Socioeconomic Position and Stage of Cervical Cancer in Danish Women Diagnosed 2005 to 2009

To reduce social disparities in cancer survival, it is important to understand the mechanisms by which social position influence cancer prognosis and how social and biologic aspects interact. Cancer of the cervix is the sixth commonest cancer among women in Denmark (1), with about 380 cases diagnosed each year (2). In 2008, a Danish nationwide population-based cohort study revealed that relatively fewer women with lower social position, as measured by education or income, survived after cervical cancer than women with higher social position; however, this study did not take factors such as stage of cancer into consideration (3, 4). As the timing of diagnosis is closely related to survival from cervical cancer (5), the relations between social factors and cancer stage must be studied to determine factors relevant to disparities in cancer survival. Cancers in less-advantaged women might be diagnosed at a more advanced stage because of low screening uptake, delay in seeking health care, and poor access to specialist care (5, 6). Furthermore, women with a lower socioeconomic position tend to have more comorbid conditions, which has also been associated with survival from cancer (4, 7). This might either increase surveillance or, conversely, decrease individual resources, the attention to abnormal gynecologic symptoms and cause delayed diagnosis.

Most of the previous studies of the effects of social factors on timing of cervical cancer diagnosis were conducted in the United States, used area-based measures of socioeconomic position, or focused on racial or ethnic differences (8–13). We conducted a nationwide population-based study, with individual information on social indicators, to investigate the influence of socioeconomic position on stage at diagnosis of cervical cancer and, in addition, whether comorbidity and time since last Papanicolau (Pap) smear mediated this potential association. This study will contribute to knowledge on the social disparities in stage of cervical cancer in a European setting with free public access to health care.

The study population was identified in the Danish Gynaecological Cancer Database, established in 2005, which covers 96% of gynecologic cancers cases in Denmark (14). We retrieved 1,651 cases of cervical cancer between 1 January 2005 and 31 December 2009 and then excluded 4 women for whom the date of diagnosis was not available, 5 for whom there was no information on cancer stage, and 10 for whom the histologic type of tumor was not recorded. Furthermore, we excluded 33 women who were born before 1920 and 18 who were from Greenland but treated in Denmark or had immigrated to Denmark fewer than 2 years before diagnosis or emigrated from Denmark more than 2 years before diagnosis. In addition, we excluded 10 women for whom no information was available on the 3 main socioeconomic variables and 22 women who were under 25 years of age and therefore considered not to have established their final income or educational level. This left 1,549 women.

From the Danish Gynaecological Cancer Database, we obtained the date of diagnosis, clinical cancer stage, tumor histology, and treating hospital, entered into the database by gynecologists at hospitals for the clinical information and by a gynecologic pathologist for the pathology. Staging was carried out according to Federation Internationale des Gynaecologistes et Obstetristes (FIGO) recommendations. Stage 1a1 cancers were staged from a cone biopsy, whereas more advanced stages were evaluated during general anesthesia by a gynecologist and a gynecologic oncologist together (14). Histologic data for 177 patients were retrieved from the nationwide Danish Pathology Register.

Cancer stage was divided into early (FIGO stages Ia1-Ib2) and advanced (FIGO IIa-IVb). Tumors were grouped histologically into squamous cell carcinomas, adenocarcinomas, and other types (including sarcomas and mixed types).

Data on Pap smear testing were obtained from the Danish Pathology Register. We identified all cervical cytologic examinations and corresponding dates from 1997 until 6 months before the date of diagnosis (later registrations were considered diagnostic tests). In the Danish population-based screening programme, women aged 23 to 50 are invited to screening free of charge every 3 years and, from age 51 to 65 years, every 5 years (before 2007, the National Board of Health's recommendations were every 3 years for women aged 23 to 59). “Last Pap smear test” was divided into 3 groups according to time of last smear: 6 months to 4 years, 4 to 8 years, and more than 8 years before diagnosis.

All somatic diagnoses other than cervical cancer were obtained from the files of the Danish National Patients Register, which contains information on all hospitalizations (since 1978) and outpatients visits (since 1994) in Denmark (15). Diagnoses were coded into a modified Danish version of the International Classification of Diseases version 8 (ICD-8) from 1978 to 1993 and thereafter into ICD-10. On the basis of this information, a Charlson comorbidity index, which covers 19 selected conditions scored from 1 to 6 by degree of severity, was calculated (16). The clinical conditions were summed until 1 year before the cancer diagnosis and grouped into 0 (none), 1, and 2+ for the analysis.

Socioeconomic data were retrieved by linking the 10-digit personal identification numbers (assigned to all residents of Denmark) of the study population to the population-based Integrated Database for Labor Market Research run by Statistics Denmark, which contains data on each individual and is updated each year. Socioeconomic information for 2 years before the date of cancer diagnosis was included to minimize possible reverse effects of early disease symptoms on socioeconomic position. Three socioeconomic indicators were selected to cover different aspects of social influence on health: knowledge-related assets, material resources, and social support (17, 18).

Level of education was based on highest attained education and categorized into short (7 or 9 years of mandatory primary school education for patients born before or after 1958, respectively), medium (8 or 10–12 years, latest grade of primary school, secondary school, or vocational education), and higher (≥12 years of education). Disposable income was based on household disposable income (after taxation and interest) per person and adjusted for the number of persons in the household and deflated to the 2,000 value of the Danish crown (DKK); it was categorized into lowest (1st quartile: ≤106,086 DKK), middle (2nd–3rd quartile: 106,086–197,539 DKK), and highest (4th quartile: ≥197,539 DKK). One thousand DKK equals approximately 135 Euro. Cohabitation status was categorized as living with a partner (married or cohabiting) and living without a partner (single, widowed, or divorced). Cohabiting was defined as, in the absence of marriage, 2 people of the opposite sex, over the age of 16 years, with a maximum 15 years of age difference, living at the same address with no other adult in residence.

Using logistic regression models, we analyzed the relations between the socioeconomic factors and cancer stage at diagnosis. With a diagram of hypothesized causal relations, confounders and mediators were identified (Fig. 1). A 4-step model was used. In model 1, associations between each socioeconomic factor and stage of cancer were estimated with adjustment for age. In model 2, associations between each socioeconomic variable were estimated with adjustment for relevant confounders according to the causal diagram (education level was adjusted for age; cohabitation status was adjusted for age and education level; and disposable income was adjusted for age, education level, and cohabitation status). In model 3, comorbidity was also entered; and in model 4, last Pap smear test was additionally included, to analyze whether comorbidity or time since the last smear mediated the association between social factors and cancer stage at diagnosis. In each of the analyses, individuals with missing data on one of the variables relevant for the specific analysis were excluded.

Interactions between socioeconomic variables, age, and comorbidity were tested for one pair at a time with the Wald test statistics, and no significant interactions were found. All the models were adjusted for a cluster effect of hospital department with generalized estimating equations, with the exchangeable working correlation structure and robust variance estimates. Linearity of age was tested. For the dichotomized outcome variable FIGO stage I versus II–IV, we ensured that the statistical estimate of stage II versus I was very close to the estimates for stage III versus I and of stage IV versus I. Disposable income was also entered into the models categorized into deciles instead of quartiles to ensure that the 1st and 3rd quartiles were valid cut-points for this variable.

A subgroup analysis was done for women aged 31 to 61 (n = 991) with model 4, to determine whether including only women who had received invitations to screening for at least 8 years changed the effects of last Pap smear on cancer stage. As screening has been shown to be less effective in detecting adenocarcinomas (19), additional analyses stratified on main histologic type were done to analyze the size of mediation of last Pap smear separately for squamous cell carcinomas and other types (n = 1,262) and adenocarcinomas (n = 287). Finally, to investigate the relations between education, last Pap smear, and cancer stage further, we estimated the effect of level of education on last Pap smear and carried out a stratified analysis of the effect of education in each stratum of last Pap smear. The analyses were carried out in SAS 9.1 with the proc genmod procedure, and P < 0.05 was considered significant.

The descriptive statistics show that more women aged more than 60 years, with short education, living without a partner, with a comorbidity score of ≥1 and with no Pap smear test within 8 years had advanced cervical cancers at diagnosis rather than early-stage cancers (Table 1). The analysis with stepwise inclusion of mediators (Table 2) showed that all the socioeconomic and disease-related factors included were significantly associated with cancer stage after adjustment for age (model 1). After adjustment for confounders (model 2), the OR for advanced stage cancer was 2.40 (1.67–3.45) for women with short education and 1.76 (1.44–2.16) for those with medium education compared with women with higher education. For women who lived without a partner compared with those who cohabited, the OR was 1.31 (1.10–1.55), and the OR for increasing age was 1.07 (1.06–1.08) per year. The association between disposable income and stage was not significant.

When comorbidity was included as a mediator (model 3), the association between level of education, cohabitation status, and advanced cancer was slightly attenuated but remained statistically significant. The adjusted risk for advanced stage cancer of women with a comorbidity score of ≥2 was 1.51 (1.18–1.94) when compared with those having no comorbidity (Table 2).

When last Pap smear was included as a mediator (model 4), the risk estimates associated with length of education decreased but remained associated with stage (OR = 2.00; 1.33–3.01 for short education and 1.68, 1.31–2.14 for women with medium education when compared with women with higher education). The estimate for women living without a partner was reduced to 1.23 (1.04–1.44), and the association with age decreased to OR = 1.05 (1.05–1.06) per year. The adjusted OR for women with a last smear test >8 years previously was 2.99 (2.70–3.30) when compared with those who had a smear test 6 months to 4 years before diagnosis.

The results for the subgroup of women aged 31 to 61 were similar to the main results, for example, the adjusted risks for advanced cancer were OR = 2.03 (1.50–2.74) for women with short education and 1.55 (1.04–2.30) for those with medium education as compared with higher education (data not shown). The mediation of last Pap smear was slightly lower in the group of women with adenocarcinomas rather than squamous cell carcinomas or other types (data not shown).

The risk for no recent smear test (>8 years ago) versus recent smear test (6 months–8 years ago) was OR = 2.49 (1.73–3.57) and OR = 1.44 (1.05–1.99) for short and medium compared with higher education (data not shown). A relation between level of education and cancer stage was found in each strata of last smear test, with the strongest association in the group of women with >8 years since last smear (Table 3).

In this study of all women with cervical cancer diagnosed in Denmark in 2005 to 2009, shorter education, living without a partner, and older age were strongly related to a diagnosis of advanced cancer. These associations were partly explained by time since the last Pap smear and, to a lesser extent, by comorbidity.

Education has been linked to level of resources, cognition, and reception of health education messages (7, 17); thus, the increased risk for advanced cancer of women with shorter education might be related to insufficient knowledge about cervical cancer and the prevention opportunities. Danish studies of cancers at other sites also found that shorter education was associated with more advanced cancer (20–23), whereas studies of cervical cancer patients in New York City (n = 2,930) and Florida (n = 852) showed no association with education after adjustment for income and marital status (9, 10). However descriptive statistics from the Florida study (9) and from a study in 5 U.S. states and 5 metropolitan areas (13) showed that women with short education were more likely to have regional spread cancer than women with higher education (37% vs. 51% and 31% vs. 38%, respectively). All 3 studies included area-based information on education (median educational level or percentage with high school education) which compared with individual socioeconomic measures, is likely to result in lower estimates of relative risk for health outcomes (24). As our study was population-based, included individual-level data, and was based on an explicit model of causation between social and disease factors, we consider that the results are valid for estimating the individual effect of education.

Women living without a partner were also more likely to have a diagnosis of advanced cancer, implying that social support from a close social relation, like a partner, can result in early diagnosis. A partner might encourage discussion of symptoms related to cancer, screening, health care seeking, and communication with health professionals. In addition, a sexual relationship could lead to earlier symptom recognition. Although cohabitation status was associated with stage in studies of other cancer sites (22, 23), the results for cervical cancer were conflicting. Ferrante and colleagues (9) found that unmarried women were at increased risk for late-stage cancer (OR = 1.63, 1.18–2.25), and Patel and colleagues (n = 7,997) reported that more patients who were single (17.0%), separated, or divorced (17.9%) or widowed (28.4%) had a cancer diagnosed at a late stage than married women (12.6%; P < 0.000, unadjusted; ref. 12). In contrast, Mandelblatt and colleagues, Barry and colleagues, and Goodwin and colleagues found no association between marital status and stage of cancer at diagnosis (8, 10, 18). We measured the effect of cohabitation status, which might be a more relevant indicator of daily social support, as it also includes nonformalized relationships rather than marital status (25), thereby minimizing misclassification of social support from a partner.

When education and cohabitation were taken into account, disposable income was not significantly associated with cancer stage in this study. In other Danish cancer patients, income was related to stage at diagnosis of cancers of the breast, rectum, and lung but not non-Hodgkin lymphoma or colon cancer (20–23) and also the results for cervical cancer and income were divergent in studies conducted in the United States. In 4 metropolitan areas (n = 32,305) women living in low- or middle-income neighborhoods were more likely to have advanced cervical cancer at diagnosis (OR = 1.43 and 1.25; P = 0.01) than those in high-income areas (11), whereas Ferrante and colleagues and Mandelblatt and colleagues found no association with median family income (9, 10). McCarthy and colleagues, Barry and colleagues, and Singh and colleagues found that living in areas with more than 60%, 40%, and 20% poor people increased the risk for advanced cervical cancer by about 20%, 50%, and 20%, respectively (8, 13, 26). None of the studies included measurements of disposable income or individual data on income, and those that found an association did not adjust for education. Our study shows that in a society with free access to most health care services, education but not disposable income seems to be a strong predictor of the timing of cervical cancer diagnosis.

We found that age was strongly related to stage at diagnosis, with a 5% risk increase per year. Similar effects of age (OR = 1.03–1.09 per year) were reported in other studies (8, 10, 11). Reacting to symptoms, attending screening, seeking health care, and communicating with health professionals may be challenging for women in older ages and thus may result in delayed diagnosis.

Comorbidity also increased the risk for advanced cancer in our study. Studies of Danish patients with breast cancer or non-Hodgkin lymphoma showed no association between Charlson comorbidity index score and cancer stage at diagnosis (20, 23), and Ferrante and colleagues also found no association for cervical cancer patients (9), whereas comorbidity reduced the risk for advanced lung cancer in Danish patients (21). Comorbidity measured by the Charlson comorbidity index includes conditions in organs situated far from the gynecologic area (7, 27), thus clinical management of such conditions might not lead to earlier detection of cervical cancer. Having a serious comorbid condition might reduce individual resources when it comes to health care seeking. However, comorbidity mediated only a small proportion of the social disparities in cancer stage in our study, indicating that the effect is only partly socially determined.

The timing of the last Pap smear test was strongly associated with cancer stage, explaining some of the socioeconomic disparities in stage. Lower education or income and being unmarried or widowed have been related to lower screening participation in other studies (28–30); however, much of the difference in stage by educational level was not attributable to time since last smear in this study. Likewise, a study in New Zealand found that only a small proportion of the elevated risk for advanced cervical cancer of Maoris and of women in deprived areas (nonsignificant) was attributable to lack of screening participation (31).

The relation between educational level and stage stratified according to time since last smear showed a similar trend of increased risk for women with shorter education in each stratum. Thus, education seem to be a strong predictor of the timing of diagnosis and, besides, differences in screening participation, differences in health care seeking, reporting of symptoms, treatment or referral to a specialist care may exist by educational group. Patient delay has been shown to account for a large proportion of the total delay in cancer diagnosis and has been linked to low symptom awareness and knowledge of disease, failure to interpret problematic symptoms and incorrect self-diagnosis (32), factors that may be associated with a lower educational level. In addition, it has been suggested that education is associated with poorer adherence to care plans in women with abnormal smear findings (33), but such information was not available from the Pap smear data for this study. Furthermore, lifestyle factors such as smoking and unhealthy dietary habits are more frequent among groups with shorter education and may indirectly affect cancer stage through more comorbidity that might lead to lower screening uptake and/or patients delay. In addition, lifestyle factors such as smoking may also directly increase the progression of cervical tumors, leading to a more advanced cancer at diagnosis (6, 9, 34).

The strengths of the study include linkage of nationwide, population-based administrative register data by the unique Danish personal identification number, which minimized selection bias and misclassification of both disease-related and socioeconomic information. Most previous studies on this topic have had area-based measures, with a risk for misclassifying socioeconomic status (24).

Our cohort was relatively young, and, although we excluded women below 25 years, the youngest members of the cohort might not have had a stable income. However, this would mainly have affected the differences in risk between the lower- and middle-income groups and not the risk relative to the higher-income group. Another limitation was that we could not separate smear tests for screening from those for symptoms; we therefore could not measure the effect of screening itself. Furthermore, we had data on smears only back to 1997, thus some of the social differences in stage among women with a last smear >8 years previously might be due to not measured differences in smear testing before 1997. However, as the estimates of mediation effect of smear test in the subgroup of women aged 31 to 61, who were invited for screening, were similar to those in the main analyses, the bias due to residual mediation of this factor was considered to be small.

As hospitalizations and outpatient visits were the source for calculating the Charlson comorbidity index, patients who were treated only by their general practitioner scored 0. The conditions included in the index are, however, mostly severe (e.g., chronic obstructive lung disease and diabetes), and such patients would at some time have been in- or outpatients at hospital.

In conclusion, shorter education, living without a partner, and older age increased the risk for a diagnosis of advanced rather than early-stage cervical cancer. The disparities in cancer stage were due partly to longer time since last Pap smear test and, to a lesser extent, to comorbidity. It is important to provide better information about screening and symptoms of cervical cancer to disadvantaged groups, to optimize screening invitation strategies and to raise the awareness of general practitioners to gynecologic problems in this patient group.

S.K. Kjær has received a commercial research grant and honoraria from speakers bureau of Sanofi Pasteur MSD, and has been a consultant and on the advisory board of Sanofi Pasteur MSD. C. Johansen has been a consultant and is on the advisory board of The National Board of Health.

Conception and design: E. Ibfelt, S.K. Kjær, C. Johansen, C. Høgdall, B.L. Frederiksen, M. Osler, and S.O. Dalton

Development of methodology: E. Ibfelt, S.K. Kjær, C. Johansen, C. Høgdall, B.L. Frederiksen, and S.O. Dalton

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Ibfelt, C. Høgdall, and S.O. Dalton

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Ibfelt, S.K. Kjær, C. Høgdall, M. Steding-Jessen, K. Frederiksen, B.L. Frederiksen, M. Osler, and S.O. Dalton

Writing, review, and/or revision of the manuscript: E. Ibfelt, S.K. Kjær, C. Johansen, C. Høgdall, M. Steding-Jessen, K. Frederiksen, B.L. Frederiksen, M. Osler, and S.O. Dalton

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Ibfelt, C. Johansen, C. Høgdall, and M. Steding-Jessen

Study supervision: S.K. Kjær, C. Johansen, B.L. Frederiksen, M. Osler, and S.O. Dalton

The authors thank the Danish Gynaecological Cancer Group for the basic clinical data from its database.

The study was financed by grants from the Danish Cancer Society (grant no. SU08005) and from the Danish Medical Research Council (grant no. 271-08-0380).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.