Cancer Causes & Control

, Volume 19, Issue 7, pp 775–782

Gynecological cancer survivors’ health behaviors and their associations with quality of life


    • Institute of Health and Biomedical Innovation, School of Public HealthQueensland University of Technology
    • Cancer and Population Studies GroupQueensland Institute of Medical Research
  • Elizabeth G. Eakin
    • Cancer Prevention Research Centre, School of Population HealthUniversity of Queensland
  • Monika Janda
    • Institute of Health and Biomedical Innovation, School of Public HealthQueensland University of Technology
  • Diana Battistutta
    • Institute of Health and Biomedical Innovation, School of Public HealthQueensland University of Technology
Original Paper

DOI: 10.1007/s10552-008-9140-y

Cite this article as:
Beesley, V.L., Eakin, E.G., Janda, M. et al. Cancer Causes Control (2008) 19: 775. doi:10.1007/s10552-008-9140-y



To evaluate health behaviors (smoking, physical activity, fruit and vegetable intake) and body mass index of gynecological cancer survivors and their association with quality of life.


We collected cross-sectional mail survey data from 802 gynecological cancer survivors (56% response rate) The questionnaire included validated measures of health behaviors, quality of life, clinical status, and demographics.


Depending on gynecological cancer subtype, 61–68% were insufficiently active or sedentary, 19–44% were obese, 31–41% had less than two daily serves of fruit, 82–89% had less than five serves of vegetables and 6–21% smoked. Significantly more endometrial than other cancer survivors were obese (44%; 95% CI 41–47%); smoking was more prevalent among cervical cancer survivors (21%; 95% CI 19–23%). Using general linear modeling we identified a positive linear trend of physical activity level with quality of life (p = 0.039). Being insufficiently active or sedentary was also associated with not meeting the guidelines for fruit and vegetable consumption, with smoking and with overweight/obesity.


This research demonstrates the importance of integrating physical activity into follow-up of gynecological cancer survivors. Interventions to improve physical activity and test the causal effect of exercise on quality of life are a future research priority.


Gynecologic carcinomaQuality of lifePhysical activityRisk factors


Gynecological cancers account for approximately 18% of all female cancers worldwide [1], and are a major source of mortality and morbidity [2]. A range of site-specific difficulties can persist after treatment, including sexual problems (12%), bladder dysfunction (11%), vaginal problems (e.g., recurrent infections) (10%), and lower limb lymphedema (10%) [3, 4]. With the exception of ovarian cancer, most gynecological cancer diagnoses are associated with good survival rates; the 5-year survival rate for cervical cancer is 75% and endometrial cancer is 83%, compared to 55% for ovarian cancer [5]. These rates are generally increasing over time in Australia [5] but have remained stable in the US since 1999 [6]. Considering this, research concerning the quality of long-term survival is imperative.

Related to this survivorship focus, there is a growing literature which documents the prevalence of lifestyle behaviors among cancer survivors [79] and their positive influence on quality of life (QoL) [1025] and more recently on recurrence and survival [2628]. A substantial number of reports suggest that cancer survivors adopt lifestyle changes in the hope of achieving improved health after treatment [29, 30], although data from the small number of population-based surveys suggest that their levels of health behaviors largely approximate those of the general population, and thus leave substantial room for improvement [79]. However, most of this work has focused on breast cancer, with more recent studies of colorectal and prostate cancer [30, 31].

Very few studies have evaluated health behaviors or their relationship with important outcomes such as QoL, in women diagnosed with gynecological cancers. Two studies examined the relationship between physical activity and QoL, but did not consider other health behaviors. These indicated that 70% of endometrial and 69% of ovarian cancer survivors were not meeting public health physical activity guidelines [13, 21]; however, for those who were sufficiently active, positive associations were found with QoL. These studies both utilized a US sample and one [21] did not use a representative sampling frame; rather they invite a sample of physicians to recruit their ovarian cancer patients, of which 28% (n = 150) did not respond.

The aim of this research was to determine, using population-based sampling, the prevalence of Australian women with gynecological cancer meeting guidelines [32, 33] for four health behaviors (smoking, physical activity, fruit and vegetable intake) plus overweight/obesity. We also examined the associations of these health behaviors with QoL and explored the overall health behavior profile of women who are insufficiently active or sedentary given the existing evidence for the importance of physical activity for cancer survivorship and survival [2628].


Procedures and participants

Ethical approval was obtained from all participants’ treating hospitals and from Queensland University of Technology. Women were randomly sampled from the Queensland Gynaecological Cancer Registry (QGCR), within strata of gynecological cancer type (cervical, endometrial, ovarian and other) and survival phase post-diagnosis (3 months–1 year, beyond 1–3 years, and beyond 3–5 years). Eligible women were 18 years or older, English speaking, and able to complete the questionnaire.

Once women had been selected from the QGCR, their treating doctors were approached for permission to contact these women (Fig. 1). Patients then were mailed an initial letter, signed by their doctor, and an information sheet, consent form, questionnaire and reply-paid envelope. Non-respondents were sent a follow-up letter at 2 weeks, were telephoned at 4 weeks, and were mailed a final reminder letter after 6 weeks (including a replacement questionnaire package).
Fig. 1

Flow of participant recruitment

The QGCR is a quasi-population-based registry covering approximately 85% of Queensland gynecological cancer patients. A comparison to summary data from the state-based registry, which has universal coverage of persons with cancer, indicated that ovarian cancer patients were slightly under-represented in our sampling frame.

Questionnaire measures

Socio-demographic characteristics

Information was obtained on respondents’ age, marital status, education level, employment, gross annual household income, number of children and geographical location.

Physical activity and BMI items

Eight standard questions from the Active Australia Survey [34] were used to classify women into those who were sufficiently active (150 min per week using the sum of walking, moderate activity and vigorous activity (weighted by two) over five or more sessions), insufficiently active (1–149 min per week), and sedentary (no physical activity), as per the National Physical Activity Guidelines for Australians [33]. Body mass index (BMI, kg/m2) was derived from self-reported weight and height, and grouped into underweight (BMI < 18.5), normal weight (18.5–24.9), overweight (25.0–29.9), and obese (≥30).

Fruit and vegetable items

Using validated items from the Australian National Health Survey [35], respondents were asked to report on average in the last month, how many servings of fruits or vegetables they had eaten each day. These were categorized into whether women met the Australian recommendations [32] of two servings of fruit or five servings of vegetables per day.

Cigarette smoking item

Women were asked if they were a current smoker, ex- or life-long non-smoker.


The Functional Assessment of Cancer Therapy-General (FACT-G) questionnaire [36] was used to measure QoL. This is a reliable and valid measure that has 28 items across four subscales: physical; social/family; emotional; and functional well being. The reliability for subscales ranges from 0.82 to 0.88 [36].

Morbidity items

Women were asked about physical and emotional co-morbidities they had ever been diagnosed with and morbidities that resulted from their gynecological cancer treatment. We grouped the following co-morbidities into none, one, two, three, four, or more co-morbidities: Kidney disease or problem, heart disease, diabetes, hypertension/high blood pressure or stroke, serious respiratory disease or problem, high cholesterol, hearing loss, HIV/AIDS-related illness, arthritis, depression, anxiety/panic attacks, eating disorders, fatigue, or other. We also asked whether they had been diagnosed with lymphedema or had undiagnosed swelling of unknown cause [4], whether their gynecological cancer had made sexual relations too difficult or uncomfortable and whether they were fitted with a stoma.

Medical history

The following characteristics were extracted from the QGCR hospital charts: diagnosis and histological type of cancer, time since diagnosis, time since last treatment, cancer stage at diagnosis, type of surgery, receipt of chemotherapy, radiotherapy, hormone therapy, and lymph node removal.

Statistical methods

Percentages (and 95% confidence intervals, CI, for key outcomes) were used to describe women with each cancer type who met the recommendations for servings of fruits or vegetables, level of physical activity, and smoking status and by BMI class. Absolute group differences of greater than 8% were considered clinically meaningful. This difference between groups reflects a similar predefined clinical difference as outlined in the FACIT Effect Size Tables [37]. Statistical support for these reported differences at the bivariate level was based on likelihood ratio chi-square tests. In order for estimated percentages to reflect the Queensland population of gynecological cancer survivors, sampling weights were defined and applied to descriptive statistics to redress the unequal probabilities of selection within the cancer type and time since diagnosis strata.

A multivariable general linear model was used to test for differences in QoL between gynecological cancer survivors meeting and not meeting public health guidelines [32, 33]. Due to limited numbers of women in some categories and data attrition in the multivariable model, stratified analysis by cancer subgroups was not possible. As QoL was not normally distributed, a natural log transformation of QoL was used in these models to ensure that model-estimated geometric means reflected median averages rather than means. In addition to health behaviors, the model was fitted with demographic and clinical covariates including age, marital status, education level, employment, household income, number of children, geographical location, cancer type, time since diagnosis, stage, surgery, chemotherapy, radiotherapy, remission status, lymphedema status, sexual dysfunction, stoma status, and number of co-morbidities. Multicollinearity was tested for suspected highly correlated independent variables; none was detected.

Results are reported for all clinically meaningful differences, with a statement of whether or not the difference also had support of statistical significance. The FACIT effect size tables [37] indicate that a meaningful group difference in the FACT-G score of greater than 7.9 is considered substantial, but that the minimally important difference might be somewhat smaller. Courneya et al. [13] in their study of endometrial cancer survivors reported a clinically meaningful difference of 5.4 and other FACT scales for breast, colorectal and lung cancer use minimal important differences of five to eight points [38]. Therefore, differences above five in this exploratory study were considered clinically meaningful.

Logistic regression modeling was used to determine the other health behavior correlates of women who were insufficiently active or sedentary. Variables included in this model were the same as within the general linear model. Reference groups were chosen so that odds ratios were greater than 1.0, for ease of interpretation. Clinically meaningful correlates were identified as those with odds ratios above 1.7 and with informative 95% confidence intervals. Depending on the prevalence, this odds ratio approximately equated to an absolute difference of 5–8%, which is a comparable meaningful difference used in the FACT scale of 1–108. Throughout, statistical significance was based on two-tailed hypotheses and the conventional 5% level.

Multivariable models were subject to missing data. However, standard procedures for missing quality of life data were employed (i.e., overall quality of life scores were accepted as long as overall item response rate was greater than 80%) and covariates that had missing values that accumulated to around 10% or more of the sample were recoded to include a missing category in the model, so that selected cases included in the analysis was maximized for power. Furthermore, a missing values sensitivity analysis was conducted to determine whether those women who dropped out of the final models were any different from those women in the models. Results showed that the models included younger, more educated, and more employed, women (data not shown).



In total 802 women returned completed surveys; 56% of 1,420 eligible patients listed on the QGCR (Fig. 1). However, we excluded 27 women who were treated within the last month from the following analyses, as participation in health behaviors and quality of life is likely to be very different during the treatment period. Respondents were representative of the sampling frame, except that older women (70 years and older) were under-represented (Table 1).
Table 1

Characteristics of a sample of 775 gynecological cancer survivors in Queensland, Australia, 2004


% (n)

Age (years)a


9.5 (74)


16.8 (130)


27.0 (209)


27.1 (210)

70 +

19.6 (152)

Marital statusa


8.0 (62)

Married/living with partner

65.5 (508)


11.0 (85)


14.6 (113)


0.9 (7)

Employment statusa

In full time paid work

17.8 (138)

In part time or casual paid work

16.5 (128)

Home duties only—no paid work

16.6 (129)


30.7 (238)

Unable to work due to illness or disability

8.4 (65)

Other or missing

9.9 (77)

Type of gynecological cancerb


25.2 (195)


31.1 (241)


27.7 (215)

Other c

16.0 (124)

Time since diagnosisb

3–12 months

23.7 (184)

>1–3 years

41.4 (321)

>3–5 years

34.8 (270)

Time since treatmenta

1–2 months

3.2 (25)

3–12 months

25.9 (201)

>1 year

60.4 (468)


7.7 (60)


2.7 (21)


I and II

71.4 (553)

III and IV

23.6 (183)

No stage/don’t know

5.0 (39)



1.7 (13)

Vaginal or laparoscopic

15.4 (119)

Open abdominal

73.2 (567)

Open bowel resection

7.5 (58)


2.3 (18)



35.9 (278)


64.1 (497)


No radiotherapy

73.9 (573)

Internal brachytherapy

5.3 (41)

External beam radiotherapy ± brachytherapy

20.8 (161)

aPatient self-reported data

bQueensland Gynaecological Cancer Registry data

c‘Other’ includes: vulval, vaginal, peritoneal, and fallopian tube cancer

Health behaviors by type of gynecological cancer

Table 2 outlines the prevalence of health behaviors and BMI by cancer type.
Table 2

Prevalencesa of health behaviors, and body mass index, within gynecological cancer type in Queensland, Australia, 2004


Type of gynecological cancer

Cervical % (n)

Endometrial % (n)

Ovarian % (n)

Otherb % (n)





    Current smoker

20.9 (51)

6.1 (29)

9.6 (30)

10.4 (12)



39.3 (96)

29.4 (141)

28.7 (90)

34.8 (40)


    Never smoked

39.8 (97)

64.5 (309)

61.8 (194)

54.8 (63)


Fruit intake



    Inadequate servings

41.4 (94)

32.6 (156)

31.0 (94)

38.8 (45)


    Adequate servings (2/day)

58.6 (133)

67.4 (322)

69.0 (209)

61.2 (71)


Vegetable intake



    Inadequate servings

86.5 (192)

82.4 (392)

78.6 (228)

89.4 (101)


    Adequate servings (5/day)

13.5 (30)

17.6 (84)

21.4 (62)

10.6 (12)


Physical activity




19.1 (47)

23.6 (118)

26.8 (86)

29.0 (36)


    Insufficiently active

42.3 (104)

42.2 (211)

37.1 (119)

39.5 (49)


    Sufficiently actived

38.6 (95)

34.2 (171)

36.1 (116)

31.5 (39)


Body mass index




4.6 (10)

2.5 (11)

3.1 (9)

8.5 (9)


    Normal weight

53.2 (116)

26.7 (116)

39.0 (114)

32.1 (34)



23.4 (51)

27.1 (118)

28.4 (83)

31.1 (33)



18.8 (41)

43.7 (190)

29.5 (86)

28.3 (30)


aSampling weights were applied to redress unequal selection probabilities across sampling strata to estimate prevalences that reflect the Queensland population

b‘Other’ includes: vulval, vaginal, peritoneal, and fallopian tube cancer

cp-values are significance values associated with likelihood ratio chi-square tests

d‘Sufficient’ is defined as 150 min per week (using the sum of walking, moderate activity and vigorous activity (weighted by two)) over five or more sessions

Cervical cancer survivors were more likely to be current smokers (21%; CI 19–23%) compared with endometrial (6%; CI 5–7%), ovarian (10%; CI 8–12%) and ‘other’ gynecological cancer survivors (10%; CI 8–12%) (p < 0.001). Between 59 and 69% of women consumed adequate servings of fruit, and 11–21% consumed adequate servings of vegetables. In particular, fewer cervical cancer survivors meet the fruit recommendations than endometrial or ovarian cancer survivors (p = 0.046), and fewer ‘other’ cancer survivors meeting the vegetable recommendations than ovarian cancer survivors (p = 0.023). About a third of women (32–39%) were sufficiently active for health, 37–42% were insufficiently active and 19–29% were sedentary. Significantly more endometrial cancer survivors were obese (44%; CI 41–47%) compared to ovarian (30%; CI 27–33%), ‘other’ (28%; CI 25–31%), and cervical cancer survivors (19%; CI 17–21%) (p < 0.001).

The association of gynecological cancer survivors’ health behaviors with QoL

After adjustment for demographic and clinical variables, as well as other health behaviors, there was a positive linear trend in the association of level of physical activity with QoL (Table 3). The association of physical activity with QoL was clinically and statistically significant, with a median difference of 6.1 between sedentary and sufficiently active women (p = 0.039). There was no association between QoL and smoking, fruit consumption, vegetable consumption, or BMI class.
Table 3

Association of gynecological cancer survivors’ health behaviors with quality of life in Queensland, Australia, 2004



Quality of lifea

95% confidence interval


Physical activity level








    Insufficiently active





    Sufficiently activeb


















Fruit intake



    Inadequate servings





    Adequate servings (2/day)





Vegetable intake



    Inadequate servings





    Adequate servings (5/day)





Body mass index



    Under or normal weight















aGeometric means which approximate estimated median quality of life, are based on multivariable models adjusted for main effects presented as well age, marital status, education level, employment, household income, number of children, geographical location, type of cancer, time since diagnosis, stage, surgery, chemotherapy, radiotherapy, remission status, lymphedema status, sexual dysfunction, stoma status, and co-morbidities

b‘Sufficient’ is defined as 150 min per week (using the sum of walking, moderate activity and vigorous activity (weighted by two)) over five or more sessions

Characteristics of insufficiently active or sedentary women

Being insufficiently active or sedentary was significantly correlated with other poor health behaviors (Table 4). The odds of women being insufficiently active or sedentary were significantly higher in women who smoked, had inadequate fruit or vegetable intakes or who were obese.
Table 4

Health behavior correlates of insufficiently active or sedentary a gynecological cancer survivors



Adjusted ORb (95% CI)







2.05 (1.01–4.16)






Fruit intake



    Inadequate serves


1.66 (1.06–2.59)


    Adequate serves (2/day)




Vegetable intake



    Inadequate serves


1.94 (1.09–3.47)


    Adequate serves (5/day)




Body mass index (BMI)



    Under or normal weight






1.50 (0.93–2.43)




1.97 (1.18–3.29)


n: participant numbers; OR: odds ratio; CI: confidence interval

a‘Insufficiently active or sedentary’ is defined as less than 150 min per week (using the sum of walking, moderate activity, and vigorous activity (weighted by two)) or more than 150 min in less than five sessions

bOdds ratios are based on multivariable models adjusted for smoking, fruit intake, vegetable intake, BMI, age, marital status, education level, employment, household income, number of children, geographical location, type of cancer, time since diagnosis, stage, surgery, chemotherapy, radiotherapy, remission status, lymphedema status, sexual dysfunction, stoma status, and co-morbidities


This study adds to the limited number of population-based studies that examine the role of health behaviors in cancer survivorship and their relationship with improved QoL after cancer treatment. It is one of the first studies to examine health behaviors and QoL in gynecological cancer survivors, and it is the only study to report behavior prevalences by all three of the main gynecological cancer subtypes. Our study confirms the low levels of physical activity previously reported in endometrial [13] (30–34%) and ovarian [21] cancer survivors (31–36%). It further shows that the rates of physical activity are similar across cervical, endometrial, and ovarian cancer survivors (32–38%). In general, the percentage of gynecological cancer survivors meeting the physical activity guidelines is similar to that of a mixed group of cancer survivors reported in a study evaluating Australian National Health Survey data [9]. However, the percentage of cervical cancer survivors who were sufficiently active among our sample (39%) is higher than reported in a previous study of mixed cancer survivors (28%) [9]. This is probably due to cervical cancer survivors being younger on average, with physical activity decreasing with age [34].

Only 15% of gynecological cancer survivors met the Australian recommendations of five servings of vegetables per day, while 64% met the recommendation for two servings of fruit. American guidelines for fruit and vegetables are different, recommending five servings a day combined of fruit and vegetables, precluding direct comparisons. The other Australian population-based study [9] of cancer survivors categorized vegetables as “four or more” due to pre-coded response options, also making a comparison difficult.

This is one of the first studies to report the prevalence of health behaviors other than physical activity in the gynecological cancer population. Of particular concern were the higher prevalences of smoking in cervical cancer survivors (21%) and overweight/obesity in endometrial cancer survivors (71%), compared to other gynecological cancer survivors. These are both risk factors for the respective cancers, and it seems this disparity in cancer risk behaviors is continued into cancer survivorship. These subgroup differences are consistent with findings from American population-based samples [7, 13], although the actual prevalence of smoking in cervical cancer survivors was higher in the American study (46%). It is unclear why rates are lower in our study, especially given similar population levels of smoking in both countries.

Few studies have evaluated the relationship between health behaviors of gynecological cancer survivors and QoL; those that have are limited to physical activity and body weight. Our study confirms the positive association between QoL and meeting physical activity guidelines reported by Courneya et al. [13] in a study of endometrial cancer survivors. Stevinson et al. [21] observed a stronger association among ovarian cancer survivors. Courneya et al. [13] and our study both utilized population-based sampling from a registry, whereas Stevinson et al. [21] recruited through physicians. The difference in effect size may be due to the differences in sampling methods and their impact on representatives, or it may be that type of gynecological cancer is an effect-modifying variable. A study which included breast, prostate and colon cancer survivors showed effect modification by cancer types of the relationship of QoL to physical activity, such that there was no relationship within breast cancer survivors, a weak relationship within colorectal cancer survivors and a moderate relationship within prostate cancer survivors [20].

From a QoL perspective, our results suggest that targeting women who are insufficiently active or sedentary may be beneficial for improving well-being. Alternatively, the association found in this cross-sectional study may reflect the notion that women with a better quality of life are more able to exercise. Assuming the former and given that insufficient physical activity was correlated with being obese, not eating adequate servings of fruits or vegetables and with smoking, it is likely that a targeted approach may also provide opportunity to focus on improving other health behaviors. Cancer survivors may perceive their vulnerability of cancer recurrence to be decreased if they meet more than one lifestyle recommendation [20].

Despite the important information gained from this study, there are a number of limitations that need to be acknowledged. Firstly, while the cross-sectional design was appropriate to determine point prevalences and their differences between groups, causality was not possible to infer. Secondly, this was a Queensland, Australian-based study and, as such, may be limited in its generalizability. However, our findings were in general consistent with the international literature in this area. Third, the slight under-representation of ovarian cancer survivors in our sampling frame may mean that the overall prevalence of gynecological cancer survivors who are overweight or obese or who are smokers may be inflated, as these characteristics are less prevalent among ovarian cancer survivors. However, we overcome this by presenting prevalences by cancer subtypes. Fourth, given women 70 years and older were less likely to participate in the survey and older women are less likely to be physically active or smoke and more likely to eat fruit [39], overall prevalence’s of physical activity and smoking may be inflated and the prevalence for adequate serves of fruit may be understated. Finally, we did not measure alcohol consumption, another health behavior of importance.

In conclusion, the results from this study provide preliminary estimates of the prevalence of a variety of cancer-related risk factors among gynecological cancer subtypes of survivors in the Australian population. Furthermore, our results suggest that physical activity is an important independent correlate of QoL. Despite the evidence, efforts to encourage physical activity are not a routine part of the cancer treatment or rehabilitation process. Education by cancer clinicians and cancer support organizations about the benefits of physical activity, especially for conditions such as fatigue, anxiety, depression, and lymphedema, are important, as previous research has demonstrated the value of such advice [40]. An intervention trial designed to test the causal effect on QoL of exercise in gynecological cancer survivors is warranted and, if successful, would provide a stronger level of evidence for the promotion of exercise within the supportive care framework.


This study was funded by the Cancer Council Queensland and Queensland University of Technology. It was also made possible by access to participants via the Queensland Gynaecological Cancer Registry at Royal Brisbane and Women’s Hospital.

Copyright information

© Springer Science+Business Media B.V. 2008