The impact of hypothetical PErsonalised Risk Information on informed choice and intention to undergo Colorectal Cancer screening colonoscopy in Scotland (PERICCS)—a randomised controlled trial

Following a successful demonstration pilot [1] a quantitative faecal immunochemical test for haemoglobin (FIT) was introduced as the primary test in the Scottish Bowel Screening Programme (SBoSP) in late 2017, replacing the guaiac faecal occult blood test (gFOBT). The advantages of FIT over gFOBT are well accepted, including being more user-friendly, specific for human haemoglobin, and importantly, providing an estimation of faecal haemoglobin concentration (f-Hb) rather than a binary positive/negative result. Since f-Hb is related to severity of disease [2], the numerical FIT result is an important risk factor for colorectal cancer (CRC). Indeed, a recent study in a symptomatic population referred for colonoscopy identified f-Hb as by far the most powerful predictor of all risk factors studied including age, gender, symptoms, family history and lifestyle factors [3]. However, in the SBoSP, FIT is currently used as dichotomous test, with the threshold for a positive result set at f-Hb of ≥ 80 μg Hb/g faeces to mimic the 2% positivity rate observed with the previous gFOBT-based screening algorithm. Although not as strongly predictive as f-Hb, increasing age and male sex have also been documented to show association with CRC prevalence [4]. Therefore, there is potential for f-Hb, along with age, sex and other risk factors to be used to provide a personalised risk of harbouring CRC, and thus empower people to make a truly informed decision about having a colonoscopy, by weighing the risks and disadvantages of the procedure (e.g. bleeding, bowel wall perforation, emotional distress) against their personalised risk of missing a cancer. This is particularly important in CRC screening, as currently about half of all cancers in the screened population in Scotland are diagnosed in the interval between screens, indicating that, at the current threshold, the test is only about 50% sensitive for CRC. Reducing the threshold would result in fewer cancers being missed but would increase the chance of a negative colonoscopy and the impact on the colonoscopy service may be unsustainable.

There is currently no existing evidence on the effects of personalised risk information on uptake of colonoscopy following first line screening for CRC. A Cochrane systematic review examined the effect of personalised risk communication for informed decision-making in uptake of medical screening compared to general information [5]. Overall, providing a numerical risk score or a categorised risk (e.g. ‘low’, ‘medium’, ‘high’) increased informed choice (odds ratio 3.65, 95% CI 2.13 to 6.23, for random effects) and screening uptake (odds ratio 1.15 95% CI 1.02 to 1.29). However, the included studies covered a wide range of screening tests and thus the results were heterogeneous. Of the studies involving CRC, only one [6] used a calculated numerical risk score, leading to greater knowledge but non-significant lower intention and uptake; it did not report changes in informed choice. Three studies used a categorised risk score; these did not assess knowledge or informed choice, but indicated a small, significant increase in uptake of screening [7‐9]. The authors concluded that the evidence that personalised risk communication increases screening uptake is weak. Further, although some included colonoscopy, all of the reviewed studies involving CRC screening related to first-line screening only. Subsequent to the commencement of our study, results have been published from the CRISP-Q [10] study where patients in GP waiting rooms in Australia were given five different hypothetical average and increased risk presentations and asked to decide for each, based on the risk presented, whether they would choose either no screening, a faecal occult blood test, or a first-line screening colonoscopy. Trends existed for selection of more appropriate screening options both for the risk presentation showing an absolute risk of CRC along with a government recommendation and the one showing an ‘expected frequency tree’. Very recently published results from a Danish RCT found that a web-based decision aid aimed at screening-naïve individuals with lower educational attainment did not affect informed choice or knowledge of CRC screening but there were trends towards improved uptake and positive attitudes towards CRC screening compared with the control group [11].

Currently, participants in the SBoSP are offered colonoscopy when they have f-Hb ≥ 80 μg Hb/g faeces. This threshold for positivity is much higher than that used in other countries. Any measurable f-Hb confers some risk of neoplasia, and sensitivity for cancer detection increases as the f-Hb threshold is lowered [12]. It is likely that those screening participants receiving a letter informing them that they have had a negative screening test result are largely unaware that around half of all CRC in the screened population arise in those with f-Hb below the threshold of 80 μg Hb/g faeces. Thus, we aimed to investigate the impact of providing the screened population with novel, hypothetical personalised risk information designed to enable a completely transparent informed choice as to whether or not to undergo colonoscopy following screening with FIT. In addition to the effect on level of informed choice, we also investigated the potential impact that providing screening participants with such information would have on colonoscopy services as well as emotional responses such as increased anxiety about the screening process.

Two thousand seven hundred sixty-seven people were invited to participate in the main study phase, with 434 (15.7%) people indicting they would be willing to participate. Of the 434 people agreeing to participate, 145, 144 and 145 were randomised into the numerical risk arm, categorical risk arm and the positive result letter arm, respectively. Three hundred nine (71.0%) participants completed and returned the materials, 100 from the numerical risk group (69.0%), 104 from the categorical risk group (72.2%) and 104 from the control group (71.7%). Of those who completed and returned the study materials, 91.0%, 93.3% and 90.4% in the respective study arms had previously participated in the SBoSP. Demographic details of those invited and responding to the study invite are shown in Table 2.

Participants’ responses to whether or not they would choose to have a colonoscopy in each of the risk scenarios are shown in Table 3. In the two study arms receiving risk information, the proportion of participants who said they would intend to undergo colonoscopy decreased as risk level was lowered. This fall in intention was statistically significant in the numerical risk study arm between both of the reductions in risk (1 in 40 cf. 1 in 1600: p = 0.0002 and 1 in 1600 cf. 1 in 3500 p = 0.003) and in the categorical risk study arm between the moderate and lowest risk scenarios (p < 0.0001). Although the number of participants in the categorical study arm who said they would undergo colonoscopy if they were told they were in the moderate risk group was 10% greater than in the equivalent group in the numerical arm (88.8% vs. 78.7%), this difference was not statistically significant (p = 0.067). Mean informed subscore of the Decisional Conflict Scale and numbers of participants with scores associated with implementing decisions (scores < 25) for each study arm are also shown in Table 3. No significant differences in mean informed subscore were observed between the study arms.

For the two additional 7-point Likert scale statements regarding intention to undergo colonoscopy (e.g. “If I received information that my risk of bowel cancer was 1 in 40 then I would intend to have a colonoscopy” and “If I was told that I had a 1 in 40 chance of having bowel cancer then I would definitely choose to have a colonoscopy”), all three groups showed a high level of agreement with the statements for the highest risk scenarios (all mean scores ≥ 6.5 out of 7) and even in the lowest risk groups in the numerical and categorical arms, participants still responded positively. Mean scores in each study arm for these items are shown in Table 4.

Mean knowledge and attitude scores for each study arm are shown in Table 5. All participants who completed the knowledge questions (97, 102 and 104 in the numerical risk, categorical risk and positive result letter arms, respectively) recorded a score of ≥ 5, indicating adequate knowledge, with the exception of one in the categorical risk group and two in the positive result letter group. No significant differences in total knowledge or attitude scores were observed between the three study arms (p > 0.05). This would be expected as all groups received the same information materials, with only presentation of risk differing. Table 6 shows the number of participants in each study arm who correctly answered each item in the knowledge section of the questionnaire.

For the highest risk groups only (risk of 1 in 40 in the numerical group, highest risk in the categorical group and the positive result letter group), Table 7 shows the number of participants in each study arm who were deemed to have made a fully informed, partially informed or uninformed choice about whether or not to have a colonoscopy. One person in the numerical study arm and one person in the positive result letter study arm made a fully informed choice not to have a colonoscopy, while the remainder of those who were classed as fully informed said they would take up the offer of colonoscopy. No significant difference was found when comparing the level of informed choice between the three study arms (p > 0.05).

The mean anxiety score for those in the positive result letter arm (36.8, 95% CI 33.9 to 39.8) was statistically significantly higher than for those in the numerical study arm (32.2, 95% CI 29.8 to 34.6, p = 0.02) and those in the categorical study arm (33.8, 95% CI 30.1 to 35.4, p = 0.04).

The proportion of subjects who reported that they strongly agreed with the statement that the information was easy to understand was 61.2%, 72.5% and 66.3% in numerical risk group, the categorical risk group and positive result letter group, respectively. The differences between the study arms were not statistically significant (p > 0.05). No participants across any of the study arms showed any disagreement with this statement. 19.1% of those in the numerical risk group, 24.0% of those in the categorical risk group and 29.6% of those in the positive result letter group agreed to some extent that they found the information presented distressing (p > 0.05), although less than 5% in each arm strongly agreed.

One hundred twenty-four (40.3%) of all participants reported that they had previously had a colonoscopy, with 43 (43.0%) in the numerical risk study arm, 41 (39.4%) in the categorical risk arm and 40 (38.5%) in the positive result letter arm. Logistic regression analysis showed that previous experience of colonoscopy did not appear to have an impact on intention in the highest and moderate risk groups for the two risk information groups or for the positive result letter study arm (p > 0.05). In the lowest risk group in the categorical risk study arm, however, people were more likely to report that they would accept the offer of a colonoscopy if they had previously had the test (odds ratio [OR] 3.13, 95% CI 1.20 to 4.02). Overall, previous colonoscopy experience was associated with a positive score for attitude towards bowel screening and colonoscopy (OR 1.81, 95% CI 1.11 to 2.95), but not when assessed for the individual study arms (p > 0.05). Previous experience of colonoscopy did not show association with STAI anxiety scores. Even when classifying participants according to a STAI score cut-off of > 40 (considered to indicate clinically significant symptoms of anxiety [25]), the ORs to indicate a relationship with previous colonoscopy experience were not statistically significant for any of the three study arms, nor overall (p > 0.05).

Thirty participants took part in telephone interviews, with 10 from each of the three study arms. All participants across all groups said that they found the scenario letters easy-to-understand acceptable and would not put them off from future participation in the screening programme. In general, participants had a good understanding of how increasing age and the amount of blood found in the screening test sample act as risk factors for CRC, although many were not clear on male sex as a risk factor. Participants in the numerical and categorical risk information groups were asked about their understanding of the different risk groups they were presented with, their reasons for deciding whether or not they would accept the offer of colonoscopy and how they felt the risks of the procedure compared to the benefits. Of those who completed a telephone interview, four participants in the numerical risk information group and six participants in the categorical risk information group had indicated on their returned materials that they would accept the offer of colonoscopy in all three risk scenarios. Explanations included “Although the medium and low scenarios had a much reduced chance of developing bowel cancer, if I was offered a colonoscopy I would accept, as I wouldn’t think it was being offered unless there was justifiable reason for doing the procedure” (numerical risk group) and “For peace of mind, even although you were low risk you would say ‘let’s know for certain’” and “I feel that if there was even the slightest risk of developing bowel cancer, the discomfort of a colonoscopy is a small price to pay for early detection” (categorical risk group). Two participants in the numerical risk group reported that they would like to be told either way whether or not they should have a colonoscopy, with one saying “I would like to know ‘yes’ you have bowel cancer or ‘no’ you do not have bowel cancer. I feel that telling me I have a 1 in 400 or 1 in 4000 chance of getting bowel cancer is as much use to me as being told my chances of dying as a result of a car crash or being hit by a bus; just something else to worry about!”. Conversely, two participants in the categorical risk group reported that they would like to be provided with a numerical level of risk, saying “I feel if percentages of risk and recovery were given, I could have made a more informed decision” and “I would have appreciated more hard facts at this stage, actual percentages of those in each of the risk groups subsequently found to have bowel cancer”. The majority of participants across all three study arms expressed that they felt that the risks associated with colonoscopy were outweighed by the benefits of having a colonoscopy. No differences in themes identified in the qualitative data were found between different age quintiles, sexes or SIMD quintiles.

The questionnaire also allowed participants to record their views on the information presented using free text. In addition to the participants who completed telephone interviews, a further 172 patients (59, 64 and 49 in the numerical risk arm, categorical risk arm and positive result letter arm, respectively) took this opportunity. Twenty-eight (47.5%), 18 (28.1%) and 30 (61.2%) participants, respectively in each study arm, made comments giving positive feedback regarding the ease of understanding and acceptability of the materials provided. Participants in the risk information arms also spoke positively about making an informed choice. One participant in the numerical risk arm said “Though I would be concerned about any future risk of cancer, I prefer an honest approach to the situation”. Similarly, those in the categorical risk arm wrote comments such as “Gave me a clear understanding i.e. risk. Would prefer to know as much as possible of my situation and risk level”. However, one participant in the categorical risk arm said, regarding a previous screening experience, “I didn’t receive a booklet regarding risks - glad I didn’t, I now feel more anxious and concerned if I need a third colonoscopy! Sometimes ignorance is bliss!” and another in the positive result letter arm said “Some people will think it’s embarrassing or worry too much given too many details”. A few participants commented on the use of colours in the infographics, with one participant saying it was helpful (categorical risk group) while one in each of the risk information groups said the use of red was “frightening”. Some participants in the categorical risk arm gave feedback that they were not clear on the levels of risk provided to them, for example “I think I understand what is meant by ‘lowest risk’ and ‘highest risk’, but ‘moderate risk’ is more difficult. Can you provide more information about this category? I think I would err on the side of caution and have a colonoscopy - if only to reduce ‘moderate risk’ to ‘lowest risk’” and “The letters would suggest everyone, regardless of risk will be offered a colonoscopy. However, if there is a “not currently at risk” group then this should be incorporated”. Three participants also commented that their responses to the materials may have been different if they received them in real life, rather than hypothetically.

When given the choice whether or not to have a colonoscopy, the majority of participants reported that they would take up the offer, regardless of the level of risk that they are provided with. Responses to the open-ended questions in the questionnaire and during telephone interviews demonstrated a theme that many people consider the risk of CRC as being more serious than the risks of colonoscopy, even when the latter is statistically more likely. Existing data on FIT in the SBoSP indicates that 2.5%, 38.8% and 58.7% of people would be in the highest, moderate and lowest risk groups, respectively. Combining these proportions with the findings of this study suggests that if all bowel screening participants were offered an informed choice to have colonoscopy, over two thirds of participants would opt to have the test. This would represent an increase in the current number of screening colonoscopies performed in Scotland from approx. 14,000 per annum to approx. 400,000 per annum. Clearly, an additional demand of this magnitude on colonoscopy services would not be manageable, with only a very small proportion of cancers and pre-cancers detected.

There are currently no accepted measures of informed choice in CRC screening. Our study has shown that bowel screening participants generally respond very positively to receiving personalised risk information using infographics. An encouraging finding was that providing more information does not appear to make people more anxious, with mean anxiety scores lower in those in the groups receiving risk information compared with the control arm. The proportion of participants who reported decisional conflict was very low in both groups who received risk information, for all three levels of risk. However, responses to the intention question used as a proxy for behaviour showed a greater level of uncertainty was apparent with the lowest risk groups (1 in 3500 or lowest risk); over a fifth of participants reported that they were unsure whether or not they would decide to have a colonoscopy. The finding that those at lowest risk were not adequately reassured that colonoscopy was not required indicates that providing participants with a fully informed choice in this setting may be inappropriate if it is not accompanied by some form of advice. Similarly, the qualitative results revealed that there was some uncertainty around the meaning of the categorical risk groups, with ‘moderate risk’ in particular perhaps lacking clarity. This feedback also suggests that providing risk information alongside more explicit advice on whether or not colonoscopy is required is likely to be a more pertinent approach. This may also have the desirable effect of increasing colonoscopy uptake in those at highest risk above its current rate of 77%. The majority of participants in the two risk information groups were classified as being fully informed, although this was similarly true of the positive result letter study arm; being representative of current practice, this is reassuring.

A major strength of the study is that development of the study materials was achieved using iterative feedback provided by a PPI group who were selected to be representative of the population eligible for bowel screening in Scotland. They unanimously agreed on the layout, content and readability of the final materials giving the study team confidence that the materials would provide an informed choice about whether or not to undergo a colonoscopy. Indeed, the vast majority of participants in all study arms and across all risk categories recorded scores on the informed subscore of the Decisional Conflict Scale which are associated with implementing decisions. In addition, final recruitment of participants was positive, with targets of 100 in each study arm being met. A further strength is that study invitations were weighted towards those in demographic groups previously shown to be least likely to participate. Although participation was still lowest in those in the most deprived quintiles, this would have been more pronounced without the selective invitation process. Comparable scores were recorded across the three study arms for knowledge of CRC screening and colonoscopy. This would be expected since all three received the same information materials, but it is reassuring that a similar, high level of understanding was demonstrated by the three groups.

The main limitation of the study is we cannot account for the likely gap between intention and behaviour. Indeed, free text comments in the study questionnaire included indications that the thought process when deciding whether or not to undergo colonoscopy may be different in a “real life” scenario. To fully assess the impact of offering a more informed choice on uptake of screening colonoscopy would require incorporation of a full scale RCT into the screening programme invitation process. However, the results of this study suggest that even with adjustment for the intention-behaviour gap, a huge increase in demand on the colonoscopy resource would still occur.

Since the study invitation was separate to the actual screening invitation, with the cohort including people who had not previously taken part in screening, the results are applicable to the overall population eligible for screening. However, a further potential limitation of the study is participant bias, in that those who accepted the study invitation and then completed and returned the materials are those who are more likely to be more health conscious and participate in screening. However, in actual practice, the offer of informed choice following FIT would only be made to those who had completed the test, so we believe this would be an acceptable, if not desirable, bias. The finding that 40% of participants had previously had a colonoscopy may also represent participation bias. It is likely that a reasonable proportion of the screening population as a whole, given the age range, will have had previous experience of colonoscopy but we do not have data available for comparison to measure the significance of this as a study limitation.

Numerical risks were calculated according to three subgroups of f-Hb only. In reality, the level of risk given would be further refined according to age and gender. The use of only three numerical risk levels (1 in 40, 1 in 1600 and 1 in 3500) allowed intra-participant comparison between all responses to the three scenarios in this study arm. For the same reason, all participants in the numerical and categorical risk arms received all three levels of risk within their scenario booklet (1 in 40, 1 in 1600 and 1 in 3500 or highest, moderate and lowest risk, respectively). However, this opportunity to consider each risk scenario by contrasting with the other two may have had an effect on their reported intention. Therefore, intention may differ if providing personalised risk in reality, where participants only receive one level of risk.

It should also be noted that was only possible to present risk scenarios using existing SBoSP data from the numbers of screen-detected and interval cancers. Since the screening programme aims to detect both CRC and its pre-cursors, the calculated risk scenarios would be an underestimate of the risk of all screen-relevant lesions.

Some participants reported that they would like to consult a medical professional about their decision whether or not to accept the invitation for a colonoscopy. Indeed, a new BMJ guideline for CRC screening [26] is the first to avoid a blanket recommendation for CRC screening in those above a certain age and instead suggests a personalised and risk-based approach should be made between the individual and the clinician. The guideline recommends that using a shared decision tool such as the QCancer® calculator (qcancer.​org/​15yr/​colorectal/) which incorporates age, sex, ethnicity, BMI, smoking status and family history of gastrointestinal cancer to estimate a personal CRC risk over the next 15 years.

Another theme identified in the qualitative analysis was that people thought that colonoscopy would not be offered to them unless it was required. The trust that the UK population holds in the NHS may be a factor contributing to the majority of people reporting that they would intend to have a colonoscopy if it was offered, regardless of risk. If the same protocol was followed in other countries, different results may occur with different healthcare systems to that of the UK. Furthermore, it is important to note that the f-Hb threshold of 80 μg Hb/g faeces used for positivity in the SBoSP is much higher than that used in other countries. Therefore, if a similar approach to ours was adopted in other countries, the moderate risk scenario would cover a much narrower f-Hb range than that of 1 to 79 μg Hb/g faeces in this study. Given our results, it is possible that the blanket offer of colonoscopy to all screening participants may contradict the key principles of screening laid out in the seminal work by Wilson and Jungner [27] and the UK National Screening Committee [28]. These principles outline that in screening programmes the costs and risks (physical and psychological) should be balanced against the benefits and that a suitable cut-off should be defined and agreed. Alternative approaches to improve transparency of information offered to screening participants could also be explored, such as providing those with a negative result with their numerical f-Hb, so that they themselves are aware of the proximity of their screening result to the threshold. This idea is supported by the findings of a recent study from Italy demonstrating that those with two consecutive negative screening test results, but a cumulative f-Hb above the cut-off for positivity had a significantly increased risk of a future diagnosis of advanced neoplasia [29]. An indication of the individual’s proximity to the positivity threshold could be provided alongside strategies applying more intelligent use of FIT, such as varying the length of the screening interval according to f-Hb, with those closest to the threshold offered the opportunity to be screened more regularly than those with very low or undetectable f-Hb.

With a growing interest in informed choice based on levels of risk in screening programmes, our finding that offering fully informed choice may not be feasible when colonoscopy capacity is limited is applicable to CRC screening programmes worldwide. However, the response to the materials was very positive, suggesting that offering risk information to those in the lowest and moderate risk groups along with advice that it is not recommended that they undergo colonoscopy at this time may be an option. A further full-scale RCT would be required to examine the actual uptake of colonoscopy in reality.