PDF hosted at the Radboud Repostory of the Radboud Unversty Njmegen The followng full text s a publsher's verson. For addtonal nformaton about ths publcaton clck ths lnk. http://hdl.handle.net/2066/23963 Please be advsed that ths nformaton was generated on 2018-01-19 and may be subject to change.
Granulocyte-M acrophage Colony-Stmulatng Factor Assocated Wth Inducton Treatment of Acute M yelogenous Leukema: A Randomzed Tral by the European Organzaton for Research and Treatment of Cancer Leukema Cooperatve Group By Robert Zttoun, Stefan Sucu, Franco Mandell, Theo de Wtte, Josef Thaler; Perre Stryckmans, Marcel Hayat, Marc Peetermans, Monque Cadou, Gabrel Solbu, Mara Concetta Pett/ and Roel Wllemze Purpose: To assess the value o f granulocyte-m acrophage colony-stm ulatng factor (GM-CSF) for nducton treatm ent of acute m yelod leukem a (AML)/ both for prm ng o f leukemc cells and fo r acceleraton o f hematopoetc recovery* Patents and Methods: GM-CSF was admnstered 5 fj, g / k g /d b y contnuous ntravenous (IV) nfuson durng nducton therapy w th daunorubcn (DNR) (days 1 to 3) and cytarabne (ARA-C) (days 1 to 7). A total of 102 patents w ere random zed onto fo u r arm s, as follow s: (1) GM-CSF 24 hours before and durng chem otherapy (arm + / ); (2) GM-CSF after chem otherapy untl day 28 or recovery of polym orphonuclear leukocytes (PMNs) (arm - / + ) ; (3) GM- CSF before, durng, and after chem otherapy (arm + /+ ); or (4) no GM-CSF (arm Stoppng rules w ere appled n case o f an n ta l WBC count greater than 30 x 109/L or a secondary ncrease of crculatng blast cells. Analyses were perform ed accordng to the ntenton-to-treat prncple. He m a t o p o e t c g r o w t h factors <hgfs), especally granulocyte colony-stmulatng factor (G- CSF) and granulocyte-macrophage colony-stmulatng factor (GM-CSF) are currently used n cancer chemotherapy to accelerate the recovery of hematopoess, although ther real mpact n terms of cost-effectveness s stll debated. They are also ncreasngly used for moblzaton of perpheral-bood progentor cells n vew of autologous transplantaton. In acute myeod leukema (AML), leukemc blast cels express receptors for HGFs, wth nterndvdual varablty n number, type, and affnty of receptors.1 Treatment From the Servce d Hématologe, Hôtel-Deu, Pars; Insttuí Gustave Roussy, Vllejuf, France; European Organzaton for Research and Treatment of Cancer Data Center; Insttute Jules Bordet, Brussels; Unversty of Antwerpen, Antwerpen, Belgum; Unverstà La Sapenza, Rome, Italy; St Radboud Hosptal, Njmegen; Leden Unversty, Leden, the Netherlands; and Unverstätklnk, Innsbruck, Austra. Submtted September 6, 1995; accepted February 2, 1996. Supported by Sandoz SA-Scherng Plough, Basel, Swtzerland, and by grants no. 5U10-CA11488-19 through 2UI0-CA11488-25 from the Natonal Cancer Insttute, Bethesda, MD. Address reprnt requests to Robert Zttoun, MD, Servce d Hématologe, Hôtel-Deu, 1, place du Pavs Notre Dame, 75181 Pars Cedex, France. 1996 by Amercan Socety of Clncal Oncology. 0732-183X/96/1407-0024$3.00/0 Results: The complete remsson (CR) rates were 77% (arm / )7 72% (arm + / - ), 48% (arm - / + ), and 46% (arm + /+ ). Patents randomzed to receve GM-CSF after nducton (arms - / + and +/+) had a sgnfcantly lower CR rate (P =.008) and a trend tow ard accelerated recovery of neutrophls, but no few er nfectons or nducton deaths. The lower CR rate appeared to be related to an ncreased resstance rate, w th persstent leukem a. The man sde effects of GM-CSF w ere flud retenton and hypotenson. Concluson: GM-CSF adm nstered durng nducton treatment of AML w th a DNR/Ara-C combnaton dd not provde any clncal beneft. Furthermore, there was a sgnfcant decrease n the CR rate w th more persstent leukema when GM-CSF was adm nstered durng the hypoplastc phase after the chem otherapy courses. J Cln Oncol 14:2750-2159. 1996 by Amercan Socety of Clncal Oncology. of AML wth HGFs theoretcally has two types of potental therapeutc effects as follows: (1) they can recrut leukemc cells nto cycle and thus enhance ther senstvty to chemotherapy2,3; therefore, ther admnstraton before and durng chemotherapy could lead to ncreased leukemc cell kll. (2) when admnstered after cytotoxc courses, they can stmulate normal hematopoetc progentors and thus accelerate hematopoetc recovery and reduce the morbdty and eventual mortalty from nfecton. The presence of receptors for HGFs n both normal and leukemc cells could challenge these therapeutc effects by two possble adverse consequences: (1) when gven before and durng chemotherapy, recrutment of normal plurpotent stem cells n cycle may ncrease the fracton exposed to cycle-dependent cytotoxc drugs and result n more prolonged marrow aplasa; (2) when admnstered after chemotherapy courses, stmulaton of resdual leukemc clones may occur, wth a rsk of resstance to nducton treatment or early relapse. Whether gven durng or after nducton courses, HGFs mght prevent chemotherapy-nduced apoptoss of leukemc cells.4,5 Several plot studes that combned GM-CSF durng nducton treatment of AML have ndcated that t could result n an hgher complete remsson (CR) rate when compared wth hstorcal controls.6'9 These plot studes were followed by prospectve randomzed trals. In two consecutve studes, Ohno et al10,11 have shown that G- CSF accelerates the recovery of neutrophls, wth a trend 2150 Journal of Clncal Oncology, Vo 14, No 7 (July), 1996: pp 2150-2159
GM-CSF FOR INDUCTION OF ACUTE MYELOID LEUKEMIA 2151 for less documented nfectons, whle the CR rate was not sgnfcantly ncreased. Results of other cooperatve trals that used GM-CSF or G-CSF have been recently publshed.12"16 Unfortunately, the desgns vary among studes, whch makes comparsons dffcult. In addton, no attempts were made to assess separately the two possble bologc effects: the prmng effect by admnstraton before and durng chemotherapy courses, and the acceleraton of normal hematopoetc recovery by postchemotherapy admnstraton. The European Organzaton for Research and Treatment of Cancer (EORTC) Leukema Cooperatve Group has decded to study prospectvely the value of these two types of admnstraton of GM-CSF durng the nducton phase n prevously untreated AML patents. PATIENTS AND METHODS The present tral (GM-CSF amendment) was desgned as an extenson of a large randomzed study, AML 8, of the EORTC and Gruppo Italano Malatte Ematologche Malgne dell Adulto (GIMEMA) Leukema Cooperatve Groups. In ths study, newly dagnosed patents aged 15 to 45 years were elgble for AML 8A, whch compared the followng three post-cr strateges: allogenec bone marrow transplantaton (BMT), autologous BMT, and short ntensve chemotherapy consoldaton. The results of the AML 8A study* not ncludng the amendment wth GM-CSF, have been recently publshed.17patents aged 46 to 60 years, n the meantme, were entered onto the AML 8B protocol, whch compared the same ntensve chemotherapy consoldaton and standard postremsson regmens of the EORTC and GIMEMA groups. The nducton course used n these two trals conssted of a combnaton of daunorubcn (DNR) and cytarabne (ARA-C), and was repeated once f a partal remsson (PR) was reached by day 28. Response to nducton treatment was classfed as CR f there were ^ 5 % bone marrow blast cells wth a normal blood count, PR f there were near normal blood counts and 6% to 25% bone marrow blasts, early death f death occurred before completon of the frst nducton course, death durng hypoplasa, and resstant dsease f there was persstent dsease n survvng patents. Patents In the present GM-CSF amendment protocol, patents were treated exactly as n the man study AML BA or SB, except for the randomzaton of GM-CSF durng the nducton phase. Only a lmted number of plot centers were allowed to partcpate n ths amendment, by enrollng all of ther patents consecutvely hosptalzed for AML, whereas the other centers contnued to treat patents accordng to the man protocols. Selecton crtera were almost the same as n the man studes. All patents aged 15 to 60 years wth prevously untreated AML were elgble on the bass of morphology crtera accordng to the French-Amercan-Brtsh (FAB) classfcaton.18patents wth blast crss of chronc myelod leukema, AML that supervened after other myeloprolferatve dseases, or after a myelodysplastc syndrome of more than 6 months duraton were excluded, as were patents wth a severe concomtant dsease or coexstent and progressve malgnancy. However, contrary to the man study, patents wth a poor performance status (World Health Organzaton [WHO] score > 2) or serum creatnne concentraton greater than 1.5 tmes the upper lmt of normal were excluded from the amendment. Informed consent was requred before randomzaton, accordng to local rules. All smears were centrally revewed for elgblty at dagnoss and for response. Treatment Protocol The desgn of the treatment protocol s shown n Fg 1. Patents receved nducton chemotherapy treatment that combned DNR 45 mg/m2/d by ntravenous (IV) push on days 1 to 3 and Ara-C 200 mg/nr7d by contnuons IV nfuson on days 1 to 7 and were randomzed to one of the four followng arms: (1) control arm wthout GM-CSF (-/-); (2) GM-CSF startng 24 hours before nducton chemotherapy and contnung untl completon of the chemotherapy course on day 7 (+/ ); (3) GM-CSF started mmedately after completon of the chemotherapy nducton course on day 8 and contnung untl day 28» or for a shorter tme n case of earler neutrophl recovery wth a polymorphonuclear leukocyte (PMN) count ^ 0.5 X 10lJ/L ( /+); or (4) GM-CSF started 24 hours before the chemotherapy nducton course and contnung untl day 28 or a PMN count 0.5 X Ï09/L (+/-H). Patents who dd not acheve a CR after one course, but met crtera that corresponded to a PR on day 28, receved a second dentcal nducton course, wth GM-CSF admnstered or not accordng to randomzaton arm. GM-CSF was provded by Sandoz/Scherng Plough (Basel, Swtzerland) as recombnant human Eschercha col-derved sterle lyophlzed powder. It was reconsttuted wth sterle water, dluted wth 0.9% sodum chlorde to a maxmum 50 ml total volume, and admnstered by contnuous IV nfuson through a central ven catheter at a daly dose of 5 ^g/kg. Accordng to the protocol, admnstraton of GM- CSF was delayed n patents randomzed to arms 2 and 4 n case of an ntal blast count greater than 30 X 10 VL, and was subsequently Fg 1. Protocol desgn. DNR 45 mg/m2/d on days 1 to 3, Ara- C 200 mg/m2/d by contnous IV nfuson on days 1 to 7. GM-CSF 5 fzgjkg/d by contnous IV nfuson. GM-CSF delayed f crculatng blast count > 30 x 109/L, started when < 20; Interrupted f, after chemotherapy, blasts are no longer seen and subsequently are > 1 x 109/L or persst and ncrease twofold. DNR ARA-C GM-CSF + /- - /+ +/ + Day % 0 O N o G M -C S F 0 1 2 3 4 5 6 7 8 * 28 (or PMN > 0.5)
2152 ZITTOUN ET AL started, after ntaton of chemotherapy, when the crculatng blast count decreased to less than 20 X 10y/L. GM-CSF was also dscontnued durng chemotherapy f the blast count ncreased to 50 X 10*7 L, and was rentroduced when the count decreased to less than 20 X 109/L. In addton, after the end of the chemotherapy course, GM- CSF was stopped f the crculatng blast cells were not or no longer detectable and subsequently reappeared to values ^ 1 X 109/L, or f they perssted and then ncreased by at least twofold. Furthermore, GM-CSF was stopped at any tme n case of severe sde effects, or allergc or anaphylactod reactons. To avod flud retenton, patents receved furosemde from the begnnng, plus IV nfuson of albumn n case of marked hypoalbumnema. Body weght was assessed daly; GM-CSF was nterrupted n case of weght ncrease by more than 5 kg, and rentated only f the body weght decreased by at least 3 kg. No GM-CSF was admnstered durng salvage treatment (combnng manly nlermedate-dose Ara-C and darubctn or amsacrne) n patents resstant to the nducton therapy or durng post-cr treatments. Statstcal Methods Randomzaton was centrally performed at the EORTC Data Center n Brussels usng the mnmzaton technque, wth age and treatment center beng used as stratfcaton factors. The 2 X 2 factoral desgn allowed evaluaton of two expermental groups: /+, GM-CSF admnstered after nducton chemotherapy, whatever the treatment appled durng nducton (e, wth or wthout GM-CSF); and +/*, GM-CSF admnstered before and durng nducton chemotherapy, whatever the treatment appled after nducton (e, wth or wthout GM-CSF). These two groups were compared wth the two correspondng control treatment groups through an a posteror stratfcaton; /+ to /, GM-CSF or not after nducton chemotherapy, whatever the treatment appled durng nducton (e, wth or wthout GM-CSF); and +/ to /*, GM-CSF or not durng nducton chemotherapy, whatever the treatment appled after nducton (e, wth or wthout GM-CSF), respectvely. In the treatment evaluaton, the followng end ponts were used: response to the frst nducton course, response to one or two nducton courses, overall response to nducton or salvage treatment, event-free survval (EFS) and overall survval. EFS was defned as the tme from evaluaton of nducton untl relapse or death n CR; patents who dd not reach CR after nducton were consdered as treatment falures at tme 0. Duraton of survval was calculated from the date of randomzaton untl death, whatever the cause of death. n all analyses, the ntenton-to-treat prncple was used, e, all patents were kept n the treatment arm allocated by randomzaton. All patents have been montored n the same way, rrespectve of treatment arm. The relatonshp between treatment group and response to chemotherapy course (categorzed as CR and no CR) was tested for statstcal sgnfcance usng the x1 test wth contnuty correcton.19each of the two comparsons (GM-CSF durng nducton, no v yes; and GM-CSF after nducton, no v yes) has been adjusted mutually by each other. For a 2 X 2 contngency table (treatment v response), the odds rato (OR) and ts 95% confdence nterval (Cl) was calculated usng the Confdence Interval Analyss program.20if the lower lmt of the 95% Cl s greater than 1, then the true OR s (wth 95% chances) greater than 1, e, the expermental group has a sgnfcant adverse mpact on the response rate. The Kaplan-Meer19 method was used to construct EFS and survval curves. Comparson between the groups for treatment outcome was tested for statstcal sgnfcance usng the two-sded log-rank test. The relatve rsk of death n the expermental versus the control group was estmated va the OR method, along wth ts 95% Cl.21 For multvarate analyses wth bnary outcome, the lnear logstc model19was used, and for EFS analyss, Cox s proportonal hazards model19was used. The ntal am of the study was to randomze 600 patents n order to delect an mprovement from 65% to 75% n terms of CR rate (OR = 0.63, alpha = 0.05, beta = 0.15). Due to the rsk of yeldng worse results n the GM-CSF-contanng groups, partcularly n the /+ group, close montorng of the tral was performed. After 103 patents had been recruted onto the tral, and the frst 93 were evaluated for response, t appeared that the estmated OR for the comparson /+ versus / was sgnfcantly (P =.01) greater than 1, and that ts correspondng 99.9% Cl dd not contan the ntal targeted value (e, 0.63). It was therefore decded to stop the tral prematurely, RESULTS Characterstcs of Treatment Arms and Groups A total of 103 patents from eght centers were ncluded from December 1990 to November 1992. All patents were elgble for entry onto the study. One patent randomzed to the +/+ arm, who had severe complcatons from the central venous access devce that allowed only pallatve chemotherapy wthout GM-CSF, was not assessable for response. The man characterstcs of the remanng 102 elgble and assessable patents are lsted n Table 1. The control group s characterzed by slghtly fewer unfavorable prognostc features, such as WBC count greater than 100 X 109/L and fever at dagnoss, and more patents wth Auer rods present n blast cells. Cytogenetcs were performed n 58 patents and showed an even dstrbuton n the four treatment arms when grouped nto good, ntermedate, and poor prognostc categores accordng to Keatng s classfcaton.21 Four patents randomzed to the + /, -/+, and +/+ arms dd not receve GM-CSF because of leukostass, vascults and lung nfltraton, skn vascults, and leukemc pleurts (Table 2). All four patents receved the chemotherapy accordng to the AML 8 protocol and were kept for analyss n ther treatment arm allocated by randomzaton. Table 2 also lsts the number of patents n whom the start of GM-CSF was, accordng to the protocol, delayed or prematurely stopped for ncreased blast cells or toxcty. Due to an ntal hgh WBC count, the start of GM-CSF was postponed n 20 patents. In fve patents who receved GM-CSF before nducton chemotherapy (arms +/ and -1-/+), an ncreased WBC count by twofold to 15-fold was observed on day 1 or 2. Ths ncrease durng GM-CSF dd not seem to correlate wth response (three n CR and two wth resstance). In addton, GM-CSF was temporarly nterrupted n eght patents and prematurely stopped n 43. The reasons for premature stoppng of GM-CSF are also lsted n Table
GM-CSF FOR INDUCTION OF ACUTE MYELOID LEUKEMIA 2153 Table 1. Patent Characterstcs at Dagnoss by Treatment Arm Treatment Arm Characterstc -/- (n = 26) +/- In = 25) -/+ (n = 27) +/ +(n 24) Age, years Medan 45 45 42 42 Range 26-59 22-58 17-59 2)-55 Sex rato (male/female) 0.86 1.08 1.08 1.18 Performance status grade 2 v 0 or 1 3 4 3 4 Fever > 38 C at entry 4 8 ó 8 WBC count (X 107L) 30-99 8 7 7 9 Al o o 0 2 2 3 Auer rods postve 12 10 10 9 FAB subfype M1/M2/M3 2/7/1 4 /9 /2 4 /6 /0 2/7/1 M4/M5/M6/M7 7/8/1 2 /6 /2 7 /8 /2 9/4/0/1 Cytogenetc rsk group (good/ntermedate/poor) 3/9/4 2 /6 /6 3/7/8 3/4/3 2, The more frequent reasons were early PMN recovery (16 patents) and ncreased crculatng blast cells (11 patents), whch led to stoppng GM-CSF on days 16 to 26 (medan, 22.6) and 11 to 24 (medan, 19.3)* respectvely. The other reasons were sde effects assumed to be due to GM-CSF (manly fever, flu-lke syndrome, flud retenton, and cardac falure), and, n sx cases, severe nfecton assocated wth serous clncal problems. Response to Inducton and Salvage Treatments Response to nducton treatment, accordng to randomzaton, s lsted n Table 3. The hghest CR rate after the frst nducton course was observed n the control arm. The three treatment arms showed a trend for a lower CR rate and an ncreased resstance rate, especally n the two arms /+ and +/-K The effect of GM-CSF was analyzed by treatment group (Table 4) after the frst nducton course or after the whole nducton treatment (one or two courses). The dfferences between CR rates of the groups randomzed to receve or not receve GM-CSF durng nducton chemotherapy were not sgnfcant, whether one consders the results after the frst cycle only (51.0% v 60.4%) or the overall results (59.2% v 62.3%). In contrast, the CR rate was sgnfcantly lower for the group randomzed to receve GM-CSF (group /+) durng the postchemotherapy perod, compared to the group / ; 43.1% versus 68.6% {P =.015) after the frst course and 47.1% versus 74.5% (.P.008) after one or two courses. These dfferences were manly attrbutable to an ncreased resstance rate n the group /+ compared wth */. On the bass of the treatment protocol, admnstraton of GM-CSF was delayed n case of an ntal WBC count greater than 30 X 109/L. However the CR rate n the group / + was stll sgnfcantly lower than n the group «/ after adjustment for WBC count. If one takes nto consderaton several factors that may nfluence the CR rate after nducton, such as age, sex, FAB cytology subtype, fever, WBC count, platelet count, performance status, Auer rods, and treatment group, the lnear logstc model showed that treatment group ( /- v */+) was the most predctve varable (P =.005), fol- Varable Table 2. Patents n Whom GM-CSF Was Not Gven, Delayed, or Prematurely Stopped by Treatment Arm Treatment Arm h/~-{n «25) /+ (n = 27} +/+ (n 24) No. % No. % No. % Total IN - 76) GM-CSF not gven 1 4 2 7 1 4 4 GM-CSF delayed 9 36 11 46 20 GM-CSF prematurely stopped 2 8 21 78 20 83 43 Man reason for nterrupton Early PMN recovery 11 41 5 21 16 Increase of blast cells 4 15 7 29 11 Supposed toxcty 2 8 3 11 2 8 7 Cardac falure 1 4 2 8 3 Severe nfecton/other 2 7 4 17 6
2154 ZITTOUN ET AL Table 3, Response to Inducton Treatment by Treatment Arm Treatment Arm +/- /+ +/+ Varable No. % No. % No. % No. % Frst course CR 20 76.9 15 60.0 12 44.4 10 4 1.7 PR 0 0 2 8.0 1 3.7 2 8.3 Resstance 4 15.4 7 28.0 13 48.1 10 41.7 Death 2 7.7 1 4.0 1 3.7 2 8.3 1 or 2 courses CR 20 76.9 18 72.0 13 48.1 11 45.8 Resstance 4 15.4 6 24.0 13 48.1 n 45.8 Death 2 7.7 1 4.0 1 3,7 2 8.3 Total 26 25 27 24 lowed by the presence of Auer rods (P =.02) and fever, whch was margnally mportant {P.09), After salvage treatment wthout GM-CSF -was admnstered to patents wth resstant dsease, the overall CR rate of the group /+ was stll lower, but not sgnfcantly (P =.11), than that of the group */ : 66.7% versus 80.4%. A smlar nonsgnfcant trend (P,34) was observed for the comparson +/ versus / : 67.3% versus 77.4%. EFS and Survval Table 5 lsts treatments gven after the completon of nducton treatment. Of 62 patents who acheved a CR, 12 receved standard mantenance and 16 receved one or two cycles of ntensve chemotherapy consoldaton, whereas 20 have been transplanted. There were a few more patents allografted n the +/ arm, but n general there was a good dstrbuton of postnducton treatments among the four arms. Table 5 also ndcates that fewer relapses have been reported n the control arm ( / ), and more n the GM-CSF-contanng arms. The hghest rate of death n frst CR has been reported n the +/ arm, a fndng whch s probably due to the hghest ncdence of allografts. The lmted number of patents n the treatment arms and groups precludes any vald comparson of the duraton of CR and of dsease-free survval. The EFS tme from evaluaton of the last nducton course was sgnfcantly shorter (log-rank P =.02) for the */+ group than for the / group, whereas the dfference between groups / and -f/ was not sgnfcant (P =,16) (Fg 2). Usng Cox s model, after adjustment for the presence of Auer rods, whch was the only sgnfcant Table 4. Response to Inducton Treatment by Treatment Group Treatment Group Varable /- /+ -/ +/ No. % No. % No. % No. % Frst course CR 35 68.6 22 43.1 32 60.4 25 51,0 PR 2 3.9 3 5.9 1 1.9 4 8.2 Resstance 11 21.6 23 45.1 17 32.1 17 34.7 Death 3 5.9 3 5.9 3 5.7 3 6.1 P*.015.41 OR 2.93 1.53 95% Cl 1.29-6.61 0.68-3.46 1 or 2 courses CR 38 74.5 24 47.1 33 62.3 29 59.2 Resstance 10 19.6 24 47.1 17 32.1 17 34.7 Death 3 5.9 3 5.9 3 5.7 3 6.1 Pm.008,86 OR 3.31 1.18 95% Cl 1.43-7.64 0.51-2.70 Total 51 100 51 100 53 100 49 100
GM-CSF FOR INDUCTION OF ACUTE MYELOID LEUKEMIA 2155 Table 5. Post-CR Treatment and Outcome After Inducton Treatment by Treatment Arm Treatment Arm Varable + /- -/+ + /+ Total No. % No. % No. % No. % No. % Treatment Toxcty -+ no 1 0 1 2 4 treatment Standard mantenance 4 4 1 3 12 Intensve consoldaton 11 7 5 3 16 Autologous BMT 1 2 4 2 9 Allogenec BMT 3 5 2 1 11 Outcome Contnued CR 14 70.0 5 27.8 5 38.5 5 45.4 29 46.8 Bone marrow relapse 5 25.0 8 44.4 7 53.8 5 45.4 25 40.3 CNS relapse 0 0.0 5.6 0 0.0 0 0.0 1 1.6 Death n frst CR 1 5.0 4 22.2 1 7.7 1 9.1 7 11.3 Total 20 100 18 100 13 100 11 100 62 100 prognostc factor for EFS (P =.015), smlar results were obtaned regardng the treatment groups. The overall survval rate from randomzaton was 32% at 3 years, wth a medan estmaton of 15 months and a medan follow-up duraton of 34 months. The total numbers of deaths n the four arms were 10, 17, 19, and 17, respectvely. A trend for a hgher death rate (log-rank P -.07) was observed n the /+ group versus */ group (OR - 1.51; 95% Cl, 0.92 to 2.49). For the comparson 47 group versus / group, the dfference was smaller (log-rank P =.37; OR = 1.26; 95% Cl, 0.76 to 2.07). Tme o Recovery of Neutrophls The tme from start of nducton to recovery of a neutrophl count greater than 0.5 X 109/L was studed by treatment arm and group, for patents who acheved a CR after the frst nducton course. Comparson among the four treatment arms showed no sgnfcant dfference (P =.28). However, a trend was observed wth a shorter duraton of neutropena n the +/+ arm and a longer duraton n the control arm (Fg 3). In addton, there was a nearly sgnfcant nverse correlaton between treatment arm (arm 1 to 4) and duraton of neutropena (log-rank test for lnear trend, P =.09), Toxcîes The man toxctes observed by treatment arm are lsted n Table 6, whch also provdes the number of clncally or mcrobologcally documented nfectons n the four randomzed arms. Flud retenton, weght gan, and hypotenson were manly observed n the two arms wth postchemotherapy admnstraton of GM-CSF. On the other hand, the number of nfectons was not reduced, and the number of days wth fever, antbotcs, and tme spent n the hosptal n the three treatment arms were not lower than n the control arm. Platelet and RBC transfusons were smlar n the four arms (data not shown). DISCUSSION The am of the present tral was to mprove the outcome of nducton treatment and the EFS n AML, by admnstraton of GM-CSF, usng varous schedules. However, our results appeared dsappontng: none of the combned modaltes of GM-CSF and chemotherapy was superor to the standard DNR/Ara-C regmen. Wth the present dose schedules of GM-CSF, t s unlkely that the CR rate n AML could be mproved by addton of GM-CSF. On the contrary, there are ndcatons that GM-CSF, when admnstered durng the postnducton chemotherapy perod, could ncrease the rsk of resstance to nducton chemotherapy. Unfortunately, ths concluson s based on a lmted number of patents, snce the tral had to be stopped prematurely. Confrmatory studes would be useful, but admnstraton of GM-CSF after chemotherapy courses to patents wth resdual leukemc cells should be dscouraged. However, ths precauton mght reduce the chance of confrmaton of our negatve results. The control arm contaned, to a certan extent, more favorable prognos tc factors n comparson wth the other randomzed treatment arms. Random bas may occur n lmted seres, but wth such central randomzaton, systematc bases are avoded and the valdty of statstcal tests guaranteed. A posteror stratfcaton by mportant prognostc factors has been performed to adjust for possble mbalances of known factors between the treatment groups, wthout changng the ntal concluson. Statstcal analyss was performed on the bass of the ntenton-to-treat prncple. In several patents random-
2156 ZITTOUN ET AL % 100 - p r 90-0 b a 80 - b 11» A N O 51 32 51 41./+ Treatment group t 70 Logrank P = 0.02 y» - 50 40 30 20 1 T L «U JLL J L.1LIJ l_l_p 10 0 12 18 24 r 30 36 42 48 (months) Number of patents at rsk Fg 2. EFS by treatment group. (A) G ro u p */- (arms - / % 100 P r0 b ab9 90 51 51 B 30 17 23 12 17 9 13 7 11 4 N 53 49 8 3 O 34 39 3 J- 1./+ Treatment group - and +/~) v / + (arms / + and + /+ ); (B) group -/-(a rm s - / - and - / + ) v + / (arms +/ - and +/+). N, number of patents; O, observed number of events (no CR after nducton courses, relapse or death n frst CR). 1! * t y 70 60 Logrank P 0.16 50 40! _ L 30 L_ ) 20-1 LI_ 10 0 T 6 I 12 18 24 ~r 30 36 ~r 42 48 (months) Number of patents at rsk : 53 26 20 15 12 9 7 3 -/. 49 21 15 11 8 6 4 1 +/. zed to the +/ and +/+ arms, GM-CSF was delayed, accordng to the protocol, because of a hgh ntal WBC count. Also, once started, GM-CSF was prematurely stopped n 41 of 51 patents randomzed to the /+ or +/+ arms, manly because of ncreased blast cells or early PMN recovery. In addton, four patents dd not receve the growth factor at all, because of persstng sgnfcant levels of crculatng blast cells or symptoms related to leukostass. Despte these precautonary measures, the CR rate was sgnfcantly lower when GM-CSF was admnstered durng the postnducton perod, and ths dfference remaned hghly sgnfcant after adjustment for ntal WBC count. All smears were centrally revewed, especally for response. Ths revew, and the subsequent hematologc evaluatons allowed us to rule out a msnterpretaton of
GM-CSF FOR INDUCTION OF ACUTE MYELOID LEUKEMIA 2 1 5 7 100 I Fg 3. Tme to recovery of PMN count s: 0.5 x 109/L n the A randomzed arms, durng the frst nducton course, for patents who acheved a CR after ths course. Medan tme n days; 24.5 (arm 22.2 (arm + / - ), 22.0 (arm /+), and 19.5 (arm +/+). N, number of patents; O, observed number of patents who reached a PMN count > 0,5 x 109/L. p r 90-0 b a 80 1 b j 70 y 60 H 50-40 - 30-20 - T L 1 I ~ 1 TL I?L N Treatment arm O 20 2 0 15 15 + /- 12 12 -/+ 10 10 + / + Logrank P = 0.09 (trend) 10-0 T 2 responses. Such rsk of erroneous evaluaton had been ponted out, wth respect to possble cytologc consequences of the admnstraton of HGFs at the bone marrow level: an underestmaton of the CR rate may result from stmulaton of early normal progentor cells that could smulate leukemc blast cells, or by a transent stmulaton of leukemc cells, whch could dsappear after stoppng the growth factor, as reported by Büchner et al.7 On the other hand, stmulaton of mature granulopoess could dlute the resdual leukemc cells and lead to an overestmaton of the CR rate. In fact, these cytologc modfcatons could manly lead to confuson between CR and PR, but our observaton of a lower response rate s stll found when PR and CR are consdered together, wth a lower total response rate n patents allocated to receve GM-CSF durng the post nducton courses. It s frequently assumed that the admnstraton of GM-CSF before nducton chemotherapy courses m g h t stmulate prolferaton of leukemc cells and/or nduce recrutment nto the cycle. Ths bologc effect has been observed by cell knetcs methods n some studes2,22,23 and could enhance leukemc cell kll, Cell knetc studes were not performed n our study. An ncrease of the W B C count cannot be smply attrbuted to stmulaton o f leukemc prolferaton* A notceable ncrease was observed only n fve patents, and dd not correlate wth an unexpectedly hgh CR rate, or, conversely, resstance rate, n ther study, Ohno et al11 dd not observe a greater ncrease of bone marrow blasts n patents who receved G-CSF 2 days before the start of nducton chemotherapy than n those who receved placebo. In fact, the optmum tm ng of admnstraton HGFs for eventual prmng o f the leukemc cells remans largely unknown. The adm nstraton of GM-CSF durng several days before the star o f Table 6. Man Sde Effects, Number of Documented Infectons, and Supportve Care by Treatment Arm Treatment Arm - / - (n = 26) + /- (n = 25) - / + (n = 27) +/-t- (n = 24) Varable No. % No. % No, % No. % Bone pan grade 3/4 0 0 0 0 2 7 2 8 Flud retenton ó 23 8 32 11 41 12 5 0 Weght gan (s= 5 kg} 2 8 0 0 4 15 4 1 7 Hypotenson 1 4 2 8 ó 22 5 Cardac 0 0 0 0 2 7 5 Infecton 19 73 19 76 22 81 92 22 Na. of days wth fever Medan 5.5 7 7 10.5 Range 0-63 0-19 0-26 3-3 0 No. of days of antbotcs Medan 12 18 14 1 8 Range 0-29 0-37 0-32 0-5 Ó 21 21
2158 ZITTOUN ET AL chemotherapy could nduce a hyperleukocytoss, wth eventual pulmonary nfltrates.23 An early start of GM- CSF, 4 to 7 days before nducton chemotherapy, mght explan the relatvely low CR rate n patents reported by Estey et al,24 when compared wth a group of matched hstorcal controls. However, these patents also receved GM-CSF durng the postchemotherapy perod, whch, accordng to the present study, may ncrease the rsk of persstent leukema. An nterestng combnaton of GM- CSF wth tmed-sequental chemotherapy has been proposed by Archmbaud et al.25 In ths combnaton, GM- CSF was admnstered durng a short perod of 5 days between two short courses of chemotherapy to ncrease the recrutment and cell kll of resdual leukemc cells. Our study showed a trend for earler recovery of PMNs n patents who receved GM-CSF. A sgnfcantly shorter duraton of neutropena was observed n most other studes wth GM-CSF13 14 26 or G-CSF.10 15 27 However, n our seres, as n most others, ths slghtly accelerated recovery dd not result n a sgnfcant reducton of the rate of documented nfectons or mortalty durng hypoplasa, or of the number of days wth fever, duraton of admnstraton of antbotcs, and duraton of stay n the hosptal. Only n the Eastern Cooperatve Oncology Group (ECOG) study was there a sgnfcant decrease of grade 4 to 5 nfectons, wth a trend for lower therapy-related mortalty and a hgher CR rate. The reasons for ths dscrepancy wth the other studes of GM-CSF12,14,16 reman to be explored, takng nto account the slghtly hgher daly dose (250 /xg/m2 v 5 (g/kg) and the use of a yeast- derved growth factor. Also n ths study, Rowe et al13 dd not observe an ncreased rate of early relapse n patents who receved GM-CSF durng nducton and con- REFE 1. Löwenberg B, Touw IP: Hematopoetc growth factors and ther receptors n acute leukema. Blood 81:281-292, 1993 2. Cannstra SA, DCarlo J, Groshek P, et al: Smultaneous admnstraton of granulocyte-macrophage colony-stmulatng factor and cytosne arabnosde for the treatment of relapsed acute myelod leukema. Leukema 5:230-238, 1991 3. Agletta M, De Felce L, Stacchn A, et al: In vvo effect of granulocyte-macrophage colony-stmulatng factor on the knetcs of human acute myelod leukema cels. Leukema 5:979-984, 1991 4. Kostnen P, Wang C, Curts JE, et al: Granulocyte-macrophage colony-stmulatng factor and nterleukn 3 protect leukemc blast cells from Ara-C toxcty. Leukema 5:789-795, 1991 5. Lotem J, Sachs L: Hematopoetc cytoknes nhbt apoptoss nduced by transformng growth factor b 1 and cancer chemotherapy compounds n myelod leukemc cells. Blood 80:1750-1757, 1992 6. Bettelhem P, Valent P, Andreef M, et al: Recombnant human granulocyte-macrophage colony-stmulatng factor n combnaton wth standard nducton chemotherapy n de novo acute myelod leukema. Blood 77:700-711, 1991 soldaton; however, the rsk of promotng the regrowth of leukemc cells was reduced n ths protocol, wth GM- CSF beng started only on day 11 and n patents wth no resdua] blood or bone marrow blast cells. Another study wth postve results n favor of a combnaton wth growth factor n AML was reported by Dombret et al,15 who used G-CSF n elderly AML patents from day 9 untl day 28 or earler hematopoetc recovery. These nvestgators observed a sgnfcantly hgher CR rate n patents who receved the growth factor, wthout reducton of the mortalty rate from nfecton. The hgher CR rate was related to less resstance of leukema, especally n patents wth adverse prognostc factors. Ths result led to the hypothess of an antleukemc effect of G-CSF by stmulaton of termnal dfferentaton of resdual leukemc cells. If confrmed, our results mght ndcate a dfferent, adverse effect of GM-CSF. The combnaton of GM-CSF wth antleukemc cytotoxc drugs mght have other dsadvantages over G-CSF wth regard to ntracellular metabolsm of Ara-C: the n vvo formaton of cytarabne trphosphate (Ara-CTP) s decreased by GM-CSF,28 and there s a decreased senstvty to Ara-C of leukemc clonogenc cells after exposure to ths growth factor 4 despte the observed ncreases n labelng ndex and of numbers of nucleosde transporters.22 Further randomzed studes to compare the varous HGFs at dfferent dose schedules are warranted. Our results ndcate that outsde such controlled prospectve studes, the use of HGF should be avoded n AML, at least durng the nducton perod, thus supportng the recent recommendatons of an expert panel.29 ACKNOWLEDGMENT We thank Murelle Dardenne (Data Manager at the EORTC Data Center) for her effcent work. JCES 7. Büchner T, Hddeman W, Koengsmann M, et al: Recombnant human granulocyte-macrophage colony-stmulatng factor after chemotherapy n patents wth acute myelod leukema at hgher age or after relapse. Blood 78:1190-1197, 1991 8. Bemell P, Kmby E, Hast R: Recombnant human granulocytemacrophage colony-stmulatng factor n combnaton wth standard nducton chemotherapy n acute myelod leukema evolvng from myelodysplastc syndromes: A plot study. Leukema 8:1631-1639, 1994 9. Valent P, Sllaber C, Gessler K, et al: Treatment of de novo acute myelogenous leukema wth recombnant granulocyte macrophage-colony-stmulatng factor n combnaton wth standard nducton chemotherapy: Effect of granulocyte macrophage-colony-stmulatng factor on whte blood cell counts. Med Pedatr Oncol suppl 2:18-22, 1992 10. Ohno R, Tomonaga M, Kobayash T, et al: Effect of granulocyte colony-stmulatng factor after ntensve nducton therapy n relapsed or refractory acute leukema. N Engl J Med 323:871-877, 1990 11. Ohno R, Naoe T, Kanamaru A, et al: A double-blnd con
GM-CSF FOR INDUCTION OF ACUTE MYELOID LEUKEMIA 2159 trolled study of granulocyte-colony-stmulatng factor started two days before nducton chemotherapy n refractory acute myelod leukema. Blood 83:2086-2092, 1994 12. Wtz F, Harousseau JL, Cahn JY, et al: GM-CSF durng and after remsson nducton treatment for elderly patents wth acute myelod leukema. Blood 84:231a, 1994 (suppl 1) 13. Rowe JM, Andersen JP, Mazza JJ, et al.: A randomzed placebo-controlled phase III study of granulocyte-macrophage colonystmulatng factor n adult patents (> 55 to 70 years of age) wth acute myelogenous leukema (AML): A study of the Eastern Cooperatve Oncology Group (E1490). Blood 86:457-462, 1995 14. Stone RM, George SL, Berg DT, et al: Granulocyte-macrophage colony-stmulatng factor after ntal chemotherapy for elderly patents wth prmary acute myelogenous leukema. N Engl J Med 332:1671-1677, 1995 15. Dombret H, Chastang C, Fenaux P, et al: A controlled study of recombnant human granulocyte colony-stmulatng factor n elderly patents after treatment for acute myelogenous leukema. N Engl J Med 332:1678-1683, 1995 16. Hel G, Chadd L, Hoelzer D, et al: GM-CSF n a double blnd randomzed placebo controlled tral n therapy of adults patents wth de novo acute myelod leukema (AML). Leukema 9:3-9» 1995 17. Zttoun RA, Mandell F, Wllemze R, et al: Autologous or allogenec bone marrow transplantaton compared wth ntensve chemotherapy n acute myelogenous leukema. N Engl J Med 322:217-223, 1995 18. Bennett JM, Catovsky D, Danel MT, et al: Proposed revsed crtera for the classfcaton of acute myelod leukema: A report of the French-Amercan-Brtsh cooperatve group. Ann Intern Med 103:620-625, 1985 19. Buyse M, Staquet M, Sylvester R: Cancer Clncal Trals: Methods and Practce. Oxford, Unted Kngdom, Oxford Unversty, 1984 20. Gardner MJ, Altman DG (eds): Statstcs Wth Confdence, Belfast, Unted Kngdom, Br Med J Pub, The Unversty Press, 1989 21. KeaLng MJ, Smth TL, Kantarjan H, et al.: Cytogenetc pattern n acute myelogenous leukema: A major reproductble determnant of outcome. Leukema 2:403-412, 1988 22. Wley JS, Cebon JS, Jameson GP, et al: Assessment of prolferatve responses to granulocyte-macrophage colony-stmulatng (GM-CSF) n acute myelod leukema usng a fluorescent lgand for the nucleosde transporter. Leukema 8:181-185, 1994 23. Wley JS, Cebon JS, Jameson GP, et al: Cytokne prmng of acute myelod leukema may produce a pulmonary syndrome when assocated wth a rapd ncrease n perpheral blood myeloblasts. Blood 82:3511-3512, 1993 24. Estey E, Thall PF, Kantarjan H, et al: Treatment of newly dagnosed acute myelogenous leukema wth granulocyte-macrophage colony-stmulatng factor (GM-CSF) before and durng cont nuous-nfuson hgh dose Ara-C + daunorubcn: Comparson to patents treated wthout GM-CSF. Blood 79:2246-2255, 1992 25. Archmbaud E, Fenaux P, Reffers J, et al: Granulocyte-macrophage colony-stmulatng factor n assocaton to tmed-sequental chemotherapy wth mtoxantrone, etoposde and cylarabne for refractory acute myelogenous leukema. Leukema 7:372-377, 1993 26. Buchner T, Hddeman W, Rottman R, et al: Multple course chemotherapy wth or wthout GM-CSF prmng and longterm admnstraton for newly dagnosed AML. Proc Am Soc Cln Oncol 12:985, 1993 (abstr) 27. Estey E, Thall P, Andreef M, et al: Use of granulocyte colonystmulatng factor before, durng and after Fludarabne plus Cytarabne nducton therapy of newly dagnosed acute myelogenous leukema or myelodysplastc syndromes: Comparson wth Fludarabne plus Cytarabne wthout granulocyte stmulatng factor. J Cln Oncol 12:671-678, 1994 28. Gandh V, Du M, Kantarjan HM, et al: Effect of granulocytemacrophage colony-stmulatng factor on the metabolsm of arabnosy Icy tosne trphosphate n blasts durng therapy of patents wth chronc myelogenous leukema. Leukema 8:1463-1468, 1994 29. Amercan Socety of Clncal Oncology recommendatons for the use of hemopoetc colony-stmulatng factors: Evdence-based, clncal practce gudelnes. J Cln Oncol 12:2471-2508, 1994