R&D costs for Gleevec

In a bid to generate sympathy for its defeat in the Supreme Court of India over efforts to evergreen patent protection for the cancer drug imatinib (trade names Gleevec or Glivec) in developing countries, Novartis has been making increasingly spectacular claims as regards its investments in the development of the drug. This blog reviews the independent evidence regarding these claims, which suggest that Novartis outlays on early research were relatively small, relative to investments by the NIH and the Leukemia Society, but that Novartis did play an important role when the products reached Phase II testing that led to the FDA approval of the drug on May 10, 2001. Depending upon the assumptions regarding the per patient costs for the Phase II trials, Novartis would have spent from $10 to $24 million out-of-pocket on the three Phase II trials that involved 1028 patients. After adjusting for the risks of failures and the opportunity cost of capital, the Novartis investments in the Phase II trials are estimated at $38 to $96 million.

I invite others to suggest better data and estimates, or advise me of errors.

Early research on Gleevec

Much of the early work on the development of Gleevec was done by Brian Druker of the Oregon Health and Science University and other academic researchers. Druker’s academic lab, working in a partnership with a scientist working for Novartis, identified the compound STI-571. They first used the compound to cure cancer in mice. Dr. Drucker oversaw the first clinical trial, and other early Phase I clinical trials. He also participated in each of the clinical trials cited in the FDA approval of the drug for treatment of Chronic Myeloid Leukemia. While he was developing the drug his laboratory’s funding sources were:

  • 50% National Cancer Institute
  • 30% Leukemia and Lymphoma Society
  • 10% Novartis
  • 10% Oregon Health and Science University

In a book review published in JAMA, Dr. Arnold S. Relman from the Channing Laboratory at the Harvard Medical School wrote:

Novartis was not “the innovative force.” Not only was all the basic research done in academic institutions, but so were the initial clinical investigations that showed STI 571 to be specifically effective against CML cells in vitro and in vivo. In fact, it took a few years for Brian Druker, the investigator most responsible for these latter studies, to convince Novartis that it should invest in a crash program to develop Gleevec and to undertake large-scale clinical trials.

Relman, A. (2003). Book Review: Magic Cancer Bullet: How a Tiny Orange Pill Is Rewriting. Medical History. JAMA, 290: 2194-2195

The NIH funding specifically associated with the early testing of Gleevec included grants from the National Cancer Institute (CA65823, to Dr. Druker, and CA32737, to Dr. Sawyers). As noted above, the Leukemia Society and others were also supporting this work. Additional details on the background on the early development of Gleevec are found in this 2001 NIH Timeline on Gleevec development. http://www.cancer.gov/newscenter/newsfromnci/2001/gleevectimeline Druker’s NEJM paper reporting the early trials is available here http://www.nejm.org/doi/full/10.1056/NEJM200104053441401. There is a discussion of the NIH role in Gleevec development in this June 21, 2001 NIH testimony: http://www.hhs.gov/asl/testify/t010621a.html. Michael Palmedo’s 2002 email exchanges with Druker are reported here: http://cptech.org/ip/health/gleevec/drucker.html.

Bottom line: The NIH and charities like the Leukemia Society supported most but not all of the early research on Gleevec.

The three Phase II trials cited in the 2001 approval

After Druker established that STI 571 was a promising drug, Novartis was convinced to invest in further testing, and most importantly, this included three Phase II trials with a total of 1028 patients.

The outlays on the Phase II trials can be estimated by multiplying the number of patients by industry averages for per patient costs of trials.

In his widely cited 2003 paper on drug development costs, DiMasi’s average cost per patient across Phase I-III trials was $23,572 per patient. (J.A. DiMasi et al. Journal of Health Economics 22 (2003) 151–185). Joe has described that to me as the fully loaded costs including company overheads and analysis. But let’s also look at some other data points. In 2002, Robert Kramer, the BMS Vice President for Oncology Drug Discovery and New Business Ventures was quoted as estimating the average cost of oncology trials to be $10,000 per patient. (Parexel Pharmaceutical R&D Statistical Sourcebook 2002, page 117). In 1999 the National Institutes of Health reported the costs of oncology trials funded by the NIH DCP Cooperative Group.

NIH DCP Cooperative Group Treatment Trials: cost per patient, 1993 to 1999
FY93 $ 3,861
FY94 $ 4,384
FY95 $ 4,692
FY96 $ 5,297
FY97 $ 4,919

Bottom line: If you use the per patient cost from the 2003 DiMasi paper, the Novartis outlays on the Phase II trials are estimated to be $24 million. If you take the BMS estimate, the cost is $10 million. If you use the NIH cost numbers, it shrinks further.

Orphan Drug Designation

Novartis has obtained Orphan Drug designations for seven indications, all of which have received marketing FDA approvals, including for chronic myelogenous leukemia, for the May 10 , 2001 FDA marketing approval. The Orphan Drug designation gives Novartis the right to take a 50 percent tax credit for qualifying trials for that indication, subject to certain conditions, including that testing occurs after the date of the Orphan designation and before the FDA approval for that indication.

Date of Designation and the Orphan Indication

  1. 01-31-2001 Treatment of chronic myelogenous leukemia
  2. 11-01-2001 Treatment of gastrointestinal stromal tumors
  3. 08-25-2005 Treatment of idiopathic hypereosinophilic syndrome including acute and chronic eosinophilic leukemia
  4. 09-09-2005 Treatment of systemic mastocytosis without the D816V c-kit mutation
  5. 10-05-2005 Treatment of myeloproliferative disorders/myelodysplastic syndromes associated with platelet-derived growth factor gene re-arrangements
  6. 10-11-2005 Treatment of Philadelphia-positive acute lymphoblastic leukemia
  7. 12-19-2005 Treatment of dermatofibrosarcoma protuberans

Bottom line: Novartis claims that it did not benefit from the credit in its 2001 approval, and the tax credit does not enter into the cost estimates for the May 10, 2001 approval. The credit has likely been a major subsidy for Novartis going forward from the initial 2001 approval, for testing following the six subsequent Orphan designations.

Risks of Failure

One might want to adjust the Novartis outlays for risk. For Orphan designated products as a whole during the period 1990 to 2000, there were 687 designations and 159 approvals — or a rough success rate of 23 percent, compared to designations. (Looking at all designations and approvals through to the present, that success rate falls to 15.4 percent).

But note that these risks decline as products pass various benchmarks, and for products that enter Phase II testing, the risks are lower.

In some earlier Tufts estimates, around the time of the Gleevec approval, slightly more than half the products that enter Phase II testing fail at that stage.

In DiMasi’s 2003 paper, he assumed an overall success rate of 21.5 percent for products that entered clinical testing, and a 71 percent chance that products entering clinical testing would reach Phase II.

“Although average cost estimates for investigational drugs are interesting in their own right, we are mainly interested in developing estimates of cost per approved new drug. To do so, we need an overall clinical approval success rate. Our statistical analysis of compounds in the Tufts CSDD database of investigational drugs that met study criteria yielded a predicted final clinical success rate of 21.5%. Applying this success rate to our estimates of out-of-pocket and capitalized costs per investigational drug results in estimates of cost per approved new drug that link the cost of drug failures to the successes.”

DiMasi estimated the probability that a product that entered clinical testing would entry Phase II to be 71 percent. From this data, the chances of FDA approval from Phase II are the overall rate divided by the chance of entering Phase II, or .215 / .71 = .303. For a Phase III trial, the odds of success in DiMasi’s 2003 paper were .215 / .314 = .685.

DiMasi’s sample was not focused on orphan products, or products that had received an FDA priority review. By simply using the .303 success assumption for Phase II trials, the risk adjusted costs of the Phase II trials would be:

Risk Adjusted Cost for 1028 patents in Phase II trials

Cost per patient Risk adjusted cost @.303 success rate
DiMasi @ $23572 $80 million
BMS: @ $10000 $34 million

Bottom line: The risk adjusted cost of the Phase II trials can be estimated at somewhere between $80 million and $34 million, with the higher number inclusive of allocations for hefty overheads and analysis costs.

Cost of capital

DiMasi and other industry consultants often consider an additional adjustment for the costs of capital, which is reasonable, assuming people understand what is going on. In general, you take outlays and increase them by some rate of return percentage every year, to address the opportunity cost of capital, since costs are incurred over time. Unfortunately, many people, including nearly all financial press journalists have a hard time presenting any of the adjustments (risk or capital cost) in a way that does not lead to a type of mental double counting (more than double, actually), since whatever number you finally get is almost always reconverted in people’s minds as simple out of pocket expenses (which is a big mistake).

To do a cost of capital adjustment, you have to know how many years to adjust for opportunity costs. In his 2003 paper, DiMasi reports that “the start of clinical testing to marketing approval in our timeline for a representative drug averaged 90.3 months.” DiMasi reported “Capitalizing out-of-pocket costs to the point of marketing approval at a real discount rate of 11%.”

Note that Gleevec began testing in the 1998 and was first approved for marketing on May 10, 2001. If clinical testing started in June 1998, that is less than three years, or just 35 months, more than 2.5 times as fast as the “representative drug” in the DiMasi paper. The Phase II trials were started after 1998. I don’t have the start and stop dates for the trials, it probably makes sense to increase the Phase II outlays by somewhere between 11 and 20 percent, assuming one uses DiMasi’s 11 percent real return discount rate.

Bottom line: After taking into account both the risk of failures and the cost of capital at an 11 percent rate rate of return, the adjusted costs of the Phase II trials can be estimated at somewhere between $96 million and $38 million, with the higher number inclusive of allocations for hefty overheads and analysis costs and the higher end of the cost of capital adjustment.

Note that Novartis sales for Gleevec in 2012 were $4.675 billion, or $390 million per month. In 2012, Novartis realized more than $100 million in Gleevec sales every 13 days.

I discuss this further, and compare the Gleevec facts to the 2003 DiMasi study, here:

Some older references

A Note on Dr. Brian Druker’s Involvement in the Research and Development of Gleevec. 2002.

Norvartis Outlays on R&D R&D For Glivec: Evidence suggests outlays are substantially below average costs estimated by Tufts University Study, September 22, 2000

The 2001 NIH Timeline on Gleevec development

Relman, A. (2003). Book Review: Magic Cancer Bullet: How a Tiny Orange Pill Is Rewriting. Medical History. JAMA, 290: 2194-2195


James Love

James Love is the Director of Knowledge Ecology International. Previously, he was an economist for the Center for Study of Responsive Law where he also directed the Consumer Project on Technology and the Taxpayer Assets Project, Senior Economist for the Frank Russell Corporation, and held lecturer positions at Rutgers and Princeton Universities. His KEI webpage is https://keionline.org/jamie.