Objective: To examine the influences of recall period, current body weight, weight gain and loss, and weight variability on the accuracy of long-term recall of past weight.
Design and subjects: Comparison of recalled weights around 25 y old with weights measured at age 25y in health checkup records among 2453 Japanese men (age: 34-61 y).
Measurements: Deviations between reported and measured weights were examined as to the three indexes: actual error (reported-measured), percent error (actual error/measured x 100), and absolute value of the percent error. Weight variability was defined as (1) the coefficient of variation of weight (CV) and (2) the root mean square error around the slope of weight on age (RMSE).
Results: Recalled weight strongly correlated with measured weight (r=0.849). Correlation coefficients decreased as age or the elapsed time since age 25y increased. Recalled weight (mean=58.3 kg) was slightly greater than measured weight at age 25y (mean=57.0 kg, mean actual error=1.28 kg). Subjects with a current body mass index (BMI) of less than 28.6 kg/m(2) overestimated their past weights, whereas those with BM1 of 28.6 kg/m(2) or over underestimated it. Subjects with weight loss since age 25y underestimated their past weights, whereas those with stable weight or gain overestimated it. There were monotonic increases in the three indexes of deviation across the CV quartile categories. Concerning the effect RMSE, a similar trend was observed.
Conclusion: These results indicate that past body weights over a long period seem to be recalled with good accuracy. However, it should be kept in mind when using recalled weight in an epidemiologic study that accuracy of recall is influenced by age or elapsed time, current BMI, weight gain and loss, and weight variability.