|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Journal of Clinical Oncology, Vol 26, No 13 (May 1), 2008: pp. 2093-2098 © 2008 American Society of Clinical Oncology. DOI: 10.1200/JCO.2007.11.9479 Nomogram for the Prediction of Having Four or More Involved Nodes for Sentinel Lymph Node–Positive Breast Cancer
From the Departments of Radiation Oncology and Surgery, Massachusetts General Hospital; and the Departments of Surgery and Radiation Oncology, Brigham and Women's Hospital, Boston, MA Corresponding author: Angela Katz, MD, Department of Radiation Oncology, Massachusetts General Hospital, 100 Blossom St, Cox 301, Boston, MA 02114; e-mail: abkatz{at}partners.org
Purpose The standard of care for patients with a positive (+) sentinel lymph node (SLN) is axillary dissection; however, for various reasons, some SLN+ patients do not undergo dissection. The purpose of this study was to define possible predictors of having four or more involved nodes to provide information for clinicians and patients making decisions about adjuvant chemotherapy and radiation. Patients and Methods We reviewed the records of 402 patients with invasive breast cancer and one to three involved SLNs who underwent completion axillary dissection at two academic cancer centers. None of these patients received neoadjuvant chemotherapy. Factors associated with having four or more involved axillary nodes (SLNs and non-SLNs) were evaluated by univariate and multivariate logistic regression analysis. Results Eighty-seven patients had four or more positive nodes. On multivariate analysis, having four or more SLNs was associated with tumor histology, primary tumor size, lymphovascular space invasion, extranodal extension, the number of involved SLNs, the number of uninvolved SLNs, and the size of the largest SLN metastasis. A nomogram to predict the probability of having four or more nodes based on patients’ pathologic data was developed from the multivariate logistic regression model. A separate previously published data set of 206 SLN+ patients treated at a community hospital in another city was used to validate this model. Conclusion Patients with a low probability of having four or more nodes can be identified from known pathologic features. The nomogram developed will be helpful to clinicians making adjuvant treatment recommendations.
The standard of care for patients with positive sentinel lymph nodes (SLNs) is axillary dissection; however, for various reasons, some SLN+ patients are referred for adjuvant therapy without a dissection. The appropriate treatment of these patients is unclear. This study evaluates potential predictors of having four or more involved nodes in a large group of patients with positive SLNs who underwent completion axillary dissection to develop a nomogram to predict the likelihood of having four or more positive nodes. The nomogram will be helpful to surgeons and radiation and medical oncologists who are making treatment decisions such as whether to dissect or radiate the axilla and whether to offer immediate reconstruction or the extent of systemic therapy needed, as well as to patients in understanding their prognosis.
The records of all patients with invasive breast cancer and at least one positive SLN treated at Massachusetts General Hospital and Brigham and Women's Hospital (Boston, MA) were retrospectively reviewed. From this group, 402 patients with invasive breast carcinoma and one to three positive SLNs who underwent completion axillary dissection, defined as removal and examination of six or more axillary nodes (SLN plus non-SLNs), were identified and evaluated for the current study (Massachusetts General Hospital, n = 248; Brigham and Women's Hospital, n = 154). Because the terminology used by surgeons in their operative reports varied, the decision was made to exclude patients with fewer than six nodes examined, because this was felt to be the minimum number of nodes needed to adequately assess the extent of nodal involvement in the axilla.1 None of these patients received neoadjuvant systemic therapy. Institutional review board approval was obtained before record review. All patients were clinically node negative. Definitive treatment consisted of mastectomy for 186 patients (46%) and breast-conserving surgery for 215 patients (53%; surgery for one patient was unknown, < 1%). Isosulfan blue and/or radioisotope were injected in the subareolar or peritumoral region of the breast. SLN mapping was performed by tracing the blue-stained lymphatics and using a handheld gamma-detection probe. A median of two SLNs were identified per procedure. Fourteen percent of the patients (n = 57) had fewer than 10 total nodes (SLN plus non-SLN) removed, 70% (n = 279) had 10 to 20 nodes removed, and 16% (n = 66) had more than 20 nodes removed. SLNs were examined by frozen section intraoperatively at the discretion of the surgeon and were then fixed in formalin, embedded in paraffin, sectioned at 2-mm intervals and entirely submitted for examination. Because keratin staining was performed only at Massachusetts General, only patients who were SLN+ on hematoxylin and eosin (H&E) staining were included in this analysis. Non-SLNs from completion axillary dissections were examined by routine H&E staining at one level. Associations between having four or more involved nodes with regard to patient, tumor, and treatment-related factors were evaluated using Fisher's exact tests and univariate and multivariate logistic regression analysis. All P values were two tailed, with .05 or lower considered significant. A nomogram was developed on the basis of a multivariable logistic regression model with a combination of continuous variables (eg, age, tumor size), discrete numerical variables (eg, the number of involved sentinel lymph nodes), and categoric variables (eg, lymphovascular space invasion, lobular histology versus other histologies.) We used a forward stepwise selection method with likelihood-ratio test to select a subset of all analyzed factors. In the final model/nomogram, only the factors that are statistically significant at the .05 level remain. To overcome some of the limitations of the stepwise selection method, we tested the final model against smaller models (ie, with a subset of factors included in the nomogram) using the likelihood-ratio test. The goodness-of-fit of the multivariate model/nomogram was evaluated using the Hosmer-Lemeshow test, grouping of the data into 10 quantiles, using a classification table, and calculating the model's sensitivity, specificity, positive predictive value, negative predictive value, and area under the receiver operating characteristic (ROC) curve. The scale parameters (ie, the constants) of the nomogram were obtained from the corresponding coefficients of the multivariate logistic regression model by multiplying them by the same normalization factor. The normalization factor was chosen to make the scale parameters of the nomogram easy to use for back-of-the-envelope calculations. The nomogram was validated using a separate, previously published data set of 206 patients who met the eligibility criteria of the current study but who were treated at Sibley Memorial Hospital, a community hospital in Washington, DC. This data set was chosen because of its ready availability to the investigators and its utility in testing the model in a subset of patients treated in a different setting than large academic centers such as Massachusetts General Hospital and Brigham and Women's Hospital. The validation procedure consisted of calculating the probability of having four or more involved nodes for all patients in the validation data set using the nomogram and analyzing the goodness-of-fit by using Pearson and Hosmer-Lemeshow tests, and by analyzing sensitivity, specificity, positive predictive value, negative predictive value, and area under the ROC curve.
Patient, tumor, and treatment characteristics are listed in Table 1. Histology was invasive ductal carcinoma for 77% of patients, invasive lobular carcinoma for 11%, mixed ductal and lobular carcinoma for 9%, and other for 2%. The median tumor size was 1.8 cm (range, 0.1 to 10.5 cm). Fifty-three percent of patients had involved SLNs on frozen section, and 76% of patients had only one involved SLN. The median size of the largest SLN metastasis on H&E staining was 5 mm (range, < 0.2 to 65 mm). Twenty-two percent of the patients had micrometastases (  2 mm). One hundred twenty-two patients (30%) had extranodal extension.
Potential Predictors of Having More Than Four Involved Nodes Eighty-one patients had four or more positive axillary lymph nodes (SLN plus non-SLN). On univariate logistic regression analysis (Table 2) for the 402 patients who underwent completion axillary dissection after identification of an involved SLN, the proportion of patients with four or more total involved nodes increased with increasing primary tumor size, histologic grade, number of involved SLNs, and size of the largest SLN metastasis, and decreased with increasing number of uninvolved nodes. The presence of lymphovascular space invasion or extranodal extension was also associated with presence of four or more involved nodes, as was invasive lobular histology or having a positive SLN identified on frozen section. There were only 28 patients 70 years of age or older. Of these, 46% had four or more involved nodes. Patients who underwent mastectomy tended to have more advanced disease and were more likely to have four or more involved nodes.
On multivariate logistic regression analysis (Table 2), increasing primary tumor size, invasive lobular histology, the presence of lymphovascular space invasion, the presence of extranodal extension, increasing number of involved SLNs, macroscopic size of largest SLN metastasis, and decreasing number of uninvolved SLNs were correlated with an increased probability of having four or more involved nodes. Figure 1A shows the ROC curve corresponding to the multivariate logistic model applied to a subset of 344 patients for whom the values or status of all seven risk factors were known. The area under the ROC curve is 0.83, indicating potentially promising predictive power of the multivariate logistic regression model. This multivariate model was applied to a separate, previously published data set of 206 patients who met the eligibility criteria of the current study but who were treated at Sibley Memorial Hospital.2 Figure 1B shows the ROC curve corresponding to the multivariate logistic model applied to the Sibley Memorial Hospital data set. The area under the ROC curve is 0.81, indicating the potentially promising predictive power of the multivariate logistic regression model in a separate data set of patients treated at a different institution.
Development and Validation of Nomogram On the basis of the results of the multivariable analysis, a formula was developed to calculate the probability of having four or more involved nodes using the known pathologic features of patients’ tumors and SLN(s) (Fig 2). A total score is calculated using primary tumor size, number of involved SLNs, micro or macro size of the largest SLN metastasis, the presence of lymphovascular space invasion, the presence of extranodal extension, histology, and the number of negative SLNs. This total score can then be used to assign a probability of having four or more involved nodes for individual patients using the scale at the bottom of the figure.
We investigated whether a simpler nomogram (with less parameters) can be almost as good as the derived nomogram with seven parameters. The likelihood-ratio tests indicated that excluding any one of the seven parameters results in statistically significant degradation of the goodness-of-fit and the predictive power of the nomogram. We calculated the nomogram's classification table, sensitivity, and negative predictive value using a 5% cutoff. That is, we evaluated the power of the nomogram to correctly classify the patients into two groups; those with fewer than four involved nodes and those with four or more involved nodes based on whether the predicted probability was more or less than 5%. Of the 118 patients with the nomogram-based calculated probability of 5% or less, 113 had indeed three or fewer involved nodes resulting in negative predictive value of 94.7%. Of the 57 cases with four or more involved nodes 54 were classified as "positive" (ie, had the calculated nomogram-based probability greater than 5%) resulting in the sensitivity of the nomogram of 97.5%. The results of Pearson and Hosmer-Lemeshow tests of goodness-of-fit were not significant (.9 and .4 respectively), indicating good fit. Table 3 presents the observed and predicted rates of having four or more involved lymph nodes for patients in four risk groups from both the Massachusetts General Hospital and Brigham and Women's Hospital data sets and the validation data set from Sibley Memorial Hospital. The nomogram predictions are consistent for both data sets, although they somewhat underestimated the actual number of observed cases in the validation data set. The results of Pearson and Hosmer-Lemeshow tests of goodness-of-fit are not significant (.6 and .5 respectively) again indicating good predictive power of the nomogram when applied to an independent validation data set.
The use of SLN mapping has spared many breast cancer patients the need for an axillary dissection, and thus the risk of long-term complications such as lymphedema, numbness, chronic pain, and reduced range of motion. Although many physicians still consider an axillary dissection to be mandatory for patients with a positive SLN, others have questioned whether it is necessary to subject all SLN+ patients to its risks. For almost half of patients with positive SLNs, there are no additional involved axillary nodes identified on completion axillary dissection.3,4 These patients might receive no therapeutic benefit from dissection, and because most oncologists now recommend systemic therapy for all node-positive patients, determining the exact number of involved nodes is now of less importance in making treatment decisions than it was in the past. On the basis of the controversial prognostic and therapeutic benefits of axillary dissection, and amid concerns about its potential morbidity, many surgeons elect not to perform completion axillary dissections on selected SLN positive patients. These patients are often seen by radiation oncologists, who must decide whether radiotherapy to the regional nodes is warranted and to what extent, and by medical oncologists, who now must choose from among the many effective systemic therapy regimens available with little information about the extent of axillary involvement. In addition, patients want accurate estimates of their potential for cure. The nomogram that we developed will therefore be of value to surgeons, medical oncologists, radiation oncologist, and patients in discussing treatment options and their potential outcomes. Our data suggest that many SLN+ patients have a low probability of having four or more involved axillary lymph nodes. The likelihood of having four or more involved nodes can be predicted on the basis of known pathologic features of the patients’ tumors and SLN(s). If these patients underwent axillary dissection, they would have undergone a standard level I to II axillary dissection, and many would not have required treatment to the axillary apex or supraclavicular fossa. Previously published data have shown that failure in the supraclavicular fossa or axillary apex is rare for patients with fewer than four involved nodes or other high-risk features, such as extranodal extension.5-7 It must be pointed out that rates of supraclavicular failure cited in the literature might under-report the true incidence of this outcome, because the area is difficult to examine and patients who develop metastatic disease are often censored from analyses of locoregional control. Yet, many radiation oncologists continue to offer comprehensive nodal radiation, including radiation to the axillary apex and supraclavicular fossa, to all SLN+ patients. In essence, this is the same treatment that would be offered if the patient had undergone no surgical evaluation of the axilla at all. This treatment philosophy was the safest approach in the era when we had limited data concerning outcome after SLN mapping; however, as more data concerning the significance of SLN metastases begin to emerge, individualization of treatment may be appropriate. There are several obvious limitations to this study. As with any retrospective investigation, there is the issue of selection bias. Only SLN+ patients who underwent a completion axillary dissection are included in the present analysis. Patients who did not undergo axillary dissection, including 73 treated on a clinical trial investigating the use of axillary radiation to the regional nodes without dissection, are not included. The exclusion of these patients is a potential source of bias in our analysis. In addition, the exact threshold of nodal involvement used to recommend radiation therapy to the axillary and supraclavicular nodes remains somewhat controversial, and some clinicians prefer to treat all node-positive patients with regional nodal radiation. Finally, because the terminology used by surgeons in their operative reports varied, the decision was made to exclude patients with fewer than six nodes examined, as this was felt to be the minimum number of nodes needed to adequately assess the number of nodes positive in the axilla. This decision introduces some bias, because some of the patients might have had more of an axillary sampling than a true axillary dissection. In our previous analysis of 224 patients who underwent SLN mapping and completion axillary dissection without prior chemotherapy, we found that having invasive lobular histology, more than one involved SLN, lymphovascular space invasion, or a SLN macrometastasis was associated with having four or more involved nodes.2 These results confirmed the findings of a similar report from The University of Texas M.D. Anderson Cancer Center (Houston, TX).8 In their analysis of 265 patients treated with mastectomy, SLN mapping, and completion axillary dissection without neoadjuvant chemotherapy, they found that lack of drainage at SLN mapping, involvement of more than one SLN, and lymphovascular space invasion were independent predictors of an increased risk of having four or more involved axillary nodes. In their patient population, the 106 (40%) of 265 patients who lacked these three features had less than a 2% risk of involvement of four or more axillary nodes. These two studies, combined with previously published data that demonstrated that for many patients, most of level I and II of the axilla, the areas addressed by a standard axillary dissection, can be encompassed within modified tangential radiation fields,9 lead to the hypothesis that patients at low risk of having four or more nodes might be adequately treated with modified high tangential fields rather than comprehensive nodal radiation. Chagpar et al10 also recently published a study of SLN+ patients that aimed to develop a predictive model for the presence of four or more nodes in an attempt to predict which patients might need postmastectomy radiation therapy and, therefore, which patients should not undergo immediate reconstruction after mastectomy. Their patient cohort included 1,133 SLN+ patients, half of whom were used in an initial data set to develop the predictive model and half of whom were used to validate the model. Their model included T stage, number of positive SLNs, and proportion of SLNs that were positive. Our nomogram uses many of the same variables, and adds to the model by taking into account factors such as lymphovascular space invasion and the size of the largest SLN metastasis. In addition, our nomogram has been validated in both large, academic settings and a smaller, community practice setting. The addition of a supraclavicular/axillary apex field to tangential breast radiation has been demonstrated to increase treatment-related morbidity. The risk of lymphedema rises from 2% to 8% to approximately 9% to 24% with the addition of a supraclavicular/axillary apex field for patients treated with axillary radiation alone.6,11,12 For patients undergoing surgery and radiation, these rates are even higher.12 Higher rates of pneumonitis and brachial plexopathy have also been demonstrated in patients treated with a third field when compared with those treated with tangents alone.6,11,12 Very little information is available concerning the morbidity of comprehensive nodal radiation after identification of an involved sentinel node. A prospective trial conducted at Massachusetts General Hospital and Brigham and Women's Hospital that included comprehensive nodal radiation for SLN+ patients was reported in abstract form in 2005 and demonstrated less than 1% lymphedema in the 73 patients treated with a median follow-up of 32 months.13 However, the latency of complications from radiotherapy is relatively long, and ranges from 2 years for fibrosis to 3 to 5 years for lymphedema or brachial plexopathy, and 7 to 14 years for arm weakness.14 If we hypothesize that patients with less than a 5% chance of having four or more involved nodes do not need treatment to the supraclavicular fossa/axillary apex, then 34% of patients in the current data set and 37% of patients in the validation data set could have been spared the morbidity of comprehensive nodal radiation. In summary, our data suggest that patients with a low probability of having four or more nodes can be identified using the known pathologic features of their primary tumor and SLN(s). The nomogram that we developed will be of use to oncologists and patients in discussing treatment options, including the need for axillary dissection, the extent of axillary radiation required and the optimal systemic therapy regimen, as well as prognosis. Further investigation concerning the appropriate treatment for patients with a positive SLN who, for various reasons, do not undergo axillary dissection is warranted.
The author(s) indicated no potential conflicts of interest.
Conception and design: Angela Katz, Barbara L. Smith, Mehra Golshan, Andrzej Niemierko, Wendy Kobayashi, Alexandra Kelada, Levi Rizk, Michele Gadd, Michelle Specht, Alphonse G. Taghian Administrative support: Wendy Kobayashi, Alexandra Kelada, Levi Rizk Provision of study materials or patients: Barbara L. Smith, Mehra Golshan, Michele Gadd, Michelle Specht, Alphonse G. Taghian Collection and assembly of data: Angela Katz, Andrzej Niemierko, Wendy Kobayashi, Rita Abi Raad, Alexandra Kelada, Levi Rizk, Julia S. Wong, Jennifer R. Bellon Data analysis and interpretation: Angela Katz, Andrzej Niemierko, Wendy Kobayashi, Julia S. Wong, Jennifer R. Bellon Manuscript writing: Angela Katz, Barbara L. Smith, Mehra Golshan, Andrzej Niemierko, Julia S. Wong, Jennifer R. Bellon, Michelle Specht, Alphonse G. Taghian Final approval of manuscript: Angela Katz, Barbara L. Smith, Mehra Golshan, Andrzej Niemierko, Wendy Kobayashi, Rita Abi Raad, Julia S. Wong, Jennifer R. Bellon, Michele Gadd, Michelle Specht, Alphonse G. Taghian
Supported in part by Grants No. CA21239 and CA50628 from the National Institutes of Health (A.N.) and in part by the Tim Levy and Blanche Montesi Funds for Breast Cancer Research, and the National Cancer Institute (NCI)/Avon Foundation supplement to the NCI Specialized Program of Research Excellence (SPORE) award P50 CA89393, Dana-Farber SPORE in Breast Cancer (A.G.T.). Presented at the 48th Annual Meeting of the American Society for Therapeutic Radiation and Oncology, November 5-9, 2006, Philadelphia, PA. Authors’ disclosures of potential conflicts of interest and author contributions are found at the end of this article.
1. Rose C, Botnick LE, Weinstein M, et al: Axillary sampling in the definitive treatment of breast cancer by radiation therapy and lumpectomy. Int J Radiat Oncol Biol Phys 9:339-344, 1983[Medline] 2. Katz A, Niemierko A, Gage I, et al: Factors associated with involvement of four or more axillary nodes for sentinel lymph node-positive patients. Int J Radiat Oncol Biol Phys 65:40-44, 2006[Medline] 3. Chu KU, Turner RR, Hansen NM, et al: Do all patients with sentinel node metastasis from breast carcinoma need complete axillary node dissection? Ann Surg 229:536-541, 1999[CrossRef][Medline] 4. McMasters KM, Giuliano AE, Ross MI, et al: Sentinel-lymph-node biopsy for breast cancer–not yet the standard of care. N Engl J Med 339:990-995, 1998 5. Katz A, Strom EA, Buchholz TA, et al: Locoregional recurrence patterns after mastectomy and doxorubicin-based chemotherapy: Implications for postoperative irradiation. J Clin Oncol 18:2817-2827, 2000 6. Recht A, Pierce SM, Abner A, et al: Regional nodal failure after conservative surgery and radiotherapy for early-stage breast carcinoma. J Clin Oncol 9:988-996, 1991[Abstract] 7. Taghian A, Jeong JH, Mamounas E, et al: Patterns of locoregional failure in patients with operable breast cancer treated by mastectomy and adjuvant chemotherapy with or without tamoxifen and without radiotherapy: Results from five National Surgical Adjuvant Breast and Bowel Project randomized clinical trials. J Clin Oncol 22:4247-4254, 2004 8. Shahar KH, Hunt KK, Thames HD, et al: Factors predictive of having four or more positive axillary lymph nodes in patients with positive sentinel lymph nodes: Implications for selection of radiation fields. Int J Radiat Oncol Biol Phys 59:1074-1079, 2004[CrossRef][Medline] 9. Schlembach PJ, Buchholz TA, Ross MI, et al: Relationship of sentinel and axillary level I-II lymph nodes to tangential fields used in breast irradiation. Int J Radiat Oncol Biol Phys 51:671-678, 2001[CrossRef][Medline] 10. Chagpar AB, Scoggins CR, Martin RC, II, et al: Predicting patients at low probability of requiring postmastectomy radiation therapy. Ann Surg Oncol 14:670-677, 2007[CrossRef][Medline] 11. Coen JJ, Taghian AG, Kachnic LA, et al: Risk of lymphedema after regional nodal irradiation with breast conservation therapy. Int J Radiat Oncol Biol Phys 55:1209-1215, 2003[CrossRef][Medline] 12. Halverson KJ, Taylor ME, Perez CA, et al: Regional nodal management and patterns of failure following conservative surgery and radiation therapy for stage I and II breast cancer. Int J Radiat Oncol Biol Phys 26:593-599, 1993[Medline] 13. Gadd M, Harris J, Taghian A, et al: Prospective study of axillary radiation without axillary dissection for breast cancer patients with a positive sentinel node. Breast Cancer Res Treat 94:S12, 2005 14. Johansson S, Svensson H, Denekamp J: Dose response and latency for radiation-induced fibrosis, edema, and neuropathy in breast cancer patients. Int J Radiat Oncol Biol Phys 52:1207-1219, 2002[CrossRef][Medline] Submitted March 30, 2007; accepted November 21, 2007.
Related Editorial
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2008 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
