Big Data: Using Data Analytics to Enhance Drug Safety

The use of big data in the health care industry, the largest sector of the economy, offers many benefits for the overall economy, businesses, and the health and well-being of people. In a first-of-its-kind study, SMU Cox ITOM Professor Vishal Ahuja and co-authors show how the use of data can validate life-changing health care findings. Additionally, this new approach using big data demonstrates how regulatory agencies can advance their governance missions by monitoring public safety with more robust, real-time analytical methods.

The Food and Drug Administration (FDA) issues black box warnings (BBWs), it's most stringent, to signify that a drug or device carries a significant risk of serious or life-threatening adverse effects. The authors studied a controversial black box warning in 2007 on a diabetes drug, rosiglitazone, marketed as "Avandia," that led to large declines in the drug's use.

Manufactured by GlaxoSmithKline, rosiglitazone belongs to a class of medications widely used to lower blood glucose in patients with type 2 diabetes. The BBW followed a publication that showed, using meta-analysis of past clinical trials, rosiglitazone to be associated with a 43% increased risk of heart attack and a 64% increased risk of cardiovascular mortality. "We studied this drug because of the high prevalence of the disease – roughly 10% of Americans have diabetes and another third of the U.S. adult population has prediabetes," says Ahuja. "Diabetes drugs are a very large market, and the U.S. economy has benefitted immensely from new blockbuster drugs coming onto the market by firms competing."

Enhanced Pharmacovigilance

The authors' approach is complementary to the FDA's existing detection systems. Their proposed template for regulatory agencies can be used to assess associations between a drug and specific adverse reactions of interest. An evidence-based decision support system that the authors coined as "Data Analytics for Robust and Enhanced Drug Decision-making" or DAREDD, employs "big data" in a sophisticated manner. It can be used retrospectively to evaluate a warning that has already been issued or a prospectively to determine whether a drug deserves a BBW.

The authors say, "Deficiencies exist in the process of collecting, reviewing, and analyzing the evidence used to issue BBWs, highlighting a gap in evidence-based medicine." By using large data sources in novel ways, such as observational studies, and advanced analytics, better pharmacovigilance can be achieved. "Use data in a sophisticated manner to catch adverse events, but validate what you find about the drug using data," suggests Ahuja. "Or if a drug already has a warning, continuously monitor it, like every six months to affirm the decision. This is a complementary approach, a thought process which can be used to further investigate."

Ahuja explains, "Regulatory agencies like the FDA, the Bureau of Consumer Protection, and those related to transportation safety, value erring on side of caution. It's the nature of agencies. We suggest the FDA continue to follow their practices using meta-analysis and post-marketing surveillance to detect adverse events like heart attack, for example, but use caution before jumping the gun [on a black box warning]."

The Study and Rosiglitazone

The authors evaluated the above-mentioned FDA warning using a large and unique dataset from U.S. Department of Veterans Affairs (VA). The dataset consisted of a retrospective cohort of over half a million patients with diabetes from 2003-2008; this resulted in close to 10 million observations, a far larger sample than previously used studies by the FDA. They conclude that the warnings were not warranted. (In 2013, the FDA reversed its decision and lifted some of the prescribing and dispensing restrictions for rosiglitazone.)

With drugs being approved faster in the last twenty years, BBWs have also increased. After the 2007 study* questioning rosiglitazone's safety, a BBW was issued by the FDA; the United States Department of Veterans Affairs (VA) also removed the drug from its national formulary. The warnings impacted patient care, given the revisions to standards of diabetes care and management that followed. Within 30 months following the FDA safety alert, rosiglitazone use declined by as much as 75%, with larger declines occurring within the VA. From December's 2006 peak annual sales of US $3.3 billion, the drug sales plummeted following the BBW, dropping to $1.2 billion in 2009.

The authors' vigorous analysis does not support the decision by the FDA to issue the BBW on rosiglitazone. In fact, they find that the drug is associated with a statistically significant (albeit small) reduction in the likelihood of acute myocardial infarction and death from cardiovascular events. Further, rosiglitazone users are less likely to die (from any cause), although they are more likely to experience stroke. Their study offers insight into how rosiglitazone use affects patient health outcomes in “real world” clinical settings.

Beyond health care

The novel approach proposed by the authors extends to products outside of medicine like children toys, food, and household cleaners. "The bigger question is how do regulatory agencies decide if and when to issue product recalls or warnings?" Ahuja questions. "Take food safety, for example, products under regulation can use a data-driven approach, given that so much is monitored and quantified." Ahuja mentions that many agencies rely on self-reporting but that is cherry picking information that could lead to erroneous conclusions.

“Given consumers’ extreme sensitivity around product safety," Ahuja expresses, "firms manufacturing, selling, or marketing consumer products, whether its Mattell, Kraft, or Pfizer, may want to conduct their own product safety evaluation, from a business standpoint.

With increasingly large amounts of data now available, together with advancement in computing, an approach such as ours may help enhance consumer safety and public health,” he concludes.

The paper "Enhancing FDA’s Decision Making using Data Analytics" by
Vishal Ahuja of Cox School of Business, Southern Methodist University; John Birge of University of Chicago; Elbert S. Huang of University of Chicago; Min-Woong Sohn of University of Virginia; and Chad Syverson of University of Chicago, is a working paper.

*S.E. Nissen and K. Wolski. "Effect of rosiglitazone on the risk of myocardial infarction and death from cardiovascular causes," New England Journal of Medicine, 356(24): 2457–2471, 2007.

Written by Jennifer Warren.