SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. References

A program to ensure an equivalent standard of care for all patients with breast cancer was initiated in the Stockholm area in the mid 1970s. As part of an evaluation of this program, social gradients in clinical stage at presentation and survival were analyzed among patients diagnosed during 1977 through 1997. The patients (n = 15,021) were selected from a database covering about 88% of all diagnosed breast cancer cases in the region. Putative associations were analyzed between clinical stage, survival and different socioeconomic indicators (level of education, income and occupation). There were significant social differences (p < 0.01) in distribution of clinical stage as well as in total and stage-specific survival. High income, more skilled work and a high level of education were all associated with clinically less advanced tumors and hence better survival. However, stage-specific survival differences were mostly generated by differences in nonbreast cancer mortality. The results indicate social inequalities regarding awareness of the disease and/or access to early detection. Social gradients in nonbreast cancer mortality were also found to influence observed survival. In contrast, we observed no significant social differences in stage-specific breast cancer mortality. © 2006 Wiley-Liss, Inc.

Ethnic and socioeconomic effects on breast cancer incidence and survival have been documented in several countries.1, 2, 3, 4, 5, 6, 7 Social gradients in incidence with higher risks among women with a high socioeconomic status have typically been attributed to life-style, particularly differences in distribution of known risk factors associated with reproductive pattern such as age at menarche, age at menopause, age at first full-term pregnancy and use of hormone replacement therapy.1, 2 Social gradients in survival have also been reported, which might reflect inequality regarding awareness of the disease or access to early detection. Typically, some survival difference appears to persist after adjustment for stage, which may reflect social differences in treatment received or yet unknown host factors.3, 4, 5, 6, 7

A comprehensive program to ensure a good and equivalent standard of care for all patients with breast cancer, irrespective of, for instance, domicile or socioeconomic status, was initiated in the Stockholm area in the mid 1970s. The program is organized by the Stockholm Breast Cancer Study Group (SBCSG) and is based on an extensive, multiprofessional collaboration between all those involved in the diagnosis, treatment and follow-up of breast cancer patients in the region. As part of the program, the SBCSG issues clinical practice guidelines concerning methods for primary diagnosis, primary local and systemic therapy, and routines for follow-up and care of patients with recurrent disease. The guidelines have been updated on several occasions over the years. Implementation of the program is facilitated by the fact that almost all health care in Sweden is tax-funded and available to everybody at a relatively low cost. To achieve a widespread implementation of the program, 5 breast cancer detection and treatment centers were set up in the late 1970s at departments of surgery and oncology. They serve as referral centers for all women with suspicious breast lesions as well as for women who have been selected at breast cancer screening. The centers are also responsible for treatment and follow-up of diagnosed breast cancer cases. Typically, every woman with a newly diagnosed breast cancer is discussed at both pre- and postoperative, multiprofessional clinical conferences. About 90% of all breast cancer patients reported from the Stockholm area to the Swedish Cancer Registry are seen at these centers and are consequently reported to a breast cancer database kept by the SBCSG.

A recommendation that breast cancer screening with mammography alone should be part of routine health care was issued by the Swedish Board of Health and Welfare in 1986. In Stockholm, population-based screening was initiated in 1989 and covers women aged 50–69 who are invited to bi–annual screening with a two-view mammography.

As part of an evaluation of the mentioned regional breast cancer program in Stockholm, we analyzed social gradients in clinical stage at presentation and survival among c. 15,000 patients diagnosed during 1977 through 1997. Such putative gradients are of particular interest since the overarching aim of the breast cancer program is to provide good and equal care to all patients. To our knowledge, this is the largest study of social gradients in stage at presentation and survival among breast cancer patients in Scandinavia.

Material and methods

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. References

In summary, the study included 15,021 breast cancer patients aged below 75 diagnosed in the Stockholm area during 1977 through 1997. The patients were selected from a clinical breast cancer database that covered about 88% of all cases reported to the Swedish Cancer Registry from the region. The analysis focused on the putative association between clinical stage at presentation and the lethality of the disease (as measured by stage-specific overall survival and cause-specific mortality) and different socioeconomic indicators (level of education, income and occupation). The study was based on computerized record linkages between the Stockholm Breast Cancer Database 1977–1997, and various population based registers (Swedish Cause-of-Death Registry 1977–1998, Swedish National Censuses 1970, 1975, 1980, 1985 and 1990, Register of the Total Population 1995 and 1997, and the National Register of Education 1997). The linkages were done using the 12-digit identification number, which is unique to all persons living in Sweden.

The study was approved by the Karolinska Institute's Research Ethics Committee.

Setting

The Stockholm region is mainly an urban area in Sweden with a population of about 1.8 million. Over the past 3 decades, primary treatment of women with breast cancer in Stockholm has become somewhat more centralized, particularly with regard to breast cancer surgery. For instance, breast surgery in the late 1970s was performed at more than 15 different centers ranging from small, community hospitals to large, university centers. In the mid 1990s, the number of departments of surgery performing breast cancer surgery had decreased to 10. Radiotherapy and medical oncology was practiced at 3 separate centers in the mid 1970s, compared to 2 centers in the mid 1990s. As mentioned, breast cancer care in the region is coordinated by a comprehensive breast cancer program initiated in the mid 1970s. The program is organized by the SBCSG and is based on a set of clinical practice guidelines implemented at 5 breast cancer detection and treatment centers. Population-based screening with mammography was initiated in the region in 1989. The selected age group was 50–69. The first round was completed in 1991, and subsequent rounds were performed during the 1990s. The participation rate has varied between 70 and 75%. In addition, about 10–15% of the female population participated in screening activities outside of the organized program. Attendance to screening in relation to socioeconomic status has so far not been evaluated in the region.

Stockholm breast cancer database and cancer registry

As part of the regional breast cancer program, a database was initiated in late 1976 aiming to prospectively include all women with primary breast cancer diagnosed in the region. It is based on reports from clinicians collaborating with the SBCSG. The information concerns stage of disease at primary diagnosis, type of surgery, histopathological data, hormone receptor status and follow-up data. The completeness of registration is checked through record linkages with the Regional Cancer Registry of the Stockholm–Gotland area, which is part of the Swedish Cancer Registry.

Clinicians and pathologists are obliged by law to report all new cases of cancer to the Cancer Registry. The completeness of registration in Stockholm was estimated at 98% of all diagnosed breast cancer cases.8 The Cancer Registry, however, includes only little clinical information on each registered case, and does not include follow-up information.

During 1977 through 1997, a total of 20,492 cases of breast cancer were included in the Breast Cancer Database compared with 23,226 cases recorded in the Regional Cancer Registry. This indicates that the coverage was 88% in the database.

Cause-of-death registry

This registry includes computerized information on all deceased individuals registered as Swedish residents at the time of death. It is based on medical death certificates and includes information on date and cause of death (underlying and contributory cause) coded according to the International Classification of Diseases, Injuries and Causes of Death (ICD).

For the purposes of this study, the Breast Cancer Database was matched to the Cause-of-Death Registry for the period 1977–1998.

National censuses 1970, 1975, 1980, 1985 and 1990

These registers were based on mandatory questionnaires sent to every household in Sweden. The available information included data on, for instance, marital status, household size, housing, employment, occupation and income (data on income was not available from the 1980 census). Information on level of education was available only in the censuses performed in 1970 and 1990.

Register of the total population (RTB)

This is a continuously updated database of all individuals residing in Sweden. The register also includes data on income and country of birth.

National register of education

This is a nationwide register covering all individuals living in Sweden aged 16–74. It was started in 1985 and is supplemented each year with new data on the highest attained level of education.

Socioeconomic data

Three different categorical variables were used as socioeconomic indicators:

Highest attained level of education. This was classified as low (compulsory school, 9 years or less), medium (upper secondary school, 10–12 years) or high (university level, more than 12 years). These data were available from the censuses performed 1970, 1990 and from the National Register of Education 1997.

Annual individual total income. This included all sources of income subject to taxation, that is, also social benefits. To obtain a common and over time comparable measure, income was stratified into 3 groups (low, medium and high), based on quartiles for each year of diagnosis with those in the highest and lowest quartile defined as the high and low group, respectively.

Occupation was derived from registered type of occupation defined as nonskilled manual work, skilled manual work, low level white collar work, medium level white collar work, high level white collar work or not gainfully employed. Data on occupation were unavailable from the census 1975.

Eligible women and record linkages

Eligible for this study were women aged 75 or less, with a primary diagnosis of a unilateral breast cancer reported to the breast cancer database during 1977 through 1997, with no previous history of breast cancer and no other malignancy 5 years prior to their breast cancer. The total number of such women was 15,021. Their median age was 58 (range 18–75 years).

The patients were matched to the mentioned censuses to obtain information on the 3 socioeconomic indicators. Data for women diagnosed after 1990, for whom no match was possible (last census was the one in 1990), were instead obtained through matches with the RTB 1995 and 1997. Since data on level of education were only available from the censuses 1970 and 1990, matching was also done against the Register of Education 1997.

Out of the 15,021 women from the breast cancer database, a total of 14,355 (95.6%) were found in the 1970 census, 14,511 (96.6) in the 1975 census, 14,566 (97.0%) in the 1980 census, 13,912 (92.6%) in the 1985 census and 12,779 (85.1%) in the 1990 census. The youngest women were aged 18, which precluded missing data due to age since all individuals aged above 15 years were obliged to fill out the census questionnaires. In 9 women, no match to a census was obtained, probably due to migration and/or early death between censuses.

The data were obtained for each individual primarily from the census closest in time to the primary diagnosis of breast cancer. In cases with missing data from that census, the information was retrieved from another available source closest in time to primary diagnosis. In cases where data on level of education were missing 1990 or 1997, the 1970 Census Data were used if the woman was aged above 35 based on the assumption that educational level probably does not change much at higher ages.

In total, data were missing or impossible to classify on income, socioeconomic status and educational level in 1.9, 7.6 and 10.1%, respectively.

Clinical data

Information from the breast cancer database was recorded on clinical stage at primary diagnosis according to the TNM system. Stage I refers to a tumor with a size less than 2 cm and no clinical evidence of tumor involvement of the ipsilateral, axillary nodes, stage II represents a primary tumor 2–5 cm or clinically involved axillary nodes, stage III refers to tumors larger than 5 cm or with more advanced nodal involvement, stage IV indicates the presence of distant metastases at clinical presentation. Data on clinical stage were missing in 380 cases (2.5%). There was a marked shift in stage distribution during the studied period (Fig. 1). The proportion of women with stage I tumors increased from 35.7% during 1977–1979 to 62.1% during 1995–1997. Conversely, the proportion of women with stage II–IV tumors decreased. The increase of stage I tumors was most pronounced for the period 1989–1991, that is, when population-based screening for women aged 50–69 was introduced.

thumbnail image

Figure 1. Distribution of clinical stage according to period of diagnosis.

Download figure to PowerPoint

Data on hormone receptor status are prospectively recorded in the breast cancer database and are available in most registered cases. However, as receptor status is generally considered to be of only minor, if any, prognostic significance, it was not considered relevant to include such data in the analyses presented here.

Follow up

Information on date of death and underlying cause of death was obtained through linkage with the Cause-of-Death Registry. The total number of deaths was 4,380. In 362 cases (8.3%), the underlying cause was missing. They were classified as a breast cancer death if a local or distant recurrence had been reported to the breast cancer database before death (n = 162), or as a nonbreast cancer death if no such recurrence had been reported. A total of 2,700 of all deaths (61.6%) were thus classified as breast cancer deaths.

The median follow-up time was 8 years (range: 1–22 years).

After the mentioned register linkages, the personal identification numbers were deleted from the files.

Statistical methods

Putative associations between clinical stage and the socioeconomic indicators were evaluated with Pearson's chi-squared test.9 Cases with missing data were excluded. Overall survival and cumulative, cause-specific mortality was estimated according to Kaplan-Meier.10 Distributional differences between subgroups were evaluated with the log-rank test.11 All statistical tests were two-tailed and p < 0.05 was considered statistically significant.

Results

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. References

Clinical stage

There were strong, statistically significant associations between clinical stage and all 3 socioeconomic indicators (level of education, income and occupation) with negative correlation coefficients indicating that high level of education, high income and more skilled type of occupation, were all associated with clinically less advanced tumors (Table I). For instance, among women with a high level education, almost 60% had a stage I tumor, and 7% were diagnosed with a stage III or IV tumor (Fig. 2). In contrast, the corresponding proportions among those with a low educational level were 48 and 11%.

thumbnail image

Figure 2. Distribution of clinical stage according to level of education.

Download figure to PowerPoint

Table I. Correlation Coefficients
 Age at diagnosisTime-period of diagnosisIncomeLevel of educationOccupation
  • *

    p < 0.001.

Clinical stage0.05*−0.19*−0.08*−0.09*−0.13*
Age at diagnosis −0.05*−0.26*−0.27*−0.51*
Time-period of diagnosis  0.11*0.23*0.14*
Income   0.33*0.53*
Level of education    0.46*

All 3 socioeconomic indicators were highly correlated with each other, as well with age and time of diagnosis (Table I).

The socioeconomic differences in stage distribution tended to be less prominent after 1989, that is, after the introduction of population-based screening (Fig. 1).

Overall survival

Overall survival was significantly (p < 0.001) related to all 3 socioeconomic indicators. For instance, the estimated overall survival at 10 years was 75% among those with a high educational level, compared to 61% among those with a low educational level (Fig. 3a). Stage-specific overall survival was also significantly related all 3 socioeconomic indicators (p < 0.01) although the differences tended to be less extreme among those with stage I–II tumors. For instance, results relative to level of education indicated differences between those with a high versus low educational level in the order of 7–10% at 10 years among stage I–II patients (Figs. 3b and 3c). Among patients with stage III tumors, the corresponding difference was 23% (Fig. 3d).

thumbnail image

Figure 3. Overall survival according to level of education. The log-rank p-value is indicated, (a) all patients, (b) stage I patients, (c) stage II patients and (d) stage III–IV patients.

Download figure to PowerPoint

Stage- and cause-specific mortality

The cumulative nonbreast cancer mortality among patients in clinical stages I–III was significantly higher among patients with a low level of education, a low income or nonskilled type of occupation. Results relative to level of education are displayed graphically in Figures 4–6. The cumulative nonbreast cancer mortality at 10 years among patients with a high level of education, for instance, was about 5–6% lower among those with stage I–III tumors, compared with those with a low education level (p < 0.001).

In contrast, the stage-specific analyses of cumulative breast cancer mortality revealed no significant social differences among patients with stage I and II disease. Results relative to level of education are displayed graphically in Figures 4 and 5. Among patients with clinical stage III disease, there was a significant difference favoring those with a high educational level (Fig. 6, p = 0.006). However, in a multivariate Cox-analysis, taking into account age and time of diagnosis, this difference became less pronounced and was no longer statistically significant (Table II).

thumbnail image

Figure 4. Cumulative breast cancer and nonbreast cancer mortality among stage I patients according to level of education.

Download figure to PowerPoint

thumbnail image

Figure 5. Cumulative breast cancer and nonbreast cancer mortality among stage II patients according to level of education.

Download figure to PowerPoint

thumbnail image

Figure 6. Cumulative breast cancer and nonbreast cancer mortality among stage III patients according to level of education.

Download figure to PowerPoint

Table II. COX-Analysis of Deaths Due to Breast Cancer Among Patients with Stage III Tumors
Level of educationNo. of patientsNo. of breast cancer deaths (%)CrudeAdjusted (A)Adjusted (B)
Rel. hazard95% CIpRel. hazard95% CIpRel. hazard95% CIp
  1.  (A): Adjusted for age at diagnosis; (B): Adjusted for age and time-period of diagnosis.

Low509246 (48.3)1.00Ref. 1.00Ref. 1.00  
Medium342139 (40.6)0.800.65–0.980.030.840.68–1.040.120.860.70–1.060.17
High18555 (29.7)0.600.45–0.800.0010.700.52–0.950.020.770.57–1.050.10

Cause-specific mortality among patients with stage IV tumors was not analyzed because of the relatively low number of such patients and because deaths among all such patients, for the purposes of this study, were classified as due to breast cancer.

Discussion

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. References

We found statistically significant social differences in clinical stage at presentation (Fig. 2) and overall survival (Fig. 3a) among c. 15,000 breast cancer patients diagnosed in the Stockholm area during1977 through 1997. High income, skilled work and high level of education were all associated with clinically less advanced tumors (Table I) and, consequently, better overall survival. This observation accords with several previous studies.3, 4, 5, 6, 7 It was beyond the scope of this study to assess the relationship between socioeconomic factors and risk of the developing breast cancer. Previous studies, however, have indicated that risk of the disease is related to high socioeconomic status.1, 2

We also observed significant social differences in stage-specific overall survival (Figs. 3b3d). The differences were substantial and comparable to or even greater than the effect of adjuvant systemic therapy versus no such treatment. However, in the analyses of cause-specific mortality, we found no significant social differences in stage-specific, cumulative breast cancer mortality among patients with stage I and II disease, who constituted more than 90% of all patients (Figs. 4 and 5). Instead, the observed differences in overall survival were mostly generated by differences in nonbreast cancer mortality: patients with a low level of education had significantly higher nonbreast cancer mortality. It is noteworthy that in this material nonbreast cancer deaths accounted for as much as 38% of the total mortality at a median follow-up of 8 years. Among patients with stage III disease (Fig. 6), the crude analysis of breast cancer mortality indicated a strong prognostic effect of level of education with a 40% relative reduction in hazard for those with a high versus a low level. However, in a multivariate Cox-analysis taking into account age and time of diagnosis, the reduction in hazard related to a high level of education decreased to 23%, which was not statistically significant (Table II).

Our observations do not accord with most previous studies in this area, which typically have shown that stage explains some but not all of the socioeconomic differences in breast cancer specific survival.6, 7, 12, 13 Differences not explained by clinical stage have been hypothesized to be related to host characteristics (for instance, obesity, alcohol use, smoking, impaired immune function or psychosocial factors) making the patient more vulnerable or less tolerant of treatment. It has also been suggested that such factors may increase tumor aggressiveness. Another possibility is inequalities in access to modern treatment.

This study had several strengths. It was large and population based. Information on clinical stage had been collected prospectively. Socioeconomic data were evaluated using census data and other population-based official registers. Such registers also ensured an almost complete and nonbiased follow up. Data on cause-specific mortality were based on officially recorded underlying causes of death, so nonrandom misclassification was probably not an issue. However, it might be seen as a weakness that we used clinical instead of histopathological stage. It is well-known that clinical stage is an unreliable indicator of extent of disease compared to histopathological data. However, for the purposes of this study, it was considered more appropriate to use clinical stage since histopathological data are unavailable among patients with inoperable disease. Also, preoperative radiation or systemic neoadjuvant therapy obscures histopathological information on tumor size and axillary nodal status. Information in the breast cancer database indicated that about 12% of registered cases had not undergone primary surgery. Use of histopathological stage would tend to reduce misclassification, but would exclude more advanced cases. Our results indicated that such cases are more prevalent among those with a low socioeconomic status, so use of histopathological data would introduce bias. In contrast, there is no reason to suspect that misclassification using clinical stage is nonrandom. Because of that, our estimates of social differences in stage distribution are, if anything, conservative.

A problem inherent in all studies of health effects of social deprivation is that the size and composition of groups defined according to socioeconomic indicators vary over time and may thus not be comparable. For instance, we found strong correlations not only between stage and socioeconomic indicators, but also between these indicators and period of diagnosis. For example, over time a decreasing proportion of the population belonged to the group with a low educational level or who had nonskilled manual work.

Level of education is a frequently used socioeconomic indicator since it typically varies less in adult life than, for instance, income or type of occupation. However, we found that all 3 socioeconomic indicators were strongly correlated with each other. Consequently, conclusions about the relationship between socioeconomy and clinical stage were roughly the same irrespective of which indicator was used.

The outcome of studies of socioeconomic differences in cancer care are probably largely specific to a particular health care environment, so differences between, for instance, countries are not surprising. Sweden is often regarded as an ethnically and socioeconomically relatively homogeneous society. Almost all health care is tax-funded and aims to provide care to all on equal terms. In addition, there is a relative uniformity of regional treatment programs, which might be thought to minimize differences in diagnostic accuracy, staging and primary treatment, all of which are important determinants of stage-specific survival. However, to date, there are only limited data on these issues for the whole of Sweden, so our results for the Stockholm area cannot be extrapolated to the rest of the country. In fact, a recent Swedish study covering regions outside of Stockholm showed a significantly higher risk of death due to breast cancer after adjustment for tumor characteristics and age, among women with a primary breast cancer diagnosed in 1993 with low versus high socioeconomic status.7 One possibility for the discrepancy between those results and the current study is that a comprehensive breast cancer program was first started in Stockholm, although in more recent years, similar programs were initiated in the whole of Sweden. The overarching aim of the regional breast cancer program in Stockholm was to ensure a good and equivalent standard of care for all patients. During the study period, we found that 88% of all breast cancer cases in the region reported to the Swedish Cancer Registry was covered by the program in that they had been reported to the breast cancer database from one of the 5 collaborating breast cancer centers in the region.

The observed social gradients in stage at presentation indicate inequality regarding awareness of the disease or access to early detection. In 1989, a population-based program for breast cancer screening with mammography every 2 years was introduced in the region, for women aged 50–69 years. One may speculate that nonattendance to screening was higher among socially deprived groups and, therefore, may have contributed to the persisting social differences. In addition, the screening program only concerned women aged above 50. Approximately 25% of breast cancer in Sweden occurs among younger women. No study from the Stockholm area has to date been published on screening attendance versus socioeconomic status. However, the current results suggest that the introduction of the service screening program tended to attenuate the social differences in stage at presentation as such differences were substantially less prominent after 1989 than in previous periods.

Since we observed no differences in stage-specific cumulative breast cancer mortality, it would appear that the regional breast cancer program largely achieved the goal of equal care. However, this does not rule out all types of social inequalities in terms of treatment or follow-up, for instance, differential access to breast conserving surgery, or small to moderate differences in the use of adjuvant systemic therapy. Randomized trials have indicated that the extent of primary surgery has little impact on overall survival, but may have implications for the patient's quality of life. Adjuvant systemic therapy has unequivocally been shown to improve survival. However, despite the large number of patients included in this analysis, we cannot rule the possibility of some differences in this regard as they would have a limited effect on overall survival. Other important aspect of quality from the patient's and her family's perspective include waiting times, information, rehabilitation and patient involvement in clinical decision-making. Such aspects were not part of this evaluation.

A more detailed analysis of the social differences in nonbreast cancer mortality was beyond the scope of this paper. However, such differences have been reported from several western countries, as well as from Sweden.14 They have typically been attributed to life-style factors such as smoking and diet. Inequalities in awareness of disease and access to primary and secondary preventive measures have also been suggested. A recent report from the Swedish National Board of Health showed that these inequalities do not appear to decline over time.15

References

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. References
  • 1
    Faggiano F, Partanen T, Kogevinas M, Bofetta P. Socioeconomic differences in cancer incidence and mortality. IARC Sci Publ 1997; 138: 65176.
  • 2
    van Loon AJ, Brug J, Goldbohm RA, van den Brandt PA, Burg J. Differences in cancer incidence and mortality among socio-economic groups. Scand J Soc Med 1995; 23: 11020.
  • 3
    Coleman MP, Babb P, Sloggett A, Quinn M, de Stavola B. Socioeconomic inequalities in cancer survival in England and Wales. Cancer 2001; 91: 20816.
  • 4
    Auvinen A, Karjalainen S, Pukkala E. Social class and patient survival in Finland. Am J Epidemiol 1995; 142: 10894.
  • 5
    Kogevinas M, Porta M. Socioeconomic differences in cancer survival: a review of the evidence. IARC Sci Publ 1997; 138: 177206.
  • 6
    Kaffashian F, Godward S, Davies T, Solomon L, McCann J, Duffy SW. Socioeconomic effects on breast cancer survival: proportion attributable to stage and morphology. Br J Cancer 2003; 89: 16936.
  • 7
    Lagerlund M, Bellocco R, Karlsson P Tejler G, Lambe M. Evidence of socio-economic differences in breast cancer survival – a population-based cohort study (Sweden). Cancer Causes Control 2005; 16: 41930.
  • 8
    Mattsson B, Rutqvist LE, Wallgren A. Undernotification of diagnosed cancer to the Stockholm cancer registry. Int J Epidemiol 1985; 14: 649.
  • 9
    Armitage P. Statistical methods in medical research. Oxford: Blackwell Scientific Publications, 1980. 20712.
  • 10
    Kaplan E L, Meier P. Non-parametric estimation from incomplete observations. J Am Stat Assoc 1958: 53; 45781.
  • 11
    Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 1959: 22; 71948.
  • 12
    Thomson CS, Hole DJ, Twelwes CJ, Brewster DH, Black RJ. Prognostic factors in women with breast cancer: distribution by socioeconomic status and effect on differences in survival J Epidemiol Community Health 2001; 55: 30815.
  • 13
    Newman LA, Bunner S, Carolin K, Bouwman D, Kosir MA, White M, Schwartz A. Ethnicity related differences in the survival of young breast carcinoma patients. Cancer 2002; 95: 217.
  • 14
    Kunst AE, Groenhoff F, Andersen O, et al. Occupational class and ischemic heart disease mortality in the United States and 11 European countries. Am J Public Health 1999; 89: 4753.
  • 15
    Folkhälsorapport 2005 (In Swedish). The National Board of Health and Welfare, Stockholm 2005.