Five Tips to Improve Your Oncology Forecasting Efforts

Oncology continues to dominate pharmaceutical pipelines, accounting for over 30% of global clinical development activity1. As cancer therapies become more personalized and data-intensive, the need for accurate oncology forecasting has intensified. The right forecasting model can help avoid costly missteps: errors in predicting demand or patient segmentation that may result in millions in lost revenue or delayed access to life-saving treatments. Moreover, with next-gen AI forecasting tools now entering the landscape, the question remains: are pharma organizations following the right forecasting approach to keep pace with the future of oncology?

What is oncology forecasting and why it’s critical to get it right

Oncology forecasting involves predicting future trends and outcomes in cancer research, treatment, and market dynamics. This process relies on data analysis and predictive models to estimate cancer incidence , treatment efficacy, market performance, patient survival, and more. With these insights, healthcare providers, pharmaceutical companies, and policy makers are better equipped to make informed decisions. Oncology forecasting can , thus, help enhance patient care and shape strategic initiatives.

Navigating the oncology market can be a challenging task for forecasters. Unlike other therapeutic areas, forecasting oncology treatments and drugs require a very unique approach due to the complexity of the disease.

Since oncology treatments are designed to target a particular patient population depending on the line of therapy, presence of biomarkers, or tumor types, forecasting models should factor in every critical element, such as patient identification, the likelihood of treatment switching and discontinuation, and time of therapy for different patients.

However, it is not always easy to develop detailed and accurate forecasts in this space, especially when the oncology environment is a rapidly evolving one. Hence, there is a significant need to adopt oncology forecasting best practices for better accuracy, increased reliability, and model robustness.

Top 5 oncology forecasting tips

Here is a list of 5 essential tips to help forecasters navigate the shifting sands of oncology trends in treatments and build an effective forecasting approach.

1. Building a monthly forecasting model

Before building an oncology quantitative forecasting model, it is important to understand the level of data granularity that users demand on an immediate and mid-to-long term basis. Annual models, albeit easier to build and maintain, do not answer key business questions like monthly sales.

It can also be difficult to adapt to an event like a data readout, where changes in forecasting output are needed at a monthly level by business users. These challenges make annual forecasting models inflexible with low precision. On the other hand, a monthly model can offer an ideal time granularity for forecasting because it incorporates oncology-specific dynamics based on available monthly data.

2. Identifying the right target patient pool

Forecasting in oncology is different from other therapeutic areas because of the significant need to follow patients through different stages, lines, and treatments as they progress through the disease. As important as this is to do, inaccurate identification of the target patient pool has been a common pitfall in oncology forecasting. Forecasters should split the population into smaller and more specific segments, and accurately model them based on incidence, recurrence, diagnosis, treatment, and other important factors to maximize the accuracy of forecast outputs.

Recent oncology trends also show increased reliance on real-world data to refine patient segmentation, particularly when evaluating rare cancer subtypes or biomarker-defined groups. Integrating these inputs into oncology treatment forecasting frameworks improves reliability.

3. Understanding dynamic patient flows

Forecasters must be able to model patients through the different stages of the disease as cancer therapy models have become more complex. They need to assess the advancement of each patient segment, understand how patients move between the lines of therapy, analyze dosing regimens, rates of progression, remission and discontinuation, patient dependency on old and new drugs or therapies, and more. A holistic understanding of the disease space, as opposed to a myopic one, is critical for forecasters to model such complex and dynamic patient flows .

Moreover, AI forecasting tools now leverage patient-level simulation models to accurately represent treatment transitions, such as line switches and re-initiations to enable forecasters to predict real-world therapy flows with greater clarity and granularity.

4. Understanding dosing patterns

Understanding the dosing patterns of your target patient population is crucial before building a quantitative forecasting model. In oncology, drug dosing is influenced by patient-specific characteristics such as body surface area, age, weight, and more. Each of these inputs need to be modelled differently, because patient segmentation and the associated granularity heavily depend on drug dosing specifications.

Duration of Treatment (DoT), Persistency Curve, and cohort models can be considered for oral targeted therapies that carry a fixed dosing approach. These factors can also be considered for monoclonal antibody (mAbs) therapies that have a weight-based dosing approach.

5. Estimating bolus demand

In addition to estimating the first-time users of your new product, forecasters must also look at 2 additional streams where patients could potentially flow in from:

1) Patients transferred from one product to another midway before completing the line of therapy

2) Patients re-initiating the treatment after temporarily discontinuing it due to tolerability issues

This is called the Bolus demand, and it forms a critical part of any forecasting model. Forecasters often tend to underestimate the demand from these channels, but it accounts for an average of 30% of the total product demand in the first 6-to-9 months of launch.

These are our top five strategies that we believe will shape the effectiveness of oncology forecasting.

How AI forecasting models are reshaping oncology forecasting

As oncology forecasting becomes more complex, traditional models like linear trend analysis, ARIMA (autoregressive integrated moving average), or Holt-Winters often fall short. These methods struggle with noisy or irregular data, require extensive feature engineering, and lack flexibility to adapt to evolving treatment dynamics or patient behavior.

In contrast, advanced AI forecasting approaches—such as deep learning models like the Temporal Fusion Transformer (TFT)—offer greater accuracy and scalability. These models can ingest diverse data types, support multi-step forecasting, and adapt in near real-time as new data becomes available. Their ability to continuously retrain and monitor forecast accuracy makes them increasingly valuable for modeling oncology's rapidly shifting landscape.

The way forward

Forecasters need to stay aligned with the evolving dynamics and complexities of the oncology trends, while proactively avoiding common pitfalls that can compromise the integrity of their forecasting models. As oncology continues to be one of the fastest-growing therapeutic areas, these tips offer a structured approach to implementing best practices, and bring better accuracy and more insights to your forecasts.

Additionally, integrating AI forecasting capabilities can support better decision-making by improving model adaptability, scalability, and accuracy. As forecasting needs grow, adopting the right mix of domain expertise, structured methodology, and AI-driven tools will be key to staying ahead.

References:

1. Grand View Research. Oncology Clinical Trials: Current Dynamics And Pipeline Outlook [Internet]. San Francisco (CA): Grand View Research; 2024 [cited 2025 Jul 22]. Available from: https://www.grandviewresearch.com/market-trends/oncology-clinical-trials-current-dynamics-pipeline-outlook