The Effects of Colorectal Cancer Screening on Health Attitudes and Practices¹

Complacency about health, leading to deterioration in health behaviors, has been hypothesized to represent an undesirable consequence of negative screening results (1). However, surprisingly few studies have investigated the impact of screening on health behaviors. The specter of adverse behavioral effects was raised in the Telemark Study when the results showed trends toward “improvement in smoking category”³ and greater body mass index gains in the group who had a negative flexible sigmoidoscopy screening result compared with those who had polyps detected (2), although neither effect was statistically significant. Lower smoking cessation and smoking reduction rates were also observed in a small group of older adult smokers who had received a clear computed tomography scan for early-stage lung cancer compared with those receiving an abnormal result (3), although both groups had higher than expected quit rates, and the results were interpreted as showing that screening participation was a catalyst to positive health behavior change. Again, the association between screening outcome and smoking behavior only approached significance.

In this study, we evaluated attitudes toward and practice of health behaviors before and after screening in a large sample of adults attending for CRC⁴ screening as part of the United Kingdom FS Trial. Participants were a population sample of average-risk adults ages 55–64 years, identified from Family Health Services Authority registers for participating general practices (primary health care in the United Kingdom) in 14 centers across the United Kingdom. Local research ethics approval was obtained for each of the participating centers. Those eligible for inclusion in the trial were written to by their general practitioner with an information leaflet on CRC and FS screening and were asked whether, if invited, they would accept the offer of screening. The information leaflet described the aims of FS screening as the detection and removal of premalignant polyps and did not refer to possible causes of CRC. The leaflet was received before the prescreening questionnaire. Respondents who replied saying they were interested in screening were randomly allocated to screening or usual care. The design of the main FS trial has been described elsewhere (4).

Screening outcomes were described as negative when either no pathological specimens were detected or when pathological analysis of specimens showed no significant pathology (65%); lower risk when the pathology detected was considered minor [e.g., 1–2 small adenomatous polyps (<1 cm) with a tubular histology and mild to moderate dysplasia or <20 hyperplastic polyps; 30%]; and higher risk when 3 or more adenomas or 20 or more hyperplastic polyps were detected, or significant pathology was found (e.g., adenomatous polyps that were either large (>1 cm) or had tubulovillous or villous histology or severe dysplasia; 5%), and a colonoscopy was recommended (see Ref. 5 for additional details). Those eligible to receive the postscreening questionnaire were people who had attended FS screening (and colonoscopic screening, if recommended) and who had received a nonmalignant screening outcome.

Health attitudes and behaviors were assessed by questionnaire in a randomly selected subset (n = 5942) of the individuals in the screening arm of the trial in trial centers in Oxford, Portsmouth, and Swansea. Questionnaires were sent before the screening appointment and 3 months after screening. The attitude items were based on the measures used in the European Health Survey (6) and were as follows: “in general, how important do you feel the following health measures are? to avoid fatty foods; to eat plenty of fruit; to take regular exercise; for women to have a cervical smear test at least every 5 years; for women to have a breast screen (mammogram) at least every 3 years?”

Health behaviors were assessed as follows: “about how many servings of fruit do you eat (fresh, frozen or canned)”, “about how many servings of vegetables do you eat (including salad but excluding potatoes),” “do you take regular exercise each week,” and “do you smoke cigarettes at all nowadays” (all items from Ref. 6; Table 1). Response options for each question are shown in the table footnote.

Demographic details were also collected.

A total of 4644 of 5942 (78%) of prescreening questionnaires were returned. A total of 4329 (73%) of those invited, attended screening, of whom 3789 had completed the prescreening questionnaire. Of the latter, 20 were not eligible for follow-up, either because they were diagnosed with cancer (n = 16), or they did not have a colonoscopy (n = 4; Fig. 1). The response rate for the postscreening questionnaire among attenders who had completed the prescreening questionnaire was high (93%; 3535 of 3769), although slightly lower in the higher risk outcome group (85%) than the other groups (both 94%; χ² = 27.0, df = 2, P < 0.001).

Overall, 60% of those originally invited to participate in this study completed both questionnaires and attended screening (3535 of 5942), and they are referred to here as completers. The limited demographic information on the full sample shows a higher percentage of men than women completed the study (61 versus 58%), but there was no difference in age between completers and noncompleters. Additional information was available on those who had filled out the prescreening questionnaire, showing that completers were less socioeconomically deprived (indexed by educational qualifications and car and home ownership), reported higher fruit and vegetable consumption, and were more likely to report taking regular exercise and to describe themselves as nonsmokers than noncompleters. Among completers, the majority were white (98%) and married (81%) and approximately half had some educational qualifications (54%). The average number of daily servings of fruit (1.47, SD: 1.05) and vegetables (1.54, SD: 0.99) were broadly comparable with the national average in the United Kingdom (7), although well below the recommended five servings a day. Reported rates of doing regular exercise (66.6%) were higher, and smoking rates (13.9%) were lower than population figures for this age group (8), suggesting a measure of health selection into the trial.⁵

This study addressed the issue of whether health behaviors and attitudes became less positive in the group who received a negative (clear) screening outcome compared with the other two groups. No specific hypotheses were made about how the lower risk and higher risk positive groups might differ from one another in their health attitudes and behaviors. Results are reported here from participants for whom both pre- and postscreening data were available.

The analyses were also repeated for all those eligible for follow-up by carrying forward the prescreening score for cases with missing postscreening data (an intention-to-treat analysis). In this analysis, it was assumed that nonresponders had not changed their behaviors and beliefs from baseline to give a more conservative estimate of behavior change after screening and compensate for the slightly higher level of noncompletion of the postscreening questionnaire in the higher risk outcome group. This resulted in an addition of 234 cases, putting the overall response rate up marginally to 63% but did not alter the pattern of results so will not be discussed further.

The effect of screening outcome on continuous variables was assessed using repeated measures ANOVA, with time (pre-versus postscreening) as the within-subjects variable and outcome group and gender as between-subjects variables. A significant group-by-time interaction was taken as evidence for differences in behavior change after screening. The effect of screening outcome on the binary variables was assessed using logistic regression with prescreening health behavior scores, gender and screening outcome as the predictors, and postscreening health behavior scores as the dependent variable. Change over time was assessed in the continuous variables by examining the main effect of time within the ANOVAs and in the binary variables with conditional logistic regression.

Eating fruit [F_(13,440) = 33.9, P < 0.001] and taking exercise [F_(13,422) = 53.3, P < 0.001] were rated as more important after screening than before, but ratings for the importance of avoiding fatty foods and attending cervical and breast cancer screening did not change. There were no significant interactions with screening outcome group.

Fruit and vegetable intake increased [F_(13,445) = 41.4, P < 0.001; F_(13,435) = 17.5, P < 0.001] exercise increased (odds ratio: 1.91; 95% CI: 1.62–2.25; P < 0.001), and smoking rates decreased (odds ratio: 0.67, 95% CI: 0.46–0.98, P = 0.04) from before to after screening (Table 1). There was no evidence that change in health behavior related to screening outcome. Fruit intake and vegetable intake showed no time by screening outcome interaction [F_(23,445) < 1, ns and F_(23,435) < 1, ns, respectively]. Exercise status did not vary by screening outcome, controlling for pre-screening exercise (Wald = 0.38, df = 2, ns). The change in smoking status was more variable across outcome groups (Table 1). Among smokers at baseline, the percentages describing themselves as nonsmokers at follow-up across the three outcome groups were 15.5% (clear outcome), 10.5% (lower risk outcome), and 16% (higher risk outcome; χ² = 2.70; df = 2; ns).

Among nonsmokers at baseline, the percentages saying they were smoking at follow-up were 1.1% (clear), 2% (lower risk), and 2.6% (higher risk; χ² = 4.82; df = 2; P = 0.09), i.e., the opposite direction to a complacency effect. The effect of screening outcome approached significance as a predictor of postscreening smoking controlling for prescreening smoking and gender (Wald = 5.59; df = 2; P = 0.06), but the pattern of results shows this was towards lower quit rates in the lower risk outcome group compared with the other two groups (P = 0.106 for the quadratic contrast within logistic regression).

The present findings can be interpreted as broadly reassuring. Three of five health attitudes changed in a positive direction, and all four health practices showed a modest improvement after attendance at screening. Although the changes were small, they do at least provide some evidence against the view that screening leads to a less healthy lifestyle, at least in the short term. Furthermore, there was no evidence of complacency in the clear outcome group compared with those who had polyps detected.

In the Telemark study, there was a trend toward improved smoking status among those with higher risk polyps (2), and this finding was based on a comparison between participants with negative results and those with any polyp detected. However, in the latter study, everyone with polyps was referred for colonoscopy. In the present study, colonoscopy referrals were restricted to the higher risk group. This raises the possibility that the experience of more extensive medical interventions, rather than the detection of polyps per se, might be the factor that influenced smoking reduction in the Telemark study. In this study there was no difference between the clear outcome and higher risk outcome groups in the likelihood of smoking postscreening controlling for prescreening smoking status, but the sample size among those referred for colonoscopy was small, and larger scale studies may detect a significant difference. In addition, the present study looked at quit rate rather than reduction in number of cigarettes smoked and the former arguably provides a better indicator of positive health change. Although Hoff et al. (2) mention a significant change in quit rates in the abstract of their article, no statistic is reported in the text.

One limitation of this study and the other two discussed in this article is the absence of an unscreened control group. Given the concern over health complacency after a clear screening outcome and its potential effect on all-cause mortality raised by the Telemark study (2), there is an urgent need for future research to establish whether any differential smoking cessation or reduction rates come from lower quitting or smoking reduction in lower risk groups (a complacency effect) or higher quitting or smoking reduction in higher risk groups, which may represent the healthy screening catalyst effect. The use of an unscreened control group would also show whether changes in health beliefs and behaviors in the population differ from changes shown by the screening participants as a whole.

Generalization of these results to other populations is limited by the fact that it was carried out among 55–64-year-old British adults who had consented to take part in a trial of a new screening program and that only 60% of those invited to take part in the study completed it. Those completing the study reported better health behaviors than noncompleters. Although this is consistent with the findings of higher rates of health behaviors among those attending screening compared with nonattenders (9, 10, 11), less healthy groups may respond differently to screening outcomes and, in addition, may represent a subgroup in whom any deterioration in health behaviors would be more detrimental to health. Noncompleters also had higher levels of socioeconomic deprivation, and it is possible that responses to screening outcomes may differ in relation to understanding of the meaning of the different screening outcomes.

An additional limitation concerns the measures of health attitudes and behaviors, which were self-report and single items. This was a limitation imposed by the need to avoid a long and complex questionnaire, which would have compromised response rates, particularly among those with lower levels of education. However, the association between prescreening ratings of health behaviors and subsequent findings at sigmoidoscopy screening provide evidence of their validity (12, 13).

Overall, however, these results could be taken as reassuring in relation to hypothesized undesirable behavioral effects of screening and encouraging in relation to using the screening context to promote cancer-preventive health behaviors (14).

We thank David Boniface for his help with the statistical analysis in this paper.