Effect of rurality on screening for breast cancer : a systematic review and meta-analysis comparing mammography

Introduction: The lower breast cancer survival rate observed among rural women may be related to differences in screening access and utilization. We evaluated existing evidence for rural and urban differences in mammography service use in adult women. Methods: A systematic search was conducted on 4 April 2012 and updated on 1 November 2012, which yielded 28 studies for inclusion. Results: The rural population was less likely to have mammographic breast screening, and this difference was consistent in various areas of the USA as well as across a number of other countries. Meta-analyses using random effects models showed that women residing in rural areas were less likely than urban women to have ever had a mammogram (odds ratio (OR)=0.74, 95% confidence interval (CI)=0.62–0.89) or to have an up-to-date mammogram (OR=0.59, 95%CI=0.49–0.70). Conclusions: Mammography is currently the best tool for the early detection and diagnosis of breast cancer. The rural disadvantage this review has identified may contribute to the lower breast cancer survival among women living outside urban areas.


Introduction
Breast cancer is the most prevalent cancer in women and disparities in prognosis based on area of residence have been observed [1][2][3] .Rural areas are characterized by a low population density and residents often have poorer access to health care than urban areas 4 .While the incidence of breast cancer has been observed to be lower in women residing in rural areas, in those diagnosed with breast cancer, death rates are higher 3,[5][6][7][8] .
Rural-urban differences in the breast cancer experience of women have been systematically reviewed 9 .Across the 41 studies reviewed, rural women were likely to report difficulties in breast cancer health service access, such as a greater distance to breast cancer specialists and treatments.Treatment procedures also differed, with a rural disadvantage noted in access to, and use of, breast cancer services after diagnosis.However, this review did not consider prediagnosis screening differences between women residing in rural and urban areas.
Despite recent evidence questioning the efficacy of mammograms across all age groups 10 , current guidelines suggest that women over the age of 50 years should have a mammogram every 2 years because this is currently the best tool available for the early detection of breast cancer 11 .Most developed countries do have population-based breast screening programs in place, where women over 50 years of age are encouraged to have a mammogram every 2 years 11 .However, whether the urban-rural differences in the use of mammograms is country-specific remains uncertain.The aim of this systematic review is to compare all existing evidence on the prevalence of mammograms between women residing in rural and urban areas.

Search terms and databases
Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were used to perform the systematic review.A review protocol for search and inclusion criteria was determined in advance and its completion was documented.The search strategy was developed by JL and SM and was implemented by JL.
The literature search was carried out on 4 April 2012 and updated on 1 November 2012.The databases searched (from earliest available date) were PubMed (from 1951), CINAHL (from 1982), Embase (from 1966), Sociological Abstracts (from 1952), Cochrane (from 1996), Web of Science (from 1898), and RURAL (from 1966).Search terms were ('breast cancer' or 'breast neoplasms' or 'breast tumor' or 'breast tumour') and ('rural' or 'rural population' or 'rural health' or 'rural health services' or 'rural health care' or 'urban rural differences') and ('mammography' or 'mammogram' or 'breast examination' or 'physical examination' or 'early detection of cancer' or 'cancer detection').Limits were applied on the search to only include peer-reviewed publications in English with a sample of adult women, with no restrictions on the date of publication.Lists of references from retained articles were screened for additional material.Where full-text articles could not be retrieved through databases, authors were contacted and the articles requested.

Eligibility criteria and study selection
Studies to be included underwent assessment by two independent reviewers (JL and SM).Where discrepancies existed between the reviewers, papers were referred to a third reviewer (DM).Studies were included in the systematic review if they met the following eligibility criteria: (1) the population included adult women; (2) the outcome measure was clinic record or self-report of mammography; (3) quantitative statistics of rural versus urban comparisons were present; (4) the study included an observation on the screening rates in a non-intervention sample; (5) quality was 6/10 or higher, based on the assessment described below.Cross-sectional studies were included.Baseline crosssectional observations from intervention studies before the intervention had been carried out are included.Studies were included in the meta-analysis if required statistics were available.

Quality assessment
Quality and risk of bias of each study was assessed by adapting the Newcastle-Ottawa Scale (NOS) for quality assessment of non-randomized studies in meta-analyses 12

Data extraction and analysis
Data extraction was performed on studies that met eligibility criteria.The parameters that were assessed included study country, sample size, sample age, rural-urban definition employed, outcome measure, control variables, and screening rates across rural and urban areas.For the metaanalysis, data extracted on screening rates included the number of participants who had mammograms in the rural and urban sample out of the total number of participants in the rural and urban sample.In studies that included more than two classifications of the rurality of area, data on the most urban and most rural area were extracted and analyzed.In the case where these data were not presented in the article, the data were obtained by contacting the study authors.The individual studies were weighted by their sample size.
Results of whether women had ever had a mammogram and whether women had an up-to-date or recent mammogram were reviewed separately.The overall effects of residing in rural areas on mammogram rates from the studies were pooled and estimated using the Cochrane Collaboration Review Manager software, RevMan v4.1 (http://tech.cochrane.org/Revman).The Mantel-Haenszel method was employed.Heterogeneity of studies was assessed by calculation of the χ 2 and I 2 statistics.A random effects model was used when the test of heterogeneity was significant.Risk of bias across studies was assessed by examination of funnel plots.Results were presented as odds ratios (OR) and 95% confidence intervals (CI).Where an adequate number of studies were present, sub-analyses were conducted stratified by country, age of sample, year of data (separate analysis for studies with data from 1990 to 1999 and 2000 to 2009), and definition of the outcome measures.

Study selection
Figure 1 shows a flow chart of included papers.The search resulted in 745 articles, of which 119 met the eligibility criteria from screening the titles and abstracts.From the full text examinations of these articles, 28 studies were found to meet the inclusion criteria for the systematic review.This resulted in a total of 1 308 126 observations across the studies.Of these, a total of 20 studies (17 provided the data; 3 were requested from authors) had data available for, and were included in, the meta-analysis.
Ten of the studies included women aged 40 years or older and seven studies were conducted on women in a specific age

'Ever had a mammogram'
Eight studies compared urban and rural women in relation to whether they had ever had a mammogram (Table 2).Five studies, which were from the USA, Canada, and Australia, found that the proportion of women who had ever obtained a mammogram was greater in the urban population than in the rural population 20,25,31,32,37 .One study from the USA observed no rural and urban differences in women aged 60 years or older 17 .Contrastingly, two studies, one from Northern Ireland 40 and one from Korea 35 , found that women residing in more rural areas were more likely to have ever had a mammogram than women residing in more urban areas.
Six studies had data for inclusion in the meta-analysis (Fig3) 20,25,31,32,35 .This resulted in 225 447 rural and 189 416 urban observations.Heterogeneity tests showed that the studies were heterogeneous, τ 2 =0.05, χ 2 (5)=109.28,p<0.001,I 2 =95%.Therefore, a random effects model was fitted.The test for overall effect was significant, Z=3.29, p<0.001.Pooled effects results showed that the rural population had significantly lower odds of ever having a mammogram than the urban population (OR=0.74,95%CI=0.62-0.89).For sensitivity analyses by country, only the USA and Canada had an adequate number of included studies for a separate analysis.The effect was stronger when only including studies from the USA 20,25 and Canada 31,32 (OR=0.65,95%CI=0.49-0.86).

'Up-to-date mammogram'
There were 23 studies that compared the prevalence of upto-date mammography service use between the rural and urban populations (Table 3).Most of the studies defined 'upto-date' as having had a mammogram in the past 1 or 2 years.
One study compared women who had a mammogram in the past 3 years 19 .All studies observed that women residing in more urban areas were more likely to have an up-to-date mammogram than women residing in more rural areas.

Discussion
The aim of the present study was to review the existing literature on rural and urban differences in mammogram use.Our analyses indicate strong and consistent evidence of differences in mammogram uptake between women residing in rural and urban areas in a number of developed countries.These differences remained consistent even when the mammographic services offered in each country may not have been directly comparable in terms of cost and availability.The rural disadvantage on the ever use of a mammogram was only observed in the USA, Canada, and Australia.However, the observation that women residing in rural areas were less likely than their urban counterparts to have had a recent mammogram was found across all the included studies.These findings are consistent with a previous review showing urban-rural disparities in breast cancer treatment and experience 9 .
Although the focus of the present paper was on comparisons of rural and urban women across many different systems, it is important to note that countries differ in the preventive public health services that are available.The importance of this work is that, within a country, differences in mammogram use are likely to play a key role in differential rates of early detection and diagnosis of breast cancer 11,42 .The data in this study therefore provide evidence for policy makers when deciding on allocation of screening resources.
A number of limitations should be considered when interpreting the findings of this paper.Firstly there is heterogeneity among studies, although this is to be expected because of the diversity of study settings.However, the metaanalysis results were consistent with the systematic review findings, and so the observed heterogeneity was unlikely to change the interpretation of the findings.While there are criticisms of wide variation of the definition of 'rurality' in previous research 9 , in the current review, most studies employed a national level standard definition.Internationally, the standard classification methods of rurality were similar and were commonly based on distance to major service areas and population density.
Another discrepancy in the reviewed studies was the age of the sample.Some studies included women who were 18 years and older, while others examined only women who were 60 years and older.Nevertheless, the urban and rural differences in mammogram use were generally consistent across studies of different age groups.Most studies examined breast screening service use in women from 40 years of age onwards, as population health breast screening programs are often targeted to women of this age group.Our meta-analysis on studies that include women from 40 years revealed a rural disadvantage in mammography use.
Studies that focused on particular age groups have merit over studies that simply include women across all ages, as the guidelines and public health recommendations and services available to women in different age groups are different 11,42 .Our search strategy was limited to English language publications, thus we may have overlooked studies that were indexed in other languages, or that appeared in electronic databases that were not accessed.Cultural variations in screening or assessment could affect results.A strength of the included studies was that many employed large national-level data sources, which allowed for adequate power.In addition, many of the reviewed studies controlled for socio-demographic variables, a potentially important factor as rural populations are often characterized by lower socio-economic status 43 .. While there are numerous studies on rural and urban differences in breast screening conducted in the USA and Canada, research in other countries is rare, although our review showed that the rural disadvantage in mammogram screening is likely to be an issue across several countries.
Research comparing rural and urban differences in breast screening rates in the population has only received research attention in recent years.To date, no studies have followed rural and urban women to ascertain if they consistently follow breast screening guidelines over time.As long-term adherence to breast screening guidelines is recommended, longitudinal studies to examine how urban and rural women engage in mammographic screening are warranted.

Conclusions
This review showed a difference in the use of mammograms in rural compared to urban women across several countries.Although there were limitations in terms of different healthcare systems in which the work was undertaken, a rural disadvantage in up-to-date mammogram use was observed internationally.However, it is clear from the analysis that more studies on rural and urban comparisons on breast screening are needed in countries other than the USA and Canada.Mammographic screening is an important step in the early detection and diagnosis of breast cancer, and future studies could also explore influences, such as physical or psychosocial factors, that may affect mammogram use in rural areas.
by the University of Newcastle and the University of Queensland.We are grateful to the Australian Government Department of Health and Ageing for funding.The funding source has no role in conducting the study.We thank Roslyn Boyd, Lars Eriksson, and Lavinia Pitic for their advice and assistance on systematic reviews.43.Australian Institute of Health and Welfare.Rural, regional and remote health: indicators of health system performance.Canberra: Australian Institute of Health and Welfare, 2008.
. The development of the NOS involved panel review and critical review by experts.The NOS has strong face validity and is endorsed by the Cochrane Collaboration.Ratings were made on a dichotomous scale on whether the study met the criteria on the selection of study groups, comparability of groups, and the assessment of outcomes.The 10 criteria assessed in the current systematic review were (1) representativeness of the rural and urban sample; (2) selection of the rural and urban sample; (3) definition and ascertainment of rural and urban areas; (4) report of the rural and urban sample sizes; (5) comparability of the rural and urban sample; (6) report of mammography events; (7) effect size described; (8) assessment of mammography; (9) definition of an up-to-date mammography; and (10) adequacy of time-frame period.The total quality score ranged from 0 to 10, with higher scores within the range indicative of higher quality.

Figure 1 :
Figure 1: Flow chart of included papers identified in the review of rural-urban comparisons on breast cancer mammogram screening.

range from 40
years (such as 40-75 and 40-80).Participants' age in the remaining studies varied from >18 years to 60 years or older.Quality was high, with a mean study quality evaluation score of 7.74 (standard deviation (SD)=1.13).Overall, the studies employed a population-representative sample for the rural and urban groups.Most studies controlled for key factors, such as age and year of data, and the studies controlled for a range of socio-economic variables when such data were available.Most studies clearly stated the use of a standard rural-urban classification method.One study from the USA used data from 198720 .Eight studies used data from 1991 to 1999, and 19 studies used data from 2000 to 2009.Funnel plots were examined and indicated that publication bias was unlikely (Fig2).

Figure 2 :
Figure 2: Funnel plot of the six studies included for the outcome of 'ever had a mammogram' and the 16 studies included for the outcome of 'up-to-date mammogram'.

Figure 3 :
Figure 3: Meta-analysis results of 'ever had a mammogram' between the rural and urban populations.

Figure 4 :
Figure 4: Meta-analysis results of 'up-to-date mammogram' between the rural and urban populations.

Table 1 :
Characteristics of included studies that examined rural and urban differences in breast cancer screening (n=28)

Table 2 :
Studies comparing whether the women had ever had a mammogram between the rural and urban populations

Table 3 :
Studies comparing the prevalence of 'up-to-date mammogram' between the rural and urban populations *p<0.05.U>R, urban prevalence greater than rural prevalence.R>U, rural prevalence greater than urban prevalence.