- Статьи
- Science and technology
- Take a calculation: the digital platform predicts tumor growth with high accuracy
Take a calculation: the digital platform predicts tumor growth with high accuracy
Scientists have conducted a study that will help drug developers select the optimal dose and evaluate the effectiveness of new cancer drugs in the early stages of clinical trials. By comparing various mathematical models that are widely used to predict the dynamics of the tumor process, they identified those that most accurately predict how the neoplasm will respond to treatment. The results of the study will be used in the Oncomonitor digital platform. Experts noted the importance of such tools for launching new drugs on the market, but told Izvestia that further research needs to be conducted on a larger number of nosologies.
Models for predicting tumor development
Scientists at the Center for Mathematical Modeling in Drug Development at Sechenov University conducted the first comparative study of five of the most widely used models for predicting the dynamics of the tumor process in drug selection.
As the scientists told Izvestia, one of the most difficult tasks in the development of new antitumor drugs is to select the optimal dose and evaluate the effectiveness of the drug in the early stages of clinical trials, when there is still very little data. To solve these problems, the developers use mathematical models that describe how the tumor behaves under the influence of the drug and predict the further course of the disease. However, until now it was not clear which of the existing ones copes with this task best.
The researchers conducted their study based on clinical data from 381 patients with non-small cell lung cancer (NSCLC). Using advanced statistical analysis, they identified the three most reliable models — BiExp, LExp, and TGI — and determined their scope of application. The TGI model of tumor growth inhibition (Claret) has become a leader in describing current data and short-term forecasting of neoplasm dynamics. While LExp has demonstrated the stability of predictions for long-term forecasting (up to 16 months).
All three models assessed the objective response of the tumor well (one of the key indicators of drug effectiveness), but none of them was able to accurately predict the moment of development of neoplasm resistance — this is a challenge for future research.
"Our work provides scientists and pharmaceutical companies with a methodological basis for selecting the optimal model that allows them to more accurately predict the dynamics of tumor size and facilitate decision—making during clinical trials," said Anna Mishina, one of the authors of the work, a junior researcher at the Center for Mathematical Modeling in Drug Development. "This is an important step towards personalizing therapy and accelerating the development of new anti—cancer drugs.
The results of the study will be used in the Oncomonitor digital platform. This tool will allow doctors and scientists to make more accurate forecasts of survival and select the optimal therapy for patients with cancer, the scientists added.
Mathematical modeling in drug testing
Choosing the optimal dose in the treatment of oncological diseases is an important task. Too low a dosage will not lead to the desired therapeutic effect, and excessive will cause excessive toxic effects. Drugs for the treatment of oncology have severe side effects and affect the body, which is already severely weakened by cancer, which causes new problems, Stanislav Stragnov, head of the Laboratory for the analysis of public health indicators and digitalization of healthcare at MIPT, told Izvestia.
This study will help pharmaceutical companies and scientists to get specific recommendations on the choice of tools in the early stages of drug development. This, in turn, can save time and resources by working on the most promising options. The models accurately assess the objective response of the tumor, which is important for making decisions about the dosage and design of the third phase of trials, added Maxim Kotov, market expert at NTI Helsnet, oncologist at the N.N. Petrov National Research Medical Center of Oncology.
— The main problem, honestly noted by the authors, is that not a single model predicted the moment of resistance development. This is a serious limitation for practical oncology, since it is the progression of therapy that determines the need to change treatment tactics. The second point is that mathematical platforms remain a tool for population—based rather than personalized forecasting," he noted.
According to him, the individual variability of the response associated with the molecular profile of the tumor, the microenvironment and the immune status of the patient is not sufficiently taken into account in such models, since this requires a large amount of data.
Improving forecasts already at the early stages of clinical trials (Phase I–II) can reduce pharmaceutical companies' costs by 30-40%, which means hundreds of millions of dollars in savings on a single drug, says Maxim Kolyasnikov, associate professor at the UrFU Institute of Economics and Management and the Department of Future Technologies at MIPT. It will also accelerate the launch of effective drugs on the market, which is critical for cancer patients, where the bill is for months of life. The study can indeed become a methodological basis for the Russian drug development platform, especially in the context of sanctions pressure and the need for import substitution of Western modeling systems.
— However, competitiveness requires a realistic assessment: the sample of 381 patients with NSCLC is relatively modest compared to the FDA or EMA studies operating on thousands of cases. Global pharmaceutical giants (Roche, Pfizer, Novartis) They are already using similar approaches, investing tens of millions in their own pharmacometrics platforms with AI components. Critically, none of the models predicted resistance, which is a key problem in cancer therapy, which limits its practical value," Maxim Kolyasnikov explained to Izvestia.
To compete at the global level, scaling of samples, validation on multiple nosologies and, most importantly, evidence of a real improvement in clinical outcomes in prospective studies are needed, the expert concluded.
The results of the scientists' work are published in the scientific journal CPT: Pharmacometrics and Systems Pharmacology.
Переведено сервисом «Яндекс Переводчик»
