Multi-arm clinical trials with finite response and non homogeneous urn function

• Autores: José Antonio Moler Cuiral , Fernando Plo , Miguel San Miguel Marco
• Localización: VIII Journées Zaragoza-Pau de Mathématiques Appliquées et de Statistiques / coord. por Manuel Pedro Palacios Latasa , David Trujillo, Juan José Torrens Iñigo , Monique Madaune-Tort , María Cruz López de Silanes Busto , Gerardo Sanz Sáiz , 2003, ISBN 84-7733-720-9, págs. 407-414
• Idioma: inglés
• Enlaces
  • Texto completo (pdf)
• Resumen

  • An adaptive design for a clinical trial with prognostic factors and more than two treatments is described using a generalised urn model in a random environment. Patients arrive sequentially and treatments are applied according to a function of the urn composition. This function may change at each stage. Patient's response is immediate and discrete, with a finite number of possible values. The evolution of the urn composition is expressed by a recurrence equation that fits the Robbins-Monro scheme of stochastic approximation. In this setting, we obtain asymptotic properties for the performance of each treatment and we illustrate the application of the rule with an example.