Optimal use of sentinel lymph node biopsy versus axillary lymph node dissection in patients with breast carcinoma

A decision analysis

Authors


Abstract

BACKGROUND

There are no data available from randomized controlled trials that compare the efficacy of sentinel lymph node (SLN) biopsy with Level I/II axillary lymph node dissection (ALND) in patients with breast carcinoma. We performed a formal decision analysis to determine whether SLN biopsy is appropriate, compared with ALND, for patients with T1, T2, and T3 tumors and to quantify the relative value of these two procedures in the management of patients with breast carcinoma.

METHODS

All clinically relevant outcomes were modeled for both SLN biopsy and ALND. The probabilities of complications and outcomes were derived using data from the University of Louisville Breast Cancer Sentinel Lymph Node Study and from extensive review of previous studies. Utilities were assigned by the authors, incorporating values from the literature whenever possible.

RESULTS

The expected utility of SLN biopsy was higher than the expected utility for ALND for T1 and T2 tumors that were 4.0 cm or smaller. There was no clear preference for either procedure with tumors that were larger than 4.0 cm. The T1 and T2 results were robust to sensitivity analysis.

CONCLUSIONS

The results of this decision analysis suggest that SLN biopsy is preferred over ALND for patients with breast tumors that are 4.0 cm or smaller. Patients should be aware of the potential for false-negative results in SLN biopsy, but this risk is outweighed by the decreased morbidity associated with the procedure itself. Cancer 2002;95:478–87. © 2002 American Cancer Society.

DOI 10.1002/cncr.10696

Level I/II axillary lymph node dissection (ALND) has long been the standard for determining the lymph node stage in patients with breast carcinoma, but the procedure can be associated with significant morbidity. Sentinel lymph node (SLN) biopsy, a diagnostic test used to determine the presence of metastases in the regional lymph nodes, is becoming increasingly accepted as a minimally invasive alternative to axillary dissection. The procedure identifies the first draining lymph node(s) to receive lymphatic drainage from the primary tumor site and accurately reflects the status of the remaining axillary lymph nodes. As with any new diagnostic modality however, it is important to evaluate the data critically and to determine how the procedure should best be utilized.

There are no data available from randomized controlled trials that compare the efficacy of SLN biopsy with the efficacy of ALND. There are significant trade-offs between the two surgical procedures and there is uncertainty as to the appropriate clinical strategy. The morbidity associated with ALND is significantly higher than with SLN biopsy. However, there is a definitive risk of false-negative results with SLN biopsy, which could lead to understaging and incorrect adjuvant therapy decisions.1 There is uncertainty as to whether SLN biopsy should be offered to all patients with invasive breast carcinoma, especially because patients with larger tumors (T2, T3) have a high rate of axillary lymph node metastases. Patients with tumors larger than 1 cm in diameter often receive adjuvant therapy regardless of axillary lymph node status. Furthermore, some have argued that patients with very small tumors (T1a, T1b) have such a low risk of axillary metastases that axillary staging procedures can be avoided altogether.2–6 It is for these patients with very small tumors, however, that the finding of lymph node metastasis would potentially make a critical difference in recommended therapy — often the difference between adjuvant chemotherapy or no chemotherapy at all.

Formal decision analysis is an appropriate method of quantifying the relative value of these two procedures in the management of patients with breast carcinoma.7, 8 Decision analysis explicitly considers each pertinent health outcome and the consequences associated with each outcome. We performed an extensive review of the literature and analyzed data from a large, multiinstitutional study to provide the estimates necessary to answer the question of whether ALND or SLN biopsy should be offered to patients with T1, T2, and T3 tumors.

MATERIALS AND METHODS

Model Design

The software package DATA 4.0 (TreeAge Software, Williamstown, MA) was used to model the formal decision analysis. The comparison of SLN biopsy to ALND is illustrated as a decision tree (Fig. 1). The two main branches of the tree represent the surgical options of SLN biopsy or ALND for axillary staging in patients with biopsy-proven invasive breast carcinoma. Subbranches represent possible clinical consequences of either SLN biopsy or ALND. Chance events were modeled as dichotomous branch points, taking into account the potential complications of each procedure. Individual decision trees were constructed and subsequent decision analyses conducted for seven tumor size breakdowns: T1a (< 0.5 cm), T1b (0.5–1.0 cm), T1c (> 1.0–2.0 cm), T2 (> 2.0–3.0 cm), T2 (> 3.0–4.0 cm), T2 (> 4.0–5.0 cm), and T3 (> 5.0 cm).

Figure 1.

Schematic representation of decision tree for the use of sentinel lymph node biopsy compared with axillary lymph node dissection.

The major risks/outcomes considered crucial to this decision analysis were 1) anesthetic/operative mortality; 2) major anesthetic/operative morbidity of myocardial infarction (MI), pulmonary embolism (PE), or stroke; 3) anaphylactic reaction (nonfatal) to blue dye; 4) postoperative complications of numbness, paresthesia, or prolonged pain; 5) chronic disabling lymphedema; and 6) lymph node metastases, including both true-positive and false-negative results.

We identified four major outcome groupings with accompanying potential side effects, resulting in four distinct patterns of risk: 1) the SLN is found and is negative, resulting in no additional procedure; 2) the SLN is found and is positive, resulting in subsequent ALND; 3) the SLN is not found and the patient proceeds immediately to ALND; and 4) ALND is performed without SLN biopsy.

Data and Assumptions

SLN identification rate, SLN false-negative rate, and prevalence of axillary metastasis

The SLN identification rate (defined as the proportion of patients with successful SLN localization), SLN false-negative rate (the proportion of patients with lymph node metastases in whom the SLN is incorrectly found to be negative), and the prevalence of axillary metastasis were derived from results of the University of Louisville Breast Cancer Sentinel Lymph Node Study, a prospective, nonrandomized study (Table 1). This study is by far the largest study of its kind, with a broad multiinstitutional experience, representing a large proportion of the available data on SLN biopsy.9, 10 We used seven types of prevalence of lymph node metastasis, which correspond with data for each of seven tumor stages as defined above. We elected, however, to use only one false-negative rate (8%) and one SLN identification rate (93%) across all tumor stages. This decision was based on the fact that although there was some variability in the SLN false-negative and identification rates by tumor size category, these differences were not statistically significant and sample sizes for many of the subgroups were relatively small.9 Our results are largely commensurate with those published in a host of other series. All patients in the University of Louisville study had clinical T1 or T2 biopsy-proven breast carcinoma and underwent SLN biopsy followed by completion level I/II axillary dissection as validation of the SLN procedure. No patients received preoperative chemotherapy or radiotherapy. Data for T3 tumors are derived from the subset of patients with clinical stage T2 tumors that were subsequently found to be larger than 5 cm on final pathologic examination.

Table 1. Probabilities and Prevalences Derived from the University of Louisville Breast Cancer Sentinel Lymph Node Study9
T stageNo.Axillary metastases (%)
T1a15512.3
T1b45918.5
T1c88230.4
T2 (> 2.0–3.0 cm)39044.1
T2 (> 3.0–4.0 cm)11058.2
T2 (> 4.0–5.0 cm)4566.7
T34477.3

Extensive literature review was performed using electronic databases to extrapolate probabilities of events (Table 2). Search terms included breast cancer, axillary lymph node dissection, sentinel lymph node biopsy, complications, morbidity, mortality, anesthetic complications, and lymphedema.

Table 2. Point Estimates and Plausible Ranges of Outcome Probabilities
Complication variableALND baseline value (%)ALND plausible range (%)SLN baseline value (%)SLN plausible range (%)References
  1. ALND: axillary lymph node dissection; SLN: sentinel lymph node.

Mortality from general anesthesia0.010.0001–0.10.010.0001–0.111
Major morbidity from general anesthesia0.920.05–2.00.920.05–2.011
Allergic reaction to isosulfan blue dye1.50–1.512
Anaphylactic reaction to blue dye10.6–2.513–15
Wound infection; seroma/hematoma; numbness; prolonged pain; weakness; axillary vein thrombophlebitis150–21.140–916–24
Lymphedema206–3010–321, 25–28

Assumptions

Certain assumptions have been made for the purposes of these analyses. First, there are no differences in life expectancy between patients undergoing the SLN and ALND procedures.29, 30 Second, there is a recognized 2–3% risk of a false-negative result associated with level I/II ALND31, 31 because level III lymph nodes are not removed (the so-called “skip metastasis”). In the current analysis, we disregard any false-negative results attributable to ALND, as it is considered to be the gold standard and because level III dissection is not part of the routine operation performed for breast carcinoma patients. Third, based on the results of many previous studies, SLN biopsy is performed optimally with the injection of both blue dye and radioactive colloid for identification of the SLN,33–37 and we presume that all SLN biopsies are performed using this technique. Fourth, although SLN biopsy can be performed using local anesthesia, this decision analysis is modeled with the assumption that all procedures are done under general anesthesia. Because we also assume that completion ALND will be performed for all patients with histologically positive SLN, a second-stage procedure necessarily carries all the attendant risks of two separate trips to the operating room. The use of intraoperative imprint cytology or frozen section analysis of the SLN is not taken into account. Finally, anesthesia/operative morbidity and mortality rates are based on generally healthy patients without any excessive cardiac or noncardiac risk factors.

Utilities

Utilities are a quantitative measure of patients', society's, or health care providers' attitudes concerning the desirability of a given health state.7 In this decision analysis, the utilities represent the relative quality of life of various health states in this context, ranging from 0.0 to 1.0, with 0.0 representing the worst possible outcome (death for those without lymph node metastasis) and 1.0 representing the best possible outcome in this context (no side effects and no lymph node metastases). Utility values were assigned by the authors based on our expert experience and extensive review of the available literature on patient preferences in ALND.20, 38–41

Although selected utilities (including combinations of outcomes) are specifically listed in Table 3, a few utilities are worthy of brief discussion. A true-positive diagnosis of lymph node metastasis was assigned a quality of life estimate of 0.60; a major anesthetic/operative side effect of MI, PE, or stroke was estimated to be 0.50; and a false-negative diagnosis — a missed lymph node metastasis — was assigned a utility of 0.40. Given that the possibility of a false-negative diagnosis is a problem unique to SLN biopsy and not ALND (ALND is considered the gold standard), its obviously low utility value was designed as a “worse case scenario,” subjecting SLN biopsy to more rigorous scrutiny. That is, a person with unrecognized lymph node metastasis would not have the opportunity to benefit from adjuvant therapy and would suffer a much lower quality of life in the long run.

Table 3. Threshold Values for T2(> 4.0–5.0 cm) Stage Tumor: Selected Utilities
VariableBaselinePlausible rangeThreshold valueSensitive?
  1. MI: myocardial infarction; PE: pulmonary embolism; TP: true-positive result; FN: false-negative result.

No lymph node metastasis and no complications1.00.95–1.00.926No
Numbness, paresthesia, or prolonged pain0.900.80–0.99NANo
Blue dye anaphylaxis0.850.75–0.95NANo
Lymphedema0.750.65–0.850.870No
Lymph node metastases (TP)0.600.50–0.700.744No
MI, PE, or stroke0.500.40–0.60NANo
MI, PE, or stroke and lymph node metastases (TP)0.350.25–0.45NANo
Lymph node metastases (FN)0.400.30–0.500.272No
Lymph node metastases (TP) and MI, PE, or Stroke and lymphedema0.340.24–0.44NANo
MI, PE, or stroke and lymph node metastasis (TP) and lymphedema and blue dye anaphylaxis and numbness, paresthesia, or prolonged pain0.330.23–0.43NANo
Lymph node metastasis (FN) and MI, PE, or stroke0.290.19–0.39NANo
Lymph node metastasis (FN) and MI, PE, or stroke and lymphedema0.280.18–0.38NANo
Death for those with lymph node metastasis0.010.00–0.03NANo
Death for those without lymph node metastasis0.000.00–0.02NANo

The decision analysis performed for each of the seven tumor stages uses quality of life as the utility (or value) of the various possible health outcomes and probability estimates for the chance events from literature estimates. The “folded back” analysis provides the expected utility (EU) for each of the two procedures (SLN biopsy and ALND) and answers the question: “On average, which procedure is optimal given the values or utilities explicitly designated for each outcome?”

The goal of formal decision analysis is to provide an optimal decision that is consistent with stated values. That is, the consequences of each action are entered explicitly into the analysis. The optimal decision is that action that maximizes, on average, the quality of life of the patients. The second stage of formal decision analysis is to evaluate the sensitivity of this optimal decision to variations (or imprecision) of the various quantitative estimates for both the probabilities and the utilities.

RESULTS

EU

The EU of SLN biopsy was higher than the EU for ALND for patients in six of the seven tumor size groupings, given the stated point estimates for probabilities of the various chance outcomes and the utilities for the various health states. Table 4 presents a comparison of the EUs for SLN biopsy and ALND by tumor size. As the stage of the tumor progresses toward T3, the differences in average quality of life between SLN biopsy and ALND decrease. Therefore, ALND may be considered a reasonable alternative to SLN biopsy for patients with tumors larger than 4.0–5.0 cm and may be preferable to SLN biopsy for patients with T3 tumors. It is also evident that patients with smaller tumors have a better advantage in quality of life, on average, with SLN biopsy compared with ALND. The differences seen in EUs between SLN biopsy and ALND are very small for tumors larger than 4.0 cm.

Table 4. EU of SLN Biopsy Compared with ALND
Tumor stageSLN EUALND EUOptimal decision
  1. EU: expected utility; SLN: sentinel lymph node; ALND = axillary lymph node dissection.

T1a0.9330.891SLN biopsy
T1b0.9070.869SLN biopsy
T1c0.8580.828SLN biopsy
T2 (> 2.0–3.0 cm)0.8010.781SLN biopsy
T2 (> 3.0–4.0 cm)0.7430.732SLN biopsy
T2 (> 4.0–5.0 cm)0.7080.702SLN biopsy/toss up
T30.6640.665ALND/toss up

Sensitivity Analysis

Threshold (sensitivity) analyses were performed to assess the robustness of the model. The value of each probability and utility in the model was varied within a plausible range to determine if there was a point at which the preferred surgical option changed from SLN biopsy to ALND. In this way, the degree of uncertainty associated with an analytic result is evaluated.

For all tumors up to 4.0 cm, there were no thresholds within the plausible range of probabilities and utilities where the optimal clinical strategy changed to favor ALND.The optimal decision in favor of SLN biopsy in each analysis was robust to changes in the estimates. A focus on tumors larger than 4.0–5.0 cm in diameter, which represents the upper range of T2 tumors for which SLN biopsy is typically considered, illustrates this robustness (Table 5 for probabilities and Table 3 for utilities).

Table 5. Threshold Values for T2 (> 4.0–5.0 cm) Stage Tumor: Probabilities
VariableBaselinePlausible rangeThreshold valueSensitive?
  1. SLN: sentinel lymph node; MI: myocardial infarction; PE: pulmonary embolism; ALND: axillary lymph node dissection.

Anesthesia/operative mortality0.00010.00001–0.0010.018No
Percent lymph node metastases (prevalence)0.6670.567–0.7670.756Yes
SLN identification rate0.930.88–0.980.075No
SLN false-negative rate0.080.03–0.200.131Yes
SLN true-positive rate1.000.95–1.000.678No
SLN complications: MI, PE, or stroke0.00920.005–0.020.050No
Blue dye anaphylaxis0.010.001–0.0250.108No
SLN complications: numbness, paresthesia, or prolonged pain0.040.001–0.090.233No
SLN chronic lymphedema0.010.001–0.020.087No
ALND complications: numbness, paresthesia, or prolonged pain0.150.001–0.21NANo
ALND chronic lymphedema0.200.06–0.300.121Yes

The point estimate for the prevalence of lymph node metastases in T2 (> 4.0–5.0 cm) tumors was 0.667, or 66.7%. Sensitivity analysis reveals that the threshold for changing the optimal decision to ALND is 0.756. If the prevalence of axillary lymph node metastases for T2 (> 4.0–5.0 cm) tumors were known to be as high as 75.6%, then the optimal decision would be to go straight to ALND. The point estimate for the proportion of false-negatives in this stage tumor was 0.08 or 8%. Our sensitivity analysis also reveals that any surgeon reporting a false-negative rate of 13.1% or higher should forgo SLN biopsy. The point estimate for the identification rate of the SLN is 0.93 or 93%, and the threshold is 0.075. In other words, if surgeons could only find the SLN 7.5% of the time, then the optimal decision would be ALND.

The thresholds for anesthesia/surgical mortality; anesthesia/surgical morbidity of MI, PE, or stroke; blue dye (nonfatal) anaphylaxis; complications of numbness, paresthesia, or prolonged pain; and the true-positive rate (assumed 1.0 or 100%) were either nonexistent or well beyond any plausible range (Table 5). The crossover threshold for chronic lymphedema associated with ALND is 0.121 or 12.1%. If the rate of chronic lymphedema following ALND for T2 (4.1–5.0 cm) tumors is less than 12.1%, ALND may be a reasonable option. The decision to choose SLN biopsy over ALND in T2 (4.1–5.0 cm) tumors was not sensitive to any changes in the utility estimates (Table 3).

One of the more crucial estimates in these analyses is that of prevalence of axillary metastases. The estimates from our own data range from 0.123 (12.3%) in T1a tumors to 0.773 (77.3%) in T3 tumors (Table 1). In each of the six tumor stage analyses in which SLN biopsy was the preferred choice, a sensitivity analysis was performed to estimate the threshold at which the prevalence of axillary metastases would be high enough to support a change from SLN biopsy to ALND as the optimal procedure. For each tumor stage, the prevalence would have to be above 0.756 (75.6%) to warrant a change in decision. If the prevalence of lymph node involvement were that high, SLN biopsy should be avoided to prevent the high rate of follow-up surgeries for completion ALND. The available data would suggest that even for T2 (> 4.0–5.0 cm) tumors, lymph node involvement in 75% of patients is unrealistic. The rate of lymph node metastasis in T3 tumors, however, does approximate 75%.

In summary, the decision to choose SLN biopsy over ALND is a robust decision, given one, two, and three-way sensitivity analyses. This result is similar in estimation and identical in inference for all T1 and T2 tumors. The optimal decision for SLN biopsy was robust across the plausible ranges of probability estimates and quality of life assessments. For T3 tumors, our analyses would suggest that SLN biopsy and ALND would, on average, result in a very similar quality of life for patients.

DISCUSSION

ALND has been performed routinely in patients with invasive breast carcinoma for over 100 years. The status of the axillary lymph nodes remains the most powerful predictor of survival and recurrence in breast carcinoma patients. SLN biopsy is becoming increasingly accepted as a minimally invasive alternative to ALND for axillary lymph node staging. However, certain limitations of the technique, including failure to identify the SLN in some patients, false-negative results, and the possibility of a second operation for completion ALND when the SLN is positive, have tempered enthusiasm for this procedure at some institutions.

In the absence of randomized controlled trials to guide our decisions whether or not to perform SLN biopsy, decision analysis provides an excellent alternative to help answer this question. A wealth of prospective nonrandomized data are currently available, but the only trial actually studying the risks and benefits of SLN biopsy compared with standard ALND has just begun.42 Even the ongoing National Surgical Adjuvant Breast and Bowel Project (NSABP) Trial B-32, in which patients are randomized to SLN biopsy plus ALND versus SLN biopsy alone (with ALND for patients who have positive SLN) will not provide new insight on this issue because all patients receive SLN biopsy as part of their treatment. Therefore, in the absence of level I evidence from randomized, controlled trials, we must rely on other ways of evaluating the existing data to make treatment decisions. Decision analysis is uniquely suited to this task.

The results of the current formal decision analysis are compatible with the hypothesis that SLN biopsy, when performed using optimal technique with acceptable outcomes, is the preferred choice for maximizing quality of life for patients with breast carcinoma. SLN biopsy maximizes the EU of patients with T1 or T2 breast carcinoma undergoing a procedure for axillary lymph node staging. For patients diagnosed with larger tumors, the advantage of SLN biopsy is less distinct than for patients with smaller tumors. For patients with T3 tumors, SLN biopsy and ALND offer similar average outcomes (EUs). This is a noteworthy result given the rigorous scrutiny applied to SLN biopsy via the low utilities assigned to outcomes such as blue dye (nonfatal) anaphylactic reaction and false-negative results.

Complications

ALND is associated with considerable morbidity, including pain, numbness, impaired arm mobility, seromas, hematomas, and most importantly, chronic lymphedema. However, many maintain that routine ALND should be continued because the value of the information obtained from pathologic lymph node analysis outweighs the associated morbidity. The morbidity of ALND may be decreased by limiting the extent of dissection. Yet, ALND limited to level I lymph nodes or lymph node sampling is not sensitive or accurate in staging the axilla.43

The most distressing and chronic complication is lymphedema and its resultant deformity, discomfort, and disability. Although quantification of the impact that lymphedema has on patients is difficult, Velanovich and Szymanski38 found a demonstrable decrease in quality of life. Lymphedema is related directly to the removal of axillary lymph nodes. Various methods exist to quantitate lymphedema, the traditional method being the comparison of upper and lower arm circumferences. The most commonly seen definition is a 2-cm or greater difference between arm circumferences. Petrek and Heelan25 performed an extensive review of the literature. They cited 35 relevant reports with lymphedema complications in 6–30% of patients, depending on the length of follow-up, the type of surgery for the primary tumor, and the definition of lymphedema. Overall, they noted a 20% incidence of objective lymphedema. Clinically evident lymphedema that interferes substantially with patients' quality of life likely has a lower incidence, but may still be as high as 15%.

Although lymphedema is the most dreaded complication of ALND, other complications are also possible. Roses et al.16 reported nonlymphedema complications including decreased upper extremity range of motion, numbness/paresthesias, seroma, pain, wound infection, and nerve injury. A significant portion of patients reported numbness or paresthesias in the early postoperative period, but only 18% had unresolved complaints at time of follow-up. Numerous other studies on the morbidity of ALND have been performed.17–24 The incidence of morbid events varies in the literature and is likely the result of differing definitions of what actually constitutes a complication, as well as different methods and length of follow-up.

Because SLN biopsy for breast carcinoma is a relatively new procedure, there are little data by which to quantify the associated complication rate. A comparison among 70 patients undergoing either SLN biopsy or ALND found negligible morbidity in the SLN group when the following factors were assessed: numbness, pain, arm mobility, arm stiffness, subjective lymphedema, arm strength, and impact on daily living.21 Sener et al.28 indicated that the incidence of lymphedema following SLN biopsy may be as high as 3%. Because of the relative paucity of complications reported for SLN biopsy in breast carcinoma patients, we also used some data from the melanoma literature for this decision analysis.26, 27

One potential complication seen with SLN biopsy but not ALND is the allergic reactions to the isosulfan blue dye used for lymphatic mapping. According to the manufacturer's package insert,12 mild allergic reactions with localized swelling at the site of administration and mild pruritis of the hands, abdomen, and neck occur with a frequency of 1.5%. Leong et al.13 reported three cases of anaphylaxis following intradermal injection with isosulfan blue for melanoma (an incidence of approximately 1%). Others14, 15 have published case reports of anaphylaxis with the use of peritumoral blue dye injection for breast carcinoma lymphatic mapping in two patients and with the subcutaneous injection of isosulfan blue for lymphangiography in one patient with squamous cell carcinoma of the cervix.

The level of radiation exposure associated with radioactive colloid for lymphatic mapping is extremely low. Technetium sulfur colloid has a half-life of approximately 6 hours. Breast (injection site) and lypmh node specimens from breast carcinoma patients are typically nonradioactive after 68 and 33 hours, respectively.44 Exposure levels are so low that monitoring devices, shielding precautions, and personal dosimetry are not required during the procedure.45

SLN Identification Rate and False-Negative Rate

The majority of patients with early breast carcinoma do not have axillary lymph node involvement. SLN biopsy spares patients with negative lymph nodes the need for ALND. There is also evidence that SLN biopsy may be more sensitive for detection of small lymph node metastases. A more thorough histologic analysis of the lymph node(s) most likely to contain metastases is performed compared with the standard analysis of ALND specimens, in which only one or two histologic sections from the center of each lymph node are examined. An important caveat to SLN biopsy is the potential for false-negative results. At what point does the chance of incorrect staging from a SLN procedure outweigh the potential morbidity of a level I/II ALND?

Based on a recent review of the current literature, the SLN false-negative rate ranges from 0% to 16.7%.33 In our large multicenter study33 comprised mostly of surgeons just beginning to perform this procedure, the false-negative rate was 7.9%. We believe that this number provides an accurate average value for inexperienced surgeons performing this technique. However, the false-negative rate declines sharply when the technique is perfomed by experienced surgeons.35, 36, 46, 47 Therefore, it is likely that the false-negative rate will be less than or equal to 5% for experienced surgeons, which is far below the threshold estimate of 13.1% in this study.

Decision-Making

An argument has been made that ALND may not be indicated for patients with early-stage breast carcinoma. However, even patients with T1a tumors have a 3–29% incidence of developing lymph node disease,3, 48 although the most commonly cited figures range from 10% to 20%.2, 4, 9, 49, 50 In fact, these patients are the optimal candidates for a minimally invasive procedure that yields highly accurate staging. A decision analysis performed by Parmigiani et al.6 suggested that routine ALND for patients with early-stage breast carcinoma may not be necessary. It is important to note that their study assumed that axillary radiation provides equally effective treatment as ALND in patients who did not undergo the procedure. However, axillary radiation results in rates of lymphedema comparable to rates of lymphedema for ALND and incurs other complications such as brachial plexopathy not usually seen with a level I and II axillary dissection. Importantly, no axillary therapy is needed in the majority of breast carcinoma patients because 70% of patients have no evidence of axillary lymph node metastasis.1, 22, 29, 30 Axillary radiation has not been accepted widely as an alternative to ALND because it is applied to all patients and ignores lymph node staging information, which may alter adjuvant therapy decisions for many patients.

The decision to perform SLN biopsy or ALND may differ based on the stage of the primary tumor. We performed separate decision analyses for seven tumor stages: T1a, T1b, T1c, T2 (> 2.0–3.0 cm), T2 (> 3.0–4.0 cm), T2 (> 4.0–5.0 cm), and T3 tumors. The threshold analyses indicate that the optimal decision in favor of SLN biopsy is robust. The decision analysis was sensitive to the probability of lymphedema following ALND for tumors larger than 4.1 cm. If ALND could be performed with a lymphedema rate of less than 12.1%, ALND may be preferred over SLN biopsy. The ALND lymphedema threshold result for tumors 3.1–4.0 cm was 0.077 (7.7%) and 0.031 (3.1%) for tumors 2.1–3.0 cm (data not shown), indicating that for tumors smaller than 4.0 cm, SLN biopsy is the optimal decision. Not surprisingly, the average advantage in postprocedural quality of life for SLN biopsy was smallest for patients with larger tumors (and hence a higher prevalence of lymph node metastases). Although our institutional bias has been to perform ALND for patients with T3 tumors, this decision analysis has prompted us to consider SLN biopsy more frequently for patients with clinical stage T3N0 breast carcinoma. Indeed, SLN biopsy may be appropriate for patients with T3 tumors who are highly motivated to avoid ALND. However, this analysis also indicates that it is very reasonable to offer ALND rather than SLN biopsy for patients with a high risk of lymph node metastasis (> 4.0 cm tumors).

Velanovich5 developed a decision-making model for T1 tumors and found an advantage in quality-adjusted life expectancy (QALE) for avoiding ALND until the probability of axillary metastases is greater than15%, although the difference in QALE was small. Jackson et al.4 also reported a decision analysis on the effect of avoiding ALND in low-risk women with invasive breast carcinoma. However, the option of SLN biopsy was not considered in either analysis. The ability to obtain axillary lymph node staging information with a procedure that carries the morbidity of a simple lymph node biopsy has changed the risk-benefit analysis. It is unrealistic to expect that the morbidity of a more extensive procedure such as ALND would ever be considered comparable to that of SLN biopsy. The decreased morbidity associated with SLN biopsy is an essential component of its attractiveness.

The debate regarding the therapeutic value of ALND is ongoing. The best randomized trials show no survival advantage for ALND. However, a metaanalysis reported by Orr51 suggested a 5.4% survival advantage in favor of ALND. This unresolved controversy really does not substantially affect the current analysis. We assumed that all patients with positive SLN would receive completion ALND. The value of completion ALND following a positive SLN biopsy is currently being addressed by the American College of Surgeons Oncology Group Trial Z0011.

In many respects, the analysis was intentionally biased against SLN biopsy as the preferred option. For example, one of our main assumptions was that all procedures were performed under general anesthesia and that intraoperative pathologic analysis of the SLN was not performed. Therefore, a positive SLN result necessitates a second operation with all the attendant anesthesia risks. This decision analysis was robust enough to withstand two separate general anesthetics if SLN biopsy and completion ALND are performed. SLN biopsy can only become more favored if the procedure is performed using locoregional anesthesia. Furthermore, the increasing use of intraoperative frozen section analysis at many institutions may obviate a second surgical procedure for the majority of patients with a positive SLN. We also deliberately assigned a low EU for a false-negative SLN biopsy result, assuming that a false-negative result would be highly detrimental in most cases. Of course, for many patients with tumors latger than 1.0 cm, the adjuvant therapy decisions may not change even with a positive lymph node. The impact of false-negative results on recurrence and survival in breast carcinoma patients is the subject of ongoing clinical trials.

Utilities

The fact that decisions for T1 and T2 tumors were not sensitive to plausible ranges in variation of utilities may surprise many. In fact, the very idea of subjective values or utilities concerns many critics of formal decision analysis. Certainly, this formal decision analysis combines objective estimates of probability with subjective estimates of the quality of life of numerous health states. For most of the utilities, a threshold was never reached in which the decision would change from performing an SLN biopsy to performing an ALND as the optimal strategy. This means that even though the assignment of utilities was subjective in most cases, the results would have been the same even if we had chosen drastically different utilities.

One reason is that many of the actual possible health outcome states are extremely rare. For example, only a small proportion of patients experience multiple negative side effects, such as MI, PE, or stroke; chronic lymphedema; blue dye anaphylaxis; numbness, paresthesia, or prolonged pain; and/or lymph node metastases. Second, many of these complications are possible for both SLN and ALND patients. Even huge variations in the utility estimates do not bring about a change in optimal decisions. It is the very strength of formal decision analysis that a detailed decision tree, consisting of scientific estimates of the probability of each event coupled with explicit (albeit subjective) values expressing the desirability of the consequences of each event, results in a robust optimal decision.

Limitations

There are some practical criticisms of this decision model. When in doubt, we chose assumptions that would bias the analysis against SLN biopsy. Financial factors were not taken into account in these analyses, although SLN biopsy does not appear to be associated with significantly higher economic costs.52

The advantage in postprocedural quality of life, on average, for SLN biopsy is based on the context of the analysis. It is important to note that the “on average” perspective of EU is exactly that; we cannot say what might be best for any one patient, unless, of course, we assess that patient's values toward risk and the pertinent health outcomes. Generalization of these results is related to accepting the structure of the analysis. In this study, a number of complications (MI, PE, stroke, blue dye anaphylaxis, lymphedema) were considered pertinent, although other more minor side effects were excluded from explicit analysis. This study focused on the complications and outcomes that seemed most clinically relevant, given the context of optimal staging of lymph node metastases. A different research architecture could produce different results. However, given the structure of these decision problems, plausible variations both in the estimates of the probabilities and utilities had no effect on the final outcome.

The information derived from this decision analysis is potentially very helpful to both clinicians and their patients. Patients may vary in their beliefs and values, and as such, utilities may differ among patients. Utilities are a quantitative measure of patients' attitudes toward the risks and benefits of treatment. Although decision analysis provides a comprehensive view of the decision-making process for a population of patients, it is important to take individual treatment goals and options into account. Ultimately, the choice to undergo SLN biopsy versus ALND is a personal one and individual treatment goals and options must be considered.

An important benefit of SLN biopsy is that it can avoid the complications of ALND in the majority of breast carcinoma patients. However, there is the risk of a false- negative SLN result, potentially leading to incorrect staging and decisions about adjuvant therapy. Our decision analysis suggests that SLN biopsy has a greater EU than ALND for T1 and T2 tumors with clinically negative lymph nodes. However, the two procedures are virtually equivalent for the subgroup of T2 tumors larger than 4.0 cm. The EU of ALND slightly exceeded that of SLN biopsy for patients with T3 tumors. Therefore, for tumors smaller than 4.0 cm, SLN biopsy is the preferred approach in most cases. For tumors larger than 4.0 cm, SLN biopsy and ALND may be considered nearly equivalent options. In such cases, individual patient preferences should be taken into account, considering the prevalence of lymph node metastases, the implications of a false-negative result, and the potential morbidity of ALND.

Ancillary