Objective There is uncertainty whether cancer screening affects participant incentives for favourable lifestyle. The present study investigates long-term effects of colorectal cancer (CRC) screening on lifestyle changes.
Design In 1999–2001, men and women drawn from the population registry were randomised to screening for CRC by flexible sigmoidoscopy (‘invited-to-screening’ arm) or to no-screening (control arm) in the Norwegian Colorectal Cancer Prevention trial. A subgroup of 3043 individuals in the ‘invited-to-screening’ and 2819 in the control arm, aged 50–55 years, randomised during 2001 had their lifestyle assessed by a questionnaire at inclusion and after 11 years (42% of cohort). The outcome was 11-year changes in lifestyle factors (body weight, smoking status, physical exercise, selected dietary habits) and in total lifestyle score (0–4 points, translating to the number of lifestyle recommendations adhered to). We compared outcomes in the two randomisation arms and attendees with positive versus negative findings.
Results Total lifestyle scores improved in both arms. The improvement was smaller in the ‘invited-to-screening’ arm (score 1.43 at inclusion; 1.58 after 11 years) compared with the control arm (score 1.49 at inclusion; 1.67 after 11 years); adjusted difference −0.05 (95% CI −0.09 to −0.01; p=0.03). The change in the score was less favourable in screening attendees with a positive compared with negative screening result; adjusted difference −0.16 (95% CI −0.25 to −0.08; p<0.001).
Conclusions The present study suggests that possible unfavourable lifestyle changes after CRC screening are modest. Lifestyle counselling may be considered as part of cancer screening programmes.
Trial registration number NCT00119912.
- COLORECTAL CANCER SCREENING
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
The available evidence on possible influence of cancer screening on health behaviour is inconclusive.
A positive cancer screening result has been found to trigger improvement in health behaviour during the first years after screening.
Long-term effects of cancer screening on lifestyle factors are not known.
What are the new findings?
Invitation to colorectal cancer screening has a small long-term adverse effect on lifestyle expressed by a score based on body weight and smoking status, physical exercise and dietary habits.
Individuals with a positive screening result may be inclined to a less favourable lifestyle compared with those who tested negative at screening.
How might it impact on clinical practice in the foreseeable future?
The small lifestyle change after colorectal screening is unlikely to have impact for an individual, but may affect mortality in a population.
These results suggest lifestyle counselling to be included as a part of CRC screening.
Screening for colorectal cancer (CRC) by flexible sigmoidoscopy or faecal occult blood testing reduces CRC mortality.1 ,2 When planning cancer screening programmes, a possible effect of screening on future lifestyle should be investigated.3 If invitation and participation to screening or a negative screening test result would provide a health certificate effect and reduce incentives for a healthy lifestyle, then lifestyle counselling in cancer screening should be considered.4 ,5 However, the literature on lifestyle changes as a consequence of screening is limited and inconclusive, and further randomised-controlled studies are needed. Until now, no large, long-term follow-up trial has been performed to assess the effect of cancer screening on future health-related behaviour.
Any undesirable change in lifestyle in a population increases the number of individuals at high lifestyle-related mortality risk.6 In recent years, large population-based studies have shown that even small changes in lifestyle (although insignificant on an individual level) can affect mortality on a population level.7 ,8 Although the effect of screening on lifestyle changes is currently unknown, lifestyle counselling in conjunction with attendance to cancer screening programmes has been proposed.9 ,10 Thus, it is important to investigate long-term effects of cancer screening on lifestyle on the population level.
The Norwegian Colorectal Cancer Screening Prevention trial (NORCCAP) investigates the effect of flexible sigmoidoscopy screening on CRC incidence and mortality. All individuals aged 50–64 years in two counties in Norway were randomised to once-only flexible sigmoidoscopy screening (performed in 1999–2001) or no screening.11 A subset of the randomised individuals in each arm was invited to participate in a lifestyle survey. We have previously reported less favourable lifestyle changes 3 years after screening in the arm invited to screening.12 The screening arm improved less on their smoking habits, did not improve in physical exercise and increased their body weight more than the control arm. This finding supported the hypothesis that screening may reduce incentives to take responsibility for one's own health.12
The present paper reports on lifestyle changes 11 years after screening in the NORCCAP trial, comparing individuals randomised to screening versus controls, screening attendees versus non-attendees and screening attendees with negative versus positive screening results, respectively. The present study is the first large-scale population-based trial on the long-term effect of cancer screening on future lifestyle behaviour.
Study design and study population
The present lifestyle study is a substudy within the NORCCAP trial. The study design and recruitment of the NORCCAP trial have been previously published.11 In the NORCCAP trial, individuals were drawn from the Norwegian Population Register and randomised to a screening arm (hereafter called ‘invited-to-screening’) or a control arm (allocation ratio 1:1). The screening took place between January 1999 and December 2001. The study was approved by the Regional Research Ethics Committee and the Norwegian Data Protection Authority.
All individuals enrolled in the NORCCAP trial in 2001 were invited to participate in the present lifestyle study. Men and women born in 1946–1950 (age 50–55 years at inclusion), living in Telemark county (mixed rural and urban population) or Oslo city (urban population) were eligible. All participants provided written informed consent. Trial protocol is given in the online supplementary Appendix. The trial was registered at ClinicalTrials.gov (identifier NCT00119912).
At the time of inclusion, all individuals were asked to complete a questionnaire about their present lifestyle. Individuals in the ‘invited-to-screening’ arm who attended screening completed the questionnaire at the screening centre before bowel preparation for the screening examination. Individuals in the ‘invited-to-screening’ arm who did not attend screening and individuals in the control arm received the questionnaire by mail. The invitation letter to the latter two groups briefly gave information on the main NORCCAP trial. In 2004 and 2012, all individuals who completed the questionnaire in 2001 received an identical questionnaire as in 2001 by mail (figure 1). The same one-page, self-report questionnaire was used at the three rounds of lifestyle assessment in 2001, 2004 and 2012 (see online supplementary figure S1).13 The questions asked were based on questionnaires used in previous national surveys.14 ,15
Individuals ‘invited-to-screening’ were offered a once-only flexible sigmoidoscopy screening examination. Individuals randomised to the control arm received no screening. Details of the screening examination and criteria for diagnosis were described previously.11 ,16 Individuals with a positive screening result (any polyp ≥10 mm or any histologically verified neoplasia, irrespective of size) at flexible sigmoidoscopy screening were offered a colonoscopy. Findings at the flexible sigmoidoscopy screening examination (and subsequent colonoscopy for those with positive screening) were classified according to the most severe pathological finding.
In the questionnaire, individuals were asked to report their body height and weight and the number of weekly hours in occupational activity. Present smoking status, frequency of physical exercise and consumption of selected food items were assessed, giving predefined answering alternatives. We calculated body mass index (BMI, kg/m2) by the reported body weight (kg) divided by the square of reported body height (m). For analyses, we transformed the replies of the physical exercise questions (‘How often do you exercise lightly or moderately/vigorously with a duration of at least 20 min?’) into exercise sessions per day. Further, we collapsed the two physical exercise variables into one variable; total physical exercise (range 0–2.0 times per day). We recoded food items consumption into frequency of daily intakes. The three variables daily servings of fruit and berries, daily servings of raw vegetables and daily servings of boiled vegetables were collapsed into one variable; total consumption of fruit, berries and vegetables (range 0–9.0 servings per day).
We created a total lifestyle score based on adherence to international and national public health recommendations from the WHO,17 ,18 World Cancer Research Fund,19 the Nordic Nutrition Recommendations20 and the Norwegian dietary recommendations.21 In order to classify the study participants according to these recommendations within the limits of the questionnaire, one point was given for each of the following criteria: normal body weight (BMI<25.0 kg/m2),17 ,19 ,20 non-smoking,18 physical activity at least once per day17 ,19 and a healthy diet. For the healthy diet, participants achieved one point if reporting all the three dietary habits: (1) consumption of fruit, berries and vegetables with at least five servings per day;17 ,19 ,20 (2) consumption of fatty fish at least once per week20 and (3) consumption of meat other than poultry for dinner maximum three times per week (according to recommendation of consumption of red and processed meat <500 g per week19 ,21). Hence, the total lifestyle score ranged from zero to four points, indicating the number of lifestyle recommendations adhered to by the individual.
We analysed lifestyle changes as the difference between 2001 and 2012. The single lifestyle variables and the total lifestyle scores in 2001 and in 2012, and the change in these (Δ2001–2012) were compared between the ‘invited-to-screening’ and the control arm. Similarly, lifestyle changes were compared between those who attended and those who did not attend the screening examination (in the ‘invited-to-screening’ arm), and between individuals with positive and negative screening results (among those who attended screening). We used paired samples t test to compare the 2001 and 2012 lifestyle factors measured on a continuous scale and the McNemar's test to compare the change in smoking status (yes/no). To evaluate the group-wise change over time, we used the independent samples t test.
We tested group-wise differences in lifestyle changes measured by continuous variables by univariate and multivariable regression analyses (analysis of covariance (ANCOVA)). The multivariable ANCOVA analyses of single lifestyle variables were adjusted for sex, area of residency (Telemark/Oslo), recent change in chronic disease status reported in 2012 (yes/no), 2001 value of the dependent variable and 2001 values of other single lifestyle variables. In the multivariable ANCOVA analyses of the total lifestyle score, we adjusted for the same variables except the single lifestyle variables included in the score variable. We fit a multivariable logistic regression model to analyse the proportion of individuals reporting smoke cessation from year 2001 to 2012, adjusting for baseline smoking status, and the same covariates as used in the ANCOVA analysis of single lifestyle factors. We compared changes in total lifestyle scores from 2001 to 2012 in the two lowest scores (zero and one) since the possible impact on population mortality is greatest among groups with low adherence to lifestyle recommendations.7 We tested the statistical significance of changes of dichotomous lifestyle score categories within groups with McNemar's test and difference in changes between groups with logistic regression analysis, adjusting for the dichotomous baseline score category, sex, area of residency and recent change in chronic disease status reported in 2012.
We included individuals with partly missing data. In the adjusted analyses, the proportion of individuals with missing information on one or several questions varied between 13% and 20%. For sensitivity testing, we analysed the data using partial imputation of missing data, but this did not alter the results (data not shown). We also conducted the abovementioned analyses including only individuals who completed all questionnaires (in 2001, 2004 and 2012), but the results remained the same.
For data quality control, we compared the reported smoking status in the lifestyle questionnaire in 2001 with the smoking status registered by the endoscopist at the screening centre (screening attendees only). A slightly higher number of individuals reported daily smoking in the medical history interview (n=873) than in the questionnaire (n=859) (Pearson's correlation coefficient in daily smoking status between the two registering methods; r=0.952).
All p values are two-tailed. p Value <0.05 is considered as statistically significant. The data were analysed by IBM SPSS Statistics software, V.184.108.40.206 (IBM Corporation, Armonk, New York, USA).
Altogether, 13 961 individuals were included; 6961 in the ‘invited-to-screening’ arm and 7000 in the control arm. Of these, 8484 (61%) responded to the questionnaire in 2001. In 2012, the responders from 2001 who were still alive and attainable (7985 individuals) were approached, and 5915 (74%) individuals responded. A total of 53 individuals were excluded due to illogical answers in 2012. This left a total of 5862 individuals eligible for analyses, 3043 in the ‘invited-to-screening’ arm and 2819 in the control arm (figure 1). The two arms (responding to the questionnaire) differed with regards to baseline (2001) age, area of residency, weekly employment, and smoking, and response to 2004 lifestyle questionnaire (table 1). We have no socioeconomic or lifestyle data at baseline in the non-repliers group. Age, sex and baseline area of residency were similar between year 2012 questionnaire repliers and non-repliers. There was a statistically higher proportion of baseline current smokers and subjects with a low number of baseline weekly employment hours among non-repliers (n=2012) than repliers (data not shown). The non-repliers had a statistically higher reported baseline mean BMI and lower levels of mean total physical exercise, diet score and total lifestyle score than repliers at baseline. These comparisons between year 2012 repliers and non-repliers did not differ between the ‘invited-to-screening’ and control arms, with the exception that in the control arm, a higher proportion of the non-repliers were from Telemark than Oslo (not shown).
Single lifestyle variables and the total lifestyle score in 2001 and in 2012 are shown in table 2. Change in the lifestyle variables from 2001 to 2012 and difference in changes between the trial arms are shown in table 3. Smoking became less prevalent from 2001 to 2012 (table 2), but between the trial arms there was no difference in the proportion of baseline smokers reporting non-smoking in 2012 (table 3). The total lifestyle score increased in both arms (table 2 and figure 2A), but the adjusted improvement was smaller (difference −0.05 units; 95% CI −0.09 to −0.01, p=0.03) in the ‘invited-to-screening’ compared with the control arm (table 3 and see online supplementary figure S2A). Except for physical exercise, which increased less in the ‘invited-to-screening’ than the control arm, there was no difference in lifestyle factors between the two arms (table 3).
We found no difference in lifestyle factors between those who attended and those who did not attend the screening examination (in the ‘invited-to-screening’ arm, data not shown).
Change in the lifestyle variables from 2001 to 2012 and difference in changes between the positive and negative screening result groups are shown in table 4. In the group with positive screening result, the proportion of daily and occasional smokers was reduced from 43.0% in 2001 to 30.3% in 2012. In the group with negative screening result, this reduction was from 33.4% to 17.9%. The baseline smokers’ adjusted odds of reporting non-smoking in 2012 was 48% lower for those with positive versus negative screening result (OR 0.52, 95% CI 0.36 to 0.75) (table 4). The adjusted improvement in the total lifestyle score (figure 2B) was significantly less in the group with positive compared with a negative screening result (−0.16 score units; 95% CI −0.25 to −0.08, p<0.001) (table 4 and see online supplementary figure S2B).
Table 5 shows the number of individuals in the dichotomous total lifestyle score categories ‘zero to one’ (low) and ‘two to four’ (high) by screening arm and screening result allocation in 2001–2012. After adjustment for baseline lifestyle score category, sex, area of residency and recent change in chronic disease status reported in 2012, the odds of being in the high lifestyle score category was 19% lower in the ‘invited-to-screening’ versus control arm (OR 0.81, 95% CI 0.71 to 0.92, p=0.002), and 35% lower in the positive versus negative screening result group (OR 0.65, 95% CI 0.50 to 0.85, p=0.002).
This study shows that overall adherence to lifestyle recommendations increased in our large sample of 50-year-old to 55-year-old Norwegian men and women from year 2001 until 2012. Invitation to screening had a small negative long-term effect on total lifestyle score, supporting the health certificate effect of cancer screening.5 Furthermore, the apparently less favourable total lifestyle change after CRC screening was more pronounced in screening attendees with a positive than negative screening result. Although the difference in the long-term lifestyle change between the randomised arms was very much smaller than what is considered as clinically relevant for an individual, it may play a role on a population level when exposed to screening.
The strength of our study is its population-based, randomised design, its large size, long follow-up time and the high response rate (61% at baseline, and 74% of these also responded to the 11-year questionnaire). The short questionnaire might not be adequate for assessment of absolute levels in lifestyle factors in individuals, but appears suitable for assessing changes at a group level. Alcohol intake registration was sacrificed from the questionnaire due to limited quality of self-reported alcohol intake and space restrictions. It would, however, be expected that adherence to the recommendation of moderate alcohol consumption is related to the total lifestyle score.22 The setting for filling in the questionnaire (while waiting for the screening examination at the screening centre for screening attendees versus mailed questionnaire filled in at home for the control arm) may have affected the response rate that was slightly higher for the ‘invited-to-screening’ than the control arm. We have no data on general screening behaviour in terms of participation in other health screening programmes (applicable only to women since screening programmes for breast and cervical cancer are the only ones available in Norway).
Mean self-reported 11-year weight gain of 1.5 kg (0.14 kg annually) in the present Norwegian population is half or less than that of the annual weight gain in large recent European and US cohort studies.23 ,24 In our study, the weight gain was the only observed unfavourable change in lifestyle variables from 2001 to 2012, largely to be considered a physiological effect of ageing. An overall reduction in the proportion of daily smokers, the overall increase in physical exercise and consumption of fruit, berries, vegetables and fatty fish suggest secular trends of lifestyle in the community. Long-term improvement in adherence to dietary guidelines has also been observed in Australian adults between 1992 and 2007.25 Although the level of adherence to all four recommendations in our study was generally low, an absolute reduction of 5.7% and 8.0% was seen in the poorest lifestyle categories in ‘invited-to-screening’ and control arms, respectively (table 5). Reducing the number of individuals with poor adherence to lifestyle recommendations might have the greatest overall impact on population mortality.7
Although overall lifestyle score improved in both arms, we observed a relative and undesirable association between CRC screening and several single lifestyle factors 3 years after the screening examination.12 Eight years later, 11 years after screening, we still observe a significant difference between those ‘invited-to-screening’ and those not invited, although the observed difference today is smaller than after 3 years of follow-up. This supports our hypothesis of a health certificate effect of CRC screening, reducing long-term incentives for a healthy lifestyle. However, the difference is modest and mainly derived by the difference in physical exercise change between the arms. There are no previous long-term follow-up studies on multiple lifestyle changes after randomisation to cancer screening or control arm. The Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO) investigated long-term effects of lung cancer screening on smoking quitting rates among daily smokers and relapse rates among former smokers, and found that cancer screening was not associated with any of these.26 Consistently, the results on smoking in the present study show no difference in smoking prevalence between those ‘invited-to-screening’ and those not invited after a long-term follow-up. Randomised trials with 1 or 2 years’ follow-up after lung cancer screening have shown either no difference between screening and control arms27 or a lower smoking abstinence rate in the screening arm.28 It should be noted that there was a higher proportion of baseline current smokers in the ‘invited-to-screening’ than control arm (35% vs 30%, respectively) in the present study, and that this difference remained during the 11-year follow-up. As smoking is associated with other unhealthy lifestyle habits, this might explain the difference in improvement in physical activity as well as total lifestyle improvement between the study arms.
We found no difference in lifestyle changes between screening attendees and non-attendees in the ‘invited-to-screening’ arm. However, considering the low number of screening non-attendees with 11-year lifestyle change data (n=147), the present study lacks power to conclude on differences between attendees and non-attendees.
Two important questions arise from the slightly less favourable change in the total lifestyle score in the ‘invited-to-screening’ arm. First, is the smaller lifestyle improvement due to the screening examination, or does it reflect selection bias? Second, does the smaller lifestyle improvement have impact on health outcomes such as cardiovascular mortality? Most of the lifestyle questionnaire responders in the ‘invited-to-screening’ arm attended the screening examination (95%), whereas in total, attendance rate in the NORCCAP trial was 65%.29 The literature suggests that screening attendees report a more favourable lifestyle than non-attendees.30 Similarly, lifestyle questionnaire responders randomised to the control arm may consist of more health-conscious individuals than the background population.31 We, therefore, expect that the respondents in the ‘invited-to-screening’ and control arms equally consist of more health-conscious individuals than the background population.
The smallest clinically important change in lifestyle to affect mortality and morbidity is unknown. However, the association between adherence to lifestyle recommendations and endpoints such as all-cause and cardiovascular mortality has been studied thoroughly. A meta-analysis from 2012,8 including 531 804 individuals from 15 studies, found a 26% reduction in all-cause mortality for individuals adhering to one lifestyle recommendation compared with none (risk ratio 0.74, 95% CI 0.67 to 0.81) and 42% reduction for individuals adhering to two recommendations compared with none (risk ratio 0.58, 95% CI 0.50 to 0.67). This indicates that even small changes in lifestyle may significantly affect mortality, suggesting that the small difference in adherence to lifestyle recommendations found in the present study can affect all-cause mortality on a population level. Consequences of lifestyle change on mortality are expected to be largest in the lowest end of the overall lifestyle scale. Of the single lifestyle factors, we found significant difference in long-term change between the randomisation arms only for physical exercise. This difference translates into as little as an increase of approximately 1 min daily exercise more in the control arm. However, all lifestyle factors contributed to the difference in the total lifestyle score change between the arms. It is difficult to quantify single preventive efforts. We cannot predict which single lifestyle change will have the largest impact on mortality. The literature emphasises the importance of the number of lifestyle recommendations adhered to without quantifying the impact of single recommendations.
Attendees in cancer screening receiving a positive test result may be more compliant for lifestyle improvement than those receiving a negative result.32 However, at 11-year follow-up, we found a less reduction in the proportion of daily smokers in the group with positive than negative screening result. Improvement in overall adherence to lifestyle recommendations was smaller in the group with positive screening result during the 11 years of follow-up, mainly derived from the lower smoking cessation rate in the screening positive result group. A higher proportion of baseline smokers in the positive screening result group should be noticed. Less long-term lifestyle improvement might be expected in the group with higher baseline smoking rate. Contrasting our results, the PLCO trial did not find any long-term effect of lung cancer screening result on smoking cessation.26 The 35% lower odds of having a high lifestyle score 11 years after screening for the positive screening result group compared with the negative screening result group implies a higher future mortality risk for those testing positive in screening. When screen-positives additionally had less favourable lifestyle characteristics at baseline, this emphasises the need for lifestyle counselling in screening programmes. There was, however, a temporary relative improvement in lifestyle score for screen-positives 3 years after screening compared with those having no findings at screening (figure 2B). This suggests that the hypothesised health certificate effect of a negative screening result may be limited to the first few years after screening, but not seen at 11-year follow-up. The possible effects on lifestyle when exposed to repetitive screening rounds are unknown.
In conclusion, this study on lifestyle characteristics suggests a modest long-term health certificate effect of invitation to CRC screening. We found a significant, but small, undesirable association with long-term lifestyle change expressed by body weight, smoking status, physical exercise and dietary habits after invitation to CRC screening. The present study suggests that individuals with a positive screening result may be inclined to a less favourable lifestyle in a long term compared with those testing negative at screening. A health certificate effect of a negative screening result could not be confirmed at 11-year follow-up. These results suggest that lifestyle counselling should be included as a part of CRC screening.
We thank the members of the NORCCAP steering committee, Eva Skovlund, Jörn Schneede, Tor Iversen, Morten H Vatn, Kjell Magne Tveit, Tor Jac Eide and Jon Lekven. We also thank Odd A Aalen for help with the statistical analysis.
Correction notice Figure 2 has been corrected since published Online First.
Contributors GH is the principal investigator of the main NORCCAP trial. PB and GH wrote the grant application. IKL conducted data collection in 2001 and 2004. PB conducted the 2012 data collection and the data analysis, and wrote the first draft of the manuscript. ML and EB participated in the data analysis. All authors contributed to data interpretation and the writing and editing of the manuscript. All authors approved the final version of the manuscript.
Funding This study was funded by South-Eastern Norway Regional Health Authority, Grant 2012094 for Paula Berstad; and the Astri and Birger Torsted's funds for cancer research.
Competing interests None.
Patient consent Obtained.
Ethics approval The study was approved by the Regional Research Ethics Committee and the Norwegian Data Protection Authority.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Study protocol is available from the corresponding author. Please contact the corresponding author to discuss deidentified data requests.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.