Journal of the New Zealand Medical Association, 30-November-2007, Vol 120 No 1266
Women with breast cancer in Aotearoa New Zealand: the effect of urban versus rural residence on stage at diagnosis and survival
Hayley Bennett, Roger Marshall, Ian Campbell, Ross Lawrenson
Urban/rural health inequalities have been identified and prioritised in several countries during the past decade. Australia, the United States, and Canada have all found that people living in regional and remote areas have higher mortality rates than people living in urban and suburban areas.1–5 The major contributors to excess mortality in regional and remote areas appear to be circulatory disease, injury (intentional and unintentional), respiratory disease, and cancer.1,2
With respect to cancer, both later stage at diagnosis and poorer survival have been shown in rural residents. Campbell et al investigated over 60,000 patients in Scotland diagnosed with one of six common cancers, and increasing distance from cancer centre was found to be associated with poorer survival. Moreover, a follow-up study by the same research group found that those living further from cancer centres were more likely to present with advanced cancer at diagnosis, and this was thought to account for most of the rural survival disadvantage.6,7
Similar results have been found for rural residents with cancer in the United States, France, and Australia.8–10 Regarding breast cancer specifically, a systematic literature review revealed at least three studies showing rural residents to be more likely to be diagnosed at a later stage of breast cancer than their urban counterparts.11–13 There was only one study that suggested a survival disadvantage for rural women with breast cancer.14
In New Zealand (NZ), while there are concerns about access for rural patients, little national research compares health outcomes of urban and rural residents. Two of the only such studies done in NZ (on breast and upper gastrointestinal cancer) did not show any difference in cancer outcome by urban/rural residence.15,16
This study set out to further explore the question of possible urban/rural health disparity, looking at a recent cohort of women with breast cancer, and using a number of different classification systems for urban and rural residence.
Women with a diagnosis of breast cancer (ICD-10 code C50) in the New Zealand Cancer Registry (NZCR) between 1 January 1998 and 31 December 2002 were identified (n=11,340); 13 women appeared in the dataset twice because they were registered for a second breast cancer.
Data on date of diagnosis, stage at diagnosis, age, ethnicity, domicile code, and date of death (where death occurred) was supplied by the New Zealand Health Information Service (NZHIS). Domicile code was then matched to an urban/rural residence group and a deprivation score using the NZ Deprivation Score (NZDep) 1996.17
Urban and Rural were classified in three different ways. Firstly the standard Statistics New Zealand classification, based on population size, was used to divide the cohort into main urban areas (≥30,000 population), secondary urban areas (10–29,999 population), minor urban areas (1000–9999 population), and rural areas (<1000 population).18
The second classification used was the Urban-Rural Profile Classification, which is based on a comparison between a person's residential address and their workplace address. Main urban areas remain the same, but two other urban categories are defined by the percentage of residents who are employed in a main urban area (see Figure 1).
Satellite urban communities have 20% or more of their working population employed in main urban areas, whilst independent urban communities have less than 20% of their workers employed in main urban areas. The previously defined rural areas are split into four categories, again on the proportion of residents employed in urban areas. This classification thus defines urban and rural areas on the strength of social and economic ties to urban centres, and aims to capture more of the heterogeneity that exists in rural areas.
A rural area with high urban influence (with many residents working in a nearby urban area) is likely to have ready access to urban amenities and services, whereas a rural area with low urban influence will be much more isolated from such services.18
Figure 1. Urban-Rural Profile Classification18
The third method for classifying urban or rural residence was to use distance from major cancer centre. Circles were drawn on a large map in bands of distance around the six major cancer centres (Auckland, Hamilton, Palmerston North, Wellington, Christchurch, and Dunedin). Each case's domicile code was located in one of the distance bands of 0–10, 11–25, 26–50, 51–100, 101–200 and >200 kilometres.
Ethnicity, age, and socioeconomic status have all been shown to affect cancer stage at diagnosis and/or survival,19–22 and were thus deemed to be important confounding variables requiring adjustment in multivariate regression analysis.
Adjustment for cancer stage was also done in the survival analysis. NZCR prioritised ethnicity codes were collapsed into three groups: Māori, Pacific, and non-Māori/non-Pacific—any indication of Māori ethnicity means the individual is prioritized to the Māori ethnic group; age was also entered as a categorical variable in 15-year age brackets. One of the age brackets (50–64 years) was the eligible age-range for BreastScreen Aotearoa during the time of this study. Deprivation score was included in the models as a continuous variable
Logistic regression was used to investigate the effect of urban/rural residence on stage at diagnosis. Stage was dichotomised into “early stage” (localised disease with no nodal involvement = 0) and “late stage” (regional/remote disease = 1). Hosmer-Lemeshow goodness of fit tests confirmed that the logistic model was an adequate fit for the data for all analyses.23 Cox proportional hazards regression was used for survival analysis.24 Survival time was the time from breast cancer diagnosis to death (in days) for cases who died. For women still alive, survival time was deemed to be censored at February 2006 (in accordance with the most up to date mortality data in the NZHIS data extract). Stata® version9 was the statistical package used.25
In the 5 years spanning 1998–2002 there were 11,340 cases of female breast cancer registered in the NZCR. Table 1 shows that approximately three-quarters of women were ≥50 years of age. Most of the cohort were of non-Māori/non-Pacific ethnicity (89.2%); 8.2% were identified as Māori and 2.7% as Pacific.
Stage at diagnosis was fairly evenly split with 45.5% at “early stage” (localised disease) and 37.1% at “late stage” (regional and remote disease). For 17.4% of cases the stage was recorded as not stated/not known, and this was similar for both rural and urban women. In the 4406 women aged 50–64 years, who would have been eligible for BreastScreen Aotearoa, 53.9% were categorised as "early stage", 34.6% as "late stage", and only 11.5% as stage not stated/not known.
Main urban areas were home to 71.6% of the study cohort according to the population size-based classification, while 18% lived in secondary and minor urban areas, and just over 10% lived in rural areas. In comparison, using the Urban-Rural Profile Classification, 73.3% lived in main urban areas, 3.3% in satellite urban communities, 13.3% in independent urban communities, and 10.2% collectively in rural and remote areas.
Just under a third of the cohort lived within 10 km of a cancer centre. Another third lived between 11–50km, 15% lived between 51–100km, and the remaining 22.3% lived over 100 km away from a cancer centre.
In the analysis of the effect of urban/rural residence on breast cancer stage at diagnosis, the 1977 cases with stage not stated/not known were excluded (thus leaving 9363 cases).
The number of cases in the highly rural/remote, and distance >200 km categories was very small, and thus these categories were analysed together with the rural area with low urban influence and 101–200 km distant groups respectively.
Table 1: Characteristics of the study cohort
*3 cases unable to be assigned an Urban-Rural Population Category (because no domicile code recorded in registry); †37 cases unable to be assigned an Urban-Rural Profile Category and ‡39 cases unable to be assigned a distance (3 with no domicile code and other women living in island, oceanic, or inlet areas).
Table 2 shows the results of three separate logistic regressions. Urban/rural residence, however classified, did not have any significant effect on the odds of late stage at diagnosis for women diagnosed with breast cancer in NZ. There was also no trend to increasing odds with increasing rurality or distance.
Table 3 shows the results of Cox Proportional Hazards Regression. There was no statistically significant effect associated with place of residence, except for those women living 51–100 km distant from a cancer centre, who appeared to have better survival. This isolated finding is dubious, particularly as it is unsupported by a trend to better or worse survival amongst the other distances.
Table 2. Odds ratios for late stage diagnosis by residence in urban and rural area
Table 3. Hazard ratios by residence in urban and rural area
The results of this study show that urban/rural residence, however classified, does not affect stage at diagnosis, or survival for women with breast cancer in NZ. This suggests that access to, and effectiveness of screening services, primary care, and specialist cancer services is equitable regardless of whether women live in cities, towns, or rural areas. However it is noted that stage at diagnosis and survival do not solely relate to access/effectiveness of care. For example, a low-grade tumour is more likely to be diagnosed at an early stage regardless the timeliness or effectiveness of care.
Previous cancer research in NZ that has included a geographic analysis has similarly not found differences in stage at diagnosis or survival by urban/rural residence. Armstrong and Borman found no difference in stage of breast cancer by urban/rural group for either Māori or non-Māori women.15 Gill and Martin found that increasing distance from cancer centre was not associated with poorer survival from upper gastrointestinal cancer.16 Our study, which used three different measures of urban/rural residence, is consistent with this previous research.
Several factors may explain why urban/rural residence does not affect breast cancer stage at diagnosis. The nationally coordinated breast screening programme (BreastScreen Aotearoa: BSA) encourages equitable screening for rural and urban women. BSA has mobile/outreach services, and is accompanied by stringent quality standards that help to ensure quality and consistency throughout the country.26
Equitable access to early diagnosis of breast cancer in primary care for both urban and rural women is supported by the existence of formal guidelines for early diagnosis of breast cancer in primary care.27 Furthermore the natural history of most breast cancers is relatively indolent,28 meaning that any small delay in detection of breast cancer for rural women would be unlikely to substantially affect the stage at diagnosis. Superimposed on these factors, is the very high awareness of breast cancer in the community (arising from strong public health messages, the popular media, and women’s health advocacy groups) which is likely to prompt both rural and urban women to seek early medical advice for breast complaints.
The lack of urban/rural difference in survival for women with breast cancer may also be explained by a number of factors. Specialist cancer services are configured in a way that attempts to balance centralisation with local access. The six regional cancer centres provide overall regional coordination, as well as providing clinics in hospitals outside their main centre. So while initial oncology appointments and radiotherapy must be done at the main centre, follow-up appointments and chemotherapy may be done at peripheral hospitals.29
Furthermore, geographic distances that must be covered to reach a major hospital (for cancer treatment) are not as vast in New Zealand as in some countries in which survival differences have been found (e.g. Australia).8 Non-governmental organisations (such as the Cancer Society) also work hard to ensure that assistance is available to support rural women to be able to attend specialist treatment (e.g. volunteer drivers, petrol vouchers).
There are some acknowledged limitations to this study. For a very small percentage of rural addresses, the automated domicile coding software can have difficulty assigning a domicile code. For example, a non-specific rural address (such as R.D. 2) may be geocoded to the rural mail delivery centre of the nearest urban area, which will have an urban domicile code. Thus a small number of rural cases may have been categorised into an urban group, possibly leading to a slight underestimation of the effect of rural residence on stage at diagnosis/survival.30
A second limitation is that we analysed deaths from all causes in the survival analysis, rather than just breast cancer deaths. Given that breast cancer is on average a relatively slowly progressive cancer that predominantly affects middle to older aged women, the amount of 'noise' from non-breast cancer related deaths may be significant in the survival analysis. Lead time bias may also have influenced survival times, particularly as women in the age-range eligible for screening (50–64 years during the time of this study) were more likely to have their cancer detected earlier than those ineligible for screening.
Finally, the time frame of this study is limited, and the cohort of women diagnosed in 2002 will only have had four years of follow up. It may be useful to re-analyse survival data at a later date.
This study did not show an urban/rural disparity in stage at diagnosis or survival for women with breast cancer in NZ. The results may assist policy-makers in deciding where to focus resources for breast cancer in the future. Since urban/rural residence does not appear to have a major influence, resources could perhaps be directed toward other factors, such as ethnicity, which are known to affect stage at diagnosis and survival for women with breast cancer in NZ.22
Competing interests: None.
Author information: Hayley Bennett, Public Health Medicine Registrar, Australasian Faculty of Public Health Medicine, Hamilton; Roger Marshall, Biostatistician, School of Population Health, University of Auckland, Auckland; Ian Campbell, Breast and General Surgeon, Senior Lecturer, Waikato District Health Board, Hamilton; Ross Lawrenson, Professor and Head of Waikato Clinical School, University of Auckland, Hamilton
Acknowledgement: The Ministry of Health’s New Zealand Health Information Service (NZHIS) is acknowledged as the source of the NZCR data.
Correspondence: Dr Hayley Bennett, Public Health Medicine Registrar, 46 Casey Avenue, Hamilton. Email: firstname.lastname@example.org
issue | Search journal |
Archived issues | Classifieds
| Hotline (free ads)
Subscribe | Contribute | Advertise | Contact Us | Copyright | Other Journals