Incorrect Analyses of Radiation and Mesothelioma

The recent article, “Analyses of Radiation and Mesothelioma in the US Transuranium and Uranium Registries,” by Gibb et al.¹ examines seven mesothelioma deaths among a small population of 329 deceased registrants in the US Transuranium and Uranium Registries (USTUR). Using the National Institute for Occupational Safety and Health’s Life Table Analysis System with the referent group being all deaths in the United States,^2,3 the study finds a proportionate mortality ratio (PMR) of 62.40 (P < .05) for mesothelioma, the highest PMR ever observed and more than an order of magnitude higher than any other published studies.¹

Mesothelioma has been coded as a separate underlying cause of death (C45.0-C45.9) since 1999 when International Classification of Diseases, 10th revision (ICD-10)⁴ was implemented. Before 1999, mesothelioma was coded as malignant neoplasm of pleura, unspecified (ICD-9:163.9); malignant neoplasm of bronchus and lung, unspecified (ICD-9:162.9); malignant neoplasm without specification of site (ICD-9:199); and so on. Life Table Analysis System has no death data in the mesothelioma disease category for the United States before 1999^2,5 and therefore does not support PMR analysis for mesothelioma deaths before 1999. More than 80% of USTUR deaths occurred before 1999 including six of seven USTUR mesothelioma cases.^6,7

Life Table Analysis System calculates PMR as a ratio of weighted sums of the proportion of deaths from a specific cause in the exposed versus the comparable weighted sum in the unexposed (often the US population deaths). Adjustment for age, race, gender, and calendar time is accomplished by stratification and indirect standardization.^2,3 The formula to calculate the PMR is presented below.

where the variables are defined as follows:

• P_i1= the i^th stratum-specific proportion in the observed cohort (the exposed cohort)

• P_i0 = the i^th stratum-specific proportion in the reference population (unexposed population)

• W_i = the i^th stratum-specific number of observed deaths in the exposed cohort.

Because all strata before 1999 have zero P_i0 and only 20% (one fifth) of USTUR deaths contribute the weighted sum of the denominator in the formula, the PMR for mesothelioma by Gibb et al. is greatly overestimated. If one assumes that the age distributions of deceased registrants (who were overwhelmingly White and male) are similar before and after 1999, and the stratum-specific proportions for mesothelioma in the reference populations are the same before and after 1999, the PMR for mesothelioma is a five-fold overestimate.

The USTUR death data are not compatible and therefore should have not been compared with the Life Table Analysis System US death data in the PMR analysis for mesothelioma because there is no specific code for it before 1999. The analyses were conducted incorrectly from the beginning, resulting in an artificially high–reported PMR for mesothelioma. Caution should be exercised when using Life Table Analysis System or similar analytic software for mortality studies in which the rules for coding cause of death are different over the time frame of a study.