Introduction

Governmental and charitable health agencies have made major efforts to change behavior in the sun (Boldeman et al., 1993; Marks, 1994; McWhirter et al., 2000), with some evidence that these primary prevention approaches to melanoma have had an impact on rising mortality (Giles et al., 1996). The difficulties of changing behavior are acknowledged however (Marks, 1994), particularly when that behavior is widely perceived as having positive benefits on health and appearance, such as suntans. The issue is more important still in families at risk of melanoma, and we describe here a study of reported attitudes to the sun, and behavior in the sun, in UK families. We have studied attitudes to sun-avoidance advice within such melanoma families to inform the counseling process, but insights from these data may be applicable to health education in more common cancer family syndromes where environmental determinants are currently less understood.

Results

Questionnaires were sent to a total of 242 relatives and 170 (70.2%) were completed (Table 1). Data were collected from 48 melanoma cases, 78 people who had a first-degree relative with melanoma, and 44 with an affected second-degree relative. Out of 185 GPs contacted, 18 refused to seek a suitable control. Three people declined and 24 never returned their questionnaires, giving a total of 140 matched controls (75.7% response rate). Of the 170 relatives, 58 were genotyped and 23 of those were subsequently found to have germ line CDKN2A mutations, but they and we were unaware of their gene status at the time the questionnaires were sent. Fifty-nine percent (26/44) of melanoma cases were atypical mole syndrome (AMS)-affected compared with 32% (25/77) of first-degree relatives and 15% (6/41) of second-degree relatives.

Table 1 - Demographic and phenotypic characteristics of the study participants.

Full table

Among relatives, greater knowledge about melanoma was correlated with a greater belief in the individual's view of their ability to avoid melanoma (correlation coefficient=0.31, P<0.0001). A greater belief in their ability to avoid melanoma (BAM (beliefs about melanoma questionnaire)) was correlated with a lower chance locus of control (C-HLoC) and lower trait anxiety (P=0.02 and P=0.04, respectively). Greater knowledge about melanoma was also correlated with a lower P-HLoC (coefficient =-0.18, P=0.02). A higher I-HLoC was correlated with lower state anxiety and lower trait anxiety (P=0.004 and P=0.0001). State anxiety and trait anxiety were also highly correlated (P<0.0001).

The first seven items in Table 2 refer to knowledge and belief scores. Of these seven items, only the distributions of scores for BAM differed significantly between the melanoma cases, first-degree, and second-degree relatives (data not shown ANOVA, P<0.0001). Second-degree relatives had similar belief scores to controls, but the melanoma cases and first-degree relatives had greater belief in their ability to avoid melanoma. Compared to all relatives, controls had a lower knowledge about melanoma (P=0.006), a higher P-HLoC (P=0.04) and lower knowledge about preventing melanoma (P=0.001). The distributions of the remaining knowledge and belief scores in Table 2 did not differ between relatives and controls. The intraclass correlations of the scores range between 0.06 and 0.18, which indicates that there is little correlation between scores in relatives from the same family.

Table 2 - Distribution of questionnaire responses in relatives and control and results of regression analyses.

Full table

Sun-protection Behavior In Relatives And Controls

The last nine items in Table 2 summarize the differences in reported sun-exposure preferences of relatives and controls. Relatives reported significantly more sun-protection behaviors than controls (P=0.001), specifically the use of sunscreen and clothing to cover the skin. Relatives were less likely to report a preference for a darker tan (P<0.0001): 31% of the relatives reported that they preferred to avoid a tan compared with 19% of the controls. Relatives were significantly less likely to report a history of sunburn than controls (P=0.03) and there was no significant difference in the number of episodes of sunburn in the previous year (ordinal regression P=0.08, data not shown). The proportion of variance owing to the family effect ranged between 0.13 and 0.19, indicating that there is some dependence in behaviors within families but it is not strong.

It was found that relatives checked their moles more frequently than did the controls (Table 2, P=0.007). However, there was no difference in mole-checking behavior between relatives of melanoma cases and controls (ordinal regression P=0.4, data not shown). Sixteen percent of relatives who had three or more cases of melanoma in the family reported sunburn compared with 39% of relatives with two cases or fewer (²(1)=9.9, P=0.002)). Relatives with three or more cases were also less likely to tan than those with two cases or fewer (43% compared with 60%, ²(1)=4.1, P=0.04). No differences were seen for the other sun-exposure behaviors.

Predictors of Behavior In The Sun

Sunburn and suntans

Hair color and Fitzpatrick skin type were not predictive of sunburn (Fisher's exact test, P=0.9 for both comparisons) and CDKN2A-mutation positivity was not associated with a history of sunburn in the previous year (²(1)=0.1, P=0.7). AMS-affected relatives were less likely to have been sunburnt during the previous year (n=161, odds ratio=0.32, 95% CI (0.13, 0.81), P=0.02). Although this relationship was no longer significant after excluding the melanoma cases (n=117, odds ratio=0.43, 95% CI (0.16, 1.13), P=0.09), the estimated odds ratio was still similar. Table 3 shows output from the univariate logistic regression predicting getting sunburnt in relatives and controls.

Table 3 - Univariate logistic regression predicting getting sunburnt in relatives and controls.

Full table

Both relatives and controls who became sunburnt were more likely to be male and younger (Table 3). Among relatives, getting sunburnt was rarer in those with higher belief in their ability to avoid melanoma but more common in those with a higher C-HLoC. Controls were four times more likely to have been sunburnt the previous year compared with melanoma cases (P=0.01). Nonetheless, among those who had a personal history of melanoma and their first-degree relatives, 6/48 (12.5%) and 26/78 (33%), respectively, reported sunburn in the preceding year (Table 3).

It was notable that getting sunburnt and a tan were significantly positively related in relatives (²(1)=9.4, P=0.002) and marginally so in the controls (²(1)=3.9, P=0.05). In both relatives and controls, getting a tan was associated with younger ages (data not shown), but was rarer among relatives with higher belief in their ability to avoid melanoma (odds ratio=0.82, 95% CI (0.73, 0.91), P<0.0001). Furthermore, first-degree relatives and controls were three times more likely to have had a tan in the previous year compared with melanoma cases (P=0.04 and P=0.01, respectively). Apart from age, in the controls there were no significant determinants of acquiring a tan.

Degree of desired tan

The degree of desired tan was not associated with hair color, AMS status, or CDKN2A status, but preference for a darker tan was higher in relatives of darker skin types (ordinal regression, P<0.0001). The first part of Table 4 shows factors predicting degree of desired tan. In both relatives and controls, the preferred tan was darker for those with a lower KAM (knowledge about melanoma questionnaire) score, a lower BAM score, and for younger people. Also, in both groups, the preference for a darker tan was associated with a higher C-HLoC. There were no associations between preferred color of tan and state anxiety, gender, I-HLoC, or P-HLoC. First-degree relatives and controls were more likely to prefer a darker tan than those with a personal history of melanoma.

Table 4 - Univariate ordinal regression predicting degree of desired tan and sun-protection behaviors in relatives and controls.

Full table

Sun-protection behavior

In relatives, the number of sun-protection behaviors was not associated with hair color or CDKN2A status, but was associated with fairer skin (ordinal regression P=0.02, data not shown). Further, more sun-protection behaviors were used by AMS-affected relatives (P<0.0001), and this relationship remained significant after excluding melanoma cases (P=0.03). Among relatives, sun-protection behaviors were more commonly reported by those with higher KAM score and BAM scores (Table 4). First-degree relatives and controls used fewer sun-protective behaviors than melanoma cases (P=0.02 and 0.001, respectively), though when adjusting for AMS status, the reduction seen among first-degree relatives became nonsignificant (P=0.1).

Discussion

This is early evidence of the determinants of behavior in multiple-case melanoma families. Weaknesses of the study are that we were unable to examine controls. The controls were identified by family doctors, and bias in selection by those doctors cannot be excluded. Questionnaires were returned over a protracted period and there may have been some variations according to time at which the questions were answered, which we were unable to address, although the questions were addressed to the preceding year in all cases.

The families knew they were at some increased risk because of their family history, but they had not undergone formal gene testing. It has been suggested that, pending more knowledge about gene testing and CDKN2A in particular, sun-protection advice should be delivered to all relatives rather than attempting to identify those at greater risk (Kefford et al., 1999). Well-recognized models of health psychology, however, suggest that preventive behaviors are only adopted when the individual perceives the risk to itself, as significant (Becker et al., 1977; Hunt et al., 2000). Furthermore, decisions about behavior related to health are complex and are made in the context of the individual's experiences, age, and health beliefs (Leventhal et al., 1998; Hunt et al., 2000) and the experiences of their family (Davidson et al., 1989). There is particular interest now, therefore, in understanding the relationships between behavior, experience of cancer in the family, perception of risk, gender, age, and personality, which are explored in this study.

Members of families at risk of melanoma worldwide are advised strongly against excessive exposure to the sun and particularly against sunburn. It was therefore chastening to determine that 31% of relatives reported sunburn during the previous year, compared with 41% of controls. We did not collect phenotypic information from our controls. Therefore, it is not possible to exclude the possibility that our families were fairer skinned than controls. However, this probably does not explain the high prevalence of sunburn in the families, as within the relatives there was no association between skin type or hair color and the history of sunburn. In summary, there was a surprising and disappointingly high prevalence of sunburn in relatives. A Swedish group had previously reported similar "risky" behaviors in a patient population with the AMS (Brandberg et al., 1996) and more recently, low usage of protective measures was reported in two studies of first-degree relatives of melanoma patients (Manne et al., 2004; Azzarello et al., 2006).

Within families, subjects were more likely to have been sunburnt in the previous year, the more distant they were from the melanoma cases. There is some evidence that those with the AMS, even excluding melanoma cases, were also less likely to be sunburnt (OR=0.43, 95% CI (0.16, 1.13)) and better behaviors were reported by members of families with larger numbers of cases. It is postulated that these individuals would have perceived themselves to be at increased risk on the basis of their family or personal experience and that this "explained" in part their better behaviors.

Gender and age were also important. Males were more likely to report sunburn in both the families and the controls, as has been reported before in Australia (Hill et al., 1992) and the UK (Melia and Bulman, 1995). This study suggests that education of men from melanoma families should be particularly addressed, and it is important to determine the effect of genetic counseling on behavior in this group.

Despite the fact that relatives showed little difference from controls in reported levels of sunburn, they did report significantly more sun-protective behaviors than controls. The explanation for the disparity is not clear. One explanation might be that it is easy to get burnt even when trying to take protective measures. It may imply that relatives' understanding of the level of protection needed was insufficient. Getting sunburnt and acquiring a tan were significantly correlated, and although relatives were significantly less likely to prefer a dark tan than controls, 55% of relatives had tanned during the previous summer. Another explanation then for the occurrence of sunburn in families was that, as in the general population, the desire for a tan is strong even when there is knowledge of particular risk. It is possible that families even share the view previously reported in the general population, that sunburn is "part of acquiring a tan" (Melia and Bulman, 1995). Again the desire for a tan was reduced in older relatives, melanoma patients, closer relatives of melanoma cases, and those with a greater belief in their ability to prevent melanoma. Clearly, therefore, there is the potential to improve behavior.

The determinants of protective behaviors in relatives were greater knowledge about melanoma, high BAM scores, personal history of melanoma, and the AMS. Surprisingly, there appeared to be no effect of number of melanoma cases in the family (P=0.93), which may suggest that the number may seem less significant to families than it is to clinicians. There is some evidence from studies on other cancers that healthy individuals may view family history as less significant than do clinicians (Jacobs, 2002), although that contrasts with experience in clinical practice. There was a nonlinear relationship between protective behaviors and anxiety trait, which was predicted. That is, relatives with low or high levels reported fewer protective behaviors than those with an "optimal" intermediate anxiety trait score. Relatives with lower P-HLoC were more likely to have higher knowledge about melanoma, which was predictive of "better" behaviors. Thus, endogenous psychological traits may influence behavior as was predicted.

What then were the determinants of high BAM scores, that is, reported belief in their ability to prevent melanoma? Overall, greater knowledge about melanoma correlated with high BAM scores, and the female controls had higher BAM scores than the male controls (linear regression, P=0.02). Most interestingly, melanoma patients had significantly better belief scores than their relatives and the controls (linear regression, P<=0.001 in all cases), and first-degree relatives had better belief scores than the second-degree relatives and the controls (P=0.01 in both comparisons). Although we cannot exclude the possibility that the BAM questionnaire was effectively measuring knowledge rather than confidence, this observation may also imply that exposure to health-care professionals does improve confidence by increased knowledge even after the diagnosis of melanoma. The clear relationship between knowledge, belief, protective behaviors, and reduced levels of sunburn in these families again argues strongly that education is helpful. It supports the view that gene testing (when appropriately supported by data on risk) may promote better behaviors rather than promoting nihilism.

Among relatives those with greater knowledge about melanoma risk were more likely to use protective behavior, whereas in controls there was no relationship between knowledge and behavior. Thus, it is insufficient in terms of motivating people to adopt health-protective behaviors merely to educate: they have to perceive themselves to be at risk as discussed by Becker in his Health Beliefs Model (Becker et al., 1977).

The determinants of health-protecting behavior are complex. They include an individual's perception of their own risk, their view of the seriousness of developing that disease, and his/her belief that a given behavior will be preventive. Behaviors such as electing to have genetic testing have been linked to taking a long-term view which is reported to correlate with higher levels of education (Gurmankin Levy et al., 2006). The long-term value of the behavior is balanced with the shorter-term negative effects. Knowledge of increased risk may be insufficient to motivate improved behaviors in those at increased risk of cancer generally, and sun protection may be particularly unacceptable in comparison with other health-promotional behaviors, such as exercise and weight reduction which have some obvious perceived additional short-term benefits at least for some. Recent studies showing similarly "poor" behaviors in other "at risk" groups were a study of survivors of childhood cancer in whom sun exposure and physical inactivity were the dominant risk factors reported (Tercyak et al., 2005) and another of renal transplant recipients in whom suntanning remained common (Donovan et al., 2004).

Materials and Methods

This study comprised individuals from families who had taken part in studies of familial melanoma since 1989 (Newton Bishop et al., 1994). Entry criteria for families were two or more cases of melanoma (16 two-case families, 5 three-case, 5 four-case, 1 five–case, and 1 six-case) or 1 melanoma case with multiple relatives with the AMS (Newton et al., 1993) (16 families). Invitations were sent to all relatives who had participated in the research. These families had all been referred to the research group by hospital clinicians who had counseled the probands about risk and sun protection. Counseling about risk and sun protection (by JNB) also took place after initial data collection in the research clinic: the whole family was counseled together in one room. Relatives were therefore considered to be "informed" about their risks in broad terms (family-based rather than individually directed). They were therefore educated about their need to reduce sun exposure by sun avoidance and use of clothing in preference to the use of sunscreen, before participation in the study reported here.

Controls were selected by asking the relatives' own family doctors to identify a person living in their geographical area, of the same sex, and who was nearest in date of birth to the family member. Past history of any cancer was a reason for exclusion. We were not funded to examine these persons who lived throughout the UK. The study received local ethical committee approval, and all participants gave informed written consent.

The following five postal questionnaires were used to assess anxiety, beliefs, behavior, knowledge, and attitudes. Two measures were well-known standardized instruments (state–trait anxiety inventory (STAI) and multidimensional health locus of control (MHLC)) and three (BAM, Current Behavior Questionnaire (CBQ), and KAM) were developed specifically for this study, but were based on established questionnaires used and evaluated in studies of women with breast cancer (Fallowfield et al., 1990).

STAI

The STAI is a 40-item questionnaire designed to evaluate state anxiety (20 items) and trait anxiety (20 items). The scores are summed and divided by 20 to produce average state anxiety and trait anxiety scores (Trask et al., 2001).

MHLC

The MHLC (Wallston et al., 1978) comprises 18 items, each item measuring one of three subscales: I-HLoC, P-HLoC, and C-HLoC. These measure the degree of belief that one's health is influenced by one's own behavior, other people (such as doctors, friends, and family), or "chance", respectively.

BAM

The BAM questionnaire consists of 10 items measured on a four-point scale. For example, statement one is "If more people had their moles checked regularly there would be fewer deaths from malignant melanoma". These scores were summed to produce total scores ranging from 0 to 30 such that a high score indicates stronger beliefs in ability to avoid melanoma.

CBQ

This consists of 14 items designed to evaluate mole checking frequency and sun-protection behaviors used on holiday that year. One point was given for each of four behaviors (Table 2), and these scores were then summed to produce a total sun-protection score ranging from 0 to 4 (used all 4). The participants were also asked "How deep a tan do you like to get?" and whether they had had sunburn in the previous summer sufficient to make them "red and sore".

KAM

The KAM questionnaire has 27 items that evaluate general knowledge about melanoma, such as the signs and symptoms of melanoma and two about prevention of melanoma. The first 27 items are measured on a true/false scale, the first item being "UVA is the part of sunlight that is most harmful". These scores were summed to produce a total score ranging from 0 (the least knowledge) to 27.

Data were collected from the relatives on their nevus phenotype (AMS status), hair and eye color, and Fitzpatrick skin type by nurse examination. Controls were not examined.

The principal outcome measures were reported sunburn, suntans, and use of protective measures.

Statistical methods

Among relatives, Spearman's correlation coefficients were produced for all pair-wise combinations of the STAI, MHLC, BAM, and KAM questionnaire variables. Differences in these scores between relatives were determined using one-way ANOVA. The scores were then compared between all relatives and controls using linear or ordinal regression. Similarly, various sun-protection measures from the CBQ were compared between relatives and controls using ordinal or logistic regression. The analyses of relatives allowed for the dependence between members of the same family by using robust standard errors to account for correlations within families. Intraclass correlation coefficients were calculated to determine the correlation of scores between relatives in a family.

Predictors of getting sunburnt and a suntan were identified using logistic regression, and predictors for the number of sun-protection behaviors were identified using ordinal regression. These analyses were performed separately for relatives and controls. All regression models included adjustment for age and sex, and analyses of relatives using robust standard errors. Random effects logistic models were used to calculate the proportion of total variance due to the family effect. Hair color, eye color, skin type, and AMS status were studied as possible predictors of getting sunburnt, getting a suntan, and the number of sun-protection behaviors. The questionnaire variables that were studied as possible predictors were the STAI, MHLC, BAM, and KAM scores.

Conflict of Interest

The authors state no conflict of interest.

References

1. Azzarello LM, Dessureault S, Jacobsen PB (2006) Sun-protective behavior among individuals with a family history of melanoma. Cancer Epidemiol Biomarkers Prev 15:142–145 | Article | PubMed |
2. Becker MH, Maiman LA, Kirscht JP, Haefner DP, Drachman RH (1977) The Health Belief Model and prediction of dietary compliance: a field experiment. J Health Soc Behav 18:348–366 | Article | PubMed | ChemPort |
3. Boldeman C, Ullén H, Månsson-Brahme E, Holm L-E (1993) Primary prevention of malignant melanoma in the Stockholm Cancer Prevention Programme. Eur J Cancer Prev 2:441–446 | PubMed | ChemPort |
4. Brandberg Y, Jonell R, Broberg M, Sjoden PO, Rosdahl I (1996) Sun-related behaviour in individuals with dysplastic naevus syndrome. Acta Derm Venereol 76:381–384 | PubMed | ChemPort |
5. Davison C, Frankel S, Smith G (1989) Inheriting heart trouble: the relevance of common-sense ideas to preventative measures. Health Educ Res 4:329–340 | Article |
6. Donovan JC, Rosen CF, Shaw JC (2004) Evaluation of sun-protective practices of organ transplant recipients. Am J Transplant 4:1852–1858 | Article | PubMed |
7. Fallowfield L, Rodway JA, Baum M (1990) What are the psychological factors influencing attendance, non-attendance and re-attendance at a breast screening centre? J R Soc Med 83:547–551 | PubMed | ChemPort |
8. Giles GG, Armstrong BK, Burton RC, Staples MP, Thursfield VJ (1996) Has mortality from melanoma stopped rising in Australia? Analysis of trends between 1931 and 1994. BMJ 312:1121–1125 | PubMed | ChemPort |
9. Gurmankin Levy A, Micco E, Putt M, Armstrong K (2006) Value for the future and breast cancer-preventive health behavior. Cancer Epidemiol Biomarkers Prev 15:955–960 | Article | PubMed |
10. Hill D, White V, Marks R, Theobald T, Borland R, Roy C (1992) Melanoma prevention: behavioural and nonbehavioural fctors in sunburn among an Australian urban population. Prevent Med 21:654–669 | Article | ChemPort |
11. Hunt K, Davison C, Emslie C, Ford G (2000) Are perceptions of a family history of heart disease related to health-related attitudes and behaviour? Health Educ Res 15:131–143 | Article | PubMed | ChemPort |
12. Jacobs LA (2002) Health beliefs of first-degree relatives of individuals with colorectal cancer and participation in health maintenance visits: a population-based survey. Cancer Nurs 25:251–265 | Article | PubMed |
13. Kefford RF, Newton Bishop JA, Bergman W, Tucker MA (1999) Counseling and DNA testing for individuals perceived to be genetically predisposed to melanoma: a consensus statement of the Melanoma Genetics Consortium. J Clin Oncol 17:3245–3251 | PubMed | ISI | ChemPort |
14. Leventhal H, Leventhal E, Contrada R (1998) Self regulation, health and behaviour: a perceptual cognitive approach. Psychol Health 13:717–734
15. Manne S, Fasanella N, Connors J, Floyd B, Wang H, Lessin S (2004) Sun protection and skin surveillance practices among relatives of patients with malignant melanoma: prevalence and predictors. Prev Med 39:36–47 | Article | PubMed |
16. Marks R (1994) Melanoma prevention: is it possible to change a population's behaviour in the sun? Pigment Cell Res 7:104–106 | Article | PubMed | ChemPort |
17. McWhirter JM, Collins M, Bryant I, Wetton NM, Newton Bishop J (2000) Evaluating 'Safe in the Sun', a curriculum programme for primary schools. Health Educ Res 15:203–217 | Article | PubMed | ChemPort |
18. Melia J, Bulman A (1995) Sunburn and tanning in a British population. J Public Health Med 17:223–229 | PubMed | ChemPort |
19. Newton Bishop JA, Bataille V, Pinney E, Bishop DT (1994) Family studies in melanoma: identification of the atypical mole syndrome (AMS) phenotype. Melanoma Res 4:199–206 | Article | PubMed | ChemPort |
20. Newton JA, Bataille V, Griffiths K, Squire JM, Sasieni P, Cuzick J et al. (1993) How common is the atypical mole syndrome phenotype in apparently sporadic melanoma? J Am Acad Dermatol 29:989–996 | PubMed | ISI | ChemPort |
21. Tercyak KP, Donze JR, Prahlad S, Mosher RB, Shad AT (2006) Multiple behavioral risk factors among adolescent survivors of childhood cancer in the Survivor Health and Resilience Education (SHARE) Program. Pediatr Blood Cancer 47:825–830 | Article | PubMed |
22. Trask PC, Paterson AG, Hayasaka S, Dunn RL, Riba M, Johnson T (2001) Psychosocial characteristics of individuals with non-stage IV melanoma. J Clin Oncol 19:2844–2850 | PubMed | ISI | ChemPort |
23. Wallston K, Wallston B, DeVellis R (1978) Development of the multidimensional health locus of control (MHLC) scales. Health Educ Monogr 6:161–170

Acknowledgments

We are grateful to the relatives who took the time to complete these questionnaires and for their continued support of research. Stephanie Hamer and Patricia Mack contributed to the data collection for this study and we are grateful to them for their help. We are also grateful to the GPs who assisted us with the identification of controls: more work for already busy people. The study was funded from a private bequest, from the estate of an anonymous individual who died of cancer and by the ICRF, now Cancer Research UK.