Occupational Mortality in Relation to Exposure

Methods of Cohort Analysis: Appraisal by Application to Asbestos Mining

“…We do not share Dr Jacobsen's enthusiasm for the Comparative Mortality Figure. Although a ratio of CMFs is indeed a ratio of weighted averages of deaths rates, the coefficient of variation of the CMF cannot be less than that of the corresponding SMR (Liddell, 1960), and we are not aware of any evidence that the covariance between numerator and denominator of a ratio of CMFs is less or greater than that for a ratio of SMRs. In much of our experience, SMR and CMF have been sufficiently close to make them interchangeable in practice.…”

“…These desirable properties hold for the ratio of any two directly standardized summary statistics using the same standard. They do not apply to internal-analysis analogues of the SMR, such as that used in Table 4; to the alternative index suggested by Liddell (1960); or to the 'inverse SMR' discussed by Kerridge. (Kerridge did not recommend the method when the age distribution of the risk set is known, as of course would be the case in any prospective study.)…”

“…We remain of the view (Liddell, 1975) that the evaluation of hypotheses can best be carried out by a posteriori methods. The two such methods we have examined produced closely similar results and both can be used for hypothesis evaluation.…”

“…We thank our colleagues-particularly Professor Alison D. McDonald and Dr Graham W. Gibbs-for allowing us to use the Quebec mortality data for this paper. We are particularly grateful to a small group of experts who reviewed methods with us during a two-day meeting at McGill University in June 1974; we have drawn frequently on their report briefly summarized by Liddell (1975). Most of the extensive computing required was ably performed by Mr Mario Rodrigues.…”

“…Once observed deaths and expected deaths have been obtained, it has been common to sum them over all eras and ages at death, and to form a ratio, which can be seen to be the Standardized Mortality Ratio (SMR), with all the disadvantages of standardization, and particularly those associated with the "indirect" method (Liddell, 1960). On theoretical grounds, there is some justification for standardizing by the "inverse" method (Kerridge, 1958), but the SMR is the more easily manipulated.…”