Page 103 - Strategies for non-invasive managementof high-grade cervical intraepithelial neoplasia - prognostic biomarkers and immunotherapy Margot Maria Koeneman

Potential predictors
Potential predictors of response were pre-selected based on expert opinions and published
studies reporting factors that influence the natural course of CIN, as is advised in methodological
guidelines for prognostic modeling. This results in higher external validity than selecting factors
that are significantly related to outcome in the same dataset [11, 12]. Evidence regarding potential
predictors of disease regression in CIN 2 is scarce. Moreover, most studies examined immunological
and histological biomarkers and not simple clinical parameters [10]. We aimed to develop a
widely applicable prediction model, and therefore limited the potential predictors to readily
available variables from medical history, such as patient demographics and characteristics of the
cervical cytology and biopsy specimens. The final set of potential predictors that was considered
for the model included age at diagnosis, smoking, the result of the PAP smear immediately before
the CIN 2 diagnosis, the number of individual CIN 2 lesions, and the existence of concomitant 5 CIN 1 and CIN 2 in the same biopsy. Younger age is associated with an increased rate of human
papilloma virus (HPV) clearance after treatment [13]. Older age is associated with a higher risk of progression or persistence than younger women and with a higher incidence of residual disease after non-radical surgical excision of high-grade CIN [14, 15]. The severity of pre-treatment cytological abnormalities is associated with residual disease after non-radical surgical treatment of high-grade CIN [14]. Milder cytological abnormalities may indicate that a CIN 2 lesion is low- risk (more often hr-HPV negative). Indeed, it was recently shown that the risk of cancer in women with a test result of HPV negative abnormal squamous cells of unknown significance (ASCUS) was comparable to that in women with a negative PAP test result [16]. Smoking is associated with an increased risk of cervical cancer, a higher risk of developing high-grade CIN in patients with abnormal cervical cytology, and disease persistence after surgical treatment [17-19]. The amount of individual CIN 2-lesions was interpreted as comparable to a larger lesion size, which has been associated with non-regression of high-grade CIN [1]. Finally, concomitant CIN 1 and CIN 2 in the same biopsy is a frequent event. Evidence of its prognostic value with regard to disease regression is lacking, but such lesions may be low-risk, with a natural history similar to that of low-grade lesions.
Sample size
At least ten events should be collected for each potential predictor that is to be evaluated in the multivariable regression analysis [20]. An event is defined as the least frequent outcome status. A recent prospective randomized controlled trial evaluating spontaneous regression of CIN 2 after 24 months showed disease regression in 40% of patients [21]. In order to develop a model with five potential predictors, at least 50 events would be required, which correlates to a sample size of at least 125 (50/0.4).
Data collection, data quality and missing data
Patients were recruited from the hospital pathology database. Patient data regarding demography and medical history were extracted from medical records by a trained senior medical student (FL). In case of doubt, data were interpreted by a second person (MK, medical doctor). Data were checked for completeness and inconsistencies (MK). Any inconsistencies or incomplete data were double-checked and corrected or completed if possible. Missing data were imputed using a
A prediction model for CIN2 prognosis
101