Success Stories

Advanced Analytics | Oncology | Promo

Forecasting in CAR T- Cell Therapy Assists in Setting Client Expectations for Patient Deliveries

Project Background

Chimeric antigen receptor (CAR) T-cell therapy is a cell-based gene therapy used to attack cancer by using the patient’s T cells, which have been altered in the lab, in order to find and destroy the patient’s cancer cells. It involves extraction of the patient’s blood (apheresis), the subsequent lab work, and then infusion back into the patient of the altered T cells. Our client had recently launched two products used in CAR T-cell therapy – one for multiple myeloma and one for large cell lymphoma. They tasked the analytics team at KMK with projecting their patient participation at the various stages of the process, starting with apheresis and going through infusion.

Unlike medications which can be taken multiple times over a period of time, CAR T-cell therapy is a once-and-done process necessitating a different forecasting approach than that used for traditional medications which are based on an epidemiology-based approach. To forecast the potential number of patients using the CAR T-cell therapy, we needed to look at how many patients were projected to start with the therapy and how many patients would drop off from the initial extraction to the infusion. We also needed to integrate this data with the potential general market share and demonstrate the projected market percentage the client could anticipate versus its competitors in the therapeutic space.

Challenges in the CAR T- Cell Therapy Space

Cell Therapy is an evolving treatment area, and as such, requires working with research and secondary data that is not necessarily proven. For instance, we obtained market research from two different data sources which offered conflicting numbers. Secondary data is also limited as many companies block the data one can normally draw upon in a source like Symphony in order to avoid competition. As a result, we could only use data derived from these sources as a guideline versus an absolute, which made forecasts far more challenging overall to develop.

KMK's Methodology

We developed a three-step approach to this assignment:

Define the problem statement

For this assignment, the main problem statement was identifying the number of patients that would be within the CAR T-Cell therapy within any given period

Analyze the problem statement and work to find the best process to solve it

We knew that the final number of patients going through infusion at the end of a given month would be a function of the number of patients who underwent apheresis during the previous month or the beginning of that month. Since our final patients were a lagging indicator of those undergoing initial apheresis, we could use that number of patients to forecast what the final number of patients undergoing the total CAR T process would be. Using this as our assumption, we used basic mathematical and statistical analysis to develop a forecast.

To provide a measure of confidence in our projections we did pressure testing of the assumptions that went into the forecast. For example, when we projected that X number of X percent of patients would be dropping off between apheresis and infusion, we had to keep looking at historical data to make sure that the assumption was correct and made adjustments over time to reflect actual occurrence.

We also worked with secondary data from sources like Symphony and EMR data such as PharMetrics to test out market assumptions on current market share or a prospective market share based on market research. For instance, based on clinical trials in the two therapeutic categories, we knew that our client’s therapy was going to be used as the third therapeutic approach, or in a certain section of the third approach. Therefore, we were able to use secondary data to see what percentage of patients starting with the first treatment approach ended up in the third approach. Although these numbers were not usable as absolutes due to the restricted availability of secondary data, we were able to forecast the percentage of patients going from first to third therapies, and then within the third level of therapy, what percentage of patients would potentially use our client’s CAR T-Cell therapy.  

We were also able to perform primary data analytics, using actual CAR T data retrieved from various data sources such as dashboards. We knew how many patients were enrolled, how many actually underwent apheresis, and then how many patients were infused. Using these data absolutes were we able to perform various analytics. For example, each week we projected how many patient infusions we expected by the end of the month using the known process time from apheresis to infusion and the number of patients enrolled and/or had undergone apheresis.  

Develop the actual solution and then make sure that solution was actually implementable, or at least usable, to the client.

To deliver a monthly forecast, we would start with weekly cohorts to see how many patients underwent apheresis every week, and then took into consideration the cancellations and the number of patients who went through infusion.

For example, if we wanted to see how many patients would be infused in a given month, we would take into account the amount of time from the apheresis date to the end of month date and drawing upon historical data, project the number of patients to be infused at different times in the month. For example, if the shortest time to infusion is 20 days and the highest is 80 days, we could project the % of patients to be infused within various time frames. These projections would also take into account the number of patients we anticipated to cancel based on whether or not they had crossed the statistical threshold to proceed to infusion.

Using this process, we developed a high estimate and a low estimate which formed the basis of our solution, providing a monthly forecast range of total patients to undergo the complete CAR T-Cell therapy. Knowing the number of patients who had already been infused for the month, combined with the forecasted numerical range of patients we anticipated to go through the complete process, we were able to provide the client with weekly updates on forecast so they could set revenue and resource allocation expectations for the end of every month.

Project Outcome

With our weekly and monthly analysis, we were able to answer management questions on patient participation so they could set expectations and plan based on forecasted ranges of patients undergoing apheresis and completing their therapeutic infusions, ultimately helping them to save money in their implementation of the therapeutic offerings. Additionally, when pressure testing the assumptions of the forecast, we gathered a significant amount data which allowed us to provide ad hoc analytics, answering additional management concerns as time went on.

Want to learn more?

Get accurate target call list for rare diseases.

Contact our expert Advanced Analytics team today to break down the barrier between data and insights.