Anyone who reads me regularly knows research is expensive. Trials of just a few hundred participants (or less) can run into the millions of dollars. Therefore, most of the big advancements in medicine are usually based on studies of far fewer people than you might expect. Even hugely successful drugs have only been studied in thousands of patients.
Unfortunately, this means that uncommon events are very unlikely to be discovered in such trials. We need bigger ones. But those cost money, and no one seems to want to fund them. John Ioannidis has an idea:
The conduct of very large, simple trials (mega-trials) is uncommon and faces several challenges, in particular cost and difficulty in patient recruitment. However, these challenges can be overcome when interventions are widely used by hundreds of thousands of individuals and when there is a potential profit to accommodate the trial cost. Accordingly, every licensed intervention with annual sales that exceed $1 billion, ie, a blockbuster, should have at least 1 trial performed with at least 10 000 patients randomized to the intervention of interest and as many randomized either to placebo (if deemed to be a reasonable choice) or to another active intervention that is the least expensive effective intervention available. The comparison drug can be a generic drug with a well-established effectiveness and safety profile.
A data-intensive mega-trial of 20 000 randomized participants with 4 years of follow-up costs an estimated $420 million, but streamlining the design, monitoring, data collection, and outcomes could save 90% of that cost. For a blockbuster with $2 billion annual sales, 1-month sales ($167 million) would suffice to conduct a mega-trial of 80 000 participants.
It’s an interesting idea. Let’s face it, though. What we’re talking about it a tax, or fee, on successful drugs to see if they have adverse effects:
The one outcome that should routinely be collected is death. For 80% power to detect a 10% relative risk reduction, sufficient follow-up is needed to achieve a 13.2% death rate in the control group with 10 000 participants per study group, or 3.3% with 40 000 participants per group. Furthermore, information also could be collected on major clinical end points that have not been adequately studied in previous trials of the blockbuster…
Besides mortality outcomes, mega-trials can focus on other specific major questions of interest. For example, for mental-health interventions, these trials would provide sufficient power to address the effects of drug interventions on important outcomes such as suicide attempts, hospitalizations, and job loss instead of relying solely on subjective scales and also would help define the spectrum of disease severity at which these treatments are effective, another issue debated endlessly based on small trials and meta-analyses thereof. Another 3 of the 24 top blockbusters are tumor necrosis factor-blocking agents (adalimumab [Humira], etenarcept [Enbrel], and infliximab [Ramicade]). Mega-trials would have provided data on the association between these drugs and the potential increased risk of serious infections and cancer and could have helped to resolve debates that still remain unsettled after almost a dozen conflicting meta-analyses.
Ioannidis also has some suggestions as to who could run such studies, and how a system could be set up. I encourage you to read the whole (short) piece. I’m sure the pharmceutical industry will howl, and others will start screaming about how much it already costs to bring a drug to market. But this is the kind of research that the government can’t afford to do, yet is still necessary to prevent some of the bad drug outcomes and recalls we’ve had in the past. At least it’s someone’s trying. I wouldn’t mind hearing some other ideas.
[Cross-posted at The Incidental Economist]