RICERCA ITALIANA     

Statistical methods for learning in clinical research

Politecnico di Torino

Abstract

In this proposal the scientific expertises of several statisticians active
in the field of biomedical research address the requests
for new methodology coming from several centers of clinical applications.

As a matter of fact, each unit is characterised as working in strict
collaboration with one or more national or international clinical research centers.
In particular, preferential relationship are as follows:
the Torino Politecnico Unit with Istituto Tumori Toscano, with Novartis Pharma
and with the International Breast Cancer Study Group;
the Bologna Unit with Prima Facoltà di Medicina e Chirurgia, Università di Roma;
the Milano Unit with Istituto Nazionale Tumori
and with the European network of excellence BIOPATTERN,
the Novara Unit with Facoltà di Medicina e Chirurgia,
Università del Piemonte Orientale.

The goal is to study several statistical issues that arise
in clinical contexts and that can be collectively described as
learning problems in biomedical research.

Large scale phase III clinical trials are the basic tool for
comparative demonstration of safety and efficacy of a new therapy. In
addition to them, there exist a variety of other kinds of clinical
trials, and in general of biomedical studies, meant at exploring some
important learning issues.

The learning process can take place >>>

Principal Investigator

Mauro Gasparini Politecnico di TORINO

Research Objectives

Clinical trials have been called "a success story" for our society,
since "They represent one of the few areas in which the effects of new
technology or new programs on humans are studied using modern
statistical principles of experimental design" (Simon 2006). This is
the starting point of this project, which focuses on some
methodological aspects of the statistics used in clinical trials and
in more general biomedical studies.

The same aspects are also present in other fields of modern
experimentation, like quality management in industry or genomics in
biological research. But, since clinical trials are a mature and
widely participated field, basic fundamental concepts and correct
statistical practices can be taken for granted, and our proposal can
focus on more specialized inferential procedures which deserve more
methodological thought.

Clinical trials are the stepping stones for clinical
development, comparative phase III trials being the most central ones.
In the case of a pharmaceutical company, phase III trials are mainly
meant to get approval from the Health Authorities to market the new
drug, whereas in the case of independent trials a similar "phase III"
exercise is done so that the new method gains general recognition and
can be recommended for medical practice.

In >>>

First Results

This proposal has not purely scientific, but also very applied goals
and, as such, is mainly aimed at the production of two kinds of scientific output:

1) methodological papers aimed at fostering the progresses
of contemporary statistical methods;

2) applied papers aimed at trasferring to clinical settings
the progresses acquired at the theoretical level.

Papers of type 1) will exhibit joint work of personnel from all units,
either within a single unit or aross several of them.

Papers of type 2) will document the strict relationship each unit
entertains with centers of clinical research, in particular:

- the Torino Politecnico Unit with Istituto Tumori Toscano, with Novartis Pharma
and with the International Breast Cancer Study Group;

- the Bologna Unit with Prima Facoltà di Medicina e Chirurgia, Università di Roma;

- the Milano Unit with Istituto Nazionale Tumori and with the European network of excellence BIOPATTERN;

- the Novara Unit with Facoltà di Medicina e Chirurgia,
Università del Piemonte Orientale.

Therefore, expected results will also be the general improvements of the
methodology level of biomedical research in those centers and in
general in the Italian scientific and economic system.
The interaction with care-producing centers and drug-producing companies
>>>

Timescale

24 months

National and international background

Dose finding is typical of phase I in that when a new drug is
candidate for clinical development, the most important step is finding
a suitable dose or dose range for it, typical by methods of dose
escalation. Dose escalation is a trade-off between efficacy and
safety, and poses challenging statistical questions.
It is often the case that the significant portions of the efficacy and
the safety curves are on the same range, interfering with one another
and making the dose escalation process a delicate inferential
procedure. Recent bibliographical references are Ting (2006) e Chevret (2006).

Certain clinical trials, especially adaptive ones, pose a critical
dilemma for investigators, known as the individual-versus-collective
ethics. It is necessary to minimize potential harm to the patients
presently under care and maximize the experimental information, two
goals which are sometimes in conflict. The issue is discussed in
Royall (1991).

In an adaptive experiment, the assignment of patients get modified as
the experiment progresses, on the basis of the information gathered up
to that point. Covariates are commonly used in a design-adaptive way,
to correct a possible imbalance of the allocations during the trial,
but Hu and Rosenberger (2006) in their recent book suggest an
alternative response-adaptive utilization. The >>>