Feasibility of targeted early detection for melanoma: a population-based screening study

The feasibility of targeted screening for cutaneous malignant melanoma in the UK using a postal questionnaire and invitation to screening by a consultant dermatologist was investigated in a population based cross-sectional survey. A total of 1600 people aged 25–69 years, stratified by the social deprivation score of wards within one general practice, were randomly selected from a population of 8000.1227 (77%) returned the questionnaire and 896 (56%) attended the screening clinic. Uptake was lower for men (P< 0.001), those aged under 50 (P< 0.001), people from deprived areas (P< 0.001) and skin types III and IV (men only, P< 0.001). Twenty per cent of women and 10% of men felt nervous about attending the clinic, but only 4% were worried by the questionnaire. The level of agreement between the self- and dermatologist's assessments of risk factors was best for hair colour (Kappa = 0.67, sensitivity 73% and specificity 98%). People tended to under-report their level of risk. Over 95% knew about at least one major sign, but 54% reported incorrect signs of melanoma. Targeted screening for melanoma in the UK will be hampered by difficulties in accurately identifying the target population. Strategies to improve skin self-awareness rather than screening should be developed and evaluated. © 2000 Cancer Research Campaign

A mole was defined as 'a skin mark which is usually dark brown but can be black or light brown. It appears before the age of 35 years and does not appear to change colour in the sun. Moles can be flat (not raised above the skin surface) or raised (felt above the skin surface)'. Only moles ≥ 2mm or more in diameter were counted. A freckle was defined as 'a light brown mark with an irregular edge that appears or goes darker in the sun, and is usually less than 2 mm in diameter. ' Moles and freckling were illustrated in photographs and diagrams. People were asked about their knowledge of the early signs of melanoma based on a recommended checklist (MacKie, 1990). Demographic details included date of birth, sex, age when left full-time education and living with a partner or spouse. Participants were asked whether they were pleased or concerned to receive the questionnaire, and to be invited to the screening clinic.
At the clinic, held 1 day per week from 14:30 to 18:30 h, the dermatologist asked about skin type, natural hair colour, a personal history of skin cancer and a family history of melanoma. A whole body examination was conducted excluding areas covered by underwear. Data on moles and freckling were collected using the same definitions as for the questionnaire. Anyone with a possible skin cancer was asked to see their GP so that they could be referred formally for excision/biopsy.
Descriptive statistics were used to study the uptake rate, and prevalence of risk factors in relation to various demographic factors in STATA. The accuracy with which people self-report risk factors for melanoma was assessed using measures of sensitivity (the proportion of people reporting a risk factor out of all those reported to have the risk factor by the dermatologist), specificity (the proportion of people who do not report a risk factor out of all those reported not to have the risk factor by the dermatologist), and Kappa statistics (the level of agreement between the individual and dermatologist taking into account agreement occurring by chance). The values of Kappa range from 0 (no agreement) to 1 (total agreement) and within this range, values < 0.4 are considered to represent poor agreement, 0.4-0.74 fair to good agreement and ≥ 0.75 excellent agreement (Landis and Koch, 1997). Logistic regression was used to study the relation of demographic factors to uptake rates, and to agreement of self-assessed risk factors to the dermatologist's observations.
Knowledge of the early signs of melanoma was studied in a multiple choice question containing nine possible correct statements interspersed with three incorrect statements. A score of correct answers ranging from 0 to 12 was created by scoring 1 for each of the nine correct signs that were ticked and scoring 1 for each of the three incorrect signs that were not ticked.

RESULTS
A total of 1766 people were randomly selected from the GP list. Of these, 164 were found to have moved or left the practice, and two had died. Out of 1600, 1277 (77%) returned the questionnaire and, of these, 896 (56% of 1600) also attended the clinic. Compared with those who had only returned the questionnaire (Q+C-), and those who did neither (Q-C-), those who had both returned the questionnaire and attended the clinic (Q+C+) had higher proportions of women than men (P < 0.001), those aged 50 or more than younger adults (P < 0.001) and people from the least deprived than more deprived areas (P < 0.001, Table 1). The most frequent reasons for non-attendance to the clinic were being too busy (47%), believing they had no risk of skin cancer (27%), and inconvenience of clinic time (18%). Only 1.5% objected to study. Thirty-eight people of Black or Brown ethnic group were excluded from subsequent analyses.
Comparing Q+C+ with Q+C-in regression analyses, men were more likely to attend the clinic if they were aged 50 or more (odds ratio (OR) 1.04, 95% confidence interval (CI) 1.02-1.06), or had skin type I or II (OR 1.71, 95% CI 1.26-2.32). Women were less likely to attend if they came from deprived areas (OR 0.8, 95% CI 0.70-0.93).
Only 4% found the questionnaire distressing. The proportion of people who were nervous about the thought of coming for a skin check was low (Q+C+ 15% and Q+C-17%), but it was twice as high in women than men (20% and 10% respectively, P < 0.001).
Twice as many in Q+C+ as in Q+C-wanted to find out more about their personal risk of skin cancer (74% vs 35%, P < 0.001). The levels of agreement between self-reports of risk factors and those recorded by the dermatologist were poor to fair for skin type, freckling and mole counts (Tables 2 and 3). The level of agreement was highest for hair colour when the reporting of red hair was compared with other hair colours as one group (Kappa 0.67) ( Table 2). Counting moles on the head and neck was more accurate among those with none or few freckles (Kappa 0.32) compared with those having moderate or many freckles (Kappa 0.17). The prevalence of freckling, and level of agreement for this factor or for mole counts did not vary with time of year.
There was a tendency for people to report the highest risk category less frequently than the dermatologist. Overall, 4% selfreported skin type I, 7% many freckles and 7% many moles. A high sensitivity could be achieved, although at a loss to specificity, if the self-reports of the two highest risk categories for each factor were compared with the dermatologist's assessment of the highest risk category (Tables 2 and 3). For example, for skin type, sensitivity increased from 24% to 96%, and specificity decreased from 100% to 63%, by comparing those self-reporting skin types I and II with those recorded as skin type I by the dermatologist. However, the prevalence of this potential target group for screening was high (48%).
In regression analyses restricted to those reported by the dermatologist to have skin type I (number in analysis 158), self-reports was more likely to agree with this assessment for males than females (OR 3.1, 95% CI 1.4-6.8) after allowing for the effects of age, social deprivation, age left education and living with a partner. For those with many freckles and those with many moles, the agreement was unrelated to demographic factors.
Ninety-three per cent knew about 'growing', 41% 'having different colours' and 35% 'having a ragged outline' as signs of melanoma. Seventy-six per cent also knew about 'changing shape'. Poor discriminatory signs 'bleeding, oozing or crusting', and 'a change in sensation such as itching or pain' were reported by 74% and 69% respectively. Fifty-four per cent reported one or more incorrect signs: scaly, hairy or raised above the skin surface as signs of melanoma. In regression analyses to study the relation between a score of correct answers (range 0-12) and demographic factors, significantly lower scores were found in men compared with women (regression coefficient -0.72, 95% CI -1.03 to -0.42) and in those who had left education by 16 (regression coefficient -0.22, 95% CI -0.37 to -0.07).
Eighteen people with a suspicion of skin cancer were advised to see their GPs for biopsy under the normal pattern of care. Twelve basal cell carcinomas and two melanomas were histologically confirmed. Both melanomas were thin (Breslow thickness < 0.76 mm) and were found on the backs of men aged more than 50 years.

DISCUSSION
Cost-effective screening requires a good uptake rate, a practical and accurate method for identifying the target population, and a choice of a target population that ensures a high yield of melanomas with consideration of both economic and psychological costs.  This study has shown that overall a good response (77%) can be achieved from a postal questionnaire but this varies according to demographic factors. The uptake to a screening clinic was lower (56%), but this could be improved by focusing on an older age group, among whom the incidence of melanoma is higher, offering a wider choice of clinic times and providing health education to increase awareness about skin cancer, particularly in more deprived areas (Eiser et al, submitted for publication). In this paper, 54% of people reported incorrect signs for melanoma.
The response rates to the questionnaire and uptake of screening were lower than that in a study conducted in an affluent general practice population in Wessex (84% and 89% respectively) (Little et al, 1995). The social mix, wider choice of appointment times, and use of the GP surgery may account for the higher response in Wessex.
The accuracy with which risk factors for melanoma could be identified by a postal questionnaire was disappointing, and was lower than in Wessex or Cheshire. One of the most important risk factors for melanoma is having a large number of naevi but there was a general reluctance for people to put themselves in the extreme group having a large number of moles. One option for targeted screening might be to select people who report moderate to high risk, to ensure that all high risk people are included in an intervention. However, this would increase the proportion of people falling into the target group (e.g. 4% and 42% self-reporting skin types I and II respectively).
The postal questionnaire raised anxiety about skin cancer in only a small proportion of cases (< 4%). Although about 15% of people said that they were nervous about attending for a skin check this did not seem to be related to attendance rates. It was not feasible to study anxiety over a longer period. The long-term psychological effects of screening would be important in future research testing new strategies.
The detection rate of melanomas (113 per 100 000) was very high compared with the incidence rate in 25-to 69-year-olds in England and Wales (8.9 per 100 000). Another study offering screening by a dermatologist in the workplace showed a similar high detection rate (Curley et al, 1993). Self-selection bias and the high detection expected in the prevalence round of screening may explain this finding.
The options for improving the early detection of melanoma include general population professional screening, targeted screening of a high risk group, self-screening and skin awareness. General population screening would not be cost-effective in the UK because of the low incidence. Targeted screening by GPs has been proposed (Little et al, 1995;Jackson et al, 1998) but this may benefit only a small proportion of melanomas and has considerable implications for workload, and psychological and economic costs (Keeley, 1995;Sinclair, 1998). Advice on skin awareness could be provided to encourage regular self-skin checks through a practice nurse (Ringborg et al, 1991). This advice should aim to improve awareness of the early signs of melanoma, demonstrated to be lacking in this survey. Recognition of melanoma by the general public might be improved by using a chart with photographs representing a range of benign, borderline and malignant lesions. The effectiveness of this approach would need to be evaluated in terms of self-referral rates to GPs, and sensitivity and specificity of detecting melanomas.
The results of previous feasibility studies for targeted screening of melanoma in the UK may have been too optimistic. Future strategies to improve and maintain both a high level of early detection by the general public and accurate recognition of suspicious lesions by GPs need to be evaluated in terms of workload, psychological outcome and economic costs.