Cancer worries and uptake of breast, cervical, and colorectal cancer screening: A population-based survey in England

Being a little worried about cancer in general seems to facilitate uptake of cancer screening, an effect that has been observed across breast,^1–3 colorectal,^3–6 and prostate⁷ cancer screening. However, worry about cancer does not always motivate early detection behaviour,^8–11 and some studies find mixed effects. Using the Health Belief Model as a framework, a large population-based study in Singapore found that worry about one’s perceived susceptibility to colorectal cancer was associated with increased participation in colorectal cancer screening, but conversely, fear of diagnosis (measured as a potential barrier) was associated with decreased participation.⁶ Similarly, a study of older adults in the UK found that, while being worried about cancer was associated with increased intentions to participate in flexible sigmoidoscopy screening, feeling uncomfortable when thinking about cancer decreased actual screening attendance among participants intending to be screened.⁴

This mixed evidence for the effects of cancer worry on screening uptake may partly be explained by the use of heterogeneous measures which commonly lack specificity. The construct variously called ‘cancer worry’, ‘cancer fear’, or ‘cancer anxiety’ has been poorly defined and operationalised. Most population-based surveys to date have measured general feelings of worry about cancer, leaving the precise object of cancer worry unspecified.^1,12,13 To better understand the way in which cancer worry might affect screening participation, it may be important to distinguish general cancer worry (an unspecified worry about cancer broadly) from specific cancer worry (concerning a specified object of worry). Consedine et al.¹² suggest that because general cancer worry may be more difficult to ‘pin down’, it may (in many cases) motivate screening attendance in an attempt to reduce a feeling of diffuse anxiety. On the other hand, they suggest that specific worries, such as worry relating to pain, can be regulated by avoiding screening, and thus act as deterrents. Indeed, a narrative review of quantitative studies found that those measuring specific worries about cancer mainly described deterring effects.¹³ More recently, a systematic review of qualitative research suggested that the behavioural effects may differ depending on the focus of worry.¹⁴ Although general worry about getting cancer appeared to motivate screening uptake, specific fears of diagnosis, treatment, or death were described as barriers in some studies included in the review.^15–20

Specific cancer worries appear not only to differ in their behavioural effects but also their prevalence. In our previous study,²¹ about two-thirds of UK adults reported that they would worry about the emotional impact and threat to life if diagnosed, while half would worry about the financial implications and treatment. Fewer (one quarter) believed they would worry about the social impact, for example, on their important relationships. Sociodemographic variation in cancer-related worries was also observed, corroborating previous findings that women and younger people worry more about cancer.^22–24 In addition, those from ethnic minority backgrounds were found to be more worried about the potential social consequences than those from White ethnic backgrounds, but were no more likely to worry about treatment or death.²¹ Other studies have observed higher cancer worry among those of a lower socio-economic status.^22–24 However, uptake of cancer screening shows the opposite pattern, being lower among socio-economically deprived and ethnic minority groups and higher (for colorectal cancer screening) in women.^25–27

To date, the behavioural associations of general cancer worry versus specific cancer worry have not been evaluated systematically in a single sample or across different cancer screening programmes. Cancer worry is potentially modifiable, so an in-depth understanding of which worries are associated with uptake may contribute to an evidence base, which could usefully inform interventions designed to optimise informed uptake of screening programmes. We carried out further analyses of data reported in our previous study.²¹ We examined the association of general cancer worry and specific worries about the consequences of diagnosis with self-reported uptake of colorectal, breast, and cervical screening.

Data had been collected as part of a population-based survey of 2048 adults aged 18–70 living in England in April and May 2016. This study used data from a subsample (n = 1694). Participants were interviewed at home by a market research agency (TNS International) as part of a weekly omnibus survey using computer-assisted personal interviewing. All participants gave verbal consent. Sampling points for recruitment were defined based on the 2001 Census small area statistics and the Postcode Address File and selected using stratified random probability sampling. Quota sampling based on age, gender, children in the home, and working status was used at each location. Weights were supplied by TNS to adjust for non-response bias, which were derived from the latest mid-year census estimates to ensure the distribution of sample characteristics was nationally representative. Institutional ethical approval was obtained (UCL REC: 5711/002).

Analyses were restricted to three subsamples eligible for the three NHS screening programmes.^28–30 For breast cancer screening (by mammogram), we included women aged 50–70 (n = 413). For cervical screening (by liquid-based cytology (Pap) test), we included women aged 25–64 (n = 808). For colorectal cancer screening (by faecal occult blood test), we included men and women aged 60–70 (n = 473).

Uptake of colorectal, breast, and cervical screening was self-reported. Responses were dichotomised for analysis as ‘regular uptake’ (those who had participated every time they were invited) and ‘irregular uptake’ (those who had not participated every time or had never participated). We grouped ‘never’ and ‘irregular’ attenders together because they comprise a non-adherent group. This also provided greater statistical power.

The frequency of general cancer worry was measured with an item adapted from the Health Information National Trends Survey³¹ (for an example study see Moser et al.³): ‘How often do you worry about your chances of getting cancer?’. The item was scored on a 5-point Likert scale and recoded into three categories for analysis (‘never’, ‘occasionally/sometimes’, and ‘often/very often’) to identify any curvilinear associations with screening participation, as have been found previously.^8–10,32 Collapsing categories was necessary to provide a sufficient number of cases. In addition, we measured 12 specific worries about cancer, which were adapted from an existing scale of worries about recurrence in breast cancer survivors.³³ As previously described,²¹ the adapted items form two subscales: (i) worries about the emotional and physical consequences of a cancer diagnosis, and (ii) worries about the social implications of a cancer diagnosis. Sum scores for these two subscales were created for participants with complete responses on all six subscale items. All items were preceded by the phrase: ‘If you were diagnosed with cancer, would you worry that…’ (e.g. ‘…you would require surgery?’), and scored on a 4-point Likert scale: ‘not at all’, ‘slightly’, ‘quite a bit’, and ‘extremely’ to measure their intensity. Respondents who had previously been diagnosed with cancer were excluded, to avoid distress. ‘Don’t know’ or ‘refused’ responses were coded as missing.

Sociodemographic variables consisted of age, gender, ethnicity, social grade, and marital status and were measured with single items. Ethnicity was recorded using the 2011 Census question³⁴ and was recoded as ‘White’ versus ‘Non-White’ due to small numbers in the Non-White categories. Social grade was used as an indicator of socio-economic status and was recoded into four categories using the National Readership Survey grades which are based on the occupation of a household’s chief income earner.³⁵ Marital status was recoded as ‘married/living as married’, ‘single’, and ‘widowed/divorced/separated’.

We further excluded those with missing data on the variables of interest. We report the characteristics of each subsample using weighted and unweighted data. In addition, we report the weighted response distribution for general cancer worry and the weighted means and standard deviations for each of the 12 specific cancer worries and their two subscales. We then used logistic regression analysis (using unweighted data) to examine the association of general cancer worry and the specific cancer worry subscales with screening participation, in unadjusted analyses and analyses adjusted for sociodemographic factors.

For descriptive analyses, adjustment weights were assigned to each case, so that respondents with characteristics that were under-represented (relative to the national population) were given a higher weight than those from groups that were relatively over-represented. This aimed to ensure the distribution of sample characteristics was nationally representative, so that proportions and means were more reliable estimates for the wider population. Weighted data were not used in regression analyses because these associations were considered to be independent of sample composition and were adjusted for demographic factors in multivariate models. All analyses were carried out using SPSS version 22.0 and an alpha level of 0.05.

After excluding those with a cancer diagnosis (N = 38 (9%) for breast, N = 40 (5%) for cervical, and N = 52 (11%) for colorectal) and missing data (N = 52 (13%) for breast, N = 97 (12%) for cervical, and N = 53 (11%) for colorectal), the samples for analyses consisted of N = 323 for breast screening, N = 671 for cervical screening, and N = 368 for colorectal screening. The weighted and unweighted sample characteristics are presented in Table 1.

Consistent with the age ranges for the respective screening programmes, the mean age of the cervical screening subsample (M = 44 years, SD = 12) was younger than for the other two subsamples (breast M = 60 years, SD = 6; colorectal M = 65 years, SD = 4). Thirteen per cent of the cervical screening subsample came from an ethnic minority background, with a lower proportion for the breast (7.7%) and colorectal cancer screening subsamples (5.9%), consistent with 2011 Census data for these age ranges.^34,36 Across all three screening programmes, about a third (31–33%) of participants were from social grade AB (the highest grade), a quarter (22–26%) from social grade C1, 17–22% from social grade C2, and 21–28% from social grade DE (the lowest grade). About two-thirds (64–72%) were married. Just over half of the colorectal screening subsample was female (52%). Self-reported regular screening uptake in the three subsamples was highest for breast screening (83%), followed by cervical screening (76%), and colorectal screening (68%).

With regards to general cancer worry, we found that a substantial minority of participants (29–44%) never worried about their chances of getting cancer (Table 2). The majority (53–62%) occasionally or sometimes worried about cancer, while only 4–10% worried often or very often. From the crude percentages, regular breast and colorectal cancer screening participation was most frequently reported by those who worried about their chances of getting cancer occasionally or sometimes. For example, regular participation in breast screening was reported by 85% of those who occasionally or sometimes worried, compared with 80% of those who never worried and 70% of those who worried often or very often (Table 3). The same pattern was observed for cervical screening, although the differences between the groups were less pronounced. However, unadjusted and adjusted logistic regression analyses did not show any significant associations of general cancer worry with self-reported screening participation. The Nagelkerke R Square values indicated that approximately 9, 6, and 10% of variance in breast, cervical, and colorectal cancer screening uptake, respectively, was explained by the adjusted models.

Worry about the emotional and physical consequences of a cancer diagnosis was not associated with self-reported participation in breast or cervical cancer screening. However, higher worry was associated with increased odds of reporting regular participation in colorectal screening (odds ratio (OR) (unadjusted) = 1.46, 95% confidence interval (CI) 1.09–1.95), and this association remained significant after adjusting for demographic factors (OR(adjusted) = 1.41, 95% CI 1.04–1.90; Table 3). The adjusted models explained approximately 8, 6, and 11% of variance in uptake for breast, cervical, and colorectal cancer screening, respectively. The response distributions for all specific worry items are reported in Online Supplement 1.

Higher worry about the social consequences of diagnosis was associated with decreased odds of reporting regular participation in breast (OR = 0.63, 95% CI 0.43–0.92) and cervical screening (OR = 0.77, 95% CI 0.62–0.97) in unadjusted analyses. However, these associations were no longer statistically significant after adjusting for sociodemographic factors (OR = 0.69, 95% CI 0.45–1.04 for breast and OR = 0.82, 95% CI 0.65–1.03 for cervical screening; see Table 3), even though the associations with sociodemographic covariates remained unchanged (results not reported). In addition, those reporting worry about the social consequences of diagnosis were no more or less likely to have participated regularly in colorectal cancer screening. R square statistics indicated that 9, 6, and 9% of variance in uptake of breast, cervical, and colorectal cancer screening was explained by the adjusted models.

This is the first population-based study to compare the associations between general and specific cancer worries and uptake of breast, cervical, and colorectal cancer screening. In our samples, general cancer worry was not related to regular participation, but there was some evidence that specific worries about the consequences of diagnosis were. Worrying about the emotional and physical consequences increased the likelihood of having regularly participated in colorectal screening. Worries about the social implications were associated with a reduced likelihood of having regularly attended breast and cervical screening but only in unadjusted analyses. Thus, it is possible that specific worries about cancer may have different patterns of association with uptake of the different screening programmes.

Previous studies have often found that a low level of general worry about cancer is associated with higher screening attendance,^2,3,13 but not always,^8–11 and we did not find this association in our sample. Our findings seem to suggest a curvilinear association between general cancer worry and screening uptake for breast and colorectal cancer screening. However, any analyses of those worrying about cancer ‘often or very often’ may have been restricted by the very small number of people in this category (4–10%), which echoes the low prevalence of high cancer worry in the general population.^10,22,37 Most studies that have reported significant associations of general cancer worry and screening uptake have included much larger samples,^2,3 and we therefore cannot rule out an association. Future studies should consider oversampling those with high levels of cancer worry.

Despite the relatively small sample sizes, we found that one of the two classes of specific cancer worry was associated with previous screening participation. Previous studies have concluded that while general cancer worry may facilitate cancer screening uptake, specific worries may act as deterrents.¹² Contrary to this, our findings suggest that the association of specific worries with screening uptake may not be uniformly negative and in some cases, may increase uptake. Specifically, those worrying about the emotional and physical consequences of diagnosis were more likely to have regularly participated in colorectal cancer screening, but the same was not true for breast or cervical screening. We also found preliminary evidence that fears of the social consequences of a cancer diagnosis may be associated with lower uptake of breast and cervical screening. However, these results were only statistically significant in unadjusted analyses and should be tested using a larger sample.

We therefore hypothesise that the direction of association between specific cancer worries and screening uptake may also depend on the nature of the cancer for which screening is carried out. For example, we speculate that concern about the physical consequences of diagnosis could be particularly salient for colorectal cancer screening because the bowel is a vital organ, and the physical health implications (i.e. colostomy) may be perceived as more severe. It is possible that people realise that they can avoid the most negative consequences by catching it early through screening. Further research is needed to investigate the root causes of worries about the emotional and physical consequences of a colorectal cancer diagnosis, to better understand why these may motivate participation in screening.

Our findings contribute to the debate about the role of specific cancer worries in screening uptake. While preliminary, they may help to direct further research which investigates both the type of specific worries that could affect informed participation, as well as the direction of their effect. A useful theoretical framework from which to build further evidence may be Witte’s Extended Parallel Process Model, which proposes that the way in which an individual behaves in response to a fearful stimulus (i.e. health threat) is partly a function of the perceived efficacy (self-efficacy and response efficacy) of the different behavioural responses for dealing with the threat.³⁸ Our study begins to clarify the nature of those fears that may affect screening behaviour, and we propose that worries affecting response efficacy (e.g. invasive treatments for screen-detected cancers) and self-efficacy (e.g. concern that social implications are not manageable) could be studied to better understand their relationship with screening uptake. Future research would also do well to use comprehensive measures which incorporate other specific sources of worry, in particular, worries about the screening test itself, which have been shown to deter screening participation.⁷

Our study had several limitations. The study design was cross-sectional, which means we cannot infer the direction of the associations we observed between worry and screening uptake. Cancer worry is likely to be a dynamic construct, influenced by previous screening behaviour. Our findings are therefore preliminary, but should direct the design of further studies that prospectively investigate the behavioural effects of general and specific cancer worry. As mentioned above, some of our analyses were also limited by relatively small sample sizes, particularly those adjusting for sociodemographic differences. This meant that we could not explore the association between each specific worry and screening uptake individually. Furthermore, although we have taken care to include measures of the most important worries about cancer, the subscales for the specific cancer worries have not been validated, and do not specify a particular cancer site (e.g. colon). Other sources of worry could be included in future studies (e.g. about screening test procedures) in relation to specific cancer types. In addition, we only measured frequency of general cancer worry using a single item and compared this with measures of intensity of specific worries. This was intended to increase comparability with previous studies, which have often used frequency measures. However, future studies should also measure the intensity of general cancer worry and use multiple item measures to improve reliability. Finally, screening uptake was self-reported during face-to-face interviews, which means it may have been over-reported due to social desirability bias. Indeed, figures for screening uptake were higher than national figures for breast (71% nationally³⁹ versus 83% in our study), cervical (72%⁴⁰ versus 76%, respectively), and colorectal (58%⁴¹ versus 68%) cancer screening. However, a recent study looking at the accuracy of self-reported colorectal screening uptake found that respondents were nearly 95% accurate when reporting ‘ever’, ‘repeat’, or ‘never’ participation.⁴²

This is the first study to suggest that specific worries about cancer may be associated with cancer screening uptake differently across breast, cervical, and colorectal cancer screening programmes. Furthermore, these specific worries have potential to both motivate and deter screening behaviour. Further research is needed to improve knowledge of the specific worries associated with participation in each programme, to understand how they affect participation, and by which mechanism. That knowledge could then be used to design targeted cancer communication strategies and interventions to optimise informed participation in screening.

We acknowledge the intellectual contribution made by Professor Wardle to the design of this study before her death in October 2015.