliable markers are needed for assessing disease progression and outcome. This study aims to assess the reliability of blood-count-derived inflammatory markers as disease predictors and to identify prognostic factors for disease severity. Patients fulfilling the inclusion criteria were enrolled in this study. Patients were divided into three study groups according to disease severity measured by the Body Surface Area (BSA) score: mild, moderate, and severe psoriasis. White blood cell count (WBC), neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), derived neutrophil-to-lymphocyte ratio (d-NLR), systemic immune index (SII), systemic inflammation response index (SIRI), and aggregate index of systemic inflammation (AISI) positively were correlated with disease severity (p < 0.005). d-NLR, NLR, and SII are independent prognostic factors for mild and moderate psoriasis (p < 0.05). d-NLR is the only independent prognostic factor for all three study groups. Moderate psoriasis is defined by d-NLR values between 1.49 and 2.19. NLR, PLR, d-NLR, MLR, SII, SIRI, and AISI are useful indicators of systemic inflammation and disease severity in psoriasis.

Multitask Learning with Convolutional Neural Networks and Vision Transformers Can Improve Outcome Prediction for Head and Neck Cancer Patients

Neural-network-based outcome predictions may enable further treatment personalization of patients with head and neck cancer. The development of neural networks can prove challenging when a limited number of cases is available. Therefore, we investigated whether multitask learning strategies, implemented through the simultaneous optimization of two distinct outcome objectives (multi-outcome) and combined with a tumor segmentation task, can lead to improved performance of convolutional neural networks (CNNs) and vision transformers (ViTs). Model training was conducted on two distinct multicenter datasets for the endpoints loco-regional control (LRC) and progression-free survival (PFS), respectively. The first dataset consisted of pre-treatment computed tomography (CT) imaging for 290 patients and the second dataset contained combined positron emission tomography (PET)/CT data of 224 patients. Discriminative performance was assessed by the concordance index (C-index). Risk stratification was evaluated using log-rank tests. Across both datasets, CNN and ViT model ensembles achieved similar results. Multitask approaches showed favorable performance in most investigations. Multi-outcome CNN models trained with segmentation loss were identified as the optimal strategy across cohorts. On the PET/CT dataset, an ensemble of multi-outcome CNNs trained with segmentation loss achieved the best discrimination (C-index: 0.29, 95% confidence interval (CI): 0.22-0.36) and successfully stratified patients into groups with low and high risk of disease progression (p=0.003). On the CT dataset, ensembles of multi-outcome CNNs and of single-outcome ViTs trained with segmentation loss performed best (C-index: 0.26 and 0.26, CI: 0.18-0.34 and 0.18-0.35, respectively), both with significant risk stratification for LRC in independent validation (p=0.002 and p=0.011). Further validation of the developed multitask-learning models is planned based on a prospective validation study, which has recently completed recruitment.

Regression Models for Lifetime Data: An Overview

Two methods dominate the regression analysis of time-to-event data: the accelerated failure time model and the proportional hazards model. Broadly speaking, these predominate in reliability modelling and biomedical applications, respectively. However, many other methods have been proposed, including proportional odds, proportional mean residual life and several other “proportional” models. This paper presents an overview of the field and the concept behind each of these ideas. Multi-parameter modelling is also discussed, in which (in contrast to, say, the proportional hazards model) more than one parameter of the lifetime distribution may depend on covariates. This includes first hitting time (or threshold) regression based on an underlying latent stochastic process. Many of the methods that have been proposed have seen little or no practical use. Lack of user-friendly software is certainly a factor in this. Diagnostic methods are also lacking for most methods.

Transportability Methods for Time-to-Event Outcomes: Application in Adjuvant Colon Cancer Trials

PURPOSE

Differences in the benefits of treatment on 5-year overall survival have been observed in 12 randomized phase III colon cancer adjuvant clinical trials from the ACCENT group. We investigated the reasons for these differences by incorporating the distribution of the observed covariates from each trial.

MATERIALS AND METHODS

We applied state-of-the-art transportability methods on the basis of causal inference, and compared them with a conventional meta-analysis approach to predict the treatment effect for the target population. Prediction errors were defined to evaluate whether the identifiability conditions necessary for causal inference were satisfied among the 12 trials, and to measure the performance of each method.

RESULTS

In the one-trial-at-a-time transportability analysis, the ranks of prediction errors for the target population were mostly consistent with the discrepancy in treatment effects among the 12 trials across the three models. The overall prediction errors between the leave-one-trial-out transportability method and the conventional individual participant data meta-analysis approach were very similar, and more than 40% lower than the overall prediction errors from the one-trial-at-a-time transportability method.

CONCLUSION

The discrepancy in treatment effects among the 12 trials is unlikely to arise from the choice of model specification or distribution of observed covariates but from the distribution of unobserved covariates or study-level features. The ability to quantify heterogeneity among the 12 trials was greatly reduced in both the leave-one-trial-out transportability method and the conventional meta-analysis approach compared with the one-trial-at-a-time transportability method.

Semi-parametric accelerated failure-time model: A useful alternative to the proportional-hazards model in cancer clinical trials

The accelerated failure-time (AFT) model has been long recognized as a useful alternative to the proportional-hazards (PH) model. Semi-parametric AFT model has been known since 1981. Its use has been hampered by the difficulty in solving the estimating equations for the model's coefficients. In recent years, however, important developments have taken place regarding the methods of solving the equations. In this article, we briefly review the developments, focusing mainly on rank-based estimation. We conduct a simulation study that directly focuses on the applicability of the model in the context of (cancer) clinical trials. We also investigate the robustness of the AFT model to the omission of covariates. Finally, we conduct a meta-analysis of multiple clinical trials in gastric cancer to illustrate the benefits of the use of the model in practice.

Elevated bilirubin, alkaline phosphatase at onset, and drug metabolism are associated with prolonged recovery from DILI

BACKGROUND & AIMS

Although most drug-induced liver injury (DILI) cases resolve after the offending medication is discontinued, time to recovery varies among patients, with 6 -12% developing a chronic disease. Herein, we investigated clinical factors and drug properties as potential risk determinants that influence the time course for DILI recovery and developed a model to predict its trajectory.

METHODS

We applied an accelerated failure time model to 294 cases collected by the International Drug-Induced Liver Network Consortium (iDILIC). Factors included in the multivariate recovery score model were selected through univariate analysis. The model was externally validated using 257 cases from the Spanish DILI Registry and 191 cases from the LiverTox database.

RESULTS

Higher serum bilirubin and alkaline phosphatase (ALP) at DILI onset, a longer time to onset, and non-significant drug metabolism were associated with a longer recovery and were included in the recovery score model. We defined high- and low-risk groups based on the scores assigned by the model. The estimated probability of recovery by 6 months was 0.46 (95% CI 0.26-0.61) for the high-risk group and 0.93 (95% CI 0.58-0.99) for the low-risk group in the iDILIC. Model performance was validated in both validation sets. The high- and low-risk cases identified by the model showed a significantly different time course for recovery, with a majority of low-risk cases recovering sooner.

CONCLUSION

The trajectory of biochemical recovery from DILI is predicted by the extent of drug metabolism, serum bilirubin and ALP at DILI onset. The model can be used to compute an estimated DILI recovery and, when a significant delay is predicted, clinicians may consider additional investigations such as histologic evaluation or extended follow-up.

LAY SUMMARY

In this study, we investigated whether drug properties and clinical factors are associated with the time it takes to recover from drug-induced liver injury (DILI). We found that total bilirubin, alkaline phosphatase level at DILI onset, time to onset, and extent of drug metabolism were consistently associated with recovery time. Using these factors, we built a model to predict the trajectory of recovery from DILI and validated this model in 2 independent cohorts. Our findings offer important insights into the factors influencing the trajectory of recovery from DILI. Additional investigations and longer follow-ups can be planned in those for whom a delayed recovery is predicted.

Cure and death play a role in understanding dynamics for COVID-19: Data-driven competing risk compartmental models, with and without vaccination

Several factors have played a strong role in influencing the dynamics of COVID-19 in the U.S. One being the economy, where a tug of war has existed between lockdown measures to control disease versus loosening of restrictions to address economic hardship. A more recent effect has been availability of vaccines and the mass vaccination efforts of 2021. In order to address the challenges in analyzing this complex process, we developed a competing risk compartmental model framework with and without vaccination compartment. This framework separates instantaneous risk of removal for an infectious case into competing risks of cure and death, and when vaccinations are present, the vaccinated individual can also achieve immunity before infection. Computations are performed using a simple discrete time algorithm that utilizes a data driven contact rate. Using population level pre-vaccination data, we are able to identify and characterize three wave patterns in the U.S. Estimated mortality rates for second and third waves are 1.7%, which is a notable decrease from 8.5% of a first wave observed at onset of disease. This analysis reveals the importance cure time has on infectious duration and disease transmission. Using vaccination data from 2021, we find a fourth wave, however the effect of this wave is suppressed due to vaccine effectiveness. Parameters playing a crucial role in this modeling were a lower cure time and a signficantly lower mortality rate for the vaccinated.

A 2D analysis of correlations between the parameters of the Gompertz-Makeham model (or law?) of relationships between aging, mortality, and longevity

When mortality (μ), aging rate (γ) and age (t) are treated according to the Gompertz model μ(t) = μ₀e^γt (GM), any mean age corresponds to a manifold of paired reciprocally changing μ₀ and γ. Therefore, any noisiness of data used to derive GM parameters makes them negatively correlated. Besides this artifactual factor of the Strehler-Mildvan correlation (SMC), other factors emerge when the age-independent mortality C modifies survival according to the Gompertz-Makeham model μ(t) = C+μ₀e^γt (GMM), or body resources are partitioned between survival and protection from aging [the compensation effect of mortality (CEM)]. Theoretical curves in (γ, logμ₀) coordinates show how μ₀ decreases when γ increases upon a constant mean age. Within a species-specific range of γ, such "isoage" curves look as nearly parallel straight lines. The slopes of lines constructed by applying GM to survival curves modeled according to GMM upon changes in C are greater than the isoage slopes. When CEM is modeled, the slopes are still greater. Based on these observations, CEM is shown to contribute to SMC associated with sex differences in lifespan, with the effects of several life-extending drugs, and with recent trends in survival/mortality patterns in high-life-expectancy countries; whereas changes in C underlie differences between even high-life-expectancy countries, not only between high- and low-life-expectancy countries. Such interpretations make sense only if GM is not merely a statistical model, but rather reflects biological realities. Therefore, GM is discussed as derivable by applying certain constraints to a natural law termed the generalized Gompertz-Makeham law.

Applying parametric models to survival data: tradeoffs between statistical significance, biological plausibility, and common sense

Parametric models for survival data help to differentiate aging from other lifespan determinants. However, such inferences suffer from small sizes of experimental animal samples and variable animals handling by different labs. We analyzed control data from a single laboratory where interventions in murine lifespan were studied over decades. The minimal Gompertz model (GM) was found to perform best with most murine strains. However, when several control datasets related to a particular strain are fitted to GM, strikingly rigid interdependencies between GM parameters emerge, consistent with the Strehler-Mildvan correlation (SMC). SMC emerges even when survival patterns do not conform to GM, as with cancer-prone HER2/neu mice, which die at a log-normally distributed age. Numerical experiments show that SMC includes an artifact whose magnitude depends on dataset deviation from conformance to GM irrespectively of the noisiness of small datasets, another contributor to SMC. Still another contributor to SMC is the compensation effect of mortality (CEM): a real tradeoff between the physiological factors responsible for initial vitality and the rate of its decline. To avoid misinterpretations, we advise checking experimental results against a SMC based on historical controls or on subgroups obtained by randomization of control animals. An apparent acceleration of aging associated with a decrease in the initial mortality is invalid if it is not greater than SMC suggests. This approach applied to published data suggests that the effects of calorie restriction and of drugs believed to mimic it are different. SMC and CEM relevance to human survival patterns is discussed.

FOLFOX or CAPOX in Stage II to III Colon Cancer: Efficacy Results of the Italian Three or Six Colon Adjuvant Trial

Purpose Given the cumulative neurotoxicity associated with oxaliplatin, a shorter duration of adjuvant therapy, if equally efficacious, would be advantageous for patients and health-care systems. Methods The Three or Six Colon Adjuvant trial is an open-label, phase III, multicenter, noninferiority trial randomizing patients with high-risk stage II or stage III colon cancer to receive 3 months or 6 months of FOLFOX (fluorouracil, leucovorin, and oxaliplatin) or CAPOX (capecitabine plus oxaliplatin). Primary end-point is relapse-free survival. Results 3,759 patients were accrued from 130 Italian sites, 64% receiving FOLFOX and 36% CAPOX. Two-thirds were stage III. The median time of follow up was 62 months and 772 relapses or deaths have been observed. The hazard ratio (HR) of the 3 months versus 6 months for relapse/death was 1.14 (95% CI, 0.99 to 1.32; P [for noninferiority] = .514) and the CI crossed the noninferiority limit of 1.20. However, the absolute difference in 3-year RFS was 1.9% (95% CI, -0.7% to 4.4%). Counter-intuitively, while the RFS curves were similar for stage III (HR, 1.07; 95% CI, 0.91 to 1.26) and for CAPOX treated patients (HR, 0.98; 95% CI, 0.77 to 1.26), they were not for stage II and for FOLFOX treated patients, with HR of 1.41 (95% CI, 1.05 to 1.89) and 1.23 (95% CI, 1.03 to 1.46), respectively, favoring the 6 months of treatment. Conclusion The Three or Six Colon Adjuvant trial failed to formally show noninferiority of 3 versus 6 months of treatment to the predefined margin of 20% relative increase. The results depended on the adjuvant regimen and risk. For CAPOX, 3 months were as good as 6 months; for FOLFOX, 6 months added extra benefit. Counter-intuitively, the low-risk patients benefitted more than the high-risk population from the 6-month duration. The choice of regimen and duration should depend on patient characteristics and be balanced against the extra toxicity of longer therapy.

Assessment of the prognostic and predictive utility of the Breast Cancer Index (BCI): an NCIC CTG MA.14 study

BACKGROUND

Biomarkers that can be used to accurately assess the residual risk of disease recurrence in women with hormone receptor-positive breast cancer are clinically valuable. We evaluated the prognostic value of the Breast Cancer Index (BCI), a continuous risk index based on a combination of HOXB13:IL17BR and molecular grade index, in women with early breast cancer treated with either tamoxifen alone or tamoxifen plus octreotide in the NCIC MA.14 phase III clinical trial (ClinicalTrials.gov Identifier NCT00002864; registered 1 November 1999).

METHODS

Gene expression analysis of BCI by real-time polymerase chain reaction was performed blinded to outcome on RNA extracted from archived formalin-fixed, paraffin-embedded tumor samples of 299 patients with both lymph node-negative (LN-) and lymph node-positive (LN+) disease enrolled in the MA.14 trial. Our primary objective was to determine the prognostic performance of BCI based on relapse-free survival (RFS). MA.14 patients experienced similar RFS on both treatment arms. Association of gene expression data with RFS was evaluated in univariate analysis with a stratified log-rank test statistic, depicted with a Kaplan-Meier plot and an adjusted Cox survivor plot. In the multivariate assessment, we used stratified Cox regression. The prognostic performance of an emerging, optimized linear BCI model was also assessed in a post hoc analysis.

RESULTS

Of 299 samples, 292 were assessed successfully for BCI for 146 patients accrued in each MA.14 treatment arm. BCI risk groups had a significant univariate association with RFS (stratified log-rank p = 0.005, unstratified log-rank p = 0.007). Adjusted 10-year RFS in BCI low-, intermediate-, and high-risk groups was 87.5 %, 83.9 %, and 74.7 %, respectively. BCI had a significant prognostic effect [hazard ratio (HR) 2.34, 95 % confidence interval (CI) 1.33-4.11; p = 0.004], although not a predictive effect, on RFS in stratified multivariate analysis, adjusted for pathological tumor stage (HR 2.22, 95 % CI 1.22-4.07; p = 0.01). In the post hoc multivariate analysis, higher linear BCI was associated with shorter RFS (p = 0.002).

CONCLUSIONS

BCI had a strong prognostic effect on RFS in patients with early-stage breast cancer treated with tamoxifen alone or with tamoxifen and octreotide. BCI was prognostic in both LN- and LN+ patients. This retrospective study is an independent validation of the prognostic performance of BCI in a prospective trial.

Mining the ACCENT database: a review and update

The database of the Adjuvant Colon Cancer End Points (ACCENT) Group was assembled to address questions in early stage colon cancer that could be best answered by information pooled across many similar trials. Today, the ACCENT database contains individual patient-level data from over 33,000 patients enrolled onto 25 adjuvant colon cancer trials conducted between 1977 and 2008. Since its flagship analysis of 3-year disease-free survival as a surrogate endpoint for 5-year overall survival in 2005, the ACCENT group has produced many noteworthy scientific findings addressing a variety of clinical questions, which we describe here. Additionally, we provide an overview of the history, collaboration, construction, principles, and future of the ACCENT database, as it has set a precedent for multi-trial database creation in other types of cancer.