Aspetti critici negli studi clinici: problemi di metodo e applicazione

A clinical study is useful to clarify some planned questions and it follows a well defined scheme (the study design). This PhD thesis has been divided in five main sections following a scheme guiding the reader in the analysis of critical aspects it's possible to find, first in data management, then in the study design and during the development and the final phases of a clinical study. Section A, Problems in database management, proposes a chapter about Record Linkage and a second chapter about Missing Values. Record Linkage consists in a series of statistical and data processing techniques for data integration and for the union of more datasets on the basis of patients' identifying variables. The problem of missing values is frequently present in a study: when a database contains missing values it's possible to turn to specific systems useful to calculate a suitable value for the unit without information, through statistical programs. In section B, Study design, first clinical trials are analysed, then surveys are described and divided in cohort studies and case-control studies. In a cohort study (also called follow-up or prospective study) one or more groups of subjects are defined in accordance with the exposure to risk factors for one or more diseases, or not. Subjects are prospectively followed to study the disease(s) incidence and to observe if the disease(s) is(are) correlated with the etiological factors. A case-control study (or retrospective study) provides a research strategy to investigate possible factors preventing or causing a particular disease. The method implies a comparison between patients with the disease (cases) and a control group (healthy subjects). The comparison ends to find out factors which can be different in the two groups, explaining the presence of the disease in patients. Section C concerns Variables treatment and Outcome variable classifications. In the first part, types of variables, their measurement scales, their reduction in indicators and their transformations are analysed. In the second part the classifications of the outcome variable is described. These classifications may regard a specific disease (ICD IX and X), the patient's symptoms (ICD) or a disease effect, i.e. an impairment, a disability or a handicap (ICIDH and ICF). Sections D and E deal with choice of the multivariate analysis method. Section D aims to give a synthesis of the main analysis methods, dividing them in accordance with the variables symmetry. If there is a distinction between outcome variable and covariates, a multiple regression (stepwise or logistic) analysis method is suggested; discriminant analysis is useful if there are more outcome variables. If variables are all on the same level, factors analysis, principal components analysis, correspondences analysis and cluster analysis are proposed in conformity with the specific analysis aims. In section E logistic regression analysis is developed, pointing out the main aspects and considering the conditional logistic regression analysis for a Matched Case-Control Study design. Every chapter is organized in two parts: in the first one a synthesis of theory and practical methods are given to explain how literature deals with the specific problem in a clinical study; in the second one a specific application is presented to show how the author solved the methodology problems in a particular clinical-genetic study carried out at the Children's Hospital of Oulu University - Finland. In this Institute a substantial work on a preterm infants database creation has been performed. The database collected all the clinical and genetic information about preterm infants born in one of the three main hospitals in Northern-central Finland, Oulu, Tampere and Seinäjoki. Through this final database containing all patients from the three hospitals, a clinical-genetic case-control study (match 1-1) on the association between Respiratory Distress Syndrome (RDS) and single nucleotide polymorphisms (SNPs) has been carried out using a conditional logistic regression analysis. The project was directed by Prof. M. Hallman.