PSA screening: correlating noise with noise?
============================================

* Maurice McGregor

Linda Perron and colleagues1 conclude that the fall in prostate cancer mortality observed since 1995 is not the result of increased screening activity. They base this on a comparison of the size of the change in prostate cancer incidence rates in 1993 with the size of the change in prostate cancer mortality rates in 1999. Although there clearly is a relation between the overall provincial numbers for the 2 (see Figure 1 in their article), when examined within 2-year birth cohorts, or within regions, none can be found. Accordingly, they conclude that the increase in diagnosis could not have led to the subsequent fall in mortality. 

However, examination of their figures suggests that (presumably due to the small numbers in each cohort or region) the differences in incidence and mortality are more likely due to noise than to any biologically meaningful phenomenon. To test this possibility, they could look for any relation between the incidence, or mortality, in any relevant cohort or region in the preceding or subsequent year. In its absence, any variation observed can be presumed to be noise, and because noise is unlikely to correlate with noise, any conclusion based on the absence of correlation is unjustified. 

**Maurice McGregor** Department of Medicine McGill University Health Centre Montreal, Que.

## Reference

1.  1. Perron L, Moore L, Bairati I, Bernard PM, Meyer F. PSA screening and prostate cancer mortality. CMAJ 2002;166(5):586-91.
    
    [Abstract/FREE Full Text](http://www.cmaj.ca/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NDoiY21haiI7czo1OiJyZXNpZCI7czo5OiIxNjYvNS81ODYiO3M6NDoiYXRvbSI7czoyMjoiL2NtYWovMTY3LzQvMzQwLjEuYXRvbSI7fXM6ODoiZnJhZ21lbnQiO3M6MDoiIjt9)