Prevalence and sociodemographic correlates of beliefs regarding cancer risks

Prevalence and Sociodemographic Correlates ofBeliefs Regarding Cancer Risks

Kevin Stein, PhD1

Luhua Zhao, MS1

Corinne Crammer, PhD1

Ted Gansler, MD, MBA2

1 Behavioral Research Center, American CancerSociety, Atlanta, Georgia.

2 Department of Health Promotion, AmericanCancer Society, Atlanta, Georgia.

BACKGROUND. Inaccurate beliefs about cancer risk may contribute to unhealthy

lifestyle behaviors and poor adherence to recommended screening and preven-

tion guidelines. To address this issue the current study assessed the prevalence

and sociodemographic correlates of scientifically unsubstantiated beliefs about

cancer risk in a representative sample of the US population.

METHODS. Nine hundred fifty-seven US adults with no history of cancer were

surveyed by telephone. The survey included 12 statements about cancer risk, risk

factors, and prevention that were framed to be contrary to the consensus of cur-

rent scientific evidence.

RESULTS. Participants were inconsistent in their ability to identify the statements

as false, and appraisal accuracy was associated with several sociodemographic

characteristics. Five of the 12 misconceptions were endorsed as true by at least a

quarter of the respondents, and uncertainty was higher than 15% for 7 state-

ments. At the same time, more than two-thirds of the participants were able to

identify 7 statements as false and, on average, respondents endorsed fewer than

3 statements as true. Respondents who were male, older, non-White, less edu-

cated, and of lower income were most likely to hold inaccurate beliefs.

CONCLUSIONS. A notable percentage of the participants in this study hold beliefs

about cancer risk at odds with the prevailing scientific evidence. Because the

population segments with the least accurate knowledge also bear the greatest

burden of cancer, areas for public education and intervention efforts are identi-

fied. Cancer 2007;110:1139–47. � 2007 American Cancer Society.

KEYWORDS: health knowledge, attitudes, practice, health education, patient edu-cation, neoplasms, attitude to health.

N umerous social theoretical frameworks have been advanced to

explain and predict health-related behaviors.1–4 Although such

health-behavior theories may differ with respect to their fundamen-

tal constructs, most include consideration of beliefs regarding one’s

susceptibility to disease, knowledge of potential risk factors, and

understanding of the associations between health-related behaviors

and health outcomes. Such theories hold that engaging in a healthy

lifestyle is partially predicated on an accurate assessment of risk fac-

tors and understanding the relation between such risks and one’s

behaviors. Applied to cancer, limited awareness of proven cancer

risk factors represents an obstacle to positive health outcomes.5–7

Without awareness of such risk factors, individuals may engage in

unhealthy lifestyles and may not adhere to recommended cancer

screening guidelines. Others may hold inaccurate beliefs about fac-

tors that have little or no scientific evidence supporting their rele-

vance to cancer risk. Previous research suggests that undue concern

regarding such factors may distract some individuals’ attention from

Address for reprints: Kevin Stein, PhD, AmericanCancer Society, Behavioral Research Center, 250Williams St. NW, Atlanta, GA 30303; Fax: (404)929-6832; E-mail: [email protected]

Supported by an Intramural funding of the Ameri-can Cancer Society as well as outside fundingfrom the Discovery Health Channel and Preven-tion Magazine.

The authors thank Eric Nehl, MS, and S. JaneHenley, MSPH, for advice regarding data analysisand Elizabeth Ward, PhD, for advice concerningdesign of the epidemiologist survey.

Received January 29, 2007; revision receivedApril 13, 2007; accepted April 24, 2007.

ª 2007 American Cancer SocietyDOI 10.1002/cncr.22880Published online 26 July 2007 in Wiley InterScience (www.interscience.wiley.com).

1139

documented risk factors and result in lifestyle beha-

vior decisions detrimental to their health.8,9 Thus,

educating individuals about factors that increase

cancer risk has been a goal of cancer prevention pro-

grams in the US.10 To address these issues, we

undertook a study to assess the accuracy of the pub-

lic’s beliefs regarding cancer risk, risk factors, and

prevention, and to identify the sociodemographic

factors associated with these beliefs.

MATERIALS AND METHODSIn late 2002 the American Cancer Society (ACS), Pre-

vention magazine (published by Rodale Press), Dis-

covery Health Channel, and Princeton Survey

Research Associates collaboratively designed and

conducted a survey of cancer-related beliefs that was

intended to further research in the area as well as

identify topics for public education projects and pro-

gramming. A previous publication reported results

concerning beliefs about cancer treatment and the

survey methods are described in detail in that

article.9

The survey instrument included standard demo-

graphic questions as well as statements about cancer

risk, screening, diagnosis, and treatment. The inten-

tion was to construct survey items that represented

common misconceptions about cancer. Thus, each

statement was framed to be contrary to the available

scientific evidence. Survey items were developed

with input from healthcare professionals and educa-

tors, including staff associated with the ACS National

Cancer Information Center, based on their experi-

ence with the public’s attitudes and beliefs about

cancer.

The current analyses focus on the 12 survey

items (Table 1) concerning overall cancer risk, speci-

fic risk factors for cancer, and cancer prevention.

Respondents could agree, disagree, or express uncer-

tainty about the truth of each statement by choosing

1 of 3 response options: ‘‘true,’’ ‘‘false,’’ or ‘‘don’t

know.’’ Although a comprehensive review of the rele-

vant epidemiological literature is beyond the scope

of this article, Table 1 includes citations to key refer-

ences relevant to each statement.

Recognizing that some survey items in this study

had not been the subject of conclusive scientific

study, the authors conducted a survey of 10 cancer

epidemiologists at the American Cancer Society to

assess their views regarding the evidence for each

statement. To characterize their views more precisely,

the response options for the epidemiologists differed

from those offered to the study participants. The epi-

demiologists’ response categories were ‘‘true,’’ ‘‘true

only in rare circumstances,’’ ‘‘difficult to evaluate

because of absent or limited data but likely to be

true,’’ ‘‘difficult to evaluate because of absent or lim-

ited data but likely to be false,’’ ‘‘false,’’ and ‘‘I’m not

familiar enough with this topic to answer this ques-

tion.’’ After each epidemiologist returned the survey

follow-up interviews explored their responses.

The sample of US adults was drawn using stand-

ard list-assisted random digit dialing method. Of the

3338 working telephone numbers (excluding fax,

business, and nonworking phones) called, 2497 were

successfully reached (contact rate 5 75%), and 1254

individuals agreed to participate (cooperation

rate 5 50%). Of the 1254 respondents, 1070 were

eligible (adults without language barriers), and 1002

completed the telephone survey (completion

rate 5 94%). The overall response rate of 35% was

calculated as the product of the contact rate (75%),

the cooperation rate (50%), and the survey comple-

tion rate (94%). Of the 1002 respondents, 45 reported

a previous cancer diagnosis. Because this study

focused on cancer risk beliefs among the general

population, individuals with a personal history of

cancer were removed from the analyses, leaving a

sample of 957 US residents with no history of cancer.

The sample was then weighted to match national

parameters (obtained from the March 2001 Current

Population Survey) for sex, age, education, race, His-

panic origin, and census region. Because of the com-

plex sampling design, we used a design effect

(51.10) to adjust the standard error for tests of statis-

tical significance.

To create an indicator of overall health literacy

regarding cancer risk, we calculated a combined

endorsement score by summing responses for all 12

statements. Each false response was assigned a zero

value, and each true response was assigned a value

of 1; don’t know and refused responses were consid-

ered missing and not included in the calculation of

this score. Thus, each participant’s score is an integer

that could vary from 0 (responded false to all 12

statements) to 12 (responded true to all 12 state-

ments), with higher scores corresponding to lower

cancer risk health literacy. Mean values were calcu-

lated for participants in various sociodemographic

categories (gender, race, education, etc), and multi-

variate linear regressions examined associations

between the combined endorsement score and

sociodemographic characteristics. Multivariate logis-

tic regressions examined associations between

responses to each of the 12 statements and sociode-

mographic variables. Confidence intervals were

adjusted by the design effect. As with the computa-

tion of the combined endorsement score, don’t know

1140 CANCER September 1, 2007 / Volume 110 / Number 5

and refused responses were considered missing and

not included in these analyses.

RESULTSSample CharacteristicsAs presented in Table 2, 54.1% of the weighted sam-

ple consisted of individuals between 35 and 64 years

old, 72.6% were non-Hispanic White, 51.4% were

female, 86.3% had at least a high school education,

and 56.2% reported a positive family history of can-

cer. The vast majority of participants said they were

very or somewhat informed about cancer.

Endorsement of Cancer Risk StatementsThe average combined endorsement score for the

sample overall and by each sociodemographic vari-

able is presented in Table 2. Multivariate linear

regression analyses indicated that the combined

endorsement score (mean of 2.74 for all participants

combined) differed by several sociodemographic

characteristics. Characteristics associated with lower

health literacy (higher combined endorsement

scores) included male gender, non-White race, His-

panic ethnicity, income less than $30,000, and less

than a high school education. Paradoxically, com-

pared with respondents who considered themselves

‘‘very informed’’ about cancer, respondents rating

themselves as ‘‘somewhat informed’’ or ‘‘not very

informed’’ had lower combined endorsement scores

(Table 2).

The associations between sociodemographic char-

acteristics and each cancer risk statement are dis-

played in Tables 3 through 5. In general, the pattern

was consistent across statements and in line with

associations noted between sociodemographic vari-

ables and the combined endorsement score, although

some variability across statements was observed. For

example, although lower income was generally asso-

ciated with greater likelihood of endorsing statements

as true, an exception was found for the belief that

underwire bras cause cancer. Similarly, those with a

college education were more likely than high school

graduates to believe that low tar cigarettes are safer,

although lower education was generally associated

with higher endorsement of other statements.

Appraisal of each cancer risk statement for study

participants is presented in Table 1. Five of the 12

statements were endorsed as true by more than 20%

of the participants, with the most frequently

endorsed misconception being ‘‘The risk of dying

from cancer in the United States is increasing,’’

TABLE 1Distribution of Responses to Statements About Cancer Risk: Cancer Myths and Cancer Facts Survey 2002

Statements

% General public No. of cancer epidemiologists

True

Not

true

Don’t

know True

True only

in rare

circumstances

Difficult

to evaluate;

likely true

Difficult

to evaluate;

likely false False

Not

familiar with

this topic

1. The risk of dying from cancer in the

United States is increasing.11–1367.7 22.5 9.8 0 0 0 0 10 0

2. Living in a polluted city is a greater risk for lung

cancer than smoking a pack of cigarettes a day.17,1838.7 42.5 18.8 0 0 0 0 10 0

3. Some injuries can cause cancer later in life.32,33 37.2 41.9 20.9 0 1 0 0 5 4

4. Electronic devices, like cell phones can cause

cancer in the people who use them.3429.7 45.7 24.7 0 0 0 8 1 1

5. What someone does as a young adult has little effect

on their chance of getting cancer later in life.3524.8 68.2 7.1 0 0 0 1 9 0

6. Long-time smokers cannot reduce their

cancer risk by quitting smoking.1716.2 78.1 5.7 0 0 0 0 10 0

7. People who smoke low-tar cigarettes have less chance of

developing lung cancer than people

who smoke regular cigarettes.18

14.7 74.5 10.8 0 0 0 1 8 1

8. Personal hygiene products, like shampoo, deodorant,

and antiperspirants, can cause cancer.3613.7 71.0 15.3 0 0 0 5 4 1

9. Getting a mammogram, or using a special X-ray machine

to detect breast cancer, can cause cancer of the breast.26,3710.2 73.8 16.1 0 2 0 1 5 2

10. Getting a base tan or base coat at a tanning salon will provide

protection from skin cancer when you go outside in the sun.388.4 78.4 13.2 0 0 0 0 7 3

11. Underwire bras can cause breast cancer.39 6.2 62.9 30.9 0 0 0 1 9 0

12. You cannot get skin cancer from using a tanning booth.38 6.2 75.5 18.3 0 0 0 3 6 1

Cancer Risk Beliefs/Stein et al. 1141

which was believed by almost 68% of the sample.

Four statements were endorsed as true by 10% to

20% of the sample, and 3 statements were endorsed

as true by less than 10% of the participants. With

respect to refuting the statements, over two-thirds of

the sample correctly identified 7 statements as false,

and each of the remaining 5 statements was judged

to be incorrect by at least 20% of the participants.

Uncertainty was relatively high for several of the

statements, with more than 15% of the sample

responding don’t know for seven statements, and

20% choosing this same option for 3 statements.

Uncertainty was lower than 10% for 3 statements.

In addition to showing the appraisal of each

statement by the study participants, Table 1 also dis-

plays the tabulation of responses to our survey of 10

cancer epidemiologists. Examination of these data

shows that none of the 12 statements were endorsed

as true by any of the epidemiologists. For the most

part, the epidemiologists identified the statements as

false or as likely to be false (but difficult to evaluate

because of absent or limited data), although the

TABLE 2Association Between Sociodemographic Characteristics and Combined Endorsement Score

Characteristic Count %*

Mean endorsement

scorey

Multivariate model design effect: 1.10

Beta

Adjusted

std err t test P{

Total 957 2.74

Sex

Women 483 51.5 2.50 Referent Referent

Men 474 48.5 3.00 0.6313 0.1142 5.53 <.001

Race

White 727 72.6 2.53 Referent Referent

Black 93 10.9 3.12 0.4603 0.1910 2.41 <.05

Hispanic 85 10.7 3.53 0.9233 0.1918 4.81 <.001

Other 39 4.5 3.57 0.9343 0.2779 3.36 <.01

Age

<35 295 30.9 2.81 Referent Referent

35–64 520 54.1 2.65 20.0380 0.1303 20.29

651 133 14.2 2.97 0.0744 0.1922 0.39

Education

<High school 80 13.4 3.73 0.6321 0.1823 3.47 <.001

High school graduate 359 37.0 2.96 Referent Referent

Some college 204 24.2 2.46 20.4468 0.1494 22.99 <.01

College graduate 311 25.2 2.16 20.6624 0.1540 24.30 <.001

Income

<$30,000 259 29.7 3.19 0.3840 0.1714 2.24 <.05

$30–50,000 173 18.3 2.56 Referent Referent

$50–100,000 265 26.6 2.42 20.0053 0.1716 20.03

$100,0001 102 9.4 2.31 20.0322 0.2313 20.13

Family history of cancer

Yes 545 56.2 2.59 Referent Referent

No 403 42.8 2.93 0.1800 0.1170 1.53

Cancer knowledge

Very informed 334 33.8 2.89 Referent Referent

Somewhat informed 509 53.1 2.63 20.3045 0.1240 22.45 <.05

Not very informed 91 10.5 2.79 20.4754 0.2014 22.36 <.05

Not at all informed 15 1.6 3.36 20.0147 0.4791 20.03

Census region

South 329 35.8 2.94 Referent Referent

Northeast 173 19.7 2.69 20.0498 0.1619 20.31

Midwest 256 23.2 2.53 20.1963 0.1532 21.28

West 199 21.3 2.69 20.1239 0.1577 20.79

* Percentages may not sum to 100 because of nondisplayed unknown/refused category.y Unknown/refused responses to each statement were excluded in the analyses.{ P values adjusted for design effect.


‘‘true, but only in rare circumstances’’ response was

used 3 times. For 7 of the statements 1 or more of

the epidemiologists reported that they were not fa-

miliar with this topic.

DISCUSSIONThis study assessed the prevalence of belief in or

uncertainty regarding several scientifically unsub-

stantiated statements about cancer risk in a sample

of 975 US adults with no personal history of cancer.

Overall, the results indicate that the participants in

this study were able to identify several incorrect

statements about cancer risk as being false. Specifi-

cally, the overall endorsement score across all 12

statements—the mean of the individual scores of all

respondents—was less than 3. This suggests that, on

average, participants in this study believed less than

TABLE 3Association Between Sociodemographic Characteristics and Beliefs About Cancer Risk

Statement 1*,y Statement 2*,y Statement 3*,y Statement 4*,y

Risk of dying increasing

City pollution worse than

smoking Injuries cause cancer

Electronic devices cause

cancer

%{ OR (CI)§ %{ OR (CI)§ %{ OR (CI)§ %{ OR (CI)§

Sex

Women 75.6 1.00 (referent) 47.4 1.00 (referent) 41.3 1.00 (referent) 35.7 1.00 (referent)

Men 74.3 1.03 (0.80–1.32) 47.9 1.16 (0.92–1.46) 52.8 1.85 (1.46–2.34) 42.8 1.37 (1.07–1.75)

Race

White 73.0 1.00 (referent) 46.1 1.00 (referent) 44.1 1.00 (referent) 36.3 1.00 (referent)

African American 79.3 1.01 (0.66–1.56) 47.7 0.96 (0.66–1.40) 54.2 1.74 (1.19–2.54) 43.8 1.23 (0.82–1.85)

Hispanic 84.0 1.52 (0.94–2.44) 51.4 1.21 (0.82–1.78) 63.6 2.62 (1.75–3.93) 53.2 1.73 (1.17–2.57)

Other 76.4 1.06 (0.57–1.96) 60.5 1.74 (0.97–3.12) 36.7 0.85 (0.47–1.56) 43.6 1.30 (0.71–2.36)

Age

<35 79.7 1.00 (referent) 41.0 1.00 (referent) 42.6 1.00 (referent) 45.0 1.00 (referent)

35–64 72.6 0.74 (0.55–1.00) 49.3 1.42 (1.09–1.85) 47.2 1.44 (1.10–1.89) 39.7 0.87 (0.66–1.13)

651 73.0 0.50 (0.32–0.77) 56.1 1.52 (1.03–2.23) 56.8 2.15 (1.44–3.21) 21.9 0.36 (0.23–0.58)

Education

<High school 85.0 1.34 (0.84–2.13) 67.1 1.99 (1.37–2.90) 58.0 1.29 (0.89–1.86) 48.4 1.47 (0.98–2.22)

High school graduate 81.1 1.00 (referent) 50.6 1.00 (referent) 47.0 1.00 (referent) 41.3 1.00 (referent)

Some college 70.2 0.51 (0.37–0.71) 41.6 0.73 (0.54–0.99) 47.8 1.08 (0.79–1.46) 38.4 0.84 (0.61–1.15)

College graduate 64.3 0.51 (0.37–0.72) 38.3 0.66 (0.48–0.90) 39.5 0.78 (0.57–1.08) 32.9 0.66 (0.47–0.91)

Income

<$30,000 82.8 1.12 (0.80–1.58) 56.4 1.31 (0.93–1.85) 52.5 1.45 (1.02–2.06) 39.8 1.02 (0.71–1.47)

$30–49,999 76.6 1.00 (referent) 45.5 1.00 (referent) 40.9 1.00 (referent) 38.6 1.00 (referent)

$50–99,999 69.3 0.74 (0.51–1.07) 46.1 1.03(0.72–1.46) 43.5 1.37 (0.96–1.96) 43.4 1.32 (0.92–1.90)

�$100,000 57.9 0.48 (0.30–0.77) 33.1 0.61 (0.37–0.98) 44.0 1.41 (0.88–2.26) 29.6 0.78 (0.47–1.28)

Family history

Yes 74.9 1.00 (referent) 44.6 1.00 (referent) 48.3 1.00 (referent) 38.8 1.00 (referent)

No 74.9 0.99 (0.76–1.28) 51.3 1.37 (1.08–1.73) 45.0 0.75 (0.59–0.96) 40.5 1.04 (0.81–1.34)

Cancer knowledge

Very informed 74.3 1.00 (referent) 53.8 1.00 (referent) 47.8 1.00 (referent) 35.2 1.00 (referent)

Somewhat informed 75.9 1.11 (0.85–1.47) 46.1 0.73 (0.57–0.94) 44.4 0.90 (0.70–1.16) 41.6 1.31 (1.00–1.72)

Not very informed 73.6 0.70 (0.44–1.10) 37.6 0.45 (0.30–0.68) 53.4 1.20 (0.81–1.79) 44.0 1.14 (0.75–1.74)

Not at all informed 77.7 1.02 (0.34–3.05) 29.6 0.30 (0.11–0.85) 57.4 1.27 (0.49–3.30) 32.2 0.84 (0.75–1.74)

Census region

South 81.3 1.00 (referent) 50.8 1.00 (referent) 50.9 1.00 (referent) 42.0 1.00 (referent)

Northeast 69.6 0.59 (0.41–0.85) 49.8 1.06 (0.76–1.47) 44.5 0.84 (0.60–1.16) 42.9 1.10 (0.78–1.54)

Midwest 70.1 0.56 (0.40–0.79) 44.1 0.86 (0.63–1.17) 42.8 0.79 (0.57–1.08) 34.6 0.77 (0.56–1.07)

West 74.2 0.76 (0.53–1.09) 44.3 0.90 (0.65–1.24) 47.1 0.88 (0.63–1.21) 37.6 0.82 (0.58–1.15)

OR indicates odds ratio; CI, confidence interval.

* Statements are fully defined in Table 1.y Unknown and/or refused responses to each statement were excluded in the analyses.{ Percentage of participants who endorsed the statement as true.§ Odds ratio for multivariate analysis adjusted for sociodemographic characteristics; 95% confidence intervals are adjusted for design effect.


a quarter of these cancer risk statements to be true,

which may be encouraging for health educators.

Indeed, endorsement rates for most statements were

relatively low, providing further evidence that the

general public is able to correctly refute many erro-

neous cancer risk beliefs. At the same time, however,

endorsement rates for several statements were rela-

tively high, and belief in some of these statements

could adversely affect health behaviors. For example,

almost 4 in 10 believed that ‘‘living in a polluted city

is a greater risk for lung cancer than smoking a pack

of cigarettes a day.’’ This belief could affect smoking

behaviors by reducing concern about the risk asso-

ciated with tobacco. Statements such as this, which

have higher levels of endorsement and relate directly

to unhealthy behaviors, should be addressed by pub-

lic health education. In addition, several sociodemo-

graphic characteristics (gender, race, education, and

TABLE 4Association Between Sociodemographic Characteristics and Beliefs About Cancer Risk


Young adult behaviors

Smokers can’t reduce

their risk

Low tar cigarettes

are safer

Hygiene products cause

cancer


Sex


Men 28.6 1.47 (1.14–1.90) 17.3 1.13 (0.85–1.50) 19.5 1.65 (1.23–2.23) 21.3 2.02 (1.48–2.77)

Race



Hispanic 37.4 2.46 (1.65–3.68) 23.7 1.81 (1.17–2.79) 23.1 1.91 (1.23–2.98) 19.4 1.30 (0.81–2.10)

Other 51.9 5.33 (2.91–9.79) 35.6 3.94 (2.18–7.10) 20.9 1.47 (0.74–2.91) 34.2 2.07 (1.08–3.96)

Age, y


35–64 26.8 1.69 (1.26–2.28) 15.5 0.97 (0.70–1.34) 17.1 1.56 (1.09–2.22) 14.6 0.65 (0.46–0.92)

651 40.1 2.87 (1.89–4.33) 25.0 1.54 (0.98–2.42) 24.9 2.47 (1.52–4.02) 15.5 0.59 (0.35–0.99)

Education

<High school 49.0 2.04 (1.43–2.93) 25.6 1.20 (0.79–1.81) 25.3 1.64 (1.06–2.54) 24.3 2.04 (1.29–3.24)


Some college 23.1 0.77 (0.56–1.07) 15.6 0.82 (0.57–1.18) 12.2 0.95 (0.62–1.44) 12.6 0.72 (0.47–1.10)


Income

<$30,000 33.6 1.14 (0.79–1.63) 22.9 1.23 (0.82–1.85) 18.3 1.34 (0.85–2.12) 13.6 0.92 (0.56–1.49)


$50–99,999 22.7 1.03 (0.71–1.50) 12.7 0.79 (0.50–1.23) 14.4 0.98 (0.61–1.56) 15.3 1.17 (0.73–1.89)

�$100,000 19.0 0.86 (0.51–1.47) 6.8 0.43 (0.21–0.91) 16.8 1.03 (0.56–1.86) 19.8 1.46 (0.79–2.69)

Family history


No 29.1 1.05 (0.81–1.35) 21.3 1.47 (1.10–1.97) 20.5 1.59 (1.17–2.15) 18.8 1.36 (0.99–1.87)

Cancer knowledge




Not at all informed 43.4 1.45 (0.55–3.80) 20.3 0.87 (0.28–2.74) 14.6 0.74 (0.19–2.79) 25.8 1.00 (0.32–3.11)

Census region


Northeast 26.2 0.93 (0.66–1.32) 16.5 1.18 (0.78–1.79) 18.7 1.13 (0.75–1.69) 14.6 0.89 (0.57–1.39)

Midwest 23.4 0.83 (0.59–1.16) 19.7 1.48 (1.02–2.15) 19.7 0.89 (0.59–1.34) 14.2 1.11 (0.74–1.68)

West 23.0 0.71 (0.50–1.01) 16.0 1.00 (0.67–1.50) 15.8 0.90 (0.60–1.36) 18.1 1.26 (0.83–1.93)


* Statements are fully defined in Table 1.y Unknown/refused response to each statement were excluded in the analyses.{ Percent of participants who endorsed the statement as true.§ Odds ratio for multivariate analysis adjusted for sociodemographic characteristics; 95% confidence intervals are adjusted for design effect.


income) were associated with greater likelihood of

endorsement (both in terms of the overall mean

endorsement score and across individual items), sug-

gesting avenues for developing targeted messages for

particular at-risk populations.

Notably, the 2 most commonly endorsed state-

ments were also among the statements that were

unanimously identified as false by all 10 epidemiolo-

gists. Specifically, 68% of the respondents believed

that the risk of dying from cancer in the US is

increasing. This statement is clearly false, as the age-

standardized cancer death rate has been decreasing

since the early 1990s,11,12 and the 5-year relative sur-

vival rate for all cancers combined has improved

steadily over the last 30 years.13 Although these

trends clearly reflect the decline in the cancer death

TABLE 5Association Between Sociodemographic Characteristics and Statements About Cancer Risk


Mammograms cause

cancer Base tans are protective

Underwire bras cause

cancer

No cancer from tanning

booths


Sex


Men 17.4 3.32 (2.27–4.86) 11.9 2.08 (1.40–3.08) 12.9 2.14 (1.32–3.46) 7.8 1.18 (0.76–1.83)

Race



Hispanic 16.7 1.37 (0.79–2.38) 15.4 2.47 (1.40–4.35) 5.7 0.96 (0.38–2.43) 17.4 3.23 (1.77–5.92)

Other 15.8 1.12 (0.49–2.60) 20.3 2.95 (1.39–6.25) 38.5 10.51 (4.69–23.6) 9.3 1.62 (0.56–4.66)

Age


35–64 10.3 0.69 (0.46–1.02) 7.3 0.80 (0.51–1.26) 5.8 0.34 (0.20–0.59) 6.8 0.96 (0.59–1.56)

651 11.8 0.61 (0.34–1.10) 17.9 1.78 (1.00–3.18) 12.9 0.89 (0.43–1.86) 10.7 1.27 (0.63–2.56)

Education

<High school 18.2 1.11 (0.67–1.86) 15.3 0.86 (0.50–1.48) 15.5 0.93 (0.48–1.77) 17.8 2.54 (1.43–4.52)


Some college 9.7 0.62 (0.38–1.00) 5.6 0.48 (0.27–0.84) 4.3 0.23 (0.11–0.49) 8.3 1.17 (0.68–2.03)


Income

<$30,000 18.3 2.93 (1.67–5.14) 13.9 3.06 (1.58–5.92) 8.3 0.93 (0.46–1.88) 11.6 2.12 (1.08–4.18)


$50–99,999 5.5 1.07 (0.54–2.12) 4.5 1.11 (0.51–2.41) 4.7 1.04 (0.46–2.32) 3.4 0.82 (0.36–1.85)

�$100,000 13.7 2.63 (1.26–5.49) 5.8 1.40 (0.53–3.67) 16.4 6.29 (2.58–15.3) 5.6 1.84 (0.68–4.95)

Family history of cancer


No 14.8 1.42 (0.98–2.04) 12.9 1.44 (0.97–2.14) 10.5 0.88 (0.54–1.45) 9.0 1.25 (0.80–1.94)

Cancer knowledge




Not at all informed 35.9 2.38 (0.86–6.62) 20.6 1.31 (0.40–4.34) 31.5 1.81 (0.52–6.25) NAk NAk

Census region


Northeast 8.3 0.62 (0.35–1.10) 8.5 0.79 (0.59–2.07) 10.0 1.13 (0.59–2.07) 7.6 1.53 (0.82–2.85)

Midwest 17.0 0.59 (0.33–0.92) 8.3 0.62 (0.36–1.08) 7.7 0.85 (0.40–1.45) 7.5 1.44 (0.80–2.61)

West 15.5 1.23 (0.80–1.95) 8.5 0.63 (0.37–1.09) 8.2 0.62 (0.26–1.08) 8.1 1.35 (0.72–2.50)


* Statements are fully defined in Table 1.y Unknown/refused response to each statement were excluded in the analyses.{ Percent of participants who endorsed the statement as true.§ Odds ratio for multivariate analysis adjusted for sociodemographic characteristics; 95% confidence intervals are adjusted for design effect.k Not estimable because of the low number of endorsements.


rate, at the time of this survey the actual number of

deaths from cancer had been steadily increasing over

several decades because of the growth and aging of

the US population.11 Thus, some respondents might

have based their answer on the absolute number of

cancer deaths, as opposed to either relative survival

rates or age-standardized death rates. The second

most commonly endorsed statement—that living in a

polluted city is a greater risk for lung cancer than

smoking a pack of cigarettes a day—was believed by

39% of the respondents, with an additional 19%

uncertain. Despite frequent public education mes-

sages about risk associated with tobacco use, media

coverage of carcinogenic environmental exposures

such as pollution may impact the public’s attitude

about relative risks associated with these factors.

Another possible explanation for this belief may lie

with personal responsibility for the exposure.

Although smoking behavior was not examined in the

present study, our results are consistent with studies

demonstrating that individuals who engage in beha-

viors like smoking or unprotected sun exposure tend

to underestimate their own health risks associated

with these choices despite knowledge of the risk in

general, a phenomenon described as self-exempting

optimistic bias.14–16

Two other misconceptions about smoking, each

endorsed by about 15% of the respondents, are that

there is little reduction in cancer risk when long-

term smokers quit, and that low tar cigarettes are

less dangerous than regular cigarettes. On the con-

trary, studies have found that quitting smoking is

indeed associated with a reduced risk of lung cancer,

even among long-term smokers,17 and research has

not demonstrated any significant difference in cancer

risk between smokers of regular and low-tar cigar-

ettes.18 These misconceptions could have an obvious

impact on behavioral choices, with significant health

implications. Future public education and smoking

cessation programs should include attempts to dispel

such inaccurate beliefs. Although such misconcep-

tions are of obvious concern, the remaining 6 mis-

conceptions in this study were endorsed as true by

less than 15% of the sample, and of these 6, 3 were

endorsed by less than 10% of the respondents. This

finding offers the encouraging suggestion that belief

in these misconceptions is relatively low and may

not require significant public education efforts.

In addition to noting substantial belief in several

statements for the sample as a whole, we also found

associations between certain sociodemographic vari-

ables and the combined endorsement score. When

examining the endorsement of each statement indi-

vidually, a consistent finding was that males were

more likely to believe the statements to be true than

were females, as noted for 8 misconceptions. Some

research indicates that males may be less attentive to

and less likely to seek medical information than are

females19,20 and thus may be less well informed, as

suggested by these data. Lower educational levels

were also significantly associated with higher com-

bined endorsement scores, which is consistent with

most prior studies of health literacy and with endor-

sement of 10 of the 12 statements.21–23 These find-

ings suggest that public education and community

intervention projects may be most effective and effi-

cient if targeted to the groups with the most miscon-

ceptions about cancer risk.

One surprising result was that those claiming to

be ‘‘very informed’’ about cancer were significantly

more likely to endorse 4 of the statements, compared

with those rating themselves as having lower cancer

literacy. This finding is consistent with previous

research demonstrating that people tend to overrate

their own abilities and reach judgments with too much

confidence.16 The practical implication for health edu-

cation is that individuals who feel confident about

their cancer knowledge may not ask relevant questions

of their healthcare providers, and may therefore miss

opportunities to obtain accurate information.

Although these data raise intriguing questions

about the public’s accuracy in judging cancer risk

associated with certain factors, several limitations of

this research should be noted. First, the degree to

which these data can be considered representative of

the US population is somewhat limited by the 35%

response rate. However, the use of weights increased

the concordance with national sociodemographic

characteristics and a response rate in this range is

consistent with other national studies from which

conclusions about health behaviors are drawn.24

A second limitation relates to the lack of strong

scientific evidence regarding some of the statements

in this study. In support of including such state-

ments, we assert that many important health deci-

sions regarding risks and prevention are made by the

general public in a context of limited or inconsistent

evidence. In fact, research suggests that perceived

ambiguity regarding cancer risk factors can increase

worry about cancer and reduce perceived prevent-

ability.25 Although the 10 epidemiologists were not

unanimous in conclusively rejecting all of the state-

ments, none endorsed any of the statements as true.

Furthermore, 10 of the 12 statements were rated as

definitely false or likely to be false (but difficult to

evaluate given current scientific knowledge) by the

epidemiologists. The variation among the epidemiol-

ogists’ responses, as assessed by qualitative follow-up


interviews, concerned rare exceptions, the quality

and quantity of available evidence, and the consis-

tency of evidence. Because all 10 epidemiologists

work together closely, their responses may not repre-

sent the full range of opinion within the epidemiolo-

gic community. It should be acknowledged that some

evidence indicates that mammograms may actually

result in a slight increase in the risk of developing

breast cancer.26 Most experts, however, believe that

the slight risk associated with mammography is

more than offset by the considerable reduction in

breast cancer mortality gained through early detec-

tion. Finally, the beliefs included in this survey were

not selected based on estimates of the mortality

attributed to related behaviors, but may still be use-

ful as a proxy for cancer-related health literacy.

In sum, the key conclusions from this study are

that beliefs in several scientifically unsubstantiated

cancer risk statements are relatively common among

the participants in this study, and that the prevalence

of such beliefs varies by certain sociodemographic

characteristics. Such beliefs may play a role in cancer

disparities or influence actual health-related beha-

viors and adherence to screening guidelines, because

previous research suggests that knowledge, attitudes,

and beliefs about health and risk factors for disease

contribute to the development and maintenance of

disparities in heath outcomes.27,28

Notwithstanding these results, it is important to

recognize that individual beliefs are frequently not

the most influential determinants of health behavior.

For example, 1 of the most powerful predictors of

cancer screening utilization is the healthcare provi-

der’s recommendation. Screening is also strongly

influenced by healthcare access issues, such as hav-

ing health insurance and a regular source of primary

care.29 Likewise, the contribution of individual

knowledge and beliefs to health disparities is influ-

enced and often substantially limited by the broader

socioeconomic context (eg, tobacco marketing tar-

geting minority youths and the limited availability of

healthful food choices and safe venues for physical

activity in certain neighborhoods).30,31 In addition,

the extent to which beliefs about risk factors influ-

ence actual health behaviors is an important topic

that should be addressed by future research. Public

education programs and interventions to address

and convincingly refute commonly held misconcep-

tions regarding cancer risks might increase the adop-

tion of healthy attitudes, beliefs, and, most

important, behaviors. Such educational and interven-

tion programs should be culturally informed and ac-

cessible to all individuals, with special attention

placed on reaching the highest risk populations.

REFERENCES1. Glanz K, Rimer BK. Theory at a Glance: A Guide for Health

Promotion Practice. Bethesda, Md: US Dept. of Health &

Human Services, Public Health Service, National Cancer

Institute; 1997.

2. Rosenstock IM, Strecher VJ, Becker MH. Social learning

theory and the Health Belief Model. Health Educ Q.

1988;15:175–183.

3. Bandura A. Social Foundations of Thought and Action: A

Social Cognitive Theory. Englewood Cliffs, NJ: Prentice-

Hall; 1986.

4. Prochaska JO, Redding CA, Evers KE.The transtheoretical

model and stages of change. In: Glanz K, Lewis FM, Rimer

BK, eds. Health Behavior and Health Education: Theory

and Practice. San Francisco, Calif: Jossey-Bass; 1997:60–84.

5. Boulware LE, Cooper LA, Ratner LE, LaVeist TA, Powe NR.

Race and trust in the health care system. Public Health

Rep. 2003;118:358–365.

6. Davis TC, Williams MV, Marin E, Parker RM, Glass J. Health

literacy and cancer communication. CA Cancer J Clin.

2002;52:134–149.

7. Nielsen-Bohlman L, Panzer AM, Kindig DA, eds, for the

Institute of Medicine Committee on Health Literacy. Health

Literacy. A Prescription to End Confusion. Washington, DC:

National Academies Press; 2004.

8. Breslow R, Sorkin J, Frey C, Kessler L. Americans’ knowl-

edge of cancer risk and survival. Prev Med. 1997;26:170–

177.

9. Gansler T, Henley SJ, Stein K, Nehl EJ, Smigal C, Slaughter

E. Sociodemographic determinants of cancer treatment

health literacy. Cancer. 2005;104:653–660.

10. US Dept. of Health & Human Services. Healthy People2010: Understanding and Improving Health. 2nd ed.Washington, DC: US Government Printing Office; 2000.

11. Jemal A, Siegel R, Ward E, et al. Cancer statistics, 2007. CACancer J Clin. 2007;57:43–66.

12. Ries LAG, Harkins D, Krapcho M, et al. SEER Cancer Statis-tics Review, 1975–2003. National Cancer Institute, Be-thesda, Md. Updated 2006. Available at URL: http://seer.cancer.gov/csr/1975_2003/ Accessed on Nov. 26, 2006.

13. American Cancer Society. Cancer facts and figures, 2006.

Atlanta, Ga: American Cancer Society; 2006.

14. Branstrom R, Kristjansson S, Ullen H. Risk perception, op-timistic bias, and readiness to change sun related beha-viour. Eur J Public Health. 2006;16:492–497.

15. Weinstein ND, Marcus SE, Moser RP. Smokers’ unrealistic

optimism about their risk. Tob Control. 2005;14:55–59.

16. Dunning D, Heath C, Suls JM. Flawed self-assessment.

Implications for health, education, and the workplace.

Psychol Sci Public Interest. 2004;5:69–106.

17. Taylor DH Jr, Hasselblad V, Henley SJ, Thun MJ, Sloan FA.

Benefits of smoking cessation for longevity. Am J Public

Health. 2002;92:990–996.

18. Harris JE, Thun MJ, Mondul AM, Calle EE. Cigarette tar

yields in relation to mortality from lung cancer in the Can-

cer Prevention Study II prospective cohort, 1982–8. BMJ.

2004;328:72–80.

19. Lorence DP, Park H, Fox S. Assessing health consumerism

on the Web: a demographic profile of information-seeking

behaviors. J Med Syst. 2006;30:251–258.

20. Rutten LJ, Squiers L, Treiman K. Requests for information

by family and friends of cancer patients calling the

National Cancer Institute’s Cancer Information Service.

Psychooncology. 2006;15:664–672.


21. Pohls UG, Fasching PA, Beck H, et al. Demographic and

psychosocial factors associated with risk perception for

breast cancer. Oncol Rep. 2005;14:1605–1613.

22. Ralston JD, Taylor VM, Yasui Y, Kuniyuki A, Jackson JC, Tu

SP. Knowledge of cervical cancer risk factors among Chinese

immigrants in Seattle. J Community Health. 2003;28: 41–57.

23. Viswanath K, Breen N, Meissner H, et al. Cancer knowledge

and disparities in the information age. J Health Commun.

2006;11(suppl 1):1–17.

24. National Center for Chronic Disease Prevention and Health

Promotion. Behavioral Risk Factor Surveillance System

Summary Data Quality Report. Centers for Disease Control.

Updated Aug. 25, 2006. Available at URL: http://www.

cdc.gov/brfss/technical_infodata/quality.htm. Accessed on

Oct. 24, 2006.

25. Han PK, Moser RP, Klein WM. Perceived ambiguity about

cancer prevention recommendations: relationship to per-

ceptions of cancer preventability, risk, and worry. J Health

Commun. 2006;11(suppl 1):51–69.

26. NCRP Scientific Committee 72 on Radiation Protection in

Mammography. A Guide to Mammography and Other

Breast Imaging Procedures: Recommendations of the

National Council on Radiation Protection and Measure-

ments. NCRP No. 149. Bethesda, Md: National Council on

Radiation Protection and Measurements; 2004.

27. Coleman MP, Babb P, Sloggett A, Quinn M, De Stavola B.

Socioeconomic inequalities in cancer survival in England

and Wales. Cancer. 2001;91(suppl):208–216.

28. House JS, Williams DR.Understanding and reducing socio-

economic and racial disparities in health. In: Smedley BD,

Syme SL, eds. Promoting Health: Intervention Strategies

From Social and Behavioral Research. Washington, DC:

National Academy Press; 2001:81–124.

29. Smith RA, Cokkinides V, Eyre HJ. American Cancer Society

Guidelines for the Early Detection of Cancer, 2005. CA

Cancer J Clin. 2005;55:31–44.

30. Kushi LH, Byers T, Doyle C, et al. American Cancer Society

guidelines on nutrition and physical activity for cancer

prevention: reducing the risk of cancer with healthy food

choices and physical activity. CA Cancer J Clin. 2006;56:

254–281.

31. Campaign for Tobacco Free Kids. Tobacco Company Mar-

keting to African Americans. Campaign for Tobacco Free

Kids. Updated 2006. Available at URL: www.tobaccofreekids.

org/research/factsheets/pdf/0208.pdf. Accessed on Jan. 25,

2007.

32. Inskip PD, Mellemkjaer L, Gridley G, Olsen JH. Incidence

of intracranial tumors following hospitalization for head

injuries (Denmark). Cancer Causes Control. 1998;9:109–

116.

33. Merzenich H, Ahrens W, Stang A, et al. Sorting the hype

from the facts in testicular cancer: is testicular cancer

related to trauma? J Urol. 2000;164:2143–2144.

34. Schuz J, Jacobsen R, Olsen J, Boice JD Jr, McLaughlin JK,

Johansen C. Cellular telephone use and cancer risk: update

of a nationwide Danish cohort. J Natl Cancer Inst. 2006;

98:1707–1713.

35. Oliveria SA, Saraiya M, Geller AC, Heneghan MK, Jorgensen

C. Sun exposure and risk of melanoma. Arch Dis Child.

2006;91:131–138.

36. Mirick DK, Davis S, Thomas DB. Antiperspirant use and

the risk of breast cancer. J Natl Cancer Inst. 2002;94:1578–

1580.

37. Narod SA, Lubinski J, Ghadirian P, et al. Screening mam-

mography and risk of breast cancer in BRCA1 and BRCA2

mutation carriers: a case-control study. Lancet Oncol.

2006;7:402–406.

38. Gallagher RP, Spinelli JJ, Lee TK. Tanning beds, sunlamps,

and risk of cutaneous malignant melanoma. Cancer Epide-

miol Biomarkers Prev. 2005;14:562–566.

39. Singer SR, Grismaijer S. Dressed to Kill. Pahoa, Hawaii:

ISCD Press; 2005.


Documents

Prevalence and sociodemographic correlates of beliefs regarding cancer risks