Socio-economic differences in incidence rates of cancers of the male genital organs in Finland, 1971–95
Article first published online: 22 OCT 2002
Copyright © 2002 Wiley-Liss, Inc.
International Journal of Cancer
Volume 102, Issue 6, pages 643–648, 20 December 2002
How to Cite
Pukkala, E. and Weiderpass, E. (2002), Socio-economic differences in incidence rates of cancers of the male genital organs in Finland, 1971–95. Int. J. Cancer, 102: 643–648. doi: 10.1002/ijc.10749
- Issue published online: 22 NOV 2002
- Article first published online: 22 OCT 2002
- Manuscript Accepted: 23 AUG 2002
- Manuscript Revised: 11 FEB 2002
- Manuscript Received: 31 AUG 2001
- prostate cancer;
- testicular cancer;
- scrotum cancer;
- penile cancer;
We studied social class variation in the incidence of cancers of the prostate, testis, penis and scrotum among 1.1 million Finnish men (45–69 years of age) during 1971–95. The incidence of prostate cancer (6,972 cases) was increasing during the study period; the highest at all the times occurred in Social Class I (highest social class), 40–50% higher than in Social Class IV (lowest). The social class gradient was strongest in localized disease but there was some variation in incidence of non-localized prostate cancer. A total of 174 testicular cancer cases were diagnosed during the study period. In the early 1970s, the incidence of testicular cancer in Social Class I was 5-fold compared to Social Classes III and IV. Thereafter, the incidence rate decreased in Social Class I, but increased in the lower classes. The positive social class gradient was similar for seminomas and non-seminomas. For penile cancer (n = 128), the incidence decreased over time and social class variation was small. Only 6 cases of scrotum cancer were observed. In testicular cancer the strong positive social class association in the early 1970s is disappearing along with converging incidence trend slopes in different social classes. The difference diminished to less than 2-fold in the 1990s. Reasons for this observation remain open. © 2002 Wiley-Liss, Inc.
Social class has been described as a risk marker for different cancer sites in many countries. The associations may be explained considering social class as an indicator of life-style, not as a ‘causal’ risk factor per se. Previous studies have identified associations with high socioeconomic status and 2 cancers of the male genital tract: prostate and testis.1 Studies on social differences on the risk of cancers of penis and scrotum are scarce. We conducted a register linkage study in Finland to access the association between social class and cancers of the male genital tract and how these associations have been changing over time.
MATERIAL AND METHODS
The cohort comprised participants at the official census organized by Statistics Finland on the last day of 1970.2 All residents in Finland were expected to complete a comprehensive questionnaire including demographic and socioeconomic questions, among others. The response rate was 98%.3
A social class classification with 4 ordinary classes was chosen. This classification results from sociological studies in Finland,4 and was formed on the basis of occupation, industrial status and industry groupings.2 Financially dependent persons (e.g., students) were classified by the occupation of their supporter.
The 4 social classes were defined as follows:
- IManagers and other higher administrative or clerical employees, farmers owning more than 50 hectares of land;
- IILower administrative or clerical employees, small-scale entrepreneurs, farmers owning 15–49.9 hectares of land;
- IIISkilled and specialized workers, farmers owning 5–14.9 hectares of land; and
- IVLaborers, farm and forestry workers, institutions' inmates, farmers owning <5 hectares of land, pensioners whose former occupation is unknown.
Persons with unknown social class (1.5% of the total population or 1.0% of the economically active population; mainly farmers and fishermen) were included in the Social Class III, because their behaviors in terms of general mortality and morbidity resembles that of Social Class III.5
The follow-up started on January 1, 1971 and ended at emigration, death or on December 31, 1995, whichever was first. Statistics Finland provided a file with dates of death for deceased persons for the entire cohort during the study period. Person-years for each social class were counted by 5-year birth cohorts (1906–10, …, 1941–45) and the follow-up period was divided into 5-year groups (1971–75, 1976–80, …, 1991–95).
The Finnish Cancer Registry (FCR) has collected data on incident cancer cases in Finland since 1953. Finland has an all-inclusive national health care, which is supplemented by a private health care (used by a very small proportion of the population). All institutions with hospital beds, medical practitioners and pathological laboratories notify the FCR about cancer cases identified. Notification of diagnosed cancer cases to the FCR is a legal requirement. Moreover, Statistics Finland reports to the FCR whenever cancer is mentioned on the death certificate. If there are only laboratory notifications or death certificate information from a cancer case, the Registry requests clinical data from the treating hospital(s).
Linkages between the census, emigration, mortality and Cancer Registry files were carried out using the personal identifiers. These unique 11-digit codes given to each resident in Finland since 1967 are used in all main personal registers in the country. The magnitude of mistakes due to incorrect personal identification numbers is <0.01% and has a negligible effect on cancer risk estimates.6 Among eligible cancer cases, 2.3% had to be excluded because the patients were not found in the 1970 Population Census.
Cancer risk estimates were calculated for strata defined by social class, calendar period and birth cohort. The analyses were restricted to birth cohorts 45–64 years of age at the beginning of each 5-year observation period because these are the age categories within the economically active Finnish population during 1970 census that could be followed throughout the 25-year observation period of our study. Those under follow-up in the first period (1971–75) were born in 1906–25 and those in the last period (1991–95) in 1926–45. We did not present data on men aged 25–45 years because almost no prostate cancer cases were diagnosed in these ages. Whenever age is mentioned, it means the age at the beginning of each follow-up period, although the persons at the end of each 5-year follow-up period were actually 5 years older. For each stratum, 3 types of indicators of cancer frequency were calculated:
- 1Observed number of cases.
- 2Incidence rate: observed number of cases divided by the stratum-specific number of person-years, given per 100,000 person-years. In the figures and text, age-adjusted totals are weighted averages of the 4 age-group specific incidence rates. The weights were based on the age distribution of all person-years in the study series.
- 3Standardized incidence ratio (SIR): ratio of the observed and expected number of cases. The expected number of cases was achieved by multiplying the stratum-specific number of person-years by the period and birth cohort-specific incidence rate of the reference population, which was the total economically active Finnish male population. Confidence intervals (CI) were defined assuming that the observed number of cases followed a Poisson distribution.
A non-commercial computer program developed by the FCR was used to calculate incidence rates and SIR.
We present results for primary cancers of the prostate (by stage), testis (seminoma and non-seminoma separated), penis and scrotum.
The number of men included in the study was 1,074,383: 9% in Social Class I, 24% in Social Class II, 54% in Social Class III and 13% in Social Class IV.
In the present series, 6,972 prostate cancer cases were diagnosed during 1971–95 in the age groups 45–64 years. Of these, 1,965 were reported as non-localized and 3,233 as localized at diagnosis, whereas for the remaining 25% (1,774 patients) the stage was unknown.
The incidence of prostate cancer in this age range increased during the study period (Fig. 1). In all periods and birth cohorts the incidence was highest among men classified as belonging to Social Class I and lowest in men in Social Class IV (p < 0.0001).
For non-localized disease the patterns over social classes is not as obvious as for localized disease (Table I). For non-localized disease the SIR for Social Class I was 1.15 (95% CI = 1.00–1.32) and for Social Class IV 0.92 (95% CI = 0.81–1.03) (p = 0.014, for the difference between Social Classes I and IV). For localized disease the SIR for Social Class I was 1.28 (95% CI = 1.15-1.41) and for Social Class IV 0.85 (95% CI = 0.77–0.93) (p < 0.0001, for the difference between Social Classes I and IV).
|Social class||All cases||Localized disease||Non-localized disease||Deaths*|
|I||781||1.24 (1.15–1.32)||373||1.28 (1.15–1.41)||206||1.15 (1.00–1.32)||307||1.18 (1.06–1.32)|
|II||1680||1.04 (0.99–1.09)||818||1.09 (1.02–1.17)||441||0.96 (0.88–1.06)||706||1.02 (0.95–1.10)|
|III||3643||0.98 (0.95–1.01)||1643||0.95 (0.91–1.00)||1058||1.01 (0.95–1.07)||1578||0.98 (0.93–1.03)|
|IV||868||0.86 (0.80–0.92)||399||0.85 (0.77–0.93)||260||0.92 (0.81–1.03)||429||0.93 (0.84–1.02)|
The relative differences in prostate cancer incidence between social classes in localized prostate cancer remained similar over calendar periods but in non-localized cancer they almost totally disappeared in the 1990s (Tables II, III). The proportion of “unknown” prostate cancer stage varied remarkably little over time (24.3% 1971–75, 26.11% in 1976–80, 25.2% in 1981–85, 22.1% in 1986–90, 27.9% in 1991–95).
|I||40||1.30 (0.93–1.76)||47||1.17 (0.86–1.55)||48||1.09 (0.80–1.44)||84||1.30 (1.04–1.62)||154||1.38 (1.17–1.61)|
|II||109||1.18 (0.97–1.42)||135||1.14 (0.96–1.34)||138||1.14 (0.96–1.33)||189||1.16 (1.00–1.33)||247||0.97 (0.85–1.09)|
|III||234||0.94 (0.82–1.06)||283||1.00 (0.89–1.12)||261||0.97 (0.85–1.09)||329||0.90 (0.81–1.00)||536||0.97 (0.89–1.05)|
|IV||78||0.88 (0.70–1.10)||67||0.75 (0.58–0.95)||64||0.85 (0.66–1.09)||82||0.89 (0.71–1.11)||108||0.87 (0.71–1.04)|
|I||26||1.34 (0.88–1.97)||26||1.17 (0.76–1.71)||44||1.38 (1.01–1.86)||38||0.87 (0.62–1.20)||72||1.17 (0.91–1.47)|
|II||55||0.95 (0.72–1.24)||62||0.95 (0.73–1.22)||75||0.87 (0.69–1.09)||117||1.08 (0.89–1.28)||132||0.95 (0.79–1.12)|
|III||152||0.98 (0.83–1.15)||151||0.97 (0.82–1.13)||204||1.06 (0.92–1.21)||250||1.03 (0.91–1.16)||301||1.00 (0.89–1.12)|
|IV||53||0.98 (0.73–1.28)||53||1.08 (0.81–1.41)||40||0.76 (0.54–1.03)||50||0.83 (0.61–1.09)||64||0.96 (0.74–1.22)|
By the end of 1997, 3,020 of the patients diagnosed with prostate cancer in 1971–95 had died from prostate cancer. The social class distribution of those deaths resembled that of non-localized prostate cancer (p-value for the difference in mortality between Social Class I and IV = 0.013) (Table I).
A total of 174 testicular cancer cases were diagnosed under the study period. In the early 1970s, the incidence of testicular cancer in the highest social class was 5-fold compared to Social Class IV (p-difference between Social Classes I and IV = 0.016; p-trend = 0.03). Thereafter the incidence rate decreased in Social Class I, but increased in the lower classes, leaving the relative difference between the extreme classes 1.6-fold in the early 1990s (Fig. 2; p-difference between social class I and IV = 0.62). The positive social class gradient was similar for seminomas (n = 130) and non-seminomas (n = 44) (Table IV).
|I||24||1.80 (1.15–2.67)||12||2.70 (1.40–4.72)||36||2.02 (1.42–2.80)|
|II||34||1.08 (0.75–1.51)||11||1.04 (0.52–1.87)||45||1.07 (0.78–1.44)|
|III||59||0.86 (0.65–1.10)||14||0.60 (0.33–1.01)||73||0.79 (0.62–1.00)|
|IV||13||0.79 (0.42–1.36)||7||1.23 (0.49–2.53)||20||0.91 (0.55–1.40)|
There was very little difference between social classes in the incidence of penile cancer (n = 128). For Social Class I the SIR was 0.95 (95% CI = 0.48–1.70); for Social Class II 0.94 (95% CI = 0.63–1.36); for Social Class III 1.03 (95% CI 0.80–1.30) and for Social Class IV 1.02 (95% CI = 0.62–1.60). Thus, the extreme social class-specific SIRs varied between 0.94 (Class II) and 1.03 (Class III). The incidence decreased in all social classes over the study periods: the incidence rates for all social classes combined declined from 1.4/105 in 1971–75 to 1.0/105 in 1991–95.
Only 6 cases of scrotum cancer were observed during the period studied: the observed numbers for Social Class I–IV were 0, 4, 2 and 0 and the expected numbers 0, 1.48, 3.17 and 0, respectively.
Our results show an increasing incidence of prostate cancer over time and consistently higher incidence in higher social classes. The variation in incidence of prostate cancer in different population groups has often been explained by differences in the diagnostic activity, e.g., frequency of thin-needle biopsies. PSA testing has been increasingly used as a diagnostic tool in Finland since the late 1980s. A PSA-based screening trial including 80,458 randomly selected Finnish men was running between 1996 and 1999.7 Therefore, organized screening has no effect on the incidence in the observation period of the present study. The increasing prostate cancer incidence could partially be explained by the increase in non-organized PSA screening (and other) way of diagnosis.
The variance of incidence rates attributable to varying diagnostic procedures has been especially connected with cancer of very old men, whereas our results refer to cancer diagnosed in ages <70 years. We were able to study social class differences also in older ages in our cohort among those born in 1906–30, the oldest men (born 1906–10) reaching an age of 89 years in the last year of our follow-up period (1995). In this age range there were 14,436 cancers of the prostate and the SIRs for Social Classes I–IV were: 1.24 (95% CI = 1.18–1.31), 1.09 (1.05–1.12), 0.97 (0.95–0.99) and 0.84 (0.80–0.87). For localized prostate cancer the respective SIRs were: 1.28 (95% CI = 1.19–1.38), 1.07 (1.02–1.13), 0.98 (0.94–1.01) and 0.82 (0.77–0.88) and for non-localized tumors were 1.16 (95% CI = 1.04–1.30), 1.13 (1.06–1.21), 0.96 (0.92–1.01) and 0.84 (0.77–0.93). These patterns are almost identical to the patterns concerning ages 45–64, indicating that diagnostic practices or variables connected to social class did not markedly vary by age.
Prostate cancer mortality or incidence was not strongly associated with socio-economic status in a recent review of the literature.1 Excesses in high social strata were described in Colombia and possibly Turkey.1 Weak negative gradients have been observed in the UK (1970–72) and Spain.1 In Finland, previous reports described a positive social gradient for incidence 1971–855 but not for mortality 1969–75.8
Although it is easy to believe that diagnostic activities strongly contribute to this social class pattern of localized prostate cancer, the 25% higher incidence for non-localized prostate cancer in Social Class I than in Social Class IV might indicate that there is also some difference in the underlying etiological factors. Little is known about the etiology of prostate cancer. Advanced age, ethnic group (blacks) and family history of prostate cancer are established risk factors for prostate cancer, but none of these could explain the social-class gradient observed in our study. There is some evidence that high plasma testosterone or dihydrotestosterone levels, or low levels of sex hormone binding protein (SHBG) may elevate risk. Polymorphisms in the vitamin receptor gene and in the androgen receptor gene, as well as mutations in the BRCA1 gene have also been related to risk in some studies. IGF-I levels have been quite consistently associated with prostate cancer risk. Among the dietary factors, there is some evidence of a protective effect on prostate cancer of diets rich in tomatoes, vitamin E and selenium supplements, fructose-rich foods such as fruit and low consumption of fat, meat and calcium-rich foods.9 The World Cancer Research Fund/American Institute for Cancer Research evaluation of Food, Nutrition and the Prevention of Cancer10 considers vegetables as possibly decreasing prostate cancer risk, total fat, saturated/animal fat, meat and milk and dairy products as possibly increasing risk and alcohol, vitamin C, coffee and tea as possibly not associated with risk. Men in higher social classes in Finland consume more fruit and vegetables than men from lower classes, which is against the hypothesis of dietary etiology. The only dietary hypothesis factor of these that fits with our findings is meat consumption, which has been highest in higher social classes.
Data on the effect of body mass index/obesity on prostate cancer risk has been evaluated recently by the International Agency for Research on Cancer.11 The conclusion, confirming the previous evaluation of the Word Cancer Research Fund publication,10 was absence of an important association between elevated body mass and the risk of prostate cancer. The same evaluation concludes that there is evidence suggesting that physical activity may protect against prostate cancer.11 Occupational physical activity is most probably heavier among blue collar workers, but leisure time physical activity is more common among higher social classes than lower social classes in Finland.12
Non-localized stage is a large, crude strata in which diagnostic differences can still explain the SES gradient in incidence. Those in lower socio-economic (SES) classes may die more often of competing causes of death without having their prostate cancer diagnosed in both localized and non-localized disease. This is a consequence of lower SES being associated with more general mortality compared to higher SES and prostate cancer being generally a relative indolent cancer in which competing causes of death play a significant role.
The pattern observed for mortality is almost identical with the pattern of non-localized disease. This was expected because non-localized stage and poor survival are strongly linked: the 5-year relative survival rates for patients with localized prostate cancer in Finland in 1985–94 was 84% and for non-localized 29%.13
In our study men belonging to the higher social classes groups in the 1970s presented a far higher testicular cancer incidence than men belonging to low social classes. The social class gradient, however, diminished over time. The reasons for this phenomenon are unknown. One could speculate whether some exposure associated with World War II may play a role: the incidence was highest among men who were around 20 years of age during the war and were exposed to special conditions (e.g., exceptional cold) at the front.
As summarized by Akre14 a positive association with high socio-economic status was noted in men diagnosed as early as in 192115 and has been evident in many16, 17, 18, 19, 20, 21, 22, 23, 24 but not all25, 26, 27, 28, 29, 30, 31 studies since then. The proportion of positive studies seems to have decreased over time, possibly due to the true decrease in social class variation also in other countries than Finland.14 In studies discriminating between histological subgroups of testicular cancer, some have found the effect of socio-economic class restricted only to seminoma,21, 23 or to non-seminoma,17 whereas others reported, in line with the present study, a similar pattern irrespective of histology.20, 22, 31
Risk factors for testicular cancer include age (with a peak around age 30), ethnicity (whites have a higher incidence than blacks), family history of the disease, some genetic disorders, birth defects as cryptorchidism and maldescendent testis and testicular trauma.14 In our study, age and ethnicity are irrelevant and the genetic etiology does not fit with the temporal changes. The association of birth defects and social class is poorly known. An increased testicular cancer incidence among patients with AIDS and renal transplant patients suggests an immune-mediated etiology32, 33, 34, 35, 36, 37 that could be social-class dependent.
At least 1 previous study38 found an inverse gradient between penile cancer risk and social class. In our analysis, penile cancer did not show any social class variation, although prevalence of smoking, which has been consistently found as a risk factor for penile cancer,39, 40, 41 is much higher in low social classes in Finland than in upper classes. Other strong risk factors for penile cancer, such as phimosis and chronic inflammatory conditions,42 are not clearly related to social class.
Number of sexual partners seems to be associated with increased penile cancer risk, suggesting an association with sexually transmitted infectious agents and penile cancer.42 Indeed, there is some evidence of an association between past infection with HPV 16 and risk of penile cancer.43, 44, 45, 46 We do not have data about the prevalence of HPV 16 by social class in this population. History of circumcision, particularly neonatal circumcision that may be related to viral load of infectious HPV in the penis and has been associated with a decreased risk of penile cancer in men,39, 42, 47 is very rare in Finland.
In ecological studies, geographical clustering of penile and cervical cancers suggest a common etiology for these tumors.48, 49 Cervical cancer in the wives is not consistently associated with cancer of the penis in the husbands,42, 50, 51, 52, 53 however, which suggests that the etiology of the 2 cancers overlaps only partially. The differences between the flat social class pattern of penile cancer in the present study and the strong negative social class gradient of the respective pattern of cervical cancer54 also speak against a parallel etiology. The social class classification system, however, may not capture relevant measures of the socio-economic gradient and thus we cannot conclude definitively that penile cancer is not associated with social class.
Cancer of the scrotum was too rare to allow conclusions about social class variation in incidence. We were unable to identify studies on scrotum cancer and social class, besides those reporting scrotum cancer in specific occupations, mainly case-reports (i.e., chimney sweeps, occupations dealing with polycyclic hydrocarbons or mineral oils). The etiology of this cancer remains unknown.
The accuracy of the occupational codes (on which the social class coding is mainly based) in Finnish censuses is high: the net error in the occupational codes in the 1980 Census was less than 2%.55 The occupational stability between Censuses of 1975 and 1980 and between Censuses of 1980 and 1985 was 85–86% in both genders,55 the highest (96–97%) appearing among high social class workers. Low stability was characteristic for blue-collar workers; even if they change work, they do not change their social class. Mainly the same individuals change occupation repeatedly.
The coverage and accuracy of the death and emigration files in Finland are very high. The completeness of the Finnish Cancer Registry is over 99% for malignant solid tumors diagnosed in Finland56 and false positive diagnoses are not registered as cancers.57 Because of the high coverage and accuracy of the linked files, the cancer risk estimates by social class can be considered valid.
Most countries lack linkable population-based registries on cancer incidence. Previous studies on social class variation in cancer risk are mostly based on mortality, instead of incidence. This may cause bias, because the principles in definition of the underlying cause of death vary by time, region and perhaps even by social class or occupation. Mortality from competing causes of death and survival of cancer patients also varies by social class.58, 59 A positive social class gradient in the incidence rates can change to a negative one in mortality, due to better survival in upper social classes, as demonstrated with the Finnish data for rectal cancer.4 In the present series we saw the social class variation for prostate cancer incidence to be about 2-fold compared to the respective variation in mortality rates. In this case, the difference in the incidence and mortality patterns seemed to be entirely attributable to more a favorable stage distribution in higher social classes. It has been demonstrated that there is socio-economic variation in the survival rates, even within cancers of similar stage.59
The 1970 census was the first computerized census in Finland and was therefore used for our study. Social class was assessed before the cancer diagnoses, i.e., the bias caused by a downward drift in the social hierarchy as a result of disease60 should have been avoided. Because occupational stability in Finland has been high, the cross-sectional information from 1970 should reflect the social class circumstances from several decades. Social class can be defined in many alternative ways. Different determinants of social class tend to give parallel results in patterns of general mortality61 or cancer incidence rates.62, 63, 64
In conclusion, our data indicate a marked positive social class variation in incidence of prostate cancer, which is consistent over time. The incidence of testicular cancer in highest social class was in the early 1970s 5-fold that of the lowest classes, but the difference diminished to less than 2-fold in the 1990s. Reasons for this observation remain open. Penile cancer did not show any variation by social class.
We thank Drs. L. Määttänen from the Finnish Cancer Registry and M. Lehtinen from the Public Health Institute, Finland, for their helpful comments concerning efforts on screening and viral infection.
- 1IARC. Social inequalities and cancer. Lyon: IARC, 1997. 65–176.
- 2Central Statistical Office of Finland. Population Census 1970: occupation and social position. vol. IX. Helsinki: Official Statistics of Finland VI C, 1974. 104.
- 3Occupational mortality in 1971–75. Studies No. 54. Helsinki: Central Statistical Office of Finland, 1979..
- 4Social structures of the Finnish society. Social Review No. 6. Helsinki: Ministry of Social Affairs and Health, 1966..
- 5Cancer risk by social class and occupation. A survey of 109,000 cancer cases among Finns of working age. Contributions to epidemiology and biostatistics. vol 7. Basel: Karger, 1995..
- 6Use of record linkage in small-area studies. In ElliottP, CuzickJ, EnglishD, SternR, eds. Geographical and environmental epidemiology: methods for small-area studies. Oxford: Oxford University Press, 1992. 125–31..
- 10American Institute for Cancer Research. Food, nutrition and cancer: a global perspective. Washington, DC: AICR, 1997. 310–23.
- 11IARC. Handbooks of cancer prevention. Weight control and physical activity. vol. 6. Lyon: IARC, 2002.
- 12National Institute of Public Health. Health behaviour among Finnish adult population, Spring 1998. Helsinki: NIH, 1998. 13.
- 14Etiological insights into the testicular cancer epidemic. Stockholm, Sweden: Karolinska Institutet, 1999. 11–25..
- 43IARC. Monographs on the evaluation of the carcinogenic risks to humans. vol. 64. Human papillomaviruses. Lyon: IARC, 1995.
- 46Etiology of squamous cell carcinoma of the penis and epidemiology of human papillomavirus infection. Scand J Urol Nephrol 2000; 34: Supplement 205: 1–7., , , .
- 55Occupational mobility in Finland 1975/1980/1985. Studies 160. Helsinki: Central Statistical Office of Finland, 1989..
- 58Inequality in the face of death. Socio-economic mortality differences in Finland in 1971–85. Studies 176. Helsinki: Central Statistical Office of Finland, 1990 (in Finnish)., , .
- 60Class and health: research and longitudinal data. London: Tavistock, 1986..
- 61The health divide. In: TownsendP, DavidsonN, eds. Inequalities in health. Harmondsworth: Penguin Books, 1988. 221–381..
- 63Measurement of women's socioeconomic position in the Finnish Journal of Social Medicine—a review. J Soc Med 1991; 28: 159–65..
- 64The indicators of women's social class and their connection on health differences and the use of health services. J Soc Med 1993; 30: 369–78., , .