Big Data: The Panacea for Pharma’s Ills?

Two words are driving innovation within the pharmaceutical industry these days: “big data.” Start-up companies like Datavant are receiving millions of dollars to harvest, organize, interpret and, hopefully, protect large amounts of data from a variety of stakeholders in the healthcare space (1).

For example, the Open Medicine Institute, a digitally focused research clinic, serves as a free private repository for the collection of all patient information. This generates vast amounts of raw data that need to be curated—a situation also facing pharmaceutical development and manufacturing. For the pharma industry, this is an opportunity as well as a curse.

While the term “big data” has been around for a long time, an actual definition is elusive since its definition continues to fluctuate. In general, big data refers to extremely large datasets, both structured and unstructured, acquired daily by businesses. These large sets of data can be analyzed to reveal patterns or trends to gain insight into problems, provide support for decision-making and review performance objectively, often in real time. This data can also be actively measured so that outcomes and reporting can be adjusted as more information is obtained.

Now, with the broad use of electronic health records, unstructured data from diagnostic equipment and its own clinical trials, the pharmaceutical industry has a treasure trove of information from which to choose. By collecting these various datasets, pharma has an opportunity to gain novel insights that can enhance and accelerate the drug development process, especially in clinical trials (2), particularly as the manufacturing side explores the possibilities of big data (3).

Big Data Meets Clinical Trials

Getting a drug product to market takes, on average, 10–12 years and requires hundreds of millions of dollars. Only one in one-thousand proposed new drug products completes this journey. Many potentially viable drugs are lost, abandoned or discarded, not necessarily due to efficacy or product safety, but because the high cost to survive a clinical trial cannot be justified. Some of these “lost” products include orphan indications, rare disease medicines and novel formulations. The loss of these products is due to a number of things—safety outcomes (phase I) effectiveness (phase II) and effectiveness plus adverse reactions (phase III) (4).

Being able to predict how a compound might act would maximize positive outcomes, minimize errors and facilitate clinical trial enrollment. Big data could mitigate costs and shorten timelines for clinical trials, providing more justification for a drug’s development.

Forecast with Predictive Modeling

Predictive modeling, or the ability to forecast the potential for a drug to be safe and effective for selective subsets of the population, is currently being used by a number of companies. When evaluating potential compounds with clinical data, predictive modeling could identify potentially successful molecules that would be both target-specific and safe and effective.

The use of big data to validate biomarkers in order to map treatment outcomes is a refined form of this type of modeling. In areas where the stakes of life and death are high, such as cancer treatments, this is especially true. A recent study used a Bayesian model to map cancer outcomes based on the genetic composition of the biomarkers found in “The Cancer Genome Atlas” database (5). By forecasting the potential success of the treatments, appropriate patient populations can be selected, therapies targeted, and dosages titrated to create a more patient-specific treatment.

Active Real-Time Monitoring

Currently, clinical trials are closely monitored with adverse events reported in a timely manner as dictated by best practice guidance. Data is reviewed for accuracy and a risk-based monitoring plan is created for each clinical study. Even with these precautions in place, however, there are still gaps in clinical trials due to the limitations in the ability of the monitor to provide 24-hour vigilance (6). Nowhere is this more apparent than in large global trials, where multiple time zones must be factored in. The opportunity to be alerted to safety or performance outcome signals could potentially allow for immediate action on poor study-related processes or multiple adverse events where intervention might be needed.

Hidden Relationships Revealed

Big data can link comorbidities or associations with other diseases, as in the case of a study of U.S. military veterans that showed a correlation between periodontal disease and rheumatoid arthritis (7). The U.S. Veterans Health Administration conducted a retrospective cohort study on 25 million patients to measure the relationship between periodontal disease and rheumatoid arthritis. The outcome indicated that patients who suffered from periodontal disease were 1.4 times more likely to have rheumatoid arthritis compared to other dental patients.

Post-Market Assessment and Safety

Big data modeling can also be used to review drugs that are currently on the market and look for new indications. By repurposing a pre-existing drug, a lower barrier to market entry is attained due to shorter approval time. In addition, using big data to review currently approved medications can be used to look for safety trends to prevent adverse events, off-label use or abuse.

Breaking Barriers to Big Data

While the advantages of applying big data in the current marketplace appear obvious, in reality, significant barriers still exist. Big data analytics are complicated and require a diverse set of skills for longterm support.

Identifying what to target can also be confusing, as there are many ways to implement big data. Devising a strategic plan requires both time and effort. Planning teams need to have fundamental research skills as well as statistical and analytical expertise. They will need a good understanding of both existing systems and the big data process.

Stakeholder positions can sometimes become entrenched. Those stances must be understood and addressed, and alternative arrangements or organizational restructuring must occur to prevent push back from stakeholders.

Cybersecurity remains high on the list of concerns, despite the fact that a good deal of data is encrypted. And U.S. and European medical privacy concerns must be considered.

Conclusion

While there are many advantages to using big data to improve outcomes or support clinical trials in drug development, there are a few pitfalls to consider.

Helping to uncover hidden relationships, maximize outcomes and minimize adverse events will expedite drug approval. As such, working with big data is certainly the future of the drug development process. And, if big data holds such major implications for drug development, how much more does it offer manufacturing?

References

1. “Datavant Acquires Universal Patient Key and Closes $40M Financing Round.” PR Newswire (April 30, 2018). www.prnewswire.com/news-releases/datavant-acquires-universal-patient-key-and-closes-40m-financinground-300638719.html (accessed Sept. 5, 2018)
2. Copping, R., and Li, M. “The Promise and Challenge of Big Data for Pharma.” Harvard Business Review (Nov. 29, 2016). https://hbr.org/2016/11/the-promise-and-challenge-of-bigdata-for-pharma (accessed Sept. 5, 2018)
3. D’Alessandro, M. “Case Study: Using Information Technology-Based Tools to Optimize Yields in Vaccine Production.” Presented at the 2015 PDA/FDA Vaccines Conference, Dec. 2, 2015, Bethesda, Md.
4. “How do I go about getting a drug approved?” FDA Basics for Industry, online. https://www.fda.gov/ForIndustry/FDABasicsforIndustry/ucm238040.htm (accessed Sept. 5, 2018)
5. Zhu, Y., et al. “Zodiac: A Comprehensive Depiction of Genetic Interactions in Cancer by Integrating TCGA Data.” Journal of the National Cancer Institute 8 (2015) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4554190/pdf/djv129.pdf (accessed Sept. 5, 2018)
6. Guidance for Industry: Oversight of Clinical Investigations – A Risk-Based Approach to Monitoring, U.S. FDA, August 2013, https://www.fda.gov/downloads/Drugs/Guidances/UCM269919.pdf (accessed Sept. 5, 2018)
7. Grasso, M.A., et al. “Using Big Data to Evaluate the Association between Periodontal Disease and Rheumatoid Arthritis.” AMIA Annual Symposium Proceedings Archive. (2015): 589–593. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4765642/ (accessed Sept. 5, 2018)