Trouver un article spécifique

Validation of nomograms to predict the risk of positive margins following breast-conservative surgery north african tunisian breast cancer population

  • S. Hidar
  • F. Hachani
  • R. Briki
  • S. Chechia
  • S. Bouguizane
  • M. Bibi
  • A. Khlifi
  • H. Khaïri

Mots clés:

marges chirurgicales
cancer du sein


Breast-conserving surgery (BCS) is the treatment of choice for both early stage invasive and in situ breast carcinoma. It has been clearly demonstrated that except for local recurrence, BCS provides similar outcomes (disease free survival and overall survival) to modified radical mastectomy [1, 2]. There is a general agreement that the most important risk factor for local recurrence after BCS is close or positive margins. [3]. After BCS, re- excision of positive margins can reach up to 50% or even more [4]. Positive margins may also indicate radical surgery with its consequences on anxiety, esthetic results, adjuvant radiotherapy and medical costs. Therefore, it seems useful to identify patients at risk of positive margins to individualize surgical plan for breast cancer patients who are scheduled for BCS. To achieve this, nomograms and scoring systems based on clinical and pathological variables have been developed [5, 6]. In 2012, Shin H-C et al.[5] developed, calibrated and externally validated a nomogram using 5 variables that were preoperatively determined to predict positive resection margins. More recently a preoperative nomogram was developed by Pleijhuis R.G et al. [6] with this aim. Both nomograms were externally validated but not on a North African population in whom data suggest some specific features that may limit systematic extrapolation of Western data.[7, 8]. We therefore we aimed to assess and compare the accuracy of these two nomograms in our population.


Standard of care for early stage breast carcinoma includes lumpectomy. The most important risk factor for local recurrence is close or positive margins. Identification of at risk group for this situation could allow individualiza

Matériel et Méthode:

A retrospective review of our database of breast conserving surgery was performed. Data included all patients who underwent initial BCS at F. Hached university teaching hospital (Sousse, Tunisia) performed by a single operator (SH) between 2007 and 2014. Similarly to reference studies [5, 6], only women with preoperative histological confirmed diagnosis, T1 or T2, M0 tumors and without neoadjuvant chemotherapy were included. Analyzed variables were those included in the above mentioned nomograms: pre operative MRI (when performed), MRI or US tumor size, presence of micro calcifications, preoperative N and T stages (clinically and radiological size were reported) breast density (when reported), palpability, suspicion of multifocality, Estrogen Receptor status, presence of DCIS (at preoperative biopsy and or on postoperative specimen), histological type (WHO classification), presence of lobular component on needle biopsy and mSBR grade by Elston and Ellis. Positive surgical margin was defined as microscopically malignant cells at any of the specimen margin(s). For each nomogram, patients with missing data required for probability estimation were excluded. We used the nomogram provided by Shin H-C et al. and the online advanced version of the nomogram provided by Pleijhuis R.G et al. ( - accessed April 2014). Both tools allow probability estimation of positive margins. Since absence of DCIS in the core needle biopsy results in high false negative findings Pleijhuis R.G et al. nomogram’s provides a probability interval in such cases with higher and lower limits. Data are reported as frequencies and percentages, means and standard deviations or their 95% confidence intervals (CI) when appropriate. For both models a Receiver Operating Curve (ROC) was drawn and the Area under the Curve (AUC) was calculated to assess its discriminative power. Fisher’s exact test was also used. Statistical analyses were conducted by SPSS 13.0 and statistical significance was set at p


The data of one hundred fourteen patients were analyzed with Pleijhuis’s nomogram and 100 for Shin’s one. Positive margins were present in 15.8% (18/114). Table 1 summarizes characteristics of the study population. Characteristics (n=114) Age 51.3± 11.5 Preoperative MRI (%) 20 (17.5) Breast density (%) 0-25% 13(11.4%) 25-50% 40(35.1%) 50-75% 38(33.3%) 75-100% 14(12.3) Unkonw 9(7.9) Pathology (%) Invasive ductal 95(83.3) Invasive lobular 11(9.6) Others 8(7) Ductal carcinoma In Situ 7(6.1) S.D : Standard Deviation Table 1 : Descriptive characteristics of the study group. (n=114) When measured by the AUC, predicted probabilities for Pleijhuis’s lower limit, Pleijhuis’s higher limit and Shin’s nomograms were respectively 0.817 (95%CI 0.72-0.913), 0.832 (95%CI 0.728-0.935) and 0.713 (95%CI 0.568-0.819). Interestingly, No patient with a higher limit 50% (14%:14/100patients) allowed detection of 55% (10/18) of patients with a positive margin. Finally, a limit > 42% (36/100 patients) was needed to achieve a detection rate of 55% (10/18) of patients with a positive margin.


Despite controversies surrounding definition of positive margins, there is a general agreement that this parameter is the most important predictive factor of local recurrence after breast conserving surgery[3]. Several authors created and validated nomograms and scoring systems to calculate specific risks in patients with breast cancer conditions e.g non sentinel node involvement when sentinel node is involved [9] or minimally involved [10]. Probability of invasive cancer in patients with DCIS preoperatively diagnosed [11, local recurrence [11] or ipsilateral breast cancer development [12]. The objective of these tools is to act as an aid decision for breast cancer management but there is an absolute need to validate them in the targeted population before clinical use. Despite progressive westernization of North African lifestyle population, breast cancer in these populations still displays some specific features compared to western population. Proportion of triple negative breast cancer is higher [13], affected patients are younger, tumor size at diagnosis is larger and genome profiling expresses some differences. [7]. Furthermore recent data demonstrated that North African immigrant patients in Western Europe display different immunohistochemical profile of the molecular classes[8]. From a statistical point of vu comparing nomogram accuracy is essentially based on the Area Under the Curve (AUC). An AUC 0.8 represents excellent discrimination.[14]. The aim of our study was to validate and compare two nomograms in our North African population. Both were find to be accurate but the AUC of Pleijhuis’s nomogram tend to be higher than the AUC in shin’ s and al. model (0.817 (95%CI 0.72- 0.913) and 0.832 (95%CI 0.728-0.935) versus 0.713 (95%CI 0.568-0.819 respectively). The two above mentioned nomograms [5, 6] were recently developed and arise from two distinct geographic area (Nederland and Korea). Pleijhuis’s nomogram [6] was developed, calibrated and externally validate in a Dutch population (n=1516) that underwent conservative breast cancer surgery between 2004 and 2011. An extracted statistical formula is implemented in a web- based calculator ( Expanded and basic versions are available; the last is used when preoperative core needle biopsy result is not available Shin’s nomogram [5] was developed and validated in Korea (n=1597) .The results were published in 2012 and a graphical calculator is provided for probability estimation. In our opinion Pleijhuis’s nomogram has advantages when compared to Shin’s nomogram: First, it can be calculated even when some data are missing which is not the case for shin’s nomogram. Secondly it’s web based and friendly to use. Considering a threshold ents (12.2%); this might lead to misestimating nomogram accuracy. Finally, in our population, only a small proportion of patients were preoperatively diagnosed with DCIS (6/114- 5.2% ) and previous studies [5, 6, 15] demonstrated DCIS as strong predictor of positive margins. This low incidence might have overestimated nomograms accuracy. On the other side, our study has some strength, it assessed real life patient and to the best of our knowledge, it is the first to compare and to validate nomograms for predicting positive surgical margins following conservative breast surgery in North African population. We conclude that both nomograms can be of help in clinical practice when assessing Tunisian - North African population or newly arrived immigrants from this area in western countries.