The Impact of Physician Screening on Melanoma Detection FREE

Cutaneous melanoma continues to be a significant public health problem, with the incidence rising rapidly in the United States and worldwide.¹ Despite much progress and novel approaches in the treatment of advanced melanoma, survival and prognosis has not notably improved over the past 3 decades.^2- 3 The overall 10-year disease specific survival for advanced-stage melanoma is below 50% vs 100% for in situ disease and 88% for stage IA melanoma (ie, <1 mm, without ulceration).⁴ Therefore, early recognition of melanoma is the best means to ensure a favorable prognosis.

A favorable prognosis for melanoma relies heavily on a timely diagnosis. Fortunately, most cutaneous melanomas are visible on the skin surface and thus are amenable to early detection. Important strategies aimed at improving early diagnosis of melanoma include physician-based screening, educating patients on melanoma recognition, and developing better imaging tools. Patient involvement in detection is imperative because patients have the most opportunities in examining their own skin, and in fact, previous studies have shown that most melanomas are self-detected.^5- 8 Studies have demonstrated that patient-detected lesions are significantly thicker than those detected by a dermatologist.^6- 10 However, given the lack of randomized controlled trials demonstrating a positive impact of population-based screening on mortality, the current position of the US Preventive Services Task Force (USPSTF) is that there is “insufficient ” evidence to recommend screening for melanoma.¹¹ Working on the premise that physician-based screening and patient self-screening are vital in the detection of early melanoma, we compared melanoma characteristics in patients new to our practice vs established patients in the pigmented lesion clinic (PLC) at Memorial Sloan-Kettering Cancer Center (MSKCC). Specifically, we explored histologic tumor characteristics and patterns of melanoma detection in both groups.

We retrospectively reviewed patient records and biopsy logs of 2 pigmented lesion specialists (A.C.H. and A.A.M.) at MSKCC from January 1998 through December 2008. The study included all lesions with a histopathologic diagnosis of melanoma or nevus that were diagnosed by A.C.H. and A.A.M. during the 10-year period. All biopsy specimens were reviewed by dermatopathologists with expertise in the evaluation of melanocytic tumors. We excluded from our study melanomas or nevi that were diagnosed after a biopsy performed at an outside facility and those patients who presented for a second opinion or management. Using these criteria, we identified 527 melanomas in 394 patients.

The study population is from the PLC at MSKCC, a tertiary referral center for patients at high risk for developing melanoma. As part of standard practice, each patient visit at the PLC is accompanied by a total body skin examination. All clinically suspicious lesions get evaluated with a handheld dermoscope. In individuals with many nevi, we also perform total body photography (TBP) that is used by both clinicians and patients to assist in the detection of new and/or changing nevi. Nevi that deviate from the benign dermoscopic patterns are subjected to short-term mole monitoring via sequential digital dermoscopy with a standard interval follow-up of 3 months. During the follow-up examination, if the clinical or dermoscopic features suggest malignancy, that lesion is removed for biopsy. Also during visits to the PLC, patients are provided with brochures with illustrations on how to perform skin self-examination (SSE) and are taught how to identify potential melanoma lesions using “ABCD ” (A  =  asymmetry; B  =  border irregularity; C  =  color variegation; D  =  diameter >6 mm) and “ugly duckling ” methods. Frequency of follow-up is individually based on tumor and patient characteristics. In general, all patients are seen in follow-up once at 6 months and yearly thereafter. Patients with a new first primary melanoma and new subsequent primary melanoma are seen in follow-up every 3 months for 1 year, then every 6 months for 4 years and yearly thereafter.

For the purposes of the study, patients were divided into 2 main groups —established patients, defined as patients who have received professional services in a PLC at MSKCC for at least 3 months, and new patients, defined as patients new to our practice. The following data were collected for each patient: age at melanoma diagnosis, sex, personal history of melanoma, presence or history of clinically diagnosed and/or histologically confirmed dysplastic nevi, number of nevi (based on total body examination), history of nonmelanoma skin cancer, and family history of melanoma. For each patient, the number of primary tumors was recorded. The dermatology clinic notes of the 394 patients were reviewed, and for each histologically proven melanoma, data were abstracted regarding whether melanoma was diagnosed in a new patient or established patient; whether a patient had undergone TBP prior to melanoma diagnosis, who first identified a suspicious lesion (patient [performing SSE], dermatologist, or other); clinical reason for biopsy (new lesion, change noted in an existing lesion, or concerning or atypical lesion); and lesion signs and symptoms of self-detected melanomas. To report melanoma location, we broadly subdivided anatomic areas into head and neck, anterior trunk, posterior trunk, upper limbs, and lower limbs.

Pathology reports of 527 melanomas were reviewed and the following tumor histopathologic characteristics were abstracted: Breslow depth of invasion (also referred to as tumor thickness measured in millimeters); tumor histogenetic type (superficial spreading, nodular, lentigo maligna melanoma, acral lentiginous melanoma or other); Clark level (also referred to as level of invasion according to depth of penetration of the dermis); mitotic index (expressed as the number of mitoses per square millimeter); and presence or absence of ulceration, vascular invasion, lymphatic invasion, regression or coexisting nevus. Unfortunately, the tumor histogenetic type was not available for most melanomas (76%), and as a result, we do not report melanoma subtypes.

Comparisons between melanomas diagnosed in new patients vs established patients in relationship to Breslow tumor thickness and histologic tumor characteristics were completed. To characterize melanomas diagnosed with and without photographically assisted follow-up, patients who had TBP prior to their melanoma diagnosis were identified. Comparisons of dermatologist-detected melanomas with self-detected lesions in both new and established patients were completed. The cause for patient concern in self-detected melanomas was identified. Benign to malignant biopsy ratio of melanocytic lesions was estimated by dividing the total number of biopsied melanocytic nevi by the total number of melanomas diagnosed between 1998 and 2008. All lesions with a histopathologic diagnosis of a melanocytic nevus were included in the “benign ” category, while all histologically proven melanomas were included in the “malignant ” category.

Descriptive frequencies, means, geometric means, and medians were used to describe the study population. Graphical assessments were used to assess the distribution of study variables. Because tumor thickness was not normally distributed, a transformation using the natural log was performed. The differences in prognostic factors between the respective groups were evaluated using the χ² test and the 2-sample t test. Linear regression models were used to assess the association between log tumor thickness and melanoma detection patterns while controlling for potential confounding variables. All tests were 2-tailed. All statistical analyses were performed with Stata version 10 (StataCorp, College Station, Texas).

Over the 10 years of evaluation, a total of 527 melanomas were identified in 394 patients. Patient and melanoma characteristics are presented in Table 1. The study population comprised 269 men (68.3%) and 125 women (31.7%), with a mean age of 61.8 years (range, 17-91 years). On total body examination, of 394 patients, 91 (23.1%) had more than 50 nevi, while more than 100 nevi were documented in 64 patients (16.2%). Clinical or histologically confirmed dysplastic nevi were determined in 174 patients (44.2%). A total of 54.6% of study population had a personal history of melanoma (n  =  215). In addition, 73 patients (18.5%) had personal history of multiple primary melanomas ranging from 2 to 8 primary tumors per patient. Among the 394 patients, 183 (46.4%) had a history of nonmelanoma skin cancer. The overall prevalence of positive family history of melanoma was 18.8% (n  =  74).

Characteristics of the melanomas are presented in Table 1. Of the 527 melanomas, 323 (61.3%) were in situ, 171 (32.4%) were invasive, and the remaining 33 lesions (6.3%) were metastatic. The mean and median Breslow depth of invasive melanoma was 0.61 mm (range, 0.1-4.5 mm) and 0.35 mm, respectively. Of 171 invasive melanomas, 143 (83.6%) were thin, with a Breslow thickness less than 0.76 mm, 13 (7.6%) were between 0.76 and 1.5 mm, and the remaining 15 melanomas (8.8%) had tumor thickness greater than 1.5 mm. Among 494 primary melanomas, 134 (27.1%) had coexisting nevus identified on histologic evaluation. Of 494 primary melanomas, 119 (24.1%) were located on upper limbs, 115 (23.3%) on posterior trunk, 98 (19.8%) on head and neck areas, 85 (17.2%) on anterior trunk, and the remaining 77 (15.6%) on lower limbs.

Table 2 compares the histologic characteristics of the primary melanomas discovered in established patients with those discovered in patients new to our practice. Of 494 primary melanomas, 169 (34.2%) were discovered in 154 new patients presenting for skin cancer screening vs 325 (65.8%) that were discovered in 225 established patients during follow-up. Established patients had greater percentage of in situ disease (70% vs 57%; P   <  .001) and thinner invasive melanomas (mean Breslow thickness, 0.45 mm vs 0.82 mm; P  =  .002) compared with melanomas detected in new patients. Of the 99 invasive melanomas diagnosed in established patients during follow-up, 89 (89.9%) were thin, with Breslow thickness of less than 0.76 mm, 6 (6.1%) were between 0.76 and 1.5 mm, and only 4 (4.0%) were thicker than 1.5 mm, 2 of which were diagnosed in patients who had not had interval follow-up for 24 months and 36 months. Of the 4 thick melanomas, 2 had superficial spreading with a nodular component, 1 was desmoplastic, and 1 was nodular melanoma. In new patients, 54 (75.0%) of the invasive melanomas had a Breslow thickness of less than 0.76 mm, while 8 (11.1%) were between 0.76 and 1.5 mm and 10 lesions (13.5%) were thicker than 1.5 mm.

In addition, melanomas discovered in established patients were less likely to have negative prognostic attributes such as ulceration and dermal mitoses when compared with melanomas in new patients. None of the invasive melanomas in established patients were ulcerated, while 9 invasive melanomas (12.5%) in new patients had ulceration identified on histologic examination. No mitotic index of 1 or greater was identified in 84 invasive melanomas in established patients vs 45 invasive melanomas (62.5%) in new patients. Furthermore, 5 invasive melanomas (6.9%) discovered in new patients had mitotic index greater than 6.

Of 354 melanomas in established patients, 110 lesions (31.1%) were discovered in patients who had undergone TBP prior to their melanoma diagnosis, compared with 244 lesions (68.9%) discovered in patients without prior TBP. Persons who had undergone TBP were more likely to have a personal and family history of melanoma. They were also more likely to be younger and have a clinical history of dysplastic nevi. However, the majority of melanomas in both groups were in situ and thin lesions ( <0.76 mm). Also, there were no statistically significant differences in histologic tumor characteristics between melanomas discovered with or without prior TBP (data not shown).

The majority of primary melanomas (n=373 [75.5%) were dermatologist detected, while patients detected 91 (18.4%) (Table 3). The remaining 30 lesions (6.1%) were identified by a nondermatology physician (n  =  12), spouse (female, n  =  8), staff during skin cancer screening (n  =  6), nurse (n  =  2), parent (n  =  1), and a medical photographer (n  =  1). Of 91 self-detected primary melanomas, 52 (57.1%) were in situ and 39 (42.9%) were invasive, while 254 (68.1%) of dermatologist-detected melanomas were in situ and the remaining 119 (31.9%) were invasive. Dermatologist-detected invasive melanomas were thinner than patient-detected lesions (geometric mean tumor thickness, 0.3 mm vs 0.5 mm; P   <  .05).

Dermatologists detected 107 newly diagnosed melanomas (63.3%) in new patients vs 266 (81.8%) in established patients during follow-up visits (Table 4). The majority of self-detected melanomas in new patients were invasive (23 of 41 [56%]), while 34 of 50 self-detected lesions (68%) in established patients were in situ. Geometric mean tumor thickness of self-detected melanomas in established patients was 0.38 mm vs 0.71 mm in new patients. However, after adjustment for age, sex, personal history of melanoma, and total body nevus count, the difference in thickness was not statistically significant for either group (P  =  .43). The geometric mean tumor thickness of physician-detected melanomas in established patients and new patients was 0.30 mm and 0.36 mm, respectively.

The majority of self-detected melanomas were changing lesions (58 of 91 [64%]), 29 [32%] were new, and the remaining 4 melanomas were “concerning ” lesions but were neither new nor changing). Of 91 self-detected melanomas, the data on self-reported signs and symptoms were available for only 60 lesions: 31 in situ melanomas and 29 invasive melanomas. The most common self-reported signs with patient-detected melanomas were color and size change (n  =  14), color change only (n  =  12), and size change only (n  =  9) (Figure 1). Most patient-detected melanomas, which were derived from a new lesion, were asymptomatic (23 of 29 [79%]). In addition, self-detected melanomas, which patients perceived to have arisen from a new lesion, were significantly thinner than those that had arisen from a changing lesion (data not shown). Comparing patient-detected malignant melanoma in situ with invasive melanoma, we found that color and size change (n  =  7) and bleeding and size change (n  =  5) were most commonly reported with self-detected invasive melanomas vs in the cases of self-detected malignant melanoma in situ, color change alone (n  =  12), color and size change (n  =  7), and size change alone (n  =  7) were the most frequent signs (Figure 1). Self-detected melanomas with reported bleeding (n  =  3), rapid growth (n  =  2), and size change and bleeding (n  =  5) were associated with greater tumor thickness. Of 10 self-detected melanomas, in which bleeding was one of the presenting signs, 9 were invasive.

The number of nevi biopsies has decreased over time (Figure 2). The overall benign to malignant biopsy ratio over the 10-year period for both physicians was 5.4:1, which translates to approximately 5 nevi biopsied to 1 melanoma diagnosed. In addition, the majority of biopsied nevi demonstrated significant architectural or cytological atypia on histologic evaluation.

The purpose of this study was to compare melanomas in established patients at the PLC vs patients new to our practice. Our results suggest that physician screening leads to detection of thinner melanoma. Established patients had a greater percentage of in situ disease, had thinner invasive melanomas, and were less likely to present with negative prognostic attributes such as ulceration and dermal mitoses compared with new patients. The importance of physician screening is also aided by the observation that the majority of melanomas in our cohort were physician detected. Importantly, the detection of early melanoma was not the result of the increase in the biopsy of benign lesions.

Identifying thinner melanoma in high-risk patients is analogous to the observation that second and third primary tumors in patients with multiple primary melanomas are often thinner than the first primaries.^8,12- 13 Although we recognize that high-risk patients may have more frequent physician skin examinations and may be more vigilant in performing SSE, we strongly believe that the PLC setting contributes to earlier detection of melanoma in our cohort. In addition to total body skin examination, in our practice we rely heavily on dermoscopy and TBP to aid the skin examination. Although discussing the diagnostic strength of dermoscopy in the evaluation of skin lesions and the utility of TBP in patients with many moles is beyond the scope of this research, others have reported that integrating TBP and dermoscopy into patient surveillance improves the accuracy, specificity, and sensitivity for the detection of early melanoma.^14- 19 In fact, in our experience we had many cases of early melanomas that did not exhibit “classic ” clinical features of melanoma and appeared benign by the naked-eye clinical examination alone but were noted as new or changed lesions compared with baseline TBP findings and had features suggestive of melanoma on dermoscopic evaluation. In addition, making digital photographs more widely available to the patient may help the patient identify new and/or changing moles more efficiently.^20- 21 Importantly, using dermoscopy and TBP, we have not experienced the increase in the excision of benign nevi. To the contrary, we have observed a decreased number of nevi biopsies during the 10-year study period. It is interesting to note that for both physicians, working independently, the mean benign to malignant ratio was 5.4:1. This observation also lends credence to an opinion expressed by A.A.M. and A.C.H. that the rate of biopsies performed in patients with and without TBP may initially be similar, diverging from each other over time with fewer biopsies being performed in patients with TBP.^22- 24

Compared with patients with newly diagnosed melanoma presenting from routine care, for which 84% were nonphysician detected,⁶ in the PLC setting 76% of melanomas were physician detected. It is important to point out that the majority of melanomas diagnosed in our clinic were in situ and thin ( <0.76 mm). Early melanomas may have clinical features not readily recognized by patients or general dermatologists, and this, in part, may explain the high rate of dermatologist-detected melanomas in our study. In fact, in the previous studies that addressed melanoma detection patterns and reported higher rates of nonphysician detected melanoma, mean tumor thickness was significantly greater than melanoma thickness observed in our cohort. For instance, Brady et al⁶ reported median tumor thickness of 1.13 mm (in situ lesions were included in this calculation). Similarly, mean tumor thickness reported by Schwartz et al⁸ was 1.02 mm. In comparison, mean tumor thickness of invasive melanomas in our study population was 0.61 mm. Similar to other investigators, who reported that physicians are more likely to detect thinner lesions,^6- 10,25 the mean tumor thickness of dermatologist-detected melanomas in our study was significantly less than that of self-detected lesions.

We must not discount that patients place an important role in melanoma recognition, since 18% of all melanomas were self-detected. In self-detected melanomas, the primary complaint for seeking evaluation was changes in a pre-existing lesion (64%). The most common signs and symptoms reported with self-detected malignant melanoma in situ were color and size change, color change alone, and size change alone, whereas bleeding and rapid growth were more commonly reported with self-detected invasive melanomas. Similarly, others investigators reported that a change in color and increase in size were associated with thinner lesions, while thicker melanomas had bleeding associated with a lesion.^8,26

Our findings have implications for promoting physician and self-screening and for emphasizing the important role that a specialized PLC setting may play in the early diagnosis of melanoma. It is generally accepted that secondary prevention, aimed at early melanoma detection, may have the greatest immediate impact on melanoma mortality.^27- 29 Many melanomas can be self-detected, and therefore, educating patients on recognizing melanoma signs and symptoms by performing SSE can have a notable impact on early diagnosis of melanoma. For example, Carli et al³⁰ found that performing SSE was strongly associated with early diagnosis (SSE performers vs nonperformers, 0.77mm vs 0.95 mm).³⁰ Similarly, Berwick et al³¹ reported that the risk of advanced disease and mortality among patients with melanoma may be reduced with performing SSE. Finally, the experience at the Lawrence Livermore National Laboratory has shown that combination of targeted screening conducted by a dermatologist and promotion of self-examination has the potential to reduce melanoma-related mortality.³² Therefore, it is crucial to emphasize that a combined strategy of physician detection and patient participation must continue to be implemented to ensure early melanoma diagnosis.³³

This study has several limitations. First, the data are retrospective. Second, the study population was drawn from patients at a tertiary referral center; as such, they are representative of neither the general population nor the general dermatology population. Third, the tumor histologic type was not available for most melanomas in the study. Fourth, the results do not address whether screening and detection of early melanoma improve patients' survival.

In conclusion, this study suggests that physician screening leads to higher rates of physician-detected melanoma and detection of thinner melanoma. High-risk individuals may benefit from PLC setting. Prospective studies are needed to assess the survival benefit of screening.