A community challenge to predict clinical outcomes after immune checkpoint blockade in non-small cell lung cancer.

BACKGROUND: Predictive biomarkers of immune checkpoint inhibitor (ICI) efficacy are currently lacking for non-small cell lung cancer (NSCLC). Here, we describe the results from the Anti-PD-1 Response Prediction DREAM Challenge, a crowdsourced initiative that enabled the assessment of predictive models by using data from two randomized controlled clinical trials (RCTs) of ICIs in first-line metastatic NSCLC. METHODS: Participants developed and trained models using public resources. These were evaluated with data from the CheckMate 026 trial (NCT02041533), according to the model-to-data paradigm to maintain patient confidentiality. The generalizability of the models with the best predictive performance was assessed using data from the CheckMate 227 trial (NCT02477826). Both trials were phase III RCTs with a chemotherapy control arm, which supported the differentiation between predictive and prognostic models. Isolated model containers were evaluated using a bespoke strategy that considered the challenges of handling transcriptome data from clinical trials. RESULTS: A total of 59 teams participated, with 417 models submitted. Multiple predictive models, as opposed to a prognostic model, were generated for predicting overall survival, progression-free survival, and progressive disease status with ICIs. Variables within the models submitted by participants included tumor mutational burden (TMB), programmed death ligand 1 (PD-L1) expression, and gene-expression-based signatures. The best-performing models showed improved predictive power over reference variables, including TMB or PD-L1. CONCLUSIONS: This DREAM Challenge is the first successful attempt to use protected phase III clinical data for a crowdsourced effort towards generating predictive models for ICI clinical outcomes and could serve as a blueprint for similar efforts in other tumor types and disease states, setting a benchmark for future studies aiming to identify biomarkers predictive of ICI efficacy. TRIAL REGISTRATION: CheckMate 026; NCT02041533, registered January 22, 2014. CheckMate 227; NCT02477826, registered June 23, 2015.

A harmonized resource of integrated prostate cancer clinical, -omic, and signature features.

Genomic and transcriptomic data have been generated across a wide range of prostate cancer (PCa) study cohorts. These data can be used to better characterize the molecular features associated with clinical outcomes and to test hypotheses across multiple, independent patient cohorts. In addition, derived features, such as estimates of cell composition, risk scores, and androgen receptor (AR) scores, can be used to develop novel hypotheses leveraging existing multi-omic datasets. The full potential of such data is yet to be realized as independent datasets exist in different repositories, have been processed using different pipelines, and derived and clinical features are often not provided or  not standardized. Here, we present the curatedPCaData R package, a harmonized data resource representing >2900 primary tumor, >200 normal tissue, and >500 metastatic PCa samples across 19 datasets processed using standardized pipelines with updated gene annotations. We show that meta-analysis across harmonized studies has great potential for robust and clinically meaningful insights. curatedPCaData is an open and accessible community resource with code made available for reproducibility.

OSCAR: Optimal subset cardinality regression using the L0-pseudonorm with applications to prognostic modelling of prostate cancer.

In many real-world applications, such as those based on electronic health records, prognostic prediction of patient survival is based on heterogeneous sets of clinical laboratory measurements. To address the trade-off between the predictive accuracy of a prognostic model and the costs related to its clinical implementation, we propose an optimized L0-pseudonorm approach to learn sparse solutions in multivariable regression. The model sparsity is maintained by restricting the number of nonzero coefficients in the model with a cardinality constraint, which makes the optimization problem NP-hard. In addition, we generalize the cardinality constraint for grouped feature selection, which makes it possible to identify key sets of predictors that may be measured together in a kit in clinical practice. We demonstrate the operation of our cardinality constraint-based feature subset selection method, named OSCAR, in the context of prognostic prediction of prostate cancer patients, where it enables one to determine the key explanatory predictors at different levels of model sparsity. We further explore how the model sparsity affects the model accuracy and implementation cost. Lastly, we demonstrate generalization of the presented methodology to high-dimensional transcriptomics data.

Blood Cell Responses Following Heavy Alcohol Consumption Coincide with Changes in Acute Phase Reactants of Inflammation, Indices of Hemolysis and Immune Responses to Ethanol Metabolites.

Aberrations in blood cells are common among heavy alcohol drinkers. In order to shed further light on such responses, we compared blood cell status with markers of hemolysis, mediators of inflammation and immune responses to ethanol metabolites in alcohol-dependent patients at the time of admission for detoxification and after abstinence. Blood cell counts, indices of hemolysis (LDH, haptoglobin, bilirubin), calprotectin (a marker of neutrophil activation), suPAR, CD163, pro- and anti-inflammatory cytokines and autoantibodies against protein adducts with acetaldehyde, the first metabolite of ethanol, were measured from alcohol-dependent patients (73 men, 26 women, mean age 43.8 ± 10.4 years) at baseline and after 8 ± 1 days of abstinence. The assessments also included information on the quantities of alcohol drinking and assays for biomarkers of alcohol consumption (CDT), liver function (AST, ALT, ALP, GGT) and acute phase reactants of inflammation. At baseline, the patients showed elevated values of CDT and biomarkers of liver status, which decreased significantly during abstinence. A significant decrease also occurred in LDH, bilirubin, CD163 and IgA and IgM antibodies against acetaldehyde adducts, whereas a significant increase was noted in blood leukocytes, platelets, MCV and suPAR levels. The changes in blood leukocytes correlated with those in serum calprotectin (p < 0.001), haptoglobin (p < 0.001), IL-6 (p < 0.02) and suPAR (p < 0.02). The changes in MCV correlated with those in LDH (p < 0.02), MCH (p < 0.01), bilirubin (p < 0.001) and anti-adduct IgG (p < 0.01). The data indicates that ethanol-induced changes in blood leukocytes are related with acute phase reactants of inflammation and release of neutrophil calprotectin. The studies also highlight the role of hemolysis and immune responses to ethanol metabolites underlying erythrocyte abnormalities in alcohol abusers.

Impact of Physical Activity on the Characteristics and Metabolic Consequences of Alcohol Consumption: A Cross-Sectional Population-Based Study.

Sedentary lifestyle and excessive alcohol drinking are major modifiable risk factors of health. In order to shed further light on the relationships between physical activity and health consequences of alcohol intake, we measured biomarkers of liver function, inflammation, lipid status and fatty liver index tests in a large population-based sample of individuals with different levels of physical activity, alcohol drinking and other lifestyle risk factors. The study included 21,050 adult participants (9940 men, 11,110 women) (mean age 48.2 ± 13.3 years) of the National FINRISK Study. Data on physical activity, alcohol drinking, smoking and body weight were recorded. The participants were classified to subgroups according to gender, levels of physical activity (sedentary, low, moderate, vigorous, extreme), alcohol drinking levels (abstainers, moderate drinkers, heavy drinkers) and patterns (regular or binge, types of beverages preferred in consumption). Serum liver enzymes (GGT, ALT), C-reactive protein (CRP) and lipid profiles were measured using standard laboratory techniques. Physical activity was linearly and inversely related with the amount of alcohol consumption, with the lowest alcohol drinking levels being observed in those with vigorous or extreme activity (p < 0.0005). Physically active individuals were less frequently binge-type drinkers, cigarette smokers or heavy coffee drinkers than those with sedentary activity (p < 0.0005 for linear trend in all comparisons). In the General Linear Model to assess the main and interaction effects of physical activity and alcohol consumption on biomarker status, as adjusted for anthropometric measures, smoking and coffee consumption, increasing levels of physical activity were found to be associated with more favorable findings on serum GGT (p < 0.0005), ALT (p < 0.0005 for men), cholesterol (p = 0.025 for men; p < 0.0005 for women), HDL-cholesterol (p < 0.0005 for men, p = 0.001 for women), LDL-cholesterol (p < 0.03 for men), triglycerides (p < 0.0005 for men, p < 0.03 for women), CRP (p < 0.0005 for men, p = 0.006 for women) and fatty liver index (p < 0.0005). The data support the view that regular moderate to vigorous physical activity may counteract adverse metabolic consequences of alcohol consumption on liver function, inflammation and lipid status. The role of physical activity should be further emphasized in interventions aimed at reducing health problems related to unfavorable risk factors of lifestyle.

Alcohol Drinking Patterns and Laboratory Indices of Health: Does Type of Alcohol Preferred Make a Difference?

Although excessive alcohol consumption is a highly prevalent public health problem the data on the associations between alcohol consumption and health outcomes in individuals preferring different types of alcoholic beverages has remained unclear. We examined the relationships between the amounts and patterns of drinking with the data on laboratory indices of liver function, lipid status and inflammation in a national population-based health survey (FINRISK). Data on health status, alcohol drinking, types of alcoholic beverages preferred, body weight, smoking, coffee consumption and physical activity were recorded from 22,432 subjects (10,626 men, 11,806 women), age range 25-74 years. The participants were divided to subgroups based on the amounts of regular alcohol intake (abstainers, moderate and heavy drinkers), patterns of drinking (binge or regular) and the type of alcoholic beverage preferred (wine, beer, cider or long drink, hard liquor or mixed). Regular drinking was found to be more typical in wine drinkers whereas the subjects preferring beer or hard liquor were more often binge-type drinkers and cigarette smokers. Alcohol use in all forms was associated with increased frequencies of abnormalities in the markers of liver function, lipid status and inflammation even at rather low levels of consumption. The highest rates of abnormalities occurred, however, in the subgroups of binge-type drinkers preferring beer or hard liquor. These results demonstrate that adverse consequences of alcohol occur even at moderate average drinking levels especially in individuals who engage in binge drinking and in those preferring beer or hard liquor. Further emphasis should be placed on such patterns of drinking in policies aimed at preventing alcohol-induced adverse health outcomes.

Tumor microenvironment as a metapopulation model: The effects of angiogenesis, emigration and treatment modalities.

Tumors consist of heterogeneous cell subpopulations that may develop differing phenotypes, such as increased cell growth, metastatic potential and treatment sensitivity or resistance. To study the dynamics of cancer development at a single-cell level, we model the tumor microenvironment as a metapopulation, in which habitat patches correspond to possible sites for cell subpopulations. Cancer cells may emigrate into dispersal pool (e.g. circulation system) and spread to new sites (i.e. metastatic disease). In the patches, cells divide and new variants may arise, possibly leading into an invasion provided the aberration promotes the cell growth. To study such adaptive landscape of cancer ecosystem, we consider various evolutionary strategies (phenotypes), such as emigration and angiogenesis, which are important determinants during early stages of tumor development. We use the metapopulation fitness of new variants to investigate how these strategies evolve through natural selection and disease progression. We further study various treatment effects and investigate how different therapy regimens affect the evolution of the cell populations. These aspects are relevant, for example, when examining the dynamic process of a benign tumor becoming cancerous, and what is the best treatment strategy during the early stages of cancer development. It is shown that positive angiogenesis promotes cancer cell growth in the absence of anti-angiogenic treatment, and that the anti-angiogenic treatment reduces the need of cytotoxic treatment when used in a combination. Interestingly, the model predicts that treatment resistance might become a favorable quality to cancer cells when the anti-angiogenic treatment is intensive enough. Thus, the optimal treatment dosage should remain below a patient-specific level to avoid treatment resistance.

Modelling of killer T-cell and cancer cell subpopulation dynamics under immuno- and chemotherapies.

Each patient's cancer has a unique molecular makeup, often comprised of distinct cancer cell subpopulations. Improved understanding of dynamic processes between cancer cell populations is therefore critical for making treatment more effective and personalized. It has been shown that immunotherapy increases the survival of melanoma patients. However, there remain critical open questions, such as timing and duration of immunotherapy and its added benefits when combined with other types of treatments. We introduce a model for the dynamics of active killer T-cells and cancer cell subpopulations. Rather than defining the cancer cell populations based on their genetic makeup alone, we consider also other, non-genetic differences that make the cell populations either sensitive or resistant to a therapy. Using the model, we make predictions of possible outcomes of the various treatment strategies in virtual melanoma patients, providing hypotheses regarding therapeutic efficacy and side-effects. It is shown, for instance, that starting immunotherapy with a denser treatment schedule may enable changing to a sparser schedule later during the treatment. Furthermore, combination of targeted and immunotherapy results in a better treatment effect, compared to mono-immunotherapy, and a stable disease can be reached with a patient-tailored combination. These results offer better understanding of the competition between T-cells and cancer cells, toward personalized immunotherapy regimens.

Cost-effective survival prediction for patients with advanced prostate cancer using clinical trial and real-world hospital registry datasets.

INTRODUCTION: Predictive survival modeling offers systematic tools for clinical decision-making and individualized tailoring of treatment strategies to improve patient outcomes while reducing overall healthcare costs. In 2015, a number of machine learning and statistical models were benchmarked in the DREAM 9.5 Prostate Cancer Challenge, based on open clinical trial data for metastatic castration resistant prostate cancer (mCRPC). However, applying these models into clinical practice poses a practical challenge due to the inclusion of a large number of model variables, some of which are not routinely monitored or are expensive to measure. OBJECTIVES: To develop cost-specified variable selection algorithms for constructing cost-effective prognostic models of overall survival that still preserve sufficient model performance for clinical decision making. METHODS: Penalized Cox regression models were used for the survival prediction. For the variable selection, we implemented two algorithms: (i) LASSO regularization approach; and (ii) a greedy cost-specified variable selection algorithm. The models were compared in three cohorts of mCRPC patients from randomized clinical trials (RCT), as well as in a real-world cohort (RWC) of advanced prostate cancer patients treated at the Turku University Hospital. Hospital laboratory expenses were utilized as a reference for computing the costs of introducing new variables into the models. RESULTS: Compared to measuring the full set of clinical variables, economic costs could be reduced by half without a significant loss of model performance. The greedy algorithm outperformed the LASSO-based variable selection with the lowest tested budgets. The overall top performance was higher with the LASSO algorithm. CONCLUSION: The cost-specified variable selection offers significant budget optimization capability for the real-world survival prediction without compromising the predictive power of the model.