Statistics from Altmetric.com
Battersby et al present a method of performing equity audit where data on incidence, deprivation, and surgical resection rates of non-small cell lung cancer are compared.1 Deprivation was measured using the Index of Multiple Deprivation (IMD) 20002 and all analyses were performed at the primary care trust (PCT) level. Battersby et al report no statistically significant associations between their measure of deprivation and age and sex standardised incidence of non-small cell lung cancer. This is highly unusual and in contrast with findings from a large number of different populations.3–9 Without clear evidence that there is something exceptional about the population studied by Battersby et al, the lack of association between deprivation and incidence of lung cancer is likely to be an artefact.
Two possible explanations of Battersby et al’s failure to find an association between deprivation and incidence of lung cancer are possible. Firstly, calculating deprivation at the PCT level may be highly inaccurate. The IMD 2000 is a ward level variable and the authors of this study use a weighted average to determine IMD 2000 scores for PCTs. However, with an average of almost 22 wards per PCT in England and an average PCT population of over 128 000 in the study, it is not clear that “the use of a weighted average IMD score is an appropriate way of including deprivation in the analysis”.1 Secondly, with only 17 data points, it is probable that Battersby et al’s study lacks sufficient power to detect any but the strongest of associations.
Although the focus of Battersby et al’s article is primarily on a methodological technique, rather than on the particulars of the example data used, the failure to detect an association between deprivation and incidence of lung cancer suggests some problem with this technique. Furthermore, presentation of these results to non-specialised audiences, as is intended, may give the impression that deprivation is not an important determinant of incidence of lung cancer—which the majority of other evidence suggests it is.
Although public health practice requires methods that are quick, easy, and cheap to perform, this should not be at the expense of accuracy to such a degree that important, strong, and known associations are overlooked.
Dr Adams is correct in her assertion that the lack of statistically significant association between the Index of Multiple Deprivation 2000 (IMD 2000) score and the incidence of non-small cell lung cancer is unusual. However, we would dispute the suggestion that calculating deprivation at primary care trust (PCT) level is inaccurate.
We suggest that the use of population weighted average scores is an appropriate method of calculating PCT deprivation scores. The IMD is a proportion and, therefore, the construction of an IMD score for aggregated populations requires population weighting. This is an accepted methodology and is widely applied.1
We did report that crude incidence of non-small lung cancer correlated with the IMD 2000 score. Moreover, when looking at mortality from all lung cancers during a similar period, there was good correlation between IMD 2000 score and age specific death rates for lung cancer, both for men r = 0.76 (95% CI 0.44 to 0.91), and women r = 0.59 (95% CI 0.15 to 0.83). It was only the relation between IMD 2000 score and the age-sex standardised incidence of non-small cell lung cancer that, although present, did not reach statistical significance.
It seems more probable that this lack of statistically significant association can be explained by the fact that the range of deprivation scores across PCTs in the east of England is comparatively narrow. IMD 2000 population weighted average scores vary from 7.1 to 38.0 across PCTs in Norfolk, Suffolk, and Cambridgeshire. This compares with a much wider range of 4.4 to 61.3 across local authorities in England.2
A further weakening of the association may also be attributable to the exclusion of small cell lung cancers (about 13% of all lung cancers), which are almost exclusively attributable to smoking. However, this effect is likely to be small.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.