Socioeconomic status and lifestyle behaviours in cancer survivors: smoking and physical activity

Abstract

Purpose

Smoking cessation and increased physical activity (pa) have been linked to better outcomes in cancer survivors. We assessed whether socioeconomic factors influence changes in those behaviours after a cancer diagnosis.

Methods

As part of a cross-sectional study, a diverse group of cancer survivors at the Princess Margaret Cancer Centre (Toronto, ON), completed a questionnaire about past and current lifestyle behaviours and perceptions about the importance of those behaviours with respect to their health. The influence of socioeconomic indicators on smoking status and physical inactivity at 1 year before and after diagnosis were assessed using multivariable logistic regression with adjustment for clinico-demographic factors.

Results

Of 1222 participants, 1192 completed the smoking component. Of those respondents, 15% smoked before diagnosis, and 43% of those smokers continued to smoke after. The proportion of survivors who continued to smoke increased with lower education level (p = 0.03). Of the 1106 participants answering pa questions, 39% reported being physically inactive before diagnosis, of whom 82% remained inactive afterward. Survivors with a lower education level were most likely to remain inactive after diagnosis (p = 0.003). Lower education level, household income, and occupation were associated with the perception that pa had no effect or could worsen fatigue and quality of life (p ≤ 0.0001).

Conclusions

In cancer survivors, education level was a major modifier of smoking and pa behaviours. Lower socioeconomic status was associated with incorrect perceptions about pa. Targeting at-risk survivors by education level should be evaluated as a strategy in cancer survivorship programs.

INTRODUCTION

Given the growing number of cancer survivors worldwide, the importance of promoting healthy lifestyle behaviours has become increasingly salient. Smoking is an established risk factor for cardiorespiratory disease and a number of cancers¹, including leukemia² and cancers of the urogenital system³. Continued smoking in cancer patients has been associated with negative outcomes, including increased chances of a second malignancy⁴, reduced quality of life⁵, poor treatment response⁶, and increased mortality⁷. More recently, greater physical activity (pa) in cancer survivors has been linked to improved survival⁸ and quality of life⁹ and to less fatigue¹⁰; it has also been described as both safe and beneficial during and after cancer treatment¹¹.

A diagnosis of cancer has been described as a “teachable moment” with respect to modification of lifestyle behaviours¹². However, studies have suggested that approximately 50%–60% of patients continue to smoke^13–15 and close to 80% remain physically inactive¹⁶ after diagnosis. A thorough understanding of the factors influencing changes in those behaviours after diagnosis is needed to best inform survivorship program development for a number of cancer groups. Thus far, studies examining smoking cessation have focused primarily on survivors of lung and head-and-neck cancers; those studies have pointed to the level of nicotine dependence¹⁷ and psychosocial factors such as social environment^18,19 and mental health^13,20 as predictors of continued smoking in cancer patients. In contrast, most studies evaluating post-diagnosis pa in cancer patients have involved survivors of breast and colorectal cancers. Those studies suggested that ethnicity²¹, social support^22,23, and cancer-related beliefs²⁴ were influential factors.

Socioeconomic status (ses) has been associated with smoking^25,26 and physical inactivity^25,27 in general populations. After a diagnosis of cancer, patients might receive more specific counselling and might have greater access to resources and programs. With those possibilities in mind, it is not clear whether the factors that influence health behaviours before diagnosis, particularly those found to be important in healthy general populations, also affect such behaviours after diagnosis. When considering a behavioural change model such as the Health Belief Model^28,29 or the Health Action Process Approach²⁸, a higher ses might reflect more resources and fewer barriers in making improvements in lifestyle habits. Evidence suggests that socioeconomic indicators such as occupation, income, and education are not always interchangeable³⁰, and it is thus important to determine which of those factors are the most influential in the context of behavioural change.

The present cross-sectional study focused on ambulatory cancer patients representing various disease sites, disease progression statuses, and stages of treatment. In this group of survivors, we examined whether ses influenced modification of smoking and pa habits, and which socioeconomic indicators were most influential. We further explored how the indicators were associated with perceptions about the effect of smoking and pa on health outcomes.

METHODS

Population and Questionnaire

After approval for the study was granted by the University Health Network Research Ethics Board, participants were approached at outpatient clinics at the Princess Margaret Cancer Centre (Toronto, ON) between May 2012 and August 2013. After providing informed consent, they were asked to complete a one-time questionnaire about their lifestyle behaviours. Recruitment was based on convenience sampling and was directed such that at least 50 participants were recruited from each major cancer grouping (for example, lung cancer, gastrointestinal cancers, and so on). Eligible participants had to meet these criteria:

• ■ Be more than 18 years of age.

• ■ Be able to communicate in English

• ■ Have no significant cognitive impairment

• ■ Have a histologically confirmed diagnosis of a solid or hematologic malignancy

• ■ Be expected by their oncologists to live for at least 12 months

• ■ Have been diagnosed at least 6 months before recruitment

A recruitment goal was to have the median time since diagnosis fall somewhere between 24 months and 30 months (a long enough period to check for substantive behavioural change). To ensure that the goal was met, the median was checked with every additional 200 patients recruited. All procedures accorded with the ethical standards of the institutional or national research committee (or both) and with the 1964 Helsinki Declaration and its later amendments (or comparable ethical standards).

Within the single questionnaire, patients provided information about sociodemographic characteristics, height and weight, patient-reported Eastern Cooperative Oncology Group performance status, and their “health in the last month” (rated as poor, fair, good, very good, or excellent). Survivors were asked to report their smoking and pa habits at that time and also to recall the nature of the same behaviours 1 year before diagnosis. Separate sets of questions elicited patient perceptions of the influence of smoking and pa on their fatigue, quality of life, and chance of survival at 5 years. Clinical data—including date of diagnosis, cancer site, cancer extent at diagnosis (local vs. distant metastatic), and prognosis—were abstracted from each participant’s electronic medical record.

Definitions of Smoking and PA Outcomes

An “ever-smoker” was defined as a patient who reported having smoked more than 100 cigarettes in his or her lifetime. Ever-smokers were further divided into those who reported being or not being smokers 1 year before diagnosis, with the smokers being further subcategorized as either continuing to smoke or having quit at the time of questionnaire completion.

The pa questions were adapted from the Godin Leisure-Time Exercise Questionnaire³¹. Patients were asked to indicate, for 1 year before diagnosis and at the time of questionnaire completion, the number of times per week they engaged in strenuous, moderate, or mild pa and the number of minutes for each pa session. Occupational activities were not considered pa. “Physically inactive” was defined as a report of zero minutes of any mild, moderate, or strenuous exercise in a typical 7-day period; individuals who were physically inactive at baseline were subcategorized as either physically inactive or physically active at the time of questionnaire completion after diagnosis.

Definitions and Categorization of Other Variables

The ses indicators used in our study were education, household income, and occupation. Education and household income were reported categorically, and patients had the option not to answer those questions. Highest level of education was classified based on the categories of having completed a professional or graduate degree, a university or college undergraduate degree, or a high school degree, or not having completed high school. Household income was classified as high ($100,000), moderate ($60,000–$99,999), low (<$60,000), or “prefer not to answer”—categories that provided an even distribution of participants. For occupation, participants recorded the job at which “they worked the most in their life,” and responses were categorized based on the International Standard Classification of Occupations (isco-08). Body mass index was calculated for each time point (1 year before diagnosis and at the time of questionnaire completion) based on patient-reported height and weight, and was categorized appropriately as underweight, normal, overweight, or obese³².

To assess their perceptions, participants were asked to rate the effects that smoking and pa had on their quality of life, 5-year survival, and fatigue. Those questions used a Likert scale from 1 (make much worse) to 7 (make much better), with 4 being a neutral stance. For smoking, responses were dichotomized as having no effect or making better (4–7) or as making worse (1–3); for pa, responses were dichotomized as having no effect or making worse (1–4) or as making better (5–7). Barriers to pa were assessed with the statement “I find it difficult to spend more time performing physical activity because,” followed by a list that included “I am not sure what to do” and “I have no access to gym or exercise equipment” for which respondents indicated yes or no. The questions were developed for the purpose of the purposes of the present study.

Statistical Analysis

Univariable logistic regression models were used to examine associations of clinical, socioeconomic, and other sociodemographic factors with smoking and physical inactivity 1 year before diagnosis and at the time of questionnaire completion. Subsequent multivariable models used a backward selection algorithm, with entry of all variables significant at p < 0.10 into the univariable analyses. For the multivariable analyses that considered smoking and pa outcomes in addition to the sociodemographic variables (education, household income, and occupation), these additional variables were considered: sex, age at diagnosis, ethnicity, time since diagnosis, body mass index, cancer site, cancer extent at diagnosis, Eastern Cooperative Oncology Group performance status, and health in the preceding month. In the multivariable smoking analyses, years smoked and number of cigarettes per day smoked were also considered; in the multivariable pa analyses, smoking status before diagnosis was also considered. Covariates that resulted in p ≤ 0.05 were retained in the model. All statistical procedures were performed using the SAS software application (version 9.2: SAS Institute, Cary, NC, U.S.A.). All tests were 2-sided, and statistical significance was defined at p ≤ 0.05. Records with missing or incomplete predictor or outcomes data were not included in the models.

RESULTS

Of 2185 eligible patients approached, 1456 (67%) consented to participate, of whom 1222 (84%) had complete information for the smoking or pa questions, forming the sample for analysis. Participants were surveyed at a median of 26 months after diagnosis. Table i describes the sociodemographic and clinical characteristics of the study population. Disease site–specific rates of continued smoking and physical inactivity are summarized in Table ii.

TABLE I.

Sociodemographic and clinical characteristics of the study population

Variable	Value	Missing (n)
Patients	1222
Female sex (%)	53	0
Age at follow-up (years)
Median	59	8
Interquartile range	19
Married or live with partner (%)	72	9
White ethnicity (%)	82	31
Education (%)
Professional/Masters/PhD	18	18
University or college	43
High school	31
Less than high school	8
Household income (%)
High (>$100,000)	28	42
Moderate ($60,000–$99,999)	22
Low (<$60,000)	25
Prefer not to answer	25
Occupation (%)
Professional	39	35
Managers	11
Technicians, associate professionals	17
Service and sales workers	12
Manual occupations	12
Not classified	8
Months since diagnosis
Median	26	9
Interquartile range	56
Body mass index (%)
At baseline
Obese	22	107
Overweight	38
Normal	38
Underweight	2
At follow-up
Obese	19	93
Overweight	34
Normal	43
Underweight	4
Cancer site
Breast	16	9
Gastrointestinal	12
Genitourinary	14
Gynecologic	9
Head, neck, and thyroid	14
Hematologic	19
Lung	6
Skin and other cancers	9
Unknown	1
Cancer extent at diagnosis
Local (solid tumour)	69	58
Distant metastatic (solid tumour)	11
Hematologic	20
ECOG performance status
0	48	34
1	37
2–4	14
Health in preceding month
Excellent	11	20
Very good	25
Good	39
Fair	22
Poor	3

TABLE II.

Cancer disease site–specific rates of smoking and physical inactivity before and after diagnosis

Cancer site	Smoking rate (%) in relation to diagnosis		Physical activity (%) in relation to diagnosis
	1 Year before (smokers/ever-smokers)	After (still smoking)	1 Year before (physically inactive/all)	After (still inactive)
Breast	34	42	33	77
Gastrointestinal	37	53	40	81
Genitourinary	15	55	37	82
Gynecologic	35	59	44	87
Head, neck, and thyroid	42	50	39	88
Hematologic	29	23	37	76
Lung	45	27	42	89
Skin and other cancers	34	47	37	86

Socioeconomic Factors Associated with Smoking Before and After Diagnosis

Of 1192 participants analyzed for smoking, 47% were ever-smokers, of whom 32% reported smoking 1 year before diagnosis (Figure 1). Table iii outlines the socioeconomic factors associated with smoking before and after diagnosis. At baseline, the smoking rates among ever-smokers were highest in participants with lung (45%) and head-and-neck and thyroid cancer (42%), and lowest in the participants with genitourinary cancers (15%). Our multivariable model demonstrated that, compared with participants having high household incomes, those with moderate and low household incomes were, respectively, 1.5 and 2.7 times as likely to be smokers (p = 0.006, Table iii). Older age at diagnosis (p < 0.001) and cancer site (p = 0.002) also remained significant in the model, but education and occupation level were not associated with baseline smoking rates (Table iii).

FIGURE 1.

Frequency of the lifestyle behaviours of tobacco smoking and physical activity at 1 year before diagnosis and at questionnaire completion after diagnosis. Subgroup analyses focused on the shaded boxes.

TABLE III.

Socioeconomic factors associated with continued smoking and physical inactivity before and after diagnosis

Socioeconomic variable	Comparator	1 Year before diagnosis							After diagnosis
		Patients (%)	Univariable			Multivariablea			Patients (%)	Univariable			Multivariableb
			OR	95% CI	p Valuec	aOR	95% CI	p Valuec		OR	95% CI	p Valuec	aOR	95% CI	p Valuec
			All ever-smokers (n= 556)								Survivors smoking before diagnosis (n=182)
Associated with smoking		Smoking	Smoking (n=182) vs. not smoking (n=374)						Smoking	Continued smoking (n=78) vs. quit smoking (n=104)
Education
Professional/graduate degree		20	Reference		0.04				31	Reference		0.03	Reference		0.03
University or college		34	2.09	1.1 to 3.9					32	1.05	0.3 to 3.4		0.92	0.3 to 3.0
High school		35	2.21	1.2 to 4.1					49	2.14	0.7 to 6.8		1.99	0.6 to 6.4
<High school		41	2.84	1.3 to 6.1					64	3.85	1.0 to 15		3.50	0.9 to 14
Household Income
High		25	Reference		0.005	Reference		0.006	41	Reference		0.64
Moderate		29	1.24	0.7 to 2.2		1.50	0.8 to 2.7		35	0.78	0.3 to 2.1
Low		44	2.35	1.4 to 3.9		2.70	1.5 to 4.7		47	1.27	0.6 to 2.9
Prefer not to answer		32	1.42	0.8 to 2.4		1.87	1.0 to 3.4		37	0.85	0.3 to 2.1
Occupation
Professionals		25	Reference		0.003				33	Reference		0.46
Managers		29	1.22	0.6 to 2.3					29	0.83	0.3 to 2.8
Technicians		38	1.15	0.7 to 2.0					44	1.60	0.6 to 4.2
Service and sales		38	1.81	1.0 to 3.2					48	1.87	0.7 to 4.8
Manual occupations		46	2.47	1.4 to 4.3					46	1.70	0.7 to 4.1
Not classified		49	2.80	1.4 to 5.6					55	2.44	0.8 to 7.1
			All survivors (n=1114)							Survivors physically inactive before diagnosis (n=427)
Associated with physical inactivity		Inactive	Physically inactive (n=427) vs. active (n=687)						Inactive	Currently physically inactive (n=350) vs. active (n=77)
Education
Professional/graduate degree		26	Reference		<0.0001	Reference		<0.0001	75	Reference		0.007	Reference		0.003
University or college		33	1.43	1.0 to 2.1		1.37	0.9 to 2.0		75	1.0	0.5 to 2.1		1.37	0.9 to 2.0
High school		45	2.36	1.6 to 3.4		2.22	1.5 to 3.3		87	2.10	1.0 to 4.6		2.42	1.1 to 5.4
<High School		63	4.92	2.8 to 8.5		4.05	2.3 to 7.2		94	5.09	0.5 to 2.1		5.89	1.5 to 22
Household Income
High		27	Reference		<0.0001				76	Reference		0.15
Moderate		38	1.66	1.2 to 2.4					76	1.05	0.5 to 2.1
Low		40	1.86	1.3 to 2.6					84	1.69	0.8 to 3.5
Prefer not to answer		48	2.49	1.8 to 3.5					86	1.95	1.0 to 3.9
Occupation
Professionals		32	Reference		<0.0001				74	Reference		0.08
Managers		27	0.77	0.5 to 1.2					84	1.94	0.7 to 5.4
Technicians		44	1.68	1.2 to 2.4					87	2.45	1.2 to 5.1
Service and sales		44	1.68	1.1 to 2.5					81	1.57	0.7 to 3.4
Manual occupations		53	2.36	1.6 to 3.6					89	2.78	1.2 to 6.7
Not classified		38	1.31	0.8 to 2.1					85	2.09	0.8 to 5.8

^aFor smoking, adjusted for age at diagnosis and cancer site; for physical inactivity, adjusted for age and ethnicity.

^bFor smoking, adjusted for time since diagnosis; for physical inactivity, adjusted for sex. For smoking, other non–socioeconomic status variables considered were sex, age, ethnicity, and body mass index. For physical inactivity, other non–socioeconomic status variables considered were age, ethnicity, time since diagnosis, body mass index, cancer site, cancer extent, Eastern Cooperative Oncology Group performance status, health in the preceding month, smoking status, years smoked, and number of cigarettes smoked 1 year before diagnosis.

^cSignificant values shown in boldface type.

OR = odds ratio; CI = confidence interval; aOR = adjusted odds ratio.

Among participants who were smokers before diagnosis (n = 182), 43% indicated that they were still smoking at the time of questionnaire completion. The highest rates of continued smoking after diagnosis were found for participants with gynecologic (59% of all baseline smokers) and genitourinary (55%) cancers; the lowest rates were found for participants with lung (27%) and hematologic (23%) cancers. After adjusting for time since diagnosis (p = 0.035), the final multivariable model found that education level was the socioeconomic factor most closely associated with continued smoking after a cancer diagnosis. Compared with participants having a graduate-or professional-level education, those with only a high school education and those who did not graduate from high school were, respectively, 2.0 and 3.5 times as likely to continue smoking (p = 0.03, Table iii).

Socioeconomic Factors Associated with Physical Inactivity Before and After Diagnosis

Of 1106 participants analyzed for pa, 39% reported being physically inactive 1 year before diagnosis (Figure 1). Table iii outlines the socioeconomic factors associated with physical inactivity before and after diagnosis. Education was the only socioeconomic variable that remained significant in the multivariable model; compared with participants having a graduate or professional education, those with a high school education or a less than high school education were, respectively, 2.2 and 4.1 times as likely to be inactive (Table iv). Older age (p < 0.0001) and non-white ethnicity (adjusted odds ratio: 1.89; 95% confidence interval: 1.36 to 2.62; p = 0.0002) also remained in the model.

TABLE IV.

Socioeconomic factors associated with perceptions and barriers to physical activity (PA) among cancer survivors

Factor	Comparator	Perception: “What effect does PA have ...”									Potential barriers: “It is difficult to spend more time performing PA ...”
		On fatigue [no effect or worsens (n=263) vs. lessens (n=830)]			On quality of life [no effect or worsens (n=102) vs. improves (n=1030)]			On 5-year survival [no effect or worsens (n=124) vs. improves (n=971)]			Because of uncertainty about what to do [yes (n=104) vs. no (n=991)]			Because of lack of access to gym or equipment [yes (n=224) vs. no (n=873)]
		OR	95% CI	p Value a	OR	95% CI	p Valuea	OR	95% CI	p Valuea	OR	95% CI	p Valuea	OR	95% CI	p Valuea
Education
Professional/Masters/PhD		Reference		0.0001	Reference		0.0001	Reference		0.11	Reference		0.006	Reference		0.92
University or college		1.55	1.0 to 2.4		1.33	0.7 to 2.7		0.96	0.6 to 1.7		1.18	0.7 to 2.2		0.95	0.6 to 1.4
High school		1.77	1.1 to 2.8		2.55	1.3 to 5.1		1.29	0.7 to 2.2		0.95	0.5 to 1.8		1.07	0.7 to 1.6
<High school		3.98	2.2 to 7.3		4.75	2.1 to 11		2.10	1.0 to 4.5		3.02	1.4 to 6.4		0.93	0.5 to 1.8
Household income
High		Reference		<0.0001	Reference		<0.0001	Reference		0.03	Reference		0.40	Reference		0.29
Moderate		1.45	0.9 to 2.2		2.63	1.1 to 6.0		0.86	0.5 to 1.6		1.56	0.9 to 2.8		0.95	0.6 to 1.5
Low		2.48	1.7 to 3.7		5.32	2.5 to 9.8		1.81	1.1 to 3.0		1.58	0.9 to 2.9		1.38	0.9 to 2.1
Prefer not to answer		1.81	1.2 to 2.7		4.58	2.2 to 9.8		1.47	0.9 to 2.5		1.33	0.7 to 2.4		1.18	0.8 to 1.8
Occupation
Professionals		Reference		0.002	Reference		<0.0001	Reference		0.16	Reference		0.59	Reference		0.79
Managers		1.03	0.6 to 1.7		1.17	0.5 to 2.8		0.67	0.3 to 1.5		0.85	0.4 to 1.8		0.78	0.5 to 1.3
Technicians		1.84	1.2 to 2.7		2.21	1.2 to 4.2		1.23	0.7 to 2.1		0.93	0.5 to 1.7		0.91	0.6 to 1.4
Service and sales		1.83	1.2 to 2.9		2.42	1.2 to 4.8		1.49	0.8 to 2.7		1.60	0.9 to 2.9		1.23	0.8 to 2.0
Manual occupations		2.17	1.4 to 3.4		4.44	2.4 to 8.3		1.76	1.0 to 3.2		1.16	0.6 to 2.3		1.05	0.6 to 1.7
Not classified		1.24	0.7 to 2.2		1.10	0.4 to 3.0		0.77	0.3 to 1.8		1.30	0.6 to 2.7		1.05	0.6 to 1.8

^aSignificant values shown in boldface type.

Although 31% of all patients were physically inactive at the time of questionnaire completion (Figure 1), the percentage of those who remained inactive among those who had been physically inactive before diagnosis (n = 430) was 82%. Lower education level (p = 0.003) and male sex (adjusted odds ratio: 2.39; 95% confidence interval: 1.39 to 4.12; p = 0.002) were factors associated with continued physical inactivity after diagnosis. Compared with participants having a graduate or professional degree, those who had a high school education and who did not complete high school were, respectively, 2.4 and 5.9 times as likely to remain inactive after diagnosis (Table iii).

SES, Barriers, and Perceptions

We explored relationships between the socioeconomic indicators and the perceptions of participants about pa and smoking. Most participants believed that pa could lessen fatigue, improve quality of life, and improve their chance of survival in 5 years. Those with higher education levels, household incomes, and levels of occupation were more likely to indicate that pa would lessen fatigue and improve quality of life (Table iv).

We also explored associations between the socioeconomic variables and potential barriers to pa. Specifically, 9% of participants reported that “being unsure of what to do” was a barrier to pa. Patients were more likely to report this barrier if their education level was lower (p = 0.006); however, household income (p = 0.40) and occupation (p = 0.59) were not significant predictors of that perception (Table iv).

A parallel analysis in relation to smoking perceptions was also conducted. No significant relationships were observed between socioeconomic factors and perceptions about smoking and its effect on fatigue, quality of life, or the chance of survival in 5 years (Table v).

TABLE V.

Socioeconomic factors associated with smoking perceptions among cancer survivors

Factor	Comparator	Perception: “What effect does smoking have ...”
		On fatigue [has no effect or makes better (n= 171) vs. makes worse (n =377)]			On quality of life [has no effect or makes better (n= 140) vs. makes worse (n= 422)]			On 5-year survival [has no effect or makes better (n=161) vs. makes worse (n= 386)]
		OR	95% CI	p Value	OR	95% CI	p Value	OR	95% CI	p Value
Education
Professional/Masters/PhD		Reference		0.53	Reference		0.68	Reference		0.91
University or college		1.00	0.6 to 1.7		1.04	0.6 to 1.8		0.94	0.6 to 1.6
High school		1.33	0.8 to 2.2		0.83	0.3 to 2.1		1.11	0.7 to 1.9
<High school		1.02	0.4 to 2.4		1.04	0.6 to 1.8		1.01	0.4 to 2.4
Household income
High		Reference		0.46	Reference		0.66	Reference		0.51
Moderate		0.95	0.6 to 1.6		0.82	0.5 to 1.5		1.15	0.7 to 2.0
Low		0.97	0.6 to 1.6		1.02	0.6 to 1.8		1.03	0.6 to 1.8
Prefer not to answer		1.36	0.8 to 2.2		1.19	0.7 to 2.0		1.42	0.9 to 2.4
Occupation
Professionals		Reference		0.66	Reference		0.97	Reference		0.82
Managers		0.86	0.5 to 1.6		0.90	0.5 to 1.7		0.82	0.4 to 1.5
Technicians		0.99	0.6 to 1.&		1.07	0.6 to 1.9		0.88	0.5 to 1.5
Service and sales		0.93	0.5 to 1.7		1.28	0.7 to 2.4		1.05	0.6 to 1.9
Manual occupations		1.46	0.8 to 2.6		1.08	0.6 to 2.0		1.35	0.8 to 2.4
Not classified		1.37	0.7 to 2.8		1.09	0.5 to 2.4		0.93	0.4 to 2.0

DISCUSSION

The promotion of smoking cessation and pa has become a priority for health care practitioners managing cancer survivors. In the present report, we examined a broad cross-section of cancer patients representing various disease sites and cancer progression statuses. The rates of continued smoking (43%) and continued physical inactivity (82%) were similar to those reported in earlier studies in mixed cancer populations^13,14,16,33.

Our results suggest that, although other socioeconomic variables such as household income might be associated with lifestyle behaviour at diagnosis, education was the strongest independent predictor of continued smoking and continued physical inactivity after a cancer diagnosis. Those findings are consistent with significant results^16,34–36 and nonsignificant trends³⁷ previously reported in studies that did not examine all 3 socioeconomic indicators. This relationship between education and post-diagnosis behaviour modification could also be reflective of a general phenomenon among patients with chronic disease³⁸. Our observations that younger age and lower income are associated with baseline smoking is reflective of known associations in the general population^26,39. Similarly, older, more educated, and non-white patients are more likely to be physically inactive^39,40.

Although lower education was a strong predictor of unhealthy lifestyle behaviours at follow-up, household income and occupation did not reach significance. Those 3 indicators, albeit related, are not necessarily redundant predictors of key health outcomes; and although most studies examining lifestyle behaviours will adjust for ses, the specific indicator or indicators included in the analyses are not consistent across reports. However, the most influential socioeconomic indicators might vary by country and centre. Canadian patients might experience less of a financial burden related to their cancer than do American patients⁴¹, and that difference could partly account for household income appearing, in our study, to be less important in predicting behaviour change after diagnosis.

We found that patients with greater education and household income, and a higher occupation level, were more likely to appreciate the potential benefits of pa on their fatigue and quality of life; however, that association was not apparent for perceptions about smoking. Individuals of a lower ses might engage in detrimental health behaviours because of a lack of knowledge or incorrect beliefs about the associated health risks⁴², but that relationship might be mediated by other factors as well⁴³. The specific mechanisms are particularly important to determine when developing interventions aimed at promoting smoking cessation and improvement in exercise habits among cancer survivors of various socioeconomic levels. Thus far, several education-based interventions such as telephone counselling⁴⁴, motivational interviewing⁴⁵, and an oncologist’s recommendation⁴⁶ have proved successful for encouraging pa in cancer patients, and further research on those types of approaches should consider examining whether they are equally effective among less educated participants. In contrast, many counselling approaches to smoking cessation have been less effective^47,48, suggesting that additional mechanisms could play important roles.

The foregoing results can also be interpreted in the context of established models for behaviour modification. The Health Action Process Approach and the Health Belief Model emphasize how an individual’s personal barriers and health perceptions influence their ability to modify behaviour^28,29, and our data suggest that ses could be an additional secondary mediator from that perspective. However, when considering the broader framework of a socio-ecological model⁴⁹, our findings could further reflect the importance of community or institutional factors in the cancer population, because those factors can be associated with ses. In addition to considering variables at the individual level, future research should explore whether community factors such a housing, access to facilities, workplaces, and local health resources influence the behaviour patterns of cancer survivors.

Socioeconomic disparities in cancer survival still exist in universal health care systems^50,51. Given the present work, it might be worth determining whether health behaviours contribute to those disparities and whether education is a significant predictor of survival. Thus far, most Canadian studies considering cancer survival have used income as the primary ses indicator and have not incorporated smoking or physical inactivity as mediating factors^50,52. Education differences in cancer survival have been identified in the United States⁵³ and continental Europe⁵¹, but could also exist in Canada and the United Kingdom.

Factors that might limit the generalizability of our work include its single-centre focus, with a sample population skewed toward more socioeconomically advantaged individuals. Ideally, studies examining the importance of ses in health outcomes should incorporate population-level data, which are not currently available in Ontario for the pre-and post-diagnosis smoking and pa outcomes that we examined. To capture a large and broad cross-section of cancer patients at our centre, we opted to administer a single questionnaire that elicited outcomes at both time points. That approach introduces a concern relating to recall bias, and it is unclear whether the reports of smoking and pa outcomes overestimated or underestimated the true values. Studies addressing the issue have found that recall of past pa behaviour up to 30 years into the past is generally reliable, but accuracy of recall for the rigour of the activity can vary^54–56. An analogous study examining recall in smoking data indicated that, although 20-year recall was reliable for smoking status, recollection of smoking amounts were less reliable⁵⁷. That background considered, we attempted to minimize bias in our study by only considering broad outcome measures such as smoking versus not smoking, and physically inactive versus active.

CONCLUSIONS

Our results suggest that, at a comprehensive cancer centre in a universal health care system, the population shows socioeconomic disparities with respect to smoking and levels of physical inactivity. Specifically, cancer patients with less education are more likely to continue smoking and to remain physically inactive after diagnosis. Health care providers should take those factors into consideration when designing survivorship care programs.

CONFLICT OF INTEREST DISCLOSURES

We have read and understood Current Oncology’s policy on disclosing conflicts of interest, and we declare that we have none.