Random errors in the measurement of 10 cardiovascular risk factors.

Random errors in the measurement of 10 commonly investigated cardiovascular risk factors (systolic and diastolic blood pressure, blood cholesterol, blood glucose, pulse rate, body mass index (BMI), cigarette consumption, passive smoking, alcohol intake and physical exercise) were assessed in a general population cohort (n = 2517) and a workforce cohort (n = 8008). Random errors were estimated from regression dilution ratios (lower ratios imply greater random error, and a ratio of one implies no random error). All of the risk factors, except for BMI (which had regression dilution ratios of 0.93 and 0.98 in the two cohorts), were measured with substantial levels of random error. Particularly low regression dilution ratios were observed for physical exercise (0.28 and 0.39) and pulse rate (0.47 and 0.56). For each of these risk factors, with the possible exception of BMI, associations with long-term average values could be importantly biased toward the null unless appropriate corrections are made.