- © 2004 by American Society of Clinical Oncology
Cost Minimization Study of Image-Guided Core Biopsy Versus Surgical Excisional Biopsy for Women With Abnormal Mammograms
- From the Division of General Internal Medicine, and Department of Medicine, The Lynn Sage Comprehensive Breast Center; the Department of Surgery, Northwestern University Feinberg School of Medicine; the VA Midwest Center for Health Services and Policy Research, Lakeside Division; the Division of Hematology/Oncology, Department of Medicine; the Center for Healthcare Studies; and the Robert H. Lurie Comprehensive Cancer Center, Northwestern University Medical Center, Chicago IL
- Address reprint requests to Charles Bennett, MD, PhD, VA Lakeside Medical Center, Division of General Internal Medicine, Department of Medicine, 400 E Ontario, Suite 204, Chicago IL 60611; e-mail: cbenne{at}northwestern.edu
Abstract
Purpose To describe the clinical and economic consequences of image-guided core biopsy versus surgical excisional biopsy of mammographically identified breast lesions.
Patients and Methods Clinical and economic data were collected for 1,121 patients undergoing core biopsies and 501 patients undergoing surgical biopsies between 1996 and 1998. Lesions were classified according to mammographic degree of suspicion and type of radiographic abnormality. Costs were measured from the societal perspective. A decision analytic model was constructed, with probabilistic sensitivity analysis.
Results Lesions diagnosed via core versus surgical biopsy were less likely to be masses (39% v 55%), less likely to be classified as high cancer suspicion (17% v 26%), and less likely to be treated with a single procedure (74% v 81%; P < .001 for each). Cancers diagnosed by a surgical biopsy were less likely to have had a single operative procedure (33% v 84%) and were associated with higher total costs whether mastectomy ($2,775 v $1,849) or lumpectomy ($2,112 v $1,365) was used. Sensitivity analysis showed core biopsy optimal in 95.4% of trials. Core biopsy was favored for low-suspicion lesions, calcifications, and masses, and overall for patients who underwent lumpectomy alone.
Conclusion Image-guided core biopsy can be cost-saving compared with surgical biopsy, particularly when the mammographic abnormality is classified as low suspicion or consists of calcifications or masses. Moving to a policy in which core biopsy is the preferred approach in these settings has the potential to result in significant cost savings.
INTRODUCTION
Stereotactic core biopsy has been shown to be a useful alternative to surgical biopsy in the evaluation of nonpalpable mammographic lesions of intermediate to high suspicion.1-7 The benefits of this procedure can include less disfigurement and recovery time, lower potential for complications, and lower immediate costs.
Because 60% to 90% of biopsies for mammographic lesions result in a benign diagnosis, the less invasive procedure appears optimal.8-11 However, there is still controversy over the value of stereotactic core biopsy in highly suspicious lesions or lesions of certain types (such as clusters of calcifications).12,13 Some believe that in lesions likely to be cancer, or those a core biopsy is more likely to miss, the core biopsy adds an additional procedure and is neither a benefit to the patient nor cost-saving. There is concern about its ability to completely characterize malignancies and allow for definitive surgical treatment and about the consequences of missing a diagnosis of early breast cancer. Although a stereotactic biopsy is less invasive and less costly than a surgical biopsy initially, it is uncertain whether the procedure is less costly overall when these concerns are factored in.
A few studies have addressed this question. In one, stereotactic core biopsies were performed on 182 patients with mammographically evident lesions, and data from clinical follow-up were collected.14 Of these patients, 42 required surgical excision. This study found a cost savings of 52% to 55% by performing the stereotactic as opposed to the surgical biopsy. However, it only considered the costs of diagnosis, did not follow up on patients with negative results to determine if there were missed diagnoses, and did not break down the cost differences by severity of diagnosis. A retrospective study of 52 consecutive patients diagnosed with invasive cancer, comparing costs for those with stereotactic core biopsies to those having surgical biopsies, determined that patients in the surgical biopsy group had positive margins more often and required re-excision more frequently.9 Median total costs through the time of definitive procedure were $1,000 less per patient for the stereotactic core biopsy group. However, this study was very small and included only patients diagnosed with invasive cancer. Although both of these studies have approached this question, neither has had the sample size or study design to appropriately conclude which of the procedures has lower cost.
This is an issue with important national economic ramifications. It is estimated that approximately 1 million breast biopsies were performed in the United States in 2000, of which at least 300,000 were for nonpalpable lesions.15 If core biopsy can be cost-saving, the net impact on the use of health resources could be tremendous.
The purpose of this study was (1) to prospectively measure and characterize the lesions found through mammography, the potential surgical course, and the related costs, and (2) to use this observational data to perform a cost-minimization study of the biopsy alternatives, considering all downstream costs from the time of biopsy through definitive surgical treatment. A decision analytic model of the outcomes of all biopsy patients seen at the Lynn Sage Breast Center during a 2-year period was constructed, in order to compare these total costs.
PATIENTS AND METHODS
Patients
Patients were selected for an image-guided core biopsy or surgical biopsy in a nonrandomized fashion after completion of the diagnostic imaging workup. Treatment selection reflected the radiologist's assessment of the suitability of the patient for image-guided breast biopsy, as well as the preference of the referring physician. In the absence of medical contraindications to image-guided breast biopsy, no standard criteria were used to select the type of biopsy procedure. Patients with diffuse abnormalities that would require mastectomy were preferentially selected for core biopsy. Core biopsy was performed using stereotactic guidance and a 14-gauge biopsy needle for calcification lesions and architectural distortions. Ultrasound guidance was used for mass lesions that could be visualized by ultrasonography. Patients with a core biopsy diagnosis of atypical ductal hyperplasia, radial scar, or a lack of concordance between the mammographic appearance of the lesion and the core biopsy diagnosis underwent surgical excision of the lesion to confirm the absence of malignancy. Papillary lesions were selectively excised based on clinician-pathologist judgment.
All surgical biopsies were performed as outpatient procedures in the operating room using local anesthesia and intravenous sedation. All lumpectomies and re-excisions were performed in a similar fashion. Mastectomies and axillary dissections were done under general anesthesia with a 23-hour period of in-hospital observation. Drains were used after mastectomy and axillary surgery, but not after lumpectomy, re-excision, or diagnostic biopsy.
In this study, patients who did not undergo axillary dissection included those with intraductal carcinoma, microinvasive carcinoma (defined as a single focus of invasion within 2 mm of the basement membrane), or pure tubular carcinoma less than 1 cm in size. In addition, older patients in whom the findings of axillary dissection would not change the approach to adjuvant systemic therapy, who had no clinical evidence of nodal metastases, and who were believed to be at low risk for nodal metastases on the basis of tumor size and histologic features did not undergo axillary dissection.
At the time of biopsy, the attending radiologist recorded the type of mammographic abnormality (ie, mass, calcification, or architectural distortion). All lesions were Breast Imaging and Reporting Data System (BIRADS) category 4 or 5,16 but because of the widely varying levels of breast cancer risk encompassed by BIRADS category 4, this group was subdivided into four risk categories (Table 1) to allow a more precise estimation of risk. All lesions were assigned a risk score of A to D based on mammographic appearance; this score is strongly related to cancer risk and incidence (Table 1). In patients with multiple lesions, each lesion was given an individual risk score. Following institutional review board approval, information regarding patient age, tumor histology, type of surgical procedure, and the number of surgical procedures to complete local therapy was obtained from the breast cancer database. All pathology records for each patient within 1 year of the time of breast cancer diagnosis were reviewed to ensure complete recording of the number of surgical procedures. Patients who underwent biopsies at the Lynn Sage Breast Center but had definitive local therapy elsewhere were excluded from analysis.
Mammographic Classification
Decision Model
Because there is no evidence to suggest a difference in treatment outcomes between surgical and core biopsies, this was developed as a cost-minimization study.17 To model the various possible clinical courses and their costs, and to be able to analyze the uncertainties surrounding all of the measures, a decision tree was constructed.18 This incorporated all reasonable decisions and chance events related to the consequences of an abnormal mammogram, as practiced in our institution.
The analysis was a comparison of surgical biopsy with core biopsy (Fig. 1 and 2). The tree represents the actual flow of events that occurred in the care of these cohorts. Each arm has the possibility of three initial readings: invasive cancer, ductal carcinoma-in-situ (DCIS), or benign. Benign itself could be a final diagnosis or considered a technical miss based on either a second biopsy procedure within one month or a later biopsy resulting from a suspicious mammogram within 1 year of the benign diagnosis.
Decision tree for surgical biopsy. For each branch, the first number represents low risk (Breast Imaging and Reporting Data System [BIRADS] 4A, 4B, and 4C) patients; the second number represents high risk (BIRADS 4D and 5) patients. bx, biopsy; DCIS, ductal carcinoma-in-situ; w/o, without; LN, lymph node; neg, negative; pos, positive; w/, with.
Decision tree for core biopsy. For each branch, the first number represents low risk (Breast Imaging and Reporting Data System [BIRADS] 4A, 4B, and 4C) patients; the second number represents high risk (BIRADS 4D and 5) patients. bx, biopsy; DCIS, ductal carcinoma-in-situ; w/o, without; LN, lymph node; w/, with; pos, positive; neg, negative; addl, additional; mar, margins.
The third level of the tree represents the treatment possibilities for each diagnosis. Among patients who had a surgical biopsy, for invasive cancer they are mastectomy or lumpectomy (with or without lymph node dissection), or the possibility that the biopsy itself was inclusive (with or without lymph node dissection; Fig 1). Patients receiving a lumpectomy had tumor margins evaluated as negative or positive. The lumpectomy was definitive when the margins were negative, or further re-excision or mastectomy was performed when the margins were positive. Patients with a diagnosis of DCIS underwent either a re-excision or mastectomy without lymph node dissection, unless the biopsy demonstrated negative margins. Patients with confirmed DCIS and negative margins required no further surgery. Those patients whose disease was determined to be invasive had axillary node dissection, and if the biopsy margins were positive, re-excision or mastectomy. Patients with a benign result had a follow-up mammogram at 1year post biopsy. A technical miss could lead to re-excision, axillary dissection, or mastectomy with or without node dissection.
The third level of the tree given a stereotactic core biopsy (Fig 2) was similar except that (1) a simple lumpectomy with negative margins sometimes led to lymph node dissection; (2) the biopsy was not considered inclusive; (3) DCIS with positive margins sometimes ultimately resulted in mastectomy or a second re-excision; (4) technical misses sometimes ultimately resulted in an initial surgical excision followed by a re-excision, a second core biopsy, or lumpectomy; and (5) patients with a benign result had follow-up mammograms at 6 months and 1 year.
As patients accrued onto the study, the actual probabilities for each chance outcome were empirically determined and used for the analysis. A sample of patients traveling through each branch was directly studied to determine actual costs at each point, as discussed further herein. Once sufficient data were accrued, the tree was analyzed to compare costs of the various combinations of biopsy approach and surgical approach.
In addition to the empirically derived tree, costs were analyzed by (1) considering only patients receiving breast-conserving (lumpectomy alone) surgery, and (2) subgroup based on degree of suspicion (low [BIRADS 4 category A-C] v high [BIRADS 4, category D and BIRADS 5]) and on radiographic abnormality type (architectural distortion, calcification, or mass).16
Assumptions
An assumption was made of equal effectiveness of both arms with respect to cancer treatment. Because of this, the only outcome measured was cost from the time of initial mammographic diagnosis to completion of surgical diagnosis and treatment. It was also assumed that the costs beyond surgery (such as radiation or chemotherapy) would not be different between the two arms.
The time horizon was 1 year. All patients who did not have an open biopsy would have mammograms performed at 6-month intervals; for those who had an open biopsy, mammograms would be performed annually. No women had both DCIS and invasive cancer at the same time. The procedures did not have serious adverse effects. No change in prognosis occurred if a false-negative stereotactic biopsy occurs and a positive biopsy occurs at the next mammographic screening.
Data Collection
Information was collected on all patients seen at the Lynn Sage Breast Center for a surgical or core biopsy from September 1, 1996, through August 31, 1998. A monthly printout of each patient's age, biopsy procedure, lesion type, degree of suspicion, and pathologic diagnosis was prepared from the Breast Center's Management Information System database. Missing data were provided by chart review. Follow-up information on surgery performed was obtained from the Northwestern Memorial Hospital Pathology Department. This information was also used to verify biopsy-related data. The information was combined into a database that was reviewed by two of the authors (L.V. and M.M.) for clinical relevancy. These clinical data were used to define the probabilities for each node of the decision tree.
Patient billing records were collected for a sample of patients from the various branches of the decision tree, and a mean cost per procedure was determined. The sum of the mean costs of the procedures was used as the baseline outcome measures in the decision tree.
Costs were measured from a societal perspective. Only direct costs related to inpatient care were considered, and they included core biopsy, surgical biopsy, lumpectomy with or without re-excision, lumpectomy with or without lymph node dissection, mastectomy with or without lymph node dissection, and lymph node dissection alone. Specific examples include the costs of the procedures, pathology readings, anesthesia, blood work, radiology testing, ECG, medications for the procedures, recovery rooms costs, and inpatient room costs. Costs were derived by application of the institution's cost-to-charge multiplier.
Sensitivity Analysis
There are two primary sources of uncertainty in the data used in this analysis: procedure costs and probability of any particular event. Sensitivity analysis on costs was performed by a series of Monte Carlo analyses. Each of these entailed recalculating the decision tree 1,000 times, each time randomly choosing a cost based on the measured distributions, resulting in a percentage of times in which each strategy would be least expensive. Because the SD for each cost was very large, we used a uniform distribution of the entire cost range for each variable. This can be considered a conservative assumption in that its impact would be to undermeasure the differences between each arm.
The uncertainty of probabilities results from there being many branches through which only a small number of patients traveled. To address this, after the baseline conclusions were reached the probabilities which would potentially lead to the opposite conclusion were biased by using the upper end of their 95% confidence interval, keeping other probabilities in their original proportion. For branches with no patients, the probability was set to 1 for the branch that could lead to the opposite conclusion. Monte Carlo analyses of the costs were also run on these biased trees.
RESULTS
Characteristics of Cohort
During the 2 years, 1,307 core biopsies were performed on 1,121 patients and 544 surgical biopsies on 501 patients (Table 2). The mean age of the patients was 53 years in the core biopsy group and 55 years in the surgical biopsy group (Table 2). There was a higher percentage of calcifications biopsied in the surgical group (52% v 40%) and a higher percentage of masses biopsied in the core group (55% v 39%). Lesions diagnosed via core biopsy were of lower suspicion than those diagnosed with surgical biopsy. Of lesions diagnosed by surgical biopsy, 26% were ductal carcinoma-in-situ or invasive cancer, compared with 20% of lesions diagnosed by core biopsy (P < .01).
Descriptive Statistics of Data Set
Overall, 81% of surgical biopsy versus 74% of core biopsy lesions had a single procedure for diagnosis and/or therapy (P < .001). Of those lesions diagnosed as cancer, only 33% of those diagnosed surgically underwent a single operative procedure, compared with 84% of those diagnosed by core biopsy (P < .001; Table 3). These differences remained consistent when stratified by degree of suspicion of the lesion, or lesion type. In comparisons of definitive surgery, those lesions treated by mastectomy or lumpectomy with axillary node dissection were significantly more likely to be treated by one surgical procedure after a core biopsy compared to a surgical biopsy (Table 3). However, for lesions treated with lumpectomy alone, there was no significant difference in the percentage of patients requiring only one surgical procedure between biopsy groups. Overall, lesions diagnosed by core biopsy were more likely to require additional surgery after an attempt at definitive local therapy was completed, 15.7% for core biopsy versus 2.1% for surgical biopsy (P ≤ .001). Of the 1,307 lesions diagnosed by core biopsy, 74 were considered technical misses and required additional procedures because of this. In this study, no patients in the core biopsy group required subsequent excision of a mammographic abnormality initially diagnosed as benign during the follow-up period.
Percentage of Cancerous Lesions Treated with a One-Stage Surgical Procedure
Individual Procedure Costs
For each procedure performed, the mean, standard deviation (SD), and range of costs in the sample of billing records are shown in Table 4.
Costs
Cost Analysis
Baseline.
The total cost of diagnosis and surgical treatment was $1,849 for core biopsy versus $2,775 for surgical biopsy. Monte Carlo analysis showed core biopsy to be the optimal approach in 95.4% of trials. When the probabilities were biased to favor surgical biopsy, the cost was $2,297 for core biopsy and $2,458 for surgical biopsy, still favoring core biopsy. Monte Carlo found core biopsy to be optimal in 53.5% of trials.
Core biopsy was favored for low suspicion lesions ($1,218 v $2,374), calcifications ($1,652 v $2,523), and masses ($1,895 v $3,265). For each of these, Monte Carlo favored core biopsy in 99.8%, 93.7%, and 100% of trials, respectively. Surgical biopsy was favored for high suspicion lesions ($3,900 v $4,983) and architectural distortion ($2,425 v $2,915), with Monte Carlo showing the same results in 99.5% and 83.8% of trials, respectively. With probabilities biased to favor the alternative approach, core remained optimal for low-suspicion lesions and became optimal for high-suspicion lesions.
Breast-conserving surgery.
When considering only those patients who underwent lumpectomy alone, the results favor core biopsy—$1,365 versus $2,112, favored in 86.4% of trials. When the probabilities were biased to favor surgery, core remained the least expensive approach—$1,900 versus $1,945—with core favored 47.6% of the time, indicating that the costs are virtually equal.
DISCUSSION
Overall, lesions diagnosed as cancer by core biopsy are more likely to require a single surgical procedure than those diagnosed by surgical biopsy. Consequently, total costs were $926 less for the core biopsy group. This is consistent with other literature that has shown a cost savings of $740 to $1,000 per patient using core versus surgical biopsy.13,19-21 In this study similar or greater savings were found with core biopsy when mammograms were interpreted as low-suspicion or the lesion type was calcification or mass. Savings were also found for patients treated with both breast-conserving approaches and mastectomy. The only subgroups in which the surgical biopsy approach was less expensive were high-suspicion lesions or when the lesion type was architectural distortion.
In interpreting our findings, several factors should be considered. Our analysis represents methodologic improvements over prior cost studies. We had the advantage of prospective data collection and consideration of all downstream breast procedures as well as associated costs. Other models have included cost estimates that ended with the diagnosis of breast cancer, did not include subsequent surgical treatment of the cancer, or did not have comprehensive follow-up of persons who had negative biopsies.13,19 Our sample size was four times larger than that included in the study of Fahy et al,13 and comprehensive follow-up for at least 1 year was carried out. Although detailed cost information was obtained from only a sample of the women in our study, sensitivity analyses supported the robustness of the cost estimates. Also, the model included a conservative assumption that costs followed a uniform distribution, rather than a normal distribution with the measured mean and SD. This assumption leads to low estimates of the frequency in which one strategy would dominate the other; given that in most cases the optimal strategy was chosen in more than 90% of trials, the true differences would likely be stronger.
There are limitations to our study. First, the probabilities and costs were measured at a single institution, raising the question of generalizability. It is possible that other providers might vary in their approach to breast cancer diagnosis. Costs could vary in different geographical or hospital settings. However, the consistency in the key results when biasing the probabilities in an opposite direction indicate a robustness that should result in similar conclusions, even with a large amount of variability in practice patterns. Second, because the clinical patterns of care in our study were not observed as part of a randomized trial, selection bias may have existed at either the physician or patient level in deciding whether a patient underwent surgical or image-guided core biopsy. Physician belief during the study period that core biopsy was particularly effective for the diagnosis of lower-suspicion abnormalities is evidenced by the fact that 83.2% of the lesions diagnosed by core biopsy had a suspicion of cancer of 70% or less, compared with 73.6% of those in the surgical biopsy group. However, this cohort study was designed as a descriptive analysis, rather than a treatment study, and it was assumed that there would be no difference in the ultimate clinical outcome based on the biopsy approach. As an observational study, the results reflect the costs based on real-world clinical practice.
The distribution of mammographic abnormalities in an individual practice (calcifications v masses) has the potential to impact on the exact cost savings that will be achieved with image-guided biopsy. It is generally accepted that for abnormalities that can be visualized by ultrasound, ultrasound-guided core biopsy is preferred to stereotactic guidance.20,22,23 The ultrasound-guided procedure avoids the use of ionizing radiation, and no breast compression is necessary, making it more comfortable for the patient, and the procedure time is shorter.23 In general, ultrasound guidance is useful only for patients with mass lesions, given that calcifications and architectural distortions are not reliably imaged with this technique. The cost of ultrasound guided biopsy is less than that of stereotactic biopsy,20 so practices with a greater percentage of mass lesions than the 50% in our study will realize greater savings than those reported here.
A proportion of patients who undergo surgical biopsy do so because of lesions that are not amenable to stereotactic biopsy, usually because of the position of the lesion within the breast or patient inability to cooperate with the procedure. The exact percentage of cases that fall into this category will vary with the experience of the physician performing the biopsy procedure. However, Liberman and Sama15 have demonstrated that the majority of lesions that could not be sampled for technical reasons with the 14-gauge biopsy needle can be biopsied using an 11-gauge directional vacuum-assisted biopsy device. The number of patients with mammographic abnormalities who require surgical excision due to technical inability to sample the lesion is currently less than 1% at our center, so this is unlikely to substantially alter our results.
Since the time of this study, sentinel node biopsy, rather than axillary dissection, is rapidly emerging as the initial procedure for axillary staging in clinically node-negative patients.24 It is possible that sentinel node biopsy is a lower cost procedure than axillary dissection for the node-negative patients who constitute the majority of mammographically diagnosed breast cancers, but an initial cost analysis does not demonstrate a clear impact of sentinel node biopsy on the cost of breast cancer treatment.25 Even if further studies do show a significant cost savings for sentinel node biopsy, the procedure is not performed in the absence of a definitive diagnosis of cancer and usually is done before the lumpectomy. Thus, a second separate procedure with its attendant costs would still be required for the majority of patients with invasive cancer. In fact, the lower morbidity of sentinel node biopsy has led to its use in some patient subsets for whom axillary dissection was not routinely performed. This includes patients with microinvasive cancer, tubular carcinoma, and some older women who did not undergo axillary dissection in this study. In patients with lymph node metastases who undergo axillary dissection, sentinel node biopsy is an added cost. Given these considerations, in the absence of an actual cost analysis, the impact of sentinel node biopsy on the cost savings described here is uncertain, but it is unlikely to alter the advantage seen for core biopsy.
The diagnosis of relatively low-suspicion mammographic abnormalities remains the major induced cost of screening.26 The failure to diagnose breast cancer and the delay in diagnosis of breast cancer are major liability issues for radiologists. This has resulted in biopsy thresholds in the United States that are quite low, with cancer diagnosed in approximately 20% of screen-detected abnormalities undergoing biopsy.27 Although substantial cost savings could be realized if thresholds were set higher, that issue was not addressed by the design of this study nor by its analysis, and higher thresholds are not current practice in the United States. However, a recent comparison of the outcome of mammographic screening in the United States and the United Kingdom demonstrated that more than twice as many women in the United States as the United Kingdom underwent open rather than image-guided core biopsy, although the rates of cancer diagnosis did not differ.28 National figures for surgical biopsy rates versus image-guided core biopsies are not available. However, in the study of Smith-Birdman et al,28 which included 1.6 million mammograms performed between 1996 and 1999, approximately one third of the biopsies were surgical. It is estimated that if all eligible women in the United States were to follow the American Cancer Society's recommendations for screening mammography, approximately 1 million breast biopsies would be recommended per year for mammographically detected abnormalities.29 From a policy perspective, the aggregate effect of increasing the proportion of core biopsies performed could be large. In a health care environment that is increasingly focused on value, physicians and policy makers will have to consider the clinical and economic implications of alternative approaches to breast biopsies. Our model provides some of the relevant background information for these efforts.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
Footnotes
-
Supported by Department of the Army Medical Research and Development Command grant 17-96-2-6013.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
- Received June 30, 2003.
- Accepted March 15, 2004.
