|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Originally published as JCO Early Release 10.1200/JCO.2004.12.015 on August 9 2004 © 2004 American Society of Clinical Oncology. Thin Primary Cutaneous Malignant Melanoma: A Prognostic Tree for 10-Year Metastasis Is More Accurate Than American Joint Committee on Cancer StagingFrom the Melanoma Program of the Abramson Cancer Center, Department of Biostatistics and Epidemiology, Department of Medicine, Department of Dermatology, Department of Pathology and Laboratory Medicine, and Department of Surgery, University of Pennsylvania School of Medicine, Philadelphia, PA Address reprint requests to Phyllis A. Gimotty, PhD, Department of Biostatistics and Epidemiology, 631 Blockley Hall, 423 Guardian Dr, Philadelphia, PA 19104-6021; e-mail: pgimotty{at}cceb.upenn.edu
PURPOSE: The majority of invasive primary melanomas are thin (  1.00 mm). Since the current staging system imperfectly predicts outcome in patients with such lesions, we sought to develop a more effective classification scheme to better identify both patients at high risk of metastasis who are candidates for further staging and therapy and those with little risk. PATIENTS AND METHODS: This prospective cohort study included 884 patients who had thin invasive melanomas. A tree-structured analysis of 10-year metastasis was used to develop a new classification scheme. RESULTS: The overall 10-year metastasis rate was 6.5% (95% CI, 4.8% to 8.1%). The prognostic tree defined four risk groups: high-risk: men with vertical growth phase (VGP) lesions that had mitotic rates (MRs) greater than 0, and for whom the 10-year metastasis rate was 31% (22% to 42%; n = 90); moderate-risk: women with VGP lesions that had MRs greater than 0 and for whom the rate was 13% (9% to 18%; n = 136); low-risk: patients with VGP lesions that had MR of 0 for whom the rate was 4% (2% to 7%; n = 247); and minimal-risk: patients with invasive lesions without VGP for whom the rate was 0.5% (0% to 1.2%; n = 411). Survival curves differed significantly among the four groups (P < .001). CONCLUSION: Growth phase, mitotic rate, and sex are important prognostic factors for patients with thin melanomas, and they identify subgroups at substantial risk for metastasis. After validation in other populations, the proposed prognostic tree will be useful in the design of clinical trials and clinical management.
Thickness, measured from the top of the epidermal granular layer to the deepest invasive melanoma cell, is a powerful prognostic factor for cutaneous invasive melanomas that are apparently confined to the primary tumor site. Thin ( 1.00 mm) invasive lesions represent the majority of primary cutaneous melanomas. During the last decade, nearly 70% of invasive melanomas reported to the Surveillance, Epidemiology, and End Results (SEER) Program were in this thickness range.1 Thin lesions confined to the primary tumor site are categorized either as stage IA (T1a) or IB (T1b) tumors in the 2002 American Joint Committee on Cancer (AJCC) staging classification.2,3 Stage IA encompasses thin lesions in which the level of invasion is limited to Clark's levels II and III and in which ulceration is absent; stage IB includes lesions characterized by Clark's level IV and V invasion or ulceration regardless of level. Thin lesions are, in general, associated with a good prognosis. However, some patients with thin melanomas develop regional metastases, disseminated disease, and/or death from their disease. The 10-year mortality rates associated with stage IA (T1a) and IB (T1b) tumors are statistically, if not clinically, different (12% and 17%, respectively).2 To better guide both management decisions and the design of clinical trials, it is critical to investigate whether there are additional prognostic factors that can better distinguish among patients with thin melanomas who are highly unlikely to develop metastatic disease and those whose likelihood of metastasis is sufficiently high to prompt interventions beyond surgery at the primary site (eg, sentinel lymph node mapping and biopsy). Clark's level and ulceration are the only two prognostic factors used in the 2002 AJCC staging system for thin lesions. The purpose of this study was to investigate additional prognostic factors and develop a classification scheme for patients with thin melanomas to further identify patients with clinically and statistically significantly different levels of risk for a metastatic event within 10 years of definitive surgical excision of the primary lesion. Our objective was to develop a classification scheme that provided good sensitivity for identification of those at high risk, as well as good specificity for identification of those at low risk.
Between 1972 and 1991, 2,188 consecutive patients with invasive primary melanoma were seen at the University of Pennsylvania's Pigmented Lesion Clinic (PLC). Of these, 1,232 had melanomas that were 1.0 mm in thickness, and 1,114 patients were eligible for this study (Fig 1). Eligibility criteria were a diagnosis of an invasive primary melanoma without evidence of metastasis and a visit to the PLC within 1 year of definitive surgery. All lesions were retrospectively reviewed without knowledge of the patients or their clinical outcomes by one of three pathologists (W.H.C., D.E.E., R.E.), and all were categorized in standard fashion.4 Invasive radial growth phase (RGP) lesions had no evidence of dermal proliferation (no mitoses and no tumor nests larger than those in the epidermis). Vertical growth phase (VGP) lesions had dermal nests or tumor cell aggregates larger than in the epidermis and/or mitotic figures in the dermis. Mitotic rate was reported as the number of mitoses per mm2 in the dermal VGP, and VGP tumor-infiltrating lymphocytes (TILs) were classified as brisk, nonbrisk, or absent. Additional histopathologic attributes available included thickness, Clark's level, regression, ulceration, and microscopic satellites.4,5 Clinical features available for all patients included age, sex, and the anatomic site of the primary lesion. Based on commonly used definitions, including those used in the AJCC staging system, several binary variables were created—age at diagnosis, either  60 years or less than 60 years6-8; anatomic site, either extremity or axial/subungual/volar, including the head, neck, trunk, palms and soles, and nail beds; mitotic rate, either equal to 0 or more than 04; TIL, either brisk, nonbrisk, or absent4; thickness, either 0.01 to 0.75 mm or 0.76 to 1.00 mm6-12; and Clark's level, either II/III or IV/V.3
The patients excluded had one or more of the following pathologic characteristics missing (n = 149): thickness (n = 13); Clark's level (n = 10); RGP regression (n = 95); VGP mitotic rate (n = 93); and TIL (n = 77). An additional 81 patients who were either lost to follow-up (n = 33) or who had died of causes unrelated to melanoma (n = 48) were only included in the survival analysis. Consequently, this prospective cohort included 884 patients with complete clinical and histologic data, and at least 10 years of follow-up or a metastatic event within 10 years of diagnosis. Follow-up has been continuously maintained since 1972, and for patients alive at the time of last follow-up (n = 736), the median length of follow-up was 17 years. The first recurrences subsequent to definitive melanoma surgery were classified as local recurrence (a recurrence within the scar at the primary site), regional metastasis (in transit dermal or subcutaneous metastases and/or nodal involvement), disseminated disease (nonregional skin and/or nodes and/or visceral metastases), and "mixed" (both regional and disseminated). For this study, local recurrences were subcategorized as "type 1" if they consisted only of RGP, "type 2" if they consisted of both RGP and VGP, and "type 3" if they consisted of a nodule without apparent RGP. Documented melanoma-specific death without information on the extent of metastatic disease was classified as disseminated disease. Metastatic events included regional metastasis, disseminated disease, mixed, and melanoma-specific death, and excluded local recurrence. The 10-year metastasis rates were defined as the proportion of patients who had a metastatic event within 10 years of definitive treatment. Metastatic failure times were computed as the time between definitive surgery of the primary lesion and the diagnosis of a metastatic event. Patients without metastatic events and patients who died from causes unrelated to melanoma more than 10 years after definitive surgery were censored at the date of last follow-up. A multivariate logistic regression analysis of 10-year metastasis was used to identify the significant independent prognostic factors among the 11 factors used to develop the prognostic tree. Variables without significant contributions to the multivariate model based on the likelihood ratio statistic were sequentially removed to obtain the final multivariate model. Kaplan-Meier survival curves were computed for the metastatic failure times, and the log-rank statistic was used to evaluate differences among survival curves for the risk groups. The Kaplan-Meier survival curves were then used to estimate the metastasis-free probabilities; 95% CIs were computed using Greenwood's formula. The logistic regression models and survival analyses for this article were generated using SAS (SAS Institute, Cary, NC) System for Windows.13 The prognostic tree for 10-year metastasis was developed using a recursive partitioning algorithm implemented in classification and regression tree (CART) software (Salford Systems, San Diego, CA).14 This algorithm sequentially divides a group of patients into subgroups that become progressively more homogeneous. At the first step, one factor among all potential factors is selected to split the entire group of patients into two subgroups that are each more homogeneous with respect to the clinical outcome than the original group. At the second, as well as subsequent steps, a factor is selected (again from all potential factors), using the same process to subdivide the current group. The process stops when no further improvement in homogeneity can be made by splitting using the available factors. Eleven factors were included—age, sex, anatomic site, thickness, level, growth phase, regression, mitotic rate, TIL, ulceration, and microscopic satellites. To minimize overfitting the data, groups with fewer than 100 patients were not further split. The tree was then pruned to develop final risk groups, between which the 10-year metastasis rates differed significantly. Sensitivity and specificity of predictions using the prognostic tree, as well as AJCC stage, were respectively defined as the proportion of patients predicted to have events who had a 10-year metastasis, and the proportion of patients predicted not to have events who were disease-free for a least 10 years following treatment. A simulation study was done to assess the reproducibility of the prognostic tree. Two hundred fifty random (bootstrap) samples consisting of 800 observations were obtained using simple random sampling from the 884 observations, and 250 prognostic trees were developed from these samples. The reproducibility of the factor selection process was characterized by the proportion of trees wherein the most frequently selected factor was observed.
The internal validation of the prognostic tree included new patients who met the study eligibility criteria and were seen between 1991 and April 15, 1995 (the start date of a study to investigate the role of sentinel lymph node biopsy in patients with thin melanoma).15 These patients had metastatic events or were followed for at least eight years following the excision of their primary lesion. The observed metastasis rates from the validation sample (n = 144) were compared to the estimates of the eight-year metastasis rates from the Kaplan-Meier curves (n = 965) for each risk group using the goodness-of-fit
Table 1 presents the clinical and histopathologic characteristics, including AJCC stage, for the 884 study patients. The median age of this group was approximately 45 years (mean, 45.6 years; range, 4 to 81 years). The most common histologic type was superficial spreading melanoma (87%), followed by lentigo maligna melanoma (6%), acral-lentiginous melanoma (1%), and nodular melanoma (3%), and the remaining 3% could not be classified.
Within 10 years of definitive excision of their primary lesions, 57 patients had metastatic disease, including one patient who had a local recurrence (type 2) who died of melanoma 7.5 years later (Table 2). Forty of these patients died from melanoma (69%), and four died from other or unknown causes (7%); the average length of follow-up after the first metastatic event for the remainder of the patients (n = 12) was 15.3 years (range, 6 to 22.7 years). Of note, 20 additional patients had first events more than 10 years after definitive surgery, including seven local recurrences (four of type 1, two of type 2, and one of type 3) and 13 metastatic events associated with seven deaths.
The overall 10-year metastasis and melanoma-specific death rates were 6.5% and 4.5%, respectively. Table 3 presents prognostic factor–specific 10-year metastasis rates and the association of each characteristic with 10-year metastasis (unadjusted odds ratios). While eight factors were significantly associated with 10-year metastasis in the univariate analysis, only four—sex, growth phase, mitotic rate, and TIL—were significant in the multivariate analysis: the odds of metastasis was 3.1-fold higher for males than females, 42-fold higher for VGP lesions compared with RGP lesions, 7.7-fold higher for lesions with a mitotic rate (MR) greater than 0 compared with those with an MR of 0, and 2.34-fold higher for lesions with brisk/nonbrisk TIL compared with those without TIL.
The prognostic tree is presented in Figure 2. Mitotic rate was the first prognostic factor selected to split the entire sample, resulting in two groups characterized by a MR of 0 and an MR greater than 0. The patients in the latter group all had VGP lesions since the presence of mitoses is one defining characteristic of VGP lesions. At the second step, the first group (MR = 0) was split on growth phase, and the second group (MR > 0) was split on sex, producing four groups that had minimal-, low-, moderate-, and high-risk of 10-year metastasis. The 10-year metastasis rates were 0.5%, 4.1%, 12.5%, and 31.1% in the risk groups representing significantly different levels of risk.
At the third step (not shown), regression was selected to split the minimal risk group (invasive RGP lesions), and those with regression (n = 91) had a 10-year metastasis rate of 2.2% compared with those without regression (n = 320) who had no events. The average length of follow-up for the 320 patients without regression was 18.1 years (range, 10.3 to 31.7 years). TIL was the factor selected to split both the low- and moderate-risk groups. In the low-risk group, the 10-year metastasis rate was 5.6% (n = 108) for those without TIL and 2.9% (n = 139) for those with TIL. In the moderate risk group, the rate was 18.6% for those (n = 43) without TIL and 9.7% for those (n = 93) with TIL. However, within each of these three risk groups (minimal, low, and moderate), there was no statistically significant difference between the two 10-year rates in the subgroups. The high-risk group was not split. In the simulation used to investigate the reproducibility of the selection of prognostic factors, mitotic rate and growth phase were the only two factors selected at the first step. Mitotic rate was the most frequently selected prognostic factor for the primary split (80% of the prognostic trees generated). At the second step, growth phase was the most frequently selected factor for those with an MR of 0, and sex was the most frequently selected factor for those with MR greater than 0. For both groups, the most frequently selected factor appeared in approximately 30% of the trees generated (among the 200 trees that split first on mitotic rate), and all other prognostic factors were observed less frequently. Figure 3 presents the receiver-operator characteristic curves (ROC) and the corresponding sensitivities and specificities for a decision rules based on the risk groups and AJCC stage. The discrimination between those who had a 10-year metastasis and those who did not is better achieved using the risk groups rather than stage. The estimated area under the ROC using the risk groups and stages are 0.85 and 0.59, respectively.
As previously noted, 20 first events were observed after 10 years of follow-up. Of the 13 metastatic events, one was in the minimal-risk group, three were in the low-risk group, six were in the moderate-risk group, and three were in the high-risk group. Of the seven local recurrences, four were in the minimal-risk group, and one was in each of the other three risk groups. The Kaplan-Meier curves for the metastasis-free interval for patients in the four risk groups are significantly different (P < .001) and are presented in Figure 4. The predicted probabilities associated with a metastasis-free interval of 5, 8, 10, 15, 20, and 25 years estimated from the Kaplan-Meier curves, are presented in Table 4. The predicted probabilities for the two higher-risk groups are substantially lower than those for the minimal- and low-risk groups.
The 8-year metastasis rates for the "learning" cohort and the internal validation cohort are presented in Figure 5. There was no significant difference in 8-year metastasis rates between the two cohorts (P < .05).
Melanomas evolve in stepwise fashion, first proliferating in the epidermis as in situ radial growth phase lesions, then invading shallowly the dermis as invasive radial growth phase lesions with little or no initial capacity for proliferation, and subsequently proliferating and expanding as VGP lesions that acquire the capacity for metastasis.4,16,17 While most invasive RGP lesions are thin (97.6% among 1,008 patients with RGP lesions seen at the PLC between 1972 and 1991), lesional thinness is an imperfect surrogate for lack of aggressiveness: 54% of our thin lesions had VGP with a melanoma-specific metastasis rate of 11.6%. Thin lesions are currently classified as either stage IA or IB based on the 2002 AJCC staging system, without regard for growth phase and using the uncommon variables of deeper level and the presence of ulceration to identify lesions of higher risk. As noted, thin lesions represent the majority of invasive primary melanomas diagnosed in this country. With informative prognostic factors that better reflect tumor progression, high-risk patients can be identified, and can be candidates for trials of additional staging procedures (eg, sentinel node biopsy) and adjuvant therapies. We have developed a classification scheme for prognosis based on a large series of consecutive cases with extensive clinical and histopathologic data and prospective long-term follow-up. Because prognostic models have not been widely adopted in clinical practice, we sought to build a model that would be clinically credible by virtue of a number of criteria, including being biologically based, using only variables that are readily obtained from skilled pathologists using routine histology, testing all of the potentially relevant prognostic factors, having a transparent and accessible structure, and being user friendly (ie, providing a simple table of risk predictions without requiring calculation).18 We have previously published a prognostic model for patients with melanomas of all thicknesses that found that growth phase, thickness, mitotic rate, TIL, sex, anatomic site, and regression were independent prognostic factors.4 Because thin melanomas are considered to represent low-risk disease19 and because patients with such lesions are generally less intrusively staged and less frequently followed up, other investigators have sought prognostic factors in patients with thin lesions that might associate with subgroups with poor prognosis who might be otherwise managed. Thickness,20-22 level,2,3,21 ulceration,2,3,20,22,23 microscopic satellites,2,3 regression,21,22,24 VGP,23,25-27 mitotic rate,23,25 age,20 TIL,21 and anatomic site21 have been found to be associated with poor prognosis. In this study, we examined these and other factors and found in the final multivariate analysis that only sex, growth phase, mitotic rate, and presence of TIL were independent prognostic factors for 10-year metastasis. These pathology attributes have been demonstrated to be reproducible in the case of growth phase28,29 and TIL.30 Since in this model, mitoses and TIL are scored as present or absent, their ascertainment is likely to be quite accurate. Current practice guidelines31 encourage inclusion in the pathology note of each of these attributes and regression. Hence, they can be readily included in reports when the model is adopted in a given institution. We have observed some differences between our study patients and the patients used in the validation of the AJCC classification.3 The melanoma-specific death rate that we observed, consistent with the Australian experience20 and SEER,1 is much lower than that reported for both T1a and T1b lesions from the AJCC cohort. While ulceration is an important factor in the 2002 AJCC staging system, our assessment of it as a prognostic factor was limited by the small number of lesions that were observed with ulceration (n = 14; 1.6%) in our study population. Ulceration is also unusual in thin melanomas in the general population: 1.2% in SEER data.1 It is likely that these differences reflect differences in patient selection. CART methodology was used develop a classification tree for prognosis.32 Three prognostic factors were identified and used to describe a prognostic tree with four different risk groups. Mitotic rate, a measure of tumor proliferation in the dermis, was identified as the best single prognostic factor in the development of the tree. Using only mitotic rate to predict 10-year metastasis has sensitivity and specificity of approximately 78%. This result is consistent with two studies of patients with thin lesions in which the mitotic rate was shown to be a significant prognostic factor.23,25 Different prognostic factors were selected for groups of patients with different mitotic rates, growth phase (MR = 0), and sex (MR > 0), emphasizing that the relationship between metastasis and sex or growth phase is not the same for the two groups defined by mitotic rate. The resulting risk groups can be used to define "high-" and "low-" risk groups that have higher sensitivity and specificity compared with AJCC stage. An advantage of CART is apparent here. While the corresponding multivariate logistic regression model with the four binary prognostic factors, mitotic rate, sex, growth phase, and TIL would result in 16 risk groups, some with indistinguishable risks, CART allowed the development of a simple classification with four groups defined by three prognostic factors. There are several challenges to developing prognostic models for patients with thin lesions since most patients are cured by surgery at the primary site. Consequently, a large sample of patients is needed to develop clinically useful prognostic models. Of 884 patients, there were only 57 patients who developed a metastasis within 10 years of definitive treatment, and an additional 13 patients who developed metastasis between 10 and 25 years after diagnosis. To avoid overfitting the data with a statistical model, it is recommended that there be at least 10 events for each factor in a model.33 Our final multivariate model had four significant prognostic factors, and our prognostic tree had two splits that resulted in four groups. By focusing on the most important prognostic factors, the estimated probabilities for groups defined using the prognostic variables are more likely to be representative of similar patient populations.34 The proposed prognostic tree includes three of the four variables identified in the multivariate analysis. The fourth variable, the presence of TIL, was selected at the third step in the development of the tree; however, some of the resulting risk groups had only a few observations, and the confidence intervals for the estimated probabilities of a 10-year metastasis overlapped. These two methods of analysis are complementary, and together, they provide strong evidence for the role of these factors in prognosis. In studies of larger cohorts, TIL is a strong candidate variable to further refine prognostication. In thicker melanomas, the prognostic significance of TIL has been demonstrated,4,27,35 and in one study of thin melanomas, TIL was the only attribute independently associated with metastasis.21 Another important consideration for a prognostic model's potential as a clinically effective tool is its validation. As part of this study, we conducted two types of validation. First we examined reproducibility of the prognostic tree by simulation assuming that the data we observed are descriptive of all patients with thin melanoma, and demonstrated that mitotic rate, sex, and growth phase are important factors in distinguishing patients who are likely to have 10-year metastasis. Second, we examined the model's predictions in a subsequent cohort of patients seen at the PLC, and found patterns of 8-year metastasis that are consistent with the "learning" model. Our model will need external validation to confirm its accuracy and generalizability. However, it has several important attributes that suggest it will be clinically effective: it was derived from a prospectively defined and rigorously followed inception cohort that closely resembles patients captured in SEER, it is internally valid, and it uses attributes that are readily and accurately ascertainable and suggested for widespread use. With the publication of the revised AJCC staging system, it is important to continue to investigate new prognostic factors that will improve the ability of the clinician and clinical trialist to classify patients based on metastatic risk and survival. Given the high incidence of thin invasive melanomas, it is particularly important to have a better understanding of prognostic factors and their interactions in prognostic models specific to these patients. This study demonstrates that our proposed prognostic classification of thin lesions based on mitotic rate, growth phase, and sex can distinguish among those with a range of risks for metastasis. Patients in our "moderate-" and "high-" risk groups have an incidence of metastasis that is comparable to those with lesions of 1 to 2 mm in thickness,2 a group that is generally staged with sentinel node biopsies. Once validated, the model can be used in several ways. It can be used to "protect" patients with a very low likelihood of metastasis after surgery at the primary site from further intervention. It is notable that patients with invasive RGP lesions and no histologic evidence of regression (approximately 36% of our thin cases) had no metastases with long-term follow-up (in the one patient with metastasis > 10 years after treatment, information on regression was missing). It can also be used to identify those for further intervention. The model's clinical effectiveness can be tested in, for example, its ability to select a population of patients with a significant rate of involved sentinel nodes. This study also addresses an additional issue with implications for patient management. Our data highlight and extend the observation36,37 that thin lesions with more than minimal risk (including RGP lesions with regression) have recurrences (Table 2) well beyond 10 years (Fig 3). These include recurrences that may be treated surgically with curative intent (eg, local recurrences [particularly of what is likely persistent RGP, a type 1 recurrence] and regional nodal disease). While all patients with melanoma are candidates for lifelong follow-up (at least yearly)31 for the development of additional primary lesions (which are usually of lower risk than the first), their examination should also include inspection and palpation of the primary site and regional skin and nodes.
The authors indicated no potential conflicts of interest.
We thank all of the patients who have been seen at the Pigmented Lesion Clinic (PLC) and given their consent for use of their data in research studies, as well as the investigators (Drs W.H. Clark Jr  , E.E. Bondi, L.P. Bucky, L.S. Callans, B. Chang, K.T. Flaherty, D.L. Fraker, A.C. Halpern, R. Hamilton, D. Hershock, D.D. Larossa, S.R. Lessin, D. Low, P. Van Belle, and J. Wolfe) and staff (R. Holmes, S. Hotz, N. Lowden, I. Matozzo, M. Price, M. Synnestvedt, and J. Thompson) of the PLC for their contributions over the last three decades to the Melanoma Core Database on which this report is based. Deceased
Supported in part by the following grants: Prediction and Modification of Melanoma Risk (CA-75434; D.G.) and SPORE on Skin Cancer (CA-093372; Meenhard Herlyn, principal investigator). Presented in part at the 38th Annual Meeting of the American Society of Clinical Oncology, Orlando, FL, May 18-21, 2002. Authors' disclosures of potential conflicts of interest are found at the end of this article.
1. Surveillance, Epidemiology, and End Results (SEER) Program (www.seer.cancer.gov) Public-Use Data (1973-2000), National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch, released April 2003, based on the November 2002 submission.
2. Balch CM, Soong SJ, Gershenwald JE, et al: Prognostic factors analysis of 17,600 melanoma patients: Validation of the American Joint Committee on Cancer melanoma staging system. J Clin Oncol 19:3622-3634, 2001
3. Balch CM, Buzaid AC, Soong SJ, et al: Final version of the American Joint Committee on caner staging system for cutaneous melanoma. J Clin Oncol 19:3635-3648, 2001
4. Clark WH, Elder DE, Guerry D, et al: Model predicting survival in Stage I melanoma based on tumor progression. JNCI 81:1893-1904, 1989 5. Greene F, Page D, Fleming I, et al: AJCC Cancer Staging Handbook Sixth Edition, New York, NY, Springer, 2002
6. Schuchter L, Schultz DJ, Synnestvedt M, et al: A prognostic model for predicting 10-year survival in patients with primary melanoma: The Pigmented Lesion Group. Ann Intern Med 125:369-375, 1996 7. Sahin S, Rao B, Kopf AW, et al: Predicting ten-year survival of patients with primary cutaneous melanoma. Cancer 80:1426-1431, 1997[CrossRef][Medline] 8. Averbook BJ, Fu P, Rao JS, et al: A long-term analysis of 1018 patients with melanoma by Cox regression and tree-structured survival analysis at a major referral center: Implications on the future of cancer staging. Surgery 132:589-604, 2002[CrossRef][Medline] 9. Garbe C, Buttner P, Bertz J, et al: Primary cutaneous melanoma: Indentification of prognostic groups and estimation of individual prognosis for 5093 patients. Cancer 75:2484-2491, 1995[CrossRef][Medline] 10. Huang X, Soong S, McCarthy WH, et al: Classification of localized melanoma by the exponential survival trees method. Cancer 79:1122-1128, 1997[CrossRef][Medline] 11. Dong XD, Tyler D, Johnson JL, et al: Analysis of prognosis and disease progression after local recurrence of melanoma. Cancer 88:1063-1071, 2000[CrossRef][Medline] 12. Beahrs OH, Henson DE, Hutter RVP, et al: American Joint Committee on Cancer manual for staging cancer. Philadelphia, PA, JB Lippincott, 1988 13. SAS Institute Inc: SAS/STAT User's Guide, Version 8. Cary, NC, SAS Institute Inc, 1999 14. Steinberg D, Colla P: CART: Tree-structured non-parametric data analysis, San Diego, CA, Salford Systems, 1995 15. Bedrosian I, Faries MB, Guerry D, et al: Incidence of sentinel node metastasis in patients with thin primary melanoma (< or = 1 mm) with vertical growth phase. Ann Surg Oncol 7:262-267, 2000[CrossRef][Medline] 16. Elder DE, Guerry D, Epstein MN, et al: Invasive malignant melanomas lacking competence for metastasis. Am J Dermatopathol 6:55-61, 1984 (suppl) 17. Guerry D, Synnestvedt M, Elder DE, et al: Lessons from tumor progression: The invasive radial growth phase of melanoma is common, incapable of metastasis, and indolent. J Invest Dermatol 100:342S-345S, 1993[CrossRef][Medline]
18. Wyatt JC, Altman DG: Commentary: Prognostic models: Clinically useful or quickly forgotten? BMJ 311:1539-1541, 1995
19. National Institutes of Health: Consensus Development Conference: Diagnosis and treatment of early melanoma. JAMA 268:1314-1319, 1992 20. McKinnon JG, Yu XQ, McCarthy WH, et al: Prognosis for patients with thin cutaneous melanoma. Cancer 98:1223-1231, 2003[CrossRef][Medline] 21. Mansson-Brahme E, Carstensen J, Erhardt K, et al: Prognostic factors in thin cutaneous malignant melanoma. Cancer 73:2324-2332, 1994[CrossRef][Medline] 22. Finley JW, Gibbs JF, Rodriguez LM, et al: Pathologic and clinical features influencing outcome of thin cutaneous melanoma: Correlation with newly proposed staging system. Am Surg 66:527-531, 2000[Medline] 23. Oliveira Filho RS, Ferreira LM, Biasi LJ, et al: Vertical growth phase and positive sentinel node in thin melanoma. Braz J Med Biol Res 36:347-350, 2003[Medline] 24. Gromet MA, Epstein WL, Blois MS: The regressing thin malignant melanoma: A distinctive lesion with metastatic potential. Cancer 42:2282-2292, 1978[CrossRef][Medline] 25. Cook MG, Spatz A, Brocker E, et al: Identification of histological features associated with metastatic potential of thin (< 1.0 mm) cutaneous melanoma with metastases: A study on behalf of the EORTC Melanoma Group. J Pathol 197:188-193, 2002[CrossRef][Medline] 26. Lefevre M, Vergier B, Balme B, et al: Relevance of vertical growth pattern in thin level II cutaneous superficial spreading melanomas. Am J Surg Pathol 27:717-724, 2003[CrossRef][Medline] 27. Massi D, Franchi A, Borgognoni L, et al: Thin cutaneous malignant melanomas (< or =1.5 mm): Identification of risk factors indicative of progression. Cancer 85:1067-1076, 1999[CrossRef][Medline]
28. A nationwide survey of observer variation in the diagnosis of thin cutaneous malignant melanoma including the MIN terminology: CRC Melanoma Pathology Panel. J Clin Pathol 50:202-205, 1997 29. McDermott NC, Hayes DP, al-Sader MH, et al: Identification of vertical growth phase in malignant melanoma: A study of inter-observer agreement. Am J Clin Pathol 110:753-757, 1998[Medline]
30. Busam KJ, Antonescu CR, Marghoob AA, et al: Histologic classification of tumor-infiltrating lymphocytes in primary cutaneous malignant melanoma: A study of inter-observer agreement. Am J Clin Pathol 115:856-860, 2001 31. National Comprehensive Cancer Network: Melanoma: Clinical practice guidelines in oncology. Available at http://www.nccn.org/professionals/physician_gls/PDF/melanoma.pdf 32. Breiman L, Friedman JH, Olshen RA, et al: Classification and Regression Trees. Boca Raton, FL, Chapman and Hall, 1984 33. Harrell FE, Lee KL, Califf RM, et al: Regression modeling strategies for improved prognostic prediction. Stat Med 3:143-152, 1984[Medline] 34. van Houwelingen HC: Validation, calibration, revision and combination of prognostic survival models. Stat Med 19:3401-3415, 2000[CrossRef][Medline] 35. Clemente CG, Mihm MC Jr, Bufalina R, et al: Prognostic value of tumor infiltrating lymphocytes in the vertical growth phase of primary cutaneous melanoma. Cancer 77:1303-1310, 1996[CrossRef][Medline] 36. Tsao H, Cosimi AB, Sober A: Ultra-late recurrence (15 years or longer) of cutaneous melanoma. Cancer 79:2361-2370, 1997[CrossRef][Medline] 37. Kalady MF, White RR, Johnson JL, et al: Thin melanomas: Predictive lethal characteristics from a 30-year clinical practice. Ann Surg 238:528-536, 2003[Medline] Submitted December 2, 2003; accepted April 20, 2004.
Related Article
Related Editorial
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2004 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
2 statistic. 
