BLACKBURN and JACOBS A SCHEMATIC OF POPULATION IP ) VERSUS INDIVIDUAL ( 1 ) CORRELATIONS A RISK B RATE MEAN LEVEL VALUE INFERENCE : AN ENVIRONMENTAL FACTOR

AN IMPORTANT SOURCE of current misunderstanding and controversy about diet-disease relationships is the inappropriate extrapolation of evidence from group data to the individual and vice versa. In this editorial piece the difference is illustrated between correlations of risk factors and disease found in individuals and those found between populations. An attempt is made to clarify the causal inference possible when these correlations are concordant and discordant. An analytic approach is encouraged in which both sources of information are considered, i.e., relationships among individuals and between populations, as well as their consistency with other data, before arriving at inference about the meaning of diet-heart, diet-cancer, or, for that matter, any statistical association. It is further suggested here that factors that determine the average and usual population levels of risk characteristics are likely to be the major determinants of mass disease. These influences on population risk may be the same as, or different from, those factors that predominantly influence risk factor level and the risk factor-disease relationship among individuals. We demonstrate that the causes of individual and population correlations are mathematically and conceptually independent. From an understanding of these ideas, the fallacy may be avoided of rejecting important causal influences in populations when relationships appear weak or absent in an individual patientas well as the inappropriate application of group data to the individual, which is called " the ecological fallacy."' 2 *