The improvement in outlook for children with acute lymphoblastic leukaemia (ALL) has been largely as a result of the decreased incidence of relapses. However, a number of children, who might otherwise be cured, still tragically die from infections and other complications of treatment. We have examined the pattern of treatment-related deaths (TRD) in three consecutive Medical Research Council (MRC) United Kingdom ALL (UKALL) trials to determine whether there has been any change in the causes of TRD and in the factors influencing the risk of death.
In the last 20 years, the survival rate for children with acute lymphoblastic leukaemia (ALL) has markedly improved, largely owing to a decrease in relapses. However, children still die from complications of treatment and these are potentially preventable. We have analysed data from three large consecutive national protocols for ALL from 1980 to 1997 [Medical Research Council United Kingdom ALL (MRC UKALL) trials VIII, X and XI] to compare the incidence and causes of treatment-related deaths (TRD). The percentage of TRD has fallen from 9% to 2% (UKALL VIII to XI), largely as a result of a decrease in fatal infections. Deaths during induction have fallen from 3% to 1%, the main causes of death being bacterial, followed by fungal infection, while other causes, chiefly haemorrhage, have not declined. Remission deaths also decreased from 6% to 1%, particularly those deaths due to measles and pneumocystis carinii. More guidelines for surveillance and treatment of infections have been included within progressively more intensive protocols. Risk factor analysis showed increased TRD in patients with Down's syndrome, high leucocyte count and older age in UKALL XI. In contrast, the introduction of blocks of intensification was not associated with an increased death rate. While improved supportive care has reduced the incidence of TRD, there is still scope for further reduction by prompt treatment of suspected infection. Maintenance of herd immunity remains of vital importance in avoiding deaths from measles.
PATIENTS and METHODS
Patients All patients entered into the MRC UKALL VIII, X and XI trials (September 1980 to 1997) were included in this study. In UKALL XI, there was no early randomization and no absolute requirement to register on the trial until d 28, hence, a number of children who died early were not registered, although there had been an intention to enrol on this study. Extensive efforts have been made to collect data on these cases and include them in the analyses to avoid any bias as a result of their omission. Details of all patients were obtained from data held at the Clinical Trial Service Unit in Oxford that included notification forms, death registration and post-mortem findings.
Protocols The details of the protocols have been described elsewhere (Eden et al, 1991; Chessells et al, 1995; Wheeler et al, 1996; Richards et al, 1998), and Table I summarizes and compares the main features of each trial and their differences. All trials have closed and have at least 21/2 years follow-up. Deaths were restricted to those during therapy or deaths occurring in a period of up to 6 months from stopping treatment.
|UKALL VII||UKALL X||UKALL XI|
|Supportive care guidelines||General||Detailed||Detailed|
|Induction Anthracycline||Randomized||Yes||Pre-1992: Yes|
|(Arm B)||Post 1992 No|
|Early (week 5) intensification||No||Randomized||Pre-1991: randomized (D)|
|(Arm B + D)||Post 1991: Yes|
|Second (week 20) intensification||No||Randomized||Pre-1992: (C + D)|
|(Arm C + D)||Post 1992: Yes|
|Third (week 35) intensification||No||No||Pre-1992: No|
Post 1992: Randomized
|Cranial radiotherapy||Yes||Yes||CNS + ve (24 Gy)|
|(18 Gy)||(18 or 24 Gy)||High count: randomized|
|High-dose methotrexate||No||No||Randomized: < 50 WCC; vs. IT MTX,|
> 50 WCC; vs. RT
|Continuation therapy||Randomized||2 years||2 years|
|CR rate 95%|
|CR rate 98%|
|CR rate 99%|
Statistics Death rates were analysed using chi-squared tests with the application of Fisher's exact test when necessary. All P-values are two-tailed. Multivariate analysis was undertaken via logistic regression to test for independence and interactions. Statistical analysis was performed using SAS software.
The total number of treatment-related deaths, excluding failed remission or relapse, during induction and first remission in each trial are shown in Fig 1 (expressed as a percentage of children entered in each trial). Infective deaths were more common than non-infective deaths in UKALL VIII and X, whilst in UKALL XI they were approximately equal. Figure 2 shows, by trial, the categories of infective agent causing death.
The number and type of deaths occurring during the initial 4 weeks (of treatment induction) are shown in Fig 3. The overall induction TRD rate was 2·8%, 2·4% and 1·2% in UKALL VIII, X and XI respectively. Further details of types of infection and non-infective causes are seen in Tables II and III. The leading cause of death during the first 4 weeks was sepsis due to bacteria, with Gram-negative pseudomonal and colifom species most common, especially in UKALL VIII and X. A number of fungal deaths (both Candida and Aspergillus spp.) occurred during induction in the trials, most notably in UKALL X. Recognition of the risk appears to have subsequently reduced the incidence of death, despite identical or more intensive therapy in UKALL XI. Non-infective deaths included haemorrhage (mainly cerebral) and a single fatal thrombotic event. Tumour lysis syndrome (TLS) causing death was relatively infrequent with a maximum of two cases per trial. Other causes included leucostasis with associated haemorrhage, encephalopathy and liver failure. In five cases, the cause was unknown.
|Induction||Early block||Second block||Late block||BMT||Continuation|
Type of organism
Deaths in first remission
The overall pattern in the MRC UKALL trials of deaths in first remission is shown in Fig 4. The TRD rate during remission was 6%, 2·6% and 1·3% in UKALL VIII, X and XI respectively. Intensification of treatment has been associated with improved disease control but a rise in infective episodes. There were eight deaths associated with the early intensification block at week 5, three in UKALL X (all infective) and five deaths in UKALL XI (four infective and one case of adult type respiratory distress syndrome). The second (week 20) intensification block had 10 associated deaths, seven in UKALL X (all infective) and three deaths in UKALL XI (two infective and a case of disseminated necrotizing leucoencephalopathy). There was only one death following the third (week 35) intensification block, from a fungal infection in UKALL XI. During continuation therapy, a secular change in the infective pattern of deaths could be seen in all trials with viral infections (particularly measles) in UKALL VIII + X and pneumocystis pneumonitis in UKALL VIII, being more important causes than in the later trials. There were only two infective deaths recorded after completion of chemotherapy, both in UKALL XI and occurring within 2 months of the end of treatment (one bacterial and one fungal due to Candida spp.).
Bone marrow transplant-related deaths
There were two recorded deaths following allogeneic matched unrelated donor (MUD) bone marrow transplant (BMT) in UKALL VIII for which no cause was recorded. Of eight deaths during allogeneic BMT in UKALL X [five matched sibling donor (MSD), three (MUD)], six were infective in origin (two fungal, two viral and two with pneumocystis pneumonitis) and the others were from graft-vs.-host disease (GVHD) in unrelated BMT. The 10 BMT (seven MSD, three MUD) deaths in UKALL XI comprised five infections (three bacterial, two adenoviral), GVHD in two (MUD) cases, graft failure in one (MSD) case and multiorgan failure in two cases.
Risk factors for treatment-related deaths
The results of the univariate analysis of the risk of death during induction and first remission are shown in Tables IV and V. Those patients undergoing BMT in first remission were excluded from the analysis, in order to eliminate bias, as it is well established that BMT is a strong risk factor in TRD in it's own right. Sex was not a significant risk factor for toxic death except in UKALL X, in which there was an excess of deaths in female patients during induction. Age below 10 years was a risk factor for induction TRD in UKALL VIII, but in XI the risk factor was for those aged more than 10 years who were at risk in both induction and first remission. A leucocyte count of greater than 50 × 109/l at diagnosis was a significant risk in the UKALL VIII and XI during induction. Children with Down's syndrome were at greater risk of treatment-related death in UKALL X and XI. There were no children with Down's syndrome treated on UKALL VIII. T immunophenotype was a significant risk factor for treatment-related deaths during induction in UKALL VIII and in both induction and first remission in XI, but was not independent from a high leucocyte count. Further multivariate analysis, using logistic regression with continuous variables, confirmed that the trend for improvement in both induction and first remission death rates over trials was independently significant. For induction deaths, leucocyte count and sex were also significant, although the sex effect was only seen in UKALL X. For first remission deaths, Down's syndrome was an independent risk factor and there was no interaction between age and trial, with older children at lower risk in UKALL VIII and higher risk in UKALL XI.
|Male||453||12||12·6||NS||922||10||22·3||P < 0·001||1200||14||14·2||NS|
|< 2||111||3||3·0||NS||163||4||3·9||NS||162||6||1·9||P < 0·001|
|WCC ( × 109/l)|
|< 50||688||13||19·1||P = 0·001||1270||28||30·7||NS||1632||9||17·0||P < 0·001|
|Common† + others||557||13||16·3||P = 0·007||1349||32||33·5||NS||1736||11||16·0||P < 0·001|
|No||No patients with||1579||36||37·6||P = 0·05*||2018||5||4·9||NS|
|No||308||3||7·8||P = 0·015|
|< 2||107||7||6·4||P = 0·04||154||4||3·2||NS||155||2||1·5||NS|
|WCC ( × 109/l)|
|Common† + others||543||34||32·9||NS||1291||27||27·6||NS||1682||11||13·6||P = 0·025|
|No||No patients with||1491||30||31·7||P = 0·03*||1937||12||14·7||P = 0·003*|
|Early + late||289||5||5·3|
In UKALL VIII after the first year, there was an induction randomization for anthracycline administration on the first 2 d of induction (Eden et al, 1991): group B children received daunorubicin, group A did not. Group B had a significantly increased risk of toxic death during induction, but not first remission, compared with group A. Continuation therapy of 3 years duration compared with 2 years randomized, as in UKALL VIII (Eden et al, 1991), did not show a significantly increased risk of treatment-related death during the third year. In the later two trials, which included randomization to receive varying numbers and timing of intensification blocks, there was no significant relationship between treatment allocation and risk in first remission.
Despite the introduction of progressively more intensified treatment in the three trials, the incidence of treatment-related deaths during induction and remission has fallen from 9% to approximately 2%. This improvement has been as a result of a decrease in infective deaths, while those owing to non-infective causes, particularly bleeding, have not changed. These findings reflect improvements in the management of the febrile neutropenic child and in the prevention of non-bacterial infections, particularly measles and Pneumocystis carinii.
The pattern of infective deaths has changed, with bacterial causes becoming less frequent and, in particular, a reduction in fatal infections owing to Gram-negative organisms such as Pseudomonas spp. and Escherichia coli. A number of factors have contributed to this improvement. Studies by the EORTC group (EORTC group, 1990; Hann et al, 1997) have confirmed the reduction in bacteraemias owing to Gram-negative organisms in recent years and the now more prevalent Gram-positive bacteraemias tend not to have such an aggressive clinical course (an exception being the Streptococcus viridans group). There has been widespread acceptance of guidelines for management of febrile neutropenia, in particular the prompt empirical use of the newer anti-pseudomonal penicillins in combination with aminoglycosides. Patients who were randomized to receive daunorubicin in induction in UKALL VIII had a significantly increased risk of death in induction and omission of this drug in UKALL XI may have contributed to a reduction in infections.
Deaths from fungal infections, both Candidia and Aspergillus species, occurred during both induction and intensification, but were less common in UKALL XI. This is presumably owing to the widespread acceptance of the need for empirical amphotericin therapy in patients with persistent febrile neutropenia (EORTC group, 1989) and recognition of hepatosplenic candidiasis (Carstensen et al, 1990; Sallah et al, 1999) on recovery from neutropenia.
The most gratifying change has been the reduction in fatal non-bacterial infections. MRC UKALL VIII was associated with a significant number of fatal pneumonitis episodes (Darbyshire et al, 1985) and the introduction of routine prophylaxis with oral cotrimoxazole resulted in a reduction in Pneumocystis carinii pneumonitis, as in other trials. Measles in the immune-suppressed child causes an interstitial pneumonitis for which there is no effective treatment (Gray et al, 1987; Kernahan et al, 1987) and the introduction of routine childhood measles, mumps and rubella (MMR) immunization in 1988 has fortunately resulted in the virtual abolition of this devastating complication. However, the recent drop in uptake of MMR vaccine (Wakefield et al, 1998; DeStafano et al, 2000) is of great concern, and immunization of siblings and social contacts remains extremely important if cases are to be avoided. There has also been a reduction in deaths owing to chickenpox and herpes simplex viruses with the standard management of passive immunization and acyclovir therapy.
The non-infective death rate has remained constant during the two decades, with only a small reduction in the main causes. Cerebral haemorrhages, associated with high leucocyte counts and leucostasis, usually occurred within the first 48 h. Fatal tumour lysis was seen in one or two children in each trial. More active measures to prevent these complications, such as leukapheresis and exchange transfusion, are of unproven benefit but have been used in some series (Wald et al, 1982; Bunin & Pui, 1985).
A potential weakness of this study was the failure to register all cases at diagnosis in UKALL XI, but extensive efforts were made, facilitated by the registration of over 90% of all childhood cancer cases with the UKCCSG, to trace any unreported early deaths. This is the only recent study of treatment-related death in a large multicentre setting, but it compares favourably with a previous report from a single centre in the UK (Atra et al, 1993).
In this large series, we were able to extend the previous analysis of factors influencing the risk of treatment-related deaths. The risk factors for death during induction were not consistent throughout the three trials, a pattern which may be owing to chance or could reflect changes in treatment, mainly the use of daunorubicin. Children with Down's syndrome were vulnerable during induction therapy in UKALL X, as previously reported (Wheeler et al, 1996), and during continuing therapy in both UKALL X and XI. This is presumably related to their relative immune deficiency (Robison et al, 1984; Levitt et al, 1990; Anneren et al, 1992; Dordelmann et al, 1998). As the results of therapy in Down's syndrome are otherwise favourable, great care must be taken to reduce infective deaths. High initial leucocyte count is a significant risk factor during induction and is associated particularly with early cerebral haemorrhage. The mechanism of greater toxic death with T-ALL is probably related to presenting high white cell count and does not seem to be an independent effect. The effects of age on remission deaths between trials were contradictory and the reason for this is unclear.
The major obstacle to cure for most children with leukaemia remains relapse of their disease. The evidence that treatment-related mortality has been reduced in consecutive trials provides encouragement that further intensification of therapy may produce a further decrease in relapse without an associated increase in treatment mortality.
MEMBERS OF THE WORKING PARTY (DURING UKALL VIII, X and XI TRIALS)
O. B. Eden (Chairman), C. C. Bailey, P. R. H. Barbor, A. Barrett, C. Barton, V. Broadbent, M. Caswell, S. C. Cartwright, J. M. Chessells, A. W. Craft, P. J. Darbyshire, S. I. Dempsey, J. Durrant, P. Emmerson, D. I. K. Evans, K. Foreman, J. J. Fenelly, D. A. G. Gatton, B. Gibson, A. Goodman, R. Gray, C. Haworth, I. M. Hann, M. Hewitt, F. G. H. Hill, J. Kernahan, D. J. King, S. E. Kinsey, J. Kohler, M. E. M. Jenny, I. J. Lewis, J. S. Lilleyman, M. Madden, J. R. Mann, J. Martin, T. J. McElwain, S. T. Meller, C. Mitchell, P. H. Morris-Jones, A. Oakhill, J. Peto, M. Radford, J. K. H. Rees, M. M. Reid, S. M. Richards, R. Shannon, O. P. Smith, R. F. Stevens, G. P. Summerfield, A. Thomas, E. N. Thompson, D. A. Walker, H. Wallace, D. Webb, K. Wheatley, A. Will, K. P. Windebank and A. Vora.