Peter Mansell explores how pharmaceutical forecasters are developing new methods and new metrics to enable them to make better guesses
Forecasting for pharmaceuticals is never going to be the most exact of sciences.
“Informed, intelligent guesswork” is one practitioner's definition of long-term forecasts.
So, uncertainty is the forecaster's natural habitat: addressing markets in flux, prone to political expediency and with different regulatory requirements, pricing and reimbursement systems, distribution structures, demographics, epidemiologies, income levels, and competitor activity.
“If it's to make an investment decision, you know you're going to be precisely wrong, so you need to be roughly right,” comments Nick Guthrie, head of forecasting at AstraZeneca in the UK.
Nor are the conditions for being “roughly right” getting any easier.
In what has traditionally been a rather sluggish and conservative industry, the last few years have seen a sharp acceleration of radical change.
This extends from the way industry discovers and develops its products to its sales and marketing strategies, global reach, reliance on outsourcing, and relationships with patients and payers as well as the pressures on healthcare systems that are ushering in new conceptions of drug value.
Most health authorities realize that, with ageing populations and limited resources, “treating the disease is no longer really going to be an option in the long term,” points out Nic Talbot-Watt, managing director of Black Swan Analysis.
There is a corresponding shift to preventive healthcare and to medicines premised on longer-term outcomes and health economic arguments, such as heading off expensive hospital stays.
What this means for forecasters is that they “start to see the wider disease maps,” Talbot-Watt adds.
And that “starts things moving from the static, prevalence-based disease models that we've been used to for the last 12 years and into something slightly more dynamic, where you've got events that can really disrupt the patient population and were never really considered before.”
Among the key disruptive elements are the growing importance of emerging markets in pharmaceutical business strategies; the increasing prevalence of health technology assessment and its impact on pricing, reimbursement and market access; and the gradual emergence of technologies that are nudging the market, at least in developed countries, towards stratified medicine and niche segmentation.
Aggravating the inherent uncertainties of emerging markets is a basic lack of data, particularly on new drugs.
As a result, forecasting is heavily reliant on primary research.
Further complications include a larger proportion of patients funding their medicines out of pocket (which muddies pricing strategies) and, in some cases, the need for special distribution channels.
While a market such as China is potentially huge on the basis of population alone, there may be a marked discrepancy between the subset of that population actually buying a Western brand—more than likely in the private market—and the total opportunity. (For more on China, see “Cracking the Chinese pharma market”.)
As Guthrie observes, the level of exposure to emerging markets will also depend on what stage of its lifecycle a drug has reached.
“If you're sitting in a brand that's on the market and trying to grow … then things aren't that much different,” he says.
“The US is still a big market and it's important to do well there. Europe is still quite important, as is Japan.”
If, however, “you're sitting in a mature brand, something where growth in established markets has been achieved, then you're going to be very much looking at new markets—the Brazils, the Chinas, the Mexicos and the Russias. And life’s very different there.” (For more from Nick Guthrie on emerging markets, see "Strategic approaches to emerging markets".)
One difference is working out how the brand fits into the local pricing and reimbursement infrastructure.
“If you go to Brazil or Mexico, you've got to think, 'Am I accessing the whole market or just part of it? Am I playing the public-pay market or the private-pay market? Or am I going to try and play in both and, if so, how?'”
With China, Guthrie adds, “You're looking at a private-pay market unless you're going to try to get on an essential medicines list. Am I then prepared to price to get on that—which is a very different game—or am I going to charge a premium and go for private pay? Then what is the impact of the inevitable generic going to be?”
(For more on the emerging markets, see “Getting into the Indian pharma market”, “The Middle East: A pharma market in the making”, “Reassessing Russia's pharma market”; “Breaking into the Brazilian pharma market”; and “How to get ahead in 'pharmerging' markets”.)
Compensating for lack of epidemiological data
All of this has to be managed with limited access to what, in other markets, are the raw materials of forecasting.
“The way we've always done new opportunities is to look at the epidemiology and say, How many people have got the illness? How many are diagnosed? How many are treated? How many of those treated am I going to get? How many are my competitors going to get?” Guthrie notes.
“And that's fine unless you're talking about markets like Russia or Brazil, and to some extent China, where you actually don't know how many patients there are, you don't have good sources of epidemiology.”
There also need to be “additional lines within that cascade. That may be the number diagnosed, but how many of those can actually afford my therapy?”
How the forecaster compensates for lack of epidemiological data will vary according to the therapeutic area and size of investment, Guthrie says.
If there is current treatment, “then there are market data. If there's not even current therapy, then what's my best surrogate? Is the Spanish epidemiology that we've got close enough?”
HTA and the need to show a viable cost-benefit equation is something else forecasters need to be thinking about, although not so much in emerging markets, which are more of a price-volume equation.
(For more on HTAs, see "Market access: How to meet both marketing authorisation and HTA needs" and "HTAs go global: What it means for market access‟.)
In Europe, though, where there is growing price contraction, or in the US, where comparative effectiveness is starting to seep into the mix, assumptions that a new product with better efficacy and/or safety will automatically command a premium price or be taken up as the standard of care may no longer hold water.
As companies try to incorporate HTA earlier in the R&D cycle, that further complicates projections of what will come out of the pipeline, how it will perform in the marketplace, or whether it can even satisfy more rigorous internal or external demand for cost-benefit advantages.
“I think companies are now aware they need a strong health economics package when they get to pricing and reimbursement,” Talbot-Watt says.
“The trouble is that health economists can only work—and I wouldn't call it forecasting—in a very defined way. Traditionally, they tend to work with clinical trial data and basically run probabilities, various different factors, through models based solely on those data. But you can't extrapolate outside that clinical trial.”
(For more on comparative effectiveness, see "How to build value through comparative effectiveness research" and "Health data and comparative effectiveness".)
New value equations
Even forecasters, though, may find the new value equations difficult to pin down.
“I think what has changed is our definition of a meaningful benefit,” Guthrie comments.
“That meaningful benefit just used to be clinical efficacy, and it was determined by the physician. Now we're talking more about the benefit being defined by the payer, including the customer who's taking the medicines—the patient as payer.”
Moreover, Guthrie says, “it has always been difficult to anticipate the future interpretation of benefits. We can do market research analysis … but the payer environment can change rapidly, because governments can intervene. And whereas people's values tend not to change quickly, government values and government intervention can.”
Also on the horizon, if not already carving out new market segments in categories such as oncology, is the trend towards personalized medicine.
One overarching uncertainty is how, and to what extent, personalized therapies will develop beyond what is still largely a conceptual stage.
As Guthrie points out, currently available targeted medicines are largely blockbuster oncology products using mechanisms that cut across tumour types.
By contrast, “if you look at the truly personalized medicine, they're modest products.” (For more on personalized medicine, see "Personalized medicine: A kick-start for innovation?".)
And adding in companion diagnostics creates a parallel track for forecasters.
“If you're thinking about personalized healthcare, you've got two uptake curves: the uptake curve for the diagnostic and the uptake curve for your product.”
(For more on companion diagnostics, see "Personalized medicine: Regulating companion diagnostics".)
What is also unclear is how much these new dynamics open up space for progressive forecasting tools that go beyond traditional order-of-entry models to incorporate new markets, stratified medicine, HTA, or shifts in therapeutic strategy within categories.
New models as a strategic imperative
It may be some time yet before vendors and consultants can persuade mainstream pharmaceutical companies, at least, that new models are a strategic imperative.
“It's not about models,” comments one forecasting manager, who stresses instead the fundamentals of understanding the market, factoring in key uncertainty drivers and “making your best guess.”
Moreover, he notes, forecasts are only useful if they support business plans.
In that respect, they must be transparent and easily communicated to management, whose interest levels may otherwise dip very quickly.
Guthrie shares this wariness: “Essentially, what we do is try to find an existing trend, which we project into the future, and then we add intelligence, events, all the things that are going to happen to change or disrupt that trend,” he says.
“When you're looking at a new opportunity, the epidemiology is the trend you're using as the basis … I haven't seen any new tools that do anything other than that in a different way.”
For Talbot-Watt, though, industry is now at “the tipping point of where these new sorts of techniques are needing to be used, because people are seeing that, especially with things like critical care forecasts, the models they have aren't actually able to cut it anymore.”
Niche players, diagnostics companies or the larger biotechs are paving the way, she adds: “They're the ones that are actually seeing their models breaking down … and they've already been wrestling with some of these problems for quite a while.”
These companies are “starting to move towards dynamic flow models. They're looking at better patient segmentation and data identification. And those two things together are really helping them drive their business case. Because they don't have large pharma budgets, they have to be very specific, very agile, and very reactive to get what's going on in their markets.”
(For more on dynamic flow models, see "Forecasting: How to get patient flow analysis right” and "Forecasting: Patient flow modeling comes of age".)
Big pharma, with its traditional focus on primary care and chronic disease, “hasn't needed these techniques yet, but I think in the next couple of years they will start to turn,” Talbot-Watt says.
One driver is the need to value new types of products either moving through pharma pipelines or absorbed through acquisition.
She also believes the knowledge base for forecasting has stagnated.
While the “big platform engines” took some of the legwork out of forecasting, knowledge “dropped away because people didn't have to produce forecasts anymore. And these days it's quite hard to find people who are experts or know a good forecast when they see it.”
Communicating to senior management
While new approaches such as dynamic patient models are starting to make their mark, the challenge is communicating their value to senior management, Talbot-Watt says.
“They are used to prevalence-based models and are inclined to say, "Why do we need to do anything different?" It just means people have to dust off their brain cells and think very logically about what's happening in their market.”
“When I've built dynamic models in areas where the patients are very tightly defined,” she adds, “if you communicate that properly—rather than just saying, 'Actually, your patient population is only half what you thought it was,' you say: ‘This is how your label is developing, this is the patient group that is now eligible’ and you explain exactly why that is. Then they actually do understand.”
Whether and to what extent these new techniques catch on, the pharmaceutical market is becoming increasingly polarized in a way that will continue to throw up challenges for forecasters.
“I think one of the bigger differences is that we have a current market that’s very big and whose growth is uncertain,” Guthrie comments.
“And we have some future markets that are not currently huge but whose growth is forecasted to be quite significant. So the today versus tomorrow dynamic is probably bigger now than it’s ever been.”