Responses to provision of personalised cancer risk information: a qualitative interview study with members of the public

To enable us to include individuals with a range of backgrounds and of different ages we used three recruitment strategies: 1) posters in the waiting rooms of ten Primary Care Practices across Cambridgeshire; 2) letters of invitation and the participant information leaflet sent via the lead of the group to a local patient participation group; and 3) letters of invitation and the participant information leaflet sent by Practice staff to 100 randomly selected patients aged between 18 and 65 years and registered at one Primary Care Practice in Cambridgeshire. By sending letters of invitation to a local patient participation group and to individual patients at one Primary Care practice, we were able to reach potential participants who may not have accessed healthcare during the recruitment period. For those responding to the posters or invitation letters via the patient participation group, no age limits were applied. The letters of invitation to individual patients were sent out part way through recruitment and the decision to limit to an upper age of 65 years was made to purposively sample younger people. In all cases those interested in taking part in the study were invited to contact the research team and all those who completed the pre-interview questionnaire and agreed to an interview were included in the study. Letters of invitation to patients registered at the Primary Care Practice were sent out in batches of 20 and we continued recruitment until we reached data saturation.

Prior to the interviews, participants provided their address and were posted a screening questionnaire asking about their risk factors for cancer. This included questions about age, sex, family history of cancer, medical history and lifestyle factors such as smoking, alcohol consumption, diet, weight and physical activity. Responses were used to i) calculate an individual’s risk of developing the cancers specified below; and ii) contextualise findings from the subsequent interviews. All those with a previous diagnosis of cancer were advised at this stage that the risk algorithms being used had not been tested in those with a previous diagnosis of cancer and so might either underestimate or overestimate their risk. Socioeconomic status was computed using the participants’ postcode and the English indices of deprivation 2010 available online [33].

To calculate the absolute and relative risk of individual cancers for each participant we adapted the risk algorithms used in the “YourDiseaseRisk” models [13] for 11 cancers (female breast, colorectal, stomach, bladder, prostate, kidney, melanoma, ovarian, lung, pancreatic, and uterine) for the UK population. These models were chosen as they have a consistent approach across multiple cancer sites. Full details of the method used to adapt the algorithms are published elsewhere [34]. In brief, we first estimated the prevalence in the UK population for each of the risk factors related to each cancer type and used that data alongside the “YourDiseaseRisk” algorithms to calculate an individual’s risk of developing each of the cancers relative to the average UK population.

To enable calculation of the 10-year estimated absolute risk for all 11 cancers we used the “Current Probability” method [35] based on data from the Cancer Research UK website [36] for age- and sex-specific cancer incidence and data from the Office of National Statistics [37] for age- and sex-specific death rates from all causes. This allowed us to estimate the average age- and sex-specific UK 10-year absolute risk for each cancer and we multiplied this by each individual’s relative risk to obtain the estimated 10-year risk for individuals.

To enable communication of the risk to participants we developed a web-based tool that presented the estimated risk for each cancer either individually or alongside others. In order to choose the formats in which to present the risk we first reviewed published literature [27, 38‐40] and the internet to compile an inventory of different formats currently in use for cardiovascular disease and cancer. Informed by best practice guidance for communication of risk [41, 42] we then met with patient and public representatives to discuss a range of formats and choose those to use for the study.

The chosen formats are shown in Fig. 1. They are: a) absolute 10-year percentage risk on a vertical thermometer scale in grey-scale; b) absolute 10-year percentage risk on a vertical thermometer scale coloured from green to red; c) absolute 10-year percentage risk as a coloured bar chart shaded; and d) relative 10-year risk on a qualitative scale from ‘Low risk’ to ‘High risk’. All participants were shown the formats in the same order for either breast or colorectal cancer to allow us to assess the impact of colour over a grey-scale image. In all formats participants were given the opportunity to select changes to their lifestyle such as those shown in panel e of Fig. 1 and see the impact of those changes on the calculated risk of that cancer. They were then given the option to view up to six cancers of their choice from those available simultaneously. Participants with a previous diagnosis of cancer were not given the option to see their risk of that specific cancer.

The one-to-one face-to-face interviews were conducted by one of two female researchers (JUS, an academic GP or BS, a psychologist; both with PhDs and experience of conducting qualitative research) and, depending on the participant’s preference, held either in the participant’s home (n = 11) or in a room at the University of Cambridge (n = 16). The interviews were semi-structured and based on a schedule covering their views on receiving personalised risk information in general, of different representations of that risk, the potential for motivating lifestyle change, and the timing and site of delivery of the information (see Additional file 1). Each lasted between 20 and 50 min and at the end of each interview the participants were offered a summary of their risk of the individual cancers and Cancer Research UK leaflets on how to reduce their risk and how to recognise signs and symptoms of cancer. The background of the researchers was not discussed during the interviews.

The interviews were audio-recorded and transcribed verbatim and analysed using thematic analysis [43] and the one sheet of paper (OSOP) method [44]. After approximately half of the interviews had been completed, the transcripts were first read and re-read by three researchers (a UK academic GP, a health psychologist and a social scientist). A coding frame (available from the authors) based on the transcripts and the interview schedule and informed by relevant concerns in the literature was developed by those same researchers. One researcher (a UK academic GP) then coded all the transcripts with the aid of NVivo software (QSR International, version 10). Once coding was complete, we created a written summary of the data for each code using the OSOP (one sheet of paper) method where every section of data relevant to that code from all the interviews was noted. These were discussed amongst the wider research team to identify the key themes first within each code and then across the entire dataset and any outlying views. These summaries were added to as further data were collected and once all data had been collected and summarised we then explored these themes in depth, looking for any patterns across different ages, sexes, current lifestyle habits, and personal and family history. Throughout the text illustrative quotes are identified by sex and age group.

The first reaction of almost all participants to being presented with their 10-year absolute risk of either breast or colon cancer in grey-scale was that it was low and they were either reassured or happy with the results and not concerned or worried.

“Well, it’s not even 1%. I just think that’s really good, out of 100%, I don’t really need to worry about colon cancer, that’s what I would say.” (F,30–40).

In some cases these impressions were clearly related just to the fact that the number was small or to the position on the scale rather than an appreciation of the meaning of the number itself.

“Good, I’m relieved. Just seems a low number. I don’t know whether it is or not, but, is it a low number?” (M,>60).

“Okay, that’s quite low then... ‘cos it’s at the very bottom of the scale. Yeah, it’s at the bottom so quite happy with that.” (F,20–30).

Only one participant thought the risk was fairly high and higher than they had expected.

When asked what would constitute a high risk for them there was a wide range of responses from 0.5% to 60%. The most common were 10% (n = 7), 20% (n = 4), 5% (n = 3), and 1% (n = 3) which together accounted for over half of the responses. There was no association with age or deprivation.

Particularly noticeable was the finding that over half were unclear about why they had chosen the number they had. Some individuals were explicit in saying that they just liked the sound of the numbers they chose whilst others found it difficult to explain their reasoning.

“I don’t know, it’s just something that, yeah, the number that just came into my head” (F,30–40).

A small number of participants related it to something in daily life or other diseases. In general, those who related it to something specific tended to choose lower, more accurate, numbers than those who didn’t.

“1% is quite a high risk for many things that, if you don't have to take it you wouldn't take it. I mean, would you choose to cross the road if one time in a hundred it happened? No, you'd find a different way.” (M,30–40).

“I would have probably compared it to cardiovascular disease risk which would be, currently it’s 10% over 5 years isn’t it I think so I guess I would have viewed it as the same figure for cancer.” (M,>60, history of cancer).

Most participants felt personalised cancer risk information would help others to make changes to their lifestyle.

“I definitely think it's helpful and I think it's gonna really motivate people to make changes because, yeah, cancer's scary and if you see that your risk can drop it, yeah, I think it might really motivate people to make changes.” (F,30–40).

A number also reported intentions to change their own behaviour after receiving their personalised cancer risk. For some this was because seeing their risk reaffirmed plans for changes they had already been considering. For others the motivation came directly from now wanting to lower their risk.

“... for me it reaffirms that my choices and changes that I’d like my lifestyle and it’s actually in the right direction and it will certainly help me to stick to them and continue with these choices for the perceived future.” (M,30–40).

“It motivates me now to decrease my risk.….I will try to start lifestyle changes now, yes….I’d like to lose some weight, and eat, you know, healthily.” (F,30–40).

There were also several who described a balance with some things being easier to achieve and so possible at a lower risk and some that would require a higher risk in order to motivate them.

“Reduce red meat, I guess I could try, maybe I would, so I would think, okay, so that’s my risk but it says less than three servings per week, I could do that I think or at least get a, go towards that, you know….. Eat the vegetables no, I’m afraid not.” (F,40–60).

However, for many, receiving their personalised cancer risk did not influence their intention to change their behaviour. Reasons for this fell into three groups (Table 2). The first, and most common, was that the risk was perceived to be so low that reducing it further was not a priority. The second, who tended to be those with the lowest risk, were reassured that they were already living a healthy lifestyle and so did not feel they needed to make any changes. The third group were older participants who felt that it was too late to start worrying about cancer, either because they felt it was simply too late to change or because they accepted that they had to die from something.

All participants were supportive of offering cancer risk information more widely and would recommend it to their family and friends. There was a clear distinction between offering and providing though, with recognition that it should be up to the individual and may not be appropriate for everyone, particularly those who are already very health-conscious.

When asked how they would prefer to receive this information outside the context of a research study the majority expressed a preference for face-to-face with the others preferring it online or online with the option to discuss it face-to-face. Reasons participants gave for preferring face-to-face delivery were that it provided an opportunity to ask questions and confirm their interpretation and that looking at it at home alone for the first time had the potential to be scary and cause panic. Some also felt receiving the risk in person was more likely to motivate people to make changes to their lifestyle.

“I think face-to-face is better you know in a general sense for any communication, just from my experience I think whatever you do, face-to-face is much more, it’s more likely to make it have an effect I think.” (M,>60, history of cancer).

Many mentioned their Primary Care physician (PCP) or other primary care healthcare professionals as the most appropriate people to deliver the information. However, several were concerned that PCPs would not have the time or should focus on other aspects of care. There was a general view though that, if the information was not being given face to face by the their PCP or another health care professional, it needed to be endorsed by the PCP to provide credibility.

These findings must, additionally, be interpreted with consideration of the limitations of the study. The main limitation is that, by definition, the participants were a small, self-selected group interested in their cancer risk as they had responded to invitations to take part in a research study to help the development of personalised cancer risk tools. We attempted to reduce this bias by using three different recruitment strategies and approximately half of the final sample reported either not being concerned about their cancer risk prior to the study or having not really thought about it. Nevertheless we cannot exclude the possibility that the self-selected nature of the sample impacted on the results. The sample also included participants with a wide range of ages and three participants with a history of cancer. This was a deliberate decision at the time of recruitment in order to capture the views of all those to whom cancer risk may be provided in the future, including those with a diagnosis of cancer where there is increasing evidence that lifestyle change can improve survival [53‐55] as well as prevent a second cancer. Whilst we did not see a clear difference in the views across the different groups, it remains possible that the views of younger participants may change as they get older and those with a personal history of cancer may interpret risk quite differently depending on the cancer they survived. Those with a previous history of cancer were also told that the risk scores had not been tested in people with experience of cancer and so the meaning of the estimated risks to them may have been different from those without a history of cancer. The effect of the presentation of risk on intention to change behaviour may also have been limited by the characteristics of the participants. Most were from areas of low deprivation, were highly educated and were already achieving the physical activity recommendations. Instead of focusing on their own lifestyle, many instead also discussed how useful they thought it would be for other people. This response may have been supported by knowing they were taking part in a study to discuss different ways of presenting cancer risk.