Regulation effects on the adoption of new medicines
Pharmaceutical
products typically face a number of regulatory hurdles; evidence on the
quality, safety and efficacy of new molecules is estimated to take around ten years
of pre-clinical and clinical research time. Following review of the new product
dossier by a regulatory authority such as the Food and Drug Authority (FDA) in
the USA or the European Medicines Agency (EMA) in the EC, marketing
authorization is established, which defines the relevant patient population and
therapeutic use.
Lags in the
adoption of innovative pharmaceutical products are then the result of different
influences in different countries, but of great importance are local price and
reimbursement regulations. Several studies in the literature have addressed
delays attributable to drug review processes generally, while more recent
studies have emphasized price controls and variations in reimbursement schemes.
The first way,
as used by, uses treatment dummies to identify whether price controls exist at
the time of launch. Similarly, Heuer et al. use dichotomous variable approach
to identify both direct price regulations (international price comparisons, therapeutic
value/cost-effectiveness and pharmaceutical contribution to the economy) and
indirect price regulations (profit control and reference pricing). The latter
is used in a discrete choice analysis to test how different pricing and
reimbursement schemes affect the probability of launch for NCEs approved by the
centralized EMA procedure within the former EU15 during 1995–2004. estimates a
discrete-time survival model using data in 28 countries over 1980–2000 using a
ranking of price bands and regulation dummies to indicate whether prescription
budgets, reference pricing, price freezes and controls affect launch times.
Studies using this first definition identify a significant effect of price
controls on the probability of launch. Countries with the highest probability
of launch impose the lowest regulation on prices and indirect price controls do
not affect launch delays significantly for on-patent drugs further observes
that launch in a price-controlled country significantly reduces the likelihood
of introducing products in additional markets.
Treatment
dummies and price ranking control for regulation only approximately and are
potentially inaccurate given the dynamic and multidimensional nature of
regulation.
Price ranking,
for example, may be highly heterogeneous with respect to therapeutic subgroups
or across time. In addition, treatment dummies frequently exhibit multicollinearity
with country effects. Partly in reaction to these criticisms, there is a more
recent, preliminary body of literature, which has incorporated product-specific
data on actual prices to identify the impact of regulation empirically.
These studies
differ broadly in how they define product prices and tend to emphasize the
formulation of firm’s expectations over how price and reimbursement controls
will affect price on entry. proxy expected price by the lagged average price
per standard unit (SU) for the therapeutic
class (ATC3) in quarters 3 and 4 prior to the first global launch, in an
attempt to capture a firm’s expectation over the impact of price controls.
While use the average competitor prices in the therapeutic class (at the finer
ATC4 level) prior to local launch as a measure of this expectation. In terms of
specification, use the continuous time Cox proportional hazard (PH) model,
whereas the latter study uses discrete-time implementation of the PH model by
complementary log–log regression. Findings from the second category of
studies, which use explicit definitions of expected product price, suggest that
the hazard of launch is positively related to expected price, once again implying
that price controls have a negative impact on launch timings.
In addition to
regulatory market barriers, late entry may reflect strategic firm behavior to
avoid the effects of price spillovers due to reference pricing and parallel
trade.
proxy such
spillover effects through overall market size, identifying a significant market
size effect, whereas conclude that total volume of drugs in a therapeutic
subgroup is not a significant factor affecting launch time.
Intercontinental
Medical Statistics (IMS) data are used with quarterly sales data, in US$, over
the period 1999 (Q1)–2008 (Q3). The data used relate to standard unit (SU)
sales of new molecules in 13 different ATC1 therapeutic categories during 1999 Q1–2008
Q3. The dataset comprises 20 countries, which
represent the major pharmaceutical markets in the OECD (plus South Africa). Each product is identified by the molecule name, IMS
generic classification, global and local launch dates, therapeutic class (ATC4) and breakdown of sales by the distribution channel
(retail versus hospital).
Spain, Turkey,
Belgium, Greece, Portugal, Spain and South Africa have only retail channel
data; in Sweden, retail and hospital sales are combined.
The global
launch date of a given molecule defines the onset of risk of subsequent launches
in other markets. The launch dates are recorded monthly. The unit of analysis is
molecule–country pairs. The time to launch for each molecule j–country k
pair is defined as the difference between the global launch date of
molecule j and the local launch date of molecule j in country k.
The dataset is expanded to define monthly time intervals following the global
launch date until the local failure (launch or censoring) to account for the
interval-censored nature of the launch timing data. Our empirical strategy
takes advantage of the variation in launch dates, which is attributable to the various
expectations held by producers over the price and sales volume attainable in individual
markets, the regulatory environment and the degree of market competition.
The molecule
set is restricted to molecules that have launched in at least ten markets, which
is a more stringent measure of global importance compared to prior studies.
Prior studies
at best consider either molecules that have launched in the USA or UK, and our
analysis is more complete in this respect. Our sample contains molecules that launched
after 1999, and the total number of molecules in this set is 22,397, with a median
time to launch of 14 months.
The analysis
uses ex-manufacturer price levels ignoring any marketing discounts and mark-ups
across wholesalers and retailers, and we focus on the regulated price.
The price for
all molecules is calculated by dividing the ex-manufacturer total revenue by
volume in SU sales. For all remaining markets, IMS data include retail
prescription, pharmacy and hospital data. Obviously, given the range of
discounts and co-payments that apply across these different sectors, our
calculated price will only ever proxy the true selling prices, but the
ex-manufacturing price is the price at which national prices are negotiated.
Moreover, estimated country fixed effects should account for some of the
variation in country-specific discounts. Quarterly average price is assumed for
each month in a given quarter, and the price calculated essentially estimates a
volume weighted average price for each molecule across all products with the
same active ingredient.
Further
(confounding) variables are defined using the IMS data on sales, including an
Herfindahl-Hirschman Index of global market competition, firm size as proxied by
sales volume and product quality as established by molecule characteristics.
OECD
statistical extracts were obtained for additional data on GDP per capita. Sales data were deflated using GDP deflators from
the International Monetary Fund World Economic Outlook Database 2008.
Our main result
for the impact of price regulation on the timing of the launch of a new
molecule is significant and strongly robust across these specifications. In all
regression specifications the estimates for our measure of the impact of price
regulation, the expected price, after controlling for volume are highly
significant (p=0.001).
A unit increase
in the log expected launch price and the log of expected market size increases
the probability of launch by 0.003 and 0.002, respectively. This is close in
value to 0.0053, the marginal effect of expected price for superior molecules.
Standard error estimates of expected price are slightly lower because we
cluster by molecule–country rather than by molecule alone.
The latter is
expected given the presence of price (regulation) benchmarking across countries
with are specific for each molecule.
With respect to
other effects, for competition, a unit increase in the log of IHH reduces
the hazard rate by 0.005 in the quadratic specification and by 0.004 in the semi-parametric
one, which implies the more competitive the subgroup, the higher is the
likelihood of quick launch. In other words, and in common with many other industries,
the higher the concentration, the lower is the likelihood of rapid launch. Firm
heterogeneity, proxied by the number of countries a
firm has launched in, is found to be highly significant; a unit increase in the
log number of countries a firm has launched in (equivalent to multiplying
geographical reach by 2.72) reduces the probability of delay by 0.01. With
respect to molecule characteristics, a unit increase in the log molecule sales
globally increases the hazard of launch by 0.004. The extent of global reach,
as expected, was found to have a significantly positive effect on the
probability of launch with a marginal effect of 0.059. The only caveat of our
approach is that it might introduce a bias for older molecules as they have had
more observed time to launch in more markets, although we do control for time
since first global launch. The effect of country income, as given by log GDP
per capita ($), is positive but not significant, and is therefore excluded from
some specifications.
Finally, time may
affect regression estimates in several ways. First, macroeconomic trends in the
sector may have an impact on price levels, so we account for this by including
dummies for each calendar year in all regressions. Second, time captures information
about the relative innovativeness of new molecules. When a new molecule is
about to launch, it represents incremental (or breakthrough) innovation
compared to the molecules in its therapeutic
subclass. The longer the time lapse from global launch, the higher is the
probability that new competitors will enter to compete against the molecule
lowering its comparative therapeutic advantage. The impact of time elapsed since
first global launch is therefore captured by interacting expected price, as
well as volume with the time since global launch. A dummy variable (first
launch before 1999) is also included to test whether the hazard of launch is
statistically different for molecules that launched globally after 1999
compared to the ones that launched first globally during (1993 and 1999).
Remember that the set of molecules was restricted to the ones that first
launched after the establishment of the EU in 1993 and that all the failures
(i.e. local launches) are post-1999. Therefore, molecules with first global launch
pre-1999 are left-truncated. Left-truncation is dealt with by omitting the
subject from all binary outcome analyses during the truncation period since the
subject could not have failed during that period.
Time
interactions of price and volume are significantly negative, which suggests that
the impact of price and volume decays over time following the global launch of the
molecule. Molecules that launched first before 1999 have a significantly lower hazard
rate compared to molecules that launched after 1999; the marginal effect is in
the range of −0.018 to −0.014 depending on the precise model
specification.
Parameter
estimates of t and t2 suggest concave duration
dependence, while the hazard of launch initially increases and then decreases,
which is in contrast to prior findings of Danzon and
Epstein (2008) who observed that hazards
first decrease then increase with time since global launch. This might be
because the molecules in this analysis are more recent, and hence potentially
more innovative and have a higher extent of global reach overall (all molecules
have launched in at least 10 markets).
Given that we
use proxies for a number of our confounding variables, we carry out a number of
robustness checks. With respect to competition effects, we carry out robustness
checks by controlling for the number of substitute molecules and investigate
whether generic competition is significant. We consider only quadratic duration
specification for robustness checks as base case estimates suggest the fit of
quadratic and semi-parametric specifications are comparable. Intermolecular
competition is found to be more influential on the decision of entry, as
compared to the extent of generic competition proxied by the number of substitute molecules with generic
competition. This is consistent with findings of Kyle (2007) that the number of competitor molecules in the
same ATC4 significantly increases the hazard of launch, while the number of
molecules with generic competition has no significant effect on the launch
decision of new molecules.
Robustness
checks were carried out by controlling for log firm sales in 2007; total and
local numbers of firm molecules firms have launched to control for economies of
scope. All scale and scope variables are robustly positive and significant.
Portfolio diversity (number of prior molecules launched) is associated with
quicker launch, which is in contrast to findings of Kyle
(2007). We find no evidence of advantage
through domestic launch.
In the
robustness checks on molecule characteristics, we further proxied therapeutic importance
using the total number of markets in which a molecule has launched, i.e. global
extent of launch.
Finally, we
also aimed to explicitly test for the potential impact of price interdependency
across country markets. We therefore restricted the country set to EU countries.
We find strong evidence that external reference pricing slows adoption of
innovation. Launch in a high-priced EU market increases the conditional
probability of launch by 0.042 compared to launch in a lower-priced EU market.
This effect increases to 0.051 for molecules that first launched after 1999,
suggesting an increase in the strategic importance of price in the timing of
entry.
From a
strategic perspective, firms may risk the loss of competitive innovative edge as
delays increase the chance of facing further competition later in time.
This suggests a
second firm strategy, which involves pursuing convergence of prices in the EU
market following launch to avoid knock-on effects due to parallel trade and external
referencing, even if at the expense of foregoing some short-term local profits in
some markets. We test for this strategy, by controlling for the extent of
deviation between expected local price and the average EU price for the
launching molecule. The absolute difference between the local expected price
and average EU price significantly decreases the hazard of launch; the sign of
this difference remains insignificant. Launch and pricing strategies are
multimarket optimization decisions; the trend to drive prices closer across
different geographies may potentially reduce global prices.
Thus, to summarize,
regardless of the precise time duration specification, and controlling for a
large number of confounding effects, price regulatory controls on reimbursement
have a strong effect on time to launch. Across a range of specifications and definitions,
we also find weak competition increases time-to-entry, while larger market size,
higher therapeutic importance and the greater the number of markets a firm operates
in reduces time-to-entry. We further find that within the confines of the EC market where, although individual
countries have their own price and reimbursement authorities, there is
considerable cross-referencing of pharmaceutical prices and parallel importing
of pharmaceutical products, price regulatory spillover effects appear to have
an impact on launch times.
(Abstract from Joan Costa-Font · Alistair
McGuire · Nebibe Varol article)
تعليقات
إرسال تعليق