IADR Abstract Archives
Dynamic Nomograms Predict Malignant Transformation In Oral Potentially Malignant Lesions
Objectives: To design a dynamic nomogram to facilitate prediction of malignant transformation of oral potentially malignant lesions.
Methods: A complete database of 590 oral potentially malignant lesions from 495 patients with single lesion disease and 95patients with multiple lesions diagnosed and followed up at Newcastle Upon Tyne Hospitals Trust (UK) between 1996-2014. All patients received at least 5 years of follow up. The r package ‘rms’ was used to fit a logistic regression model with internal validation and calibration using the bootstrap. The R package DynNom was then used to produce a dynamic nomogram. Outcome measures were continuing disease/malignant transformation versus resolution of lesions. All patients underwent incisional biopsy and subsequent laser surgery treatment for their potentially malignant lesion. The model used predictors of age, gender, clinical appearance of lesion, site, incisional biopsy pathology, laser excision pathology to determine clinical outcome.
Results: 152 lesions in this study underwent malignant transformation (n=100 ;16.9%) or further disease (n=52;8.8%), 438 (74.0%) lesions remained disease free. The dynamic nomogram (bias corrected indices: Somers D = 0.75; C index = 0.87) was able to predict further disease or malignant transformation with a sensitivity of 88%, specificity of 93%, negative predictive value of 94% and positive predictive value of 86%.
Conclusions: Malignant transformation of oral potentially malignant lesions can be accurately predicted using a dynamic nomogram and creation of nomograms may help clinical decision making for potentially malignant disorder patients. This nomogram now needs to be validated on larger groups of patients at different clinical units.
Methods: A complete database of 590 oral potentially malignant lesions from 495 patients with single lesion disease and 95patients with multiple lesions diagnosed and followed up at Newcastle Upon Tyne Hospitals Trust (UK) between 1996-2014. All patients received at least 5 years of follow up. The r package ‘rms’ was used to fit a logistic regression model with internal validation and calibration using the bootstrap. The R package DynNom was then used to produce a dynamic nomogram. Outcome measures were continuing disease/malignant transformation versus resolution of lesions. All patients underwent incisional biopsy and subsequent laser surgery treatment for their potentially malignant lesion. The model used predictors of age, gender, clinical appearance of lesion, site, incisional biopsy pathology, laser excision pathology to determine clinical outcome.
Results: 152 lesions in this study underwent malignant transformation (n=100 ;16.9%) or further disease (n=52;8.8%), 438 (74.0%) lesions remained disease free. The dynamic nomogram (bias corrected indices: Somers D = 0.75; C index = 0.87) was able to predict further disease or malignant transformation with a sensitivity of 88%, specificity of 93%, negative predictive value of 94% and positive predictive value of 86%.
Conclusions: Malignant transformation of oral potentially malignant lesions can be accurately predicted using a dynamic nomogram and creation of nomograms may help clinical decision making for potentially malignant disorder patients. This nomogram now needs to be validated on larger groups of patients at different clinical units.