aNon-upstaged defined as pTis and pN0 (n = 28 335); upstaged defined as pT > is, pN > 0, or pM = 1 (n = 7915).
Customize your JAMA Network experience by selecting one or more topics from the list below.
Chavez de Paz Villanueva C, Bonev V, Senthil M, et al. Factors Associated With Underestimation of Invasive Cancer in Patients With Ductal Carcinoma In Situ: Precautions for Active Surveillance. JAMA Surg. 2017;152(11):1007–1014. doi:10.1001/jamasurg.2017.2181
What factors are associated with underestimation of invasive cancer in patients presenting with low-risk ductal carcinoma in situ that would preclude active surveillance?
In this cohort study using the National Cancer Database, 22.2% of patients with a clinical diagnosis of non–high-grade ductal carcinoma in situ were found to have invasive carcinoma based on final pathologic findings at surgical excision. Factors that were significantly associated with an upgraded diagnosis of invasive carcinoma included younger age, negative hormone receptor status, more comorbidities, higher annual income, diagnosis in a more recent year, and treatment at an academic facility.
When selecting patients for active surveillance of ductal carcinoma in situ, consideration should be given to sociodemographic and biological factors that may be associated with underlying invasive cancer.
Recent recognition of the overdiagnosis and overtreatment of ductal carcinoma in situ (DCIS) detected by mammography has led to the development of clinical trials randomizing women with non–high-grade DCIS to active surveillance, defined as imaging surveillance with or without endocrine therapy, vs standard surgical care.
To determine the factors associated with underestimation of invasive cancer in patients with a clinical diagnosis of non–high-grade DCIS that would preclude active surveillance.
Design, Setting, and Participants
A retrospective cohort study was conducted using records from the National Cancer Database from January 1, 1998, to December 31, 2012, of female patients 40 to 99 years of age with a clinical diagnosis of non–high-grade DCIS who underwent definitive surgical treatment. Data analysis was conducted from November 1, 2015, to February 4, 2017.
Patients with an upgraded diagnosis of invasive carcinoma vs those with a diagnosis of DCIS based on final surgical pathologic findings.
Main Outcomes and Measures
The proportions of cases with an upgraded diagnosis of invasive carcinoma from final surgical pathologic findings were compared by tumor, host, and system characteristics.
Of 37 544 women (mean [SD] age, 59.3 [12.4] years) presenting with a clinical diagnosis of non–high-grade DCIS, 8320 (22.2%) had invasive carcinoma based on final pathologic findings. Invasive carcinomas were more likely to be smaller (>0.5 to ≤1.0 cm vs ≤0.5 cm: odds ratio [OR], 0.73; 95% CI, 0.67-0.79; >1.0 to ≤2.0 cm vs ≤0.5 cm: OR, 0.42; 95% CI, 0.39-0.46; >2.0 to ≤5.0 cm vs ≤0.5 cm: OR, 0.19; 95% CI, 0.17-0.22; and >5.0 cm vs ≤0.5 cm: OR, 0.11; 95% CI, 0.08-0.15) and lower grade (intermediate vs low: OR, 0.75; 95% CI, 0.69-0.80). Multivariate logistic regression analysis demonstrated that younger age (60-79 vs 40-49 years: OR, 0.84; 95% CI, 0.77-0.92; and ≥80 vs 40 to 49 years: OR, 0.76; 95% CI, 0.64-0.91), negative estrogen receptor status (positive vs negative: OR, 0.39; 95% CI, 0.34-0.43), treatment at an academic facility (academic vs community: OR, 2.08; 95% CI, 1.82-2.38), and higher annual income (>$63 000 vs <$38 000: OR, 1.14; 95% CI, 1.02-1.28) were significantly associated with an upgraded diagnosis of invasive carcinoma based on final pathologic findings.
Conclusions and Relevance
When selecting patients for active surveillance of DCIS, factors other than tumor biology associated with invasive carcinoma based on final pathologic findings may need to be considered. At the time of randomization to active surveillance, a significant proportion of patients with non–high-grade DCIS will harbor invasive carcinoma.
Ductal carcinoma in situ (DCIS) is a noninvasive neoplastic lesion of the breast, but it displays an unpredictable risk for developing into invasive cancer1 that can rarely lead to death.2 Widespread implementation of screening mammography has been linked to the increasing incidence of DCIS in asymptomatic women.3,4 Ductal carcinoma in situ is detected in 0.5 to 3.6 cases per 1000 women screened,5 and the annual incidence of DCIS in the United States approaches 60 000.4 It is unclear what proportion of these cases represents overdiagnosis of pathologic findings that would not otherwise progress to an adverse outcome. Because multimodal treatment is recommended for DCIS after identification, overdiagnosis results in overtreatment.6 To address overdiagnosis of breast cancer, purported to represent as many as 30% of new breast cancer cases,3 updated mammography screening guidelines recommend increasing the age to initiate screening as well as increasing the screening interval.7,8
Although controversy exists regarding the degree to which screening mammography prevents breast cancer deaths,3,5 the harms of overtreatment are well accepted. To address overtreatment of DCIS detected by screening mammography, the concept of active surveillance has been proposed to monitor the breast in which DCIS has been diagnosed and to intervene only in the event of progression to invasive disease.6,9-11 Three clinical trials have been developed to evaluate the outcomes of active surveillance of DCIS in patients with a low risk of progression.9,10,12 The details of each trial are compared in Table 1.9,10,12 All 3 trials use a randomized clinical study design in which women with low- to intermediate-grade DCIS diagnosed without an attempt to clear margins by either core and/or surgical biopsy are assigned to receive conventional care (surgery, with or without radiotherapy or endocrine therapy, followed by routine imaging surveillance) or active surveillance (imaging follow-up only), with a main outcome measure of the occurrence of ipsilateral invasive breast cancer.
A considerable barrier to active surveillance of DCIS is the risk of understaging invasive cancer when DCIS is found based on core biopsy. A meta-analysis of the underestimation of invasive cancer in the finding of DCIS based on diagnostic biopsy reported a significantly increased risk of understaging with the use of a smaller-gauge, non–vacuum-assisted device; high-grade DCIS; and the presence of a mass.13 These factors, which are available preoperatively, were incorporated into the exclusion criteria of the active surveillance trials, but the meta-analysis also showed that 20% of women with core biopsy–proven, non–high-grade DCIS (low and intermediate grades) still had invasive cancer when surgically treated.13 In a risk analysis predicting disease-specific cumulative mortality for active surveillance of DCIS compared with usual surgical and adjuvant care, the variable with the greatest risk of mortality was understaging invasive cancer at the time of diagnosis.11
We sought to determine the factors associated with an upgraded diagnosis of invasive cancer in patients with non–high-grade DCIS that would preclude active surveillance by applying active surveillance study criteria to relevant cases in the National Cancer Database (NCDB), a joint project of the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society.
We performed a retrospective cohort review using data from the NCDB from January 1, 1998, to December 31, 2012. We selected a study cohort based on combining selection criteria available in the NCDB from the protocols of the LORIS (Low Risk DCIS) trial,10 the LORD (Low Risk DCIS) study,9 and the COMET (Comparison of Operative to Monitoring and Endocrine Therapy)12 trial. The study cohort included women 40 years or older with clinical stage 0 (cTisN0M0), non–high-grade (low and intermediate grade) DCIS histologic findings. All patients underwent definitive cancer-directed surgery to confirm the final pathologic diagnosis that was recorded as non-upgraded (pTis) or upgraded (pT1-4). Cases missing data on final American Joint Commission on Cancer pathologic tumor, node, or metastasis stage from definitive cancer-directed surgery were excluded from the study. Variables that were analyzed included year of diagnosis, age group, race/ethnicity, estrogen receptor (ER) status, Charlson/Deyo Score,14 insurance status, annual income, and facility type. Racial/ethnic groups were identified by medical record or surname.15 The Charlson/Deyo Score is a weighted score derived from the sum of the scores for each of the comorbid conditions listed in the Charlson Comorbidity Score.14 Commission on Cancer facility type designation requires that a community cancer program treat 100 to 499 new cancer cases yearly, a comprehensive community cancer program treat 500 or more new cancer cases annually, and an academic or research program treat 500 or more new cancer cases annually and include research and training programs.16 The Loma Linda University institutional review board approved this study and waived the need for patient consent.
In the NCDB, grade17 and tumor size18 are reported based primarily on pathologic findings when available, with tumor size supplemented by radiographic and clinical findings secondarily. To select non–high-grade DCIS, we limited the analysis to cases with low- and intermediate-grade DCIS. For patients whose diagnosis was upgraded to invasive carcinoma based on final surgical pathologic findings, the pathologic stage and receipt of adjuvant radiotherapy, chemotherapy, and endocrine therapy were noted.
The χ2 and Fisher exact tests were used to compare proportions, as well as frequencies and percentages, whereas the t test and the Mann-Whitney Wilcoxon test were used to compare nonparametric quantitative variables (α = .05, assuming normal distribution and homogeneity of variance). Odds ratios (ORs) and 95% CIs were calculated for each covariate in the unadjusted and adjusted model using logistic regression. Missing variables were excluded from statistical analyses, which were conducted from November 1, 2015, to February 4, 2017, using SAS, version 9.4 (SAS Institute Inc). P < .05 was considered significant.
Between 1998 and 2012, there were 2 807 541 cases of malignant neoplasms of the breast in the NCDB. Case selection criteria yielded 37 544 women 40 years or older with non–high-grade DCIS who underwent definitive surgical therapy with pathologic stage available in the study cohort (mean [SD] age, 59.3 [12.4] years) (Figure). Removing cases with missing pathologic stage eliminated 35 313 cases from analysis. Overall, 8320 patients with non–high-grade DCIS (22.2%) had invasive carcinoma based on final surgical pathologic findings.
Demographic, clinical, and pathologic features of the study population are listed in Table 2. Most patients were younger than 80 years, with 18 283 of 36 250 (50.4%) between 40 and 59 years. Younger age (40-59 years) was associated with increased odds of an upgraded diagnosis to invasive carcinoma when compared with those 60 to 79 years of age (OR, 0.84; 95% CI, 0.77-0.92; P < .001). A total of 29 361 of 37 189 patients (79.0%) were non-Hispanic white, 4527 (12.2%) were non-Hispanic black, 1688 (4.5%) were Hispanic, and 1613 (4.3%) were other race/ethnicity. There was no difference in the rates of diagnoses of DCIS upgraded to invasive carcinoma among the racial/ethnic groups. Most tumors (20 616 of 22 774 [90.5%]) were ER positive and demonstrated a protective effect against an upgraded diagnosis (OR, 0.39; 95% CI, 0.34-0.43; P < .001).
For annual income categories, only the highest income category remained at risk for an upgraded diagnosis compared with the lowest income category on multivariate analysis (>$63 000 vs <$38 000: OR, 1.14; 95% CI, 1.02-1.28; P = .02) (Table 2). Most patients (27 438 of 31 737 [86.5%]) did not report any comorbidities; those with more comorbidities had increased odds of an upgraded diagnosis (Charlson/Deyo Score of 2 vs 0: OR, 1.28; 95% CI, 1.03-1.60; P = .03; Charlson/Deyo Score of 1 vs 0: OR, 1.14; 95% CI, 1.02-1.26; P = .02). Insurance status was not associated with upgrade rates.
There was an increase in the number of non–high-grade DCIS cases after 2007, with 22 133 of 37 544 patients (59.0) who received a diagnosis between 2008 and 2012 and 15 411 patients (41.0%) who received a diagnosis between 1998 and 2007, as well as an increase in the odds of an upgraded diagnosis for the most recent group (OR, 1.64; 95% CI, 1.50-1.80; P < .001) (Table 2). Comprehensive community cancer centers treated most patients (22 882 of 37 536 [61.0%]), while 10 463 patients (27.9%) were seen in academic research centers and 4191 patients (11.2%) were seen in community centers. Patients treated in an academic or comprehensive cancer center had higher odds of an upgraded diagnosis than did those treated in a community cancer center (OR, 2.08; 95% CI, 1.82-2.38; P < .001).
Based on postoperative findings, invasive carcinomas were more likely than DCIS to be smaller than 1 cm (5795 of 7268 [79.7%] vs 9568 of 16 052 [59.6%]; P < .001) and lower grade (2681 of 8320 [32.2%] vs 8700 of 29 224 [29.8%]; P < .001) (Table 2).
Table 3 shows the tumor characteristics and adjuvant treatment for 8320 patients for whom invasive carcinoma was identified based on final pathologic findings. Most invasive cancers were ER positive (6192 [74.4%]). Although 552 patients (6.6%) had ERBB2 (formerly HER2) amplification, most cases (5733 [68.9%]) did not report ERBB2 findings. The final pathologic tumor stage was T1 for 6388 patients (76.8%), with 1051 (12.6%) reporting microinvasion. Of the patients with node-positive carcinoma, 237 of 862 (27.5%) had micrometastatic neoplasms. The final pathologic stage was stage I for 7074 patients (85.0%), stage II for 862 patients (10.4%), stage III for 104 patients (1.3%), and stage IV for 13 patients (0.2%).
In our analysis of NCDB data, 22.2% of women presenting with a clinical diagnosis of non–high-grade DCIS were found to have invasive carcinoma at definitive surgery. This overall upgrade rate is comparable to that in the published literature. In a meta-analysis of 7350 cases from 52 studies that evaluated factors associated with invasive cancer based on final surgical pathologic findings when DCIS was diagnosed initially, the pooled risk of understaging was 25.9%. Core biopsies with high-grade DCIS had a significantly higher risk of underestimation at 32.3% compared with a 21.1% risk for non–high-grade (low- or intermediate-grade) DCIS.13 To estimate survival outcomes with an active surveillance protocol for DCIS, Ryser et al11 developed a risk projection model incorporating rates of progression of DCIS and invasive carcinoma and screening parameters using an 18.9% probability of understaging invasive cancer. The numbers of patients needed to treat to avoid 1 breast cancer death when diagnosed at 40 years was 28.3, when diagnosed at 55 years was 67.3, and when diagnosed at 70 years was 157.2. The variable with the greatest effect on model outcome was the probability of underestimating invasive cancer at the time of diagnosis. Our higher upgrade rate of 22.2% would estimate a greater mortality risk for active surveillance and decrease the number needed to treat to prevent 1 breast cancer death.
We found that younger age was associated with approximately a 20% higher risk of underestimation of invasive cancer, while ER-negative status was associated with approximately a 60% higher risk. There was no significant difference in upgrade rates based on race or ethnicity. Although young age, hormone receptor negativity, and black race are known risk factors for poorer outcomes with DCIS,2,4,19 these variables have been less consistently studied as predictors of understaging of invasive cancer at diagnosis. Yen et al20 corroborated a 2-fold higher odds of upstaging (OR, 2.19; 95% CI, 1.11-4.34; P = .02) with age younger than 55 years compared with older women. However, other studies have failed to show a significant association between patient age and risk of understaging at diagnosis.13,20-27 The effect of hormone receptor status of DCIS at initial diagnosis has rarely been reported; Park et al25 did not find a significant association between ER status and risk of underestimation of invasive cancer. The COMET trial uniquely mandates hormone receptor–sensitive DCIS as inclusion criteria.12 To our knowledge, the effect of race and ethnicity on the likelihood of upgrading of DCIS has not been previously reported.
Other factors assessed in this study were not specific to tumor biology. Patients with more comorbidities were at higher risk of understaging (14% higher odds with 1 comorbidity and 28% higher odds with >1 comorbidity). The presence of multiple comorbidities could preclude optimal imaging or biopsy technique and decrease diagnostic accuracy. Treatment at an academic or research center or a comprehensive community cancer center was associated with up to a 2-fold higher rate of upgrading to invasive cancer than was treatment at a community cancer program. This finding could reflect higher volumes of definitive cancer therapy at these institutions, as CoC facility type classification designates hospital case volume.16 The association of higher annual income with upgrade could mirror the higher incidences of breast cancer and mammography screening uptake that have been documented in women of higher income levels.28,29
More than half of the patients with DCIS (59.0%) received a diagnosis in the last 5 years of the study period (2008-2012). In 2008, the CoC required reporting of clinical stage by registrars30; this mandate may have contributed to the increase in the number of cases staged as cTisN0M0. However, the increasing incidence of DCIS during the mammography screening era has been well recognized.3-5 That later years of study were associated with a 64% increased odds of understaging could be attributed to greater use of percutaneous needle biopsy for diagnosis, which has been shown to be associated with understaging.20 The association of study year with upgrade rates was corroborated by Brennan et al,13 with cases diagnosed in 2001 and after having the highest odds of upgrade compared with cases diagnosed in earlier years.
A major limitation with the use of NCDB data during this study period is that tumor size and grade primarily reflect final pathologic findings.17,18 Because tumor size is recorded from the surgical pathologic report, we could not evaluate tumor size as a preoperative predictor of underestimation of invasive cancer in this study. More than half of occult invasive cancers (57.2%) were 0.5 cm or less. For patients whose diagnosis was not upgraded to invasive cancer, the extent of DCIS based on final pathologic findings was larger, with 4.7% of patients having a tumor size larger than 5 cm. On January 1, 2016, the CoC mandated reporting of both pathologic and clinical tumor size data in the NCDB30; subsequent studies could evaluate tumor size as a preoperative variable.
Similarly, tumor grade is recorded in the NCDB based on the most definitive pathologic findings available. We limited case selection to clinical TisN0M0 with low- or intermediate-grade histologic findings to reflect the patient population screened for active surveillance. However, in doing so, cases upgraded to high-grade invasive carcinoma based on final pathologic findings were potentially eliminated, as were cases of high-grade DCIS upgraded to low- or intermediate-grade invasive carcinoma. Because the grade on core biopsy reflects the grade on final pathologic findings in most cases (71.1%), with correlation improved for high grade (83.8% for high grade, 66.0% for intermediate grade, and 68.7% for low grade),31 we thought it a reasonable assumption for the majority of patients to use the grade variable in the NCDB. Furthermore, genomic analyses have shown that, when progression occurs, low-grade in situ lesions lead to low-grade invasive lesions and high-grade in situ lesions lead to high-grade invasive lesions; dedifferentiation from low-grade tumors to high-grade tumors occurs rarely and is marked by differences in both the frequency and the mechanism of 16q loss.32 We evaluated all cTisN0M0 cases in this data set unselected by grade and found upgrading in only 5.4% (5446 of 101 075). Grade assignment could also be affected by lack of documentation of the source of pathologic findings (eg, core biopsy, excisional biopsy, or partial or total mastectomy specimen) and interobserver variability.31,33 Although the meta-analysis by Brennan et al13 demonstrated a significant difference on univariate analysis between upgrade rates of high-grade vs non–high-grade DCIS diagnosed preoperatively, other single-institution series failed to show significant differences in upgrade rates based on DCIS grade in multivariate analyses.25,27
As expected, the majority of invasive carcinomas identified at surgery were hormone sensitive (74.4% ER positive), small (54.4% T1a or T1mi), and node negative (88.0% N0). However, these cases represent potential missed, or at least delayed, diagnoses in a population undergoing active surveillance. A significant proportion of patients with a diagnosis of invasive cancer received adjuvant therapies. Several studies have attempted to predict outcomes of active surveillance retrospectively. Pilewskie et al34 stratified patients by LORIS criteria after breast-conserving surgery for pure DCIS and found that the 10-year risk of any ipsilateral breast tumor recurrence was lower for patients who met LORIS criteria (10.3%) than those who did not (15.4%), but the difference was not statistically significant (P = .08). In this study, all patients received standard surgical care, and upgrade to invasive carcinoma at the time of definitive surgery was excluded. Although the rate of progression to ipsilateral breast tumor recurrence was low overall, the authors acknowledged that patients not receiving initial excision would likely have higher rates of invasive cancer. Sagara et al35 reviewed cases of DCIS in the Surveillance, Epidemiology, and End Results database and compared the 2% of patients who did not receive surgery with the 98% of patients who received surgery. Although there were significant differences in patients with intermediate-grade and high-grade DCIS, there was no significant difference in 10-year breast cancer–specific survival in patients with low-grade DCIS who did or did not undergo definitive cancer surgery (98.6% vs 98.8%; hazard ratio, 0.85; 95% CI, 0.21-3.52; P = .83). Overall, the 10-year breast cancer–specific survival for those who did not receive surgery was 93.4% compared with 98.5% for those who underwent surgery (P < .001). Grade reporting in the Surveillance, Epidemiology, and End Results database is subject to similar considerations in the present study,17 and with active surveillance of DCIS, final pathologic grade is not available from a surgical specimen.
In addition to the limitations stated, we could not determine presenting symptoms or method of diagnosis. Factors associated with increased rates of upgrade to invasive cancer after an initial diagnosis of DCIS include the presence of a mass lesion detected during physical examination or imaging, bloody nipple discharge, and the use of needle biopsy for diagnosis.13,20-27 Data on the use of initial breast imaging or breast imaging results are not available in the NCDB. The use of needle biopsy for diagnosis rather than surgical excision could have contributed to increased rates of upgrade observed in recent years as percutaneous biopsy technology became widespread, academic and research or comprehensive community cancer programs adopted new technology, patients with comorbidities underwent less invasive procedures, and higher-income patients had access to more screening and percutaneous biopsy procedures. Moreover, risk factors for breast cancer were not available for review. Because many of the exclusion criteria for the active surveillance trials (Table 1) were not available in the NCDB and the study spans a time frame in which diagnostic technology has improved, the present analysis could overestimate projected upstaging in active surveillance protocols.
Successful use of active surveillance for patients with DCIS to diminish harmful effects of overtreatment resulting from overdiagnosis will depend on accurate diagnosis and diligent follow-up. One study documented that only 34% of patients with DCIS underwent annual mammography 5 years after breast conservation and 15% reported having annual mammography at 10 years.36 Novel imaging technology, such as digital breast tomosynthesis or abbreviated and functional magnetic resonance imaging, and algorithms to optimize screening indications and intervals are being studied to improve diagnostic accuracy and decrease false positives.37 Refinement of molecular assays38 to predict which DCIS are obligate and nonobligate precursors to invasive disease could be used to help stratify risk of progression and would be particularly useful in the unavoidable event of understaging at diagnosis.
Our study showed that approximately 1 in 5 women presenting with non–high-grade DCIS will have an underlying invasive carcinoma and adds to the burgeoning evidence for social determinants of health outcomes. In addition to biological factors, sociodemographic factors, such as treatment at academic or research centers and higher annual income, were associated with an increased risk of understaging at diagnosis. The use of active surveillance must be implemented cautiously because the patients at greatest risk for harboring underlying invasive carcinoma are also most likely to be offered entry into an active surveillance protocol.
Accepted for Publication: April 1, 2017.
Correction: This article was corrected on January 17, 2018, to fix errors in the numbers in the Figure and the numbers (percentages) in Table 3.
Corresponding Author: Sharon S. Lum, MD, Division of Surgical Oncology, Department of Surgery, Loma Linda University School of Medicine, 11175 Campus St, Coleman Pavilion 21111, Loma Linda, CA 92350 (email@example.com).
Published Online: July 12, 2017. doi:10.1001/jamasurg.2017.2181
Author Contributions: Dr Lum had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Chavez de Paz Villanueva, Bonev, Garberoglio, Lum.
Acquisition, analysis, or interpretation of data: Chavez de Paz Villanueva, Senthil, Solomon, Reeves, Namm, Lum.
Drafting of the manuscript: Chavez de Paz Villanueva, Bonev, Lum.
Critical revision of the manuscript for important intellectual content: Chavez de Paz Villanueva, Senthil, Solomon, Reeves, Garberoglio, Namm, Lum.
Statistical analysis: Chavez de Paz Villanueva, Lum.
Administrative, technical, or material support: Chavez de Paz Villanueva, Senthil, Lum.
Study supervision: Chavez de Paz Villanueva, Senthil, Solomon, Garberoglio, Namm, Lum.
Conflict of Interest Disclosures: None reported.
Disclaimer: The data used in the study are derived from a deidentified National Cancer Database file. The American College of Surgeons and the Commission on Cancer have not verified and are not responsible for the analytic or statistical methods used, or the conclusions drawn from these data, by the investigator.
Meeting Presentation: This paper was presented at the 88th Annual Meeting of the Pacific Coast Surgical Association; February 18, 2017; Indian Wells, California.
Create a personal account or sign in to: