Efficacy of Skin Self-Examination Practices for Early Melanoma Detection

Melanoma incidence in the United States has increased ∼200% from 1973 to 2002 (1, 2); similar increases have been reported in other countries (3). Given that melanoma outcomes differ strongly by tumor thickness at diagnosis, maximizing early detection is critical (4). Melanomas are most commonly detected by patients themselves (5, 6); therefore, considerable effort has been directed to develop interventions to increase the proficiency and practice of skin self-examination (SSE; refs. 7-9).

A minority of individuals practice regular SSE, with estimates of ∼20% to 25% in the United States and Australia (9, 10). The efficacy of SSE in detecting thinner melanomas and reducing morbidity and mortality has not been extensively studied (1, 11). In 1996, Berwick et al. reported that SSE could reduce melanoma mortality by as much as 63% (12). A 2003 study found that regular performance of SSE was associated with a significantly reduced likelihood of having tumors >1 mm thick at diagnosis [covariate-adjusted odds ratio (OR), 0.65; 95% confidence interval (95% CI), 0.45-0.93], although details regarding the thoroughness and frequency of SSE were not obtained (13). Improved understanding of the effectiveness of SSE has been hampered by variable study definitions of SSE, including the number or percent of body sites examined and the frequency and method of examination (14-17) and the small number of studies examining the reported benefits of techniques to supplement SSE, such as the use of photographs (8, 18, 19).

In particular, little attention has been given to examining specific SSE behaviors that are associated with early detection and tumor thickness at diagnosis. Recognizing this, we developed and administered a questionnaire to evaluate the specific skin examination-associated behaviors and practices that newly diagnosed melanoma patients had engaged in before their diagnosis. We examined the effect of three SSE practices on melanoma severity (as measured by tumor thickness) at diagnosis. Our aim was to obtain a more accurate, detailed description of the associations between different SSE practices and successful early detection of melanoma.

Approval for melanoma case ascertainment was obtained from the institutional review boards of Stanford University Medical Center, Veterans Affairs Palo Alto Health Care System, and University of Michigan. Patients were surveyed in the melanoma clinics of these institutions from May 17, 2006 to May 31, 2008. Individuals ages ≥18 years with a recent diagnosis of invasive cutaneous melanoma were eligible for this study. Patients with in situ melanoma or mucosal, genital, perianal, and ocular primary melanoma were excluded.

Patients were surveyed consecutively at Stanford University Medical Center and Veterans Affairs Palo Alto Health Care System. Because of the high proportion of patients with thin melanoma at University of Michigan, we attempted to survey all University of Michigan patients with melanoma >2 mm and randomly selected one of every three patients with melanoma ≤2 mm. Eligible patients were contacted by phone or approached by a study coordinator in the clinic before their initial consultation to discuss participation. If interested, they were asked to complete a self-administered survey before the consultation. Among all cases diagnosed at the three sites during this study period, only 14.3% of patients refused to be surveyed. Overall, 342 patients agreed to be surveyed. Twenty-one of these patients were excluded from these analyses because they did not have complete education or SSE practice data; 321 were included. All questionnaires were completed within 3 months of melanoma diagnosis based on the date of biopsy.

Data obtained on the questionnaire that was used in this analysis included patient demographics, education, previous history of melanoma, and information on SSE-associated practices within the last year. We sought to develop a comprehensive assessment of optimal SSE practice. SSE was measured in three ways: (a) routine examination of 13 specific body areas, (b) frequency of mole examination, and (c) use of a picture aid illustrating a melanoma tumor. The first measure asked patients to identify which of the following areas of their skin they routinely examined: scalp, face, neck, shoulder, front of arms, back of arms, chest, stomach, upper back, lower back, front of legs, back of legs, and bottom of feet. Patients were categorized as routinely examining their skin on 0, 1 to 4, 5 to 8, or 9 to 13 of these body areas. This measure was also dichotomized by whether patients routinely examined their skin on any areas at all. The second SSE-related measure assessed the frequency with which patients carefully examined all of their moles (including the ones on their back), categorized as every 1 to 2 months, every 6 months, every year, and never. The third item assessed whether patients ever used a picture of melanoma in a poster or handout to help them look at their skin.

In addition to these three measures, a series of composite variables combining data from two and all three of the SSE measures was also created. One composite variable constructed after initial data analysis assessed thoroughness of SSE by combining data on whether patients routinely examined their skin and whether they used a melanoma picture to assist in SSE. Patients reporting the use of a melanoma picture/poster to aid in SSE were placed in the High SSE thoroughness group regardless of the number of areas they reported routinely examining. Patients who did not use such a picture aid but who did routinely examine their skin on one or more of the 13 evaluated body areas were placed in the Intermediate SSE thoroughness group. Patients who did not use a picture aid and did not check any of their skin routinely were placed in the Low SSE thoroughness group.

Patients were clinically staged at diagnosis according to the American Joint Committee on Cancer 2002 primary tumor characteristics before sentinel lymph node biopsy (20). A dermatopathologist at the academic center confirmed the histologic diagnosis and Breslow depth of all cutaneous melanoma specimens.

We conducted both linear and logistic multivariable regression analyses to assess the effect of SSE practices on tumor thickness. Multivariate linear regressions used log-transformed melanoma thickness as a continuous outcome variable. Log transformation was carried out to correct for the right-skewed distribution of melanoma thickness. Geometric mean tumor thicknesses were calculated by exponentiating the β parameters obtained from each model (21, 22). Ratios of these geometric mean tumor thicknesses (and 95% CIs) were created for each SSE measure by comparing the geometric mean for each category of the SSE measure to the geometric mean of the least frequent or thorough SSE category in each measure. Multinomial logistic regression (generalized logit) models, with tumor thickness categorized as ≤1, 1.01 to 4, and >4 mm, were used to calculate ORs (and 95% CI) for having thicker tumors versus a thinner tumor (≤1 mm) by category of the SSE measures (23). In these models, each category of a SSE measure was compared with the most frequent or thorough SSE category in the measure.

Decisions to include covariates as potential confounders in the models were based on a significant association between the covariate and the independent (SSE) and dependent (tumor thickness) variables and either the statistical significance of the covariate’s inclusion in the models or a change of 10% in the size of the SSE measure’s β parameter on inclusion of the covariate. Potential confounders examined included sociodemographic variables, previous history of melanoma, and histologic subtype. Education was dichotomized as 4-year college degree or postgraduate degree versus less than a college degree. Statistical interaction between SSE measures and covariates was assessed using partial F tests and by the significance of interaction terms entered into the models (24). Analyses were conducted using SAS 9.1. All tests of significance were two-sided; significance was defined as P < 0.05.

Table 1 presents demographic information on the 321 patients included in these analyses. Patient age ranged from 18 to 99 years; 59% of respondents were male. Approximately 95% of the sample were White, and 46% of the sample had received a 4-year college degree or further education. Median tumor thickness was 1.30 mm; mean thickness was 2.08 mm (SD, 2.47 mm). Most patients (53%) self-detected melanoma compared with detection by physicians (19%), spouse/partners (16%), and family members/friends/other (12%).

The numbers of patients engaging in the three SSE practices measured are also presented in Table 1. Approximately 83% of patients said they routinely examined the skin on any areas of their body, including 26.8% who reported that they routinely examined their skin on at least 9 of 13 possible areas. About 50% of patients reported that they carefully examined their moles (including the ones on their back) at least once per year; 15% examined them every 1 to 2 months. Finally, 29% of patients reported that they had ever used a melanoma picture as an aid in performing SSE.

Table 2 shows the ratios of geometric mean tumor thicknesses calculated by category of the demographic, SSE, and skin characteristic variables. Models were adjusted for age, gender, and education level; entry of other variables (including tumor histology and previous history of melanoma) did not change the pattern of results for any model.

The SSE measure of the number of body areas on which a patient routinely examined his/her skin showed an inconsistent association with tumor thickness. Patients reportingthat they routinely examined their skin on 1 to 4, 5 to 8, and 9 to 13 areas of their body (out of 13) had geometric mean ratios of 0.68, 0.73, and 0.77 when compared with those reporting they did not routinely examine any of their skin. When this measure was dichotomized, patients reporting that they routinely examined their skin on at least some of the measured body areas had an adjusted geometric mean tumor thickness ratio of 0.73 (95% CI, 0.65-0.82) when compared with patients who did not routinely examine any of their skin.

How often a patient carefully examined their moles was not associated with decreased tumor thickness. In contrast, use of a melanoma picture as an aid in SSE was significantly associated with reduced tumor thickness. Patients who had used a picture as a SSE aid had an adjusted geometric mean tumor thickness ratio of 0.75 (95% CI, 0.62-0.92) compared with those who never had.

The majority of composite variables created by combining data from two or more SSE measures were less strongly associated with tumor thickness than the SSE measure assessing whether or not a patient used a melanoma picture to assist in SSE (data not shown). The composite “SSE thoroughness” variable, however, had the strongest association with tumor thickness of all measures assessed. There was a statistically significant, graded association between increasing SSE thoroughness and reduced tumor thickness. Patients in the Intermediate SSE thoroughness category (e.g., those who did not use a melanoma picture to aid SSE but did routinely examine some areas of their skin) had an adjusted geometric mean tumor thickness ratio of 0.71 (95% CI, 0.62-0.82) when compared with those in the Low thoroughness category (e.g., those who did not use a melanoma picture aid to SSE and did not examine any of their skin routinely). Patients in the High thoroughness category (e.g., patients who used a melanoma picture to aid in SSE) had an adjusted geometric mean tumor thickness ratio of 0.58 (95% CI, 0.48-0.70) when compared with those in the Low group.

Table 3 presents the results of age-adjusted multinomial logistic regressions models that provide the OR for having a melanoma tumor thickness at diagnosis of 1.01 to 4 mm (intermediate) versus ≤1 mm (thin) and for having a tumor thickness of >4 mm (thickest) versus ≤1 mm. The SSE variable evaluating how many areas of skin were routinely examined was not associated with an increase in the likelihood of having thicker tumors, except for in patients who reported they did not routinely examine any areas of their skin. When compared with patients who did routinely examine at least some of their skin, these patients had increased odds of having a melanoma of >4 versus ≤1 mm in thickness (OR, 2.78; 95% CI, 1.17-6.63); their odds of having a melanoma of between 1 and 4 mm was modestly increased (OR, 1.71; 95% CI, 0.88-3.32).

As in the linear regressions, how often patients carefully examined their moles was not associated with tumor thickness. Patients who had never used a melanoma picture to help with SSE in the past had higher odds of having a thicker tumor when compared with those who had (OR, 2.02; 95% CI, 0.86-4.74 for >4 versus ≤1 mm tumor thickness). The composite Thoroughness variable was also associated with increased odds of a thicker tumor: when compared with patients in the High SSE thoroughness category, patients in the Intermediate and Low categories were 1.54 (95% CI, 0.63-3.77) and 4.66 (95% CI, 1.56-13.93) times more likely to have a >4 mm tumor. Patients in the Low thoroughness category were also more than twice as likely to have an intermediate thickness tumor (1.01-4 mm) as those in the High category (OR, 2.22; 95% CI, 1.02-4.84).

Age was the only variable that consistently met the criteria for a potential confounder (as described in Materials and Methods) in the logistic models; inclusion of other potential confounders did not meaningfully change the pattern of any results.

Reported rates of SSE vary not only by the population in which SSE practices are assessed but also according to how these practices are defined. Studies differ widely in their definitions of SSE and typically do not break down the overall practice of SSE into specific behaviors that can be more reliably measured. In a random digit dialing survey of Rhode Island adults, Weinstock et al. found that 58% of subjects reported examining their skin “deliberately and systematically,” but only 9% met their criteria for thorough self-examination. This criteria required examination “always” or “almost always” of each of eight skin areas (arms and face, chest and front of legs, side of body, back of legs, upper back and tops of shoulders, sides of legs and bottoms of feet, middle and lower parts of back, and back of thighs), except one, which could be examined at least “sometimes” (14). Carli et al. showed that 45.8% of their subjects in an Italian cohort reported performing SSE (20.4% regularly and 25.4% occasionally), although SSE criteria were not defined and were determined by the individual patients themselves (13). In an Australian telephone survey of 3,110 residents, Aitken et al. defined SSE as having “deliberately checked the skin on your whole body,” including “skin front and back,” and found that only 25.9% of participants had performed SSE within the last year (10).

Our study is unique in its comparison of different SSE practices in recently diagnosed melanoma patients and the linkage of performance of each practice to melanoma tumor thickness. The three SSE practices had different rates of performance; whereas 82% of subjects routinely examined the skin on at least some of the 13 body areas evaluated, only 49% carefully examined all their moles at least once a year, and even fewer (29%) used a picture or other visual aid in their practice of SSE. Such variation may point to actual differences in self-examination behaviors: individuals who routinely examine most of their skin may not necessarily examine all of their moles regularly and vice versa. Differences in response to the three questions might also reflect that these measures evaluated different aspects of SSE practice: the first and third measures asked about thoroughness or completeness of SSE practice (how many of a possible 13 body areas were routinely examined and whether a picture was used as an aid), whereas the second measure focused on SSE practice frequency (how often were moles examined). Much lower rates for the SSE measure inquiring about use of a melanoma picture aid are likely due to the fact that such use is prone to be a component of only a carefully performed SSE, which is not as common (14).

The three SSE measures also had different associations with tumor thickness. Whether or not patients routinely examined the skin on at least some of the 13 body areas recorded was associated with decreased tumor thickness (although the actual number of body areas patients routinely examined was not clearly associated with mean thickness). Frequency of mole examination, however, showed little association with tumor thickness. Use of a melanoma picture to assist with SSE was also associated with thinner tumors at diagnosis. A composite measure assessing SSE thoroughness by combining the first and third measures was the strongest predictor of tumor thickness. As this composite measure was constructed retrospectively, after initial analysis of the data, these results require confirmation in future studies. Nevertheless, the overall pattern of results suggests that studies evaluating SSE practice using only one measure may not accurately predict its effect on melanoma outcomes.

As noted, although patients who routinely examined their skin on some of the measured body areas had significantly thinner tumors than those who did not, examining a higher number of body areas was not associated with a decrease in tumor thickness. The proportion of skin routinely examined may simply not be a good surrogate for efficacy of SSE. One possible explanation is that looking at a few sites that have a high index of suspicion may be as effective as looking all over one’s body. Another potential explanation is that patients who examine more body sites may not be focusing on areas where melanomas are more likely to occur, that is, back of the body. Physicians also play a role in examining body sites that are harder for patients to see (25).

Our findings may also have been affected by errors and biases potentially inherent in responses to these measures. Questions requiring self-report of the frequency or completeness of SSE practices may have been prone to more error than specific yes/no questions evaluating concrete SSE practices (e.g., use of a melanoma picture to assist with SSE), as the former questions forced individuals to accurately quantify their behaviors or place them on a continuum. Such response errors were likely nondifferential and thus would have biased results toward the null, providing one possible explanation for the greater predictive strength of the more simple yes/no SSE measures. Responses may also have been subject to differential bias, that is, a patient diagnosed with a thin melanoma associated with a good prognosis may have recalled his/her SSE practices differently compared with a patient diagnosed with a thick, more deadly melanoma. Again, differential bias would be less likely to affect responses to simple, objective yes/no questions. Questions inquiring about objective SSE practices may prove more effective in documenting overall SSE performance in future studies.

A secondary set of analyses examining patient characteristics that contribute to an increased risk of having thickest (>4 mm) versus thinnest (≤1 mm) melanoma showed that the 17% of patients who reported never examining their skin on any body areas had a higher likelihood of having the thickest tumors. Similarly, patients in the Low category of the composite SSE thoroughness measure created had strongly increased odds of having thick tumors. Thus, certain risk factors such as never examining one’s skin not only have an effect on mean/median melanoma tumor thickness but also appear to significantly increase the odds of having very advanced melanoma at diagnosis.

Limitations of this study include the fact that all SSE practices were patient-reported and not confirmed by an outside source. As noted previously, recall of SSE practices might also have been affected by the patients’ recent melanoma diagnoses. It was also not possible to assess any specific reasons patients may have had for engaging in more time-intensive SSE practices (e.g., using a picture as a SSE aid); some might have done so due to a specific concern such as an unusual mole or a doctor’s previous observation. Strengths included completion of all surveys within 3 months of diagnosis, three separate measures of SSE practice, and the association of these practices with dermatopathologist-confirmed tumor thickness.

In conclusion, our results suggest that SSE is composed of several distinct behaviors and that consideration of these individual behaviors is important in determining SSE efficacy. The SSE practices showed different associations with melanoma thickness, although a composite variable combining information from two of these measures was the best predictor of tumor thickness. Evaluating multiple aspects of SSE (alone or in combination) may increase the predictive ability of studies examining the effect of SSE on outcomes such as tumor thickness. Finally, differences in associations between the three SSE measures and melanoma thickness suggest a complexity in SSE practices that should be addressed in future studies.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Mitzi Rabe, R.N., for her role as study coordinator at the University of Michigan site.