Epidemiological data Regression

Surprisingly, no clinical trials have specifically addressed the contribution of hypertension to the development of peripheral arterial disease. Most of the data comes from multiple regression analysis of many parameters looking for correlations, or subgroup analysis of trials designed primarily to look at coronary artery disease. However, Kannel cidA.. (6) has shown that epidemiological data analyzed in this way accurately predicts disease risk in a variety of American population samples, and in elderly as well as young coronary candidates. 

Note Seafood consumption is based on disappearance data calculated as catch plus imports minus exports. Prevalence rates of major depression are from the Epidemiological Catchment Area Study, with the exception of Japan, as described in Hibbeln (1998). Results of a simple persons linear regression that includes all countries is (r = -0.85. p < 0.0005) and that excludes Japan (r=-0.74, p <0.03). These data do not demonstrate a causal relationship between seafood consumption and lower prevalence rates of major depression. 

Epidemiological studies provide another means of indicating a cause/effect relationship. In retrospective studies it is often not possible to characterize the exposure which may have occurred many years previously. In prospective studies, however, data are kept on a number of variables including the primary exposures of interest and what may be secondary exposure variables and demographic variables. In this way, one can perform regressions to determine significant correlations between exposure and effect. It is extremely critical that epidemiology studies be carefully constructed so that sufficient variables are considered to assure that the exposure correlations are not disputed. 

In another study (Vahter and Friberg 1988), QC samples consisting of both external (unknown) and internal (stated) concentrations were distributed to laboratories participating in the epidemiology research. In this study, the maximum acceptable deviation between the regression anaiysis of reported results and reference values was set at Y=X (0.05X+0.2) for a concentration range of 0.3-5.0 pg Cd/i. it is reported that only 2 of 5 laboratories had acceptable data after the first QC set, and only 1 of 5 laboratories had acceptable data after the second QC set. By the fourth QC set, however, all 5 laboratories were judged proficient. 

---