Effect of a pre-screening survey on attendance in colorectal cancer screening: A double-randomized study in Finland

Introduction

Colon and rectum cancers together are the third most common cancer site by incidence among Finnish men and women combined, with over 2500 new cases annually. ¹ Colorectal cancer (CRC) incidence and mortality can be reduced by repeated faecal occult blood testing, followed by colonoscopy for test positives.^2,3 A population-based programme for CRC screening by biennial faecal occult blood testing for 60–69 year olds has been running in Finland since 2004. The screening is voluntary for the municipalities to organize. An experimental design has been incorporated for public health evaluation, using individual level randomization of the target population into screening and control groups. ⁴

For a cancer screening programme to be effective from the mortality-lowering aspect, a high participation rate is crucial. The Finnish CRC screening has achieved high attendance: from 2004 to 2007, overall attendance was 71%. Participation was better in women (78%) than in men (62%)^5,6 and in older age groups than in younger people.

Screening programmes provide a useful setting in which to conduct epidemiological surveys, but there is a risk of lowering attendance if participants are asked to submit to extra investigations. In evaluating the effects of CRC screening, information on life style, quality of life, and costs is also needed. In Finland, the “Psychosocial effects of colorectal cancer screening study” was incorporated into the CRC screening programme in 2010. The study utilizes a prospective, randomized and controlled study design to ensure an unbiased setting. Life style and quality of life are assessed through a self-administered postal questionnaire at two stages, at baseline one year before the first possible screening invitation, and a similar questionnaire one year after the first invitation (Figure 1). A comparison can then be made on how different life style related factors change among the invited and the control population, and the possible effect of screening on health habits can be evaluated. OPEN IN VIEWER

We here attempt to determine whether sending a short (7-paged) or long (10-paged) questionnaire on life style and quality of life one year before the first ever screening invitation has an effect on uptake of screening. We also investigate whether this effect varies between different groups according to background demographic factors.

The study design was approved by the Helsinki and Uusimaa hospital district Ethics Committee (15.12.2009, 420/13/03/00/2009).

Methods

Results

In May 2010, 10,648 randomly selected individuals from the study population were mailed the study questionnaires. The rest (n = 21,303) were not contacted (Figure 2). A reminder letter was sent to non-responders after four weeks in June 2010. In December 2010, the questionnaires were resent to those who still had not responded. After exclusions (empty return forms, wrong person responded, returned refusals, same person responded twice) the final survey response rate was 55.1% (n = 5,870) (Figure 2). The response rate in women was higher than that in men (60.5% and 49.8%, respectively). OPEN IN VIEWER

In January 2011, the target population of CRC screening born in 1951 (n = 32,072) was randomized 1:1 into screening arm (n = 16,037) or control arm (n = 16,035). Of the 2010 survey study population 467 individuals had either died or moved to a municipality not involved in colorectal cancer screening programme or abroad, and were no longer in the target population. There were 31,484 individuals randomized twice, first for the survey and later for screening, of whom 15,748 were invited for screening (Figure 2). The overall attendance rate for screening in the 60 year-olds in 2011 was 59.0% (n = 9,292). The attendance rate of 60-year-olds had shown a downward trend in the course of CRC screening in Finland (Figure 3). OPEN IN VIEWER

Of the 5185 individuals who had received the survey in 2010 and were invited for CRC screening in 2011, 2,935 (56.6%) attended screening. Of the 10,563 individuals who had not received the questionnaire in 2010, 6,357 (60.2%) attended screening (Figure 2). Screening attendance for those who had received the shorter questionnaire was 57.3%, and for the longer one 56.0% (Table 1).

OPEN IN VIEWER

Table 1.

Number of individuals invited, attending and attendance proportions for colorectal cancer screening 2011 in Finland among those who had been sent a survey study questionnaire in 2010 and those with no preceeding questionnaire.

	short questionnaire			long questionnaire			no questionnaire
	invited	attenders	Attendance rate (%)	invited	attenders	Attendance rate (%)	invited	attenders	Attendance rate (%)	P-value
sex										<0.001
man	1288	618	48.0	1322	613	46.4	5272	2700	51.2
woman	1285	855	66.5	1290	849	65.8	5291	3657	69.1
marital status1										<0.001
partner	1596	993	62.2	1605	977	60.9	6630	4308	65.0
no partner	968	477	49.3	1001	484	48.4	3922	2047	52.2
mother tongue2										0.211
Finnish	2429	1399	57.3	2438	1373	56.2	9856	5951	60.4
Swedish	109	61	56.0	132	68	51.1	563	336	59.7
residence										0.008
urban	1569	866	55.2	1540	866	56.2	6355	3763	59.2
rural	1004	607	60.5	1072	596	55.6	4208	2594	61.6
area										0.017
South	480	259	54.0	455	252	55.4	1939	1163	60.0
Middle	889	516	58.0	878	522	59.5	3639	2211	60.8
West	522	297	56.9	491	268	54.6	2005	1190	59.4
East	314	185	58.9	373	223	59.8	1388	851	61.3
North	368	216	58.7	415	197	47.5	1592	942	59.2
total	2573	1473	57.3	2612	1462	56.0	10563	6357	60.2	<0.001

Table 2 shows the absolute (difference in percent of attendance between the groups) and relative decrease (the absolute difference divided by attendance in those with no previous questionnaire) in screening attendance, by length of the questionnaire and demographic factors. Willingness to attend screening, especially among men, was reduced among those who had received a survey (the relative decrease in attendance was 7.8% in men and 4.2% in women). Marital status or habitation appeared not to have any major effect on screening adherence. In the Swedish speaking minority, the pre-screen questionnaire decreased consecutive screening attendance more than in the Finnish speaking group (relative decrease of 10.7% vs. 6.1%). There were some differences in the decrease of screening attendance by geographical area: in Eastern Finland, having received the survey questionnaire reduced future screening attendance only slightly (2.6%), but in Northern Finland the decrease was considerable (10.5%).

OPEN IN VIEWER

Table 2.

Attendance to colorectal cancer screening in those with no previous questionnaire (No Q-group) and absolute and relative difference between those who had been sent a short (short Q), long questionnaire (long Q), or any questionnaire (any Q) before.

	Attendance rate (%)	Absolute attendance difference (%)			Relative attendance difference (%)
	No Q-group	short Q	long Q	any Q	short Q	long Q	any Q
total	60.2	−2.9	−4.2	−3.6	−4.8	−7.0	−6.0
sex
man	51.2	−3.2	−4.8	−4	−6.3	−9.4	−7.8
woman	69.1	−2.6	−3.3	−2.9	−3.8	−4.8	−4.2
marital status
partner	65	−2.8	−4.1	−3.5	−4.3	−6.3	−5.4
no partner	52.2	−2.9	−3.8	−3.4	−5.6	−7.3	−6.5
mother tongue
Finnish	60.4	−3.1	−4.2	−3.7	−5.1	−7.0	−6.1
Swedish	59.7	−3.7	−8.6	−6.4	−6.2	−14.4	−10.7
residence
urban	59.2	−4.0	−3.0	−3.5	−6.8	−5.1	−5.9
rural	61.7	−1.2	−6.1	−3.7	−1.9	−9.9	−6.0
area
South	57.6	−3.6	−2.2	−2.9	−6.3	−3.8	−5.0
Middle	61.7	−3.7	−2.2	−3.0	−6.0	−3.6	−4.9
West	59.4	−2.5	−4.8	−3.6	−4.2	−8.1	−6.1
East	61.0	−2.1	−1.2	−1.6	−3.4	−2.0	−2.6
North	59.0	−0.3	−11.5	−6.2	−0.5	−19.5	−10.5

The longer questionnaire was more detrimental for the future screening attendance than the shorter one. This was most obvious in Northern Finland, where the short questionnaire did not affect screening uptake much, but the longer one lowered uptake considerably (19.5%) (Table 2).

Discussion

There were two important considerations in our study design: firstly, the value of a valid control group, and secondly, the necessity for all conclusions to be made based on a randomized design.

For several reasons, organized screening with personal invitations may be an ideal setting for incorporating different kinds of survey initiatives. Population-based screening has the advantage of reaching the total population of a pre-defined target group. Well-established screening may already have gained support within the population, which could also help to diminish the barriers to participation in studies embedded within the screening setting. It is important, however, that combining any extra research into routine medical practice should not have a detrimental effect on screening participation, as this is vital in achieving the primary aim of screening - in this instance, cancer-specific mortality reduction. Two reviews on the determinants of participation in CRC screening^9,10 have suggested that participation patterns depend on local circumstances, highlighting the need to acquire information on programme participation.

Our results suggest that sending a 7- or 10-paged questionnaire to potential participants one year prior to the first screening invitation had an effect on screening uptake. Screening attendance rate was 60.2 % among those who were not sent the questionnaire, and 56.6% among questionnaire-receivers. In addition, the longer questionnaire had a more detrimental effect on screening attendance than the shorter one. Sending a questionnaire may adversely affect future screening attendance, and the longer the questionnaire, the greater this effect.

The strengths of our study include the randomized and controlled design, with no exclusions in the invited population-based sample.

Inclusion of a self-administered questionnaire accompanying the invitation to routine breast cancer screening did not affect the screening uptake in a randomized setting in the Netherlands in 1992, nor in southern England in 1994–1996.^11,12 It is noteworthy that these studies only contacted target individuals once, whereas in our study target individuals were contacted 1–3 times (questionnaire, reminder, resending of questionnaire). In the Dutch and UK studies, conducted over a decade ago, target groups consisted only of women aged 50–70 ¹¹ or 50–64. ¹² Our study included only one age cohort of 60-year-olds.

In general, women and married individuals tend to participate in surveys more actively than men and single or divorced people. The same pattern can be observed in screening attendance, and has been evident since the launch of the Finnish CRC screening programme in 2004. ¹³ The youngest age group has the lowest attendance in Finland, and there has been a downward trend for many years. Compared with 2010, attendance rates in 2011 were worse, even without the extra lowering effect of survey (60.2% vs. 62.1% in 2010; men 51.2% vs. 53.4%, women 69.1% vs. 70.8%). It could be argued that the survey manifested a strengthening of an existing trend.

The absolute differences in attendance rates by various background factors were mostly modest. For example, the absolute difference in attendance rates by marital status was 2.9% in women and 4% in men. After correcting for attendance relative to those who had not received a survey, these rates changed to 4.2% and 7.8%. Overall, the absolute differences in screening attendance rates were not large, but after correcting for attendance relative to those with no survey, the relative differences were larger and had more variability. Using the same method to assess the effect of marital status showed that sending a questionnaire decreased attendance at screening more in those without a partner than in those with a partner (6.5% vs. 5.4%).

In 2011 the Finnish CRC screening programme carried out 9,462 primary tests in 60-year-olds, resulting in 343 positive results, equating to 3.6% of participants. We estimate that, if there had been no survey in 2010, the overall screening attendance rate would have been similar to that among those who did not receive the pre-screening questionnaire, ie. 60.2%. This would mean that 192 more individuals would have attended screening (60.2% out of 16,037 invited, instead of 59.0% now reached in the 60-year old population). We therefore calculated that, because of the survey, an estimated seven positive test results were not produced (3.6% x 192). Out of 343 test positives in 2011, 86 (25%) were diagnosed as having an adenoma (n = 76) or CRC (n = 10) in the follow-up colonoscopy. With seven test positives, there would have been two more expected adenomas. However, if we had sent a questionnaire to all of the cohort and gained an overall attendance rate of 56.6% (attendance in those who were actually sent the questionnaire in 2010), the loss of people attending would have been close to 577 individuals (56.6% out of 16,037 invited instead of 60.2%). Especially in the subgroups where the relative decrease in attendance was most marked (eg. in Northern Finland), this could also affect screening results, as non-attendance may be selective. Even if no substantial harm was caused, on this occasion, to our target population, we do not generally recommend a policy of combining surveys with screening programmes.

We will monitor the future CRC screening behaviour of this study group. The same invited individuals were re-invited to CRC screening in 2013, after a two year interval. In the previous year (2012) the study group had been subjected to the follow-up questionnaire of “Psychosocial effects of colorectal cancer screening”. Extending monitoring until 2015 (when there will no longer be any further pre-screening questionnaire) will ultimately disclose whether or not the effect of questionnaires on screening attendance is long-term.

Why does sending and receiving a questionnaire lower future screening uptake? Though the questionnaires were mostly about general life style in terms of diet, physical activity, smoking and alcohol use, the accompanying covering letter referred to the study ”Psychosocial effects of colorectal cancer screening” as a whole. In this way, the questionnaire was linked to screening. The screening invitation may then have been perceived by the recipients as burdensome when dealing with the same matter again, even though it was clearly stated that participation in the survey was completely voluntary and it would not affect screening randomization. In previous studies on colorectal cancer screening participation, sending a pre-screening notification two weeks prior the screening invitation has had a positive effect on screening participation.^14,15 We therefore suggest that the nature of the pre-screening contact, not the contact per se, is pivotal. Requesting effort on the part of the recipient may be harmful for future screening willingness.

Simultaneous mailing in the same envelope is the standard method to incorporate an extra survey into routine health activity. It is reasonable to assume that such extra effort reduces the willingness to participate even more than was observed in our study with a one year lag between these two activities (survey and screening). Therefore, in a real life situation, the observed effect may be sufficient to preclude incorporating any extra material in an activity like a screening invitation.

In conclusion, sending a pre-screening survey reduced attendance in colorectal cancer screening. This risk of lowering participation should be taken into account when planning to exploit the screening context in conducting epidemiological surveys or other intervention studies.