Individualized survival prediction for patients with oropharyngeal cancer in the human papillomavirus era

Advancing Head and Neck Cancer Survival Prediction via Multi-Label Learning and Deep Model Interpretation

A comprehensive and reliable survival prediction model is of great importance to assist in the personalized management of Head and Neck Cancer (HNC) patient treated with curative Radiation Therapy (RT). In this work, we propose IMLSP, an Interpretable Multi-Label multi-modal deep Survival Prediction framework for predicting multiple HNC survival outcomes simultaneously and provide time-event specific visual explanation of the deep prediction process. We adopt Multi-Task Logistic Regression (MTLR) layers to convert survival prediction from a regression problem to a multi-time point classification task, and to enable predicting of multiple relevant survival outcomes at the same time. We also present Grad-Team, a Gradient-weighted Time-event activation mapping approach specifically developed for deep survival model visual explanation, to generate patient-specific time-to-event activation maps. We evaluate our method with the publicly available RADCURE HNC dataset, where it outperforms the corresponding single-modal models and single-label models on all survival outcomes. The generated activation maps show that the model focuses primarily on the tumor and nodal volumes when making the decision and the volume of interest varies for high- and low-risk patients. We demonstrate that the multi-label learning strategy can improve the learning efficiency and prognostic performance, while the interpretable survival prediction model is promising to help understand the decision-making process of AI and facilitate personalized treatment. The project website can be found at https://github.com/***.

Vision Transformer-Based Multilabel Survival Prediction for Oropharynx Cancer After Radiation Therapy

PURPOSE

A reliable and comprehensive cancer prognosis model for oropharyngeal cancer (OPC) could better assist in personalizing treatment. In this work, we developed a vision transformer-based (ViT-based) multilabel model with multimodal input to learn complementary information from available pretreatment data and predict multiple associated endpoints for radiation therapy for patients with OPC.

METHODS AND MATERIALS

A publicly available data set of 512 patients with OPC was used for both model training and evaluation. Planning computed tomography images, primary gross tumor volume masks, and 16 clinical variables representing patient demographics, diagnosis, and treatment were used as inputs. To extract deep image features with global attention, we used a ViT module. Clinical variables were concatenated with the learned image features and fed into fully connected layers to incorporate cross-modality features. To learn the mapping between the features and correlated survival outcomes, including overall survival, local failure-free survival, regional failure-free survival, and distant failure-free survival, we employed 4 multitask logistic regression layers. The proposed model was optimized by combining the multitask logistic regression negative-log likelihood losses of different prediction targets.

RESULTS

We employed the C-index and area under the curve metrics to assess the performance of our model for time-to-event prediction and time-specific binary prediction, respectively. Our proposed model outperformed corresponding single-modality and single-label models on all prediction labels, achieving C-indices of 0.773, 0.765, 0.776, and 0.773 for overall survival, local failure-free survival, regional failure-free survival, and distant failure-free survival, respectively. The area under the curve values ranged between 0.799 and 0.844 for different tasks at different time points. Using the medians of predicted risks as the thresholds to identify high-risk and low-risk patient groups, we performed the log-rank test, the results of which showed significantly larger separations in different event-free survivals.

CONCLUSION

We developed the first model capable of predicting multiple labels for OPC simultaneously. Our model demonstrated better prognostic ability for all the prediction targets compared with corresponding single-modality models and single-label models.

Comparing Individualized Survival Predictions From Random Survival Forests and Multistate Models in the Presence of Missing Data: A Case Study of Patients With Oropharyngeal Cancer

Background

In recent years, interest in prognostic calculators for predicting patient health outcomes has grown with the popularity of personalized medicine. These calculators, which can inform treatment decisions, employ many different methods, each of which has advantages and disadvantages.

Methods

We present a comparison of a multistate model (MSM) and a random survival forest (RSF) through a case study of prognostic predictions for patients with oropharyngeal squamous cell carcinoma. The MSM is highly structured and takes into account some aspects of the clinical context and knowledge about oropharyngeal cancer, while the RSF can be thought of as a black-box non-parametric approach. Key in this comparison are the high rate of missing values within these data and the different approaches used by the MSM and RSF to handle missingness.

Results

We compare the accuracy (discrimination and calibration) of survival probabilities predicted by both approaches and use simulation studies to better understand how predictive accuracy is influenced by the approach to (1) handling missing data and (2) modeling structural/disease progression information present in the data. We conclude that both approaches have similar predictive accuracy, with a slight advantage going to the MSM.

Conclusions

Although the MSM shows slightly better predictive ability than the RSF, consideration of other differences are key when selecting the best approach for addressing a specific research question. These key differences include the methods' ability to incorporate domain knowledge, and their ability to handle missing data as well as their interpretability, and ease of implementation. Ultimately, selecting the statistical method that has the most potential to aid in clinical decisions requires thoughtful consideration of the specific goals.

Identification of an individualized therapy prognostic signature for head and neck squamous cell carcinoma

BACKGROUND

Head and neck squamous cell carcinoma (HNSCC) are the most common cancers in the head and neck. Therapeutic response-related genes (TRRGs) are closely associated with carcinogenesis and prognosis in HNSCC. However, the clinical value and prognostic significance of TRRGs are still unclear. We aimed to construct a prognostic risk model to predict therapy response and prognosis in TRRGs-defined subgroups of HNSCC.

METHODS

The multiomics data and clinical information of HNSCC patients were downloaded from The Cancer Genome Atlas (TCGA). The profile data GSE65858 and GSE67614 chip was downloaded from public functional genomics data Gene Expression Omnibus (GEO). Based on TCGA-HNSC database, patients were divided into a remission group and a non-remission group according to therapy response, and differentially expressed TRRGs between those two groups were screened. Using Cox regression analysis and Least absolute shrinkage and selection operator (LASSO) analysis, candidate TRRGs that can predict the prognosis of HNSCC were identified and used to construct a TRRGs-based signature and a prognostic nomogram.

RESULT

A total of 1896 differentially expressed TRRGs were screened, including 1530 upregulated genes and 366 downregulated genes. Then, 206 differently expressed TRRGs that was significantly associated with the survival were chosen using univariate Cox regression analysis. Finally, a total of 20 candidate TRRGs genes were identified by LASSO analysis to establish a signature for risk prediction, and the risk score of each patient was calculated. Patients were divided into a high-risk group (Risk-H) and a low-risk group (Risk-L) based on the risk score. Results showed that the Risk-L patients had better overall survival (OS) than Risk-H patients. Receiver operating characteristic (ROC) curve analysis revealed great predictive performance for 1-, 3-, and 5-year OS in TCGA-HNSC and GEO databases. Moreover, for patients treated with post-operative radiotherapy, Risk-L patients had longer OS and lower recurrence than Risk-H patients. The nomogram involves risk score and other clinical factors had good performance in predicting survival probability.

CONCLUSIONS

The proposed risk prognostic signature and Nomogram based on TRRGs are novel promising tools for predicting therapy response and overall survival in HNSCC patients.

Head and neck cancer predictive risk estimator to determine control and therapeutic outcomes of radiotherapy (HNC-PREDICTOR): development, international multi-institutional validation, and web implementation of clinic-ready model-based risk stratification for head and neck cancer

BACKGROUND

Personalised radiotherapy can improve treatment outcomes of patients with head and neck cancer (HNC), where currently a 'one-dose-fits-all' approach is the standard. The aim was to establish individualised outcome prediction based on multi-institutional international 'big-data' to facilitate risk-based stratification of patients with HNC.

METHODS

The data of 4611 HNC radiotherapy patients from three academic cancer centres were split into four cohorts: a training (n = 2241), independent test (n = 786), and external validation cohorts 1 (n = 1087) and 2 (n = 497). Tumour- and patient-related clinical variables were considered in a machine learning pipeline to predict overall survival (primary end-point) and local and regional tumour control (secondary end-points); serially, imaging features were considered for optional model improvement. Finally, patients were stratified into high-, intermediate-, and low-risk groups.

RESULTS

Performance score, AJCC8thstage, pack-years, and Age were identified as predictors for overall survival, demonstrating good performance in both the training cohort (c-index = 0.72 [95% CI, 0.66-0.77]) and in all three validation cohorts (c-indices: 0.76 [0.69-0.83], 0.73 [0.68-0.77], and 0.75 [0.68-0.80]). Excellent stratification of patients with HNC into high, intermediate, and low mortality risk was achieved; with 5-year overall survival rates of 17-46% for the high-risk group compared to 92-98% for the low-risk group. The addition of morphological image feature further improved the performance (c-index = 0.73 [0.64-0.81]). These models are integrated in a clinic-ready interactive web interface: https://uic-evl.github.io/hnc-predictor/ CONCLUSIONS: Robust model-based prediction was able to stratify patients with HNC in distinct high, intermediate, and low mortality risk groups. This can effectively be capitalised for personalised radiotherapy, e.g., for tumour radiation dose escalation/de-escalation.

Predicting local persistence/recurrence after radiation therapy for head and neck cancer from PET/CT using a multi-objective, multi-classifier radiomics model

Objectives

Accurate identifying head and neck squamous cell cancer (HNSCC) patients at high risk of local persistence/recurrence (P/R) is of importance for personalized patient management. Here we developed a multi-objective, multi-classifier radiomics model for early HNSCC local P/R prediction based on post-treatment PET/CT scans and clinical data.

Materials and methods

We retrospectively identified 328 individuals (69 patients have local P/R) with HNSCC treated with definitive radiation therapy at our institution. The median follow-up from treatment completion to the first surveillance PET/CT imaging was 114 days (range: 82-159 days). Post-treatment PET/CT scans were reviewed and contoured for all patients. For each imaging modality, we extracted 257 radiomic features to build a multi-objective radiomics model with sensitivity, specificity, and feature sparsity as objectives for model training. Multiple representative classifiers were combined to construct the predictive model. The output probabilities of models built with features from various modalities were fused together to make the final prediction.

Results

We built and evaluated three single-modality models and two multi-modality models. The combination of PET, CT, and clinical data in the multi-objective, multi-classifier radiomics model trended towards the best prediction performance, with a sensitivity of 93%, specificity of 83%, accuracy of 85%, and AUC of 0.94.

Conclusion

Our study demonstrates the feasibility of employing a multi-objective, multi-classifier radiomics model with PET/CT radiomic features and clinical data to predict outcomes for patients with HNSCC after radiation therapy. The proposed prediction model shows the potential to detect cancer local P/R early after radiation therapy.

Locoregional recurrence prediction in head and neck cancer based on multi-modality and multi-view feature expansion

Objective. Locoregional recurrence (LRR) is one of the leading causes of treatment failure in head and neck (H&N) cancer. Accurately predicting LRR after radiotherapy is essential to achieving better treatment outcomes for patients with H&N cancer through developing personalized treatment strategies. We aim to develop an end-to-end multi-modality and multi-view feature extension method (MMFE) to predict LRR in H&N cancer. Approach. Deep learning (DL) has been widely used for building prediction models and has achieved great success. Nevertheless, 2D-based DL models inherently fail to utilize the contextual information from adjacent slices, while complicated 3D models have a substantially larger number of parameters, which require more training samples, memory and computing resources. In the proposed MMFE scheme, through the multi-view feature expansion and projection dimension reduction operations, we are able to reduce the model complexity while preserving volumetric information. Additionally, we designed a multi-modality convolutional neural network that can be trained in an end-to-end manner and can jointly optimize the use of deep features of CT, PET and clinical data to improve the model’s prediction ability. Main results. The dataset included 206 eligible patients, of which, 49 had LRR while 157 did not. The proposed MMFE method obtained a higher AUC value than the other four methods. The best prediction result was achieved when using all three modalities, which yielded an AUC value of 0.81. Significance. Comparison experiments demonstrated the superior performance of the MMFE as compared to other 2D/3D-DL-based methods. By combining CT, PET and clinical features, the MMFE could potentially identify H&N cancer patients at high risk for LRR such that personalized treatment strategy can be developed accordingly.

Development of a prognostic model of overall survival in oropharyngeal cancer from real-world data: PRO.M.E.THE.O

Objective

The PRO.M.E.THE.O. study (PredictiOn Models in Ent cancer for anti-EGFR based THErapy Optimization) aimed to develop a predictive model (PM) of overall survival (OS) for patients with locally advanced oropharyngeal cancer (LAOC) treated with radiotherapy (RT) and cetuximab (Cet) from an Italian dataset.

Methods

We enrolled patients with LAOC from 6 centres treated with RT-Cet. Clinical and treatment variables were collected. Patients were randomly divided into training (TS) (80%) and validation (VS) (20%) sets. A binary logistic regression model was used on the TS with stepwise feature selection and then on VS. Timepoints of 2, 3 and 5 years were considered. The area under the curve (AUC) of receiver operating characteristic of 2, 3 and 5 year and confusion matrix statistics at 5-threshold were used as performance criteria.

Results

Overall, 218 patients were enrolled and 174 (79.8%) were analysed. Age at diagnosis, gender, ECOG performance, clinical stage, dose to high-risk volume, overall treatment time and day of RT interruption were considered in the final PMs. The PMs were developed and represented by nomograms with AUC of 0.75, 0.73 and 0.73 for TS and 0.713, 0.713, 0.775 for VS at 2, 3 and 5 years, respectively.

Conclusions

PRO.M.E.THE.O. allows the creation of a PM for OS in patients with LAOC treated with RT-Cet.

Accounting for not-at-random missingness through imputation stacking

Not-at-random missingness presents a challenge in addressing missing data in many health research applications. In this article, we propose a new approach to account for not-at-random missingness after multiple imputation through weighted analysis of stacked multiple imputations. The weights are easily calculated as a function of the imputed data and assumptions about the not-at-random missingness. We demonstrate through simulation that the proposed method has excellent performance when the missingness model is correctly specified. In practice, the missingness mechanism will not be known. We show how we can use our approach in a sensitivity analysis framework to evaluate the robustness of model inference to different assumptions about the missingness mechanism, and we provide R package StackImpute to facilitate implementation as part of routine sensitivity analyses. We apply the proposed method to account for not-at-random missingness in human papillomavirus test results in a study of survival for patients diagnosed with oropharyngeal cancer.

Development and Assessment of a Model for Predicting Individualized Outcomes in Patients With Oropharyngeal Cancer

IMPORTANCE

Recent insights into the biologic characteristics and treatment of oropharyngeal cancer may help inform improvements in prognostic modeling. A bayesian multistate model incorporates sophisticated statistical techniques to provide individualized predictions of survival and recurrence outcomes for patients with newly diagnosed oropharyngeal cancer.

OBJECTIVE

To develop a model for individualized survival, locoregional recurrence, and distant metastasis prognostication for patients with newly diagnosed oropharyngeal cancer, incorporating clinical, oncologic, and imaging data.

DESIGN, SETTING, AND PARTICIPANTS

In this prognostic study, a data set was used comprising 840 patients with newly diagnosed oropharyngeal cancer treated at a National Cancer Institute-designated center between January 2003 and August 2016; analysis was performed between January 2019 and June 2020. Using these data, a bayesian multistate model was developed that can be used to obtain individualized predictions. The prognostic performance of the model was validated using data from 447 patients treated for oropharyngeal cancer at Erasmus Medical Center in the Netherlands.

EXPOSURES

Clinical/oncologic factors and imaging biomarkers collected at or before initiation of first-line therapy.

MAIN OUTCOMES AND MEASURES

Overall survival, locoregional recurrence, and distant metastasis after first-line cancer treatment.

RESULTS

Of the 840 patients included in the National Cancer Institute-designated center, 715 (85.1%) were men and 268 (31.9%) were current smokers. The Erasmus Medical Center cohort comprised 300 (67.1%) men, with 350 (78.3%) current smokers. Model predictions for 5-year overall survival demonstrated good discrimination, with area under the curve values of 0.81 for the model with and 0.78 for the model without imaging variables. Application of the model without imaging data in the independent Dutch validation cohort resulted in an area under the curve of 0.75. This model possesses good calibration and stratifies patients well in terms of likely outcomes among many competing events.

CONCLUSIONS AND RELEVANCE

In this prognostic study, a multistate model of oropharyngeal cancer incorporating imaging biomarkers appeared to estimate and discriminate locoregional recurrence from distant metastases. Providing personalized predictions of multiple outcomes increases the information available for patients and clinicians. The web-based application designed in this study may serve as a useful tool for generating predictions and visualizing likely outcomes for a specific patient.

Potential Added Value of PET/CT Radiomics for Survival Prognostication beyond AJCC 8th Edition Staging in Oropharyngeal Squamous Cell Carcinoma

Accurate risk-stratification can facilitate precision therapy in oropharyngeal squamous cell carcinoma (OPSCC). We explored the potential added value of baseline positron emission tomography (PET)/computed tomography (CT) radiomic features for prognostication and risk stratification of OPSCC beyond the American Joint Committee on Cancer (AJCC) 8th edition staging scheme. Using institutional and publicly available datasets, we included OPSCC patients with known human papillomavirus (HPV) status, without baseline distant metastasis and treated with curative intent. We extracted 1037 PET and 1037 CT radiomic features quantifying lesion shape, imaging intensity, and texture patterns from primary tumors and metastatic cervical lymph nodes. Utilizing random forest algorithms, we devised novel machine-learning models for OPSCC progression-free survival (PFS) and overall survival (OS) using “radiomics” features, “AJCC” variables, and the “combined” set as input. We designed both single- (PET or CT) and combined-modality (PET/CT) models. Harrell’s C-index quantified survival model performance; risk stratification was evaluated in Kaplan–Meier analysis. A total of 311 patients were included. In HPV-associated OPSCC, the best “radiomics” model achieved an average C-index ± standard deviation of 0.62 ± 0.05 (p = 0.02) for PFS prediction, compared to 0.54 ± 0.06 (p = 0.32) utilizing “AJCC” variables. Radiomics-based risk-stratification of HPV-associated OPSCC was significant for PFS and OS. Similar trends were observed in HPV-negative OPSCC. In conclusion, radiomics imaging features extracted from pre-treatment PET/CT may provide complimentary information to the current AJCC staging scheme for survival prognostication and risk-stratification of HPV-associated OPSCC.

Epigenetic prediction of complex traits and mortality in a cohort of individuals with oropharyngeal cancer

BACKGROUND

DNA methylation (DNAm) variation is an established predictor for several traits. In the context of oropharyngeal cancer (OPC), where 5-year survival is ~ 65%, DNA methylation may act as a prognostic biomarker. We examined the accuracy of DNA methylation biomarkers of 4 complex exposure traits (alcohol consumption, body mass index [BMI], educational attainment and smoking status) in predicting all-cause mortality in people with OPC.

RESULTS

DNAm predictors of alcohol consumption, BMI, educational attainment and smoking status were applied to 364 individuals with OPC in the Head and Neck 5000 cohort (HN5000; 19.6% of total OPC cases in the study), followed up for median 3.9 years; inter-quartile range (IQR) 3.3 to 5.2 years (time-to-event-death or censor). The proportion of phenotypic variance explained in each trait was as follows: 16.5% for alcohol consumption, 22.7% for BMI, 0.4% for educational attainment and 51.1% for smoking. We then assessed the relationship between each DNAm predictor and all-cause mortality using Cox proportional-hazard regression analysis. DNAm prediction of smoking was most consistently associated with mortality risk (hazard ratio [HR], 1.38 per standard deviation (SD) increase in smoking DNAm score; 95% confidence interval [CI] 1.04 to 1.83; P 0.025, in a model adjusted for demographic, lifestyle, health and biological variables). Finally, we examined the accuracy of each DNAm predictor of mortality. DNAm predictors explained similar levels of variance in mortality to self-reported phenotypes. Receiver operator characteristic (ROC) curves for the DNAm predictors showed a moderate discrimination of alcohol consumption (area under the curve [AUC] 0.63), BMI (AUC 0.61) and smoking (AUC 0.70) when predicting mortality. The DNAm predictor for education showed poor discrimination (AUC 0.57). Z tests comparing AUCs between self-reported phenotype ROC curves and DNAm score ROC curves did not show evidence for difference between the two (alcohol consumption P 0.41, BMI P 0.62, educational attainment P 0.49, smoking P 0.19).

CONCLUSIONS

In the context of a clinical cohort of individuals with OPC, DNAm predictors for smoking, alcohol consumption, educational attainment and BMI exhibit similar predictive values for all-cause mortality compared to self-reported data. These findings may have translational utility in prognostic model development, particularly where phenotypic data are not available.

Hippo Pathway and YAP Signaling Alterations in Squamous Cancer of the Head and Neck

Head and neck cancer affects the upper aerodigestive tract and is the sixth leading cancer worldwide by incidence and the seventh by cause of death. Despite significant advances in surgery and chemotherapy, molecularly targeted therapeutic options for this type of cancer are scarce and long term survival rates remain low. Recently, comprehensive genomic studies have highlighted the most commonly altered genes and signaling pathways in this cancer. The Hippo-YAP pathway has been identified as a key oncogenic pathway in multiple tumors. Expression of genes controlled by the Hippo downstream transcriptional coactivators YAP (Yes-associated protein 1) and TAZ (WWTR1, WW domain containing transcription regulator 1) is widely deregulated in human cancer including head and neck squamous cell carcinoma (HNSCC). Interestingly, YAP/TAZ signaling might not be as essential for the normal homeostasis of adult tissues as for oncogenic growth, altogether making the pathway an amenable therapeutic target in cancer. Recent advances in the role of Hippo-YAP pathway in HNSCC have provided evidence that genetic alterations frequent in this type of cancer such as PIK3CA (phosphatidylinositide 3-kinase catalytic subunit alpha) overexpression or FAT1 (FAT atypical cadherin 1) functional loss can result in YAP activation. We discuss current therapeutic options targeting this pathway which are currently in use for other tumor types.