Socioeconomic position and survival after lung cancer: Influence of stage, treatment and comorbidity among Danish patients with lung cancer diagnosed in 2004–2010

Abstract

Background. To address social inequality in survival after lung cancer, it is important to consider how socioeconomic position (SEP) influences prognosis. We investigated whether SEP influenced receipt of first-line treatment and whether socioeconomic differences in survival could be explained by differences in stage, treatment and comorbidity.

Material and methods. In the Danish Lung Cancer Register, we identified 13 045 patients with lung cancer diagnosed in 2004–2010, with information on stage, histology, performance status and first-line treatment. We obtained age, gender, vital status, comorbid conditions and socioeconomic information (education, income and cohabitation status) from nationwide population-based registers. Associations between SEP and receipt of first-line treatment were analysed in multivariate logistic regression models and those with overall mortality in Cox regression models with stepwise inclusion of possible mediators.

Results. For both low- and high-stage lung cancer, adjusted ORs for first-line treatment were reduced in patients with short education and low income, although the OR for education did not reach statistical significance in men with high-stage disease. Patients with high-stage disease who lived alone were less likely to receive first-line treatment. The socioeconomic difference in overall survival was partly explained by differences in stage, treatment and comorbidity, although some differences remained after adjustment. Among patients with high-stage disease, the hazard ratio (HR) for death of those with low income was 1.12 (95% CI 1.05–1.19) in comparison with those with high income. Among patients with low-stage disease, those who lived alone had a 14% higher risk for dying (95% CI 1.05–1.25) than those who lived with a partner. The differences in risk for death by SEP were greatest in the first six months after diagnosis.

Conclusion. Socioeconomic differences in survival after lung cancer are partly explained by social inequality in stage, first-line treatment and comorbidity. Efforts should be made to improve early diagnosis and adherence to first-line treatment recommendations among disadvantaged lung cancer patients.

As elsewhere, the prognosis of lung cancer is poor in Denmark, with a five-year survival of 12% in 2012. The incidence of this cancer is associated with socioeconomic position (SEP) [1], but relatively few studies have investigated the association between SEP and lung cancer survival [2–7]. In a nationwide Danish study, male lung cancer patients with high SEP, measured as education, income, affiliation to the work market and housing size, had a short-term survival benefit over men with lower SEP [8]. This study did not, however, include information on stage, treatment or the health status of the patients, all of which are potentially influenced by SEP [9]. In further investigations based on data from a nationwide clinical lung cancer database, short education and living alone were associated with a higher risk for advanced-stage non-small cell lung cancer (NSCLC) [10], whereas early-stage NSCLC patients with low income or living alone were less likely to receive surgical treatment even after stage, age, hospital capacity and comorbidity were taken into account [11]. In a recent systematic review and meta-analysis, lung cancer patients of low SEP were less likely to receive surgery and chemotherapy in both universal and non-universal healthcare systems, regardless of stage or histology [12]. Thus, in addition to earlier diagnosis, better general health status and access to treatment may explain some of the observed better short-term survival of socially advantaged lung cancer patients. We therefore investigated the association between SEP, measured as education, income and cohabiting status, on receipt of first-line treatment and overall survival in a population-based cohort of all patients with lung cancer diagnosed in Denmark in 2004–2010.

Material and methods

The Danish Lung Cancer Registry was established in 2001, and, since 2003, registration covers more than 90% of lung cancer cases in Denmark [13]. From the files of the Registry, we identified 33 834 people born in 1920–1978 in whom lung cancer was diagnosed in 2004–2010, when Eastern Cooperative Oncology Group performance status and treatment were registered. Information on histology, stage and performance status was available for 19 281. We excluded 5076 patients with performance status > 2.

Stage and first-line treatment

We grouped the lung cancers into low and high stage according to the TNM classification. Low stage was defined as T1-3N0M0 or T0-3N1M0, thus including a single category from stage IIIa, and high stage as T0-3N2-3M0, T4N0-3M0 or TxNxM1. We defined first-line treatment according to TNM status and performance status as shown in Appendix, to be found online at http://informahealthcare.com/doi/abs/10.3109/0284186X.2014.1001037 and categorised receipt of first-line treatment as a binary variable (yes, no).

Socioeconomic and demographic factors

By means of the unique personal identification number issued to all residents of Denmark, we obtained information on the education, income and cohabitation status of the lung cancer patients by data linkage to the population-based registers on education and income maintained by Statistics Denmark [14,15]. Highest attained educational level was categorised into short (i.e. mandatory education of up to 7 or 9 years for people born before or after 1 January 1958, respectively), medium (between 8–10 and 12 years, last grades of primary, secondary and vocational education) and higher education (> 12 years).

Disposable income was defined as household income after taxation and interest per person, adjusted for the number of people in the household and deflated according to the 2000 value of the Danish crown according to a formula from the Danish Ministry of Finance. We categorised disposable income according to quartiles of the national gender-specific disposable household income per person, into low (1st quartile), medium (2nd–3rd quartile) and high (4th quartile). Cohabitation status was defined as living alone or with a partner, irrespective of marital status.

Patients for whom there was no information on any socioeconomic variable were excluded (912 people), as were 248 patients aged < 30 or ≥ 85 years at the time of lung cancer diagnosis, who might either still have been in the educational system or for whom information on education was missing. A total of 13 045 patients were included in the analyses.

Comorbid disorders

The Charlson Comorbidity Index [16] provides an overall score for comorbidity on the basis of composite values weighted by level of severity assigned to 19 selected conditions, with scores from 1 to 6; the scores were grouped as 0, 1, 2 and ≥ 3. The comorbidity index was generated by linking the personal identification number of each person to the files of the Danish National Patient Register for a full history of diseases leading to hospitalisation from 1978 and, from 1995, outpatient visits up to the year preceding the lung cancer diagnosis. The Register includes dates of admission and discharge and diagnoses coded according to Danish modified versions of the International Classification of Diseases (ICD) version 8 (ICD-8) and, from 1994 onwards, ICD-10 [17].

Statistical methods

All analyses were performed separately for patients with low- and high-stage cancers, as the influence of SEP might differ according to the stage of disease and the expected outcome of treatment.

Multiple logistic regression models were used to examine the influences of education, income and cohabiting status on the likelihood of receiving first-line treatment, measured as odds ratios (ORs) and respective 95% confidence intervals (CIs). To account for possible clustering in hospital wards, we used generalised estimating equations with the exchangeable working correlation structure and robust variance estimates.

A two-step procedure was used. In the first model, each socioeconomic variable was entered alone, with adjustment for age, period and gender. In the second model, individual socioeconomic variables were additionally adjusted for variables further upstream in the causal pathway (cohabiting status was adjusted for education and income, and income was adjusted for education) and for comorbidity.

Tests for interaction (effect modification) between covariates were performed with the Wald test statistic. Interactions between education and gender, comorbidity and age, respectively, as well as between comorbidity and gender and age, respectively, were investigated. As we found an interaction between education and gender in patients with high-stage cancer, separate analyses were performed by gender.

Survival time was defined as the time between diagnosis and death from any cause (overall survival) or the end of follow-up (31 December 2011), whichever occurred first. To assess the influences of education, income and cohabitation status while adjusting for potential confounders (age, gender and period) and further stepwise adjustment for potential mediating factors (stage at diagnosis, receipt of first-line treatment, comorbidity and performance status), we used multiple Cox proportional hazards models to estimate hazard ratios (HRs) and 95% CIs for overall mortality. Separate analyses were conducted for patients with low- and high-stage cancer and, in further models, also by time since diagnosis into 0–½, ½–1, 1–2, 2–3 and ≥ 3 years. Tests for interactions revealed no statistically significant interactions. Statistical analyses were performed with SAS 9.3. A p-value less than 0.05 was considered significant, and all statistical tests were two-sided.

Results

The mean age of the 13 045 lung cancer patients was 68 years (5–95% range, 51–81), and the distribution of education, income and cohabitation status was similar in patients with low- and high-stage cancer, as were most other demographic or clinical variables, except for performance status: 12% of patients with low-stage cancer and 23% of patients with high-stage cancer had a score of 2 (Table I). Patients with lower SEP were less likely to receive first-line treatment (Table II). In general, adjustment for comorbidity influenced the ORs only slightly. Adjusted ORs for receipt of first-line treatment were statistically significantly lower in patients with low-stage cancer with short education (OR, 0.76; 95% CI 0.61–0.95) than in those with higher education. Female patients with high-stage cancer and those with medium or short education had statistically significantly lower adjusted ORs than patients with higher education (0.71; 95% CI 0.57–0.89 and 0.76; 95% CI 0.62–0.94, respectively). The ORs for patients with low income in both stage groups were statistically significantly lower (OR, 0.80; 95% CI 0.64–0.99 and 0.84; 95% CI 0.73–0.96, respectively) than those with a high income. Receipt of first-line treatment was statistically significantly less likely among patients with high-stage cancer who lived alone (OR, 0.83; 95% CI 0.79–0.88) (Table II).

Table I. Descriptive characteristics of 13 045 patients aged 30–84 years with lung cancer diagnosed 2004–2010 in Denmark.

	Low-stage lung cancer		High-stage lung cancer
Characteristic	N	%	N	%
Age at diagnosis (years)
30–54	320	8	946	10
55–64	1017	27	2674	29
65–74	1493	40	3461	37
75–84	944	25	2190	24
Gender
Male	2026	54	4942	53
Female	1748	47	4329	47
Period of diagnosis
2004–2005	618	16	1340	14
2006–2007	1150	30	2996	32
2008–2009	1281	34	3254	35
2010	725	19	1681	18
Education
Short	1525	40	3587	39
Medium	1780	47	4518	49
Higher	469	12	1166	13
Income
Low	1179	31	2808	30
Medium	1981	52	4919	53
High	614	16	1544	17
Cohabitation status
Living with partner	2464	65	6058	65
Living alone	1310	35	3213	35
Charlson Comorbidity Index score
0	1637	43	4507	49
1–2	1475	39	3378	36
≥ 3	662	18	1386	15
Performance status
0	2001	53	3098	33
1	1337	35	4023	43
2	436	12	2150	23
Histology
Non–small cell lung cancer	3559	94	7714	83
Small cell lung cancer	215	6	1557	17
Stage
Ia	1186	9
Ib	1587	42	–
IIa	101	3	–
IIb	781	21	–
IIIa	119	3	1094	12
IIIb			2773	30
IV			5404	58

Table II. Odds ratios for receipt of first-line treatment by stage group in 13 045 patients with lung cancer diagnosed 2004–2010 in Denmark.

	Low stage				High stage
	OR^a (95% CI)		OR^b (95% CI)		OR^a (95% CI)		OR^b (95% CI)
Education					Men
Short	0.75	(0.60–0.94)	0.76	(0.61–0.95)	0.87	(0.73–1.04)	0.88	(0.74–1.06)
Medium	0.90	(0.74–1.09)	0.90	(0.74–1.09)	0.98	(0.85–1.15)	1.00	(0.86–1.15)
Higher	1		1		1		1
					Women
Short					0.74	(0.61–0.89)	0.76	(0.62–0.94)
Medium					0.70	(0.56–0.87)	0.71	(0.57–0.89)
Higher					1		1
Income
Low	0.73	(0.59–0.91)	0.80	(0.64–0.99)	0.79	(0.69–0.91)	0.84	(0.73–0.96)
Medium	0.86	(0.68–1.07)	0.91	(0.73–1.15)	0.89	(0.78–1.01)	0.94	(0.82–1.07)
High	1		1		1		1
Cohabitation status
Living alone	0.95	(0.82–1.10)	0.98	(0.84–1.15)	0.82	(0.77–0.87)	0.83	(0.79–0.88)
Living with partner	1		1		1		1

CI, confidence interval; OR, odds ratio; low-stage, T0-3N0-1M0; high-stage, T4N0M0; T0-3N2-3M0; TxNxM1.

^aAdjusted for age, gender and period; ^bAdjusted for age, gender, period and comorbidity; ORs for income further adjusted for education; ORs for cohabitation status further adjusted for education and income.

In patients with low-stage cancer, the HRs for death adjusted for age, gender and period were statistically significantly higher for patients with low income or living alone than for patients with high income or living with a partner (Table III). These HRs were attenuated after further adjustment for stage, receipt of first-line treatment, comorbidity and performance status but remained significantly increased for patients with low-stage cancer who lived alone (adjusted HR 1.14; 95% CI 1.04–1.25). Among patients with high-stage cancer, the pattern was similar, with increased HRs for patients with short education, medium or low income and living alone (Table III). Adjustment for potential mediating factors reduced the HRs for education and cohabitation towards the null, but the HRs for medium and low income remained virtually unchanged and statistically significantly increased (HR 1.08; 95% CI 1.02–1.15 and 1.12; 95% CI 1.05–1.19).

Table III. Hazard ratios for overall death by stage group in 13 045 patients with lung cancer diagnosed 2004–2010 in Denmark.

	Low stage				High stage
	HR^a	95% CI	HR^b	95% CI	HR^a	95% CI	HR^b	95% CI
Age at diagnosis
(per year)	1.04	1.03–1.04	1.02	1.02–1.03	1.02	1.01–1.02	1.01	1.01–1.01
Gender
Male	1.20	1.10–1.31	1.20	1.11–1.27	1.18	1.13–1.24	1.24	1.19–1.30
Female	1		1		1		1
Period of diagnosis
2004–2005	1.35	1.20–1.67	1.29	1.11–1.51	1.08	1.00–1.17	1.04	0.93–1.16
2006–2007	1.15	0.98–1.34	1.09	0.97–1.23	1.03	0.97–1.11	1.00	0.90–1.11
2008–2009	1.01	0.87–1.19	1.00	0.86–1.17	0.99	0.93–1.06	0.98	0.90–1.06
2010–2011	1		1		1		1
Education
Higher	1		1		1		1
Medium	1.09	0.95–1.27	1.03	0.92–1.16	1.05	0.98–1.13	1.00	0.94–1.05
Short	1.13	0.97–1.31	1.00	0.88–1.15	1.11	1.03–1.19	1.05	0.98–1.13
Income
High	1		1		1		1
Medium	1.23	1.06–1.42	1.14	0.99–1.32	1.10	1.03–1.17	1.08	1.02–1.15
Low	1.29	1.10–1.51	1.18	0.99–1.42	1.14	1.06–1.23	1.12	1.05–1.19
Cohabitation status
Living with partner	1		1		1		1
Living alone	1.20	1.09–1.32	1.14	1.04–1.25	1.07	1.02–1.12	1.01	0.97–1.05
Charlson Comorbidity Index
0	1		1		1		1
1–2	1.21	1.09–1.35	1.13	1.02–1.26	1.02	0.97–1.07	0.98	0.94–1.01
≥ 3	1.57	1.39–1.77	1.39	1.20–1.61	1.15	1.08–1.22	1.02	0.95–1.10
Performance status
0	1		1		1		1
1	1.76	1.59–1.93	1.55	1.38–1.75	1.48	1.41–1.56	1.44	1.36–1.54
2	2.55	2.24–2.91	2.25	1.69–2.98	2.17	2.04–2.30	2.03	1.77–2.33
Stage
Ia	1		1
Ib	1.81	1.61–2.03	1.77	1.60–1.96
IIa	1.97	1.49–2.61	1.49	1.16–1.91
IIb	2.74	2.42–3.11	2.16	1.89–2.47
IIIa	2.52	1.99–3.20	1.78	1.49–2.12	1		1
IIIb					1.37	1.26–1.48	1.34	1.26–1.43
IV					2.06	1.92–2.22	2.00	1.87–2.13
Receipt of first-line treatment
No	1		1		1		1
Yes	0.52	0.48–0.57	0.66	0.59–0.75	0.83	0.79–0.86	0.84	0.77–0.90

^aAdjusted for age, gender and period, HRs for income further adjusted for education and HRs for cohabitation adjusted for education and income; ^bAdjusted as in 1 and further for stage, first-line treatment, Charlson Comorbidity Index and performance status.

The adjusted HRs were generally highest for the first six months after diagnosis, with statistically significant HRs for low income of 1.53 (95% CI 1.09–2.12) in patients with low-stage lung cancer, and for medium and low income (1.14; 95% CI 1.03–1.26 and 1.22; 95% CI 1.09–1.36) and living alone (1.09; 1.01–1.16) in patients with high-stage lung cancer (Supplementary Table I, to be found online at http://informahealthcare.com/doi/abs/10.3109/0284186X.2014.1001037). The raised HRs associated with lower SEP were attenuated with time; however, indications of social inequality in survival remained both for patients with low- and high-stage cancer until 2–3 years after diagnosis.

Discussion

In this nationwide cohort study, we observed social inequality in receipt of first-line lung cancer treatment for patients with both low- and high-stage disease, as socially disadvantaged patients were less likely to receive the recommended treatment. Differential treatment, stage, performance status and comorbidity partly explained the social inequality in survival from lung cancer. After adjustment, patients with high-stage lung cancer and medium-to-low income had an 8–12% higher HR of dying than patients with high income, and patients with low-stage cancer living alone had a 14% higher HR of dying than patients living with a partner. The differences in risk for death by SEP were greatest in the first six months after diagnosis, supporting the hypothesis that stage at diagnosis and access to first-line treatment play a role in social inequality in overall survival after lung cancer.

The strengths of our study include the population-based approach, with a cohort of all lung cancer patients treated in Denmark in 2004–2011. The availability of detailed clinical prognostic and treatment information, including performance status and receipt of first-line treatment, from the Lung Cancer Registry is clearly an advantage over previous studies, as is the availability of individual information on SEP and comorbidity. We used the Charlson Comorbidity Index score for the presence of serious chronic disease based only on hospital in- and outpatients; this might be a limitation, as the score will not include serious chronic disorders managed exclusively in the primary healthcare sector. Furthermore, using only registry information on discharge diagnosis obviated assessment of the severity of disorders in individuals. As a result, for example, chronic obstructive pulmonary disease that is mild and well managed has the same score as a severe debilitating disorder. Inclusion of performance status as a measure of general health status assessed clinically at the time of diagnosis is a strength and would reduce any residual mediation from misclassification of comorbidity scores. A recent systematic review of a number of population-based cohort studies in which the Charlson Comorbidity Index was used to measure comorbidity found that patients with comorbid conditions had less advanced NSCLC but were nevertheless less likely to receive guideline-recommended treatment [18]. A recent Danish cohort study, also based on the Danish Lung Cancer Registry, showed that having a comorbidity score of ≥ 3 affected the survival of NSCLC patients who had undergone surgery by as much as one stage-increment [19].

With an inclusive definition of first-line treatment, receipt of treatment was more frequent in patients with both low- and high-stage cancer who had higher education and income and who lived with a partner, indicating social inequality in access to treatment regardless of whether the intent of treatment is curative or palliative. Contrary to our expectation, this was explained to only a minor degree by the prevalence of comorbidity. In our dataset, some 46% of patients did not receive the recommended first-line treatment for the specific stage of disease and performance status. Given the nature of the data, we cannot determine why this was the case. Treating physicians are in principle responsible for which treatment is offered to each lung cancer patient. The considerations include a number of clinical factors and an overall assessment of the patient's comorbid conditions and general health status. In principle, a decision on the actual treatment plan is made in collaboration with the patient, and the process is complex. A clinician might recommend the treatment that offers the best clinical gain (survival and quality of life), whereas the patient might balance the advantages and disadvantages—the expected benefit of treatment and the side effects, values and trust in the healthcare system [20]. Studies show that physicians communicate differently with patients according to age, race, education and income. Patients who are less well educated and older might have a more passive communication style, and the physician might misperceive their desire and need for information or demand for treatment [21,22]. Potential and probably not exclusive explanations include physician perceived patient vulnerability, i.e. psychological disorder, mental fragility, alcohol or drug abuse, that is not picked up by the comorbidity index and performance status but may increase the risks for treatment complications. Thus, the first-line treatment recommended for the stage-specific disease is not offered, or, alternatively, the patient did not opt for the recommended treatment. As the reasons for not initiating treatment are not reported in the Danish Lung Cancer Registry, the extent to which receipt of treatment is a result of the clinician's or the patient's choice could not be investigated.

Our finding that social inequality was attenuated after adjustment for stage, treatment and comorbidity is in line with those of a number of population-based studies of social inequality and its possible mediation in lung cancer survival in both universal [3,4,6] and non-universal healthcare systems [2,5,7]. Studies of NSCLC in Canada (N = 10 063) and the USA (N = 19 702) showed that area-based SEP was independently associated with survival even after adjustment for surgery, race and marital status [2] or age [4]. Conversely, a Swedish study of 3370 NSCLC patients found that adjustment for stage and treatment eliminated the inequality in survival by education, except among women with early-stage disease [3]. Two studies included all lung cancer patients and a measure of comorbidity, as in the present study [5,6]. Using area-based SEP, the authors of a study of 15 582 lung cancer patients in the UK found no significant inequality in the prognosis of early-stage NSCLC after adjustment for stage, treatment and comorbidity, whereas affluent patients with advanced-stage NSCLC and all patients with small-cell lung cancer, regardless of stage, had better survival than poorer patients [6]. In a study of 76 086 lung cancer patients in the USA, adjustment for stage, treatment and comorbidity accounted for part of the association between area-based SEP and survival, although patients living in low-SEP communities still had a slightly worse prognosis than affluent patients (HR 1.05; 95% CI 1.02–1.09) [5].

We used both education and income as measures of SEP, as they reflect partly overlapping but specific aspects of SEP [23]. Education can be considered a measure of overall knowledge and the ability to communicate and access health care; this indicator was indeed more strongly associated with lung cancer stage at diagnosis [10], whereas receipt of first-line treatment was associated with both income and education. For survival, especially during the first six months after diagnosis of lung cancer, income may be a stronger marker of vulnerability than education, because it is a measure of overall resources during an especially vulnerable period, when access to care is of major importance.

We also used cohabitation status as a measure of SEP. Cancers in patients who live alone are more likely to be diagnosed at a later stage, and they are less likely to receive first-line treatment, especially palliative treatment, and have worse survival than their cohabiting counterparts. Living with a partner has been shown consistently to be associated with better cancer outcomes [24,25], perhaps due to structural and emotional support in navigating the healthcare system throughout diagnosis, treatment, rehabilitation and survivorship.

Various approaches have been proposed to address social disparities in cancer outcomes, including ‘patient navigation’. Nurses or lay volunteers assist patients by identifying and overcoming barriers to high-quality, timely health care, e.g. by helping keep track of appointments, interpreting medical information and providing social support [26]. Most navigation programmes target transitions in care, i.e. to another treatment phase or healthcare sector, which represent challenging breaks in care continuity and potentially affect the compliance of vulnerable patients. The aim of such programmes is to help underserved populations gain access to better cancer care (see, e.g. [27,28]). In lung cancer where evidence of unequal access to treatment is mounting, we are aware of one small navigation programme that was tested among lung cancer patients, with improved timeliness of treatment initiation [29].

The incidence of lung cancer is characterised by social inequality, mainly due to differences in smoking rates by SEP. Therefore, efforts to reduce social differences in smoking will reduce the social inequality in the incidence not only of lung cancer but also the severe comorbid conditions that are highly prevalent in this patient group. The present study suggests that social inequalities in stage at diagnosis, comorbidity and receipt of treatment contribute to the poorer outcomes in the large group of patients who have short education and low income and live alone. Social inequality might affect other outcomes after lung cancer, such as access to supportive care and rehabilitation [30], with potential differences in the prevalence of late effects and need for support [31,32]. Our findings indicate that earlier detection is not enough to reduce the social inequality in survival after lung cancer; efforts are needed to ensure optimal treatment of lung cancer patients with a low social position. In a free, equal-access healthcare system like that in Denmark, the survival rates of patients with a high SEP should set the level for survival of all lung cancer patients. Strategies to meet this aim include differential attention by healthcare professionals to vulnerable patients by systematic focus on how these patients navigate in the complex healthcare system, perhaps in structured, standardised navigation programmes, and monitoring adherence to treatment guidelines for patients treated with either curative or palliative intent.

Supplementary material available online

Supplementary Appendix, Supplementary Table I to be found online at http://informahealthcare.com/doi/abs/10.3109/0284186X.2014.1001037.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.