Comparative Effectiveness of Sentinel Lymph Node Biopsy for Ductal Carcinoma In Situ

Because of the non-invasive nature of ductal carcinoma in situ (DCIS), patients treated with available therapies have excellent outcomes and very low rates of breast cancer mortality. Considerable debate exists as to how the DCIS lesion should be treated, although there is a movement toward less intensive intervention by the identification of patient subsets with favorable prognoses. Some prospective studies have found that the rate of ipsilateral invasive cancer occurrence is still high after receiving breast conserving surgery (BCS) alone, even among patients with favorable pathologic characteristics. Such findings argue against active surveillance for DCIS treatment. However, evidence exists that older DCIS patients have a lower rate of ipsilateral recurrence because DCIS among older patients tends to be indolent. Identifying suitable subgroups among this lower risk group who may be safe to receive a less aggressive treatment could change the current practice pattern of aggressive treatment.

Even when DCIS patients opt to receive a less intensive treatment such as BCS without radiation therapy, they and their providers need to decide whether to undergo sentinel lymph node biopsy (SLNB). A systematic review has shown that evidence gaps exist regarding the benefits of SLNB for DCIS. Given that the likelihood of axillary recurrence is low among DCIS patients who received radiation therapy, routine SLNB is not recommended for DCIS patients. Of note, radiation therapy can control axillary disease if present. If the investigators plan to empower DCIS patients to choose less intensive management options, such as BCS forgoing radiation therapy (RT), it will be crucial for patients and providers to understand the role of SLNB.

The overarching goals of this study are to compare side effects and outcomes between receiving SLNB vs. not receiving SLNB among older DCIS patients who received BCS. With this data, the investigators also aim to identify sub-populations for whom less intensive treatments may be appropriate. Using the Surveillance, Epidemiology, and End Result (SEER)-Medicare linked data, our project's overarching aims are:

Among older women with DCIS who have received BCS as their first surgery, to compare the outcomes of receiving sentinel lymph node biopsy (SLNB) vs. not receiving SLNB within 6 months of DCIS diagnosis:

Aim 1: The investigators will determine associations between SLNB and acute/subacute side effects, including lymphedema, pain, and limitation of movement of upper extremity from the first BCS to 9 months post-diagnosis Aim 2: The investigators will determine associations between SLNB and long-term outcomes, including breast cancer specific mortality, ipsilateral invasive breast cancer diagnosis, subsequent mastectomy as treated recurrence, and lasting side effects, from > 9 months' post-diagnosis to death or the end of the study period.

Our study is a retrospective cohort study with the study population being DCIS patients older than 67 years (hereafter referred to as older women) who were enrolled in a fee-for-service Medicare program and resided in the SEER areas from 1998 to 2011 (2001 to 2013 for Aim 2) and who were followed up to 2012 (2015 for Aim 2). The investigators selected age 67 years as a cut-off value because the investigators plan to use two years of claims data to identify patient comorbidities and control for them in our statistical models, and data is first available at age 65. Given the nature of our observational study design, the investigators will apply standard multivariate analyses and propensity score methodology to reduce the influence from confounders. The investigators will control for patient demographics, comorbidities, functional status, tumor characteristics, and prior healthcare utilization. Using distance to the nearest provider that uses SLNB for DCIS or surgeon's tendency in using SLNB for stage I/II breast cancer, the investigators also plan to conduct instrument variable analyses if necessary. Stratifying patients by key DCIS characteristics (including grade, comedonecrosis, and tumor size) and their predicted life expectancy (given their age and comorbidities), the investigators also hope to identify patient subgroups who may safely forgo SLNB.

In comparisons of baseline characteristics between intervention and control groups, the investigators will conduct standard descriptive statistics using chi-square tests for categorical variables and t-tests for continuous variables. The investigators will tabulate the frequencies of outcomes of interests by the intervention vs. control group. For multivariable analyses, the investigators will apply proportional hazards models to test whether the intervention is associated with better outcomes.

the investigators plan to control for this issue using propensity score matching methodology. Prior literature has suggested inclusion of either all measured variables or those variables that are associated with treatment selection (SLNB status) when conducting an analysis using the propensity score method. Specifically, our approach to matching will be based on the Mahalanobis distance calculated using age, race, residence in a metropolitan county, comorbidity, prior influenza vaccination or prior visit to a primary care physician (both as proxies for access to care), income, preoperative MRI use, and tumor characteristics. Tumor characteristics include size, grade, comedonecrosis, and estrogen receptor status. By incorporating these factors in matching, the investigators expect to substantially decrease bias and balance the risk for outcomes of interest between the SLNB and non-SLNB groups. The difference in outcomes between the control and intervention groups will be estimated in a Kaplan-Meier curve. The investigators will estimate the relative risk in the propensity score matched sample using the standard method for matched-pair data. The investigators will use the Cox proportional hazards models to investigate the association between various factors (grade, tumor size, and estrogen receptor status) and outcomes.