Page 108 - Strategies for non-invasive managementof high-grade cervical intraepithelial neoplasia - prognostic biomarkers and immunotherapy Margot Maria Koeneman

Chapter 5
Discussion
This study describes the development and internal validation of a prediction model for spontaneous regression of CIN 2 lesions. The model consists of four simple clinical parameters, including smoking habits, the result of the last PAP smear, concomitant CIN 1 and CIN 2 in one biopsy, and multiple CIN 2 lesions in different biopsies. This prediction model has reasonable discriminative capacity and is accurate, especially in the prediction of disease regression. As such, it can be applied to reassure patients that observational management can be a good option when the model indicates a high probability of regression.
The prediction model was designed for CIN 2, but not CIN 3, because overtreatment is more common for CIN 2 than for CIN 3. This is due to the fact that CIN 2 has a higher spontaneous regression rate than CIN 3 [5]. Indeed, international guidelines also advocate conservative treatment of younger women with CIN 2.[7, 8] In previous studies on the conservative treatment of CIN 2, no women progressed beyond CIN 3 within 12 months [6, 23, 24]. This was confirmed in our study, which showed no disease progression beyond CIN 3 within 24 months of follow-up. These findings indicate that observational management of CIN 2 with adequate follow-up is safe and may prevent overtreatment. It must be noted, however, that discussion has evolved as to whether CIN 2 and CIN 3 are truly two grades of the same pathological condition. Additional studies on epigenetic changes may subdivide CIN lesions into early or advanced lesions, carrying a lower or higher risk of malignant transformation, regardless of the conventional histopathological diagnosis.[25] However, such tests are currently not available in clinical practice. Since the differentiation between CIN 2 and CIN 3 is still based on histopathology, this study focused on CIN 2 because of its high spontaneous regression rate and the safety of conservative management.
The prediction model aims at widespread applicability by using simple clinical parameters as predictors. Two previous models were developed to predict the risk of progression and regression of CIN 1-2 and CIN 2-3 lesions, but included several biomarkers for which additional immunohistochemical staining was necessary.[1, 26] Both models have good predictive values, but utilize predictors that are not readily available and require additional tests. This reduces the clinical applicability of these models. For the same reason, HPV genotyping was not included in our model as a predictor. Indeed, the presence of high-risk HPV types in CIN 2 lesions are strongly predictive for non-regression, but HPV genotyping is not routinely performed at all medical institutions.[2, 27, 28] Although most countries and institutions will apply HPV-based cervical cancer screening in the future, cervical cytology will be used in clinical practice outside screening programs and in those institutions where HPV-based cervical cancer screening cannot be applied. Our prediction model is based entirely on simple clinical parameters and is consequently widely applicable in different patient populations.
Interestingly, age was dropped from the prediction model during backward stepwise deletion. As mentioned before, previous research suggests an increased probability of disease progression of LSIL with increasing age and that HPV clearance is more common in younger patients. In our cohort, age did not influence the outcome (data not shown). Although age is not part of
106