A Systematic Review on Reporting of Methods in National Surveys about Adults’ Attitudes to Lifestyle and Environmental Risk Factors for Cancer

Abstract

The risks of developing cancer are significantly affected by our lifestyle and environment. While there are several uncontrollable risk factors, we can modify our lifestyle and our environment to reduce the increased threat of cancer. This systematic review aims to evaluate the methodological assessment used to evaluate attitudes about cancer risk factors among the general population. Two researchers independently screened the articles for inclusion and Critical Appraisal Skills Programme (CASP) checklists were used to assess the methodology of the included studies. Thirty-one manuscripts met the inclusion criteria with a majority of them focusing on attitudes to several cancer risk factors and six on specific cancer location risk factors. This systematic review highlights the diversity of notions used around attitudes and methods used in the method of administering the survey, as well as the format of the questions and the response scales. It is thus difficult to compare data between different countries. However, cancer is a global problem. Harmonizing methods could allow a comparison of data between countries. Recommendations to this effect are suggested.

1. Introduction

Cancer is a significant global public health problem, with approximately 18 million new cases and more than 9.6 million deaths from the disease in 2018 [1]. The number of cancer cases has been increasing steadily over the last 10 years worldwide, with a 33% increase in the number of cases between 2005 and 2015 [2]. Research has shown that certain risk factors may increase a person’s chances of developing cancer [3]. Around the world, researchers have estimated the proportion of cancer cases attributable to lifestyle and environmental factors [4,5]. Between 30 and 50% of cancers may be preventable by modifying or avoiding key lifestyle and environmental risk factors [6]. The modifiable risk factors include: smoking (active and passive), alcohol, diet, overweight, obesity, infections, occupational exposures, ultraviolet radiation, ionizing radiation, physical activity, hormones, breastfeeding, outside air pollution, arsenic, and benzene [7].

The risks of developing cancer are significantly affected by lifestyle choices, which are particularly influenced by our social environment. There are several uncontrollable risk factors, but we can modify our lifestyle and act at our environmental level to reduce the increased threat of cancer [8]. Primary prevention and health promotion through lifestyle and environmental interventions remain the main way to reduce the global burden of cancer [9]. It is widely agreed that protective behavior can prevent many cancers, and knowledge is a necessary predisposing factor for such behavioral change [10]. In the public health research literature, notions of knowledge, awareness, belief, representation, perception, or attitude are sometimes juxtaposed. Although these words are frequently used interchangeably without posing a problem, they sometimes appear to have different intended meanings [11]. Whereas Trevethan (2017) recommended that “knowledge” be used to refer to information that is detailed and factual and that awareness be associated with personally relevant information, other authors [11,12,13] have used the term “attitude” which seems to include all of these concepts. For these reasons, in our article, we will systematically use the term attitude.

Several national surveys have been conducted to better understand attitudes about cancer risk factors, notably environmental factors (e.g., Health Information National Trends Survey, Baromètre du Cancer). For both decision-makers and researchers, it is important to know attitudes about cancer risk factors among the general population to guide public health policies and health communication strategies and to design lifestyle and environmental interventions. It is therefore necessary to identify national surveys reporting people’s attitudes about cancer risk factors, in particular, to point out the methodological similarities and divergences to provide an international working basis for future surveys about people’s attitudes about cancer risk factors, and to identify possible elements of comparability between studies and countries. To our knowledge, the reporting of methods used by national surveys to evaluate attitudes about cancer risk factors has not been systematically assessed.

To fill this gap, the French National Cancer Institute (INCa), which is the preeminent health and science agency in charge of cancer control in France, piloted a systematic review to evaluate the methodological assessment used to evaluate attitudes about cancer risk factors among the general public. Based on the highlighted results, recommendations to assess the perception of cancer risk factors can then be suggested.

2. Materials and Methods

According to French regulations, IRB approval is not required for this study.

2.1. Search Strategy

A search in the Prospero database (https://www.crd.york.ac.uk/PROSPERO/ (accessed on October 2020)) showed that no literature review is currently in progress on this subject (search terms: risk factors AND cancer AND knowledge OR awareness OR perception). This systematic review was guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) [14].

A search was performed from 1 May 2021 using the following databases: PubMed, PsychINFO, and PsycARTICLES, limiting the search to national surveys of adults (>18 years at diagnosis) published in English or French in peer-reviewed journals. Search terms are presented in Box 1 and the research equations used in PubMed (see Appendix A) were adapted for the other databases.

Box 1. Search terms used in the databases.

Representation* OR Awareness OR Perception* OR Knowledge* OR belief* OR attitude*

AND

risk factor or risk factors

AND

neoplas* OR cancer* OR tumor OR tumors OR tumour OR tumours OR malign* OR adenocarcinoma* OR carcinoma*

2.2. Inclusion and Exclusion Criteria

Qualitative, quantitative, or mixed design studies were included if they fulfilled the following criteria: (a) assessed environmental and lifestyle risk factors for cancer among the general public; (b) conducted on people aged 18 and over; (c) written in English or French; (d) published in peer-reviewed journals.

Predefined exclusion criteria were: (a) reviews, case–control studies, protocol or pilot studies and studies that evaluated cancer risk factors and not cancer risk factor awareness or the awareness of risk factors for specific cancers (e.g., breast cancer); (b) unpublished papers due to the lack of peer-review oversight; (c) studies on cancer patients or their families, or people at risk of developing cancer; (d) protocol studies or studies that described and/or evaluated the effectiveness of an intervention on knowledge and lifestyle for cancer prevention; (e) cohorts that were not representative of a defined population; (f) studies focusing on a subgroup of the main sample.

2.3. Study Selection and Data Extraction

Duplicate articles were removed. Title and abstract screening was conducted according to the eligibility criteria. We obtained full articles for all titles that appeared to meet the inclusion criteria. Full-text reviewing was conducted by two different researchers (Stephane Faury and Jerome Foucaud) to establish a final list of eligible studies. Data extraction was performed by S.F. and checked by J.F. In cases of disagreement, a third researcher (Bruno Quintard) was requested for an opinion, and a consensus was reached among the researchers (Stéphane Faury, Jérôme Foucaud, and Bruno Quintard). Additional papers were searched using the reference list of eligible studies. Data collection was performed with the aim of including: authors, year, country, sample, study design and methods, objective(s), main results, and limitation of the method. Extraction included all reporting items used in the studies. Items other than the main preventable cancer risk factors were classified into several categories by Philémon Aurouet, and a consensus was reached among the other authors on this categorization.

2.4. Critical Appraisal of Study Quality

The methodological quality of included studies was independently assessed by two researchers (S.F. and J.F) using the Critical Appraisal Skills Programme (CASP) checklists [15,16,17]. In cases of discrepancies, the manuscripts were discussed verbally. CASP checklists consist of three sections (Section A: “Are the results of the trial valid?”; Section C: “What are the results?”; Section C: “Will the results help locally?”). As the CASP checklist does not provide a total score for each study, based on Lamore et al. (2019) [18], we chose to classify the studies as either (1) a low-quality study (i.e., participants not recruited in an acceptable way and weak results), (2) a medium-quality study (i.e., participants recruited in an acceptable way and weak to moderate results) or (3) a high-quality study (i.e., participants recruited in an acceptable way and strong results).

3. Results

The initial search yielded 1417 records. After the removal of duplicates, the titles and abstracts of 1185 records were screened; 1129 records were identified as clearly non-relevant and were thus excluded: 56 were retained for full-text analysis. Finally, 31 articles [19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49] were included in this systematic review without any disagreement (i.e., inter-rater agreement = 100%). For two articles [22,43], we contacted the study authors to obtain additional information but did not receive any responses. Figure 1 presents a flow diagram of the research article selection process.

3.1. Study Design and Participant Characteristics

The 31 cross-sectional studies included in the review were published between 1992 and 2021 and conducted around the world (see

Table 1. Countries of included studies.

Country of Publication 1	Number	References
Australia	6	[19,21,23,24,29,35]
United-Kingdom	4	[29,41,44,49]
United States	4	[22,25,30,48]
India	3	[36,39,40]
Canada	2	[29,47]
Danish	2	[29,33]
France	2	[34,38]
Swedish	2	[29,33]
Turkey	2	[32,45]
Ethiopia	1	[20]
Iran	1	[28]
Ireland	1	[43]
Japan	1	[31]
Mexico	1	[37]
Morocco	1	[26]
New-Zealand	1	[42]
>>Palestine/Gaza	1	[27]

¹ Some studies included multiple countries.

). The majority of the studies focused on attitudes to cancer risk factors (n = 24 out of 31; [20,21,23,24,26,27,28,29,30,31,33,34,35,36,37,38,39,40,41,43,44,45,46,47,48]) and six others focused on attitudes to risk factors for many specific cancers (for example, for breast, cervical, prostate and colon cancer) [19,22,25,32,42,49]. In total, 112,904 participants (from 358 to 19,076 participants), mostly women, were included in these studies. A description of the study included in this review is presented in Appendix B.

3.2. Quality Assessment of the Included Studies

The results of the quality assessment are summarized in

Table 2. CASP checklist results for assessing the methodological quality of the included studies.

Cohort Studies	Q1	Q2	?	Q3	Q4	Q5a	Q5b	Q6a	Q6b	Q7	Q8	Q9	Q10	Q11	Q12
Baghurst et al. (1992) [19]								NA	NA
Bantie et al. (2021) [20]							+/−	NA	NA
Buykx et al. (2015) [21]								NA	NA
Breslow et al. (1997) [22]								NA	NA
Cameron et al. (2010) [23]								NA	NA
Cotter et al. (2013) [24]								NA	NA
Daley (2007) [25]
El Rhazi et al. (2014) [26]								NA	NA
Elshami et al. (2020) [27]								NA	NA
Feizi et al. (2010) [28]								NA	NA
Forbes et al. (2013) [29]								NA	NA
Hawkins et al. (2010) [30]								NA	NA
Inoue et al. (2006) [31]								NA	NA
Karadeniz et Çetinkaya (2021) [32]								NA	NA
Lagerlund et al. (2015) [33]								NA	NA
Lamore et al. (2019) [34]								NA	NA
MacTiernan et al. (2014) [35]								NA	NA
Oswal et al. [36]							+/−	NA	NA
Perez-Contretras et al. (2004) [37]								NA	NA
Peretti-Watel et al. [38]								NA	NA
Puri et al. (2009) [39]								NA	NA
Raj et al. (2012) [40]
Redeker et al. (2009) [41]								NA	NA
Reeder et al. (2003) [42]						+/−	+/−	NA	NA
Ryan et al. (2015) [43]
Sanderson et al. (2009) [44]								NA	NA
San Turgay et al. (2005) [45]								NA	NA
Schliemann et al. (2020) [46]								NA	NA
Shi et al. (2020) [47]
Vanderpool et al. (2010) [48]								NA	NA
Wardle et al. (2001) [49]								NA	NA

Question for cohort study assessment: Q1 = “Did the study address a clearly focused issue?”; Q2 = “Was the cohort recruited in an acceptable way?”; ? = “Is it worth continuing?”; Q3 = “Was the exposure accurately measured to minimize bias”; Q4 = “Was the outcome accurately measured to minimize bias?”; Q5a = “Have the authors identified all important confounding factors ?”; Q5b = “Have they taken account of the confounding factors in the design and/or analysis?”; Q6(a) = “Was the follow-up of subjects complete enough?”; Q6(b) = “Was the follow-up of subjects long enough?”; Q7 = “What are the results of this study?”; Q8 = “How precise are the results?”; Q9 = “Do you believe the results?”; Q10 = “Can the results be applied to the local population?”; Q11 = “Do the results of this study fit with other available evidence?”; Q12= “What are the implications of this study for practice?”. Answer key: ; = yes or strong; ; = no or weak; +/− = uninterpretable; NA = not applicable.

. Very high inter-rater agreement was obtained. We are unable to answer questions Q6a and Q6b as to whether the follow-up of subjects was complete enough.

Four cohort studies were classified in the “low-quality” [25,40,43,47] group because the participants might not have been recruited in an acceptable way and this may compromise the extent to which the findings can be generalized (see answers to Q2). The aims of the Ryan et al. study [43] and the Shi et al. study [47] were to assess public perception of the risk factors for cancer. However, participants were recruited through social media platforms, which represents a major limitation for the representativeness of the general population. For the Raj et al. study [40], clear information on participant recruitment is not presented. The Daley [25] study was designed to ascertain college students’ knowledge about risk factors. Recruitment was performed in one university: The author distributed surveys to undergraduate students at a large public university in the Northeastern United States. Only students present on the day that the surveys were handed out were surveyed; as such, absentees were not given another chance to fill them out. Seventeen studies were classified into the “medium-quality” group [19,22,23,24,26,27,31,32,35,36,39,41,42,45,46,48,49] because (1) exposure was not accurately measured to minimize bias (see answers to Q3); (2) the pilot test was not used to establish the reliability and validity of the questionnaire for new questions (see answers to Q4); (3) the most important confounding factors were not identified and/or taken into account in the design and/or analysis (see answers to Q5a and Q5b); (4) the studies had no precise results (see answers to Q8); (5) the design and methods of the study were sufficiently flawed to make the results unreliable (see answers to Q9); (6) the results cannot be applied to the local population (see answers to Q10). Finally, ten cohort studies were classified into the “strong-quality” group [20,21,28,29,30,33,34,37,38,44].

3.3. Objective(s) of Study about Cancer Risk Factors?

In the public health research literature, the terms “knowledge”, “awareness”, “belief”, “representation”, “public perception”, or “attitude” are sometimes juxtaposed. In this systematic review, twelve articles focused on “awareness” [20,26,28,31,33,40,41,43,44,46,47,49], eight on “knowledge” [21,22,25,27,30,36,37,39], six on “perception” [19,23,34,35,38,42], one on “knowledge and belief” [24], one on “awareness and belief” [29], two on “knowledge and attitude” [32,45] and one on “perception and belief” [48] (see Appendix B).

The majority of studies focused on several cancer risk factors [20,22,25,26,27,28,30,31,32,33,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49]. Six studies focused on specific cancer risk factors: dietary and environmental factors [19,34], overweight and obesity-related cancer risk factors [23,34], alcohol factors [21,24] and age factors [29].

3.4. Instrument Administration

Different methods were used to administer questionnaires: interview (n = 12) [20,22,27,28,31,32,36,39,40,41,44,49]; computer-assisted telephone interview (CATI) (n = 9) [29,30,33,34,35,38,42,46,48]; online survey (n = 3) [21,43,47]; self-reported (n = 2) [37,45]; telephone surveys [23,24]; sent questionnaire (n = 1) [19]; distributed questionnaire for one survey [25] (description in Appendix B). The data are missing for one study [26].

3.5. How Are Attitudes to Risk Factors Assessed?

Some studies used open-ended questions [27,30,39,42,42,44]. Close-ended questions were used in many studies but differed according to their pre-defined responses. Some articles used a list of cancer risk factors [19,22,25,41,43,49]. Others used Likert scale multiple-choice responses about the importance of risk factors (for example, “probably not very important” to “very important” or “very likely” to “very unlikely”) [19,23,34,41]; the extent to which factors increase or decrease cancer risk [19,35]; and the degree of agreement [24,29,33,38,43,46,48]. Some studies proposed dichotomous questions (yes/no; true/false; risk factor/preventive factor) [20,21,26,32,36,43]. For one study, the responses were the attributable proportion of cancer causes [31]. For one dichotomous question (yes/no), participants who answered “yes” were then asked an open-ended question [23]. For five studies, data about how attitudes to risk factors are assessed are missing [28,37,39,40,45]. Three studies used both open-ended and closed-ended questions [23,27,39]. A complete description of studies’ attitudes’ assessment methods is presented in Appendix B. Four studies used validated standardized measures: the Awareness and Beliefs about Cancer (ABC) questionnaire [33,46,47] and the Cancer Awareness Measure (CAM) questionnaire [27].

The ABC questionnaire is a reliable, validated instrument for measuring knowledge and beliefs about cancer (see

Table 3. Items in CAM and ABC questionnaires.

Topic	CAM	ABC
Smoking	Smoking any cigarettes at all Exposure to another person’s cigarette smoke	Smoking? Exposure to another person’s smoke?
Alcohol	Drinking more than one unit of alcohol a day	Drinking more than one unit of alcohol a day. One unit of alcohol is equivalent to a single measure of spirits, a third of a pint of normal-strength lager or beer, or a small glass of wine.
Diet	Eating less than five portions of fruit and vegetables a day Eating red or processed meat once a day or more	Eating less than five portions of fruit and vegetables a day. Eating red or processed meat once a day or more. By processed meat, I mean meat that is smoked, salted or chemically preserved.
Obesity/overweight	Being overweight (BMI over 25)	Being obese.
UV	Getting sunburnt more than once as a child	Getting sunburnt more than once as a child.
		Using a sunbed.
Age	Being over 70 years old	Being over 70 years old.
?	Having a close relative with cancer	Having a close relative with cancer.
Infections	Infection with HPV (Human Papillomavirus)	Infection with HPV, Human Papillomavirus.
Physical activity	Doing less than 30 min of moderate physical activity five times a week	Low physical activity.
Ionizing radiation		Exposure to radiation such as radioactive materials, X-rays or radon.

) [50]. The core measure includes 32 ‘core’ items (open-ended and closed-ended questions): (1) awareness of cancer symptoms, (2) awareness of cancer outcomes, (3) help-seeking intentions, (4) beliefs about cancer, (5) beliefs about barriers to symptomatic presentation and (6) estimated age at which people are most likely to develop cancer. The optional modules are modules on cancer screening and awareness of risk factors for cancer [50]. The 13 risk factors for cancer are: smoking, exposure to another person’s smoke, drinking more than one unit of alcohol a day, eating less than five portions of fruit and vegetables a day, eating red/processed meat, obesity, sunburn in childhood, being over 70 years old, having a close relative with cancer, infection with human papillomavirus (HPV), low physical activity, using sunbeds, and exposure to ionizing radiation.

The CAM questionnaire is a validated standardized measurement for cancer awareness in the general population (see Table 3) [51]. The CAM consists of 47 items (open-ended and closed-ended questions): (1) warning signs; (2) seeking medical advice; (3) barriers to seeking medical advice; (4) risk factors; (5) cancer and age; (6) most common cancer; (7) awareness of NHS screening programs. One open-ended question has been designed to enquire about risk factors: “What things do you think affect a person’s chance of getting cancer?” and 11 closed-ended questions have been designed to measure a respondent’s level of agreement with the 11 risk factors: “These are some of the things that can increase a person’s chance of developing cancer. How much do you agree that each of these can increase a person’s chance of developing cancer?”. The 11 risk factors are: smoking any cigarettes at all, exposure to another person’s cigarette smoke, drinking more than one unit of alcohol a day, eating less than five portions of fruit and vegetables a day, eating red or processed meat once a day or more, being overweight (BMI over 25), getting sunburnt more than once as a child, having a close relative with cancer, being over 70 years old, infection with HPV, and performing less than 30 min of moderate physical activity five times a week.

The number of cancer risk factor reporting items used in the studies varied greatly, ranging from 1 to 128 items used [19,21,29]. Questions focused on: cancer-reducing strategies [23,24,30,36,41,42] and/or things that cause a person to develop cancer or increase their chances of developing cancer [19,20,20,22,23,24,25,26,27,28,29,31,32,33,34,35,37,38,39,40,41,42,43,44,45,46,47,48,49].

Most of these studies used items related to the main preventable cancer risk factors [1]: smoking [19,20,22,23,25,26,27,28,31,32,33,35,36,37,38,39,40,41,43,45,46,47,49], alcohol [19,20,21,24,25,26,28,31,32,33,35,36,37,38,39,40,41,43,45,46,47], diet [19,20,22,23,25,26,27,28,31,32,33,34,35,36,37,39,40,41,43,45,46,47,49], overweight and obesity [19,20,23,25,26,27,28,31,32,33,34,35,40,41,43,46,47,49], infections [19,25,28,32,33,37,40,41,46,47,49], occupational exposure [25,31], ultraviolet radiation [19,20,23,27,27,28,32,33,35,38,41,45,46], ionizing radiation [19,28,32,33,37,46], insufficient physical activity [19,23,26,27,31,32,33,34,35,36,38,41,43,46], hormones [19,25,28,41,49], breastfeeding [26,34,43], outside air pollution [19,31,35,38,45], and arsenic or benzene [32].

A large number of items were used to assess cancer risk attitudes among the general population with regard to other environmental and endogenous factors, along with specific health aspects. Most of these items were related to probable risk factors or assessed common beliefs about non-presumptive factors. Other environmental factors that were assessed were related to: diet, such as specific foods (33 items: such as fatty foods [23,26,35,45,49], salt [26,34,35,43], charred meat or fish [31,35], green tea [26,43], vitamins and minerals [35,43]) and processed food ingredients (additives and preservatives [19,28,31,35,49], artificial sweeteners [19,32], food coloring [26]); exposure to toxins, such as chemicals (14 different items such as: antiperspirant [25,32,41], chemical substances in general [19,25,28], pesticides [19,24,31], aerosols [19,43], and hair dyes [19,32]), pollution (industrial [38,40], environmental [39], and general pollution [49]), and radiation [28]; water treatments [19]; electromagnetic fields (high-voltage power lines [19,32,35,41,49], mobile phones [32,35], mobile phone relay stations [38], and microwave ovens [19]); other waves (loud music [35], computer screen [19], TV screen [19], fluorescent light [19]); medical procedures (breast implant [25], Pap smear [25], tooth filling [19], and vasectomy [25]) and medication (aspirin [19], cough medicine [19], sleeping pills [19]); animal-related causes [19]; and finally, bacteria and parasites [19,28,31].

Endogenous factors were assessed through three main item categories: age, expressed as aging [20,22,32,33,41,49], being 35 or 50 years old [45], and being 70 years old [27,28,29,46,47]; heredity, expressed as having close relatives with cancer [27,33,41,45,46,49], a family history of cancer [19,20,22,32,40,47]; and genetics [25,31,43].

Specific health aspect factors were also assessed and grouped into four categories: sexual and reproductive health/behaviors, such as having many/multiple sexual partners [19,22,24,25,40,41,49], having sex at a young age [25,40], early puberty [40], not using condoms [25] and poor sexual/genital health [40]; mental health, consisting of stress [19,28,31,35,38,41,49], good sleep hygiene [19,36], resentment caused by a personal or professional disappointment [38], general worries [19] or painful experiences [38]; women’s health, regarding late menopause [25,40], nulliparity [40], pregnancy [25], abortion [25], having a first child after 30 [25]: and finally, general lifestyle and behavioral responsibility [19,48].

Finally, 13 items could not be classified, such as: wearing underwire bras [25,43], wart or mole irritation [45], poverty [25] or luck [43].

4. Discussion

This systematic review aims to evaluate the methodological assessment used to evaluate attitudes about cancer risk factors among the general public.

The results of this systematic review show that different methods were used. The main method used to administer questionnaires is the interview. The advantage of interview-administered surveys is that respondents had the opportunity to seek clarification if they did not understand a question [28]. Online surveys [21,43,47] are also used but the findings may not be representative of people who do not access the Internet [21]. However, Connor et al.’s [52] study compares risk factors between data collected online and face-to-face. Comparisons of data collected using face-to-face interviews and online surveys revealed minor differences between samples [52].

In 2007, the NHS Cancer Reform Strategy emphasized the importance of raising awareness of early warning signs and risk factors of cancer among the general population. The CAM tool [51] was developed to help measure levels of cancer awareness, explore risk factors for poor cancer awareness, and develop and evaluate interventions to promote cancer awareness [53]. CAM data were first collected in 2008 and subsequently every two years up to 2014. In 2014, the CAM tool was modified to include additional questions. A trend analysis was conducted comparing data from 2008–2014. In 2017, data were collected online and face-to-face. Results were compared to gain an understanding of differences according to the data collection method. Based on these results, the CAM survey was moved online in 2019. Only the 2008 version of CAM has been validated. In this systematic review, of the 20 studies published in 2010 or after, only two [27,47] used the CAM questionnaire and only one used it to evaluate the attitude about cancer risk factors [27]. Elshami’s study used 8 items of the 11 items proposed by the CAM questionnaire. Items about drinking more than a unit of alcohol a day, eating red or processed meat once a day or more, and HPV infections are excluded.

Internationally, there are variations in cultural attitudes to cancer (e.g., public education about cancer, delivery of healthcare, etc.), and these may shape attitudes about cancer. Simon et al. [50] only found one validated measure of cancer awareness: CAM, which does not include items on beliefs or attitudes and has not been assessed internationally. An internationally valid measure of attitude about cancer is essential to take this research forward [50]. The ABC questionnaire was designed for this purpose. In this systematic review, three studies used the ABC questionnaire [29,33,46,47]. Item selection for the ABC tool was informed by theoretical frameworks outlining processes of patient delay, the English Department of Health’s National Awareness and Early Diagnosis pathway and existing surveys such as the CAM questionnaire [50]. The optional module on awareness of risk factors for cancer included 13 items: 11 from the CAM questionnaire and two further items (using a sunbed and exposure to ionizing radiation).

In this systematic review, of the 15 studies published in 2013 or after, only four [29,33,46,47] used the ABC questionnaire and only three used it to evaluate attitudes toward cancer risk factors. Only one used the 13 items of the original version [33]. Shi et al. [47] used 10 of the 13 items; the physical activity risk factor was omitted from their final questionnaire due to an error when transcribing the survey into the web survey tool. The items regarding using a sunbed and exposure to ionizing radiation were excluded. Thus, except for the physical activity items, the items used in Shi et al. [47] to measure attitudes about cancer risk factors are the 11 common items between the CAM questionnaire and the ABC questionnaire. Forbes et al. [29] used the ABC questionnaire to assess the recognition of cancer symptoms.

The ABC tool was designed to be administered by telephone interview in order to be practical for data collection across diverse geographic areas [33]. The CAM tool was designed to be administered as an interview, either face-to-face or over the telephone. The CAM tool may be used on the Internet, or as a ‘self-reported’ survey without supervision (e.g., by mail) but these options will provide lower-quality data [53]. Thus, the CAM questionnaire can be used in many study designs, whereas the ABC questionnaire seems to be used only for telephone interviews.

The ABC tool is a reliable and valid measure of cancer awareness and beliefs. Validated versions have been developed for six countries (UK, Australia, Canada, Sweden, Denmark and Norway) and in five languages (English, Canadian French, Swedish, Danish and Norwegian). The CAM questionnaire was developed for the UK population, although some authors (e.g., Elshami et al. [27]) translated the CAM questionnaire from English to another language.

Given that most focused on general cancer risk factors, a wide variety of items were observed among the studies. Assessing environmental/behavioral and endogenous cancer risk factors, these studies mainly addressed attitudes among the general population towards preventable risk factors. Smoking, alcohol, diet, and overweight/obesity were assessed by around two-thirds of the included studies. However, on account of the diversity of methodologies used, and the heterogeneity of these studies, it is not possible to compare their results. Although it is necessary to understand attitudes among the general population to the main cancer risk factors, some studies did not include all main risk factors: For instance, Elshami et al. [27] used the CAM survey but removed the item on alcohol even though it is the second most important preventable risk factor. In the studies retrieved in this review, there appears to be a lack of homogeneity in the items used to compare the assessment of attitudes on the main preventable cancer risk factors. Nevertheless, Shi et al. [54] used a 10-item version of the ABC questionnaire that ultimately resembles the CAM questionnaire. Furthermore, while they are not responsible for the most attributable fraction, some of the main preventable cancer risk factors were not sufficiently included in the surveys in the studies, such as infections, occupational exposures, and ionizing radiation. This list of established cancer risk factors is based on the current state of epidemiological knowledge and may be subject to change as exposure evaluation methods develop. Environmental synergies such as “cocktail effects” need to be better documented and cancer awareness reporting surveys should take these environmental questions into consideration. Some studies proposed unestablished and suspected environmental cancer risk factors. There is a need for better homogeneity and priority setting in terms of including suspected cancer environmental risk factors when measuring attitudes among the general population.

Few studies used open-ended questions. Closed-ended questions have a limited set of possible answers and questions are often good for surveys, because higher response rates are obtained. Prompted questions receive significantly higher recognition than open-ended questions that rely on recall [35], but closed-ended questionnaires prevent respondents from qualifying or justifying their responses [38]. Moreover, the closed-ended question format is also challenging for researchers because it may be assumed that they know more about a survey topic than they do. Closed-ended questions come in a multitude of forms. In this systematic review, we observed different formats such as the Likert scale multiple-choice question responses about the importance of risk factors (for example, “probably not very important” to “very important” or ‘to very likely’ to “very unlikely”); the extent to which factor increase or decrease cancer risk; and the degree of agreement. Some studies proposed dichotomous questions: Yes/No; True/False; Risk factor/Preventive factor. There are almost as many studies as there are response methods, which makes it difficult to compare the results because the information levels are not the same.

This systematic review highlights the diversity of methods used both in the method of administering the survey as well as the format of the questions and the response scales. It is thus difficult to compare data between different countries. However, cancer is a global problem. Harmonizing methods could make it possible to compare data between countries. This could also make it possible to build prevention campaigns based on international data. In accordance with the findings of this systematic review, we present a schematic representation of the steps researchers should follow to build a study to evaluate attitudes about cancer risk factors among the general public, along with recommendations:

(1)

Regarding the terms used, in studies included in this systematic review, to evaluate attitudes about cancer risk factors among the general public. Some studies used awareness (n = 12), others knowledge (n = 8) or perceptions (n = 6). Authors do not define the term used, so we do not know whether different terms are used to refer to the same concept. We recommend that authors define the concept used and what they want to evaluate. If the term is defined, it will be possible to make a comparison between studies.

(2)

Regarding “How questions about risk factors are determined”, the authors used previously standardized and validated tools (e.g., ABC, CAM), or the questions were based on prior cancer research, international literature and other cancer-related population surveys. We also recommend presenting the work teams in the article, including, in addition to researchers and health experts, representatives of patient organizations. To build a questionnaire, it is also important to know the concerns of the target population concerning risk factors and not just researchers’ concerns.

(3)

Regarding question format: We suggest that for some questions for which the findings need clarification, closed-ended and open-ended formats can be asked for the same question. For example, in the Cancer Barometer 2015, almost 10% of respondents were of the view that some cancers are contagious, but the closed-ended question used does not make it possible to elaborate on this result. In 2020, the same closed-ended question was asked, but for participants who responded “Agree”, an open-ended question was asked to find out why they think that some cancers are contagious. In this systematic review, Cameron et al. [23] used one dichotomous question (Yes/No), but participants who responded “Yes” were then asked an open-ended question [23]. Moreover, a combination of open-ended and closed-ended questions is often used to compare spontaneous responses and respondents’ choices when lists of responses are provided. For example, in Cancer Barometer 2020, firstly, an open-ended question is asked: “What do you think the three main causes of cancer are?”, followed by a closed-ended question with 17 risk factors (Likert scale for each item: Certainly, probably, probably not, certainly not, don’t know).

(4)

In this systematic review, few studies mentioned whether pilot testing was used or not. We recommend that pilot testing be used to establish the reliability and validity of the questionnaire for new questions.

(5)

Regarding the method used to administer the questionnaire: It is important that the method include all participants so that the sample is representative of the general population (e.g., people who do not access the Internet; illiterate people, etc.). For Connor et al. [52], there are minor differences between data collected using face-to-face interviews and online surveys [52]; combining methods is possible.

This review has certain limitations. Firstly, even though our search was extensive, we cannot be certain that all relevant articles were included. Secondly, studies published in sources other than peer-reviewed journals were not included. We can assume that some authors have presented their work in non-scientific journals. Thirdly, we only included studies with a representative cohort of a defined population. Indeed, some design studies are excluded such as qualitative studies. Fourthly, studies focused on specific cancer risk factors were not included. Fifthly, our reporting item review might not be exhaustive as studies may have included items not mentioned in the article in their questionnaire items.

5. Conclusions

This systematic review highlights the diversity of methods used: method of administering the survey, question format, and response scales. It is thus difficult to compare data between different countries. However, cancer is a global problem. Harmonizing methods could allow a comparison of data between countries. Moreover, to assess trends in attitudes among the general population as well as the effectiveness of communication campaigns around the prevention of cancer risk factors, longitudinal studies should be considered.