American Journal of Epidemiology Vol. 142, No. 11: 1136-1146
Copyright © 1995 by The Johns Hopkins University School of Hygiene and Public Health
research-article |
Exploring the Role of Diet in Modifying the Effect of Known Disease Determinants: Application to Risk Factors of Liver Cirrhosis
1Institute of Statistical and Mathematical Sciences, University of Milan Milan, Italy
2Institute of Mathematics and Statistics, University of Ancona Ancona, Italy
3Division of Gastroenterology, Ospedale Maunziano Umberto I Torino, Italy
Reprint requests to Prof. Giovanni Corrao, Istituto di Scienze Statistiche e Matematiche "Marcello Boldrini," Università degli Studi di Milano, Via Conservatorio, 7, Milano 67100, Italy
The analysis of the combined effects of nutritional factors with other putative disease determinants in log-linear or logistic models is methodologically complicated by the strong multicollinearity between nutritional factors, resulting in poor precision in estimating the parameters. Furthermore, the generally used multiplicative structure is not always the most appropriate for describing the resulting joint effect of two or more factors on the disease risk. The authors addressed such problems in a case-control study assessing the interactions between alcohol intake, chronic hepatitis C virus (HCV) infection, and nutrient intake on the risk of liver cirrhosis. During the period from November 1989 to May 1990, 282 patients admitted to the medical departments of the hospitals of the Province of L'Aquila (central Italy) were enrolled: 115 cirrhotic patients aged 24–82 years (78 of whom were males) hospitalized because of liver decompensation, and 167 control patients aged 25–84 years (100 of whom were males) admitted to the same hospitals for acute diseases unrelated to alcohol intake, infection with hepatotropic viruses, and nutrition. No dose-effect relation was found between the intake of any nutrient and the risk of cirrhosis using classical methods. The analysis of principal components showed, however, that a pattern of higher lipid but lower protein and carbohydrate intakes was significantly associated with the risk of cirrhosis. The Breslow and Storer parametric family of relative risk functions showed that a multiplicative structure was the most adequate to describe the joint effect of nutritional pattern with alcohol intake and/or chronic HCV infection, whereas an additive structure best described the joint effect of chronic HCV infection and alcohol intake. In conclusion, the analysis of principal components and the Breslow and Storer family are useful tools to explore the role of diet on disease risk when precise pathogenic knowledge is not available. As an original finding, the authors suggest that a higher lipid intake, combined with lower protein and carbohydrate intakes, modifies multiplicatively the risk of cirrhosis associated with alcohol intake and/or chronic HCV infection. Am J Epidemiol 1995;142:1 136–46.
alcohol drinking; analysis; hepatitis C viruses; infection; liver cirrhosis; logistic models; nutrition; risk