Rritish journal o/ Hac~matoloyg.199 1 , 78. 187-1 96

ADONIS

0 0 0 7 10489 1 0 0 1 2 3D

Results of Medical Research Council Childhood Leukaemia Trial UKALL VIII (Report to the Medical Research Council on behalf of the Working Party on Leukaemia in Childhood) 0. B. EDEN,' J. s. L I L L E Y M A N . ' S . KICHARDS,j M . P. S H A W 'A N D J . P E T 0 4 'Royal Hospitalfor Sick Children, Edinburgh. 'The Children's Hospital, Shefield, 3The Clinical Trials Service Unit, Oxford, and 41nstitute of Cancer Research, London

Received 7 September 1990; accepted for publication 28 Ianuary 1991

Summary. During the 1971)s. despite apparently similar treatment, the prognosis for children with lymphoblastic leukaemia (ALL) improved more in some countries, notably the United States and West Germany, than in others. To find out why, the first phase of the United Kingdom (UK) Medical Research Council (MRC) childhood ALL trial, UKALL VIII, was designed to see whether similar results to the IJnited States Children's Cancer Study Group (CCSC) could be obtained in the U.K. using an identical protocol (CCC 162). Protocol 162 was one of a series of regimens devised by the American Children's Cancer Study Group in the 19 70s and was used specifically for their average risk patients (all children with ALL with an initial white cell count up to 5 0 x 1 O'/l except those aged 3-6 years with white cell counts under 10 x 10"/l).One arm ( 1 A ) of their study was adopted by the MRC for all children in the U.K. aged 0-14 years with confirmed ALL. Eight hundred and twenty-nine consecutive patients were entered between 1980 and 1984. The first 199 patients formed a single arm study as per the original protocol 162 (arm 1 A), but the subsequent 6 30 children were randomized to receive or not two doses of daunorubicin on the first 2 d of induction. This randomization was an attempt to answer the

important question as to whether event-free survival was influenced by the use of four rather than three induction agents. A second randomization between 2 and 3 years continuing therapy was also introduced at this stage as it had been by the CCSC in their protocol. With a minimum follow up period of more than 5 years, disease-free survival for the whole group is 55%, a considerable improvement on all previous UKALI, trials. Results for patients directly comparable with those in CCSG 162 ('average risk' patients) and their American counterparts were similar. Daunorubicin was associated with more early deaths but improved disease-free survival for those achieving remission. More children relapsed who stopped treatment after 2 years than those who continued for 3 . but this was balanced by increased treatment mortality in the third year. The fact that for UKALL VIII the results were similar to those of the CCSG suggests that previous MRC protocols were not sufficiently sustained and intensive, particularly during the maintenance phase of treatment. Daunorubicin may have provided better disease control. albeit with increased toxicity, and 3 years of maintenance therapy did not confer any advantage in long-term survival over 2 even though it may have reduced the rate of early relapse.

From 1970 the Working Party on Childhood Leukaemia of the IJnited Kingdom Medical Research Council (MRC) con-

ducted a series of therapeutic trials for acute lymphoblastic leukaemia (ALL).The broad principles of therapy, consisting of a 4-week induction phase with prednisolone. vincristine and asparaginase. central nervous system directed treatment (cranial irradiation and a course of intrathecal methotrexate injections) and continuing therapy for 2-3 years with oral 6mercaptopurine and methotrexate with periodic pulses of steroids and intravenous vincristine, were established as early as 1972. By this time a substantial proportion of longterm survivors was accumulating and clinically identifiable prognostic subgroups had become apparent. In Trials 11-VII various aspects of 'standard' therapy were tested (including

Members of the Working Party during IJKALL VIII: Professor T. J . McElwain. Dr C. C. Bailey, Dr P. R. H. Barbor. Dr V. Broadbent. Dr S . C. Cartwright. Dr A. W. Craft, Professor J . M. Chessels. Dr S. I. Dempsey. Dr J . Durrant, Dr 0. B. Eden, Dr P. M. Emerson, Professor D. A. C. Galton. Dr 1. M. Hann. Dr F. G. H. Hill, Dr J . Kernahan. Dr J. S. Lilleyman. llr J . R. Mann. Dr J . Martin, Dr P. Morris-Jones, Dr A. Oakhill. Professor J . Peto. Dr J. K . H. Kees. Dr M. Kadford. 1x R. F. Stevens, Dr G. P. Sumnierfield and Dr E. Thompson. Correspondence: Dr 0. B. Eden. Koyal Hospital for Sick Children. Edinburgh.

187

0.B. Eden et al

188

!

I

Bone marrows

i

1

Radinllieraov

111

W

e -4

I

e

2

k

3

’

4

~

I

5 6 induclion

I

-+ 7

I

8

I

1

I

2

~

3

4

I

I

I

I

5 6 7 8 hja1nIe naric e

I

9

~

1

I

1011lZll

I

I

~

Fig 1. Schematic representation of UKALL VIII study. UKALL VIII Trial varied only by the randomization to receive (arm 9) or not (arm A) daunPmycin 45 mg/m2 i.v. on days 1 and 2 of the induction. From 1 9 8 3 length of continuing therapy was randomized between 2 and 3 years. N.B.6 Bone marrow at Day 14 was not mandatory for the study and early part oftrial but as evidence accumulated of its prognostic significance it was requested from 1 9 8 3 onwards.

additional agents during induction, maintenance and pulsed continuing therapy) but in general the results were disappointing with little improvement for the 1470 patients enteredbetween 1972 and 1979 (MRC, 1982a, b, 1986a, b). Whenever there was a major deviation from ‘standard’ therapy results were generally inferior. This stagnation in therapeutic progress was not being experienced elsewhere. In particular, better results were being observed by the U.S. Children’s Cancer Study Group (CCSG) in North America (Hammond et a/, 1986: Poplack & Reaman. 1988) and also by the BFM group in West Germany (Riehm et al, 1983). To find out why the U.K. results were falling behind, in 1 9 8 0 the MRC sought permission to adopt a CCSG protocol for use in Britain. The idea was to treat all children (ages 0-14 inclusive) with ALL, irrespective of prognostic features, with an identical protocol to one simultaneously in use elsewhere. Because treatment regimens on both sides of the Atlantic appeared superficially similar, it was not clear at the time whether the discrepancy in U.K. results was due to an unrecognized difference in therapy or a different U.K. incidence of subtypes of ALL. The first phase of UKALL VIII, the single arm ‘study’, was designed to answer this question. Subsequently two unrelated randomizations were added: the giving or not of daunorubicin on the first 2 d of treatment, and a comparison of 2 versus 3 years overall length of therapy. This report summarizes the results.

PATIENTS AND METHODS

UKALL VIII ran from 1 September 1980 to 3 1 December 1984 and was open to all children with acute lymphoblastic leukaemia aged 0-14 years inclusive (Fig 1). Individual physicians diagnosed their patients on clinical, peripheral blood and bone marrow appearances (standard May-Grunwald-Giemsa (MGG), periodic acid schiff (PAS), sudan black, esterase and acid phosphatase stains were all required). They were also required to complete a data file on any patient entered and submit unstained marrow smear slides to a three-man review panel (J.S.L., I.H.. R.S.). Slides were centrally stained with MGG and PAS stains. Each patient was by consensus opinion of the panel given a diagnosis (ALL confirmed or refuted) and a French-American-British (FAB) classification whenever the material was adequate for analysis. Initial presenting data required included patient age. sex. initial white cell count, haemoglobin level, platelet count, blast cell immunophenotype, cytogenetic data on blasts, liver, spleen and node size, presence or not of mediastinal mass and immunoglobulin levels. Initial cerebrospinal fluid (CSF) cell count and cytospin results were also recorded. The protocol for the initial single arm study was identical to the CCSG protocol 162 arm 1A. The immediately obvious differences between this and previous UKALL Trials were:

1. A course of L-asparaginase was given as intramuscular (im.) injections commencing on day 4 and given three times

I

l

Childhood Leukaemia Trial UKALL Vlll a week for 3 weeks, a longer course than previously used and the drug was given ism. instead of i.v. 2. Daily mercaptopurine in full dose was continued throughout the CNS therapy phase (it had previously been reduced to 50% at this stage in an attempt to reduce lymphopenia and immunosuppression). 3. No gaps were permitted during induction and between induction and continuing therapy. 4 . Mercaptopurine (6MP) and methotrexate were commenced at full target dosage (6MP 75 mg/mL) and only reduced in response to neutropenia or thrombocytopenia (to 50% target dose if the neutrophil count fell to below 1.0 x 1 0 y / l or platelets to below 100 x 10y/l: to 0% if the neutrophil count went below 0.5 x 10y/lor platelets below 50 x 1 0”/1). Previously the criteria for dose reductions were less specific and more at the physician’s discretion. 5. In the continuing therapy phase vincristine and prednisolone were always given irrespective of peripheral blood counts. Previously this had not always been so. 6. Intrathecal methotrexate dosages were calculated on an age rather than surface area basis: Age (yr) Dose (mg) 3 12.5 7. For the first time infants under 1 year were entered into the study and trial. 8. Prophylactic co-trimoxazole was introduced after the first 6 months to combat the high initial incidence of pneumonitis. 9. More emphasis was placed upon clinicians complying with the protocol, with specific brand forms of drugs being used and restrictions on any deviations. 10. The use of 1800 cGy cranial irradiation (10 fractions in 2 weeks) compared with the previous standard 2400 cGy. After trial entry the only exclusions from analysis were those where the diagnosis of ALL was not sustained by the independent review panel. All other patients were included in analyses even if protocol deviants. From 1 September 198 1 a single randomized variable was introduced for patients to receive (arm B) or not (arm A ) a dose of daunorubicin (45 mg/m2 i.v.) on each of the first 2 d of treatment. Treatment on arm A was consequently identical to that in the study. At the beginning of 1983 in line with a decision by the CCSG a second randomized variable was introduced to decide duration of continuing therapy between 2 and 3 years. This randomization took place at 2 years from diagnosis after continuing full remission had been confirmed by marrow review. The trial was closed to entry of new patients in December 1984. All phases of UKALI, VIII recruited all types of ALL. but due to alternative U.K. studies there was a relative lack of T cell ALL (identified by clinical, radiographic and immunophenotyping) before 1983, a lack of B cell ALL (surface immunoglobulin positive blasts) after 1982 and a lack of infants after 1983. The responsible physician was also required to complete

189

progress sheets for all patients including follow-up blood counts and marrow reports confirming remission achieved at day 14 or day 28 and CSF cell counts at each lumbar puncture (LP). Clinicians were also bound to record therapy given, toxicity experienced and relapses, deaths, or other major events. At the time of this analysis (July 1990) the minimum documented follow-up time was over 5 years. STATISTICAL METHODS All the main analyses were performed using the log rank survival method with a cut-off date of 3 1 October 1989 (Peto et al, 1977). All P-values quoted are one sided. The primary analysis was of disease-free survival. i.e. time to relapse or death. Patients who died without going into remission were counted as having an event on day 1. Other analyses looked at other endpoints: (a)CNS relapse, (b)testicular relapse, (c)haematological relapse, (d)death, (e) death in first remission, ( f ) relapse, (g) disease-free survival. excluding non-remitters. In the case of the first three, patients who relapsed in some other site had their survival time censored at that point. Both isolated and combined relapses were looked at in each case. Relapses occurring at different sites within 30 d of each other were considered to be combined. Patients who died without remission were excluded from the analyses of (a).(b) and (e) and counted as having an event on day 1 for (c) and (f). When examining treatment differences, only randomized patients were considered. Analyses were stratified by the prognostic features found to be important, and when comparing 2 versus 3 years’ treatment. also by initial treatment. RESULTS 199 children ( 11 6 boys, 83 girls) were entered into the first non-randomized phase (the study). and 630 ( 3 3 7 boys, 293 girls) into the subsequent trial giving a total of 829 in UKALL VIII altogether. 322 received daunomycin (arm B) and 308 (arm A) did not. Three patients who were considered not to have ALL were excluded.

lnduction mortality (see Table I) Deaths during induction, including those failing to remit. constituted 4.6%of all patients. The principal cause of death was septicaemia complicated by haemorrhage. The commonest bacteria causing death were Gram negative, but overall the commonest identified infecting organisms were Gram positive (57% of all infections). Infection was less frequent in arm A of the trial compared with the study, but a higher risk of gram negative septicaemia and more prolonged neutropenia were noted in arm B. During the later stages of the study and at the beginning of the trial there was a gradual change from the use of E. coli L-asparaginase to the Erwinia product (due to reducing availability of the E. coli product in the U.K.) with a clear decrease in morbidity and mortality. This observation forms the basis of a separate report (Eden et al, 1990). Remission mortality (Table I and Table 11) Early in the study, the single most important cause of death during remission was interstitial pneumonitis (Darbyshire et

190

0.B. Eden et al Table I. UKALL VIII 1980-84: events by treatment arm

Total no.

Study

A

B

199

308

322

Failed induction (non-remitters+ deaths in induction) Died in remission Total relapsed Second tumour as first event Disease-free survival at 5 years Relapse-free survival at 5 years Remitters disease-freesurvival at 5 years

Table [(a). Analyses by randomized treatment in UKALL VIII Trial. All stratified by age. sex and white blood count.

Disease-free survival Relapse Disease-freesurvival (remitters only) Survival

No. of events/ no. of patients

Statistical calculations for arm B

Ann A

Arm B

0-E

Var ( 0 - E )

P-value

Odds ratio (95% confidence interval)

142/308 128/308

141/322 114/322

-6.0 -11.5

64.3 54.7

NS 0.06

0.9 (0.7. 1.2) 0.8 (0.6.1 . 1 )

133/299 91/308

122/303 104/322

- 10.9 5.1

58.2 43.2

0.08

0.8 (0.6.1 . 1 )

NS

1.1 (0.8,1 . 5 )

Arm A: No daunomycin; Arm B: daunomycin 45 mg/m2 i.v. days 1 and 2.

Table It. UKALL VIII deaths in first remission ( n = 53)

Pneumonitis Measles Pneumonia Septicaemia Other sepsis (meningitis) Bleeding (incl. 1 post liver biopsy) Other viral Chickenpox Herpes encephalitis Flu Other Encephalitis/encephalopathy Post BMT

Fibrosing alveolitis Road traffic accident Unknown

10 9 5 5 3 5 3 2 1 1 2 2 2 1 2

al. 1985). This prompted the introduction of prophylactic co-trimoxazole and no further deaths occurred from this cause in patients taking the drug. Although the peak incidence of death in remission was in the first 6 months of therapy, especially during episodes of marked lymphopenia. some risk did appear to persist throughout treatment and one death occurred 3 months after completion of therapy.

Disease-free survival (DFS) Fig 2 (top line) shows disease-free survival for all patients entered into UKALL VIII study and trial. Fig 2 also compares the results of UKALL VIII with those of the previous trials UKALL 11-VII for both standard and nonstandard therapy. The immediately observed improved disease-free survival appears to have been sustained (MRC. 1986a). Results from patients matched by risk group to those treated by the USCCSG on protocol 162 1A are virtually identical (Poplack & Reaman. 1988). Fig 3 shows DFS by treatment and Table I shows the different types of events. There was no significant difference in disease-free survival between the two randomized treatment arms (Table Ia). There was a significantly increased risk in arm B compared with A for ‘first month’ and remission deaths (44/322 versus 22/308. P 5 blast cells/mm’). superficial node size, initial platelet count, immunoglobulin levels and day 14 marrow cellularity gave no significant prediction of outcome. It is important to note that those with blasts in the CSF received extra therapy, and were few in number (12 with blast cells but < 5/mm’ and 11 with > 5 blast cells/mm’).

192

0. B. Eden et al 100

75

50 a

25

0 0

1

2

3

5

4

6

TIME IN YEARS

Fig 4. Disease-free survival from the time ofsecond randomization. Interrupted line=continuing 3 years of treatment, n = 203. 77% DFS (SE3). Continuous line=stopping at 2 years. n=203. 73% DFS (SE3).

Table IV. Disease-free survival from stop/continue randomization

Variable

Total

Observed events

Expected

O/E

Stop at 2 years Continue for 3 years

203 203

52 45

47 49

1.09 0.9 1

Odds ratio (95%confidence interval) 0.83 (0.56. 1.24)

Odds ratio for males only 0.78 (0.46. 1.30).females only 0.97 (0.52. 1.81).

Table V. UKALL VIII prognostic indicators: disease-free survival P-values for trend Variable ('P-value for heterogeneity)

Total analysed

WBC Age* Sex FAB (L3 excluded) Mediastinal mass Spleen size Liver size Cell type (C, T. N)* Cell type (C, T)* Haemoglobin Philadelphia chromosome Chromosomes* High hyper v. other Pseudo v. other Day 14 RM %blasts

829 829 829 73 3 814 827 829 61 1 584 81 3 552 281 281 281 577

Stratified by: Unstratified

WBC

+Sex

+Age

0.04

0.06

0.2 0.04 0.06 0.008

0.2 0.2 0-006

0.03

0.04

0.0005

0.003

< 0~00005 0.001 0~0001 0.06 0.008 0.0002 0.003 < 0~00005 0~0001 0.02 0.0007 0.05 0.004 0.03 04002

0.05 5% cells 'block' positive:

t

> 10% cells showing

vacuoles

Table VIII. Isolated CNS relapses: % (no. of relapses/no. in group)

Girls BOYS

'LOW'

'Average'

'High'

2.8% (2/71) 6.4% (8/125)

7.7% ( 1 7/221) 7.0% (171244)

9.1% ( 6 / 6 6 ) 12.5% (8.64)

Low=aged 3-6, WBC < I 0 average=all patients WBC 0-50 except 'low': high = all with WBC > 50.

Central nervous system relapse Table VIIl shows the isolated C N S relapse rate to have been approximately 7%. The single most important predictor of CNS relapse was initial white cell count (Fig 7) (P-value for trend 5 blasts/mm3 have so far shown further C N S relapse. Table VIII suggests that only girls with a white count < 10 x 109/1aged 3-6 truly may be considered to be at low risk for CNS disease. lsolated testicular relapse The rate of this complication was reduced compared with previous MRC Trials (Eden et al, 1978) and the only significant correlation found was with the presence of mediastinal mass (P=0*002)and T cell disease. Second tumours To date, eight possible second malignancies have been reported: two with non-Hodgkin's lymphoma? (true or late nodal relapse), four secondary non-lymphoblastic leukaemias (three monosomy 7), one Hodgkin's disease and one cortical glioma. Further analysis will be necessary to confirm the nature of these events, the incidence of which was evenly distributed between the study and two arms of the trial. Protocol compliance (a) Induction. No gaps in therapy were permitted during the first weeks of treatment and strict rules were applied for any dosage reduction. In assessment of the outcome for patients who had 7 or more days delay in therapy at this stage or 10% (or greater reduction) in dosage, the relapse rate was 42% for the 2 4 non-compliers compared with 28% for the others (P