Prognostic significance of risk group stratification in elderly patients with acute myeloid leukaemia

Authors


Professor Anders Wahlin, Haematology/Medicine, University Hospital of Northern Sweden, S-901 85 Umeå, Sweden. E-mail:anders.wahlin.us@vll.se

Abstract

Prognostic factors were studied in a series of 211 acute myeloid leukaemia (AML) patients over 60 years of age, treated at a single centre. The patients were allocated into three risk groups based on cytogenetics, occurrence of antecedent haematological disorder and leucocyte count. Only 3% had low-risk features, 39% had intermediate- and 58% had adverse-risk features. Complete remission (CR) was achieved in 43% of all patients. In multivariate analyses, the number of cycles needed to achieve CR and the risk group were significantly associated with the duration of CR. Median survival time for the entire cohort of patients was only 107 d. Advanced age, low induction treatment intensity, treatment during earlier years and adverse-risk group were associated with shorter overall survival times. Risk group classification may help selection of elderly patients with a good chance of benefiting from intensive treatment to actually receive such treatment, while sparing others with a low probability of survival benefit from toxic treatment. Low intensity induction treatment reduces the chance of obtaining complete remission, produces inferior survival times and should consequently be avoided when the aim is to obtain complete remission. In elderly AML patients, introducing age and re-evaluation of intermediate and good prognosis patients regarding response to induction treatment may improve the risk group classification.

Progress has been made in the understanding of the pathogenesis of acute myeloid leukaemia (AML) during the last two decades, contributing to improved outcome for patients with AML (Löwenberg et al, 1999). Cytogenetics has become the single most useful factor for predicting outcome to induction and post-remission therapy (Arthur et al, 1989; Stasi et al, 1993; Dastugue et al, 1995; Raimondi et al, 1999). According to a recent report, 50% of adult patients with de novo AML with favourable cytogenetic abnormalities may be expected to be disease-free at 5 years, while less than 10% of those with adverse cytogenetic aberrations may be cured (Bloomfield et al, 1998). The means of accomplishing the improvement have been intensification of therapy with the use of high-dose cytarabine, autologous and allogeneic stem cell transplantation in selected cases, paralleled by improvements in supportive care.

The biology of AML has been less extensively studied in the elderly than in younger patients. According to a recent South Western Oncology Group (SWOG) report, 27% of patients in leukaemia trials were over 65 years of age although patients above 65 years comprised 63% of all leukaemia cases (Hutchins et al, 1999). In Sweden, the median age of patients with AML is 67 years. The recent improvement in outcome has primarily benefited younger patients below 60 years of age. The prevalence of poor-risk cytogenetic abnormalities and secondary AML that has evolved from myelodysplastic syndrome (MDS) is considerably higher among elderly patients than younger patients. Elderly patients tolerate intensive chemotherapy, including high-dose cytarabine, less well. Even with the use of intensive chemotherapy the chance of achieving remission is lower in the elderly and, if remission is achieved, the relapse rate is higher than in younger patients. Allogeneic transplantation can be carried out successfully in selected patients up to about 65 years of age (Deeg et al, 2000), but the transplantation-associated risks have generally been regarded as too high to make allogeneic transplantation an option in patients older than 60 years of age during the period of time that we studied.

The prognostic significance of pretreatment karyotype has lead to the recent development of programmes stratifying patients with AML into two or three risk groups based on their leukaemia karyotype. The prognostic information is often used for selecting younger patients at low risk of relapse to receive only chemotherapy and those at high risk of relapse for transplantation. However, very little is known about the prognostic significance of the same factors applied to elderly patients with AML. For several years our department has been the only centre treating AML irrespective of age in a defined geographical area. The organization of the leukaemia care in our region has enabled us to retrospectively study the prognostic significance of risk factors in a cohort of patients that is representative of elderly patients with AML.

Patients and methods

Patients All 211 patients over 60 years of age who were diagnosed with AML at our centre between 1982 and 1998 were studied. Most patients were referred from district hospitals, all of which lack facilities for treatment of AML throughout the period of time. The Swedish Cancer Registry, to which all cases have to be reported by legislation, contained 288 reported AML cases in this age category who were diagnosed in the catchment area during the study period, so 73% of the cases were referred to our centre and included in the study. Patients fulfilling the diagnostic criteria of AML (Bennett et al, 1985) or refractory anaemia (Bennett et al, 1982) with a bone marrow blast cells percentage exceeding 20% [refractory anaemia (RA) with excess blasts in transformation (RAEBT)] were included, irrespective of a prior history of myelodysplasia or other antecedent haematological malignancy, cytotoxic treatment or radiation.

Cytogenetics Chromosomal preparations were made at diagnosis from bone marrow, either directly or after 24–48 h of culture. Standard techniques using colchemid and hypotonic treatment, methanol–acetic acid fixation, conventional Leishman staining and G-banding were applied. Cytogenetic examinations were not routinely repeated after the diagnostic examination.

Risk group classification Risk group stratification was performed retrospectively, based on cytogenetic findings at diagnosis, occurrence or absence of antecedent haematological disorder and leucocytosis, as outlined below. Risk group 1, favourable prognosis, was cytogenetically characterized by the demonstration of any of three aberrations; inv(16), t(8;21) or t(15;17), irrespective of the concomitant occurrence of any other cytogenetic aberration. Central nervous system (CNS) leukaemia or trilinear dysplasia was not allowed. In cases with t(8;21), leucocyte counts exceeding 20 × 109/l were not allowed. Risk group 2, intermediate prognosis, included those with normal karyotype, or 11q23 abnormalities, +8, del(7q) as a single aberration, or other miscellaneous structural or numerical defects not encompassed by the favourable- or adverse-risk groups. Risk group 3, adverse prognosis, was defined by the occurrence of any of the following: presence of −5, del(5q), −7, t(3;3) or inv(3)(q21q26), t(6;9)(p23;q34), t(9;22), CNS leukaemia, antecedent haematological disorder, or a leucocyte count exceeding 100 × 109/l. Myeloproliferative disorders, myelodysplasia and lymphoproliferative disorders were all defined as antecedent haematological disorder if diagnosed and biopsy-verified at least 2 months before leukaemia diagnosis.

Chemotherapy Several different protocols of varying intensity were used for induction or palliation therapy. Each induction protocol was allocated into one of three intensity categories; high, intermediate or low.

High intensity protocols were:

DC, daunorubicin 45–50 mg/m2 I.V./d for 3 d in combination with continuous infusions of cytarabine 100–200 mg/m2 I.V./d for 7 d.

IC, idarubicin 10–12 mg/m2 I.V./d for 3 d in combination with continuous infusions of cytarabine 100–200 mg/m2 I.V./d for 7 d.

MC, mitoxantrone 12 mg/m2 I.V./d for 3 d in combination with continuous infusions of cytarabine 100 mg/m2 I.V./d for 7 d.

ACE, amsacrine 125 mg/m2 I.V./d for 5 d, plus etoposide 110 mg/m2 I.V./d for 5 d, in combination with continuous infusions of cytarabine 100 mg/m2 I.V./d for 5 d.

DC, MC or IC were used as first options during different periods of time in patients regarded as being eligible for high-dose induction therapy. ACE was used as an alternative to those protocols in patients with cardiac insufficiency or a history of heart problems. Intermediate intensity protocols consisted of the same drug combinations as the high intensity protocols, but at modestly reduced dosages. The most common reduction was to reduce the number of anthracyclin infusions by 1 d and the number of cytarabine infusions by 2 d. Low intensity protocols consisted of oral treatment with only hydroxyurea or thioguanine (21 cases), no consistent cytotoxic treatment (10 cases), low-dose cytarabine (seven cases), or ACE with amsacrine and etoposide for 2 d only (one case).

Two or three consolidation cycles were to be given according to the protocols in use, but in many cases the number of cycles actually given was smaller.

Definition of response Response to induction therapy was evaluated by bone marrow aspiration after each cycle, usually 4 weeks after the start of the cycle. In remission, bone marrow samples were obtained when clinically indicated, e.g. when haemoglobin or platelet levels started to fall from a steady level, or when changes were seen in the white blood cell (WBC) count or the differential count. Complete remission (CR) required a bone marrow cellularity > 25% with less than 5% blast cells, absence of blasts in the peripheral blood, neutrophils > 1 × 109/l and platelets > 50 × 109/l. Relapse was defined by the reappearance of more than 5% blasts in the bone marrow.

Definition of survival and response duration Overall survival was estimated from the date of diagnosis until the date of death. The duration of the response, defined as time in first CR, was measured from the date of complete remission until the time of relapse. Patients were censored for CR duration at the last date known to be alive and for death in first CR as a result of causes unrelated to leukaemia or leukaemia treatment.

Statistics A Cox regression model was applied to analyse the influence of prognostic factors on survival time and duration of response (Cox, 1972). Univariate and multivariate analyses were carried out. To avoid assumptions regarding linearity, all covariates were defined as categorical in the Cox regression model (Table I). Life table estimates were calculated according to Kaplan and Meier (Peto et al, 1977). The numbers of patients achieving CR in different groups were compared using Mantel–Haenszel tests (Mantel & Haenszel, 1959). A P-value < 0·05 was regarded as statistically significant.

Table I.  Variables included in Cox analyses.
VariableCategory
  1. See text for definitions of AHD, drug combination, cytogenetic class, and risk group.

Date of diagnosis1982–1989, 1990–1998
Serum lactate dehydrogenase (LDH, U/l)< 480, 480–1200, > 1200
FAB classM1, M2, M3, M4, M5, other
SexMale, female
Age group (years)60–69, 70–79, 80–89
WBC (× 109/l)< 10, 10–40, > 40
Antecedent haematological disorder (AHD)Yes, no
Intensity of induction cycle no. 1Low, intermediate, high
Drug combination in induction cycle no.1ACE, DC, other combination
Number of induction cycles to reach CR0 (not attained), 1, 2, 3
Cytogenetic risk class (based exclusively on cytogenetic findings)Favourable, intermediate, adverse
Risk group (based on cytogenetics, WBC, and AHD combined)Favourable, intermediate, adverse

Results

Patient characteristics

Two hundred and eleven patients > 60·0 years of age were diagnosed during the period 1982–98 (Table II). The median age of the patients was 72·8 years (range 60–90); 118 were men and 93 were women. Morphological classification of leukaemia subtype showed French–American–British (FAB) classes M1 and M2 to be the most common types. There were six cases of M0, 46 M1, 90 M2, 6 M3, 17 M4, 16 M5, 13 M6 and 17 myelodysplastic syndrome (RAEBT). Seventy-nine patients (37%) had antecedent haematological disorders; 58 myelodysplastic syndrome [18 RAEB, 16 RA, 9 sideroblastic anaemia (RAS), 8 RAEBT, 7 chronic myelomonocytic leukaemia (CMMoL)], 3 myelofibrosis, 2 myeloproliferative disorder, 2 non-Hodgkin's lymphoma, 1 polycythaemia vera, 1 essential thrombocythaemia, 1 multiple myeloma and 11 poorly defined myelodysplasia.

Table II.  Univariate analysis of factors affecting overall survival.
Clinical featurenP-valueExp(β)95% CI for exp(β) (range)
  1. AHD denotes antecedent haematological disorder. ref., reference group. See text and Table I for definitions.

Age (years)2110·000  
 60–6978ref.  
 70–79890·0001·9001·368–2·638
 80–89440·0002·5661·735–3·794
Sex2110·332  
 Female93ref.  
 Male1180·3321·1520·866–1·532
AHD2110·000  
 No132ref.  
 Yes790·0001·8861·401–2·539
Date of diagnosis2110·001  
 1982–1989100ref.  
 1990–19981110·0010·6110·460–0·813
FAB class2110·845  
 M146ref.  
 M2900·4321·1610·800–1·683
 M360·5400·7250·260–2·025
 M4170·3701·3020·731–2·320
 M5160·8880·9570·523–1·754
 Other360·7211·0860·690–1·711
WBC (× 109/l)2050·686  
 < 10113ref.  
 10–40450·8810·9730·676–1·399
 > 40470·4331·1520·809–1·641
LDH (U/l)1810·166  
 < 48041ref.  
 480–1200950·2020·7780·529–1·144
 > 1200450·7191·0850·697–1·688
Drugs in induction2110·098  
 ACE69ref.  
 DC770·9610·9920·707–1·392
 Other650·0651·3970·979–1·993
Treatment intensity2110·000  
 High131ref.  
 Intermediate440·0002·1531·507–3·077
 Low360·0003·4092·289–5·077
Number of cycles to CR2110·000  
 0121ref.  
 1700·0000·1080·072–0·161
 2 or 3190·0000·1480·086–0·254
Cytogenetics1800·000  
 Favourable70·0270·2730·087–0·862
 Intermediate134ref.  
 Adverse390·0021·8011·241–2·613
Risk group1980·000  
 Favourable60·0770·2800·068–1·146
 Intermediate79ref.  
 Adverse1130·0002·1641·577–2·970

Using cytogenetic classification only, seven patients were classified as cytogenetically favourable, 134 were intermediate and 39 were adverse. Sixteen patients had WBC counts > 100 × 109/l.

The combined information from cytogenetics, WBC and antecedent haematological disorders was used to form the risk groups. In 13 cases inadequate or missing cytogenetic information or other data did not permit risk group stratification, which accordingly could only be performed in 198 patients. Six patients were allocated to risk group 1, 77 to risk group 2 and 115 to risk group 3.

Non-referred patients

As stated in Patients and methods, 77 cases of AML diagnosed at the referring hospitals were not referred to our centre. From Swedish Cancer Registry data we calculated their median age as 77 years and their median survival time to 27 d. There was only one long-term survivor, who had received high intensity treatment at the district hospital. We did not perform validation of the diagnosis in these cases and figures concerning these patients are not included in the analyses performed in the present study.

Complete remission rate

Intensive induction treatment was used in all six patients (100%) in the favourable-risk group, 57/79 (72%) in the intermediate- and 62/113 (55%) in the adverse-risk group. CR was attained in 90 patients (43%). The CR rate was highest among patients in the favourable-risk group with four out of six patients entering CR. The CR rate was significantly higher (59%) in the intermediate-risk group than in the adverse-risk group (30%, P < 0·001). Patients without antecedent haematological disorders (AHD) had a significantly better chance of obtaining CR, with 52% entering CR in this group versus only 27% among patients with AHD (P < 0·001). The CR rate was 71% among patients with favourable cytogenetics, 51% among patients with intermediate cytogenetics and 23% in the cytogenetically adverse group (P < 0·001). The WBC count at diagnosis did not predict for obtaining CR (P = 0·31). The CR rate was 58% among patients who were treated with intensive induction therapy, 32% with intermediate intensity and 0% with low intensive or no chemotherapy (P < 0·001).

Overall survival

Age, antecedent haematological disorder, date of diagnosis, intensity of induction treatment, number of cycles to CR, cytogenetics and risk group were statistically significant in the separate univariate Cox regression models (Table II). Sex, FAB class, WBC count, serum lactate dehydrogenase (LDH) or drug combination used for induction had no significant impact on overall survival in univariate analyses (Table II). The median survival times by risk group were > 691 d for the favourable-risk group, 321 d for the intermediate-risk group and 66 d for the adverse-risk group (Fig 1). The differences were significant (P < 0·001). Median survival time for the entire cohort of 211 patients was only 107 d owing to the large fraction of patients with adverse-risk features.

Figure 1.

Kaplan–Meier estimate of overall survival by risk group. Differences between groups are significant (P < 0·001). Tick marks represent patients still under observation.

In a multivariate analysis we used the same variables as for the univariate analyses with the exceptions of the number of induction cycles to CR, antecedent haematological disorder, WBC and cytogenetics (Table III). The reasons for these exclusions are that the number of cycles needed to reach CR is unknown at diagnosis and the risk group classification is based upon the other three variables. Age, date of diagnosis, treatment intensity and risk group were all statistically significantly associated with overall survival time in the multivariate analysis. Advanced age, treatment during earlier years, low induction treatment intensity and adverse-risk group were all associated with shorter survival times. Sex, FAB class or induction drug combination were not predictive of survival.

Table III.  Multivariate analysis of factors affecting overall survival.

Clinical feature

P-value

Exp(β)
95% CI for exp(β)
(range)
  1. Categories within brackets are reference groups. AHD, antecedent haematological disorder. See text and Table I for definitions.

Age (years)
 (60–69)0·017  
 70–790·0131·6881·118–2·547
 80–890·0101·9861·180–3·341
Sex
 Female0·993  
 Male0·9930·9980·709–1·405
Date of diagnosis
 1982–19890·008  
 1990–19980·0080·6150·430–0·881
FAB class
 M10·234  
 M20·2150·7510·478–1·181
 M30·0905·6380·761–41·745
 M40·6351·1880·583–2·420
 M50·2750·6620·316–1·388
 Other0·2980·7580·449–1·278
LDH (U/l)
 (< 480)0·014  
 480–12000·4000·8280·533–1·286
 > 12000·0711·5980·961–2·657
Drugs in induction
 ACE0·378  
 DC0·3640·8260·547–1·247
 Other0·5161·2180·672–2·211
Treatment intensity
 High0·050  
 Intermediate0·0311·6501·046–2·602
 Low0·1301·7010·855–3·384
Risk group
 Intermediate0·000  
 Favourable0·1430·2270·031–1·648
 Adverse0·0002·2291·547–3·211

Disease-free survival

Ninety patients (43%) achieved remission, but one of them died the same day. The variables age, antecedent haematological disorder, LDH, cytogenetics and risk group were statistically significant in the separate univariate Cox regression models (Table IV). The median disease-free survival by risk group was > 1564 d for patients with favourable characteristics, 474 d for intermediate characteristics and 237 d for patients with adverse-risk features (Fig 2). The differences were significant (P < 0·001).

Table IV.  Univariate analysis of factors affecting disease-free survival.
Clinical featurenP-valueExp(β)95% CI for exp(β) (range)
  1. AHD denotes antecedent haematological disorder. ref., reference group. See text and Table I for definitions.

Age (years)890·031  
 60–6944ref.  
 70–79360·0102·0131·178–3·440
 80–8990·9081·0540·436–2·548
Sex890·740  
 Female42ref.  
 Male470·7401·0890·657–1·806
AHD890·010  
 No69ref.  
 Yes200·0102·1381·199–3·814
Date of diagnosis890·287  
 1982–198932ref.  
 1990–1998570·2870·7520·446–1·271
FAB class890·683  
 M121ref.  
 M2340·9390·9760·515–1·848
 M330·2840·3310·044–2·500
 M470·9281·0520·351–3·153
 M5100·9150·9500·371–2·443
 Other140·2771·5340·710–3·314
WBC (× 109/l)870·361  
 < 1044ref.  
 10–40200·1971·5060·808–2·807
 > 40230·31313780·740–2·568
LDH (U/l)770·010  
 < 48016ref.  
 480–1200410·0090·4200·218–0·808
 > 1200200·8790·9460·461–1·941
Drugs in induction890·909  
 ACE41ref.  
 DC330·8970·9650·559–1·666
 Other150·6630·8460·399–1·795
Treatment intensity890·180  
 High75ref.  
 Intermediate140·1801·6070·831–3·109
 Low0   
 Number of cycles to CR890·141  
 169ref.  
 2 or 3200·1411·5260·869–2·681
Cytogenetics810·001  
 Favourable50·0890·1790·025–1·302
 Intermediate67ref.  
 Adverse90·0023·2231·541–6·743
Risk group840·000  
 Favourable/intermediate51ref.  
 Adverse330·0002·9021·692–4·977
Figure 2.

Kaplan–Meier estimate of disease-free survival by risk group. Differences between groups are significant (P < 0·001). Tick marks represent patients still under observation.

The variables were entered in a multivariate regression analysis. LDH, number of cycles needed to achieve CR and risk group were statistically significant (Table V). However, the biological significance of the LDH level was impossible to interpret because patients with modestly increased levels had the best outcome. In addition, the drug combination used in induction treatment was nearly significant (P = 0·06). The need for more than one induction course and adverse-risk group predicted for a short time in remission.

Table V.  Multivariate analysis of factors affecting CR duration.

Clinical feature

P-value

Exp(β)
95% CI for exp(β)
(range)
  1. Categories within brackets are reference groups. AHD, antecedent haematological disorder. See text and Table I for definitions.

Age (years)
 (60–69)0·119  
 70–790·2271·5820·752–3·331
 80–890·4300·6080·177–2·094
Sex
 Female0·108  
 Male0·1080·5750·293–1·130
Date of diagnosis
 1982–19890·424  
 1990–19980·4240·7330·342–1·570
FAB class
 M10·609  
 M20·1270·5020·207–1·217
 M30·9770·0000·000
 M40·7320·7730·178–3·364
 M50·1760·4370·132–1·449
 Other0·6590·7940·284–2·216
LDH (U/l)
 (< 480)0·001  
 480–12000·0050·2850·120–0·679
 > 12000·6751·2450·446–3·473
Drugs in induction
 ACE0·062  
 DC0·0570·4610·208–1·023
 Other0·3781·6460·544–4·978
Treatment intensity
 High0·999  
 Intermediate0·9991·0000·356–2·803
Number of cycles to CR
 10·017  
 2 or 30·0172·8561·204–6·775
Risk group
 Favourable/intermediate0·000  
 Adverse0·0006·7462·918–15·592

Risk group

The risk group variable was composed of cytogenetics, antecedent haematological disorder and WBC count. The multivariate regression analyses of overall survival and disease-free survival reported in Table III and Table V were repeated with the variables antecedent haematological disorder, WBC count and cytogenetics in lieu of risk group. Antecedent haematological disorder was statistically significant both for overall survival (P = 0·006) and CR duration (P = 0·028), but cytogenetics was statistically significant only for overall survival (P = 0·0037). The WBC count was nearly significant for CR duration (P = 0·054) but not for overall survival (P = 0·70).

Discussion

The 211 patients reported in this study can be regarded as fairly representative for AML in this age category as they comprised 73% of the entire cohort of elderly patients with AML in the geographical area. However, some patients, 27% of the entire cohort of elderly AML, were not referred to our centre. The median overall survival time for patients who were not referred was very short, 27 d, indicating that these patients were judged to be too ill to have any chance of benefiting from leukaemia treatment or being transported to our centre. During the study period, 66% of all AML cases at our centre were diagnosed in patients over 60 years of age, reflecting the true age incidence of AML in the Swedish population.

The prevalence of antecedent haematological disorders, 37%, was high among our patients, but this figure also includes non-MDS malignant blood disorders. Twenty-seven per cent of the patients had MDS prior to AML. Antecedent haematological disorder was found to be associated with a poor chance of attaining remission, brief response and poor survival. Patients with preceding MDS, myeloproliferative disorders and preceding cytotoxic therapy have long been known to have an inferior response to therapy and survival compared with those with de novo AML (Mertelsmann et al, 1980; Hoyle et al, 1989).

In the 1990s, many AML studies confirmed the prognostic importance of specific cytogenetic lesions (Stasi et al, 1993; Dastugue et al, 1995; Leith et al, 1997; Grimwade et al, 1998). In the present study, cytogenetics was highly significantly associated with the CR rate, disease-free survival and overall survival, confirming the report by Leith et al (1997).

Leucocytosis was not associated with poor survival or brief response duration, in spite of the use of the leucocyte count as one of the factors in our risk group classification, as well as in similar classifications in present use in Great Britain and Germany. Although leucocytosis correlates with a poor prognosis in some reports (Dutcher et al, 1987), the correlation is not very strong, possibly because fatal complications of leucocytosis are rare in AML except in promyelocytic leukaemia, in which the risk of disseminated intravascular coagulation is correlated with the WBC count. Only six patients in the present series had promyelocytic leukaemia. The relationship between WBC count and tumour burden is presently unknown and other factors including chemoresistance seem to be more important for outcome.

The principal aim of the present investigation was to evaluate the prognostic impact of patient characteristics on response rate, response duration and survival, in particular the prognostic value of our present risk group classification. We have used the same risk group classification since 1997 for patients below the age of 60 years, primarily for treatment decisions concerning autologous and allogeneic transplantation, but its prognostic value, if any, for patients over 60 years is unknown. The classification is based on the WBC count, absence or presence of antecedent haematological disorder and type of cytogenetic aberration. A few cytogenetic aberrations are known to be prognostically favourable, others have an adverse impact and a third group is composed of lesions with an intermediate impact. The three cytogenetic groups in the present study were based on reports from other large studies concerning the prognostic importance of karyotypically defined aberrations (Arthur et al, 1989; Stasi et al, 1993; Dastugue et al, 1995; Leith et al, 1997; Grimwade et al, 1998).

Separation of the patients into three risk groups was feasible in 94% of the patients, although the proportions of the groups differ from those seen in younger patients. Only 3% had low-risk features, 40% had intermediate- and the proportion of patients with adverse-risk features was large, 57%. The differences in outcome between the risk groups were highly significant, measured both as overall survival and response duration. The median survival times were at least 23 months for the favourable-risk group, 11 months for the intermediate-risk group and 2 months for the adverse-risk group. These differences reflect large differences in the chance to enter remission as well as the duration of remission.

Other factors known to be prognostically important in AML in general were studied. In agreement with other reports, increasing age was found to be associated with poor survival, but age was not found to be statistically significant for the duration of response. Thus, advanced age seems to reduce the chance to enter CR, but not the quality of the response once attained. No significant differences in outcome between the two sexes were seen and there was no correlation between morphological (FAB) AML class and outcome.

The intensity of the induction treatment was found to be important for the outcome. Patients treated with high intensity, i.e. induction chemotherapy in doses generally recommended for younger patients, had a significantly higher chance to achieve complete remission and better survival times than patients treated with reduced doses. Similar observations have been reported previously (Tucker et al, 1990; Arlin et al, 1991). There were no differences in outcome between the induction regimens. The similarity in outcome was surprising considering the fact that one of the regimens, consisting of amsacrine, etoposide and cytarabine, was used preferentially in patients with cardiac problems or antecedent haematological disorders.

The number of induction cycles needed to attain remission was prognostic for response duration, confirming previous reports about the importance for outcome of time to CR (Estey et al, 2000).

The present patient series was diagnosed and treated over several years. Although the outcome for most elderly patients with AML is still dismal, we observed a significant improvement in overall survival from the 1980s to the 1990s. The improvement is probably the result of better supportive care and better adaptation of different treatment options to the individual patient, improving the chance of attaining CR (Mandelli et al, 1998), because there was no difference in CR duration over time.

In summary, the most important information gained from the present study is that risk group stratification founded on cytogenetics is also feasible and informative in elderly patients with AML. Inclusion of the leucocyte count into the risk group stratification did not improve the prognostic value. The use of risk group stratification yields knowledge about the chance of obtaining durable remission and survival time in the individual patient which can be used for treatment decisions, i.e. selection of patients with a good chance of benefiting from intensive treatment to actually receive such treatment while sparing others with a low probability of survival the effects of toxic treatment. Reducing the intensity of the induction treatment reduces the chance to obtain complete remission and produces inferior survival times; it should be avoided when the aim is to obtain complete remission. For treatment decisions in elderly AML, the risk group classification may be improved by introducing age and re-evaluation of patients with intermediate or good prognosis with regard to response to induction treatment. The proposed strategy would make it easier to abstain from or stop treatment when the patient's chance of benefiting from treatment is very small and to use intensive therapy in adequate doses in those patients who can benefit from such treatment.

Acknowledgment

This study was supported by Lion's Research Fund in Umeå.

Ancillary