Developing targeted therapeutics is a smart way to increase the woeful success rates in clinical development and the JP Morgan Healthcare Conference, as described in “Pulse of J.P. Morgan 2014″ from Fierce Biotechnology, showed that many companies are adopting this strategy. However, as this blog has previously discussed, developing targeted products, and especially orphan drugs, means smaller markets and smaller revenue potential. The high cost of traditional development methods has been a drag on development of all types of drugs, including small molecules with blockbuster potential. The cost is simply unsustainable with the strategic shift to targeted therapeutics and smaller markets. With payers balking, there is a limit to how much of an R&D investment in a targeted medicine a developer can recover through aggressive pricing. Furthermore, success rates may be higher for targeted biologics, but substantial risk remains. We will have to use our ingenuity to make targeted therapeutics pay off.
One thing we have to recognize is that for all the talk about Big Data, we need to be able to learn much more from Small Data – data collected in early-phase trials involving small populations. For example, we can we use n-of-1 designs and crossover designs to learn more about patient response and better plan next steps. We also need to get better at utilizing incoming data to continually reevaluate whether it makes sense to proceed to the next step.
The central idea of Small Data is that we use each bit of new information as an addition to a knowledge base that helps us project where we’re headed. Used in this way, Small Data enables a big step toward continuous development. But in order to make the most of Small Data, we need to change the way we think about clinical development and run programs. Rather than viewing clinical programs in terms of discrete, prespecified decision points, we need a more flexible approach that allows decisions as soon as the data and our best techniques allow.
To make this work, some things in clinical development have to change. We need:
- Immediate, actionable data, not just raw data or batched data to use as a starting point for analysis
- An understanding that each new piece of information is precious because it can tell us where we’ve been and provide a sense of variability in the results
- Recognition that there is no hard and fast rule for determining when enough information is enough
- Modeling and simulations to project based on current information where we are likely to end up – the predictive probability of different outcomes.
When enough information is enough depends on what comes next. If our choices are all or nothing, go or no-go, we need much more reassurance before deciding to continue. This tilts methodology toward Big Pharma preferences and Big Data as a basis for decision-making. But if the next steps involve adaptive methods, the bar for progression may be lower because we can build in multiple checkpoints rather than making a huge irrevocable commitment of R&D resources.
In the end, the benefit of greatly reducing the need for absolute assurance before proceeding to the next step is most notable for phase II (either PoC or dosing). A more nuanced approach with more decision points along the way provides better control of program budgets and reduces the risk of failure along the way.