J CUnEptdemtol Vol. 44, No. 8, pp. 807415, Printed in GreatBritain.All rightsreserved

08954356/91$3.00+ 0.00

1991

copyright0

1991 Pergamon Press

plc

THE ROLE OF SOCIOECONOMIC FACTORS IN THE SURVIVAL OF PATIENTS WITH COLORECTAL CANCER IN SAARLAND/GERMANY HERMANN BRENNER,‘*’ ANDREAS MIELCK,~* ROLAND KLEIN~ and HARTWIG ZIEGLER* ‘Forschungs- und Geschiftsstelle Epidemiologie der Universitat Ulrn, D-79 Ulm, *Saarllndisches Krebsregister, Statistisches Landesamt, D-6600 Sasarbriicken, ‘Institut fiir Medizinische Informatik und Systemforschung (MEDIS) der GSF, D-8042 Neuherberg and 4Kinderklinik der

Universitiit des Saarlandes, D-6680 Neunkirchen, Deutschland (Received in revised form 20 September

1990)

Abstract-The role of socioeconomic factors in the survival of patients with colorectal cancer was assessed using data from the cancer registry of Saarland/Germany, and census information. Among 2627 patients with colorectal cancer diagnosed from 1974 to 1983, patients from communities in the lowest of three categories defined by socioeconomic factors showed significantly lower survival rates than patients from other communities. After adjustment for potential biological and other sociogeographic risk factors in multivariate analyses, relative hazard of death associated with low socioeconomic status @ES) compared with high SES was estimated to be 1.22 (95% CI: 1.01-l .47) for colon cancer and 1.32 (95% CI: 1.09-l .60) for rectum cancer. The results are in agreement with earlier studies from North America, Hawaii and Sweden and indicate that an attempt to improve health care services and acceptance and possibly other relevant general living conditions in socioeconomically less privileged communities may be a rewarding approach towards increasing survival of patients with colorectal cancer.

Neoplasms Germany

Colorectal

Registries

INTRODUCTION

Studies on the role of socioeconomic factors in the survival of patients with colorectal cancer have yielded inconsistent results: Earlier findings of more favourable survival rates associated with higher socioeconomic status @ES) in studies from North America [l-3] could not

always be confirmed in more recent studies when other explanatory variables, such as race, age at diagnosis, stage, severity of disease and *Address all correspondence to: Andreas Mielck, GSFMEDIS, Ingolstidter LandstraDe 1, D-8042 Neuherberg, Deutschland. CE ws-F

807

Survival

Socioeconomic factors

initial course of treatment were controlled for simultaneously in multivariate analysis [4-6]. However, in a large study conducted in Hawaii, survival rates of patients with low SES were about 20% lower than survival rates of patients with higher SES for both colon and rectum cancer after adjustment for race, age, sex and stage, although these results just failed to reach statistical significance [4]. In a study from Ohio/U.S.A., Chirikos and Horner [A found a pronounced effect of economic status on survival of white male patients with digestive system cancer, particularly colorectal cancer, which was independent of

HEIUUANN BRENNER et al.

808

age, stage and initial course of treatment. They concluded that “differences in immunologic status, tumour characteristics and follow-up treatment may account for these economic effects” [7, p.2101. In later analyses on the same data, Homer and Chirikos [8] also found substantial survival differences for gastrointestinal cancer between patients from urban areas and those from rural areas, which persisted after adjustment for SES and a variety of demographic and medical explanatory variables. Survival differences related to social class were also found in a large study recently reported from Sweden for a variety of cancer sites including colon and rectum cancer, but stage at diagnosis, the most important predictor of prognosis, was not controlled for, and therefore the reasons for the observed survival differences were not clear [9]. To our knowledge, the role of socioeconomic factors in the survival of colorectal cancer has so far not systematically been investigated in populations from central Europe. With regard to health policy, social and regional differences in survival associated with stage at diagnosis and therapy, as well as differences which are independent of these factors, are of significance. In addition, it is important to know whether social factors offer a suited basis for regionalization in health care policy instead or along with merely administrative or geographical regionalization. In this study, which is based on data of the population-based cancer registry of Saarland/ Germany, the effects of social and regional factors on the survival of patients with colorectal cancer are simultaneously assessed in order to identify possible health policy approaches to lower mortality from colorectal cancer in a central European population.

MFTHODS

Within Germany, Saarland is the only state for which reliable population-based cancer registration has been carried out for the last two decades [lo, 111. Saarland covers a small, partly highly industrialized, though quite heterogeneous, area in the southwest of Germany. The total population is about 1.05 million persons. The population-based cancer registry of Saarland was established in 1966, and from the early 1970s on, it achieved high levels of internationally recognized completeness [lo]. Mortality follow-up of cancer patients is routinely done by sophisticated computer-assisted record linkage with death certificates [l 11. Our analyses are based on cases of malignant tumours of the colon and rectum (ICD-9 positions 153 and 154) first diagnosed between 1 January 1974 and 31 December 1983 in patients aged 45-74 years at the time of diagnosis. The reasons for excluding patients above age 74 years at the time of diagnosis are the increasing unreliability of cancer registration as well as the increasing rates of competing causes of death in the elderly. Colorectal cancer before age 45 years, on the other hand, is extremely rare. Patients whose cancer was not histologically verified or for whom no TNM classification of stage at diagnosis, the most important predictor of prognosis, could be determined were also excluded. All records were made anonymous, reviewed and coded in a standardized way by one single physician (RX.) who was unaware of the social status and place of residence of the patients. Regional and social factors were measured at the community level (50 communities) based on the 1970 census and an additional investigation of area-use carried out in 1981. Two distinct social variables identified by factor

Table 1. Factor analysis of socioeconomic variables Factor loadings

Residential area per inhabitant Traffic area (area of streets, roads, places) per inhabitant Proportion of Catholics Proportion of inhabitants with no more. than 9 yr school education Proportion of blue collar workers among inhabitants aged 15-65 yr Mean number of persons per household Source: [12].

Factor 1 “SES”

Factor 2 “Urbanity”

-0.34 -0.33

-0.80 -0.68

-0.46

-0.13

-0.71

0.15

-0.97 -0.58

0.12 -0.63

Socioeconomic Factors in the Prognosis of Colored

analysis [ 121have been considered (see Table 1): the first factor, which will be referred to as socioeconomic status (SES) from now on, is mainly characterized by the proportion of blue collar workers among persons aged 15-65 years and the proportion of persons with no more than 9 years schooling; the second factor, denoted “urbanity”, is largely characterized by residential area density, traffic area density and mean number of persons per household. Patients were a priori classified by three approximately equally-sized categories (highmiddl+low) of these social factors according to their community of residence at the time of diagnosis. “Natural breaks” in the factor values were used as cutpoints. Furthermore, combinations of the six administrative districts of Saarland were used as the basis for a regional factor defined by community of residence (Appendix A): Region 1 comprises the metropolitan area of Saarbrticken, the capital and by far the largest community in Saarland (about 200,000 inhabitants); Region 2 is the southeast of Saarland, which includes the only medical school (Homburg/Saar) and university hospitals of Saarland; and Region 3 covers mainly rural areas in the north and west. Appendix B displays the numbers of inhabitants as well as the regional and social variables for the 50 communities of Saarland. In addition, the following individual-level categorical predictor variables were considered: Sex Tumor site (categories: colon = ICD-9 posn 153, rectum = ICD-9 posn 154) Age at diagnosis (categories: 45-54, 55-64, 65-74) Stage at diagnosis (categories: localized = Dukes A or B, regional = Dukes C, distant = Ml) Calendar period of diagnosis (categories: 19741978, 1979-1983). Univariate analyses were carried out using life table methods [13]. The Cox proportional hazards model [14] was used for simultaneous adjustment for multiple predictor variables including the assessment of interaction effects. Categories of both nominal and ordinal variables were represented by dummy variables in multivariate analyses in order to avoid unjustified assumptions on the shape of predictor-hazard relationships.

Cancer

809

RESULTS

Overall, 6420 cases of colorectal cancer were registered from 1974 to 1983. 2315 cases were excluded due to the restrictions on age, another 1478 cases were excluded due to missing histological verification or TNM classification. Among the remaining 2627 patients, there were 1465 persons with colon cancer (622 men, 843 women) and 1162 persons with rectum cancer (616 men, 546 women). Figure I shows the unadjusted survival curves for all patients by categories of SES and stage at diagnosis. As expected, stage at diagnosis was found to be a most powerful determinant of prognosis. However, for each stage at diagnosis, prognosis was worse for patients from low SES communities compared with patients from middle or high SES communities. Table 2 summarizes the results of basic univariate survival analyses. The median survival time for patients from low SES communities was 28 months compared with 37 and 39 months for patients from high and middle SES communities (p-value for log rank test x0.05). Site of tumour, sex, age, stage and year of diagnosis also appeared to be important predictors of prognosis. In order to identify the potential for confounding, the distribution of potential biological and sociological predictor variables by levels of socioeconomic factors is given in Table 3. No meaningful association could be observed between SES and site of tumour, sex, age and year of diagnosis. However, there was an interesting and important association between SES and stage at diagnosis: distribution of stage at diagnosis was less favourable in patients from high SES communities compared with patients from both middle and low SES communities. SES, urbanity and region are also correlated. Table 4 shows the results of the multivariate survival analysis, in which all of the included variables are adjusted for each other. Due to violation of the proportional hazards assumption, tumour site and stage at diagnosis were controlled for by blocking. Relative risk estimates for categories of SES were very similar for single categories of the control variables, the tests for interaction of SES with other variables did not reveal any significant pattern. Patients from low SES communities had a clearly higher hazard of death than patients from high SES communities for both colon

810

HEMANNBWNNERer al.

1.0

Local stage

Regional

stage

b

0.9 0.8

0.8 0.7

% z

0.7

a 0.6 9" 2 n 0.5 2 '2 0.4

SES=low

cz 0.3

0.3

0.2

0.2

-.-._ .-.-.-.SES=low

0.1

0.1 1

I 0

I

12 24

I

I

I

I

I

36

46

60

72

64

Months

1.0 r

I

I

96108

J 0

120

I

I

II

12

24

36

of follow-up

Distant

48

Months

1.0

stage

0.9

0.9

0.8 t

0.8

I

I

I

60

72

84

I

I

96106

I 120

of follow-up

All stages

0.7 S+ *g 2 0.6 n 2 P 0.6 2 '2 0.4 z

SE.S=low

0.3 0.2 0.1

L 0

11 12

24

I

I

36

48

Months

11 60

1 72

04

11

I

96108

120

of follow-up

I 0

12

I I I I I I I I 24

36

48

Months

60

72

84

96108

I 120

of follow-up

Fig. 1. Unadjusted survival curves of colorectal cancer patients by stage of diagnosis and SES. Saarland, ICD-9 positions 153 + 154, 1974-1983.

(+22%) and rectum (+ 32%) cancer, after adjusting for all other variables. Differences between high and middle SES communities were smaller and not statistically significant. Furthermore, prognosis was poorer for males than for females and for the elderly compared with 45to 54-year-old patients. As results were quite

consistent for both rectum and cancer patients, a pooled analysis including all patients simultaneously was also performed in order to obtain more precise estimates. Figure 2 shows theoretical survival curves for colorectal cancer patients by SES after simultaneous adjustment for all other variables.

Socioeconomic Factors in the Prognosis of Colorectal Cancer Table 2. Univariate survival analysis-median survival time for levels of predictor variables; patients with colorectal cancer (ICD-9 positions 153 and 154) Saarland, 19741983 Predictor variable SES High Middle Low Tumour site Colon Rectum Sex Men Women Age olr) 45-54 55-64 65-74 Stage Local Regional Distant Year of diagnosis 1974-1978 1979-1983 Urbanity Urban Middle Rural Region Saarbrilcken area Northwest Southeast

Median survival time (months)

p-Value for log rank test

31 39 28

Germany.

The role of socioeconomic factors in the survival of patients with colorectal cancer was assessed using data from the cancer registry of Saarland/Germ...
776KB Sizes 0 Downloads 0 Views