by Yan Xu

It’s time for Canada to leverage “big data” for stronger drug safety

“In God we trust; all others [must] bring data.” – W. Edwards Deming

Baycol, Vioxx. Avandia. These medications became household names not for their widespread prescription worth billions of dollars in sales, but for their well-publicized safety concerns that have touched the lives of millions of patients and their families.

More than 80 new patented medications enter into Canadian market each year. Post-marketing drug safety, however, remains a concern: at least 41 drugs have been withdrawn from Canadian market between 1963 and 2004 – averaging one product withdrawal per year. These are agents that have been tested in rigorous Phase III clinical trials. Yet despite these safeguards, over 27,000 seniors are admitted to hospitals in Canada each year due to adverse drug-related events.

So what gives? The reality is that results of clinical trials are often difficult to generalize to the population-at-large and cannot reliably predict the “real-world” outcomes of a particular agent that is approved for market entry. Because clinical trials are experiments designed to answer specific questions for a specific agent, they follow strict criteria for selection of participants into the study. For example, while it has been know that renal function is often suboptimal in patients with heart failure and heart attacks, over half of clinical trials evaluating these diseases excluded patients with kidney disease.

Furthermore, the multi-national nature of today’s clinical trials means pooling of data from countries with diverse medical practice and quality. For example, the recent Randomised Evaluation of Long-term Anticoagulation Therapy (RE-LY) trial compared blood-thinner warfarin, which has been the standard of care in prevention of stroke in certain heart rhythm conditions, to a novel agent dabigatran. In warfarin therapy, it is imperative that patients spend an adequate time in therapeutic range to ensure treatment effect while avoiding adverse effects. An international study, the trial was conducted in 44 countries with average time in therapeutic range that varied from 44% in Taiwan to 77% in Sweden. It can be difficult to determine whether results of the overall trial, which takes into account all trial participants of all countries, will be reflective of the clinical practice pattern and needs of a specific region.

Another consideration is that clinical trials are inherently underpowered to detect rare, but important, adverse events. For example, to detect a 3-fold increase of an infrequent (yet potentially life-threatening) adverse event that occurs in 1 in 1,000 individuals, over 7,800 patients need to be enrolled in a trial. However, less than 10% of trials for medications approved between 2000 and 2010 studied over 5,000 patients. Of the 24 blockbuster drugs that generated $2 billion in sales, only one drug (clopidogrel) had a randomized trial with over 10,000 participants in the experimental arm.

Finally, some drugs are approved based on a priority review, which has been associated with increased risk of safety concerns after they enter the market. Currently, the our pharmaceutical safety monitoring system continues to rely heavily on patient, healthcare professional and industry reports to the drug regulators (Health Canada), which only captures 1% of the actual number of adverse effects experienced by those exposed to the drugs. Real-time detection of drug-related adverse events, therefore, is more important than ever, where administrative databases that capture the healthcare-related diagnoses, utilization and outcomes of an entire population offer the key to improving our drug safety efforts.

Several initiatives stand out: the National Prescription Drug Utilization Information System (NPDUIS) is a pan-Canadian database that includes claims accepted by public drug programs, providing information on use of prescribed medications; however, a critical limitation is its inability to be linked to other administrative health databases. The Canadian Network for Observational Drug Effect Studies is a national consortium of drug safety researchers that seeks to provide answers to timely drug-safety questions using multiple population-based databases with coverage of over 40 million people. These are great steps forward, but the databases are often limited to medications covered by the public payer.

With the substantial role that private insurers play in Canada’s prescription medications and differences between public and private drug formularies, there is a pressing need for databases that encompass all medications for all patients. An example is the PharmaNet system in British Columbia, the implementation of which has been associated with reduced risks of inappropriate prescriptions and has been used in several safety studies. Roll-out of similar initiatives in other provinces recommended by a recent Senate report, such as the Medication Management System in Ontario, will promote evidence-based decision-making that generate population health gains and cost savings in avoidable hospitalizations and resource utilization. However, full implementation and use drug safety research is years away in most jurisdictions.

Improving drug safety begins with better data. The time is ripe for investment in population-based monitoring strategies that pay dividends in maximizing drug effectiveness and safety, leading to better health for all.

The comments section is closed.


Yan Xu


Yan Xu is a medical student at Queen’s University. He was the former coordinator of a student-directed seminar, “Ethics of International Service-Learning”, at the University of British Columbia.

Republish this article

Republish this article on your website under the creative commons licence.

Learn more