Modelization of bevacizumab effect on tumour perfusion assessed by Dynamic Contrast Enhanced Ultrasonography
M. Fontanilles(1), F. Pilleul(2), O. Colomban(1), B. Tranchand(1,5), A.M. Schott(3), J.A. Chayvialle(4), M. Tod(1)
(1) EA 3738 Ciblage Thérapeutique en Oncologie, faculty of medecine Lyon Sud, University Claude Bernard Lyon 1 of Lyon, Lyon, France; (2) Hospices Civils de Lyon, service de l’imagerie digestive, hôpital Edouard Herriot, Lyon, France ; (3) Hospices Civils de Lyon, Département d’Information Médicale et Recherche Clinique, Lyon, France ; (4) Hospices Civils de Lyon, service d’hépatogastroentérologie, hôpital Edouard Herriot Lyon, France; (5) Centre Léon Bérard, Lyon, France
Objectives: Bevacizumab (Avastin®) is an angiogenesis inhibitor. Its clinical efficiency in the treatment of solid tumour in association with classical chemotherapy has been demonstrated by several authors [2]. The clinical use of bevacizumab would be improved if a predictive factor of its efficacy was established. The aim of this work is to assess the role of Dynamic Contrast Enhanced Ultrasonography (DCEUS) as an early predictor of response to chemotherapy with bevacizumab. DCEUS is a medical imagery by ultrasonography which uses a contrast agent containing microbubbles (Sonovue®). This agent is strictly intravascular and its kinetics (intensity function of time) allows an assessment of tumour vascularisation.
Methods: 13 patients with metastatic colorectal cancer, treated by 5FU, irinotecan, leucovorin and bevacizumab entered this study. The imagery analysis performed on the hepatic metastases resulted in three curves of intensity at day 0, 21 and 49 plus 3 measures of tumour diameters (scanner) at month 0, 2 and 4 were obtained. The work is divided into two parts (1) estimation of hemodynamic parameters in tumour and (2) building of a predictive model of response. (1) The hemodynamic parameters were estimated from the curve of intensity using a multivessel model adapted from the model published by Krix and al[1]. The model was fitted using Adapt II. The estimated parameters were: the slope of increase m (~velocity: ν), the maximum intensity N0 ( ~ blood volume) and the elimination rate constant Ke. The secondary parameters were: blood flow f ( ~ N0*ν) and perfusion P (~ f / Tumour volume). (2) The predictive model of tumour volume variation at month 4(M4) was built by multivariate linear regression. The explicative variables are the estimated and secondary parameters, plus their relative variation between day 0 and day 21 or 49
Results: The modified Krix model fitted very well the sonographic data. The median tumoral perfusion at D0, D21 and D49 was 3.87E-02, 3.74E-02 and 4.38E-02 mL/s/tumour volume respectively. The best predictive model incorporated N0, P, the variation of both blood flow and perfusion between D0 and D49 as variables to explain the relative variation of tumour size at M4: Δtumour size at M4 = -0.0063*N0 + 0.19* P + 0.132*Δ f - 0.0978*Δ P Each parameter contributes significantly (p < 0.05) to the model with a standard error acceptable. The overall p value is < 10-6 and the R-squared is 0.96. No significant relation has been found between parameters at D21 and tumour size at M2 or M4.
Conclusions: A model able to predict early the treatment response by using a non invasive medical imagery was established. The analysis provided relevant information for understanding the pharmacological action of bevacizumab on metastatic vascularisation. The inhibitor of VEGF seems to reduce blood flow but to improve the action of associated chemotherapy by increasing the perfusion of the tumour. The study is ongoing in order to validate the model.
References:
[1] Krix M, et al. Ultrasound Med Biol. 2003; (10):1421-30.
[2] Hurwitz H, et al N Engl J Med. 2004; 350(23):2335-42