The outcome of children requiring admission to an intensive care unit following bone marrow transplantation


Dr Marks Bone Marrow Transplant Unit, Bristol Royal Hospital for Sick Children, St Michael's Hill, Bristol BS2 8BJ.


We report the results of a retrospective study of the role of intensive care unit (ICU) admission in the management of 367 children who underwent bone marrow transplantation (BMT) at a tertiary referral institution. 39 patients (11%) required 44 ICU admissions for a median of 6 d. 70% received marrow from unrelated donors, half of which were mismatched; 80% had leukaemia and two-thirds were considered high-risk transplants. Respiratory failure was the major reason for admission to ICU. 75% of admissions required mechanical ventilation (for a median of 5 d) and 20 patients had lung injury as defined by the criteria of the Seattle group. None of 11 patients with proven viral pneumonitis survived (P = 0.06) and only one of 20 patients with lung injury survived (P < 0.01). Six of seven patients with a primary neurological problem survived (P < 0.001); these appear to represent a good outcome group. Age, the presence of graft-versus-host disease, the use of inotropes, isolated renal or hepatic impairment, and paediatric risk of mortality (PRISM) score were not predictive of outcome. In total, 12 patients (27% of admissions) survived and were discharged from hospital 30 d or more after admission and eight (18%) survived >6 months. ICU admission can be beneficial to selected children post-BMT but it may be less useful in proven viral pneumonitis. Where mechanical ventilation is required, the duration of this support should be limited unless there is rapid improvement.

High-dose therapy and stem cell transplantation are used increasingly to treat malignant and non-malignant diseases in children. Studies of adult bone marrow transplant recipients have shown that 16–40% of patients will require admission to an intensive care unit (ICU) in the post-transplant period ( Rubenfeld & Crawford, 1996; Paz et al, 1993 ). Reasons for admission include respiratory failure, septic or cardiogenic shock, hepatorenal failure, or acute intracranial events. A large proportion of these patients require mechanical ventilation which is associated with a poor outlook, with long-term survival post-ventilation in only 3–4%. The Seattle group have recently described the outcome of a large cohort of mechanically ventilated (predominantly adult) bone marrow recipients and found that younger age, a lower APACHE III score, and a shorter time from transplantation to intubation were predictive of survival ( Rubenfeld & Crawford, 1996). There were no survivors among a group of patients who had lung injury and either required >4 h of vasopressor support or had sustained hepatic and renal failure. The same study also demonstrated improved survival in a more recent time cohort, despite the confounding influence of worsening case mix and severity.

The purpose of this study was to employ similar techniques to analyse a paediatric cohort of BMT patients with the aim of identifying prognostic variables which were predictive of outcome. Thus, life-prolonging intensive therapy would not be administered to children with no chance of survival, with the aim of lessening the physical and emotional burden on children and their parents.


We conducted a retrospective study of all children (defined as age 17 or younger) who underwent bone marrow transplantation at the Bristol Royal Hospital for Sick Children between January 1987 and August 1997. The hospital is a tertiary referral centre for bone marrow transplantation and specializes in unrelated donor transplantation ( Oakhill et al, 1996 ). The case notes of 367 children were reviewed to determine which children required admission to the ICU at the United Bristol Healthcare Trust following their admission for bone marrow transplant (BMT). The commonest indications for transplantation were active lymphoblastic leukaemia (ALL: 53%), acute myeloid leukaemia (AML: 14%) and severe aplastic anaemia (8%). 222 transplants (60%) were from unrelated donors. There were 18 second transplants. Some patients were admitted to the ICU on more than one occasion. Re-admissions were only considered as separate events if the reason for the second admission was different and if at least 7 d had elapsed following the previous ICU discharge.

There were no strict criteria for ICU admission. Typical indications included respiratory failure, septic shock, and acute neurological events ( Table I). Patients with post-operative complications requiring a short ICU admission and those admitted after a single seizure were excluded from analysis. Each patient's severity of illness was assessed during the first 24 h following admission to the ICU. The paediatric risk of mortality (PRISM) score ( Pollack et al, 1988 ), which is derived from 14 physiological and laboratory parameters, was used to calculate the predicted risk of mortality. In addition, organ failure was analysed as a predictor of outcome. Renal impairment was defined as a serum creatinine >100 μmol/l and hepatic impairment as a bilirubin level >40 μmol/l. We defined lung injury as a requirement for an inspired concentration of oxygen (FiO2) of >60%, or a positive and inspiratory pressure (PEEP) value of >6 cm water, for >24 h. In addition we stratified our patients into either high-risk or standard risk groups for transplant outcome. Standard-risk patients included autologous transplants and either sibling or fully-matched unrelated donor transplants for acute leukaemia in first or second complete remission (CR1/CR2), chronic myeloid leukaemia in first chronic phase or thalassaemia. High-risk transplants included second transplants, transplants for osteopetrosis, unrelated donor transplants for severe asplastic anaemia or advanced leukaemia, haplo-identical transplants, and transplants from mis-matched unrelated donors.

Table 1. Table I. Clinical characteristics of children admitted to ICU following bone marrow transplantation.Thumbnail image of

Survivors were defined as those children who were alive 30 d after their ICU discharge and who were subsequently discharged from the hospital ( Rubenfeld & Crawford, 1996). In addition we assessed longer-term survival (>6 months post-ICU discharge). Survivor and non-survivor groups were compared statistically. Continuous variables were analysed using the Student t-test and categorical variables using the Mantel-Haenszel chi-square test. Not significant (NS) implies a P value of >0.10.


Patients and reasons for ICU admission

During the study period, 39 patients (10.6% of transplanted children) had a total of 44 admissions to the Paediatric ICU following bone marrow transplantation. Two cases admitted on day −1, who subsequently received their transplant on the ICU, were included in the analysis. The clinical characteristics of these admissions are outlined in Table I. There were more male admissions than female. The median age was 8 years (range 0.3–17.5). Leukaemia was the diagnosis in 80% (35/44) of admissions. 31 admissions (70%) followed unrelated donor transplantation. The median time interval from marrow infusion to ICU admission was 36 d (range −1 to 1005). 23 admissions (52%) were for respiratory failure, with septic shock and acute neurological events being the other main contributors ( Table II). 23/43 evaluable (allogeneic) admissions (54%) had active GVHD at the time of ICU admission.

Table 2. Table II. Reason for admission to ICU.Thumbnail image of

Degree of support

3 Table III outlines the degree of ICU support required, together with details on risk factor categories. Of the 44 admissions, 33 (75%) resulted in mechanical ventilation for a median duration of 5 d. In total 29/367 patients were ventilated (8%). Inotrope support was required in 14 (32%) of 44 cases. Four patients also received nitric oxide but no patient received high-frequency oscillatory ventilation or extracorporeal membrane oxygenation (ECMO).

Table 3. Table III. ICU support and risk factors.Thumbnail image of

Lung injury

Twenty patients (45%) met our criteria for lung injury. Nine patients had proven viral pneumonitis (CMV, four; RSV, three; parainfluenza, two) and none survived. (Two further patients with viral pneumonitis did not meet the criteria for lung injury as one declined ventilation and the other patient died within 24 h of ventilation). The reasons for admission were as follows: proven viral pneumonitis, nine; idiopathic pneumonitis, three; sepsis and respiratory failure, four; fluid overload, two; hepatic failure, one; obliterative bronchiolitis, one.

Survival and predictive factors

Twelve (27%) of admissions resulted in short-term survival (>30 d) and eight (18%) survived 6 months or more. The presence of lung injury was predictive of poor outcome ( Tables IV and V). Only one of 20 (5%) patients with lung injury survived compared to 11/24 (46%) without this risk factor. In particular, there were 11 admissions with proven viral pneumonitis, none of whom survived (0/12 v 11/32, P= 0.06). Seven patients had lung injury and hepatic impairment; there was a trend towards this being associated with poor outcome (P= 0.08).

Table 4. Table IV. Correlation of categorical risk factors with survival.Thumbnail image of
Table 5. Table V. Correlation of continuous variables with survival.Thumbnail image of

Twelve patients survived to be successfully discharged from hospital 30 d following ICU discharge. Of these patients, six (50%) were admitted to ICU with repeated seizures, three of whom required mechanical ventilation. (In fact only one patient with a primary neurological reason for admission did not survive; this may represent a good prognosis group.) Four survivors had sepsis, one requiring mechanical ventilation. The other two survivors included a patient with probable cyclophosphamide-induced cardiac dysfunction (not ventilated) and a patient with respiratory failure secondary to fluid overload in the treatment of haemorrhagic cystitis. This patient required mechanical ventilation for 7 d and met our criteria of lung injury. With the exception of this last patient, no survivor was mechanically ventilated for >2 d.

A number of potential risk factors were not found to be predictive of outcome. These include age (within the paediatric cohort), high versus standard risk BMT, the use of inotropes, isolated renal or hepatic impairment or the presence of GVHD. The mean PRISM scores in survivors and non-survivors were 21 and 34 respectively, but this difference was not significant (P= 0.21). Four of 12 admissions that resulted in survival versus 10/32 non-surviving admissions had a high risk (>30%) PRISM score (P= NS).


In this study we present the results of a retrospective analysis of the outcome of ICU admission in 367 children at a single centre who underwent BMT in a 10-year period. Of these, 11% required ICU admission and only 8% mechanical ventilation (compared to 25% in the Seattle and 12% in the Philadelphia studies respectively ( Rubenfeld & Crawford, 1996; Paz et al, 1993 ). In order to assess the role of ICU admission in the management of children undergoing BMT we chose to analyse nearly all patients admitted to ICU regardless of indication. Only patients with brief post-operative and neurological problems were excluded (see Methods). Our results identified a successful outcome in 27% of cases with 18% long-term survival (>6 months), supporting the valuable role of ICU care in certain patients. Stratifying our patients into high and low risk categories did not appear to correlate with outcome. Similarly, the use of the PRISM score, which does not take into account diagnosis, was not predictive of outcome in our patient group. Risk factors which did appear predictive of poor outcome were lung injury and proven viral pneumonitis. A primary neurological reason for admission was associated with a favourable outcome.

There are multiple adult series reporting nearly 2000 patients who were mechanically ventilated post transplant but there is a paucity of data concerning the outcome of paediatric patients. The Johns Hopkins group reported 318 patients transplanted between 1978 and 1988 (before ganciclovir was available) with an intubation rate of 6% and a survival of 9% ( Nichols et al, 1994 ). Todd et al (1994 ) described 54 intubated children (1973–90) with 11% survival to hospital discharge and a 9% 6-month survival. Although these data are from an earlier era of transplantation, many findings were in accord with our study. No patient with proven viral pneumonitis survived ventilation. Parenchymal lung disease without a proven pathogen and multiorgan failure were also associated with a poor outcome. Of the six survivors in their study, five received ventilatory support for <72 h. A diagnosis of leukaemia was an adverse factor, in contrast to our series. However, it should be noted that these studies predominantly concern matched sibling donor transplants with only a small proportion of unrelated donor transplants.

Our experience indicates that a requirement for mechanical ventilation post-BMT should not contraindicate admission to the ICU. 33 patients required ventilation and five (15%) patients were successfully discharged from hospital, with four of these patients surviving long term. However, of the 11 patients with proven viral pneumonitis, 10 were mechanically ventilated (one patient declined). None of these survived. It is difficult to make definitive recommendations based upon only 11 patients, but if the prognosis of children with proven viral pneumonitis who also require mechanical ventilation post-BMT is so poor, then perhaps a decision not to admit them to the ICU may seem justified. However, this must be tempered by the knowledge that small numbers of children and adults have survived this complication with the assistance of mechanical ventilation (A. J. Cant, personal communication; P. Ljungman, personal communication).

In addition to proven viral pneumonitis, our series identified the onset of lung injury as a related, significant, poor prognostic factor. Only one of these patients survived: a patient with fluid overload. Thus, the presence of lung injury, as identified by our criteria, during mechanical ventilation — even in the absence of a proven viral aetiology — has been shown to be associated with a poor short-term prognosis. Our data also seem to argue for brief trials of mechanical ventilation in post-transplant patients.

It is obvious that our study has significant limitations. Analyses of relatively small numbers of patients treated over 10 years with a variety of reasons for ICU admission do not lend themselves to making firm recommendations. Our results cannot necessarily be applied to other centres as different protocols, diseases, types of transplant and post-BMT complications may be critical in determining outcome. It is also possible that innovations in therapy in the future may improve results. Our numbers were insufficient to carry out a statistically valid time cohort analysis in order to determine if outcome has improved. In this study we were not able to consider the possible impact of nitric oxide, high frequency oscillatory ventilation or extracorporeal lung support. It is certainly not our intention to discourage well-conducted trials of new therapies in critically ill patients.

The practice of withholding life-sustaining therapy in children has been recently reviewed by the Royal College of Paediatrics and Child Health ( Anonymous, 1997). A situation of ‘no chance’ was described as when treatment delays death without alleviating suffering. Schneiderman et al (1990 ) proposed a quantitative approach to medical futility. They suggested that if experience suggests that the chance of success is <1%, then medical intervention may be considered futile. In practical terms, not recommending ICU admission to the parents of a child who is deteriorating post transplant can be difficult. In situations of less than total certainty, many parents may opt for a trial of ICU therapy including mechanical ventilation, even when the chances of success are very small. In fact, many paediatricians might argue that the demonstration that everything possible has been done is helpful to parents in coming to terms with the fact that their child is dying. However, in the ‘no chance’ situation a decision to change from life-sustaining treatment to palliative treatment may be in the best interests of the child.

In summary, ICU admission can be beneficial to a selected group of children following BMT. Larger studies are required to draw more definitive conclusions. However, in children with proven viral pneumonitis, a consensus decision to withhold mechanical ventilation may be justified. When mechanical ventilation is required, unless there is some improvement within 72 h it would seem reasonable to restrict the duration of this support.


We thank Mr R. Thorne and Mrs H. Hawkins for their invaluable assistance with data collection and the nurses of the BMT unit and Paediatric ICU for their expert care of the patients.