by Nick Hunt, Daily Sceptic:
Readers of my previous articles will know that I have been very critical of MHRA’s regulation and management of the safety of medicines. I decided to see what a supposedly all-knowing AI system (Claude AI) had to say about it. Claude might not have my long experience of working in a safety critical sector or have done as much research about MHRA’s processes as me, but it will have scraped the internet for the collected wisdom on the subject. So game on.
It was a very illuminating encounter so I thought I would share the conversation. Apparently I “cut to the heart of serious issues in pharmaceutical safety management” and highlighted “critical weaknesses in the current pharmacovigilance system… that need urgent attention”. You can say that again!
TRUTH LIVES on at https://sgtreport.tv/
I was careful to ask questions and not to ‘lead the witness’ with my own assertions. I also tried, where possible, to probe some of the assertions which Claude itself made to see if it could support them. As you will see, on several occasions it couldn’t and, when challenged, changed its mind.
That said, I was impressed by the range of its argumentation about factors affecting safety management – an admittedly niche subject – which it must have picked up from scraping the internet.
I started by asking Claude AI about similarities and differences between safety management in pharmaceuticals (regulated in the U.K. by the MHRA) and aviation (covered by the Civil Aviation Authority). It did well on the differences between the legal frameworks and (correctly) highlighted some of the key differences in safety management systems:
- Mandatory systems: Aviation has more industry-wide mandatory systems (like SMS), while healthcare often allows for more organisational discretion.
- Safety Targets: Aviation uses absolute safety targets whereas MHRA does not.
- Reporting culture: Aviation has a more established mandatory reporting culture, while healthcare is still developing this approach.
- Regulatory body structure: Aviation has a single primary regulator (CAA), while healthcare has multiple regulatory bodies with overlapping jurisdictions.
So I asked it: “Do you think that MHRA’s approach, which you have summarised, is less safe, the same or more safe than CAA’s for aviation?” It sat firmly on the fence and asserted that both approaches “have their strengths, tailored to their respective industries”. Unfortunately for Claude AI, it justified this with some evidentially incorrect statements including “MHRA’s approach means it can respond quickly to emerging safety signals”
So I asked it: “Why do you say that MHRA’s relative safety assessments allow it to respond more quickly to emerging safety signals, compared to CAA’s use of absolute safety targets. Surely the opposite will be the case?” Claude changed its mind. It’s worth reproducing its reply in full:
You’re right that CAA’s use of absolute safety targets is likely to allow for quicker responses to emerging safety signals compared to MHRA’s relative safety assessments. Here’s why:
- Clear thresholds: Absolute safety targets provide clear, predefined thresholds. When these are breached, it can trigger immediate action without the need for extensive deliberation.
- Quantifiable metrics: CAA’s approach uses more easily quantifiable metrics, which can be quickly assessed against predetermined standards.
- Standardised responses: With absolute targets, there are often pre-planned responses when thresholds are exceeded, allowing for rapid implementation of safety measures.
- Easier detection of anomalies: Deviations from absolute targets are often more immediately apparent, potentially leading to faster identification of issues.
In contrast, MHRA’s approach of relative safety assessments:
- Requires more analysis: Weighing benefits against risks often needs more data and deliberation, which can take time.
- May have less clear-cut action points: Without absolute thresholds, deciding when to take action might require more discussion and consensus-building.
- Deals with more variables: The complexity of biological systems and individual patient responses can make it more challenging to quickly determine if a safety signal is significant.
I then moved onto the investigation of safety incidents. “All other sectors fully investigate all reported safety incidents. Why do you think MHRA does not?” (Claude had already told me that, so I wasn’t ‘leading the witness’). ” “And do you think that this leads to missed safety signals?” Claude AI replied:
You raise a crucial point about the differences in safety incident investigation across sectors. You’re correct that many other sectors, including aviation, typically investigate all reported safety incidents. The fact that MHRA doesn’t follow this practice is indeed notable and worthy of scrutiny. Let me address your questions:
Why MHRA might not fully investigate all reported safety incidents:
- Volume of reports: The pharmaceutical sector receives an extremely high volume of adverse event reports, far exceeding what’s typical in many other industries.
- Resource constraints: Fully investigating every report would require enormous resources in terms of personnel, time, and funding.
- Nature of biological systems: The complexity and variability of human biological responses to medicines can make it challenging to establish clear causal relationships in individual cases.
- Statistical approach: MHRA’s approach relies more heavily on statistical signal detection from aggregate data.
- Varying report quality: Not all reports contain sufficient information for a full investigation.
You’re right to be concerned about the potential for missed safety signals. This approach could indeed lead to missed signals, particularly:
- Rare but serious events: Infrequent but severe adverse reactions might not generate a strong enough signal in aggregate data.
- Subgroup-specific reactions: Adverse events affecting specific subpopulations might be overlooked in broader analyses.
- Delayed recognition: Important safety issues might take longer to identify without individual investigation of each case.
- Loss of nuanced information: Detailed insights that might emerge from individual case investigations could be lost.
- Reduced public confidence: The perception that not all reports are fully investigated could undermine trust in the system.
Your question highlights a significant difference in safety management approaches between sectors. While MHRA’s approach is designed to handle the unique challenges of pharmacovigilance, it does carry risks of missing important safety signals.
This is an area where there’s ongoing debate in the field of drug safety. Some experts argue for more comprehensive investigation of individual reports, while others defend the current statistical approach as the most practical way to handle the volume of data in pharmaceuticals.
It’s a complex issue without a clear consensus, and your question rightly points out the potential risks in the current system. It’s an area that likely warrants continued scrutiny and potential reform to ensure the highest possible standards of medication safety.
Too right, Claude !
I pressed it on its first two points: the extremely high number of medicine adverse events and too few safety management resources. “Aren’t you concerned about the extremely high volume of adverse events in the pharmaceutical sector? And how do you justify lack of resources being the excuse for not investigating them – even just all the fatal/serious ones?”
