Editor's Choice | This Week in BMJ | Press releases
BMJ No 7106 Volume 315 Letters Saturday 23 August 1997 Intersalt dataCorrection for regression dilution bias in Intersalt study was misleadingEditor,We were surprised by the response of Alan R Dyer and colleagues in their commentary on our article on correction for regression dilution bias in the Intersalt study.(1) They fail to address the substantive points adequately while restating evidence on the association between dietary salt and blood pressure, which was not the concern of our article. A random 8% of participants in the Intersalt study had repeat 24 hour urine collections and blood pressure measurements taken on average three weeks after the initial measurement. The repeat 24 hour urine collections were used to estimate the degree to which single 24 hour collections were inadequate indicators of usual sodium excretion. The reliability coefficient calculated from these repeat measures was used to greatly inflate the estimated strength of association between sodium excretion and blood pressure. However, if changes in urinary sodium excretion and blood pressure between the first and the repeat measurements tend to coincide then the association between sodium excretion and blood pressure would not have been underestimated to the degree to which the correction method assumes. We cited papers that show that changes in sodium excretion and blood pressure do indeed coincide, for measurements taken one to two months apart. As N E Day shows [letter above], if the correlations between changes in sodium excretion and changes in blood pressure are even of moderate strength then the correction method used by the Intersalt team is highly misleading. Dyer and colleagues state that 'to the best of our knowledge, there are no data showing physiological day to day parallel fluctuations of sodium or blood pressure.' They miss the point that fluctuations in sodium excretion and blood pressure do not need to coincide on a day to day basis to invalidate their method. If the fluctuations correspond over a three week period, which was the interval in their validation study, then this will render their correction method fallacious. What is most odd about their response is that the Intersalt investigators do indeed have the data that can test this point, as they measured both blood pressure and sodium excretion in their validation study. How can they then state that no such data exist? We ask that they report the correlations between the changes in sodium excretion and the changes in blood pressure over the three week period to allow other investigators to evaluate whether one of the central assumptions of their correction method is met. The Intersalt investigators seem not to appreciate the difficulties of correcting for measurement imprecision in the presence of confounders. Body mass index is only a proxy for the various aspects of body composition that confound the association between sodium excretion and blood pressure. The mathematical construct of weight divided by the square of height is clearly not a perfect measure of these. The appropriate reliability coefficient is that between body mass index and the relevant aspects of body composition, not between body mass index measured on two occasions. The importance of measurement error in confounders is, fortunately, well recognised by other investigators.(2) All of the conclusions of our paper remain unchallenged by the response by Dyer and colleagues; indeed, they are strengthened by the demonstration that a research team that uses these methods to essentially triple the magnitude of their association between sodium excretion and blood pressure does so without a clear appreciation of the implications of these correction methods. The use of epidemiology to inform public health policy would be furthered better by improving study design(3) than by using obfuscating and potentially erroneous statistical 'corrections,' the assumptions of which have not been adequately tested. George Davey Smith Professor of clinical epidemiology University of Bristol, Andrew N Phillips Reader in epidemiology and biostatistics University Department of Primary Care and Population Sciences, References 1 Davey Smith G, Phillips A N. Inflation in epidemiology: 'The proof and measurement of association between two things' revisited. [With commentary by A R Dyer et al.] BMJ 1996;312:1659-64. 2 Marshall J R, Hastrup J L. Mismeasurement and the resonance of strong confounders: uncorrelated errors. Am J Epidemiol 1996;143:1069-78. 3 Phillips A N, Davey Smith G. The design of prospective epidemiological studies: more subjects or better measurements? J Clin Epidemiol 1993;46:1203-11.
Home | Current issue | Past issues | Classified ads | Career Focus | Feedback Collections | About this site | About the BMJ | BMA | Medline
|