| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, Institute for Technology Assessment, Massachusetts
General Hospital, 101 Merrimac St., 10th Fl., Boston, MA 02114.
2 Division of Abdominal Imaging and Interventional Radiology, Massachusetts
General Hospital, Boston, MA.
3 Surgical Oncology, Massachusetts General Hospital, Boston, MA.
4 Gastrointestinal Unit, Institute for Technology Assessment, Massachusetts
General Hospital, Boston, MA.
Received February 18, 2008;
accepted after revision June 2, 2008.
Address correspondence to P. V. Pandharipande.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. A decision-analytic Markov Model was developed to estimate quality-adjusted life expectancy and lifetime costs among 61-year-old women with clinically node-negative early breast cancer. Three axillary staging strategies were compared: MR lymphangiography alone, combined MR lymphangiography–SLN biopsy, and SLN biopsy alone. The model incorporated treatment decisions, outcome, and costs consequent to axillary staging results. An incremental cost-effectiveness analysis was performed to compare strategies. The effect of changes in key parameters on results was addressed in sensitivity analysis.
RESULTS. In the base-case analysis, combined MR lymphangiography–SLN biopsy was associated with the highest quality-adjusted life expectancy (13.970 years) and cost ($63,582), followed by SLN biopsy alone (13.958 years, $62,462) and MR lymphangiography alone (13.957 years, $61,605). MR lymphangiography–SLN biopsy and SLN biopsy both were associated with higher life expectancy and cost relative to those of MR lymphangiography. MR lymphangiography–SLN biopsy, however, was associated with greater overall life expectancy and greater added life expectancy per dollar than was SLN biopsy. SLN biopsy alone therefore was not considered cost-effective, but MR lymphangiography and MR lymphangiography–SLN biopsy remained competing choices. Preference of MR lymphangiography strategies was most dependent on the sensitivity of MR lymphangiography and SLN biopsy and on the quality-of-life consequences of SLN biopsy and axillary lymph node dissection, but otherwise was stable across most parameter ranges tested.
CONCLUSION. From a cost-effectiveness perspective, MR lymphangiography strategies for axillary staging of early breast cancer are preferred over SLN biopsy alone. The sensitivity of MR lymphangiography is a critical determinant of the cost-effectiveness of MR lymphangiography strategies and merits further investigation in the care of patients with early breast cancer.
Keywords: breast cancer • cost-effectiveness analysis • MR lymphangiography • sentinel lymph node biopsy
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Axillary node status can be a critical determinant of treatment recommendations and survival among patients with preoperative stage I–II breast cancer without palpable adenopathy [7]. In centers of expertise, SLN biopsy, because of its minimally invasive nature, has emerged as the standard of care for axillary staging in this population [2–6]. Even so, SLN biopsy has the disadvantages of a false-negative rate of 8–9% [4, 8]; risk of complications such as seroma, neuropathy, and lymphedema; and high cost driven by intraoperative and postoperative histologic findings [9, 10].
MR lymphangiography is a high-resolution imaging technique in which differentiation of benign and malignant lymph nodes is based on differential uptake of lymphotropic nanoparticles [11–16]. Compared with SLN biopsy, MR lymphangiography is less invasive and expensive and therefore may have a favorable role in axillary staging of early breast cancer. The performance characteristics of axillary MR lymphangiography in the care of breast cancer patients have been reported in four studies [11, 14–16], three of which predominantly included patients with category T1 and T2 lesions [14–16]. MR lymphangiography had a sensitivity range of 82% (nine of 11 patients) [15] to 100% (six of six patients) [14] and a specificity range of 80% (four of five patients) [16] to 100% (16 of 16 patients) [14] with axillary lymph node dissection as the standard of reference.
The purpose of this study was to com pare MR lymphangiography axillary staging strategies with use of SLN biopsy alone from perspectives of effectiveness, cost, and cost-effectiveness in a hypothetical cohort of postmenopausal women with a common profile of early clinically node-negative breast cancer. A decision-analytic Markov Model was developed to estimate and compare quality-adjusted life expectancy and costs consequent to three potential axillary staging strategies: MR lymphangiography alone, combined MR lymphangiography and SLN biopsy, and SLN biopsy alone. Standard incremental cost-effectiveness analysis was used to evaluate strategies from a cost-effectiveness perspective [17].
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Cost-effectiveness analysis was performed under guidelines issued by the Panel on Cost-Effectiveness in Health and Medicine [17]. The analysis was performed with a quasisocietal perspective. We included costs of disease management regardless of who incurred them, but we did not include certain costs to the patient, such as time costs. Cost values used in the analysis were specific to the United States.
Quality-adjusted life expectancy and lifetime costs were compared for three axillary staging strategies: MR lymphangiography alone, combined MR lymphangiography and SLN biopsy, and SLN biopsy alone. A no-staging strategy also was included in which patients underwent breast conservation surgery without axillary staging. This strategy was included only for comparison and was not considered clinically reasonable. Quality-adjusted life expectancy was measured in quality-adjusted life years (QALYs) with use of health state–specific utilities from 0 (equivalent to death) to 1 (perfect health) to weight yearly quality of life [17]. An annual discount rate of 3% was applied to future costs and quality-adjusted life expectancy [17].
Strategies were compared in a standard incremental cost-effectiveness
analysis [17]. Strategies with
fewer QALYs and greater costs than others were strongly dominated and
eliminated. The other strategies were aligned in order of increasing
quality-adjusted life expectancy and costs, and incremental cost-effectiveness
ratios were computed (
cost / QALYs). Strategies were further
eliminated by weak dominance if they had an associated incremental
cost-effectiveness ratio higher than the next more effective and expensive
strategy, and incremental cost-effectiveness ratios were recomputed. The
remaining strategy list constituted the primary analysis result.
Decision Tree and Model Structure
A decision-analytic Markov Model was constructed to estimate
quality-adjusted life expectancy and lifetime costs consequent to each staging
strategy. The primary (base-case) analysis incorporated best available model
input estimates (Tables 1,
2,
3). Stability of results over
changes in model estimates was evaluated in secondary (sensitivity) analysis.
The model was constructed with the TreeAge Pro 2007 program (TreeAge
Software). A decision tree schematic is shown in
Figure 1.
|
|
|
|
Cohort Characteristics
The selected cohort age (61 years) was the median age of breast cancer
diagnosis in the United States
[18]. It was assumed that
patients had chosen breast conservation surgery with radiation therapy for
primary tumor treatment. Primary cancers were assigned a common profile of
features: 1–2 cm (category T1c), grade 1, estrogen receptor/progesterone
receptor–positive, and HER2/neu negative. It was assumed that
patients had clinically node-negative status (no suspicious palpable axillary
adenopathy). The prevalence of axillary malig nancy (28%) was specific to 1-
to 2-cm tumors [19].
In our experience and in keeping with international consensus guidelines, axillary malig nancy in patients with the described tumor fea tures would strongly favor use of systemic chemotherapy and hormonal therapy, and patients without axillary malignancy would undergo hormonal therapy alone [7]. Because in our analysis MR lymphangiography is assumed to have slightly lower sensitivity in the detection of axillary malignancy than does SLN biopsy [4, 8, 14–16], the selected study population was biased against MR lymphangiography. Specifically, false-negative axillary staging results were penalized in two ways: first, systemic chemo therapy was forgone, and second, axillary malig nancy was not resected. For other tumor profiles, such as estrogen receptor/progesterone receptor–negative tumors, axillary status may not sub stantially influence the choice of adjuvant therapy [7]. In our analysis, incorporated tumor features ensured both described penalties for false-negative staging results.
Strategies Modeled
For MR lymphangiography, it was assumed that patients underwent both
unenhanced MRI and injection of a lymphotropic nanoparticle contrast agent 24
hours before contrast-enhanced MRI
[12–15].
For combined MR lymphangiography–SLN biopsy, preoperative MR lym
phangiog raphy was performed first. Abnormal findings were followed by
axillary dissection and normal findings by SLN biopsy. Breast con servation
surgery, SLN biopsy, and axillary dissection were generally assumed to be con
current. For SLN biopsy, it was assumed that frozen-section,
permanent-section, and immuno histochemical analyses were performed
[9].
Positive staging results were followed by two-level axillary dissection. For a subset of patients (41%) with axillary malignancy detected at SLN biopsy, malignancy was assumed to be occult during intraoperative analysis but detectable 1–2 days after surgery at permanent-section or immunohistochemical analysis [9]. It was assumed that only 50% of these patients returned to the operating room for axillary dissection, according to institutional practice and because of the questionable added value of axillary dissection in this setting [20].
Key Model Input Parameters
Strategy test performance characteristics— Model probability
estimates are included in Table
1. To date, the performance characteristics of MR lymphangiography
for axillary staging have been reported in three studies
[14–16]
with a pre dominance of patients with clinical category T1 (
2 cm) and T2
(2–5 cm) tumors. In all three studies, axillary dissection and
subsequent histo logic analysis were used as the standard of reference. Two
studies [14,
15] were prospective with
consecutively enrolled patients and with MR lymphangiography readers blinded
to histo patho logic results. The third study
[16] was prospective, but
consecutive patient enrollment and blinding of readers of MR lymphangiography
to histologic findings were not specifically reported in the study methods. In
these studies, the per-patient sensitivity and specificity of MR
lymphangiography in detection of axillary malignancy were 82% (nine of 11) and
100% (seven of seven) [15],
100% (five of five) and 80% (four of five)
[16], and 100% (six of six)
and 100% (16 of 16) [14].
We planned to incorporate the sensitivity and specificity of MR lymphangiography in our model through a function defined by a summary receiver operating characteristic (ROC) curve analysis [21] if sensitivity and specificity in these studies were significantly different from one another when the Fisher's exact test was applied (p < 0.05). The rationale for this practice was that if readers in each study [14–16] operated at different positivity thresholds, then a summary ROC curve could provide added information about the trade-offs between the sensitivity and specificity of axillary MR lymphangiography. However, the Fisher's exact test results did not show that sensitivity and specificity differed across the three studies (p = 0.71, p = 0.18). Therefore, a summary ROC curve was not generated, and pooled MR lymphangiography sensitivity and specificity values (91%, 20 of 22; 96%, 27 of 28) were used for the base-case analysis [14–16].
A fourth study [11] of the performance of MR lymphangiography in axillary staging of breast cancer was identified, but the investigators in that study specifically excluded patients with stage I breast cancer who were appropriate candidates for SLN biopsy. The study therefore was included only in sensitivity analysis (see Sensitivity Analysis). The study was prospective with con secutively enrolled patients, and MR lym phangio graphy readers were blinded to histopathologic results. Axillary dissection was used as the standard of reference [11].
In a series specific to 1- to 2-cm breast tumors [8], the sensitivity of SLN biopsy was 92% (572 of 624 patients). A specificity of 100% was assumed for SLN biopsy, indicating that false-positive results could not occur at SLN biopsy. For combined MR lymphangiography–SLN biopsy, conditional independence between MR lymphangiography and SLN biopsy was assumed. That is, the sensitivity of SLN biopsy after a false-negative MR lymphangiography finding was estimated to be the same as the sensitivity in any other setting (92%) [8].
Markov Model—After breast conservation surgery and axillary staging, the lifetime course of the cohort was modeled with a Markov Model that had a 1-year cycle length. We constructed this Markov Model in keeping with widely accepted methods for Markov modeling in medicine [22]. Four health states were defined (Fig. 2): receiving adjuvant therapy, postadjuvant therapy, metastatic breast cancer, and death.
|
Markov transition probabilities were estimated as follows. Cancer-specific mortality estimates for each axillary staging result (true-positive, true-negative, false-negative) were derived from a validated Web-based prediction tool for breast cancer–specific survival (Adjuvant! Online, see Appendix 1) [24, 26, 27]. Specifically, this tool was used to obtain predictions of cancer-specific survival as a function of features of the primary tumor, axillary node status, and adjuvant therapy chosen. Survival estimates were converted to mortality estimates for use in our analysis (Table 2). The yearly probability of death (0.6) among patients with metastatic disease was derived from the literature [28]. Given test result–dependent cancer-specific mortality estimates [24, 26, 27] and a yearly mortality estimate for patients with metastatic disease [28], transition probabilities from postoperative to metastatic disease states could be calculated through calibration for each test result possibility.
We made two further assumptions that merit specific mention. First, we assumed that all patients with positive MR lymphangiography or SLN biopsy findings underwent axillary dissection, at which time true nodal status was determined. Patients with false-positive MR lymphangiography findings underwent axillary dissection but not further adjuvant chemotherapy. The prob ability of transition to metastatic disease among these patients was therefore identical to that of patients with true-negative findings. Second, a 2% absolute increase in 10-year cancer-specific mortality was incorporated for false-negative axillary staging results to account for lack of removal of malignant lymph nodes. Using a meta-analytic approach, Orr [29] estimated a 5.4% reduction in survival rate if prophylactic axillary node dissection were to be omitted in the treatment of patients with clinically node-negative cancer. This estimate, however, was not specific to a time interval and was con sidered too high for a more contemporary patient population with small tumors, as suggested by Orr.
In a separate decision analysis, Orr et al. [30] concluded that there was no benefit to axillary dissection in the treatment of patients with node-positive disease who did not undergo adjuvant chemotherapy. As a compromise between these extremes, we included a 2% cancer-specific increase in 10-year cancer-specific mortality for lack of removal of malig nant axillary lymph nodes. In sensitivity analysis, this value was varied from 0% to 16%, which includes the highest estimated penalty among trials used in Orr's meta-analysis [29], to determine the effect on model results.
Health-state utilities were derived from the literature (Table 1). Postoperative utilities were specific to adjuvant therapy course. Utilities applied to time spent receiving adjuvant chemotherapy [31] or tamoxifen [32] and to time spent with metastatic disease [33] were obtained from studies in which the visual analog scale elicitation method among health care professionals was used. For time spent in the postadjuvant therapy health state, a utility was applied from a study in which the standard-gamble elicitation method was used among both patients and health care professionals [34].
Because of a lack of pertinent published quality-of-life data, the effects of reduced quality of life specific to the staging process, such as lymphedema after SLN biopsy or axillary dissection, could not be reliably incorporated in the base-case analysis. Although it is known that the relative morbidity of axillary dissection is greater than that of SLN biopsy and that MR lym ph angiography is likely to be more easily tolerated than SLN biopsy, to our knowledge, the relative magnitude of these differences has not been explicitly quantified in the literature in the form of utilities. The effect of long-term reduction in quality of life due to SLN biopsy or axillary lymph node dissection was assessed in sensitivity analysis to determine whether staging strategy preference, from a cost-effectiveness perspective, was affected by chronic adverse effects of these procedures (see Sensitivity Analysis).
Costs—Model costs are included in Table 3. All costs were converted to 2006 dollars with use of the Medical Care component of the U.S. Consumer Price Index. The cost of MR lymph angiography was not available because MR lymphangiography is still performed primarily in investigational settings. The cost of MR lymph angiography was estimated by multiplying the cost of unilateral breast MRI with gadolinium administration (Current Procedural Terminology 2006 code 76093) by 1.5 because a patient would be imaged twice but in shorter sessions than two full MRI examinations with IV contrast en hancement [12–16].
Shared costs of breast conservation surgery with or without SLN biopsy and axillary dissection were taken from a study by Gemignani et al. [9]. An isolated breast conservation surgery cost was not reported explicitly in that study. This cost was estimated based on reported subcosting data for shared procedures [9]. Radiation therapy costs were derived from the literature [35].
Tamoxifen costs were based on a dose of 20 mg/d for 5 years [36]. Systemic chemotherapy included costs of four cycles of paclitaxel, doxorubicin, and cyclophosphamide. Infusion and physician visit costs were included [35, 36]. Yearly health-state costs were derived from the literature [37, 38].
Sensitivity Analysis
The effect on results of uncertainty in model assumptions and parameters
was evaluated in one-way and two-way sensitivity analyses.
One-way sensitivity analysis—The tested param eter ranges for one-way analysis are shown in Tables 1, 2, 3. Results were considered stable if strategy ranking or dominance remained constant across the range considered. If strategy ranking or dominance changed, results were considered sensitive to that parameter. To test result stability under the assumption of conditional independence (between MR lymphangiography and SLN biopsy) in the combined strategy, the sensitivity of SLN biopsy after false-negative findings at MR lymphangiography was varied from 50% of its base-case value to 100%.
MR lymphangiography data from a study that was initially excluded (because of exclusion of patients with stage I disease) [11] were incorporated into pooled sensitivity and specificity of MR lymphangiography to assess the effect on results. That study showed a sensitivity of 100% (23 of 23 patients) and specificity of 80% (eight of 10 patients). In combination with results of other studies [11, 14–16], MR lymphangiography had a pooled sensitivity of 96% (43 of 45 patients) and specificity of 92% (35 of 38 patients). These values were tested in sensitivity analysis.
Sensitivity analysis ranges for transition probabilities from nonmetastatic to metastatic disease states (Table 1) were calculated on the basis of ranges assigned to each test result–dependent mortality estimate derived from Adjuvant! Online (Table 2). Specifically, the upper and lower values of the transition probability ranges in Table 1 were derived through calibration described earlier (see Markov Model).
Sensitivity analysis was not performed on the isolated cost of SLN biopsy because the costs of SLN biopsy were modeled with shared costs of breast conservation surgery and axillary dissection [9]. The cost of each SLN biopsy setting considered (breast conservation surgery with SLN biopsy, breast conservation surgery with SLN biopsy and concomitant axillary dissection, and breast conservation surgery with SLN biopsy and axil lary dissection performed at a later time) was varied in sensitivity analysis.
Two-way sensitivity analysis—Two-way analysis was performed to capture important changes in strategy preference that could occur only if two parameters were varied simultaneously. The preferred strategy was defined as that which yielded the greatest number of QALYs without being strongly or weakly dominated. Three parameter sets were tested with two-way analysis. Two included MR lymphangiography sensitivity and MR lymphangiography specificity, and MR lymph angiography sensitivity and SLN biopsy sensitiv ity. In these two-way analyses, each individual para meter was varied across the ranges specified in Table 1.
Two-way analysis also was used to assess the effects on strategy preference of long-term quality-of-life reduction attributable to SLN biopsy or axillary lymph node dissection. For this analy sis, lifetime post–SLN biopsy and post–axillary lymph node dissection utilities, respectively, were introduced into the model and simultaneously varied from 0.95 to 1. Specifically, each original Markov health-state utility value in the decision model was multiplied by a utility value ranging from 0.95 to 1 if SLN biopsy or axillary lymph node dissection was performed. For patients who underwent both procedures, only the utility value for axillary lymph node dissection was introduced. In this way, a parameter space was generated that defined strategy preference for each possible combination of post–SLN biopsy and post–axillary lymph node dissection utilities.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
|
In the incremental cost-effectiveness analysis, SLN biopsy was eliminated by weak dominance because of its higher incremental cost-effectiveness ratio compared with that of combined MR lymphangiography–SLN biopsy, the next more expensive and higher-life-expectancy strategy. The incremental cost-effectiveness ratio of MR lymph angiography relative to the no-staging strategy was $37,244/QALY. The incremental cost-effectiveness ratio of combined MR lymphangiography–SLN biopsy compared with MR lymphangiography was $153,007/QALY.
Weak dominance of SLN biopsy is depicted in Figure 3, which shows an efficiency frontier within cost-effectiveness space. An efficiency frontier is a standard graphic means of delineating strategies preferred from a cost-effectiveness perspective. Diminishing marginal returns in life expectancy are seen for higher-cost strategies; strategies that are weakly dominated (SLN biopsy) fall below the frontier.
Sensitivity Analysis Results
One-way sensitivity analysis—In one-way sensitivity
analysis, dominance of SLN biopsy by MR lymphangiography–based
strategies was robust across most of the parameter ranges evaluated. This
result was sensitive to changes in MR lymphangiography and SLN biopsy
sensitivity. SLN biopsy, however, remained weakly or strongly dominated (by
combined MR lymphangiography–SLN biopsy and MR lymphangiography,
respectively) throughout most of the MR lym phangiography and SLN biopsy
sensitivity ranges tested (Fig.
4).
|
Incorporation of additional MR lymphangiography data (from the study that excluded cases of stage I cancer [11]) into pooled MR lymphangiography sensitivity and specificity resulted in strong dominance of SLN biopsy by MR lymphangiography. Specifically, MR lymphangiography had greater associated quality-adjusted life expectancy than SLN biopsy (13.964 compared with 13.958 years) at lower lifetime cost ($61,922 compared with $62,462).
Results were insensitive to costs of MR lymphangiography from 50% to 150% of the base-case estimate. Results were sensitive to shared costs of SLN biopsy, axillary dissection, and breast conservation surgery; however, SLN biopsy remained weakly dominated by MR lymphangiography strategies over most of each cost range tested (Table 5).
|
Results were insensitive to changes in all other parameters evaluated in one-way sensitivity analysis, including prevalence of axillary malignancy, probability of detection of axillary malignancy only at permanent-section or immunohistochemical analysis, probability of return (in this scenario) for completion axillary dissection, radiation and adjuvant therapy costs, test result–dependent mortality probability estimates derived from Adjuvant! Online [26], Markov transition probabilities, yearly health-state costs, and patient health-state utilities. Results were stable across the tested range of increased 10-year cancer-specific mortality attributed to lack of axillary dissection in the setting of axillary malignancy [29].
Two-way sensitivity analysis—Results were sensitive to the introduction of utilities in which the long-term quality-of-life consequences of SLN biopsy and axillary lymph node dissection were estimated (Fig. 5). For most post–SLN biopsy and post–axillary lymph node dissection utility estimates considered, MR lymphangiography alone was preferred. If both utilities were very high, combined MR lymphangiography–SLN biopsy was preferred. If post–SLN biopsy utilities were very high but post–axillary lymph node dissection utilities were relatively slightly lower, SLN biopsy alone was preferred.
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The quality-adjusted life expectancy gain from combined MR lymphangiography–SLN biopsy compared with SLN biopsy alone was 4 days, and the added cost was $1,120. A question may be raised whether this gain is too small to merit added expenditures of $1,120 per patient. In regard to this question, it is important to note that the presented gains and losses associated with each staging strategy in our analysis should be considered as averaged over a population. Some patients may not benefit from MR lymphangiography–SLN biopsy; others may benefit much more than a few days. In addition, although the MR lymphangiography–SLN biopsy strategy is more expensive, it enables a gain of more quality-adjusted life expectancy per dollar than does SLN biopsy and is therefore considered a higher-value strategy than SLN biopsy alone.
A further question may be raised about how our results, specifically the presented strategy incremental cost-effectiveness ratios and the weak dominance of SLN biopsy, can be used to aid health care funding decisions. Strategy preference at the societal level ultimately depends on societal willingness to pay. Commonly accepted willingness to pay thresholds in the United States range from $50,000/QALY to $100,000/QALY [39–42]. This range is linked to reported incremental cost-effectiveness ratios for hemodialysis, an intervention widely considered a benchmark for societal willingness to pay [39–42]. If this range were applied to the situation in our study, MR lymphangiography would be considered cost-effective compared with no axillary staging (incremental cost-effectiveness ratio, $37,244/QALY). The incremental cost-effectiveness ratio of combined MR lymphangiography–SLN biopsy relative to MR lymphangiography ($153,007/QALY) would exceed the hemodialysis range but is much lower than that of SLN biopsy relative to MR lymphangiography ($729,286/QALY) (Table 4). This result indicates that if U.S. societal expenditures beyond the cost of MR lymphangiography for axillary staging are feasible and accepted, as is currently the case, combined MR lymphangiography–SLN biopsy should be preferred over SLN biopsy alone. If not, MR lymphangiography alone should be preferred.
It is important to note that willingness-to-pay thresholds vary substantially across nations. The costs incorporated in our analysis and the discussion of our results are specific to a U.S. societal context. Nonetheless, the framework and analysis described provide the groundwork for similar analyses of the cost-effectiveness of MR lymphangiography across nations and economic states.
Another potential noninvasive strategy for axillary staging of early breast cancer is FDG PET. Using Canadian cost data, Sloka et al. [43] found PET cost-effective for axillary staging, but SLN biopsy and MR lymphangiography were not assessed. In a metaanalysis, Sloka et al. found the sensitivity and specificity of PET in axillary staging of breast cancer to be 83.4% and 85.6%, respectively. These values are lower than those for the pooled MR lymphangiography data [14–16] in our analysis, and PET is generally more expensive than MRI [44]. Under these assumptions, PET would be strongly dominated by MR lymphangiography. However, the number of studies of PET in breast cancer staging far exceeds the number of studies of MR lymphangiography [11, 14–16, 43, 45]. Further primary investigations of MR lym phangiography are clearly needed to better establish the performance characteristics of MR lymphangiography in this setting.
The primary limitations of this study relate to the requisite simplifications used to reduce a complex human disease into a biologically and clinically plausible model. These simplifications were addressed by selection of the most relevant patient and cohort characteristics, health states, and outcomes and by evaluation of model uncertainty through sensitivity analysis. Nonetheless, some important limitations merit specific mention.
In our Markov Model, the only end points considered were quality-adjusted life expectancy and lifetime costs. Estimation of test result–dependent transition probabilities from postoperative to metastatic disease states necessitated a simplified four-state model, which limited incorporation of complex postoperative courses, such as those of patients with recurrence of axillary malignancy. Use of this model may have reduced adverse effects that resulted from false-negative results in our base-case analysis. However, the possibility of worse long-term outcome due to less invasive staging was evaluated in sensitivity analysis. In that analysis, the yearly probability of development of metastatic disease in patients with occult axillary malignancy (due to false-negative staging) was varied substantially (Tables 1 and 2). Throughout the range considered, SLN biopsy remained weakly dominated by MR lymphangiography strategies.
In addition, relative quality-of-life differences specific to axillary staging methods (MR lymphangiography, SLN biopsy, combined MR lymphangiography–SLN biopsy, and axillary dissection) were not incorporated in the base-case analysis because of lack of pertinent utility data in the literature. Adverse effects of MR lymphangiography have been reported to occur in as many as 28% of patients (43 of 152). These effects include headache, back pain, vasodilatation, and hives and are typically minor temporary events [46]. Adverse effects of SLN biopsy, reported to occur in as many as 25% of patients after SLN biopsy alone and in 70% of patients undergoing subsequent axillary dissection, include both short-term and long-term sequelae, such as seroma, wound infection, neuropathy, and lymphedema [5, 47]. The effects of this limitation, that is, lack of incorporation of utilities specific to the process of axillary staging, are difficult to predict without primary patient data. This limitation resulted in two opposite effects on the favorability of MR lymphangiography strategies. Because MR lymphangiography is relatively noninvasive, short-term postprocedure quality of life would likely be higher compared with SLN biopsy or axillary dissection. However, false-positive MR lymphangiography findings lead to axillary dissection, which is associated with greater morbidity than is SLN biopsy alone.
We elected not to prospectively estimate and integrate utility penalties for each staging technique in our base-case analysis because such empiric estimation would have been overly subjective. However, in two-way sensitivity analysis, we assessed the effect of introducing and varying long-term quality-of-life penalties associated with SLN biopsy and axillary lymph node dissection, respectively. We found that from a cost-effectiveness perspective, preference of axillary staging strategies was highly sensitive to long-term quality-of-life sequelae attributable to these procedures (Fig. 5). Minor differences in estimated quality-of-life consequences had substantial implications for strategy choice. Further primary investigation of quality of life specific to each staging strategy is absolutely critical before policy-level decisions are made regarding routine use of MR lym phangiography in early breast cancer.
Quality adjustment of life expectancy was limited by the heterogeneity of studies from which health-state utilities were derived. Ideally, incorporated utilities should be from a single study in which the standard-gamble technique is used to elicit subject preferences [48, 49]. For a true societal perspective, subjects would ideally be representatives of the general community, not patients or health care professionals [17]. Further investigation of long-term posttreatment quality of life in a uniform contemporary series would be beneficial to further confirm model stability.
We did not incorporate use of aromatase inhibitors in our analysis. Although aromatase inhibitors are increasingly used as adjuvant hormonal therapy in the care of postmenopausal patients with hormone receptor–positive breast cancer [50], current predictions of long-term outcome in the management of early breast cancer are based primarily on the use of tamoxifen [51]. As of this writing, aromatase inhibitors had not been incorporated into the Adjuvant! Online survival prediction tool used in our analysis (for the 5-year postsurgical phase of adjuvant therapy). This limitation was considered reasonable because tamoxifen is a clinically viable option and because incorporation of aromatase inhibitors was unlikely to substantially affect our results. Specifically, in our model, all patients were treated with adjuvant hormonal therapy regardless of findings at MR lymphangiography or SLN biopsy. Therefore, substitution of aromatase inhibitors for tamoxifen would have resulted in negligible effects on preference of staging strategy in the incremental cost-effectiveness analysis, even when potential increases in efficacy and cost were accounted for [50].
The population considered in this analysis represented patients with a common profile of clinically node-negative early breast cancer. Within this subset, there was far greater complexity inherent to patient and tumor characteristics than could be incorporated. These factors might have affected choice of adjuvant therapy and consequent patient outcomes. For example, complementary DNA microarray gene expression signature profiling is increasingly used for breast cancer risk assessment, the findings of which can influence adjuvant therapy recommendations beyond tumor size, hormone receptor status (HER2/neu overexpression), and other conventional characteristics [52, 53]. Furthermore, the patient populations for which the performance characteristics of MR lym phangiography have been reported are small and heterogeneous [11, 14–16]. Large-scale studies of the performance of MR lym phangiography in breast cancer will be critical for assessment of the reproducibility of the sensitivity and specificity of MR lym phangiography reported to date in smaller series. As the performance characteristics of MR lymphangiography become more firmly established for each primary tumor stage, cost-effectiveness analysis for axillary staging strategies should include a broader spectrum of patients and tumors. This step would enable optimal tailoring of future staging recommendations to patient and tumor risk.
Finally, the cost of MR lymphangiography was prospectively chosen to represent 1.5 times the cost of unilateral breast MRI with contrast administration. Although this estimate was thought reasonable in the long term, initial costs may be higher, as they are with the introduction of any new technology. It is difficult to accurately predict the cost of MR lymphangiography during its anticipated transition to clinical practice. However, our results were robust to the substantial cost variability of MR lymphangiography. SLN biopsy remained weakly dominated by MR lymphangiographic strategies when the cost of MR lymphangiography was varied from 0.5 to 1.5 times the prospective estimate (Table 5). The cost of MR lymphangiography therefore is an unlikely primary determinant of the relative cost-effectiveness of MR lymphangiography staging strategies.
To our knowledge, this study was the first cost-effectiveness analysis of MR lymphangiography for axillary staging of early breast cancer. Our results are not intended to promote current changes in patient-level decisions. Instead, the results of the analysis serve to generate an economic framework for facilitating population-level consideration of MR lymphangiography–based axillary staging strategies from a cost-effectiveness perspective. Our results, first, establish the potential for MR lymphangiography–based strategies as routine alternatives to SLN biopsy from a cost-effectiveness perspective and, second, emphasize the need for large-scale investigation of the performance of MR lymphangiography in early breast cancer. Finally, the results of our analysis emphasize the importance of and need for uniform quality-of-life assessments specific to MR lymphangiography, SLN biopsy, and axillary lymph node dissection, which should precede policy-level decisions regarding use of MR lymphangiography in the management of early breast cancer.
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  C. S. Poon A Brief Commentary on Cost-Effectiveness Analysis in Radiologic Research Am. J. Roentgenol., November 1, 2008; 191(5): 1320 - 1322. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
