Inequalities in colorectal cancer screening participation in the first round of the national screening programme in England

Background: Introduction of organised, population-based, colorectal cancer screening in the United Kingdom using the faecal occult blood test (FOBT) has the potential to reduce overall colorectal cancer mortality. However, socio-economic variation in screening participation could exacerbate existing inequalities in mortality. Methods: This study examined FOBT uptake rates in London, England in relation to area-level socio-economic deprivation over the first 30 months of the programme during which 401 197 individuals were sent an FOBT kit. Uptake was defined as return of a completed test kit within 3 months. Area-level deprivation in each postcode sector was indexed with the Townsend Material Deprivation Index. Analyses controlled for area-level household mobility, ethnic diversity and poor health, each of which was associated with lower return rates. Results: The results showed a strong socio-economic gradient in FOBT uptake, which declined from 49% in the least deprived quintile of postcodes to 38% in the middle quintile and 32% in the most deprived quintile. Variation in socio-economic deprivation between sectors accounted for 62% of the variance in return rates, with little attenuation as a result of controlling for ethnic diversity, household mobility or health status. Conclusion: These results highlight the need to understand the causes of socio-economic gradients in screening participation and address barriers that could otherwise increase disparities in colorectal cancer survival.

Reducing health inequalities through equality of access to health care is a cornerstone of the UK Government's health policy, and has been re-emphasised in the Department of Health (2007). However, associations between socio-economic status (SES) and health have persisted through major changes in social conditions and provision of medical care, and it has been argued that innovations in health technology may widen inequalities if people with more knowledge, money or power are better able to harness the beneficial effects (Link et al, 1998;Baker and Middleton, 2003;Coleman et al, 2004).
The UK Breast and Cervical Cancer Screening Programmes were adopted more enthusiastically in more affluent areas (Gatrell et al, 1998;Webb et al, 2004;Dailey et al, 2007), and at least in the case of cervical cancer, this has had a measurable impact on disparities in mortality (Link et al, 1998). As part of the National Health Service (NHS), cancer screening in the United Kingdom incurs no direct financial costs, so affordability cannot explain variation in uptake. However, both breast and cervical screening require attendance for a clinic appointment, and therefore factors such as pressures of time, transport problems or discomfort in interacting with medical services and health professionals, could lie behind SES differences in participation.
Whether the same SES differential will be seen for the latest addition to the UK Cancer Screening Programmes, biennial bowel cancer screening using the faecal occult blood test (FOBT), is an important question. In the Bowel Cancer Screening Programme (BCSP), all 60 -69-year olds are sent an invitation to participate around a week before the test kit is mailed to the home address. The kit is accompanied by step-by-step instructions on how to complete the test at home and a hygienic freepost envelope is provided to return the samples to the laboratory for processing. Individuals are informed of the result and advised if further tests are needed (repeat FOBT or colonoscopy) within 2 weeks.
Faecal occult blood test has the potential to reduce colorectal cancer mortality by 16%  but SES differences in participation could increase existing differentials in survival (Mitry et al, 2008). The BCSP has some features that ought to minimise social inequalities: it incurs no direct or indirect financial cost to the individual, no time off work is required, and the test is delivered direct to the home, is self-administered and returned in a 'freepost' envelope. This did not prevent SES differences in the UK trial of FOBT screening (Whynes et al, 2003) or the second round of the UK colorectal screening pilot (Weller et al, 2007). However, as inequalities in participation in the cervical screening programme *Correspondence: Dr C von Wagner; E-mail: c.wagner@ucl.ac.uk have narrowed over time (Baker and Middleton, 2003), it is possible that this will occur in the context of the BCSP.
This study assessed FOBT participation in relation to neighbourhood socio-economic characteristics in the London area in the first 30 months of the national roll out of the BCSP. Data were based on test kit return rates for 401 197 individuals aged 60 -69, resident in 808 postcode sectors, who were sent FOBT kits between October 2006 and January 2009.

Implementation of the screening programme in London
In 2006, the Department of Clinical Biochemistry at Northwick Park and St Marks Hospitals became the London 'Hub' for the BCSP, responsible for sending out and analysing FOBT kits, and supported by six screening centres carrying out follow-up investigations. The analyses reported here are based on 401 197 first invitations sent out by the Hub across 808 postcode sectors between October 2006 (when the programme went 'live') and January 2009. Kits were recorded as returned if they were received by the screening centre within 3 months.
Owing to the way the programme is organised, not all eligible individuals would have been invited during the first 30 months of the programme. We therefore divided the number of invitations sent to a postcode sector by the number of people in the age range living in that sector (data from April 2009) to estimate screening coverage. On average, 497 invitations were sent per postcode sector, giving an average of 68% coverage (s.d. ¼ 17.0) during the study period.

Area-level deprivation
Postcode sectors are areas defined by the first inward digit of the postcode and contain an average of 3000 addresses. Socioeconomic deprivation for each postcode sector was indexed with the Townsend Material Deprivation Index (Townsend et al, 1988). This is based on four area-level indicators from the most recent Census: levels of unemployment among those who are economically active, owner-occupancy (% of homes owned by the occupier), car ownership (% with a car) and home overcrowding (% with no communal living space or less than one bedroom per single adult, couple or pair of children) (Mackenzie et al, 1998). Higher Townsend scores indicate greater deprivation.

Other area-level measures
We used 2001 Census data (Office for National Statistics, 2001Statistics, , 2004 to include the following additional factors in the analyses: the proportion of people in each postcode sector belonging to nonwhite ethnic groups (i.e. all ethnic groups other than 'white British', 'white Irish' and 'white Other'), levels of reported poor health (% of people describing their health as 'not good') and household mobility (% of people who had reported moving out of an area within 12 months before the 2001 Census). Household mobility may cause patient lists to be inflated by records of patients no longer residing in the area.

Analyses
Analyses were carried out using SPSS, version 14.0 (SPSS, Inc., Chicago, IL, USA). Postcode sectors were grouped into quintiles of Townsend scores to examine the association between area-level deprivation and uptake using the w 2 test for linear trend. Ethnic diversity and area mobility variables were transformed to correct for skewed distributions: [log 10 (ethnic diversity þ 6); log 10 (area mobility þ 1)]. Linear regression analysis using SPSS Complex Samples was used to examine the independent effect of deprivation after controlling for other area-level factors.
As the percentage of the eligible population who had been sent a kit during the first 30 months of the programme varied from 2% to 97% across the 808 postcodes, and there was a marginally significant trend for variation in coverage across deprivation quintiles (P ¼ 0.07), data were weighted by coverage in further analyses.

RESULTS
The percentage of individuals returning kits increased from 32% in the most deprived quintile, through 34%, 38% and 42%, to 49% in the least deprived quintile (see Figure 1). An important observation here is that the results indicate a gradient across quintiles (linear trend: Po0.001) rather than a sudden 'drop-off' in uptake for the most deprived quintile. Univariate regression analysis (see Table 1) demonstrated that area deprivation accounted for 62% of the variance in response rates across sectors.
Test kit return was also lower for postcode sectors with higher ethnic diversity, area mobility or poor health (all Pso0.001), characteristics that were all associated with area deprivation. Regression analyses were used to examine the extent to which deprivation was associated with screening uptake independently of ethnic diversity, poor health or mobility. The results (see Table 2) showed that associations between each of these neighbourhood characteristics and uptake were substantially reduced after including deprivation in the model. Meanwhile the b coefficient for deprivation was reduced only slightly by the addition of any of these predictors and was actually increased by controlling for poor health.

DISCUSSION
These analyses demonstrate a strong SES differential in uptake of the FOBT-based BCSP in London. Residents in the most affluent   areas were significantly more likely to return the test kits than those in slightly less affluent (though nonetheless still well-to-do) areas, and this pattern continued linearly through the quintiles of deprivation. Controlling for area mobility as an indicator of unreliable address records made no difference to the gradient. Uptake in the most affluent quintile was 50% higher than in the most deprived quintile, showing as much of a gradient as was seen in the early days of the breast or cervical screening programmes (Majeed et al, 1995;Baker and Middleton, 2003). While longitudinal analysis of cervical screening uptake showed a narrowing of the gradient over time (Baker and Middleton, 2003), and this may also turn out to be true for the bowel screening programme, there is a need for immediate action to reduce inequity in delivery. The magnitude of the deprivation effect is particularly worrying given that the greater economic and socio-cultural mix in London might have been expected to dilute the association between indicators of area-level deprivation and screening uptake. Other area-level indicators that are known to be associated with deprivation and are of particular relevance in London, such as ethnic diversity of area mobility, did not explain the socioeconomic variation. Some of the shortcomings associated with our measures (discussed in more detail below) should be considered when interpreting this finding. The measure of ethnic diversity was relatively simplistic and data regarding area mobility would have benefited from a more up-to-date source, but notwithstanding these limitations, it remains an important finding of the present analysis that there was a strong and linear gradient of deprivation on uptake in bowel cancer screening in London that could not be explained by ethnic diversity or area mobility. Consequently, strategies such as translating the materials to other languages or addressing culturally specific attitudes to screening, important as these are in their own right, may only achieve a small impact on the deprivation effect.
The above finding should not detract from the observation that ethnic diversity contributed independently to lower uptake, similar to the effect observed in the English pilot study of the BCSP, which found lower uptake in areas with higher proportions of residents from the Indian subcontinent (Whynes et al, 2003). In this ethnic group, there was evidence that people had more difficulty in completing the test and were more likely to perceive it as unhygienic, embarrassing and distasteful, pointing to some possible avenues for intervention.
The finding that areas with a larger proportion of the population reporting poor health achieved higher uptake accords with individual-level studies identifying perception of good health and absence of symptoms as barriers to early detection (Ogedegbe et al, 2005) including uptake of FOBT (Ioannou et al, 2003). However area-based measures are less easily interpreted because of the possibility that high concentrations of individuals with poor health may influence the provision or quality of health-care services (for example, in the form of health promotion efforts). Nevertheless, the fact that controlling for poor health increased the observed SES gradient highlights the need to emphasise to the public that screening is for everyone, and not just people with symptoms (Brotherstone et al, 2006). This may have implications for the information materials that accompany the test kit.
It has been argued that innovations in health technology can widen inequalities if people with more knowledge, money or power are better able to harness the beneficial effects (Link et al, 1998;Baker and Middleton, 2003;Coleman et al, 2004). While it is not immediately obvious that 'knowledge, money or power' explain SES differences in performance of a comparatively simple home test that has no associated cost and requires no social interaction, the fact that the gradient matches that observed for other forms of screening and other health behaviours, highlights the importance of understanding broader influences on engagement with preventive health behaviours, particularly in relation to cancer prevention.
In the meantime, there is an urgent need to identify midstream or downstream strategies that can ameliorate inequalities in participation in the BCSP. Explanations put forward for social differentials in health behaviours often focus specifically on the lowest SES group (e.g. severe difficulties in health literacy), and although addressing these may not affect the whole gradient, they could contribute to better health outcomes in the most disadvantaged sector (von Wagner et al, 2009). One potential explanatory factor that has been shown to have a graded association with SES is the perceived value of early detection of cancer (Wardle et al, 2004). Fatalism about cancer has been shown to be associated with SES (Powe and Finnie, 2003;Wardle et al, 2004) and could contribute to reluctance to engage with screening if early detection is perceived as creating a longer period of suffering before inevitable death. Strategies to promote a more positive view of cancer outcomes could have a role here. Another factor is future orientation. There is some evidence that SES is related to the relative value attached to future (vs present) gains and losses (Wardle and Steptoe, 2003). If the perceived benefit of better survival a long way ahead is more highly discounted by lower SES groups, it could be insufficient to offset the unpleasantness of collecting faecal samples in the short term. There is no doubt that some people find the idea of completing this test at least unpleasant if not 'disgusting' (Chapple et al, 2008), particularly because touching faecal material is taboo in all societies. The numbers of toilets or private places in the home to store the samples could have a role in compliance; which could be related to SES, but interestingly, there is also some evidence for a social gradient in the 'disgust' response (Rozin et al, 2000). Modifying the sample collection procedure by, for example, introducing alternative testing modalities requiring fewer samples, could contribute to reducing the SES gradient in test participation.

Limitations
There were limitations to this study. Area-based measures of socio-economic deprivation may be out-of-date because they are usually based on census data collected at 10-yearly intervals, but this should serve only to underestimate the deprivation effect. There can also be large variations in levels of personal deprivation within an area; which are not identified with area-based measures (Drever, 1995) making it impossible to draw definite conclusions about associations between uptake and individual SES. The use of smaller geographical units (e.g. Lower Super Output Areas) as well as commercial geo-demographic classification systems, such as ACORN or MOSAIC that supplement census level data with up-to-date demographic and lifestyle data, would give a finer resolution. We also lacked data on sex or age, which might have influenced or interacted with factors in our analysis such as perceived health status, but these require individual-level data that were not available for these analyses. Variability in coverage may have compromised the extent to which we could generalize the current findings although we did control for coverage in our analysis. It was also reassuring to find that excluding postcode sectors with less than 50% coverage (N ¼ 201) did not alter the effect of area-based socio-economic deprivation. Finally, this analysis was restricted to test kits sent out in London, which, as noted above, has a very heterogeneous population. Future studies addressing inequalities in bowel cancer screening will benefit from data from the 'hubs' delivering the programme throughout the country, which will provide information about additional issues such as differences between urban and rural areas.

CONCLUSIONS
This analysis has demonstrated a striking and linear association between area-level deprivation and the proportion of individuals returning an FOBT kit as part of the UK BCSP; which is not explained by differences in ethnic diversity, area mobility or health status. It does not provide explanations, but does emphasise the urgency of constructing more comprehensive multilevel explanatory models. In the meantime, the fact that arealevel characteristics predict participation provides an opportunity to address midstream factors that maintain disparities in uptake through tailored educational and instructional materials to offset some of the negative effects of deprivation.