A simulation model for small-area cancer incidence rates

Cancer epidemiologists are often asked by members of the interested public about possible associations between suspected carcinogens and apparently increased small-area cancer incidence rates. Frequently, no systematic incidence differences can be demonstrated. Nevertheless, it is necessary to address public concerns about suspected cancer clusters. To facilitate explanations about the large random variation of small-area tumor incidence, we implemented a software simulation tool in R.

Under the assumption of no cancer causes other than chance, the tool simulates a small village population with an average number of five inhabitants per house and allows graphical visualisation of ten streets with 100 houses. Published age-specific incidence and mortality data are used for event sampling based on the binomial distribution. Program parameters include sample size, age distribution, cancer incidence, and mortality rates.

On average, 22 percent (2.2/10) of all houses per street have been inhabited by at least one cancer patient during the last five years in our simulated small village. A situation where all (10) houses in a street have been inhabited by at least one cancer patient during the last five years appears to be very rare (less than one in a million streets).

Our software tool can be used effectively for numerical and graphical visualisation of small-area tumour incidence and prevalence rates due to chance alone. The explanation of basic epidemiological concepts to members of the public can help to increase public motivation and support for population-based cancer registration. Our simulation tool can be used to support this goal.