Risk Assessment in Oncology Clinical Practice

November 1, 2003

Myelosuppression and neutropenia represent the major dose-limitingtoxicity of cancer chemotherapy. Chemotherapy-induced neutropeniamay be accompanied by fever, presumably due to life-threateninginfection, which generally requires hospitalization for evaluationand treatment with empiric broad-spectrum antibiotics. The resultingfebrile neutropenia is a major cause of the morbidity, mortality, andcosts associated with the treatment of patients with cancer. Furthermore,the threat of febrile neutropenia often results in chemotherapydose reductions and delays, which can compromise long-term clinicaloutcomes. Prophylactic colony-stimulating factor (CSF) has been shownto reduce the incidence, severity, and duration of neutropenia and itscomplications. Guidelines from the American Society of Clinical Oncologyrecommend the use of CSF on the basis of the myelosuppressivepotential of the chemotherapy regimen. The challenge in ensuring theappropriate and cost-effective use of prophylactic CSF is to determinewhich patients would be most likely to benefit from it. A number ofpatient-, disease-, and treatment-related factors are associated with anincreased risk of neutropenia and its complications. A number of clinicalpredictive models have been developed from retrospective datasetsto identify patients at greater risk for neutropenia and its complications.Early studies have demonstrated the potential of such models toguide the targeted use of CSF to those patients who are most likely tobenefit from the early use of these supportive agents. Additional prospectiveresearch is needed to develop more accurate and valid riskmodels and to evaluate the efficacy and cost-effectiveness of modeltargeteduse of CSF in high-risk patients.

ABSTRACT: Myelosuppression and neutropenia represent the major dose-limitingtoxicity of cancer chemotherapy. Chemotherapy-induced neutropeniamay be accompanied by fever, presumably due to life-threateninginfection, which generally requires hospitalization for evaluationand treatment with empiric broad-spectrum antibiotics. The resultingfebrile neutropenia is a major cause of the morbidity, mortality, andcosts associated with the treatment of patients with cancer. Furthermore,the threat of febrile neutropenia often results in chemotherapydose reductions and delays, which can compromise long-term clinicaloutcomes. Prophylactic colony-stimulating factor (CSF) has been shownto reduce the incidence, severity, and duration of neutropenia and itscomplications. Guidelines from the American Society of Clinical Oncologyrecommend the use of CSF on the basis of the myelosuppressivepotential of the chemotherapy regimen. The challenge in ensuring theappropriate and cost-effective use of prophylactic CSF is to determinewhich patients would be most likely to benefit from it. A number ofpatient-, disease-, and treatment-related factors are associated with anincreased risk of neutropenia and its complications. A number of clinicalpredictive models have been developed from retrospective datasetsto identify patients at greater risk for neutropenia and its complications.Early studies have demonstrated the potential of such models toguide the targeted use of CSF to those patients who are most likely tobenefit from the early use of these supportive agents. Additional prospectiveresearch is needed to develop more accurate and valid riskmodels and to evaluate the efficacy and cost-effectiveness of modeltargeteduse of CSF in high-risk patients.

Despite the recent introductionof a number of new agents forcancer treatment, chemotherapyremains the principal methodof systemic treatment in many of thenearly 1.3 million cancer patients inthe United States each year.[1,2]Along with surgery and radiotherapy,chemotherapy plays a major role in thetreatment of many common malignancies-including lung, breast, colon,and ovarian cancer-as well as thehematologic malignancies. The importanceof the dose-response relation ofchemotherapy agents has long beenrecognized.[3] Long-term studies haveshown the importance of maintainingchemotherapy dose to patient survival,especially in responsive and potentiallycurable malignancies.[4-8]Myelosuppression, specifically neutropenia,is the major dose-limitingtoxicity of cancer chemotherapy, constrainingthe regimens, doses, andschedules that can be used. Consequently,avoiding or reducing the riskof chemotherapy-induced neutropeniacontinues to be an important considerationin the treatment of patients withcancer.Chemotherapy-InducedNeutropeniaBecause neutrophils represent amajor defense against infection, neutropeniacan create a favorable environmentfor bacterial invasion andmultiplication, with the potential forrapidly spreading life-threatening infection.[9] The relative absence ofneutrophils also reduces the symptomsand signs of infection, with fever oftenbeing the only indication of infectionin a patient with neutropenia. Becauseof the risk of life-threateninginfection and sepsis, patients with febrileneutropenia must be treated aggressively.[9,10] Hospitalization forrapid evaluation and treatment withintravenous antibiotics remains thestandard of care for the majority ofthese patients.[11]Hospitalization for febrile neutropeniarepresents a considerable financialburden in the care of patients withcancer. The current costs of cancercare are estimated to be more than$150 billion per year, with the financialconsequences of hospitalizationaccounting for the largest single componentof these costs. Two economicanalyses have placed the mean lengthof hospitalization for febrile neutropeniaat 10 days and the average costof hospitalization at more than$20,000.[12,13] Substantial indirectand out-of-pocket costs have also beenreported in association with neutropeniaand its complications.[14] Neutropeniaalso has negative effects on patients'quality of life. When patientsare hospitalized for febrile neutropeniatheir well-being suffers from beingseparated from family.[15] Patientshospitalized with febrile neutropeniamay be anxious about infection, invasivemedical procedures, ineffectivetreatments, and death.[15] Further,preliminary research suggests an associationbetween the depth of theabsolute neutrophil count (ANC) nadirand a decline in quality oflife.[16,17] A neutropenia-specificquality-of-life instrument developed tostudy how neutropenia can affect specificquality-of-life domains such asfatigue and worry is undergoing furthervalidation.[18]Perhaps of greatest importance, thecomplications of neutropenia can leadto reductions in chemotherapy doseintensity, adversely affecting longtermpatient outcomes. The occurrenceof neutropenia often leads physiciansto reduce the chemotherapy dose ordelay subsequent cycles of treatment.The net result is the delivery of lowerrelative dose intensity.[19-21] In potentiallycurable cancers, such as earlystagebreast cancer and non-Hodgkin'slymphoma, there is considerable evidencethat such chemotherapy dose alterationsare associated with lower disease-free and overall survival.[4-6,8]The extent to which chemotherapydose intensity was modified in majorrandomized clinical trials has beenexplored in a systematic review of theliterature on chemotherapy in patientswith early-stage breast cancer andnon-Hodgkin's lymphoma. Data onrelative dose intensity delivered werereported inconsistently and in variousways in the trials analyzed in thisstudy, with 40% of the trials not reportingdose intensity information atall.[22] Studies of practice patternshave also shown the extent to whichchemotherapy dose reductions and delaysoccur in practice. A survey ofpractice patterns in more than 20,000patients treated with chemotherapy inthe adjuvant setting for breast cancerfound that 56% of patients weretreated with a relative dose intensityless than 85% of the reference standard,increasing to 67% of patientsaged 65 years and older.[20] Severalfactors may influence the relative doseintensity delivered, including bothplanned and unplanned dose reductionsand treatment delays (Table 1).The Role of Colony-StimulatingFactorProphylactic colony-stimulatingfactor (CSF) is associated with a reducedrisk of febrile neutropenia,documented infection, and the needfor dose reduction or treatment delay,often enabling the delivery of full doseintensity.[23-25]Clinical practice guidelines for theuse of the CSF have been developedby the American Society of ClinicalOncology (ASCO).[26] These guidelinescurrently support the routine useof prophylactic CSF in patients treatedwith a chemotherapy regimen with anaverage risk of febrile neutropenia of40% or greater. These guidelines weresupported by early economic analysesbased on a limited assessment of thedirect medical costs of hospitalizationfor febrile neutropenia at a single institution.[27] This model found thatCSF use is associated with a reductionin cost when the risk of febrileneutropenia is above the 40% threshold.The risk threshold is the breakevenpoint at which the added cost ofthe CSF is offset by the reduction inthe costs of hospitalization for febrileneutropenia. The costs of hospitaliza-

tion for febrile neutropenia have beenfurther studied both within single-institutionand in multi-institution studies.Incorporating updated cost estimatesinto this economic model providesthreshold risk estimates closerto 20% (Figure 1). The ASCO CSFguidelines are currently being revisedin consideration of this and other additionalclinical and economic information.Risk thresholds for the cost-savinguse of CSF have been further evaluatedin sensitivity analyses and areseen to depend on the assumedbaseline risk of febrile neutropenia. Itis clear that some chemotherapy regimensare more likely to cause seriousmyelosuppression than others, butmuch about the patterns and incidenceof chemotherapy-induced neutropeniawith the regimens remains unknown.A survey of the literature for the ratesof neutropenia in randomized clinicaltrials of chemotherapy found that therates of myelosuppression and itscomplications often were not reported.[22] When they were reported,the rates of hematologic toxicity withthe same and similar regimens variedgreatly, making it difficult to assess theactual risk.The actual risk has been shown tovary with a large number of treatment-,disease-, and patient-relatedcharacteristics. The use of a specifiedrisk threshold for a given regimen toselect patients for first-cycle CSF useassumes an average population riskprofile. There clearly is a need to "individualize"the risk by considering aspecific patient's risk factors. This canbe most reliably and reproduciblydone by using a risk model to selectpatients for CSF prophylaxis. Whetherto use CSF in accordance with a costminimizationmodel with a specifiedrisk threshold is an economic decisionthat is based on population risk, and itdoes not take into account the possibleclinical benefits in individual patientsin whom neutropenic complicationscould be avoided.[27]The optimal cost-effective use ofCSF hinges on using it in those patientswho are most likely to benefitfrom it, and being able to predict whothose patients are. In the absence ofreliable values for chemotherapy regimen-specific risks, clinicians oftenuse various patient characteristics toget an idea of a patient's risk for neutropeniaand its complications. TheASCO guidelines define "special circumstances"as patient populations inwhom the risk is high and primaryprophylaxis with CSF is recommended.These include extensive priortreatment, prior radiation therapy toconsiderable marrow-producing bone,and risk factors for infection, includ-

ing open wounds, low performancestatus, and reduced immune function.However, the list of these special circumstancesis not comprehensive, noris the list routinely used in selectingpatients for first-cycle CSF management.The lack of uniform standardscan leave oncologists uncertain aboutwhich patients are at greatest risk forneutropenic complications.A "watch and wait" approach isoften used, in which patients who experiencedsevere or febrile neutropeniain the first cycle are given CSF afterchemotherapy in later cycles. Such"secondary prophylaxis," while reasonablein lower-risk settings, is not adesirable option in those at greaterrisk, given the life-threatening natureof febrile neutropenia and demonstrationwith several chemotherapy regimensthat the greatest risk of the initialepisode of febrile neutropenia occursearly in the treatment course.[29-31] Initiating CSF later in the treatmentcourse forces most patients to experiencethe very complication theyare intended to prevent. Clearly, a newstrategy is needed to better select patientsfor early CSF management.Risk Models for NeutropeniaConsiderable research has been directedtoward identifying patient-, disease-,and treatment-related factorsassociated with the risk of neutropeniaand its complications. The focushas moved from the uncertainty of reportedrisk associated with variousregimens in highly selected populationsto a more focused assessment ofrisk in the individual patient based onhis or her clinical condition and treatmenthistory.[28] The purpose of clinicalpredictive models is to define theindependent contributions of the variousrisk factors for neutropenic complicationsand reduced dose intensity.By identifying patients at greater risk,such models can be used to target theCSFs more efficiently and cost-effectivelyto those most likely to benefitfrom such supportive care. Predictiverisk models more closely reflect thesituation in clinical practice, in whichpatients are selected for treatment onthe basis of their individual prognosticand predictive factors, while basingsuch decisions on explicit, objective,and reproducible measures thatcan be systematically applied and validated[28] (Figure 2).Risk models for neutropenic eventshave been studied most extensively inchemosensitive, potentially curablemalignancies with full-dose chemotherapy,namely early-stage breastcancer and non-Hodgkin's lymphoma.[32-35] Those risk factorsfound to be statistically significant inmultivariate models in two or morestudies are indicated in Figure 3.Clinical prediction models evaluaterelations between risk factors andclinical outcomes, such as hospitalizationfor febrile neutropenia or reducedchemotherapy dose intensity. Riskmodels can be based on pretreatmentcharacteristics alone (unconditional)or include the initial results with treatment(conditional). Some obviouspretreatment factors are advanced age,low performance status, comorbidities,and prior treatment. Additionalrisk factors found to be usefulin risk models in patients with non-Hodgkin's lymphoma include serumalbumin levels, lactate dehydrogenaselevels, bone marrow involvement, andsoluble tumor necrosis factor receptorlevels.[33-35]The most common type of conditionalmodel for predicting later neutropeniccomplications incorporatesthe patient's early hematologic responseto chemotherapy, which canidentify patients at increased riskfor subsequent neutropenic eventswho are likely to benefit from CSFsupport.[32,37] Conditional modelsare most useful with regimensthat are associated with a relativelylow average risk in the early cycles oftreatment, while unconditionalmodels are more useful when theearly risk of neutropenic events is relativelyhigh.An example of an unconditionalrisk model has recently been presentedthat identified five independentfactors associated with an increasedrisk of hospitalization for febrile neutropenia:age 65 years or greater, serumalbumin levels 3.5 g/dL or less,planned average reduced dose intensity80% or more, pretreatmentANC 1.5 109/L or less, and hepaticdisease.[36] This model was thenused to classify patients as beingat low, medium, or high risk accordingto how many risk factors werepresent.[36]Economic evaluations have shownthe ability of clinical prediction modelsto increase cost savings and thecost-effective application of CSF byenabling targeted use in patients atgreatest risk and most likely to benefit.[29,38] For instance, using a validatedconditional model for identifyingpatients treated with adjuvant chemotherapyfor early-stage breast cancerwho are at increased risk for futureneutropenic events that are likelyto compromise dose intensity, Silberand colleagues showed that a strategyof targeting G-CSF in the 50% of patientsat greatest risk was associatedwith an average cost-effectiveness ofapproximately $34,000 per life-yeargained.[38]Evaluating Risk ModelsMost of the risk models for chemotherapy-induced neutropenia thathave been developed to date have beenretrospective in design, fail to provideclearly defined hypotheses, and lackimportant data elements for adequateanalysis. Clinical predictive modelsshould be developed from studies thatminimize bias by stating the hypothesisin advance and providing clearinclusion and exclusion criteria. Anysubgroup analyses that are to be performedshould be stated in advance,and any differences in the outcomesbetween subgroups should be assessedby testing the interaction betweenthe prognostic factor and thevariable that defined the subgroupsrather than by separate analyseswithin the subgroups. Studies shouldbe designed to limit the potential formissing data, ideally to less than 10%.The issues of multiple comparisonsshould be considered when severalprognostic factors or cut points areevaluated, and tests of significanceshould be adjusted accordingly. Thepitfalls of stepwise multiple regressionmodels, including model instability,and exaggeration of coefficient estimates(and P values) should be acknowledged.The majority of the studies to datehave not been independently validatedin a similar but separate patient population.The issue of multiple testing orsmall sample sizes within subgroupsmust be properly addressed when apparentdifferences between subgroupsare observed.A risk model should ideally be developedfrom a prospective, hypothesis-driven study in a large representativepatient population. For this purpose,a prospective registry in differenttumor types in a large communitybasedoncology setting has been created.The registry will make it possibleto collect comprehensive patient data,including cycle-by-cycle informationon hematologic function. The purposeof the registry is to develop a powerfulrisk model for neutropenia that isreliable in routine clinical practice.[22,39]ConclusionsNeutropenia and its complicationsremain a major challenge in the managementof patients with cancer treatedwith cytotoxic chemotherapy. Colonystimulatingfactor is effective in reducingneutropenia and its complicationsacross a broad range of disease categoriesand treatment regimens. TheAmerican Society of Clinical Oncologyhas developed clinical practiceguidelines for the use of CSF that weresupported by early economic modelsbased on the average risk of febrileneutropenia with various chemotherapyregimens. While updated coststudies support lower cost-savings riskthresholds in the average setting, currentefforts have shifted the focus totreatment-, disease-, and patient-specificrisk factors for neutropenia andits complications.Early studies of predictive modelsfor identifying patients at increasedrisk for neutropenic complicationsshow promise for clinical decisionmaking and the effective targeting ofCSF use toward those most likely tobenefit. Such models have been shownto further reduce the cost associatedwith CSF support in patients treatedwith chemotherapy for cancer and tobe cost-effective in responsive andpotentially curable malignancies byenabling the delivery of full-dose-intensitytherapy on time. A multi-institutioneffort is currently under way toestablish a large prospective registryof patients treated with cytotoxictherapy in order to develop accurateand valid risk models to guide cliniciansin the optimal use of CSF.[39,40]

Disclosures:

The author(s) have no significant financial interest or other relationship with the manufacturers of any products or providers of any service mentioned in this article.

References:

1.

Edwards BK, Howe HL, Ries LA, et al:Annual report to the nation on the status ofcancer, 1973-1999, featuring implications ofage and aging on US cancer burden. Cancer94:2766-2792, 2002.

2.

Yancik R, Ries LA: Aging and cancer inAmerica: Demographic and epidemiologic perspectives.Hematol Oncol Clin North Am 14:17-23, 2000.

3.

Frei E 3rd, Canellos GP: Dose: A criticalfactor in cancer chemotherapy. Am J Med69:585-594, 1980.

4.

Bonadonna G, Valagussa P, Moliterni A,et al: Adjuvant cyclophosphamide, methotrexate,and fluorouracil in node-positive breastcancer: The results of 20 years of follow-up. NEngl J Med 332:901-906, 1995.

5.

Budman DR, Berry DA, Cirrincione CT,et al: Dose and dose intensity as determinantsof outcome in the adjuvant treatment of breastcancer. The Cancer and Leukemia Group B. JNatl Cancer Inst 90:1205-1211, 1998.

6.

Kwak LW, Halpern J, Olshen RA, et al:Prognostic significance of actual dose intensityin diffuse large-cell lymphoma: Results ofa tree-structured survival analysis. J Clin Oncol8:963-977, 1990.

7.

Epelbaum R, Faraggi D, Ben-Arie Y, etal: Survival of diffuse large cell lymphoma. Amultivariate analysis including dose intensityvariables. Cancer 66:1124-1129, 1990.

8.

Lepage E, Gisselbrecht C, Haioun C, etal: Prognostic significance of received doseintensity in non-Hodgkin’s lymphoma patients:Application to LNH-87 protocol. The GELA(Groupe d’Etude des Lympomes de l’Adulte).Ann Oncol 4:651-656, 1993.

9.

Bodey GP, Buckley M, Sathe YS, et al:Quantitative relationships between circulatingleukocytes and infection in patients with acuteleukemia. Ann Intern Med 64:328-340, 1966.

10.

Schimpff SC: Empiric antibiotic therapyfor granulocytopenic cancer patients. Am J Med80:13-20, 1986.

11.

Hughes WT, Armstrong D, Bodey GP,et al: 2002 guidelines for the use of antimicrobialagents in neutropenic patients with cancer.Clin Infect Dis 34:730-751, 2002.

12.

Kuderer NM, Cosler L, Crawford J, etal: Cost and mortality associated with febrileneutropenia in adult cancer patients (abstract998). Proc Am Soc Clin Oncol 21:250a, 2002.

13.

Gandhi SK, Arguelles L, Boyer JG: Economicimpact of neutropenia and febrile neutropeniain breast cancer: Estimates from twonational databases. Pharmacotherapy 21:684-690, 2001.

14.

Calhoun EA, Chang C-H, Welshman EE,et al: Evaluating the total costs of chemotherapy-induced toxicity: Results from a pilotstudy with ovarian cancer patients. The Oncologist6:441-445, 2001.

15.

Lyman GH, Kuderer NM: Filgrastim inpatients with neutropenia: Potential effects onquality of life. Drugs 62(suppl 1):65-78, 2002.

16.

Fortner BV, Stolshek B, SchwartzbergLS, et al: Decline in absolute neutrophil count(ANC) is associated with lower quality of lifein cancer patients receiving docetaxel (abstract2808). Proc Am Soc Clin Oncol 21:247b, 2002.

17.

Okon TA, Fortner BV, Schwartzberg L,et al: Quality of life (QOL) in patients withgrade IV chemotherapy-induced neutropenia(CIN) (abstract 2920). 21:275b, 2002.

18.

Calhoun EA, Chang C-H, Welshman EE,et al: A neutropenia-specific quality of life instrument:Rationale for the development of theFACT-N (abstract 1498). Proc Am Soc ClinOncol 21:375a, 2002.

19.

Link BK, Budd GT, Scott S, et al: Deliveringadjuvant chemotherapy to women withearly-stage breast carcinoma: Current patternsof care. Cancer 92:1354-1367, 2001.

20.

Lyman GH, Dale D, Crawford J: Incidenceand predictors of low dose intensity inadjuvant breast cancer chemotherapy: A nationwidestudy of community practices. J ClinOncol. In press.

21.

Chrischilles E, Link B, Scott S, et al:Factors associated with early termination ofCHOP, and its association with overall survivalamong patients with intermediate-grade non-Hodgkin’s lymphoma (NHL) (abstract 1539).Proc Am Soc Clin Oncol 21:385a, 2002.

22.

Dale D, Crawford J, Lyman GH:Myelotoxicity and dose intensity of chemotherapy:Reporting practices from randomizedclinical trials. J Natl Compr Cancer Netw1:440-454, 2003.

23.

Lyman GH, Kuderer NM, DjulbegovicB: Prophylactic granulocyte colony-stimulatingfactor in patients receiving dose intensivecancer chemotherapy: A meta-analysis. Am JMed 112:406-411, 2002.

24.

Crawford J, Ozer H, Stoller R, et al: Reductionby granulocyte colony-stimulating factorof fever and neutropenia induced by chemotherapyin patients with small-cell lung cancer.N Engl J Med 325:164-170, 1991.

25.

Trillet-Lenoir V, Green J, Manegold C,et al: Recombinant granulocyte colony stimulatingfactor reduces the infectious complicationsof cytotoxic chemotherapy. Eur J Cancer29A:319-324, 1993.

26.

Ozer H, Armitage JO, Bennett CL, et al:2000 update of recommendations for the useof hematopoietic colony- stimulating factors:Evidence-based, clinical practice guidelines.American Society of Clinical Oncology GrowthFactors Expert Panel. J Clin Oncol 18:3558-3885, 2000.

27.

Lyman GH, Lyman CG, Sanderson RA,et al: Decision analysis of hematopoieticgrowth factor use in patients receiving cancerchemotherapy. J Natl Cancer Inst 85:488-493,1993.

28.

Lyman GH: A predictive model for neutropeniaassociated with cancer chemotherapy.Pharmacotherapy 20(7 pt 2):104S-111S, 2000.

29.

Lyman GH, Kuderer NM, Crawford J,et al: Economic impact of pegfilgrastim usebased on the risk of febrile neutropenia (FN)in NHL patients treated with CHOP (abstract2384). Proc Am Soc Clin Oncol 22:593, 2003.

30.

Gomez H, Hidalgo M, Casanova L, etal: Risk factors for treatment-related death inelderly patients with aggressive non-Hodgkin’slymphoma: Results of a multivariate analysis.J Clin Oncol 16:2065-2069, 1998.

31.

Caggiano V, Stolshek B, Delgado D, etal: First and all cycle febrile neutropenia hospitalizations(FNH) and costs in intermediategrade non-Hodgkin’s lymphoma (IGL) patientson standard-dose CHOP therapy (abstract1810). Blood 98:431a, 2001.

32.

Silber JH, Fridman M, DiPaola RS, etal: First-cycle blood counts and subsequentneutropenia, dose reduction, or delay in earlystagebreast cancer therapy. J Clin Oncol16:2392-2400, 1998.

33.

Intragumtornchai T, Sutheesophon J,Sutcharitchan P, et al: A predictive model forlife-threatening neutropenia and febrile neutropeniaafter the first course of CHOP chemotherapyin patients with aggressive non-Hodgkin’s lymphoma. Leuk Lymphoma37:351-360, 2000.

34.

Tompkins KA, Imrie KR: Neutropenicevents in patients receiving CHOP chemotherapyfor large cell non-Hodgkin’s lymphoma:Predicting who is at risk (abstract1427). Blood 98:337a-338a, 2001.

35.

Voog E, Bienvenu J, Warzocha K, et al:Factors that predict chemotherapy-inducedmyelosuppression in lymphoma patients: Roleof tumor necrosis factor ligand-receptor system.J Clin Oncol 18:325-331, 2000.

36.

Lyman GH, Morrison VA, Dale DC, etal: Risk of febrile neutropenia among patientswith intermediate-grade non-Hodgkin’s lymphomareceiving CHOP chemotherapy. LeukLymphoma 44(12):2069-2076, 2003.

37.

Blay JY, Chauvin F, Le Cesne A, et al:Early lymphopenia after cytotoxic chemotherapyas a risk factor for febrile neutropenia.J Clin Oncol 14:636-643, 1996.

38.

Silber JH, Fridman M, Shpilsky A, et al:Modeling the cost-effectiveness of granulocytecolony-stimulating factor use in early-stagebreast cancer. J Clin Oncol 16:2435-2444,1998.

39.

Lyman GH, Crawford J, Dale DC, et al:Predicting the risk of chemotherapy-inducedneutropenia (CIN) in patients with breast cancer:Rationale for prospective risk model development(abstract 537). Breast Cancer ResTreat 76(1):S126, 2002.

40.

Dale DC, Wolff D, Agboola O, et al: Developmentof a risk model for neutropeniccomplications based on a prospective nationwideregistry (abstract 2229). Proc Am Soc ClinOncol 22:554, 2003.