Long-term Survival Outcomes of Invasive Micropapillary Carcinoma of the Breast: A Population-based Study

Background: Whether or not invasive micropapillary carcinoma(IMPC) histology is an independent prognostic factor for breast cancer remains controversial. Moreover, the relationship between different molecular subtypes and survival outcomes of IMPC and invasive ductal carcinoma (IDC) is still unknown. Methods: Using the SEER database to identify breast cancer patients, we retrospectively analyzed 959 IMPC and 174591 IDC cases diagnosed between 2010-2016 with non-metastatic diseases that underwent surgery. Specically, we compared long-term outcomes of breast cancer-specic survival (BCSS) and overall survival (OS). Results: Relative to IDC patients, IMPC patients were younger at diagnosis and had more moderate and poorly differentiated tumors (93.2% vs. 78.5%), more T3 and T4 tumors (11.0% vs. 6.9%), a higher percentage of nodal involvement (48.9% vs. 30.9%) and AJCC stage III patients (11.9% vs. 6.9%), and presented a higher proportion of HR positivity (91.2% vs. 82.3%) and HER2 positivity (22.3% vs. 16.9%). IMPC had a better BCSS (P=0.039) but showed no signicant difference in OS (P=0.095) compared with IDC. In multivariate Cox analysis, IMPC histologic type was an independent favorable prognostic factor for both BCSS (HR=0.509, P=0.002, 95%CI: 0.335-0.775) and OS (HR=0.637, P=0.003, 95%CI: 0.475-0.854). After the case-control matched analysis using the propensity score matching method, IMPC still had a better BCSS (P=0.001); however, we observed no signicant difference in OS (P =0.385). While different molecular subtypes have different impacts on survival outcomes, no signicant differences were observed in BCSS and OS between IMPC and IDC in relation to Luminal B, HER2-enriched, and Triple-negative subtypes. However, in relation to the Luminal A subtype, IMPC had better BCSS (HR= 0.399, P=0.001, 95%CI: 0.226−0.703) and OS (HR=0.508, P=0.001, 95%CI: 0.345−0.746). In the case-control cohort, IMPC still had a better BCSS (HR= 0.423, P=0.005, 95%CI: 0.233−0.770), but no signicantly difference was observed in OS (HR=0.767, P=0.22, 95%CI: 0.502−1.172) in Luminal A subtype. Conclusion: Relative to IDC, IMPC presents better long-term survival outcomes, and the survival benets are conned to the Luminal A subtype.


Introduction
Invasive micropapillary carcinoma (IMPC) is a relatively rare histologic subtype of invasive breast cancer, accounting for approximately 1.0-8.4% of all breast cancer cases [1][2][3][4][5] .IMPC was rst described in 1993, and classi ed as an independent breast tumor by the World Health Organization in 2003 [6,7] .Pathologically, IMPC is known for a high incidence of lymphovascular invasion (LVI) and lymph node (LN) metastasis, and typically shows a more aggressive behavior than invasive ductal carcinomas (IDC). [1,[8][9][10][11] However, the utility of IMPC histology as an independent prognostic factor for breast cancer remains controversial. Several studies indicate that there is no signi cant difference in survival outcomes between IMPC and IDC patients when matched for lymph node status [1,12] .There is still controversy about the comparison of breast cancer-speci c survival (BCSS) and overall survival(OS) between IMPC and IDC [8,9,13] . Moreover, some studies have demonstrated similar BCSS and OS between IMPC and IDC.
However, few previous studies have shown a detailed comparison based on breast cancer molecular subtype. This is likely because the breast cancer molecular subtypes have only been proposed in the past 20 years. Therefore, we conducted this study to compare the survival of IMPC and IDC in different molecular subtypes of breast cancer utilizing the Surveillance, Epidemiology, and End Results (SEER) database.

Materials And Methods
We retrospectively analyzed the SEER database released in 2019, including cases of primary breast cancer diagnosed between 2010-2016. Our inclusion criteria were female patients with histologically con rmed unilateral invasive ductal carcinoma (IDC) or invasive micropapillary carcinoma (IMPC), who were treated with surgery and without metastasis diseases at diagnosis. We excluded patients with more than one primary cancer or unknown data for a parameter of interest. The recorded patient characteristics included age at diagnosis, race, laterality, histologic grade, tumor size (T stage), lymph nodes positivity (N stage), staging (using AJCC 7th edition), hormonal receptor (HR) status, HER2 status, molecular subtype, and type of surgical procedure. Borderline estrogen receptor(ER) or progesterone receptor(PR) status was considered positive, and HR-positive was de ned as ER and/or PR positive. Borderline HER2 status was de ned as missing data. Comparisons of the characteristics between IDC and IMPC were performed using Pearson's Chi-square test. A univariate analysis using Kaplan-Meier survival curves was utilized to estimate BCSS and OS outcomes, and log-rank tests were performed to compare the two groups.
Potential prognostic variables, including histologic subtype, age, race, laterality, grade, T stage, N stage, HR status, HER2 status, and surgery, were calculated using a Cox proportional hazard model for univariate and multivariate analyses. Subgroup analysis based on molecular subtype was performed to compare survival outcomes; for visualization purposes, the interaction effects of each molecular subtype on BCSS and OS were displayed using forest plots. To further diminish the effects of baseline difference between two groups, we applied the propensity score matching method. Speci cally, each IMPC patient was matched to 3 IDC patients with similar characteristics, and survival outcomes of the matched groups were compared using the above methods repeatedly. Statistical analyses were performed using R statistical software (version 3.6.1, R Project for Statistical Computing, Vienna, Austria). A two-sided P value of<0.05 was considered statistically signi cant.

Survival outcomes
For the total study population, the median follow-up was 39 months. A variety of potential prognostic variables were investigated for univariate and multivariate Cox analyses. Among these, age, race, tumor grade, T stage, N stage, HR status, and surgical type were found to be independent prognostic factors for both BCSS and OS. In multivariate Cox analysis, IMPC histologic type was an independent favorable prognostic factor for both BCSS (HR=0.509, P=0.002, 95%CI: 0.335-0.775) and OS (HR=0.637, P=0.003, 95%CI: 0.475-0.854) ( Table 2 and 3). According to the comparisons between IMPC and IDC for the total population, IMPC displayed better BCSS (P=0.039), but showed no signi cant difference in OS (P=0.095) (Figure1). A 1:3 matched case-control analysis was conducted to control for the effects of baseline differences between IMPC and IDC patients. A total of 959 IMPC patients were matched to 2877 IDC patients based on age, race, laterality, grade, T stage, N stage, HR status, HER2 status, molecular subtype, and surgery type. In each characteristic, no signi cant difference was observed between IMPC and the matched IDC cohort (Table 4). After the matched analysis, IMPC still had a better BCSS (P=0.001) but shown no signi cant difference in OS (P=0.385) compared with IDC ( Figure 4).
In the subgroup analysis based on molecular subtype, different molecular subtypes were shown to have varying impacts on survival outcomes. For the total study population, no signi cant differences in BCSS and OS were observed between IMPC and IDC patients relative to Luminal B, HER2-enriched, and triplenegative subtypes. In comparing within the Luminal A subtype, IMPC had better BCSS (

Discussion
IMPC is a rare type of breast cancer. Previous studies have shown that IMPC is more aggressive than IDC, with a worse prognosis. Moreover, IMPC typically has a high association for lymphovascular invasion (LVI) and lymph node (LN) metastasis. Many researchers consider IMPC a high-risk factor. However, previous studies have also suggested that IMPC has better long-term survival than IDC [14] . IMPC has favorable BCSS and OS rates [15] . Few studies have focused on the correlation between IMPC and survival due to a relatively low incidence. Moreover, there remains great controversy regarding this topic. Since the concept of breast cancer molecular subtype has been put forward in the past 20 years, most previous studies have not analyzed the survival based on molecular subtype. Therefore, we designed this study to analyze the relationship between survival of IMPC and IDC based on the molecular subtypes of breast cancer.
Using a Kaplan-Meier analysis to compare the survival of the two cancer types, we found differences in BCSS between the IMPC and IDC groups. However, we observed no effects on OS. This is consistent with the previous results of Chen et a [3,4] l. Moreover, research from Yoon GY et al. was in agreement that the IMPC group showed a worse total recurrence-free survival(RFS) but found no signi cant difference in OS as well [20] . In their research, the IMPC group showed worse total RFS (hazard ratio [HR]=1.63, P=0.016), local RFS (HR=2.86, P=0.042), and distant RFS (HR=1.85 P=0.018), but there was no signi cant difference in OS(HR=1.30, P=0.335). Other studies also suggested a lack of difference in BCSS and OS between the two groups in stage I breast cancer patients [14] . Hongliang Chen found that IMPC was correlated with aggressive clinical characteristics, such as larger tumors, more positive lymph nodes, and more advanced stage, relative to IDC. Additionally, a higher rate of ER/ PR positivity was observed in IMPC. These results are consistent with the research reported here. Further more, in a case-control analysis, the Hongliang Chen research still showed that IMPC was an independent favorable prognostic factor for BCSS (HR = 0.410, P < 0.001, 95% CI: 0.293-0.572) and OS (HR = 0.497, P < 0.001, 95% CI: 0.387-0.637). In subgroup analysis, IMPC routinely shows better survival outcomes compared with IDC, except for AJCC stage I and histologic grade I disease. However, data from Hao et al. suggested that there was a lack of difference in prognostic power between the two groups after the application of dependency matched analysis [21] . The survival analysis revealed no signi cant reduced overall survival (p = 0.752) or disease-free survival (p = 0.578) in IMPC patients. Moreover, multivariate Cox regression analysis revealed that IMPC was not an independent prognostic factor for disease-free survival (hazard ratio [HR] = 0.944; 95% con dential interval [CI], 0.601-1.481) or overall survival (HR = 0.727; 95% CI, 0.358-1.478).
We further analyzed breast cancer molecular subtype dependent prognosis. We found that BCSS and OS were different between the IMPC and IDC group only in Luminal A subgroup, but not in Luminal B, triplenegative, and HER2-enriched subgroups. Lewis et al. also concluded that the prognosis of Triple-negative IMPC is as poor as that of IDC [22] .
Considering the unbalanced data between the IMPC and IDC groups, propensity score matching (PSM) was used to adjust for potential baseline confounding between groups. After PSM, we still observe a difference in BCSS, but not OS, between the IMPC group and IDC group. Analysis by molecular subtype found that patients with Luminal A alone had differences in BCSS between the IMPC group and IDC group, yet no difference in OS. There is no difference in BCSS and OS between Luminal B, triple-negative, and HER2-enriched subtypes. Therefore, clinical characteristics consistency leads to a similar prognosis between the IMPC group and IDC, except for the Luminal A subtype. In our understanding, the Luminal A breast cancer subtype has a good prognosis, and most cases require only endocrine therapy without chemotherapy. In conclusion, the prognosis of IMPC of Luminal A is better than that of IDC.

Conclusion
In conclusion, IMPC is correlated with a favorable survival outcome compared with IDC. In addition,