1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Modelling the constraints on the global pharmaceutical industry
Geoffrey D Hobbs Bryan R Deane
8a Alma Road SmithKline Beecham Pharmaceuticals
Reigate SB House
Surrey, RH2 ODA, UK Great West Road
: Brentford
Middlesex, TW8 9BD, UK
Abstract
Advances in all fields of medical technology have driven rapidly growing expectations of medical
care over the last half century. The rate of growth of this demand for health-care has consistently
exceeded GDP growth and, in many countries, the health-care bill has been absorbing an ever greater
proportion of government spending. Governments th h the world are
that this growth is The phar ical industry, whose products account t for about 5
to 10% of the health-care spend, has become an early target for a wide variety of cost containment
measures. The industry's growth has been based on a cycle of growing sales from ever more effective
new products, fuelling a substantial re-investment in high-risk, long-term research and development,
leading to further advances and new product introductions. Historically the overall sales growth has
been founded on both price-related and volume-related factors.
The purpose of this study was to explore the relevance of a System Dynamic modelling approach to
the p ially complex i ions between the phar ical industry, the medical
and related p i and the and pay who fund health-care. A prototype model
has been constructed and used to create a variety of scenarios describing alternative futures, in which
the regulators impose constraints on either price or volume increases. Not surprisingly, the
vary, d ding on the type and severity of the constraints and the industry's
responses “to them. For example, under price control, in ‘keting by the
industry to promote volume growth might accelerate the reduction in the industry's profit margins,
leading to a fall in the proportion of sales income devoted to R&D and a fall in the number of new
products being developed. This scenario suggests a transformation of the industry towards a high
volume, low margin, non-innovative "commodity" industry.
Experiments with the prototype suggest that the System Dynamics modelling approach can help to
explore in an insi; way the p ially complex i ions of the various groups involved in
health-care delivery. There is extremely broad scope for further development of the approach, which
should ideally be targeted at specific issues within the system.
Industry, page 45
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Modelling the Constraints on the Global Pharmaceutical Industry
1 Introduction
Over the last fifty years, dramatic advances have been made in all aspects of the practice of medicine,
including surgical applied ry and supportive care as well as the introduction of
new medicines. At the same time, the exp ions of the edi d populati of the world have
risen dramatically. For any patient it is almost assumed that ever more heroic (and expensive)
measures will be taken until either recovery is complete, or medical practice has exhausted its
repertoire. This expectation has ensured that demand for health-care expenditure has grown faster than
GDP in most developed countries for at least the last three decades. That populations are aging is both
testimony to the medical success story, and further fuel for health-care demand.
Governments throughout the world are i ingly d that the ing costs of health: i
claiming an ever greater share of government expenditure, are unsustainable. In the USA for example,
it has been esti that health: for the disposal of about 14% of GDP in 1992; about
half of this was in the public sector, which was growing at about 13% pa (compared with about 3%pa
GDP growth). Where, traditionally, decisions at the point of delivery (usually the doctor-patient
interface) have been d with risk/b already cost is being factored in as
another issue in the decision of whether or how to treat. Debate will undoubtedly continue as to what
extent such decisions should be taken on purely medical grounds and to what extent it should be seen
in a wider socio-economic context. One thing is certain: governments’ ability to fund health-care is
not unlimited. Undoubtedly, budgets for health ding will be lied at some level.
Although the medicines bill accounts for only 5 to 10% of overall health-care expenditure in most
countries, this has become an early target for cost i While the p
industry is already highly regulated on the supply side, many of these regulations, (covering, for
example, the quality of preclinical and clinical testing and manufacturing) tend to drive up industry's
costs and therefore contribute to the inflating prices of newly-developed drugs. For several years there
have been examples of direct controls on drug prices and/or drug company Profitability, and
are now and impl ing a variety of d d. i
@ (including p ibing budgets, eference pricing systems, mandatory price controls or cuts, limited
reimbursement lists, generic and therapeutic substitution and varying patient co-payment contributions).
For the global pharmaceutical industry, two aims are paramount: to contribute to real advances in
medical treatment and to provide shareholders with an increasing (in real terms) flow of dividends.
The pursuit of these goals has resulted in a rapid growth in sales, driven partly by volume growth and
partly by rising prices. The opportunity for achteving sales growth through annual price increases is
now, however, severely limited. In such a climate the only mechanism leading to rising average prices
is the continual introduction of innovative new products at premium prices. While this may be critical
for the sustained growth of individual companies, in a highly fragmented global industry, the impact
of any one product introduction on global sales is minimal, at least in the first few years.
Other important drivers of global growth are volume-related. These can be classified as relating to
Pp growth, to g need for med: amongst certain groups within the population and
to increasing access of the popul. to ci Population growth clearly leads to a potential
increase in the volume of med d. I ing need ip factors such as the
increasing proportion of aging people. needing more medical care per person, and the emergence or
spread of new or difficul diseases, including AIDS. h ing access depends upon economic
Industry, page 46
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
development allowing more people access to better medical care; while this is particularly relevant to
developing countries, it is also an issue in, for example, the US, where bringing the millions of
uninsured into a health-care system should in theory expand the potential market volume.
The growth aspirations of the ph ical industry and the cost containment objectives of
governments are fundamentally in conflict. Each might be expected to adopt measures aimed at
securing success in its own objectives; each will respond to minimise the damage to its own aims.
It would be surprising if the outcome were as either anticipated. At worst, the result might be
seriously damaging - either to the industry or to the broader soci ic aims of gi or
to both.
Therefore the purpose of the study described here was to explore the relevance of a System Dynamics
modelling ap to the p ially complex i between the p
industry, the | medical and related p ions, and the v and pay who fund health-care,
much of it on behalf of g: The model described here has been constructed using "iThink"
software and has been used to create many scenarios describing alternative futures, some of which are
discussed below. At this stage of the project, the model should be considered a prototype designed
to examine whether further devel of this type of app might offer a p
learning tool (2,3) which would provide valuable insights into some of the issues which currently
concer governments and, of necessity, the industry. For this reason, the scenarios described below
should not be interpreted as quantitative forecasts or as the basis for policy recommendations.
2 Mapping the model's structure
2.1 The feedback loop structure
The aim, in developing the model, was to identify the most important mechanisms, described in terms
of feedback loops, driving the growth of the pharmaceutical industry and then to superimpose the
mechanisms that might limit its growth and bring the system into a dynamic equilibrium. The model,
in its most basic form, is constructed from three positive (reinforcing) and two negative (balancing)
feedback loops (Fig 1). The rei ing loops are all i
Figure 1. Model structure
with the pharmaceutical industry's growth aspirati: The balancing loops are iated with the
efforts of the funding organisations 6 constrain growth. The growth of population is included as an
exogenous driving force. Additi y i have been added to increase the models
realism.
Industry, page 47
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
The R&D loop represents the industry's core activity: the expenditure of part of its sales income on
R&D in order to add to the portfolio of products it offers in the world's markets -a portfolio which
generates a volume of drug usage which, in turn, creates further sales value and hence more R&D.
Sales income is also used to finance marketing effort which in part promotes the more intensive use
of existing drugs (eg by treating a wider range of illnesses -volume growth loop 1) and in part by
ing their ibility (eg by ical market ge - volume growth loop
2).
The two control loops represent the funders’ responses to costs rising faster than they wish. They can
attempt to limit growth by price control, by volume control or by a combination of both.
2.2 Industry driven growth
The core RED loop is displayed. in outline only, in Figure 2. It is assumed that the industry allocates
a percentage of its sales income to R&D and that this, when divided by the R&D expenditure incurred
for every successful new product launched, creates a flow of new products which adds to the portfolio
of existing products. In reality the R&D exp d with a 1 product is incurred
over the 10-15 years prior to launch but this process is not captured in the models described here. The
price at which a new product is launched is in general higher than that of existing products and
therefore causes the average price to rise - a driving mechanism contributing to the growth of sales
value. The model does not differentiate between products within the portfolio and assumes that each
generates an average annual volume sales at an average annual price. This i is a gross simplification
which will be addressed in future versions of the model by di: between of
products. Products are eventually di d, at a rate d on their rate of launch a product-
lifetime ago.
———
oot ne Portolo
—> launches a incontinued
[nap |
i oe
_/ a
Sales
R&D/sales —e er foe volume
Launch \
price
Volume per
Sales product pe
value
Figure 2. Core R&D Loop
The two volume-growth loops are shown in more detail in Figures 3 and 4. In both it is assumed that
the industry has a long-term target growth rate which can be for either sales volume or sales value.
Figure 3 shows schematically the first of these volume-growth loops and captures the effect of
marketing effort (in excess of that needed to maintain product usage at current levels) on the average
volume of drugs prescribed per patient per product per year.
Indusin.. page 4
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Portfolio
product —————m of
launches products ‘5
R&D
spend patients
Volume per
‘Alii product pa
growth
pi 4
Volume per
Target patient per
growth product pa
rate $$» Markating
ad
Figure 3. Volume growth loop 1
Marketing effort is measured in money terms and represents the costs of the sales force, promotional
li etc. which p the medical p ion to prescribe the product to all relevant patients
and for all relevant illnesses i in a range of appropriate formulations. The concept of treated patients
has been introduced to capture the fact that only a proportion of the population is ill at any particular
time (measured in terms of disease prevalence) and, of this, a fraction only will have access to the full
portfolio of drugs. The effects of aging can be captured by increasing disease prevalence.
The second volume-growth loop (Fig. 4) describes the other of marketing expendi -
to bring about the expansion of product availability to geographic areas outside the country or region
of original launch. Pharmaceutical companies rarely achieve simultaneous worldwide launches because
of variations in registration timescales or, for many, because the upfront costs of concurrent global
development and launch are prohibitive. It is usually a process of progressive territorial extension.
New Pontete
—_—_——»
eincnee products" Population
spend ar Patients
|
NL Sales et i
value f
iN Access
7 Volume
“, Actual control
g loop 2
Figure 4, Volume growth loop 2
In the regional version of the model these loops are Teplicated (with different parameters) and
contribute volume and value sales to the p ies which prise the global
industry. It is assumed that R&D is carried out on a global rather than regional or national basis.
Industry, page 49
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
2.3 Cost containment
As already described in the introduction, the isations that fund ical hi have
a wide range of methods available to them for controlling costs. These methods vary from country
to country (particularly between the USA, where health-care is predominantly in the private sector, and
many other countries, where public provision is the norm) but fall broadly into two categories. The
first includes a variety of methods for controlling prices. The second comprises a portfolio of policies
for limiting volume. In the multi-regional model these control loops are replicated.
Conflict arises because of the disparity between the annual drugs bill with which the funders are
confronted and the budget that has been made available to them either by governments or, in the USA,
by the health-i i from indivi and ies. The absolute size of
these budgets has been determined by history but _ growth 1 rates are now increasingly seen in the
context of overall economic growth. It is th ion that future health-care
budgets will be tied closely to the rate of growth of GDP. It is also assumed, at least in the first
instance, that the share of health-care exp d for by p i will be held
broadly constant.
In both control loops (Figs. 5 and 6) the divergence between budget and actual expenditure is
lated, with some al for the inevitable delays, into a desired expenditure. In the first loop
this is then converted into a price adjustment mechanism which operates on the average sales price
in the region. Some additional control is also imposed on launch prices.
Pri
fee a
VY
‘expenditure:
Figure 5. Price control loop
Volume per
patient per
New goo Fogalle ~ product pa
product
launches products Volume per
tontee Volume
constraint
RaD Soles
spend volume Desired
XO ~, a
Sales
Budget. ws
value phe
Druge
bill Petia
Figure 6. Volume control loop
Industry, page 50
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
In the second loop the desired expenditure is d into an adj ism which controls
the volume per patient per product pa (the average annual dose) on the assumption that the funders
cannot directly influence the number of patients or the number of products in the portfolio. A more
sophisticated representation might allow for a reduction in the number of products for which
reimbursement is available, or for an increase in the percentage of the cost borne by the patients
themselves and the deterrent effect this would have on people seeking treatment in the first place (ie
indirectly limiting access).
It is not only funders who will experience financial pain. If the controls prove to be severe, either by
intention or by accident, the pharmaceutical industry will also experience pain in the form of reduced
profits. The model represents these effects, ing profit by ‘ing R&D and i
expenditures, production costs as well as un-attributable administrative costs from sales income. What
happens to profits will influence the industry’ 's policies on R&D expenditure, marketing and pricing
q ly of new prod In lar a cap is imposed on the proportion of sales revenue that
can be spent on marketing and there is assumed to be a lower acceptable limit for profit margin at
which R&D expenditure would be reduced to protect the remaining profits.
3 Validation of core assumptions
A set of core assumptions was chosen to ensure that the model broadly reproduced historical data over
the decade 1980-1990. Data is sparse at this level of aggregation but estimates exist for total real sales
value, ratio of R&D expenditure to sales, the number of products on the market, the number of new
chemical entities and major line-extensions launched annually, the average R&D cost to the industry
(and its rate of growth) for every successful product launched and the time taken to develop a product
(4). From these it has been possible to estimate the values of variables for which past data are not
directly available and to establish an internally i set of i ptions which
constitute the historic foundation for the model.
A regional version of the model divides the global market into two - that in the Industrial World (N
America, W Europe, Japan and Australasia) and that in the Developing World (the rest).
Since the models are seen as prototypes to explore a principle rather than to produce a quantitatively
accurate representation of reality, only broad agreement has been sought between the model results
and the historical data.
The key historical features replicated by the models are:
(a) A global market value of about $180bn in 1990, which has doubled in real terms (ie
after allowing for general inflation) since 1980.
(b) An increase in real pnces of 3-4% pa during the 1980s.
(c) An average of 11% of sales spent on R&D.
(d) A mean R&D expenditure per successful product that has increased over the preceding
decade at a rate of about 7% pa in real terms to $360m in 1990.
(e) The launch of approximately 55 new products (new chemical entities plus major line
, extensions) in 1990, a figure that has declined from 60-65 in 1980.
(f) A total portfolio of some 3600 products currently prescribed worldwide, each with
average sales of $50m pa.
(g) Sales in the developing world rising from 12% of the world market in 1980 to 16%
in 1990.
(h) Average sales prices in the developing world substantially lower (20% assumed) than
those in the industrial world due, at least in part to lower usage of more expensive
drugs.
Industry, page 51
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
4 Scenarios
A wide range of scenarios describing likely states of the future market are being explored with the
models. These range from a naive extrapolation of current trends to scenarios in which complex
interactions are taking place between the industry and the controlling authorities.
41 Unrestrained growth
The "base case" against which all other scenarios can be compared assumes that the future will be
similar to the 1980s, ie that the balance of cost constraint and growth aspirations will remain
unchanged and that the growth of the phar ical market will i d. Sales in the
industrial world continue to grow in real terms by about 6.5% pa and in the developing world by 9-
10% pa. By 2020, the developing world's share of the market will be 30-35% and the global market
will be growing at about 7.5% pa. This is likely to be 3-4% pa faster than world GDP, resulting in
a pharmaceutical expenditure, in 2020, of some 2.5-3.5% of GDP compared to 1% today. The number
of new products launched annually continues to decrease. If the pharmaceutical industry holds its unit
costs of manufacture constant in real terms (as is assumed rather simplistically), by 2020 it would be
earning a substantially higher profit margin than it does now. If other health-care costs also continue
to rise at similar rates, this leads to a picture of the future with which many funding authorities would
be uncomfortable.
The model can be used to examine the sources of the global sales growth. Of the 6.8% pa growth in
the mid-1990s, 1.3% pa comes from population growth, 1.5% pa from increasing access (mainly in
the developing world), 1.0% pa from increasing volume per capita (of which 0.4% pa is attributable
to the growth of the number of available products) and 3.0% pa comes from increasing real prices.
4.2 Imposition of Price Control
Two versions of this scenario have been created: the first, in which the pharmaceutical industry
accepts the control without response; the second, in which the industry responds in an attempt to
maintain its ility by raising penditure and hence p ing the growth of sales
volume.
In both scenarios, sales growth is constrained such that each region's expenditure on drugs is pulled
into line with a drugs budget that is, in its turn, constrained to grow at no more than a hypothetical
GDP growth rate - assumed to be 3% pa in the industrial world (where controls are phased in over
a 2-3 year period from 1990) and 7% pa in the developing world (control phased in from 2000). This
is achieved by progressively reducing the average sales price of existing drugs. A limited degree of
control is also imposed on the price-premium at which new products are launched.
In the non-reactive scenario the industry accepts the consequences of price control. After an initial
transient during which the authorities impose price reductions to bring the situation under control and
then relax in the light of the consequences, price rises in the industrial world are pulled back by about
2:5 percentage points, slightly more than offsetting the rate of price rise that has been imposed by the
industry historically. The only price rise being permitted is that caused by adding new higher (but
nevertheless constrained) price products to the portfolio. Similar consequences follow in the
developing world, but the amount of long-term price control needed turns out to be smaller (about -1%
pa) because the industrial world’s restraint on launch prices carries over into the developing world and
curbs sales growth there anyway. The effects on global average prices are shown, in comparison with
the unrestrained scenario in Figure 7.
Industry, page 52
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Brice control No ladastry response
Figure 7. Effect of price control; no industry counter response
Figure 8. Industry sales and profit under price control; no
industry counter response
The consequences for the industry's finances are serious but not devastating (Fig. 8). Sales and profits
continue to grow, but at sharply reduced rates (for profits from 10% pa to a long-term 6% pa). The
first round of price restraint, in the industrial world, stabilises profit margins at a (hypothetical) plateau
at about 23-24% in the 1990s. The second round of price control holds the profit margin at 25% for
about 5 years before it rises slowly towards 30% by 2020. The R&D:sales ratio is maintained
throughout and the R&D spend therefore suffers the same diminution of growth as do sales. As a
result the number of new products launched each year falls from 55 in 1990 to under 20 in 2020.
In the second scenario, the industry responds in both regions by raising the proportion of sales income
devoted to marketing and drives up its volume growth (Fig. 9). The response in thedeveloping world
has been chosen to be particularly marked (marketing: is capped at 25% in the Developed World
and to 30% in the Developing). The authorities respond in turn by tightening the screw further on
prices. They still achieve their budgetary targets but, this time by reducing long-term real price growth
to -3% pa (Fig. 10). The effect on the industry's finances is much more dramatic. Profit margins
collapse and profits fall in real terms from 2002 (Fig. 11). Although the industry is not forced into
loss, from about 2010, because the profit margin has dropped below the critical level of 10%, the
industry can no longer fund its R&D efforts; it reduces its R&D:sales ratio and the rate of launch of
new products drops rapidly towards zero (Fig. 12).
Industry, page 53
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Fate of volume increase (pa)
Figure 9. Effect of price control; Industry promotes volume Figure 10. Effect of price control; Industry promotes volume
growth growth
Prot (son) o New products launched pa
ry
ca 0
«
aol
2
°
‘1900 1985-1900 +1905 ~—~—~2000 ~+~—«-2008.-—=—aa10~—«oTs 2020 1980 1985 —=«1990~=~«NGS ~~~ ~=«COSSSCOSC NSO
Your Yer
Pree contro NolIndustry response ne Price contro NoIndastry responte
= Price contol Indesry promotes volume == Price control Industry promotes volume
Figure 11. Industry profit under price control; industry promotes Figure 12. New product launches under price control; industi
volume growth promotes volume growth
In both these scenarios the industry undergoes a dramatic long-term transformation. It becomes a bulk
supplier of drugs which undertakes little R&D and creates few new products, ie a high-volume/low-
price non-innovative industry. What is more, in the second, the industry compounds its problems
through the adoption of inappropriate policies. It seems that no response would have been better than
the wrong response.
4.3 Imposition of volume control
The same objectives of budget growth are ascribed to the controlling authorities as in the previous
scenarios. However, in this case, the budgetary pressure is relieved by holding down volume growth
and leaving prices to follow an unconstrained path. In the first scenario, the industry passively accepts
the restraint. The impact on profit is similar to, but less severe than that under price control; (real)
profit growth is suppressed during the five years or so that it takes for the new policies to get a grip,
but then recovers, albeit to a rate some 3% pa below the pre-intervention level. The industry's ability
to finance its R&D is not impaired but the lower sales growth causes a similar decline in the rate of
Jaunch of new products. The main effect of the controls is to reduce significantly the volume of drugs
prescribed per person (by as much as 30% in the industrial world). This scenario, therefore, has social
Industry, page 54
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
implications in that in the long term the market is one characterised by high-price and low-volume and
is therefore one in which access to treatment is being rationed and the cost of treatment is high for
those who receive it.
The industry has two ways in which it can try to offset the effects of the controls. It can increase
marketing expenditure in order to promote volume usage of its products - thereby entering into direct
conflict with the authorities - or it can seek to raise prices. In the first case the (model) authorities
just bear down harder still on volume, the industry achieves no higher growth but incurs the additional
costs of the marketing effort, thereby further depressing its profitability. If it puts up prices, the
industry does protect its profit margin to some degree. However, the authorities squeeze volume to
such an extent that there must inevitably be a deterioration in the quality of health-care being provided
- with the associated political risk to both the authorities and the industry.
4.4 Imposition of both price and volume control
have been in which the ities impose a mix of price and volume controls and
the industry responds with its own mix of pricing and marketing policies. Different outcomes for the
industry and for society at large can be obtained depending on the balance between the authorities’
price and volume policies and the industry's choice of response. In broad terms, these compromise
futures lie between the extremes described in sections 4.2 and 4.3 above. It is in the creation of such
scenarios that real learning starts to take place and the model begins to reward the investment that has
been made.
5 Conclusions
The prototype models have demonstrated that it is possible to create a simplified but realistic
of the global phar ical industry and the interactions between its growth aspirations
and the authorities’ cost containment objectives. Although not too much weight should be put on the
preliminary quantitative results described above, the exercise has shown the likely consequences of
the world continuing along the path taken in the past decade and can explore the effects on the
industry of different types and strengths of cost containment policies. It can demonstrate not only the
direct consequences of these constraints on the industry, but also the industry's possible reactions and.
thus the ultimate effectiveness of both the authorities' and the industry's strategies. The complex
feedback nature of the total system will often lead to ‘pected or
The models have not yet been exhaustively explored. They contain control and response mechanisms
yet to be ined, eg i d patient co-p: and their q in terms of icti
access. There are many more that could and should be incorporated eg the pressure to use generic
rather than branded drugs.
An important area lying outside the current boundaries of the model is the interaction between drug
consumption nd other forms of medical intervention. The industry has always argued that ill-
ions in expendi on drugs will ly result in
in the health-care system.
Currently the models are non-specific in terms of the issues that they address. In this form, they could
be developed as a "microworld" - perhaps as a learning tool for senior management or as an aid to
constructive dialogue between the industry and the authorities. They could also be developed to
provide the and ing of a scenario planning process. More
productively, the ap could be developed in a more focused, perhaps country-specific way, in
the context of a clearly defined issue or problem.
Industry, page 55
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
6 References
1 Dower, M. 1993. A year to retrench, restructure and rethink. Scrip Magazine January 19+
2-5.
2. de Geus, A.P. 1992. Modelling to predict or learn? European Journal of Operational
Research 59: 1-5.
3. Morecroft, J.D.W. 1992. Executive knowledge, models and learning. European Journal of
Operational Research 59: 9-27.
4. Derived from a variety of i and p ical industry sources.
Industry, page 56
