To evaluate the results of autologous stem cell transplantation (ASCT) in a large population of adults with acute lymphoblastic leukemia (ALL) in first complete remission (CR), we performed an individual data-based overview of the last three trials from the LALA group. Overall, 349 patients with ALL prospectively randomized in the consecutive LALA-85, -87, and -94 trials to receive either ASCT or chemotherapy as post-CR treatment were analyzed. Eligibility criteria were 15–50-year-old patients without sibling donors in both LALA-85/87 trials and 15–55-year-old patients with high-risk ALL and no sibling donors in the LALA-94 trial. Intent-to-treat analysis, which compared 175 patients from the ASCT arm to 174 patients from the chemotherapy arm, showed that ASCT was associated with a lower cumulative incidence of relapse (66 vs 78% at 10 years; P=0.05), without significant gain in disease-free or overall survival. Despite a possible lack of statistical power, a nested case–control analysis performed in 85 patient pairs adjusted for time to transplant and prognostic covariates confirmed these intent-to-treat results in patients actually transplanted. Of interest, the reduced relapse risk after ASCT translated in better disease-free survival in the 300 rapid responders who reached CR after the first induction course.
Even if complete remission (CR) may be achieved in about 80% of adult patients with acute lymphoblastic leukemia (ALL), relapse occurs frequently leading to poor long-term disease-free survival (DFS), usually estimated between 25 and 40%.1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13 Risk-adapted strategies based on initial prognostic grouping are currently applied in these patients. Intensive strategies, including allogeneic and autologous stem cell transplantation (SCT), are generally offered to high-risk patients after achievement of the first CR. Based on donor versus no-donor comparison, allogeneic SCT may result in a better outcome as compared to maintenance chemotherapy or ASCT14, 15, 16 (and Dekker et al., Blood 2001; 98: 859a, abstract). The benefit of ASCT over chemotherapy remains, however, questionable. Some retrospective reports, mostly performed in patients actually receiving ASCT and not based on the intent-to-treat principle, have suggested that ASCT may be associated with prolonged DFS.17, 18, 19, 20, 21, 22, 23, 24, 25 However, reported prospective studies enrolling either standard-risk (Truchan-Graczyk et al. Blood 2002; 100: 861a, abstract), high-risk,26, 27 or all-risk ALL patients1, 2, 28 (and Rowe et al. Blood, 2001; 98, 481a, abstract) failed to demonstrate the superiority of ASCT over chemotherapy, perhaps because of the relatively limited populations of patients randomized.
The LALA group has addressed this question in three consecutive prospective trials. Patients from all risk groups were randomly allocated to ASCT or standard chemotherapy in the first two trials (LALA-85 and LALA-87), while only high-risk patients were randomized in the more recent LALA-94 trial.2, 26, 29 As a trend in favor of ASCT was observed in LALA-87 and LALA-94 trials,2, 26 we performed the present overview on individual data of all patients randomized in the three trials. The aim of this study was to compare ASCT and chemotherapy in a large population of adults with ALL in first CR.
Patients and methods
The three consecutive LALA-85, LALA-87, and LALA-94 trials were conducted in France, Belgium, Switzerland, and Australia by the Groupe d'Etude et de Traitement des Leucémies Aiguës Lymphoblastiques de l'Adulte (GET-LALA) between August 1985 and January 2001. In these trials, a total of 349 eligible adults with newly diagnosed de novo ALL (LALA-85, 63 patients; LALA-87, 191 patients; LALA-94, 96 patients) were randomized after first CR achievement to further receive either chemotherapy (CHEMO group, 174 patients) or autologous SCT (ASCT) (ASCT group, 175 patients). All these patients were included in the present overview.
The morphological classification was completed by cytochemical and immunological analysis.30 Immunological marker panel (including surface immunoglobulin, CD1, CD2, CD3, CD3, CD5, CD7, CD4, CD8, HLA-DR, CD10, CD19, and CD20, with an arbitrary level of 20% labeled cells as threshold for positive expression) was selected to assess the B-cell precursor (BCP) or T-lineage (T) origin of the leukemic clone and its stage of differentiation. ALL expressing myeloid markers was defined by the coexpression of at least two of the three CD13, CD33, and CD34 antigens. Results were centrally reviewed by an immunological working committee. Immunophenotypes were classified as CD10+BCP-ALL, CD10− BCP-ALL, T-ALL, or missing. Immunophenotype was missing in 12% of patients from LALA-85/87 trials and 6% of patients from the more recent LALA-94 trial. Karyotypes were performed using short-term unstimulated cultures and analyzed according to the International System for Human Cytogenetic Nomenclature.31 Results were centrally reviewed by a cytogenetic working committee. Karyotype was missing in a large number of patients from LALA-85/87 trials (73 and 54% of included patients, respectively), while available in most patients from LALA-94 trial (93%). Moreover, in this recent trial, a molecular screening for BCR-ABL, E2A-PBX1, and MLL-AF4 transcripts was performed at diagnosis. Overall, cytogenetics was classified as high-risk (Ph-positive ALL, i.e. t(9;22) and/or BCR-ABL transcripts; ALL with 11q23 abnormality and/or MLL-AF4 transcripts; ALL with t(1;19) and/or E2A-PBX1 transcripts), standard-risk (non-high-risk), or missing.
Patients with ALL-L3 in the French–American–British classification were not eligible. Age eligibility was 15–50 years in LALA-85/87 trials and 15–55 years in the LALA-94 trial. A risk-adapted strategy was applied in the LALA-94 trial, while not in previous LALA-85/87 trials. As a consequence, all patients without HLA-identical sibling donor were considered for the CHEMO/ASCT randomization in LALA-85/87 trials, while only high-risk patients without donor were considered in the LALA-94 trial. These high-risk patients were defined as those who did not achieve hematological CR after the first induction course or those with BCP-ALL presenting at least one of the following factors: (1) white blood cell count (WBC)30 × 109/l; (2) CD10−; (3) high-risk cytogenetics; (4) coexpression of myeloid antigens; and (5) missing immunophenotype and/or cytogenetics. Of note, in LALA-94 but not in previous LALA-85/87 trials, patients with Ph-positive ALL diagnosed before day 35 as well as those with CNS involvement were not randomized for the ASCT/CHEMO comparison, but treated according to separate Phase 2 schedules. Two patients in whom BCR-ABL rearrangement was evidenced after day 35 were, however, randomized and kept in the population analyzed. These nonuniform eligibility criteria among trials explain the imbalance observed between the three trial populations (Table 1).
Treatments schedules are summarized in Figures 1 and 2. The LALA-85 trial was the pilot phase of the LALA-87 trial.2 From August 1985 to February 1987, 164 patients from 33 participating centers entered this trial. From October 1986 to July 1991, 634 patients from 43 participating centers then entered the LALA-87 trial. The general design of the LALA-85/87 protocol has been previously reported.2, 11, 29 Briefly, patients first received a 4-week, 4-drug standard induction course. A salvage course comprising intermediate-dose cytarabine and amsacrine was offered to patients who did not achieve CR after the induction course. According to age (>40 years or not) and presence of a sibling donor, patients in CR after induction or induction followed by salvage were allocated to receive allogeneic SCT or not. Patients not allocated to allogeneic SCT received three courses of consolidation comprising an anthracyclin, low-dose cytarabine, and L-asparaginase. Those aged 50 years or less were randomized after the second consolidation course to receive either maintenance chemotherapy or ASCT (Figure 1). ASCT conditioning regimen was high-dose cyclophosphamide and total body irradiation (TBI) (Figure 2). Autologous marrow cells were depleted using rabbit complement lysis with B or T monoclonal antibodies or mafosfamide. Maintenance chemotherapy consisted of a modified L10 regimen for eight cycles.32 Cycles 1, 3, 5, and 7 consisted of prednisone, vincristine, and anthracyclin, followed by 6-mercaptopurine and methotrexate, and dactinomycin. Cycles 2, 4, 6, and 8 consisted of prednisone, vincristine, cyclophosphamide, and carmustine, followed by 6-mercaptopurine and methotrexate, and dactinomycin. Prophylaxis of CNS relapse was based on six intrathecal injections of methotrexate and cranial irradiation.
From June 1994 to January 2002, 1000 patients from 47 participating centers entered the LALA-94 trial. Only 873 patients with a minimum follow-up of 5 months in October 2001 were considered for the present study. After an induction course combining prednisone, vincristine, cyclophosphamide, and either daunorubicin or idarubicin (according to baseline randomization), patients were stratified according to risk subgroups. Postremission therapy was allocated as follows: (1) allogeneic SCT was offered to all high-risk patients under 50 years of age with a sibling donor; (2) other high-risk patients were randomized between ASCT and chemotherapy. Before SCT or further chemotherapy, all patients received an intensive consolidation course with intermediate-dose cytarabine and mitoxantrone (HAM), followed by two courses of high-dose methotrexate and L-asparaginase (Figure 1). ASCT conditioning regimen was high-dose cyclophosphamide and TBI. Autologous blood, or in few cases marrow, stem cells were not manipulated ex vivo. Oral chemotherapy with 6-mercaptopurine and methotrexate was given as maintenance therapy during 2 years after ASCT (Figure 2). In the CHEMO arm, patients received two courses of VAD regimen followed by alternating cycles with cytarabine and cyclophosphamide or methotrexate and L-asparaginase for 24 months. Between these cycles, oral 6-mercaptopurine and methotrexate was administered. Prophylaxis of CNS relapse was based on six intrathecal injections of methotrexate and cranial irradiation in not transplanted patients. The three trials were approved by appropriate Ethics Committees and all patients gave signed informed consent.
Overall survival was calculated from the time of inclusion until death and patients alive were censored at the time of last contact. DFS was calculated from the date of CR achievement until first relapse or death in CR, and patients alive in CR were censored at the time of last contact. Cumulative incidences of relapse and death in CR were calculated from the date of first CR achievement until the date of relapse or death, respectively, when occurring as first event.33 As relapses occurred earlier in the ASCT group as compared to the CHEMO group resulting in crossing outcome curves (median time from randomization to relapse, 7 versus 11 months in ASCT and CHEMO patients, respectively; P=0.004, by the Mann–Whitney test), graphical methods showed that the proportional hazards assumption was violated by the ASCT/CHEMO covariate.34 To allow the baseline hazard functions to differ in the early as compared to the later follow-up, Cox proportional hazard models stratified on two analysis time subsets were systematically used for all comparisons of cumulative incidence of relapse, DFS, and overall survival.35 The optimal time cutoff for DFS (0–9 months, >9 months) was determined from the database. In addition, due to the differences in trials' design and eligibility mentioned above, trial was systematically introduced as covariate in all ASCT/CHEMO comparisons. For the nested case–control study, patient pairs were determined after adjustment on time to transplant, trial (LALA-85/87 or LALA-94), number of induction cycles to reach the first CR, WBC, 3-class immunophenotyping (CD10+ BCP-ALL, CD10− BCP-ALL, T-ALL), cytogenetics for 33 of these 85 pairs (t(9;22)/BCR-ABL, 11q23 abn./MLL-AF4, other), and then on age and eventually on date of entry. Early toxic deaths were defined as treatment-related deaths occurring within the first 100 days after ASCT or after the onset of the maintenance chemotherapy. Outcome data have been updated for the present study based on data available as of April 2002. The actuarial median follow-up of the 349 patients included was 10 years (11.8 and 4 years for LALA-85/87 and LALA-94 patients, respectively). For all outcome comparisons, patients not experiencing prior events were censored at 10 years. Hazard ratio (HR) was given with 95% confidence intervals (CIs). A P-value less than 0.05 was considered to indicate statistical significance. All calculations were performed using the Stata software, version 7.0 (Stata Corporation, College Station, TX, USA).
Treatment group characteristics
Main characteristics of patients included within each treatment group are indicated in Table 2. There was no great imbalance between the two groups, with the exception of median age, which was significantly higher in patients from the CHEMO group than in those from the ASCT group (P=0.03).
A subgroup of patients with ‘identified high-risk ALL’ was retrospectively isolated from the whole patient population, in order to evaluate the ASCT effect in high-risk patients only. To define this subgroup, all LALA-94 high-risk criteria were used with the exception of criteria 4 and 5 (missing immunophenotype and/or cytogenetics), since numerous LALA-85/87 patients met these criteria. Such a definition ensured that all patients from this subgroup had at least one high-risk factor, even if it was quite certain that all high-risk patients from the former LALA-85/87 trials were not considered. Characteristics of this selected subgroup of 180 patients are shown in Table 3.
Using stratified Cox models adjusted on the three trials, overall and DFS were not significantly different in the two treatment arms. At 10 years, the estimated overall survival was 22% (95% CI, 16–30%) in the CHEMO arm versus 30% (95% CI, 23% to 38%) in the ASCT arm (P=0.48) (Figure 3). Estimated 10-year DFS was 19% (95% CI, 13–26%) in the CHEMO arm versus 28% (95% CI, 21–35%) in the ASCT arm (P=0.12) (Figure 4). There was, however, a reduction in the cumulative incidence of relapse in the ASCT arm, which reached the statistical significance level. At 10 years, the cumulative incidence of relapse was 66% (95% CI, 59–73%) in the ASCT arm versus 78% (95% CI, 71–84%) in the CHEMO arm (P=0.05; HR=0.78 in ASCT patients, 95% CI 0.60–0.99) (Figure 5).
Fifteen patients died without having relapsed (11 from the ASCT arm and four from the CHEMO arm). The estimated 10-year cumulative incidence of death in first CR was 7% (95% CI, 4–12%) in the ASCT arm and 3% (95% CI, 1–7%) in the CHEMO arm (P=0.14). Early toxic deaths occurred in five patients, all from the ASCT arm. Among these five patients, two were not actually transplanted because of early severe infection, two died from post-transplant infectious events, and the remaining one died in the context of post-transplant veno-occlusive disease. Causes of death in the 10 additional patients who died later were as follows: one secondary acute myeloid leukemia (in a patient randomized in the ASCT arm but who actually received chemotherapy); one post-transplant hemolytic uremic syndrome, two post-transplant infectious, three car crash-related deaths; and unknown causes in the three remaining patients.
Results in high-risk ALL patients
In the subset of 180 patients with high-risk ALL, as described in Table 3, only nonsignificant trends in favor of the ASCT arm were observed. At 10 years, the estimated overall survival was 13% in the CHEMO arm versus 20% in the ASCT arm (P=0.78). Estimated 10-year DFS was 12% in the CHEMO arm versus 20% in the ASCT arm (P=0.10). At 10 years, the cumulative incidence of relapse was 84% in the CHEMO arm versus 76% in the ASCT arm (P=0.08). Since most LALA-94 patients belonged to this high-risk ALL subgroup and because of the different designs of LALA-94 and LALA-85/87 protocols, we further compared the outcome of patients with identified high-risk ALL from the LALA-94 (87 patients, 39 in the CHEMO and 48 in the ASCT arm) to those from the LALA-85/87 trials (93 patients, 44 in the CHEMO and 49 in the ASCT arm). The trend for a lower incidence of relapse in the ASCT arm was observed in LALA-94 patients (P=0.08), while not in LALA-85/87 patients (P=0.40), indicating a potential beneficial effect of intensive pretransplant consolidation and post-transplant maintenance.
Results in rapid versus slow responders
Of interest, differences in outcome appeared to be more marked when the 49 slow responders in whom two induction courses were needed to reach the first CR were excluded from the ASCT/CHEMO comparison. In the remaining 300 rapid responders, estimated 10-year cumulative incidence of relapse was 61% in the ASCT arm as compared to 76% in the CHEMO arm (P=0.01). In these patients, estimated 10-year DFS and overall survival were 31 and 34% in the ASCT arm as compared to 22 and 26% in the CHEMO arm, respectively (P=0.04 for DFS and 0.45 for overall survival). Numbers of patients were too small to evaluate ASCT in the subgroup of 131 rapid responders with high-risk ALL.
Nested case–control analysis
Among the 174 patients allocated to the CHEMO arm, all patients except one received only chemotherapy. This single patient received ASCT (protocol violation). Among the 175 patients allocated to the ASCT arm, 118 patients (67%) actually received autologous transplantation in first CR. The major cause for not being transplanted was early relapse (29 patients). Other causes were severe infection (seven patients), failure of stem-cell collection (six patients), patient's refusal (five patients), unrelated allogeneic bone marrow transplantation (one patient), and various other causes in nine patients. The percentage of patients from the ASCT arm who actually received ASCT was not increased in the more recent LALA-94 trial (70% for the LALA-85/87 trials and 62% for the LALA-94 trial). Of note, among the 31 LALA-94 patients transplanted, only 16 actually received the entire planned post-ASCT maintenance treatment. In the 118 patients actually transplanted, the estimated overall survival was 36% at 10 years (95% CI, 27–45%). In these patients, estimated 10-year DFS and cumulative incidence of relapse were 35% (95% CI, 26–44%) and 58% (95% CI, 48–67%), respectively.
To further compare the outcome of transplanted patients to similar patients from the CHEMO group, a nested case–control study was performed. Among the 118 patients actually transplanted, 85 patients have been matched to 85 similar patients from the CHEMO group. Main characteristics of these 85 patient pairs are detailed in Table 4. In the 85 patients from the ASCT group, median time from diagnosis to transplant was 136 days (range, 70–281). In these 170 patients, ASCT was still associated with a lower cumulative incidence of relapse (54 versus 74% at 10 years; P=0.04; HR=0.50, 95% CI 0.20–0.97) without significant increase in DFS and overall survival. Estimated 10-year DFS was 37% (95% CI, 27–48%) in ASCT patients versus 22% (95% CI, 13–32%) in CHEMO patients (P=0.10). Estimated 10-year overall survival was 39% (95% CI, 27–48%) in ASCT patients versus 25% (95% CI, 15–36%) in CHEMO patients (P=0.28).
The three successive prospective LALA-85, -87, and -94 trials asked the question of ASCT randomly compared to chemotherapy alone in adult patients with ALL in first CR. The randomization concerned all-risk patients in both LALA-85/87 trials, while only high-risk patients in LALA-94. That explains marked imbalances in patient characteristics according to trials. There was, however, no difference in patient characteristics according to the two treatment arms, except for the median age at diagnosis with younger patients in the ASCT group. None of these three trials shown any significant superiority of ASCT over chemotherapy, even if there was a trend in favor of ASCT in both LALA-87 and -94 trials.2, 26 This was the rational to perform the current overview in a larger population of 349 patients with a long follow-up (median, 10 years).
This overview also failed to demonstrate any important benefit of ASCT over chemotherapy in the whole population of patients randomized. Even if the cumulative incidence of relapse was reduced in the ASCT arm, DFS and overall survival were not significantly improved. The relatively poor DFS observed in the whole population of this overview may be related to the quite high proportion of patients with high-risk ALL included, most of them from the LALA-94 trial. As a matter of fact, this patient selection might have represented a factor interacting with the ASCT effect, as it appeared for instance that rapid rather than slow responders were those who could actually benefit from ASCT. This is reminiscent of results reported in studies evaluating the value of minimal residual disease (MRD) status in the context of ASCT, either in patients with ALL,36 Ph-positive ALL,37 or acute promyelocytic leukemia.38 In all these studies, patients with a high MRD level before ASCT did worse than those with a low MRD level. Unfortunately, we did not evaluate the value of ASCT according to MRD status in the present study, because of too many missing data.
In accordance with previous reports,14 only 67% of patients included in the ASCT arm actually received ASCT. This raises again the issue of intent-to-treat analyses versus comparisons of treatments actually received. Even if not reaching statistical significance, differences in relapse incidence, DFS, and overall survival observed between both randomization arms in the nested case–control analysis looked more impressive than those observed in the intent-to-treat analysis. These differences might have reached the significance with higher number of pairs and statistical power. Although ASCT was planned early after CR achievement, the main cause for not receiving transplantation was early relapse. Even if intensive pretransplant consolidation planned in the LALA-94 trial did not improve the feasibility of ASCT and with regards to previous remarks on the impact of MRD status, effective in vivo purging seems to be important before ASCT. The trend for a lower incidence of relapse associated with ASCT in high-risk patients from the LALA-94 trial, while not in corresponding patients from LALA-84/87 trials, underlines the value of this approach.
Optimal modalities of ASCT remain questionable in these patients. If conditioning regimen usually includes TBI, as in the allogeneic setting, the roles of ex vivo purging and post-transplant maintenance chemotherapy remain unclear. Post-transplant therapy with 6-mercaptopurine, methotrexate, and vincristine–prednisone reinductions have been reported to decrease relapse risk and improve DFS in these patients.39 Post-ASCT maintenance chemotherapy was scheduled in the LALA-94 protocol, but was not administered or prematurely stopped because of cytopenia or infections in 48% of eligible patients. Owing to the low number of patients, we were, however, not able to evaluate separately the value of the whole ASCT + post-transplant maintenance approach.
In conclusion, these results obtained through prospective trials over more than 15-year period of time still raise the question of how to consolidate the first CR in ALL adults without a donor? New targeted agents such as imatinib for Ph-positive ALL,40 rituximab for mature B-ALL, or alemtuzumab are currently being evaluated in specific patient subgroups. Importantly, patients with Ph-positive ALL are now treated with imatinib-containing regimens in specific studies that appear to highly improve the outcome of these high-risk patients.41, 42, 43 Otherwise, the choice between ASCT or chemotherapy is still open. One may consider that the use of ASCT significantly reduces the total duration of treatment. On the other hand, DFS might also be improved using more intensive chemotherapy schedules, inspired by modern pediatric protocols. This latter option has been adopted by the new French–Belgian–Swiss Group for Research on Adult Acute Lymphoblastic Leukemia (GRAALL).
We are indebted to Wim van Putten for providing Stata packages with facilities for Kaplan–Meier survival and cumulative incidence curves. This study was supported in part by Le Programme Hospitalier de Recherche Clinique (PHRC N°94-95-97.02), Ministère de l'Emploi et de la Solidarité, France; and by L'Association Contre le Cancer (National Grants ARC Nos. 6237, 9623, and 5484).
French participating centers were: Hôpital Pitié-Salpétrière, Paris; Hôpital Saint-Louis, Paris; Hôpital Edouard Herriot, Lyon; Hôpital du Haut Lévêque, Bordeaux; Hôpital Purpan, Toulouse; Hôpital André Mignot, Versailles; Hôpital Henri Mondor, Créteil; Hôpital Bicètre, Le Kremlin Bicètre; Institut Paoli Calmettes, Marseille; Hôpital Cochin, Paris; Hôpital de Hautepierre, Strasbourg; EFS, Besançon; Hôpital de l'Archet, Nice; Hôpital Necker, Paris; Hôpital Jean Bernard, Poitiers; Hôpital Michallon, Grenoble; Institut Gustave Roussy, Villejuif; INSERM U268, Villejuif; Hôpital du Bocage, Dijon; Hôpital Saint-Antoine, Paris; Centre Hospitalier, Caen; Hôpital Pontchaillou, Rennes; Centre Hospitalier de la Côte Basque, Bayonne; Centre Hospitalier, Lyon; HIA Percy, Clamart; HIA du Val-de-Grâce, Paris; Centre Hospitalier, Chambéry; Hôpital Dupuytren, Limoges; Centre Hospitalier, Avignon; Hôpital Louis Pasteur, Colmar; Centre Hospitalier, Mulhouse; Centre Henri Becquerel, Rouen; Centre Hospitalier, Clermont-Ferrand; Centre Hospitalier, Lille; Hôpital Beaujon, Clichy; Centre Hospitalier, Annecy; Centre Hospitalier, Perpignan; Centre Hospitalier Lapeyronie, Montpellier; Centre Hospitalier, Aix en Provence; Hôpital Jean Monod, Le Havre; Centre Hospitalier, Meaux; Hôpital Hôtel Dieu, Paris; Laboratoire CERBA, Val d'Oise; Hôpital Lariboisière, Paris; Hôpital Victor Dupouy, Argenteuil; Centre Hospitalier Dr Schaffner, Lens; Centre Hospitalier, Valenciennes; Centre Hospitalier Saint-Vincent, Lille; Centre Hospitalier, Roubaix; and ETS, Lille. Belgium participating centers were: Cliniques St Luc, Bruxelles; Centre Hospitalier Notre Dame et Reine Fabiola, Charleroi; Cliniques de Mont Godinne, Yvoir; ASBL, Loverval; Hôpital Saint-Joseph, Gilly; Hôpital de Jolimont, Haine St Paul; Hôpital Saint-Joseph, Mons; Hôpital de la Citadelle, Liège; and Laboratoire de Cytogénétique, Leuven. Swiss participating centers were: Centre Hospitalier Universitaire Vaudois, Lausanne; Kantonsspital, St Gallen; Universitätsspital, Zürich; Hôpital Cantonal Universitaire, Genève; Kantonsspital, Basel; Inselspital, Bern; Kantonsspital, Winterthur; Kantonsspital, Luzern, Switzerland. Australian participating centers were: Westmead Hospital, Westmead; Adelaide Hospital, Adelaide; Mater Misericordae Hospital, Newcastle; Alfred Hospital, Melbourne; Royal Adelaide Hospital, Adelaide; Peter Mac Callum Cancer Centre Institute, Melbourne; Monash Medical Centre, Melbourne; Liverpool Hospital, Sydney; St George Hospital, Sydney, Australia.