Total body irradiation–high-dose cytosine arabinoside and melphalan followed by allogeneic bone marrow transplantation from HLA-identical siblings in the treatment of children with acute lymphoblastic leukaemia after relapse while receiving chemotherapy: a Société Française de Greffe de Moelle study


Dr Pierre Bordigoni Unité de Transplantation Médullaire, Hôpital d'Enfants, Centre Hospitalier Universitaire de Nancy, rue du Morvan, 54511 Vandœuvre-les-Nancy, France.


We investigated the use of a new conditioning regimen followed by allogeneic bone marrow transplantation (BMT) for treating children with acute lymphoblastic leukaemia (ALL) after relapse within 6 months of the completion of therapy. One hundred and sixteen children with acute lymphoblastic leukaemia in second or subsequent complete remission (CR) underwent allogeneic bone marrow transplantation from HLA-identical siblings after a preparative regimen comprising total body irradiation (TBI), high-dose cytosine arabinoside and melphalan (TAM regimen). The Kaplan-Meier product-limit estimate (mean ± SE) of disease-free survival (DFS) at 7 years was 59.5 ± 9% (95% confidence interval). The estimated chance of relapse was 22.5 ± 15% with a median follow-up of 88.5 months (range 51–132). 26 patients (22.4%) died with no evidence of recurrent leukaemia, mainly from interstitial pneumonitis, veno-occlusive disease or acute graft-versus-host disease (GVHD). Three factors significantly affected DFS: acute GVHD, site of relapse and, for children in second remission after a marrow relapse, the disease status at the time of transplantation. The DFS were 59.02 ± 12.6%, 37.5 ± 19.8% and 77.4 ± 15% among patients in CR2 after a marrow relapse, in CR3 or in untreated partial marrow relapse, and in CR2 after an isolated CNS relapse, respectively. The lowest DFS was seen in children with acute GVHD grades 3–4. Two significant factors were associated with relapse: the marrow status at the time of transplantation and chronic GVHD. The relapse rate was lower among children in CR2 or with chronic GVHD. We conclude that transplantation after the TAM regimen is an effective therapy for this population with acceptable toxicity, particularly for children in second remission after a very early marrow relapse, or those with early isolated CNS involvement.

Although the prognosis for children with newly diagnosed acute lymphoblastic leukaemia (ALL) has improved ( Reiter et al, 1994 ), the likelihood of long-term disease-free survival falls to below 10–30% after a relapse ( Buchanan et al, 1988 ; Henze et al, 1991 ). This is particularly true for patients with early marrow relapse (up to 6 months after stopping front-line therapy) and/or with T-ALL ( Henze et al, 1991 ), even with the best chemotherapy regimens. There is general agreement that the survival of certain subsets of children with adverse prognostic factors who experience isolated meningeal relapse remains poor ( Baruchel et al, 1995 ; Kumar et al, 1995a , b; Ribieroet al, 1995 ; Steinherz, 1995). The main problem for these patients is the very high rate of marrow relapse, whatever the intensity of reinduction protocols. These patients should be considered as candidates for allogeneic bone marrow transplantation (BMT).

Unfortunately, BMT is limited by the lack of matched sibling donors. Only 15–20% of referred patients are given related donor BMT ( Bordigoni et al, 1996 ; Giona et al, 1994 ). The benefit of allogeneic BMT for childhood ALL in second or third complete remission (CR) after an early marrow relapse is still controversial ( Pinkel, 1995).

The best conditioning regimen for such patients is not known. A combination of cyclophosphamide (Cy) and total body irradiation (TBI) provided a 5-year disease-free survival of <40% in a Seattle study ( Sanders et al, 1985 ), and 64% was reported by the Memorial Sloan Kettering Institute after hyperfractionated TBI followed by Cy for children in CR2. This improved survival was attributed mainly to a lower relapse rate after transplantation ( Brochstein et al, 1987 ). Recently, the addition of high-dose etoposide ( Dopfer et al, 1991 ) or cytosine-arabinoside (Ara-C) ( Cahn et al, 1991 ; Coccia et al, 1988 ; Gordon et al, 1997 ; Kamani et al, 1995 ; Moussalem et al, 1995 ; Weyman et al, 1993 ; Woods et al, 1990 ) to fractionated TBI has been evaluated for its antileukaemic activity with encouraging results, but only limited numbers of patients were treated.

We have therefore carried out a prospective multicentre study to evaluate the role of a new conditioning regimen, TBI followed by high-dose Ara-C and melphalan (TAM regimen), in histocompatible related donor marrow transplantation for children with ALL in CR2 or CR3 after early medullary or isolated extramedullary relapse.


A total of 116 unselected children who presented consecutively with ALL from 21 institutions between December 1985 and December 1992, were prospectively enrolled in the TAM protocol. All the children were <18 years of age at the time of transplantation and had received marrow from HLA-identical sibling donors after a relapse while on chemotherapy.

Patients' characteristics

The original chemotherapy varied between institutions. However, most patients relapsed after CLCG-EORTC 58831–58832 ( Ferster et al, 1994 ) or Fralle 83, 87, 89 ( Schaison et al, 1990 ) protocols. 60 had isolated marrow relapses and 25 had combined marrow/other system relapses (12 CNS, seven testicular, two CNS and testicular, four various) and 31 had isolated extramedullary relapses (30 CNS, one CNS and testicular). Their characteristics are summarized in 1 Tables I and II. The remission number was defined by the number of previous relapses, either medullary or extramedullary. 92 children were grafted in CR2, 10 in CR3 and 14 in a third untreated, partial relapse (<20% blasts in the marrow smear). All patients who developed a first-event isolated CNS relapse underwent BMT in CR2. The duration of CR1 was 2–35 months (median 18.5 months). By definition, very early relapses were those that occurred within 18 months of the first remission. Early relapses occurred up to 6 months after cessation of the initial therapy. The relapses were treated by the protocols that were currently in use at each institution. They were classified as intensive or non-intensive according to the number of chemotherapy agents used: intensive, more than four systemic drugs (n = 78 patients); non-intensive fewer than five systemic drugs (n= 35 patients). The 10 boys with a prior testicular relapse were given 12–24 Gy to the testis after relapse and before transplantation. 10/45 patients with a prior CNS relapse were given 8–24 Gy cranial irradiation 2–4 weeks before beginning the conditioning regimen (isolated CNS relapse, n = 4; combined, n = 6). 10 children (isolated CNS relapse, n = 5; combined, n = 5) were given prophylactic CNS radiation (18–24 Gy) during the initial treatment. All the patients with CNS relapse were given triple intrathecal therapy (TIT) weekly for 6–12 weeks before BMT. All other children were given prophylactic TIT weekly for 4–6 weeks before BMT.

Table 1. Table I. Patient characteristics: disease-related variables.Thumbnail image of
  • a

    * At diagnosis of the disease.† Very early relapse: relapse occurred <18 months after diagnosis.‡ Early relapse: relapse occurred between 19 months after diagnosis and up to 6 months after discontinuation of front-line therapy.§ Intensive: minimum of five systemic drugs; non-intensive: fewer than five systemic drugs.

  • Table 2. Table II. Patient characteristics: transplant-related variables.Thumbnail image of

    As data were occasionally missing for one or more variables (WBC, karyotype, immunophenotype or type of reinduction chemotherapy), the totals in some cases are lower than expected.

    Conditioning regimen ( 2 Table II)

    The TAM regimen was as follows: each patient was given 12 Gy in six doses of 2 Gy, as two fractions per day for three consecutive days (n = 81 patients) or 10 Gy given in a single fraction (n = 35 patients). All children had lung shielding leaving a median lung dose of 9.2 ± 0.15 Gy (fractionated TBI) and 7.9 ± 0.1 Gy (single-dose TBI). 23/67 boys in testicular CR1 (marrow relapse, n = 19; isolated CNS relapse, n = 4) at time of transplantation were given a prophylactic booster dose to the testicles (4 Gy in one or two fractions) during TBI, depending on the treatment policy of each centre. The TBI was followed by a 1 h infusion of high-dose Ara-C (3 g/m2 every 12 h) for four (n = 16) or six doses (n = 8) (children >12 years old ) or for eight doses (n = 92) (children leqslant R: less-than-or-eq, slant 12 years old) and by melphalan (140 mg/m2) administered over 5 min the day before the graft.

    The marrow grafts were obtained following a commonly used method and were infused on day 0.

    GVHD prophylaxis, assessment and treatment ( 2 Table II)

    GVHD prophylaxis varied over the years but included: methotrexate (MTX) (n = 4 patients) or cyclosporine A (CyA) (n = 23 patients) or MTX and CyA (n = 85 patients) ( Storb et al, 1989 ) or marrow T-cell depletion (n = 4 patients).

    Diagnosis and grading of acute and chronic GVHD were carried out according to previously established criteria ( Glucksberg et al, 1974 ; Shulman et al, 1980 ). Acute GVHD was treated with systemic corticosteroids in moderate to high doses. Chronic GVHD was treated as recommended by Sullivan et al (1981 ).

    Supportive care

    All patients had indwelling central venous catheters and were nursed in laminar air-flow isolation units. Non-absorbable antibiotics were used for prophylaxis against infection. Whether intravenous immunoglobulin was prescribed or not depended on the practice of the individual centre.

    Prophylaxis against Pneumocystis carinii pneumonia (PC) consisted of trimethoprim-sulphamethoxazole administered from day −7 to day −2 and from day 30 to at least day 100. All patients were placed on parenteral hyperalimentation. Platelet transfusions were given to maintain the platelet count >20 × 109/l and packed RBC were given to maintain haemoglobin levels >8 g/dl. All patients were also given prednisolone eye drops every 2–4 h during Ara-C infusion and for 24 h after the last dose.

    The toxicity of the preparative regimen was evaluated and scored according to the criteria established by Herzig et al (1983 ).

    Statistical analysis

    The clinical and laboratory data were retrieved from the French Society of Bone Marrow Transplantation (SFGM) database which systematically and prospectively collects data on all our bone marrow transplantation recipients.

    The endpoint of analysis was the number of deaths from non-leukaemic causes and the duration of remission. These were combined to define disease-free survival. Means were compared by Student's t-test or by the Mann-Whitney U-test. Distribution was compared with the chi-squared test. Fischer's exact test and the Wilcoxon rank test were used to compare differences. Kaplan-Meier product-limit estimates with a 95% confidence interval calculated from the standard errors were used to estimate the probabilities of relapse and relapse-free survival. The log-rank procedure was used to assess the statistical significance of differences between subgroups of children with respect to DFS and probability of relapse. Multivariable analysis was performed using a Cox regression model to determine the potential factors important for risk of relapse and DFS. All patients were followed for at least 4 years after BMT.

    Informed consent

    The risks of the treatment protocols were fully explained to parents and older children conforming to the regulations of the respective centre.



    All patients were successfully transplanted, as documented by the recovery of peripheral granulocyte and platelet counts within the same time range in each group.

    Disease-free survival and prognostic factors ( 3 Tables III and IV) ( Figs 1 and 2)

    Table 3. Table III. Disease-free survival, relapse rate and treatment failure in the 116 patients studied.Thumbnail image of
  • a

    * Third complete remission or early relapse.† All patients underwent BMT in CR2.

  • Table 4. Table IV. Variables significantly associated with outcome in the 116 children studied: results of the Cox model.Thumbnail image of
  • a

    * Relative risk.

  • Figure 1.

    Fig 1. Actuarial probability of disease-free survival and relapse.

    Figure 2.

    Fig 3. Actuarial probability of relapse according to the status of the disease and the site of relapse.

    The DFS at 7 years was 59.5 ± 9% (95% confidence interval) with a median follow-up of 88.5 months (range 51–132) for the 66 surviving patients. DFS was 64.8 ± 9.8% for the patients transplanted while in CR2. Multivariate analysis indicated that the two variables associated with remission duration were the presence of acute GVHD and the site of relapse. The risk of treatment failure for patients who developed acute GVHD grades 3–4 was 3.9 times greater than that for no GVHD or grades 1–2 GVHD (61.5 ± 9.6% v 27.2 ± 26.8%; P = 0.001). Likewise, patients given a BMT for an isolated early CNS relapse had a significantly better DFS (77.4 ± 15%) than those receiving BMT for marrow relapse (51.2 ± 10.8%; P = 0.009). When multivariate analysis was restricted to patients with marrow relapse, it indicated that only the status of disease at transplant and the presence of acute GVHD were significantly associated with DFS. The children undergoing BMT in CR2 had a better DFS (59.02 ± 12.6%) than those undergoing BMT in CR3 or in a third untreated, partial relapse (37.5 ± 19.8%) (P = 0.05 ; RR = 3.6). Patients with grades 3–4 acute GVHD fared worse than those with no GVHD or grades 1–2 (P = 0.001). There was a borderline statistical difference between the DFS of children transplanted while in CR2 after marrow relapse and those transplanted after an isolated CNS relapse (P = 0.06; RR = 3.4) ( Table III, Figs 1 and 2). No other risk factor investigated, including the duration of CR1 (±18 months), type of TBI (fractionated or single dose) or the intensity of reinduction therapy, was significantly associated with DFS.

    Leukaemia relapse ( 3 Tables III and IV; Figs 1 and 3)

    Figure 3.

    Fig 2. Actuarial probability of disease-free survival according to the status of the disease and the site of relapse.

    The estimates of the chances of relapse in the 7 years after transplantation were 22.5 ± 15% (95% confidence interval), with a plateau from 51 to 132 months. 23/116 patients developed recurrent leukaemia 2–36 months (median 9 months) post-transplant (9/61 transplanted in CR2 after marrow relapse, 11/24 transplanted while in CR3 or in a third untreated partial relapse, and 4/31 transplanted in CR2 after isolated CNS relapse). The most common site of relapse was bone marrow (n = 19). Two patients developed an isolated CNS relapse at 9 and 12 months post-BMT, one patient suffered an isolated mediastinal relapse, and one presented an isolated testicular relapse at 32 months post-transplant. This latter patient was successfully retransplanted and is still in remission 8 years later. Multivariate analysis revealed two significant factors associated with relapse, the marrow status at transplantation and the development of chronic GVHD ( Table V). The 7-year probability of relapse was 15.2 ± 8.4% for children transplanted while in CR2 and 47 ± 23.4% for those transplanted during a more advanced stage of disease (marrow and isolated CNS relapses) (P = 0.007; RR = 4). Chronic GVHD had an antileukaemic effect but acute GVHD did not. Children without chronic GVHD had a relapse rate 3.7 times greater than those with chronic GVHD (P = 0.05) ( Table V). Patients without GVHD also had a risk of relapse 4 times greater than those with acute and chronic GVHD (P < 0.05) ( Table V). Multivariate analysis restricted to patients with marrow relapse showed the same two prognostic factors significantly associated with relapse. The 7-year relapse rates were 17.4 ± 11.2% in patients transplanted while in CR2, and 47 ± 23.4% for those transplanted while in CR3 or in a third untreated partial relapse (P = 0.002; RR = 9.6) ( 3 Table III).

    Table 5. Table V. Relationship between additional pre-transplant CNS radiotherapy and outcome for patients with CNS relapse.Thumbnail image of
  • a

    M: marrow; E: eyes; CNS: central nervous system; Med: mediastinum; T: testicular.

  • Other possible prognostic factors, including the first complete remission duration and the site of relapse, had no significant effect on the outcome.

    Control of prior extramedullary disease ( 5 Table V)

    (a) CNS relapses

    45 patients were given BMT for isolated (n = 31) or combined (n = 14) CNS disease. Only one had a subsequent CNS relapse. An additional CNS irradiation before the conditioning regimen had no influence on either DFS or on the probability of a further CNS relapse.

    (b) Testicular relapses

    None of the 10 boys who had any testicular relapse before BMT developed any subsequent testicular involvement. All these boys were given an additional pre-BMT testicular irradiation. Of the 19 children in marrow relapse but in testicular CR1 at BMT who had received a boost of testicular irradiation as part of TAM regimen, six have remained in CR for 77–100 months (median 86 months) post-transplantation, seven died from transplant-related complications and six had a subsequent relapse (four marrow, two CNS). Of the 28 boys in marrow relapse but in testicular CR1 at BMT, who were not given additional testicular irradiation during TBI, 17 remained in remission 51–125 months (median 84 months) post-BMT, eight died from non-leukaemic causes and three had a subsequent marrow relapse. The irradiated group appeared to fare worse than the unirradiated group. The relapse rate (51.5 ± 27.6% compared to 16.9 ± 15.6%; P = 0.04) was disturbingly high and the DFS (31.6 ± 21.4% compared to 63 ± 18.6 %; P = 0.03) low among patients given prophylactic testicular boost during TBI.

    Transplantation-associated toxicity ( 3 Tables III and 4 IV)

    Twenty-six /116 patients (22.4%) died between days 28 and 330 with no evidence of recurrent leukaemia. Nine died from interstitial pneumonia (IP), four from acute GVHD (with concomitant cytomegalovirus (CMV) pneumonia in one), five from hepatic veno-occlusive disease (VOD) (with concomitant IP in one), three from bacterial or fungal infection, and two from cardiac failure (one with concomitant renal failure 11 months after BMT, and the other with idiopathic IP). A 4-year-old boy developed a fatal acute leuco-encephalopathy 5 d following transplant. He had been given an additional 24 Gy cranial irradiation for CNS disease 2 months prior to the preparative regimen. Finally, one girl developed a glioblastoma multiforme 8 years after transplantation and died 9 months later with her leukaemia in continuous remission.

    Three patients experienced severe cardiac failure during the first few months after BMT. It was reversible in one, but fatal 3 and 11 months later in the other two children. Two patients developed reversible capillary-leak syndrome and one reversible early (day 5) toxic epidermal necrolysis.

    Sixteen patients developed interstitial pneumonitis (13.8%) with a fatality rate of 81.2 %. Nine contracted idiopathic IP (7.7%) and seven died. Seven were diagnosed with various infectious organisms, including CMV (n = 4, with PC n = 2), PC (n = 1), fungal (n = 1), adenovirus (n = 1) and four died. Other severe toxicities (grade ≥ 3) ( Herzig et al, 1983 ) of the preparative regimen included diarrhoea (in 4.4% of patients), rash (in 16% of patients), oropharyngeal mucositis (in 23% of patients), conjunctivitis (in 35% of patients) and hepatic failure (in 9% of patients).

    Seventy-five evaluable patients developed acute GVHD (66.3%) at a median time of day 22 (range 7–58 d). Grades 3 or 4 disease occurred in 9.7% of all patients. Chronic GVHD occurred in 31/ 96 (32.2%) evaluable patients.

    The incidence of non-leukaemic mortality before day 100 appeared to be associated with two risk factors, the year of transplant (<1989, 36.6%; ≥1989, 17.3%; P = 0.02, RR = 5.3) and the presence of grades 3–4 acute GVHD (72.7% with and 18.2% without; P = 0.001, RR = 16.4) ( Table IV). More deaths occurred among patients who suffered marrow relapse before BMT (marrow relapse: 25.8%; CNS relapse: 12.9%; P = 0.1, RR = 3.5) and among those who developed chronic GVHD (22.6% compared to 7.7%; P = 0.09, RR = 4.2) ( 3 Table III).

    The quality of life of the 66 long-term survivors was, in general, excellent, with Karnofsky performance scores of 90–100% in most children.


    We report the outcome of 116 children who were given allogeneic BMT following a new preparative regimen (TAM protocol) ( Cahn et al, 1991 ; Moussalem et al, 1995 ) after they had relapsed during chemotherapy. Our results are better than those reported after Cy + TBI ( Barrett et al, 1994 ; Sanders et al, 1985 , 1987; Uderzo et al, 1995a , b), especially for those patients with unfavourable prognostic factors, such as a first remission of 18 months or less. The probability of relapse was only 17.4 ± 11.2% 7 years after BMT for children transplanted while in CR2 after a marrow relapse. Our data are similar to those other groups for comparable, but smaller, cohorts of children with ALL who underwent BMT in CR2 after high-dose ARA-C ( Coccia et al, 1988 ; Gordon et al, 1997 ; Kamani et al, 1995 ) or etoposide ( Dopfer et al, 1991 ) and TBI. There is no general agreement as to the best BMT preparative regimen. The combination of Cy and TBI has now been used to treat many children with ALL and has a probability of 5-year DFS of 20–30% in patients transplanted in CR2, whose first remission lasted 18 months or less, and 30–40% for those transplanted within 18–36 months after diagnosis ( Barrett et al, 1994 ; Sanders et al, 1985 , 1987; Uderzo et al, 1995a , b). The DFS is <20% in patients with more advanced disease. This regimen has been modified to give hyperfractionated TBI (13.2 Gy) with a booster dose of irradiation to the testicles followed by Cy. Encouraging results have been reported, with 5-year DFS of 64 ± 9%, 42 ± 14% and 23 ± 11% and probabilities of relapse of 13 ± 7%, 25 ± 13% and 64 ± 16% for patients in second, third and fourth remission or relapse respectively ( Brochstein et al, 1987 ). However, recent reports from the Minneapolis team using this regimen showed no substantial advantage over either single fraction or fractionated TBI and Cy ( Weisdorf et al, 1994 ).

    High doses of Ara-C have been used as an alternative to Cy because Ara-C is an active agent in ALL, particularly for the reinduction of patients with marrow and/or CNS relapse. A paediatric pilot study ( Coccia et al, 1988 ; Gordon et al, 1997 ) reported a low relapse rate (14%) with a very long follow-up period for a group of children in second or third remission after a preparative regimen consisting of high-dose Ara-C (36 g/m2) before fractionated TBI. However, two other non-randomized studies suggest that it has no advantage over Cy-TBI ( Weyman et al, 1993 ; Woods et al, 1990 ). A recent multicentre survey ( Weyman et al, 1993 ) suggested that ARA-C/TBI is a good preparative regimen, but only for children <11 years old. Older patients suffered both more toxic deaths and more relapses that younger patients. Our data do not confirm these findings, despite a potentially more intensive preparative regimen.

    The therapeutic action of high-dose etoposide in addition to or instead of Cy in TBI-containing regimens needs to be more extensively defined in children. The only data published describe small groups of patients with a short follow-up or children in CR1 and/or mixed with adults ( Blume et al, 1993 ; Dopfer et al, 1991 ).

    Many authors have looked for prognostic factors that will predict the DFS and relapse after BMT. An increased risk of treatment failure is associated with age >10–11 years, an elevated WBC count at diagnosis, the T-cell phenotype, previous extramedullary leukaemia, a first remission lasting <18 months, and advanced stage of disease at the time of transplantation ( Barret et al, 1994 ; Butturini et al, 1987 ; Sanders et al, 1987 ; Uderzo et al, 1995a , b). Only this advanced disease state significantly affected the outcome of the leukaemia in our patients. Neither a CR1 of <18 months, a patient age >10 years, nor a history of extramedullary leukaemia were associated with a poor prognosis. GVHD can have both favourable or unfavourable effects on transplant outcome. It may reduce the risk of relapse, but increase transplant-related deaths. As in other reports ( Brochstein et al, 1987 ; Sanders et al, 1985 ; Uderzo et al, 1995a ), children who developed chronic GVHD had a significantly improved survival rate and a reduced incidence of relapse. Acute GVHD had no antileukaemic effect and patients with grades 3–4 acute GVHD had an extremely high transplant-related mortality rate.

    Very few deaths could be directly attributed to the use of Ara-C. Only three patients developed severe, but reversible, toxicity probably attributable to the combination of Ara-C and TBI (two capillary-leak syndromes and one toxic epidermal necrolysis). Ocular and skin toxicity was significantly more frequent and severe after the TAM regimen, but the incidence of other transplantation-related complications, such as mucositis, severe diarrhoea, leuco-encephalopathy and liver disease, appeared to be similar to those after Cy and TBI.

    Pulmonary toxicity from high-dose Ara-C has been reported for adults. Andersson et al (1990 ) found that respiratory failure occurred in as many as one-quarter of patients with acute myeloblastic leukaemia treated with high-dose Ara-C. Similarly, this complication has recently been described in paediatric patients ( Shearer et al, 1994 ). Most studies have established an association between the development of pulmonary insufficiency and multiple courses of Ara-C, particularly if it is given by continuous infusion. This could partly explain the somewhat higher incidence of fatal idiopathic IP in children on the TAM regimen in our series.

    There is considerable debate as to whether children with ALL in second complete remission should undergo BMT, and if so, which type. Data from a matched-pair analysis from the International Bone Marrow Transplant Registry ( Barrett et al, 1994 ) and from trials with chemotherapy, or with autologous BMT (ABMT) ( Boulad et al, 1994 ; Kersey et al, 1987 ; Uderzo et al, 1995b ), suggest that allogeneic BMT has an overall advantage over chemotherapy and ABMT for treating ALL in CR2 after ‘on therapy’ marrow relapse, whatever the duration of CR1. Recently, Bordigoni et al (1996 ) and Cavazzana-Calvo et al (1996 ) used matched cohorts to compare the results of three different procedures, HLA geno-identical BMT, autologous BMT, and partially incompatible BMT, for children with ALL in CR2 after short CR1 duration (<28 months). All the patients were given the same conditioning regimen, the TAM protocol. The DFS at 5 years were 83.7 ± 7.4%, 24.6 ± 10.2% and 30.7 ± 10.8%. The difference is statistically significant (P = 0.005). The probability of relapse was 5.5 ± 5.4%, 72.4 ± 11.2% and 32.2 ± 12.1%, respectively. The difference between the geno-identical group and the other two groups is statistically significant (P = 0.0001).

    Control of prior extramedullary disease was good. Of the 45 patients with prior CNS disease, only one had a subsequent CNS relapse. In terms of the probability of post-BMT CNS and/or medullary relapse, the differences are not significant from the group of patients without a history of CNS relapse, whether or not they were given additional cranial radiotherapy prior to the conditioning regimen. These results suggest that an additional cranial irradiation can be omitted for patients with CNS involvement at relapse, when TBI is included in the conditioning regimen. There is also no need for post-transplant CNS therapy ( Ganem et al, 1989 ).

    The isolated CNS relapse rate for the majority of current risk-group-directed protocols is <10%. Salvage treatment has included combined intrathecal or intraventricular chemotherapy and cranial or craniospinal irradiation, but this was inherent with acute and chronic neurotoxicity ( Kumar et al, 1995b ; Ribiero et al, 1995 ). There is general agreement that the survival of children who experience meningeal leukaemia is poor. The EFS is <30–40%, particularly for certain patients with adverse prognostic factors including high WBC and immunophenotype T at diagnosis, short duration of CR1 (<18–24 months) and prior CNS prophylaxis including radiotherapy. Recently published evidence of PCR-based marrow-residual-disease in most children with isolated CNS relapse ( O'Reilly et al, 1995 ) identified a group of patients at high risk of marrow relapse. There is no definitive information from the few published studies as to whether the use of BMT (autologous or allogeneic) improves the cure rate for these children ( Baruchel et al, 1995 ; Colleselli et al, 1994 ; Hoogerbrugge et al, 1995 ; Rossetti et al, 1993 ). The small number of patients has precluded controlled studies, and the optimal therapy remains unknown. Nevertheless our data on such high-risk patients indicate that allogeneic BMT is an adequate first-choice alternative. Haematological and/or testicular relapses are, rather than a new CNS event, the main obstacle to durable survival after conventional regimens. Therefore a treatment strategy must include intensive systemic therapy. The paediatric literature and our data both indicate a very low further CNS relapse (<3%) after allogeneic BMT, whether or not the patients suffered from a prior CNS disease. Hence, we recommend allogeneic transplantation in the first haematological remission for patients who develop a very early isolated CNS relapse, or who suffer from more than one isolated extramedullary relapse while on chemotherapy.

    None of the boys with testicular relapse before BMT had a subsequent testicular relapse. This is probably due to our policy of testicular irradiation before TBI. The amount of irradiation given during the conditioning regimen seems to be inadequate for males suffering from testicular relapse. Testicular irradiation before TBI is advised for all such patients ( Sanders et al, 1985 ).

    Poelman & Sanders (1992) and Brochstein et al (1987 ) suggest that testicular irradiation, in addition to TBI, could benefit children in testicular remission at BMT. Our findings are appreciably different; we found that a booster dose of irradiation to the testicles had a detrimental impact on the incidence of testicular relapse and on DFS. This could be because of the small number of children included, or because of the historical comparison used to analyse the two subgroups of patients. We are therefore not using such potentially toxic therapy.

    In summary, we have confirmed that allogeneic bone marrow transplantation from HLA-identical siblings is an effective treatment for children with ALL in second or third complete remission, particularly for those whose previous remission lasted <18 months, whatever the site of relapse. But we have not completely answered the question of the best treatment for those children, and we still need a well-designed prospective randomized trial of chemotherapy versus allogeneic BMT. The use of fractionated TBI followed by high-dose Ara-C and melphalan is an effective therapy with a very low incidence of relapse and a remarkably good leukaemia-free survival. The only two prognostic factors that significantly affected outcome were the disease status at the time of transplantation and acute GVHD grades 3–4. We need new approaches for leukaemic control in patients in the later stages of disease and more effective strategies to reduce the frequency of high-grade acute GVHD.

    The control of extramedullary disease in these patients was good, so that prophylactic testicular or CNS irradiation given in addition to TBI can be omitted. The toxicity of the TAM regimen is considered acceptable; the early toxic deaths were mostly due to common complications of allogeneic BMT.