part 10

Case Study: Optimization of Clinical Trials through Biomarker Analysis

The emerging use of advanced analytic techniques is enabling Life Sciences companies to reduce drug development costs and improve outcomes using Biomarkers, a previously underutilized data source. As objective indicators of a medical state, biomarkers are used in drug development research to investigate the impact and interaction of critical factors such as prior treatments and body systems. Biomarker analyses are being leveraged to optimize clinical trial participant selection and execution, and the success rate for drugs developed using biomarkers is 18% higher than those developed without biomarkers. Projects analyzed with biomarkers can benefit from enhanced clarity and predictability, which means that early-stage projects may be stopped earlier and those studies that progress to a later stage have increased confidence in the results.

The application of biomarkers has proven successful in the diagnosis and management of cardiovascular disease, infections, immunological and genetic disorder, and cancer. The insights derived from advanced analysis of biomarker data allows Life Sciences companies to better understand the activity of their drugs, reducing the incidence of adverse events during clinical trials as well as the cost and time of bringing their drugs to market.

Challenge

A large Pharmaceutical company desired to translate their patient clinical and outcome data into valuable insights for the development of a cancer drug.  They sought to identify and rank clinical variables that influenced tumor shrinkage and to use these variables to predict tumor response. The organization wanted to use these advanced analytics as a diagnostic tool to provide them with a better idea of the data they should be collecting and monitoring during the clinical trials. They were interested in identifying biomarkers that are correlated with the drug’s influence on tumor shrinkage in order to prepare for their next round of clinical trials with a larger set of subjects.

However, the client was having trouble assessing and predicting drug treatment outcomes by analyzing biomarker data.  It was not an organizational priority to build complex machine learning models in support of predictive analytics.  In addition, they had little capacity to extract meaningful insights from very sparse data. The company reached out to Knowledgent to help them derive meaningful insights from their biomarker data.

Approach

Knowledgent worked with the company to identify, prepare, and analyze biomarkers (e.g. cytokine panel, flow cytometry, genotyping) and clinical covariates (e.g. age, sex, and ethnicity).  We helped the client assess the completeness and utility of available datasets and to prepare the data and impute missing values to enable richer analysis.  Knowledgent and the client worked together to develop six classes of sophisticated predictive machine learning models including classification and regression trees (CARTs), random forest, linear models and support vector machines (SVMs).  These were embedded in a reusable, parameter-driven structure that will position the client to replicate the processes in the future.  Knowledgent also worked with the client to review the analytic results and to train business users on use of the models.

Results

The data science models provided the client with key insights they were seeking into the impact of biomarkers and clinical covariates on tumor shrinkage.  The models ranked the importance of each biomarker and clinical covariate variable on predicting tumor shrinkage. They also identified the marginal effect of changing a variable on the predicted outcomes.  As a result of the project, the team was able to identify the top two clinical predictors of tumor shrinkage. Additionally, they were able to demonstrate the mathematical impact of the tumor progression rate and tumor shrinkage rate on the tumor shrinkage over time.

The client also greatly benefited from the knowledge-transfer involved when transitioning the ownership of the models and their optimization-schemes to their staff.  They enhanced their analytic capabilities with the sophisticated machine-learning models and advanced analytic techniques. They have retained higher quality data from the advanced preparation and imputation methods that were used.  Additionally, they will continue to benefit from the reusable, parameterized models and trained users that will result in ease of future analyses.

Summary

Advanced analytics during clinical trial design and execution allows Pharmaceutical companies to understand drug effectiveness for individuals through improved analytic insights.  Knowledgent worked with a large Pharmaceutical company to assess the impact of patient biomarker and clinical covariate data on tumor shrinkage using predictive machine learning analytic models.  Knowledgent and the client were able to gain new insights into drug candidates and guidance for additional clinical trials and data collection.  The results will enable the client to optimize the participant selection and execution of clinical trials, increasing the efficacy of the results, reducing adverse events and shortening the time to market.  In addition to the benefits from the current project, the client company has gained important expertise and reusable analytic assets that will allow them to undertake additional analyses in the future.

Leave a Reply

Your email address will not be published. Required fields are marked *

Time limit is exhausted. Please reload CAPTCHA.