From the author:
Thank you for the opportunity to respond to the thoughtful letter from J.G. Smith [1].
The main issues she raises stem from the original paper having aims other than use for prediction in current individual or legal cases. Our main aim was to describe the epidemiology of malignant mesothelioma in Western Australia and to demonstrate how little improvement there has been in survival over the last 30 yrs, despite large changes in treatment [2]. Thus, we fitted survival models to all available data. The complicated age terms were just the best-fitting fractional polynomial fit to these data. Further analyses revealed that the u-shaped survival with age was caused by the comparatively poor survival in patients diagnosed under the age of 40 yrs, such that when these were excluded from analysis, a linear decline in age was the best fit. We also included all patients with an unknown site of mesothelioma, leading to the apparently anomalous better survival in this group.
We suggest that if a survival model is to be developed and used for individual prediction (possibly in legal cases), we would need to exclude: patients who were diagnosed at death; possibly all those with unrecorded histology; all those with unspecified site of disease; and possibly all those diagnosed before 1980 (or maybe 1990 or later).
To that end, we have spoken with J.G. Smith and suggested that we work together with the necessary subset of the data (which may soon include an additional year of accrued cases) in order to produce a revised publishable model that would have much greater utility in the context for which she wishes to use it.
Footnotes
Statement of Interest
None declared.
- ©ERS 2012