Mini-Sentinel, a huge active pharmacovigilance system that lets regulators search for safety signals in health-care-related databases that have data on over 125 million Americans, is up and running, and producing actionable information. Speaking at a Sentinel Initiative Public Workshop in January, Rachel Sherman, associate director for medical policy at the US Food and Drug Administration (FDA), said the scheme is “ahead of schedule”, nearly 5 years since Congress mandated its creation. “We can talk about reviewers using the functionality,” she added. The case studies presented at the meeting provided interesting insight into Mini-Sentinel’s capabilities.
The European Medicines Agency (EMA), meanwhile, is implementing new pharmacovigilance legislation later this year. Although its new rules have a broad scope (BOX 1), they importantly also include
measures to improve active surveillance. “The age of enhanced and active pharmacovigilance is here,” says Peter Arlett, head of pharmacovigilance and risk management at the EMA.
Taken together, the dual approaches have obvious implications for drug developers. Large databases and better use of statistics can speed up the identification of safety signals by several years (Drug Safety 33, 475–487; 2010). The Mini-Sentinel pilot project, which will eventually mature into Sentinel, could prove to be particularly powerful in identifying increased risks for relatively common events, such as heart attacks, depression and suicide. Drug-induced increases in risks are particularly difficult to assess when background rates are already high, but active surveillance should enable better estimates of real risks in specific populations. “This system gives us a way of checking out some of those risks in a way we’ve
never had before,” said Bob Temple, Director of the Office of Medical Policy at the FDA, at the meeting. “It is extremely exciting.”
Sentinel is also winning industry support, particularly in its ability to reduce uncertainty. “I think Sentinel has been a huge success,” says Elliott Levy, Senior Vice President of Global Pharmacovigilance and Epidemiology at Bristol-Myers Squibb (BMS). As the system evolves, he adds, it could feed into benefit–risk discussions for experimental drugs.
Mini-Sentinel shows it cardsRegulators have relied on passive surveillance approaches — such as case studies, emerging clinical science and spontaneous reporting of adverse events (of which the FDA currently receives several hundred thousand reports each year) — to monitor drug safety. Sentinel turns these approaches on their head by enabling regulators to actively mine and monitor, and ask questions of, distributed health-care databases
Unleashing the Mini-SentinelFirst data from the FDA’s new active surveillance programme highlight how the system might influence drug development programmes.
PHO
TOD
ISC
NEWS & ANALYSIS
NATURE REVIEWS | DRUG DISCOVERY VOLUME 11 | APRIL 2012 | 255
that are held independently by different stakeholders. Currently, the system primarily analyses insurance claims records, but it is expanding to incorporate clinical health-care records, vaccine registries and more.
Whereas past public Sentinel workshops involved in-depth discussions of infrastructure and privacy issues, the handful of case studies presented by Richard Platt, an epidemiologist at Harvard University and principal investigator on Mini-Sentinel, now demonstrate some of its live capabilities.
In a first example with Lilly’s antiplatelet agent prasugrel, he showed how Mini-Sentinel can characterize the populations of patients who are receiving a specific treatment or experiencing a health-care event. Prasugrel, which was approved in 2009 for acute coronary syndrome, is contraindicated in patients with a history of stroke or transient ischaemic attack (TIA). When Mini-Sentinel was used to assess the population of patients who were receiving the drug, to see whether prescription advice was being followed, it found that out of 7,000 new patients, 8% had a prior history of stroke and 2% had
a prior history of TIA. It also found that out of 150,000 new patients receiving a comparator, Sanofi/BMS’s antiplatelet agent clopidogrel, which does not carry similar contraindications, 17% had a history of stroke and 8% had a history of TIA.
“This is a good example of how Sentinel can provide us with real-time insight into the effectiveness of risk minimization measures,” said Levy. “This kind of information is of great interest for manufacturers,” he added. Discussions are underway to determine whether drug developers will be able to run queries on Sentinel’s distributed databases.
A second example shows that Mini-Sentinel can assess specific safety concerns arising from older approved products. After submissions to the FDA’s spontaneous reporting Adverse Event Reporting System (AERS) database suggested that a specific angiotensin II receptor blocker (ARB), olmesartan, increases the risk of developing celiac disease, the agency requested Mini-Sentinel’s data to assess the relationship. It found over 500,000 patients who had been newly placed on one of seven different ARBs, and that the proportion of
patients who were diagnosed with celiac disease after starting treatment was relatively similar between agents (between roughly 0.04 and 0.08 cases per 100 patient years). The analysis, which broke down diagnosis by age and sex as well, also showed that the average time to diagnosis with celiac disease was just a few months.
Mini-Sentinel can also monitor the effects of FDA actions on drug use. An ongoing project is examining whether the February 2010 labelling changes to long-acting beta agonists, advising against their long-term use as single agents, affected drug use and health outcomes.
A fourth case study shows that Mini-Sentinel can collect and assess accumulating safety evidence for newly approved products. In the ongoing analysis, researchers are monitoring for acute myocardial infarction in new users of BMS/AstraZeneca’s dipeptidyl peptidase 4 inhibitor saxagliptin, and comparing the incidence with that of other antidiabetics.
Full results from these and all future studies will be publicly disclosed after they are completed (on the Mini-Sentinel website).
Despite enthusiasm for these new capabilities Mini-Sentinel has key weaknesses, driven largely by the limitations of the databases it can mine. The insurance claims records that currently make up the bulk of its data are best at capturing events that are relevant to insurers. It therefore has good data on the incidence of heart attacks, for example, but lacks information on health factors, such as weight or pulmonary function, that might be important for understanding risks. Although it is expanding into clinical electronic health records, these databases have their limitations as well. “There are lots of things that just aren’t well captured by these systems,” says Platt. Ongoing methodological and validation experiments will reveal which types of outcomes and disease factors it can and can’t capture.
Consequently, he adds, Sentinel will only ever be an addition to the drug regulators’ safety toolbox, not a replacement for passive surveillance systems and clinical trials.
Benefits for drug developersAlthough increased pharmacovigilance may in some ways raise further challenges for drug developers, officials at the meeting noted that Sentinel is likely to act primarily as an accelerator for signal detection. In many instances, added Arlett, faster identification of safety signals should mean that drug developers can take earlier steps, such as labelling changes and diagnostic
Box 1 | The EMA revamps European pharmacovigilance
The European Parliament adopted new pharmacovigilance legislation to better protect patients in December 2010, and the new rules are being rolled out by the European Medicines Agency (EMA) this year. “In terms of scope, this legislation is without doubt the biggest change to the legal framework for human medicines since 1995, and possibly longer,” says Peter Arlett, head of pharmacovigilance and risk management at the agency.
To meet the high-level objectives of the new legislation, which includes the need to establish clear roles and responsibilities for regulators and companies in addressing drug safety, the agency is adopting multiple measures: •From July, the Pharmacovigilance and Risk Assessment Committee (PRAC) will meet on a monthly
basis to discuss drug safety issues. The Committee for Medicinal Products for Human Use (CHMP) will rely on these publicly disclosed decisions in taking final action.
•Newly approved drugs will require a risk management plan, which will detail the known risks, the potential risks and the plans for post-marketing safety and efficacy studies for addressing uncertainties. The agency will have legal power to take action if drug firms do not complete their post-marketing studies as per the authorization requirements.
•Regulators and companies will have a legal responsibility to measure the effectiveness of risk minimization measures — for example, through outcomes studies.
•Firms will be required to periodically submit ‘cumulative benefit–risk assessments’, rather than the current practice of submitting safety updates that discuss the events since the last report.
•Whereas firms must currently submit separate adverse event reports to each regulatory agency of each member state, reporting will be centralized through the EMA.
•The EMA will be able to charge a fee specifically for the costs of running, and of investing in, enhanced pharmacovigilance systems.
The legislation also shifts the emphasis from collecting data to detecting safety signals, says Arlett. This new focus, combined with the ability to better fund pharmacovigilance, he adds, will translate into a more active pharmacovigilance approach. It could, for example, enable greater use of European active surveillance networks such as the EU-ADR, an active surveillance network that includes electronic health record data from 30 million patients from four different countries (The Netherlands, Denmark, United Kingdom and Italy). “We have a multiheaded beast in Europe, but overall our approach does similar things as Sentinel,” says Arlett.
N E W S & A N A LY S I S
256 | APRIL 2012 | VOLUME 11 www.nature.com/reviews/drugdisc
recommendations, to avert drug safety disasters. “Ninety-nine times out of a hundred, pharmacovigilance is not about taking drugs off the market earlier, it is about taking a risk minimization strategy earlier.”
Arlett adds that drug firms should be able to benefit from the increased capabilities of the EMA and FDA. “For me, strengthening pharmacovigilance is actually good for drug development and good for innovation,” he says. The new systems won’t fundamentally change the nature of the benefit–risk hurdle, he says, but will provide regulators with assurances that certain types of data — including both outcomes data and how well risk mitigation strategies are working — will be collected post-authorization, easing the decision-making process.
Platt agreed, noting that active discussions are underway to determine how, and in which instances, Sentinel’s capabilities can be leveraged during candidate reviews. “There is no question that Sentinel can get you information faster than a conventional clinical trial would,” he adds.
Echoing these sentiments, Levy added that Sentinel will prove to be useful for
estimating background rates of very rare events, which could help regulators and companies to understand the real risk of rare safety signals that pop up in clinical trials. “Sentinel won’t erase critical safety concerns, but can help to reduce uncertainty around other concerns and simplify the approval pathway for compounds that have favourable overall benefit–risk profiles,” says Levy.
In the longer term, Sentinel might also be able to help identify unexpected efficacy signals, or lack of harm signals, for approved products. Although in most cases such signals would probably be hypothesis-generating rather than confirmatory, they could nevertheless open up intriguing avenues for exploration. Platt cautions, however, that it is much easier to identify risk than harm. “We will have to deal with this on a case-by-case basis.”
As Mini-Sentinel releases more project data, discussions with stakeholders advance and methodologies improve, the exact capabilities of the new system will come into even clearer focus. “The Sentinel of today is not the Sentinel of 25 years from now,” said Sherman at the meeting.
N E W S & A N A LY S I S
NATURE REVIEWS | DRUG DISCOVERY VOLUME 11 | APRIL 2012 | 257
