A large number of studies have shown that overall and cancer-specific survival is better in individuals with high socioeconomic status (SES) (Bentley et al, 2008; Byers et al, 2008; Halmin et al, 2008; Harris et al, 2009; Aarts et al, 2010; Cavalli-Bjorkman et al, 2011). Socioeconomic status is usually defined by factors such as maximum attained level of education, income (own/family's/neighbourhood's), and overall standard of living (Blais et al, 2006). Although strictly not contributing to SES, social network has been studied in this context and patients with few friends and family members have been shown to have shorter cancer survival than those with a large social sphere (Goodwin et al, 1987; Reynolds et al, 1994; Datta et al, 2009; Chou et al, 2010). In controlled studies on cancer survival, the relative risk for mortality of married respondents was 12% lower than that of unmarried persons (Pinquart and Duberstein, 2010).

Reasons for poorer outcome in socially deprived patients might be more advanced disease at presentation, comorbidity and treatment-related morbidity (Frederiksen et al, 2009; Hines et al, 2009; Fusco et al, 2010). Difficulty in accessing health care may also be a factor (Lejeune et al, 2010; Sankaranarayanan et al, 2010; Baade et al, 2011). Proposed explanations for the importance of social support range from psychosocial ones (more meaningful to prolong survival when surrounded by loved ones) to the benefit reaped from practical help during cancer treatment. There is also a possibility that patients with low SES and/or poorer social support are offered different treatment for cancer than more privileged patients (Dobie et al, 2006; Villingshoj et al, 2006; McBride et al, 2010; van Steenbergen et al, 2010). If patients' SES or social situation determines which treatment they are offered the reasons for this need to be more fully understood. In the present study, the setting is Scandinavia, where healthcare is tax financed and freely available to all citizens regardless of social standing. Cancer treatment is exclusively administered at government-funded hospitals and not at private clinics. Treatment guidelines are regularly updated to promote fair and equal treatment for patients of different geographic regions. If treatment were to vary by social standing in Scandinavia, it could mean that doctors take the patient's SES and/or social circumstances into account when designing cancer treatment (Lemmens et al, 2005; Tilney et al, 2009).

The aim of the present study was to explore treatment and survival in a Scandinavian, population-based, cohort of metastatic colorectal cancer (mCRC) patients stratified for indicators of SES and social support.

Materials and methods

On 1st October 2003, prospective registration of patients with mCRC referred to oncology units of three university hospitals was initiated. The aim was to use hospitals which received all referrals to an oncology department of that region, that is, hospitals in regions where there was only one oncology department providing radiation therapy and medical treatment for adult cancer patients. Uppsala University Hospital in Sweden serves a region of 280 000 inhabitants, Odense University Hospital in Denmark serves 475 000 and Haukeland University Hospital in Norway 450 000 people. All patients in these three areas who need oncologic treatment will be referred to the only university hospital in that region. A cohort consisting of all patients who were initially diagnosed with or who developed metastases from CRC and who were referred to these oncology departments until 1st August 2006 was created.

The cohort consisted of 781 patients, 296 from Norway, 323 from Denmark and 162 from Sweden. All patients had stage IV histologically confirmed adenocarcinoma of the colon or rectum. They completed questionnaires regarding medical history, family structure, maximum level of attained education and smoking habits at the first consultation. We prospectively recorded clinical characteristics, comorbid conditions, routine blood tests, treatment offered and whether the patient was included in a clinical trial or not. Missing data has been retrospectively retrieved from the patients' files. The cohort was originally set up with the primary intent to study trial inclusion, and the results from this project have been previously published (Sorbye et al, 2009).

Socioeconomic status and social support

Disease-related factors were tested against five variables: marriage/cohabitation with a partner, having children, living alone, smoking and maximum level of education. The first four variables were answered by either ‘yes’ or ‘no’, whereas educational level was split in two groups: maximum level 12 years of school (non-university), and university or the equivalent. In Scandinavia cohabitation, without being married, is common and for the purpose of this study cohabitation with a partner is considered equal to marriage. Married patients were further divided by sex, as it had been reported that men and women benefit differently from the educational level of their spouse (Jaffe et al, 2006).

Other variables

Co-morbidity was sub-classified into diabetes, cardiovascular, pulmonary and psychiatric, but for the purpose of these analyses only division into comorbidity/ no comorbidity was used. In a Dutch colon cancer study, further subdivision of comorbidity according to a modified Charlson classification did not improve the interpretation of comorbidity data (Lemmens et al, 2005). The patients were divided into two groups: those with metastatic disease at presentation (synchronous metastases) and those who had developed metastases after initially having had localised disease (metachronous metastases). If metastases were found within 90 days of the initial diagnosis, the patient was considered to have synchronous metastases. At the patients' first consultation, performance status (PS) was scored using the WHO classification and patients were divided into four groups: PS 0, 1, 2 and 3–4. The latter were grouped together as they constitute patients generally considered ineligible for palliative chemotherapy. Combination chemotherapy, that is, an irinotecan or oxaliplatin combination with 5-fluorouracil (5-FU), was administered to 350 patients (Table 1). Bevacizumab and cetuximab were grouped together under ‘antibody treatment’. Some cohort patients (n=55) were included in the randomised Nordic VII trial, where 2 out of 3 of patients were given cetuximab as first-line treatment. As trial inclusion was a factor to be analysed, we chose to limit the antibody group to patients treated outside of a clinical trial. Patients over the age of 75 were excluded from analyses on antibody treatment and secondary surgery as novel treatments were rarely used for elderly patients during this time frame. Data on age and comorbidity were used to adjust results.

Table 1 Patient characteristics of all registered patients (n=781)

Statistical methods

As the numbers of events in some categories were few, statistical analysis had to be adjusted. Combination chemotherapy, administered to 350 patients, was compared with the 100 patients receiving only 5-FU. The smallest group in this category limited the statistical model, which led to omission of analysis on possible effects on marital status. The smallest group treated (49 patients) was the group who had secondary surgery for metastases and the number of factors, which we could test with accuracy, were calculated to be three. For multivariate analyses in Table 2, the analyses for secondary surgery had to be limited to living alone, having children and educational level. Logistic regression was used for dichotomous and ordinal logistic regression for ordinal outcome variables, with multiple imputations to reduce loss of data due to missing values in the covariates. Two group comparisons were performed using exact chi-square tests for dichotomous or nominal variables, exact Mann–Whitney tests for ordinal variables, and Student t-tests for continuous variables. Overall survival analyses for all patients were performed for the time interval from the date of non-resectable metastatic disease to the date of death or censored if the patient was alive on 15 August 2008, using the Kaplan–Meier method with log-rank tests for univariate and Cox regression for multivariate analyses. Overall, rather than cancer-specific, survival was opted for since it was of interest to study influence of comorbidity on treatment and survival. A separate Cox regression analysis was done to deduce possible confounding effects of high age and comorbidity on survival in patients living alone. The proportional hazard assumptions were checked as recommended by Therneau and Grambsch, 2000.

Table 2 Multivariate analyses (ordinal regression) of disease characteristics and treatment by education and family structure

All analyses were performed with the statistical programme: R version 2.10.0 (The R Foundation for Statistical Computing, Vienna, Austria), using the R package rms for regression analyses. The function aregImpute in the R package Hmisc functions was used for multiple imputations (200 imputed data sets).

All statistical tests were two tailed, using the 5% significance level.


Patient characteristics

Baseline patient characteristics of all registered patients in the cohort are shown in Table 1. As the cohort was population-based and patients were consecutively included, a large percentage of patients were older and had poor performance status. More than half of the patients had significant comorbid conditions, as judged by the attending physicians.

Disease characteristics and treatment

Table 2 shows results of multivariate analyses with adjustment for age and comorbidity. Patients who lived alone more often presented with metastatic disease than those who lived with someone (OR 2.17, 95% CI 1.32–3.55) and were less often treated with combination chemotherapy (OR 0.19, 95% CI 0.04–0.85) and secondary surgery of metastases (OR 0.29, 95% CI 0.10–0.86). Patients with university education more often had surgery of metastases than patients with non-university education (OR 2.22, 95% CI 1.10–4.49). Smokers had poorer PS than non-smokers at baseline (OR 1.65, 95% CI 1.19–2.29). No differences in disease stage, PS and treatment given were seen with respect to marital status and having children. With the exception of surgery for liver or lung metastases, educational attainment did not influence treatment.


Median survival times stratified by educational level, comorbidity and indicators of social support are shown in Table 3. Differences in overall survival were seen between patients living alone and those living with someone (7.7 vs 11.7 months, P<0.001). Similar survival differences were seen between patients with and without comorbidity (8.0 vs 12.5 months, P=0.01). No significant differences in survival were seen with regard to smoking, marital status, educational level and having children. Cox regression analysis showed that living alone (HR 1.19, 95% CI 1.01–1.41, P=0.03) and higher age (HR 1.03, 95% CI 1.02–1.03, P<0.001) were independent prognostic factors for survival, whereas comorbidity was not. When checking proportional hazard assumptions, the global test showed significant deviations (P<0.001). No deviations were indicated for living alone (P=0.208), but for both adjustments variables age (P<0.001) and comorbidity (P=0.012). Plots of Schoenfeld residuals indicated that age may be less influential as the disease progresses while comorbidity may be a risk factor around the time of diagnosis of mCRC.

Table 3 Median overall survival in different patient groups


In this population-based mCRC cohort, patients who lived alone had more advanced disease at presentation were less likely to receive potent combination chemotherapy regimens and had poorer survival than those who lived with someone. As data has been adjusted for age and comorbidity, it is not possible to explain the above by hypothesising that patients who live alone are old and frail with poor PS. Reports that focus on treatment of cancer patients who live alone are scarce. In an American study on survival of head–neck cancer patients, men without a partner were found to have considerably shorter survival even after controlling for disease variables (Konski et al, 2006). Other studies have found that patients with Non-Hodgkin lymphoma (Frederiksen et al, 2011) and lung cancer (Dalton et al, 2011) present with later stages of their respective diseases if they live alone than if they live with someone. To our knowledge, it has not been shown previously that lowered treatment intensity might be a contributing factor to the shorter survival of patients who live alone. This is of concern because of the large number of people who live by themselves – a phenomenon which is becoming increasingly common, especially in urban areas.

It is reasonable to speculate that the absence of a partner can delay diagnosis as the patient lacks the social support needed to actively pursue medical help. It is also possible that these patients may be, consciously or subconsciously, judged by physicians as being without the support to qualify them for more intensive oncologic treatment. Knowing that patients have strong networks enables clinicians to risk treatment with potentially serious side-effects. Further, the choice between undergoing potentially life-prolonging treatment while sometimes enduring considerable toxicity, or declining treatment, has to be made by the patient. We cannot rule out the possibility that the will to have such treatment is stronger in patients who live with someone than it is in those who do not.

Colorectal cancer treatment has developed immensely in the last decade. In the 1980s it was not certain that mCRC patients had any benefit from chemotherapy at all. A few studies subsequently showed that 5-FU alone gave slightly prolonged survival, and in the 1990s addition of oxaliplatin and irinotecan improved survival even more (Glimelius and Cavalli-Bjorkman, 2012). As it is clear that combining 5-FU with oxaliplatin or irinotecan prolongs survival, we must see a connection between patients living alone receiving less combination chemotherapy and having shorter survival.

During the time frame that the cohort in this study was set up, targeted drugs in the form of antibodies were tested, mainly within the confines of clinical trials. The three Scandinavian countries have very similar treatment guidelines for mCRC patients. If the patient's PS is good enough patients in all three countries will be recommended to receive sequential treatment of 5-FU, oxaliplatin and irinotecan, in different combinations and different orders, provided toxicity is acceptable and the disease responds favourably. In some instances, antibody will be added. The only difference between the countries is that Norway does not generally recommend palliative chemotherapy for patients over the age of 75. In our cohort, patients over this age were excluded from analysis and results were thus not affected by this.

Marital status has previously been studied and as a marker for social support some groups have found advantages for married patients with regard to cancer survival (Patel et al, 2010). We were not able to verify this in our material. However, being married is not always the same as living with someone. In large register studies, married patients may be perceived as having good family support. In our cohort, some patients were married but lived apart as the spouse was in a nursing home. We also included patients living with their adult children into the group of cohabitating patients as we felt this reflected a dimension of social support. An individual evaluation of the patients' living arrangements, as was possible in this prospectively collected cohort, is likely to be a more accurate way to study the effect of social support than to simply look at marital status.

Having children can mean different things to different people – when relationships with offspring are close it means increased social support, but not all people remain in contact with adult children. Because of this discrepancy, having children, as an SES parameter, is probably not very reliable in indicating good network and support. Having a live-in-partner is more likely to be indicative of solid family structure in this sense.

Maximum attained educational level has traditionally been used as a strong indicator of SES. In the Scandinavian countries it is more likely to reflect SES than income or area of residence and has been used in a number of studies on cancer and SES (Eaker et al, 2009; Berglund et al, 2010). When using social databases to extract information on occupation, employment status, income and socioeconomic index, all factors have followed the trend set by educational level (Eaker et al, 2009; Cavalli-Bjorkman et al, 2011). Register studies in Scandinavia and elsewhere have shown that there are differences in diagnostic activity, treatment and survival by educational level (Halmin et al, 2008). One major drawback in a number of these studies has been the lack of information on comorbidity (Halmin et al, 2008; Cavalli-Bjorkman et al, 2011). There is strong evidence that the prevalence of comorbid conditions is higher in groups with lower SES (Louwman et al, 2010). Many studies on SES and cancer have been inconclusive as data could not be adjusted for comorbidity. Interestingly, in the smaller cohort of patients in the present study we could not verify differences by educational level with the exception of secondary surgery for distant metastases. It is possible that this is because we have the advantage of detailed information on patients' comorbidity and we have been able to adjust for this. We have, on 773 patients in the cohort (99%), scored and sub-classified comorbidity from an oncologists' point of view. As some comorbidity contraindicates chemotherapy while other diseases do not, this is an asset in the present study.

Compared with other countries' treatment frequencies, some of the percentages in this study can seem low. Secondary curative surgery, for instance, is usually reported to be performed in 10–15% of the patients, whereas in our study the number was 6%. As this cohort is population based and includes all consecutive patients with mCRC, it well represents the general population of stage IV CRC patients. In this cohort treatment and outcome are poorer than in clinical trials, where patients often are highly selected. Almost 20% of patients presented with PS 3 or worse, automatically disqualifying them from chemotherapy, and many patients were elderly. Overall survival is also much shorter than in other patient series. Again this reflects the population-based nature of the patient material; overall survival in patients treated with combination therapy or in a clinical trial was similar to all other series, or close to 2 years (Sorbye et al, 2009). Bevacizumab treatment was quite rare in this cohort. This is best explained by the clinical trials that were ongoing during the years 2003–2006. The trials, in all three countries, recruited more patients to treatment with cetuximab than bevacizumab. During this time frame patients were not selected for treatment with cetuximab based on their K-ras-status.

In Scandinavia regional oncologic centres work towards equal treatment for all patients regardless of background and place of habitation. National guidelines approved by the National Health authorities for both adjuvant and palliative treatment are produced in conjunction with these centres. This may in itself lessen the risk of unequal treatment, and many clinicians feel that inequality in the public health care sector is not a prevailing problem. We did not find significant differences between patients of different educational level and social situation with the exception of patients living alone. This is an interesting finding which deserves attention as it may be a factor which subconsciously influences the decisions made by oncologists for cancer patients. If patients can be offered better support, it may be possible to increase treatment intensity and prolong survival in this group.