NME Output versus R&D Expense – Perhaps there is an explanation
A popular display of innovative productivity in the Biopharmaceutical Industry has been the plot of New Molecular Entitities (NMEs) versus Total R&D spend in the U.S., Figure 1.
The response to this chart is straight forward. The number of NMEs per year was trending upward to 1996 but now in recent years then number appears to be going down. It would be good for NMEs per year to go up. Increased R&D spending has not helped. The R&D spend appears to be out of control.
Many have responded to this chart. It’s becoming a Rorschach test for the Industry. These references contain a partial listing where this graph, or a version of it, has appeared. (Ref 1, Ref 2), 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14
Tollman et al. prepared a plot of NMEs over a six year period (2002-2008) per R&D spend over a six year period (1998-2004) for the top 25 biopharmaceutical companies to identify companies with the greatest number of NMEs with the lowest R&D investment. 15
A common misrepresentation of the graph in Figure 1 is to plot NME vs R&D spend per year from 1996 (the peak year for NMEs) through to the current year, suggesting a distinct downward trend in NMEs per year. It’s more dramatic but misrepresents the fact that 1996, 1997 and 1999 were probaly anomalies (due to PDUFA, see later) and that NME output per year is more constant with perhaps a slight upward shift from around 18 per year to around 22 per year.
The increasing distance between NMEs and Pharma Spend in Figure 1 is often referred to as the “Innovation Gap”, e.g. Ref 12 .
This chart is often deployed as evidence for the industry needing to fix various things. The chart is often shown in relation to the number of high selling drugs that are nearing patent expiration and the lack of drugs to replace the loss of sales.
To this author’s awareness, no one has adequately explained this conundrum of flat productivity against an inexorable increase in R&D expense, with perhaps one exception. Bernard Munos published an interesting study late in 2009. (Ref 13)
He brings in three important measures to discuss the chart – a) the number of companies in the industry that registered at least one NME since 1950 (261 out of 4300 companies), b) the rate of NMEs per year for any given company and c) the difference in NME rate between small companies and large companies. Let’s examine each of these measures to see if they do explain Figure 1, and consider a few others.
Number of Companies.
Munos notes that of 261 that registered at least one NME since 1950, “only 32 have been in existence the entire period. The remaining 229 (88%) organizations have failed, merged, been acquired, or were created by such M&A deals, resulting in substantial turnover in the Industry. Of the 261 organizations total, only 105 exist today, whereas 137 have disappeared through M&A and 19 were liquidated.” (Ref 13) That’s a lot of change! Of course companies not only disappear, they also get created, for example in the biotech boom of the 80s. Interestingly, the number of these companies that were in existence in any given year (the red line in Figure 2c) follows fairly closely the plot of NMEs per year (the bold blue line in Figure 2c). He derived a measure called “expected NME output” that is based on the number of companies, Figure 2d. That plot overlays the plot of NMEs per year quite nicely, Figure 2c.
Figure 2 from Ref 13 . Figure 2c compares actual NMEs per year with Expected NMEs per year and number of companies per year as described in Ref 13 . Figure 2d compares Expected NME Output with Number of Companies as described in Ref 13 .
Munos postulates that mergers and acquisitions in the late 90s and early part of this decade have caused the decrease in number of companies, resulting in the decline of NMEs from the peak in 1996 and 1997. This is a potentially unique observation that would explain why NME production has declined in recent years. Of course, Munos recommends more companies are needed to replenish NME productivity.
There is a complexity hidden in the number of companies, however that challenges this simple interpretation. As we see in Figure 2c, year-on-year as the number of companies rises, the number of NMEs tends to rise, albeit with a variability in FDA approval rate. In fact the peak of NMEs/year corresponds to the peak year of number of companies. This does not square with the fact that the new companies are not likely to impact the number of NMEs in that year or for at least another 5-15 years, due to the time it takes to discover and develop new drugs. It is also unlikely that any promising projects would be scuttled in a merger or an acquisition, so M&A may not necessarily reduce the number of NMEs. There should be a time lag both in the increase and the decrease of NMEs/year due to the rise and fall in number of companies, Figure 3. Perhaps some companies do appear in the count years before they get an NME approval, but it is not clear from the discussion.
Figure 3. Expected NMEs with a ten year time delay to account for the 10+ years of R&D leading up to an NME from a new company and to account for the continuation of promising projects in a merged company. (Data from Ref 13 )
The Stable Rate of a Company’s NMEs/year
Munos postulated that companies tend to generate the same number of NMEs per year overall, with the rate remaining stable for a very long time. He first shows companies that have not engaged in dramatic M&A activity, Figure 4a. These companies in fact, have the same overall rate of NMEs per year, ~1 NME/yr. Other companies, that have engaged in significant M&A activity, show <1 NME/year, Figure 4b. With those companies there may be flat periods where no NMEs are approved, but after such flat periods a stable rate of NMEs/year gets established, often with a different slope. Wyeth showed a decrease in rate in the 70s, and BMS showed a decrease in the mid-90s, while J&J managed to ramp up the rate in the late 90s (albeit with a rather lengthy dry spell from ’98-‘04), and Pfizer showed a slight increase in rate in the 70s. Examinations of the histories of these companies do not show M&A around the periods of inactivity, so the lack of NMEs are not related to reorganization following M&A. A more likely explanation for a lack of NMEs is failure in the clinic or with approvals. The rate changes are intriguing and worth further examination. Figure 4b thus seems to contradict Munos’ hypothesis that companies tend to have a stable rate of NMEs/year.
Figure 4. Figure 4a and Figure 4b taken from Figures 2a and 2b in Ref 13 .
If it could be shown that the overall output of the industry is fairly stable, that would be a significant observation. It would suggest that there may be nothing that the Industry, as it currently operates, can do to increase NME output. As was discussed, Munos recommends more companies.
The fact that some companies have been able to increase their rate of NMEs/year for extended periods (Figure 4b) suggests that there may be organizational and process changes that can help increase productivity. J. P. Garnier, the former CEO of GlaxoSmithKline recommends an overhaul of R&D processes. 16
Impact of Company Size on NME Rate
Munos also examined the relationship of company size to NME rate. He took the top 15 companies and their predecessors as large. He found that from the 50s through the 80s, this cohort of large companies had produced ~75% of the NMEs/year, but in the 80s onwards that share has declined to ~35 % of the NMEs/year, Figure 5. In comparison the share of NMEs/year for small companies started out small, ~ 23%, but since the 80s has increased to nearly 70% of the industry output. He shows that this increase was due to an increase in the actual mean output of NMEs/year for small companies (see Figure 4b in Ref 13 ) as well as an increase in the number of small companies since the 1980s fueled by the biotech boom. He attributes the decrease in large company share of NMEs/year to the decrease in number of large companies, due to M&A in that sector. This finding does argue for more start-ups. But we’re talking about a lot of small companies – the curve in Figure 5 is fueled by 4300 biotechs.
Figure 5, Comparing the Percent of NMEs approved per year for Large Companies to Small Companies, Ref 13 .
Munos notes “by virtue of their number, small firms collectively can explore far more directions, and investigate areas that their larger, more conservative competitors avoid.” But he adds a note of caution “only a small fraction of these small companies will be rewarded with an FDA approval.”
This finding does argue for partnering between big Pharma and small companies, a direction that most big Pharma are taking. 17
The Impact of PDUFA
Munos noted that the NME peak in 1996-1999 can also be related to the enactment of the Prescription Drug User Fee Act (PDUFA) in 1992, which provided the FDA with sufficient funds to hire more staff to work on the backlog of approvals which had built up to that point. Thus the release of the backlog caused an unusual spike of approvals that tapered off in subsequent years. Let us suppose that the NMEs not approved in the low NME output years contributed to the spike of NMEs in 1996-1999. Let us subtract out those high numbers from that period and take them down to 30 NMEs/year in those years and back-fill the low NME output years from 1971 to 1980. We then get a curve, Figure 6, that is more like Munos’ “expected NME” output.
Figure 6, Hypothetically Adjusted NMEs per year obtained by taking the “extra NMEs” in 1996-1999 due to the backlog release due to PDUFA and back-filling the low years between 1971-1980.
We still have a minor increase in NMEs/year in the 90s compared to the 70s and 80s and since 2000. But any curve fitted to the adjusted NMEs would be less dramatic, in other words NME output would be relatively flat. Munos noted that any particular company displays a relatively stable NME output rate, the highest being companies that can achieve 1 NME per year, whereas many companies tend to achieve less than 1 NME per year. Thus the industry as a whole would be expected to display a rather stable rate of NMEs per year varying, as Munos observed, by the number of companies at play in any given period.
Graham evaluated the peaks and valleys in the NMEs/year, 18 and was able to develop an ARIMA model that predicts the peaks and valleys for the NMEs/year for all years except for 1996-1998 (Figure 17 in Ref 18) His analysis, however, suggests that these anomalies are due an increase in approval rates post-PDUFA, not due to an increased ability to work on a backlog of applications (p.84, Ref 18). Graham believes that the approval rate will settle into the modest increase established prior to 1996,
It should be noted that the number of NME filings has dropped from 50/year in 1995 to 24/year in 2003 (Ref 1). Thus the FDA does not appear to be the cause for the decrease in NME approvals since 1995.
The Rising Cost of R&D
So let’s get back to the logarithmic increase in R&D spend over the years. It is obvious by now that the increase in R&D spend has not influenced the relatively flat productivity of R&D.
Of course that increase is fueled by inflation and new technologies that come on board every year that require an increase in spend on capital equipment as well as an increase in expendible items needed to run the new equipment. In fact a common issue in any line department is the failure to increase the expendables budget in accordance with new equipment purchased in any given year.
But there are other aspects of this increase in R&D spend that need to be considered, which may actually better explain the flat productivity of R&D.
The Flatting of R&D Spend as a Percent of Sales
Cohen (Ref 1) found that while the R&D Spend per year is exponential, when considered as a percent of sales, it has actually leveled off since 1994, Figure 7. Sales continue to rise, but not at the rate of earlier decades due to patent expirations and the lack of new blockbuster drugs to add new sales, it would make sense drug companies would not be able to increase the percent of R&D in their budgets, especially if the companies are concerned about their own ability to create new drugs.
Figure 7, R&D Spend as a Percent of Sales, from Ref 3.
Of the dollars spent on R&D each year, Cohen also showed that the drug companies have changed their allocations to the various stages of R&D, Figure 8. Cohen pulled most of this data from past reports of the Pharmaceutical Manufacturers Association (PHRMA). We back-filled the last three available years of data from PHRMA annual profiles.
Figure 8, Pharmaceutical industry percentage allocation of R&D expenditures from 1976–2008. 1976-2002 data provided by PHRMA and analyzed in F. Cohen, 2004 (Ref 1). Data set is discontinuous. Non-clinical/preclinical is the term by PHRMA for Discovery and Preclinical Research. *Data supplied from PHRMA Profiles for 2008-2010.
This bar chart shows that in the ‘70s almost 70% of the R&D budget went into Discovery and Preclinical research. But look where the Industry has gone – since 2006 the Industry spent less than 30% of its budget on Discovery and Preclinical research. The lion’s share of the budget now goes into the clinical stages of R&D (The same trend was observed by Moses et al.) 19 ) Clinical research is getting more costly. Between the 80s and 90s Dimasi et al. estimated that capitalized clinical costs rose three times faster than preclinical (and discovery) costs. (Figure 2, Ref 4 )
Here is another telling graph, Figure 9. It shows that the overall R&D spend has leveled off since 2007.
So, if one needs more money to spend on clinical trials and one isn’t going to increase spending on R&D, it’s pretty obvious where the extra bucks will come from – Discovery and Preclinical Research!
There are a number of explanations for this dismal scenario. The PHRMA provides the following explanations. 22
a) Drug Development costs have gone up.
b) The complexity of clinical trials has increased.
c) New pharmaceutical medicines face competition after a relatively short period on the market.
d) There are earlier and more frequent patent challenges by generic companies.
e) Now days, nearly all first-in-class medicines being approved already had potential competitors in Phase II clinical testing.
f) Just two in 10 approved medicines produce revenues that exceed average R&D costs.
For all of these reasons, drug companies are forced to spend more on Development, and the only way they can do that with flat R&D budgets is to rob Discovery to pay for Development.
Where NMEs Come From – Discovery
One could argue that a lot of innovation occurs in the clinic. But an NME approval is the first in kind. It either came out the Discovery organization of the company that nursed it through clinical trials or was in-licensed from another company who discovered the drug in its own Discovery organization. The innovation that is drying up is the innovation that gives rise to new drug candidates – Innovation in Discovery.
We often hear of the pipeline running dry. You have probably seen the cartoon that shows a lot of projects at the beginning of Drug Discovery shrinking like a pipe constricting down at the point of entering preclinical development, shrinking further through the stages of clinical trials ending with a few drugs approved at the end. Here is my version, Figure 10.
Figure 10, A Hypothetical R&D Pipeline Based on Industry Attrition Rates by Stage, and One Launch per Year. http://www.portfoliomanagementsolutions.com/the-organization-of-pharmaceutical-rd/attrition/
It shows a pharmaceutical R&D pipeline based on Industry averages for attrition at each stage in the pipeline. It suggests that for one drug to launch from the pipeline a considerably larger set of projects is needed in Drug Discovery and Preclinical Research. 89% of the projects in the pipeline need to be in Discovery and Preclinical Research. Most companies want 2-3 NMEs per year, so their Discovery portfolios need to be considerably larger than that of Figure 10. But since Munos shows no one achieves more than 1 NME/year the rates of attrition are likely even higher than that shown in Figure 10.
We learned in Figure 9 that <30% of the Industry R&D budget is now allocated to fund almost 90% of the pipeline represented by Discovery and Preclinical Research. It can only do that either by dramatically lowering the cost of each project in Discovery and Preclinical Research or by totally revising its strategy for finding new drugs. In fact many drug companies have decided to get out of Drug Discovery all together. 23
Coincidentally, a number of academic labs are evolving into nonprofit Drug Discovery organizations. 24 Machin argues that the future of the drug industry may one in which major pharma exist only to fund late stage clinical trials and to market drugs. 25
NMEs per Year versus Percent of R&D Spend allocated to Discovery & Preclinical Research
In conclusion, the chart of NMEs per year and R&D Spend per year, Figure 1, while potent with imagery, actually tells the wrong story. The real story is Figure 11 which shows that NMEs per year have dropped while the percent of R&D spend on Discovery & Preclinical Research has declined.
Figure 11, Plot comparing Percent of R&D Budget allocated to Nonclinical (Discovery) and Preclinical Stages (Ref 1) versus NMEs/year (Ref 1-3). The discontinuous years shown correspond to the years shown in Ref 1. *Data supplied from PHRMA Profiles for 2008-2010. Correspondingly, certain years of NME production are not shown here (e.g. 1996 – 53 NMEs), thus the NME/year line plot differs from Figure 1.
Unfortunately, industry analysts tend to avoid Drug Discovery because the data from any company engaged in Discovery is proprietary. Since clinical trial information must be disclosed to the FDA and thereby becomes available to the public, industry analysts tend to study Drug Development and are inclined to draw conclusions from Development data. So industry analysts aren’t likely to point the finger at decreased spending in Discovery as the culprit behind a dry pipeline.
Certainly, improving success rates in the various stages of Discovery and Development will help improve productivity. Now that many of the hurdles to success have been removed with protein-based drugs, we expect to see an enhanced contribution of protein-based drugs to the productivity of the Industry. 26
We believe that the flat or decreasing productivity as measured by NMEs per year can only be remedied by a dramatic increase in Discovery funding to bring more first in kind candidates into Development. At least some companies, such as GSK, argue that they can’t spend more if they aren’t making more. Perhaps the new increase in academic discovery groups will help fill the gap. (Ref 17 Ref 12) And certainly increased partnering between big Pharma and small companies is likely to help.
 Munos notes that most drug companies aspire to produce 2 or more NMEs per year. He found that the best barely manage 1 NME/year.
- F J Cohen “Macrotrends in pharmaceutical innovation” Nature Rev. Drug Disc. 2005, 4, p78-84 ↩
- K J Kaitin “Deconstructing the Drug Development Process”, Clin Pharm Ther 2010 87 p356-361 ↩
- PHRMA “Pharmaceutical Industry Profile 2009”, www.pharma.org. ↩
- J. DiMasi, R. Hansen, H. Grabowski, L. Lasagna, “Cost of innovation in the pharmaceutical industry” J. Health Economics, 1991, 10 p107-42. (Perhaps the first use of this chart). ↩
- Price Waterhouse Coopers, “Biotech – Lifting Big Pharma’s prospects with biologics”, The MoneyTree Report, May 2000, www.pwc.com/en_GX/gx/pharma-life-sciences/pdf/biotech-final.pdf ↩
- “High Performance Drug Discovery, An Operating Model for a New Era” Accenture, 2001. ↩
- “The Changing Structure of the Pharmaceutical Industry” I. Cockburn, Health Affairs, 2004, 23, p. 10-22. ↩
- “A CBO Study – Research and Development in the Pharmaceutical Industry”, Congressional Budget Office, October 2006, Pub. No. 2589, p. 1-55. ↩
- M. Hu, K. Schultz, J. Sheu, D. Tschopp, “The Innovation Gap in Pharmaceutical Drug Discovery & New Models for R&D Success” March 12, 2007, Kellogg School of Management, www.kellogg.northwestern.edu/biotech/faculty/articles/newrdmodel.pdf ↩
- J. Sollano, J. Kirsch, M. Bala, M. Chambers and L. Harpole, “The Economics of Drug Discovery and the Ultimate Valuation of Pharmacotherapies in the Marketplace”, Clin. Pharm. Ther., 2008, 84, p. 263-266. ↩
- “Outlook 2009”, Tufts Center for the Study of Drug Development, http://csdd.tufts.edu ↩
- N. Campbell, “Mega Mergers – Are they turning Pharma Companies into Zombies”, Pharma Focus Asia, 2009, p. 8-14. ↩
- B Munos “Lessons from 60 Years of Pharmaceutical Innovation”, Nature Reviews Drug Discovery, 2009, 8, p. 959-968. ↩
- Greg Miller, “Is Pharma Running Out of Brainy Ideas?”, 2010, 239, p. 502-504. ↩
- P. Tollman, Y. Monieux, J. K. Murphy, and U. Schulze, “Identifying R&D Outliers”, Nat. Rev. Drug Disc. 2011 10, p. 653-654. ↩
- J. P. Garnier, “Rebuilding the R&D Pharma Engine”, Harvard Bus. Rev. 2008, 86, p. 68-70. ↩
- A. Mullard, “Partnering between pharma peers on the rise”, Nat. Rev. Drug Disc. 2011, 10, p. 561-562 ↩
- J. Graham, “Trends in U.S. Regulatory Approvals of Biopharmaceutical Therapeutic Entities”, M.A. Thesis, MIT Sloan School of Management, 2004. http://mit.dspace.org/handle/1721.1/30276. ↩
- H. Moses, E. R. Dorsey, D. H. M. Matheson, S. O Their, “Financial Anatomy of Biomedical Research” JAMA 2005 294 p1333-1342 ↩
- Pharmaceutical Research and Manufacturers of America, PhRMA Annual Member Survey (Washington, DC: PhRMA, 1981–2010). ↩
- Burrill & Company, analysis for PhRMA, 2005–2010 (includes PhRMA research associates and nonmembers) ↩
- PHRMA Chart Pack “Biopharmaceuticals in Perspective”, p. 21-27, 2010, www.pharma.org. ↩
- “Layoffs – They’re Happening, Big Time. http://www.portfoliomanagementsolutions.com/compelling-questions/lay-offs-in-big-pharma-are-happening-big-time-what-does-this-mean-for-portfolio-management/ ↩
- C. J. Tralau-Stewart, C. A. Wyatt, D. E. Kleyn and A. Aya, “Drug discovery: new models for industry–academic partnerships” ↩
- J. Dixon, G. Lawton and P. Machin, “Vertical disintegration: a strategy for pharmaceutical businesses in 2009?” Nature Rev. Drug Disc. 8, 435, 2009 ↩
- See “Small Molecule, Peptide and Protein-Based Drugs – The Differences and Similarities” at http://www.portfoliomanagementsolutions.com/the-organization-of-pharmaceutical-rd/small-molecule-peptide-and-protein-based drugs/ ↩