As has been reported previously [2, 3], organised cervical screening from 1991 was apparently effective in increasing the rate of fall in cervical cancer incidence and mortality in NSW women. We observed this pattern in both incidence and mortality in women born in Australia, New Zealand, the United Kingdom and Ireland, Asia and the Middle East and North Africa. For women born in Europe (other than the United Kingdom and Ireland) and the ‘Rest of the World’, however, there was no evidence of an increase in the rate of fall in incidence after 1991 and for women born in the Rest of the World mortality from cervical cancer may have increased after 1991.
While we did not attempt to relate the observed downtrends up to 1991 to change in cervical screening during this period, opportunistic screening and better treatment probably increased following the introduction of universal health insurance in 1975 with reimbursement of the cost of cervical cytology and access free-of-charge to treatment of cervical cancer precursors and cervical cancer itself. Changes in sexual behaviour might also have influenced these and the later trends. There is evidence from studies both in and outside Australia that mortality from cervical cancer in young women was increasing just before or early in the study period [24–27] due, perhaps, to the changes in sexual mores that occurred during and after the Second World War. These changes might have kept cervical cancer incidence and mortality rates higher in the first part of the period than they would otherwise have been. It is possible, also, that increasing condom use after the beginning of the epidemic of HIV infection during the 1980s  could have contributed to the downtrends in the 1990s. In addition, it is possible that the observed trends may have been influenced by differential exposure over time and between various groups of migrant women to the established co-factors of human papillomavirus (HPV) infection (parity, age at first full term pregnancy, use of tobacco and oral contraceptives) The very small fall in incidence and the increase in mortality from cervical cancer in women from the Rest of the World after 1991 may reflect recent changes in patterns of migration to Australia. Many women classified as from the ‘Rest of the World’ have come from African and Latin America countries where cervical cancer is very common . The numbers of migrants from these areas have increased in the last two decades [29, 30]. Thus the apparently less favourable trend in incidence and the unfavourable trend in mortality in this migrant group may reflect the high cervical cancer incidence and mortality of their home region, which would be expected to persist for some time after coming to Australia.
No previous study has examined trends in cervical cancer incidence and mortality in migrant women in Australia. While a number of international studies in diverse populations have documented disparities in cervical cancer incidence and mortality between migrant women and native-born women or women of different ethnic backgrounds [6–12], only a few have examined time trends in cervical cancer incidence or mortality by ethnic background or country of birth.
A New Zealand study  examined trends in cancer incidence from 1981–1986 to 2002–2004 in four groups defined by ethnicity or country of birth: Maori, Pacific Islander, Asian and European or Other. The investigators found that, compared to European or Other women, Maori, Pacific Islander, and Asian women had higher incidence rates of cervical cancer in 2002–2004. The age standardised incidence rate ratio (SRR) comparing Maori and Pacific Islander women with European/Other women fell with time, which suggests a greater rate of fall in Maori and Pacific Islander than European/Other women. However, the SRR for Asian women increased with time. While New Zealand introduced organised cervical screening in 1991, this report did not compare trends before and after this date.
McDougall et al.  examined trends in incidence of cervical cancer by ethnicity in the US between 1992 and 2003 based on information from the 13 cancer registries. They found similar falls in incidence of cervical cancer overall and in squamous cell carcinoma in four different ethnic groups: Non-Hispanic whites, Hispanic whites, African-American, and Asian or Pacific Islander, over the period of study. However falls were more pronounced among Asian or Pacific Islanders. A second US study , examined cervical cancer incidence trends in four categories: Hispanic/all races, Non-Hispanic/white, non-Hispanic/black, non-Hispanic/other using a dataset from 22 state cancer registries. Incidence of cervical cancer was significantly less in all four race/ethnic groups in 2000–2004 than 1995–1999 (rate ratio 0.83, 95% CI 0.82-0.84), with standardized rate ratios ranging from 0.75 (95% CI 0.70-0.79) for Non-Hispanic/other to 0.84 (95% CI 0.82-0.85) for non-Hispanic/white. A third US study examined the incidence and mortality of cervical cancer among Asian and non-Hispanic white women in California in 1990 to 2004. Cervical cancer incidence and mortality fell in each group during this period. The APCs in the rates were −8.7% for Vietnamese, -5.1% among Koreans, -4.6% among Filipinos, -5.4% for Chinese and −2.3% among non-Hispanic Whites . None of these studies related the trends in incidence or mortality with trends in cervical screening, which is largely opportunistic in the USA.
When examined in broad age categories, there was less evidence in older than younger NSW women that mortality fell faster after 1991 than it did up to 1991. This age group difference could be due to less screening [32–34] or less effective screening in older women  after 1991, although the first should also affect incidence trends. We have no reason to think that older women treated for cervical pre-neoplasia or cancer would have received poorer treatment after 1991 than before. However, it is important to note that the optimal interval for screening is longer in women over 50 years  when compared to younger women, and it has previously been proposed that apparently greater effects of opportunistic screening in older women could be partly due to the greater efficacy of irregular screening in older than younger women .
Our findings are based on 36 years of data from a high quality cancer registry covering a large population. The period for which cancer registry data was available included 18 years before the year of introduction of organised cervical screening program (1991) and 17 years after; thus trends before and after introduction of an organised approach could be modelled with considerable precision in a range of groups of migrant women.
We analysed trends in cervical cancer rates in relation to country of birth, but we did not have individual-level data on the age of migration, and therefore could not explicitly account for prior screening experience in the country of origin. However, cytological screening in women aged 20–24 years has been shown to have little or no impact on rates of invasive cervical cancer up to age 30 years , and therefore any effects of pre-migration screening for women who migrated as children or young adults are expected to be very limited. Even for women who migrated at ages older than 25 years, the results of a major audit and case–control study in the UK (a study which underpins IARC’s 2005 recommendations for the cervical screening interval) , found that the relative risk of invasive cancer in screened women dropped to the same level as that of unscreened women after 3.5 years in women younger than 50 years, and after 5–6 years in women over 50 years of age . In effect, the cytological screening effect ‘wears off’ relatively rapidly which is why frequent repeated screening is required with cytology. Therefore, it is unlikely that screening history before migration would have a major impact on the trends observed in this study.
We were unable to adjust for hysterectomy over the whole analysis period because hysterectomy frequencies by calendar year, age and region of birth were only collected by the NSW Admitted Patient Data Collection from 1991 onwards. However, we conducted sensitivity analyses using the NSW Admitted Patient Data to correct the populations at risk for hysterectomies and found very similar results to those in the main analysis. In addition, the age adjusted hysterectomy incidence rates in women 20-85+ years of age, estimated from the NSW Admitted Patient Data Collection , were stable from 1991 to 1997 and then fell steadily in all country of birth groups from 5.2/1,000 (Australian born) to 3.0/1,000 (Asian born) in1997 to a half to two-thirds of the 1997 values (3.1/1,000 in Australian born to 1.7/1,000 in Asian born) in 2008 (data not shown). The effect of these trends would be to reduce the observed rate of fall in cervical cancer rates in all country of birth groups in the later part of the period following introduction of organised screening. This effect, though, would be greater in older than younger women and thus might comprise part of the explanation of the less evident increase in the downtrends in cervical cancer incidence and mortality in older women after 1991 (Figures 1b-d).
We were, in addition, not able address possible confounding of trends by country of birth by, for example, trends by socioeconomic status and area of residence. Information on socioeconomic status based on the Australian Bureau of Statistics Index of Relative Socioeconomic Disadvantage (IRSD) was not available from the CCR for years before 1980 and classification of area of residence based on Accessibility and Remoteness Index for Areas (ARIA) was only available from 2000. Neither of these, however, would be expected to have had much of a confounding effect. There were only small differences in the distributions of IRSD between Australian-born and migrant women studied from 1980 and the vast bulk of both Australian born and migrant women studied from 2000 resided in major city or inner regional areas (89.8% and 98.0% respectively).
Finally, it should be noted that rates of invasive cervical cancer incidence and mortality in Australia nationally appear to have stabilised since about 2002 ; reductions prior to that time have been predominately driven by declines in invasive squamous cervical cancer, whereas the incidence of adenocarcinoma (glandular cancers) appears not to have been substantially impacted by cytological-based screening. This stabilisation effect may have resulted in our calculated average annual percent changes from 1991 to 2008 being slightly lower than if rates had not stabilised over the last 5 years of the period included in our assessment. However, in this study we have taken a broad approach to assessing the overall changes in rates after the period of interest, and focused mainly on the differentials between various groups of migrant women. We have identified some groups of migrant women from certain European and other countries that may be able to benefit, to a greater degree, from the organised cervical screening program in Australia. In the context of a current review of cervical screening recommendations and technology in Australia , further reductions in invasive cervical cancer incidence and mortality rates could also potentially be achieved in the population overall by switching to a primary screening test which is potentially more effective in detecting adenocarcinoma, such as primary HPV DNA-based screening. In the longer term, the National HPV Vaccination Program, introduced in Australia in 1997, also has the potential to further reduce rates of cervical cancer in all Australian women .