Geographic remoteness and risk of advanced colorectal cancer at diagnosis in Queensland: a multilevel study.
ABSTRACT We examine the relationships between geographic remoteness, area disadvantage and risk of advanced colorectal cancer.
Multilevel models were used to assess the area- and individual-level contributions to the risk of advanced disease among people aged 20-79 years diagnosed with colorectal cancer in Queensland, Australia between 1997 and 2007 (n=18,561).
Multilevel analysis showed that colorectal cancer patients living in inner regional (OR=1.09, 1.01-1.19) and outer regional (OR=1.11, 1.01-1.22) areas were significantly more likely to be diagnosed with advanced cancer than those in major cities (P=0.045) after adjusting for individual-level variables. The best-fitting final model did not include area disadvantage. Stratified analysis suggested this remoteness effect was limited to people diagnosed with colon cancer (P=0.048) and not significant for rectal cancer patients (P=0.873).
Given the relationship between stage and survival outcomes, it is imperative that the reasons for these rurality inequities in advanced disease be identified and addressed.
- SourceAvailable from: Fahui Wang[show abstract] [hide abstract]
ABSTRACT: Differences in late-stage cancer risk between urban and rural residents are a key component of cancer disparities. Using data from the Illinois State Cancer Registry from 1998 through 2002, the authors investigated the rural-urban gradient in late-stage cancer risk for 4 major types of cancer: breast, colorectal, lung, and prostate. Multilevel modeling was used to evaluate the role of population composition and area-based contextual factors in accounting for rural-urban variation. Instead of a simple binary rural-urban classification, a finer grained classification was used that differentiated the densely populated City of Chicago from its suburbs and from smaller metropolitan areas, large towns, and rural settings. For all 4 cancers, the risk was highest in the most highly urbanized area and decreased as rurality increases, following a J-shaped progression that included a small upturn in risk in the most isolated rural areas. For some cancers, these geographic disparities were associated with differences in population age and race; for others, the disparities remained after controlling for differences in population composition, zip code socioeconomic characteristics, and spatial access to healthcare. The observed pattern of urban disadvantage emphasized the need for more extensive urban-based cancer screening and education programs.Cancer 06/2009; 115(12):2755-64. · 5.20 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: Most colorectal cancers still present with symptoms because screening, although effective, is not yet widely practiced. A careful history and physical examination are still the usual methods for suspecting colorectal cancer and ordering appropriate investigation. Therefore, we studied the symptoms, duration, and clues to location of colorectal cancer. We reviewed both hospital and office records for 204 consecutive patients with colorectal cancer, first diagnosed after symptoms, at one regional referral center from 1983-87. We abstracted data on demographic characteristics, presence and duration of 15 symptoms, and characteristics of the tumors. The 194 patients included in the study were similar to those with colorectal cancer described elsewhere in terms of age, gender, and tumor location (58% distal to the splenic flexure), and stage (56% stage A or B). The most common symptoms were rectal bleeding (58%), abdominal pain (52%), and change in bowel habits (51%); the majority had anemia (57%) and occult bleeding (77%). The median duration of symptoms (from onset to diagnosis) was 14 wk (interquartile range 5-43). We found no association between overall duration of symptoms and the stage of the tumor. Patient age, gender, and proximal cancer location were also not associated with a longer duration of symptoms before diagnosis. We developed a rule for predicting a distal location of cancer using multiple logistic regression. Independent predictors were (odds ratio [95% CI]): Hb (1.34 for each g/dl [1.16-1.54]); rectal bleeding (3.45 [1.71-6.95]); constipation (3.16 [1.38-7.24]); and proximal symptoms (at least one of anorexia, nausea, vomiting, abdominal pain, or fatigue) (0.48 [0.20-1.02]). The rule had sensitivity of 93% and a specificity of 47%, with an area under the ROC curve of 0.79. Until prevention of colorectal cancer is more common, we must continue to rely on clinical findings for detecting this cancer. Our results will remind physicians to keep colorectal cancer on the differential diagnosis of "chronic" gastrointestinal symptoms, and our decision rule may prompt earlier investigation with colonoscopy.The American Journal of Gastroenterology 11/1999; 94(10):3039-45. · 7.55 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: The aim of this study was to determine the degree to which welfare state regime characteristics explained the proportional variation of self-perceived health between European countries, when individual and regional variation was accounted for, by undertaking a multilevel analysis of the European Social Survey (2002 and 2004). A total of 65,065 individuals, from 218 regions and 21 countries, aged 25 years and above were included in the analysis. The health outcomes related to people's own mental and physical health, in general. The study showed that almost 90% of the variation in health was attributable to the individual-level, while approximately 10% was associated with national welfare state characteristics. The variation across regions within countries was not significant. Type of welfare state regime appeared to account for approximately half of the national-level variation of health inequalities between European countries. People in countries with Scandinavian and Anglo-Saxon welfare regimes were observed to have better self-perceived general health in comparison to Southern and East European welfare regimes.Social Science [?] Medicine 07/2008; 66(11):2281-95. · 2.73 Impact Factor
Geographic remoteness and risk of advanced colorectal cancer
at diagnosis in Queensland: a multilevel study
PD Baade*,1,2,3, P Dasgupta1, J Aitken1,3,4and G Turrell2
1Viertel Centre for Research in Cancer Control, Cancer Council Queensland, Spring Hill, Brisbane, QLD 4004, Australia;2School of Public Health,
Queensland University of Technology, Brisbane, QLD 4059, Australia;3Griffith Health Institute, Griffith University, Gold Coast, QLD 4222, Australia;
4School of Population Health, University of Queensland, Brisbane, QLD 4061, Australia
BACKGROUND: We examine the relationships between geographic remoteness, area disadvantage and risk of advanced colorectal
METHODS: Multilevel models were used to assess the area- and individual-level contributions to the risk of advanced disease among
people aged 20–79 years diagnosed with colorectal cancer in Queensland, Australia between 1997 and 2007 (n¼18561).
RESULTS: Multilevel analysis showed that colorectal cancer patients living in inner regional (OR¼1.09, 1.01–1.19) and outer regional
(OR¼1.11, 1.01–1.22) areas were significantly more likely to be diagnosed with advanced cancer than those in major cities
(P¼0.045) after adjusting for individual-level variables. The best-fitting final model did not include area disadvantage. Stratified analysis
suggested this remoteness effect was limited to people diagnosed with colon cancer (P¼0.048) and not significant for rectal cancer
CONCLUSION: Given the relationship between stage and survival outcomes, it is imperative that the reasons for these rurality inequities
in advanced disease be identified and addressed.
British Journal of Cancer (2011) 105, 1039–1041. doi:10.1038/bjc.2011.356
Published online 6 September 2011
& 2011 Cancer Research UK
Keywords: colorectal cancer; stage; health inequality; rural health; socio-economic status
While colorectal (CRC) survival rates in Australia are among the
highest in the world (Coleman et al, 2011), people living outside
major cities or in disadvantaged areas have poorer prognosis (Yu
et al, 2005; Kelsall et al, 2009). Since stage at diagnosis is a major
predictor of long-term CRC outcomes (Altekruse et al, 2010), its
relationship to factors such as socio-economic status (SES) and
geographic remoteness is of particular relevance for cancer
control. With major medical centres being concentrated in densely
populated urban centres, it has been suggested that rural and
remote locations may be associated with poorer access to
screening and specialised health care (Parikh-Patel et al, 2006;
Heathcote and Armstrong, 2007). However ecological studies have
presented a mixed picture in terms of SES (Parikh-Patel et al, 2006;
Frederiksen et al, 2008; Henry et al, 2009; Booth et al, 2010) and
geographic remoteness (Fazio et al, 2005; McLafferty and Wang,
2009; Sankaranarayanan et al, 2009) in stage at diagnosis.
Most of the evidence is based on ecological studies and these are
not able to separate the area-level or individual-level influences
(Baade et al, 2010), limiting our understanding about area-level
health inequalities. To date, no Australian study has employed
multilevel methods to investigate links between geographic
individual-level factors and
MATERIALS AND METHODS
Ethical approval to conduct this study was obtained from the
University of Queensland and Queensland Health. Data for
individuals aged 20–79 years diagnosed with invasive stage
1–IV CRC (ICD-O3 codes C18 to C20, C21.8) in Queensland
between 1 January 1996 and 31 December 2007 (inclusive)
(n¼18561) with complete address information were extracted
from the Queensland Cancer Registry (QCR).
Information extracted from pathology forms (Krnjacki et al,
2008) was used to determine stage at diagnosis according to TNM
system (Sobin and Wittlekind, 2002) as described previously
(Baade et al, 2011). For multivariate analysis, localised cases
(Stages I–II) were considered as early stage (Parikh-Patel et al,
2006; Henry et al, 2009) while regional and distant cases were
categorised as ‘advanced’ based on their lower survival rates
(Altekruse et al, 2010).
Information was obtained from QCR on individual-level
variables: year and age of diagnosis, cancer site, gender, marital
status, occupation (Turrell et al, 2007) and indigenous status (see
Table 2 for categories).
Statistical Local Areas (SLAs), which are typically based on
local governments and councils and thus likely to be socio-
economically relevant to their residents, were used as the
geographical definition for area-level analysis (Baade et al,
2010). Remoteness of residence was defined using the Accessi-
bility/Remoteness Index of Australia (ARIAþ) classification
measured using theIndex
Received 6 July 2011; revised 18 July 2011; accepted 13 August 2011;
published online 6 September 2011
*Correspondence: Professor PD Baade; E-mail: firstname.lastname@example.org
British Journal of Cancer (2011) 105, 1039–1041
& 2011 Cancer Research UK All rights reserved 0007– 0920/11
Disadvantage (IRSD) (Australian Bureau of Statistics, 2006) which
categorises SLAs into five quintiles of increasing advantage from
Multilevel logistic modelling (MLwiN 2.21) using Markov Chain
Monte Carlo (Browne, 2009) approaches in MLwiN version 2.21
(University of Bristol, UK) was used. Chain convergence was
checked using Raftery-Lewis diagnostics. Models were compared
Spiegelhalter et al, 2002), with smaller values indicating better fit.
Analyses were conducted in three steps: (1) a null model
comprising individuals (Level 1) nested in SLAs (Level 2) with
no fixed effects; (2) extending to include individual-level factors
as fixed effects (Model 2); and (3) geographic remoteness
(Model 3) and neighbourhood disadvantage (Model 4) were
included separately as fixed effects to quantify how much area
variation in stage was due to these factors independent of
compositional effects, and then in combination (Model 5). Fixed
effects results are reported as odds ratios (95% CI) (Merlo et al,
2001; Eikemo et al, 2008). Significance of individual coefficients was
tested using Z test.
Overall, 57.1% of patients were male with 67.2% having colon
cancer. The mean age at diagnosis was 65 years (median¼66
years). About 44.8% were diagnosed with advanced CRC. Just over
half of the patients (57.6%) lived in major cities and around 36.5%
were in the two most affluent SES quintiles.
In the multivariable logistic regression analyses, the null model
(Model 1) indicated significant (P¼0.041) between area variations
across the SLAs (Table 1).
Based on the DIC statistic (Table 1), the model fit improved
substantially by including individual-level characteristics (Model 2)
and then geographic remoteness (Model 3). Adding area
disadvantage (Model 4) to Model 2 did not improve the fit.
Similarly, the full model (Model 5) provided a poor fit to the data
than Model 3, suggesting that Model 3 was the best-fitting model
for these data. There was no evidence for area-level interaction
(results not shown).
In this final model (Model 3), and independent of individual
factors, geographic remoteness was associated with cancer
stage (Table 2). At individual level sex, occupation, indigenous
status and anatomic site were independent predictors (Po0.001)
of advanced CRC (Table 2). Independent of area effects the
likelihood of advanced CRC was significantly higher for females
than for males; blue-collar workers vs professionals; individuals
with known indigenous status compared with unknown and
patients with colon rather than rectal cancer (Table 2).
Analyses stratified by cancer site (Model 3) (results not shown)
showed that area remoteness was significant for colon cancer
(P¼0.048) but not for rectal cancer (P¼0.873).
Table 1Random effects
Model 1 Model 2 Model 3Model 4 Model 5
Area variance and standard error
P-value for area variance
Percentage reduction in area variance from the null model
Abbreviation: DIC¼Deviance Information Criterion. Model 1: no fixed effects. Model 2: adjusted for sex, age, year, marital and indigenous status, occupation and cancer site.
Model 3: adjusted for sex, age, year, marital and indigenous status, occupation, cancer site and remoteness. Model 4: adjusted for sex, age, year, marital and indigenous status,
occupation, cancer site and area disadvantage. Model 5: adjusted for sex, age, year, marital and indigenous status, occupation, cancer site, remoteness and area disadvantage.
advanced stage colorectal cancer, Queensland, 1996–2007
Final fixed effect factors on the probability of experiencing
Fixed effects OR95% CI
Area-remoteness index of Australia
Year of diagnosis
Not in the labour force
Abbreviations: OR¼odds ratio; CI¼confidence interval.
Determinants of colorectal cancer stage
PD Baade et al
British Journal of Cancer (2011) 105(7), 1039–1041
& 2011 Cancer Research UK
This study is one of the first to consider geographical variations in
CRC stage at diagnosis after adjusting for both area- and
individual-level factors. We found significant evidence that a
person’s risk of being diagnosed with advanced CRC depends on
where they live, specifically for those diagnosed with colon cancer,
independently of the individual characteristics of the patient
themselves. The impact of geographical location, however, was
limited to rurality with no evidence that area disadvantage was
associated with stage at diagnosis.
Given the nature of our data, any discussion of the possible
reasons for the remoteness differential can only be speculative; but
these may include a relative shortage of experienced medical staff
in regional areas and greater difficulty of accessing diagnostic
Significantly higher risks of late-stage diagnosis were seen for
patients with colon vs rectal cancer, consistent with international
studies (Frederiksen et al, 2008; Sankaranarayanan et al, 2009). We
also found that the risk of advanced disease was higher in more
regional areas compared with major cities for colon cancers only.
A contributing reason for both these observations may be that
compared with colon cancer rectal cancer often presents with more
visible symptoms (Majumdar et al, 1999), thereby making patients
more likely to seek medical care and be diagnosed earlier.
The strengths of this study include the use of staged CRC cases
from a large, unselected, state-wide population-based registry.
Approximately 84% of records in our initial cohort had sufficient
information to be staged similar to that reported elsewhere (Yu
et al, 2008). We were limited to the individual-level SES variable of
occupation, since the QCR does not collect information about
education (Frederiksen et al, 2008), income (Frederiksen et al,
2008) or private insurance status (Halpern et al, 2009) known to be
associated with advanced CRC. In addition, due to the high
prevalence of advanced CRC the odds ratios may reflect an
overestimation of the relative risk.
Given the relationship between stage at diagnosis and survival
outcomes, it is imperative that the reasons for the geographical
inequities in advanced disease be identified and addressed.
This study was supported by a grant from the (Australian)
National Health and Medical Research Council (NHMRC)
(ID561700). Associate Professor Peter Baade is supported by an
NHMRC Career Development Fellowship (ID1005334); Professor
Gavin Turrell is supported by an NHMRC Senior Research
Fellowship (ID 1003710).
Conflict of interest
The authors declare no conflict of interest.
AIHW (2004) Rural, Regional and Remote Health: A Guide to Remoteness
Classifications. Australian Institute of Health and Welfare AIHW Cat.
No. PHE 53: Canberra
Altekruse S, Kosary C, Krapcho M, Neyman N, Aminou R, Waldron W, Ruh
LJ, Howlade RN, Tatalovich Z, Cho H, Mariotto A, Eisner M, Lewis D,
Cronin K, Chen H, Feuer E, Stinchcomb D, Edwards BE (2010) SEER
Cancer Statistics Review, 1975–2007. In http://seer.cancer.gov/csr/
1975_2007/ based on November 2009 SEER data submission, posted to
the SEER web site, 2010
Australian Bureau of Statistics (2006) Census of Population and Housing:
Socio-Economic Indexes for Areas (SEIFA), Australia, 2006. ABS: Canberra
Baade PD, Dasgupta P, Turrell G, Aitken JF (2011) Association between
survival and distance to closest radiation treatment facilities for people
diagnosed with rectal cancer in Queensland, Australia. Med J Aust;
e-pub ahead of print 19 September 2011, doi:10.5694/mja10.11204
Baade PD, Turrell G, Aitken JF (2010) A multilevel study of the determinants
of area-level inequalities in colorectal cancer survival. BMC Cancer 10: 24
Booth CM, Li G, Zhang-Salomons J, Mackillop WJ (2010) The impact of
socioeconomic status on stage of cancer at diagnosis and survival.
Cancer 116: 4160–4167
Browne WJ (2009) MCMC estimation in MLwin v2.1. Bristol, UK.
Coleman MP, Forman D, Bryant H, Butler J, Rachet B, Maringe C, Nur U,
Tracey E, Coory M, Hatcher J, McGahan CE, Turner D, Marrett L,
Gjerstorff ML, Johannesen TB, Adolfsson J, Lambe M, Lawrence G,
Meechan D, Morris EJ, Middleton R, Steward J, Richards MA (2011)
Cancer survival in Australia, Canada, Denmark, Norway, Sweden, and the
UK, 1995–2007 (the International Cancer Benchmarking Partnership): an
analysis of population-based cancer registry data. Lancet 377: 127–138
Eikemo TA, Bambra C, Judge K, Ringdal K (2008) Welfare state regimes
and differences in self-perceived health in Europe: a multilevel analysis.
Soc Sci Med 66: 2281–2295
Fazio L, Cotterchio M, Manno M, McLaughlin J, Gallinger S (2005)
Association between colonic screening, subject characteristics, and stage
of colorectal cancer. Am J Gastroenterol 100: 2531–2539
Frederiksen BL, Osler M, Harling H, Jorgensen T (2008) Social inequalities
in stage at diagnosis of rectal but not in colonic cancer: a nationwide
study. Br J Cancer 98: 668–673
Halpern MT, Pavluck AL, Ko CY, Ward EM (2009) Factors associated with
colon cancer stage at diagnosis. Dig Dis Sci 54: 2680–2693
Heathcote K, Armstrong BK (2007) Disparities in cancer outcomes in
regional and rural Australia. Cancer Forum 31: 70–74
HenryKA,Sherman R,RocheLM(2009)Colorectal cancerstageat diagnosisand
area socioeconomic characteristics in New Jersey. Health Place 15: 505–513
Kelsall HL, Baglietto L, Muller D, Haydon AM, English DR, Giles GG (2009)
The effect of socioeconomic status on survival from colorectal cancer in
the Melbourne Collaborative Cohort Study. Soc Sci Med 68: 290–297
Krnjacki LJ, Baade PD, Lynch BM, Aitken JF (2008) Reliability of collecting
colorectal cancer stage information from pathology reports and general
practitioners in Queensland. Aust NZJ Public Health 32: 378–382
MajumdarSR, Fletcher RH, Evans AT(1999)Howdoescolorectal cancerpresent?
Symptoms, duration, and clues to location. Am J Gastroenterol 94: 3039–3045
McLafferty S, Wang F (2009) Rural reversal? Rural-urban disparities in
late-stage cancer risk in Illinois. Cancer 115: 2755–2764
Merlo J, Ostergren PO, Broms K, Bjorck-Linne A, Liedholm H (2001)
Survival after initial hospitalisation for heart failure: a multilevel analysis
of patients in Swedish acute care hospitals. J Epidemiol Community
Health 55: 323–329
Parikh-Patel A, Bates JH, Campleman S (2006) Colorectal cancer stage
at diagnosis by socioeconomic and urban/rural status in California,
1988–2000. Cancer 107: 1189–1195
Sankaranarayanan J, Watanabe-Galloway S, Sun J, Qiu F, Boilesen E,
Thorson AG (2009) Rurality and other determinants of early colorectal
cancer diagnosis in Nebraska: a 6-year cancer registry study, 1998–2003.
J Rural Health 25: 358–365
Sobin LH, Wittlekind C (2002) TNM Classification of Malignant Tumours
6th edn. John Wiley and Sons: New York
Spiegelhalter DJ, Best NG, Carlin BP, Van Der Linde A (2002) Bayesian measures
of model complexity and fit. J R Stat Soc Ser B (Stat Meth) 64: 583–639
Turrell G, Kavanagh A, Draper G, Subramanian SV (2007) Do places affect
the probability of death in Australia? A multilevel study of area-level
disadvantage, individual-level socioeconomic position and all-cause
mortality, 1998–2000. J Epidemiol Community Health 61: 13–19
Yu XQ, O’Connell DL, Gibberd RW, Armstrong BK (2005) A population-
based study from New South Wales, Australia 1996–2001: area variation
in survival from colorectal cancer. Eur J Cancer 41: 2715–2721
Yu XQ, O’Connell DL, Gibberd RW, Abrahamowicz M, Armstrong BK
(2008) Misclassification of colorectal cancer stage and area variation in
survival. Int J Cancer 122: 398–402
Determinants of colorectal cancer stage
PD Baade et al
British Journal of Cancer (2011) 105(7), 1039–1041
& 2011 Cancer Research UK