Download Now

WhitePaper
Market Research Meets
Market Reality:
A Dynamic Approach to Discounting
Physician Preference Share
2011
Kim Morneau
MARKET RESEARCH MEETS MARKET REALIT Y
Market Research Meets Market
Reality: A Dynamic Approach
to Discounting Physician
Preference Share
As the pharmaceutical market has evolved at tremendous
speed, so too have our measures for understanding
physicians’ preference for treatment options – but the
same cannot be said of approaches for discounting
preference share.
Experience shows that the key to success here lies in a
dynamic approach grounded in both market research and
market reality…
Understanding physicians’
preference towards
treatment choices
The founding principle
Conjoint methodology
For the last three decades, one key quantitative approach
has enabled us to better understand physicians’ preference
for treatment options: conjoint methodology.
Introduced by Green and Rao in ‘71, conjoint studies allow
physicians to view various combinations of product
attributes intended to represent potential ‘target product
profiles’ (TPP’s). After viewing each TPP, physicians are
asked to indicate their level of preference, and/or rank
them based on their likelihood to prescribe.1
In doing so, they offer us some insight into how that
product will perform in-market.
2
Copyright ©2012 Ipsos. All rights reserved.
MARKET RESEARCH MEETS MARKET REALIT Y
Evaluating the different
discount methods
The second generation
Discrete choice models
This approach served us well enough until the ‘90’s, when
a new innovation fuelled the next generation of conjoint.
Thanks to the introduction of web-based data collection
and software for analysing responses, ‘Discrete Choice
Models’ were born.
Regardless of survey design, market researchers
unanimously agree that preference share does not equal
peak market share. The reason? Unlike the real world, the
test environment is devoid of any market barriers, be they
physician awareness, formulary access, competitive counter
detailing, promotional spend, or others.
Although a fairly similar animal to the conjoint design, the
discrete choice model has one key difference. While both
methodologies enable physicians to view a series of
potential TPPs for a new drug (in which product attributes
are always varied), a discrete choice model requires the
physician to choose the profile they most prefer.
So how do we account for the market barriers that will
inevitably be present inmarket?
The ‘rule of thumb’ method
Useful but limited
The most simplistic approach is ‘rule of thumb’. This sees
claimed preference share discounted by 50-70%, a figure
derived by comparing claimed preference share from
market research, and actual market share at the point of
peak sales. This is based upon finding the average variance
from a large sample of case studies.
In forcing the physician to choose a single TPP instead of
providing a rating for each, discrete choice is thought to
more accurately reflect preference among a specific set of
product attributes.
There still remains an inherent difficulty, however;
physicians rarely select a single drug to treat all patients.
Instead, they choose from a small ‘armamentarium’ that
best matches patient characteristics and product attributes.
The benefits of this approach are fairly clear; it’s easy to
use, and the calculation can be done quickly with limited
data and expense.
A fusion of the two
The choice-based model
The drawback, however, is its inability to correlate, directly,
the discount rate with influencing factors such as the level
of promotional spend or the number of competitors in the
marketplace. Additionally, it implies that all physicians will
become prescribers of the product, which we know is not
necessarily true.
In attempts to overcome this limitation, the past ten years
have seen further evolution of the conjoint methodology
into a fusion of traditional conjoint and discrete choice. This
is known as the ‘choice-based model’.
Like its predecessors, the choice-based model exposes the
physician to a series of TPP’s with varying product attributes.
After viewing each TPP, the physician completes an allocation
table. Herein, they distribute their next one hundred patients
with a certain condition among all available treatment
options in the competitive set, including the test product.
The result is what we call ‘preference share’. Specifically, this
measures the depth to which physicians are interested in
prescribing the product in question.
According to a retrospective analysis conducted by Sobel
and Brodsky, while this approach may be fairly accurate in
aggregate, it tends to be vastly incorrect for individual
observations.2
This is certainly a useful measure, but one that’s still
flawed; in spite of the constant evolution of methodologies
for predicting preference share, we still see in all conjoint
designs an overstatement of modelled preference relative
to actual market share.
Acceptance of these overstated results can have significant
consequences for an organisation. It may inappropriately
prioritise one molecule over others in a R&D pipeline,
result in the purchase of unprofitable asset, or cause an
incorrect allocation of promotional resources. The key,
then, is to find an effective approach to discounting survey
results – but, unfortunately, these have failed to evolve at
the same pace…
3
Copyright ©2012 Ipsos. All rights reserved.
MARKET RESEARCH MEETS MARKET REALIT Y
Getting more sophisticated
Measuring breadth and depth
There is a more sophisticated approach that is commonly
used in the pharmaceutical industry, which accounts for
both breadth and depth of prescribing.
This calibration typically discounts both the percentage of
physicians who are likely to prescribe the product (breadth)
and the share of patients they will prescribe it for (depth).
Usually, this calibration methodology applies different
discount factors depending on the level of commitment the
physician has indicated about adopting the product into his
or her armamentarium.
Here’s a basic example:
MD
Calibration
Factor
Preference
Share
Calibration
Factor
Claimed
Preference
Share
Peak
Share
Intention
To Prescribe
% of MDs
“Definitely will”
25%
X
0.9
X
36%
X
0.9
=
7.3%
“Probably will”
42%
X
0.4
X
26%
X
0.3
=
1.3%
“Might or might not”
19%
X
0.05
X
15%
X
0.1
=
0.1%
“Probably will not”
10%
X
0.0
X
10%
X
0
=
0%
“Definitely will not”
4%
X
0.0
X
5%
X
0
=
0%
Adjusted Peak Preference Share
8.7%
The disadvantage of this type of calibration technique is that
it takes a significant amount of data to achieve reasonably
accurate discounting factors. Experts recommend, at
minimum, preference share data from primary market
research studies going back at least five years spanning
multiple brands, therapeutic categories and physician
specialties – and, on top of this, actual market share for the
same brands from a secondary data provider.3
Given that calibration factors can vary by therapeutic class,
physician specialty and geographic location of the
respondent, we need this level of data for each country in
which we wish to calibrate preference share.
4
Copyright ©2012 Ipsos. All rights reserved.
MARKET RESEARCH MEETS MARKET REALIT Y
A new way of discounting
preference share
The inherent problem
Historic analogues and why more isn’t always better
Regardless of how sophisticated we become at deriving
calibration factors, and how much granularity we can
achieve, both of these methods suffer from the same
inherent downfall: their reliance on historic analogues to
derive a calibration factor…
A truly dynamic approach
Identifying market barriers, local and global
There is, however, a more dynamic approach to preference
share conversion – that is, to identify the barriers likely to
hinder prescribing behaviour and estimate the percent of
physicians or patients who will be impacted by each barrier
on an ongoing basis.
Some market researchers tend to believe that the larger the
number of observations used to derive conversion factors,
the better. We’ve all seen leading research companies tout
20+ years of examples, spanning 30+ therapeutic
categories, clearly inferring that their robust number of
observations will lead to a more accurate calibration factor.
In some cases, this may be true; in others, being so focused
on the past can actually skew the findings.
This approach should account not only for the global
barriers that will be present in all geographic markets and
therapeutic areas, but also the localised barriers that most
brands encounter. These local barriers are specific to the
healthcare system where the product is being marketed,
and also the distribution channels through which it is
dispensed. The table below outlines some of the most
common barriers that cause market share to be lower than
claimed preference share.
To explain: the calibration factors derived from these data
sets, even the more targeted ones, are only meaningful if
the pharmaceutical business model, and the framework in
which physicians make treatment choices, are the same
now as they were when the analogues were derived. By
contrast, we know that in many geographic markets – both
developed and emerging – the regulations for promoting
pharmaceuticals, as well as reimbursement policies, have
changed significantly over the past 10 years. For example,
the introduction and evolution of the EFPIA code in Europe,
and of the PhRMA code in the US, have significantly
restricted the sales and marketing tactics employed by
pharmaceutical manufacturers. In our experience, this, in
turn, has impacted both the percentage of physicians who
become aware of new products, and the speed at which
physicians adopt them into their treatment paradigm.
Global Barriers
Local Barriers
Physician Awareness
Product Availability
Physician Adoption
Product Affordability
Incorporating market reality
Applying real-world knowledge
In a perfect marketplace, each barrier would equal 0% thus causing no deflation to preference share. However,
with the exception of orphan drugs, branded pharmaceutical
products rarely compete in a market free of restrictions.
What’s more, the magnitude of each barrier changes
throughout the life of a drug. For example, during the first
six months that a product is available, we can see
significant changes to the rates of each barrier on a weekly
or monthly basis, especially for physician awareness and
product availability. Physician adoption and product
affordability, however, tend to evolve more slowly during
the first 18 - 24 months.
5
Copyright ©2012 Ipsos. All rights reserved.
MARKET RESEARCH MEETS MARKET REALIT Y
The 4 key barriers
A closer look
Physician awareness tends to build in a logarithmic pattern,
with the majority of brands exceeding 50% awareness by
the eighth month in-market.8 In addition to cumulative
detailing reach, our experience shows that unmet medical
needs in a therapy area also dictate how fast awareness
will build. For example, awareness of a new oncology
brand typically builds the fastest, with many physicians
aware of the brand even before the start of promotion by
the manufacturer. Barring differences in unmet medical
needs, however, there appears to be nominal variance in
awareness uptake across therapeutic classes.
To better understand this approach, let’s look at each of
the barriers in turn:
1. Physician awareness: a fundamental barrier
Physician awareness of the brand is the most fundamental
barrier to gaining market share. After all, if the physician
cannot recall the brand name unaided, it is unlikely to be
part of his or her set of treatment options when writing a
prescription. A good example of this can be seen in a
retrospective analysis of more than 300 brands launched
since the EFPIA code was introduced in Europe; less than
6% of the brands achieved physician awareness greater
than 90%.4 Similar results can be seen for brands
launched in the US since 2003 when PhRMA guidelines
were rolled-out.
If we can approximate the percentage of physicians who
are likely to become aware of the product through detailing
and other promotional avenues, we can estimate the
percentage of overstatement due to ‘imperfect awareness’.
Accordingly, preference share should be discounted by the
percentage of physicians who will not become aware of the
product. Simply put: the physician cannot prescribe the
product if he or she is unaware of it or unable to recall the
product name.
Despite restrictions on promotion, however, physicians’
cumulative spontaneous brand awareness continues to be
highly correlated to cumulative detailing reach during the
first 24 months that the brand is in-market (Pearson
Correlation = 0.942).
To put some figures to this: of the brands launched during
the past 20 years, less than 10% detailed more than 80%
of the physicians in the prescribing universe.5 In fact, on
average, marketing teams deploy enough sales force to
cover around 60% of prescribers in the marketplace, who
manage 70-80% of the patients within the disease state6.
However, this does not necessarily mean that awareness
will automatically be capped at 60-70%. It is common to
see physician awareness of a brand exceed the percentage
of physicians in the universe who received a detail by
7-10%.7 This can be attributed to the other ways in which
a physician may become aware of the product: reading
about it in a medical journal, seeing a presentation of the
data at a clinical symposium, hearing about it from a
colleague, etc.
6
Copyright ©2012 Ipsos. All rights reserved.
MARKET RESEARCH MEETS MARKET REALIT Y
2. Physician adoption: multiple influencing factors
What causes some products to be adopted within months
of launch and others to take several years? The variation in
physician adoption can be attributed, in part, to certain
product characteristics. When it comes to oncology
compounds, our experience shows that adoption rates are
impacted by many factors, including the number of
alternative treatments, tumour type, availability of a
biomarker test, therapeutic class, and more…
Although physician awareness is driven largely by detailing
reach, the rate to which physicians begin prescribing a new
brand is mainly driven by the characteristics of the product
and, to a lesser degree, by detailing frequency.
A recent analysis of oncology compounds launched in the
US and 5 European markets since 2003 showed three
distinct trajectories for physician adoption – depending on
the market environment and certain product characteristics.
On average, peak share for the first indication of these
compounds occurred around 2.4 years after launch. Others
followed a more traditional 5-year-to-peak trajectory. A
proportion of these brands, however, achieved peak share
for the initial indication in less than 18 months.9
Number of treatment alternatives: For a first
indication, the number of treatment alternatives is one of
the most influential factors on the speed of adoption.
When the number of competitive alternatives is limited to
two or less, physicians begin prescribing the new drug
quickly and peak sales are generally achieved in under two
years. However, when a new drug enters a crowded
marketplace and offers little improvement in either efficacy
or safety, physician adoption is inevitably slower. Under
these conditions, achieving peak sales can take upwards
of three and a half years.10
Tumour type: Tumour type also plays a role in
determining how quickly physicians adopt a new oncology
drug. Newly-approved drugs to treat solid tumours reach
peak share in approximately 2 years, while drugs to treat
haematological cancers take nearly a year longer.11
Availability of a biomarker test: Similarly, the
availability of a biomarker test to predict the likelihood of
response to certain drugs also impacts the rate of physician
adoption. Whilst the introduction of biomarkers into the
marketplace has allowed physicians to better segment their
patients, it also appears to slow physician adoption down.
A drug with a biomarker test typically takes about six
months longer to reach peak share than one without a
qualifying test.12 One hypothesis for this is that the
biomarker pre-empts physicians from using that drug for
all patients, and there is a time lag between testing and
treatment for those specific indications.
Source: Ipsos Healthcare Global Oncology Monitor
Therapeutic class: Finally, the rate of physician adoption
can also vary by therapeutic class. Historically, physicians
have more readily started prescribing new products for
diseases in which there is a visible symptom that can be
monitored to ensure the drug is working – and, of course,
when there are limited consequences if the product does
not perform in line with clinical trial results. After 24 months
in-market, over 40% of physicians who were aware of new
drugs for pain management and birth control had prescribed
the product for at least one patient; by contrast, only 24%
of the physicians who were aware of new respiratory and
metabolic products had ever prescribed them.13
7
Copyright ©2012 Ipsos. All rights reserved.
MARKET RESEARCH MEETS MARKET REALIT Y
End of 24 Months
% Aware
% Prescribing
Conversion Rate
Pain management
72%
33%
46%
Cardiovascular
70%
23%
33%
Gynaecology
67%
28%
42%
Metabolics
72%
17%
24%
Neuroscience
77%
24%
31%
Respiratory
75%
18%
24%
Urology
70%
22%
31%
3. Product availability: a straightforward barrier
– or is it?
Product availability, also known as distribution, refers to
the percentage of hospitals, pharmacies and physician
offices that will stock the drug. In contrast to physician
awareness and adoption, product availability is a far more
straightforward barrier, and one that pharmaceutical
manufacturers can more easily estimate.
In geographic markets where drugs are dispensed primarily
through retail pharmacies – US and Europe, for example
– distribution tends to be less of a barrier. This is because
most pharmacies are willing to stock multiple brands in a
drug class and can order additional stock within a day or
two. However, product availability plays a larger role in
countries where pharmaceuticals are primarily dispensed
through hospitals, such as China, or at the physician’s
office, as seen in Singapore. When drugs are dispensed
through institutions or private practices, it is not uncommon
to find the brand selection limited to two, sometimes three,
options per drug class. This is in part due to storage space
constraints but also, in some markets, a desire to limit the
amount of cash tied up in inventory.
How do we accurately capture the effect that limited
product availability will have on physician usage? Simply,
the percentage of expected distribution should be weighted
to account for the percentage of patients who are treated
at the institutions or physician offices that will carry the
product. For example, if a new Hepatitis B drug will be
available in 60% of the Level III hospitals in China, and
those same hospitals are known to treat 80% of all HBV
patients in China, the weighted distribution rate is 80%.
Subsequently, preference share should be discounted by
20% to account for the patients who are treated at
institutions where the new drug will not be available.
8
Copyright ©2012 Ipsos. All rights reserved.
MARKET RESEARCH MEETS MARKET REALIT Y
4. Product affordability: growing in importance
This holistic approach would be further complemented by
the inclusion of a patient component; sufferers of the
condition treated by product X can provide feedback around
their willingness to accept certain prices that are tied to the
tier statuses already being modelled. The integration of
these three factors – physician feedback, payer probabilities
and patient price elasticity – provides the most complete
picture of how product affordability will impact preference
share and a subsequent US-based forecast.
Finally, product affordability is emerging as a more
meaningful market barrier in both developed and emerging
markets. Below are just a few of the pricing and
reimbursement policies that have recently been put in place
around the world; they point to the growing concern about
the affordability of pharmaceuticals:
• Healthcare reform legislation passed and currently
being implemented in the US
Outside of the US, however, barriers created by product
affordability may be more straightforward; many
governments dictate which products will be reimbursed
under the national healthcare system and which ones will
not. Moreover, if a brand is included on a National
Reimbursement Drug List, there is typically little to no
expense to the patient and subsequently, product
affordability is 100%. By contrast, if the government has
not agreed to reimburse it, the patient bears the full
expense and product affordability may become the most
influential barrier of all.
• A freeze on prices for three years in Germany (from
August 2010)
• 2.5% price reduction on generics and introduction of
generic tenders by AIFA in Italy
• Introduction of mandatory 7.5% price cuts on
innovative drugs provided through the public health
system in Spain
In many cases, new and expensive specialty drugs are
either denied reimbursement coverage by national
healthcare systems and private insurance plans, or linked
to high out-of-pocket co-pays by the patient. In such cases,
many physicians will opt for a more affordable brand that
will cause less financial burden to the patient.
Let’s take China as an example: as of 2009, China’s National
Reimbursement Drug List (NRDL) included 2,127 medications,
of which 1140 are Western Medications.14 For these 1140
drugs, product affordability is not a barrier; the Chinese
government has negotiated pricing with each respective
manufacturer to enable the cost of the medication to be
covered by the national and/or provincial governments.
However, as multi-national pharmaceutical companies
continue to expand distribution of their branded products
into China, many will not be granted placement on the
NRDL. The desired price will exceed what the government
can pay, or acceptable generic alternatives will be available.
Then, in other cases, the product may eventually be added
to the NRDL but will have to wait until the Ministry of Health
reconvenes to review all new drug applications and amend
the list. It is therefore imperative to understand the likelihood
and timing of a new product’s inclusion on the NRDL – as
well as the proportion of patients willing to fill the
prescription if they have to pay cash for it. Without this
knowledge, preference share becomes just that: a preference
that may or may not be reflected in the marketplace.
In the US, the impact that a new drug’s formulary position
will have on prescribing is most easily approximated through
primary market research. The questionnaire is structured
such that physicians are initially shown the base-case TPP,
and instructed to assume the product will have full market
access. After exposure, physicians are asked to record their
preference share for this scenario. Subsequently, they are
asked to assume that the TPP remains the same but with the
formulary status changed to one of four scenarios: Tier 2 (no
restrictions); Tier 2 (with restrictions); Tier 3 (no restrictions);
Tier 3 (with restrictions). The four reimbursement scenarios
are rotated and the physician is asked to record his or her
preference share after viewing each scenario. In some cases,
the formulary access scenarios may be segmented with more
granularity to reflect various restrictions such as fail-first
policies or prior authorisation policies.
Using this exercise, it is possible to gauge how much the
product’s formulary position will impact physician
prescribing. This information is most useful when responses
to the four scenarios are weighted to match the formulary
access that the pharmaceutical manufacturer expects to
achieve. However, manufacturers’ expectations are, at best,
a biased estimation of a product’s likelihood to land on any
one tier. A less biased approach is actually to survey medical
officers who serve as the decision-makers around pharmacy
benefits. By gauging their estimation of a product’s
likelihood to fall on a certain tier, we can define a probability
curve to use in Monte Carlo simulations of the forecast.
Clearly, when a manufacturer suspects there is a chance
that a new drug will not be granted reimbursement they
need to be forearmed with market research that will help
them estimate the impact on sales. The best route is to
approximate the extent to which this will be a barrier to
physician preference share by conducting primary market
research among patients. We can then understand patients’
willingness to fill the prescription at various price points, and
the resulting price elasticity curve will serve as the means of
discounting preference share when the price is set.
9
Copyright ©2012 Ipsos. All rights reserved.
MARKET RESEARCH MEETS MARKET REALIT Y
Conclusions
References
A work in progress
1. Green, P.E. and Rao, V.R., 1971, “Conjoint measurement for
quantifying judgmental data” Journal of Marketing Research,
8:355-363
It seems clear that the barrier approach to discounting
preference share has significant advantages over traditional
preference share conversion methodologies; these can be
summarised as follows:
2. Sobel, K., &Bodsky, J., 2008, “Translating preference share
into market share: rules of thumb that really work”, In:
PMRG Annual National Conference 2008, 9-11 March,
Phoenix, US
1. Each of the factors causing a reduction in preference
share can be examined according to the specifics of the
individual product: the characteristics of the drug, the
level of promotional support that will be put against it,
and the competitive market situation it will be entering.
3. Ziment, 2008, “Calibration of preference share”, In: 2008
Pharmaceutical Management Science Association
Conference, Las Vegas, US
4. Data from Ipsos Healthcare Sales Force Effectiveness
Monitor, Jigsaw
2. When a range is used for each barrier, a sensitivity
analysis can be performed to determine the barriers that
are likely to have the greatest impact, either reward or
penalty, on brand share depending on the manufacturers
marketing plans.
5.Ibid
6.Ibid
7.Ibid
8.Ibid
9. Data from Ipsos Healthcare Global Oncology Monitor (US
& EU)
3. By outlining the specific assumptions around each barrier
that the brand will encounter in-market, there is
transparency around the rationale for discounting
preference share. (This may facilitate better
understanding and alignment among the broader
business team regarding why preference share is being
discounted and by how much.)
10.Ibid
11.Ibid
12.Ibid
13.Ibid
14.IMS Market Research Consulting, The 2009 Revision of the
National Reimbursement Drug List (NRDL), (Online),
Shanghai, Available at http://www.imshealth.com/
imshealth/Global/Content/Document/2009_Revision_
NRDL.pdf) (Accessed 5 August 2011)
4. All of the barriers described in this approach can be
tracked after the product has been launched – making
it easy for the market researcher to compare actual
results against the projections, and adjust market share
expectations if necessary.
Of course, while this method offers advantages over
standard conversion approaches, it is not perfect. In many
cases there will be an inter-relationship between the
various discounting barriers. For example, those physicians
not being detailed and who are unaware of the brand may
be employed at institutions where the product does not
have distribution. In a perfect world, we would be able to
estimate exactly how much overlap exists to ensure that
we discount without double counting. However, given the
data currently available to researchers, it is unlikely that a
precise prediction of all inter-relationships among all the
barriers can be known. Accordingly, the researcher may
need to rely on ranges for each of the barriers.
To conclude, incorporating the impact of market barriers
is a significant step forward in discounting preference
share. However, precise prediction will only be achieved
through continued research into, and evolution of, our
discounting approaches. Until then, the barrier approach,
with its grounding in both market research and market
reality, is an important addition to the pharmaceutical
forecaster’s arsenal.
10
Copyright ©2012 Ipsos. All rights reserved.
Contact
To learn more about how Ipsos can help you with your
Healthcare research, please visit www.ipsos.com
Or contact:
Kim Morneau
+1 617 526 0050
[email protected]
About Ipsos Healthcare
Ipsos Healthcare is a global business division focusing
on research in the pharmaceutical, bio-tech, and
medical device markets. It is also the leading provider
of global syndicated therapy monitor data. Operating
in over 40 countries, the team of 600 pharmaceutical
market research experts, marketers and client-side
brand-builders focus on delivering outcome-oriented
research for its clients. Drawing from a broad range of
qualitative and quantitative techniques, Ipsos
Healthcare offers custom and syndicated research
programmes to evaluate motivations, experiences,
interactions and influence of stakeholders forming the
multi-customer markets which increasingly drive
business success in the healthcare industry.
Copyright ©2012 Ipsos. All rights reserved.
1 2 - 0 4 - 0 9