How to accurately preoperative screen nipple-sparing mastectomy candidate—a nomogram for predicting nipple-areola complex involvement risk in breast cancer patients

The surgical treatment of breast cancer has evolved from maximal radical cure to maximal preservation of shape and function. Breast surgeons and patients pay attention to the expected survival time while taking into account the improvement of postoperative quality of life [1]. The surgical treatment of choice for most breast cancer patients is breast-conserving therapy. However, total mastectomy is still required in 30 to 40% of all breast cancer patients who undergo surgery [2, 3]. Nipple-sparing mastectomy (NSM), which evolved from skin-sparing mastectomy, is characterized by the preservation of the entire nipple-areola complex (NAC) and breast skin envelope despite the removal of the mammary tissue [4, 5]. NSM is a surgical procedure that improves patient satisfaction with cosmetic outcome [6].

However, previous studies have found that the rate of occult (pre)-malignant invasion of a clinically normal nipple in mastectomy specimens ranged from 0 to 58% [7, 8], implying that there was a risk of leaving an occult malignant tumor inside the NAC during NSM [9]. The main concern associated with NSM is the risk of local breast cancer recurrence at the retained NAC consequent to occult nipple involvement. As the oncology safety of NSM has been proven in multiple prospective and retrospective studies [1, 10‐12], increasing numbers of patients with breast cancer are selecting NSM, the indications for NSM should be thoroughly reviewed.

NSM was provided to patients with a clinically normal NAC and no skin involvement. Although anomalous NAC clinical signs (inverted nipple, bloody nipple discharge, abnormal NAC skin, and so on) are frequently thought to be significant evidence of NAC invasion, there is presently no consistent and efficient preoperative evaluation technique for clinically occult NAC invasion. Previous research has found that tumor size, tumor-to-nipple distance, lymph node status, and tumor location are all clinical characteristics that may be linked to nipple-areola complex involvement (NACi) [9, 13]. Patient selection based on clinicopathologic criteria, on the other hand, is debatable, because a clinically negative NAC symptom can rule out occult nipple involvement with a high negative predictive value, even in individuals at high risk [14]. Furthermore, in clinical practice, differing measuring methods and categorization criteria have a significant impact on NACi outcomes. Tumor nipple distance (TND) is thought to be the most closely associated with NACi; however, different assessment methods, such as ultrasound (US), mammography (MG), and magnetic resonance imaging (MRI), have varied clinical implications. [15‐17]. MRI is currently a more accurate testing approach for evaluating the abnormalities of NAC. Hirohito Seki and colleagues produced good results on the aforementioned topic and established two separate models for the NACi risk problem, the first of which was built based on clinical and imaging data and had an accuracy of 93.5%. The second model was based on MRI imaging data for prediction, with an accuracy of 96.0%, indicating that both models had a strong predictive ability [9, 18]. Patients who are unable to undertake or finish MRI exams within the effective period, on the other hand, are not in the minority. Based on this, we created a more easy, simple, and accurate visual model prediction tool using non-MRI data.

Nomograms based on clinical, imaging, and pathology data have been proven to provide more accurate predictions of tumor risk and treatment results for individual patients [19, 20]. We developed a nomogram based on preoperative indications to predict the likelihood of NACi in breast cancer patients in this study.

At present, the commonly used clinical prediction tools for NACi include clinical physical examination, imaging, and pathological evaluation. Although the criteria used to select NSM for IBR have broadened over time, the consensus has not yet been reached regarding this issue [1, 9, 15, 17, 18]. Previous studies indicated that tumor size, TND, lymph node status, lymphovascular invasion, and tumor location were closely related to NAC invasion. Pathological indicators such as pathological type, histological grade, ER/PR, HER2, and KI-67 showed great differences in different studies [25‐27]. However, predicting NACi risk using these recommendations alone is problematic since some of the factors cannot be evaluated preoperatively.

There is an urgent need for really simple and reliable prediction models that can be used in clinical practice. Hirohito Seki and colleagues have carried out a series of studies and achieved good prediction results [9, 18]. CTS, TND by MG/MRI, nipple contrast findings by MRI, tumor in a central portion, multicentric/focal lesion, and cN positive were significantly associated with the presence of NAC involvement. Part of the experimental results was validated in our study. And they developed a more accurate preoperative prediction model based on enhanced breast MRI. The experimental results showed that the model achieved sensitivity 89.4%, specificity 97.6%, PPV 89.4%, NPV 97.6%, FNR 10.6%, FPR 2.4%, and accuracy 96.0%. Unfortunately, it did not achieve the desired results owing to radiologists having great differences in the assessment of some indicators. Nevertheless, most clinicians believe that MRI may not be so necessary for patients who intend to perform NSM. In addition, the high expenses and time costs of MRI limit the popularity of the model [28]. Therefore, we have developed a prediction model based on general clinical imaging and pathological indicators. This nomogram prediction model makes the results more concise and visual, which has strong clinical operability.

Published studies show larger CTS and smaller TND are always more likely to lead to NAC positive, but different tumors in imaging, especially when combined with carcinoma in situ components, show great differences [26, 29]. However, there was no consistent conclusion in our study, which may be due to the underestimation of preoperative hollow needle aspiration in the diagnosis of CIS. To achieve better consistency, we improve the CTS/TND measurement idea in this study. MG and US measurements were recorded separately, in which CTS recorded the maximum value and TND recorded the minimum value of the two measurement methods. In the literature, the optimal cutoff distance varies. The optimal CTS cutoff value achieved with the ROC curve analysis in this study was 3.78 cm, which was consistent with earlier research [9, 30]. In this investigation, we used a threshold value of 4.0 cm for grouping. At the same time, while the optimal TND cutoff value was 1.2 cm, we used 1.0 cm for grouping, which was consistent with the majority of the literature [18, 31].

In this study, according to the binary logistic regression model, it was concluded that CTS > 4.0 cm, TND ≤ 1.0 cm, cN( +), tumor central location, and ki-67 ≥ 20% were independent predictors of NACi risk. Although we analyzed more histopathological variables, we did not find significant associations between these factors and NAC involvement. The nomogram was produced drawn on different weight coefficients, and the model accuracy rate was 0.776 in the development group and 0.843 in the validation group. This may be the first visualization model for NACi risk prediction to our knowledge, the internal validation results of both groups were good. This model can visually analyze the data well, score each independent predictor separately, and calculate predicted probability. This simple and easy visualization process greatly facilitates preoperative doctor-patient dialogue and the prediction of risk assessment.

The occult positive rate of NACi was 11.1% (64/578) in the development group and 12.5% (14/112) in the validation group, with no significant difference between the two groups. This indirectly demonstrates that IFP can accurately estimate the potential of NACi [32, 33]. The diagnostic accuracy of IFP was 92.9%, sensitivity was 64.3%, specificity was 96.9%, the false-positive rate was 3.1%, and the false-negative rate was 35.7%. We noticed that the false-negative rate was rather high, with the false-negative rate of IFP reported in prior literature ranging from 0.7 to 54.5% from 2008 to 2021 [18, 34‐36]. As a result, other surgeons did not approve that IFP was required [18, 37].

We further conducted stratified analysis according to the model risk and set NACi risk < 10% as low-risk, 10–50% as intermediate-risk, and > 50% as high-risk. According to our research data, in the low-risk group, both IFP and the model had a certain false-negative rate, and IFP was higher, and the false-positive rate of both was very low. This may explain that the model can be complementary to IFP improvements in some patients. Because a higher false-positive rate means that more NAC may be resected incorrectly, the particularly pernicious false-positive rate is significantly greater than that of reoperation due to false negatives.

In our study, the results suggested that the IFP diagnosis was negative in 5 patients, while the postoperative pathology was positive. One case of DCIS and 4 cases of lobular carcinoma in situ (LCIS) was diagnosed by postoperative pathology. For NAC diagnosed as LCIS, clinical management is relatively difficult, if LCIS is regarded as a precursor of invasive carcinoma [38]; thus, the NAC should be excised. Rachel E.K [32] approved that irrespective of whether LCIS is viewed as a “marker” or as a “precursor” lesion, recommend NAC resection in such cases since there is inadequate follow-up data recognizing that others may hold a contrary view. If LCIS is excluded, then the FNP of IFP is 7.14% in our study (versus 35.7% if LCIS is included).

The model’s value was that it could appropriately screen the absolute low-risk and high-risk groups of NAC involvement preoperatively. For patients with NACi risk < 10% and IFP ( −) can implement NSM securely, while patients with NACi risk > 50% and IFP ( +) should be recommended SSM. For patients with intermediate-risk, or NACi risk < 10% and IFP ( +), or NACi risk > 50% and IFP ( −), these three scenarios should recommend patient-physician shared decision making. This prediction model is simple to calculate, which greatly facilitates the judgment of clinical surgeons in the assessment of the positive risk of NACi.

In this study, we developed a NACiPM based on non-MRI data, an accurate and visual predictive model in NSM + IBR. The NACiPM can be used to distinguish the low, intermediate, and high risk of NAC before surgery. Combined with IFP, we can develop a decision-making system for the implementation of NSM.