Report: Pharmacovigilance Strategy Meeting, Boston, MA, USA November 2017
In less time than it took for me to travel to the meeting from New Zealand – the day was over! Hard to believe that within a short eight hours, 25 lively roundtables were conducted, two keynote speeches were delivered, and the panel discussed the learning of the day, including challenges of monitoring diverse sources, importance of real-world evidence in the context of risk benefit, and the game-changing impact of artificial intelligence.
Add to that, 95 organised 1-to-1 meetings between highly-qualified representatives of the Pharmacovigilance (PV) industry and a variety of vendors offering a range of solutions for the numerous activities required to comply with the global authority regulations….. and to top it off… numerous opportunities for a casual chat over a coffee or the splendid lunch provided.
During this well-organised meeting, there was very little down-time. The 100+ delegates followed individual itineraries and were personally ushered from room to room where they were encouraged to participate in the round-table discussions that followed six (hot) topic streams:
- Adverse Events
- Artificial Intelligence
- Benefit-Risk Management
- Signal Detection
The Opening Keynote Address was delivered by Mariette Boerstoel-Streefland, Senior Vice President, Head Global Drug Safety, Shire, who shared with the delegation her personal PV vision for 2025.
Mariette described the future as a change from “PV” to “Drug Safety”, where some form of evidence relating to risk and benefit will be necessary to judge the overall safety of an agent. She described this new approach as enabling treatment choices to be made based on the benefit of an individual drug in the context of the severity of disease, comorbid conditions, and overall treatment regime, as opposed to selecting a drug based simply on its toxicity.
She posed that data sharing and transparency would need to increase and therefore individual databases would become obsolete, instead one large cloud containing all the safety data in the post-marketing stage would exist. She qualified that the equivalent for clinical data for a drug would not be appropriate because of the sensitive nature of the information.
She questioned how far we could go in the development of artificial intelligence over human judgement when it came to decision-making where alignment and interpretation of different data were required…. but did not provide a visionary solution. She did, however, invite the audience to think beyond the current challenges saying that the industry is at an inflection point where there is a great need to evolve.
During the rest of the conference, collectively, the general hum of the chatter during breaks, the experiences discussed within roundtables, and the insights gleaned from the keynote addresses revealed some recurring themes that need to be considered and further investigated in order to be best prepared for the evolving challenges in the area of PV.
These included the disparity between the guidance and expectations of regulatory authorities, particularly relating to monitoring local literature and social media. Pains highlighted in these two areas were the unstructured nature of the information and the obvious fact that there are multiple languages to cope with.
The fact that PV will evolve and become more focussed on the risk-benefit of an agent was a common talking point. Many discussions agreed that real-world evidence would become more significant for monitoring performance of a drug outside the controlled environment of the clinical trial. The concept of a future where more tailored and targeted medicines would be passing through development pipelines, in restricted patient groups and tested over reduced time periods, was raised. Contributors to such discussions believed that this would result in fewer adverse events during the clinical stage, and that the true performance of the drug would emerge once the drug was on the market in a true-life environment; giving even more good reason for monitoring literature in the post-marketing phase.
From what I heard, I am convinced that artificial intelligence is an important innovation with regard to the problem of ever-increasing volumes of safety literature that need to be monitored and reviewed. And, while there is some comfort around allowing process tasks to be automated, there is still some concern regarding decision making, especially where inference and synthesis of information from disparate pieces of data are concerned. There is no doubt that artificial intelligence will be part of the future PV – how quickly it is adopted and trusted will depend on how much the PV industry is willing to give up its human hands-on approach. Maybe this will be the brilliant sacrificial move that allows the perfect strategic game to play out?