Cervical cancer is the third most common cancer in women worldwide, after cancers of the breast and large bowel, with approximately 470 000 new cases a year in 2000, some 10% of all cancers in women (Ferlay et al, 2001). Approximately 80% of all cases occur in developing countries, where cervical cancer is often the most common cancer in women, whereas in developed countries, cervical cancer only accounts for about 5% of all cancers in women. It has been known for 30 years from observational data that cervical cancer is likely to have an infectious component (Beral, 1974), subsequently confirmed as HPV (Bosch et al, 1994). Trials of vaccination already appear promising for primary prevention (Gravitt and Shah, 2005; Mao et al, 2006). In England and Wales, the age-standardised incidence of cervical cancer in the 1970s and 1980s remained between 14 and 16 per 100 000, although age-specific rates were influenced by strong cohort effects (those women born in the 1890s were at high risk, those born in the mid-1930s at low risk). The cervical screening programme began in the 1960s, offering cervical smear tests to detect premalignant lesions in women aged 25–64 years on a 5-yearly basis, but it was largely ineffective until the late 1980s (Murphy et al, 1988). In 1988, a national call and re-call system was established and coverage rose rapidly to approximately 85% (Quinn et al, 1999). As a result of these improvements, the incidence of invasive malignancy fell continuously after 1990 to below 9 per 100 000 in 2000 (Quinn et al, 2001; Office for National Statistics, 2003a). Mortality from cervical cancer in England and Wales had been declining steadily at approximately 1.5% each year from the early 1950s to the late 1980s. This long-term decline began before the introduction of screening. From 1990, the rate of decline in mortality increased three-fold, and by 2002 the age-standardised death rate had fallen to 29 per 100 000, barely one-third of the death rate in 1971 (83 per 100 000) (Office for National Statistics, 1999, 2003b). Only 1000 deaths were attributed to cervical cancer in 2002, and the screening programme has been credited with preventing 800 or more deaths a year (Quinn et al, 1999; Peto et al, 2004). We analysed the data for 44 090 women diagnosed with invasive cancer of the uterine cervix in England and Wales during the 14-year period 1986–1999, some 90% of those eligible for inclusion in the analyses. Approximately 4% of women otherwise eligible for analysis were excluded with zero recorded survival (date of diagnosis same as date of death). The proportion of women for whom recorded survival was zero did not vary between socioeconomic groups, however, and it was stable throughout the 1990s (data not shown). Therefore, these exclusions are unlikely to have had any substantial impact on socioeconomic gradients in survival, or on trends in the gradient. A further 3% of women with cervical cancer were excluded because it was not their first primary malignancy, and 2% because their vital status was unknown on 5 November 2002, the date when the data were extracted for analysis. Some two-thirds (67%) of cervical cancers were squamous carcinomas, and a further 16% were adenocarcinomas: both proportions were similar throughout the 1990s. Other morphological types were individually rare, but cancers with poorly specified morphology accounted for about 10%. Annual incidence rates for invasive cervical cancer in the late 1990s were 50% higher among the most deprived women (12.8 per 100 000) than among the most affluent (8.5 per 100 000), but this difference was less marked than in the early 1990s (63%). This is because incidence declined more among the most deprived women over this period (18%) than among the most affluent (11%) (Figure 1). Figure 1 Trends in the age-standardised incidence of cervical cancer in women aged 15–99 years, by deprivation group: England and Wales, 1986–1999.