In a presentation at Pharmaceutical Management Science Association’s Fall Symposium 2022, Indegene’s Shri Amrute and Steven White highlighted one of the most challenging areas in pharmaceutical forecasting — the oncology forecast. This blog covers the highlights of their presentation.
Oncology forecasting has several features making it particularly complex, including (but not limited to):
Products approved in multiple diseases, under a variety of conditions within each disease
Different regimens in different lines of therapy—and with different patterns of disease, remission, and relapse
High toxicity of some agents, leading to adjustments of regimens within a line of therapy
Cross-line reduction – where patients often cannot be treated with a given therapy more than once
Patient flow models have some distinct advantages in the oncology forecasting space. They:
capture a patient’s entire treatment process
effectively model market complexities and illuminate the mechanisms behind them
include relevant drivers and decision makers determining which path of care will be taken, and when
provide a holistic view of the oncology environment, and can incorporate a wide variety of dynamic – not static – input parameters
Creating a good patient flow model is not easy. To do one right, a forecaster must identify many input variables. To get those input variables to be as accurate as possible, the forecaster must then assemble substantial amounts of information from sources like market research, real-world data, and clinical trial results. And getting those inputs right is the single most important factor in getting a good forecast.
Steven also outlined the specific challenges of the patient flow model, and outlined situations in which it may be “too much tool” for the input data available. For example, looking at strategic models of assets or indications still several years away from the market, many of the inputs that would be required for a patient flow model would be too tentative or speculative to use. In these cases, patient models that resemble patient flow models – but are not true patient flow models – might be a better option.
After discussing the complexities in oncology forecasting and how to address them in patient flow models, Shri and Steven talked about the best practices in oncology forecasting. They emphasized the importance of maintaining a definitive data repository in conjunction with any forecasting model – but especially a patient flow model. It is altogether too common for a large organization having different departments to use different sets of data for their models. This results in models having divergent results, often adding confusion to the decision-making process. While the data in a repository may change over time, it is imperative that everyone in an organization operate from the same data source – using the same playbook, if you will. That is the only way for the organization to view its world consistently.
Shri and Steven’s message concerning oncology modeling, ultimately, was simple: Use the best, most advanced tools you have available that are appropriate to the modeling problem. These tools could be a cloud-based forecasting platform or an Excel®-based platform. These models could be patient-flow models, patient models, or something else. But ultimately a model that is inaccurate, or that the management cannot understand or support, is not useful. Forecasters should adopt the best appropriate tools they can for higher accuracy and impact.