Forecasting for orphan drugs: The data challenge
Gianclaudio Floria, senior manager for business planning and analysis at Amgen, on best practice in forecasting for orphan drug development.
By on Jul 8, 2010Gianclaudio Floria, senior manager for business planning and analysis at Amgen, on best practice in forecasting for orphan drug development.
Orphan drugs have been a resounding success from a revenue perspective, with 2006 sales reaching $32 billion and growth in the 8% per year range.
Yet orphan drug development is fraught with uncertainty, particularly with respect to data.
For example, how many individuals suffer from a disease?
Are enough patients at a stage in their illness where treatment is possible?
How easily are these patients identified and treated? How well will they comply with therapy?
The US Rare Disease Act of 2002 defines an orphan disease as one affecting fewer than 200,000 patients, about one in 1,500 Americans.
Approximately 5,000 illnesses fall under this category.
The European Commission designation is somewhat broader, comprising rare, life-threatening, or chronically debilitating diseases as well as some tropical diseases.
The US provides tax incentives and exclusivity for orphan drug developers, while the EU offers only market exclusivity.
Large companies generally stay away from orphan drugs due to the small number of patients served for any single disease.
Less pricing flexibility, particularly in the EU, is another barrier.
The few brave companies that venture into orphan markets immediately notice a dearth of reliable data, says Gianclaudio Floria, senior manager for business planning and analysis at Amgen in Zug, Switzerland.
Amgens romiplostim (Nplate) drug for idiopathic thrombocytopenic purpura (ITP), a rare platelet condition, was approved in 2009 in the EU and in 2010 in the US.
ITP is an autoimmune disease resulting in low platelet counts, and therefore bleeding disorders.
Data availability
Developers of medicines for more common conditions have a wealth of epidemiologic information and secondary data at their disposal in the published literature, from national health agencies, and through private data-collection services.
One can obtain similar information through prescription-tracking services that operate through pharmacies, registries, payer formularies, and participating physicians.
Not so with orphan diseases.
Data availability for diseases like ITP is extremely limited and not very accurate, as the methodology may be very different from study to study, Floria says.
Some studies estimate prevalence as X, others as Y, and there may be considerable variance between the two.
Nor is estimation possible through prescription monitoring.
Before approval of romiplostim, the only treatment for ITP were drugs not specifically approved for ITP, like corticosteroids, which are used for dozens of different conditions.
With such variability in relatively small patient numbers, its no wonder forecasting models employing standard market sizing (number of patients, appropriate disease stage, and treatment rate), patient uptake (class uptake percent, class share percent) duration and dose (weeks of treatment, average dose size, compliance rate), and pricing can easily under- or overestimate the true market several-fold.
Getting good numbers
The key to arriving at good numbers is to conduct sound primary research and to temper it with any available published data.
These numbers should be based on prevalence (number of individuals with the disease at a specific time) rather than incidence (number of new cases).
The former number is usually quite stable, and especially so for chronic illnesses like ITP.
With prevalence numbers in hand, researchers must correct for stage of the disease applicable to the drugs label.
For example, in ITP, romiplostim is indicated for chronic patients who have undergone removal of the spleen or those for whom splenectomy is contraindicated.
Other factors operate as well through what Floria describes as a decision-making funnel: undiagnosed patients, those under care but not yet treated, those in remission.
This model begins with the big picture, he says, but for orphan drugs you only have estimates for those who are diagnosed, and no clue about the others.
The romiplostim forecast was stress-tested using other analog orphan drugs to raise the confidence level in uptake and peak sales.
In the European region, the forecast has proved to be very solid.
We closed out 2009 extremely close to our initial estimates, except slightly higher due to the fact that one country launched earlier than we expected, says Floria.
Is this approach valid, and does it scale upwards and downwards?
According to Floria, another company outlined a similar strategy for a drug whose target indication was considerably smaller than that of romiplostim.
Of course, there are some degrees of uncertainty, but the model seems to hold well, even for much narrower orphan categories, says Floria.
As personalization and results-driven prescribing become more the norm, expect to see markets for blockbuster drugs fragmenting into much smaller markets for many more drugs for a given indication.
Drug firms need to pay attention.
For more on personalized prescribing, see Personalized medicine comes one step closer.