The old maps just don’t work as well as they used to. That’s true of both ancient maps of the world and out-of-date maps of the pharmaceutical market. According to David Godolphin, vice president of the Equinox Group, we tend to rely too much on what we already know. At the eyeforpharma Pharma Forecasting Excellence USA in October of 2007, Godolphin spoke on “Improved Forecasting through Mapping Market Structure and Patient Flow.” Thanks to access to improved data and technology, our ability to map opportunities has increased, says Godolphin; now we have the tools to see the current market more clearly and forecast the future with greater accuracy.
It was once sufficient to send the sales force out into “broad swathes of not-very-well-characterized territory,” Godolphin says. We could rely on primary care markets, blockbuster drugs and the power of detailing to ensure sales. But the market is changing. The shift is towards greater focus on specialty drugs – drugs that aren’t for the mass market. The question then becomes, where are our patients? And when will they become available for treatment? Questions like these can only be answered by sophisticated, highly targeted maps.
Differentiating market structure analysis from patient flow analysis
Market structure is much like a photograph – it shows what the market looks like now. The market share model provides detailed information about the current population, allowing for a very precise understanding of the different segments. This analysis also quantifies the current treatment mix and identifies places where market share can be increased or must be defended. It is a base from which an informed forecast can be extrapolated.
Patient flow analysis includes a temporal element; it tracks patients over time, including their treatment histories. While a prevalence-based model can determine how many people in a given area have high cholesterol, for example, patient flow analysis digs deeper, looking at what medication lines patients have already explored and exhausted. This model enables forecasters to predict changes over time, says Godolphin, “enabled by and in harmony with patient-level data.”
Making sense of market structure
Market structure analysis is very useful when, for example, a pharma company has too many variables and too many segments involved in forecasting. In one example, a client was looking at the points of intersection between diabetes and obesity. The initial picture, Godolphin says, was chaotic. There was a great deal of variance in patient groups with unmet needs, prevalence rates and treatment rates within the segments. In the initial model, there were sixteen potential populations: too much to produce an accurate forecast for a product in the early stages of development.
The solution was to trim the model, to aggregate those groups whose differences were not clear or meaningful enough to merit separate categories, and to eliminate other groups who fell outside the scope of the study. Once the “morbidly obese” and the “obese” were grouped together, and people with Type 1 diabetes and healthy people with no risk factors for Type 2 diabetes were removed, the chart was reduced from 16 segments to eight. A “rigorous analysis of prevalence and treatment rates and mix” halved the segments again, from eight down to four. A close look at those segments revealed that weight had a significant impact on Type 2 diabetes rates, and the segment of obese people with Type 2 diabetes merited continued focus. The analysis enabled a clear picture of the market and a forecast that was not impossibly complicated.
So when is market analysis the right tool? Says Godolphin, this model is valuable when there are many comorbidities, etiologies, distinctions in severity or when there’s a large gap between reported prevalence and current market size (for example, smoking cessation or obesity – both of these have large un- or undertreated populations). Market analysis can be useful when the picture is complicated by treatments that vary largely due to demographic factors such as age, ethnicity or gender.
Patient flow: a natural history model
In this case study, the challenge was to understand the potential market for new therapies for the Hepatitis C virus (HCV). There was a large reservoir of patients, many awaiting good therapies for their condition, but there was a diminishing inflow of patients. A look at markets worldwide showed that mature markets like the US had low incidence rates of new infections and relatively high treatment levels for those with the disease. Emerging markets such as China had a much different picture, with high rates of new infections and low treatment rates for the infected. The client wanted to know if there would be time to develop and market a new HCV therapy before the patient reservoir emptied and the opportunity was gone.
Says Godolphin, the complexity of the market for HCV therapies requires a model that can incorporate all the key drivers. One such complication was incidence factors – subject to radical changes and differing widely across countries, rates of infection can be dramatically different over time and distance. The relatively long delay period between infection and manifestation of clinical consequences can mean a large number of potential patients exist who are not currently under treatment. As medicines and technologies improve, so will rates of diagnosis and treatment, further complicating the picture. Finally, the model needs to account for the potential for re-treatment of some patients in the future.
The solution, according to Godolphin, is to follow patients over time using a natural history model. A simple patient-based prevalence model looks at prevalence (how many people have the disease), the diagnosed (how many people have been diagnosed with the disease) and the drug-treated (how many people are receiving treatment), and then uses that information to forecast rates over time. However, HCV incidence rates show wide variation: infection rates differ by age, for example. And there are historical variations, with people who were infected as long ago as the late 1960s just now entering treatment programs. They key here is to account for the complex ways the patient source develops.
It’s also necessary for the model to be flexible enough to accommodate changes in the eligible population. Some treated patients may be cured and fall out of the patient source; some may fail and require re-treatment at a later date. Some patients may simply become too sick to be treated. Whether patients fall out of eligibility due to cure or mortality needs to be reflected in the model.
So, what is the result of such complex modeling? Says Godolphin, you get a “detailed and credible” view of the evolution of the market. With the natural history model, the client was able to get a picture of changes in the HCV market and understand the time urgency in places like the US where treatment rates are leveling off and the number of patients needing treatment is dropping. The model also demonstrated the relatively low treatment rates in China and forecast that China’s expanding access to health care would mean a large population seeking and receiving treatment in future years.
The natural history model, Godolphin says, is valuable when new therapies could make dramatic differences in the course of the disease and the size of the patient population. The natural history model allows for time elements to be added to market structure analysis – particularly useful if the market could potentially shrink and correct timing of new therapies is critical. This model may be best when there are substantial differences in “treatment paradigms” across countries.
In another example of a patient flow model in use, a client with a therapy for rheumatoid arthritis wanted to identify sources of patients. The key for this client was in keeping track of patients’ unsuccessful use of previous therapies. By tracking patient histories, the client was able to determine when patients became available for treatment and what therapies remained open to them. By looking at the market “line-by-line,” the client got insight into market dynamics that would have been invisible in a simple prevalence view model. The third line of treatment emerged as the best place for the client to place the new therapy, and the client was able to target sales and marketing resources toward that segment.
The patient flow model is valuable for dealing with progressive diseases that have multiple lines of therapy and treatment failures; for event-driven conditions (for example, following a myocardial infarction); for identifying “low-hanging fruit” when an agent is nearing launch; for dealing with multiple settings for treatment (hospital, office, long-term care) or for handling complex diagnosis, treatment and referral patterns.
Making the most of patient-level information
Now that you have all this information, how do you use it to create an informed, accurate forecast? Patient information capture has gotten quite sophisticated, and it’s possible to have considerable accumulated data: for example, what are the drivers behind a patient’s therapy choice (disease severity, age/race/gender, treatment history, comorbidities, source of payment/copays, physician specialty)? It’s now necessary to track numerous patient characteristics and regimen choices to get the most complete, sophisticated picture of the market. Says Godolphin, we can no longer just attempt to determine how many patients are out there, now we need to identify which patients are best to target.
“Discrete choice simulation,” according to Godolphin, means following patients through time, watching the choices the patients make about treatment. A mathematical model then generates thousands or tens of thousands of virtual patients based on the patient-level information captured and projects what will likely happen to those patients over time and with treatment by new therapies. The model looks at treatment characteristics and how they impact a patient’s decision (these include efficacy, side effects, relief from symptoms as well as copay and treatment availability), as well as characteristics of patients (current treatment, disease severity, physician specialty and insurance status). Both of these streams feed into the treatment appeal of a therapy, and the model “looks to increase the appeal of treatment” at each time point. Outputs of all this data are used to guide marketing resources and sales force towards the most promising targets.
Patient flow simulation is best when patient characteristics affect treatment choices (such as previous treatments and responses), when patient-level data are available, when there are highly complex treatment regimens to be modeled, and when the question isn’t “how many potential patients are there,” but rather “which patients should we be targeting?”
With so many models available and the market paradigm in flux, it is perhaps as useful to know the right questions as it is to know the right answers. Thanks to increased information and technology, it is possible to build a better map of the marketplace. Do you know where on earth your best opportunities are?
Author: Shannon Perry, journalist, theforecaster
