MAGIC-BULLET - Model-based support of translational steps throughout drug development of antibody-drug conjugates (ADCs) in breast cancer

Breast cancer was the most common type of cancer among females in developing and developed countries in 2020. Subtypes of breast cancers can be classified by the expression of certain proteins. Human epidermal growth factor receptor 2 (HER2) is such a protein involved in normal cell growth. However, larger than normal amounts of HER2, promote the growth of cancer cells. The HER2-positive subtype accounts for approximately 20% of all breast cancers and has always been considered one of the most aggressive cancers with the worst prognosis if left untreated. However, as new treatment options could specifically target HER2, we have observed just the opposite, namely a better prognosis in cases treated with such drugs. Treatment with HER2-directed drugs has become the usual treatment for women with early breast cancer. These drugs, especially if wisely chosen for the particular patient, will become even more important worldwide because we observed approximately 2.3 million new cases in 2020 and see a predicted annual increase of over 3 million new cases by 2040.

The first HER2-directed drugs were antibodies. However, tumors can become resistant (less/not susceptible) to them limiting their efficacy. In the last decade, a new drug class called antibody-drug conjugates (ADCs) was developed. ADCs are targeted drugs that deliver chemotherapy agents to cancer cells. As usual in clinical development, many ADCs failed to prove their benefit. Successful ADCs have shown an overall increased benefit within clinical trial populations (e.g. prolonged time to cancer recurrence). Nevertheless, this benefit can vary from person to person and may be influenced by a person’s characteristics. Therefore, using the summary results of these clinical trials in clinical practice is challenging.

Against this background, decision makers in clinical development or in routine patient care could be supported by recommendations from (statistical) models. MAGIC-BULLET therefore develops statistical models to (1) predict the benefit of new drugs from preclinical experiments to clinical trials in humans and to (2) predict individual responses to already marketed drugs. While project goal (1) primarily adds value to clinical development of new drugs in the future, project goal (2) can potentially improve patient care with already available drugs.

Therefore, we consider two ADCs to demonstrate that (1) the (early) clinical response can be predicted from preclinical experiments and that (2) individualized treatment decisions in clinical practice can be derived from clinical trial data. For project goal (1), we extend existing models to predict ADC behavior in clinical trials. For project goal (2), we study how the individual response to the ADC depends on patient characteristics and how this could possibly be used (via a model) to derive individualized treatment recommendations. For this purpose, we apply several modeling options including, the creation of weighted clinical trial populations that resemble patients encountered in clinical practice or by using the clinical trial information to predict how individual patients may benefit based on their own characteristics