Haploidentical stem cell transplantation for children with acute leukaemia


Dr David I. Marks, BMT Unit, Bristol Royal Hospital For Sick Children, United Bristol Healthcare Trust, Bristol BS2 8BJ, UK.
E-mail: david.marks@ubht.swest.nhs.uk


Some children with relapsed or high-risk acute leukaemia have an improved outcome if they have an allogeneic stem cell transplant, preferably from a sibling or well-matched unrelated donor. However, some children do not have these options or there is an urgent need to proceed to transplant because of disease status. We have investigated the role of haploidentical family members as donors in 34 patients with acute leukaemia (median age 11 years, range 1–16 years). Patients were conditioned with cyclophosphamide and total body irradiation (14·4 Gy in eight fractions) and received T-cell depleted peripheral blood stem cell grafts with a median CD34 cell dose of 13·8 × 106/kg (range 4·2–35·1) and 0·7 × 104 CD3-positive cells/kg. The actuarial survival at 2 years was 26% (13–41%, 95% CI). Eight patients have survived disease-free with a median follow up of 62 months. They have good performance status and a median lymphocyte count of 1·8 × 109/l. Relapse (14 patients) and adenoviral (six patients) or fungal infections (four patients) were the major causes of death. Haploidentical stem cell transplantation can produce medium term disease-free survival in a proportion of children with high-risk or relapsed acute leukaemia. None of the nine patients with acute myeloid leukaemia not in remission have survived.

Only a quarter to a third of children with acute leukaemia who are candidates for allogeneic stem cell transplantation (SCT) have a human leucocyte antigen (HLA)-identical sibling donor. Of the remaining patients, about 60–85% have a suitable unrelated donor, depending on the degree of mismatch that is accepted and the ethnic and racial background of the child. Search times for unrelated donors vary from 2 to 6 months and, sometimes, alternative stem cell sources are considered because of clinical urgency. Alternative sources of stem cells in this group are from unrelated cord blood donors and haplotype mismatched family members. Moreover, with the promising results obtained by the Perugia group in predominantly adult patients with acute myeloid leukaemia (AML) not in remission (Aversa et al, 1998, 2005), many centres have investigated haploidentical SCT in paediatric practice where there is unstable or refractory disease (Veys et al, 2003).

In Germany, of 63 paediatric patients (the majority of whom had leukaemia) transplanted with one to three antigen-mismatched grafts, sustained engraftment was seen in 98% (15% after ‘boosts’) and acute and chronic graft-versus-host disease (GVHD) occurred in 7% and 13% respectively (Lang et al, 2004). This group used CD34 or CD133 selected peripheral blood stem cells (at a median dose of nearly 20 × 106/kg) but with a similar T-cell dose to our series. The survival in patients with AML and chronic myeloid leukaemia (CML) was 18% and was 48% in patients with acute lymphoblastic leukaemia (ALL) or non-Hodgkin lymphoma (NHL). Overall, 18% experienced lethal viral infections but newer strategies reduced this incidence to 8%. Ortin et al (2002) reported excellent survival in 16 children with malignant conditions who received partially matched related donor allografts. With a follow up of 1·5 years, 13 had survived free of disease and the incidence of grade III–IV GVHD was low. This latter group used conventional full intensity chemoradiotherapy, post-transplant ciclosporin and higher T-cell doses in the graft.

The results of treatment for children with ALL and AML are improving but relapse is still a common event. The majority of these children achieve a second remission and, depending on the duration of the first remission, may be candidates for allogeneic SCT. Some of these second remissions are of short duration or are only obtained with multiple chemotherapy regimens. Other potential candidates for this approach are children who have such high-risk features at diagnosis that they are unlikely to be cured by chemotherapy alone, children whose leukaemia never achieves remission and those with refractory relapsed leukaemia.

We report here the clinical outcome and toxicities of 34 children with acute leukaemia or related diseases treated on Medical Research Council protocols who underwent haploidentical SCT. Our protocol differed from those above, in that Campath 1H was mainly used in pretransplant conditioning and the median dose of T cells was <1 × 104/kg. This study also reports the results in nine children with AML not in remission.

Patients and methods

Patients and inclusion criteria

The pretransplant clinical characteristics of the 34 children are shown in Table I. Most patients were at high risk. We considered patients at high risk of relapse or risk of transplant-related mortality to have a high risk of treatment failure. Of the 17 AML patients, 13 were high risk for relapse. Nine patients were not in remission and were resistant to multiple courses of chemotherapy including five who received gemtuzumab ozogamicin. Two patients were in third complete remission (CR3) and two of the patients with AML in CR2 had CR1 durations <12 months.

Table I.   Patient characteristics.
  1. CP2, second chronic phase; CMV, cytomegalovirus.

  2. *Four of these patients were BFM S4.

  3. †Two of these patients had their marrow treated with Campath 1M.

Median age, years (range)11 (1–16)
 Acute myeloid leukaemia17
 Acute lymphoblastic leukaemia14*
 Myelodysplastic syndrome1
 Chronic myeloid leukaemia-CP21
 Biphenotypic acute leukaemia1
Risk status
Campath 1H ‘in the bag’7†
Clinimacs CD34 selection27
CMV seronegative donor + recipient8
CMV seropositive donor + recipient8
CMV seropositive recipient, negative donor7
CMV seronegative donor, positive recipient11

Of the 14 ALL patients, five had Philadelphia-positive ALL (two were in CR2), one was in CR3, one in CR4 and four were in CR2 (Berlin-Franfürt-Münster risk group S4, Borgmann et al, 2003). Of the remaining three patients, one relapsed at week 17 of therapy, one had minimal residual disease after consolidation and another patient with T-ALL relapsed 1 year after diagnosis.

Search strategy

None of the patients had matched sibling donors available. All patients had an unrelated donor search and, in the last 3 years, a cord blood search. Patients in remission preferentially had an A-, B-, C-, DR- and DQ-matched unrelated donor transplant or a class I single antigen mismatched donor transplant if that was available without undue delay.

Patients with AML not in remission were clinically unstable after inadequate responses to attempted induction chemotherapy and proceeded to transplant as soon as haploidentical stem cells were available. Killer cell immunoglobulin-like receptor (KIR) mismatched donors were selected wherever possible (see below) after 2003.

Selection of KIR mismatches

Killer cell immunoglobulin-like receptor mismatches were selected for the presence or absence of KIR ligands of HLA origin. The KIR ligand specificities used for matching criteria are as follows: HLA-C group 1 (C*01, C*03, C*07, C*08, C*13, C*14 for KIR genes 2DL2/3 and 2DS2), HLA-C group 2 (C*02, C*04, C*05, C*06, C*17, C*18 for KIR genes 2DL1 and 2DS1), HLA-B (Bw4 for KIR genes 3DL1/S1) and HLA-A (A*03A*11 for KIR gene 3DL2).

We determined only that a KIR ligand mismatch was present in the transplant pair and identified those donor/recipient(s) where the ligand mismatch was in the graft-versus-host (GVH) or/and host-versus-graft (HVG) directions (Ruggeri et al, 2002). KIR receptor genotyping was not performed for the purposes of this study.

Pretransplant conditioning therapy, GVHD prophylaxis

Graft processing evolved over the period of study with advances in methods of T-cell depletion. All children received cyclophosphamide 60 mg/kg intravenously (i.v.) on days −6 and −5 followed by total body irradiation 14·4 Gy in eight fractions on days −3 to 0 inclusive. All but six patients received Campath 1H 0·2 mg/kg i.v. for 5 d. Six patients with AML received antithymocyte globulin (ATG) 2·5–5 mg/kg/d from days −2 to −2 (5 d). Seven patients received ciclosporin as GVHD prophylaxis. The patients with Miltenyi CD34-selected grafts did not receive ciclosporin.

Supportive care

All patients were transplanted in positive pressure single rooms with air filtration by EU8-grade filters, and remained in protective isolation from day 0 until engraftment. Patients were free to leave the unit during conditioning. All patients received a ‘clean’ diet, which excluded fresh fruit and vegetables and tap water from day −3 until 60 d after discharge from isolation. Patients who were re-admitted to the unit for management of complications were housed in single rooms without air filtration. Intravenous access was achieved by double-lumen tunnelled central venous catheter. In the absence of line infections and provided lumens remained patent, lines were left in situ for 3–6 months, depending on venous access. Patients received oral ciprofloxacin (250 mg b.d.) from day −3 until intravenous antibiotics were commenced. Co-trimoxazole (960 mg twice or thrice weekly) was given as prophylaxis against Pneumocystis carinii pneumonia (PCP) from day 28; this was replaced by nebulised pentamidine (300 mg every 3–4 weeks) in eight patients with poor graft function or allergy. PCP prophylaxis was used routinely for 6–12 months, but continued longer in patients with chronic GVHD. Phenoxy-methyl penicillin (250 mg b.d.) was commenced at 3–6 months and continued for life. Standard antifungal prophylaxis consisted of itraconazole solution dosed to achieve levels >500 ng/l. Patients intolerant to itraconazole were given ambisome 1 mg/kg three to seven times a week. Antifungal prophylaxis was given from the day of admission (day −9) for 3 months, but continued longer if there was poor graft function or chronic GVHD. Patients at low risk of cytomegalovirus (CMV) disease (both patient and donor seronegative) received low-dose oral aciclovir (400 mg t.d.s.) from day −4 and CMV seronegative blood products. Patients at high risk of CMV disease (patient, donor or both seropositive) received high-dose aciclovir (500 mg/m2 t.d.s. i.v.) from day −4 to day +30, thereafter oral aciclovir (400 mg t.d.s.), and normal pooled human immunoglobulin (200 mg/kg) every 3 weeks, commencing on day −1. CMV prophylaxis was given routinely for 6 months, but continued if there was severe chronic GVHD. If there was evidence of CMV re-activation in blood, ganciclovir was administered for at least 2 weeks at 5 mg/kg b.d. CMV disease was treated with ganciclovir plus intravenous immunoglobulin. Patients were vaccinated against diphtheria, tetanus and polio (killed Salk vaccine) at 18 months post-transplant using primary immunisation schedules, and against measles at 2 years.

Febrile neutropenia (defined as pyrexia in a patient with an absolute neutrophil count <1 × 109/l), was treated with broad spectrum intravenous antibiotics, with the addition of intravenous liposomal amphotericin at 72 h for patients with refractory fever pending a computed tomography scan. Patients with proven or probable aspergillosis were usually treated with ambisome 3 mg/kg and with caspofungin and voriconazole as second-line therapy.


The first patient was transplanted in 1997 and accrual continued to the end of 2004 with all consecutive patients included. We elected to have a long follow-up period because some of these patients had life threatening complications far out from transplant and relapse was a common event until 1 year after transplant.

Kaplan–Meier estimates of survival were calculated from the date of BMT and subgroups were compared using log-rank tests (univariate analysis). Subgroups for age and CD34 were defined by using the median value as the cut-off point. Numbers were deemed too small for effective multivariate analysis.


Patients, demographics and diagnoses

The characteristics of the 34 patients are described in Table I. Twenty-one patients did not have a suitable unrelated donor-match and 13 patients received haploidentical stem cells because of clinical urgency.

Tissue typing, KIR mismatching

Mismatches were recorded in both the GVH and HVG directions. Of 10 possible loci for GVH mismatches, nine were mismatched at 3 loci, five at 4 loci and 20 at 5 loci. In the HVG direction, three were mismatched at 1 locus, two at 2 loci, seven were mismatched at 3 loci, six at 4 loci and 15 at 5 loci. Of the 34 patients, 18 were KIR mismatched in the GVH direction and 20 in the HVG direction. From 2003 onwards it was policy to select KIR mismatches wherever possible.

Graft composition, T-cell depletion

All patients received peripheral blood stem cell grafts. Eleven donors required more than one aphaeresis session. The median CD34 cell dose received was 13·8 × 106/kg (range 4·2–35·1). Twenty-three of 34 patients received more than 10 × 106 CD34 cells/kg.

Twenty-seven patients had ex vivo T-cell depletion, 26 with a Clinimacs device and one with CD133-positive selection. These patients received 0·3 to 5·2 × 104 CD3-positive cells per kg recipient weight in their graft (median 0·7).

Seven patients had ex vivo T-cell depletion with monoclonal antibodies, five with alemtuzamab and two with Campath 1M. These seven patients received 3·6–41·7 × 107 CD3-positive cells/kg in their graft.


Thirty-one of 34 patients (91%) engrafted at a median of 13 d post-SCT infusion (range 10–30 d). Twenty-four patients achieved a platelet count of 50 × 109/l at a median of 15 d (range 10–98 d). Eight patients took 30 d or longer to achieve this level of platelet engraftment.

Of the three patients with primary graft failure, none survived. One patient was re-grafted from the same donor with fludarabine and OKT3 and had evidence of full donor myeloid chimerism but died 5 months later of adenoviral pneumonia. The other two patients died of pulmonary disease (haemorrhage and infection) and relapse respectively.

Recovery of lymphocyte counts

Twenty-one patients survived to day 100 without disease relapse. Their absolute lymphocyte counts ranged from 0·08 to 3·6 × 109/l (median 0·6 × 109/l). The eight survivors who are all >1 year from transplant have lymphocyte counts ranging from 1 to 3·5 × 109/l (median 1·8 × 109/l).

Graft-versus-host disease

Nine of 31 evaluable patients (29%) developed acute GVHD, all of the skin. In four patients (13%) it was grade II or more. Three of 25 evaluable patients (12%) developed cGVHD (12%), all of the skin.


Ten of 34 patients died of infection, with viral infection being a particular problem. Adenovirus re-activation was the cause of death in six patients but two patients who had leukaemia present at the time of death also had adenovirus evident. One patient who died of adenovirus also had CMV re-activation and probable encephalitis. Seven patients had adenovirus isolated from stool. Four patients received cidofovir for adenovirus infection but none of these patients survived. Twenty-six patients were at risk for CMV reactivation and five had evidence of re-activation (19%). Three patients had parainfluenza III and there were single cases of echovirus, picornavirus and rhinovirus. None of these infections required intervention. Eleven patients had viruses detected in the urine (nine polyoma and two adenovirus). There were 19 proven bacteraemias. Eight had coagulase-negative staphylococci, nine various streptococcal species and one stenotrophomonas.


Thirteen patients relapsed, none of whom survived. Five patients had palliative oral chemotherapy but no patient was given donor lymphocyte infusions (DLI). Of the 24 patients transplanted in remission, three patients relapsed (13%).


Eight of 34 patients survive to date. The median follow up of survivors was 62 months (range 12–86 months). Five of the eight patients are >4 years out from transplant. The actuarial survival at 2 years was 26% (13–41%, 95% CI) (Fig 1). Survival at 4 years was 23% (10–38%). Being not in remission at transplant was significantly associated with a poorer outcome (Log rank P = 0·033). Survival was not significantly associated with recipient age, CD34 dose, KIR status in either direction, CMV status, donor or recipient sex or with the development of acute GVHD (data not shown).

Figure 1.

 Kaplan–Meier survival estimate versus time of analysis (years). Tick marks represent surviving patients.

None of the nine patients with AML not in remission survive. Six relapsed and three died of infection. Their median survival was 160 d (range 26–699 d). Twenty-two patients with AML or ALL were in remission at the time of transplant and eight survive (36%). Seven of these patients had AML in CR and two survive (28%) compared with six of 16 patients (38%) with ALL in CR who survive.

Causes of death

Fourteen of 26 patients (54%) who died had relapse as the primary cause. Adenovirus infection was the main cause in six patients and probable fungal infection in four patients. Two other patients died of pancreatitis and haemorrhage (central nervous system and pulmonary) respectively.


The place of haploidentical donors for allogeneic SCT in the management of paediatric and adult patients with leukaemia is gaining credence. Following the results published by the Perugia group (Aversa et al, 1998) many units with a track record in alternative donor transplantation began to explore the use of haploidentical SCT in patients with no other allogeneic donor option or where there was a need to proceed to transplant urgently. Results in children from Germany and St Jude's are impressive (Lang et al, 2004). More recently there is renewed interest in single or double cord blood transplants, which offer a very different graft composition and toxicity profile (Laughlin et al, 2004; Barker et al, 2005). Comparisons of umbilical cord blood transplant and haemopoietic SCT are underway but registry-based studies will always have limitations and both types of transplants are constantly evolving.

Of the 34 patients with leukaemia who received conventional full intensity conditioning, megadoses of CD34-selected cells and a median of 0·7 × 104 CD3-positive cells/kg in the graft, about a quarter could become medium term disease-free survivors. The disease state at the time of SCT was the major determinant of outcome. Patients in remission can achieve an outcome similar to that with other donor stem cell sources (Green et al, 1999). Twenty-five of these children had diseases associated with a poor survival after other forms of SCT and only three have survived. Half died of relapse and half of infection. Their diseases included: Philadelphia-positive ALL, ALL in CR2 in BFM S4, AML not in CR after course 2, CR1 < 12 months or not in CR, CML in second chronic phase and AML/ALL in CR3 or greater.

There was no evidence for efficacy in nine patients with refractory AML, but there could be a number of reasons for this observation. The use of pre-transplant Campath antibodies in some of these patients may have reduced the effect of natural killer (NK) cells in the KIR-mismatched group. The ability of KIR-mismatched haploidentical allografts to cure patients with AML not in remission is being tested by a UK national study in adults using the Perugia protocol including ATG.

How might haploidentical transplants cure patients with resistant leukaemia? High-dose chemoradiotherapy and the provision of ‘clean’ marrow play a role but there is a clear evidence that a graft-versus-leukaemia effect is important in both AML and ALL (Passweg et al, 1998). These effects are generally insufficient for DLI to be effective in overt disease but may be extremely important in patients who have achieved a minimal residual disease status. Haploidentical SCT does result in some ‘conventional’ GVHD (both acute and chronic) but there may also be some effect via NK cell alloreactivity in AML. This effect may be sufficient to cure a minority of patients with relapsed refractory leukaemia although the initial results of the Perugia group (Aversa et al, 1998) have not been verified by other groups and are in relatively small numbers of patients.

The way we currently perform these transplants requires refinement. The effective near total elimination of the immune system by a profound T-cell depletion in order to achieve engraftment and prevent severe GVHD seems a high price to pay. It is possible that pretransplant Campath 1H delayed immune reconstitution and may have contributed to the high incidence of viral deaths. Much research is ongoing into engineering grafts so that patients can respond to viruses (Peggs, 2004) and specifically kill tumour cells, but this work is in the early stages.

Infection remains a major stumbling block. In the largest series of paediatric haploidentical transplants reported so far (63 children) 16% of patients succumbed to viral infection within the first 180 d. These deaths were due to disseminated adenoviral infection, CMV disease or herpes simplex virus (Feuchtinger et al, 2005), adenovirus being the most prevalent – a feature common to all series of haploidentical SCT.

Cellular immunotherapy approaches are being developed in an attempt to combat these problems. One approach involves in vitro removal of alloreactive cells that express activation markers, such as CD25, in allogeneic reactions. This allows administration of residual cells capable of responding to many different pathogens, accompanied by a lower risk of GVHD induction than would be expected with unselected DLI. One group have reported death from infection in only one of 15 patients treated and improvement in antiviral immunity from as few as 3 × 105/kg T cells (Amrolia et al, 2005).

An alternative approach is to clone cytotoxic lymphocytes against specific viral and fungal antigens, administering pools of those clones that appear free of alloreactivity in vitro. Such a system has shown promising early results in the haploidentical setting (Perruccio et al, 2005). Unfortunately, this technique requires very large-scale laboratory culture, is prolonged (taking approximately 1 month) and is difficult to perform to modern Good Manufacturing Practice standards.

In summary, the results reported in this series are modest but they are helpful in defining future directions. We found no evidence for these transplants in children with refractory AML or resistant ALL using our protocols but are aware that our conditioning may have reduced the effect of alloreactive NK cells. The results in children in remission do indicate that medium-term disease-free survival can be achieved and careful consideration of a haploidentical SCT in selected patients can be justified.