Forecasting for complex diseases
Peter Mansell reports on how advances in molecular and cell biology are making the forecasters job more complexand more business critical.
Advanced awareness of molecular and cell biology is honing down therapeutic targets to ever lower levels of the disease cascade.
That should bring long-term benefits to patients and health systems, in the form of better targeted, more cost-effective medicines with more predictable responses and fewer side-effects.
For market forecasters, though, and specifically those concerned with epidemiological analyses, this improved specificity presents something of a headache.
As David Robinson, senior director of epidemiology services in Kantar Healths Corporate Development division, explains, the starting point of any forecast is determining the number of eligible patients for a particular therapy.
Understanding the target patient profile
That calls for a thorough understanding of the target patient profile, and often clients dont really know what that is, according to Robinson, especially when a compound is still in the early stages of development and the increasing sophistication of drug discovery techniques has multiplied the therapeutic possibilities.
Nor do these clients necessarily grasp that the patient profile will change as the drug moves through development.
In oncology, for example, targeting disease mechanisms at cellular level may mean an initial focus on five to seven different tumor types, Robinson notes.
Epidemiological forecasters need to be aware of all these options and how they may be prioritized or discarded later in the development process.
Especially when biomarkers come into play as predictors of patient response, there is too often a dearth of reliable, well-conducted epidemiological studies that enable forecasters to estimate patient subsets, Robinson says.
As a result, they sometimes have to resort to generalizations that are broader than youd like.
That is not always the case.
One example is the distinction in oncology R&D between people who have the normal (wild-type) form of the KRAS gene and those with a KRAS mutation.
KRAS gene testing has been shown to predict whether patients will benefit from EGFR-inhibiting therapies, such as cetuximab (Erbitux) for colorectal cancer.
In this area, the epidemiological data are beginning to be fairly robust, Robinson notes.
Defining commercially relevant patients
But a few years ago, there was not much out there.
And as Robinson explains, the issue goes deeper than just whether a patient has the KRAS gene mutation or not.
There is then the question of whether the ratio of KRAS mutant versus wild-type patients holds true for newly diagnosed patients and for surviving patients diagnosed four years previously.
Alternatively, a drug developer will want to know whether the mutant to wild-type ratios in newly diagnosed patients who are the most studied patient type in epidemiology can be extrapolated to those who have a recurrence of disease.
It is all about defining the commercially relevant patient segment, Robinson comments.
The forecaster needs to come up with an estimate that is both collectively exhaustive (i.e., it takes into account all potential patients) and mutually exclusive (i.e., without double-counting patients).
If a company is developing a drug for major depressive disorder, social anxiety and bipolar disorder, for instance, there will be some degree of overlap between those conditions, Robinson explains.
Increasing complexity
The bottom line is that epidemiological forecasters have expanded the breadth of the studies we review to accommodate data more related to clinical trials, Robinson says.
That, in turn, means a longer review process and higher costs for clients.
All the same, he adds, if a pharmaceutical company wants to ensure payers will accept an expensive new therapy in an environment increasingly dominated by health technology assessments, it is in that companys interest to show the therapy is not aimed at everyone.
There is going to be a lot more research into elucidating variable responses to drugs, and the hope is that it will lead to better therapeutic targets, Robinson comments.
But for epidemiological forecasters, things will only get more convoluted.
Every few months sees news of a new and apparently important biomarker, and almost every drug research study now has a genetic component, he notes.
What is more, epidemiological forecasts have to be made globally, which means also accounting for huge variations in disease incidence both from country to country and within individual countries that may have an influence on genetic susceptibility to disease.
For pharmaceutical R&D departments and for the various support functions that try to ensure those R&D efforts will have scientifically and commercially viable outcomes, that is just increasing the complexity of everything we do, Robinson says.
For more on forecasting for oncology products, see The challenge of oncology forecasting.
For more on biomarkers, see Different forecasting methods in the US and Europe.