World Health Organization classification in combination with cytogenetic markers improves the prognostic stratification of patients with de novo primary myelodysplastic syndromes

Authors


Prof. Paolo Bernasconi, Hematology Division, Fondazione IRCCS Policlinico San Matteo, 27100 Pavia, Italy.
E-mail: p.bernasconi@smatteo.pv.it

Summary

This study correlated chromosomal defects with French–American–British (FAB)/World Health Organization (WHO) classification subtypes, proposed a revised International Prognostic Scoring System (IPSS) cytogenetic grouping; and established which classification, when used with the IPSS cytogenetic categories, best predicted clinical outcome in the myelodysplastic syndromes (MDS). A higher prevalence of chromosomal defects and distinct defects were observed in patients with multi-lineage dysplasia and a blast cell percentage >10%. Abnormalities of the long arm of chromosome 3, del(7)(q31q35), trisomy 8, del(11)(q14q23), del(12p) and 20q- could be segregated from their respective IPSS cytogenetic categories and used to develop new cytogenetic subgroups. Clinical parameters, FAB/WHO classification, IPSS score and standard or revised cytogenetic categories were statistically relevant for overall survival (OS) and progression-free intervals (PFI) and were included within five distinct multivariate models compared by the Akaike Information Criterion. To predict OS, the best models included age, WHO classification and standard or revised IPSS cytogenetic categories; to predict PFI, the best model included the same variables and revised cytogenetic categories. In conclusion, (i) the WHO classification was associated with a more homogeneous cytogenetic pattern than the FAB classification, (ii) WHO classification and standard/revised IPSS cytogenetic categories were much more effective than IPSS for predicting MDS clinical outcome, (iii) revised cytogenetic subgroups predicted PFI more effectively than standard categories.

Myelodysplastic syndromes (MDS) are clonal stem cell disorders characterised by hypercellular marrow with ineffective haemopoiesis determining peripheral cytopenia. In 1982, the French–American–British (FAB) group developed a morphological classification that distinguishes five MDS subtypes: refractory anemia (RA), RA with ringed sideroblasts (rars), refractory anemia with excess of blasts (RAEB), RAEB in transformation (RAEB-t) and Chronic Myelomonocytic Leukemia (CMML) (Bennett, 2005). Recently, the World Health Organization (WHO) developed a new classification that identifies the following MDS subtypes: RA, RARS, 5q- syndrome (MDS 5q-), refractory cytopenia with multilineage dysplasia with ringed sideroblasts (RCMDS), refractory cytopenia with multilineage dysplasia (RCMD), RAEB type 1 (RAEB-1), RAEB type 2 (RAEB-2) and MDS unclassifiable (MDS U) (Jaffe et al, 2001). Consequently, patients diagnosed as RAEB-t by the FAB classification are no longer considered as an MDS by the WHO classification, as the latter system classifies patients with ≥20% marrow blast cells as acute myeloid leukaemia (AML) (Bennett, 2005). Chromosomal abnormalities are found in about 40–70% of de novo MDS patients; they also confirm the clonality of the disorder, identify peculiar morphological entities and, except for 5q- (which, as a single defect is now considered a new distinct entity by the WHO classification), are not specific for any FAB subtype. Moreover, the cytogenetic pattern has an independent prognostic significance (Pierre et al, 1989; Bernasconi et al, 1994; Olney & Le Beau, 2001).

The identification of the most relevant biological and clinical parameters affecting overall survival (OS) and progression-free interval (PFI), defined as the time interval between diagnosis and MDS/AML progression, has been and still is a primary goal in MDS patients, as the natural history of the disease varies widely (Aul et al, 2002). In the past, six scoring systems based on the FAB classification have been used (Mufti et al, 1985; Sanz et al, 1989; Aul et al, 1992; Morel et al, 1993; Toyama et al, 1993; Parlier et al, 1995); of these, three demonstrated the prognostic power of chromosomal defects by multivariate analysis (Morel et al, 1993; Toyama et al, 1993; Parlier et al, 1995). Subsequently, the International MDS Risk Analysis Workshop established that, in the MDS karyotype, blast cell percentage and amount of peripheral cytopenias are the most important prognostic variables and proposed a scoring system, the International Prognostic Scoring System (IPSS), which allows the subdivision of MDS patients into four prognostically different categories: low-risk, score = 0; intermediate (int)-1 risk, score = 0·5–1·0; int-2 risk, score = 1·5–2·0; high-risk, score ≥2·5 (Greenberg et al, 1997). However, despite its improved accuracy (Pfeilstöcker et al, 1999; Germing et al, 2000a; Sperr et al, 2001; Germing et al, 2005), the IPSS still has some pitfalls, one of the most important being the inclusion of miscellaneous, less frequent single chromosome defects and double defects in the intermediate cytogenetic category (Soléet al, 2000; Bernasconi et al, 2005). Recent evidence suggests that some of these less common abnormalities may play a very significant role in the MDS clinical outcome when large patient series are examined (Soléet al, 2005). In the meantime, it has been observed that the IPSS score and the WHO classification are equally effective in predicting MDS prognosis (Nösslinger et al, 2001; Lee et al, 2003; Howe et al, 2004; Malcovati et al, 2005).

In a previous study of 331 MDS patients, diagnosed according to the FAB classification, we demonstrated that chromosome defects, as grouped by IPSS and blast cell percentage, were the most relevant parameters for predicting OS and PFI and that the prognostic power of standard IPSS cytogenetic grouping was not ameliorated when del(7)(q31q35) was considered as a new prognostic entity (Bernasconi et al, 2005). That study could not assess the prognostic power of single abnormalities included within the intermediate cytogenetic category, because the number of patients harbouring such defects was insufficient for statistical analysis. Therefore, defining the clinical relevance of these defects was one of the main goals of the present study, which is also aimed to analyse the incidence of chromosomal defects in relation to FAB and WHO subtypes and establish which classification, in combination with IPSS cytogenetic categories, best predicts MDS clinical outcome.

Materials and methods

Patients

All of the 491 consecutive de novo MDS patients, including 331 analysed in a previous report (Bernasconi et al, 2005) and 160 newly diagnosed patients, were evaluated at the Division of Haematology, IRCCS Policlinico San Matteo, Pavia between January 1990 and December 2004. All of them were classified according to FAB criteria and 436 patients, including 276 of the 331 previously reported patients and the 160 newly diagnosed patients, had also been classified according to WHO criteria. Considering the previously described 331 patients, a revision of bone marrow morphology by two independent reviewers was possible in 327; for four patients (3 RA and one RARS with a normal karyotype), material was not available for morphological analyses. After this revision, the WHO classification could be applied to only 276 patients because 51, classified as RAEB-t and having a blast cell percentage >20%, were diagnosed as AML according to WHO criteria. No patient with secondary MDS or a white blood cell (WBC) count above 12 × 109/l was included in the study. Diagnostic procedures at the onset of disease and during follow-up as well as clinical monitoring were carried out as already described (Bernasconi et al, 2005). Moreover, because the IPSS was developed for patients undergoing treatment that does not affect the MDS natural course, and because the MDS natural course may be significantly changed by intensive chemotherapy and allogeneic bone marrow transplantation (allo-BMT), our patients submitted to these therapeutic procedures were evaluated only until the start of such treatment.

Cytogenetic studies

Chromosome studies were performed as previously reported (Bernasconi et al, 1994). Chromosome identification and karyotype description were made in accordance with the International System for Chromosome Nomenclature (Mitelman, 1995). Karyotypes were defined as complex when they included ≥3 chromosomal abnormalities.

Statistical analysis

Statistical analyses were carried out as previously described (Bernasconi et al, 2005). Continuous variables were dichotomised according to values reported in the literature (Greenberg et al, 1997). Median follow-up was computed according to the ‘reverse Kaplan Meier’ method, which calculates potential follow-up in the same way as the Kaplan–Meier estimate of the survival function, but with the meaning of the status indicator reversed. Thus, death censors the true but unknown observation time of an individual, and censoring is an end-point (Schemper & Smith, 1996). Therefore, in the present analysis censoring occurred in case of: end of study, death, start of intensive chemotherapy or allo-BMT, depending on which event occurred first. In fact, to estimate the prognostic significance of any specific chromosomal defect and the clinical relevance of the IPSS and WHO classification, all the data from our 491 MDS patients were collected until the start of intensive chemotherapy or allo-BMT, as both these therapeutic procedures may significantly change the natural course of the disease. Survival was estimated by using the Kaplan–Meier method (Kaplan & Meier, 1958). Progression-free interval was defined as the time elapsed between MDS diagnosis and MDS/AML progression. Univariate and multivariate Cox proportional hazard models were used to identify possible predictors of events. Hazard ratios (HR) and their 95% confidence intervals (95% CI) were computed; the corresponding P-values were also reported. The assumption of proportional hazards was verified.

To test the clinical impact of cytogenetics, the OS and PFI of each single defect were compared with those of the remaining set of abnormal cases with single aberrations (i.e. the total number of patients having single defects excluding patients with the defect analysed) and to those of the remaining set of abnormal cases included within the same IPSS cytogenetic category (i.e. the total number of patients with the defects included within that specific IPSS cytogenetic category excluding patients with the defect analysed). A two-sided P-value ≤ 0·004, determined by applying the Bonferroni correction, was considered statistically significant when ≥5 comparisons were made. Subsequently, cytogenetic defects with an OS or a PFI that was significantly different from those of the respective IPSS cytogenetic category were entered in to a revised IPSS cytogenetic grouping. Clinically relevant variables were included in a Cox multivariate model. Competing models were informally compared by means of Akaike Information Criterion (AIC). stata 9 (StataCorp, College Station, TX, USA) was used for computation. A two-sided P-value ≤ 0·05 was considered statistically significant.

Results

Patients

Patients’ clinical and haematological features are reported in Table I. In our series, the 2- and 5-year OS were 0·76% (95% CI, 95% CI = 0·72–0·82%) and 0·58% (95% CI = 0·51–0·64%), respectively. At the time of the analysis, 339 patients were alive and 152 had died. The 2- and 5-year PFI were 0·72% (95% CI = 0·66–0·76%) and 0·58% (95% CI = 0·51–0·64%), respectively.

Table I.   Patients’ clinical and haematological features.
Total patients491 (%)
  1. *Applied to 436 patients: FAB, French–American–British; WHO, World Health Organization; IPSS, International Prognostic Scoring System; RA, refractory anaemia; RARS, RA with ringed sideroblasts; raeb, RA with excess of blasts; RAEB-t, RAEB in transformation; MDS 5q-, 5q- syndrome; RCMD, refractory cytopenia with multilineage dysplasia; RCMDS, DCMD with ringed sideroblasts; RAEB-1, RAEB type 1; RAEB-2, RAEB type 2; MDS U, MDS unclassifiable; Int, intermediate.

Median age, years (Inter-quartile range)64·8 (54·9–71·9)
Age (years)
 <4557 (11·6)
 ≥45 ≤ 65193 (39·3)
 >65241 (49·1)
Sex
 Male295 (60)
 Female196 (40)
Haemoglobin (g/l)
 ≥100213 (43·4)
 <100278 (56·6)
Absolute neutrophil count, ANC (×109/l)
 ≥1·8239 (48·7)
 <1·8252 (51·3)
Platelet count (×109/l)
 ≥100263 (53·6)
 <100228 (46·4)
Cytopenias
 None60 (12·2)
 One204 (41·5)
 Two155 (31·6)
 Three72 (14·7)
Marrow blast cell (percentage)
 <5287 (58·5)
 5–1062 (12·6)
 11–2093 (18·9)
 21–3049 (10·0)
FAB classification
 RARS79 (16·1)
 RA209 (42·6)
 RAEB152 (30·9)
 RAEB-t51 (10·4)
WHO classification*
 RARS49 (11·2)
 RA78 (18·0)
 MDS del(5q)29 (6·6)
 RCMDS16 (3·7)
 RCMD105 (24·1)
 RAEB-162 (14·2)
 RAEB-290 (20·6)
 MDS-U7 (1·6)
IPSS
 Low115 (23·4)
 Int-1179 (36·5)
 Int-2110 (22·4)
 High87 (17·7)

Considering all 491 patients, treatment consisted of supportive care (transfusions and haemopoietic growth factors) in 372, differentiation-inducing agents in 15, immunosuppressive agents in 11, low-dose chemotherapy aimed at reducing the WBC count in 31, and various regimens of intensive chemotherapy in 52. Twenty patients were submitted to allo-BMT: 13 had previously received supportive treatment only and seven had been submitted to intensive chemotherapy, which induced a complete remission in four and a partial remission in three. Therefore, our patients were not uniformly treated. However, our results were not affected by this event and, in fact, the 52 patients submitted to intensive chemotherapy and the 20 patients submitted to allo-BMT were censored at the start of each therapeutic procedure. Moreover, the decision to administer treatment was exclusively based on either the patients’ age or on the fact that progression to a more advanced MDS or to AML had just occurred. Therefore, we are quite confident that, by excluding these patients, we did not affect the impact of cytogenetics or other clinical parameters on OS and PFI.

Cytogenetic results

At diagnosis, all 491 patients had successful cytogenetic analyses, a mandatory requirement to define their IPSS score. Clonal chromosome defects were detected in 294/491 (59·9%) patients classified according to FAB criteria and in 252/436 (57·7%) patients classified according to WHO criteria. The most frequent single defects, their overall incidence in relation to FAB/WHO subtypes and IPSS prognostic categories, and their grouping according to IPSS cytogenetic categories [Good = normal, -Y, 5q-, 20q- only; Intermediate = +8, various single and double defects, Poor = complex (≥3 defects), −7, del(7q)] are listed in Table II. For patients classified as RARS/RA according to FAB criteria, the overall frequency of chromosomal abnormalities was 50%, but when these same patients were assigned to RA/RARS and RCMDS/RCMD categories according to WHO criteria, the overall incidence of chromosomal defects was 39·3% and 50·4%, respectively. In addition, −7, del(7)(q31q35), 3q abnormalities and complex karyotypes (with ≥3 defects) had a higher frequency in patients classified as RCMDS/RCMD than in those classified as RARS/RA (16·5%vs 2·3%). Trisomy 6 was revealed in only two patients, both classified as RA and presenting a hypocellular marrow. One additional patient with marrow fibrosis, demonstrated by a trephine biopsy, was classified as RA and showed a normal chromosomal pattern.

Table II.   Patients’ cytogenetic features.
Total of patients491 (%)
  1. *Miscellaneous = other single defects occurring in ≤3 patients; numbers and percentages in parentheses represent the true numbers and the true percentages of patients having either an abnormal cytogenetic pattern or the cytogenetic defect analysed in each FAB, WHO, IPSS category; numbers and percentages in parentheses represent the true number and the true percentages of patients over the total number of patients.

  2. FAB, French–American–British; WHO, World Health Organization; IPSS International Prognostic Scoring System; RA, refractory anemia; RARS, RA with ringed sideroblasts; raeb, RA with excess of blasts; RAEB-t, RAEB in transformation; MDS 5q-, 5q- syndrome; RCMD, refractory cytopenia with multilineage dysplasia; RCMDS, DCMD with ringed sideroblasts; RAEB-1, RAEB type 1; RAEB-2, RAEB type 2; MDS U, MDS unclassifiable; Int, intermediate.

Karyotype
 Normal197 (40·1)
 Abnormal294 (59·9)
Most common single abnormalities
 +1q6 (1·2)
 3q rearrangements13 (2·6)
 del(5q)43 (8·7)
 –713 (2·7)
 del(7)(q31q35)20 (4·1)
 +822 (4·5)
 del(11)(q14q23)5 (1·0)
 del(12p)17 (3·4)
 del(17p)6 (1·2)
 del(20q)15 (3·0)
 –Y5 (1·0)
 Miscellaneous*27 (5·5)
Chromosomal defects in each FAB subtype
 RARS34 (43·0)
 RA110 (52·6)
 RAEB108 (71·1)
 RAEB-t42 (80·8)
Chromosomal defects in each WHO subtype
 RARS16 (32·6)
 RA34 (43·6)
 MDS del(5q)29 (100)
 RCMDS12 (75·0)
 RCMD49 (46·6)
 RAEB-141 (66·1)
 RAEB-267 (74·4)
 MDS-U4 (57·1)
Chromosomal defects and IPSS
 Low30 (26·1)
 Int-1100 (55·8)
 Int-282 (74·5)
 High82 (94·2)
Chromosomal defects
 Single192 (39·1)
 Double61 (12·4)
 Complex41 (8·4)
IPSS cytogenetic categories
 Good260 (53·0)
 Intermediate157 (32·0)
 Poor74 (15·0)

Prognostic factors

Univariate analysis

The clinical and biological factors effecting OS and PFI in the univariate analysis are listed in Table III. The relationships of OS and PFI to standard IPSS cytogenetic categories are shown in Fig 1. Among clinical parameters, age and gender had a clear impact on OS (P = 0·015 and P = 0·02, respectively), but not on PFI (P = 0·320 and P = 0·165, respectively). The chromosome pattern, including number and type of cytogenetic abnormalities as well as IPSS cytogenetic categories, had a significant impact on OS and risk of MDS/AML progression (Table III; Figs 1 and 2). Considering single defects, +1q, del(5q), +8, del(11)(q14q23), del(12p) and del(20q) presented an OS similar to that of normal karyotypes; whereas 3q abnormalities, −7, del(7)(q31q35), del(17p) presented a worse OS; +1q, del(5q), del(11)(q14q23) and del(12p) were also associated with a PFI similar to that of normal karyotypes, and 3q abnormalities, −7, del(7)(q31q35), +8, del(17p) and del(20q) were associated with a shorter PFI.

Table III.   Clinical and cytogenetic parameters relevant to survival and MDS/AML progression.
VariableSurvivalMDS/AML progression
Total ptsEvents (%)2-year survival (95% CI)5-year survival (95% CI) HR (95% CI)P-valueTotal ptsEvents (%) HR (95% CI)P-value
  1. Events = number and percentages of patients in whom the event (either death or MDS/AML progression) has occurred; HR, hazard ratio; 95% CI, 95% confidence intervals; NR, not reached; revised IPSS cytogenetic categories (cytog. cat.): the abnormalities included within each revised IPSS cytogenetic subgroup are provided in the text.

  2. FAB, French–American–British; WHO, World Health Organization; IPSS International Prognostic Scoring System; RA, refractory anemia; RARS, RA with ringed sideroblasts; raeb, RA with excess of blasts; RAEB-t, RAEB in transformation; MDS 5q-, 5q- syndrome; RCMD, refractory cytopenia with multilineage dysplasia; RCMDS, DCMD with ringed sideroblasts; RAEB-1, RAEB type 1; RAEB-2, RAEB type 2; MDS U, MDS unclassifiable; Int, intermediate.

Age (years)     0·015   0·31
 <45579 (16)0·84 (0·68–0·92)0·75 (0·55–0·87)5720 (35)
 45–6519353 (27)0·77 (0·69–0·83)0·62 (0·51–0·71)1·9 (0·9–3·9)0·0619360 (31)0·9 (0·5–1·5)0·743
 >6524170 (29)0·74 (0·66–0·80)0·48 (0·37–0·57)2·5 (1·2–5·0)0·0124152 (21)0·7 (0·4–1·2)0·216
Gender
 Female19645 (23)0·82 (0·74–0·87)0·62 (0·51–0·71)19649 (25)
 Male29587 (30)0·73 (0·66–0·78)0·54 (0·45–0·63)1·5 (1·0–2·2)0·02029583 (28)1·3 (0·9–1·8)0·165
Haemoglobin (g/l)
 ≥10021348 (22)0·84 (0·77–0·89)0·63 (0·52–0·72)21352 (24)
 <10027884 (30)0·70 (0·63–0·76)0·53 (0·44–0·61)1·5 (1·1–2·2)0·01227880 (28)1·3 (0·9–1·9)0·06
ANC (×109/l)
 ≥1·823948 (20)0·83 (0·77–0·88)0·66 (0·56–0·74)23949 (20)
 <1·825284 (33)0·70 (0·62–0·76)0·49 (0·39–0·58)2·0 (1·4–2·9)<0·000525283 (33)1·9 (1·3–2·7)<0·0005
Platelet count (×109/l)
 ≥10026363 (24)0·83 (0·77–0·88)0·67 (0·58–0·74)26364 (24)
 <10022869 (30)0·67 (0·59–0·74)0·45 (0·34–0·55)1·9 (1·3–2·7)<0·000522868 (30)1·7 (1·2–2·4)0·001
Cytopenias     <0·0005   <0·0005
 0607 (11)0·95 (0·82–0·98)0·80 (0·56–0·92)609 (15)
 120447 (23)0·85 (0·78–0·90)0·66 (0·56–0·74)2·0 (0·9–4·4)0·08520451 (25)1·7 (0·8–3·5)0·118
 215554 (35)0·66 (0·56–0·74)0·43 (0·31–0·55)4·2 (1·9–9·4)<0·00515548 (31)2·9 (1·4–5·9)0·003
 37224 (33)0·53 (0·35–0·67)0·39 (0·20–0·58)6·1 (2·6–14·1)<0·0057224 (33)4·3 (2·0–9·4)<0·005
Blast cell (%)     <0·0005   <0·0005
 ≤528739 (13)0·91 (0·87–0·94)0·80 (0·72–0·86)28744 (15)
 6–106227 (44)0·65 (0·48–0·78)0·24 (0·09–0·42)5·1 (3·1–8·4)<0·0056215 (24)2·3 (1·2–4·1)0·006
 11–209340 (43)0·54 (0·40–0·67)0·27 (0·13–0·43)6·3 (4·0–10·0)<0·0059346 (49)5·9 (3·8–9·0)<0·005
 21–304926 (53)0·32 (0·16–0·49)13·0 (7·7–22·0)<0·0054927 (55)9·5 (5·8–15·8)<0·005
FAB     <0·0005   <0·0005
 RARS797 (8)0·95 (0·84–0·98)0·86 (0·71–0·94)793 (4)
 RA20932 (15)0·90 (0·84–0·94)0·78 (0·68–0·85)2·3 (1·0–5·2)0·04520942 (20)7·0 (2·2–22·7)0·001
 RAEB15265 (42)0·59 (0·48–0·68)0·24 (0·14–0·36)10·8 (4·9–24·0)<0·00515260 (39)21·0 (6·5–67·4)<0·005
 RAEB-t5128 (55)0·33 (0·18–0·50)0·09 (0·01–0·31)23·1 (9·9–54·0)<0·0055127 (53)42·0 (12·6–139·8)<0·005
WHO     <0·0005   <0·0005
 RA784 (5)0·96 (0·87–0·99)0·91 (0·70–0·97)7813 (16)
 RARS490NRNR491 (2)0·1 (0·0–0·8)0·028
 RCMDS165 (31)0·86 (0·56–0·96)0·58 (0·24–0·80)4·2 (1·1–16·0)0·031162 (12)0·5 (0·1–2·3)0·378
 RCMD10523 (22)0·84 (0·73–0·90)0·65 (0·49–0·76)4·6 (1·6–13·4)0·00510520 (19)1·3 (0·6–2·7)0·422
 MDS del(5q)295 (17)0·91 (0·71–0·98)0·86 (0·62–0·95)2·1 (0·5–7·8)0·267296 (20)0·9 (0·3–2·4)0·832
 MDS-U70NRNR72 (28)2·7 (0·6–12·3)0·181
 RAEB-16227 (43)0·65 (0·48–0·78)0·24 (0·09–0·42)12·7 (4·4–36·4)<0·0056215 (24)2·1 (0·9–4·4)0·05
 RAEB-29038 (42)0·54 (0·39–0·66)0·24 (0·11–0·41)15·8 (5·6–44·4)<0·0059045 (50)5·5 (2·9–10·2)<0·005
IPSS     <0·0005   <0·0005
 Low11512 (10)0·96 (0·90–0·99)0·88 (0·76–0·94)11511 (9)
 Int-117934 (19)0·88 (0·82–0·93)0·70 (0·59–0·79)2·3 (1·2–4·5)0·01217931 (17)2·2 (1·1–4·5)0·02
 Int-211044 (40)0·58 (0·45–0·69)0·27 (0·14–0·42)8·3 (4·3–16·0)<0·00511044 (40)8·4 (4·3–16·6)<0·005
 High8742 (48)0·37 (0·24–0·51)0·08 (0·01–0·28)16·3 (8·4–31·8)<0·0058746 (53)15·9 (8·1–31·5)<0·005
IPSS cytog. cat.     <0·0005   <0·0005
 Good26049 (18)0·91 (0·86–0·94)0·74 (0·65–0·81)26043 (16)
 Intermediate15748 (30)0·67 (0·57–0·75)0·46 (0·32–0·58)2·4 (1·6–3·6)<0·00515755 (35)2·8 (1·9–4·3)<0·005
 Poor7435 (47)0·38 (0·24–0·53)0·15 (0·04–0·31)6·7 (4·3–10·5)<0·0057434 (46)6·4 (4·0–10·1)<0·005
Rev. IPSS cyto. cat.     <0·0005   <0·0005
 Good28252 (18)0·90 (0·86–0·94)0·73 (0·64–0·80)26740 (15)
 Intermediate14245 (31)0·69 (0·58–0·77)0·44 (0·31–0·57)2·5 (1·7–3·8)<0·00512844 (34)2·9 (1·9–4·5)<0·005
 Poor6735 (52)0·26 (0·13–0·41)0·11 (0·02–0·27)8·3 (5·3–13·0)<0·0059648 (50)6·5 (4·2–10·0)<0·005
Cytogenetic pattern     <0·0005   <0·0005
 Normal19737 (19)0·92 (0·86–0·95)0·72 (0·61–0·80)19726 (13)
 Abnormal29495 (32)0·66 (0·59–0·72)0·48 (0·39–0·56)2·2 (1·5–3·2)<0·0005294106 (36)3·3 (2·2–5·2)<0·005
 Single defect19258 (30)0·72 (0·63–0·79)0·50 (0·39–0·61)1·8 (1·2–2·8)0·00319270 (36)3·1 (2·0–4·9)<0·005
 +1q61 (16)0·66 (0·05–0·94)NA1·4 (0·2–10·9)0·69762 (33)3·9 (0·9–16·8)0·062
 3q abnormalities139 (69)0·28 (0·06–0·54)NA6·7 (3·1–14·0)0·002135 (38)4·4 (1·7–11·7)0·002
 del(5q)438 (18)0·88 (0·71–0·95)0·84 (0·64–0·93)0·8 (0·4–1·7)0·5944310 (23)1·6 (0·7–3·3)0·22
 −7135 (38)0·69 (0·31–0·89)0·23 (0·01–0·62)3·4 (1·3–8·7)0·010138 (61)8·1 (3·6–18·1)<0·005
 del(7)(q31q35)209 (45)0·62 (0·30–0·82)0·15 (0·01–0·48)3·9 (1·8–8·2)<0·0052011 (55)6·2 (3·1–12·8)<0·005
 +8225 (23)0·86 (0·53–0·96)NA1·7 (0·7–4·5)0·2332214 (63)6·3 (3·2–12·2)<0·005
 del(11)(q14q23)51 (20)NA0·50 (0·01–0·91)1·0 (0·1–7·7)0·95151 (20)1·6 (0·2–11·8)0·645
 del(12p)172 (11)0·79 (0·37–0·95)0·79 (0·37–0·95)0·9 (0·2–3·8)0·919172 (11)1·1 (0·2–4·9)0·834
 del(17p)62 (33)0·75 (0·12–0·96)NA4·1 (0·9–17·4)0·05263 (50)6·8 (1·8–20·5)0·003
 del(20q)154 (26)0·85 (0·54–0·96)0·61 (0·23–0·84)1·2 (0·4–3·4)0·717156 (40)2·6 (1·1–6·5)0·030
 –Y50 (0)NANANA51 (20)0·6 (0·1–4·9)0·628
 5q- plus AA.198 (42)0·48 (0·19–0·73)0·48 (0·19–0·73)3·4 (1·6–7·3)0·002195 (26)2·8 (1·0–7·3)0·035
 Miscellaneous6116 (26)0·74 (0·59–0·85)0·58 (0·38–0·73)1·7 (0·9–3·0)0·0776121 (34)3·0 (1·7–5·4)<0·005
 Complex4121 (51)0·09 (0·01–0·31)0·09 (0·01–0·31)15·4 (8·7–27·4)<0·0054115 (36)9·3 (4·8–18·0)<0·005
Figure 1.

 Overall survival and progression-free interval of myelodysplastic syndromes patients according to standard International Prognostic Scoring System cytogenetic categories.

Figure 2.

 Overall survival and progression-free interval of myelodysplastic syndromes patients according to revised International Prognostic Scoring System cytogenetic categories.

In a subsequent step, the OS and the risk of MDS/AML progression for each single defect were compared with those of the remaining set of cases with single defects. This analysis showed that 5q- was associated with a significantly longer OS (P = 0·001) and PFI (P = 0·002); in contrast 3q abnormalities were associated with a significantly shorter OS (P = 0·002) (Table IV).

Table IV.   Impact of specific chromosomal defects on survival and MDS/AML progression.
 SurvivalMDS/AML progression
Total ptsEvents (%)2-year survival (95% CI)5-year survival (95% CI) HR (95% CI)P-valueTotal ptsEvents (%) HR (95% CI)P-value
  1. *A specific single defect versus the remaining set of single defects. Considering survival the P-values for +1q, −7, del(7)(q31q35), +8, del(11)(q14q23), 12p-, 17p- and 20q- were 0·75, 0·22, 0·06, 0·75, 0·54, 0·22, 0·38 and 0·26, respectively; considering PFI the P-values for +1q, −7, del(7)(q31q35), +8, del(11)(q14q23), 12p-, 17p- and 20q- were 0·84, 0·02, 0·05, 0·02, 0·42, 0·06, 0·34 and 0·57, respectively; a specific defect versus the remaining set of chromosomal abnormalities included within the same IPSS cytogenetic category. Considering survival, the P-values for 5q- plus one additional defect and 12p- were 0·21, and 0·17, respectively; considering PFI, the P-values for 5q- plus one additional defect and 12p- were 0·57 and 0·10, respectively; NA, not available.

3q abnormalities*139 (69)0·28 (0·06–0·54)NA135 (38)
Single defects17949 (27)0·76 (0·67–0·83)0·55 (0·43–0·65)3·6 (1·7–7·4)0·00217965 (36)0·4 (0·2–0·7)0·53
5q-438 (18)0·88 (0·71–0·95)0·83 (0·64–0·93)4310 (23)
Single defects14950 (33)0·66 (0·56–0·75)0·38 (0·25–0·51)0·3 (0·1–0·7)0·00114960 (40)0·4 (0·2–0·7)0·002
20q-154 (26)0·85 (0·54–0·96)0·61 (0·23–0·84)156 (40)
Low24545 (18)0·91 (0·86–0·94)0·75 (0·66–0·82)1·3 (0·4–3·8)0·5524537 (15)2·4 (1.0–5.8)0·04
Intermediate14439 (27)0·72 (0·61–0·80)0·51 (0·37–0·64)14450 (34)
3q abn.139 (69)0·28 (0·06–0·54)NA0·3 (0·1–0·6)0·002135 (38)0·7 (0·3–1·9)0·58
Intermediate13543 (31)0·63 (0·52–0·72)0·47 (0·33–0·60)13541 (30)
+8225 (22)0·86 (0·53–0·96)NA1·3 (0·5–3·5)0·482214 (63)0·4 (0·2–0·8)0·01
Poor5426 (48)0·29 (0·14–0·47)0·14 (0·03–0·34)2·3 (1·0–4·9)5423 (42)
Del(7)(q31q35)209 (45)0·61 (0·30–0·82)0·15 (0·01–0·48)0·032011 (55)1·4 (0·6–2·9)0·34

Yet another step consisted of the comparison of the OS and the risk of MDS/AML progression for each single defect with those of the remaining set of abnormalities included within the same IPSS cytogenetic category (Table IV). This analysis showed that 3q abnormalities were associated with an OS that was statistically shorter than that of the IPSS intermediate cytogenetic category (P = 0·002) and similar to that of the poor cytogenetic category (P = 0·96), del(7)(q31q35) with an OS statistically better than that of the poor cytogenetic category (P = 0·03) and similar to that of the intermediate cytogenetic category (P = 0·17) and del(12p) with an OS similar to that of the good and intermediate cytogenetic categories (P = 0·95 and P = 0·17, respectively). Considering PFI, del(20q) was correlated with a risk of MDS/AML progression higher than that of the good cytogenetic category (P = 0·04); +8 with a PFI significantly shorter than that of the intermediate cytogenetic category (P = 0·01) and similar to that of the poor cytogenetic category (P = 0·51) and del(12p) with a PFI similar to that of the good and intermediate cytogenetic categories (P = 0·95 and P = 0·10).

Based on these results, IPSS cytogenetic categories were revised. For OS, del(11)(q14q23) and del(12p) were included within the good risk category, del(7)(q31q35) within the intermediate risk category and 3q rearrangements within the high-risk category; for the risk of MDS/AML progression, del(11)(q14q23) and del(12p) were included within the good risk category, del(20q) within the intermediate risk category and +8 within the high-risk category. These revised IPSS cytogenetic categories had a clear impact on OS and PFI (Table III). The relationships between OS and PFI to the revised IPSS cytogenetic categories are plotted in Fig 2.

Multivariate analysis

Five multivariate Cox models were developed and compared by means of the AIC (Table V). The five models were all statistically significant in determining both OS (P < 0·0005) and PFI (P < 0·0005) individually. According to the AIC, the model that included the WHO classification and the revised IPSS cytogenetic categories was the best (AIC = 942) for evaluating OS, although not much better than the model that included the WHO classification and the standard IPSS cytogenetic categories (AIC = 947), but definitely better than the model that included IPSS score and age (AIC = 1286). However, the model that included the WHO classification and the revised IPSS cytogenetic categories was definitely the best for predicting PFI (AIC = 1020), when compared with either the model that included the WHO classification and standard IPSS cytogenetic categories (AIC = 1029) or with the model that including IPSS score and age (AIC = 1339).

Table V.   Biological and clinical parameters effecting OS and MDS/AML evolution in the five multivariate models.
First modelSecond modelThird modelFourth modelFifth model
VariablesCat.SurvivalVariablesCat.SurvivalVariablesCat.SurvivalVariablesCat.SurvivalVariablesCat.Survival
HR (95% CI)P-valueHR (95% CI)P-valueHR (95% CI)P-valueHR (95% CI)P-valueHR (95% CI)P-value
  1. HR, hazard ratio; 95% CI, 95% confidence intervals; AIC, Akaike Information Criterion; Cyto cat, cytogenetic category.

  2. Revised IPSS cytogenetic categories: the abnormalities included within each revised IPSS cytogenetic subgroup are provided in the text.

Age (years)<65Age<65Age<65Age<65Age<65
≥651·6 (1·1–2·3)0·007≥651·6 (1·1–2·3)0·008≥651·5 (1·0–2·1)0·024≥651·6 (1·1–2·4)0·015≥651·6 (1·1–2·4)0·015
Cytopenias  0·0007Cytopenias  0·0002IPSS  <0·0005WHO class.  <0·0005WHO class.  <0·0005
00LowRAAR
12·3 (1·0–5·2)0·03611·9 (0·8–4·4)Int-12·3 (1·2–4·5)0·011RARSARS
23·5 (1·6–7·7)0·00223·0 (1·4–6·7)0·006Int-28·2 (4·2–15·7)<0·005RCMD3·6 (1·2–10·9)0·025RCMD3·3 (1·1–10·2)0·035
34·8 (2·1–11·1)<0·00534·9 (2·1–11·6)<0·005High16·7 (8·6–32·6)<0·005RCMDS3·1 (0·8–12·0)0·095RCMDS2·9 (0.7–11.1)0·120
         5q- syn.2·8 (0·7–11·0)0·1235q- syn.2·6 (0·7–10·1)0·150
         MDS-UMDS-U
         RAEB-110·8 (3·6–31·7)<0·005AREB-19·6 (3·3–28·3)<0·005
         RAEB-212·8 (4·4–37·5)<0·005AREB-211·7 (4·0–34·4)<0·005
Blast cell percentage  <0·0005Blast cell percent.  <0·0005    Cytopenias  0·348Cytopenias  0·413
<5<5    00
5–104·4 (2·6–7·5)<0·0055–104·4 (2·6–7·3)<0·005    12·1 (0·8–5·1)0·09711·8 (0·7–4·4)0·173
11–205·0 (3·1–8·0)<0·00511–204·8 (3·0–7·7)<0·005    22·1 (0·9–5·0)0·09621·8 (0·7–4·4)0·165
21–309·1 (5·3–15·7)<0·00521–307·4 (4·3–12·9)<0·005    32·3 (0·8–6·2)0·08632·3 (0·8–6·2)0·092
IPSS cyto. cat.  <0·0005Rev. IPSS cyto. cat.  <0·0005    IPSS cyto. cat.  <0·0005Rev. IPSS cyto. cat.  <0·0005
GoodGood    GoodGood
Inter.1.9 (1.2–2.9)0·002Inter.1·9 (1·2–2·9)0·003    Interm.1·9 (1·2–3·0)0·007Interm.1·9 (1·2–3·1)0·006
Poor4.4 (2.7–7.1)<0·005Poor4·8 (2·9–7·9)<0·005    Poor3·7 (2·1–6·6)<0·005Poor4·7 (2·7–8·3)<0·005
AIC1239AIC1236AIC1286AIC947AIC942
 MDS/AML evolution MDS/AML evolution MDS/AML evolution MDS/AML evolution MDS/AML evolution
  HR (95% CI)P-value  HR (95% CI)P-value  HR (95% CI)P-value  HR (95% CI)P-value  HR (95% CI)P-value
Cytopenias  0·031Cytopenias  0·022IPSS  <0·0005WHO class.  <0·0005WHO class.  <0·0005
00LowRARA
12·2 (1·1–4·6)0·0312·2 (1·0–4·5)0·033Int-12·2 (1·1–4·5)0·022RARS0·1 (0·01–0·9)0·046RARS0·1 (0·01–0·9)0·037
22·5 (1·2–5·1)0·0122·5 (1·2–5·1)0·013Int-28·4 (4·3–16·6)<0·005RCMD1·0 (0·4–2·2)0·965RCMD1·0 (0·4–2·2)0·985
33·2 (1·4–7·0)0·00333·4 (1·5–7·4)0·002High15·9 (8·0–31·4)<0·005RCMDS0·3 (0·07–1·5)0·164RCMDS0·3 (0·07–1·5)0·156
         5q- syn.1·3 (0·4–3·6)0·6025q- syn.1·4 (0·5–3·8)0·529
         MDS-U2·7 (0·6–12·1)0·199MDS-U3·1 (0·7–14·4)0·135
         RAEB-11·7 (0·8–3·9)0·152RAEB-11·8 (0·6–1·3)0·152
         RAEB-23·9 (1·9–7·8)<0·005RAEB-23·4 (1·7–6·8)0·001
Blast cell percentage  <0·0005Blast cell percent.  <0·0005    Cytopenias  0·192Cytopenias  0·134
<5<5    00
5–101·9 (1·0–3·5)0·035–101·9 (1·0–3·6)0·03    11·7 (0·8–3·8)0·15311·8 (0·8–4·0)0·113
11–204·4 (2·8–6·9)<0·00511–204·1 (2·6–6·4)<0·005    21·6 (0·7–3·6)0·23021·7 (0·7–3·7)0·190
21–306·3 (3·7–10·7)<0·00521–306·1 (3·6–10·4)<0·005    32·5 (1·0–5·9)0·03932·7 (1·1–6·3)0·025
IPSS cyto. cat.  <0·0005Rev. IPSS cyto. cat.  <0·0005    IPSS cyto. cat.  <0·0005Rev. IPSS cyto. cat.  <0·0005
GoodGood    GoodGood
Inter.2·1 (1·4–3·2)0·001Inter.2·0 (1·3–3·3)0·002    Interm.2·6 (1·6–4·2)<0·005Interm.3·0 (1·7–5·1)<0·005
Poor4·0 (2·4–6·0)<0·005Poor4·3 (2·7–6·9)<0·005    Poor4·1 (2·3–7·5)<0·005Poor5·1 (3·0–8·9)<0·005
AIC1324AIC1316AIC1339AIC1029AIC1020

Discussion

To the best of our knowledge, this is one of the few studies that has examined the incidence of chromosomal defects in relation to both the FAB and WHO classifications of MDS. Clonal chromosomal abnormalities were found in 59·9% of the 491 consecutive de novo MDS patients classified according to FAB criteria and in 57·7% of the 436 patients re-classified according to WHO criteria. These frequencies are similar to those reported in other studies (Pierre et al, 1989; Morel et al, 1993; Toyama et al, 1993; Bernasconi et al, 1994, 2005; Parlier et al, 1995; Soléet al, 2000; Soléet al, 2005). Considering low-risk MDS, the different incidences and types of chromosomal defects observed in our series when patients were classified according to FAB or WHO classification are in line with previous studies that underlined the morphological and cytogenetic differences between RARS/RA and RCMDS/RCMD (Rosati et al, 1996; Germing et al, 2000a,b; Malcovati et al, 2005) and have stressed the biological similarity between RCMDS/RCMD and RAEB. Recently, another report has proposed that uni-lineage dysplasia is the first step in the pathogenesis of MDS (Čermàk et al, 2005), an event that might have happened in three of our RARS/RA patients [2 with a del(7)(q31q34) and one with a complex karyotype], who progressed from uni- to multi-lineage dysplasia 1 month after diagnosis. Therefore, given that various reports (Ramos et al, 1999; Bowen et al, 2003) have emphasised the very low reproducibility of the morphological criteria used to define dysplasia and since –7, del(7)(q31q35), 3q abnormalities and complex karyotypes either occur in most patients with multi-lineage dysplasia or in those who rapidly progress from RARS/RA to RCMDS/RCMD, we suggest that these defects are, per se, sufficient to identify tri-lineage dysplasia. In addition, our study showed that the blast cell percentage threshold of 10%, used by the WHO classification to distinguish between RAEB-1 and RAEB-2, was effective in identifying RAEB patients with a higher incidence of chromosomal defects.

In our study, the karyotypic pattern, along with various clinical factors, was significantly associated with outcome in the univariate analysis (Table III) (Figs 1 and 2). Two approaches were followed to establish the prognostic power of single abnormalities. Firstly, the OS and PFI of each single defect were compared with those of the remaining set of abnormal cases with a single defect. In this analysis, 5q- was associated with a significantly longer OS and PFI (P = 0·001 and P = 0·002), whereas 3q rearrangements were associated with a significantly shorter OS (P = 0·002) (Table IV). Subsequently, the OS and PFI of each single most common karyotype defect were compared with those of the respective standard IPSS cytogenetic category to segregate the defect analysed from its own IPSS category. By applying a similar approach, Soléet al (2005) identified 1q rearrangements, del(11)(q14q23), del(12p) and del(17p) as the defects that could be segregated from the IPSS intermediate cytogenetic category. In our series, del(11)(q14q23) and del(12p) were also associated with a statistically better clinical outcome than that of the IPSS intermediate cytogenetic category (Tables III and IV). Del(11)(q14q23) is a very rare defect with a still controversial prognostic significance (Panani et al, 2004). Instead, and in line with other reports (Soléet al, 2000; Soléet al, 2005), our study did not determine the good clinical outcome of 12p- by its association with a less advanced FAB subtype, as 10 of our patients were classified as RARS/RA and seven as RAEB/RAEB-t. Moreover, progression to AML occurred in only two patients, 1 and 2 years after diagnosis.

Our data also showed that 3q rearrangements and del(7)(q31q35) could be segregated from the intermediate and poor IPSS cytogenetic groups for OS. Chromosomal 3q rearrangements, which are considered of poor prognostic significance in AML (Byrd et al, 2002) and predictive of AML progression in MDS (Soléet al, 2000), showed an OS similar to that of the IPSS poor cytogenetic category (P = 0·96), but a PFI similar to that of the intermediate cytogenetic category (P = 0·589). This data might be affected by the advanced patient age. In fact, nine subjects were >63 years old and died of complications related to marrow deficiency, whereas two of the four patients <53 years old progressed to MDS/AML. The inferior survival of del(7)(q31q35) has already been demonstrated in a previous study that, however, found no improvement in the prognostic power of the IPSS cytogenetic grouping when del(7q) was entered as a new entity (Bernasconi et al, 2005).

In our series, trisomy 8 was associated with a survival that was similar to that of normal karyotypes (P = 0·23), but with a significantly shorter PFI (P = 0·000) (Table III), and similar results were obtained when it was compared with the IPSS intermediate cytogenetic category (P = 0·48 for OS and P = 0·01 for PFI) (Table IV). However, in estimating +8 prognostic significance, one should consider that it occurs as a constitutional trisomy in about 15–20% of MDS/AML patients (Maserati et al, 2002), is the most common secondary abnormality (Paulsson et al, 2001), is often present in a clonal cell population of variable size, is associated with a heterogeneous gene expression pattern (Schoch et al, 2005) and cryptic intra-chromosomal imbalances (Paulsson et al, 2006). Therefore, the poor prognostic significance of trisomy 8 might be determined by the presence of these cryptic defects and not by the presence of +8 itself.

In addition, and in line with previous observations (Giagounidis et al, 2004; Bernasconi et al, 2005), our results confirmed that the clinical outcome of 5q- patients was not determined by the different breakpoints of the interstitial deletion, but instead by marrow blast cell percentage, which had a significant influence on OS (P = 0·009) and showed a trend towards significance on PFI (P = 0·09). In our series, the effect of this parameter on OS and PFI was evident for blast cell percentages ≥10% (P = 0·02) and ≥21% (P = 0·007). Therefore, only 5q- patients with ≤5% marrow blasts must be diagnosed as ‘5q- syndromes’.

Based on the above results, we developed a revised IPSS cytogenetics grouping, the statistical significance of which was confirmed by uni- and multi-variate analysis. These new IPSS cytogenetic subgroups, along with FAB/WHO classification, standard IPSS cytogenetic categories and IPSS scores, which are well-established prognostic indicators for OS and risk of MDS/AML evolution (Pfeilstöcker et al, 1999; Soléet al, 2000; Sperr et al, 2001; Bernasconi et al, 2005; Germing et al, 2005), were included in five multivariate models. Revised IPSS cytogenetic categories were as effective as standard IPSS categories for predicting OS (AIC = 1236 vs. 1239), but even more effective for predicting the risk of MDS/AML evolution (AIC = 1316 vs. 1324) (Table V). However, the novel finding of the present series is the demonstration that the WHO classification is much more effective than the IPSS in predicting MDS clinical outcome in both univariate and multivariate analyses. In fact, our two multivariate models that included the WHO classification and either revised or standard IPSS cytogenetic categories were the most accurate when compared with the others (Table V). Moreover, when considering PFI, the model that included revised IPSS cytogenetic categories was even better than the model including standard categories (AIC = 1020 vs. 1029).

In conclusion, (i) the WHO classification was associated with a more homogeneous cytogenetic pattern than the FAB classification, (ii) WHO and either standard or revised IPSS cytogenetic categories were much more effective than IPSS for predicting MDS clinical outcome, (iii) revised IPSS cytogenetic categories were more effective than standard categories for predicting PFI.

Ancillary