Fax: (206) 667-5948
Risk of invasive breast carcinoma among women diagnosed with ductal carcinoma in situ and lobular carcinoma in situ, 1988-2001
Article first published online: 10 APR 2006
Copyright © 2006 American Cancer Society
Volume 106, Issue 10, pages 2104–2112, 15 May 2006
How to Cite
Li, C. I., Malone, K. E., Saltzman, B. S. and Daling, J. R. (2006), Risk of invasive breast carcinoma among women diagnosed with ductal carcinoma in situ and lobular carcinoma in situ, 1988-2001. Cancer, 106: 2104–2112. doi: 10.1002/cncr.21864
- Issue published online: 27 APR 2006
- Article first published online: 10 APR 2006
- Manuscript Accepted: 19 DEC 2005
- Manuscript Revised: 15 NOV 2005
- Manuscript Received: 7 OCT 2005
- National Cancer Institute (NCI), through a contract with the Fred Hutchinson Cancer Research Center. Grant Number: K01-CA101970
- breast cancer;
- ductal carcinoma in situ;
- lobular carcinoma in situ;
- second primary cancers
Incidence rates of ductal carcinoma in situ (DCIS) and lobular carcinoma in situ (LCIS) have been rising, but little is known about which patients will develop invasive breast cancer or what types of tumors these patients may develop.
By using Surveillance, Epidemiology and End Results (SEER) data, the authors evaluated how types of invasive breast cancers diagnosed among 37,692 DCIS and 4490 LCIS patients differed and how clinical characteristics influenced subsequent breast cancer risk.
Among DCIS patients, incidence rates of ipsilateral and contralateral invasive breast cancer were 5.4/1000 person-years and 4.5/1000 person-years, respectively; and among LCIS patients, incidence rates were 7.3/1000 person-years and 5.2/1000 person-years, respectively. LCIS patients were 5.3-fold more likely than DCIS patients to develop invasive lobular carcinomas. Women whose DCIS had comedo histologic features or was poorly differentiated had 1.4-fold and 2.0-fold elevations in ipsilateral invasive breast cancer risk. Furthermore, among DCIS patients, 20-49 year-olds and black women and Hispanic white women had 1.6, 2.7, and 2.3-fold elevated risks of Stage III/IV breast cancer compared with 50-59 year-olds and non-Hispanic whites, respectively.
Screening young DCIS patients more frequently and improving the follow-up care of blacks and Hispanic whites with DCIS may reduce their risk of advanced-stage breast cancer. In addition, LCIS may be a precursor rather than just an ambiguous risk factor for invasive breast cancer, and, therefore, localized treatment for LCIS may be warranted. Given that incidence rates of DCIS and LCIS have been rising, investigations of these tumors should be continued to better understand their etiology and appropriate clinical management. Cancer 2006. © 2006 American Cancer Society.
Incidence rates of both ductal carcinoma in situ (DCIS) and lobular carcinoma in situ (LCIS) have increased 7.2-fold and 2.6-fold, respectively, from 1980-2001.1 These trends have been attributed primarily to the increasing numbers of mammograms and biopsies of suspicious lesions that are being performed.2, 3 DCIS is generally thought to be a precursor lesion of invasive breast cancer, and compared with women in the general population, women diagnosed with DCIS have a 2.0-fold to 8.6-fold higher risk of developing invasive breast cancer.4–7 Alternatively, lobular carcinoma in situ (LCIS) is viewed as a marker of an increased risk of invasive breast cancer, rather than as a true precursor lesion.8 Compared with women in the general population, women with LCIS have a 3.0-fold to 4.2-fold higher risk of invasive breast cancer.3–6 LCIS is a challenging disease to study because, unlike DCIS, it lacks clinical signs and is almost always an incidental finding in breast biopsies performed for another reason, such as a suspicious mammogram. Although LCIS has long been thought to not be associated with any specific mammographic findings, more recent data indicate that calcifications are seen in 21% to 67% of LCIS cases.9–11 Clinically, LCIS and DCIS are viewed quite differently. Unlike DCIS, LCIS is commonly viewed as a nonsurgical disease because several small studies found that LCIS patients are just as likely to develop invasive tumors in the breast ipsilateral to their LCIS as they are in their contralateral breast.12–16 On the basis of these studies, it has been argued that the only definitive surgical treatment for LCIS is a bilateral mastectomy, an unnecessary operation for an estimated 75% of LCIS patients.17
Using data on DCIS and LCIS patients diagnosed from 1988 to 2002 in 11 cancer registries that participate in the Surveillance, Epidemiology, and End Results (SEER) Program, we evaluated how types of invasive breast cancers that women with DCIS and LCIS develop differ and how characteristics of carcinomas in situ influence risk of subsequent invasive breast cancer.
MATERIALS AND METHODS
Women diagnosed with unilateral DCIS or LCIS between January 1988 and December 2002 without a prior history of any type of in situ or invasive cancer were identified through 11 population-based cancer registries in the United States that participate in the National Cancer Institute's SEER Program. January 1988 was chosen as the starting point for this analysis because beginning in 1988, all of the SEER registries began to uniformly record data on estrogen receptor and progesterone receptor status. Patient medical records are the principal source of data used by SEER. It is estimated that over 95% of all incident cancer cases in the populations under surveillance are ascertained. Additional operational details and methods used by the SEER Program are provided elsewhere.18
A total of 49,852 potentially eligible in situ carcinoma cases were identified from SEER. In situ carcinoma cases with less than 6 months of follow-up were excluded (n = 3981) to ensure that at least 6 months separated a woman's in situ carcinoma diagnosis from her invasive carcinoma diagnosis. DCIS cases were defined based on the following International Classification of Diseases for Oncology (ICDO-3) codes: 8201, 8230, 8500, 8501, 8503, 8507, and 8523 (n = 37,692); LCIS cases were defined based on 1 of the following ICDO-3 codes, 8520 and 8524 (n = 5138). DCIS cases were also grouped by subtype into the following 5 categories, DCIS not otherwise specified (NOS; ICDO-3 codes: 8500 and 8523), comedocarcinoma (ICDO-3 code: 8501), papillary (ICDO-3 code: 8503, 8507), cribiform (ICDO-3 code: 8201), and solid (ICDO-3 code: 8230). All of these categorizations are based on a scheme that has been used previously.1 Women with other types of carcinoma in situ (CIS), including tumors containing both DCIS and LCIS were excluded from this analysis because there were too few cases of these types of CIS to evaluate them separately (total n = 3041). In addition, all LCIS cases treated with a total mastectomy were excluded from all analyses (n = 648, 12.6% of the total). This exclusion was made because data on whether or not women were treated with a unilateral or a bilateral mastectomy was available from SEER only from 1998-2002. Data from 1998-2002 indicated that 65% of LCIS patients treated with a total mastectomy received a bilateral mastectomy. Thus, given the high proportion of bilateral mastectomies occurring among LCIS patients and the finding that those receiving this treatment have an extremely low risk of developing subsequent invasive breast cancer, we elected to exclude all LCIS patients treated with a mastectomy. This same criterion was not applied to DCIS patients because only 11% of DCIS patients treated with a total mastectomy received a bilateral mastectomy based on the 1998-2002 SEER data.
All cases of both DCIS and LCIS were followed for diagnosis of a subsequent invasive breast cancer. Cases of invasive breast cancer were defined as those that were diagnosed 6 months or longer after a CIS diagnosis. As a result, DCIS and LCIS cases with less than 6 months of follow-up were excluded from analysis. Vital status and follow-up information are ascertained annually through a variety of local, regional, and national sources. Follow-up duration is calculated in months using the date of diagnosis and whichever of the following occurred first, 1) date of death, 2) date last known to be alive, or 3) December 2002 (the follow-up cutoff date used in our analysis). In our analysis, subjects were censored as cases at the time they were diagnosed with invasive breast cancer. The remaining subjects were censored at 1) date of death, 2) date last known to be alive, 3) December 2002, or 4) the date they were diagnosed with a second in situ carcinoma (n = 936), whichever occurred first.
Along with age and year of diagnosis, SEER also provides information on race/ethnicity, laterality of the tumor, American Joint Committee on Cancer stage, tumor size, tumor grade, number of positive lymph nodes, number of lymph nodes examined, histology, estrogen-receptor status, and progesterone-receptor status. Information on surgical and radiation treatments administered within 4 months of diagnosis is available, but data on adjuvant hormonal therapy are not. For this study, we divided information on cancer-directed surgical treatment received for CIS into 3 categories, 1) none, 2) partial or less than total mastectomy (i.e., excisional biopsy, lumpectomy, segmental mastectomy, quadrectomy, tylectomy, wedge resection, nipple resection, and partial mastectomy, NOS), and 3) total mastectomy (i.e., modified radical mastectomy, radical mastectomy, and total mastectomy, NOS). In addition, for women who received either a partial or less than total mastectomy or a total mastectomy, we created a variable indicating whether or not lymph node resection was a part of the surgical treatment based on the data available from SEER. In our assessments of incidence rates of ipsilateral breast cancer among DCIS patients, analyses were restricted to women whose DCIS was not treated with a total mastectomy, given the very low risk of ipsilateral invasive breast cancer that these women have.
All statistical analyses were performed using Stata SE for Windows (College Station, TX). Multivariate adjusted hazard ratios (HR) and 95% confidence intervals (95% CI)) were computed using the Cox proportional hazard model.19 Two sets of analyses were performed. First, we compared the histologic types of invasive breast cancers that women with DCIS developed to the types that women with LCIS developed (2 categories were evaluated, ductal [ICDO-3 code: 8500] and lobular [ICDO-3 codes: 8520 and 8522]). Because all LCIS patients treated with a mastectomy were excluded from our main analyses, for this analysis, which compared DCIS patients to LCIS patients, we also excluded DCIS patients treated with a total mastectomy to make these 2 groups more comparable. This analysis was adjusted for age and year of CIS diagnosis, SEER registry, race/ethnicity, and whether or not cancer-directed surgical treatment for the CIS was performed.
In our second set of analyses, we evaluated how specific features of DCIS (such as age at diagnosis, race/ethnicity, treatment, histologic subtype, tumor size, and tumor grade) relate to risk of developing an invasive breast cancer, and we evaluated how features of LCIS relate to risk of developing an invasive breast cancer. (Note: the relations between LCIS radiation treatment, tumor size, and grade and risk of invasive breast cancer could not be assessed because of missing data and the relatively small number of women diagnosed with LCIS.) We also stratified these analyses by the laterality of the invasive breast cancer. Again, our analyses of factors associated with risk of ipsilateral invasive breast cancer among DCIS patients were restricted to patients whose DCIS was not treated with a total mastectomy. All hazard ratios for risk of invasive breast cancer after DCIS were adjusted for age and year of DCIS diagnosis, race/ethnicity, SEER registry, and surgery/radiation treatments for DCIS. All hazard ratios for risk of invasive breast cancer after LCIS were adjusted for age and year of LCIS diagnosis, race/ethnicity, SEER registry, and surgical treatments for LCIS. In addition, hazard ratios for DCIS and LCIS treatments were adjusted for whether or not lymph node resection was a part of the surgical treatment.
LCIS cases had a slightly younger mean age at diagnosis compared with DCIS cases (54.3 ± 11.0 vs. 58.6 ± 12.9; Table 1). LCIS patients were much more likely than DCIS patients not to have received cancer-directed surgery (8.2% vs. 2.2%) or radiation therapy (97.8% vs. 64.5%). Among patients treated with a partial or less than total mastectomy, 97.6% (3951 of 4046) of LCIS patients did not receive radiation, whereas only 45.9% (10,883 of 23,687) of DCIS patients treated with a partial or less than total mastectomy did not receive radiation (data not shown).
|DCIS (n = 37,692)||LCIS (n = 4490)|
|Age at in situ carcinoma diagnosis|
|Mean age, y (SD)||58.6||(12.9)||54.3||(11.0)|
|Year of in situ carcinoma diagnosis|
|American Indian/Alaska Native||78||0.2||9||0.2|
|Surgery for in situ carcinoma|
|Partial/less than total mastectomy||24,014||63.9||4080||91.8|
|without axillary node dissection||21,591||57.5||3980||89.5|
|with axillary node dissection||2418||6.4||100||2.2|
|Radiation therapy for in situ carcinoma|
|Surgery and radiation therapy for in situ carcinoma|
|Partial/less than total mastectomy only||10,883||29.3||3,951||89.6|
|Total mastectomy only||12,422||33.4||NA*|
|Surgery (any type) and radiation||13,047||35.1||95||2.2|
Of the 37,692 DCIS cases, 1504 (4.0%) were diagnosed with invasive breast cancer for an age-adjusted incidence rate of 8.0/1000 person-years over the entire study period (Table 2). Of the 4490 LCIS cases, 282 (6.2%) were diagnosed with invasive breast cancer for an age-adjusted incidence rate of 12.5/1000 person-years. Stratified by laterality, the incidence rate of ipsilateral invasive breast cancer among women with DCIS was 5.4/1000 person-years, and was 4.5/1000 person-years for contralateral breast cancer. LCIS patients had an incidence rate of ipsilateral invasive breast cancer that was 35% higher than that of DCIS patients (rate = 7.3/1000 person-years), but a rate of contralateral breast cancer that was only 16% higher than that of DCIS patients (rate = 5.2/1000 person-years).
|Duration of Follow-up, mo||All Cases of Invasive Breast Cancer||Ipsilateral vs. Contralateral Invasive Breast Cancer|
|No. of Subjects (%)||Cumulative No. of Cases||Total AAIR||Cumulative % Total Dx IBC||Cumulative No. of Ipsilateral Cases*||Total AAIR†‡||Cumulative No. of Contralateral Cases*||Total AAIR|
|DCIS (n= 37,692)|
|Mean [SD]||66.1 [44.0]|
|LCIS (n= 4490)|
|Mean [SD]||66.6 [44.6]|
In an analysis restricted to DCIS and LCIS patients not treated with a total mastectomy, compared with DCIS patients, LCIS patients were 0.8-fold (95% CI, 0.7-1.0) less likely to be diagnosed with invasive ductal carcinoma, and 5.3-fold (95% CI, 4.1-6.9) more likely to be diagnosed with invasive lobular carcinoma (Table 3).
|Histology of the Invasive Cancer||Invasive Cancers Diagnosed Among Women With DCIS||Invasive Cancers Diagnosed Among Women With LCIS|
|No. of Cases||% of Total||HR†||95% CI||No. of Cases||% of Total||HR†||95% CI|
We examined the effects of DCIS features on the risk of subsequent ipsilateral and contralateral breast cancer separately. Compared with women diagnosed with DCIS at age 50-59 years, women diagnosed at 20-49 years of age had an elevated risk of ipsilateral invasive breast cancer (Table 4). With respect to race/ethnicity, blacks with DCIS had increased risks of both ipsilateral and contralateral breast cancer compared with non-Hispanic whites. Women with comedo, papillary, and solid DCIS had modest increased risks, whereas women with cribiform DCIS had a reduced risk of ipsilateral breast cancer compared with women with DCIS NOS, though only the risk associated with comedo DCIS was statistically significant. Women with solid DCIS also had an elevated risk of contralateral breast cancer, but, again, this risk was within the limits of chance. Finally, compared with women with well differentiated DCIS, women with poorly differentiated DCIS had an elevated risk of ipsilateral, but not of contralateral, invasive breast cancer.
|DCIS Characteristic||All Cases||Ipsilateral Cases*||Contralateral Cases|
|HR||95% CI||HR||95% CI||HR||95% CI|
|Age at diagnosis, y†|
|Cancer directed surgery/radiation**|
|Partial/less than total mastectomy only||1.0||ref||1.0||ref||1.0||ref|
|Surgery (any type) and radiation||0.6||0.6-0.7‡||0.7||0.5-1.2||1.0||0.9-1.3|
|Tumor size, cm§††|
With respect to age, women diagnosed with DCIS between the ages of 20-49 years had 1.4-fold (95% CI, 1.1-1.9) and 1.6-fold (95% CI, 1.0-2.6) elevated risks of Stage II and Stage III/IV invasive breast cancers, respectively, compared with women diagnosed with DCIS at age 50-59 years (Table 5). Compared with non-Hispanic white women, black women had elevated risks of Stage II (HR = 1.7; 95% CI, 1.2-2.4) and Stage III/IV breast cancer (HR = 2.6; 95% CI, 1.5-4.4), and Hispanic white women had an elevated risk of Stage III/IV breast cancer (HR = 2.3; 95% CI, 1.1-4.8).
|DCIS Characteristic||Stage I||Stage II||Stage III/IV|
|HR||95% CI||HR||95% CI||HR||95% CI|
|Age at diagnosis, y*|
Age, race/ethnicity, and type of cancer-directed surgery among LCIS patients were not related to risk of invasive breast cancer (data not shown).
Before interpreting results of this study, it is important to acknowledge its limitations. First, the histologic categorizations used were based on diagnoses made by pathologists from multiple institutions, and diagnostic criteria likely varied across pathologists, thus resulting in a certain degree of misclassification error. In particular, SEER captures data on DCIS but not on atypical ductal hyperplasia, and distinguishing between these two lesions can be quite challenging given the somewhat vague criteria used for their diagnoses. This same problem is also present for the distinction between LCIS and atypical lobular hyperplasias (which are also not recorded by SEER).20 Studies evaluating the concordance between tumors classified as DCIS or LCIS by SEER and those classified through a centralized pathology review are needed to quantify the magnitude of this misclassification, as none have been reported in the literature.
Another limitation of this study is that we lacked data on factors that could influence the development of invasive carcinomas, such as use of hormonal therapy, mammography utilization, family history of breast cancer, and body mass index. Similarly, some of the differences in risk we observed across treatment groups may have been due to confounding by indication, as characteristics of a woman's initial in situ carcinoma or personal history may have affected her treatment choice. The impact of these unmeasured confounders on both directions and magnitudes of the risks that we observed is uncertain. However, it is important to acknowledge that this is an observational study, so results of this study should not influence current clinical guidelines in the absence of supportive data from randomized trials.
Despite these limitations, strengths of this study are that it is population-based, includes women diagnosed in multiple geographic regions, and includes a large sample size. This study also confirms results of various other studies, thus lending credence to the current study's findings. For example, we found that 4.0% (1504 of 37,692) of subjects in our DCIS cohort developed invasive contralateral breast cancer, which is within the range of 2% to 6% reported in 4 prior studies.21–23 In addition, the relation between comedo DCIS and risk of subsequent ipsilateral breast cancer has been noted in other studies.21, 22, 24
Four potential implications of this study, when considered in the context of other available evidence, are as follows: 1) Because young women diagnosed with DCIS had elevated risks of advanced stage breast cancer, screening them more frequently may reduce these risks; 2) Compared with non-Hispanic white women, black women and Hispanic white women with DCIS also have an elevated risk of advanced stage breast cancer, suggesting that efforts to improve follow-up care for black and Hispanic white women with a history of DCIS are needed to reduce this health disparity; 3) Rather than simply a marker of risk for invasive breast cancer, LCIS may be a precursor lesion of ILC; and 4) Given that rates of ipsilateral invasive cancer are higher than rates of contralateral cancer among women with LCIS, the treatment of LCIS as a nonsurgical disease may need to be reconsidered. Each of these implications is discussed in detail below.
We noted that women diagnosed with DCIS between the ages of 20-49 years had elevated risks of advanced stage breast cancers compared with older women. The reason for this is unclear, but it may be related to differences in the biologic aggressiveness of invasive tumors by age because younger women tend to be diagnosed with more biologically aggressive tumors. Thus, they may grow and spread more quickly during the time between their follow-up mammograms compared with the less aggressive tumors that older women develop. So our findings suggest that shorter screening intervals for women diagnosed with DCIS at a young age, or perhaps screening them with additional modalities, may be potential means of reducing their risk of advanced stage breast cancer, though these approaches require further study.
We also observed that black women and Hispanic white women with DCIS had elevated risks of advanced stage breast cancer. These findings are consistent with several studies indicating that black women in the general population have higher risks of advanced stage breast cancer and of mortality after their diagnosis than do white women.25, 26 These studies suggest that these disparities are primarily related to socioeconomic status and access to appropriate care rather than to differences in tumor biology or genetic susceptibility. A recent study observed that compared with non-African American DCIS patients, African American DCIS patients were significantly less likely to receive radiation after their lumpectomy.27 Two previous studies have also evaluated risk of second breast cancers by race. One found that black women did not have an increased risk of a second primary in situ or invasive contralateral breast cancer, though its analyses were limited by a small sample size.7 Consistent with our results, the second study observed that among women with localized invasive or in situ breast cancer, black women had an increased risk of contralateral breast cancer compared with white women.28 Thus, efforts to improve follow-up care for black women and Hispanic white women diagnosed with DCIS may be an important means of lowering their risk of subsequent invasive breast cancer.
DCIS is generally considered to be a precursor lesion for invasive breast cancer, whereas LCIS is typically thought of as a nonsurgical disease that is viewed as a risk factor for, but not a precursor of, invasive breast cancer. Our observation that women with LCIS are 5.3-fold more likely to develop invasive lobular cancer (ILC) and 0.8-fold less likely to develop invasive ductal carcinoma (IDC) compared with women with DCIS, may lead one to question the latter description. These results are also consistent with a previous report indicating that LCIS is often followed by ILC,29 and suggest that LCIS may be a precursor of ILC specifically, as one might expect based on its histopathologic characteristics. There is laboratory evidence to support this. E-cadherin is a cell adhesion molecule that is not expressed by almost all ILCs, but this molecule is expressed by almost all IDCs. Acs et al.30 found that 96 of 100 (96%) IDCs expressed e-cadherin, whereas 41 of 42 (98%) ILCs showed complete loss of e-cadherin expression. Similarly, 128 of 131 DCIS (98%) cases expressed e-cadherin, whereas 50 of 53 LCIS (94%) cases had complete loss of e-cadherin expression. Thus, e-cadherin not only distinguishes lobular from ductal carcinomas, but also loss of its expression appears to be an early event in the development of lobular carcinomas. So there is growing evidence suggesting that LCIS may be a precursor of ILC, rather than just a marker of invasive breast cancer risk.
One of the reasons for our observation that women with LCIS have a higher incidence rate of invasive breast cancer, and particularly of ipsilateral invasive breast cancer, may be because LCIS has been historically viewed as a nonsurgical disease and, thus, often not treated definitively. This is largely due to results of several studies published from 1974 to 1991 that were based on cohorts of LCIS ranging in size from 34 to 258 cases, with 6 to 54 incident invasive breast cancers. These studies found that women with LCIS were equally likely to be diagnosed with ipsilateral and contralateral invasive breast cancers. Thus, it has been argued that the only logical operation for LCIS would be a bilateral mastectomy, which would be unnecessary approximately 75% of the time.17 Our results challenge this notion, as we actually found that incidence rates of ipsilateral invasive breast cancer were higher among LCIS patients compared with DCIS patients (7.3/1000 vs. 5.4/1000, respectively), whereas rates of contralateral invasive breast cancer were more similar between LCIS and DCIS patients (5.2/1000 vs. 4.5/1000, respectively). The reason for these differences may be that DCIS patients are more likely to have definitive local surgical treatment than are LCIS patients. Specifically, it is striking that only 95 of 4046 (2%) LCIS patients in this study who received a partial or less than total mastectomy also received radiation, whereas 12,804 of 23,687 (54%) DCIS patients who received a partial mastectomy were treated with radiation. Clearly, LCIS patients are routinely not receiving definitive local treatment, but, based on our study, it appears that these women have a higher incidence rate of ipsilateral invasive breast cancer than of contralateral breast cancer, contrary to what has been reported previously. Thus, local treatment for women with LCIS may be warranted, although further studies are needed to evaluate this.
In summary, this study provides greater insight into the clinical significance of DCIS and LCIS over a recent time period in which the incidence rates of these two lesions has increased dramatically.1 With respect to DCIS, our results suggest that screening women diagnosed with DCIS at a young age more frequently may be one way to reduce their risk of subsequent advanced-stage breast cancers. Furthermore, improving follow-up and screening of black women and Hispanic white women with DCIS may be an important means of reducing their risks of advanced-stage invasive breast cancer. With respect to LCIS, our data indicate that LCIS may be a precursor lesion of ILC rather than just an ambiguous risk factor for invasive breast cancer and that localized treatment for LCIS may be warranted given that these women have much higher rates of ipsilateral invasive breast cancer, but much more similar rates of contralateral breast cancer, compared with DCIS patients. Given that there are a growing number of women who have been diagnosed with DCIS and LCIS, continued investigations of these tumors are needed to better understand their etiology and to determine what treatments may be most effective for improving overall and disease-free survivals of these women.
- 12Lobular neoplasia (lobular carcinoma in situ). In: HaagensenCD, editor. Diseases of the Breast. Philadelphia, PA: WB Saunders; 1986: 192–241..
- 18Cancer incidence and mortality in the United States, 1973-77. Natl Cancer Inst Monogr. 1981: 1–187., , , et al.
- 19Regression models and life tables (with discussion). J R Stat Soc (B). 1972; 34: 187–220..
- 26African-American ethnicity, socioeconomic status, and breast cancer survival: a meta-analysis of 14 studies involving over 10,000 African-American and 40,000 White American patients with carcinoma of the breast. Cancer. 2002; 94: 2844–2854., , , et al.