Accuracy of self-reported nevus and pigmentation phenotype compared to clinical assessment in a population-based study of young Australian adults

Abstract

Background

Awareness of individual risk may encourage improved prevention and early detection of melanoma.

Methods

We evaluated the accuracy of self-reported pigmentation and nevus phenotype compared to clinical assessment, and examined agreement between nevus counts from selected anatomical regions. The sample included 456 cases with invasive cutaneous melanoma diagnosed between ages 18-39 years and 538 controls from the population-based Australian Melanoma Family Study. Participants completed a questionnaire regarding their pigmentation and nevus phenotype, and attended a dermatologic skin examination.

Results

There was strong agreement between self-reported and clinical assessment of eye color (kappa, κ, =0.78, 95% confidence interval (CI) 0.74-0.81); and moderate agreement for hair color (κ =0.46, 95% CI 0.42-0.50). Agreement between self-reported skin color and spectrophotometer-derived measurements was poor (κ =0.12, 95% CI 0.08-0.16) to moderate (Spearman correlation r_s=-0.37, 95% CI -0.32- to -0.42). Participants tended to under-estimate their nevus counts and pigmentation; men were more likely to under-report their skin color. The r_s was 0.43 (95% CI 0.38-0.49) comparing clinical total body nevus counts with self-reported nevus categories. There was good agreement of quartile distributions of total body nevus counts with site-specific nevus counts, particularly on both arms.

Conclusions

Young adults have sub-optimal accuracy when assessing important risk characteristics including nevus numbers and pigmentation. Measuring nevus count on the arms is a good predictor of full body nevus count.

Impact

These results have implications for the likely success of targeted public health programs that rely on self-assessment of these factors.

Introduction

Melanoma is a common malignancy in fair-skinned populations (1-3). In Australia, melanoma is the most frequently diagnosed cancer among those aged 15-44 years, accounting for 23% of incident cancers and 8% of cancer deaths in this age-group (4) Melanoma is a highly preventable cancer (5, 6) and has an excellent prognosis when detected early (7). There is no formal melanoma screening program in the general population, so it is particularly important that individuals are able to accurately assess key melanoma risk factors so that those at high risk can identify themselves as such and take appropriate preventative actions, such as increasing sun-protection and having regular skin checks, from a young age.

Visibly identifiable risk factors for melanoma include nevi (8) and pigmentation indicative of a sun-sensitive phenotype such as fair skin, red or blonde hair, and blue or green eye color (9). For assessing the validity of self-reported phenotypic characteristics, the ‘gold standard’ measure is clinical assessment by dermatologists, who have expertise in classifying level of pigmentation and identifying nevi.

There are several limitations to previous validation studies of self-reported phenotype characteristics. Although young adults are considered a key target group for melanoma prevention campaigns, most previous validation studies have been conducted among older adults. Many previous studies have also used convenience or clinic-based sampling, and had small sample sizes. It is also uncertain to what degree people with and without melanoma differ in the accuracy of their self-reported phenotype, and whether this has the potential to bias risk estimates derived from case-control studies of melanoma. Several nevus validation studies have focused on assessment of larger and atypical nevi (10, 11); however, the total number of common melanocytic nevi on the body is a stronger risk factor for melanoma (8).

The aim of this study was to evaluate the accuracy of self-reported pigmentation and nevus phenotype compared to more objective measures recorded by dermatology trainees during a clinical skin examination, for 994 young adults aged 18-44 years including 456 people with melanoma (cases) and 538 people without melanoma (controls) who participated in the Australian Melanoma Family Study. We also examined predictors of disagreement. A secondary aim was to examine agreement between clinically-measured nevus counts on the whole body to that from selected anatomical regions, in order to strengthen the evidence for using localized nevus count as a proxy measure of whole body nevus counts.

Materials and Methods

Study sample

The Australian Melanoma Family Study is a multi-centre, population-based, case-control-family study of invasive cutaneous melanoma diagnosed between ages 18 and 39 years. The study design, recruitment, data collection and participant characteristics have been previously described in detail (12). Recruitment of case and control participants was locally conducted in Sydney, Melbourne, and Brisbane, Australia. Approval for the study was obtained from the ethics committees of the three coordinating centers and the cancer registries. All participants provided written, informed consent.

Case participants

Cases were identified from population-based state cancer registries, diagnosed between 1st July 2000 and 31st December 2002 at ages 18-39 years with incident, histopathologically-confirmed, first-primary invasive cutaneous melanoma. Participation was 54% of those eligible and 76% of those contactable.

Control participants: population, spouse/friend and sibling

Population controls were aged between 18 and 39 years at the time of approach and had no history of invasive or in situ melanoma. They were selected from the electoral roll (registration to vote is compulsory for Australian citizens aged 18 years and over) and were frequency-matched to cases by city, age (within 5 years) and sex. Participation was 23% of those eligible and 42% of those contactable. Eligible spouse or friend controls were a spouse, partner, or friend nominated by a case as a potential control subject. They were eligible if they were at least 18 years of age and had no history of invasive or in situ melanoma; there were no other age, sex or residency restrictions. A spouse or friend was nominated as a potential control subject by 59% of cases, and participation was 80% of those nominated. We sought to recruit all siblings of cases for whom the case permitted contact. Unaffected siblings of cases were considered as sibling controls, and at least one sibling control was identified for 91% of cases.

Self-reported pigmentation phenotype and nevus counts

Participants completed a questionnaire in which they reported their skin color (very fair, fair, olive or brown, Asian, black, other), eye color (blue or grey, green or hazel, brown or black), natural hair color at age 18 (red or auburn, fair or blonde, light or mouse brown, grey, dark brown, black), ability to tan (tanning response to repeated sun exposure in summer: deep tan, moderately tan, mild tan, no tan), propensity to burn sunburn (response to prolonged sun exposure in summer: severe sunburn with blistering, painful sunburn with peeling, mild sunburn, no sunburn), usual tanning and sunburn response to prolonged or repeated exposure of skin to sunlight in summer, the number of nevi covering their body (described pictorially as none, few, some, many), and were asked to have someone count the number of all moles on their back (using picture guides to define the area and describe moles) (12).

Clinical assessment of pigmentation and nevus phenotype

A clinical skin examination was completed by 73% of cases, 55% of population controls, 67% of spouse or friend controls and 43% of sibling controls. Skin examinations were conducted at dermatology clinics in Brisbane, Sydney, and Melbourne by dermatology trainees trained on the study protocol. A dermatologist with experience in using this protocol trained examiners in classification of lesions to minimize undercounting. Participants removed their clothing except for underpants and bra. Measurement of nevi was based on international guidelines (13). Separate counts were made for melanocytic nevi of 2-5mm and >5mm, raised nevi of >2mm, and clinically atypical nevi of >2mm, on 30 body sites (Supplementary Figure S1). Only nevi over 2mm were considered, to minimise confusion with freckles and lentigines. Natural hair color at age 18 (red, fair or blonde, brown, black) and eye color (light blue or blue or blue/grey, green or hazel, light brown or brown or dark brown) were recorded by using wig hair swatches and eye photographs.

Skin reflectance

Reflected skin color, a correlate of melanin content (14, 15), was objectively recorded (mean of 6 readings) from both the outer and inner part of the subject's left upper arm using a BYK-Gardner CGSS (Geretsried, Germany) hand-held reflectance spectrophotometer with standard reflectance at 685 nm, calibrated before each session. The multi-wavelength data quantify color using the Commission Internationale de l'Éclairage L* a* b* color space parameters: with dimension L* indicating lightness of color (smaller values indicate darker color), a* the position between red and green and b* the position between yellow and blue (smaller values indicate lighter color) (16). Inner arm L* values have been used to describe natural skin color, b* values to describe tanning, and a* values to describe erythema (14, 16-18). Spectrophotometer readings from the inner arm were also converted to individual typography angle (ITA) scores, calculated using the values of L* and b* according to the formula: ITA°=(ArcTAN((L*-50)/b*))*180/3.14159 (19), where Arc Tangent is an inverse trigonometric function expressed in radians. ITA scores have been shown to correlate with melanin content (19) and have been used to classify skin color into six categories (19, 20); we collapsed to three categories using cut-points: very fair ≥ 55°> fair ≥ 28°> olive/brown/dark (19).

Statistical analysis

Data were available for 994 participants aged 18-44 years (456 cases and 538 controls, of which 131 population-based controls, 168 spouse, partner or friend controls and 239 sibling controls), who had completed a self-report questionnaire and attended a clinical skin examination.

We compared self-reported and clinical measures using Cohen's kappa (κ) and weighted kappa (κ_w) statistics for categorical and ordinal variables (21) and intra-class correlation coefficients for continuous variables. Kappa statistics are interpreted as: <0.20 poor agreement beyond chance, 0.21-0.40 fair agreement, 0.41-0.60 moderate agreement, 0.61-0.80 good agreement and 0.81-1.00 excellent agreement (22). Box-and-whisker plots and Spearman correlation coefficients (r_s) were used to compare ordinal variables with continuous variables, e.g. when comparing self-reported nevus categories (none, few, some, many) with clinically-measured nevus counts. Spectrophotometer (L* a* b*) values and nevus counts were analyzed as continuous variables, and as quartiles based on the distribution in controls.

As there were slight differences in the categories for eye and hair color measures between self-report and clinical assessment, we harmonized the categories for analysis. We used four categories for natural hair color at age 18: red (self-report: red or auburn; clinical: red), blonde/fair (self-report: fair or blonde, grey; clinical: fair or blonde), brown (self-report: light or mouse brown, dark brown; clinical: brown), black (self-report: black; clinical: black). We used three eye color categories: blue/grey (self-report: blue or grey; clinical: light blue or blue or blue/grey), Green/hazel (self-report: green or hazel; clinical: green or hazel), brown/black (self-report: brown or black; clinical: light brown or brown or dark brown). Self-reported skin color was compared with spectrophotometer values as specified above; due to small numbers we grouped participants with olive or brown, Asian, black, or other skin color as one category “olive/dark”.

We examined agreement between clinically-measured nevus counts on the whole body to those from selected anatomical regions, including the posterior body (from head to toe on the back side of the body), the back, both arms, the right arm, and both thighs (see Supplementary Figure 1 for classification of body sites). We examined whether any personal characteristics or other factors were associated with disagreement between self-reported and clinically-measured pigmentation phenotype variables (hair, skin and eye color), using log-binomial regression model to estimate probability ratios and 95% confidence intervals (CI) after adjusting for age, sex and city of recruitment.

All participants (cases and different control groups) were combined for the main analysis, and we conducted subgroup analyses to examine whether the accuracy of self-reported measures differed by case-control status and sex. All analyses were performed using SAS Statistical Software (version 9.3, SAS Institute, Cary, NC), and statistical significance was inferred at two-sided P <0.05.

Results

Demographic and phenotype characteristics of cases and controls are shown in Table 1. Nearly all participants were of white European descent and 42% of both cases and controls had a University degree. Cases were more likely to have a sun-sensitive phenotype and more nevi than controls.

Table 1. Demographic and clinically-measured phenotype characteristics of cases and controls.

Characteristic	Cases n = 456		Controls n = 538
	N	%	N	%
Sex
Male	170	(37)	232	(43)
Female	286	(63)	306	(57)
Age at diagnosis/interview (years)1
18-29	139	(30)	120	(22)
30-44	317	(70)	418	(78)
Ethnicity
Caucasian	443	(97)	516	(96)
Other	13	(3)	22	(4)
Highest level of education
Junior secondary school	61	(13)	54	(10)
Senior secondary school	116	(26)	124	(23)
Vocational/Certificate	86	(19)	134	(25)
University	189	(42)	223	(42)
Hair Color2
Red	82	(18)	49	(9)
Blonde/Fair	197	(43)	213	(40)
Black/Brown	177	(39)	274	(51)
Eye Color2
Blue/grey	233	(51)	262	(49)
Green/hazel	133	(29)	155	(29)
Brown/black	87	(19)	120	(22)
Skin color2,3
Very fair	288	(63)	289	(54)
Fair	149	(33)	202	(38)
Olive/Brown/Dark	18	(4)	43	(8)
Total-body nevi ≥ 2 mm2 (median, 25th-75th centile)	201	107-307	91	37-191

¹All cases were < 40 years at diagnosis and all population controls were < 40 years when ascertained. Cases and controls could be up to age 44 years at interview for this analysis.

²Assessed from the clinical skin examinations.

³Skin color based on individual typography angle (ITA) scores, calculated using the values of L* and b* measured by skin reflectance and based on cut-points: very fair ≥ 55°> fair ≥ 28°> olive/brown/dark (19).

Hair, eye and skin color categories

There was excellent agreement between self-reported and clinical assessment of eye color (kappa, κ =0.78); and moderate agreement for hair color (κ =0.46) (Table 2). Agreement between self-reported skin color and spectrophotometer-derived measurements was weak (κ =0.12) when using the pre-specified ITA category cut-points. Participants tended to under-report blonde/fair and red hair, very fair skin, and blue/grey eyes, compared to clinical assessment.

Table 2. Self-report versus clinical assessment of hair, eye and skin color.

The grey-shaded boxes indicate perfect agreement between the self-reported and clinical assessment of phenotype.

The boxes surrounded by dashed lines indicate that the participant over-estimated their pigmentation phenotype.

The boxes surrounded by straight lines indicate the participant under-estimated their pigmentation phenotype.

¹Clinically-measured skin color was based on reflectance spectrophotometry L* and b* values converted to individual typography angle (ITA) scores, and categorized according to published ITA cut-points: very fair ≥ 55°> Fair ≥ 28°> Olive/Brown/Dark (19).

Predictors of disagreement for hair, eye and skin color

There were no significant differences in kappa values for hair, eye or skin color when stratified by case-control status. Men and women had similar agreement for eye and hair color, but skin color agreement appeared slightly higher for women (κ =0.17) than for men (κ =0.07). We further examined predictors of disagreement in a log-binomial regression model adjusted for age, sex and city of recruitment (Supplementary Table S1). Again, we observed less disagreement between self-reported and spectrophotometer-measured skin color for women compared with men (probability ratio 0.81, 95% CI 0.72-0.90). Agreement for eye color (p for difference = 0.01) and, to a lesser degree skin color (p for difference = 0.05), was worse for participants from Brisbane (lower latitude) than for those living in Sydney and Melbourne (higher latitudes). Age, ethnicity, country of birth, residence-based socio-economic disadvantage level, education, case-control status, family history of melanoma in a first-degree relative and nevus counts were not significantly associated with disagreement (Supplementary Table S1). Disagreement for skin color was not a predictor of disagreement for eye or hair color, and vice-versa.

Skin reflectance correlations with self-reported skin color

Table 3 shows descriptive statistics and correlations for skin reflectance values (ITA scores, L*, b* and a* scale values) on the upper inner arm compared to self-reported skin color categories. Self-reported skin color was negatively correlated with ITA scores and L* scale values, and positively correlated with b* and a* scale values; all consistent with increasing pigmentation being objectively detected in the skin for people self-reporting darker-colored skin. The strength of the correlation was moderate for ITA scores (r_s -0.37), b* scale values (r_s 0.34), and L* scale values (r_s -0.31) and weak for a* scale values (r_s 0.17). Correlations were similar or slightly lower when the outer arm values were used instead of the inner arm values (data not shown). The difference in b* values between the inner and outer arm measurements, which has been used as an indicator of the degree of tanning (18), was not correlated with skin color (r_s 0.00) in our sample.

Table 3. Self-reported skin color compared with objectively-measured skin reflectance by spectrophotometry.

Self-reported skin color	Objectively-measured skin reflectance on the upper inner arm
	ITA score		L* scale		b* scale		a* scale
	Median	25th-75th centile	Median	25th-75th centile	Median	25th-75th centile	Median	25th-75th centile
Very fair	60.1	56.5-63.4	69.2	68.1-70.4	11.0	10.0-12.2	6.0	5.3-7.0
Fair	56.1	51.4-60.6	68.0	66.3-69.4	11.9	10.7-13.4	6.6	5.6-7.5
Olive/dark	49.9	42.2-55.2	66.4	63.9-68.2	14.0	11.9-15.8	7.1	5.9-8.3
Spearman rs	-0.37 (-0.32 to -0.42) p<0.001		-0.31 (-0.26 to -0.37) p<0.001		0.34 (0.28 to 0.39) p<0.001		0.17 (0.11 to 0.23) p<0.001

As b* values have sometimes been used as an indicator of tanning, and a* values for erythema, we also compared these skin reflectance values with self-reported ability to tan and propensity to burn, respectively. The r_s for self-reported ability to tan was 0.41 (95% CI 0.36 to 0.46) p<0.001 for inner arm b* values and 0.33 (95% CI 0.27 to 0.38, p<0.001) for outer arm b* values. There was poor correlation between a* values and self-reported propensity to burn: r_s = 0.13 (95% CI 0.07 to 0.19, p<0.001) for inner arm values and 0.04 (95% CI -0.02 to 0.11, p=0.17) for outer arm values.

Nevus counts

Figure 1 box-and-whisker plot shows moderate agreement between clinical total body nevus counts and self-reported nevus density (none, few, some, many) categories (r_s = 0.43, 95% CI 0.38-0.49). There was better distinction between the ‘some’ and ‘many’ categories than for the ‘none’ and ‘few’ categories: the median clinically-measured nevus counts for each self-reported category were 50 (none), 86 (few), 170 (some), and 254 (many). There were no significant differences by case-control status or sex.

Figure 1.

Box-and-whisker plot showing agreement between clinically-measured whole-body nevus counts (2+ mm) and self-reported nevus density descriptive (none, few, some, many) categories (Spearman r_s = 0.43, 95% CI 0.38-0.49, p<0.001).

When self-reported counts of moles on the back were compared with clinically measured nevus counts on the back, the intraclass correlation coefficient (ICC) was 0.36 (95% CI 0.30 to 0.41) (Figure 2). When nevus counts on the back were categorized into quartiles and compared, the κ_w was 0.28 (95% CI 0.23 to 0.32). Agreement for nevus counts on the back did not significantly differ by case-control status or sex. On average, participants under-reported their nevus counts by more than half: the median number of nevi recorded on the back by self-assessment was 13 for cases and eight for controls; whereas clinicians recorded a median 38 nevi for cases and 19 for controls.

Figure 2.

Plot comparing self-reported and clinically measured total nevus counts on the back. Intraclass correlation (ICC) = 0.36, 95% CI 0.30 to 0.41.

Agreement between clinically measured nevus counts on selected anatomical sites

There was generally good agreement when clinically-measured total-body nevus counts were compared with site-specific nevus counts, particularly for the whole back of the body and both arms (Table 4).

Table 4. Intraclass correlations (ICC) comparing clinically-measured whole-body nevus counts (2+ mm) with site-specific nevus counts.

	Males		Females
Body Site	ICC	95% CI	ICC	95% CI
Posterior body1	0.87	0.84-0.89	0.84	0.82-0.86
Back	0.46	0.38-0.53	0.31	0.23-0.38
Both arms	0.49	0.42-0.56	0.57	0.52-0.62
Right arm only	0.26	0.17-0.35	0.32	0.24-0.39
Both thighs	0.33	0.24-0.42	0.40	0.33-0.47

¹The classification of the different body sites is shown in Supplementary Figure S1.

Discussion

An important finding of this study is that young adults living in a country with relatively high levels of ultraviolet radiation (UV) and elevated melanoma incidence rates tend to under-report their nevus counts and skin, hair and eye pigmentation phenotype. Young people are a key target group for preventive efforts as they tend to have excessive UV exposure and slow adoption of sun protective behaviors (20, 23, 24). Awareness of individual risk may encourage improved prevention and early detection of melanoma; however, these results show that young people have sub-optimal accuracy when assessing important risk characteristics including nevus numbers and pigmentation. This has implications for the likely success of targeted public health programs that rely on self-assessment of these factors. There was no evidence that case and control participants differed in their level of agreement, which suggests that differential misclassification of phenotype characteristics is unlikely to have occurred in this case-control study, and thus minimizing the potential for risk estimates to be biased. However, there was some evidence that women had better accuracy than men at reporting their own skin color, and that participants from lower latitudes had better agreement for eye color and, to a lesser degree skin color, compared with participants from higher latitudes.

Our study found excellent agreement for eye color and moderate agreement for hair color when comparing self-report with dermatologist assessment. A study of 114 participants aged 40-69 years from a cohort in Queensland, Australia, reported moderate agreement for eye color (κ = 0.58) and hair color (κ = 0.63) (25). In the United Kingdom, Melia et al (26) also found moderate agreement for self-report versus dermatologist assessment of hair color.

There was fair to moderate agreement between self-reported skin color and continuous measures of spectrophotometer-derived skin reflectance. However, when skin color reflectance values were categorized into very fair, fair and olive/dark categories the agreement was poor (κ =0.12). We used published cut-points for ITA scores (19, 20) but a range of different reflectance values and cut-points have been used in other studies (23, 27, 28) to classify skin color. Our study sample had a fairly homogenous skin type and a narrow range of skin reflectance values, which probably contributed to the poor agreement with self-reported skin color, as has been noted in other studies conducted among predominantly white populations (23). Studies with a greater diversity of skin color including non-white participants have demonstrated better correlation between self-report and reflectance spectrophotometry (15, 20).

Of the people who classified themselves as having olive or dark skin color in our study, only 21% had the same classification based on spectrophotometer assessment, whereas 54% were classified by spectrophotometry as having ‘fair’ skin and 25% as having ‘very fair’ skin. Similarly, a New Zealand study of 289 university students (20) found that 77% of participants who classified themselves as having medium skin color were in the fair skin color category when measured by spectrophotometry. In a convenience sample of 341 adults in Queensland, Australia, 47% of those categorized as fair skin color by spectrophotometry reported their own skin color as medium or olive, and agreement was summarized by a κ_w of 0.11 (95% CI 0.01–0.21) (23). Studies that have compared self-report with dermatologist (but not spectrophotometer) assessment of skin color have also found fair agreement (26, 29). However, contrary to our results of higher skin color agreement for women than men, Melia and colleagues reported better agreement for men (OR 3.1, 95% CI 1.4–6.8) (26).

Number of nevi is one of the strongest risk factors for melanoma (8) but is often problematic for people to accurately report (30), partly due to confusion with solar lentigines, freckles and seborrhoeic keratoses (11). Our study found fair to moderate agreement between self-reported and clinically assessed nevus counts. Participants underestimated their nevus counts on the back by more than half. Despite this under-reporting, our data demonstrate that participants are able to adequately rank themselves; the correlation was 0.36 (95% CI 0.30 to 0.41) comparing self-report and clinician counts on the back (κ_w 0.28 when categorized into quartiles) and 0.43 (95% CI 0.38 to 0.49) when participants classified themselves into descriptive (none, few, some, many) categories of whole-body nevus density. However, discrimination was modest between the ‘none’ and ‘few’ categories and thus consideration should be given to combining these categories in future risk analyses. Other studies that have used similar diagrams for classifying self-reports of whole-body nevus density have found comparable agreement with clinical assessment (25, 31). Interestingly, in the Queensland study (25), the number of clinically measured nevi corresponding to each descriptive nevus category was less than half that recorded for the same categories in our study. This difference might be due to the younger age of participants in our study since nevi start to regress after early adulthood (32), differences in clinician reporting, or because our sample also included melanoma cases. Nevertheless, it highlights that caution should be used when combining data for self-reported nevus categories across studies, because although the data are suitable for ranking people according to their nevus count, they should not be used to estimate absolute number of nevi. In contrast to the under-reporting of nevi in our study, a study of university students in Germany found that participants counted on average 14% more nevi on the arms than the examiners, although the discrepancy was larger for non-medical students and those with fair skin (33).

Direct comparisons with previous validation studies are hampered by differences in study design, nevus-counting protocols, nevus definition and type, statistical methods, the sex and age of participants, and whether counts were conducted on the whole body or isolated to specific anatomical sites. Other validation studies of self-reported whole body nevus counts have generally reported poor (34) or fair agreement (29), although agreement tends to be higher when classified into fewer categories (35, 36). We did not find any differences in level of agreement for nevus counts by sex or case-control status, which is consistent with some studies (31, 33, 34), however Lawson and colleagues (11) found that men were more likely than women to under-report large nevi.

For our secondary aim, we examined agreement between clinically-measured nevus counts on the whole body to that from selected anatomical regions. This comparison is relevant for epidemiologic studies, which for reasons of time, cost and feasibility, often use localized nevus count as a proxy measure of whole body nevus counts. We found excellent agreement between nevus counts on the entire back side of the body and total body nevus count. However, for studies where an extensive nevus count on the body is impractical, our results suggest that measuring nevus counts on both arms would be the best alternative, consistent with previous reports (37, 38). Nevus distribution is related to age, sex and sun exposure, however sex had little impact on anatomic site as a proxy measure for whole-body nevus count.

Strengths of our study include the large population-based sample of young adults, and comprehensive measures of pigmentation and nevus phenotype. We used spectrophotometer assessment of skin reflectance, which is free from observer bias and errors associated with recall. The clinical phenotype assessment was conducted by experienced dermatology trainees who were trained on the study protocol, and they used standardized clinical measurement tools including wig hair swatches, eye photographs and a plastic size standard for counting nevi. However, clinician measurement still has an element of subjectivity and this may have influenced our results. In addition, we did not examine inter-examiner reliability. In other studies, inter-examiner reliability has been found to range from moderate to excellent between dermatologists (11, 39), but was poor when comparing experienced dermatologists with non-medically trained interviewers (39). Another limitation was that the categories for measuring and classifying nevus and pigmentation phenotype did not always match perfectly between self-report and clinical assessment, and we had few self-reported nevus measures.

These results have implications for sun-related public health programs targeted at younger people. To improve the likely success of future public health campaigns, it would be worthwhile to develop and test novel ways of improving self-assessment and simple, objective measurement of nevus and pigmentation phenotype. For example technological advances would now enable the development of online self-assessment systems showing standardized photographs for comparison, or mobile phone apps that could detect skin color from real-time photographs and count freckles and nevi based on in-built algorithms. It is also of interest to determine to what extent people who under-estimate their nevus and pigmentation phenotype also under-estimate their susceptibility to melanoma, and how this correlates with sun-related behaviors. In the meantime, public health campaigns targeted at younger people should not rely too heavily on self-identification of high risk status based on phenotypic characteristics.

Abbreviations

OR

odds ratio

CI

confidence interval

ITA

individual topography angle