Population-Based Validation of the Prognostic Model ADJUVANT! for Early Breast Cancer
http://www.100md.com
《临床肿瘤学》
the Breast Cancer Outcomes Unit, Radiation, Medical, and Population and Preventive Oncology Programs, the Vancouver Island, Vancouver, and Fraser Valley Centres of the British Columbia Cancer Agency
the University of British Columbia, Victoria, Vancouver, and Surrey, British Columbia, Canada
The University of Texas Health Sciences Center, San Antonio, TX
ABSTRACT
PURPOSE: Adjuvant! (www.adjuvantonline.com) is a web-based tool that predicts 10-year breast cancer outcomes with and without adjuvant systemic therapy, but it has not been independently validated.
METHODS: Using the British Columbia Breast Cancer Outcomes Unit (BCOU) database, demographic, pathologic, staging, and treatment data on 4,083 women diagnosed between 1989 and 1993 in British Columbia with T1-2, N0-1, M0 breast cancer were abstracted and entered into Adjuvant! to calculate predicted 10-year overall survival (OS), breast cancer–specific survival (BCSS), and event-free survival (EFS) for each patient. Individual BCOU observed outcomes at 10 years were independently determined. Predicted and observed outcomes were compared.
RESULTS: Across all 4,083 patients, 10-year predicted and observed outcomes were within 1% for OS, BCSS, and EFS (all P > .05). Predicted and observed outcomes were within 2% for most demographic, pathologic, and treatment-defined subgroups. Adjuvant! overestimated OS, BCSS, and EFS in women younger than age 35 years (predicted – observed = 8.6%, 9.6%, and 13.6%, respectively; all P < .001) or with lymphatic or vascular invasion (LVI; predicted – observed = 3.6%, 3.8%, and 4.2%, respectively; all P < .05); these two prognostic factors were not automatically incorporated within the Adjuvant! algorithm. After adjusting for the distribution of LVI, using the prognostic factor impact calculator in Adjuvant!, 10-year predicted and observed outcomes were no longer significantly different.
CONCLUSION: Adjuvant! performed reliably. Patients younger than age 35 or with known additional adverse prognostic factors such as LVI require adjustment of risks to derive reliable predictions of prognosis without adjuvant systemic therapy and the absolute benefits of adjuvant systemic therapy.
INTRODUCTION
Adjuvant treatment recommendations for patients with stage I or II breast cancer are based on the clinician's estimates of the likely risks of breast cancer relapse and death, and the likely benefit of adjuvant therapy. These estimates are currently based on stage of disease, pathologic characteristics of the tumor, and the efficacy of treatment.1,2 The estimated benefit of treatment may be expressed as either relative risk reductions or more usefully, as absolute numbers of patients likely to benefit per 100 treated. Predicting the absolute benefit of adjuvant systemic therapy for an individual woman with stage I or II breast cancer is complex and involves integration of information about underlying prognosis, efficacy of treatment options, and estimates of competing risk. Estimates of the benefit of chemotherapy and hormonal therapy influence a woman's willingness to accept these therapies.3,4 Several tools have been developed to make individualized estimates and to convey them to patients.5-9
One of the more complex and ambitious of these tools, Adjuvant! (www.adjuvantonline.com), can be used to estimate objectively the benefit of adjuvant systemic therapy for women with stage I or II breast cancer.5,7 The risk estimates in Adjuvant! were based on 10-year observed overall survival (OS) for women age 36 to 69 years diagnosed with breast cancer between 1988 and 1992 in the United States and recorded in the Surveillance, Epidemiology and End Results (SEER) registry. Breast cancer–specific survival (BCSS) for untreated patients was projected from observed BCSS in the SEER registry adjusted for the estimated frequency of systemic therapy use and the expected efficacy of that therapy. Ten-year estimates of event-free survival (EFS) were indirectly derived from SEER mortality data. Estimates of the efficacy of adjuvant tamoxifen and chemotherapy were derived from the Early Breast Cancer Trialists' Collaborative Group 1998 meta-analysis data.1,2 The efficacy of combined chemotherapy and endocrine therapy was based on the product of the individual risk reductions from hormonal therapy and chemotherapy alone.7 The validity of the assumptions inherent in Adjuvant! and its applicability to women beyond the range of ages used to develop the model, have not been independently evaluated. This is a report of a population-based assessment of the validity of Adjuvant! based on comparison of predicted OS, BCSS, and EFS estimates with observed outcomes for women with stage I or II breast cancer.
METHODS
Patients with invasive, pathologic stage I or II breast cancer10 diagnosed in British Columbia, Canada, between 1989 and 1993 were identified from the Breast Cancer Outcomes Unit (BCOU) database of the British Columbia Cancer Agency (BCCA). The BCCA has the mandate of cancer control for a geographically defined population of approximately four million, including developing cancer management guidelines, provision of all radiation therapy, and management of the budget for all antineoplastic drugs for the province (www.bccancer.bc.ca). During the interval of this study, approximately 75% of patients with newly diagnosed breast cancer were referred to the BCCA. Nonreferred patients were generally elderly or treated by mastectomy without indications for adjuvant therapy.
The BCOU database contains prospectively recorded demographic, pathologic, staging, initial treatment, and outcome information (first local, regional, and distant relapse, and if applicable, date and cause of death) for patients diagnosed with breast cancer in British Columbia since January 1, 1989, and who were referred to the BCCA. Outcome data are reported by the treating oncologist during clinical encounters at a BCCA facility, by annual letter contact with the patient's family physician, or monthly death clearance linkage between the British Columbia Cancer Registry and Department of Vital Statistics for the province of British Columbia. For this study, males (n = 21), women younger than 20 years (n = 1) or older than 85 years (n = 63); patients with incomplete local therapy (breast-conserving surgery [BCS] without radiation therapy [RT]; n = 169), unknown tumor or nodal status (n = 356), prior or synchronous contralateral breast cancer (n = 355); or those referred at relapse or with less than 10 years of follow-up (n = 360) were excluded.
Information abstracted from the BCOU database included age, sex, menopausal status, year of diagnosis, histology (ductal, lobular, or other), grade (1, 2, 3, or unknown), tumor size (in millimeters), number of involved axillary nodes, lymphatic or vascular invasion (LVI) status (present, absent, or unknown), estrogen receptor (ER) status (positive, negative, or unknown), type of local therapy (BCS alone, BCS plus RT, mastectomy, mastectomy plus RT), and type of adjuvant systemic therapy (none, tamoxifen alone, chemotherapy alone, or chemotherapy plus tamoxifen). Because the algorithms in Adjuvant! attributed different efficacy estimates depending on the type of chemotherapy used, adjuvant chemotherapy was characterized as four cycles of doxorubicin plus cyclophosphamide; 6 months of cyclophosphamide, methotrexate, and fluorouracil; or other chemotherapy.
The study end points were OS (the likelihood of being alive 10 years after the diagnosis of breast cancer); BCSS (the likelihood of not dying of breast cancer as the primary or underlying cause within 10 years of diagnosis); and EFS (the likelihood of surviving 10 years without local, regional, or distant recurrence, a contralateral new primary breast cancer, or death from breast cancer. Adjuvant! reports relapse-free survival but that end point includes the occurrence of contralateral breast cancers so the more correct term, EFS, was used in this study.
The 10-year predicted OS, BCSS, and EFS values were determined for each patient using Adjuvant! version 5.0, available in June 2003, by investigators (P.M.R. and G.D.), blinded to actual patient outcomes. Patient age, tumor size, number of positive nodes, grade, ER status, and adjuvant systemic therapy used were entered into the model for each patient, and 10-year OS, BCSS, and EFS values were calculated. The default comorbidity assumption of "minor health problems" was used.
Observed 10-year OS, BCSS, and EFS values for each patient were determined from the BCOU database. The predicted and observed 10-year outcomes were compared for the entire study cohort, for patients aggregated by 5% prognostic intervals and for clinically relevant subgroups. An a priori assumption was to consider Adjuvant! reliable enough for clinical use if the predicted and observed OS, BCSS, and EFS values were within 2%.
The presence of LVI was an important prognostic factor in the BCOU data11 and LVI was used to select patients for systemic therapy in BC (Table 1) but was not used automatically by the prediction algorithm in Adjuvant!. Using prognostic factor impact calculator (PFIC) predictions of OS in Adjuvant!, BCSS and EFS for patient subgroups receiving different adjuvant systemic therapies were adjusted for the impact (relative risk, 1.5) and distribution (36%) of the presence of LVI in the BCOU patients. Initial and adjusted predictions were compared with observed 10-year outcomes.
Adjuvant Therapy in 1989 to 1993
Table 1 summarizes the BCCA systemic therapy guidelines in effect during the majority of the study period.12,13 RT was recommended always after BCS. After mastectomy, RT was recommended if positive nodes were bulky or multiple (> 4 or > 50% nodes involved) or if there was skin, muscle, or margin involvement. Compliance in British Columbia with breast cancer management guidelines has been demonstrated to be high.13
This study was approved by the University of British Columbia BCCA Institutional Review Board.
Statistical Analysis
For the overall data set (n = 4,083) and various clinical subsets, observed percentages were calculated for each outcome. For the same data, the average Adjuvant! predicted values were calculated. Comparisons between the observed percentages and average predicted value assumed the later was constant. An SE was therefore only calculated for the observed percentage and a t test was used to compare the observed percentage and the average predicted value. This assumed that the observed percentage was an estimate with a normal distribution, which was reasonable because the percentage was always based on a sample of 50 or more patients. A P value less than .05 was used to define significance.
Separately for each outcome (OS, BCSS, and EFS), the data were divided into 5% intervals for the predicted values (ie, 0 to 4.99%, 5% to 9.99%, and so on). Intervals containing fewer than 50 observations were combined. Observed percentages were calculated for each interval subset and were plotted against the average predicted values. A straight line was fitted to the observed percentages and compared to a line with slope = 1 and intercept = 0. That line corresponded to perfect agreement between the observed and average predicted values. Tests of whether the fitted line differed from the predicted line were based on the estimated slope differing from 1.
RESULTS
Among 5,604 women with nonmetastatic, stage I to II breast cancer diagnosed in British Columbia and referred to the BCCA during the study period, 4,083 (73%) met eligibility criteria and were the study participants. Tables 2, 3, and 4 list the distribution of demographic, pathologic, staging, and initial treatment characteristics and the Adjuvant!-predicted and BCOU-observed rates of 10-year OS, BCSS, and EFS, respectively. For the entire cohort, the predicted and observed 10-year OS, BCSS, and EFS were within 1% and were not significantly different (P > .05 for each).
The average predicted outcomes of patient subgroups across the range of prognoses by 5% intervals are shown in Figure 1. In less than 1% of patients Adjuvant! predicted a 10-year OS of less than 25%, BCSS less than 35%, or EFS less than 25%. Adjuvant! predicted 10-year OS and BCSS well (Figs 1A and 1b). The slopes of a line representing a perfect match between predicted and observed outcomes and a line fitted to the observed outcomes were not significantly different (P > .05) for OS or BCSS. Event-free survival was overestimated by Adjuvant! when the EFS was low and underestimated by Adjuvant! when the EFS was high (P < .05 for slope; Fig 1C).
For most subgroups in Tables 2 to 4, the differences between the predicted and observed outcomes were within 2% or were not significantly different (P > .05).
Subgroups of patients notable for their exception to the general reliability of Adjuvant! included women age 20 to 35 years or older than 75 years, patients with LVI in the primary site, and patients who received combined chemotherapy and endocrine therapy. Women age 20 to 35 years had predicted outcomes that were significantly more favorable than observed by 8.6%, 9.6%, and 13.6% for OS, BCSS, and EFS, respectively. Women age 76 years had predicted and observed estimates that differed by more than 2.0% but the differences were only significant (P < .05) for BCSS.
Overall, 36% of patients had LVI present. Women with LVI present had significantly more favorable predicted than observed outcomes by 3.6%, 3.8%, and 4.2% for OS, BCSS, and EFS, respectively (all P < .05). In contrast, predicted outcomes were less favorable than observed outcomes (by 0.9% to 2.6%) for patients when LVI was absent. Predicted and observed outcomes were similar for patients with unknown LVI status. The presence of LVI was associated with a 1.5-fold increase in the risk of a breast cancer event or death from any cause and a 2.3-fold increase in the risk of a breast cancer death, compared with patients without LVI. Patient subgroups receiving adjuvant systemic therapy were enriched for LVI because it was an indication for treatment in the BCCA guidelines (Table 1). LVI was present in 53%, 63%, and 62% of patients receiving tamoxifen, chemotherapy, or combined chemotherapy and endocrine therapy, respectively. In contrast, only 10% of patients treated without adjuvant systemic therapy had LVI present.
Table 5 lists the 10-year BCSS for subgroups of patients characterized by stage, selected prognostic factors, and the use of adjuvant systemic therapy. Patients who received systemic adjuvant therapy had worse predicted and observed BCSS than those who did not because the BCCA guidelines recommended treatment for higher risk patients (Table 1). In the absence of adjuvant systemic therapy, the predicted BCSS was close to the observed BCSS for most subgroups. A notable exception was node-negative patients with an LVI-positive tumor, a prognostic factor not incorporated within Adjuvant!. The overall difference between predicted and observed BCSS among patients receiving adjuvant systemic therapy was 2.0% (P < .05; Table 5). The predicted BCSS was approximately 5% more favorable than observed for patients with T1N0, ER-positive, or grade 3 tumors, or node-positive disease who received adjuvant systemic therapy. These subgroups had LVI present in 56% to 63% of patients. In contrast, T1N0 patients not receiving systemic therapy had LVI present in just 6% of patients. When adjusted for the distribution and impact of LVI, the differences between predicted and observed BCSS for these subgroups were smaller and no longer significantly different. Similar patterns were observed for EFS and OS (data not shown).
The predicted OS, BCSS, and EFS corresponded closely to the respective observed outcomes for patients treated without adjuvant systemic therapy, or with tamoxifen alone or chemotherapy alone, as shown in Tables 2 to 4 (all P = not significant except OS without adjuvant systemic therapy). However, for patients who received both adjuvant hormonal and chemotherapy, the predicted OS, BCSS, and EFS were 5.2%, 4.6%, and 6.4% higher than the observed OS, BCSS, and EFS, respectively (P < .05 for OS and EFS).
The influence of LVI status on predicted outcomes among patients receiving different systemic therapies was investigated using the PFIC in Adjuvant!. The distribution of LVI-positive patients and an LVI relative risk impact of 1.5 were entered in the PFIC for each systemic treatment–defined subgroup for OS, BCSS, and EFS. Table 6 shows the 10-year BCSS Adjuvant! predictions adjusted for LVI. Predicted 10-year BCSS increased for patients treated without adjuvant systemic therapy and who had a lower than average frequency of LVI-positive tumors. In contrast, the BCSS predictions decreased for subgroups receiving systemic therapy, which were enriched with LVI-positive cases. Differences between the adjusted predictions of BCSS and the observed outcomes for each subgroup were not statistically significant (P > .05). Similar patterns of adjustment were seen for OS and EFS (data not shown).
DISCUSSION
This study has demonstrated that the predicted outcomes of Adjuvant! were valid with the exception of a few specific subgroups of patients. Overall, the Adjuvant!-predicted and BCOU-observed OS, BCSS, and EFS were within 1% of each other at 10 years. The predictions were close to observed outcomes over a wide range of prognoses, especially for OS and BCSS. Adjuvant!-predicted 10-year natural history of moderate risk, stage I and II breast cancer in the absence of adjuvant systemic therapy corresponded well to the observed outcomes in the population-based BCOU database of the BCCA. Therefore, clinicians can confidently use Adjuvant! to derive an objective estimate of the baseline risks of breast cancer recurrence and mortality in the absence of systemic therapy for an individual woman.
A strength of the current analysis is the use of population-based data set that was distinct from the SEER registry and US mortality data that were used as the basis for Adjuvant!. The close match between predicted and observed outcomes for untreated patients also can be interpreted as a validation of the population-based representativeness of the BCOU data or at least that the BCOU is comparable to the SEER registry data but with the added strength of having relapse information captured prospectively.
The present study identified two important subgroups of women with adverse prognostic factors not automatically accounted for in Adjuvant!. Adjuvant! predictions were more favorable than observed for patients age 20 to 35 years. Patients age 35 years and younger have higher locoregional and systemic recurrence risks and poorer survival than older patients,14-16 in particular for those with ER-positive tumors.17 Because this is an area of controversy, Adjuvant! directs the user to review a discussion on this topic in the help files and suggests that the user may want to consider the option of making an adjustment of 1.5-fold increased risk of poorer outcome for the ER-positive patients younger than 36 years. However, Adjuvant! did not automatically account for the effect of young age. A manual adjustment using the PFIC reduced the apparent discrepancy, but predictions by Adjuvant! were still overly optimistic compared with the observed outcomes for BCSS and EFS.
Reflecting these differences in outcome for young patients, some breast cancer management guidelines use age 35 years as an independent indication for systemic therapy among women with node-negative breast cancer.18,19 In the light of poorer OS, BCSS, and EFS observed in the BCOU data for women age 20 to 35 years, these recommendations seem justified. A new revision of Adjuvant! will incorporate young age as an independent and readily quantifiable adverse prognostic factor.
The presence of LVI in the primary site is a recognized prognostic factor14,19-21 but it is not widely used in North America. In British Columbia, the presence of LVI conferred a 1.75 relative risk of recurrence for patients with node-negative breast cancer and a 1.2 relative risk of recurrence for patients with node-positive disease.13,22 Prognostic factors such as LVI, young age, or HER-2 overexpression can be incorporated into the algorithm of Adjuvant! by using the optional PFIC. This Bayesian tool requires a priori estimates of the percentage of patients expressing the feature (36% had LVI present in the current study), and a relative risk for the high-risk compared with the low-risk group (approximately 1.5 with LVI present). Using the PFIC tool transformed the predicted 10-year BCSS from 87.2% and 77.2% for LVI-negative and LVI-positive disease, respectively, to 89.2% and 71.4%, which were closer to the observed 10-year BCSS values of 88.4% and 73.4% for LVI-negative and LVI-positive patient subgroups.
Similar adjustments for the impact of LVI on the predicted outcomes for patients receiving different adjuvant systemic therapies resulted generally in a reduction in the differences between predicted and observed outcomes among the various systemic treatment subgroups to 2% or less or not significantly different from each other. Adjuvant! was consistently more optimistic for OS, BCSS, and EFS by approximately 5% absolute, in comparison with observed outcomes in British Columbia for patients treated with both tamoxifen and chemotherapy. However, after adjusting for the influence of LVI status, a factor that increased the risk of relapse and was also used as a selection factor for adjuvant systemic therapy, there were no significant differences between predicted and observed outcomes. By incorporating the ability to adjust predictions for factors such as LVI, estimates by Adjuvant! of the absolute benefit to be derived from adjuvant systemic therapies alone or in combination can also be considered to be clinically useful.
The estimates made by Adjuvant! for histologies other than ductal and lobular were on average too optimistic. This was due to medullary cancer patients having a better outcome than would have been suggested by their high histologic grade (data not shown), as has been recognized previously.23 In this validation, Adjuvant! was running in a default mode and did not take into account special histologic subtypes. In the opening sequence of Adjuvant! the user is alerted to additional information about special subtypes in the help files. The discussion in the help files suggests that for medullary cancer, estimates of breast cancer–specific events should be adjusted by a factor of 0.5. This adjustment would have reduced the apparent discrepancy.
In the era of this study, adjuvant chemotherapy was largely doxorubicin plus cyclophosphamide for four cycles or cyclophosphamide, methotrexate, and fluorouracil for 6 months. Validation of the ability of Adjuvant! to estimate the impact of current, more intensive chemotherapy, taxane-containing regimens, and the adjuvant use of aromatase inhibitors, is needed. In addition, it is not clear how nodal status as assessed by sentinel node mapping and biopsy should be incorporated into Adjuvant! Elucidation of such issues will require long-term follow-up of patients in population-based cohorts receiving current therapy.
Adjuvant! makes reliable predictions for the average patient and has extensive help files that describe the nuances and cautions about its use for particular subgroups of patients. In addition, it is necessary to use appropriately the PFIC to adjust estimates for important prognostic factors such as the presence of LVI or HER-2 overexpression that may be available in the pathology report. When adjusting for a prognostic factor not presently considered by Adjuvant!, the user needs to enter representative relative risk and prevalence values into the PFIC to use this tool correctly.
In summary, Adjuvant! is a useful tool to aid physicians and patients in making adjuvant systemic treatment decisions for early-stage breast cancer. Adjuvant! can be used confidently to provide estimates of 10-year EFS, BCSS, and OS risks without adjuvant systemic therapy. The predictions of the absolute benefits of adjuvant systemic therapy were generally reliable but were overly optimistic for subgroups enriched with adverse prognostic factors not accounted for by the model. An optional PFIC adjusts estimates by an appropriate magnitude and in the correct direction. The PFIC should be used manually to adjust the prognostic estimates for patients age 35 years, with LVI present in the primary tumor, and potentially for other factors for which the relative risk and frequency of occurrence in the population is known. In addition, it is clear that Adjuvant! should be used as an aid to decision making and not as a replacement for good clinical judgment.
Authors' Disclosures of Potential Conflicts of Interest
The following authors or their immediate family members have indicated a financial interest. No conflict exists for drugs or devices used in a study if they are not being evaluated as part of the investigation. Leadership Position: Peter M. Ravdin, Adjuvant Inc; Greg J. Davis, Adjuvant Inc. Consultant/Advisory Role: Peter M. Ravdin, Adjuvant Inc; Brian D. Norris, AstraZeneca, Pfizer, Roche; Greg J. Davis, Adjuvant Inc. Stock Ownership: Peter M. Ravdin, Adjuvant Inc. Honoraria: Brian D. Norris, AstraZeneca, Pfizer, Roche. Research Funding: Brian D. Norris, Roche. For a detailed description of these categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and Disclosures of Potential Conflicts of Interest found in Information for Contributors in the front of each issue.
Acknowledgment
We thank Lisa Kelly for assistance in preparation of the manuscript and Boon Chua, MD, for helpful comments on an early draft.
NOTES
Supported by grant No. 014452 from the Canadian Breast Cancer Research Alliance. Chris Bajdik, PhD, is supported by a scholar award from the Michael Smith Foundation for Health Research.
Presented in part at the 40th Annual Meeting of the American Society of Clinical Oncology, New Orleans, LA, June 5-8, 2004.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Early Breast Cancer Trialists' Collaborative Group: Tamoxifen for early breast cancer: An overview of randomized trials. Lancet 351:1451-1467, 1998
Early Breast Cancer Trialists' Collaborative Group: Polychemotherapy for early breast cancer: An overview of the randomized trials. Lancet 352:930-942, 1998
Ravdin PM, Siminoff LA, Harvey JA: Survey of breast cancer patients concerning their knowledge and expectations of adjuvant therapy. J Clin Oncol 16:515-521, 1998
Siminoff LA, Fetting JH, Abeloff MD: Doctor-patient communication about breast cancer adjuvant therapy. J Clin Oncol 7:1192-1200, 1989
Ravdin PM: A computer program to assist in making breast cancer adjuvant therapy decisions. Semin Oncol 23:43-50, 1996 (suppl 2)
Loprinzi CL, Thorne SD: Understanding the utility of adjuvant systemic therapy for primary breast cancer. J Clin Oncol 19:972-979, 2001
Ravdin PM, Siminoff LA, Davis GJ, et al: Computer program to assist in making decisions about adjuvant therapy for women with early breast cancer. J Clin Oncol 19:980-991, 2001
Lundin J, Lundin M, Isola J, et al: A web-based system for individualized survival estimation in breast cancer. BMJ 326:29, 2003
Whelan T, Sawka C, Levine M, et al: Helping patients make informed choices: A randomized trial of decision aid for adjuvant chemotherapy in lymph node-negative breast cancer. J Natl Cancer Inst 95:581-587, 2003
Fleming ID, Cooper JS, Henson DE, et al: (eds): AJCC Cancer Staging Manual (ed 5). Philadelphia, PA, Lippincott-Raven, 1997
Ragaz J, Spinelli J, Speers C, et al: Adverse impact of lymphangitic and vascular invasion in early breast cancer: Results from the British Columbia breast cancer outcomes unit. Breast Cancer Res Treat 76:S42, 2002 (suppl 1; abstr 118)
Sawka C, Olivotto I, Coldman A, et al: The association between population-based treatment guidelines and adjuvant therapy for node-negative breast cancer. Br J Cancer 75:1534-1542, 1997
Olivotto A, Coldman AJ, Hislop TG, et al: Compliance with practice guidelines for node-negative breast cancer. J Clin Oncol 15:216-222, 1997
Fourquet A, Campana F, Zafrani B, et al: Prognostic factors of breast recurrence in the conservative management of early breast cancer: A 25-year followup. Int J Radiat Oncol Biol Phys 17:719-725, 1989
Nixon AJ, Newberg D, Hayes DF, et al: Relationship of age to pathologic factors of the tumor and prognosis for patients with stage I or II breast cancer. J Clin Oncol 12:888-894, 1994
Rosner D, Lane WW: Predicting recurrence in axillary node-negative breast cancer patients. Breast Cancer Res Treat 25:127-139, 1993
Aebi S, Gelber S, Castiglione-Gertsch M, et al: Is chemotherapy alone adequate therapy for young women with oestrogen-receptor positive breast cancer Lancet 355:1869-1874, 2000
Senn HJ, Thurlimann B, Goldhirsch A, et al: Comments on the St. Gallen Consensus 2003 on the primary therapy of early breast cancer. Breast 12:569-582, 2003
Levine M: Clinical practice guidelines for the care and treatment of breast cancer: Adjuvant systemic therapy for node-negative breast cancer (summary of the 2001 update). CMAJ 164:213, 2001
Lee AK, Delellis RA, Silverman ML, et al: Prognostic significance of peritumoral lymphatic and blood vessel invasion in node-negative carcinoma of the breast. J Clin Oncol 8:1457-1465, 1990
Gasparini G, Weidner N, Bevilacqua P, et al: Tumor microvessel density, p53 expression, tumor size and peritumoral lymphatic vessel invasion are relevant prognostic markers in node-negative breast carcinoma. J Clin Oncol 12:454-466, 1994
Weir L, Speers C, D'yachkova Y, et al: Prognostic significance of the number of axillary lymph nodes reviewed in patients with node-negative breast cancer. J Clin Oncol 20:1793-1799, 2002
Pedersen L, Zedeler K, Holck S, et al: Medullary carcinoma of the breast: Prevalence and prognostic importance of classical risk factors in breast cancer. Eur J Cancer 31:2289-2295, 1995(Ivo A. Olivotto, Chris D.)
the University of British Columbia, Victoria, Vancouver, and Surrey, British Columbia, Canada
The University of Texas Health Sciences Center, San Antonio, TX
ABSTRACT
PURPOSE: Adjuvant! (www.adjuvantonline.com) is a web-based tool that predicts 10-year breast cancer outcomes with and without adjuvant systemic therapy, but it has not been independently validated.
METHODS: Using the British Columbia Breast Cancer Outcomes Unit (BCOU) database, demographic, pathologic, staging, and treatment data on 4,083 women diagnosed between 1989 and 1993 in British Columbia with T1-2, N0-1, M0 breast cancer were abstracted and entered into Adjuvant! to calculate predicted 10-year overall survival (OS), breast cancer–specific survival (BCSS), and event-free survival (EFS) for each patient. Individual BCOU observed outcomes at 10 years were independently determined. Predicted and observed outcomes were compared.
RESULTS: Across all 4,083 patients, 10-year predicted and observed outcomes were within 1% for OS, BCSS, and EFS (all P > .05). Predicted and observed outcomes were within 2% for most demographic, pathologic, and treatment-defined subgroups. Adjuvant! overestimated OS, BCSS, and EFS in women younger than age 35 years (predicted – observed = 8.6%, 9.6%, and 13.6%, respectively; all P < .001) or with lymphatic or vascular invasion (LVI; predicted – observed = 3.6%, 3.8%, and 4.2%, respectively; all P < .05); these two prognostic factors were not automatically incorporated within the Adjuvant! algorithm. After adjusting for the distribution of LVI, using the prognostic factor impact calculator in Adjuvant!, 10-year predicted and observed outcomes were no longer significantly different.
CONCLUSION: Adjuvant! performed reliably. Patients younger than age 35 or with known additional adverse prognostic factors such as LVI require adjustment of risks to derive reliable predictions of prognosis without adjuvant systemic therapy and the absolute benefits of adjuvant systemic therapy.
INTRODUCTION
Adjuvant treatment recommendations for patients with stage I or II breast cancer are based on the clinician's estimates of the likely risks of breast cancer relapse and death, and the likely benefit of adjuvant therapy. These estimates are currently based on stage of disease, pathologic characteristics of the tumor, and the efficacy of treatment.1,2 The estimated benefit of treatment may be expressed as either relative risk reductions or more usefully, as absolute numbers of patients likely to benefit per 100 treated. Predicting the absolute benefit of adjuvant systemic therapy for an individual woman with stage I or II breast cancer is complex and involves integration of information about underlying prognosis, efficacy of treatment options, and estimates of competing risk. Estimates of the benefit of chemotherapy and hormonal therapy influence a woman's willingness to accept these therapies.3,4 Several tools have been developed to make individualized estimates and to convey them to patients.5-9
One of the more complex and ambitious of these tools, Adjuvant! (www.adjuvantonline.com), can be used to estimate objectively the benefit of adjuvant systemic therapy for women with stage I or II breast cancer.5,7 The risk estimates in Adjuvant! were based on 10-year observed overall survival (OS) for women age 36 to 69 years diagnosed with breast cancer between 1988 and 1992 in the United States and recorded in the Surveillance, Epidemiology and End Results (SEER) registry. Breast cancer–specific survival (BCSS) for untreated patients was projected from observed BCSS in the SEER registry adjusted for the estimated frequency of systemic therapy use and the expected efficacy of that therapy. Ten-year estimates of event-free survival (EFS) were indirectly derived from SEER mortality data. Estimates of the efficacy of adjuvant tamoxifen and chemotherapy were derived from the Early Breast Cancer Trialists' Collaborative Group 1998 meta-analysis data.1,2 The efficacy of combined chemotherapy and endocrine therapy was based on the product of the individual risk reductions from hormonal therapy and chemotherapy alone.7 The validity of the assumptions inherent in Adjuvant! and its applicability to women beyond the range of ages used to develop the model, have not been independently evaluated. This is a report of a population-based assessment of the validity of Adjuvant! based on comparison of predicted OS, BCSS, and EFS estimates with observed outcomes for women with stage I or II breast cancer.
METHODS
Patients with invasive, pathologic stage I or II breast cancer10 diagnosed in British Columbia, Canada, between 1989 and 1993 were identified from the Breast Cancer Outcomes Unit (BCOU) database of the British Columbia Cancer Agency (BCCA). The BCCA has the mandate of cancer control for a geographically defined population of approximately four million, including developing cancer management guidelines, provision of all radiation therapy, and management of the budget for all antineoplastic drugs for the province (www.bccancer.bc.ca). During the interval of this study, approximately 75% of patients with newly diagnosed breast cancer were referred to the BCCA. Nonreferred patients were generally elderly or treated by mastectomy without indications for adjuvant therapy.
The BCOU database contains prospectively recorded demographic, pathologic, staging, initial treatment, and outcome information (first local, regional, and distant relapse, and if applicable, date and cause of death) for patients diagnosed with breast cancer in British Columbia since January 1, 1989, and who were referred to the BCCA. Outcome data are reported by the treating oncologist during clinical encounters at a BCCA facility, by annual letter contact with the patient's family physician, or monthly death clearance linkage between the British Columbia Cancer Registry and Department of Vital Statistics for the province of British Columbia. For this study, males (n = 21), women younger than 20 years (n = 1) or older than 85 years (n = 63); patients with incomplete local therapy (breast-conserving surgery [BCS] without radiation therapy [RT]; n = 169), unknown tumor or nodal status (n = 356), prior or synchronous contralateral breast cancer (n = 355); or those referred at relapse or with less than 10 years of follow-up (n = 360) were excluded.
Information abstracted from the BCOU database included age, sex, menopausal status, year of diagnosis, histology (ductal, lobular, or other), grade (1, 2, 3, or unknown), tumor size (in millimeters), number of involved axillary nodes, lymphatic or vascular invasion (LVI) status (present, absent, or unknown), estrogen receptor (ER) status (positive, negative, or unknown), type of local therapy (BCS alone, BCS plus RT, mastectomy, mastectomy plus RT), and type of adjuvant systemic therapy (none, tamoxifen alone, chemotherapy alone, or chemotherapy plus tamoxifen). Because the algorithms in Adjuvant! attributed different efficacy estimates depending on the type of chemotherapy used, adjuvant chemotherapy was characterized as four cycles of doxorubicin plus cyclophosphamide; 6 months of cyclophosphamide, methotrexate, and fluorouracil; or other chemotherapy.
The study end points were OS (the likelihood of being alive 10 years after the diagnosis of breast cancer); BCSS (the likelihood of not dying of breast cancer as the primary or underlying cause within 10 years of diagnosis); and EFS (the likelihood of surviving 10 years without local, regional, or distant recurrence, a contralateral new primary breast cancer, or death from breast cancer. Adjuvant! reports relapse-free survival but that end point includes the occurrence of contralateral breast cancers so the more correct term, EFS, was used in this study.
The 10-year predicted OS, BCSS, and EFS values were determined for each patient using Adjuvant! version 5.0, available in June 2003, by investigators (P.M.R. and G.D.), blinded to actual patient outcomes. Patient age, tumor size, number of positive nodes, grade, ER status, and adjuvant systemic therapy used were entered into the model for each patient, and 10-year OS, BCSS, and EFS values were calculated. The default comorbidity assumption of "minor health problems" was used.
Observed 10-year OS, BCSS, and EFS values for each patient were determined from the BCOU database. The predicted and observed 10-year outcomes were compared for the entire study cohort, for patients aggregated by 5% prognostic intervals and for clinically relevant subgroups. An a priori assumption was to consider Adjuvant! reliable enough for clinical use if the predicted and observed OS, BCSS, and EFS values were within 2%.
The presence of LVI was an important prognostic factor in the BCOU data11 and LVI was used to select patients for systemic therapy in BC (Table 1) but was not used automatically by the prediction algorithm in Adjuvant!. Using prognostic factor impact calculator (PFIC) predictions of OS in Adjuvant!, BCSS and EFS for patient subgroups receiving different adjuvant systemic therapies were adjusted for the impact (relative risk, 1.5) and distribution (36%) of the presence of LVI in the BCOU patients. Initial and adjusted predictions were compared with observed 10-year outcomes.
Adjuvant Therapy in 1989 to 1993
Table 1 summarizes the BCCA systemic therapy guidelines in effect during the majority of the study period.12,13 RT was recommended always after BCS. After mastectomy, RT was recommended if positive nodes were bulky or multiple (> 4 or > 50% nodes involved) or if there was skin, muscle, or margin involvement. Compliance in British Columbia with breast cancer management guidelines has been demonstrated to be high.13
This study was approved by the University of British Columbia BCCA Institutional Review Board.
Statistical Analysis
For the overall data set (n = 4,083) and various clinical subsets, observed percentages were calculated for each outcome. For the same data, the average Adjuvant! predicted values were calculated. Comparisons between the observed percentages and average predicted value assumed the later was constant. An SE was therefore only calculated for the observed percentage and a t test was used to compare the observed percentage and the average predicted value. This assumed that the observed percentage was an estimate with a normal distribution, which was reasonable because the percentage was always based on a sample of 50 or more patients. A P value less than .05 was used to define significance.
Separately for each outcome (OS, BCSS, and EFS), the data were divided into 5% intervals for the predicted values (ie, 0 to 4.99%, 5% to 9.99%, and so on). Intervals containing fewer than 50 observations were combined. Observed percentages were calculated for each interval subset and were plotted against the average predicted values. A straight line was fitted to the observed percentages and compared to a line with slope = 1 and intercept = 0. That line corresponded to perfect agreement between the observed and average predicted values. Tests of whether the fitted line differed from the predicted line were based on the estimated slope differing from 1.
RESULTS
Among 5,604 women with nonmetastatic, stage I to II breast cancer diagnosed in British Columbia and referred to the BCCA during the study period, 4,083 (73%) met eligibility criteria and were the study participants. Tables 2, 3, and 4 list the distribution of demographic, pathologic, staging, and initial treatment characteristics and the Adjuvant!-predicted and BCOU-observed rates of 10-year OS, BCSS, and EFS, respectively. For the entire cohort, the predicted and observed 10-year OS, BCSS, and EFS were within 1% and were not significantly different (P > .05 for each).
The average predicted outcomes of patient subgroups across the range of prognoses by 5% intervals are shown in Figure 1. In less than 1% of patients Adjuvant! predicted a 10-year OS of less than 25%, BCSS less than 35%, or EFS less than 25%. Adjuvant! predicted 10-year OS and BCSS well (Figs 1A and 1b). The slopes of a line representing a perfect match between predicted and observed outcomes and a line fitted to the observed outcomes were not significantly different (P > .05) for OS or BCSS. Event-free survival was overestimated by Adjuvant! when the EFS was low and underestimated by Adjuvant! when the EFS was high (P < .05 for slope; Fig 1C).
For most subgroups in Tables 2 to 4, the differences between the predicted and observed outcomes were within 2% or were not significantly different (P > .05).
Subgroups of patients notable for their exception to the general reliability of Adjuvant! included women age 20 to 35 years or older than 75 years, patients with LVI in the primary site, and patients who received combined chemotherapy and endocrine therapy. Women age 20 to 35 years had predicted outcomes that were significantly more favorable than observed by 8.6%, 9.6%, and 13.6% for OS, BCSS, and EFS, respectively. Women age 76 years had predicted and observed estimates that differed by more than 2.0% but the differences were only significant (P < .05) for BCSS.
Overall, 36% of patients had LVI present. Women with LVI present had significantly more favorable predicted than observed outcomes by 3.6%, 3.8%, and 4.2% for OS, BCSS, and EFS, respectively (all P < .05). In contrast, predicted outcomes were less favorable than observed outcomes (by 0.9% to 2.6%) for patients when LVI was absent. Predicted and observed outcomes were similar for patients with unknown LVI status. The presence of LVI was associated with a 1.5-fold increase in the risk of a breast cancer event or death from any cause and a 2.3-fold increase in the risk of a breast cancer death, compared with patients without LVI. Patient subgroups receiving adjuvant systemic therapy were enriched for LVI because it was an indication for treatment in the BCCA guidelines (Table 1). LVI was present in 53%, 63%, and 62% of patients receiving tamoxifen, chemotherapy, or combined chemotherapy and endocrine therapy, respectively. In contrast, only 10% of patients treated without adjuvant systemic therapy had LVI present.
Table 5 lists the 10-year BCSS for subgroups of patients characterized by stage, selected prognostic factors, and the use of adjuvant systemic therapy. Patients who received systemic adjuvant therapy had worse predicted and observed BCSS than those who did not because the BCCA guidelines recommended treatment for higher risk patients (Table 1). In the absence of adjuvant systemic therapy, the predicted BCSS was close to the observed BCSS for most subgroups. A notable exception was node-negative patients with an LVI-positive tumor, a prognostic factor not incorporated within Adjuvant!. The overall difference between predicted and observed BCSS among patients receiving adjuvant systemic therapy was 2.0% (P < .05; Table 5). The predicted BCSS was approximately 5% more favorable than observed for patients with T1N0, ER-positive, or grade 3 tumors, or node-positive disease who received adjuvant systemic therapy. These subgroups had LVI present in 56% to 63% of patients. In contrast, T1N0 patients not receiving systemic therapy had LVI present in just 6% of patients. When adjusted for the distribution and impact of LVI, the differences between predicted and observed BCSS for these subgroups were smaller and no longer significantly different. Similar patterns were observed for EFS and OS (data not shown).
The predicted OS, BCSS, and EFS corresponded closely to the respective observed outcomes for patients treated without adjuvant systemic therapy, or with tamoxifen alone or chemotherapy alone, as shown in Tables 2 to 4 (all P = not significant except OS without adjuvant systemic therapy). However, for patients who received both adjuvant hormonal and chemotherapy, the predicted OS, BCSS, and EFS were 5.2%, 4.6%, and 6.4% higher than the observed OS, BCSS, and EFS, respectively (P < .05 for OS and EFS).
The influence of LVI status on predicted outcomes among patients receiving different systemic therapies was investigated using the PFIC in Adjuvant!. The distribution of LVI-positive patients and an LVI relative risk impact of 1.5 were entered in the PFIC for each systemic treatment–defined subgroup for OS, BCSS, and EFS. Table 6 shows the 10-year BCSS Adjuvant! predictions adjusted for LVI. Predicted 10-year BCSS increased for patients treated without adjuvant systemic therapy and who had a lower than average frequency of LVI-positive tumors. In contrast, the BCSS predictions decreased for subgroups receiving systemic therapy, which were enriched with LVI-positive cases. Differences between the adjusted predictions of BCSS and the observed outcomes for each subgroup were not statistically significant (P > .05). Similar patterns of adjustment were seen for OS and EFS (data not shown).
DISCUSSION
This study has demonstrated that the predicted outcomes of Adjuvant! were valid with the exception of a few specific subgroups of patients. Overall, the Adjuvant!-predicted and BCOU-observed OS, BCSS, and EFS were within 1% of each other at 10 years. The predictions were close to observed outcomes over a wide range of prognoses, especially for OS and BCSS. Adjuvant!-predicted 10-year natural history of moderate risk, stage I and II breast cancer in the absence of adjuvant systemic therapy corresponded well to the observed outcomes in the population-based BCOU database of the BCCA. Therefore, clinicians can confidently use Adjuvant! to derive an objective estimate of the baseline risks of breast cancer recurrence and mortality in the absence of systemic therapy for an individual woman.
A strength of the current analysis is the use of population-based data set that was distinct from the SEER registry and US mortality data that were used as the basis for Adjuvant!. The close match between predicted and observed outcomes for untreated patients also can be interpreted as a validation of the population-based representativeness of the BCOU data or at least that the BCOU is comparable to the SEER registry data but with the added strength of having relapse information captured prospectively.
The present study identified two important subgroups of women with adverse prognostic factors not automatically accounted for in Adjuvant!. Adjuvant! predictions were more favorable than observed for patients age 20 to 35 years. Patients age 35 years and younger have higher locoregional and systemic recurrence risks and poorer survival than older patients,14-16 in particular for those with ER-positive tumors.17 Because this is an area of controversy, Adjuvant! directs the user to review a discussion on this topic in the help files and suggests that the user may want to consider the option of making an adjustment of 1.5-fold increased risk of poorer outcome for the ER-positive patients younger than 36 years. However, Adjuvant! did not automatically account for the effect of young age. A manual adjustment using the PFIC reduced the apparent discrepancy, but predictions by Adjuvant! were still overly optimistic compared with the observed outcomes for BCSS and EFS.
Reflecting these differences in outcome for young patients, some breast cancer management guidelines use age 35 years as an independent indication for systemic therapy among women with node-negative breast cancer.18,19 In the light of poorer OS, BCSS, and EFS observed in the BCOU data for women age 20 to 35 years, these recommendations seem justified. A new revision of Adjuvant! will incorporate young age as an independent and readily quantifiable adverse prognostic factor.
The presence of LVI in the primary site is a recognized prognostic factor14,19-21 but it is not widely used in North America. In British Columbia, the presence of LVI conferred a 1.75 relative risk of recurrence for patients with node-negative breast cancer and a 1.2 relative risk of recurrence for patients with node-positive disease.13,22 Prognostic factors such as LVI, young age, or HER-2 overexpression can be incorporated into the algorithm of Adjuvant! by using the optional PFIC. This Bayesian tool requires a priori estimates of the percentage of patients expressing the feature (36% had LVI present in the current study), and a relative risk for the high-risk compared with the low-risk group (approximately 1.5 with LVI present). Using the PFIC tool transformed the predicted 10-year BCSS from 87.2% and 77.2% for LVI-negative and LVI-positive disease, respectively, to 89.2% and 71.4%, which were closer to the observed 10-year BCSS values of 88.4% and 73.4% for LVI-negative and LVI-positive patient subgroups.
Similar adjustments for the impact of LVI on the predicted outcomes for patients receiving different adjuvant systemic therapies resulted generally in a reduction in the differences between predicted and observed outcomes among the various systemic treatment subgroups to 2% or less or not significantly different from each other. Adjuvant! was consistently more optimistic for OS, BCSS, and EFS by approximately 5% absolute, in comparison with observed outcomes in British Columbia for patients treated with both tamoxifen and chemotherapy. However, after adjusting for the influence of LVI status, a factor that increased the risk of relapse and was also used as a selection factor for adjuvant systemic therapy, there were no significant differences between predicted and observed outcomes. By incorporating the ability to adjust predictions for factors such as LVI, estimates by Adjuvant! of the absolute benefit to be derived from adjuvant systemic therapies alone or in combination can also be considered to be clinically useful.
The estimates made by Adjuvant! for histologies other than ductal and lobular were on average too optimistic. This was due to medullary cancer patients having a better outcome than would have been suggested by their high histologic grade (data not shown), as has been recognized previously.23 In this validation, Adjuvant! was running in a default mode and did not take into account special histologic subtypes. In the opening sequence of Adjuvant! the user is alerted to additional information about special subtypes in the help files. The discussion in the help files suggests that for medullary cancer, estimates of breast cancer–specific events should be adjusted by a factor of 0.5. This adjustment would have reduced the apparent discrepancy.
In the era of this study, adjuvant chemotherapy was largely doxorubicin plus cyclophosphamide for four cycles or cyclophosphamide, methotrexate, and fluorouracil for 6 months. Validation of the ability of Adjuvant! to estimate the impact of current, more intensive chemotherapy, taxane-containing regimens, and the adjuvant use of aromatase inhibitors, is needed. In addition, it is not clear how nodal status as assessed by sentinel node mapping and biopsy should be incorporated into Adjuvant! Elucidation of such issues will require long-term follow-up of patients in population-based cohorts receiving current therapy.
Adjuvant! makes reliable predictions for the average patient and has extensive help files that describe the nuances and cautions about its use for particular subgroups of patients. In addition, it is necessary to use appropriately the PFIC to adjust estimates for important prognostic factors such as the presence of LVI or HER-2 overexpression that may be available in the pathology report. When adjusting for a prognostic factor not presently considered by Adjuvant!, the user needs to enter representative relative risk and prevalence values into the PFIC to use this tool correctly.
In summary, Adjuvant! is a useful tool to aid physicians and patients in making adjuvant systemic treatment decisions for early-stage breast cancer. Adjuvant! can be used confidently to provide estimates of 10-year EFS, BCSS, and OS risks without adjuvant systemic therapy. The predictions of the absolute benefits of adjuvant systemic therapy were generally reliable but were overly optimistic for subgroups enriched with adverse prognostic factors not accounted for by the model. An optional PFIC adjusts estimates by an appropriate magnitude and in the correct direction. The PFIC should be used manually to adjust the prognostic estimates for patients age 35 years, with LVI present in the primary tumor, and potentially for other factors for which the relative risk and frequency of occurrence in the population is known. In addition, it is clear that Adjuvant! should be used as an aid to decision making and not as a replacement for good clinical judgment.
Authors' Disclosures of Potential Conflicts of Interest
The following authors or their immediate family members have indicated a financial interest. No conflict exists for drugs or devices used in a study if they are not being evaluated as part of the investigation. Leadership Position: Peter M. Ravdin, Adjuvant Inc; Greg J. Davis, Adjuvant Inc. Consultant/Advisory Role: Peter M. Ravdin, Adjuvant Inc; Brian D. Norris, AstraZeneca, Pfizer, Roche; Greg J. Davis, Adjuvant Inc. Stock Ownership: Peter M. Ravdin, Adjuvant Inc. Honoraria: Brian D. Norris, AstraZeneca, Pfizer, Roche. Research Funding: Brian D. Norris, Roche. For a detailed description of these categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and Disclosures of Potential Conflicts of Interest found in Information for Contributors in the front of each issue.
Acknowledgment
We thank Lisa Kelly for assistance in preparation of the manuscript and Boon Chua, MD, for helpful comments on an early draft.
NOTES
Supported by grant No. 014452 from the Canadian Breast Cancer Research Alliance. Chris Bajdik, PhD, is supported by a scholar award from the Michael Smith Foundation for Health Research.
Presented in part at the 40th Annual Meeting of the American Society of Clinical Oncology, New Orleans, LA, June 5-8, 2004.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Early Breast Cancer Trialists' Collaborative Group: Tamoxifen for early breast cancer: An overview of randomized trials. Lancet 351:1451-1467, 1998
Early Breast Cancer Trialists' Collaborative Group: Polychemotherapy for early breast cancer: An overview of the randomized trials. Lancet 352:930-942, 1998
Ravdin PM, Siminoff LA, Harvey JA: Survey of breast cancer patients concerning their knowledge and expectations of adjuvant therapy. J Clin Oncol 16:515-521, 1998
Siminoff LA, Fetting JH, Abeloff MD: Doctor-patient communication about breast cancer adjuvant therapy. J Clin Oncol 7:1192-1200, 1989
Ravdin PM: A computer program to assist in making breast cancer adjuvant therapy decisions. Semin Oncol 23:43-50, 1996 (suppl 2)
Loprinzi CL, Thorne SD: Understanding the utility of adjuvant systemic therapy for primary breast cancer. J Clin Oncol 19:972-979, 2001
Ravdin PM, Siminoff LA, Davis GJ, et al: Computer program to assist in making decisions about adjuvant therapy for women with early breast cancer. J Clin Oncol 19:980-991, 2001
Lundin J, Lundin M, Isola J, et al: A web-based system for individualized survival estimation in breast cancer. BMJ 326:29, 2003
Whelan T, Sawka C, Levine M, et al: Helping patients make informed choices: A randomized trial of decision aid for adjuvant chemotherapy in lymph node-negative breast cancer. J Natl Cancer Inst 95:581-587, 2003
Fleming ID, Cooper JS, Henson DE, et al: (eds): AJCC Cancer Staging Manual (ed 5). Philadelphia, PA, Lippincott-Raven, 1997
Ragaz J, Spinelli J, Speers C, et al: Adverse impact of lymphangitic and vascular invasion in early breast cancer: Results from the British Columbia breast cancer outcomes unit. Breast Cancer Res Treat 76:S42, 2002 (suppl 1; abstr 118)
Sawka C, Olivotto I, Coldman A, et al: The association between population-based treatment guidelines and adjuvant therapy for node-negative breast cancer. Br J Cancer 75:1534-1542, 1997
Olivotto A, Coldman AJ, Hislop TG, et al: Compliance with practice guidelines for node-negative breast cancer. J Clin Oncol 15:216-222, 1997
Fourquet A, Campana F, Zafrani B, et al: Prognostic factors of breast recurrence in the conservative management of early breast cancer: A 25-year followup. Int J Radiat Oncol Biol Phys 17:719-725, 1989
Nixon AJ, Newberg D, Hayes DF, et al: Relationship of age to pathologic factors of the tumor and prognosis for patients with stage I or II breast cancer. J Clin Oncol 12:888-894, 1994
Rosner D, Lane WW: Predicting recurrence in axillary node-negative breast cancer patients. Breast Cancer Res Treat 25:127-139, 1993
Aebi S, Gelber S, Castiglione-Gertsch M, et al: Is chemotherapy alone adequate therapy for young women with oestrogen-receptor positive breast cancer Lancet 355:1869-1874, 2000
Senn HJ, Thurlimann B, Goldhirsch A, et al: Comments on the St. Gallen Consensus 2003 on the primary therapy of early breast cancer. Breast 12:569-582, 2003
Levine M: Clinical practice guidelines for the care and treatment of breast cancer: Adjuvant systemic therapy for node-negative breast cancer (summary of the 2001 update). CMAJ 164:213, 2001
Lee AK, Delellis RA, Silverman ML, et al: Prognostic significance of peritumoral lymphatic and blood vessel invasion in node-negative carcinoma of the breast. J Clin Oncol 8:1457-1465, 1990
Gasparini G, Weidner N, Bevilacqua P, et al: Tumor microvessel density, p53 expression, tumor size and peritumoral lymphatic vessel invasion are relevant prognostic markers in node-negative breast carcinoma. J Clin Oncol 12:454-466, 1994
Weir L, Speers C, D'yachkova Y, et al: Prognostic significance of the number of axillary lymph nodes reviewed in patients with node-negative breast cancer. J Clin Oncol 20:1793-1799, 2002
Pedersen L, Zedeler K, Holck S, et al: Medullary carcinoma of the breast: Prevalence and prognostic importance of classical risk factors in breast cancer. Eur J Cancer 31:2289-2295, 1995(Ivo A. Olivotto, Chris D.)