Comparison of Overall Survival Between Invasive Lobular Breast Carcinoma and Invasive Ductal Breast Carcinoma: A Propensity Score Matching Study Based on SEER Database

Objective Invasive lobular carcinoma (ILC) and invasive ductal carcinoma (IDC) account for most breast cancers. However, the overall survival (OS) differences between ILC and IDC remain controversial. This study aimed to compare nonmetastatic ILC to IDC in terms of survival and prognostic factors for ILC. Methods This retrospective cohort study used data from the Surveillance, Epidemiology and End Results (SEER) Cancer Database (www.seer.cancer.gov). Women diagnosed with nonmetastatic ILC and IDC between 2006 and 2016 were included. A propensity score matching (PSM) method was used in our analysis to reduce baseline differences in clinicopathological characteristics and survival outcomes. Kaplan-Meier curves and log-rank test were used for survival analysis. Results Compared to IDC patients, ILC patients were diagnosed later in life with poorly differentiated and larger lesions, as well as increased expression of estrogen receptors (ERs) and/or progesterone receptors (PRs). A lower rate of radiation therapy and chemotherapy was observed in ILC. After PSM, ILC, and IDC patients exhibited similar OS (HR=1.017, p=0.409, 95% CI: 0.967–1.069). In subgroup analysis of HR-negative, AJCC stage III, N2/N3 stage patients, or those who received radiotherapy, ILC patients exhibited worse OS compared to IDC patients. Furthermore, multivariate analysis revealed a 47% survival benefit for IDC compared to ILC in HR-negative patients who received chemotherapy (HR=1.47, p=0.01, 95% CI: 1.09–1.97). Conclusions Our results demonstrated that ILC and IDC patients had similar OS after PSM. However, ILC patients with high risk indicators had worse OS compared to IDC patients by subgroup analysis.


INTRODUCTION
Invasive lobular carcinoma (ILC) is the second most common histological subtype of invasive breast cancer and contributes to 5%-15% of all breast cancer cases (1)(2)(3)(4). The incidence of ILC is increasing, especially in women older than fifty (5,6). At present, most of the systemic therapy decisions for ILC are derived from randomized clinical trials based on invasive ductal carcinoma (IDC), which may explain why the St Gallen International Expert Consensus guidelines and the National Comprehensive Cancer Network (NCCN) still recommend that ILC be treated with the same treatment paradigms as IDC; however, ILC has many unique features.
Compared to IDC, ILC has better prognosis, being almost invariably positive for hormone receptors, with low histological grade, negative for HER2 and with a generally good response to endocrine therapy (1,3). In general, ILC tends to grow in a multicentric or multifocal pattern, which makes resection for negative margins difficult when performing breast-conserving surgery (7)(8)(9)(10). In addition, ILC patients tend to be at risk for distant recurrence for greater than 5-10 years (11) and have an unusual metastatic site with involvement of the gastrointestinal tract, pelvic organs and peritoneal sites (12,13). Therefore, the clinical response of ILC has unique aspects and deserves special attention. ILC is considered to be less chemo-sensitive for either adjuvant or neoadjuvant chemotherapy compared to IDC, which is thought to be mediated by molecular features, such as lowgrade and estrogen receptor (ER) positivity (1,3,4,9,14,15).
There are conflicting data regarding the prognosis between ILC and IDC. Long-term outcomes for ILC have been reported as worse (16)(17)(18)(19), no different (20)(21)(22)(23), and better (3,4,24,25) than for IDC. Whether differences exist with respect to diseasefree survival (DFS) and overall survival (OS) between ILC and IDC is still controversial. Furthermore, differences in prognosis between ILC and IDC with the same molecular subtypes are not clear. Our study used a large database to further compare survival between IDC and ILC across a range of subgroups and to identify prognostic factors for early breast cancer patients with ILC.

SEER Database
This study cohort employed data from the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) 18 tumor registry database that was updated in November 2016. The SEER program registries contain population demographics, tumor characteristics, nodal staging, surgery information, vital status, and follow-up information from 18 geographic regions with more than 3 million patients, covering approximately 28% of the U.S. population. The database emphasizes quality control and stipulates a less than five percent error rate (26). We defined IDC patients according to the International Classification of Diseases (ICD) histological code 8500/3 and ILC patients according to the ICD histological code 8520/3.
We were granted permission to access the cancer files from the SEER program along with other treatment information, including chemotherapy and radiation therapy with a reference number of 19952-Nov2018. The requirement for informed consent was waived because personal information of patients was not involved.

Case Selection
We first identified 288,216 IDC patients and 30,190 IDC patients from 2006 to 2016 according to the following criteria: female breast cancer patients aged 18-90 years old; primary cancer only; unilateral and known laterality; ductal and/or lobular carcinoma; availability of detailed information about grade, T/N stages, ER status, and PR status; and availability of detailed data about survival ( Table 1). We used PSM to equate the two groups, and the final cohort consisted of 58,398 patients, 29,199 IDC patients, and 29,199 ILC patients.

Data and Statistical Analysis
Differences in the characteristic variables between ILC and IDC were compared by Chi-square test. The multivariate relationship of tumor characteristic variables and survival outcomes were examined by Cox proportional hazard model. Statistical significance was set at 0.05. OS was used as the survival endpoint in this survey and was analyzed using the Kaplan-Meier method. OS was defined as the time between confirming breast cancer to any cause of death. The log-rank test was utilized to calculate the hazard ratio (HR) and 95% confidence interval (95% CI) for OS. Our study utilized the propensity score matching method to diminish baseline differences in clinicopathological characteristics and survival outcome. Cases were 1:1 matched between ILC and IDC patients in accordance with age, histological grade, tumor stage, nodal stage, ER status, PR status, and so on. Statistical analysis was performed using Statistical Product and Service Solutions (SPSS) software (version 22; SPSS Inc., Chicago, USA). The propensity score matching method was calculated using the "MatchIt" package in R software (version 3.6.2, Synergy Software, Inc., Essex Junction, VT, USA).

Patient Characteristics Between ILC and IDC
The SEER tumor registry database was used to identify 1,097,908 patients diagnosed with ILC and IDC. After selecting patients based on specific inclusion and exclusion criteria, the remaining 318,406 patients were included in our research. Table 1 shows the patient selection process. Finally, 30,190 patients (9.48%) were assigned to the ILC group and 288,216 patients (90.52%) were assigned to the IDC group. Table 2 summarizes the clinical characteristics of the ILC and IDC groups.
People diagnosed with ILC tended to be older (median age of 63 years old in the ILC group versus 59 years old in the IDC group; p<0.0001), exhibit poorly differentiated and larger lesions, be ER/PR positive and were administered less radiation therapy and chemotherapy.
Given the surgical procedures, ILC had a higher percent of mastectomy compared to IDC cases (50.0% versus 39.2%, respectively). A lower rate of radiation therapy and chemotherapy was observed in the ILC group (55.3% versus 56.7%; 34.0% versus 45.5%, respectively; p<0.0001).

Survival Outcomes Between ILC and IDC Group
Due to significant differences in clinical characteristics between ILC and IDC groups, our research used the propensity score matching method, based on age, histological grade, tumor stage, nodal stage, ER status, PR status, surgery type, chemotherapy, and radiation therapy, to reduce discrepancies in survival outcomes between the two groups. Each ILC patient was matched to one IDC patient. As shown in Table 3, both ILC and IDC groups comprised 29,199 patients with similar baseline clinicopathological characteristics for further analysis.
The median OS was not reached in either group. During follow up, 2,999 patients (10.28%) in the ILC group died and 3,057 patients (10.46%) in the IDC group died. Based on comparison of the unmatched population database, IDC patients exhibited better OS (HR=1.045, P=0.025, 95% CI: 1.007-1.085) compared to ILC patients. Figure 1
We stratified patients by treatment to further validate the differential prognosis by whether the patients received chemotherapy or not between the ILC and IDC cases. Among patients who received chemotherapy, the ILC group exhibited poorer OS compared with the IDC group in HR-negative patients (p=0.010, HR=1.47, 95% CI 1.09-1.97, Figure 4B). In contrast, the ILC and IDC groups presented similar OS in HRpositive patients (p=0.871, HR=0.99, 95% CI 0.91-1.09, Figure  4A). Among patients who did not receive chemotherapy, the ILC group had a similar OS compared to IDC group in both HRpositive and HR-negative patients (p=0.865, HR=1.00, 95% CI 0.94-1.06; p=0.839, HR=1.04, 95% CI 0.74-1.45, respectively. Figures 4C, D).
HER2 status is a very important prognostic and predictive factor in breast cancer. Therefore, we extracted patients with available HER2 status from the matched population, We further generated Kaplan-Meier survival curves and conducted a pairwise comparison between the two different HER2 statuses in IDC and ILC groups. We found that HER2 status was not a prognostic indicator for OS between IDC and ILC groups (HER2+: 92.99% vs. 94.60%, p>0.05; HER2-: 94.36% vs. 94.41%, p>0.05; Figures 5A, B).

DISCUSSION
ILC is the most common subtype of breast cancer that tends to have poor prognosis due to higher rates of lymph node metastasis compared to patients with IDC and tends to present with larger tumor size (27)(28)(29). However, ILC is also associated with favorable prognosis for patients with lower histologic grade, hormone receptor positivity and HER2 negative status (1,3). Thus, the prognosis of ILC compared to IDC is still controversial. Our study was based on a retrospective analysis of a relatively large cohort of data retrieved from the population-based SEER database. After adjustment for patient and treatment characteristics, we found that the survival of ILC and IDC was equivalent in this analysis of outcomes for with two large patient cohorts. However, compared to IDC, we identified negative hormone receptor and positive lymph node statuses as adverse predictors for ILC.
In this study, we observed several differences in demographic and tumor characteristics between ILC and IDC in accordance with prior studies. ILC patients were older and had larger tumors than IDC patients. Moreover, ILC tumors were more often hormone receptor positive and of lower histological grade than IDC. The characteristics of these patients tended to be similar to prior studies reported for ILC (4,11,15,(30)(31)(32). The lower rate of chemotherapy can be attributed to the positivity of hormone receptors and low histological grade for most ILC patients, while the lower rate of BCS can be attributed to relatively larger tumor size. Furthermore, we found that the rate of lymph node positivity was higher in ILC in agreement with previous literature (15,29,32).
HER2 status is a very important prognostic and predictive factor in breast cancer. However, the majority of ILCs lack HER2 overexpression or amplification, and previous studies have reported that HER2 positivity occurs in 3%-5% of ILCs (30,33). In our study, the rate of HER2 positivity was 4.85%, which is in accordance with data published previously. Furthermore, we found that HER2 status was not a prognostic indicator for OS HER2+ patients (HER2+: 92.99% vs. 94.60%, p>0.05; HER2-: 94.36% vs. 94.41%, p>0.05).
Previous studies comparing survival between IDC and ILC have reported conflicting results. Identifying 1034 patients The effect of neoadjuvant chemotherapy on ILC has been reported by several retrospective studies, with pCR rates ranging from 0-11% for ILC compared to 9%-25% for IDC (1,4,14,15,34). A total of 1,051 patients with ILC were included in a pooled analysis by Loibl et al. (22,35,36), and the end point was pathological complete response (pCR). There was a 6.2% pCR for ILC and 17.8% in HR negative and high-grade subgroups. A previous study also provided evidence that biologically aggressive ILCs (HR-and G3) achieved a higher pCR rate. However, the pCR rate is still lower in HR-negative ILC compared to HRnegative IDC (1). pCR, a surrogate endpoint marker, has been correlated with improved long-term outcomes, such as diseasefree or overall survival (37)(38)(39), which could explain our result that HR-negative ILC exhibited worse OS than IDC. Furthermore, Huober et al. conducted a retrospective analysis to identify factors predicting relapse based on the GBG metadatabase that includes five neoadjuvant trials (19). Their study demonstrated that the prognosis of patients with ILC was worse than IDC, even if all reached pCR. The reason for this result is that the majority of patients included in this study had triple negative breast cancer or HER2 positive breast cancer, in agreement with our study that ILC patients with HR-negative status have worse prognosis. As mentioned above, ILC patients do not respond as well to neoadjuvant chemotherapy compared to IDC patients. This may explain a general assumption that ILC patients also have a poor response to adjuvant chemotherapy compared to IDC, but there is currently no data to support this notion. However, several studies have attempted to evaluate the effect of chemotherapy in ILC and reported no benefits followed by chemotherapy. Truin et al. (40) retrospectively examined the effect of chemotherapy in a multicenter cohort of postmenopausal patients with pure or mixed type ILC and found additional beneficial effects (10-year OS 66% vs 68%, p=0.45). Marmor et al. (41) also reported a similar result in patients with ER-positive, HER2-negative, stage I/II ILC who received endocrine therapy, demonstrating that they did not benefit from the addition of adjuvant chemotherapy. However, by analyzing 2,318 patients with ILC, Nonneville et al. (42) reported that ILC patients exhibited significant differences in DFS or OS benefit from chemotherapy in high-risk patients, such as those who were lymph node positive or presented with lympho-vascular invasion. In agreement with previous results, we identified negative HR, N2/N3, stage III, and receiving radiotherapy as predictors for long-term adverse outcomes in ILC. According to guidelines of the St Gallen International Expert Consensus and the National Comprehensive Cancer Network (NCCN), ILC patients are recommended the same systemic treatment as IDC patients. Systemic therapies include chemotherapy, radiotherapy, and endocrine therapy. Although previous studies have suggested that the sensitivity to radiotherapy is similar for ILC and IDC patients (43), the prognosis for ILC patients is still worse than IDC patients. Our explanation is that patients who receive radiotherapy tend to present positive lymph nodes and a larger tumor size, which are risk factors of worse survival. Chemotherapy is recommended for these patients with high risk. However, ILC patients are less sensitive to chemotherapy compared to IDC patients, which means that the treatment strategy of ILC may not completely be referred to that of IDC. ILC patients have their unique characteristics and might require additional stronger systemic treatments to improve survival. Previous studies analyzed the effect of adjuvant chemotherapy in HR-positive    different outcomes in ILC, for example, pleomorphic ILC with aggressive clinical features has a worse prognosis than classical ILC (7,30). Second, our study is a retrospective cohort study performed using the SEER database, so there is selection bias and missing data.
In conclusion, our research demonstrated that there was no significant difference in overall survival between ILC and IDC patients after matching for several known covariates. In the subgroup analysis stratified by negative hormone receptor and positive lymph node status, we found that the prognosis of ILC was significantly worse compared to IDC. Among patients who received chemotherapy, the ILC group had worse prognosis compared to the IDC group in HR-negative patients, whereas the ILC and IDC groups exhibited similar OS in HR-positive patients. These results may indicate that we should enhance treatment among these special subgroups to prolong survival.