METHODOLOGICAL ISSUES IN WEIGHT CYCLING

Recent studies have suggested that weight changes may be related to di sease risk independent of weight status. A critical step in testing th is assertion is the measurement of weight change and so-called ''weigh t cycling.'' However intuitive the concept of weight cycling may appea r, research in this area is hampered by complex methodological issues. This article discusses various measures of nominal weight cycling, in cluding the standard deviation, coefficient of variation, regression t echniques, and cycles. A cycle is a sequence of a gain followed by a l oss or vice versa. The various measures are compared in seven hypothet ical cases created to illustrate their strengths and weaknesses. Super ior performance of the cycles measure over the coefficient of variatio n, number of fluctuations, and simple regression methods is argued. Th e linkage of the cycles measure with the statistical theory of runs al so provides a basis for testing the significance of weight fluctuation s or other variables that may cycle, such as blood lipids, etc. The cy cles measure and runs test provide a viable definition for identifying weight cycling and a tool for evaluating the critical amount of weigh t gained and/or lost in relationship to risk.