Last week I attended the 5th Annual GCP Compliance conference in Washington, D.C. A theme of the conference was risk-based monitoring (RBM). I was there to listen and learn, as well as meet some of the people at the sharp end of making this a reality.
If you don’t know, RBM refers to a strategy for clinical site monitoring that moves the industry away from its unofficially adopted “100% source data verification on site, visit once every 4 to 8 weeks” model. This model was never mandated by regulatory authorities but caught on, probably because no-one got in trouble for doing it. This is in effect a very expensive and time consuming QC program. Two types of check are performed – source data review to confirm that the data look plausible, and source data verification to confirm the data were transcribed correctly from source. The use of EDC, central labs, IRT etc. means that most of the data are available centrally in close to real time for data review (e.g. to look for implausible results). In some cases (think where central lab data or electronic health records are integrated directly into the EDC), the source data are not at site, the data are on someone’s servers somewhere, rendering source data review and verification at site redundant. Moreover, the available data show that transcription error rates are low (3%) and much of those are caught by real-time or near real-time QC checks built into the EDC or through existing central data review.
The pitch for risk-based monitoring is that the traditional model is inefficient. It treats all sites and data points as equally important in terms of attention paid in review. We should be able to save money by performing a risk analysis on the study protocol and sites’ profiles, developing a targeted monitoring plan. Visits might be less frequent, and/or triggered by specific events or anomalous data, source data verification can be less than 100%. At the conference speakers told us that site monitoring was 30% of the trial costs and we could shave off 20% of that number.
So why isn’t this about monitoring? It is, but it’s really about a much bigger shift to a quality by design (QbD) approach to clinical drug and medical device development. QbD (which originated in manufacturing) ensures all the individual parts of your manufacturing process meet requirements, and through piloting ensures that they work together and produce the correct output (i.e. the final product that meets requirements. With this approach you should not need to QC all of the product at the end of the production line (where addressing problems is at its most expensive). In other words, if you make sure the machinery making your light bulbs works the right way, you don’t need to check every light bulb produced to see if it works. Various bodies including regulatory authorities have pushed for these ideas to be adopted in pharmaceutical manufacturing and now the clinical trial world. For sure there is an element of reducing costs but we should see the cost savings in the broader context of streamlining the drug development process, not just the headline cost of monitoring on a particular trial. It is also an acknowledgement that the current method isn’t that effective; for example from my point of view trials that collect health data that aren’t needed, or collect the data in a way that unnecessarily burdens patients raise ethical concerns.
Why isn’t it about risk? Several speakers stated we should stop talking about it being “risk-based”; perhaps agree a different terminology. This is partly because many people feel this implies a trade-off where errors will be acceptable in certain areas because they have been determined as low-risk. My answer is to go back to QbD. If you design your program and protocols to avoid problems, and qualify and train your sites in order to reduce the occurrence of remaining possible problems, you are reducing risk of failure, e.g. to patient safety, prospectively. Thus you can adopt a less comprehensive monitoring plan. Conversely if we do not design for failure we cannot afford to reduce QC, onsite or otherwise.
This area is in its infancy, but it is fascinating. It mandates a culture shift from many groups in the development process, from the protocol designers and clinical scientists, to data management and clinical monitoring, sponsor and CRO. If done right it will also mean sites will have to absorb change and meet new challenges with less on-site support and double-check of their work.
But will we increase errors in low-risk areas? Perhaps in the short term but we have to get the big stuff right first before we can focus on the little things. We might never know what happened to the Malaysia Airlines flight that disappeared over the Indian Ocean, but we do know that flying in commercial aircraft is much safer than any other form of transport. Aircraft manufacturers and airlines are very familiar with this approach to design and manufacture: before you design, work out all the things that could go wrong and focus preventive efforts on the big ticket items. A broken seat back that does not recline won’t force an emergency landing because it doesn’t endanger anyone (except as one of the speakers, Beat Widler, said to me the broken seat back could ground the aircraft if it belongs to the pilot!).