ARTICLESCost-Minimization Analysis of ProphylacticGranulocyte Colony-Stimulating Factor AfterInduction Chemotherapy in Children WithNon-Hodgkins Lymphoma

Carole Rubino, Agne`s Laplanche, Catherine Patte, Jean Michon*

Background: Colony-stimulating factors promote the prolif-eration of certain bone marrow cell populations. The pri-mary objective of treating patients with these factors pro-phylactically following chemotherapy is to reduce the risk ofinfection, thereby minimizing the need for hospitalizationand parenteral antibiotics. The use of colony-stimulating fac-tors in children is widespread, despite the absence of con-clusive supportive data and the high cost of these drugs.Consequently, the costbenefit ratio of using prophylacticcolony-stimulating factors is an important issue in cancertherapy. Methods: During the period from January 1994through June 1996, 149 afebrile children with newly diag-nosed non-Hodgkins lymphoma were randomly assignedeither to receive granulocyte colony-stimulating factor (G-CSF) (lenograstim; 5 g/kg body weight per day subcutane-ously) or not to receive it (the control) at the end of the firsttwo courses of induction chemotherapy with cyclophospha-mide, vincristine, prednisone, doxorubicin, and methotrex-ate. A cost-minimization analysis was performed to assessthe cost of chemotherapy in each group and to quantify howmuch could be economized by prescribing G-CSF. Results:The total cost for induction chemotherapy was $29 765 inthe G-CSF-treated group and $30 774 in the control group,indicating that the treatment strategy with G-CSF wasslightly less expensive than the strategy without G-CSF(mean difference = $1009; 95% confidence interval = $1474to $3492). Conclusion: Treatment with G-CSF followingchemotherapy in children with non-Hodgkins lym-phomapreviously shown to be of limited clinical benefitalso does not appear to reduce the costs of chemotherapy. [JNatl Cancer Inst 1998;90:7505]

The main objective of prophylactic treatment with colony-stimulating factors (CSFs) is to reduce the risk of infection. Withsuch hematologic support, the duration of hospitalization and theneed for parenteral antibiotics can be minimized. The AmericanSociety of Clinical Oncology (1,2) recommends that primaryadministration of CSFs be reserved for adults, when febrile neu-tropenia is likely to occur in 40% or more of the patients. This

recommendation is also considered applicable to children, al-though there are no conclusive data to support its use. Currently,only a few clinical controlled trials of prophylactic CSFs inchildren have been published; most concern acute lymphoblasticleukemia, and all have yielded disappointing results (37). Re-sults concerning the incidence of febrile neutropenia are diver-gent: A statistically significant difference was observed in onestudy (6) but not in three other studies (3,4,7). Despite its finan-cial burden, granulocyte colony-stimulating factor (G-CSF) isnonetheless widely used in children. In such a context, it isimportant for oncologists to know whether adding CSFs to in-duction chemotherapy is clinically effective compared with che-motherapy alone and whether the cost of such treatment is jus-tified.

To determine whether the adjunction of CSFs was clinicallyand economically beneficial, we initiated in 1993 a randomizedtrial testing prophylactic lenograstim, a glycosylated recombi-nant human G-CSF, in children with non-Hodgkins lymphomareceiving chemotherapy. G-CSF was administered subcutane-ously (5 mg/kg body weight per day) at the end of the first twocourses of chemotherapy inducing neutropenia. The results ofthis clinical trial have been described elsewhere (8). In brief,G-CSF did not reduce either the rate of febrile neutropenia or theincidence of severe infection, and it was apparently of limitedclinical benefit. However, the duration of febrile neutropeniawas 2 days shorter in the group treated with G-CSF. In thiscontext, a cost-minimization analysis was warranted to deter-mine the cost of the procedure in each group and to quantify howmuch could be economized by prescribing G-CSF. We presenthere the results of the economic analysis of this clinical trial.

*Affiliations of authors: C. Rubino, A. Laplanche (Departement de Biostatis-tique et dEpidemiologie), C. Patte (Departement de Pediatrie), Institut GustaveRoussy, Villejuif, France; J. Michon, Departement de Pediatrie, Institut Curie,Paris, France.

Correspondence to: Agne`s Laplanche, M.D., Departement de Biostatistique etdEpidemiologie, Institut Gustave Roussy, rue Camille Desmoulins, 94805Villejuif Cedex, France. E-mail: laplanch@igr.fr

See Notes following References.

Oxford University Press

750 ARTICLES Journal of the National Cancer Institute, Vol. 90, No. 10, May 20, 1998

by guest on March 9, 2015

http://jnci.oxfordjournals.org/D

ownloaded from

Subjects and MethodsPatients and Treatment Protocol

During the period from January 1994 through June 1996, 149 children withnewly diagnosed non-Hodgkins lymphomas, treated at 28 centers of the FrenchSociety of Pediatric Oncology, were enrolled in the clinical trial. To be eligiblefor inclusion, participants had to be 17 years of age or younger and to have adiagnosis of either non-Hodgkins B-cell, T-cell, or anaplastic large-cell lym-phomas. In addition, each patient had to be undergoing induction chemotherapywith two courses of COPAD(M) (cyclophosphamide at a dose of 0.5 or 1 g/m2on days 24; vincristine at 2 mg/m2 on day 1 day 6; prednisone at 60 mg/m2on days 17; doxorubicin at 60 mg/m2 on day 2 high-dose methotrexate at 3or 8 g/m2 on day 1). Children with fever, with known infection, in need ofabsorbable antibiotics to decontaminate the gastrointestinal tract, or undergoingG-CSF treatment were excluded. Children infected with human immunodefi-ciency virus or with immunodeficiency or a previous cancer were also excludedfrom the study. All parents gave their written informed consent before treatmentrandomization. The study protocol was approved by the institutional ethics com-mittee and by the local institutional review board in each participating center, asrequired.

Randomization and treatment. Children were randomly assigned either toreceive or not to receive subcutaneous G-CSF with the use of a randomizationprocedure stratified according to the type of lymphoma (B-cell, T-cell, or ana-plastic large-cell). In addition, patients with B-cell lymphomas were stratifiedaccording to the prognostic subgroup (low, middle, or high risk of recurrence),which determined the chemotherapy protocol administered. Each participatingphysician was expected to contact the data manager in charge of the clinical trialon the first day of the first COPAD(M) course to ascertain the group to whicheach patient was to be assigned. Lenograstim (Granocyte) was supplied byBellon Laboratory (Montrouge, France) in the current packaging of 263 mg pervial initially designed for treatment in adults. Lenograstim was administeredsubcutaneously at a dose of 5 mg/kg body weight per day (equivalent to a doseof 150 mg/m2 per day) up to a maximum of 263 mg (i.e., one vial). Treatmentwas started on day 7 of each COPAD(M) course for a minimum of 6 days anda maximum of 15 days, but it was discontinued when the absolute neutrophilcount was above 5 105 cells/mL over a 48-hour period or when the absolutewhite blood cell count exceeded 2 107 cells/mL at any time during the courseof treatment.

Clinical follow-up and end points. Patients were monitored throughout thetwo consecutive induction courses of COPAD(M). Body temperature was re-corded twice a day, and blood cell counts were determined every 2 days for allpatients (hospitalized patients and outpatients). Our primary clinical end pointwas the incidence of febrile neutropenia, which was defined as an absoluteneutrophil count of 5 105 cells or fewer per milliliter and a rectal temperaturegreater than 38 C monitored two or more times within 2 hours. The secondaryclinical end points were the duration of febrile neutropenia, the incidence ofsevere infections [i.e., World Health Organization grade 3 or 4 (9)], the incidenceof grade 4 mucositis, and the duration of fever, neutropenia, and thrombocyto-penia.

Statistical analysis. Based on our previous experience, the incidence of fe-brile neutropenia without G-CSF administration was estimated to be approxi-mately 90%. It was necessary to examine a total of 150 patients to detect aminimum difference of 20% (from 90% to 70%) (type I error 4 0.05; power 480%; two-sided test) (10). Results were expressed as percentages or medianvalues for the consumption of medical resources or as mean values for cost data.The final concern for decision-makers is the total cost, which can be derived onlyfrom an analysis of the mean values (11). Therefore, when considering cost data,we used mean values. The consumption of medical resources by the two groupswas compared by use of nonparametric tests. We performed an intention-to-treatanalysis. All tests are two-sided.

Economic Analysis

A cost-minimization analysis was performed to assess the cost of the twoconsecutive induction courses of COPAD(M) with and without G-CSF. Allpatients who had received these two treatment courses were included, and costswere computed on an intention-to-treat basis to reflect treatment in routinepractice (11). We decided to consider this analysis from a health care point ofview. Therefore, both total inpatient costs and costs of G-CSF treatment admin-istered at home were taken into account. The evaluation method used 1996

health care costs standardized in the center that had the highest patient accrual,i.e., Institut Gustave Roussy, a hospital specializing in cancer treatment. Allcosts were converted into U.S. dollars at a rate of six French francs per dollar.

Resources used for cost measurement. The following economic end pointswere measured prospectively in physical units, for each child, from the first daythrough the last day of inductive chemotherapy: the length of the hospital stay forinpatients, the number of days G-CSF was administered, the number of days ofparenteral antibiotic or antifungal therapy, and the number of platelet and packederythrocyte transfusions.

The daily number of laboratory tests (e.g., hematologic, bacteriologic, andbiochemical) and the consumption of drugs for induction chemotherapy, paren-teral antibiotics, antifungal agents, and drugs other than G-CSF were not re-corded for each child in the study. These data were collected from medicalrecords from the Institut Gustave Roussy study patients. This information wasused to estimate the daily mean number of laboratory tests and the daily con-sumption of other drugs during the inpatient hospital stay.

Unit cost determination. The following four sources were used for unit costs:1) Hospitalization. Data from the accounts department of the Institut Gustave

Roussy were used to compute a per diem cost for an inpatient in the pediatricsdepartment. This per diem cost comprised the salaries of medical and nonmedi-cal staff, secretarial and housekeeping expenses, and the cost of office andmedical supplies. Overhead costs (approximately 25% of the total per diem cost)were included in this cost.

2) G-CSF (lenograstim). The acquisition prices paid by the central pharmacyof the Institut Gustave Roussy (negotiated prices and not national prices) wereused as unit drug prices.

3) Laboratory costs. To determine laboratory expenses, the 1996 NationalHealth Service unitary point value for all laboratory tests was applied to thecurrent quotation of laboratory tests (12).

4) Blood products. The official 1996 prices for blood products (platelet trans-fusion and erythrocyte transfusion) were directly fixed by the government eachyear (13).

The unit costs are summarized in Table 1, A, and the source of cost variationfor the sensitivity analysis is shown in Table 1, B.

Cost computation. All costs were calculated for each patient separately. Thecost of the least expensive formulation was used to compute an individual drugprice when several sources were available. Hospitalization, laboratory tests, andother drug consumption costs were calculated by multiplying their per diem costby the duration of the inpatient stay. The cost of G-CSF for each child was basedon the individual dose prescribed and the duration of the treatment. For paren-

Table 1, A. Summary of unit costs of resources used after inductionchemotherapy in children with non-Hodgkins lymphoma

Resources used Cost in 1996 U.S. dollars

1) HospitalizationDaily inpatient cost* 729Daily other drug costs* 32

2) Granulocyte colony-stimulating factorLenograstim, 263 mg (per vial) 116

3) LaboratoryDaily laboratory cost* 59

4) Blood productsPlatelet transfusion 738Red blood cell transfusion 328

Table 1, B. Source of cost variation for sensitivity analysis

Resources used Cost in 1996 U.S. dollars

1) HospitalizationPediatric daily inpatient cost

Institut Curie 424Ile de France area 498

Intensive care unit inpatient cost 14502) Granulocyte colony-stimulating factor

Lenograstim, 105 mg (per vial) 54

*This cost concerns only the inpatient stay.Cost for one episode.

Journal of the National Cancer Institute, Vol. 90, No. 10, May 20, 1998 ARTICLES 751

by guest on March 9, 2015

http://jnci.oxfordjournals.org/D

ownloaded from

teral antibiotics, the duration of the prescription and the average number ofantibiotics taken per day of treatment were recorded prospectively. Because thetype of antibiotic had not been recorded, the cost estimated for antibiotics wasbased on the protocol of the Institut Gustave Roussy (i.e., ceftazidime, vanco-mycin for the first 48 hours, and, if necessary, the adjunction of amikacin oramikacin plus teicoplanin). The cost of antibiotherapy for each child took intoaccount his or her body weight. We assumed that amphotericin B was theantifungal treatment and that a dose of 1 mg/kg body weight per day wasadministered. The cost of antifungal treatment was calculated for each patient bymultiplying the duration of antifungal treatment by the cost of the daily dose ofamphotericin B. The cost of chemotherapy was calculated separately for each ofthe three chemotherapy protocols: one for low-risk B-cell lymphomas; one formiddle-risk B-cell lymphomas, T-cell lymphomas, and anaplastic large-cell lym-phomas; and one for high-risk B-cell lymphomas. This calculation took intoaccount the body surface area.

The number of transfusion episodes was recorded for each child, instead of thenumber of erythrocyte and platelet units actually used. The price of each trans-fusion was the cost of the average number of blood units given per transfusionepisode. The number of blood units given was abstracted from the InstitutGustave Roussy study patients medical records. Costs of erythrocyte and plate-let transfusions were calculated for each child by multiplying this price by thenumber of platelet or erythrocyte transfusions each patient had undergone.

Sensitivity analysis. Sensitivity analysis is a process through which the ro-bustness of an economic model is assessed by examining changes in the resultsof the analysis when key variables change over a specified range (14,15). Thefollowing three sources of cost variation were incorporated into our pharma-coeconomic study for the purposes of sensitivity analysis:

1) Lenograstim supplied by the Bellon Laboratory (263-mg vials, allowingtreatment of a patient weighing about 50 kg at a dose of 5 mg/kg body weight perday) was not appropriate for treating children. We used the cost of the adultdosage in the end-point cost analysis ($116, Table 1, A). However, the BellonLaboratory has recently provided a new packaging of 105 mg per vial especiallyadapted for the treatment of children. The dose contained in this new packagingcosts less for children weighing less than 21 kg. In the sensitivity analysis, wetook into account the acquisition cost of this packaged pediatric dose ($54, Table1, B).

2) We used 1996 health care costs incurred in the center that had the largestpatient accrual (i.e., Institut Gustave Roussy, $729 per day, Table 1, A). How-ever, there are wide variations in per diem room costs between hospitals. Inthe sensitivity analysis, we allowed for variations in the per diem room costsproportional to the Indice Synthetique dActivite point value (16) observed in theInstitut Curie (ranking second for patient accrual) and in the Ile de France area.The Indice Synthetique dActivite is comparable to the fixed Prospective Pay-ment System per case in the United States. Cases are classified into categoriescalled Diagnosis-Related Groups (DRGs). The price fixed for each case belong-ing to a particular DRG is determined according to a weighing of the DRG in anational scale of costs and adjustments for hospital location (urban or rural), costof living in an area, etc. The per diem room costs thus calculated were $424 forthe Institut Curie and $498 for the Ile de France area (Table 1, B).

3) Few patients required hospitalization in an intensive care unit for a shortperiod, but inpatient costs for intensive care were evaluated according to thehospital cost in the pediatric care unit. The daily hospital cost per patient inintensive care in the Robert Debre Hospital ($1450 cost calculated by analyticaccounting) was used in the sensitivity analysis (Table 1, B).

Results

A total of 149 patients were included in 28 centers (AppendixTable 1): 75 in the G-CSF group and 74 in the control group.Two patients in the control group were excluded from theeconomic analysis: One patient died of a brain hemorrhage (un-related to chemotherapy) during the first course of chemo-therapy, and the other patient was excluded because of incom-plete data.

Medical Outcomes

The incidence of febrile neutropenia during the two courses,95% in the G-CSF group and 99% in the control group, did notdiffer significantly between the two groups, whereas the mediantotal duration of febrile neutropenia was 2 days shorter (P

lenograstim by the G-CSF group was almost balanced by theexpense incurred for an additional 2 days of hospitalization bythe control group.

Sensitivity Analysis

A sensitivity analysis was performed to test the robustness ofthe economic results. The overall results remained nearly thesame for each of the three sources of cost variation tested (Fig.1). G-CSF became more expensive only when the per dieminpatient cost weighted with the value of the Indice SynthetiquedActivite point at the Institut Curie (the center ranking secondaccording to patient accrual) was taken into account with theadult dosage and the cost of hospitalization in the pediatric careunit.

The results of the two treatment procedures were practicallyequivalent (control group cost was $321 less than the G-CSFgroup cost) after exclusion of four outliers (one in the G-CSFgroup and three in the control group). A patient was consideredas an outlier if his or her hospitalization exceeded 50 days duringthe two COPAD(M) courses.

Discussion

The statistical comparisons concerning economic items inthis study should be interpreted with caution because the numberof subjects was determined through a clinical criterion and didnot take into account the wide variation in costs (11). The studyis therefore too underpowered to detect differences in cost. Therobustness of the economic results was assessed by the sensi-tivity analysis. The total mean cost for induction chemotherapywas $29 765 with G-CSF and $30 774 without G-CSF. As aresult, the G-CSF treatment strategy was slightly less expensivethan the treatment strategy without G-CSF with a mean differ-ence of $1009 (95% confidence interval 4 $1474 to $3492).The difference between the cost of these two procedures ismainly imputable to the difference in the hospital stay and theuse of G-CSF. The cost of 2 additional hospitalization days forthe control group practically balanced out the G-CSF acquisitionprice. In the sensitivity analysis, the results were reversed onlywhen the per diem hospital cost per patient at the Institut Curiewas taken into account with the adult packaged dose and the costof hospitalization in a pediatric unit. In a similar manner, afterexclusion of children with an inpatient stay exceeding 50 days,the procedure with G-CSF became more expensive.

The per diem hospital cost used was high because the pedi-atric oncology department of the Institut Gustave Roussy alsoincludes a bone marrow transplantation unit. In addition, thetime of hospitalization represented a major part of the total costsand was the first factor to influence the difference in cost be-tween the G-CSF and control groups. As a result of these factors,the cost difference may have been overestimated.

Only one study (4) has reported economic end points forchildren. This double-blinded clinical trial concerned 167children with newly diagnosed acute lymphoblastic leukemiatreated with prophylactic G-CSF versus a placebo during in-duction chemotherapy. In that study, cost evaluation was basedon a fixed daily cost for G-CSF, with antibiotic and antifungaltreatment (costs for a child weighing 40 kg or with a body

Table 3. Consumption of medical resources in granulocyte colony-stimulatingfactor (G-CSF)-treated and control patients with non-Hodgkins lymphoma

Resource

Treatment group

Two-sidedP

G-CSF(n 4 75)

Control(n 4 72)

Inpatient length of stay, No.of days, median (range)

31 (954) 33 (978) .15

G-CSF use, No. of days,median (range)

17 (127) 0 (024)

Parenteral antibiotic 92% 99% .12No. of days, median (range) 13 (045) 16 (039) .07

Antifungal therapy 31% 44% .08No. of days, median (range) 0 (027) 0 (031) .09

Red blood cell transfusion 85% 82% .58No. of episodes, median (range) 2 (08) 2 (08) .85

Platelet transfusion 36% 28% .29No. of episodes, median (range) 0 (011) 0 (023) .30

Table 4. Mean cost per patient for two courses of induction chemotherapy with and without granulocyte colony-stimulating factor (G-CSF) and the differencebetween the two groups*

Treatment group

Resource

G-CSF (n 4 75) Control (n 4 72)Cost difference,Mean cost Range Mean cost Range

Inpatient 22 103 656139 366 24 675 729056 862 2572Other drugs 975 2901739 1089 3212509 114Laboratory 1784 5303177 1991 5884589 207G-CSF\ 2175 1205792 114 02780 2061Chemotherapy 312 97520 327 100736 15Parenteral antibiotics\ 1019 05671 1295 07302 276Parenteral antifungal agents\ 8 087 16 0199 8Platelet transfusions\ 738 08118 613 08856 125Red blood cell transfusions\ 651 02621 655 02621 4

Total cost 29 765 959759 129 30 774 927080 649 1009

*Including outpatient G-CSF costsThe G-CSF group underwent 17 days of treatment.Costs (1996) in U.S. dollars.Cost difference 4 the control group cost minus the G-CSF group cost.\A cost equal to zero was observed in the patients who did not receive the considered treatment.

Journal of the National Cancer Institute, Vol. 90, No. 10, May 20, 1998 ARTICLES 753

by guest on March 9, 2015

http://jnci.oxfordjournals.org/D

ownloaded from

surface area of 1.3 m2). No difference in the incidence of fe-brile neutropenia was reported, but the length of the inpatientstay due to febrile neutropenia was 4 days shorter (P 4 .011)in the G-CSF group and the total costs incurred for the twotreatment procedures were similar. Only three medico-economicstudies (1719) have been devoted to adult patients in a con-trolled clinical trial investigating prophylactic G-CSF duringchemotherapy without bone marrow transplantation. The costof G-CSF was not taken into account in one study (17); in theother two studies (18,19), the primary reason for using G-CSFwas to intensify chemotherapy. Results of a double-blindedclinical trial comparing G-CSF versus a placebo during in-duction chemotherapy for patients with solid tumors or lym-phomas have recently been published (20). Some cost end pointswere included in the discussion, but the methodology usedto determine costs was not described. The authors concludedthat G-CSF therapy multiplied standard outpatient costsper cycle of chemotherapy (cyclophosphamide, doxorubicin,vincristine, and prednisone or cyclophosphamide, doxorubicin,

and fluorouracil) twofold to threefold, but this conclusion isnot substantiated by economic data. Results of a double-blinded,multi-institutional clinical trial comparing granulocytemacrophage colony-stimulating factor (GM-CSF) versus aplacebo as adjunct therapy in 117 adults with acute myelo-genous leukemia were recently published (21). Patients treatedwith GM-CSF had fewer serious infections, but the durationof hospitalization and the number of transfusions were notsignificantly different between the two treatment groups. Theeconomic analysis was based on a model using resource con-sumption by seven patients. It estimated that patients treatedwith GM-CSF had lower costs of care than patients receiving aplacebo.

Daily and subcutaneous administration of prophylactic CSFsis an additional painful procedure. When used to treat children,it should therefore be restricted to those likely to benefit. Wewere unable to show a significant clinical benefit with the use ofG-CSF. The economic advantage reaped was too limited to pro-pose systematic G-CSF treatment in this context.

Appendix Table 1

Participating centers of the French Society of Pediatric Oncology Participating physicianNo of patients

(% of total)Institut Gustave Roussy, Villejuif Dr. C. Patte 20 (13)Institut Curie, Paris Dr. J. Michon 15 (10)Institut de Puriculture, Strasbourg Professor P. Lutz 9 (6)Hpital Purpan CHRU, Toulouse Dr. A. Robert 8 (5)Groupe hospitalier Pellegrin, Bordeaux Dr. Y. Perel 7 (5)Centre Lon Brard, Lyon Dr. D. Frappaz 7 (5)Hpital denfants CHU, Nancy Dr. C. Schmitt 7 (5)Hpital de la mre et lenfant, Nantes Dr. F. Mechinaud-Lacroix 7 (5)Hpital St. Louis, Paris Dr. A. Baruchel 7 (5)Hpital Huriez, Lille Dr. P. Vic 6 (4)Hpital denfants de la Timone, Marseille Dr. C. Coze 6 (4)Htel Dieu, Clermont Ferrand Professor F. Democq 5 (3)Hpital denfants CHRU, Dijon Dr. G. Couillault 5 (3)Hpital Jean Bernard, Poitiers Dr. F. Millot 5 (3)Hpital Sud CHRU, Rennes Dr. C. Bergeron 5 (3)Hpital Amiens Nord CHU, Amiens Dr. B. Pautard 4 (3)Hpital Debrousse, Lyon Dr. Y. Bertrand 4 (3)CHU, Caen Dr. P. Boutard 3 (2)Hpital Michallon CHU, Grenoble Dr. D. Plantaz 3 (2)Hpital Dupuytren CHU, Limoges Professor L. de Lumley 3 (2)Hpital Lenval enfants, Centre Lacassagne, Nice Dr. A. Thyss 3 (2)Hpital Clocheville, Tours Professor J. P. Lamagnre 3 (2)Hpital Robert Debr, Paris Dr. P. Rohrlich 2 (1)Centre Robert Debr CHU, Angers Dr. X. Rialland 1 (

References(1) American Society of Clinical Oncology. Recommendations for the use of

hematopoietic colony-stimulating factors: evidence-based, clinical practiceguidelines. J Clin Oncol 1994;12:2471508.

(2) American Society of Clinical Oncology. Update of recommendations forthe use of hematopoietic colony-stimulating factors: evidence-based clini-cal practice guidelines. J Clin Oncol 1996;14:195760.

(3) Dibenedetto SP, Ragusa R, Ippolito AM, Lo Nigro L, Di Cataldo A,DAmico S, et al. Assessment of the value of treatment with granulocytecolony-stimulating factor in children with acute lymphoblastic leukemia: arandomized clinical trial. Eur J Haematol 1995;55:936.

(4) Pui CH, Boyett JM, Hughes WT, Rivera GK, Hancock ML, Sandlund JT,et al. Human granulocyte colony-stimulating factor after induction chemo-therapy in children with acute lymphoblastic leukemia. N Engl J Med1997;336:17817.

(5) Kalmanti M, Lydaki E, Dimitriou H, Kalmanti T, Vlachonicolis I, Kam-bourakis A, et al. Effect of granulocyte colony-stimulating factor on che-motherapy-induced neutropenia in children with cancer. Pediatr HematolOncol 1994;11:14755.

(6) Welte K, Reiter A, Mempel K, Pfetsch M, Schwab G, Schrappe M, et al.A randomized phase-III study of the efficacy of granulocyte colony-stimulating factor in children with high-risk acute lymphoblastic leukemia.BerlinFrankfurtMunster Study Group. Blood 1996;87:314350.

(7) Calderwood S, Romeyer F, Blanchette V, Chan H, Doyle J, Greenberg M,et al. Concurrent RhGM-CSF does not offset myelosuppression from in-tensive chemotherapy: randomized placebo-controlled study in childhoodacute lymphoblastic leukemia. Am J Hematol 1994;47:2732.

(8) Patte C, Michon J, Laplanche A, Eyer D, Robert A, Perel Y, et al. Resultsof a randomised trial on prophylactic G-CSF during induction treatment ofnon-Hodgkins lymphoma. International Society of Pediatric OncologyXXIXth Meeting. Istanbul, September 1997.

(9) World Health Organization. Handbook for reporting results of cancer treat-ment. WHO offset publication. Geneva: WHO, 1979, No. 48.

(10) Casagrande JT, Pike MC, Smith PG. An improved approximate formula forcalculating sample sizes for comparing two binomial distributions. Biomet-rics 1978;34:4836.

(11) Drummond M, Menzin J, Oster G. Methodological issues in economicassessments of new therapies: the case of colony-stimulating factors. Phar-macoEconomics 1994;6(suppl 2):1826.

(12) Nomenclature des actes de biologie medicale. Union des caisses nationalesde Securite Sociale. Paris: Caisse Nationale de LAssurance Maladie desTravailleurs Salaries, 1994.

(13) Journal Officiel de la Republique Franaise. Paris: Direction des JournauxOfficiels, April 3, 1996.

(14) Drummond MF, Stoddart GL, Torrance GW. Methods for the economicevaluation of health care programs. Toronto: Oxford Medical Publications,1987.

(15) Agro KE, Bradley CA, Mittmann N, Iskedjian M, Ilersich AL, EinarsonTR. Sensitivity analysis in health economic and pharmacoeconomic stud-ies. An appraisal of the literature. PharmacoEconomics 1997;11:7588.

(16) Ministre du Travail et des Affaires Sociales. Mesure et analyse delactivite hospitalire MCO. Paris: Direction Regionale des Affaires Sani-taires et Sociales dIle de France, 1995.

(17) Souestre E, Qing W. Economic analysis of lenograstim in the correction ofneutropenia following chemotherapy for non-Hodgkins lymphoma. Phar-macoEconomics 1994;6(suppl 2):3643.

(18) Mapelli V, Schulenburg JM, Laaser U, Allhoff PG, Rossi F. Economicevaluation of lenograstim (glycosylated rHuG-CSF) in the treatment ofinflammatory breast cancer for Germany and Italy. PharmacoEconomics1994;6(suppl 2):2735.

(19) Drummond M, Davies L. Economic evaluation of lenograstim for prophy-laxis of chemotherapy-induced neutropenia in patients with small cell lungcancer. PharmacoEconomics 1994;6(suppl 2):4452.

(20) Hartmann LC, Tschetter LK, Habermann TM, Ebbert LP, Johnson PS,Mailliard JA, et al. Granulocyte colony-stimulating factor in severechemotherapy-induced afebrile neutropenia. N Engl J Med 1997;336:177680.

(21) Bennett CL, Golub R, Waters TM, Tallman MS, Rowe JM. Economicanalyses of phase III cooperative cancer group clinical trials: are theyfeasible? Cancer Invest 1997;15:22736.

NotesSupported in part by a Programme Hospitalier de Recherche Clinique grant

(PHRC nb 94009). Bellon Laboratory (Montrouge, France) supported the studywith a grant and lenograstim.

We thank Nolle Dupouy (Epidemiology Department) for data management,Emmanuelle Prot (Accounts Department) and Philippe Bourget (Hospital Phar-macy) for the evaluation of costs, Dr. Valerie Lapierre (Transfusion Unit), Pro-fessor Marcel Assicot, and Dr. Cyrille Tancrde (Department of Clinical Biol-ogy) for their help in retrieving cost data, and Lorna Saint-Ange for editing. Weare also indebted to Dr. Isabelle Durand-Zaleski (Henri Mondor Hospital, Paris)for her helpful advice.

Manuscript received October 28, 1997; revised March 9, 1998; acceptedMarch 10, 1998.

Journal of the National Cancer Institute, Vol. 90, No. 10, May 20, 1998 ARTICLES 755

by guest on March 9, 2015

http://jnci.oxfordjournals.org/D

ownloaded from