Economic Analysis of Targeting Chemotherapy Using a 21-Gene RT-PCR Assay in Lymph-Node"Negative, Estrogen-Receptor"Positive, Early-Stage Breast Cancer

, ,
The American Journal of Managed Care, May 2005, Volume 11, Issue 5

Objective: To appraise the economics of a recurrence score(RS), based on an assay that predicts distant recurrence-free survivalin lymph-node-negative (LN-), estrogen-receptor-positive(ER+) patients with early-stage breast cancer receiving tamoxifen.

Study Design: Cost-utility analyses using a decision analyticmodel.

Methods: Using a Markov model, we forecast overall survival,costs, and cost effectiveness of using the RS in patients classified ashaving low or high risk of distant recurrence based on NationalComprehensive Cancer Network (NCCN) clinical guidelines. Datafrom a large multicenter clinical trial (NSABP B-14) were analyzedto derive risk classification based on guideline criteria and RSassignments. Efficacy of adjuvant chemotherapy (CT) on distantrecurrence-free survival (DRFS) was based on published metaanalysesof CT trials. The analysis took a societal perspective, consideringsurvival, quality of life, and relevant costs.

Results: Fifty-three patients (8%) were classified as having lowrisk of distant recurrence by NCCN guidelines and the RS reclassified15 of these patients (28%) to an intermediate/high-risk group.The remaining 615 patients (92%) were classified at high risk ofdistant recurrence by NCCN guidelines and the RS reclassified 300of these patients (49%) to a low-risk group. Among a hypotheticalcohort of 100 patients, RS is predicted on average to increase quality-adjusted survival by 8.6 years and reduce overall costs by$202 828. RS was cost saving in more than two-thirds of probabilisticsimulations, with cost effectiveness most influenced by thepropensity to administer CT based on RS results, and by the proportionof patients at low risk as defined by NCCN guidelines.

Conclusions: The RS predicts more accurately than currentguidelines recurrence risk in LN-, ER+ patients with early-stagebreast cancer. If applied appropriately, the assay is predicted toincrease quality-adjusted survival and save costs.

(Am J Manag Care. 2005;11:313-324)

More than 210 000 women in the United Statesare diagnosed each year with breast cancer.1Breast cancer remains among the most commoncancers in women, and the most common cause ofdeath among women between the ages of 40 and 79.1

Several large randomized clinical trials demonstratedthe benefit of hormonal therapy in patients with estrogen-receptor-positive (ER+) early-stage breast cancer(ESBC).2 An important decision for a patient withlymph-node-negative (LN-), ER+ ESBC is whether toalso undergo adjuvant chemotherapy after primary surgeryto prevent or delay distant recurrence.3-7Consensus guidelines endorse the addition of adjuvantchemotherapy for LN-, ER+ cancer for patients up to 70years old, or older if they are medically fit.8-11 Expertsalso recommend against routine use of adjuvantchemotherapy for small tumors (<1 cm) or for smalltubular or mucinous tumors.12

Enhanced public health efforts to detect breast cancer,such as mammographic screening, have increasedearly-stage detection.13,14 The success of this campaignhas naturally resulted in physicians and patientsincreasingly facing a complex question: do the benefitsof adjuvant chemotherapy outweigh the medical risksand known adverse effects on quality of life?15 That thisquestion is difficult to answer is supported by recentevidence showing wide variation in the propensity toprescribe adjuvant chemotherapy, a variation that cannotbe explained by characteristic risk factors such asage, tumor size, and histology.10,16-21 An active area ofoncology research therefore is identifying additionalreliable predictors of recurrence in ESBC that wouldassign risk more accurately and help guide the decisionto prescribe adjuvant chemotherapy.22-27

type

P

P

A 21-gene reverse transcriptase-polymerase chainreaction (RT-PCR) assay (OncoDX Breast CancerAssay, Genomic Health, Inc, Redwood City, Calif) generates an individualized "recurrence score" that isderived from a proprietary algorithm. The recurrencescore has been prospectively validated as a predictor of10-year distant recurrence-free survival (DRFS) inpatients with LN-, ER+ ESBC.28 The investigatorsobtained tissue samples from 668 LN-, ER+ patientswho were treated with tamoxifen in the NationalSurgical Adjuvant Breast Cancer Project (NSABP) B-14clinical trial from 1982 through 1988 and whose outcomeshave been tracked over time by NSABP sites. Therecurrence score accurately classified patients intolow and high risk of DRFS (< .001).28 When therecurrence score was examined together with age andtumor size in a multivariate analysis, only recurrencescore remained a significant predictor of DRFS at 10years (< .001).28 The study also confirmed the relativelypoor reliability of current risk stratification.Twelve of 53 (22%) patients initially classified byNational Comprehensive Cancer Network (NCCN) criteriaas low risk were reclassified by the recurrencescore as intermediate risk and had a 10-year DRFSequal to 82% (95% confidence interval [CI] 60%-100%);3 (6%) patients were reclassifiedas high risk and had a 10-yearDRFS equal to 57% (95% CI 1%-100%) (Figure 1).29 Conversely,300 of 615 (49%) patientsassigned by NCCN criteria as highrisk for recurrence were reclassifiedby the recurrence score aslow risk and had a 10-year DRFSequal to 92% (95% CI 89%-95%).

Although the recurrence scorehas been subjected to clinical validation,questions necessarilyarise about its affordability andfactors that would influence itsappropriate use. The purposes ofthis study were (1) to estimatethe incremental benefits, costs,and cost effectiveness of using therecurrence score to better assignrisk of DRFS associated withESBC, and (2) to assess the factorsthat most influence potentialbenefits and efficient use of therecurrence score.

METHODS

This evaluation focused on therecurrence score and not otherRT-PCR or genomic assays underdevelopment, because the analyses depend critically onthe quality of evidence and detailed findings of the validationstudies described herein. We present DRFS,overall survival (OS), and relevant costs associated withuse of the recurrence score to reclassify risk of recurrencecompared with risk classification using currentNCCN guidelines alone. Outcomes were evaluated usinga common framework for health economic appraisals,called the Markov model (Figure 2), which provides aconvenient way of modeling prognosis.30 In a Markovmodel, a patient may be in 1 of a finite number of statesof health, and events of interest are modeled as transitionsfrom one state to another. For each state, analystsassign a utility as an adjustment factor for quality of life.Utility weights typically range from 0 to 1, in which 0represents a state as bad as death, 1 represents perfecthealth, and the values between 0 and 1 representdegrees between these extremes. The contribution tototal utility, commonly referred to as quality-adjustedlife years (QALYs), of a particular state consists of thelength of time spent in a state multiplied by the utilityof that state.

Two scenarios were considered involving representativepatients with LN-, ER+ ESBC who were expected toreceive 5 years of hormonal therapy. In scenario 1,patients were classified by NCCN as low risk (eg, tumorsize <1 cm), would therefore not receive chemotherapy,and would have a predicted 10-year probability ofrecurrence of 7.8% based on analyses of NSABP B-14data (Table 1). In scenario 2, patients were classified byNCCN as high risk (eg, tumor size >2 cm), would thereforebe recommended to receive chemotherapy, andwould have a predicted 10-year probability of recurrenceof 21.9%.

The recurrence score was assumed to reclassifyrecurrence risk independent of NCCN risk criteria, withthe probability of reclassification based on results of theNSABP B-14 data. In the base case, we assumed that allpatients assigned as intermediate/high risk by the recurrencescore would undergo chemotherapy and allpatients assigned as low risk by the recurrence scorewould not receive chemotherapy. We explored in sensitivityanalysis the implications of different probabilitiesof chemotherapy use based on risk classification by therecurrence score. The model tracked rates of recurrenceand death, as shown in the outcome subtree ofthe decision tree in Figure 2.

Probability of Risk Reclassification

The NSABP provided us with analyses of B-14 data todetermine the risk of recurrence based on NCCN criteriaand on the recurrence score results (Figure 1 andTable 2).29 For the 53 (7.9%) women assigned as havinglow risk for recurrence based on NCCN criteria, NSAPBB-14 showed a 28% (95% CI 16.8%-42.4%) probability ofreclassification to intermediate/high risk (15 of 53women) with the recurrence score. For the other 615(92.1%) women assigned as having high risk for recurrencebased on NCCN criteria, NSABP B-14 showed a49% (95% CI 44.5%-52.6%) probability of reclassificationto low risk (300 of 615 women) based on therecurrence score.

Risk of Recurrence and Death

Annual risks of recurrence and survival wereobtained from published meta-analyses of clinical trials(Table 1).31 The relative risk reduction of distant recurrenceassociated with chemotherapy plus tamoxifenversus tamoxifen only is approximately 30%, and isassumed to be the same regardless of NCCN risk classification.31 Studies are ongoing to assess the correlationof tumor gene expression and response to chemotherapy.35 Gianni et al recently reported that gene expressionprofiles of paraffin-embedded core biopsy tissue predictedresponse to chemotherapy in patients with locallyadvanced breast cancer,36 which was subsequentlyconfirmed in an NSABP study.37 In the base case model,we assumed a 15% risk reduction associated withchemotherapy if the recurrence score falls in the low-riskcategory and a 45% risk reduction associated withchemotherapy if the recurrence score falls in the intermediate/high-risk category.

After recurrence, the chemotherapeutic regimen,probability of response, and risk of further disease progressiondepend on Her2/neu status, assumed to be positivein 25% of cancers.32 Regardless of response, therate of disease progression is lower for patients withHer2/neu-positive tumors receiving combinationtrastuzumab/paclitaxel than for patients with Her2/neunegativetumors receiving paclitaxel monotherapy. Theannual probability of death after progression is approximately40%.32

Quality of Life

not

had

Previous studies have reported a relatively highutility of 0.98 after initial chemotherapy for ESBC(Table 3).38 To account for emerging data on late negativeeffects of early treatment, such as potential recurrenceor late effects of hormonal and chemotherapy,we lowered this value in sensitivity analyses.45,46 Weused published estimates of the negative impact onutility for recurrence (recently used by Elkin et al toassess implications of newer strategies for patientswith advanced breast cancer).32 Large variabilityexists on how breast cancer survivorsreport their perceptions ofthe value of chemotherapy therapies.39,47 For example, in a recentstudy, among survivors receivingchemotherapy, 61% reported theywould have chosen chemotherapy ifit offered at least a 6% reduction inbreast cancer mortality.39 By contrast,97% of survivors who received chemotherapy wouldchoose chemotherapy if it offered atleast a 6% reduction in breast cancermortality.39 Cole et al showed theimplications of chemotherapy onquality-adjusted survival over a range of utilities, using0.5 as an illustrative value.31 We found in preliminaryanalyses that use of chemotherapy in low-riskpatients yields no gain in QALYs when the utility ofchemotherapy is set to 0.5 for 6 months of treatment.We applied this utility to chemotherapy, suchthat findings of the base case model would be consistentwith current guideline recommendations for useof chemotherapy in the adjuvant setting.

Costs

Cost of the assay was based on the manufacturer'ssuggested retail price of $3450. Drug acquisition costsfor chemotherapy were based on 2004 Redbook prices(Table 3).40 Estimates of adjuvant chemotherapy costsvary depending on the type of regimen used. For example,regimens that include taxanes are more costly thanregimens using only anthracycline and cyclophosphamide(also known as AC) or methotrexate and 5-fluorouracil (also known as CMF). Surveys of medicaloncologists have demonstrated that taxane-based regimensare used more frequently in patients withHer2/neu-positive tumors.48 We included costs associatedwith infusion, patient time, use of colony-stimulatingfactors to prevent myelosuppressive complications, andmanagement of chemotherapy-related side effects.Oncologists have reported that approximately 50% ofpatients with ESBC receive an eythropoietic-stimulatingfactor or a myeloid colony-stimulating factor.48

We reviewed the literature to determine the cost ofcancer recurrence surveillance in patients with ESBC.Simon et al. retrospectively evaluated the costs of routinesurveillance in patients treated for ESBC, includingoutpatient visits, laboratories, and imaging studies,49which remain the approach recommended by guidelines.41 They reported a cost of $1396 over 5 years in1993 dollars. This results in an average annual cost,inflated to 2004 dollars, of $421.

A search of the medical literature (MEDLINE, CANCERLIT)revealed several studies that reported thecosts in the last year of life for elderly Medicare beneficiaries,varying between $23 000 and $37 000.42-44 Weapplied an average cost of $30 000 for end-of-life care.

Other Policy Assumptions and Analyses

The model takes a societal perspective, including relevantcosts and outcomes associated with breast cancer.We forecasted outcomes over the patient's lifetime,as recommended in guidelines for economic appraisalsof medical technologies.42-44 In sensitivity analyses, weexplored the implications of shorter forecast durations.The cycle length of the model was 1 year, and 10-yearrecurrence rates were used to estimate annual rates inthe model assuming a constant hazard rate over 10years. Several factors may have influenced the effect ofrecurrence score testing on QALYs and costs. We conducteda series of sensitivity analyses to explore whatfactors had the greatest influence on these end points,and on costs per QALY gained. Although we examinedeach variable, we focused on several variables that wepredicted a priori would have more influence on theseend points, such as (1) the propensity to changechemotherapy decision based on recurrence score, (2)relative risk reduction of chemotherapy plus tamoxifenversus tamoxifen only, (3) cost of chemotherapy, and(4) the proportion of patients defined by NCCN criteriaas low risk. Using standard methodologies, we discountedboth costs and survival by a fixed annual rate of 3%.50

Incremental cost utility was estimated according to astandardized method described in published guidelinesas the difference in cumulative costs between testingand no testing with the recurrence score, divided bythe difference in years of quality-adjusted survivalbetween the 2 approaches. We show outcomes for riskscenarios first separately and then applied to a cohort of100 patients.

We conducted extensive 1-and 2-way sensitivityanalyses on the model assumptions. Ranges used inthese analyses are shown in Tables 1 and 3. Last, weconducted a probabilistic analysis (Monte Carlo) todetermine the effect of uncertainty in all variables onthe cost-utility ratio for the cohort. A technicalappendix with detailed information on these simulationsis available to readers upon request of the correspondingauthor.

RESULTS

Cost Effectiveness ofChemotherapyBased on NCCNGuidelines

Administration of adjuvantchemotherapy inhigh-risk patients basedonly on NCCN guidelineswas forecast to increaseoverall survival by 1.53years and increase QALYsby 0.65 years. Adjuvantchemotherapy increasesoverall costs for breastcancer by $13 878, resultingin a cost utility equalto $21 428 (Table 4). Inlow-risk patients, adjuvantchemotherapy wasforecast to increase overallsurvival, but only by0.37 years compared withuse of chemotherapy inhigh-risk patients. However,when adjusting foradverse consequences onquality of life ofchemotherapy, quality-adjustedsurvival withchemotherapy is lowerthan without chemotherapy.The model findingstherefore are consistentwith current NCCNguidelines that adjuvantchemotherapy improvesoverall survival andQALYs for patients at highrisk of recurrence, but hasa net negative effect onQALYs for patients at lowrisk of recurrence.

Cost Effectiveness of Reclassifying PatientsUsing the Recurrence Score

Using the recurrence score to reclassify NCCN-definedlow-risk patients to intermediate/high risk wasprojected to add $15 776 per patient for chemotherapy.Another $12 190 in testing costs was expected perpatient reclassified as intermediate/high risk (Table 5).Chemotherapy administered to patients at intermediate/high risk has an expected increase of 1.86 years inoverall survival. The cost per QALY gained of using therecurrence score to reclassify NCCN-defined low-riskpatients was calculated as $31 452.

Reclassifying a patient with NCCN-defined high-riskESBC to low risk based on the recurrence score willsave costs by foregoing administration of chemotherapy.An average of 2 NCCN-defined low-risk patientsmust be tested to reclassify 1 patient; hence, the assaycost per patient reclassified was $7073. Because thesepatients are now classified as having a low risk ofrecurrence, avoiding chemotherapy is predicted toincrease quality-adjusted survival and save $8947 (SD$302) in lifetime costs.

Among a group of 100 patients with LN-, ER+ESBC (7.9% of whom are NCCN-classified as lowrisk for distant recurrence), use of the recurrencescore was expected to reclassify 2 patients from low tointermediate/high risk and 45 patients from high tolow risk. Assuming that patients at intermediate/highrisk receive chemotherapy and patients at low risk donot receive chemotherapy, testing will increaseQALYs in this cohort by 8.6 years (Table 6). Use ofthe recurrence score in 100 patients will cost$345 000; the cost of adjuvant chemotherapy willdecrease by 46%, from $1.63 million to $876 000.Overall costs were projected on average to decline by5%, from $4.32 million to $4.12 million, for a net savingsof $202 828.

Sensitivity Analyses

In 1-way sensitivity analyses (Table 7), recurrencescore testing was beneficial and cost saving using alternativeassumptions for most of the variables. The 2exceptions were (1) the propensity to not usechemotherapy if the recurrence score reclassifiedpatients from high to low risk and (2) the proportion ofpatients tested with the recurrence score who were lowrisk by NCCN criteria. If only 50% of high-to-low reclassifiedpatients were to forego chemotherapy, then thetest would be cost increasing; the cost per QALY gainedwould be $17 234. Testing only NCCN-defined high-riskpatients has a minimal effect on QALYs, but resultsin larger cost savings. Testing only NCCN-defined low-riskpatients will have larger benefit in QALYs and becost increasing, with a cost utility of $31 529.

The QALYs gained with recurrence score testingwere relatively unchanged from the base case except forsituations in which the effect of chemotherapy on qualityof life was substantial (eg, when setting loss in utilityassociated with chemotherapy as low as 0.1). Costsunrelated to chemotherapy drug acquisition, such asthose for infusions, management of emesis and otheradverse events, and use of colony-stimulating factors,had the largest effect on differences in costs. Cost savingsassociated with recurrence score testing wereattentuated in scenarios with low nonchemotherapydrug costs, such as if few patients receive colony-stimulatingfactors. Alternatively, if nonchemotherapy drugcosts are 25% higher, cost savings are even greater.Including the cost of patient travel and lost work productivity,equal to $17 per infusion,32 also makes recurrencescore testing more cost saving.

The base case evaluation used NCCN criteria toassess risk of distant recurrence. Other guidelineshave been published and, for the most part, containsimilar classification criteria, including tumor size,age, histology, and hormonal status.9,51 Analyses ofNSABP B-14 data using thecriteria of other guidelinestherefore would have relativelylittle impact on thecost utility of the recurrencescore assay.

The probability of beingreclassified as low risk andnot receiving chemotherapyis a more significant driverof cost effectiveness thanthe probability of beingreclassified as intermediate/high risk and receivingchemotherapy. The costeffectiveness of the recurrencescore in a hypotheticalcohort of patients wascalculated as less than$50 000 per QALY gained if the propensity to usechemotherapy when recurrence score reclassifies apatient as low risk is at least 30%. Recurrence score iscost saving when the propensity not to use chemotherapyin reclassified low-risk patients exceeds 60%.

In probabilistic analyses, more than two thirds ofsimulations showed the recurrence score to improveQALYs and save costs. The upper 95th percentile ofcost utility equaled $16 874.

DISCUSSION

Our study projects the benefits, costs, and cost effectivenessof a genomic test based on empirical evidenceof its clinical effectiveness, and has 3 primary findings.First, reclassifying patients who were NCCN-defined aslow risk to intermediate/high risk by the recurrencescore was projected to increase overall survival, quality-adjustedsurvival, and costs. The average gain in OS perreclassified patient was estimated to be 1.86 years.Total costs were estimated to increase by about $25 000($12 190 to identify intermediate-/high-risk patientsand at least $15 000 for chemotherapy, offset by savingsof $2344 because of lower risk of recurrence). The costutility of recurrence score testing for this cohort was$31 452 per QALY gained. Second, reclassifying NCCN-definedhigh-risk patients as low risk by the recurrencescore was cost saving; the added $7073 testing to identify1 reclassified patient was offset by the $15 000 savingsfor not using chemotherapy. Third, in a populationof 100 patients with characteristics similar to those ofthe NSABP B-14 participants, more than 90% of whomwere NCCN-defined as high risk, using the 21-gene RTPCRassay was expected to improve quality-adjustedsurvival and save costs.

Our analyses applied efficacy and costs assumptionsconcerning patterns of prescribing chemotherapy basedon physician surveys conducted in 2003. These analysesare important steps to calibrate the effects ofchemotherapy on quality of life and to validate use ofthe methodology to evaluate the 21-gene RT-PCRassay.52

No threshold of cost effectiveness has been universallyadopted as a standard that policymakers must usein deciding whether to adopt or fund new medical technologies.53,54 Other factors influencing funding decisionsinclude community and practioners'perceptions ofmedical necessity,55 quality of evidence, disparities ofcare,56 and impact of funding on overall budgets, especiallywhen addressing competing organizational objectives.In 2000, Earle and colleagues summarized theliterature on ranges of cost-effectiveness ratios foralmost 90 assessments of cancer interventions.38 Theyfound a large range in cost-effectiveness ratios, withsome cost-saving technologies, such as obtaining a biopsyfrom a 50-year-old man with elevated prostate-specificantigen levels,57 to other technologies havingcost-effectiveness ratios exceeding $1 million per QALYsaved, such as immune globulin for chronic lymphocyticleukemia.58 Sixty-four percent of cancer interventionshad a cost-effectiveness ratio of $50 000 or less.38Commenting on thresholds of cost effectiveness in general,Ubel et al recently argued that technologies withcost-effectiveness ratios of less than $100 000 wouldrepresent reasonable public investments.59 Arguingfrom a broad perspective of public investment in societalgoods, such as health, education, environment, anddefense, Garber and Phelps stated that interventionswould be considered reasonable public investments ifcost-effectiveness ratios were less than twice averageannual income.60 By any of these standards, the estimatedcost effectiveness of this prospectively validatedgenomic assay—when appropriately used to improveclassification of risk in patients with ESBC—is withinthe accepted range of other generally accepted healthcaretechnologies funded in the United States.

The commercial version of the assay has been availablesince early 2004, and more remains to be learnedabout its optimal use. Bast and Hortobagyi61 recentlyhighlighted several important questions about the assaythat require further research prior to its widespreadadoption in clinical practice, such as, how well does therecurrence score predict prognosis in women notreceiving hormone therapy? and, does the assay predictresponse to chemotherapy? Studies have been completed62and are ongoing to assess the prognostic andpredictive ability of the recurrence score—specifically,how the assay performs for patients with ER-or LN+disease and whether the test can predict response tochemotherapy. Preliminary data suggest that the recurrencescore has a positive correlation with the relativerisk reduction of chemotherapy, a finding that wouldmake the test even more useful: a high-risk score predictsgreater benefit of chemotherapy and viceversa.37,63

Our study complements Bast and Hortobagyi's editorialby revealing other factors that are likely to influencethe societal benefits and costs of the test. Wefound that these outcomes were highly influenced by(1) how patients and physicians interpret and use theresults of the assay and (2) whether the assay will beused in all patients with LN-, ER+ ESBC or restrictedto a subset of patients. For example, a hypotheticalscenario in which use of this assay could be inefficientis one in which patients reclassified from NCCN-definedhigh risk to low risk by the recurrence scorewere to rarely change their decision to receivechemotherapy. As shown here, it will be important tofollow trends of testing and treatment patterns usingavailable clinical databases and registries, and updateeconomic analyses in light of relevant clinical evidenceand data on evolving practice patterns.Moreover, because of the small percentage of patientswho were classified by NCCN as low risk, additionalevidence from other studies is needed on DRFS amongsuch women who are reclassified to high risk by therecurrence score. Our sensitivity analyses capturedthe uncertainty in this variable.

Newer agents being developed for adjuvant andneoadjuvant settings, such as extended use of aromataseinhibitors, use of trastuzumab, and dose-denseregimens, appear likely to further increase the cost ofchemotherapy. Our sensitivity analyses show that ascosts associated with chemotherapy increase, so doesthe cost effectiveness of the test, further highlightingthe need for more predictive approaches than are currentlyavailable to classify recurrence risk. Also, the 21-gene RT-PCR assay is a first-generation test. The fieldof molecular diagnostics is progressing rapidly withdevelopment of second-generation assays that willinclude more genes and that may further improve theprognostic and predictive accuracy of the assay. Fornow, the assay has been validated in patients with LN-,ER+ ESBC. This cohort was important to study initiallybecause, with enhanced screening, more patients arebeing diagnosed with ESBC. Implications of the assayfor patients with LN+ and ER-cancers has yet to bedetermined.

We deliberately omitted a number of factors thatmay further influence the clinical utility and economicsof the assay. For example, we did not include anadjustment in quality of life to account for the effect thetest may have on decision making for patients andphysicians. Patients have reported a strong desire forinformation to facilitate decision making and to buildconfidence in their decisions.63 How the assay mightaffect patients' attitudes about their decisions has notbeen assessed prospectively, but should be monitoredas use of the assay increases. Other questions worthpursuing include: Do patients feel the assay helpedthem to better understand the risks and benefits ofchemotherapy? Does the assay result influence theirdecision to undergo chemotherapy? Do they feel moreor less confident about their decision to receive orforego chemotherapy because of the assay result?

The assay also may have implications for otheraspects of management of ESBC, such as method andinterpretation of lymph node staging, interpretation ofimmunohistochemistry of lymph nodes, and intensity ofposttreatment cancer surveillance. With the databasescurrently available, assessing such potential impacts isdifficult. As genomic assays like the one studied herebecome more widely available, prospective studies willprovide information on how they influence variousaspects of quality of care, patient safety, and costs.

Experts widely perceive that adherence to guidelinesis significantly less than 100%.10,20,21,64,65 While notassessed here, a low score on the recurrence score mayconceivably boost patients' and physicians' confidencein foregoing chemotherapy when NCCN criteria alsodefine patients as having low risk of recurrence.

In summary, the recurrence score for predictingrisk of distant recurrence from LN-, ER+ ESBC wasshown prospectively to more accurately forecast outcomesthan existing guidelines. The recurrence score,derived from a 21-gene RT-PCR assay—one of the firstcommercially available genomic assays—shows potentialto align adjuvant chemotherapy decisions moreclosely with risk, thereby improving quality-adjustedsurvival and providing more efficient use of resourcesfor managing breast cancer. The results of this analysisshould help in developing guidelines to assure optimaland efficient use of the assay.

Acknowledgments

The authors thank Katherine Robertus, MPH, and 4 anonymous refereesfor review of the manuscript. We also thank Genomic Health and theNSABP with providing findings of analyses of NSABP B-14 study based onNCCN guidelines.

From The SPHERE Institute/Acumen, LLC, Burlingame, Calif (JH); the Department ofVeterans Affairs, Palo Alto, Calif (JH); the Department of Medicine, Stanford UniversitySchool of Medicine, Stanford, Calif (JH); the Department of Humanities and Social Sciences,Albany College of Pharmacy, Albany, NY (LEC); and the Department of Medicine,University of Rochester School of Medicine and Dentistry and the James P Wilmot CancerCenter, University of Rochester Medical Center, Rochester, NY (GHL).

This study was sponsored in part by Genomic Health, Inc., Redwood City, Calif. Dr.Hornberger's recent and current funding includes projects sponsored by county, state, andfederal government agencies (NIH, AHRQ, VA, HRSA, OTA), professional medical societies(Renal Physicians Association), non-profit agencies and foundations (Picker Foundation,Salt Lake City Health Access Program), and for-profit companies (Roche Pharmaceuticals,Genomic Health, Inc, Genentech, Inc). He is an Adjunct Clinical Professor of Stanfordattending at the VA, and derives no income from this or any other clinical activity. DrLyman receives grant funding from federal agencies and industry including GenomicHealth, Amgen, GlaxoSmithKline, and Ortho-Biotech.

Address correspondence to: John Hornberger, MD, MS, 1415 Rollins Road, Suite 110,Burlingame, CA 94010. E-mail: jhornberger@acumen-llc.com.

CA Cancer J Clin.

1. Jemal A, Tiwari R, Murray T, et al. Cancer statistics, 2004. 2004;54:8-29.

Lancet.

2. Early Breast Cancer Trialists' Collaborative Group. Polychemotherapy for earlybreast cancer: an overview of the randomised trials. 1998;352:930-942.

Breast.

3. Muss HB. Adjuvant therapy for older women with breast cancer. 2003;12:550-557.

Breast.

4. Bergh J. Best use of adjuvant systemic therapies II, chemotherapy aspects: doseof chemotherapy-cytotoxicity, duration and responsiveness. 2003;12:529-537.

Breast.

5. Cardoso F, Piccart MJ. The best use of chemotherapy in the adjuvant setting.2003;12:522-528.

Breast.

6. Pritchard KI. The best use of adjuvant endocrine treatments. 2003;12:497-508.

Breast.

7. Piccart MJ, Sotiriou C, Cardoso F. New data on chemotherapy in the adjuvantsetting. 2003;12:373-378.

J Clin

Oncol.

8. Goldhirsch A, Glick JH, Gelber RD, Coates AS, Senn HJ. Meeting highlights:International Consensus Panel on the Treatment of Primary Breast Cancer. SeventhInternational Conference on Adjuvant Therapy of Primary Breast Cancer. 2001;19:3817-3827.

J Clin

Oncol.

9. Goldhirsch A, Wood WC, Gelber RD, et al. Meeting highlights: updated internationalexpert consensus on the primary therapy of early breast cancer. 2003;21:3357-3365.

Tumori.

10. Palazzi M, De Tomasi D, D'Affronto C, et al. Are international guidelines forthe prescription of adjuvant treatment for early breast cancer followed in clinicalpractice? Results of a population-based study on 1547 patients. 2002;88:503-506.

Breast.

11. Senn HJ, Thurlimann B, Goldhirsch A, et al. Comments on the St. GallenConsensus 2003 on the Primary Therapy of Early Breast Cancer. 2003;12:569-582.

Arch Intern Med.

12. Li CI, Moe RE, Daling JR. Risk of mortality by histologic type of breast canceramong women aged 50 to 79 years. 2003;163:2149-2153.

Cancer.

13. Ernst MF, Voogd AC, Coebergh JW, Roukema JA. Breast carcinoma diagnosis,treatment, and prognosis before and after the introduction of mass mammographicscreening. 2004;100:1337-1344.

Ann Intern Med.

14. McCarthy EP, Burns RB, Coughlin SS, et al. Mammography use helps toexplain differences in breast cancer stage at diagnosis between older black andwhite women. 1998;128:729-736.

N Engl J Med.

15. Kassirer JP, Pauker SG. The toss-up. 1981;305:1467-1469.

Lancet.

16. Davies C, McGale P, Peto R. Variation in use of adjuvant tamoxifen. 1998;351:1487-1488.

J Clin Oncol.

17. Buban GM, Link BK, Doucette WR. Influences on oncologists' adoption ofnew agents in adjuvant chemotherapy of breast cancer. 2001;19:954-959.

Ann

Oncol.

18. Stiggelbout AM, de Haes JC, van de Velde CJ. Adjuvant chemotherapy innode negative breast cancer: patterns of use and oncologists' preferences. 2000;11:631-633.

J Cancer Res Clin Oncol.

19. Nagel G, Rohrig B, Hoyer H, Wedding U, Katenkamp D. A population-basedstudy on variations in the use of adjuvant systemic therapy on postmenopausalpatients with early stage breast cancer. 2003;129:183-191.

J Clin Oncol.

20. Harlan LC, Abrams J, Warren JL, et al. Adjuvant therapy for breast cancer:practice patterns of community physicians. 2002;20:1809-1817.

Med Care.

21. Bickell NA, McEvoy MD. Physicians' reasons for failing to deliver effectivebreast cancer care: a framework for underuse. 2003;41:442-446.

Lancet.

22. Brenton JD, Caldas C. Predictive cancer genomics—what do we need? 2003;362:340-341.

CAP Today.

23. Titus K. Reclassifying cancer, guided by genomics. 2001;15:1, 14-16, 18. Available at: http://www.cap.org. Accessed December 14, 2004.

Cancer Cell.

24. Ince TA, Weinberg RA. Functional genomics and the breast cancer problem.2002;1:15-17.

Medscape General Medicine: Hematology-Oncology

25. Mariani SM. Functional genomics: improving cancer prognosis and drug development.2003;5(1). Availableat: http://www.medscape.com/viewarticle/450095. Accessed December 14, 2004.

Cancer Lett.

26. Bisca A, D'Ambrosio C, Scaloni A, et al. Proteomic evaluation of core biopsyspecimens from breast lesions. 2004;204:79-86.

Science.

27. Garber K. Genomic medicine. Gene expression tests foretell breast cancer'sfuture. 2004;303:1754-1755.

N Engl J Med.

28. Paik S, Shak S, Tang G, et al. A multigene assay to predict recurrence oftamoxifen-treated, node-negative breast cancer. 2004;351:2817-2826.

29. Paik S, Shak S, Tang G, et al. Risk classification of breast cancer patients by theRecurrence Score assay: comparison to guidelines based on patient age, tumorsize, and tumor grade. Abstract presented at: 27th Annual San Antonio BreastCancer Symposium; December 8-11, 2004; San Antonio, Tex. Abstract 104.

Med Decis Making.

30. Sonnenberg FA, Beck JR. Markov models in medical decision making: a practicalguide. 1993;13:322-338.

Lancet.

31. Cole BF, Gelber RD, Gelber S, Coates AS, Goldhirsch A. Polychemotherapyfor early breast cancer: an overview of the randomised clinical trials with quality-adjustedsurvival analysis. 2001;358:277-286.

J Clin Oncol.

32. Elkin EB, Weinstein MC, Winer EP, et al. HER-2 testing and trastuzumab therapyfor metastatic breast cancer: a cost-effectiveness analysis. 2004;22:854-863.

J Clin Oncol.

33. Desch CE, Hillner BE, Smith TJ, Retchin SM. Should the elderly receivechemotherapy for node-negative breast cancer? A cost-effectiveness analysis examiningtotal and active life-expectancy outcomes. 1993;11:777-782.

N Engl J

Med.

34. Hillner BE, Smith TJ. Efficacy and cost effectiveness of adjuvant chemotherapyin women with node-negative breast cancer. A decision-analysis model. 1991;324:160-168.

Lancet.

35. Chang J, Wooten E, Tsimelzon A, et al. Gene expression profiling for the predictionof therapeutic response to docetaxel in patients with breast cancer. 2003;362:362-369.

J Clin Oncol.

36. Gianni L, Zambetti M, Clark K, et al. Gene expression profiles of paraffin-embeddedcore biopsy tissue predict response to chemotherapy in patients withlocally advanced breast cancer. Abstract presented at: American Society of ClinicalOncology (ASCO) 40th Annual Meeting; June 4-8, 2004; New Orleans, La.Abstract 501. 2004;22(suppl): Abstract 501. Available at:http://www.asco.org. Accessed December 14, 2004.

37. Paik S, Shak S, Tang G, et al. Expression of the 21 genes in the RecurrenceScore assay and prediction of clinical benefit from tamoxifen in NSABP study B-14and chemotherapy in NSABP study B-20. Abstract presented at: 27th Annual SanAntonio Breast Cancer Symposium; December 8-11, 2004; San Antonio, Tex.Abstract 24.

J Clin Oncol.

38. Earle C, Chapman R, Baker C, et al. Systematic overview of cost-utility assessmentsin oncology. 2000;18:3302-3317.

J Clin Oncol.

39. Love N, Ravdin P, Hortobagyi G, Grana G, Paley MF, Poltorack L, for theBreast Cancer Update Working Group. Influence of prior therapy on breast cancersurvivors' preferences for adjuvant systemic therapy in hypothetical scenarios. Abstractpresented at: American Society of Clinical Oncology (ASCO) 40th AnnualMeeting; June 4-8, 2004; New Orleans, La. Abstract 591. 2004;22(suppl): Abstract 591. Available at: http://www.asco.org. Accessed December 14, 2004.

2003 Drug Topics Red Book: The Pharmacist's Trusted Companion for More

Than a Century.

40. Montvale, NJ: Thomson Healthcare; 2003.

Breast Cancer, Version 1.

41. National Comprehensive Cancer Network. ClinicalPractice Guidelines in Oncology, v.1.2005. Jenkintown, Penn: NationalComprehensive Cancer Network, Inc; 2005. Available at: http://www.nccn.org/professionals/physician_gls/PDF/breast.pdf. Accessed April 5, 2005.

Health Care Financ Rev.

42. Fireman BH, Quesenberry CP, Somkin CP, et al. Cost of care for cancer in ahealth maintenance organization. 1997;18(4):51-76.

Health Aff (Millwood).

43. Hogan C, Lunney J, Gabel J, Lynn J. Medicare beneficiaries' costs of care inthe last year of life. 2001;20(4):188-195.

Health Serv Res.

44. Hoover DR, Crystal S, Kumar R, Sambamoorthi U, Cantor JC. Medical expendituresduring the last year of life: findings from the 1992-1996 Medicare currentbeneficiary survey. 2002;37:1625-1642.

Med Care.

45. Nattinger AB. Quality of care for breast cancer. 2003;41:341-343.

Cancer.

46. Kornblith AB, Herndon JE 2nd, Weiss RB, et al. Long-term adjustment of survivorsof early-stage breast carcinoma, 20 years after adjuvant chemotherapy.2003;98:679-689.

Med Decis Making.

47. Jansen SJ, Kievit J, Nooij MA, Stiggelbout AM. Stability of patients' preferencesfor chemotherapy: the impact of experience. 2001;21:295-306.

48. OncoSurvey.com. Breast cancer [survey]. Summary 4th quarter 2003. Detailsof survey design available at: http://www.oncosurvey.com/aboutoncosurvey.html.Accessed December 14, 2004.

Breast Cancer Res Treat.

49. Simon MS, Stano M, Hussein M, Hoff M, Smith D. An analysis of the cost ofclinical surveillance after primary therapy for women with early stage invasivebreast cancer. 1996;37:39-48.

Cost-Effectiveness in Health and

Medicine.

50. Lipscomb J, Weinstein M, Torrence G. Time preference. In: Gold M,Weinstein M, Lipscomb J, Torrence G, eds. New York, NY: Oxford University Press; 1996:214-246.

Adjuvant Therapy for Breast Cancer.

51. NIH Consensus Development Program. NIH Consensus Statement Online 2000 November 1-3;17(4):1-23. ConsensusStatement 114. Available at: http://consensus.nih.gov/cons/114/114_intro.htm.Accessed December 14, 2004.

Value Health.

52. Weinstein MC, O'Brien B, Hornberger J, et al. Principles of good practice fordecision analytic modeling in health-care evaluation: report of the ISPOR TaskForce on Good Research Practices—Modeling Studies. 2003;6:9-17.

Health Aff (Millwood).

53. Reinhardt UE, Hussey PS, Anderson GF. US health care spending in an internationalcontext. 2004;23(3):10-25.

Health Aff (Millwood)

54. Garber A. Cost effectiveness and evidence evaluation as criteria for coveragepolicy. [serial online]. May 14, 2004;web exclusive:W4-284-W4-296. Available at: http://content.healthaffairs.org. Accessed December 14,2004.

Health Aff (Millwood).

55. Singer SJ, Bergthold LA. Prospects for improved decision making about medicalnecessity. 2001;20(1):200-206.

Unequal Treatment:

Confronting Racial and Ethnic Disparities in Health Care.

56. Smedley BD, Stith AY, Nelson AR, eds, and Committee on Understandingand Eliminating Racial and Ethnic Disparities in Health Care. Washington, DC:National Academy Press; 2003.

Invest Radiol.

57. Gottlieb RH, Mooney C, Mushlin AI, Rubens DJ, Fultz PJ. The prostate:decreasing cost-effectiveness of biopsy with advancing age. 1996;31:84-90.

N Engl J Med.

58. Weeks JC, Tierney MR, Weinstein MC. Cost effectiveness of prophylactic intravenousimmune globulin in chronic lymphocytic leukemia. 1991;325:81-86.

Arch Intern Med.

59. Ubel PA, Hirth RA, Chernew ME, Fendrick AM. What is the price of life andwhy doesn't it increase at the rate of inflation? 2003;163:1637-1641.

J Health Econ.

60. Garber AM, Phelps CE. Economic foundations of cost-effectiveness analysis.1997;16:1-31.

N Engl J Med.

61. Bast RC Jr, Hortobagyi GN. Individualized care for patients with cancer—awork in progress. 2004;351:2865-2867.

62. Habel L, Quesenberry C, Jacobs M, et al. A large case-control study of geneexpression and breast cancer death in the Northern California Kaiser Permanentepopulation. Abstract presented at: 27th Annual San Antonio Breast CancerSymposium; December 8-11, 2004; San Antonio, Tex. Abstract 3019. Available at:http://www.abstracts2view.com/sabes. Accessed April 5, 2005.

Cancer.

63. Bruera E, Willey JS, Palmer JL, Rosales M. Treatment decisions for breast carcinoma:patient preferences and physician perceptions. 2002;94:2076-2080.

Ann Oncol.

64. Roila F, Ballatori E, Patoia L, et al. Adjuvant systemic therapies in women withbreast cancer: an audit of clinical practice in Italy. 2003;14:843-848.

Br J Cancer.

65. Bloom BS, de Pouvourville N, Chhatre S, Jayadevappa R, Weinberg D. Breastcancer treatment in clinical practice compared to best evidence and practiceguidelines. 2004;90:26-30.