OBJECTIVE: To illustrate the advantages of structural equation models in biomedical research using the complex example of cystic fibrosis. MATERIAL AND METHODS: 595 blood samples from 312 patients were tested. The model studied the effects of age, BMI and clinical condition on seven major latent variables: pulmonary function, lipid oxidation status, vitamins A and E, glutathione, carotenoids, two essential fatty acids and arachidonic acid. RESULTS: The model confirmed previous associations: positive (fatty acids, arachidonic acid, carotenoids and vitamins with pulmonary function and with lipid oxidation) and negative (glutathione with pulmonary function). It also verified the decrease in fatty acids during bronchial exacerbation and the increase in fatty acids and lipid oxidation after antibiotic treatment. Above all, the model revealed new positive associations between lipid oxidation and carotenoid levels and between lipid oxidation and vitamin A and E levels. CONCLUSIONS: Structural equations dealt easily with the great number of outcome variables of the example. They deserve a central place in biomedical issues involving too many correlated factors to help physicians and statisticians conceive biological models that best represent reality.
OBJECTIVE: To illustrate the advantages of structural equation models in biomedical research using the complex example of cystic fibrosis. MATERIAL AND METHODS: 595 blood samples from 312 patients were tested. The model studied the effects of age, BMI and clinical condition on seven major latent variables: pulmonary function, lipid oxidation status, vitamins A and E, glutathione, carotenoids, two essential fatty acids and arachidonic acid. RESULTS: The model confirmed previous associations: positive (fatty acids, arachidonic acid, carotenoids and vitamins with pulmonary function and with lipid oxidation) and negative (glutathione with pulmonary function). It also verified the decrease in fatty acids during bronchial exacerbation and the increase in fatty acids and lipid oxidation after antibiotic treatment. Above all, the model revealed new positive associations between lipid oxidation and carotenoid levels and between lipid oxidation and vitamin A and E levels. CONCLUSIONS: Structural equations dealt easily with the great number of outcome variables of the example. They deserve a central place in biomedical issues involving too many correlated factors to help physicians and statisticians conceive biological models that best represent reality.
Authors: James V Pottala; Gemechis D Djira; Mark A Espeland; Jun Ye; Martin G Larson; William S Harris Journal: Comput Math Methods Med Date: 2014-04-15 Impact factor: 2.238