Extended Methods section

Extra Table
 

Extended Methods section

A detailed description of the data collection, exclusion criteria, and analytical methods for the estimation of trends in cancer survival has been published.3 Briefly, data for 2.9 million adults aged 15-99 with cancer diagnosed at some point during the 20 year period 1971-90 were derived from population based registries covering the whole of England and Wales, and updated from the NHS Central Register with details of deaths occurring up to 31 December 1995. The analyses covered 47 different cancers, which account for 93% of all diagnoses of invasive malignant disease (excluding non-melanoma skin cancer). In situ malignancies were also excluded. Patients were categorised by type of cancer, sex, age at diagnosis (15-39, 40-49, 50-59, 60-69, 70-79, and 80-99 years), and deprivation category of their census enumeration district of residence at diagnosis (quintiles of the Carstairs score4 for Great Britain around the 1981 and 1991 censuses). Relative survival up to five years after diagnosis was estimated for patients categorised by time of diagnosis (1971-5, 1976-80, 1981-5, and 1986-90) by means of a STATA algorithm based on methodology developed by Estève et al.5 Specially constructed life tables for each sex, single year of age at death, and deprivation categoryw1 were used to adjust for differences in background mortality at each age between deprivation groups and over time. Relative survival rates (the ratio of the survival actually observed among cancer patients and the survival that would be expected if they had the same death rates as the general population) were used as the basic outcome measure. A relative survival rate of 100% would imply that the cancer patients had the same survival rate as the general population from which they are drawn w2; lower survival rates reflect the excess mortality among cancer patients.

We carried out further analyses of these results to estimate how many cancer deaths were avoided within five years of diagnosis among patients with cancer diagnosed during 1986-90 as a result of their survival rates being higher than those of patients who had cancer diagnosed during 1981-5. We used relative survival rates for each cancer by age, sex, calendar period and (for common cancers) deprivation category.

We defined the observed excess mortality for patients with cancer diagnosed during 1986-90 as the number of deaths observed within five years of diagnosis over and above the number of deaths that would have occurred if the patients had experienced the same mortality as that of the corresponding sex and age group of the general population during the same calendar period. This is estimated by multiplying the complement of the relative survival rate for each category of sex and age by the corresponding number of patients with cancer diagnosed during 1986-90.3 The excess mortality for each cancer is then the sum of these estimates across all subgroups.

The expected excess cancer mortality among patients with cancer diagnosed during 1986-90 is estimated in similar fashion, but using the relative survival rate for patients diagnosed in 1981-5. This represents the excess deaths that would have been expected among patients diagnosed during 1986-90 if there had been no improvement in relative survival over the previous five years.

The difference between the observed and expected excess mortality for each cancer represents the number of deaths that were avoided within five years of diagnosis among patients diagnosed during 1986-90 as a result of improvements in survival, adjusted for age and sex. Avoided deaths are reported both as the number of deaths and as a percentage reduction in the excess deaths that would have been expected if survival had not improved. Avoided deaths were calculated separately for each cancer: results are presented for selected cancers, for all other cancers combined, and for all cancers combined.

w1 Coleman MP, Babb P, Mayer D, Quinn MJ, Sloggett A. Cancer survival trends in England and Wales 1971-1995: deprivation and NHS region (CDROM). London: Office for National Statistics, 1999.

w2 Estève J, Benhamou E, Raymond L. Statistical methods in cancer research. Vol IV. Descriptive epidemiology. Lyon: International Agency for Research on Cancer, 1994. (IARC Scientific Publications No 128.)

Deaths from selected cancers within five years of diagnosis among adults in England and Wales with cancer diagnosed during 1986-90 that were avoided due to improved survival compared with patients with cancer diagnosed in 1981-5. Values are numbers (percentages)*
 

Extra Table
 

*Percentage of expected number of cancer deaths, shown as zero if <0.5%.

†Breast cancer in men and 34 other cancers.