In England, 1-year relative survival improved dramatically for all cancers analysed, with the exception of cervical cancer, between 1991 and 2006. The lack of improvement in cervical cancer survival has been found in previous studies and may be attributable to stage migration following the introduction of effective screening.18 19
Since 1998, cervical screening has reached at least 85% of eligible women in England. Near-universal screening means that many more cancers are detected at an in situ stage, and are removed before they become invasive. This has led to a dramatic decline in the incidence of invasive cancers.20–22
There is mounting evidence from other countries with near-universal screening programmes that the invasive cancers that are now identified tend to occur in women who have not attended screening, or to be particularly aggressive cancers with poor prognosis.22–26
Given the improvement in multidisciplinary team working and chemotherapeutic advances for cervical cancer since the 1990s it is more likely that screening rather than treatment explains the plateau in survival for this cancer.19
The results of this study show clear geographical inequality in survival. Cancer networks that had consistently low outlying survival estimates compared with the England estimates were clustered across northern England and the East Midlands, while results for southern and western cancer networks were more consistently comparable with, or above, the England estimates. This broadly confirms the findings of previous studies that have pointed to the north–south divide in cancer survival in England.5 6 27 28
The largest increase in the proportion of cancer networks that were low outliers occurred in the period 2001–6, following the NHS National Cancer Plan. This increase was observed in men only. Three of the four cancers in men were poor prognosis cancers, compared with three of the six in women, and it is possible that this played a role in the sex difference observed here. To explain why the proportion of low outliers increased for men but not women, it is also worth noting that women had a higher degree of variation at baseline for at least two cancers: the IQR for cancers of the oesophagus and colon was considerably higher in women ().
The lack of a significant reduction in geographical inequalities in cancer survival following the NHS National Cancer Plan echoes previous research showing there has been little or no reduction in the deprivation gap in short-term survival by socioeconomic status in England in this period.29
Given the large rise in overall survival, however, it is notable that the geographical inequalities have not widened more significantly, as shown by the rather stable proportion of cancer networks that were low survival outliers over time. Furthermore, even though the number of lower outliers remained stable over time, the distribution of these outliers has shifted, leading to a less-pronounced north–south divide in survival.
Although regional disparities in survival from most of these cancers did not fall over time, breast cancer is the exception, with a distinct convergence of the cancer network-specific survival estimates around the England value in the later period. The reduced variation can partly be explained by the fact that 1-year survival from breast cancer in 2001–6 is reaching the ceiling of 100%. Even accounting for this, it is clear that variation reduced, as can be seen by the relatively small number of outliers outside the (very narrow) control limits on the funnel plots in the final period. This is an important finding given that the improving outcomes guidance (IOG) issued in the mid-1990s was implemented earliest for this cancer. Multidisciplinary team working was universally established for breast cancer by the time it was assessed by the Audit Commission in 2001. At that time, implementation of the IOG for lung, bowel and the gynaecological cancers was uneven across the country.30
If this reasoning is correct, it may be hypothesised that the reduction in geographical inequalities for breast cancer may soon be followed by a similar convergence in regional survival for other cancers, provided the guidance issued in the manual for cancer services 31
(which has now replaced the IOG) is implemented with the same dedication and with equivalent resources as for breast cancer. We may even predict that regional disparities in survival from bowel cancer will similarly decline in the near future, since the National Cancer Peer Review of 2004–7 established that multidisciplinary teams in this sector had caught up with the breast cancer standard.32
A further issue to consider is the impact of screening on regional variation in survival through earlier diagnosis. Three of the six cancers included in this analysis were the subject of screening programmes. National programmes of cervical and breast cancer screening were effective throughout this study period, and two rounds of the colorectal screening pilot were conducted in 2000 and 2003–5 in some primary care trusts (PCT) in the Arden cancer network (N12).
Given that the proportion of eligible women attending screening for cervical cancer has remained stable since the late 1990s at over 80% in nearly 90% of health authorities,21
it is unlikely that the screening programme was responsible for the reduction in geographical inequalities in survival during 1991–2006.
Breast cancer screening expanded during the period of study in terms of the number of eligible women reached and the definition of eligible women. By 2006, coverage was over 70% of eligible women in 90% of PCTs and screening was contributing to the reduction of breast cancer mortality.33 34
However, less than one third of invasive breast cancers are screen detected.35
It is therefore unlikely that expansion of the screening programme played an important role in rising regional equality in breast cancer survival.
The colorectal screening pilot was rolled out in several PCT in the Coventry and Warwickshire area in 2000 and 2005–7. These PCT became part of the Arden cancer network when it was formed in 2002.36
Survival from cancer of the colon did improve very slightly in Arden relative to other cancer networks between 1996–2000 and 2001–6, but so did survival for patients with lung cancer and in women with oesophageal and cervical cancer. It is too early to say whether the launch of the national bowel cancer screening programme in 2007 will reduce geographical inequalities in survival.37
It is unlikely that variation in the quality of registration data has contributed to the regional inequalities identified in cancer survival, at least since the mid-1990s. Registration quality in England is determined by cancer registries (of which there are eight) rather than cancer networks. The annual reports of the United Kingdom Association of Cancer Registries show that quality is relatively homogenous between the registries today (with the exception of bladder cancer, see below).38
The proportion of eligible patients excluded from analysis through the quality control conducted for this analysis was largely driven by the number of DCO registrations and patients with zero survival (making up 80.7% of all exclusions). We showed in that this proportion was indeed correlated with survival in 1991–5, but not in the later calendar periods, and that regional diversity in the proportion of DCO has decreased.
The cancer sites included in this study were those incorporated for survival analysis in the annual compendium of clinical and health indicators for public health monitoring in England.39
Bladder and prostate cancer are also included in those indicators but were excluded here. Results for bladder cancer were excluded because wide changes in pathological coding, and diversity between cancer registries in the implementation of those changes, have led to non-interpretable trends in survival. Age-standardised survival from prostate cancer proved to be very difficult to estimate because of small numbers of deaths in some age groups at the cancer network level, so results were not reported.
We acknowledge that presenting data grouped by calendar period disguises fluctuations in trends and may not accurately portray the timing of change; however, the small sample sizes in the analysis by cancer network precluded the estimation of annual relative survival. Further methodological developments are in progress to enable age standardisation for small populations, in order that survival estimates can be produced for sparse data.5
What is already known on this subject
There is a north-south divide in cancer survival in England, with lower survival in the north. Policy changes under the Labour Government, including the NHS Cancer Plan (2000), aimed to reduce geographic inequality in survival. It is necessary to assess whether these policy changes have reduced geographic inequality in survival in England. This study uses the most suitable unit for analysis: cancer networks, which were formed to oversee the process.
What this study adds
There has been a lessening of the north-south divide in cancer survival but the overall level of geographic inequality has remained stable despite policy change. Breast cancer is the exception, showing a reduction in geographic variation in survival. If guidance is implemented with the same consistency for other cancers, geographic inequality in cancer survival may similarly decline.