Journal of Cancer Survivorship

, Volume 5, Issue 1, pp 27–34

Lifestyle behaviors in Massachusetts adult cancer survivors

Authors

    • Section of General Internal Medicine, Department of MedicineBoston University School of Medicine
  • Joshua Nyambose
    • Massachusetts Comprehensive Cancer Prevention and Control Program, Bureau of Community Health and PreventionMassachusetts Department of Public Health
  • Tracy A. Battaglia
    • Women’s Health Unit, Section of General Internal Medicine and Women’s Health Interdisciplinary Research Center, Evans Department of MedicineBoston University School of Medicine
Article

DOI: 10.1007/s11764-010-0162-6

Cite this article as:
Linsky, A., Nyambose, J. & Battaglia, T.A. J Cancer Surviv (2011) 5: 27. doi:10.1007/s11764-010-0162-6

Abstract

Introduction

Adoption of healthy lifestyles in cancer survivors has potential to reduce subsequent adverse health. We sought to determine the prevalence of tobacco use, alcohol use, and physical inactivity among cancer survivors overall and site-specific survivors.

Methods

We performed a cross-sectional analysis of the Massachusetts Behavioral Risk Factor Surveillance System, 2006–2008, and identified 1,670 survivors and 18,197 controls. Specific cancer sites included prostate, colorectal, female breast, and gynecologic (cervical, ovarian, uterine). Covariates included age, gender, race/ethnicity, education, income, marital status, health insurance, and physical and mental health. Gender stratified logistic regression models associated survivorship with each health behavior.

Results

4.9% of men and 7.7% of women reported a cancer history. In adjusted regression models, male survivors were similar to gender matched controls, while female survivors had comparable tobacco and alcohol use but had more physical inactivity than controls (OR 1.5; 95% CI, 1.2–1.8). By site, breast cancer survivors were more likely to be physically inactive (OR 1.5; 95% CI, 1.1–2.0) and gynecologic cancer survivors were more likely to report current tobacco use (OR 1.8; 95% CI, 1.2–2.8).

Conclusions and Implications for Cancer Survivors

Specific subgroups of cancer survivors are more likely to engage in unhealthy behaviors. Accurate assessment of who may derive the most benefit will aid public health programs to effectively target limited resources.

Keywords

CancerSurvivorsLife styleBehavior

Background

With continued improvements in detection and treatment, more Americans are surviving a diagnosis of cancer. Defining a survivor as someone with any history of a cancer diagnosis, regardless of time since diagnosis, there were nearly 12 million American survivors in 2007 [1]. These survivors are at risk for recurrence, secondary cancers, and other medical problems, including cardiovascular disease and diabetes [2].

Risk for subsequent health problems may result from cancer treatments, genetics, or lifestyle behaviors [3]. Estimates indicate that one third of cancer deaths are related to tobacco and another third are due to physical inactivity and dietary habits [4]. While a cancer diagnosis is conceptualized as a “teachable moment,” behavior change made post-diagnosis is often not maintained [3]. As such, tertiary prevention via adoption and maintenance of healthy behaviors and avoidance of unhealthy habits has potential to reduce adverse health consequences and improve quality of life for survivors [3, 5].

Studies of national survey data suggest that, as a group, cancer survivors have similar lifestyles as those without a cancer history [68]; however, certain survivors are at increased risk for unhealthy behaviors. Younger survivors are more often current smokers [6, 7], yet concurrently are more physically active [6]. Meanwhile, prostate cancer survivors have a higher prevalence of moderate-to-heavy alcohol use [8].

Prior studies have conflicting findings regarding unhealthy behaviors in cancer survivors, in part due to differences in methodology and unmeasured confounders, such as physical and mental health status. Furthermore, regional variation may be obscured with national level analyses. Since many public health interventions are organized at the state level, detailing survivors’ behaviors within an individual state may better direct limited resources and public health planning. Therefore, we sought to determine the prevalence of unhealthy lifestyle behaviors of Massachusetts cancer survivors by conducting a cross sectional analysis using data from the Behavioral Risk Factor Surveillance System (BRFSS). Because patient characteristics associated with specific cancer sites differ, we also assessed whether survivors of specific cancers were more or less likely to engage in unhealthy lifestyles.

Methods

Data source

The BRFSS is an annual state-based cross-sectional telephone survey of adults ≥18 years, established and coordinated by the Centers for Disease Control and Prevention (CDC) [9]. A range of information is collected, including health conditions, alcohol and drug use, sexual behaviors, and environmental exposures. Massachusetts added cancer survivorship questions to the BRFSS Survey in 2006 to track behaviors of survivors.

Participants

There were 20,900 Massachusetts residents who were asked “Have you ever been diagnosed with any type of cancer?” in 2006–2008. The 18,197 respondents who reported no history of cancer comprise the non-cancer controls. The 2,339 subjects who responded affirmatively were then asked to name the cancer site (up to three). Individuals who refused or did not know cancer history or cancer site were excluded. We further removed those who did not report age and gender, reported non-melanoma skin cancer only, or had inconsistent gender/cancer combinations, yielding 1,670 cancer survivors in our study.

Cancer survivors were further classified by primary site of cancer. We focused specifically on the sites with the highest frequencies of responses: female breast, gynecologic, prostate, and colorectal (CRC). Gynecologic survivors were those who answered that they were diagnosed as having cervical, uterine, or ovarian cancer—a single combined answer option in the BRFSS. Respondents with other cancer sites or a history of multiple cancers formed separate subgroups of “Single—other” and “Multiple,” respectively.

Demographics and covariates

Self-reported information was obtained for age, race, ethnicity, education, employment status, marital status, health insurance, and physical and mental health status. Race and ethnicity were recoded into white non-Hispanic vs. other. We dichotomized education, marital status, and health insurance. Employment status had three levels (employed, unemployed, retired).

Physical health was assessed using the question “For how many days during the past 30 days was your physical health not good?” We categorized responses into ≥14 day vs. <14 days, which corresponds with physical activity levels [10]. Mental health was assessed by asking “For how many days during the past 30 days was your mental health not good?” Frequent Mental Distress was considered present if the subject reported ≥14 days, consistent with CDC standards and other studies [11, 12].

Behavioral outcomes

Tobacco use

Smoking behavior was measured using combinations of current and former cigarette use. “Current smokers” reported ≥100 lifetime cigarettes and current use. “Former smokers” reported ≥100 lifetime cigarettes without current use. Subjects with <100 lifetime cigarettes were classified as “Never smokers.” All analyses assessing former tobacco use excluded never smokers since they could not become either a former or current smoker.

Alcohol use

Heavy alcohol use was defined as an average intake in the past 30 days of 60 drinks for a man and 30 drinks for a woman, consistent with general recommendations in the absence of universal guidelines for alcohol consumption [4].

Physical inactivity

Respondents were asked “During the past month...did you participate in any physical activities such as running, calisthenics, golf, gardening or walking for exercise?” If they answered negatively, they were designated physically inactive.

Statistical analyses

To account for the complex sample survey design, all analyses were weighted to reflect the probability of selection of a telephone number, the number of adults in a household, and differences in participation by gender and age to provide Massachusetts state-level estimates. All sample sizes reported are unweighted and all percentages are weighted. Due to the imbalance in opportunity for men and women to be included in the study population (two female cancers, one male cancer, one gender neutral cancer) all analyses were stratified by gender. Within this stratification, analyses were conducted on survivors as a group and by subgroups of specific cancer sites.

We used descriptive statistics and chi-square tests to examine demographics and differences in categorical baseline characteristics. Differences in age were tested with t-tests. We assessed for colinearity of characteristics with non-parametric methods.

We determined both unadjusted prevalence and then age adjusted prevalence of each behavioral outcome. Adjusted rates were calculated using the direct method to the year 2000 Census Massachusetts population for each behavioral outcome for cancer survivors overall and site-specific survivors. Age adjusted logistic regression models determined the odds of each outcome for survivors and site-specific survivors compared to controls. We then modeled each behavior adjusting for age, race/ethnicity, education, and physical and mental health status, as these factors have potential to impact the adoption of health behaviors. Missing data were excluded from analysis.

Statistical significance was set at alpha = 0.05. All analyses were performed using version 9.2 of SAS software (SAS Institute Inc, Cary, North Carolina). This study was approved by the Institutional Review Boards of the MA Department of Public Health and Boston University Medical Center.

Results

Population characteristics

Cancer survivors

Among the unweighted study sample of 19,867 respondents, there were more women than men (64% vs. 36%). In gender stratified samples, the weighted prevalence of having any cancer history was 4.9% for men and 7.7% for women (Table 1). Compared to same gender controls, survivors of any cancer were older, more likely to be white non-Hispanic, not be currently employed, have health insurance and report more poor physical health days (Table 2). There were no differences in poor mental health days for men or women. Among women, cancer survivors were less likely to have any college education.
Table 1

Gender stratified Massachusetts cancer survivor distribution, 2006–2008

 

Men (n = 6,709) n (% ± SE)

Women (n = 11,488) n (% ± SE)

No history of cancer

6,709 (95.1 ± 0.3)

11,488 (92.3 ± 0.3)

Any cancer

516 (4.9 ± 0.3)

1,154 (7.7 ± 0.3)

Specific cancers

 Colorectal

52 (0.5 ± 0.1)

72 (0.5 ± 0.1)

 Prostate

244 (2.2 ± 0.2)

 Breast

479 (3.1 ± 0.2)

 Gynecologica

275 (2.1 ± 0.2)

 Single other site

188 (1.9 ± 0.2)

264 (1.8 ± 0.2)

 Multiple sites

32 (0.3 ± 0.1)

64 (0.3 ± 0.1)

All percentages are weighted. Totals may not equal 100% due to rounding.

Data source: MA Behavioral Risk Factor Surveillance System, 2006–2008.

Abbreviations: SE standard error

aGynecologic cancer includes cervical, uterine, and ovarian.

Table 2

Baseline characteristics by cancer survivorship and gender, 2006–2008

Characteristic

Men

Women

Non-cancer controls (n = 6709) % (SE)

Survivors (n = 516) % (SE)

p-value

Non-cancer controls (n = 11,488) % (SE)

Survivors (n = 1,154) % (SE)

p-value

Age, mean (95% CI)

44 (43–45)

67 (65–68)

<0.001

46 (46–47)

61 (59–64)

<0.001

Race/Ethnicity

  

0.004

  

<0.001

Non-Hispanic White

81 (0.8)

90 (2.3)

 

83.5 (0.6)

94 (0.9)

 

Other

19 (0.8)

10 (2.3)

 

16.5 (0.6)

6 (0.9)

 

Missing, n

70

5

 

76

10

 

Education

  

0.42

  

0.001

College or more

68 (1.0)

65 (3.0)

 

68.3 (0.7)

61 (2.2)

 

High School or less

32 (1.0)

35 (3.0)

 

31.7 (0.7)

39 (2.2)

 

Missing, n

14

3

 

16

1

 

Employment

  

<0.001

  

<0.001

Employed

74 (0.8)

32 (3.0)

 

61 (0.7)

40 (2.2)

 

Unemployed

14 (0.8)

15 (2.3)

 

25 (0.7)

25 (2.0)

 

Retired

12 (0.5)

53 (3.1)

 

14 (0.4)

36 (2.0)

 

Missing, n

16

0

 

14

2

 

Married

  

<0.001

  

<0.001

Yes

65 (1.0)

75 (2.5)

 

61 (0.8)

54 (2.2)

 

No

35 (1.0)

25 (2.5)

 

39 (0.8)

46 (2.2)

 

Missing, n

25

1

 

39

2

 

Health Insurance

  

<0.001

  

0.017

Yes

90 (0.7)

99 (0.3)

 

95 (0.4)

97 (0.7)

 

No

10 (0.7)

1 (0.3)

 

5 (0.4)

3 (0.7)

 

Missing, n

20

0

 

18

2

 

Physical Health

  

<0.001

  

<0.001

≥14 unhealthy days

8 (0.5)

20 (2.5)

 

10 (0.4)

19 (1.7)

 

<14 unhealthy days

92 (0.5)

80 (2.5)

 

90 (0.4)

81 (1.7)

 

Missing, n

77

15

 

200

42

 

Mental Health

  

0.43

  

0.26

≥14 unhealthy daysa

8 (0.5)

9 (1.8)

 

10.5 (0.5)

12 (1.5)

 

<14 unhealthy days

92 (0.5)

91 (1.8)

 

89.5 (0.5)

88 (1.5)

 

Missing, n

89

7

 

164

17

 

All percentages are weighted. Totals may not equal 100% due to rounding.

MA Behavioral Risk Factor Surveillance System, 2006–2008.

aFrequent Mental Distress is ≥14 unhealthy days of Mental Health.

Abbreviations: SE standard error

Site-specific survivors (data not shown)

Among men, prostate cancer survivors (n = 244) and CRC survivors (n = 52) were older than controls (mean age 72 vs. 66 vs. 44 years, respectively). They were more likely to be white non-Hispanic, not be currently employed, and report more poor physical health days.

Among women, breast cancer survivors (n = 479) and CRC survivors (n = 72) were oldest (mean age 65 and 70 years, respectively) compared to gynecologic cancer survivors (n = 275) and controls (51 and 46 years, respectively). Gynecologic survivors had the lowest prevalence of any college education. Compared to the breast and CRC survivors, they had the highest prevalence of ≥14 days poor physical health. Although not statistically significant, gynecologic survivors were more likely to have Frequent Mental Distress (14%).

Prevalence of health behaviors

Cancer survivors

There was no difference in the age adjusted prevalence of heavy alcohol use or physical inactivity for male survivors compared to controls, but survivors were more likely to report former smoking (45% vs. 29%) and less likely to report never smoking (31% vs. 53%) (Table 3). Likewise, female cancer survivors were similar to controls in heavy alcohol use and physical inactivity, but they were more likely to report current smoking (24% vs. 16%) or former smoking (39% vs. 25%) and less likely to report never smoking (42% vs. 60%).
Table 3

Age adjusted prevalence of behavioral risk factors by cancer survivorship and gender, 2006–2008

Behavior

Men

Women

Non-cancer controls (n = 6,709) % (SE)

Survivors (n = 516) % (SE)

Non-cancer controls (n = 11,488) % (SE)

Survivors (n = 1,154) % (SE)

Heavy drinking

Yes

6.3 (0.5)

8.4 (4.5)

5.0 (0.3)

9.3 (2.4)

No

93.7 (0.5)

91.6 (4.5)

95.0 (0.3)

90.7 (2.4)

Missing, n

170

11

234

17

Smoking

Current

18.5 (0.8)

24.0 (6.2)

15.5 (0.6)

24.0 (3.2)

Former

29.0 (0.7)

45.1 (6.4)

25.0 (0.6)

33.6 (3.1)

Never

52.5 (0.9)

30.9 (5.8)

59.5 (0.7)

42.4 (3.6)

Missing, n

30

0

57

6

Physical inactivity

Inactive

19.9 (0.8)

31.4 (6.4)

22.5 (0.6)

27.5 (2.9)

Active

80.1 (0.8)

68.6 (6.4)

77.5 (0.6)

72.5 (2.9)

Missing, n

3

0

2

1

All percentages are weighted. Totals may not equal 100% due to rounding. Age adjusted by the direct method to the year 2000 Census Massachusetts population using the age groups 18–49 years, 50–59 years, 60–69 years, and 70 years or older. MA Behavioral Risk Factor Surveillance System, 2006–2008.

Abbreviations: SE standard error

Site-specific survivors (data not shown)

Small sample size for site-specific male survivors precluded age adjusting, but in unadjusted analyses, men with a history of prostate cancer or CRC had comparable heavy alcohol use and physical inactivity as controls. In age adusted analyses of women survivors of breast and gynecologic cancers, there were no differences from controls in heavy alcohol use or physical inactivity. Gynecologic survivors had more current smoking than controls (32% vs. 16%) and breast cancer survivors reported more former smoking than controls (43% vs. 25%).

Adjusted odds of health behaviors

Cancer survivors

In multivariable models, male survivors were no more likely to have heavy alcohol use, current or former tobacco use, or be physically inactive than controls (Table 4). Female survivors had similar alcohol and tobacco use as controls, but had greater odds of being physically inactive [Odds Ratio (OR) 1.5; 95% CI, 1.2–1.8].
Table 4

Adjusted odds of health behaviors

 

Heavy drinking OR (95% CI)

Current smoking OR (95% CI)

Former smoking OR (95% CI)

Physical inactivity OR (95% CI)

Male non-cancer control (ref)

1.0

1.0

1.0

1.0

Male survivors, any cancer

1.3 (0.7–2.3)

0.8 (0.5–1.3)

1.3 (0.8–2.1)

1.0 (0.7–1.4)

Male non-cancer control (ref)

1.0

1.0

1.0

1.0

Male survivors, CRC

1.7 (0.5–5.7)

0.5 (0.1–1.9)

3.1 (0.7–13.5)

0.7 (0.3–1.6)

Male non-cancer control (ref)

1.0

1.0

1.0

1.0

Male survivors, prostate

1.1 (0.5–2.2)

0.8 (0.4–1.6)

0.9 (0.4–1.8)

0.8 (0.5–1.2)

Female non-cancer control (ref)

1.0

1.0

1.0

1.0

Female survivors, any cancer

1.4 (0.9–2.2)

1.1 (0.8–1.5)

1.2 (0.9–1.6)

1.5 (1.2–1.8)

Female non-cancer control (ref)

1.0

1.0

1.0

1.0

Female survivors, CRC

0.7 (0.1–3.5)

1.0 (0.3–3.8)

0.8 (0.2–3.3)

1.4 (0.6–3.6)

Female non-cancer control (ref)

1.0

1.0

1.0

1.0

Female survivors, breast

1.2 (0.6–2.2)

0.8 (0.5–1.3)

1.5 (0.9–2.4)

1.5 (1.1–2.0)

Female non-cancer control (ref)

1.0

1.0

1.0

1.0

Female survivors, Gyna

1.9 (0.9–4.3)

1.8 (1.2–2.8)

0.8 (0.4–1.5)

1.3 (0.8–2.1)

Multivariable logistic regression adjusted for age, race, education, and physical and mental health status.

Former smoking models only include former and current smokers (never smokers excluded).

MA Behavioral Risk Factor Surveillance System, 2006–2008.

Abbreviations: CI Confidence Interval, CRC Colorectal, Gyn Gynecologic.

aGynecologic cancer includes cervical, uterine, and ovarian.

Site-specific survivors

Separate multivariable models for male survivors of CRC or prostate cancer found no differences from controls for any unhealthy behavior (Table 4). Models for female CRC survivors also did not detect any statistically significant differences from controls. However, breast cancer survivors continued to have greater physically inactivity (OR 1.5; 95% CI, 1.1–2.0), while gynecologic cancer survivors had higher odds of current smoking (OR 1.8; 95% CI, 1.2–2.8).

Discussion

We illustrate unhealthy behaviors of cancer survivors residing in Massachusetts. Specifically, after adjusting for physical and mental health, female survivors, especially of breast cancer, were more likely to be physically inactive. Gynecologic cancer survivors were more likely to currently smoke, even after controlling for physical and mental health status. Given the known benefits of tobacco cessation and physical activity, these results highlight specific survivor subgroups to target interventions aimed at reducing health risks.

An elevated rate of former smoking among survivors overall is similar to others’ findings and may reflect successful secondary prevention [7, 13]. Concurrently, the higher odds of current smoking in gynecologic cancer survivors (which includes cervical cancer) is consistent with previously reported cervical cancer survivors’ behavior [68] and warrants further attention. We expanded on this association by controlling for mental health since higher rates of depression are seen in cervical cancer survivors [14], and depression is associated with tobacco use [15]. We also adjusted for race to account for the higher proportion of poor health behaviors in minority survivors compared to white survivors [16]. Finally, our findings may be related to the fact that tobacco increases host susceptibility to human papilloma virus [17].

In evaluating physical inactivity, we used a relatively low threshold compared to other studies that required minimum exertion or duration. Even still, in adjusted analyses, female survivors were more likely to be physically inactive. When analyzed by cancer site, breast cancer survivors were significantly more inactive than controls. This contrasts to others’ findings where after adjustment survivors were no different from controls [7], or even more likely to be active [8].

Different definitions of physical activity may explain these discrepancies or there may be uncontrolled confounding. Both poor physical health and depressive symptoms have been associated with lower physical activity in breast cancer survivors [18]. Grimmett et al. did adjust for the presence of arthritis [13], but other comorbidities have physical limitations. We controlled for self-perceived physical health, which has been associated with physical activity restriction [10].

Limitations

Our findings should be interpreted in the context of the following limitations. The BRFSS interviews non-institutionalized individuals with land telephones and results may not generalize to institutionalized populations or those with only cellular telephones. All cancer history is self-reported and subject to inaccuracy, but previous research has demonstrated high quality of cancer reporting via the BRFSS [19]. Data on cancer stage and treatments are unknown, but it is unclear how each would affect the impact a cancer diagnosis has on behavior change. Given the observational nature of the study we can only assess association, not causation. The small sample sizes for subgroups of specific cancers, especially colorectal and prostate, may have been underpowered to show an association. However, this does not detract from those results which were statistically significant. We are limited by lack of data on temporality, including pre-diagnosis behaviors and time since diagnosis. Behaviors pre-diagnosis can predict post-diagnosis habits [5], and we cannot determine behavioral changes due to a cancer diagnosis. Additionally, behavioral changes may occur proximal to the diagnosis, but not be sustained with time [18]. Finally, the BRFSS options for cancer site combined cervical, uterine, and ovarian cancers into one response choice of gynecologic cancer, precluding further cancer type stratification. The relatively younger age of the gynecologic cancer survivors, along with the presence of a screening test and better survival, may indicate a higher proportion are cervical cancer survivors, but this needs to be addressed in future studies.

Implications for cancer survivors

This study provides guidance to clinicians and public health professionals for targeted interventions to improve healthy behaviors in cancer survivors. Recognizing patterns on the state level enables the appropriate public health agencies to direct funds and programming resources to those most in need. This is highlighted by the fact that MA has similar, but not identical, findings of cancer incidence and health behaviors as compared to a national sample [20]. Age adjusted invasive cancer incidence and mortality in MA is greater than the national rate. Further, MA state level estimates of heavy drinking for men and women are higher than national prevalence estimates, while smoking rates and physical inactivity are lower. As individual states may have different behavior profiles for survivors, programs can be suitably tailored. Perhaps most striking in our study is the high rate of current tobacco use among gynecologic survivors. Unlike age, genetics, and other immutable factors, smoking is modifiable. Its numerous associations with risk for cancer and other medical conditions make it a prime target of intervention. This subgroup is younger and investments in their health may have more time to show health and economic returns.

Greater physical inactivity in breast cancer survivors also raises concern for subsequent adverse health. Vigorous exercise is associated with lower all-cause mortality in cancer survivors [21]. Engagement in physical activity has been shown to benefit breast cancer survivors, improving health outcomes and health related quality of life [22, 23]. Even with a liberal definition of physical activity, we found that breast cancer survivors had more inactivity. These survivors, who continue to grow in number, may need additional resources targeted to increasing their activity levels.

Welcomed improvements in diagnosis and treatment will also lead to a larger population of cancer survivors, with attendant increased risk for other cancers and medical conditions. Successful secondary and tertiary prevention to promote tobacco cessation and adoption of physical activity may mitigate some of these risks. With accurate assessment of populations who may derive the most benefit from interventions, public health programs can most effectively direct limited resources. Ongoing research is warranted to determine the most effective means to initiate and maintain adoption of healthy lifestyle behaviors.

Acknowledgements

Thank you to Michael Winter, MPH for his statistical assistance.

Copyright information

© Springer Science+Business Media, LLC (outside the USA)  2010