1. Assessment of melanoma histotypes and associated patient related factors: Basis for a predictive statistical model
- Author
-
Timo Buhl, Hans Peter Bertsch, Steffen Emmert, Albert Rosenberger, Michael P. Schön, Holger A. Haenssle, and Saskia Hoffmann
- Subjects
Adult ,Male ,medicine.medical_specialty ,Skin Neoplasms ,Multinomial logistic model ,Dermoscopy ,Dermatology ,Lentigo maligna ,Single Center ,Risk Assessment ,Sensitivity and Specificity ,Young Adult ,Age Distribution ,Germany ,Prevalence ,Humans ,Medicine ,Computer Simulation ,Neoplasm Invasiveness ,Sex Distribution ,Young adult ,Head and neck ,Melanoma ,Aged ,Retrospective Studies ,Related factors ,Models, Statistical ,business.industry ,Reproducibility of Results ,Retrospective cohort study ,Middle Aged ,Prognosis ,medicine.disease ,3. Good health ,Data Interpretation, Statistical ,Female ,business ,Algorithms - Abstract
Summary Background Certain melanoma histotypes carry a worse prognosis than others. We aimed to identify patient related factors associated with specific melanoma histotypes. Patients and methods Single center study including 347 melanoma patients, prospectively assessed for 22 variables leading to a database of more than 7 600 features. Results Melanomas were histologically categorized as superficial spreading (SSM, 70.6 %), nodular (NM; 12.7 %), acrolentiginous (ALM; 4.0 %), lentigo maligna (LMM; 3.8 %), or unclassified melanoma (UCM; 8.9 %). Well recognized melanoma risk indicators (i. e. many atypical nevi, freckles, previous melanoma), were significantly associated with SSM and LMM histotypes. NM and ALM patients carried significantly less common and/or atypical nevi. NM were mostly self-detected or detected by relatives. In contrast, SSM, LMM, and ALM were most frequently detected by dermatologists. NM and UCM were preferentially located on poorly observable sites, SSM on the lower limbs, ALM on plantar sites, and LMM on the head and neck. ALM and LMM patients were significantly older than other patients. A multinomial logistic model was designed to predict a certain melanoma histotype (overall accuracy 81 %), which could be helpful to focus the attention of clinicians or may be integrated into fully automated diagnostic algorithms. Conclusions Melanoma histotypes show significant differences regarding patients’ characteristics.
- Published
- 2015
- Full Text
- View/download PDF