Abstract
Purpose
Trials for DCIS have not explored whether outcomes for patients with large disease burden requiring mastectomy are comparable to those of patients with lumpectomy-amenable disease. We aim to identify whether patients with DCIS larger than 5 cm and diffuse-type DCIS differ in breast cancer mortality (BCM) from patients with disease less than 5 cm.
Methods
Patients diagnosed with DCIS in the SEER program were assessed to identify factors prognostic of breast-cancer-specific survival using competing risks regression.
Results
44,849 patients met criteria for the cumulative incidence estimate. On competing risks cumulative incidence approximation, the 10-year estimate for BCM for each group was 1.3%, 1.3%, 2.3%, and 5.1%, respectively, and the difference among groups was significant (p = 0.017). On competing risks regression of patients with known covariates, both diffuse-type disease and disease larger than 5 cm (hazard ratio [HR] = 6.2 and 1.7, p = 0.013 and p = 0.042, respectively) were associated with increased risk of BCM. After matching, DCIS > 5 cm and diffuse disease were associated with increased BCM relative to disease < 5 cm (HR = 1.69, p = 0.04). Among patients undergoing mastectomy for disease larger than 5 cm or diffuse disease, the 10-year cumulative incidence for BCM was 0.5% among patients undergoing bilateral mastectomy and 2.4% for patients undergoing unilateral mastectomy.
Conclusion
Patients with large and diffuse DCIS represent uncommon but poorly studied DCIS subgroups with worse prognoses than patients with disease smaller than 5 cm. Further studies are needed to elucidate the appropriate treatment for these patients.
Keywords: Breast cancer, DCIS, Cancer outcomes, Epidemiology, Breast cancer mortality, Mastectomy
Introduction
Evolution of DCIS management and the “mastectomy blind spot”
Ductal carcinoma in situ (DCIS) is a commonly diagnosed neoplastic process of the breast that encompasses a heterogeneous group of lesions which vary with respect to histopathologic features and outcomes [1]. Prior to the 1980s, mastectomy was standard of care for DCIS because of the potential of DCIS to progress to invasive disease [2, 3]. After randomized trials demonstrated that mastectomy did not result in a difference in outcomes for invasive breast cancer patients relative to breast conservation therapy, it was assumed by the research community that these results would also apply to DCIS, and lumpectomy was adopted for its treatment [2, 3].
Management of DCIS was then refined in response to multiple randomized trials which demonstrated that the addition of radiation therapy (RT) to lumpectomy-reduced ipsilateral in situ and invasive recurrences, and among estrogen receptor (ER)-positive patients endocrine therapy (ET) such as tamoxifen reduced some combination of in situ and invasive recurrences in either breast [2-7]. Neither of these adjuvant therapies demonstrated a reduction in breast cancer mortality (BCM) [2, 3]. Importantly, these trials only investigated patients who were candidates for lumpectomy, not mastectomy, in spite of the fact that mastectomies represent between 27 and 56% of the surgeries received by DCIS patients [5, 6, 8-12].
Whether the results established by these randomized DCIS trials are applicable to the cohort of patients with large DCIS, particularly DCIS necessitating mastectomy, is unknown. Patients with what is referred to as “diffuse” DCIS, for example, defined by the Surveillance, Epidemiology and End Results (SEER) database as disease involving ¾ or more of the breast, are more likely to require mastectomy to achieve adequate resection of disease. Such patients may have a different risk profile for invasive recurrence and development of metastatic breast cancer and BCM than patients with smaller disease. If patients with larger- or diffuse-type disease have markedly worse outcomes from those studied in the previously mentioned randomized DCIS trials, this difference would necessitate the reconsideration of whether these entities should be treated simply as DCIS with a “high-risk” factor of size, or if an alternative treatment plan would benefit them.
Given the known excellent prognosis of patients with DCIS amenable to lumpectomy in the previously mentioned trials, it would be difficult to justify a prospective trial to answer these questions without evidence to suggest significant differences in these large and diffuse cohorts.
Aim
We seek to investigate the influence of large (> 5 cm)- and diffuse (involving > ¾ of the breast)-type DCIS on BCM in patients diagnosed with DCIS in a large national database.
Methods
This study was conducted in accordance with U.S. Common Rule. The UCSD HRPP/IRB deferred need for approval of this SEER-based study due to the use of public, de-identified data. Informed consent was not and could not be obtained for the patients studied due to the use of de-identified data.
Data Extraction
SEER 18 data were used including cases diagnosed between 1975 and 2016. For the competing risks cumulative incidence, inclusion criteria included female sex, first breast neoplastic event of pure DCIS, known follow-up time, performance of surgery as part of treatment course, known tumor size, known vital status and cause of death, and follow-up time greater than 6 months (Fig. 1). For competing risks regression, the previously stated inclusion criteria were used, as were known tumor grade, ER status, and race.
Fig. 1.
Selection criteria for cumulative incidence estimate and multivariate competing risks regression
Patients with ER-positive and borderline disease were categorized together, and patients of non-white, non-black race with non-missing race were categorized together. Lesion size was categorized as follows: ≤ 2 cm, 2 – 5 cm, > 5 cm, and diffuse.
Statistical Analysis
Competing risks cumulative incidence was approximated with BCM and non-breast mortality treated as competing risks [13]. Competing risks regression by Fine and Gray method [30] was performed to approximate hazard ratios (HR) for BCM.
P values were calculated as two-sided and statistical significance was declared for p less than 0.05. All statistical analyses were performed in R (version 3.5.1, R Foundation for Statistical Computing, Vienna, Austria) using RStudio (Version 1.1.463) and packages “tidyverse” (Version 1.3.0), “survival” (Version 3.1–7), “survminer” (Version 0.4.6), and “cmprsk” (Version 2.2–9), and matching was performed using the nearest neighbor method in a 4:1 matching ratio using the “MatchIt” package (Version 4.2.0) [14].
Results
Large (> 5 cm) and diffuse DCIS associated with higher 10-year cumulative incidence of BCM
A total of 44,849 patients met inclusion criteria for the competing risks cumulative incidence, including 34,377 patients (76.7%) with disease 2 cm or less, 8,145 (18.2%) with disease 2–5 cm, 2,246 patients (5.0%) with disease greater than 5 cm, and 81 patients (0.2%) with diffuse-type disease (Fig. 2). Within this group, median follow-up time was 70 months (interquartile range [IQR] 39–108), and 550 (1.2%) patients suffered BCM, 4,425 (9.9%) suffered non-BCM death, and 39,874 (88.9%) were alive at time of last follow-up.
Fig. 2.
Breast cancer mortality cumulative incidence among patients with DCIS categorized by tumor size
The 10-year BCM cumulative incidence was 5.1% (95% CI 1.6–16.3%) for patients with diffuse-type disease, 2.3% (95% CI 1.6–3.4%) for patients with disease larger than 5 cm, 1.3% (95% CI 1.0–1.6%) for patients with disease 2–5 cm, and 1.3% (95% CI 1.1%–1.4%) for patients with disease smaller than 2 cm. The difference among groups was statistically significant (p = 0.02).
10-year breast cancer mortality for patients with large (> 5 cm) and diffuse DCIS lower for patients undergoing bilateral mastectomy relative to unilateral mastectomy
For the 2,327 patients in this group with DCIS size larger than 5 cm or diffuse DCIS, 1,487 (63.9%) underwent mastectomy with known surgical laterality. Among them, 1,138 (48.9%) had unilateral mastectomy and 349 (15.0%) underwent bilateral mastectomy. Within this group, the median age was 52 (IQR 44–62) and median follow-up was 81 months (IQR 44–129). Among patients undergoing unilateral mastectomy, 20 (1.8%) suffered BCM and 79 (6.9%) suffered non-BCM death. Among patients undergoing bilateral mastectomy, 1 (0.3%) patient suffered BCM and 3 (0.9%) patients suffered non-BCM death. The estimated cumulative incidence of BCM at 10 years for patients underoging unilateral mastectomy was 2.4% (95% CI 1.4–3.9%) and for patients undergoing bilateral mastectomy was 0.5% (0.1–4.0%), though this difference was not statistically significant (p = 0.12) (Fig. 3).
Fig. 3.
Breast cancer mortality cumulative incidence among patients with DCIS undergoing mastectomy categorized by unilateral vs. bilateral mastectomy
Large (> 5 cm) and Diffuse DCIS Associated with Increased Risk of BCM in Multivariate Model
29,430 met inclusion criteria for the multivariate competing risks regression (Fig. 1). Among them, median age at diagnosis was 58 (IQR 49–67). Median follow-up time was 70 months (IQR 39–108). 21,830 (74.2%) patients had disease 2 cm or smaller, 5,844 (19.9%) had disease 2–5 cm, 1,712 (5.8%) had disease larger than 5 cm, and 44 (0.1%) had disease classified as diffuse. The distribution of patient and lesion-associated factors grouped by size classification is shown in Table 1. Among these patients, 1,278 (4.3%) had invasive events with median time to invasive event of 45 months (IQR 20.0–82.1 months). None of the patients with diffuse disease had subsequent invasive events, 88 (5.1%) of patients with disease > 5 cm had subsequent invasive events, 228 (3.9%) of patients with disease 2–5 cm had subsequent invasive events, and 962 (4.4%) of patients with disease 2 cm or less had subsequent invasive events. Among these patients with invasive subsequent events, 63 (4.9%) died of breast cancer mortality, 65 (5.1%) died of non-breast causes, and 1,150 (90.0%) were alive at time of last follow-up.
Table 1.
Characterization of DCIS cohorts by lesion size
| < 2 cm 21,830 (74.2%) |
2–5 cm 5,844 (19.9%) |
> 5 cm 1,712 (5.8%) |
Diffuse 44 (0.1%) |
||
|---|---|---|---|---|---|
| Age at diagnosis | > 60 | 9,374 (42.9%) | 2,363 (40.4%) | 542 (31.7%) | 14 (31.8%) |
| 40–60 | 11,537 (52.9%) | 3,112 (53.3%) | 977 (57.1%) | 26 (59.1%) | |
| < = 40 | 919 (4.2%) | 369 (6.3%) | 193 (11.2%) | 4 (9.1%) | |
| ER status | Positive or borderline | 19,036 (87.2%) | 4,657 (79.7%) | 1,287 (75.2%) | 7 (15.9%) |
| Negative | 2,794 (12.8%) | 1,187 (20.3%) | 425 (24.8%) | 37 (84.1%) | |
| PR status | Positive or borderline | 15,807 (72.4%) | 3,733 (63.9%) | 1,051 (61.4%) | 30 (68.2%) |
| Negative | 4,516 (20.6%) | 1,704 (29.2%) | 534 (31.2%) | 13 (29.5%) | |
| Missing | 1,507 (7.0%) | 407 (6.9%) | 127 (7.4%) | 1 (2.3%) | |
| HER2 status | Positive or borderline | 496 (2.3%) | 133 (2.3%) | 46 (2.7%) | 1 (2.3%) |
| Negative | 951 (4.4%) | 204 (3.5%) | 57 (3.3%) | 3 (6.8%) | |
| Missing | 20,383 (93.3%) | 5,507 (94.2%) | 1,609 (94.0%) | 40 (90.9%) | |
| Date Diagnosis | 2010–2015 | 11,851 (54.3%) | 3,493 (59.8%) | 1,032 (60.3%) | 17 (38.6%) |
| 2005–2009 | 7,505 (34.4%) | 1,840 (31.5%) | 565 (33.0%) | 22 (50.0%) | |
| Before 2004 | 2,474 (11.3%) | 511 (8.7%) | 115 (6.7%) | 5 (11.4%) | |
| Surgery extent | Lumpectomy | 17,433 (79.9%) | 3,388 (58.0%) | 523 (30.5%) | 7 (15.9%) |
| Unilateral mastectomy | 2,485 (11.4%) | 1,530 (26.2%) | 761 (44.5%) | 26 (59.1%) | |
| Bilateral mastectomy | 1,375 (6.3%) | 627 (10.7%) | 295 (17.2%) | 7 (15.9%) | |
| Mastectomy (missing unilateral vs. bilateral) | 537 (2.4%) | 299 (5.1%) | 133 (7.8%) | 4 (9.1%) | |
| Tumor grade | High | 9,512 (43.6%) | 3,430 (58.7%) | 1,090 (63.7%) | 24 (54.5%) |
| Intermediate | 9,074 (41.6%) | 1,960 (33.5%) | 522 (30.5%) | 16 (36.4%) | |
| Low | 3,244 (14.8%) | 454 (7.8%) | 100 (5.8%) | 4 (9.1%) | |
| Radiation therapy | Administered | 13,225 (60.6%) | 2,794 (47.8%) | 450 (26.3%) | 9 (20.5%) |
| Not administered | 8,605 (39.4%) | 3,050 (52.2%) | 1,262 (73.7%) | 35 (79.5%) | |
| Race | White | 16,367 (75.0%) | 4,166 (71.3%) | 1,172 (68.5%) | 32 (72.8%) |
| Black | 2,395 (11.0%) | 639 (10.9%) | 266 (15.5%) | 6 (13.6%) | |
| Non-white, non-black | 3,068 (14.0%) | 1,039 (17.8%) | 274 (16.0%) | 6 (13.6%) | |
| Outcome | Alive at last follow up | 20,424 (93.6%) | 5,500 (94.1%) | 1,613 (94.2%) | 41 (93.2%) |
| Breast cancer mortality | 163 (0.7%) | 35 (0.6%) | 20 (1.2%) | 2 (4.5%) | |
| Non-breast cancer mortality | 1,243 (5.7%) | 309 (5.3%) | 79 (4.6%) | 1 (2.3%) | |
| Median follow-up time in months (interquartile range) | 72 (40–110) | 67 (37–103) | 64 (36–102) | 87.5 (60.75–116.75) |
On multivariate competing risks regression, relative to DCIS 2 cm or smaller, patients with disease larger than 5 cm (HR = 1.69, p = 0.04) and diffuse-type disease (HR = 6.21, p = 0.01) had increased risk of BCM (Table 2). Other factors associated with increased risk of BCM include ER-negative disease (HR = 1.73, p = 0.001), and black race (HR = 2.84, p < 0.001) relative to white race. Factors associated with decreased BCM risk were age 40–60 (0.45, p < 0.001) and < = 40 (0.45, p = 0.02) relative to patients older than 60, administration of radiation therapy (HR = 0.49, p < 0.001), and mastectomy (HR = 0.62, p = 0.02) relative to lumpectomy.
Table 2.
Univariate and multivariate competing risks regression for breast cancer mortality
| Univariate | Multivariate | ||||
|---|---|---|---|---|---|
| HR (95% CI) | P | HR (95% CI) | P | ||
| Tumor size | < = 2 cm | Ref | – | Ref | – |
| 2–5 cm | 0.88 (0.61–1.27) | 0.50 | 0.85 (0.58–1.24) | 0.40 | |
| > 5 cm | 1.81 (1.13–2.87) | 0.01 | 1.69 (1.02–2.80) | 0.04 | |
| Diffuse | 5.33 (1.32–21.5) | 0.02 | 6.21 (1.47–26.23) | 0.01 | |
| Age at diagnosis | > 60 | Ref | – | Ref | – |
| 40–60 | 0.43 (0.33–0.57) | < 0.001 | 0.45 (0.34–0.60) | < 0.001 | |
| < = 40 | 0.50 (0.25–0.98) | 0.04 | 0.45 (0.23–0.89) | 0.02 | |
| ER status | Positive or borderline | Ref | – | Ref | – |
| Negative | 1.68 (1.24–2.28) | < 0.001 | 1.73 (1.24–2.42) | 0.001 | |
| Surgical extent | Lumpectomy | Ref | – | Ref | – |
| Mastectomy | 1.03 (0.76–1.38) | 0.87 | 0.62 (0.42–0.93) | 0.02 | |
| Tumor grade | High | Ref | – | Ref | – |
| Intermediate | 1.15 (0.87–1.52) | 0.31 | 1.18 (0.88–1.58) | 0.27 | |
| Low | 0.64 (0.39–1.06) | 0.08 | 0.62 (0.36–1.04) | 0.07 | |
| Radiation therapy | Not administered | Ref | – | Ref | – |
| Administered | 0.61 (0.47–0.79) | < 0.001 | 0.49 (0.36–0.69) | < 0.001 | |
| Race | White | Ref | – | Ref | – |
| Black | 2.66 (1.95–3.64) | < 0.001 | 2.84 (2.07–3.89) | < 0.001 | |
| Non-white, non-black | 0.73 (0.45–1.17) | 0.19 | 0.79 (0.49–1.27) | 0.32 | |
Multivariate Model with Matching of Large and Diffuse DCIS with DCIS Less Than 5 cm
Matching resulted in improved agreement among baseline characteristics among patients with large and diffuse disease with those having disease < 5 cm both graphically (Fig. 4) and numerically (Table 3). Large and diffuse disease were associated with significantly increased risk of breast cancer mortality in both the univariate (HR = 1.66, p = 0.04) and multivariate (HR = 1.69, p = 0.04) regressions relative to disease ≤ 5 cm (Table 4). The only other significant factors in the matched multivariate model were age 40–60 (HR = 0.37, p < 0.001) and black race (HR = 2.60, p < 0.001).
Fig. 4.
Standardized mean differences between patients with disease >5 cm or diffuse disease compared with patients having disease <=5 cm before and after matching
Table 3.
Characterization of matched cohorts of patients with disease >5 cm or diffuse disease and those with disease ≤5 cm
| ≤ 5 cm 6,968 (80%) |
> 5 cm or diffuse 1,742 (20%) |
||
|---|---|---|---|
| Age at diagnosis | > 60 | 2,284 (32.8%) | 550 (31.6%) |
| 40–60 | 4,052 (58.1%) | 995 (57.1%) | |
| < = 40 | 632 (9.1%) | 197 (11.3%) | |
| ER status | Positive or borderline | 5,276 (75.7%) | 1,316 (75.5%) |
| Negative | 1,692 (24.3%) | 426 (24.5%) | |
| PR status* | Positive or borderline | 4,275 (61.4%) | 1,076 (61.8%) |
| Negative | 2,259 (32.4%) | 543 (31.2%) | |
| Missing | 434 (6.2%) | 123 (7.0%) | |
| HER2 status* | Positive or borderline | 197 (2.8%) | 49 (2.8%) |
| Negative | 238 (3.4%) | 58 (3.3%) | |
| Missing | 6,533 (93.8%) | 1,635 (93.9%) | |
| Date Diagnosis* | 2010–2015 | 559 (8.0%) | 55 (3.2%) |
| 2005–2009 | 3,036 (43.6%) | 649 (37.3%) | |
| Before 2004 | 3,373 (48.4%) | 1,038 (59.5%) | |
| Surgery extent | Lumpectomy | 2,100 (30.1%) | 525 (30.1%) |
| Mastectomy | 4,868 (69.9%) | 1,217 (69.9%) | |
| Tumor grade | High | 4,394 (63.1%) | 1,108 (63.6%) |
| Intermediate | 2,163 (31.0%) | 531 (30.5%) | |
| Low | 411 (5.9%) | 103 (5.9%) | |
| Radiation therapy | Administered | 1,700 (24.4%) | 455 (26.1%) |
| Not administered | 5,268 (75.6%) | 1,287 (73.9%) | |
| Race | White | 4,773 (68.5%) | 1,199 (68.8%) |
| Black | 1,088 (15.6%) | 270 (15.5%) | |
| Non-white, non-black | 1,107 (15.9%) | 273 (15.7%) | |
| Outcome* | Alive at last follow up | 6,463 (92.8%) | 1,642 (94.3%) |
| Breast cancer mortality | 62 (0.9%) | 21 (1.2%) | |
| Non-breast cancer mortality | 443 (6.3%) | 79 (4.5%) | |
| Median follow-up time in months (interquartile range)* | 79 (44–117) | 65 (37–103) |
Not included among matching criteria
Table 4.
Competing risks regression results for breast cancer mortality comparing patients with disease >5 cm or diffuse disease and those with disease <=5 cm
| Univariate | Multivariate | ||||
|---|---|---|---|---|---|
| HR (95% CI) | P | HR (95% CI) | P | ||
| Tumor size | ≤ 5 cm | Ref | – | Ref | – |
| > 5 cm or diffuse | 1.66 (1.01–2.73) | 0.04 | 1.69 (1.03–2.78) | 0.04 | |
| Age at diagnosis | > 60 | Ref | – | Ref | – |
| 40–60 | 0.35 (0.22–0.56) | < 0.001 | 0.37 (0.23–0.59) | < 0.001 | |
| < = 40 | 0.46 (0.20–1.07) | 0.07 | 0.45 (0.19–1.06) | 0.07 | |
| ER status | Positive or borderline | Ref | – | Ref | – |
| Negative | 1.31 (0.82–2.1) | 0.26 | 1.40 (0.84–2.35) | 0.20 | |
| Surgical extent | Lumpectomy | Ref | – | Ref | – |
| Mastectomy | 0.80 (0.51–1.24) | 0.32 | 0.70 (0.36–1.37) | 0.29 | |
| Tumor grade | High | Ref | – | Ref | – |
| Intermediate | 1.41 (0.90–2.20) | 0.13 | 1.33 (0.83–2.14) | 0.23 | |
| Low | 0.37 (0.09–1.53) | 0.17 | 0.35 (0.08–1.47) | 0.15 | |
| Radiation therapy | Not administered | Ref | – | Ref | – |
| Administered | 0.92 (0.56–1.51) | 0.74 | 0.66 (0.32–1.36) | 0.26 | |
| Race | White | Ref | – | Ref | – |
| Black | 2.78 (1.73–4.44) | < 0.001 | 2.69 (1.67–4.33) | < 0.001 | |
| Non-white, non-black | 0.56 (0.24–1.29) | 0.17 | 0.54 (0.23–1.27) | 0.16 | |
Discussion
Here, we demonstrate that large- and diffuse-type DCIS are associated with increased risk of BCM. Patients with DCIS not amenable to lumpectomy at time of first excision were not included in the prospective randomized DCIS trials that established the use of adjuvant radiation and ET, so the optimal management for these patients has not been elucidated. Whether a patient’s disease is amenable to lumpectomy is a subjective decision based on discussion between provider and patient regarding acceptable postoperative cosmesis, as well as patient-related factors such as tumor volume relative to overall breast volume [15]. When comparing BCM rates among patients with these subtypes of disease, the group undergoing bilateral mastectomy had rates of BCM at 10 years (0.5%) that were lower than those undergoing unilateral mastectomy (2.4%), though this difference was not significant (p = 0.12).
The current standard of care for DCIS involves surgery that excises disease with margins of at least 10 mm, with or without some combination of adjuvant radiation and ET[16, 17]. For patients undergoing lumpectomy, adjuvant radiation therapy is considered standard of care due to its demonstrated reduction in all ipsilateral breast cancer recurrences, as well as invasive ipsilateral recurrences, however, since it has not been shown to reduce BCM or all-cause mortality, informed decisions with patients must be made based on patient and lesion-associated risk factors [2, 17]. Adjuvant ET has been shown to reduce all breast cancer recurrences, as well as contralateral recurrences, in patients with ER-positive disease, though similar to radiation therapy, it has not been shown to reduce BCM or overall mortality [7, 9, 17].
An important study for comparison of our results is Steven Narod’s 2015 SEER analysis of BCM after a diagnosis of DCIS. There are several key differences between the cohort used in that study and the one used for our analysis. First, Narod’s analysis was performed on patients in SEER through the year 2011, whereas ours used a more recent SEER iteration that includes patients through 2015, with over half of our patients being diagnosed after 2010 [18]. Second, due to the very large number of stage 0 breast cancer cases that were not clearly of ductal origin, rather than excluding a list of histology codes as Narod did, we only allowed specific histology codes which we felt confident were ductal origin in situ disease consistent with DCIS [18]. This excluded many cases that were used in Narod’s study. Another important and obvious difference is that we included patients with diffuse-type disease, as tumor size was of primary focus in our study, and we also included patients undergoing bilateral mastectomy, which were excluded from Narod’s study [18]. We divided non-diffuse disease into > 5 cm, which has been associated with increased risk of invasive recurrence and nodal spread, as well as ≤ 2 cm and 2–5 cm [18-20]. And finally, rather than using standard Kaplan–Meier curves and Cox regression, we used competing risks analysis since SEER includes data on both vital status and cause of death, allowing for the estimation of BCM when considering non-BCM deaths as a competing risk.
Despite these differences, several of our results regarding risk factors for BCM are similar, including black race, ER-negative disease, and tumor size. Tumor grade was not a significant factor for BCM in our study though high-grade disease was associated with increased BCM in Narod’s study [18]. Radiation therapy, which was not significantly associated with reduction in BCM in Narod’s study, did reduce risk in our study (HR = 0.49, p < 0.001). Younger age, which was associated with higher risk of BCM in Narod’s study, was associated with lower risk in our study—this is likely due to our use of competing risks analysis, as not accounting for competing risks in survival analysis results in overestimation of events of interest as older patients are more likely to die of any cause than younger patients [21]. Mastectomy, which was associated with higher risk of BCM in Narod's study, was associated with lower risk in our study [18]. This is likely related to the inclusion of patients undergoing bilateral mastectomy. This was most notable for patients with disease larger than 5 cm or diffuse disease in our study, for whom the rate of BCM was low.
While our data and results are unable to elucidate the mechanism conferring increased risk to patients with large and diffuse DCIS, several possibilities exist. The simplest explanation is that patients with large disease have a larger number of pathologic sections to be examined and an increased number of unexamined sections between slides leading to increased chance of a missed focus of invasion. If this was the case, however, there could be an anticipated increase in BCM rates with increasing lesion size, but this was not observed—patients with disease less than 2 cm and 2–5 cm had similar 10-year BCM (both 1.3%). Another possible explanation is that patients with large- and diffuse-type DCIS may have an underlying genetic predisposition to breast neoplastic processes, such that the large DCIS sizes are not secondary to field cancerization but rather to multiple foci of DCIS arising independently. This would not only explain the discrepancy in BCM for large and diffuse diseases with respect to smaller disease but also explain the appearance of improved BCM outcomes with respect to patients undergoing bilateral mastectomy compared to those undergoing unilateral mastectomy.
The use of competing risks for the estimation of BCM and the HRs of associated covariates allows for a more accurate determination of these values than non-competing survival models, such as those used in Narod’s study, because it accounts for non-censoring endpoints that preclude the occurrence of the event of interest. In our study, this was non-BCM death.
There are several important limitations to this study. The analysis is retrospective preventing the use of randomization and standardized follow-up. This also leads to higher rates of missing data for covariates of interest, thereby narrowing the number of patients who could be included in the multivariate analysis, and prevents the use of covariates of interest that are not recorded in SEER, such as surgical margins and use of adjuvant ET. SEER data also rely on the documentation of tumor registrars from many different institutions across the country, so there may be institutional variations, particularly with respect to organ-specific coding options such as the diffuse classification of breast cancer. Another notable limitation of our study is that due to recommendations provided to registrars in the SEER Inquiry System, if patients with an initial diagnosis of DCIS present with metastatic disease or suffer BCM without an intervening diagnosis of invasive breast cancer, coders are instructed to recode the initial DCIS diagnosis as invasive. This results in unilateral removal of patients from the DCIS cohort in SEER who develop metastatic disease or die of BCM [22]. Therefore, even with the use of competing risks analysis, our cumulative BCM underestimates the true incidence of BCM among DCIS patients in the US. This is unsurprising given that a meta-analysis of the prospective DCIS trials estimated the 10-year BCM rate at 3.7% for patients undergoing only lumpectomy and 4.1% for patients undergoing BCS with adjuvant radiation, rates that are comparable to the highest-risk cohort in our group, those with diffuse-type disease. While this limits the utility of the cumulative incidence estimate from our study, we do not suspect that it would markedly alter the HRs determined for large- and diffuse-type disease, as there is no apparent reason why the adherence to this recommendation would be associated with a patient’s tumor size or other lesion or patient-related covariates. Finally, SEER does not document ipsilateral in situ recurrences within 5 years after the initial diagnosis of DCIS, so we were unable to accurately quantify in situ recurrences.
Another important limitation to our study is the lack of availability of data on testing for genetic mutations among DCIS patients in SEER. If there was an association between large- and diffuse-type disease and germline mutations, especially BRCA, this might explain their propensity to develop such widespread DCIS as well as their worse prognosis as evidenced by an increased risk of BCM. It would also offer a simple explanation for why they may gain a survival benefit when undergoing bilateral instead of unilateral mastectomy. It may be reasonable to perform genetic testing on patients with large and diffuse DCIS independent of family history to identify such germline mutations, as a large or diffuse DCIS lesion with underlying BRCA germline mutation would lend increased weight to recommendation and consideration of bilateral mastectomy among this cohort.
Future studies are needed to further elucidate this work, in particular prospective studies. Perhaps the most interesting study suggested by our results would be a trial in which patients with DCIS of sufficiently large size to require mastectomy would be randomized to undergo unilateral or bilateral mastectomy. While the difference between groups among patients with DCIS larger than 5 cm or diffuse disease was not statistically significant, our study suggests there may be a BCM benefit conferred by bilateral mastectomy. If the difference was significant, an important resulting question would be whether adjuvant ET after a unilateral mastectomy provides a comparable BCM benefit to bilateral mastectomy. An important issue in designing trials regarding this cohort is that because patients with disease requiring mastectomy were not included in the prospective DCIS trials, and the estimates from our study are known underestimates for the previously stated reasons, the calculation of an appropriate sample size would be challenging. Regardless, given the relative rarity of large and diffuse DCIS, such a study would need to be multi-institutional.
Acknowledgements
We acknowledge the National Cancer Institute, specifically the SEER team, for granting us access to SEER data for this analysis.
Funding
Olivier Harismendy is supported by awards from the National Cancer Institute (U01CA196406 and P30CA023100).
Footnotes
Availability of data and material SEER data are publicly available.
Code availability Can be provided upon request.
Conflict of interst There are no conflicts of interest to disclose.
Ethics approval Deferred by UCSD IRB.
Informed consent Informed consent could not be obtained due to the use of public, de-identified data.
Consent to participate Unable to obtain due to use of public, deidentified data.
Consent for publication Unable to obtain due to use of public, deidentified data.
Human and animal participate All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
References
- 1.Punglia R, Bifolck K, Golshan M et al. (2018) Epidemiology, Biology, Treatment, and Prevention of Ductal Carcinoma In Situ (DCIS). Jnci Cancer Spectr. 10.1093/jncics/pky063. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Early Breast Cancer Trialists’ Collaborative Group (EBCTCG), Correa C, McGale P, Taylor C, Wang Y, Clarke M, et al. (2010) Overview of the randomized trials of radiotherapy in ductal carcinoma in situ of the breast. J Natl Cancer Inst Monogr. 10.1093/jncimonographs/lgq039. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Barrio AV, Van Zee KJ (2017) Controversies in the treatment of ductal carcinoma in situ. Annu Rev Med 68:197–211. 10.1146/annurev-med-050715-104920 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Fisher B, Costantino J, Redmond C, Fisher E, Margolese R, Dimitrov N et al. (1993) Lumpectomy compared with lumpectomy and radiation therapy for the treatment of intraductal breast cancer. N Engl J Med 328:1581–1586. 10.1056/NEJM199306033282201 [DOI] [PubMed] [Google Scholar]
- 5.Donker M, Litière S, Werutsky G et al. (2013) Breast-conserving treatment with or without radiotherapy in ductal carcinoma In Situ: 15-year recurrence rates and outcome after a recurrence, from the EORTC 10853 randomized phase III trial. J Clin Oncol Off J Am Soc Clin Oncol 31:4054–4059. 10.1200/JCO.2013.49.5077 [DOI] [PubMed] [Google Scholar]
- 6.Holmberg L, Garmo H, Granstrand B, Ringberg A, Arnesson L-G, Sandelin K et al. (2008) Absolute risk reductions for local recurrence after postoperative radiotherapy after sector resection for ductal carcinoma in situ of the breast. J Clin Oncol Off J Am Soc Clin Oncol 26:1247–1252. 10.1200/JCO.2007.12.7969 [DOI] [PubMed] [Google Scholar]
- 7.Houghton J, George WD, Cuzick J et al. (2003) Radiotherapy and tamoxifen in women with completely excised ductal carcinoma in situ of the breast in the UK, Australia, and New Zealand: randomised controlled trial. Lancet Lond Engl 362:95–102. 10.1016/s0140-6736(03)13859-7 [DOI] [PubMed] [Google Scholar]
- 8.Fisher B, Dignam J, Wolmark N et al. (1998) Lumpectomy and radiation therapy for the treatment of intraductal breast cancer: findings from National Surgical Adjuvant Breast and Bowel Project B-17. J Clin Oncol Off J Am Soc Clin Oncol 16:441–452. 10.1200/JCO.1998.16.2.441 [DOI] [PubMed] [Google Scholar]
- 9.Cuzick J, Sestak I, Pinder SE et al. (2011) Effect of tamoxifen and radiotherapy in women with locally excised ductal carcinoma in situ: long-term results from the UK/ANZ DCIS trial. Lancet Oncol 12:21–29. 10.1016/S1470-2045(10)70266-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.McGuire KP, Santillan AA, Kaur P et al. (2009) Are mastectomies on the rise? A 13-year trend analysis of the selection of mastectomy versus breast conservation therapy in 5865 patients. Ann Surg Oncol 16:2682–2690. 10.1245/s10434-009-0635-x [DOI] [PubMed] [Google Scholar]
- 11.Silverstein MJ, Barth A, Poller DN et al. (1990) Ten-year results comparing mastectomy to excision and radiation therapy for ductal carcinoma in situ of the breast. Eur J Cancer Oxf Engl 1995(31A):1425–1427. 10.1016/0959-8049(95)00283-o [DOI] [PubMed] [Google Scholar]
- 12.Ward EM, DeSantis CE, Lin CC et al. (2015) Cancer statistics: breast cancer in situ. CA Cancer J Clin 65:481–495. 10.3322/caac.21321 [DOI] [PubMed] [Google Scholar]
- 13.Gray RJ (1988) A class of $k$-sample tests for comparing the cumulative incidence of a competing risk. Ann Stat 16:1141–1154. 10.1214/aos/1176350951 [DOI] [Google Scholar]
- 14.Ho DE, Imai K, King G, Stuart EA (2007) Matching as nonparametric preprocessing for reducing model dependence in parametric causal inference. Polit Anal 15:199–236. 10.1093/pan/mpl013 [DOI] [Google Scholar]
- 15.Morrow M, Bucci C, Rademaker A (1998) Medical contraindi-cations are not a major factor in the underutilization of breast conserving therapy. J Am Coll Surg 186:269–274. 10.1016/S1072-7515(97)00153-1 [DOI] [PubMed] [Google Scholar]
- 16.Schmale I, Liu S, Rayhanabad J et al. (2012) Ductal carcinoma in situ (DCIS) of the breast: perspectives on biology and controversies in current management. J Surg Oncol 105:212–220. 10.1002/jso.22020 [DOI] [PubMed] [Google Scholar]
- 17.Solin LJ (2019) Management of ductal carcinoma in situ (dcis) of the breast: present approaches and future directions. Curr Oncol Rep 21:33. 10.1007/s11912-019-0777-3 [DOI] [PubMed] [Google Scholar]
- 18.Narod SA, Iqbal J, Giannakeas V et al. (2015) Breast cancer mortality after a diagnosis of ductal carcinoma in situ. JAMA Oncol 1:888–896. 10.1001/jamaoncol.2015.2510 [DOI] [PubMed] [Google Scholar]
- 19.Glencer AC, Wong JM, Hylton NM et al. (2021) Modulation of the immune microenvironment of high-risk ductal carcinoma in situ by intralesional pembrolizumab injection. Npj Breast Cancer 7:59. 10.1038/s41523-021-00267-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Yi M, Krishnamurthy S, Kuerer HM et al. (2008) Role of primary tumor characteristics in predicting positive sentinel lymph nodes in patients with ductal carcinoma in situ or microinvasive breast cancer. Am J Surg 196:81–87. 10.1016/j.amjsurg.2007.08.057 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Satagopan JM, Ben-Porat L, Berwick M et al. (2004) A note on competing risks in survival data analysis. Br J Cancer 91:1229–1235. 10.1038/sj.bjc.6602102 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.O’Keefe TJ, Wallace AM (2020) Surveillance, Epidemiology, and End Results program underestimates breast cancer-specific mortality after ductal carcinoma in situ diagnosis. Breast Cancer Res Treat 182:761–762. 10.1007/s10549-020-05733-z [DOI] [PubMed] [Google Scholar]