Tumour-free distance: a novel prognostic marker in patients with early-stage cervical cancer treated by primary surgery

Abstract

Background: Models predicting recurrence risk (RR) of cervical cancer are used to tailor adjuvant treatment after radical surgery. The goal of our study was to compare available prognostic factors and to develop a prognostic model that would be easy to standardise and use in routine clinical practice.

Methods: All consecutive patients with early-stage cervical cancer treated by primary surgery in a single referral centre (01/2007-12/2016) were eligible if assessed by standardised protocols for pre-operative imaging and pathology. Fifteen prognostic markers were evaluated in 379 patients, out of which 320 lymph node (LN)-negative.

Results: The best predictive model for the whole cohort entailed a combination of tumour-free distance (TFD) ≤ 3.5 mm and LN positivity, which separated two subgroups with a substantially distinct RR 36% and 6.5%, respectively. In LN-negative patients, a combination of TFD ≤ 3.5 mm and adenosquamous tumour type separated a group of nine patients with RR 33% from the rest of the group with 6% RR.

Conclusions: A newly identified prognostic marker, TFD, surpassed all traditional tumour-related markers in the RR assessment. Predictive models combining TFD, which can be easily accessed on pre-operative imaging, with LN status or tumour type can be used in daily practice and can help to identify patients with the highest RR.

Conflict of interest statement

The authors declare no competing interests.

Figures

Fig. 1

Study flow chart.

Fig. 2. Tumour-free distance (TFD) assessment.

a Scheme of TFD measurement as the minimal lateral distance of uninvolved stroma between the tumour and pericervical ring; b TFD measurement by ultrasound in a transversal plane; c TFD measurement on pathological specimen.

Fig. 3. Accuracy of prognostic models for recurrence risk in the whole cohort (ROC analysis).

Model 1: Minimal tumour-free distance (TFD), number of positive lymph nodes (LN); Model 2: TFD binarised, LN positivity binarised; Model 3: TFD binarised, LN positivity binarised, lymphovascular space invasion; Model 4: TFD binarised, LN positivity binarised, depth of stromal invasion; Model 5: TFD binarised, LN positivity binarised, tumour type; Model 6: largest tumour size binarised, LN positivity binarised.

Fig. 4. Kaplan–Meier recurrence-free survival (RFS) curve for Model 2 in the whole cohort.

Group description: Group 1: tumour-free distance (TFD) > 3.5**, N0**; Group 2: TFD ≤ 3.5**, N0**; Group 3: TFD ≤ 3.5**, N1**. Two patients without event are not included in the groups: minimal tu-pcf (US) > 3.5**; a number of positive LN > 0**. *Log-rank test. **Cut-off determined by ROC analysis, the criterion was the highest value of the sum of sensitivity and specificity.

Fig. 5. Accuracy of prognostic models for disease recurrence in lymph node (LN)-negative patients (ROC analysis).

Model 1: minimal tumour-free distance (TFD), tumour type; Model 2: TFD binarised, tumour type; Model 3: TFD binarised; tumour type, lymphovascular space invasion; Model 4: TFD binarised, tumour type, depth of stromal invasion; Model 5: largest tumour size binarised, tumour type; Model 6: GOG score.

Fig. 6. Kaplan–Meier recurrence-free survival (RFS) curve for Model 2 in lymph node-negative patients.

Group description (number of patients related to this characteristic is in the bracket): Group 1: other combinations not included in Group 2; Group 2: tumour-free distance ≤3.5**, adenosquamous tumour type. One patient without event is not included into the groups: Minimal tu-pcf (US) > 3.5**, adenosquamous tumour type. *Log-rank test. **Cut-off determined by ROC analysis, the criterion was the highest value of the sum of sensitivity and specificity.