J7ournal of Epidemiology and Community Health 1991; 45: 216-219 J Epidemiol Community Health: first published as 10.1136/jech.45.3.216 on 1 September 1991. Downloaded from

Socioeconomic differences in cancer survival

M Kogevinas, M G Marmot, A J Fox, P 0 Goldblatt

Abstract Two decades later, interest in survival patterns Study objective-The aim was to was renewed when large differences among ethnic investigate the relationship between groups in the USA became evident, possibly due socioeconomic status and cancer survival. to differences in the timing of cancer detection.2 Design-This was a prospective study, To obtain nationally representative figures, we linking census and vital registration records examined cancer survival for the period 1971- for an approximate 1% representative 1983 in England and Wales, in the OPCS sample ofthose enumerated in England and Longitudinal Study, a 1%/' sample of people Wales in the 1971 census. identified in the 1971 census. Setting-The study population is nationwide. Participants-The study sample consists Methods of 250 588 men and 262 484 women. During The OPCS Longitudinal Study is a cohort study 1971-81, 17 844 cases of cancer were of an approximately 1 % representative sample of registered, and of those registered, 13 532 those enumerated in England and Wales in the died during 1971-1983. 1971 census and it links census and vital Measurements and main results- registration records.>5 The present analysis is Socioeconomic status was assessed in terms based on death records incorporated from the of housing tenure. Council tenants, the low National Health Service central register, socioeconomic group, had poorer survival population characteristics from the 1971 census, than owner occupiers, the high socio- and cancer incidence from the National Cancer economic group, for the combined group of Registration Scheme. The Longitudinal Study all neoplasms, and for 11 out of 13 sample consists of 250 588 men and 262 484 neoplasms examined in males, and 12 out of women. During the years 1971-1981, 9196 men 15 neoplasms examined in females. and 8652 women were registered with cancer and Differences were found irrespective of age, of those, 7466 men and 6066 women died during cause of death and prognosis of the cancer. the follow up period (1971-1983). Survival analysis by length of follow up Study members were classified by housing indicated that council tenants were more tenure. Owner-occupiers (50% ofthe population) likely to present at a later stage than owner should be regarded as the "high" socioeconomic occupiers. status group and council tenants (30%/ of the http://jech.bmj.com/ Conclusions-Wide survival differentials population) as "low". Two smaller groups, those were observed between socioeconomic in privately rented accommodation and those in Department of have been excluded from Community Medicine, groups. Differences in survival for cancers institutions, University College ofpoor prognosis (eg, oesophagus, pancreas, comparisons here. London and lung) where treatment has little effect, Indirect methods of standardisation were used Middlesex School of to socioeconomic and Standardised Case Fatality Ratios (SFR) Medicine, Gower cannot be attributed Street, London WC1, differences in treatment. The survival were calculated using the case fatality rates of the on September 25, 2021 by guest. Protected copyright. United Kingdom differences for cancers of good prognosis whole Longitudinal Study population for the M Kogevinas corpus skin) could, in cancer in question as the standard. M G Marmot (eg, uteri, bladder, Office of Population part, be due to differences in treatment. It is Standardisation was carried out for age and period Censuses and probable that delay in seeking care is one of of follow up, to take into account possible time Surveys, Medical the major contributing causes. trends in incidence. Exact Poisson 95% Statistics Division, for Kingsway, London confidence intervals were estimated WC2 standardised ratios using computerised tables. A J Fox In the discussion of social inequalities in health For the life table analysis the SAS LIFETEST Social Statistics was Research Unit, City there has been much debate on the role of medical procedure used. University, London care. In England and Wales there are large EC1 socioeconomic differences in incidence and P 0 Goldblatt mortality from cancer. To understand the Results Correspondence to: potential importance of socioeconomic Both male and female council tenants had Dr Kogevinas, at Unit of differences in prompt detection and treatment of significantly worse survival than owner-occupiers Analytical Epidemiology, International Agency for cancer it is essential to have data on differences in when SFRs were calculated for all malignant Research on Cancer, 150 cancer survival. These have been examined less neoplasms (tables I and II). This overall figure is, Cours Albert-Thomas, 69372 Lyon Cedex 08, extensively than differences in cancer incidence. however, affected by the different distribution of France The first studies, conducted by Cohart in 1955, cancers among incident cases. Cancers with poor make a of all Accepted for publication detected an association between socioeconomic prognosis up larger proportion September 1990 status and cancer survival only for breast cancer. ' cancers in lower socioeconomic groups than in Socioeconomic differences in cancer survival 217 J Epidemiol Community Health: first published as 10.1136/jech.45.3.216 on 1 September 1991. Downloaded from

higher, especially in men. Among council tenants, Death certificates do not always show a person 48% of male cases and 20% of female were with a cancer to have died of this cancer. Patterns registered with lung, stomach, pancreatic, and observed for any cause of death were similar to oesophageal cancer, compared to 38% of male those based only on deaths from primary cancer. owner-occupiers and 16% in female. An analysis of socioeconomic differences in Of 13 sites examined in men, which comprised survival by age, showed that difference persisted 86% of the total number of cancer cases, council in all age groups. tenants had higher SFRs than owner-occupiers in 11 (table I). In women, council tenants had higher SFRs than owner-occupiers in 12 of the 15 sites Discussion examined, which comprised 85% of the total In this large, nationally representative study, number of cancer cases (table II). council tenants had higher case fatality rates than Among major sites in men, relative differences owner-occupiers in 11 out of 13 major cancers in in case fatality rates were more pronounced for men and in 12 out of 15 major sites in women. cancers of the colon, bladder and skin; the first of Less privileged socioeconomic groups have those significant at the p <005 level. Among usually been found to have worse survival than major cancers in women, the largest excesses in more privileged, although this has varied over case fatality among council tenants were for time and in different populations and cancer sites. pancreas, skin, corpus uteri, and bladder; In the OPCS Longitudinal Study population, significant for the first two. For prostate cancer relative differences were pronounced for cancers and for the leukaemias in men, and for stomach, of the testis, colon, bladder, and skin in men, and rectal, and breast cancer in women, owner- pancreas, skin, corpus uteri, and bladder in occupiers had higher case fatality rates than women. This accords, in part, with findings from council tenants. the SEER Program in the USA: the widest Table III shows median survival time for all survival differences between blacks and whites neoplasms, lung cancer and for cancers of good were for cancers of the bladder and the corpus prognosis, by housing tenure. Differences in uteri.2 6 In other studies socioeconomic length of survival were large for good prognosis differences in colon cancer are inconsistent,7-14 in cancers, especially cancer of the bladder, the cancer of the pancreas they are small,2 7 8 10 and corpus uteri, and malignant melanoma. Survival there are few published data on skin or testicular analysis by length of follow up for cancers where cancer.7 8 wide differences were observed indicated that In most studies, low socioeconomic groups owner-occupiers had lower case fatality rates than have worse survival for prostate cancer2 7-9 11 15 16 council tenants throughout long periods of follow or breast cancer, 7-9 11 14 15 17-20 contrary to the up. Case fatality rates were especially high for findings in the Longitudinal Study and three council tenants in the initial months of the follow other studies.'0 21 22 There are no obvious up period. reasons for this inconsistency. The higher case fatality of the lower Table I Standardised casefatality ratios (SFR) with 950o confidence intervals (CI) socioeconomic groups cannot be ascribed to for men, by socioeconomic status. Cancers are grouped by prognosis. differences in causes of death other than cancer. Housing tenure The socioeconomic differences were observed whether deaths from any cause, or only deaths

Owner-occupiers Council tenants http://jech.bmj.com/ from the primary cancer, were considered. Observed SFR Observed SFR (95% CI) Possible explanations of socioeconomic (950° CI) differences in survival All neoplasms 3131 92 2198 110 include: differences in (88-95) (105-114) treatment, stage of presentation, in host in tumour in Prognosis: 1-1 9% five year relative survival resistance, characteristics, and time Oesophagus 79 93 56 103 of diagnosis (lead time bias). We consider each in (74-117) (78-134) turn. Stomach 313 96 210 106 (86-108) (92-122) on September 25, 2021 by guest. Protected copyright. Pancreas 130 96 55 107 TREATMENT (81-115) (80-139 Lung 922 96 889 104 Survival differences in the Longitudinal Study (90-102) (97-111) were consistent and wide for the group of cancers Prognosis: 20-390/ five year relative survival with the best prognosis, eg, corpus uteri, testis, Colon 207 89 109 128 and (77-102) malignant melanoma, but were also evident Rectum (105-154) 175 92 111 109 for cancers with bad prognosis: oesophagus, (79-107) (90-132) pancreas, The Prostate 295 103 136 94 lung. large socioeconomic (92-116) (79-112) differences in survival for good prognosis cancers Lymphomasa 76 102 46 116 are as expected if differences in treatment (80-127) (85-154) are Leukaemias 86 105 35 96 playing a role. The differences in survival for poor (84-130) (67-134) prognosis cancers, where treatment has little Prognosis: 40-59% five year relative survival effect, suggest that socioeconomic differences in Bladder 173 91 122 111 treatment are far from being the most important (78-106) (92-132) factor. Prognosis: 60o0 + five year relative survival Larynx 26 97 29 119 (63-142) (80-171) STAGE AT PRESENTATION Other skin 183 91 103 114 (78-105) Variation in stage of cancer at presentation has Testis (93-138) 15 99 8 160 been identified as a contributing but not a (55-163) (69-315) complete cause of a survival differentials in SFRs for both groups are other above 100 because the third housing tenure group has low case fatality. studies. 2 6 7 10 16 17 23 The National Cancer 218 M Kogevinas, M G Marmot, A J Fox, P 0 Goldblatt J Epidemiol Community Health: first published as 10.1136/jech.45.3.216 on 1 September 1991. Downloaded from

Registration Scheme in England and Wales, and Table III Median survival time (years) for all hence the OPCS Longitudinal Study, does not neoplasms, lung cancer, and cancers of good prognosis by provide any information on stage. The high case socioeconomic status. Median survival is not age adjusted. fatality of council tenants in the very first period Housing tenure of follow up may indicate that compared to Owner-occupiers Council tenants owner-occupiers, a higher proportion of council (a) Men tenants presented at a late stage oftheir cancer. In All neoplasms 083 0*57 England and Wales, high socioeconomic groups Lung 027 025 may make more use of preventive health Prognosis: 40-59% five year relative survival services,24 and have better compliance with Bladder 3 52 2 81 cervical cancer screening.25 26 Diverse results Prognosis: 60% + five year relative survival Larynx 5 87 4 39 have been found for breast cancer and colorectal Other skin > 10 > 10 cancer screening.27-29 Testis > 10 > 10

HOST RESISTANCE AND TUMOUR CHARACTERISTICS (b) Women All neoplasms 290 1 83 Lower socioeconomic status may also be related to Lung 0 39 0 39 lower resistance affecting the rate of progression Prognosis: 40-59% five year relative survival of the disease.7'-2 Influences of host Breast 5 15 5 80 characteristics, eg, immune response and Cervix uteri 6 16 6-65 nutritional status, on the progression of a cancer Bladder 4 48 1-75 are well known. The degree to which such Prognosis: 60% + five year relative survival Malingnant melanoma > 10 6 25 characteristics influence socioeconomic Other skin > 10 > 10 differences in survival is still questionable. Corpus uteri > 10 7 19 Differences in tumour characteristics have been examined mainly among ethnic groups, eg, high socioeconomic groups could appear better different histological types for cancer of the not because prompt diagnosis altered the natural corpus uteri in black and white Americans,6 but history of the disease but simply because evidence is still limited. diagnosis has taken place earlier in the natural history of the disease-lead time bias. Such LEAD TIME BIAS information as is available on stage at presentation Diagnostic patterns have been shown to affect indicates that high socioeconomic groups are comparison of incidence and case fatality rates frequently diagnosed earlier and lead time could and validity of long term comparisons of survival therefore be one of the factors contributing to rates has been questioned.30 31 The survival of survival differentials.

Table II Standardised case-fatality ratios (SFR) with 95% confidence intervals (CI) CONCLUSIONS for women, by socioeconomic status. Cancers are grouped by prognosis. The prospective and record linkage character of Housing tenure the OPCS Longitudinal study provided the opportunity of examining cancer survival Owner-occupiers Council tenants differences in a large representative sample of the Observed SFR Observed SFR population of England and Wales. Wide (95% CI) (95% CI) differences were observed between socio- All neoplasms 2719 94 1601 105 http://jech.bmj.com/ (90-97) (101-111) economic groups. It is probable that delay in seeking care is one of the major contributing Prognosis: 1-19%O five year relative survival other Oesophagus 54 92 41 116 causes. Treatment, host resistance, and (70-121) (83-158) factors may also contribute. Stomach 182 102 126 96 (87-118) (80-115) Pancreas 96 96 39 145 We thank Mr M Rosato from the Social Statistics (78-118) (108-198) Research Unit, the City University, who extracted the Lung 265 94 220 106 (83-106) (93-122) tables used in this paper. on September 25, 2021 by guest. Protected copyright. (© Crown Copyright. Prognosis: 20-390o five year relative survival Colon 279 92 133 102 1 Cohart EM. Socioeconomic distribution of cancer of the (82-103) (85-121) female sex organs in New Haven. Cancer 1955; 8: 34-41. Rectum 136 104 67 85 2 Young JL, Gloeckler-Ries L, Pollack ES. Cancer patient (88-124) (66-108) survival among ethnic groups in the United States. J Natl Ovary 176 94 81 107 Cancer Inst 1984; 73: 341-52. (80-109) (85-134) 3 Fox AJ, Goldblatt PO. Longitudinal Study: socio- Lymphomas 58 93 34 116 demographic mortality differentials, 1971-1975. OPCS, (71-120) (89-174) Series LS No. 1. London: HMSO, 1982: 1-227. Leukaemias 72 96 33 115 4 Leon DA. Longitudinal Study 1971-1975. Social distribution (75-121) (79-161) of cancer. OPCS, Series LS No. 3. London: HMSO, 1988: 1-128. Prognosis: 40-590o five year relative survival 5 Kogevinas, M. Socio-economic differences in cancer survival, Breast 605 99 302 97 The OPCS Longitudinal Study 1971-1983. OPCS, Series (91-107) (87-109) LS No. 5. London: HMSO, 1990: 1-97. Cervix uteri' 72 95 72 97 6 Hankey BF, Myers MH. Black/white differences in bladder (75-120) (76-122) cancer patient survival. J Chron Dis 1987; 40: 65-73. Bladder 59 83 37 117 7 Berg JW, Ross R, Latourette HB. Economic status and (63-107) (82-161) survival of cancer patients. Cancer 1977; 39: 467-77. 8 Vagero D, Persson G. Cancer survival and social class in Prognosis: 60%o + five year relative survival Sweden. J Epidemiol Community Health 1987; 41: 204-9. Malignant melanoma 16 81 18 121 9 Lipworth L, Bennett B, Parker P. Prognosis of nonprivate (46-105) (72-191) cancer patients. J Natl Cancer Inst 1972; 48: 11-6. Other skin 125 88 71 125 10 Page WF, Kuntz AJ. Racial and socioeconomic factors in (73-105) (98-157) cancer survival. A comparison of veterans administration Corpus uteri 75 85 47 120 results with selected studies. Cancer 1980; 45: 1029-40. (67-107) (88-160) 11 Silman AJ, Evans SJW. Regional differences in survival a SFRs for both groups are below 100 because the third housing tenure group has high case fatality. from cancer. Community Med 1981; 3: 291-7. Socioeconomic differences in cancer survival 219 J Epidemiol Community Health: first published as 10.1136/jech.45.3.216 on 1 September 1991. Downloaded from

12 Wegner EL, Kolonel LN, Nomura AM, Lee J. Racial and 22 Keirn W, Metter G. Survival of cancer patients by economic socioeconomic status differences in survival of colorectal status in a free care setting. Cancer 1985; 55: 1552-5. cancer patients in Hawaii. Cancer 1982; 49: 2208-16. 23 Steinhorn SC, Myers MH, Hankey BF, Pelham VF. Factors 13 Chirikos TN, Homer RD. Economic status and associated with survival differences between black women survivorship in digestive system cancers. Cancer 1985; 56: and white women with cancer of the uterine corpus. Am 210-7. Epidemiol 1986; 124: 85-93. 14 Bonett A, Roder D, Esterman A. Determinants of case 24 Department of Health and Social Security. Inequalities in survival for cancers of the lung, colon, breast and cervix in health. Report of research working group (Black Report). South Australia. Med Aust 1984; 141: 705-9. London: DHSS, 1980. 15 Lipworth L, Abelin T, Connelly RR. Socioeconomic- 25 Cardiff Cervical Cytology Study. Enumeration and factors in the prognosis of cancer patients. Chron Dis definition of population and initial acceptance rates. 1970; 23: 105-16. 16 Polissar Epidemiol Community Health 1980; 34: 9-13. Dayal HH, L, Dahlberg S. Race, socioeconomic 26 Chamberlain J. Screening for early detection of cancer: status, and other prognostic factors for survival from general principles. In: Alderson M, ed. The prevention of prostate cancer. J7 Natl Cancer Inst 1985; 74: 1001-6. cancer. Edinburgh: Churchill-Livingstone, 1982: 227-58. 17 Bassett MT, Krieger N. Social class and black-white 27 differences in breast cancer survival. Am Public Health Hobbs P, Smith A, George WD, Sellwood RA. Acceptors 1986; 76: 1400-3. and rejectors of an invitation to undergo breast screening 18 LeMarchand L, Kolonel LN, Nomura AMY. Relationship compared with those that referred themselves. Epidemiol of ethnicity and other prognostic factors to breast cancer Community Health 1980; 34: 19-22. survival patterns in Hawaii. Natl Cancer Inst 1984; 73: 28 Calnan MW, Chamberlain J. Explaining participation in 1259-65. programmes for the early detection of breast cancer: a 19 Dayal HH, Power RN, Chiu C. Race and socioeconomic comparative analysis. Rev Epidemiol Sante Publique 1984; status in survival from breast cancer. J Chron Dis 1982; 35: 32: 376-82. 675-83. 29 Farrands PA, Hardcastle JD, Chamberlain J, Moss S. 20 Neale AV, Tilley BC, Vemon SW. Marital status, delay in Factors affecting compliance with screening for colorectal seeking treatment and survival from breast cancer. Soc Sci cancer. Community Med 1984; 6: 12-19. Med 1986; 23: 305-12. 30 Enstrom JE, Austin DF. Interpreting cancer survival rates. 21 Bain RB, Greenberg RS, Whitaker JP. Racial differences in Science 1977; 195: 847-51. survival of women with breast cancer. Chron Dis 1986; 39: 31 Doll R, Peto R. The causes of cancer. Oxford: Oxford 631-42. University Press, 1981. http://jech.bmj.com/ on September 25, 2021 by guest. Protected copyright.