What is the value of pre-surgical variables in addition to cone dimensions in predicting cone margin status?

OBJECTIVE

Previous studies have shown associations between independent pre-conization variables (e.g. smoking, age, cytological grade, menopause) and positive or negative cone margins. However, it is not clear if these pre-surgical variables add predictive value to cone dimensions in the prediction of cone margin status. This study aimed to compare different models predicting positive ecto- or endocervical margins to assess whether pre-conization variables provide significant added value compared with cone dimensions alone.

STUDY

DESIGN

One hundred and sixty-one consecutive women with high-grade cervical intra-epithelial neoplasia on cone specimens were analysed retrospectively. The sample was divided into women with positive ecto- or endocervical cone margins and women with negative ecto- or endocervical cone margins. Pre-conization clinical features, cone volume and cone length were included among the study variables. Multivariate stepwise regression analysis was used to create different models predicting incomplete cervical excision. The added value of pre-conization predictors was measured with receiver operating characteristic (ROC) curve comparisons.

RESULTS

Fifty-seven (35.4 %) women had incomplete cervical excision. Multivariate analysis showed that a positive ectocervical margin was significantly associated with low-grade cervical cytology [odds ratio (OR) = 0.25, 95 % confidence interval (CI) 0.09-0.70] and cone length (OR = 0.69, 95 % CI 0.58-0.82, criterion <9 mm). The area under the curve (AUC) of the combined model for prediction of a positive ectocervical margin was 0.78 (95 % CI 0.70-0.84, p < 0.001). A positive endocervical margin was associated with cone length (OR=0.78, 95 % CI 0.65-0.93, criterion <9 mm) and age (OR=1.07, 95 % CI 1.02-1.11, criterion ≥45 years). The AUC of the combined model for prediction of positive endocervical margin was 0.75 (95 % CI 0.66-0.82, p < 0.001). Comparison of ROC curves showed that the addition of pre-conization variables to cone length did not yield significant predictive results for either ecto- or endocervical cone margins (p = 0.228 and 0.349, respectively).

CONCLUSIONS

The addition of pre-conization clinical variables to cone dimensions did not improve the prediction of cone margin status significantly in the study cohort. Among cone dimensions, cone length was the best predictor of come margin status.