There are various reports about risk factors for finding IDC by postoperative pathological examination in cases diagnosed as DCIS by preoperative biopsy. A meta-analysis lists the following items as risk factors: biopsy devices, high-grade DCIS, tumor larger than 20 mm, palpability, and others [1]. There are also reports that list negative hormonal receptor [20, 21], HER2 overexpression [22, 23], and lymphoid infiltrate as risk factors [18, 19]. However, other reports do not identify these clinicopathologic features as risk factors [24]. No study reporting high Ki67 as a risk factor was found, but it was thought that DCIS with HER2 overexpression or high Ki67 had a potential risk of invasion because these features represented a high risk for postoperative DCIS recurrence [25–27].
In this study, the same clinicopathological factors as those previously reported were identified as risk factors for finding infiltration after surgery. However, the patients included in this study differed from the patients of previous reports in terms of clinicopathologic features. An important difference is the rate of IDC found by postoperative pathological examination. In the meta-analysis previously reported, the ratio was 25.9% (18.6–37.2%) [1], while we found a higher rate of 39.1%. We presume that this difference is due to the biopsy device. We have been using 16-gauge CNB because of the physical burden on patients. As a result, IDC was found by postoperative pathological examination at a high-rate of 47.9% in cases diagnosed by CNB, which may have affected the overall rate of IDC found by postoperative pathological examination.
Another difference is that fewer patients had the clinicopathological features listed as risk factors. For example, the palpable tumor rate was 8.8-fold higher in this study compared with a previous report [1]. Similarly, the overexpression rate of HER2 was 14.8% in this study and 28–65% in a previous report [28]. In addition, high-grade DCIS corresponded to 49.4% of cases in a previous report [1] but only 18.8% in this study. These differences may exist due to the low rate of breast cancer screening in Japan, which is lower than that in other countries at 40% [29]. The fact that various risk factors are lower than those reported previously may indicate that, in many cases, invasion already occurred at the time the patient presented to the hospital.
In this study, current-smokers were significantly less likely to find postoperative IDC compared to never-smokers and former-smokers. In addition, the risk of finding IDC decreased with increasing daily smoking and smoking duration, although these were not significant. This is supported by the results of a case-control study. In that study, the current status as a smoker was more important for the onset of DCIS than the exposure to tobacco, and the result was an inverse correlation [8]. However, experiments in vivo and in vitro have shown that tobacco components increase the malignancy of breast cancer cells [3–5]. In addition, we have shown in clinical samples that smoking can enhance HER2 expression in breast cancer and increase tumor-infiltrating lymphocyte density in the microenvironment surrounding the cancer [16, 30]. From these results, we speculate that smoking may affect DCIS during the beginning of the disease, causing it to acquire invasive ability at an early stage and resulting in IDC identified by biopsy at initial diagnosis. Our speculation is supported by a case-control analysis that listed smoking as a risk of developing luminal A IDC, while it was associated with reduced risk of developing luminal A DCIS; however, the associations were not significant [31].
Some of the limitations of this study are the higher rate of postoperative IDC detection and the fewer number of patients with risk factors than those previously reported. In addition, the data regarding smoking habits may not be entirely accurate because it was based on self-reporting. Because some studies listed younger age as a risk factor for finding invasion by postoperative pathological examination [32–34], the fact that the age distribution was disproportional between smoking and non-smoking patients may represent a small limitation.