Classification of localized melanoma by the exponential survival trees method

Pharmacogenomic biomarkers for personalized medicine

Pharmacogenomics examines how the benefits and adverse effects of a drug vary among patients in a target population by analyzing genomic profiles of individual patients. Personalized medicine prescribes specific therapeutics that best suit an individual patient. Much current research focuses on developing genomic biomarkers to identify patients, to identify which patients would benefit from a treatment, have an adverse response, or no response at all, prior to treatment according to relevant differences in risk factors, disease types and/or responses to therapy. This review describes the use of the two personalized medicine biomarkers, prognostic and predictive, to classify patients into subgroups for treatment recommendation.

Assessment of reproducibility of cancer survival risk predictions across medical centers

BACKGROUND

Two most important considerations in evaluation of survival prediction models are 1) predictability - ability to predict survival risks accurately and 2) reproducibility - ability to generalize to predict samples generated from different studies. We present approaches for assessment of reproducibility of survival risk score predictions across medical centers.

METHODS

Reproducibility was evaluated in terms of consistency and transferability. Consistency is the agreement of risk scores predicted between two centers. Transferability from one center to another center is the agreement of the risk scores of the second center predicted by each of the two centers. The transferability can be: 1) model transferability - whether a predictive model developed from one center can be applied to predict the samples generated from other centers and 2) signature transferability - whether signature markers of a predictive model developed from one center can be applied to predict the samples from other centers. We considered eight prediction models, including two clinical models, two gene expression models, and their combinations. Predictive performance of the eight models was evaluated by several common measures. Correlation coefficients between predicted risk scores of different centers were computed to assess reproducibility - consistency and transferability.

RESULTS

Two public datasets, the lung cancer data generated from four medical centers and colon cancer data generated from two medical centers, were analyzed. The risk score estimates for lung cancer patients predicted by three of four centers agree reasonably well. In general, a good prediction model showed better cross-center consistency and transferability. The risk scores for the colon cancer patients from one (Moffitt) medical center that were predicted by the clinical models developed from the another (Vanderbilt) medical center were shown to have excellent model transferability and signature transferability.

CONCLUSIONS

This study illustrates an analytical approach to assessing reproducibility of predictive models and signatures. Based on the analyses of the two cancer datasets, we conclude that the models with clinical variables appear to perform reasonable well with high degree of consistency and transferability. There should have more investigations on the reproducibility of prediction models including gene expression data across studies.

Assessment of performance of survival prediction models for cancer prognosis

BACKGROUND

Cancer survival studies are commonly analyzed using survival-time prediction models for cancer prognosis. A number of different performance metrics are used to ascertain the concordance between the predicted risk score of each patient and the actual survival time, but these metrics can sometimes conflict. Alternatively, patients are sometimes divided into two classes according to a survival-time threshold, and binary classifiers are applied to predict each patient's class. Although this approach has several drawbacks, it does provide natural performance metrics such as positive and negative predictive values to enable unambiguous assessments.

METHODS

We compare the survival-time prediction and survival-time threshold approaches to analyzing cancer survival studies. We review and compare common performance metrics for the two approaches. We present new randomization tests and cross-validation methods to enable unambiguous statistical inferences for several performance metrics used with the survival-time prediction approach. We consider five survival prediction models consisting of one clinical model, two gene expression models, and two models from combinations of clinical and gene expression models.

RESULTS

A public breast cancer dataset was used to compare several performance metrics using five prediction models. 1) For some prediction models, the hazard ratio from fitting a Cox proportional hazards model was significant, but the two-group comparison was insignificant, and vice versa. 2) The randomization test and cross-validation were generally consistent with the p-values obtained from the standard performance metrics. 3) Binary classifiers highly depended on how the risk groups were defined; a slight change of the survival threshold for assignment of classes led to very different prediction results.

CONCLUSIONS

1) Different performance metrics for evaluation of a survival prediction model may give different conclusions in its discriminatory ability. 2) Evaluation using a high-risk versus low-risk group comparison depends on the selected risk-score threshold; a plot of p-values from all possible thresholds can show the sensitivity of the threshold selection. 3) A randomization test of the significance of Somers' rank correlation can be used for further evaluation of performance of a prediction model. 4) The cross-validated power of survival prediction models decreases as the training and test sets become less balanced.

Sentinel node positive melanoma patients: prediction and prognostic significance of nonsentinel node metastases and development of a survival tree model

BACKGROUND

Completion lymph node dissection (CLND) following positive sentinel node biopsy (SNB) for melanoma detects additional nonsentinel node (NSN) metastases in approximately 20% of cases. This study aimed to establish whether NSN status can be predicted, to determine its effect on survival, and to develop survival tree models for the sentinel node (SN) positive population.

MATERIALS AND METHODS

Sydney Melanoma Unit (SMU) patients with at least 1 positive SN, meeting inclusion criteria and treated between October 1992 and June 2005, were identified from the Unit database. Survival characteristics, potential predictors of survival, and NSN status were assessed using the Kaplan-Meier method, Cox regression model, and logistic regression analyses, respectively. Classification tree analysis was performed to identify groups with distinctly different survival characteristics.

RESULTS

A total of 323 SN-positive melanoma patients met the inclusion criteria. On multivariate analysis, age, gender, primary tumor thickness, mitotic rate, number of positive NSNs, or total number of positive nodes were statistically significant predictors of survival. NSN metastasis, found at CLND in 19% of patients, was only predicted to a statistically significant degree by ulceration. Multivariate analyses demonstrated that survival was more closely related to number of positive NSNs than total number of positive nodes. Classification tree analysis revealed 4 prognostically distinct survival groups.

CONCLUSIONS

Patients with NSN metastases could not be reliably identified prior to CLND. Prognosis following CLND was more closely related to number of positive NSNs than total number of positive nodes. Classification tree analysis defined distinctly different survival groups more accurately than use of single-factor analysis.

An Integrated Tree-Based Classification Approach to Prognostic Grouping with Application to Localized Melanoma Patients

We propose an integrated tree-based approach for prognostic grouping of localized melanoma patients. This approach incorporates the survival tree model with the agglomerative hierarchical clustering to group terminal subgroups with similar prognoses together. The Brier score is used to evaluate the goodness of fit and the k-fold cross-validation test is used to evaluate the reproducibility of the scheme for prediction. The proposed approach is applied to an American Joint Committee on Cancer (AJCC) localized melanoma data set and compared with the current AJCC staging system. This approach performs more efficiently than the standard tree methods and has made improvement over the current AJCC melanoma staging system.

Molecular diagnostics in melanoma

Molecular pathology is rapidly evolving, featuring continuous technologic improvements that offer novel clinical opportunities for the recognition of disease predisposition, for identifying sub-clinical disease, for more accurate diagnosis, for selecting efficacious and non-toxic therapy, and for monitoring of disease outcome. Currently, the identification and prognosis of primary cutaneous melanoma is based on histologic factors (tumor depth and ulceration) and clinical factors (number of lymph node and/or distant metastases). However, metastasis can occur in patients with thin melanomas, and sentinel lymph node biopsy does not identify all patients at risk for distant metastasis. New markers exist that correlate with melanoma progression, which may aid in melanoma identification, prognostication, and detection of minimal residual disease/early recurrence. Moreover, not many therapeutic options exist for melanoma as no regimen prolongs survival. Emerging data with investigational therapies suggest that certain markers might play a crucial role in identifying patients who will respond to therapy or show utility in the monitoring the response to therapy. Herein, molecular diagnostics that can potentially benefit the individual melanoma patient will be discussed.

Thin primary cutaneous malignant melanoma: a prognostic tree for 10-year metastasis is more accurate than American Joint Committee on Cancer staging

PURPOSE

The majority of invasive primary melanomas are thin (< or = 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.

Refining esophageal cancer staging

OBJECTIVE

Cancer staging is dynamic, reflecting accrual of knowledge and experience in treatment. The objectives of this study were to assess current esophageal cancer staging and to determine whether refinements of classification and stage grouping are necessary.

METHODS

From 1983 through November 2000, 480 patients underwent esophagectomy without induction therapy. Depth of tumor invasion (T), regional lymph node status (N), distant status (M), number of metastatic regional lymph nodes, and histopathologic type and grade were subjected to survival-tree analysis, multivariable Cox and hazard function analysis, and residual misclassification risk analysis.

RESULTS

Inhomogenity of survival was found within and lack of distinction was found between current American Joint Committee on Cancer staging groups, supporting the need for refinement. T1 and N1 were redefined on the basis of survival differences. T1a is intramucosal cancer, T1b is submucosal cancer (P =.008), N1 is 1 or 2 metastatic regional lymph nodes, and N2 is 3 or more metastatic regional lymph nodes (P =.01). Current subclassification of M1 is not warranted (P =.9). Histopathologic type (P =.17) and grade (P =.3) minimally refined staging. Reassignment of staging groups constrained by American Joint Committee on Cancer definitions of stages 0 and IV produced less monotonic, distinctive, and homogeneous survival than free assignment of staging groups.

CONCLUSIONS

Current American Joint Committee on Cancer staging of esophageal cancer is inadequate. Refinement requires redefinition of T1, N1, and M1 classifications. Stage grouping within the constraints of American Joint Committee on Cancer definitions produces less accurate prognosis than free assignment based on survival data.

Biomarkers in melanoma: staging, prognosis and detection of early metastases

Currently, melanoma remains a surgical disease since early detection and excision of thin melanomas offers the best chance of a cure. Despite intensive clinical investigation, no effective systemic therapies exist for metastatic melanoma. Sentinel lymph node biopsy has greatly aided the staging and prognostic evaluation of primary cutaneous melanoma, however, approximately a third of patients diagnosed with metastatic melanomas present without prior regional lymph node involvement. Additional prognostic biomarkers exist which help determine the risk of advanced melanoma but the accuracy for each current marker is less than 100%. A greater understanding of the biology of melanomas and the development of new methods to identify patients with early (subclinical) metastatic disease may allow for selective and more effective therapy for patients at-risk for advanced disease. In this paper, current and novel potentially more accurate biomarkers for the staging and prognostic evaluation of melanoma patients, and for the detection of subclinical metastases are reviewed.

Use of nonphysician health care providers for skin cancer screening in the primary care setting

BACKGROUND

Nonphysician health care providers are in an optimal position to provide cancer prevention and screening services.

METHODS

We conducted a survey of primary care physicians to determine physician use and amenability to use of nonphysician health care providers to perform skin cancer screening in comparison with other cancer screening examinations.

RESULTS

A total of 1,363 eligible physicians completed the survey. Of these, 631 physicians (46%) reported a nurse practitioner or physician assistant performing at least one type of cancer screening examination on their patients. Twenty-nine and 22% of all physicians reported nurse practitioners or physician assistants performing skin cancer screening, respectively. Family physicians were more likely to use nurse practitioners and physician assistants to perform these cancer screening examinations than internists (chi(2) test, P = 0.001 for each examination). Skin examinations were performed less frequently by nurse practitioners and physician assistants than all other cancer screening examinations. A total of 73-79% of family physicians and 60-70% of internists were amenable to having a nonphysician health care provider perform one or more of these examinations.

CONCLUSIONS

Primary care physicians are currently utilizing nonphysician health care providers to perform cancer screening examinations and the majority of those surveyed are amenable to the use of these providers for such examinations. This suggests that one possible strategy for increasing skin cancer screening is through an expanded role of nonphysician health care providers.

Prognostic immunohistochemical markers of primary human melanomas

BACKGROUND

Several clinical and histological factors of primary melanomas comprise a relatively large quantity of prognostic information.

OBJECTIVE

To find immunohistochemical markers that can improve the prognostic accuracy achieved by factors that are available without extra laboratory work, i.e. mitotic rate, tumour thickness, ulceration, localization, gender and age.

METHODS

Immunohistochemical markers were determined on frozen sections. Univariate and multivariate Cox regression analyses were performed after 5-10 years follow-up.

RESULTS

Seven immunohistochemical markers were related to disease-free and overall survival in univariate Cox regression analysis: Ki-67, human leucocyte antigen (HLA) -DQ, HLA-DP, Muc 18, A-10-33, transferrin receptor, and H-2-8-10. Only Ki-67 (n = 399) and HLA-DQ (n = 452) retained prognostic significance when evaluated in multivariate analyses in several models together with tumour thickness alone and with tumour thickness, gender, mitotic rate, age, localization and ulceration.

CONCLUSIONS

Ki-67 and HLA-DQ may be useful for risk assessments in primary melanomas.

Can immunohistochemical markers and mitotic rate improve prognostic precision in patients with primary melanoma?

BACKGROUND

In addition to tumor thickness, several other prognostic parameters have been identified in primary human melanomas. Some are available readily (localization, gender, age, and ulceration). Others must be evaluated with a moderate or even substantial amount of work (mitoses and immunohistochemical markers). This study was undertaken to determine whether this extra effort is justified because it actually improves the precision of prognostic statements.

METHODS

Immunohistologic markers were determined on frozen sections from 691 biopsies of human melanomas with the immunoperoxidase method. Univariate and multivariate Cox regression analyses were performed with metastases and with death as endpoints.

RESULTS

Fifteen parameters were related to disease free survival in univariate Cox regression analysis: tumor thickness, ulceration, localization, gender, age, mitoses, and the immunohistochemical markers very late antigen (VLA)-2, human leukocyte antigen (HLA)-ABC, HLA-DR, NKI-beteb, Mel 14, intercellular adhesion molecule (ICAM-1), K-1-2, G-7-E2, and H-2-4-7. Three of the easily available parameters exhibited independent significance in multivariate Cox regression analysis: tumor thickness, ulceration, and localization. If mitotic rate was included in this model, then it had independent prognostic significance but ulceration was no longer significant. However, the model that included tumor thickness, localization, and ulceration had a slightly higher overall chi-square test score, indicating a better performance compared with thickness, localization, and mitoses. The model that included tumor thickness, localization, and mitoses could not be improved by any of the immunohistochemical markers in this study.

CONCLUSIONS

Nine immunohistochemical markers with established prognostic significance for primary human melanoma were not found to improve a prognostic model that included tumor thickness, localization, and mitoses. If mitoses was replaced by ulceration, then the model performed slightly better, although ulceration was not significant in the presence of mitoses.

Prognostic classification of malignant melanomas by combining clinical, histological, and immunohistochemical parameters

In a retrospective study the prognostic relevance of clinical, histopathological, immunohistochemical, and flow-cytometric parameters in primary malignant melanomas was evaluated using both the receiver operating characteristic ROC procedure and the logistic regression model. The proteolytic enzymes collagenase IV, cathepsin B, and cathepsin D proved to be significant prognostic factors. Combining the results obtained with these enzymes with gender, anatomic site, tumour thickness, Clark's level, ulceration, pattern of invasive growth, and presence of large round cells resulted in greatly improved discrimination between metastasized and non-metastasized cases. It is anticipated that this method could allow for precise individual prognostic characterization and in particular for identification of high-risk patients for adjuvant therapy.

Malignant melanoma

The incidence of malignant melanoma is increasing at a rate greater than any other cancer occurring in humans. In this era of managed care, patients with a suspicious pigmented lesion may first present to their primary care physician for evaluation. Therefore it is mandatory that the primary care physician be capable of distinguishing between benign and malignant pigmented lesions, know how to evaluate such patients, and know when to refer patients with suspicious or malignant pigmented lesions. Surgical removal remains the mainstay of treatment for patients with melanoma. Thus, to increase the cure rate for melanoma, both the public and nondermatologists need to be educated regarding the prevention and early detection of melanoma. Only in this way can the diagnosis of melanoma be made early before deep invasion has occurred and the patient placed at risk for systemic spread. In recent years, the surgical management of melanoma has become more conservative and rational. Limb amputation, arbitrary 5-cm margins of excision, and elective lymph node dissections are no longer performed. The recommended margins of excision are now based on objective pathologic and clinical data and are more conservative, and the sentinel node biopsy is now used to determine which high-risk patients should undergo a formal lymph node dissection. Although encouraging results are being seen with immunotherapy protocols, to date the only adjunctive therapy shown to increase survival in patients at high risk for systemic spread is alpha-interferon. With this drug, the improved survival is modest at best; it is expensive and a minority of patients can tolerate it in the doses recommended. Although response rates of 20% are seen with chemotherapy in patients with disseminated disease, these responses are short-lived, and there is no associated increased survival. Except for lentigo maligna, radiation therapy, even when its delivery is modified, still is useful only as an adjunct to surgery or for palliation.