Saturday Archive: A Day Late and Half a Million Short?
When the math is so bad you don't even notice
Note: this was originally posted on LinkedIn on June 25, 2024. On Saturdays I will be migrating work from other platforms onto FPI.
If you take a Ford Edsel and you rebuild it – say you put some airbags in, add a USB plug and a satellite radio – then it’s fair to say you’ve updated the Edsel. But it’s hard to claim that you’ve really made the Edsel good. It’s true that a car with airbags is better than the same car without airbags, but a lack of airbags wasn’t really the problem with the Edsel in the first place. The first problem was the Edsel just wasn't a good car.
So anyway, a new analysis from Tufts CSDD, “How Much Does a Day of Delay in a Clinical Trial Really Cost?”, comes to a surprising conclusion: $500,000 a day in losses due to trial delays!
That conclusion is wrong.
It is not even close to right. In fact, it’s so far off that I think it didn’t even set off anyone’s WTF Meter. But it really is very, very wrong.
And the problem isn’t with Tuft’s numbers. The problem is their numbers are the airbags on the Edsel. Their work updates a previous analysis, and while their data may be more accurate than the original, they didn’t fix any of the flaws of the original analysis.
Future revenues are not guaranteed – or even likely!
The Tufts analysis looks at average sales of marketed drugs. But we’re looking at clinical trial delays, and most drugs in clinical trials won’t make it to market. Drugs in early phase trials may have a less than 1 in 10 chance of ever seeing a dollar in revenue, and you can’t lose what you were never going to make. It’s unclear what the average probability of success of all drugs currently in trials might be, but it’s certainly under 50% - meaning the analysis is already at least 2x too high.
Lost sales are not lost revenue
The analysis looks at future sales of drugs. But of course, future sales come with future expenses (manufacturing, distribution, sales guys, TV ads!). If you delay the sales, you also delay the expenses – you’re only “losing” the profit. Now, pharma’s pretty profitable, but industry EBITDA is somewhere around 30%, meaning only about a third of the day's sales can be considered "lost" in some manner. So again, the analysis is overstating costs by at least another factor of 3x.
Profit delayed is not profit denied
Finally, we have to realize that the delay didn’t zap these future revenues out of existence – they just shifted them back a day. And a drug that takes 8 years to get through clinical trials will almost certainly launch with the same total remaining patent life as a drug that took 8 years and one day. Now, inflation is higher these days, but with the current cost of money at around 5.5%, the difference between expected value of two future revenue streams that are off by a day is much smaller than a day’s worth of revenue. My back of the envelope calculation is that a day’s delay of profits is closer to 30% of a day’s revenue (and that’s very generously assuming the drug will launch and immediately start producing full profits). So we again see another 3x difference between this analysis and reality.
There are a handful of other, smaller issues with the analysis (like not recognizing costs as a combination of fixed and marginal), but just with the three factors above we can see the Tufts estimate of costs is already approaching being 20 times too high.
And that makes way more sense! If pharma companies were losing $500,000 a day on slow trials, then literally every one of them would be running Super Bowl ads for patient recruitment.
As the Tufts team mentions in the article:
There is no question that time is extremely valuable in drug development. It is also extremely important that drug development stakeholders use more accurate measures of the value of time for decision-making and planning purposes, and for quantifying the financial return on time saved through innovative approaches and new investments.
That's exactly right - but if you ignore risk of failure, don't distinguish between sales and profits, and don't account for time-shifted revenues, you can't get to an accurate measure.