[Ip-health] The variables in Dr. Peter Bach's Drug Abacus

Jamie Love james.love at keionline.org
Fri Jun 19 12:14:10 PDT 2015

Below are the variables in Dr. Peter Bach's Drug Abacus pricing model.

I like his proxy for R&D costs, as a starting point, which is the
enrollment in the trials for the lead indication.   Eventually it will be
helpful to have some data on per-patient costs for different types of
trials, some metrics of the length of the trials,  and information about
government or charity subsidies for R&D, including the 50 percent Orphan
Drug Tax Credit (9 of 10 cancer drugs approved in 2014 qualified).

The number of human subjects enrolled in the approval trials for the first
indication (including FDA mandated confirmatory trials) was used as a proxy
for the research and development costs necessary to develop the drug. This
imperfect surrogate is reasonable as in most cases the cost of human
subjects research is a major component of overall R&D expense.

I also liked the novelty variable.   What do others think?


Sloan Kettering Drug Abacus


Selection of Drugs and Indications

This version of DrugAbacus covers 54 cancer drugs approved between 2001 and
2015 by the US FDA for the treatment of cancer. This subset of all drugs
approved during that time period begins with the 2001 approval of Gleevec
and was compiled from the work of Howard et al for drugs approved through
2013, and supplemented with all drugs with first approval for the treatment
of cancer subsequent to that time period. All characteristics of drugs are
based on information related to the first approved indication.

Current Prices

Estimated monthly prices for chemotherapy drugs are based on the allowable
Medicare charge, and calculated according to a methodology used in prior
reports.1 Since 2005 Medicare has reimbursed at 106% of the average sales
price (ASP) for Part B drugs. ASPs are reported in quarterly files released
by CMS.2 For Part D drugs, current prices are retrieved from Medicare’s
publicly available web-based “PlanFinder” tool.3 The price reported is the
“Full Cost of Drug” as reported in the PlanFinder for the Humana PDP
Enhanced plan, for a beneficiary living within ZIP code 10021. Payment
limits for prior years vary and are described briefly in the table below
and in more detail within the previously mentioned article.1

If a drug’s ASP is not available Medicare calculates the payment limit as
95% of the Average Wholesale Price (AWP). ASP and AWP for Synribo (a part B
drug approved in October 2012), Xofigo (a part B drug approved in May of
2013), Cyramza (a part B drug approved in April of 2014), Opdivo (a part B
drug approved in December of 2014) and PlanFinder listings for Gilotrif (a
part D drug approved in July 2013), Lenvima, Ibrance and Farydak (part D
drugs approved in February 2015) were not available. Estimated monthly
prices reported in the media at the time of drug approval were used for
these drugs. The reduction in reimbursement due to the 2013 ‘sequester’,
which lowered reimbursement to ASP+4.2%, is not included in these

For drugs that can be taken for an indefinite period of time, the relevant
payment limit is applied to a 12 week dosing regimen for an “average” adult
weighing 70kg, or with a body area of 1.7 meters squared, and divided by
2.77 to arrive at a monthly price (there are, on average, 2.77 months in 12
weeks). The 12 week dosing regimen is retrieved from the FDA approved label
for the drug, these labels are available via the FDA’s “Drugs at FDA”
database.4 The lowest total dosing regimen within the first FDA approved
indication for the drug is used in all cases. The prices shown are for the
listed drug only, costs for supportive care or administration fees are not
included. For drugs with as set regimen given over a pre-defined period of
time, such as the drugs Provenge or Bexxar, the monthly price is determined
by dividing the cost of the regimen by the number of months the regimen
takes to administer.

Abacus Price

That Abacus price is calculated from the following formula:   (check the
web page if the list-serve strips out the non-standard ASCII characters

Price=(β_eff 〖∙X〗_eff)(1-(β_tol 〖∙X〗_tol))(β_nov 〖∙X〗_nov)(β_(R&D)
〖∙X〗_(R&D))(β_rar 〖∙X〗_rar)(β_bur 〖∙X〗_bur)

Where each X is a measurement for one of six “domains” (efficacy,
tolerability, novelty, research & development costs, rarity and population
burden) and each β is a weight defining the importance of that domain with
respect to the drug’s price (chosen by the user within a predetermined
range of possible values). These domains, and their measures, are described
in more detail below.

The Abacus price is relevant for a treatment period equivalent to the
typical duration of treatment that was required to achieve the reported
benefit in the FDA approval trial(s). Model inputs (particularly efficacy
and tolerability) correspond to the length of treatment received by
patients in the relevant clinical trial(s). Therefore, we consider
model-calculated prices to be the price for the duration of treatment used
in the clinical trial and adjust to a monthly price via division. More
detail on treatment duration is provided below.


The efficacy of the drug is measured as the improvement in overall
survival, or a surrogate for this endpoint, attributable to the drug, as
measure in the highest level of evidence clinical trial that led to
FDA-approval for the first indication. A level-of-evidence grade is applied
to the measure of overall survival benefit, such that two drugs with
equivalent trial results will receive different efficacy grades if the
trial for one drug was of higher quality. In some cases, drugs are approved
by the FDA without evidence of an overall survival gain, on the basis of an
improvement in either progression-free survival or in response rates. We
considered the margin of gain in progression free survival to be equivalent
to the gain in overall survival if overall survival data were not
available, but the endpoint progression free survival received a lower
level of evidence rating. When other endpoints, such as response rates or
single arm trial endpoints were all that were available, these were
converted to estimates of overall survival benefit using available
literature from studies of analogous treatments.


A drug’s toxicity was characterized from the listing of the frequency and
severity of side effects experienced by patients receiving the drug,
relative to the severity of side effects those patients would otherwise
experience. The measurement for this domain consists of a combination of
two components. The effect of the drug on the probability of a given
patient experiencing severe side effects (proportion of patients in the
treatment arm experiencing grade 3 or 4 side effects, minus the proportion
in the control arm) and the effect of the drug on the probability or
discontinuing use of the drug due to severe side effects (again, difference
in proportions by trial arm).


The novelty of each drug was scored by two clinical experts, in line with
their recent related research. The experts classified each drug to one of
three groups based on its mechanism of action. The three groups are: 1)
Novel mechanism of action (score of 1), 2) Drugs with known target but
novel delivery (eg. Capecitabine) Drugs with known target but different
mechanism of targeting (eg. TKI vs. MAB, antibody drug conjugates etc)
(score of 0.5), 3) Next-in-class (score of 0). The raters assigned
conflicting scores to two drugs, Ixempra and Halaven, in each case the
more-novel score was assigned to the drug.

The number of human subjects enrolled in the approval trials for the first
indication (including FDA mandated confirmatory trials) was used as a proxy
for the research and development costs necessary to develop the drug. This
imperfect surrogate is reasonable as in most cases the cost of human
subjects research is a major component of overall R&D expense.


Rarity of the disease was determined from the projected incidence of the
disease in 2015 as per the American Cancer Society Facts & Figures Report.

Population health burden

The population health burden of the disease the treatment targets in its
first approved indication was determined from the estimated years of life
lost due to the disease in the US population, a statistic defined as the
average difference between life expectancy at death and age at death for
individuals dying from the disease in question during a specific period of
time. This statistic is reported in the SEER Cancer Statistics Review for
many diseases; where not, we estimated the years of life lost by
replicating the SEER methodology using information on the count of deaths
at various ages from the CDC’s WONDER database along with estimates for
life expectancy by age and gender, also from the CDC.


Treatment Duration: Treatment duration is taken from the FDA approved label
or trial publication. In cases where treatment duration is not available
the average progression free survival in the treatment arm of that trial is
used as a substitute.

Total Sales: Calculated at the drug level as consensus estimates for
expected total US sales in the year 2015 according to the EvaluatePharma
database, courtesy of Defined Health. Estimated sales at Abacus prices
assume that the volume of sales of each drug is not altered by a change in


Bach PB. Limits on Medicare’s ability to control rising spending on cancer
drugs. N Engl J Med. Feb 5 2009;360(6):626-633.

Center for Medicare and Medicaid Services. Medicare Part B Drug Average
Sales Price. Manufacturer reporting of

Average Sales Price (ASP) data.

McrPartBDrugAvgSalesPrice/index.html. Accessed 10/18/2013.

Center for Medicare and Medicaid Services. Medicare Plan Finder.
https://www.medicare.gov/find-a-plan/questions/home.aspx. Accessed

Food and Drug Administration. Drugs at FDA: FDA Approved Drug Products.
http://www.accessdata.fda.gov/scripts/cder/drugsatfda/index.cfm. Accessed

James Love.  Knowledge Ecology International
KEI DC tel: +1.202.332.2670, US Mobile: +1.202.361.3040, Geneva Mobile:
+41.76.413.6584, twitter.com/jamie_love

More information about the Ip-health mailing list