Public attitudes towards screening for kidney cancer: an online survey

We conducted an online population-based survey. The survey was conducted and is reported in line with the checklist for Reporting Results of Internet E-Surveys (CHERRIES) [11].

We aimed to recruit 1000 participants (50% female) through Prolific (www.​prolific.​ac), an online platform in which individuals volunteer to take part in studies and are compensated for their time at an hourly rate (£6/hour). Participants were eligible if they were between 45–79 years of age and, in line with accepted practice [12], had a Prolific approval rating ≥ 95%. The approval rating reflects the percentage of studies that participants have completed that have been approved by researchers. We chose the age range 45–79 years because it is unlikely that a future kidney cancer screening programme would screen individuals younger than 45 and we wanted to be able to capture any differences in attitudes in older groups. We did not prevent participants with a history of kidney cancer from completing the survey but their data were subsequently excluded from the analysis.

When the survey was launched, a random subset of eligible participants were invited to complete the survey via email through the Prolific platform (Additional file 1). The survey then remained open to new participants until recruitment was complete. After starting the survey, all participants had the option to ‘return’ their submission at any point. This allowed them to stop part way through the study without affecting their Prolific rating. Data from participants who returned their submission were deleted and not analysed. Cookies were not used to identify each participant computer. Duplicate entries were prevented within the Prolific platform through use of unique participant identification numbers.

The survey was developed by the study team and piloted with two patient and public representatives to check that the questions were comprehensible and being interpreted correctly. Validated questions were used wherever possible. A complete version of the survey is provided in Additional file 2. Briefly, participants first provided socio-demographic details including age, lifestyle, highest level of education, employment and socioeconomic status (assessed via the occupation of the household’s chief income earner [13, 14]). They were then asked questions about their medical screening history, perception of cancer risk, cancer worry [15], knowledge of kidney cancer symptoms and risk factors for kidney cancer [16]. Participants were then given brief summary information about kidney cancer and the estimated benefits and harms of four potential screening methods using data from the literature [5]. A series of questions followed about beliefs and attitudes concerning kidney cancer and screening [17] and whether, based on the information provided, participants would take part in screening. Finally, participants were asked about how much influence various other factors, identified from the literature [18], would have on their decision to undergo screening. Written online consent was obtained from each participant before they began the survey.

The questions and scenarios were embedded within an online questionnaire and two studies (one for men and one for women) run in parallel on the Prolific website to enable us to achieve a sample with equal numbers of men and women. All questions except those inviting free text responses were mandatory for all participants. The questions were distributed over 22 pages with between one and seven questions per page to enable completion on tablets or smartphones with small screens. It was not possible for participants to go back through the pages and change their answers. The technical functionality of the electronic questionnaire was tested by the study team. We included an instructional manipulation check to identify inattentive participants and increase the validity and reliability of the responses [19]. Participants who failed to answer this question correctly were excluded from the study.

We used descriptive statistics to characterise the participants and their overall attitudes towards, and intention to take-up, kidney cancer screening. Socioeconomic status (SES) was classified as higher (ABC1) and lower (C2DE). For awareness of kidney cancer symptoms and risk factors participants were categorised into two groups (high/low) based on whether their overall score was above or below the median (see Additional file 2 for details).

We used a linear regression model to analyse differences between the burden/ inconvenience and worry associated with each screening test. The null hypothesis of no difference between the four groups was tested using an F-test, followed by the estimation of four pairwise contrasts between the screening tests: (1) urine vs blood; (2) urine vs ultrasound; (3) urine versus CT; and (4) ultrasound vs CT. Confidence intervals (98.75%, based on a Bonferroni corrected significance threshold of 1.25%) are presented for these analyses to acknowledge that four pairwise comparisons were performed.

Logistic regression was then used to examine the influence of the different screening tests and participant characteristics on intention to take up screening. For these analyses, cancer beliefs about treatment was treated as a continuous variable from 1–5, 1 reflecting the view that most cancer treatment is worse than the cancer itself. Cancer beliefs about outcomes was similarly treated as a continuous variable, computed by combining responses from the relevant questions 1, 4 and 6 using published coefficients [17] (see Additional file 2 for details).

To enable us to compare the intention to take up each of the five screening test modalities, we included the screening test as an independent variable as well as adjusting for participant characteristics (age, sex, education level, ethnicity, self-reported general health, country of residence, income group, body mass index (BMI), smoking status, prior history of cancer, family history of kidney cancer), cancer worry, perceived risk of cancer, beliefs about cancer outcomes and treatments, the mean burden/inconvenience and the mean worry associated with all of the tests combined for each participant. The analysis was adjusted for the clustering of responses by participant using the—cluster—command in Stata. To enable comparison of participant characteristics associated with intention to take up each of the screening tests separately, we repeated the analysis for each screening test with the burden/inconvenience and worry associated with each specific test. We checked for collinearity between the covariates in the regression models by calculating the variance inflation factors (vif) for each covariate using the command –collin- in Stata. All variables had vif values less than 4.20, suggesting no evidence of substantial collinearity. We also separately assessed for interactions between cancer worry, perceived risk of cancer and beliefs about cancer outcomes, and burden/inconvenience of the screening tests and worry about the screening tests by adding interaction terms into the regression. The hypothesis being tested was that worry about the screening tests and the perceived burden/inconvenience of screening may have less of an impact on intention to take up screening among those individuals with higher cancer worry, higher perceived risk of cancer or stronger negative beliefs about the outcomes of cancer. Interactions were considered significant if p<0.05. Where the interaction was significant we additionally stratified by high (> 5) and low (≤ 5) cancer worry and estimated the association between worry about screening tests and intention to take up screening separately in these two groups. All results from these analyses are presented as odds ratios (OR) with 95% confidence intervals.

All analyses were performed using Stata Version 14.

Two members of the public contributed, through face-to-face discussions and email, to the development of the survey and participant information sheet. They also provided comments on the final manuscript prior to submission.

Table 2 shows the participants’ general cancer beliefs and their attitudes towards kidney cancer screening. The majority of participants were positive about cancer outcomes, only 6.7% expressed cancer fatalism. Attitudes towards screening were also positive, the majority (91.8%) agreed with the statement that screening could reduce the chance of dying from kidney cancer.

Intention to take-up screening was high overall (Table 3), with over 90% reporting that they would accept the invitation for screening by urine testing, blood testing or low-dose CT combined with lung screening. Intention was lower for ultrasound (89.5%) and low-dose CT of kidneys alone (78.8%).

Fig. 1 shows the pattern of responses at an individual level across the five modalities. A majority (771, 76%) reported that they would attend screening by all five modalities, 20 (2%) said they would not attend screening by any method and 28 (27%) reported they would only attend screening by the combined lung and kidney CT scan and not by the other methods. 217 (21%) participants would not attend screening by CT scan of the kidneys alone, but 181 of these 217 (83%) participants would attend kidney cancer screening by CT scan if lung cancer screening was also included.

The mean burden/inconvenience and the mean worry for each of the four screening tests are given in Table 3. Both were measured on a scale from 1–5. There were significant differences between the screening tests in both burden/inconvenience and worry, with significantly higher values for ultrasound, blood test and low-dose CT than for urine test [mean difference 0.38 (0.32–0.44) and 0.16 (0.12–0.19) for burden/inconvenience and worry respectively for ultrasound, 0.33 (0.28–0.37) and 0.11 (0.07–0.15) for blood test and 0.69 (0.62–0.75) and 0.47 (0.41–0.53) for low-dose CT]. Additionally, both burden/inconvenience and worry were significantly higher for low-dose CT than ultrasound (mean difference 0.31 (0.26–0.35) and 0.36 (0.31–0.41)).

After adjusting for participant characteristics (age, sex, education level, ethnicity, self-reported general health, country of residence, income group, BMI, smoking status, prior history of cancer, family history of kidney cancer), cancer worry, perceived risk of cancer, beliefs about cancer outcomes and treatments, and the burden/inconvenience and worry associated with the tests, and clustering of response by participant, there were significant differences between the different tests in intention to take up screening (Fig. 2). In particular, compared with intention to take up screening with a urine test, participants were half as likely to report that they intended to undergo blood [OR 0.56 (0.43–0.73)] or ultrasound [OR 0.50 (0.38–0.67)] testing compared to urine testing, and half again as likely to report that they intended to take part in a screening programme featuring a low dose CT scan [OR 0.19 (0.14–0.27)).

Participants with higher BMI and higher general cancer worry were more likely to report intention to attend screening (Fig. 3). Being more worried about the screening tests or associating more burden and inconvenience were associated with decreased odds of intending to take up screening overall in the study population. There was, however, an interaction between general cancer worry and worry about the screening tests, with worry about screening tests associated with reduced odds of intending to attend only in those with low cancer worry (cancer worry scale ≤ 5) (OR 0.67 (0.48–0.94)) and not in than those with high cancer worry (OR 0.98 (0.69–1.40)). There was no evidence of an interaction (p>0.05) for the other potential interactions assessed. Although not statistically significant as a result of the small numbers with a family history of kidney cancer, having a family history of kidney was also associated with an increased odds of intending to take up screening (OR 4.70 (0.41–53.7), with 27 of the 29 participants with a family history of kidney cancer likely to take up all four screening tests. These findings were similar when each screening option was considered separately (Additional file 3: Supplementary Table 2 and Additional file 4: Supplementary Fig. 1a-e). Notably all those with a family history of kidney cancer reported they would take up screening with an ultrasound scan, 28 out of 29 would take up screening with urine or blood testing, and 27 out of 29 would take up low-dose CT scanning.

The reported influence of other factors on intention to attend kidney cancer screening are reported in Table 4. Having symptoms of kidney cancer and a recommendation by their GP were the most influential factors.

The overall positive attitude towards a screening programme for kidney cancer mirrors the general enthusiasm for cancer screening seen in population based surveys [20] and findings of a similar study on a potential lung cancer screening programme [13]. In that study over 90% of survey respondents believed that there was benefit to lung cancer screening, with the majority agreeing that lung screening could reduce chances of lung cancer death and only a small minority endorsing avoidance of lung cancer screening due to fear of what might be found or low perceived effectiveness of screening.

The observed differences between tests in intention to take part in a screening programme, with a preference for urine testing over a blood test, ultrasound scan or low-dose CT alone, is also consistent with existing literature which reports a preference for non-invasive tests requiring no preparation and causing no pain or long-term harm [21‐23]. The greater intention seen for low-dose CT in combination with lung cancer screening likely reflects greater acceptability for combination screening, as seen in the context of a ‘One Stop’ cancer screening programme [24]. Given the low prevalence of kidney cancer, such a ‘bolt on’ screening programme may well be the most cost-effective as well as the most acceptable. It may also increase uptake of existing programmes [25].

The survey was developed with the support of patient and public representatives in order to maximise participants’ understanding and, where possible, previously validated questions were used. The main limitation is our use of an online recruitment platform. Although the platform we used has been developed specifically for research and enabled us to rapidly recruit participants of different ages and from different socio-economic backgrounds, compared with the UK and US population, the participants in this study were more likely to have university level education (46% in this study compared with 40% in the UK [26] and 35% in the US [27]) and be of White ethnicity (95% in this study compared with 86% in the UK [28] and 76% in the US [29]) and high socioeconomic status (76% in social group ABC1 in this study compared with 54% in England [30]). Uptake of other cancer screening programmes is known to be around 5–15% higher amongst these groups than those without university education, of ethnic minority groups and in lower social classes [31, 32]. The high proportion of participants expressing an intention to attend screening observed in this study may therefore be an over-estimation of intention in the wider population. Additionally, Prolific members are experienced in completing online tasks and their views may not be representative of those of the general population [33]. However, while these differences might lead to over-estimation of levels of intention and affect the absolute levels of worry and burden reported by the respondents as a whole, the differences observed between screening tests and the associations between worry and burden and other participant socio-demographic characteristics and intention to take up screening would be expected to be generalizable to other populations [34]. In the absence of an existing screening programme we were also only able to measure intention and not attendance at screening. Although there is an established relationship between screening intention and screening attendance [35], there is known to be an ‘intention-behaviour gap’ which is modified by various factors [36, 37]. The actual uptake of any future screening programme is therefore likely to be lower than the estimates of intention in this study. The uptake of any future screening programme will also likely depend on the information provided to eligible individuals. As data are currently not available on all the potential benefits and harms of kidney cancer screening, the information provided to participants in this study was based on estimates from the best available evidence and provided in less detail that would likely be included in any future screening programme. In particular, although we mentioned the possibility of both false positive and false negative findings as a result of screening, we did not provide participants with specific details on the possibility of detecting benign small renal masses or data on the growth rate of small incidentally detected tumours. With on-going research in this area, especially surrounding potential blood and urine biomarkers and approaches to the management of small renal masses, these estimates may change, making screening either more or less favourable to individuals.