Why Clinical Trials Are Becoming More Expensive
6 CommentsBy Ed Silverman // May 13th, 2010 // 10:15 am
Containing drug development costs and speeding compounds through the pipeline is always a big issue, but clinical trials are becoming more expensive anyway. Why? One answer is the increasing complexity of the studies - the number of procedures for each clinical trial rose 49 percent from the 2000 to 2003 period to the 2004 to 2007 timeframe, and the total effort per protocol jumped 54 percent.
For instance, the average number of eligibility criteria used to screen volunteers rose 58 percent, which contributed to a 21 percent decline in volunteers enrolling in trials. But the larger number of procedures per protocol dissuades volunteers from completing trials - retention rates dropped 230 percent, according to the Tufts Centers for the Study of Drug Development, which reviewed data from 8,325 study protocols that was gleaned from a database from more than 75 drugmakers.
Breaking it down further, Phase I and II protocols experienced the biggest annual growth in complexity and execution as more data is gathered in these earlier phases. Similarly, there was a significant growth in Phase IV as more post-marketing data is collected for safety and marketing purposes (here is a press release; the report requires a subscription).
Protocols in anti-infectives, immunology, oncology, CNS and cardiovascular disease had the
highest total number of procedures in Phase I studies. For therapeutic areas showing positive growth in the number of procedures in Phase II protocol from 2002 to 2007 - hematology, dermatology and oncology - the growth in work burden exceeded growth in total procedures. And across all therapeutic areas, the total number of procedures grew at a 12.5 percent compounded annual growth rate.
And what about Phase III protocols? Immunology, CNS, oncology, anti-infectives, and endocrine disorders were the most complex as measured by the total number of procedures per protocol during 2002 and 2007. The average work burden per protocol for only three therapeutic areas - CNS, oncology and antiinfectives - exceeded the aggregate average across all therapeutic areas. And the typical Phase III protocol had an average of 148 total procedures more than five times the number of
unique procedures.
Phase IV protocols in endocrine disorders saw the fastest growth in total procedures and very high
growth in work burden. And in the 2002 to 2007 timeframe, the overall growth rate in total procedures in Phase IV studies for each of the five top therapeutic areas was dwarfed by the overall
growth in work burden.
pharmavet
Here’s a simple way to bring down costs. It is the very first lesson that I learned when I wrote my first protocol in 1983. A protocol should ideally be designed to test a single hypothesis, and use the fewest and most sensitive and specific outcome measures to answer this single question. My first job was running Phase IV clinical trials at Abbott. Believe it or not, our ENTIRE budget for that year was $1 million. We were very cost-conscious and negotiated with investigators and labs to get the best price. Despite budget limitations, we turned that $1 million budget into four new indications.
Notwithstanding inflation over the years, I think that the industry can still do cost-effective clinical trials,
MedicalQuack
Participants may be wearing devices too, which is another added cost. All throughout healthcare devices are starting to enter areas with data reporting where humans did the job before.
http://ducknetweb.blogspot.com/2009/12/41-of-canadian-clinical-trials-use.html
You could be wearing a wireless blue tooth shirt, or sleeping on a pillow that does wireless. The link above states that 41% of Canadian trials are now using devices so this also adds to the bottom line cost.
Salmon
Pharmavet,
That is a good suggestion. Studying multiple secondary endpoints and then trying to cherrypick them may be counterproductive in phase III. Although in phase II it is appropriate for hypothesis generation.
I have another reason. It’s inherent in the post itself. Don’t try so hard to exclude people.
Rather than re-write something. I will simply refer to a recent post I made on this topic at SGP (Shearlings Got Plowed).
https://www.blogger.com/comment.g?blogID=4241416962008169508&postID=3882062217833509262
patrons99
Salmon - thanks for the link to the SGP thread…very informative discussion.
A concern I’ve had for quite some while about the drug approval process, is there is no long-term data as to safety. Further, there is no requirement by FDA to even start collecting data on long term safety, all cause morbidity, mortality, and survival benefit,if any.
Another trend that worries me, is in the area of biologicals. Rather than using the previous “gold standard” of a double blind, placebo controlled prospective study, FDA is now allowing active comparators as the “control” and making the tacit assumption that the active control is completely safe, since it has already been approved. This assumption is often categorically false.
BTW - I didn’t know you are physician, until I saw your comment SGP.
pharmavet
Salmon, good post on SGP. I agree that attempts to ramp up exclusion criteria are misguided. My interpretation is also this: by ramping up exclusion criteria, one can study a more “pure” population, reduce the chance of a Type II error, lower the costs (since one has probably now chosen to power the study at 80% vs 90%)and increase the probability of success. This is a fool’s errand in my opinion.
When I was doing trials with antidepressants (which have a high placebo response rate), we changed the standard entry criteria from a screening Hamilton Depression Rating score of 21 to 27, the thinking being that the sicker the patients the more robust the response to active drug. The problem was the drug itself. If you don’t have an efficacious drug to begin with, no amount of pruning of the cohort will likely improve the outcome.
Human biological variability is an accepted fact. If someone wants to purify the population, they would be better served to place a call to Charles River and order a strain of inbred lab rats.
re Patrons’ comment, I really don’t know how you would conduct a comparative safety study and get anything but descriptive rather than inferential statistics. The only way to control it might be to include a placebo arm, which I think an IRB would approve, since we don’t know at the outset whether either of the two active comparators is safer than no drug at all.
Salmon
Pharmavet,
You make additional good points and I agree with most of them.
Ideally we want to show robustness with a drug, i.e. that it will work in the general population that it’s prescribed for and not a highly selected subpopulation.
As for the inclusion criteria for antidepressant studies, without data the hypothesis seems sound. However we now know that the antidepressant trials are more likely to show efficacy if sicker patients are enrolled. Consequently, rather than showing robustness the study design you describe may do the opposite, i.e. point out that a drug is marginally effective when the population as a whole is examined.