Towards a survival risk prediction model for metastatic NSCLC patients on durvalumab using whole-lung CT radiomics.

Background: Existing criteria for predicting patient survival from immunotherapy are primarily centered on the PD-L1 status of patients. We tested the hypothesis that noninvasively captured baseline whole-lung radiomics features from CT images, baseline clinical parameters, combined with advanced machine learning approaches, can help to build models of patient survival that compare favorably with PD-L1 status for predicting 'less-than-median-survival risk' in the metastatic NSCLC setting for patients on durvalumab. With a total of 1062 patients, inclusive of model training and validation, this is the largest such study yet. Methods: To ensure a sufficient sample size, we combined data from treatment arms of three metastatic NSCLC studies. About 80% of this data was used for model training, and the remainder was held-out for validation. We first trained two independent models; Model-C trained to predict survival using clinical data; and Model-R trained to predict survival using whole-lung radiomics features. Finally, we created Model-C+R which leveraged both clinical and radiomics features. Results: The classification accuracy (for median survival) of Model-C, Model-R, and Model-C+R was 63%, 55%, and 68% respectively. Sensitivity analysis of survival prediction across different training and validation cohorts showed concordance indices ([95 percentile]) of 0.64 ([0.63, 0.65]), 0.60 ([0.59, 0.60]), and 0.66 ([0.65,0.67]), respectively. We additionally evaluated generalization of these models on a comparable cohort of 144 patients from an independent study, demonstrating classification accuracies of 65%, 62%, and 72% respectively. Conclusion: Machine Learning models combining baseline whole-lung CT radiomic and clinical features may be a useful tool for patient selection in immunotherapy. Further validation through prospective studies is needed.

Identifying Patient Populations in Texts Describing Drug Approvals Through Deep Learning-Based Information Extraction: Development of a Natural Language Processing Algorithm.

BACKGROUND: New drug treatments are regularly approved, and it is challenging to remain up-to-date in this rapidly changing environment. Fast and accurate visualization is important to allow a global understanding of the drug market. Automation of this information extraction provides a helpful starting point for the subject matter expert, helps to mitigate human errors, and saves time. OBJECTIVE: We aimed to semiautomate disease population extraction from the free text of oncology drug approval descriptions from the BioMedTracker database for 6 selected drug targets. More specifically, we intended to extract (1) line of therapy, (2) stage of cancer of the patient population described in the approval, and (3) the clinical trials that provide evidence for the approval. We aimed to use these results in downstream applications, aiding the searchability of relevant content against related drug project sources. METHODS: We fine-tuned a state-of-the-art deep learning model, Bidirectional Encoder Representations from Transformers, for each of the 3 desired outputs. We independently applied rule-based text mining approaches. We compared the performances of deep learning and rule-based approaches and selected the best method, which was then applied to new entries. The results were manually curated by a subject matter expert and then used to train new models. RESULTS: The training data set is currently small (433 entries) and will enlarge over time when new approval descriptions become available or if a choice is made to take another drug target into account. The deep learning models achieved 61% and 56% 5-fold cross-validated accuracies for line of therapy and stage of cancer, respectively, which were treated as classification tasks. Trial identification is treated as a named entity recognition task, and the 5-fold cross-validated F1-score is currently 87%. Although the scores of the classification tasks could seem low, the models comprise 5 classes each, and such scores are a marked improvement when compared to random classification. Moreover, we expect improved performance as the input data set grows, since deep learning models need to be trained on a large enough amount of data to be able to learn the task they are taught. The rule-based approach achieved 60% and 74% 5-fold cross-validated accuracies for line of therapy and stage of cancer, respectively. No attempt was made to define a rule-based approach for trial identification. CONCLUSIONS: We developed a natural language processing algorithm that is currently assisting subject matter experts in disease population extraction, which supports health authority approvals. This algorithm achieves semiautomation, enabling subject matter experts to leverage the results for deeper analysis and to accelerate information retrieval in a crowded clinical environment such as oncology.

Interim recruitment prediction for multi-center clinical trials.

We introduce a general framework for monitoring, modeling, and predicting the recruitment to multi-center clinical trials. The work is motivated by overly optimistic and narrow prediction intervals produced by existing time-homogeneous recruitment models for multi-center recruitment. We first present two tests for detection of decay in recruitment rates, together with a power study. We then introduce a model based on the inhomogeneous Poisson process with monotonically decaying intensity, motivated by recruitment trends observed in oncology trials. The general form of the model permits adaptation to any parametric curve-shape. A general method for constructing sensible parameter priors is provided and Bayesian model averaging is used for making predictions which account for the uncertainty in both the parameters and the model. The validity of the method and its robustness to misspecification are tested using simulated datasets. The new methodology is then applied to oncology trial data, where we make interim accrual predictions, comparing them to those obtained by existing methods, and indicate where unexpected changes in the accrual pattern occur.

Statistical and practical considerations in designing of immuno-oncology trials.

The novel mechanism of action of immunotherapy agents, in treatment of various types of cancer, poses unique challenges during the designing of clinical trials. It is important to account for possibility of a delayed treatment effect and adjust sample size accordingly. This paper provides an analytical approach for computing sample size in the presence of a delayed effect using a piece-wise proportional hazards model. Failing to account for an anticipated treatment delay may result in considerable loss in power. The overall hazard ratio (HR), which now represents the average HR across the entire treatment period, can remain a meaningful measure of average benefit to patients in the trial. We show that, special consideration needs to be given for the designing of interim analyses related to futility, so as not to increase the probability of incorrectly stopping an effective agent. It is shown that the weighted log-rank test, using the Fleming-Harrington class of weights, can be used as supportive analysis to better reflect the impact of a delayed effect and possible long-term benefit in a subset of the overall population.

Biomarker data visualisation for decision making in clinical trials.

OBJECTIVE: To understand how visualization of biomarker data is used for decision making in clinical trials, and identify problems with and suggest improvements to this process. METHODS: We carried out semi-structured interviews with 18 professionals involved in various aspects of developing or using visualizations of biomarker data for decision making in clinical trials. We used an inductive thematic analysis to identify implicit and explicit ideas within the data captured from the interviews. RESULTS: We identified 6 primary themes, including: how visualizations were used in clinical trials; the importance of having a clear understanding of the underlying data; the purpose or use of the visualization, and the properties of the visualizations themselves. The results show that participants' 'trust' in the visualization depends on access to the underlying data, and that there is currently no standard or straightforward way to support this access. CONCLUSIONS: Incorporating information about data provenance into biomarker-related visualizations used for decision making in clinical trials may increase users' trust, and therefore facilitate the decision making process.

Distinguishing prognostic and predictive biomarkers: an information theoretic approach.

Motivation: The identification of biomarkers to support decision-making is central to personalized medicine, in both clinical and research scenarios. The challenge can be seen in two halves: identifying predictive markers, which guide the development/use of tailored therapies; and identifying prognostic markers, which guide other aspects of care and clinical trial planning, i.e. prognostic markers can be considered as covariates for stratification. Mistakenly assuming a biomarker to be predictive, when it is in fact largely prognostic (and vice-versa) is highly undesirable, and can result in financial, ethical and personal consequences. We present a framework for data-driven ranking of biomarkers on their prognostic/predictive strength, using a novel information theoretic method. This approach provides a natural algebra to discuss and quantify the individual predictive and prognostic strength, in a self-consistent mathematical framework. Results: Our contribution is a novel procedure, INFO+, which naturally distinguishes the prognostic versus predictive role of each biomarker and handles higher order interactions. In a comprehensive empirical evaluation INFO+ outperforms more complex methods, most notably when noise factors dominate, and biomarkers are likely to be falsely identified as predictive, when in fact they are just strongly prognostic. Furthermore, we show that our methods can be 1-3 orders of magnitude faster than competitors, making it useful for biomarker discovery in 'big data' scenarios. Finally, we apply our methods to identify predictive biomarkers on two real clinical trials, and introduce a new graphical representation that provides greater insight into the prognostic and predictive strength of each biomarker. Availability and implementation: R implementations of the suggested methods are available at https://github.com/sechidis. Supplementary information: Supplementary data are available at Bioinformatics online.

Predictors of enhanced response with benralizumab for patients with severe asthma: pooled analysis of the SIROCCO and CALIMA studies.

BACKGROUND: Benralizumab is an anti-eosinophilic, anti-interleukin-5 receptor α monoclonal antibody that has been shown to significantly reduce asthma exacerbations and improve lung function for patients with severe, uncontrolled asthma. We further explored the efficacy of benralizumab for patients with different baseline blood eosinophil thresholds and exacerbation histories. METHODS: This study is a pooled analysis of the results from the randomised, double-blind, placebo-controlled SIROCCO (NCT01928771) and CALIMA (NCT01914757) phase 3 studies. In these studies, patients with severe, uncontrolled asthma were randomly assigned (1:1:1) to receive subcutaneous benralizumab 30 mg, either every 4 weeks or every 8 weeks (with first three doses given every 4 weeks), or placebo every 4 weeks. The primary endpoint was annual exacerbation rate (AER) ratio versus placebo, analysed by baseline eosinophil counts (≥0, ≥150, ≥300, or ≥450 cells per µL) and by number of exacerbations (two vs three or more) during the year before enrolment. The analyses were done in accordance with the intention-to-treat principle. FINDINGS: Of 2295 patients, 756 received benralizumab every 4 weeks, 762 received benralizumab every 8 weeks, and 777 patients received placebo. AER among patients with baseline blood eosinophil counts of at least 0 cells per µL was 1·16 (95% CI 1·05-1·28) in patients who received placebo versus 0·75 (0·66-0·84) in patients who received benralizumab every 8 weeks (rate ratio 0·64, 0·55-0·75; p<0·0001). In patients who received benralizumab every 4 weeks who had eosinophil counts of 0 or more cells per µL, AER was 0·73 (0·65-0·82); rate ratio versus placebo was 0·63 (0·54-0·74; p<0·0001). The extent to which exacerbation rates were reduced increased with increasing blood eosinophil thresholds and with greater exacerbation history in patients in the 4-weekly and 8-weekly benralizumab groups. Greater improvements in AER were seen with benralizumab compared with placebo for patients with a combination of high blood eosinophil thresholds and a history of more frequent exacerbations. INTERPRETATION: These results will help to guide clinicians when they are deciding whether to use benralizumab to treat patients with severe, uncontrolled, eosinophilic asthma. FUNDING: AstraZeneca.

Disentangling Prognostic and Predictive Biomarkers Through Mutual Information.

We study information theoretic methods for ranking biomarkers. In clinical trials, there are two, closely related, types of biomarkers: predictive and prognostic, and disentangling them is a key challenge. Our first step is to phrase biomarker ranking in terms of optimizing an information theoretic quantity. This formalization of the problem will enable us to derive rankings of predictive/prognostic biomarkers, by estimating different, high dimensional, conditional mutual information terms. To estimate these terms, we suggest efficient low dimensional approximations. Finally, we introduce a new visualisation tool that captures the prognostic and the predictive strength of a set of biomarkers. We believe this representation will prove to be a powerful tool in biomarker discovery.

Data Science Solution to Event Prediction in Outsourced Clinical Trial Models.

Late phase clinical trials are regularly outsourced to a Contract Research Organisation (CRO) while the risk and accountability remain within the sponsor company. Many statistical tasks are delivered by the CRO and later revalidated by the sponsor. Here, we report a technological approach to standardised event prediction. We have built a dynamic web application around an R-package with the aim of delivering reliable event predictions, simplifying communication and increasing trust between the CRO and the in-house statisticians via transparency. Short learning curve, interactivity, reproducibility and data diagnostics are key here. The current implementation is motivated by time-to-event prediction in oncology. We demonstrate a clear benefit of standardisation for both parties. The tool can be used for exploration, communication, sensitivity analysis and generating standard reports. At this point we wish to present this tool and share some of the insights we have gained during the development.

A multicenter validation of recombinant ß3 integrin-coupled beads to detect human platelet antigen-1 alloantibodies in 498 cases of fetomaternal alloimmune thrombocytopenia.

BACKGROUND: Fetomaternal alloimmune thrombocytopenia (FMAIT) is caused by human platelet (PLT) antigen (HPA) incompatibility. Beads coupled with recombinant ß3 integrins, displaying the biallelic HPA-1 epitopes (rHPA-1), have been shown to detect HPA-1a alloantibodies implicated in FMAIT. This report describes a multicenter validation of the beads using the results of well-characterized samples to define the optimum parameters for analysis of a large cohort of 498 clinical samples. STUDY DESIGN AND METHODS: Fifty-one blinded quality assurance (QA) samples were tested by six laboratories to standardize the rHPA-1 bead assay and to develop an algorithm for sample classification. Five laboratories retrieved samples from 498 independent FMAIT cases, previously tested by the monoclonal antibody-specific immobilization of PLT antigens (MAIPA) assay, from their local archives for testing with the rHPA-1 beads. The results were evaluated using a mathematical algorithm developed to classify the samples. RESULTS: The QA samples gave a mean concordance of 94% between the bead and MAIPA assays, while 97% concordance was observed with the FMAIT samples. Of the 15 discrepant samples, seven were positive by the beads but negative by MAIPA, while the contrary was observed for eight samples. Overall, the bead assay achieved 98% sensitivity for HPA-1a antibody detection in FMAIT and 98.7% specificity compared to the local MAIPA. CONCLUSION: The rHPA-1 bead assay is a rapid 3-hour assay for the sensitive detection of HPA-1 antibodies. Its ease of use would enable prompt detection of maternal HPA-1a antibodies in suspected FMAIT cases, which is important supportive evidence for treatment by transfusion with HPA-1b1b PLTs.

Validation of genotyping protocols for diagnostic use.

This chapter contains some advice for first validating a new genotyping method and then continuing to ensure that it performs well when adopted for routine use. It should be applicable to most of the genotyping methods described in this volume; however, because the range of techniques is so wide and these recommendations have been written with PCR amplification in mind, the reader may need to adapt some of the recommendations for non-PCR based tests.It does not include any advice on how to handle unscreened human samples within the laboratory, and it assumes that the laboratory already has established good practice for DNA amplification and the avoidance of contamination.

Establishment of the first WHO international genetic reference panel for Prader Willi and Angelman syndromes.

Prader Willi and Angelman syndromes are clinically distinct genetic disorders both mapping to chromosome region 15q11-q13, which are caused by a loss of function of paternally or maternally inherited genes in the region, respectively. With clinical diagnosis often being difficult, particularly in infancy, confirmatory genetic diagnosis is essential to enable clinical intervention. However, the latter is challenged by the complex genetics behind both disorders and the unmet need for characterised reference materials to aid accurate molecular diagnosis. With this in mind, a panel of six genotyping reference materials for Prader Willi and Angelman syndromes was developed, which should be stable for many years and available to all diagnostic laboratories. The panel comprises three Prader Willi syndrome materials (two with different paternal deletions, and one with maternal uniparental disomy (UPD)) and three Angelman syndrome materials (one with a maternal deletion, one with paternal UPD or an epigenetic imprinting centre defect, and one with a UBE3A point mutation). Genomic DNA was bulk-extracted from Epstein-Barr virus-transformed lymphoblastoid cell lines established from consenting patients, and freeze-dried as aliquots in glass ampoules. In total, 37 laboratories from 26 countries participated in a collaborative study to assess the suitability of the panel. Participants evaluated the blinded, triplicate materials using their routine diagnostic methods against in-house controls or externally sourced uncertified reference materials. The panel was established by the Expert Committee on Biological Standardization of the World Health Organization as the first International Genetic Reference Panel for Prader Willi and Angelman syndromes.

Preparation and validation of the first WHO international genetic reference panel for Fragile X syndrome.

Fragile X syndrome is the most common inherited form of mental retardation. It is caused by expansion of a trinucleotide (CGG)n repeat sequence in the 5' untranslated region of the FMR1 gene, resulting in promoter hypermethylation and suppression of FMR1 transcription. Additionally, pre-mutation alleles in carrier males and females may result in Fragile X tremor ataxia syndrome and primary ovarian insufficiency, respectively. Fragile X is one of the most commonly requested molecular genetic tests worldwide. Quality assessment schemes have identified a wide disparity in allele sizing between laboratories. It is therefore important that clinical laboratories have access to characterized reference materials (RMs) to aid accurate allele sizing and diagnosis. With this in mind, a panel of genotyping RMs for Fragile X syndrome has been developed, which should be stable over many years and available to all diagnostic laboratories. Immortalized cell lines were produced by Epstein-Barr virus transformation of lymphocytes from consenting patients. Genomic DNA was extracted in bulk and RM aliquots were freeze-dried in glass ampoules. Twenty-one laboratories from seventeen countries participated in a collaborative study to assess their suitability. Participants evaluated the samples (blinded, in triplicate) in their routine methods alongside in-house and commercial controls. The panel of five genomic DNA samples was endorsed by the European Society of Human Genetics and approved as an International Standard by the Expert Committee on Biological Standardization at the World Health Organization.

Detection of human platelet antigen-1a alloantibodies in cases of fetomaternal alloimmune thrombocytopenia using recombinant ß3 integrin fragments coupled to fluorescently labeled beads.

BACKGROUND: Testing for alloantibodies against human platelet antigens (HPAs) is essential for the clinical diagnosis of fetomaternal alloimmune thrombocytopenia (FMAIT), posttransfusion purpura, and platelet (PLT) refractoriness. Most of the methods currently used for HPA alloantibody detection rely on the availability of panels of HPA-typed PLTs and some rely on validated monoclonal antibodies (MoAbs) against the PLT glycoproteins. Recombinant ß3 integrins displaying the HPA-1a (rHPA-1a) or HPA-1b (rHPA-1b) epitopes have been produced as an alternative source of antigen. The suitability of these integrin fragments was evaluated for the development of an HPA-1a alloantibody screening assay, using Luminex xMAP technology. STUDY DESIGN AND METHODS: A 3-plex bead assay was developed by coupling biotinylated rHPA-1a, rHPA-1b, and recombinant glycoprotein VI to LumAvidin microspheres. Forty patient samples referred for FMAIT diagnostic testing, which were previously screened by the MoAb-specific immobilization of PLT antigens (MAIPA) assay, were used to assess the assay. RESULTS: The rHPA-1a- and rHPA-1b-coupled beads were able to detect HPA-1a and HPA-1b alloantibodies in all patient samples tested that were previously confirmed to contain HPA-1-specific antibodies. Furthermore, HLA Class I antibodies did not cross-react with the coupled beads. CONCLUSION: The 3-plex bead assay can be used to detect HPA-1a antibodies with sufficient specificity and sensitivity for use in the clinical setting of FMAIT. The development of other recombinant integrin fragments with the use of Luminex xMAP technology may assist in providing more rapid HPA antibody detection, enabling prompt diagnosis of alloimmune PLT disorders.

Establishment of the first World Health Organization International Genetic Reference Panel for quantitation of BCR-ABL mRNA.

Serial quantitation of BCR-ABL mRNA levels is an important indicator of therapeutic response for patients with chronic myelogenous leukemia and Philadelphia chromosome-positive acute lymphoblastic leukemia, but there is substantial variation in the real-time quantitative polymerase chain reaction methodologies used by different testing laboratories. To help improve the comparability of results between centers we sought to develop accredited reference reagents that are directly linked to the BCR-ABL international scale. After assessment of candidate cell lines, a reference material panel comprising 4 different dilution levels of freeze-dried preparations of K562 cells diluted in HL60 cells was prepared. After performance evaluation, the materials were assigned fixed percent BCR-ABL/control gene values according to the International Scale. A recommendation that the 4 materials be established as the first World Health Organization International Genetic Reference Panel for quantitation of BCR-ABL translocation by real-time quantitative polymerase chain reaction was approved by the Expert Committee on Biological Standardization of the World Health Organization in November 2009. We consider that the development of these reagents is a significant milestone in the standardization of this clinically important test, but because they are a limited resource we suggest that their availability is restricted to manufacturers of secondary reference materials.

Challenges in the prediction and modeling of oral absorption and bioavailability.

To predict the performance of a drug following oral dosing, a thorough understanding of the dissolution, uptake and metabolism of the compound is required. In this review, approaches to in silico modeling of these processes are discussed. Although oral absorption, which is limited by dissolution and passive permeation, is to some extent predictable, bioavailability, which is influenced by first-pass metabolism in the intestines and liver, is much more difficult to predict. Much of the difficulty in predicting oral absorption and bioavailability is in the experimental quantification of solubility in the gastrointestinal tract lumen, membrane permeability, plasma protein binding, metabolism and active transport, rather than the formulating of the mathematical models.