Incidence projections: Methodology

The cancer incidence projections project was undertaken by a team at the Wolfson Institute of Preventive Medicine, Queen Mary University of London, as part of a Cancer Research UK programme grant. This has resulted in the paper by Mistry et al (2011) in the British Journal of Cancer published in October 2011.1 Cancer incidence projections are useful for a range of stakeholders in order to estimate the cancer burden in the future.

The number of newly diagnosed cancers in Great Britain, broken down by cancer site, sex and five-year age-group for the years 1975 to 2007, were provided to the team at the Wolfson by the Statistical Information Team at Cancer Research UK, along with relevant population data to allow projections to 2030 for the UK to be calculated.

The cancer incidence projections do not take into account potential changes in lifestyle or treatments that could alter future cancer rates. They are based entirely on past incidence rates.

section updated 13/12/11

The Wolfson team developed a range of smoothed age-period-cohort models for each cancer site by sex to project age-specific incidence rates for five-year age-group for individual years from 2008 to 2030 based on the rates from 1975 to 2007 for Great Britain.

Age-period-cohort models take into account three main influences on the incidence rate over time:

• Age: as people get older the risk of cancer increases. As the population as a whole ages we would therefore expect more cases of cancer with no change in the underlying risk.
• Period: the period effect reflects a change in risk which happens over time, and so has similar effects on people of all ages– such as a general decrease in passive smoking as a result of the ban of smoking in public places.
• Cohort: a cohort is a group of people born at a similar time. The male cohort born in 1960s is much less likely to smoke than the male cohort born in the 1940s. The younger cohort is therefore less likely to get a smoking related cancer.

In practice there will be a relationship between age, cohort and period effects (and people born in the same cohort will be at the same age in the same period); and so it is important that all effects are included.

section updated 13/12/11

For most cancer sites: a range of smoothed age-period-cohort models were fitted for each sex based on the rates from 1975 to 2007.

For prostate and breast cancers, the wider use of PSA testing and the introduction of the NHS Breast Cancer Screening Programmes across the UK, respectively, have interrupted the time trend and so the researchers have used slightly different models.

For prostate cancer: the authors assumed that the increased use of PSA testing caused an increase in the number of prostate cancers. They modelled the observed underlying increase in prostate cancer rates from 1975 to 1991 before PSA testing was widespread, and compared this modelled increase in rates to the actual rates in 2004 to 2007 to get an inflation factor associated with PSA testing. The projections for 2008 to 2030 use the underlying 1975-1991 trends multiplied by the inflation factor for 2004-2007. Thus, it is based on the assumption that a similar level of PSA testing will be used in the future.

For breast cancer: the method used was similar to that used for prostate cancer and they used the incidence rates from before screening was offered to particular age-groups to project what would have happened to the rates going forward once the data had stabilised, which was different for different age-groups.1

For all cancers combined: to estimate the total numbers of cancers in 2030, 34 projections were made for 17 non sex-specific sites, four for female only cancers and two for male only cancers. The results of these projections were added to two further models, a male “other sites” and a female “other sites” (these were aggregates of all of the other smaller sites).

The numbers of cases in the UK have been calculated by applying the age-specific incidence rates in Great Britain to the projected UK populations by age and sex. This is reasonable because the Northern Ireland population is a small percentage of the UK population (approximately 3%).

Find out the projections results for all cancers combined and selected cancers.

section updated 13/12/11