Decision analysis to assess the efficacy of routine sentinel lymphadenectomy in patients undergoing prophylactic mastectomy

Authors


  • This article is a US Government work and, as such, is in the public domain in the United States of America.

Abstract

BACKGROUND.

Patients who have invasive breast cancer identified after prophylactic mastectomy (PM) require axillary lymph node dissection (ALND) for lymph node staging (ie, directed ALND). Because the majority of these patients will be lymph node negative, sentinel lymphadenectomy (SLND) has been advocated at the time of PM to avoid the sequelae of unnecessary ALND. The objective of this study was to compare the efficacy of 2 surgical strategies, routine SLND versus directed ALND, in PM patients.

METHODS.

A decision-analytic model was created to compare the 2 surgical strategies. Model estimates were derived from a systematic literature review. The endpoints that were examined to compare the 2 strategies were the number of SLNDs performed per breast cancer detected, the number of SLNDs attempted to avoid 1 ALND in a lymph node-negative patient with occult invasive cancer, and the number of axillary complications associated with each strategy.

RESULTS.

The prevalence of invasive cancer in patients undergoing PM was estimated at 1.9%. At this rate, 37 SLNDs were performed per 1 breast cancer detected, and 73 SLNDs were required to avoid 1 ALND in a lymph node-negative PM patient. In 1 model scenario, the probability of complications per breast cancer detected was 9-fold greater with the SLND strategy than with the directed ALND strategy (2.7 vs 0.3). The complication rates for the 2 strategies become equivalent in the model scenario when the prevalence of occult invasive cancer was projected to 28%.

CONCLUSIONS.

Routine SLND for patients undergoing PM is not warranted given the large number of procedures required to benefit 1 patient and the potential complications associated with performing SLND in all patients. Cancer 2007. Published 2007 by the American Cancer Society.

Prophylactic mastectomy (PM) has been performed for more than 40 years for patients at high risk of breast cancer. The most common indication for PM is a family or personal history of breast cancer. It has been demonstrated that PM reduces the risk of breast cancer development by 90% in BRCA1 and BRCA2 mutation carriers1, 2 and in women with a personal or family history of breast cancer.3–6 The frequency of PM also has increased as awareness of breast cancer genetics and the risk reduction afforded by surgical prophylaxis has increased. Women who are newly diagnosed carriers of BRCA mutations choose bilateral mastectomy.7 Furthermore, review of the Surveillance, Epidemiology, and End Results database confirms an increase in contralateral PM from 1.8% in 1998 to 4.5% in 2003.8

Lymph lymph node staging is not performed routinely with PM. However, patients who are diagnosed with occult invasive cancer in the PM specimen subsequently are obligated to undergo axillary lymph node dissection (ALND) for lymph node staging. Because sentinel lymphadenectomy (SLND) has become standard practice for lymph node staging of early-stage breast cancer and has demonstrated 95% accuracy in multiple studies,9, 10 its role has expanded. Because most patients with occult carcinoma will be lymph node-negative, SLND at the time of PM has been advocated by some to avoid the sequelae of unnecessary ALND.11–13 In contrast, other investigators have suggested that SLND at the time of PM is not routinely warranted.14, 15 The consideration of sentinel lymph node (SLN) surgery in noncancer surgeries also has extended to patients undergoing reduction mammoplasty.16

Associated complications have become more apparent as experience with SLND has increased. These complications have been quantified in recent prospective studies.17–21 Of significance are the reported long-term complications of axillary paresthesias (9%–22%), lymphedema (7%), and decreased range of motion in the shoulder (0%–4%). In contrast, the frequency of occult invasive cancer in PM specimens is very low (range, 0.1%–3.5%).3, 4, 11, 22, 23 The frequent complications associated with SLND and the low prevalence of occult carcinoma in PM specimens suggest that routine SLND may not be warranted in all patients undergoing this procedure.14, 24 However, patients undergoing PM are a heterogeneous population, and some subgroups have as high as a 10% probability of finding an occult invasive carcinoma.14 Given the rising demand for PM, the debate over the efficacy of routine SLN surgery in these patients, along with the heterogeneity of occult cancer risk in this cohort, standard recommendations in this area are lacking. Therefore, we sought to evaluate the efficacy of SLN surgery using a decision-analysis model comparing the outcomes of 2 strategies—routine use of SLND at the time of PM versus directed ALND only when invasive cancer is found in the PM specimen—to assist with the development of practice guidelines for this patient population.

MATERIALS AND METHODS

We designed a decision-analytic model to simulate a randomized trial comparing routine SLND (ie, PM with concomitant SLND) and directed ALND (ie, PM without simultaneous lymph node staging followed by ALND in patients who had occult invasive cancer identified in the PM specimen) (Fig. 1). In the routine SLND strategy, all patients undergoing PM had a total mastectomy in conjunction with SLND. Patients with a positive SLN, irrespective of the pathologic findings in the breast, were obligated to undergo completion ALND. For patients with a failed SLN mapping procedure (ie, with a SLN that cannot be identified) who had no carcinoma detected in the PM specimen, no further axillary staging was undertaken. If occult invasive cancer was identified in the PM specimen, then these patients returned to the operating room for ALND. In the directed ALND strategy, routine SLND was not used in patients undergoing PM. If invasive cancer was identified on final pathologic evaluation of the PM specimen, then these patients returned to the operating room for ALND to complete staging.

Figure 1.

Decision-analytic model comparing routine sentinel lymphadenectomy (SLND) and directed axillary lymph node dissection (ALND) in patients undergoing prophylactic total mastectomy (PM). SLN indicates sentinel lymph node; cx, complications.

Our model also makes the following explicit assumptions: 1) failed SLN mapping is not associated with complications, 2) all occult invasive cancers detected in the PM specimen are T1 tumors (ie, tumors that measure ≤2 cm), and 3) the sensitivity for detecting lymph node metastases is higher with SLND than it is with standard assessment of lymph nodes in the ALND specimen because of the detailed pathologic assessment techniques used.

We assumed that women undergoing PM had at least 1 of the following indications, which increased their risk of developing breast cancer: 1) a family history of breast cancer, 2) a personal history of breast cancer in the contralateral breast, 3) a BRCA mutation, and 4) lobular carcinoma in situ on previous biopsy. Occult breast cancer was defined as invasive cancer identified in the PM specimen or as tumor cells in the SLN specimen. The decision-analytic software TreeAge Pro 2006 (Williamston, Mass) was used to construct and analyze the model.

Model Parameters

A systematic review of the literature was performed. Studies identified in a search of PUBMED entries published in English from 1993 through 2006 were reviewed comprehensively. The year 1993 was selected because this is the year of the first publication on SLND. Subject headings included prophylactic mastectomy, sentinel lymph node complications AND breast cancer, axillary lymph node dissection AND breast cancer, and nodal metastasis in T1 breast cancer. Summary statistics based on patient-level data were extracted from each study. These data included the prevalence of occult invasive cancer, SLN identification rate, incidence of complications of SLND and ALND, incidence of positive SLN and of positive lymph nodes at ALND in patients with T1 tumors, and incidence of positive SLNs in patients without cancer in the PM. The data were assimilated by using weighted averages to obtain a base estimate and range for each model parameter (Table 1).

Table 1. Base Estimates and Literature-based Ranges for Model Parameters
ParameterBase estimateMinimumMaximumReferences
  1. PM indicates prophylactic mastectomy; SLN, sentinel lymph node; SLND, sentinel lymphadenectomy; ALND, axillary lymph node dissection.

Prevalence of occult invasive cancer in the PM specimen0.0190.0010.0353,10,13,21,22
SLN identification rate0.950.85126–34
Probability of a positive SLN with cancer in the PM specimen0.2860.210.47526,27,35–39
Probability of a positive SLN with no cancer in the PM specimen0.0070.0060.03510,13,22
Probability of a positive axillary lymph node with cancer in the PM specimen0.2450.200.2735,37,40–44
SLND complication rate0.0680.00.2216–20
ALND complication rate0.3140.120.6916,18–20

Outcome Measures

The numbers of breast cancers identified, ALNDs performed, and axillary complications were calculated for each strategy (routine SLND vs directed ALND) for a cohort of 10,000 high-risk women undergoing PM. For this analysis, we defined clinical outcomes of interest including 1) the average number of SLNDs performed to detect 1 patient with breast cancer, 2) the number of SLNDs attempted to avoid 1 ALND in a lymph node-negative patient with occult invasive cancer, and 3) the potential number of axillary complications for a cohort of patients treated according to each strategy. For this model, we considered only chronic complications of lymphedema, paresthesias, and decreased range of shoulder motion associated with each surgical approach.

Statistical Analyses

A base-case analysis was performed using best estimates for all model parameters and probabilities. One-way sensitivity analyses were performed to investigate the effects of changes in individual model parameters, including SLN identification rate, prevalence of invasive breast cancer in the PM specimen, probability of positive lymph nodes at SLND or ALND, and complication rates of SLND and ALND.25, 26 The threshold value for the prevalence of invasive breast cancer in PM specimens, defined in terms of the ratio of axillary complications per breast cancer detected, also was projected.

RESULTS

Model Parameters

The prevalence of occult invasive cancer in PM specimens was reported in 5 studies of 1050 patients who were treated over the last 20 years.3, 11, 14, 22, 23 One report4 with a 0.1% rate of occult cancer was excluded from the calculation of the base estimate, because the pathologic analysis of specimens has changed significantly since the time of that study (1960–1993); however, this incidence was incorporated into the sensitivity analysis. The SLN identification rate was 95%.27–35 The probability of finding a positive SLN or a positive axillary lymph node in the case of occult invasive cancer in the PM specimen was estimated from the literature using the assumption that occult cancers in PM specimens are T1 tumors. Seven studies identified the incidence of SLN metastasis in T1 tumors27, 28, 36–40; whereas for the directed ALND strategy, data were used from 7 studies of ALND for T1 tumors that were conducted before the SLN surgery era.36, 38, 41–45 Data regarding positive SLN in the absence of invasive cancer in the PM specimen were assimilated from the largest 2 studies in the literature14, 23; data from a smaller, third study11 were excluded because they were discordant with the data in the larger reports. However, we included this outlier result in the sensitivity analysis. The probability of complications from SLND and ALND in breast cancer patients was calculated from recent prospective studies of each surgical procedure.17–21

Base Estimates

The base estimate for the prevalence of finding occult invasive cancer in a PM specimen was calculated as 1.9%, and the SLN identification rate was calculated as 95%. The estimated frequency of a positive SLN for patients with cancer in the PM specimen was 29%, and that of a positive SLN for patients without cancer was 0.7%. The probability of identifying axillary lymph node metastasis at the time of ALND was estimated at 25%. SLND and ALND complication rates were calculated as 7% and 31%, respectively (Table 1).

Comparison of Lymph node Strategies

The clinical outcomes associated with routine SLND and directed ALND in a theoretical cohort of 10,000 high-risk women undergoing PM were calculated (Table 2). In this model, 255 cases of invasive breast cancer were identified in the routine SLND strategy, and 190 cases of breast cancer were identified in the directed ALND strategy. Of the 255 cancers that were identified in the routine SLND strategy, 126 patients had SLN metastasis and underwent completion ALND whereas in the directed ALND strategy, all 190 patients who had cancer identified required ALND. Only 7 lymph node-negative patients underwent ALND using the SLND strategy; these represent patients in whom the SLN was not identified at the time of surgery and final pathology revealed invasive cancer in the breast. In contrast, in patients undergoing directed ALND, most (144 patients) were identified as lymph-node negative. Although fewer patients in the SLND arm ultimately underwent ALND, the number of patients experiencing axillary complications was much higher with routine SLND (680 patients vs 60 patients). In this hypothetical cohort of 10,000 women, routine SLND resulted in 680 patient complications and avoided ALNDs in 137 lymph node-negative patients.

Table 2. Clinical Outcomes Associated With the Routine Sentinel Lymphadenectomy Strategy and the Directed Axillary Lymph Node Dissection Strategy in a Theoretical Cohort of 10,000 Patients
Outcome*Average no. of patients (rate per 10,000)
Routine SLND strategyDirected ALND strategy
  • SLND indicates sentinel lymphadenectomy; ALND, axillary lymph node dissection.

  • *

    Clinical outcomes are based on base case parameter values.

  • Lymphedema, paresthesias, or decreased shoulder range of motion.

Breast cancer identified255190
ALND performed126190
ALND performed in lymph node-negative patients7144
Axillary complications68060

By using base-case estimates, the model demonstrates that, with the routine SLND strategy, 37 SLNDs are performed for the detection of 1 patient with breast cancer in this population. Thus, in this scenario, 36 patients without breast cancer are subjected unnecessarily to the potential complications of SLND. Furthermore, using routine SLND, 73 patients are subjected to the risks of SLND to spare 1 lymph node-negative patient from an ALND. A 9-fold difference in complication rates was noted between the 2 strategies. For every 10 breast cancers detected using the SLND strategy, 27 patients would experience axillary complications; however, for every 10 breast cancers detected using the directed ALND strategy, only 3 patients would experience axillary complications (Table 3) (Fig. 2A).

Figure 2.

Comparison of lymph node staging strategies in women undergoing prophylactic mastectomy (PM). (A) The ratio of the number of axillary complications per breast cancer detected in the routine sentinel lymphadenectomy (SLND) strategy to that of the directed axillary lymph node dissection (ALND) strategy decreases as the prevalence of invasive breast cancer in the PM specimen increases. (B) The number of SLNDs required to avoid 1 ALND in a lymph node-negative patient with breast cancer decreases as the prevalence of breast cancer in the PM specimen increases. The breast cancer prevalence of 1.9% represents the base estimate. SLN indicates sentinel lymph node.

Table 3. One-way Sensitivity Analysis of Model Parameters and Associated Clinical Outcomes
ParameterAverage no. of SLNDs per breast cancer detectedNo. of SLNDs needed to avoid 1 ALND in a lymph node-negative patientRisk of axillary complications per breast cancer detected
Routine SLND strategyDirected ALND strategyRatio (routine SLND/directed ALND)
  1. SLND indicates sentinel lymphadenectomy; ALND, axillary lymph node dissection; SLN, sentinel lymph node; PM, prophylactic mastectomy.

SLN identification rate
 0.8534812.50.38.3
 0.9036772.60.38.7
 0.95 (base)37732.70.39.0
 139702.80.39.3
Prevalence of occult invasive cancer in the PM specimen
 0.00112413908.70.329.0
 0.005822785.70.319.0
 0.01571394.00.313.3
 0.019 (base)37732.70.39.0
 0.03523401.70.35.7
Probability of a positive SLN with cancer in the PM specimen
 0.2137742.60.38.7
 0.286 (base)37732.70.39.0
 0.437732.70.39.0
 0.47537722.70.39.0
Probability of a positive SLN with no cancer in the PM specimen
 0.00639732.80.39.3
 0.007 (base)37732.70.39.0
 0.0134732.40.38.0
 0.0225731.90.36.3
 0.03518731.40.34.7
Probability of a positive axillary lymph node with cancer in the PM specimen
 0.237692.70.39.0
 0.2237712.70.39.0
 0.245 (base)37732.70.39.0
 0.2737762.70.39.0
SLND complication rate
 037730.20.30.7
 0.0337731.30.34.3
 0.068 (Base)37732.70.39.0
 0.137733.80.312.7
 0.1537735.70.319.0
 0.2237738.20.327.3
ALND complication rate
 0.1237732.60.126.0
 0.237732.60.213.0
 0.314 (base)37732.70.39.0
 0.437732.70.46.8
 0.537732.80.55.6
 0.6937732.80.74.0

The results from the 1-way sensitivity analysis are summarized in Table 3, which indicates that, as the prevalence of invasive breast cancer increases, the number of SLNDs needed to detect breast cancer decreases along with the ratio of complications. The average number of SLNDs performed to detect 1 breast cancer was sensitive to changes in the SLN identification rate, the prevalence of breast cancer in the PM specimen, and the probability of a positive SLN without identification of a primary breast cancer. The most significant change was observed when the prevalence of breast cancer in the PM specimen increased from 0.001 to 0.035, resulting in a decrease in the average number of SLNDs performed to detect a single breast cancer from 124 to 23 (Table 3). The analysis was also robust with respect to the number of SLNDs needed to avoid 1 ALND performed in a lymph node-negative patient. An ALND was avoided if a patient had occult cancer identified in the PM specimen, had undergone SLND for lymph node evaluation, and was staged as lymph node negative. The number of SLNDs needed to avoid 1 ALND in a lymph node-negative patient was 1390 SLNDs when the prevalence of breast cancer was 0.001 but decreased to only 40 SLNDs when the prevalence was 0.035 (Table 3) (Fig. 2B). The complications associated with the SLND strategy were less sensitive to changes in the SLN identification rate, the prevalence of breast cancer in the PM specimens, the probability of a positive SLN without identification of a primary breast cancer, and the probability of complications with SLND. The most notable changes in the complication ratio (SLND strategy/directed ALND strategy; from 29 to 5.7) were noted with a change in the prevalence of breast cancer in PM specimens from 0.001 to 0.035. The risk of complications varied from 0% to 22% with the SLND strategy as the complication ratio varied from 0.7 to 27.3. In contrast, as the probability of complications with ALND increased from 0.12 to 0.69, the complication ratio decreased from 26 to 4. Estimating from our model, we determined that, when the prevalence of invasive breast cancer in the PM specimens reached 28%, the complication rates for both routine SLND and directed ALND became equal (Table 4).

Table 4. Clinical Outcomes Associated With Increasing Prevalence of Breast Cancer
Prevalence of breast cancerAverage no. of SLNDs per breast cancer detectedNo. of SLNDs needed to avoid 1 ALND in a lymph Node-negative patientRisk of axillary complications per breast cancer detected
Routine SLND strategyDirected ALND strategySLND to ALND complication ratio
  1. SLND indicates sentinel lymphadenectomy; ALND, axillary lymph node dissection.

0.019 (base)37732.70.39.0
0.0517281.250.314.3
0.07512190.890.313.0
0.19140.700.312.3
0.2570.400.311.3
0.28350.310.311
0.3350.300.310.96
0.5230.210.310.7

DISCUSSION

The role of SLND in patients undergoing PM continues to be debated. To our knowledge, this study is the first to attempt an estimate of the long-term clinical implications of routine SLND among patients undergoing PM. Using a decision-analytic model, we compared the effectiveness of routine SLND with directed ALND in these patients. The results of our model indicated that directed ALND was a more effective strategy than routine SLND. For the average patient undergoing PM, the likelihood of axillary complications from routine SLND outweighs the benefits. We did not consider quality of life or cost in this analysis; however, in light of the significant difference in clinical outcomes between the 2 strategies, particularly with respect to the increased burden of complications in the routine SLND strategy, the addition of costs and utilities to our assessment would only increase the magnitude of the difference and would not favor the SLND strategy.

Our data indicate that the effectiveness of the SLND strategy depends primarily on the risk of occult cancer. The threshold value of 28% incidence of invasive carcinoma that we identified from our analysis is higher than clinically observed in patients undergoing prophylactic mastectomy. However, we recently reported that subgroups of PM patients, such as those with a history of invasive lobular carcinoma or those aged >60 years, may have a risk as high as 10% for occult invasive cancer.14 At this risk level, the routine SLND strategy still is less effective in terms of the clinical outcomes we evaluated. However, the differences between the 2 strategies are less marked in cohorts at higher risk for breast cancer, and it is possible that, in these cohorts, routine SLND may be the preferred strategy if additional financial and psychosocial costs associated with a second operation for ALND are considered. For a patient who is at higher than average risk for occult invasive cancer in the PM specimen, discussion of the risks and benefits of SLND with the patient becomes more important to allow for individualized care.

Before PM, a comprehensive examination of the breast, including clinical breast examination and diagnostic mammography obtained within 6 months before surgery, is required. Any lesion that is discovered in the breast should be worked up fully before surgical intervention to reduce the odds of identifying an occult malignancy in the PM specimen. Many centers routinely obtain bilateral breast magnetic resonance images (MRIs) as part of the preoperative evaluation of patients with biopsy-proven, invasive breast cancer for the evaluation of disease extent, multifocal and multicentric lesions, and detection of contralateral breast cancers.46–49 In high-risk patients, MRI also has been recommended for those who opt to undergo surveillance.50 Although to our knowledge data are scant, routine use of breast MRI before PM is likely to improve the ability to detect occult malignancy before prophylactic surgery and to minimize the rate of incidental cancer found in the PM specimen, further decreasing the role for SLND at time of PM.

A finding of interest from the current study is that a greater number of patients are diagnosed with breast cancer (defined as invasive cancer in the breast or metastatic disease in the lymph nodes) with routine SLND. Metastatic disease in the axillary lymph nodes in the absence of cancer identified in the breast is recognized as an initial presentation of primary breast cancer. Limitations in the processing of the mastectomy specimen can result in failure to identify small invasive breast cancers that have the potential to metastasize. Therefore, patients who have SLN-only disease may represent the early detection of a true occult malignancy. Conversely, it has been suggested that benign transport of epithelial cells because of manipulation of the breast parenchyma may lead to falsely positive SLNs.51, 52 However, the low incidence of positive SLNs in the absence of an obvious breast primary tumor, along with the lack of a previous breast biopsy, suggests that, in women who undergo PM, these SLN findings are more likely to be relevant biologically. This topic remains an area of substantial controversy, and the optimal treatment for these patients is not known. Therefore, the increased detection of breast cancer lymph node metastases in this model by the use of routine SLND strategy should not be interpreted as a reason to perform routine SLND.

Our model has several limitations inherent to the assumptions made to develop it. In routine clinical practice, not all patients who have a positive SLN identified, with or without cancer in the PM specimen, will proceed to ALND. Another limitation is that the estimated complication rate for SLND in the model was based primarily on reports of patients undergoing partial mastectomy and radiotherapy, because the SLND complication rate for patients who undergo total mastectomy is not well documented. These estimates may bias the model toward the ALND strategy, possibly by overestimating the complications associated with SLND. Conversely, failed SLN mapping most likely has some morbidity secondary to the dissection performed trying to identify the SLN; therefore, our assumption of no complications likely favors the routine SLND strategy. Finally, the axillary complications vary widely from mild arm paresthesias to disabling lymphedema, and the severity of the long-term complications from lymph node staging was not incorporated into this model.

The current study also has several strengths. We used weighted averages from the extensive review of published studies to ensure precise estimates for the variables in the model. In addition, we focused on clinically relevant outcomes known to affect long-term quality of life.19, 21, 53 Because the retrospective data on long-term consequences of lymphadenectomy often are inaccurate, we used data only from prospective trials to estimate the complication rates of ALND and SLND. We also incorporated the higher sensitivity of SLND than ALND for detecting lymph node metastasis. On the basis of the results from using this model, we conclude that the routine use of SLND is not warranted in patients who are undergoing PM.

Ancillary