A practical perspective on the concordance index for the evaluation and selection of prognostic time-to-event models

Prognostic nomograms for young breast cancer: A retrospective study based on the SEER and METABRIC databases

Background

Young breast cancer (YBC) is a subset of breast cancer that is often more aggressive, but less is known about its prognosis. In this study, we aimed to generate nomograms to predict the overall survival (OS) and breast cancer-specific survival (BCSS) of YBC patients.

Methods

Data of women diagnosed with YBC between 2010 and 2020 were obtained from the Surveillance, Epidemiology, and End Results (SEER) database. The patients were randomly allocated into a training cohort (n = 15,227) and internal validation cohort (n = 6,526) at a 7:3 ratio. With the Cox regression models, significant prognostic factors were identified and used to construct 3-, 5-, and 10-year nomograms of OS and BCSS. Data from the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) database were used as an external validation cohort (n = 90).

Results

We constructed nomograms incorporating 10 prognostic factors for OS and BCSS. These nomograms demonstrated strong predictive accuracy for OS and BCSS in the training cohort, with C-indexes of 0.806 and 0.813, respectively. The calibration curves verified that the nomograms have good prediction accuracy. Decision curve analysis demonstrated their practical clinical value for predicting YBC patient survival rates. Additionally, we provided dynamic nomograms to improve the operability of the results. The risk stratification ability assessment also showed that the OS and BCSS rates of the low-risk group were significantly better than those of the high-risk group.

Conclusions

Here, we generated and validated more comprehensive and accurate OS and BCSS nomograms than models previously developed for YBC. These nomograms can help clinicians evaluate patient prognosis and make clinical decisions.

Multivariable models of outcomes with [177Lu]Lu-PSMA-617: analysis of the phase 3 VISION trial

Background

[177Lu]Lu-PSMA-617 (177Lu-PSMA-617) prolonged life in patients with metastatic castration-resistant prostate cancer (mCRPC) in VISION (NCT03511664). However, distinguishing between patients likely and unlikely to respond remains a clinical challenge. We present the first multivariable models of outcomes with 177Lu-PSMA-617 built using data from VISION, a large prospective phase 3 clinical trial powered for overall survival.

Methods

Adults with progressive post androgen receptor pathway inhibitor and taxane prostate-specific membrane antigen (PSMA)-positive mCRPC received 177Lu-PSMA-617 plus protocol-permitted standard of care (SoC) or SoC alone. In this post hoc analysis, multivariable Cox proportional hazards models of overall survival (OS) and radiographic progression-free survival (rPFS), and a logistic regression model of prostate-specific antigen response (≥50% decline; PSA50) were constructed and evaluated using C-index or receiver operating characteristic (ROC) analyses with bootstrapping validation. Nomograms were constructed for visualisation.

Findings

Patients were randomised between June 2018 and October 2019. Data from all 551 patients in the 177Lu-PSMA-617 arm were analysed in multivariable modelling. The OS nomogram (C-index, 0.73; 95% confidence interval [CI], 0.70-0.76) included whole-body maximum standardised uptake value (SUVmax), time since diagnosis, opioid analgesic use, aspartate aminotransferase, haemoglobin, lymphocyte count, presence of PSMA-positive lesions in lymph nodes, lactate dehydrogenase (LDH), alkaline phosphatase (ALP), and neutrophil count. The rPFS nomogram (C-index, 0.68; 0.65-0.72) included SUVmax, time since diagnosis, opioid analgesic use, lymphocyte count, presence of liver metastases by computed tomography, LDH, and ALP. The PSA50 nomogram (area under ROC curve, 0.72; 95% CI, 0.68-0.77) included SUVmax, lymphocyte count and ALP. Performances of the OS and rPFS models were maintained when they were reconstructed excluding SUVmax.

Interpretation

These models of outcomes with 177Lu-PSMA-617 are the first built using prospective phase 3 data. They show that a combination of pretreatment laboratory, clinical, and imaging parameters, reflecting both patient and tumour status, influences outcomes. These models are important for aiding treatment selection, patient management, and clinical trial design.

Funding

Novartis.

Integrating Clinical Variables, Radiomics, and Tumor-derived Cell-free DNA for Enhanced Prediction of Resectable Esophageal Adenocarcinoma Outcomes

BACKGROUND

The value of integrating clinical variables, radiomics, and tumor-derived cell-free DNA (cfDNA) for the prediction of survival and response to chemoradiation of resectable esophageal adenocarcinoma (rEAC) patients is not yet known. Our aim was to investigate if radiomics and cfDNA metrics combined with clinical variables can improve personalized predictions.

METHODS

A cohort of 111 rEAC patients from two centers treated with neoadjuvant chemoradiotherapy was used for exploratory retrospective analyses. Models combining the clinical variables of the SOURCE survival model with radiomic features and cfDNA, were built using elastic net regression and internally validated using 5-fold cross validation. Model performance for overall survival (OS) and time to progression (TTP) were evaluated with the C-index and the area under the curve (AUC) for pathological complete response (pCR) RESULTS: The best performing baseline models for OS and TTP were based on the combination of SOURCE-cfDNA which reached a C-index of 0.55 and 0.59 compared to 0.44-0.45 with SOURCE alone. The addition of re-staging PET radiomics to SOURCE was the most promising addition for predicting OS (C-index: 0.65) and TTP (C-index: 0.60). Baseline risk-stratification was achieved for OS and TTP by combining SOURCE with radiomics or cfDNA, log-rank p<0.01. The best performing combination model for the prediction of pCR reached an AUC of 0.61 compared to 0.47 with SOURCE variables alone.

CONCLUSIONS

The addition of radiomics and cfDNA can improve the performance of an established survival model. External validity needs to be further assessed in future studies together with the optimization of radiomic pipelines.

Validation of the Mirai model for predicting breast cancer risk in Mexican women

OBJECTIVES

To validate the performance of Mirai, a mammography-based deep learning model, in predicting breast cancer risk over a 1-5-year period in Mexican women.

METHODS

This retrospective single-center study included mammograms in Mexican women who underwent screening mammography between January 2014 and December 2016. For women with consecutive mammograms during the study period, only the initial mammogram was included. Pathology and imaging follow-up served as the reference standard. Model performance in the entire dataset was evaluated, including the concordance index (C-Index) and area under the receiver operating characteristic curve (AUC). Mirai's performance in terms of AUC was also evaluated between mammography systems (Hologic versus IMS). Clinical utility was evaluated by determining a cutoff point for Mirai's continuous risk index based on identifying the top 10% of patients in the high-risk category.

RESULTS

Of 3110 patients (median age 52.6 years ± 8.9), throughout the 5-year follow-up period, 3034 patients remained cancer-free, while 76 patients developed breast cancer. Mirai achieved a C-index of 0.63 (95% CI: 0.6-0.7) for the entire dataset. Mirai achieved a higher mean C-index in the Hologic subgroup (0.63 [95% CI: 0.5-0.7]) versus the IMS subgroup (0.55 [95% CI: 0.4-0.7]). With a Mirai index score > 0.029 (10% threshold) to identify high-risk individuals, the study revealed that individuals in the high-risk group had nearly three times the risk of developing breast cancer compared to those in the low-risk group.

CONCLUSIONS

Mirai has a moderate performance in predicting future breast cancer among Mexican women.

CRITICAL RELEVANCE STATEMENT

Prospective efforts should refine and apply the Mirai model, especially to minority populations and women aged between 30 and 40 years who are currently not targeted for routine screening.

KEY POINTS

The applicability of AI models to non-White, minority populations remains understudied. The Mirai model is linked to future cancer events in Mexican women. Further research is needed to enhance model performance and establish usage guidelines.

Evaluating dynamic and predictive discrimination for recurrent event models: use of a time-dependent C-index

Interest in analyzing recurrent event data has increased over the past few decades. One essential aspect of a risk prediction model for recurrent event data is to accurately distinguish individuals with different risks of developing a recurrent event. Although the concordance index (C-index) effectively evaluates the overall discriminative ability of a regression model for recurrent event data, a local measure is also desirable to capture dynamic performance of the regression model over time. Therefore, in this study, we propose a time-dependent C-index measure for inferring the model's discriminative ability locally. We formulated the C-index as a function of time using a flexible parametric model and constructed a concordance-based likelihood for estimation and inference. We adapted a perturbation-resampling procedure for variance estimation. Extensive simulations were conducted to investigate the proposed time-dependent C-index's finite-sample performance and estimation procedure. We applied the time-dependent C-index to three regression models of a study of re-hospitalization in patients with colorectal cancer to evaluate the models' discriminative capability.

Prognostic role of albumin-bilirubin (ALBI) score and Child-Pugh classification in patients with advanced hepatocellular carcinoma under systemic treatment

Hepatocellular carcinoma (HCC) is a lethal malignancy associated with cirrhosis and liver dysfunction. The aim of this study is to characterize a cohort of patients with advanced HCC according to liver function-related variables and evaluate the prognostic significance of Child-Pugh (CP) and albumin-bilirubin (ALBI) scores. A database of 406 HCC patients treated between 2009 and 2023 was retrospectively evaluated. Clinical and laboratory parameters were collected to classify patients into ALBI and CP scores. Survival was estimated using the Kaplan-Meier method and multivariate models were used to evaluate prognosis prediction. In this cohort, 337 (83%) patients were classified as CP-A, while 69 (17%) as CP-B. Additionally, according to ALBI score, 159 (39.2%) individuals were categorised as ALBI-1, 233 (57.4%) as ALBI-2 and 14 (3.4%) as ALBI-3. A statistically significant association between both classifications was observed (p < 0.001). CP and ALBI scores were independently associated with prognosis (Hazard ratio = 2.93 and 1.66, respectively), with better survival for patients with CP-A (versus B) and ALBI-1 (versus -2 and -3). ALBI score showed better predictive performance versus CP (c Harrell´s C index = 0.65 versus 0.62; p = 0.008) and ALBI evolution during the first month of treatment was associated with overall survival. Additionally, ALBI score was able to define distinct prognostic subgroups within CP-A patients. In conclusion, liver function scores, such as ALBI and CP, have a clinically relevant prognostic role in patients with advanced HCC under systemic treatment. ALBI score is a more granular scoring scale than CP, and enables a more precise evaluation of patients with CP-A.

Advancing Drug Safety in Drug Development: Bridging Computational Predictions for Enhanced Toxicity Prediction

The attrition rate of drugs in clinical trials is generally quite high, with estimates suggesting that approximately 90% of drugs fail to make it through the process. The identification of unexpected toxicity issues during preclinical stages is a significant factor contributing to this high rate of failure. These issues can have a major impact on the success of a drug and must be carefully considered throughout the development process. These late-stage rejections or withdrawals of drug candidates significantly increase the costs associated with drug development, particularly when toxicity is detected during clinical trials or after market release. Understanding drug-biological target interactions is essential for evaluating compound toxicity and safety, as well as predicting therapeutic effects and potential off-target effects that could lead to toxicity. This will enable scientists to predict and assess the safety profiles of drug candidates more accurately. Evaluation of toxicity and safety is a critical aspect of drug development, and biomolecules, particularly proteins, play vital roles in complex biological networks and often serve as targets for various chemicals. Therefore, a better understanding of these interactions is crucial for the advancement of drug development. The development of computational methods for evaluating protein-ligand interactions and predicting toxicity is emerging as a promising approach that adheres to the 3Rs principles (replace, reduce, and refine) and has garnered significant attention in recent years. In this review, we present a thorough examination of the latest breakthroughs in drug toxicity prediction, highlighting the significance of drug-target binding affinity in anticipating and mitigating possible adverse effects. In doing so, we aim to contribute to the development of more effective and secure drugs.

Deep learning model integrating radiologic and clinical data to predict mortality after ischemic stroke

Objective

Most prognostic indexes for ischemic stroke mortality lack radiologic information. We aimed to create and validate a deep learning-based mortality prediction model using brain diffusion weighted imaging (DWI), apparent diffusion coefficient (ADC), and clinical factors.

Methods

Data from patients with ischemic stroke who admitted to tertiary hospital during acute periods from 2013 to 2019 were collected and split into training (n = 1109), validation (n = 437), and internal test (n = 654). Data from patients from secondary cardiovascular center was used for external test set (n = 507). The algorithm for predicting mortality, based on DWI and ADC (DLP_DWI), was initially trained. Subsequently, important clinical factors were integrated into this model to create the integrated model (DLP_INTG). The performance of DLP_DWI and DLP_INTG was evaluated by using time-dependent area under the receiver operating characteristic curves (TD AUCs) and Harrell concordance index (C-index) at one-year mortality.

Results

The TD AUC of DLP_DWI was 0.643 in internal test set, and 0.785 in the external dataset. DLP_INTG had a higher performance at predicting one-year mortality than premise score in internal dataset (TD- AUC: 0.859 vs. 0.746; p = 0.046), and in external dataset (TD- AUC: 0.876 vs. 0.808; p = 0.007). DLP_DWI and DLP_INTG exhibited strong discrimination for the high-risk group for one-year mortality.

Interpretation

A deep learning model using brain DWI, ADC and the clinical factors was capable of predicting mortality in patients with ischemic stroke.

Polygenic scores for longitudinal prediction of incident type 2 diabetes in an ancestrally and medically diverse primary care physician network: a patient cohort study

BACKGROUND

The clinical utility of genetic information for type 2 diabetes (T2D) prediction with polygenic scores (PGS) in ancestrally diverse, real-world US healthcare systems is unclear, especially for those at low clinical phenotypic risk for T2D.

METHODS

We tested the association of PGS with T2D incidence in patients followed within a primary care practice network over 16 years in four hypothetical scenarios that varied by clinical data availability (N = 14,712): (1) age and sex; (2) age, sex, body mass index (BMI), systolic blood pressure, and family history of T2D; (3) all variables in (2) and random glucose; and (4) all variables in (3), HDL, total cholesterol, and triglycerides, combined in a clinical risk score (CRS). To determine whether genetic effects differed by baseline clinical risk, we tested for interaction with the CRS.

RESULTS

PGS was associated with incident T2D in all models. Adjusting for age and sex only, the Hazard Ratio (HR) per PGS standard deviation (SD) was 1.76 (95% CI 1.68, 1.84) and the HR of top 5% of PGS vs interquartile range (IQR) was 2.80 (2.39, 3.28). Adjusting for the CRS, the HR per SD was 1.48 (1.40, 1.57) and HR of the top 5% of PGS vs IQR was 2.09 (1.72, 2.55). Genetic effects differed by baseline clinical risk ((PGS-CRS interaction p = 0.05; CRS below the median: HR 1.60 (1.43, 1.79); CRS above the median: HR 1.45 (1.35, 1.55)).

CONCLUSIONS

Genetic information can help identify high-risk patients even among those perceived to be low risk in a clinical evaluation.

Architecting the metabolic reprogramming survival risk framework in LUAD through single-cell landscape analysis: three-stage ensemble learning with genetic algorithm optimization

Recent studies have increasingly revealed the connection between metabolic reprogramming and tumor progression. However, the specific impact of metabolic reprogramming on inter-patient heterogeneity and prognosis in lung adenocarcinoma (LUAD) still requires further exploration. Here, we introduced a cellular hierarchy framework according to a malignant and metabolic gene set, named malignant & metabolism reprogramming (MMR), to reanalyze 178,739 single-cell reference profiles. Furthermore, we proposed a three-stage ensemble learning pipeline, aided by genetic algorithm (GA), for survival prediction across 9 LUAD cohorts (n = 2066). Throughout the pipeline of developing the three stage-MMR (3 S-MMR) score, double training sets were implemented to avoid over-fitting; the gene-pairing method was utilized to remove batch effect; GA was harnessed to pinpoint the optimal basic learner combination. The novel 3 S-MMR score reflects various aspects of LUAD biology, provides new insights into precision medicine for patients, and may serve as a generalizable predictor of prognosis and immunotherapy response. To facilitate the clinical adoption of the 3 S-MMR score, we developed an easy-to-use web tool for risk scoring as well as therapy stratification in LUAD patients. In summary, we have proposed and validated an ensemble learning model pipeline within the framework of metabolic reprogramming, offering potential insights for LUAD treatment and an effective approach for developing prognostic models for other diseases.

Deep learning model for predicting postoperative survival of patients with gastric cancer

Background

Prognostic prediction for surgical treatment of gastric cancer remains valuable in clinical practice. This study aimed to develop survival models for postoperative gastric cancer patients.

Methods

Eleven thousand seventy-five patients from the Surveillance, Epidemiology, and End Results (SEER) database were included, and 122 patients from the Chinese database were used for external validation. The training cohort was created to create three separate models, including Cox regression, RSF, and DeepSurv, using data from the SEER database split into training and test cohorts with a 7:3 ratio. Test cohort was used to evaluate model performance using c-index, Brier scores, calibration, and the area under the curve (AUC). The new risk stratification based on the best model will be compared with the AJCC stage on the test and Chinese cohorts using decision curve analysis (DCA), the net reclassification index (NRI), and integrated discrimination improvement (IDI).

Results

It was discovered that the DeepSurv model predicted postoperative gastric cancer patients' overall survival (OS) with a c-index of 0.787; the area under the curve reached 0.781, 0.798, 0.868 at 1-, 3- and 5- years, respectively; the Brier score was below 0.25 at different time points; showing an advantage over the Cox and RSF models. The results are also validated in the China cohort. The calibration plots demonstrated good agreement between the DeepSurv model's forecast and actual results. The NRI values (test cohort: 0.399, 0.288, 0.267 for 1-, 3- and 5-year OS prediction; China cohort:0.399, 0.288 for 1- and 3-year OS prediction) and IDI (test cohort: 0.188, 0.169, 0.157 for 1-, 3- and 5-year OS prediction; China cohort: 0.189, 0.169 for 1- and 3-year OS prediction) indicated that the risk score stratification performed significantly better than the AJCC staging alone (P < 0.05). DCA showed that the risk score stratification was clinically useful and had better discriminative ability than the AJCC staging. Finally, an interactive native web-based prediction tool was constructed for the survival prediction of patients with postoperative gastric cancer.

Conclusion

In this study, a high-performance prediction model for the postoperative prognosis of gastric cancer was developed using DeepSurv, which offers essential benefits for risk stratification and prognosis prediction for each patient.

Prognostic Performance of RECIP 1.0 Based on [18F]PSMA-1007 PET in Prostate Cancer Patients Treated with [177Lu]Lu-PSMA I&T

In metastatic castration-resistant prostate cancer (mCRPC) patients treated with prostate-specific membrane antigen (PSMA)-targeted radioligand therapy (RLT), the recently proposed criteria for evaluating response to PSMA PET (RECIP 1.0) based on 68Ga- and 18F-labeled PET agents provided prognostic information in addition to changes in prostate-specific antigen (PSA) levels. Our aim was to evaluate the prognostic performance of this framework for overall survival (OS) in patients undergoing RLT and imaged with [18F]PSMA-1007 PET/CT and compare the prognostic performance with the PSA-based response assessment. Methods: In total, 73 patients with mCRPC who were scanned with [18F]PSMA-1007 PET/CT before and after 2 cycles of RLT were retrospectively analyzed. We calculated the changes in serum PSA levels (ΔPSA) and quantitative PET parameters for the whole-body tumor burden (SUVmean, SUVmax, PSMA tumor volume, and total lesion PSMA). Men were also classified following the Prostate Cancer Working Group 3 (PCWG3) criteria for ΔPSA and RECIP 1.0 for PET imaging response. We performed univariable Cox regression analysis, followed by multivariable and Kaplan-Meier analyses. Results: Median OS was 15 mo with a median follow-up time of 14 mo. Univariable Cox regression analysis provided significant associations with OS for ΔPSA (per percentage, hazard ratio [HR], 1.004; 95% CI, 1.002-1.007; P < 0.001) and PSMA tumor volume (per unit, HR, 1.003; 95% CI, 1.000-1.005; P = 0.03). Multivariable Cox regression analysis confirmed ΔPSA (per percentage, HR, 1.004; 95% CI, 1.001-1.006; P = 0.006) as an independent prognosticator for OS. Kaplan-Meier analyses provided significant segregation between individuals with versus those without any PSA response (19 mo vs. 14 mo; HR, 2.00; 95% CI, 0.95-4.18; P = 0.04). Differentiation between patients with or without progressive disease (PD) was also feasible when applying PSA-based PCWG3 (19 mo vs. 9 mo for non-PD and PD, respectively; HR, 2.29; 95% CI, 1.03-5.09; P = 0.01) but slightly failed when applying RECIP 1.0 (P = 0.08). A combination of both response systems (PCWG3 and RECIP 1.0), however, yielded the best discrimination between individuals without versus those with PD (19 mo vs. 8 mo; HR, 2.78; 95% CI, 1.32-5.86; P = 0.002). Conclusion: In patients with mCRPC treated with RLT and imaged with [18F]PSMA-1007, frameworks integrating both the biochemical (PCWG3) and PET-based response (RECIP 1.0) may best assist in identifying subjects prone to disease progression.

Relevance of different prognostic scores in primary CNS lymphoma in the era of intensified treatment regimens: A retrospective, multicenter analysis of 174 patients

OBJECTIVES

Treatment intensification (including consolidative high-dose chemotherapy with autologous stem cell transplantation [HDT-ASCT]) significantly improved outcome in primary central nervous system lymphoma (PCNSL) patients.

METHODS

We conducted a multicenter, retrospective analysis of newly diagnosed PCNSL patients, treated with intensified treatment regimens. The following scores were evaluated in terms of overall survival (OS) and progression-free survival (PFS): Memorial Sloan-Kettering Cancer Center (MSKCC), International Extranodal Lymphoma Study Group (IELSG), and three-factor (3F) prognostic score. Further, all scores were comparatively investigated for model quality and concordance.

RESULTS

Altogether, 174 PCNSL patients were included. One hundred and five patients (60.3%) underwent HDT-ASCT. Two-year OS and 2-year PFS for the entire population were 73.3% and 48.5%, respectively. The MSKCC (p = .003) and 3F score (p < .001), but not the IELSG score (p = .06), had the discriminatory power to identify different risk groups for OS. In regard to concordance, the 3F score (C-index [0.71]) outperformed both the MSKCC (C-index [0.64]) and IELSG (C-index [0.53]) score. Moreover, the superiority of the 3F score was shown for PFS, successfully stratifying patients in three risk groups, which also resulted in the highest C-index (0.66).

CONCLUSION

The comparative analysis of established PCNSL risk scores affirm the clinical utility of the 3F score stratifying the widest prognostic spectrum among PCNSL patients treated with intensified treatment approaches.

Generation of synthetic whole-slide image tiles of tumours from RNA-sequencing data via cascaded diffusion models

Training machine-learning models with synthetically generated data can alleviate the problem of data scarcity when acquiring diverse and sufficiently large datasets is costly and challenging. Here we show that cascaded diffusion models can be used to synthesize realistic whole-slide image tiles from latent representations of RNA-sequencing data from human tumours. Alterations in gene expression affected the composition of cell types in the generated synthetic image tiles, which accurately preserved the distribution of cell types and maintained the cell fraction observed in bulk RNA-sequencing data, as we show for lung adenocarcinoma, kidney renal papillary cell carcinoma, cervical squamous cell carcinoma, colon adenocarcinoma and glioblastoma. Machine-learning models pretrained with the generated synthetic data performed better than models trained from scratch. Synthetic data may accelerate the development of machine-learning models in scarce-data settings and allow for the imputation of missing data modalities.

Clinical Feature Ranking Based on Ensemble Machine Learning Reveals Top Survival Factors for Glioblastoma Multiforme

Glioblastoma multiforme (GM) is a malignant tumor of the central nervous system considered to be highly aggressive and often carrying a terrible survival prognosis. An accurate prognosis is therefore pivotal for deciding a good treatment plan for patients. In this context, computational intelligence applied to data of electronic health records (EHRs) of patients diagnosed with this disease can be useful to predict the patients' survival time. In this study, we evaluated different machine learning models to predict survival time in patients suffering from glioblastoma and further investigated which features were the most predictive for survival time. We applied our computational methods to three different independent open datasets of EHRs of patients with glioblastoma: the Shieh dataset of 84 patients, the Berendsen dataset of 647 patients, and the Lammer dataset of 60 patients. Our survival time prediction techniques obtained concordance index (C-index) = 0.583 in the Shieh dataset, C-index = 0.776 in the Berendsen dataset, and C-index = 0.64 in the Lammer dataset, as best results in each dataset. Since the original studies regarding the three datasets analyzed here did not provide insights about the most predictive clinical features for survival time, we investigated the feature importance among these datasets. To this end, we then utilized Random Survival Forests, which is a decision tree-based algorithm able to model non-linear interaction between different features and might be able to better capture the highly complex clinical and genetic status of these patients. Our discoveries can impact clinical practice, aiding clinicians and patients alike to decide which therapy plan is best suited for their unique clinical status.

Differentiation Potential of Hypodifferentiated Subsets of Nephrogenic Rests and Its Relationship to Prognosis in Wilms Tumor

Background Wilms tumor (WT) is highly curable, although anaplastic histology or relapse imparts a worse prognosis. Nephrogenic rests (NR) associated with a high risk of developing WT are abnormally retained embryonic kidney precursor cells. Methods After pseudo-time analysis using single-cell RNA sequencing (scRNA-seq) data, we generated and validated a WT differentiation-related gene (WTDRG) signature to predict overall survival (OS) in children with a poor OS. Results A differentiation trajectory from NR to WT was identified and showed that hypodifferentiated subsets of NR could differentiate into WT. Classification of WT children with anaplastic histology or relapse based on the expression patterns of WTDRGs suggested that patients with relatively high levels of hypodifferentiated NR presented a poorer prognosis. A WTDRG-based risk model and a clinically applicable nomogram was developed. Conclusions These findings may inform oncogenesis of WT and interventions directed toward poor prognosis in WT children of anaplastic histology or relapse.

Plasma NGAL levels in stable kidney transplant recipients and the risk of allograft loss

BACKGROUND

The objective of this study was to investigate the utility of neutrophil gelatinase-associated lipocalin (NGAL) and calprotectin (CPT) to predict long-term graft survival in stable kidney transplant recipients (KTR).

METHODS

A total of 709 stable outpatient KTR were enrolled >2 months post-transplant. The utility of plasma and urinary NGAL (pNGAL, uNGAL) and plasma and urinary CPT at enrollment to predict death-censored graft loss was evaluated during a 58-month follow-up.

RESULTS

Among biomarkers, pNGAL showed the best predictive ability for graft loss and was the only biomarker with an area under the curve (AUC) > 0.7 for graft loss within 5 years. Patients with graft loss within 5 years (n = 49) had a median pNGAL of 304 [interquartile range (IQR) 235-358] versus 182 (IQR 128-246) ng/mL with surviving grafts (P < .001). Time-dependent receiver operating characteristic analyses at 58 months indicated an AUC for pNGAL of 0.795, serum creatinine-based Chronic Kidney Disease Epidemiology Collaboration estimated glomerular filtration rate (eGFR) had an AUC of 0.866. pNGAL added to a model based on conventional risk factors for graft loss with death as competing risk (age, transplant age, presence of donor-specific antibodies, presence of proteinuria, history of delayed graft function) had a strong independent association with graft loss {subdistribution hazard ratio (sHR) for binary log-transformed pNGAL [log2(pNGAL)] 3.4, 95% confidence interval (CI) 2.24-5.15, P < .0001}. This association was substantially attenuated when eGFR was added to the model [sHR for log2(pNGAL) 1.63, 95% CI 0.92-2.88, P = .095]. Category-free net reclassification improvement of a risk model including log2(pNGAL) in addition to conventional risk factors and eGFR was 54.3% (95% CI 9.2%-99.3%) but C-statistic did not improve significantly.

CONCLUSIONS

pNGAL was an independent predictor of renal allograft loss in stable KTR from one transplant center but did not show consistent added value when compared with baseline predictors including the conventional marker eGFR. Future studies in larger cohorts are warranted.

NCAPH drives breast cancer progression and identifies a gene signature that predicts luminal a tumour recurrence

BACKGROUND

Luminal A tumours generally have a favourable prognosis but possess the highest 10-year recurrence risk among breast cancers. Additionally, a quarter of the recurrence cases occur within 5 years post-diagnosis. Identifying such patients is crucial as long-term relapsers could benefit from extended hormone therapy, while early relapsers might require more aggressive treatment.

METHODS

We conducted a study to explore non-structural chromosome maintenance condensin I complex subunit H's (NCAPH) role in luminal A breast cancer pathogenesis, both in vitro and in vivo, aiming to identify an intratumoural gene expression signature, with a focus on elevated NCAPH levels, as a potential marker for unfavourable progression. Our analysis included transgenic mouse models overexpressing NCAPH and a genetically diverse mouse cohort generated by backcrossing. A least absolute shrinkage and selection operator (LASSO) multivariate regression analysis was performed on transcripts associated with elevated intratumoural NCAPH levels.

RESULTS

We found that NCAPH contributes to adverse luminal A breast cancer progression. The intratumoural gene expression signature associated with elevated NCAPH levels emerged as a potential risk identifier. Transgenic mice overexpressing NCAPH developed breast tumours with extended latency, and in Mouse Mammary Tumor Virus (MMTV)-NCAPHErbB2 double-transgenic mice, luminal tumours showed increased aggressiveness. High intratumoural Ncaph levels correlated with worse breast cancer outcome and subpar chemotherapy response. A 10-gene risk score, termed Gene Signature for Luminal A 10 (GSLA10), was derived from the LASSO analysis, correlating with adverse luminal A breast cancer progression.

CONCLUSIONS

The GSLA10 signature outperformed the Oncotype DX signature in discerning tumours with unfavourable outcomes, previously categorised as luminal A by Prediction Analysis of Microarray 50 (PAM50) across three independent human cohorts. This new signature holds promise for identifying luminal A tumour patients with adverse prognosis, aiding in the development of personalised treatment strategies to significantly improve patient outcomes.

Deep learning in MRI-guided radiation therapy: A systematic review

Recent advances in MRI-guided radiation therapy (MRgRT) and deep learning techniques encourage fully adaptive radiation therapy (ART), real-time MRI monitoring, and the MRI-only treatment planning workflow. Given the rapid growth and emergence of new state-of-the-art methods in these fields, we systematically review 197 studies written on or before December 31, 2022, and categorize the studies into the areas of image segmentation, image synthesis, radiomics, and real time MRI. Building from the underlying deep learning methods, we discuss their clinical importance and current challenges in facilitating small tumor segmentation, accurate x-ray attenuation information from MRI, tumor characterization and prognosis, and tumor motion tracking. In particular, we highlight the recent trends in deep learning such as the emergence of multi-modal, visual transformer, and diffusion models.

Deep learning-assisted survival prognosis in renal cancer: A CT scan-based personalized approach

This paper presents a deep learning (DL) approach for predicting survival probabilities of renal cancer patients based solely on preoperative CT imaging. The proposed approach consists of two networks: a classifier- and a survival- network. The classifier attempts to extract features from 3D CT scans to predict the ISUP grade of Renal cell carcinoma (RCC) tumors, as defined by the International Society of Urological Pathology (ISUP). Our classifier is a 3D convolutional neural network to avoid losing crucial information on the interconnection of slides in 3D images. We employ multiple procedures, including image augmentation, preprocessing, and concatenation, to improve the performance of the classifier. Given the strong correlation between ISUP grading and renal cancer prognosis in the clinical context, we use the ISUP grading features extracted by the classifier as the input to the survival network. By leveraging this clinical association and the classifier network, we are able to model our survival analysis using a simple DL-based network. We adopt a discrete LogisticHazard-based loss to extract intrinsic survival characteristics of RCC tumors from CT images. This allows us to build a completely parametric survival model that varies with patients' tumor characteristics and predicts non-proportional survival probability curves for different patients. Our results demonstrated that the proposed method could predict the future course of renal cancer with reasonable accuracy from the CT scans. The proposed method obtained an average concordance index of 0.72, an integrated Brier score of 0.15, and an area under the curve value of 0.71 on the test cohorts.

Survival prediction of glioblastoma patients using modern deep learning and machine learning techniques

In this study, we utilized data from the Surveillance, Epidemiology, and End Results (SEER) database to predict the glioblastoma patients' survival outcomes. To assess dataset skewness and detect feature importance, we applied Pearson's second coefficient test of skewness and the Ordinary Least Squares method, respectively. Using two sampling strategies, holdout and five-fold cross-validation, we developed five machine learning (ML) models alongside a feed-forward deep neural network (DNN) for the multiclass classification and regression prediction of glioblastoma patient survival. After balancing the classification and regression datasets, we obtained 46,340 and 28,573 samples, respectively. Shapley additive explanations (SHAP) were then used to explain the decision-making process of the best model. In both classification and regression tasks, as well as across holdout and cross-validation sampling strategies, the DNN consistently outperformed the ML models. Notably, the accuracy were 90.25% and 90.22% for holdout and five-fold cross-validation, respectively, while the corresponding R2 values were 0.6565 and 0.6622. SHAP analysis revealed the importance of age at diagnosis as the most influential feature in the DNN's survival predictions. These findings suggest that the DNN holds promise as a practical auxiliary tool for clinicians, aiding them in optimal decision-making concerning the treatment and care trajectories for glioblastoma patients.

Re-exploration of prognosis in type B thymomas: establishment of a predictive nomogram model

OBJECTIVE

To explore the risk factors for disease progression after initial treatment of type B thymomas using a predictive nomogram model.

METHODS

A single-center retrospective study of patients with type B thymoma was performed. The Cox proportional hazard model was used for univariate and multivariate analyses. Variables with statistical and clinical significance in the multivariate Cox regression were integrated into a nomogram to establish a predictive model for disease progression.

RESULTS

A total of 353 cases with type B thymoma were retrieved between January 2012 and December 2021. The median follow-up was 58 months (range: 1-128 months). The 10-year progression-free survival (PFS) was 91.8%. The final nomogram model included R0 resection status and Masaoka stage, with a concordance index of 0.880. Non-R0 resection and advanced Masaoka stage were negative prognostic factors for disease progression (p < 0.001). No benefits of postoperative radiotherapy (PORT) were observed in patients with advanced stage and non-R0 resection (p = 0.114 and 0.284, respectively).

CONCLUSION

The best treatment strategy for type B thymoma is the detection and achievement of R0 resection as early as possible. Long-term follow-up is necessary, especially for patients with advanced Masaoka stage and who have not achieved R0 resection. No prognostic benefits were observed for PORT.

Predicting lung cancer survival prognosis based on the conditional survival bayesian network

Lung cancer is a leading cause of cancer deaths and imposes an enormous economic burden on patients. It is important to develop an accurate risk assessment model to determine the appropriate treatment for patients after an initial lung cancer diagnosis. The Cox proportional hazards model is mainly employed in survival analysis. However, real-world medical data are usually incomplete, posing a great challenge to the application of this model. Commonly used imputation methods cannot achieve sufficient accuracy when data are missing, so we investigated novel methods for the development of clinical prediction models. In this article, we present a novel model for survival prediction in missing scenarios. We collected data from 5,240 patients diagnosed with lung cancer at the Weihai Municipal Hospital, China. Then, we applied a joint model that combined a BN and a Cox model to predict mortality risk in individual patients with lung cancer. The established prognostic model achieved good predictive performance in discrimination and calibration. We showed that combining the BN with the Cox proportional hazards model is highly beneficial and provides a more efficient tool for risk prediction.

Endothelial Activation and Stress Index in adults undergoing allogeneic hematopoietic cell transplantation with post-transplant cyclophosphamide-based prophylaxis

BACKGROUND AIMS

Post-transplant cyclophosphamide (PTCY)-based prophylaxis is becoming widespread for allogeneic hematopoietic cell transplantation (allo-HCT) performed independently of the selected donor source. In parallel, use of the Endothelial Activation and Stress Index (EASIX)-considered a surrogate parameter of endothelial activation-for predicting patient outcomes and clinical complications is gaining popularity in the allo-HCT setting.

METHODS

We first investigated whether the dynamics of EASIX after allo-HCT differ between patients receiving PTCY and patients receiving other prophylaxis. We then investigated whether the predictive capacity of EASIX persists in PTCY-based allo-HCT. A total of 328 patients transplanted between 2014 and 2020 were included, and 201 (61.2%) received PTCY.

RESULTS

EASIX trends differed significantly between the groups. Compared with patients receiving other prophylaxis, patients receiving PTCY had lower EASIX on day 0 and higher values between day 7 and day 100. In patients receiving PTCY, higher EASIX correlated significantly with higher non-relapse mortality (NRM) and lower overall survival (OS) when measured before and during the first 180 days after allo-HCT. In addition, higher EASIX scores measured at specific time points were predictors of veno-occlusive disease (VOD), transplant-associated thrombotic microangiopathy (TA-TMA) and grade 2-4 acute graft-versus-host disease (aGVHD) risk.

CONCLUSIONS

This study demonstrates how EASIX trends vary during the first 180 days after allo-HCT in patients receiving PTCY and those not receiving PTCY and validates the utility of this index for predicting NRM, OS and risk of VOD, TA-TMA and grade 2-4 aGVHD in patients receiving PTCY.

Application of Artificial intelligence in COVID-19-related geriatric care: A scoping review

BACKGROUND

Older adults have been disproportionately affected by the COVID-19 pandemic. This scoping review aimed to summarize the current evidence of artificial intelligence (AI) use in the screening/monitoring, diagnosis, and/or treatment of COVID-19 among older adults.

METHOD

The review followed the Joanna Briggs Institute and Arksey and O'Malley frameworks. An information specialist performed a comprehensive search from the date of inception until May 2021, in six bibliographic databases. The selected studies considered all populations, and all AI interventions that had been used in COVID-19-related geriatric care. We focused on patient, healthcare provider, and healthcare system-related outcomes. The studies were restricted to peer-reviewed English publications. Two authors independently screened the titles and abstracts of the identified records, read the selected full texts, and extracted data from the included studies using a validated data extraction form. Disagreements were resolved by consensus, and if this was not possible, the opinion of a third reviewer was sought.

RESULTS

Six databases were searched , yielding 3,228 articles, of which 10 were included. The majority of articles used a single AI model to assess the association between patients' comorbidities and COVID-19 outcomes. Articles were mainly conducted in high-income countries, with limited representation of females in study participants, and insufficient reporting of participants' race and ethnicity.

DISCUSSION

This review highlighted how the COVID-19 pandemic has accelerated the application of AI to protect older populations, with most interventions in the pilot testing stage. Further work is required to measure effectiveness of these technologies in a larger scale, use more representative datasets for training of AI models, and expand AI applications to low-income countries.

Data-driven grading of acute graft-versus-host disease

Despite advances in allogeneic hematopoietic cell transplantation, acute graft-versus-host disease (aGVHD) remains its leading complication, yet with heterogeneous outcomes. Here, we analyzed aGVHD phenotypes and clinical classifications in depth in large, multicenter cohorts involving 3019 patients and addressed prevailing gaps by developing data-driven models. We compared, tested and verified these along with all conventional classifications in independent cohorts and found that data-driven grading outperformed conventional grading in Akaike information criterion and concordance index metrics. Data-driven classifications refined aGVHD assessment with up to 12 severity grades, which were associated with distinct nonrelapse mortality (NRM) and confirmed the key role of intestinal aGVHD. We developed an online calculator for physicians to implement principal component-derived grading (PC1). These results provide substantial insight into the evaluation of aGVHD phenotypes and multiorgan involvement, which relegates the exclusive reporting of overall aGVHD severity grades in transplant registries and clinical trials. Data-driven aGVHD grading provides an expandable platform to refine classification and transplant risk assessment.

A prognostic model for ovarian neoplasms established by an integrated analysis of 1580 transcriptomic profiles

Even after debulking surgery combined with chemotherapy or new adjuvant chemotherapy paired with internal surgery, the average year of disease free survival in advanced ovarian cancer was approximately 1.7 years1. The development of a molecular predictor of early recurrence would allow for the identification of ovarian cancer (OC) patients with high risk of relapse. The Ovarian Cancer Disease Free Survival Predictor (ODFSP), a predictive model constructed from a special set of 1580 OC tumors in which gene expression was assessed using both microarray and sequencing platforms, was created by our team. To construct gene expression barcodes that were resistant to biases caused by disparate profiling platforms and batch effects, we employed a meta-analysis methodology that was based on the binary gene pair technique. We demonstrate that ODFSP is a reliable single-sample predictor of early recurrence (1 year or less) using the largest pool of OC transcriptome data sets available to date. The ODFSP model showed significantly high prognostic value for binary recurrence prediction unaffected by clinicopathologic factors, with a meta-estimate of the area under the receiver operating curve of 0.64 (P  =  4.6E-05) and a D-index (robust hazard ratio) of 1.67 (P  =  9.2E-06), respectively. GO analysis of ODFSP's 2040 gene pairs (collapsed to 886 distinct genes) revealed the involvement in small molecular catabolic process, sulfur compound metabolic process, organic acid catabolic process, sulfur compound biosynthetic process, glycosaminoglycan metabolic process and aminometabolic process. Kyoto encyclopedia of genes and genomes pathway analysis of ODFSP's signature genes identified prominent pathways that included cAMP signaling pathway and FoxO signaling pathway. By identifying individuals who might benefit from a more aggressive treatment plan or enrolment in a clinical trial but who will not benefit from standard surgery or chemotherapy, ODFSP could help with treatment decisions.

A deep learning approach based on multi-omics data integration to construct a risk stratification prediction model for skin cutaneous melanoma

PURPOSE

Skin cutaneous melanoma (SKCM) is a highly aggressive melanocytic carcinoma whose high heterogeneity and complex etiology make its prognosis difficult to predict. This study aimed to construct a risk subtype typing model for SKCM.

METHODS

The study proposes a deep learning framework combining early fusion feature autoencoder (AE) and late fusion feature AE for risk subtype prediction of SKCM. The deep learning framework integrates mRNA, miRNA, and DNA methylation data of SKCM patients from The Cancer Genome Atlas (TCGA), and clusters the screened multi-omics features associated with survival prognosis to identify risk subtypes. Differential expression analysis and functional enrichment analysis were performed between risk subtypes, while SVM classifiers were constructed between differentially expressed genes (DEGs) obtained by Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression screening and risk subtype labels inferred from multi-omics data, and the predictive robustness of risk subtypes inferred from the risk subtype classification prediction model was validated using two independent datasets.

RESULTS

The deep learning framework that combined early fusion feature AE with late fusion feature AE distinguished the two best risk subtypes compared to the multi-omics integration approach with single strategy AE or PCA. A promising C-index (C-index = 0.748) and a significant difference in survival (log-rank P value = 4.61 × 10-9) were found between the identified risk subtypes. The DEGs with the top significance values together with differentially expressed miRNAs provided the biological interpretation of risk subtypes on SKCM. Finally, the framework was applied to predict risk subtypes in two independent test datasets of SKCM patients, all of which showed good predictive power (C-index > 0.680) and significant survival differences (log-rank P value < 0.01).

CONCLUSION

The SKCM risk subtypes identified by integrating multi-omics data based on deep learning can not only improve the understanding of the molecular mechanisms of SKCM, but also provide clinicians with assistance in treatment decisions.

Accounting for time dependency in meta-analyses of concordance probability estimates

Recent years have seen the development of many novel scoring tools for disease prognosis and prediction. To become accepted for use in clinical applications, these tools have to be validated on external data. In practice, validation is often hampered by logistical issues, resulting in multiple small-sized validation studies. It is therefore necessary to synthesize the results of these studies using techniques for meta-analysis. Here we consider strategies for meta analyzing the concordance probability for time-to-event data ("C-index"), which has become a popular tool to evaluate the discriminatory power of prediction models with a right-censored outcome. We show that standard meta-analysis of the C-index may lead to biased results, as the magnitude of the concordance probability depends on the length of the time interval used for evaluation (defined e.g., by the follow-up time, which might differ considerably between studies). To address this issue, we propose a set of methods for random-effects meta-regression that incorporate time directly as covariate in the model equation. In addition to analyzing nonlinear time trends via fractional polynomial, spline, and exponential decay models, we provide recommendations on suitable transformations of the C-index before meta-regression. Our results suggest that the C-index is best meta-analyzed using fractional polynomial meta-regression with logit-transformed C-index values. Classical random-effects meta-analysis (not considering time as covariate) is demonstrated to be a suitable alternative when follow-up times are small. Our findings have implications for the reporting of C-index values in future studies, which should include information on the length of the time interval underlying the calculations.

Increasing Neurofilament and Glial Fibrillary Acidic Protein After Treatment Discontinuation Predicts Multiple Sclerosis Disease Activity

BACKGROUND AND OBJECTIVES

Stable patients with multiple sclerosis (MS) may discontinue treatment, but the risk of disease activity is unknown. Serum neurofilament light chain (sNfL) and serum glial fibrillary acidic protein (sGFAP) are biomarkers of subclinical disease activity and may help risk stratification. In this study, sNfL and sGFAP levels in stable patients were evaluated before and after treatment discontinuation to determine association with disease activity.

METHODS

This observational study included patients enrolled in the Comprehensive Longitudinal Investigation in MS at the Brigham and Women's Hospital who discontinued treatment after >2 years disease activity-free. Two serum samples within 2 years, before and after treatment stop, were sent for sNfL and sGFAP measurements by single-molecule array. Biannual neurologic examinations and yearly MRI scans determined disease activity by 3 time-to-event outcomes: 6-month confirmed disability worsening (CDW), clinical attacks, and MRI activity (new T2 or contrast-enhancing lesions). Associations between each outcome and log-transformed sNfL and sGFAP levels pretreatment stop and posttreatment stop and the percent change were estimated using multivariable Cox regression analysis adjusting for age, disability, disease duration, and duration from attack before treatment stop.

RESULTS

Seventy-eight patients (92% female) discontinued treatment at a median (interquartile range) age of 48.5 years (39.0-55.7) and disease duration of 12.3 years (7.5-18.8) and were followed up for 6.3 years (4.2-8.5). CDW occurred in 27 patients (35%), new attacks in 19 (24%), and new MRI activity in 26 (33%). Higher posttreatment stop sNfL level was associated with CDW (adjusted hazard ratio (aHR) 2.80, 95% CI 1.36-5.76, p = 0.005) and new MRI activity (aHR 3.09, 95% CI 1.42-6.70, p = 0.004). Patients who had >100% increase in sNfL level from pretreatment stop to posttreatment stop had greater risk of CDW (HR 3.87, 95% CI 1.4-10.7, p = 0.009) and developing new MRI activity (HR 4.02, 95% CI 1.51-10.7, p = 0.005). Patients who had >50% increase in sGFAP level also had greater risk of CDW (HR 5.34, 95% CI 1.4-19.9, p = 0.012) and developing new MRI activity (HR 5.16, 95% CI 1.71-15.6, p = 0.004).

DISCUSSION

Stable patients who discontinue treatment may be risk stratified by sNfL and sGFAP levels measured before and after discontinuing treatment. Further studies are needed to validate findings and determine whether resuming treatment in patients with increasing biomarker levels reduces risk of subsequent disease activity.

A novel deep learning-based algorithm combining histopathological features with tissue areas to predict colorectal cancer survival from whole-slide images

BACKGROUND

Many methodologies for selecting histopathological images, such as sample image patches or segment histology from regions of interest (ROIs) or whole-slide images (WSIs), have been utilized to develop survival models. With gigapixel WSIs exhibiting diverse histological appearances, obtaining clinically prognostic and explainable features remains challenging. Therefore, we propose a novel deep learning-based algorithm combining tissue areas with histopathological features to predict cancer survival.

METHODS

The Cancer Genome Atlas Colon Adenocarcinoma (TCGA-COAD) dataset was used in this investigation. A deep convolutional survival model (DeepConvSurv) extracted histopathological information from the image patches of nine different tissue types, including tumors, lymphocytes, stroma, and mucus. The tissue map of the WSIs was segmented using image processing techniques that involved localizing and quantifying the tissue region. Six survival models with the concordance index (C-index) were used as the evaluation metrics.

RESULTS

We extracted 128 histopathological features from four histological types and five tissue area features from WSIs to predict colorectal cancer survival. Our method performed better in six distinct survival models than the Whole Slide Histopathological Images Survival Analysis framework (WSISA), which adaptively sampled patches using K-means from WSIs. The best performance using histopathological features was 0.679 using LASSO-Cox. Compared to histopathological features alone, tissue area features increased the C-index by 2.5%. Based on histopathological features and tissue area features, our approach achieved performance of 0.704 with RIDGE-Cox.

CONCLUSIONS

A deep learning-based algorithm combining histopathological features with tissue area proved clinically relevant and effective for predicting cancer survival.

Applicability and validation of different prognostic scores in allogeneic hematopoietic cell transplant (HCT) in the post-transplant cyclophosphamide era

We investigated the predictive capacity of six prognostic indices [Karnofsky Performance Status (KPS), Hematopoietic Cell Transplant-Specific Comorbidity Index (HCT-CI), Disease Risk Index (DRI), European Bone Marrow Transplantation (EBMT) and Revised Pre-Transplantation Assessment of Mortality (rPAM) Scores and Endothelial Activation and Stress Index (EASIX)] in 205 adults undergoing post-transplant cyclophosphamide (PTCy)-based allo-HCT. KPS, HCT-CI, DRI and EASIX grouped patients into higher and lower risk strata. KPS and EASIX maintained appropriate discrimination for OS prediction across the first 2 years after allo-HCT [receiver operating characteristic curve (area under the curve (AUC) > 55 %)]. The discriminative capacity of DRI and HCT-CI increased during the post-transplant period, with a peak of prediction at 2 years (AUC of 61.1 % and 61.8 %). The maximum rPAM discriminative capacity was at 1 year (1-year AUC of 58.2 %). The predictive capacity of the EBMT score was not demonstrated. This study validates the discrimination capacity for OS prediction of KPS, HCT-CI, DRI and EASIX in PTCy-based allo-HCT.

Re-evaluation of publicly available gene-expression databases using machine-learning yields a maximum prognostic power in breast cancer

Gene expression signatures refer to patterns of gene activities and are used to classify different types of cancer, determine prognosis, and guide treatment decisions. Advancements in high-throughput technology and machine learning have led to improvements to predict a patient's prognosis for different cancer phenotypes. However, computational methods for analyzing signatures have not been used to evaluate their prognostic power. Contention remains on the utility of gene expression signatures for prognosis. The prevalent approaches include random signatures, expert knowledge, and machine learning to construct an improved signature. We unify these approaches to evaluate their prognostic power. Re-evaluation of publicly available gene-expression data from 8 databases with 9 machine-learning models revealed previously unreported results. Gene-expression signatures are confirmed to be useful in predicting a patient's prognosis. Convergent evidence from [Formula: see text] 10,000 signatures implicates a maximum prognostic power. By calculating the concordance index, which measures how well patients with different prognoses can be discriminated, we show that a signature can correctly discriminate patients' prognoses no more than 80% of the time. Additionally, we show that more than 50% of the potentially available information is still missing at this value. We surmise that an accurate prognosis must incorporate molecular, clinical, histological, and other complementary factors.

Association of substance use and other psychiatric disorders with all-cause and external-cause mortality in individuals given community sentences in Sweden: a national cohort study

Background

Consistently high rates of premature mortality have been reported in individuals who receive community sentences. However, few studies have explored potential modifiable risk factors for these rates, particularly mental health. We examined the association of substance use and other psychiatric disorders with all-cause and external-cause mortality in individuals convicted of a criminal offence and given a community sentence.

Methods

We did a longitudinal cohort study of 109,751 individuals given community sentences in Sweden using population-based registers. We calculated mortality rates for all-cause and external-cause mortality, hazard ratios for the association between psychiatric disorders and mortality, and population attributable fractions to quantify the contribution of psychiatric disorders to mortality risk.

Findings

During the follow-up, 5749 (5.2%) individuals died, including 2709 (2.5%) from external causes. Individuals with pre-existing substance use and other psychiatric disorders had an increased mortality risk from any cause (aHR = 2.28 [95% CI 2.15-2.42]) and from external causes (3.11 [2.85-3.40]) compared to individuals without known psychiatric or substance use disorders. Suicide was the most common cause of death in younger persons.

Interpretation

In individuals given community sentences, substance use and other psychiatric disorders were associated with an increased risk of premature death with suicide being the leading cause of death. Community supervision represents an opportunity to provide sentenced individuals with access to evidence-based treatment targeting substance misuse and psychiatric disorders to prevent potentially preventable deaths.

Funding

Wellcome Trust.

Predictors of malignant transformation in oral leukoplakia and proliferative verrucous leukoplakia: An observational prospective study including the DNA ploidy status

BACKGROUND

This prospective observational study investigated the determinants of malignant transformation (MT) in localized oral leukoplakia (OL) and proliferative verrucous leukoplakia (PVL).

METHODS

Demographic, clinical, histological, and DNA ploidy status data were collected at enrolment. Survival analysis was performed (MT being the event of interest).

RESULTS

One-hundred and thirty-three patients with OL and 20 patients with PVL entered the study over 6 years (mean follow-up 7.8 years). The presence of OED, DNA ploidy, clinical presentation, and lesion site were associated with MT in patients with OL in a univariate analysis. In a multivariate model, OED was the strongest predictor of MT in patients with OL. Adding DNA ploidy increased the model's predictive power. None of the assessed predictors was associated with MT in patients with PVL.

CONCLUSIONS

DNA ploidy might identify a subset OL with low risk or minimal risk of MT, but it does not seem to be a reliable predictor in patients with PVL.

Social Determinants of Health Data Improve the Prediction of Cardiac Outcomes in Females with Breast Cancer

Cardiovascular disease is the leading cause of mortality among breast cancer (BC) patients aged 50 and above. Machine Learning (ML) models are increasingly utilized as prediction tools, and recent evidence suggests that incorporating social determinants of health (SDOH) data can enhance its performance. This study included females ≥ 18 years diagnosed with BC at any stage. The outcomes were the diagnosis and time-to-event of major adverse cardiovascular events (MACEs) within two years following a cancer diagnosis. Covariates encompassed demographics, risk factors, individual and neighborhood-level SDOH, tumor characteristics, and BC treatment. Race-specific and race-agnostic Extreme Gradient Boosting ML models with and without SDOH data were developed and compared based on their C-index. Among 4309 patients, 11.4% experienced a 2-year MACE. The race-agnostic models exhibited a C-index of 0.78 (95% CI 0.76-0.79) and 0.81 (95% CI 0.80-0.82) without and with SDOH data, respectively. In non-Hispanic Black women (NHB; n = 765), models without and with SDOH data achieved a C-index of 0.74 (95% CI 0.72-0.76) and 0.75 (95% CI 0.73-0.78), respectively. Among non-Hispanic White women (n = 3321), models without and with SDOH data yielded a C-index of 0.79 (95% CI 0.77-0.80) and 0.79 (95% CI 0.77-0.80), respectively. In summary, including SDOH data improves the predictive performance of ML models in forecasting 2-year MACE among BC females, particularly within NHB.

Polygenic Scores for Longitudinal Prediction of Incident Type 2 Diabetes in an Ancestrally and Medically Diverse Primary Care Network

OBJECTIVE

The clinical utility of genetic information for type 2 diabetes (T2D) prediction with polygenic score (PGS) in ancestrally diverse, real-world US healthcare systems is unclear, especially for those at low clinical phenotypic risk for T2D.

RESEARCH DESIGN AND METHODS

We tested the association of PGS with T2D incidence in patients followed within a primary care practice network over 16 years in four hypothetical scenarios that varied by clinical data availability (N = 14,712): 1) age and sex, 2) age, sex, BMI, systolic blood pressure, and family history of diabetes; 3) all variables in (2) and random glucose; 4) all variables in (3), HDL, total cholesterol, and triglycerides, combined in a clinical risk score (CRS). To determine whether genetic effects differed by baseline clinical risk, we tested for interaction with the CRS.

RESULTS

PGS was associated with incident diabetes in all models. Adjusting for age and sex only, the Hazard Ratio (HR) per PGS standard deviation (SD) was 1.76 (95% CI 1.68, 1.84) and the HR of top 5% of PGS vs interquartile range (IQR) was 2.80 (2.39, 3.28). Adjusting for the CRS, the HR per SD was 1.48 (1.40, 1.57) and HR of top 5% of PGS vs IQR was 2.09 (1.72, 2.55). Genetic effects differed by baseline clinical risk [(PGS-CRS interaction p =0.05; CRS below the median: HR 1.60 (1.43, 1.79); CRS above the median: HR 1.45 (1.35, 1.55)].

CONCLUSIONS

Genetic information can help identify high-risk patients even among those perceived to be low risk in a clinical evaluation.

Multicenter Survival Analysis and Application of an Olfactory Neuroblastoma Staging Modification Incorporating Hyams Grade

Importance

Current olfactory neuroblastoma (ONB) staging systems inadequately delineate locally advanced tumors, do not incorporate tumor grade, and poorly estimate survival and recurrence.

Objective

The primary aims of this study were to (1) examine the clinical covariates associated with survival and recurrence of ONB in a modern-era multicenter cohort and (2) incorporate Hyams tumor grade into existing staging systems to assess its ability to estimate survival and recurrence.

Design, Setting, and Participants

This retrospective, multicenter, case-control study included patients with ONB who underwent treatment between January 1, 2005, and December 31, 2021, at 9 North American academic medical centers.

Intervention

Standard-of-care ONB treatment.

Main Outcome and Measures

The main outcomes were overall survival (OS), disease-free survival (DFS), and disease-specific survival (DSS) as C statistics for model prediction.

Results

A total of 256 patients with ONB (mean [SD] age, 52.0 [15.6] years; 115 female [44.9%]; 141 male [55.1%]) were included. The 5-year rate for OS was 83.5% (95% CI, 78.3%-89.1%); for DFS, 70.8% (95% CI, 64.3%-78.0%); and for DSS, 94.1% (95% CI, 90.5%-97.8%). On multivariable analysis, age, American Joint Committee on Cancer (AJCC) stage, involvement of bilateral maxillary sinuses, and positive margins were associated with OS. Only AJCC stage was associated with DFS. Only N stage was associated with DSS. When assessing the ability of staging systems to estimate OS, the best-performing model was the novel modification of the Dulguerov system (C statistic, 0.66; 95% CI, 0.59-0.76), and the Kadish system performed most poorly (C statistic, 0.57; 95% CI, 0.50-0.63). Regarding estimation of DFS, the modified Kadish system performed most poorly (C statistic, 0.55; 95% CI, 0.51-0.66), while the novel modification of the AJCC system performed the best (C statistic, 0.70; 95% CI, 0.66-0.80). Regarding estimation of DSS, the modified Kadish system was the best-performing model (C statistic, 0.79; 95% CI, 0.70-0.94), and the unmodified Kadish performed the worst (C statistic, 0.56; 95% CI, 0.51-0.68). The ability for novel ONB staging systems to estimate disease progression across stages was also assessed. In the novel Kadish staging system, patients with stage VI disease were approximately 7 times as likely to experience disease progression as patients with stage I disease (hazard ratio [HR], 6.84; 95% CI, 1.60-29.20). Results were similar for the novel modified Kadish system (HR, 8.99; 95% CI, 1.62-49.85) and the novel Dulguerov system (HR, 6.86; 95% CI, 2.74-17.18).

Conclusions and Relevance

The study findings indicate that 5-year OS for ONB is favorable and that incorporation of Hyams grade into traditional ONB staging systems is associated with improved estimation of disease progression.

Decreased Expression of GLYATL1 Predicts Poor Prognosis in Patients with Clear Cell Renal Cell Carcinoma

Background

GLYATL1 is a member of the glycine-N-acyltransferase family, which catalyses acyl group transfer. The role of GLYATL1 in cancer is largely unknown; therefore, the potential value of GLYATL1 in clear cell renal cell carcinoma (ccRCC) was explored.

Methods

The ccRCC gene expression profiles and clinical data were obtained from the University of California Santa Cruz Xena platform. Differential expression and survival analysis were performed using R software. Samples from the TIMER public database and real-world were used for validation. The potential molecular mechanism of GLYATL1 in ccRCC was explored using gene set enrichment analysis (GSEA).

Results

GLYATL1 was downregulated, indicating a poor prognosis in ccRCC. Low expression of GLYATL1 was significantly associated with advanced stage and higher histological grade ccRCC. The differential expression of GLYATL1 was validated at the protein level using clinical samples and tissue microarray. The results of GSEA showed that multiple crucial signalling pathways including fatty acid metabolism, adipogenesis, oxidative phosphorylation and epithelial-mesenchymal transition were enriched.

Conclusion

This study demonstrated that GLYATL1 downregulation has an unfavourable impact on the survival of patients with ccRCC. The resulting data indicated that GLYATL1 could be a potential new target for ccRCC therapy, which may be helpful for the personalized treatment of such individuals.

Establishment and Clinical Application of the Nomogram Related to Risk or Prognosis of Hepatocellular Carcinoma: A Review

Hepatocellular carcinoma (HCC) is the most prevalent primary liver malignancy, accounting for approximately 90% of all primary liver cancers, with high mortality and a poor prognosis. A large number of predictive models have been applied that integrate multiple clinical factors and biomarkers to predict the prognosis of HCC. Nomograms, as easy-to-use prognostic predictive models, are widely used to predict the probability of clinical outcomes. We searched PubMed with the keywords "hepatocellular carcinoma" and "nomogram", and 974 relative literatures were retrieved. According to the construction methodology and the real validity of the nomograms, in this study, 97 nomograms for HCC were selected in 77 publications. These 97 nomograms were established based on more than 100,000 patients, covering seven main prognostic outcomes. The research data of 56 articles are from hospital-based HCC patients, and 13 articles provided external validation results of the nomogram. In addition to AFP, tumor size, tumor number, stage, vascular invasion, age, and other common prognostic risk factors are included in the HCC-related nomogram, more and more biomarkers, including gene mRNA expression, gene polymorphisms, and gene signature, etc. were also included in the nomograms. The establishment, assessment and validation of these nomograms are also discussed in depth. This study would help clinicians construct and select appropriate nomograms to guide precise judgment and appropriate treatments.

Improvement of a prediction model for heart failure survival through explainable artificial intelligence

Cardiovascular diseases and their associated disorder of heart failure (HF) are major causes of death globally, making it a priority for doctors to detect and predict their onset and medical consequences. Artificial Intelligence (AI) allows doctors to discover clinical indicators and enhance their diagnoses and treatments. Specifically, "eXplainable AI" (XAI) offers tools to improve the clinical prediction models that experience poor interpretability of their results. This work presents an explainability analysis and evaluation of two HF survival prediction models using a dataset that includes 299 patients who have experienced HF. The first model utilizes survival analysis, considering death events and time as target features, while the second model approaches the problem as a classification task to predict death. The model employs an optimization data workflow pipeline capable of selecting the best machine learning algorithm as well as the optimal collection of features. Moreover, different post hoc techniques have been used for the explainability analysis of the model. The main contribution of this paper is an explainability-driven approach to select the best HF survival prediction model balancing prediction performance and explainability. Therefore, the most balanced explainable prediction models are Survival Gradient Boosting model for the survival analysis and Random Forest for the classification approach with a c-index of 0.714 and balanced accuracy of 0.74 (std 0.03) respectively. The selection of features by the SCI-XAI in the two models is similar where "serum_creatinine", "ejection_fraction", and "sex" are selected in both approaches, with the addition of "diabetes" for the survival analysis model. Moreover, the application of post hoc XAI techniques also confirm common findings from both approaches by placing the "serum_creatinine" as the most relevant feature for the predicted outcome, followed by "ejection_fraction". The explainable prediction models for HF survival presented in this paper would improve the further adoption of clinical prediction models by providing doctors with insights to better understand the reasoning behind usually "black-box" AI clinical solutions and make more reasonable and data-driven decisions.

Pregnancy Zone Protein Serves as a Prognostic Marker and Favors Immune Infiltration in Lung Adenocarcinoma

Lung adenocarcinoma (LUAD) is a public enemy with a very high incidence and mortality rate, for which there is no specific detectable biomarker. Pregnancy zone protein (PZP) is an immune-related protein; however, the functions of PZP in LUAD are unclear. In this study, a series of bioinformatics methods, combined with immunohistochemistry (IHC), four-color multiplex fluorescence immunohistochemistry (mIHC), quantitative real-time PCR (qRT-PCR) and enzyme-linked immunosorbent assay (ELISA), were utilized to explore the prognostic value and potential role of PZP in LUAD. Our data revealed that PZP expression was markedly reduced in LUAD tissues, tightly correlated with clinical stage and could be an independent unfavorable prognostic factor. In addition, pathway analysis revealed that high expression of PZP in LUAD was mainly involved in immune-related molecules. Tumor immune infiltration analysis by CIBERSORT showed a significant correlation between PZP expression and several immune cell infiltrations, and IHC further confirmed a positive correlation with CD4+ T-cell infiltration and a negative correlation with CD68+ M0 macrophage infiltration. Furthermore, mIHC demonstrated that PZP expression gave rise to an increase in CD86+ M1 macrophages and a decrease in CD206+ M2 macrophages. Therefore, PZP can be used as a new biomarker for the prediction of prognosis and may be a promising immune-related molecular target for LUAD.

Spatial cellular architecture predicts prognosis in glioblastoma

Intra-tumoral heterogeneity and cell-state plasticity are key drivers for the therapeutic resistance of glioblastoma. Here, we investigate the association between spatial cellular organization and glioblastoma prognosis. Leveraging single-cell RNA-seq and spatial transcriptomics data, we develop a deep learning model to predict transcriptional subtypes of glioblastoma cells from histology images. Employing this model, we phenotypically analyze 40 million tissue spots from 410 patients and identify consistent associations between tumor architecture and prognosis across two independent cohorts. Patients with poor prognosis exhibit higher proportions of tumor cells expressing a hypoxia-induced transcriptional program. Furthermore, a clustering pattern of astrocyte-like tumor cells is associated with worse prognosis, while dispersion and connection of the astrocytes with other transcriptional subtypes correlate with decreased risk. To validate these results, we develop a separate deep learning model that utilizes histology images to predict prognosis. Applying this model to spatial transcriptomics data reveal survival-associated regional gene expression programs. Overall, our study presents a scalable approach to unravel the transcriptional heterogeneity of glioblastoma and establishes a critical connection between spatial cellular architecture and clinical outcomes.

Risk factors and survival prediction of young breast cancer patients with liver metastases: a population-based study

Background

The risk and prognosis of young breast cancer (YBC) with liver metastases (YBCLM) remain unclear. Thus, this study aimed to determine the risk and prognostic factors in these patients and construct predictive nomogram models.

Methods

This population-based retrospective study was conducted using data of YBCLM patients from the Surveillance, Epidemiology, and End Results database between 2010 and 2019. Multivariate logistic and Cox regression analyses were used to identify independent risk and prognostic factors, which were used to construct the diagnostic and prognostic nomograms. The concordance index (C-index), calibration plot, receiver operating characteristic (ROC) curve, and decision curve analysis (DCA) were used to assess the performances of the established nomogram models. Propensity score matching (PSM) analysis was used to balance the baseline characteristics between the YBCLM patients and non-young patients with BCLM when comparing overall survival (OS) and cancer-specific survival (CSS).

Results

A total of 18,275 YBC were identified, of whom 400 had LM. T stage, N stage, molecular subtypes, and bone, lung, and brain metastases were independent risk factors for LM developing in YBC. The established diagnostic nomogram showed that bone metastases contributed the most risk of LM developing, with a C-index of 0.895 (95% confidence interval 0.877-0.913) for this nomogram model. YBCLM had better survival than non-young patients with BCLM in unmatched and matched cohorts after propensity score matching analysis. The multivariate Cox analysis demonstrated that molecular subtypes, surgery and bone, lung, and brain metastases were independently associated with OS and CSS, chemotherapy was an independent prognostic factor for OS, and marital status and T stage were independent prognostic factors for CSS. The C-indices for the OS- and CSS-specific nomograms were 0.728 (0.69-0.766) and 0.74 (0.696-0.778), respectively. The ROC analysis indicated that these models had excellent discriminatory power. The calibration curve also showed that the observed results were consistent with the predicted results. DCA showed that the developed nomogram models would be effective in clinical practice.

Conclusion

The present study determined the risk and prognostic factors of YBCLM and further developed nomograms that can be used to effectively identify high-risk patients and predict survival outcomes.

Elevated Body Mass Index Is Associated with Improved Overall Survival in Castration-Resistant Prostate Cancer Patients Undergoing Prostate-Specific Membrane Antigen-Directed Radioligand Therapy

In patients with prostate cancer scheduled for systemic treatment, being overweight is linked to prolonged overall survival (OS), whereas sarcopenia is associated with shorter OS. We investigated fat-related and body composition parameters in patients undergoing prostate-specific membrane antigen (PSMA)-directed radioligand therapy (RLT) to assess their predictive value for OS. Methods: Body mass index (BMI, in kg/m²) and CT-derived body composition parameters (total, subcutaneous, visceral fat area, and psoas muscle area at the L3-L4 level) were determined for 171 patients scheduled for PSMA-directed RLT. After normalization for stature, the psoas muscle index was used to define sarcopenia. Outcome analysis was performed using Kaplan-Meier curves and Cox regression including fat-related and other clinical parameters (Gleason score, C-reactive protein [CRP], lactate dehydrogenase [LDH], hemoglobin, and prostate-specific antigen levels). The Harrell C-index was used for goodness-of-fit analysis. Results: Sixty-five patients (38%) had sarcopenia, and 98 patients (57.3%) had increased BMI. Relative to the 8-mo OS in normal-weight men (BMI < 25), overweight men (25 ≥ BMI > 30) and obese men (BMI ≥ 30) achieved a longer OS of 14 mo (hazard ratio [HR], 0.63; 95% CI, 0.40-0.99; P = 0.03) and 13 mo (HR, 0.47; 95% CI, 0.29-0.77; P = 0.004), respectively. Sarcopenia showed no impact on OS (11 vs. 12 mo; HR, 1.4; 95% CI, 0.91-2.1; P = 0.09). Most of the body composition parameters were tightly linked to OS on univariable analyses, with the highest C-index for BMI. In multivariable analysis, a higher BMI (HR, 0.91; 95% CI, 0.86-0.97; P = 0.006), lower CRP (HR, 1.09; 95% CI, 1.03-1.14; P < 0.001), lower LDH (HR, 1.08; 95% CI, 1.03-1.14; P < 0.001), and longer interval between initial diagnosis and RLT (HR, 0.95; 95% CI, 0.91-0.99; P = 0.02) were significant predictors of OS. Conclusion: Increased fat reserves assessed by BMI, CRP, LDH, and interval between initial diagnosis and RLT, but not CT-derived body composition parameters, were relevant predictors for OS. As BMI can be altered, future research should investigate whether a high-calorie diet before or during PSMA RLT may improve OS.

Radiologic Assessment of Lung Edema Score as a Predictor of Clinical Outcome in Children with Acute Respiratory Distress Syndrome

PURPOSE

The radiographic assessment of lung edema (RALE) score enables objective quantification of lung edema and is a valuable prognostic marker of adult acute respiratory distress syndrome (ARDS). We aimed to evaluate the validity of RALE score in children with ARDS.

MATERIALS AND METHODS

The RALE score was measured for its reliability and correlation to other ARDS severity indices. ARDS-specific mortality was defined as death from severe pulmonary dysfunction or the need for extracorporeal membrane oxygenation therapy. The C-index of the RALE score and other ARDS severity indices were compared via survival analyses.

RESULTS

Among 296 children with ARDS, 88 did not survive, and there were 70 ARDS-specific non-survivors. The RALE score showed good reliability with an intraclass correlation coefficient of 0.809 [95% confidence interval (CI), 0.760-0.848]. In univariable analysis, the RALE score had a hazard ratio (HR) of 1.19 (95% CI, 1.18-3.11), and the significance was maintained in multivariable analysis adjusting with age, ARDS etiology, and comorbidity, with an HR of 1.77 (95% CI, 1.05-2.91). The RALE score was a good predictor of ARDS-specific mortality, with a C-index of 0.607 (95% CI, 0.519-0.695).

CONCLUSION

The RALE score is a reliable measure for ARDS severity and a useful prognostic marker of mortality in children, especially for ARDS-specific mortality. This score provides information that clinicians can use to decide the proper time of aggressive therapy targeting severe lung injury and to appropriately manage the fluid balance of children with ARDS.

Should lethal arrhythmias in hypertrophic cardiomyopathy be predicted using non-electrophysiological methods?

While sudden cardiac death (SCD) in hypertrophic cardiomyopathy (HCM) is due to arrhythmias, the guidelines for prediction of SCD are based solely on non-electrophysiological methods. This study aims to stimulate thinking about whether the interests of patients with HCM are better served by using current, 'risk factor', methods of prediction or by further development of electrophysiological methods to determine arrhythmic risk. Five published predictive studies of SCD in HCM, which contain sufficient data to permit analysis, were analysed to compute receiver operating characteristics together with their confidence bounds to compare their formal prediction either by bootstrapping or Monte Carlo analysis. Four are based on clinical risk factors, one with additional MRI analysis, and were regarded as exemplars of the risk factor approach. The other used an electrophysiological method and directly compared this method to risk factors in the same patients. Prediction methods that use conventional clinical risk factors and MRI have low predictive capacities that will only detect 50-60% of patients at risk with a 15-30% false positive rate [area under the curve (AUC) = ∼0.7], while the electrophysiological method detects 90% of events with a 20% false positive rate (AUC = ∼0.89). Given improved understanding of complex arrhythmogenesis, arrhythmic SCD is likely to be more accurately predictable using electrophysiologically based approaches as opposed to current guidelines and should drive further development of electrophysiologically based methods.