scholarly journals Predictive modeling of clinical trial terminations using feature engineering and embedding learning

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Magdalyn E. Elkin ◽  
Xingquan Zhu

AbstractIn this study, we propose to use machine learning to understand terminated clinical trials. Our goal is to answer two fundamental questions: (1) what are common factors/markers associated to terminated clinical trials? and (2) how to accurately predict whether a clinical trial may be terminated or not? The answer to the first question provides effective ways to understand characteristics of terminated trials for stakeholders to better plan their trials; and the answer to the second question can direct estimate the chance of success of a clinical trial in order to minimize costs. By using 311,260 trials to build a testbed with 68,999 samples, we use feature engineering to create 640 features, reflecting clinical trial administration, eligibility, study information, criteria etc. Using feature ranking, a handful of features, such as trial eligibility, trial inclusion/exclusion criteria, sponsor types etc., are found to be related to the clinical trial termination. By using sampling and ensemble learning, we achieve over 67% Balanced Accuracy and over 0.73 AUC (Area Under the Curve) scores to correctly predict clinical trial termination, indicating that machine learning can help achieve satisfactory prediction results for clinical trial study.

PLoS ONE ◽  
2021 ◽  
Vol 16 (7) ◽  
pp. e0253789
Author(s):  
Magdalyn E. Elkin ◽  
Xingquan Zhu

As of March 30 2021, over 5,193 COVID-19 clinical trials have been registered through Clinicaltrial.gov. Among them, 191 trials were terminated, suspended, or withdrawn (indicating the cessation of the study). On the other hand, 909 trials have been completed (indicating the completion of the study). In this study, we propose to study underlying factors of COVID-19 trial completion vs. cessation, and design predictive models to accurately predict whether a COVID-19 trial may complete or cease in the future. We collect 4,441 COVID-19 trials from ClinicalTrial.gov to build a testbed, and design four types of features to characterize clinical trial administration, eligibility, study information, criteria, drug types, study keywords, as well as embedding features commonly used in the state-of-the-art machine learning. Our study shows that drug features and study keywords are most informative features, but all four types of features are essential for accurate trial prediction. By using predictive models, our approach achieves more than 0.87 AUC (Area Under the Curve) score and 0.81 balanced accuracy to correctly predict COVID-19 clinical trial completion vs. cessation. Our research shows that computational methods can deliver effective features to understand difference between completed vs. ceased COVID-19 trials. In addition, such models can also predict COVID-19 trial status with satisfactory accuracy, and help stakeholders better plan trials and minimize costs.


Blood ◽  
2019 ◽  
Vol 134 (Supplement_1) ◽  
pp. 5864-5864
Author(s):  
Amany R. Keruakous ◽  
Adam S. Asch

Background: Clinical trials, key elements of the processes that account for many of the recent advances in cancer care, are becoming more complex and challenging to conduct. The Stephenson Cancer Center (SCC) has been the lead accruer to NCI-LAP trials over the past three years, and in addition, fields investigator initiated and industry sponsored trials. To identify opportunities for continued improvement in clinical trial enrolment, we sought to identify the obstacles encountered by our clinical trial staff in these activities. Method: We conducted a survey of our research staff including all research nurses and disease site coordinators who participate in recruitment, screening, consenting, data collection and compliance. The survey, sent by email to the clinical trial list-serve at SCC (90 staff member), invited respondents to enumerate obstacles to patient participation in clinical trials. We then performed a follow up meeting with our research coordinators to clarify responses. A total of 26 responses from 90 respondents were received and tabulated by disease site. Results: The most commonly reported obstacles to enrolment were, in descending order: communication/language barriers, cultural bias, time/procedure commitment, and complexity of the trial protocol, financial logistics, comorbidities, and stringent trial criteria. Respondents identified 83 obstacles as frequently encountered obstacles to enrolment. The 83 reported obstacles were classified into 9 categories and organized by disease site as presented in tabular format (below). The most commonly identified obstacles to patient enrolment were communication and language barriers. In patients for whom Spanish is the primary language this was a universal obstacle, as there is a lack of consistent Spanish consents across the clinical trial portfolio. Cultural bias, as an obstacle was manifested as a general mistrust by prospective trial participants of experimental therapies and clinical trials. After communication and cultural bias as barriers, travel requirements and the associated expenses playing a role in patients from rural areas were identified as the most commonly encountered barrier. The complexity of trial protocols and the associated large number of clinic visits, frequent laboratory and imaging tests were also identified as common obstacles. Clinical trial complexity with strict inclusion and exclusion criteria and trial-specified biopsies were frequently cited. Implications: In this descriptive study, common barriers to patient enrolment in clinical trials were identified by clinical trial staff. Assessing barriers encountered by clinical trial staff is infrequently used as a metric for improving clinical trial enrolment, but provides important perspective. In our study, some obstacles are inherent in our patient populations, others appear to be actionable. Development of Spanish language consents and specific programs to overcome negative bias regarding clinical trials are potential areas for improvement. The complexity of clinical trial protocols and the increasingly strict inclusion/exclusion criteria, are issues that will require consideration and action at the level of the cooperative groups and industry. Disclosures No relevant conflicts of interest to declare.


2013 ◽  
Vol 31 (15_suppl) ◽  
pp. 9560-9560 ◽  
Author(s):  
Amy Pickar Abernethy ◽  
Ryan David Nipp ◽  
David Currow ◽  
Nathan Cherny ◽  
Florian Strasser ◽  
...  

9560 Background: BSC as a control arm in clinical trials is poorly defined. A systematic review was conducted to evaluate clinical trial concordance with published, consensus-based framework for BSC delivery in trials. Methods: A consensus-based Delphi panel previously identified 4 key domains of BSC delivery in trials: multidisciplinary care; supportive care documentation; symptom assessment at least as often as the intervention arm; and guideline-based symptom management. A systematic review of trials including BSC control arms assessed BSC concordance to the consensus-based domains. Databases were searched from 2002-2012 using search strings: “cancer”; “best supportive care”; “randomized” or “random allocation”; and “supportive” or “palliative.” Exclusion criteria were: no BSC arm, non-human trial, not randomized, not English, not advanced cancer, or not including anticancer therapy. Data were independently extracted by 2 reviewers and scored by 4 reviewers for conformance with consensus-based BSC framework. Results: 373 articles were retrieved, 17 retained after applying exclusion criteria. Overall, trials conformed to <18% of the consensus-based BSC standards. 35% of articles offered a detailed description of BSC. 65% reported baseline and regular symptom assessment, and 47% reported using validated symptom assessment measures. 35% reported symptom assessment at identical intervals in both experimental and BSC arms. None listed an evidence-based guideline for symptom management. None of the multicenter trials reported standardization of BSC across sites. No studies reported educating patients on symptom management or goals of anti-cancer therapy. No studies reported offering access to palliative care specialists, social workers, financial or spiritual counseling. Conclusions: Reporting of BSC in trials is incomplete, resulting in uncertain internal and external validity. Such poorly defined interventions and variation between sites is unacceptable for other aspects of a clinical trial. Unless it is truly best supportive care, such studies may risk systematically over-estimating the clinical effect of the comparator arms. Standardization of a BSC delivery framework is needed to improve trial design and data generalization.


Spinal Cord ◽  
2006 ◽  
Vol 45 (3) ◽  
pp. 222-231 ◽  
Author(s):  
M H Tuszynski ◽  
J D Steeves ◽  
J W Fawcett ◽  
D Lammertse ◽  
M Kalichman ◽  
...  

2017 ◽  
Vol 35 (15_suppl) ◽  
pp. 10031-10031
Author(s):  
Catalina Hernandez Torres ◽  
Winson Y. Cheung ◽  
Christopher J. O'Callaghan ◽  
Tina Hsu

10031 Background: Older adults (OA) age 65+ make up to 60% of all newly diagnosed cancers. However, only 22-32% of patients accrued in cooperative group studies in the 1990s were age 65+. In 2003, several studies suggested that clinical trial design, in particular the presence of strict exclusion criteria, was a major barrier to accrual of OA. The objective of this study was to determine: 1) whether there has been an improvement in accrual of OA to clinical trials led by the Canadian Cancer Trials Group (CCTG) over time; 2) clinical trial features associated with accrual of OA to clinical trials 3) whether exclusion criteria in trials initiated 2003 or after have been relaxed. Methods: All completed randomized Phase II and III CCTG-led clinical trials initiated between 1990 or later were included. Trial characteristics including tumor type, stage, treatment type, and exclusion/inclusion criteria, as well as percentage of OA age 65+ accrued were recorded. Association between percentage of OA accrued and trial characteristics were compared using the Wilcoxin rank sum test. Assessment of exclusion criteria before and after 2003 was compared using the Chi Square test or Fisher exact test. Results: A total of 68 trials were included. Most trials were phase III (73%), chemotherapy trials (48%), opened before 2003 (70.6%), advanced disease (73%) and lung cancer was the most common tumour site (17.6%). OA accrual remains low compared to OA diagnosed with cancer in Canada (41% vs. 56%, p < 0.001). There was an improvement in accrual of OA after 2003 (47.1% vs. 34.9%, p = 0.02). Tumour site, early stage disease, more restrictive performance status, requiring a new biopsy, and having a longer consent form, were associated with lower accrual of OA (p < 0.05). There was no significant loosening of exclusion with time though patients with pulmonary comorbidities were more likely to be excluded in studies initiated in 2003 or later (p = 0.006). Conclusions: OA remain under-represented in clinical trials. There has been no relaxing of exclusion criteria; however, exclusion based on comorbidities was not significantly associated with under accrual of OA in our study.


2020 ◽  
Vol 38 (15_suppl) ◽  
pp. e14100-e14100
Author(s):  
Arushi Khurana ◽  
Raphael Mwangi ◽  
Grzegorz S. Nowakowski ◽  
Thomas Matthew Habermann ◽  
Stephen M. Ansell ◽  
...  

e14100 Background: Only 3-5% of adult cancer patients in the US enroll in clinical trials. Patients with organ dysfunction are often excluded from clinical trials, regardless of specific drug metabolism or relative function of the organ. The ASCO and the US FDA recommend modernizing criteria related to baseline organ function and comorbidities. In hematological malignancies, often the disease itself is the reason for organ function derangement. In order to better inform clinical trial eligibility and improve participation in the future, we evaluated the impact of baseline organ function on the potential eligibility for clinical trial enrollment for real world patients with newly diagnosed DLBCL. Methods: Consecutive, newly diagnosed lymphoma patients were offered enrollment from 2002-2015 into the Molecular Epidemiology Resource (MER) of the University of Iowa/Mayo Clinic Lymphoma Specialized Program of Research Excellence. This analysis is based on 1270 DLBCL patients receiving immunochemotherapy. Baseline organ function parameters were identified from the exclusion criteria for hemoglobin, absolute neutrophil count (ANC), platelet count, creatinine, and bilirubin reported in recent frontline trials in DLBCL (Table). Abstracted clinical labs from the MER were used to determine the percent of patients that would be excluded based on the criteria. Results: We determined that 11-23% of MER DLBCL patients receiving standard of care frontline therapy would have been excluded in the various trials utilizing baseline organ function alone (Table). Hemoglobin and renal function had the greatest impact on exclusion. Conclusions: Current national and international (phase II and III) trials are excluding up to 23% of patients from clinical trial participation based on organ function alone in DLBCL. These data will be useful in future clinical trial development to meet ASCO recommendations to increase trial accrual, while balancing the drug toxicities and patient safety. An online tool was developed based on these results to aid future trial development. [Table: see text]


Stroke ◽  
2020 ◽  
Vol 51 (Suppl_1) ◽  
Author(s):  
Masaki Ito ◽  
Satoshi Kuroda ◽  
Hidetsugu Asanoi ◽  
Taku Sugiyama ◽  
Takafumi Shindo ◽  
...  

Background: Outcomes of stroke with cancer-related coagulopathy (Trousseau syndrome) is predominantly attributed to cancer managements; however, stroke management by anticoagulants can contribute to the best supportive care. We aimed to find predictors of the outcome by multivariate analysis, including machine-learning (ML) based feature-engineering. Methods: A single-center retrospective study using a prospective cohort was conducted between April 2011 and June 2019. Out of the cumulative total of 110 acute ischemic stroke patients with malignancy, 65 were treated with anticoagulants, including warfarin (n=19), non-vitamin K dependent oral anticoagulants (NOAC, n=40), or subcutaneous heparin injections (n=6). Cancer-related coagulopathy was defined by elevated blood D-dimer levels at the onset of stroke with malignancy. The incidence of stroke recurrence was analyzed using 40 variables by logistic regression (LR) and in-house ML programs. Results: Out of 65 instances of the cancer-related stroke, 12 (18.5%) stroke recurrences were observed during 455 ± 70 days (mean, SEM). The stroke subtypes were cardioembolism (n=2), stroke with undetermined etiology (n=23) or other determined etiology (cancer-related coagulopathy, n=40). Multivariate LR revealed significant predictors of stroke recurrence, including NOAC usage and stroke subtype. Whereas, combination of forward stepwise selection and Naïve-Bayes (NB) or support vector machine found the blood D-dimer level as an additional important predictor. Input the D-dimer level in addition to NOAC usage and stroke subtype yielded the best area under the curve (AUC) for either of LR or NB compared to input warfarin or heparin usage. AUC for the LR for these 3 variables was better than that for NB. Conclusion: This study suggests the incidence of stroke recurrence is high in this clinical situation. NOAC usage, stroke subtype, and blood D-dimer level at the onset of stroke have predictive value of the outcome.


2021 ◽  
Vol 50 (Supplement_1) ◽  
pp. i12-i42
Author(s):  
J P Renton ◽  
D McAllister

Abstract Introduction Fewer clinical trials are carried out in older people. It is unclear how representative and applicable clinical trials carried out exclusively in older people are. We compared clinical trials recruiting older people exclusively “older trials” to those recruiting adults of all ages “all age trials”, using anti-hypertensives acting on the renin-angiotensin aldosterone system (RAAS drugs) as an exemplar. Method We searched the US clinical trials register 1, to identify all trials carried out exclusively in those aged over 60. From these we selected trials of RAAS drugs. These were matched in a 2:1 ratio to trials carried out in adults of all ages. Data regarding baseline characteristics, adverse events and eligibility criteria were collected from clinical trial reports and clinicaltrials.gov. Estimated associations were calculated for age, sex and adverse events. Eligibility criteria were described and ICD- 10 coded, as appropriate. Results 71 clinical trials were carried out exclusively in older people.13 related to RAAS drugs. Participants in “Older trials” had higher mean age (73.1 and 55.9 respectively), mean difference 16.17 (CI 15.31–17.02). Older trials had fewer male participants. Participants in older trials had lower mean body mass index (BMI). A higher rate of participants in older trials experienced serious adverse events. (2.07, CI 1.55–2.75.) Few older trials had upper age limits (23.1% V 27% all age trials). All trials had exclusion criteria in multiple ICD blocks. Concurrent medications were a more common exclusion criterion in older trials (61.5% v 40.9%). Conclusions Clinical trials carried out exclusively in older people are representative in terms of age, serious adverse events and eligibility. Although there are multiple exclusion criteria for clinical trial participation in both groups, this is not prohibitive. This supports carrying out more trials exclusively in older people. References 1. NIH, US national library of medicine.ClinicalTrials.gov Available at https://clinicaltrials.gov/


2021 ◽  
pp. 135245852098511
Author(s):  
Kris Oliver Jalusic ◽  
David Ellenberger ◽  
Paulus Rommer ◽  
Alexander Stahmann ◽  
Uwe Zettl ◽  
...  

Background: Newly approved, drug-modifying therapies are associated with still unknown adverse events, although clinical trials leading to approval have strict inclusion and exclusion criteria and analyse safety and efficacy. Objectives: The aim of this study was to analyse the eligibility of multiple sclerosis (MS) patients treated in routine care into the phase III clinical trial of the respective drug. Methods: In total, 3577 MS patients with 4312 therapies were analysed. Patients with primary-progressive MS were excluded. Inclusion and exclusion criteria of phase III clinical trials in relapsing–remitting MS were adopted and subsequently applied. A comparison in clinical and sociodemographic characteristics was made between patient who met the criteria and those who did not. Results: 83% of registered patients would not have been eligible to the respective phase III clinical trial. Relapse was the single most frequent criterion not fulfilled (74.7%), followed by medication history (21.2%). Conclusion: The majority of MS patients treated in routine care would not have met clinical trials criteria. Thus, the efficacy and safety of therapies in clinical trials can differ from those in the real world. Broader phase III inclusion criteria would increase their eligibility and contribute to a better generalizability of the results in clinical trials.


2021 ◽  
Author(s):  
Jie Xu ◽  
Hao Zhang ◽  
Hansi Zhang ◽  
Jiang Bian ◽  
Fei Wang

Restrictive eligibility criteria for clinical trials may limit the generalizability of treatment effectiveness and safety to real-world patients. In this paper, we propose a machine learning approach to derive patient subgroups from real-world data (RWD), such that the patients within the same subgroup share similar clinical characteristics and safety outcomes. The effectiveness of our approach was validated on two existing clinical trials with the electronic health records (EHRs) from a large clinical research network. One is the donepezil trial for Alzheimer's disease (AD), and the other is the Bevacizumab trial on colon cancer (CRC). The results show that our proposed algorithm can identify patient subgroups with coherent clinical manifestations and similar risk levels of encountering severe adverse events (SAEs). We further exemplify that potential rules for describing the patient subgroups with less SAEs can be derived to inform the design of clinical trial eligibility criteria.


Sign in / Sign up

Export Citation Format

Share Document