Validation of the new comprehensive cytogenetic scoring system (NCCSS) on 630 consecutive de novo MDS patients from a single institution

Authors


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

Abstract

This study evaluated whether the NCCSS truly improves the prognostic stratification of 630 consecutive de novo MDS patients and established which cytogenetic grouping [NCCSS or International Prognostic Scoring System (IPSS)], when combined with the WHO classification, best predicted the clinical outcome of myelodysplastic syndromes (MDS). The frequency of chromosomal defects was 53.8%. Clinical parameters, including number of cytopenias, WHO classification, IPSS cytogenetic categories and scores, NCCSS were all relevant for overall survival (OS) and leukemia-free survival (LFS) and were included in six distinct multivariate models compared by the Akaike Information Criterion (AIC). The most effective model to predict OS included the number of cytopenias, the WHO classification and the NCCSS, whereas the model including the number of cytopenias, blast cell percentage and the NCCSS and the model including the number of cytopenias the WHO classification and the NCCSS were almost equally effective to predict LFS. In conclusion, the NCCS (i) improves the prognostic stratification of the good and poor IPSS cytogenetic categories by introducing the very good and the very poor categories; (ii) is still incomplete in establishing the prognostic relevance of rare/double defects, (ii) applied to patients who receive supportive treatment only identifies five different prognostic subgroups, but applied to patients treated with specific therapies reveals only a trend toward a significantly different OS and LFS when patients of the poor and intermediate cytogenetic categories are compared, (iii) combined with the WHO classification is much more effective than the IPSS in predicting MDS clinical outcome. Am. J. Hematol. 88:120–129, 2013. © 2012 Wiley Periodicals, Inc.

Introduction

Myelodysplastic syndromes (MDS) are a heterogeneous group of clonal stem cell disorders marked by a hypercellular marrow exhibiting dysplasia in one or more myeloid cell lines and an increased level of apoptosis which result in peripheral blood cytopenias [1]. These disorders show substantial clinical variability, ranging from stability for 10 or more years to death due to cytopenias or evolution into acute myeloid leukaemia (AML) within a few months [2]. To more precisely identify patients with markedly different clinical outcomes, over the past 15 years various prognostic scoring systems have been developed and applied [3-16]. The International Prognostic Scoring System (IPSS) [8] was that most commonly used and has been now revised [17-19]. Thus, cytogenetic abnormalities remain one of the major determinants of MDS pathogenesis, diagnosis, prognosis and guide any potential treatment decisions [12-25], however more sensitive techniques such as array Comparative Genomic Hybridization, Single Nucleotide Polymorphism arrays (SNP-a) and Next-Generation Sequencing are still used in the research setting only [26-31]. The IPSS precisely defines the prognostic value of the most frequent chromosomal defects, whereas it does not define rare or combined abnormalities, which are included within the intermediate category [8]. Thus, these latter defects, which flag the extreme cytogenetic heterogeneity of MDS, should be included within a more precise cytogenetic scoring system [21]. In the last 10 years, large patient series have provided some relevant information on rare isolated abnormalities [9-13], but very few have provided information on combined defects [10, 13]. In addition, it has become evident that the clinical outcome of patients with a complex karyotype, previously defined by the presence of ≥3 defects, worsens with an increase in the number of chromosomal defects [13, 20, 21]. These limitations have spurred the revision of the IPSS cytogenetic categories and have led to the development of the NCCSS [17-19].

The principal goals of the present study were to confirm the effectiveness of the recently developed NCCSS for predicting overall survival (OS) and leukemia-free survival (LFS) in a large series of consecutive de novo MDS patients from a single Institution and to compare different multivariate prognostic models in order to establish which model truly improves patient prognostic stratification.

Materials and Methods

Diagnosis

All of the 630 consecutive de novo MDS patients, including 491 analyzed in a previous report [12], were diagnosed at the Division of Hematology, Fondazione IRCCS Policlinico San Matteo, Pavia between January 1990 and December 2010. MDS diagnosis was made according to the WHO criteria [1]. None of the 630 patients had a white blood cell (WBC) count above 12 × 109/L. Patients with a diagnosis of chronic myelomonocytic leukaemia, those previously treated with chemo-radiotherapy for other cancers or non-malignant disorders (e.g., autoimmune diseases) and those previously exposed to environmental carcinogens were excluded from the study. Diagnostic procedures at the onset of disease and clinical monitoring during the follow-up were carried out as already described [12]. Patients were seen at our outpatient clinic every 3 months, unless there was a change in their clinical condition, that is, a MDS/AML progression. As the IPSS was developed for patients undergoing treatments that do not affect the MDS natural course, the only intervention allowed in this patients cohort was supportive care (transfusions, growth factors). Patients submitted to treatments (hypomethylating agents, lenalidomide, intensive chemotherapy, and allogeneic hematopietic stem cell transplantation [allo-HSCT]), which were commenced according to the guidelines of the Italian Society of Hematology [32] and may significantly affect MDS natural course, were evaluated with and without censoring for specific treatments.

Cytogenetic studies

Chromosome studies were performed on bone marrow cells at diagnosis, using a trypsin-Giemsa banding technique. Metaphase cells were obtained from short-term non-stimulated cultures. Whenever possible at least 20 metaphases were analyzed and 10 were fully karyotyped. Chromosome identification and karyotype description were made according to the International System for Chromosome Nomenclature [33]. Complex karyotypes, previously defined by the presence of at least three abnormalities were subdivided into different categories depending on the number of chromosome defects.

Statistical analysis

Statistical analyzes were carried out as previously described [12]. According to the IPSS [8] continuous variables were dichotomized according to values reported in the literature. 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. In this case, death censors the true but unknown observation time of an individual, and censoring is an end-point [34]. Therefore, in the present analysis censoring occurred in case of: end of study, death, start of lenalidomide, hypo-methylating agents, allo-HSCT, depending on which event occurred first. However, in a second step OS and LFS were analyzed without censoring for specific treatments. Survival was estimated by using the Kaplan–Meier method [35]. LFS was defined as the time elapsed between MDS diagnosis and AML evolution. 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.

Clinically relevant variables, those with P-values ≤0.05 in univariate analyzes, were included in a Cox multivariate model. Competing models were informally compared by the Akaike Information Criterion (AIC) and STATA 9 (StataCorp, College Station, TX) was used for computation. A two-sided P-value ≤0.05 was considered statistically significant.

Results

Patients

The clinical and haematological characteristics of our patient cohort are reported in Table 1. When censoring was performed median follow-up was 20.9 months (Inter-quartile range, IQR = 8.3–47.7); 444 patients were alive and 186 had died. Median survival time was 32.2 months (IQR = 11.5–72.6). The death rate was 10.5% per 100 person years (95% Confidence Intervals, 95% CI = 9.1–12.1). The 2- and 5-year OS were 0.77% (95% CI = 0.73–0.81) and 0.59% (95% CI = 0.53–0.64), respectively. Median time to death was 20.1 months (IQR = 9.9–49.9). A total of 162 patients experienced clinical progression, 50 progressed into a more advanced MDS and 112 into AML. The AML evolution rate was 6.8% per one hundred person years (95% CI = 5.7–8.2). The 2- and 5-year LFS were 0.81% (95% CI = 0.77–0.84) and 0.74% (95% CI = 0.69–0.78). Median time to AML was 8.2 months (IQR = 4.3–20.0).

Table 1. Clinico-Hematological Findings in Our Patient Cohort
 Total (n = 630)
CharacteristicsNo%
  1. FAB classification, French-American-British classification; WHO classification, World Health Organization classification; RARS, RA with ringed sideroblasts; RA, refractory anemia; RCMD, refractory cytopenia with multilineage dysplasia; RCMDS, RCMD with ringed sideroblasts; RAEB-1, RA with excess of blasts type 1; RAEB-2, RAEB type 2; MDS-U, unclassifiable MDS.

  2. a

    Including differentiation-inducing agents in 15, immunosuppressive agents in 12, hypomethylating agents in two, lenalidomide in two, low-dose chemotherapy aimed at reducing white blood cell count in 41.

  3. b

    With censoring.

Sex  
Male38160.5
Female24939.5
Age, years 
Median65.3
Interquartile range (IQR)56.6–72.5
FAB classification  
RA and RARS41134.8
RAEB219191
WHO classification 30.3
RA and RARS19127.5
RCMD and RCMDS17316.5
RAEB-1104 
RAEB-211518.2
5q- syndrome386.1
MDS unclassifiable91.4
Peripheral blood cytopenias  
None10015.9
One26241.6
Two18228.9
Three8613.6
Bone marrow blast cell percentage  
<541365.5
5–09815.6
11–2011918.9
IPSS score  
Low17728.1
Int-125640.6
Int-213721.8
High609.5
Therapies  
Supportive care45772.5
Intensive chemotherapies7011.1
Other therapiesa7211.5
Bone marrow transplantation314.9
Follow-upb 
Median months20.9
Interquartile range (IQR)8.3–47.7

When censoring for specific therapies was not performed, median follow-up was 22.9 months (IQR = 9.0–50.5); 397 patients were alive and 233 had died. The death rate was 12.2% per 100 person years (95% CI = 10.8–13.9). The 2- and 5-year OS were 0.74% (95% CI = 0.69–0.77) and 0.54% (95% CI = 0.49–0.59), respectively. Median time to death was 20.7 months (IQR = 10.1–43.2). A total of 207 patients experienced clinical progression, 67 progressed into a more advanced MDS and 140 into AML. The AML evolution rate was 7.9% per 100 person years (95% CI = 6.7–9.4). The 2- and 5-year LFS were 0.77% (95% CI = 0.73–0.81) and 0.70% (95% CI = 0.65–0.74). Median time to AML was 8.2 months (4.3–19.8).

Cytogenetic results

At diagnosis, all 630 patients analyzed had successful cytogenetic analyzes, which are absolutely required to exactly define their IPSS cytogenetic category, IPSS score and NCCSS category. An abnormal karyotype was revealed in 53% of patients. The most common chromosomal lesions, their relation to the FAB/WHO classification and IPSS score, as well as their grouping according to the IPSS cytogenetic categories and to the NCCSS are reported in Table 2.

Table 2. Chromosomal Characteristics of Our Patient Cohort
 Total (n = 630)
CharacteristicsNo%
  1. FAB classification, French-American-British classification; WHO classification, World Health Organization classification; RARS, RA with ringed sideroblasts; RA, refractory anemia; RCMD, refractory cytopenia with multilineage dysplasia; RCMDS, RCMD with ringed sideroblasts; RAEB-1, refractory anemia with excess of blasts type 1; RAEB-2, RAEB type 2; MDS-U, unclassifiable MDS; IPSS, International Prognostic Scoring System; NCCSS, New Comprehensive Cytogenetic Scoring System.

Karyotype  
Normal29146.2
Abnormal33953.8
Most common single abnormalities  
Der(3)(q21)/Der(3)(q26)101.6
Del(5q)558.7
Del(7q)233.6
−7142.2
Der(1;7)50.8
+8355.5
Del(11q)50.8
Del(12p)172.7
Del(20q)203.2
-Y81.3
Others single558.7
Double defects  
Del(5q) and an add. Defect172.7
Any other double defect253.9
−7/Del(7q) and an add. defect40.6
Complex (with 3 vs. ≥4 defects)46 (13 vs. 33)7.1 (2.0 vs. 5.2)
Chrom. abnormal patients/total in each FAB subtype:  
RARS and RA190/41146.2
RAEB149/21968.0
Chrom. abnormal patients/total in each WHO subtype:  
RARS and RA67/19135.0
RCMD and RCMDS83/17347.9
RAEB-166/10463.5
RAEB-283/11572.2
5q- syndrome38
U-MDS3/933.3
Chrom. abnormal patients/total in each IPSS subgroup:  
Low39/17722.0
Int-1137/25653.5
Int-2107/13778.1
High56/6093.3
IPSS cytogenetic categories:  
Good37459.4
Intermediate17427.6
Poor8213.0
NCCSS categories:  
Very good132.0
Good40564.3
Intermediate13821.9
Poor426.7
Very poor325.1

Prognostic factors

Univariate analysis

The clinical and biological parameters which had a significant impact on OS and LFS after censoring and without censoring for specific treatments are listed in Tables 3 and 4. As already reported [12], after censoring age and gender have a statistically significant influence on OS only (HR = 1.6 with P < 0.0001 and HR = 1.7 with P < 0.0001, respectively), whereas anemia (Hb < 10 g/dL), thrombocytopenia (Plts <100/µL) and neutropenia (<1.8 × 109/L) significantly affect OS (HR = 1.7 and P < 0.0002; HR = 1.9 and P < 0.0001; HR = 1.6 and P < 0.0005) and LFS (HR = 1.7 and P < 0.004; HR = 2.1 and P < 0.0001; HR = 2.4 and P < 0.0001). These variables were also strongly influenced by the bone marrow blast cell percentage.

Table 3. Biological and Clinical Parameters Influencing OS and LFS in Univariate Analysis When Censoring for Specific Therapies was Applied
 Overall survivalAML progression
CategoryTot.Events (%)At 2-year (95% CI)At 5-years (95% CI)HR (95% CI)PTot.Events (%)At 2-years (95% CI)HR (95% CI)P
  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; NA, not available. FAB classification, French-American-British classification; WHO classification, World Health Organization classification; RARS, RA with ringed sideroblasts; RA, refractory anemia; RCMD, refractory cytopenia with multilineage displasia; RCMDS, RCMD with ringed sideroblasts; RAEB-1, refractory anemia with excess of blasts type 1; RAEB-2, RAEB type 2; MDS-U, unclassifiable MDS; IPSS, International Prognostic Scoring System; NCCSS, New Comprehensive Cytogenetic Scoring System.

  2. a

    As no patient within the very good prognostic category progressed to AML, all the comparisons were made versus the intermediate risk category: the good risk category had a significant lower risk of AML progression in comparison to the intermediate risk category.

  3. b

    No patient with more than six defects.

  4. c

    Versus normal karyotypes.

Cytopenias     0.00001    0.00001
010018 (18.0)0.90 (0.80–0.95)0.77 (0.64–0.86)1006 (6.0)0.96 (0.87–0.98)
126266 (25.2)0.85 (0.80–0.90)0.67 (0.58–0.74)1.4 (0.8–2.4)NS26240 (15.3)0.86 (0.80–0.90)2.6 (1.1–6.1)0.03
218269 (37.9)0.68 (0.59–0.75)0.45 (0.35–0.54)2.8 (1.6–4.7)0.000118240 (21.9)0.74 (0.66–0.81)4.6 (1.9–10.9)0.0001
38633 (38.4)0.50 (0.36–0.64)0.35 (0.20–0.51)4.5 (2.5–8.1)0.00018626 (30.2)0.58 (0.43–0.70)9.0 (3.7–22.5)0.0001
Blast cell percentage     0.00001    0.00001
≤541389 (21.5)0.86 (0.82–0.89)0.74 (0.67–0.79)41329 (7.0)0.94 (0.91–0.96)
6–109841 (41.8)0.63 (0.49–0.74)0.25 (0.13–0.39)3.4 (2.3–6.0)0.00019826 (26.5)0.64 (0.50–0.74)6.0 (3.5–10.3)0.0001
11–2011955 (47.0)0.53 (0.40–0.63)0.18 (0.08–0.31)5.5 (3.9–7.8)0.000111957 (47.9)0.42 (0.31–0.53)14.5 (9.1–22.9)0.0001
WHO classification     0.00001    0.00001
RA12118 (14.9)0.92 (0.84–0.96)0.80 (0.66–0.89)1216 (6.6)0.94 (0.86–0.98)
RARS704 (5.7)0.98 (0.87–0.99)0.92 (0.78–0.97)0.03 (0.1–0.9)0.03702 (2.8)0.98 (0.89–0.99)0.3 (0.1–1.7)NS
RCMDS2511 (44.0)0.77 (0.53–0.90)0.60 (0.48–0.69)2.8 (1.3–5.9)0.007252 (8.0)0.10 (—)1.0 (0.2–4.9)NS
RCMD14846 (31.1)0.78 (0.69–0.84)0.59 (0.48–0.69)2.2 (1.3–3.9)0.00314813 (8.8)0.92 (0.85–0.96)1.1 (0.6–3.5)NS
5q- Syndrome388 (21.1)0.88 (0.72-0.95)0.80 (0.60–0.91)0.9 (0.4–2.2)NS384 (10.5)0.92 (0.85–0.96)1.2 (0.4–4.1)NS
MDS-u92 (22.2)0.73 (0.28–0.92)0.73 (0.28–0.92)1.9 (0.4–8.6)NS9
RAEB-110443 (41.3)0.63 (0.49–0.73)0.26 (0.14–0.40)4.7 (2.7–8.3)0.000110428 (26.9)0.64 (0.50–0.74)6.3 (2.8–13.8)0.0001
RAEB-211554 (46.9)0.52 (0.40–0.63)0.19 (0.08–0.32)7.7 (4.5–13.3)0.000111555 (47.8)0.42 (0.30–0.53)15.0 (7.1–31.8)0.0001
IPSS score:     0.00001    0.00001
Low-risk17723 (12.9)0.95 (0.89–0.97)0.83 (0.73–0.89)1776 (3.4)0.97 (0.92–0.99)
Intermed -125671 (27.7)0.82 (0.76–0.87)0.66 (0.58–0.73)2.3 (1.5-3.7)0.000125627 (10.5)0.91 (0.86–0.94)3.4 (1.4–8.3)0.0001
Intermed -213761 (44.5)0.56 (0.45–0.66)0.24 (0.14–0.36)6.9 (4.2–11.2)0.000113747 (34.3)0.58 (0.47–0.68)17.9 (7.6–42.0)0.0001
High-risk6031 (51.7)0.38 (0.22–0.54)0.10 (0.02–0.26)14.4 (8.3–25.1)0.00016032 (53.3)0.25 (0.13–0.40)45.5 (18.8–110.3)0.0001
IPSS cytogenetic categ.:     0.00001    0.00001
Good37483 (22.2)0.87 (0.83–0.91)0.70 (0.63–0.76)37437 (9.9)0.90 (0.86–0.93)
Intermediate17459 (33.9)0.73 (0.64–0.79)0.52 (0.42–0.62)1.7 (1.2–2.4)0.00117442 (24.1)0.74 (0.65–0.81)2.8 (1.8–4.3)0.0001
Poor8244 (53.6)0.40 (0.27–0.53)0.18 (0.07–0.32)5.0 (3.4–7.3)0.00018233 (40.2)0.50 (0.36–0.62)6.9 (4.3–11.2)0.0001
NCCSa     0.00001    0.00001
Very good132 (15.4)NR0.90 (0.47–0.98)13
Good40596 (23.7)0.85 (0.81–0.89)0.69 (0.62–0.74)3.7 (0.9–15.1)0.0440541 (10.1)0.90 (0.86–0.93)0.3 (0.2–0.4)0.0001
Intermediate13847 (34.1)0.73 (0.63–0.81)0.45 (0.33–0.57)6.8 (1.6–28.4)0.00813838 (27.5)0.70 (0.60–0.78)
Poor4220 (47.6)0.48 (0.29–0.65)0.25 (0.09–0.45)12.4 (2.9–53.5)0.0014219 (45.4)0.51 (0.33–0.66)2.1 (1.2–3.6)0.009
Very poor3221 (65.6)0.06 (0.04–0.24)NA61.3 (13.9–268.7)0.00013214 (43.7)NA5.0 (2.7–9.5)0.0001
Karyotype     0.0002    0.00001
Normal29162 (21.3)0.87 (0.82–0.91)0.69 (0.60–0.76)29123 (7.9)0.92 (0.87–0.95)
Abnormal339124 (36.6)0.69 (0.63–0.75)0.51 (0.43–0.57)1.8 (1.3–2.4)0.0000133989 (26.5)0.72 (0.66–0.77)3.5 (2.2–5.5)0.0001
Number of abnormalities     0.00001    0.00001
None29162 (21.3)0.87 (0.82–0.91)0.69 (0.60–0.76)29123 (7.9)0.92 (0.87–0.95)
One24777 (31.2)0.79 (0.73–0.85)0.57 (0.48–0.65)1.3 (0.9–1.8)NS24759 (23.8)0.77 (0.71–0.83)2.8 (1.7–4.6)0.0001
Two4620 (43.5)0.51 (0.33–0.67)0.47 (0.29–0.64)2.5 (1.5–4.2)0.00014610 (21.7)0.75 (0.57–0.86)3.5 (1.6–7.3)0.0001
Three136 (46.1)0.50 (0.15–0.78)0.25 (0.02–0.64)3.1 (1.3–7.3)0.008136 (46.1)0.52 (0.19–0.77)7.1 (2.9–17.4)0.0001
Four117 (63.6)0.19 (0.02–0.53)NA14.0 (6.3–31.2)0.0001113 (27.3)NA10.2 (3.0–34.3)0.0001
Five74 (57.1)0.27 (0.01–0.67)NA8.1 (2.9–22.7)0.000173 (42.8)NA13.1 (3.9–44.3)0.0001
Sixb1510 (66.8)NANA17.1 (8.4–34.5)0.0001158 (53.3)NA25.7 (11.2–59.0)0.0001
Single defects     0.00001    0.00001
3q defects106 (60.0)0.41 (0.11–0.70)0.28 (0.04–0.59)3.2 (1.4–7.4)0.007104 (40.0)0.60 (0.20–0.80)12.8 (6.9–23.6)0.0001
Del(5q)5514 (25.4)0.84 (0.69–0.92)0.71 (0.52–0.83)0.9 (0.4–2.1)NS558 (14.5)0.87 (0.73–0.94)2.0 (0.8–4.6)NS
Del(7q)2311 (47.3)0.64 (0.36–0.82)0.33 (0.09–0.59)2.8 (1.5–5.3)0.002235 (21.7)0.74 (0.43–0.89)3.0 (1.1–7.9)0.025
−7146 (42.8)0.58 (0.21–0.82)0.21 (0.01–0.58)3.3 (1.4–7.6)0.006148 (57.1)0.43 (0.16–0.68)11.2 (5.0–25.2)0.0001
Der(1;7)52 (40.0)NR0.66 (0.05–0.94)1.6 (0.4–6.7)NS51 (5.9)0.75 (0.13–0.96)2.1 (0.3–16.0)NS
+83513 (37.1)0.73 (0.52–0.86)0.37 (0.16–0.59)1.8 (1.0–3.4)0.04359 (25.7)0.77 (0.55–0.89)3.6 (1.6–7.8)0.0001
Del(11q)52 (40.0)NR0.75 (0.13–0.96)0.7 (0.2–3.0)NS5NANS
Del(12p)173 (17.6)0.82 (0.54–0.94)0.82 (0.54–0.94)0.7 (0.2–2.1)NS171 (5.9)NR0.6 (0.1–4.6)NS
Del(20q)207 (35.0)0.89 (0.64–0.97)0.59 (0.29–0.80)1.1 (0.5–2.5)NS206 (30.0)0.84 (0.59–0.95)2.8 (1.1–6.9)0.025
-Y8NANANANS8NS
Double incl. del(5q)c178 (47.0)0.53 (0.23–0.75)0.53 (0.23–0.75)2.1 (1.4–4.5)0.04172 (11.7)0.84 (0.49–0.96)1.7 (0.4–7.4)NS
Double incl. any other2511 (44.0)0.49 (0.24–0.70)0.41 (0.17–0.63)2.8 (1.4–5.4)0.001257 (28.0)0.71 (0.45–0.86)4.1 (1.8–9.7)0.001
Table 4. Biological and Clinical Parameters Influencing OS and LFS in Univariate Analysis With no Censoring for Specific Therapies
 Overall survivalAML progression
CategoryTot.Events (%)At 2-year (95% CI)At 5-years (95% CI)HR (95% CI)PTot.Events (%)At 2-years (95% CI)HR (95% CI)P
  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; NA, not available. FAB classification, French–American–British classification; WHO classification, World Health Organization classification; RARS, RA with ringed sideroblasts; RA, refractory anemia; RCMD, refractory cytopenia with multilineage displasia; RCMDS, RCMD with ringed sideroblasts; RAEB-1, refractory anemia with excess of blasts type 1; RAEB-2, RAEB type 2; MDS-U, unclassifiable MDS; IPSS, International Prognostic Scoring System; NCCSS, New Comprehensive Cytogenetic Scoring System.

  2. a

    As no patient within the very good prognostic category progressed to AML, all categories were compared to the intermediate risk category which showed a significant higher risk of AML evolution than the good risk category.

  3. b

    No patient with more than six defects.

  4. c

    Versus normal karyotypes.

Cytopenias     0.00001    0.00001
010019 (19.0)0.90 (0.80–0.95)0.75 (0.61–0.85)1008 (8.0)0.96 (0.87–0.98)
126285 (32.4)0.83 (0.77–0.87)0.61 (0.52–0.68)1.7 (1.0–2.8)0.03526250 (19.1)0.83 (0.77–0.88)2.4 (1.1–5.0)0.023
218284 (46.1)0.65 (0.57–0.72)0.43 (0.34–0.52)2.9 (1.7–4.7)0.000118247 (25.8)0.72 (0.64–0.79)3.9 (1.8–8.2)0.0001
38645 (52.3)0.43 (0.30–0.55)0.32 (0.20–0.45)4.6 (2.7–7.9)0.00018635 (40.7)0.45 (0.31–0.58)8.9 (4.1–19.3)0.0001
Blast cell percentage     0.00001    0.00001
≤5413105 (25.4)0.85 (0.80–0.88)0.71 (0.65–0.76)41339 (9.4)0.92 (0.87–0.95)
6–109849 (50.0)0.58 (0.45–0.69)0.24 (0.13–0.36)3.1 (2.2–4.4)0.00019829 (29.6)0.62 (0.49–0.72)4.7 (2.9–7.6)0.0001
11–2011979 (66.4)0.46 (0.36–0.56)0.18 (0.10–0.27)4.8 (3.6–6.5)0.000111972 (60.5)0.36 (0.26–0.46)12.3 (8.2–18.3)0.0001
WHO classification     0.00001    0.00001
RA12125 (20.6)0.90 (0.82–0.95)0.74 (0.60–0.83)12110 (8.3)0.92 (0.84–0.96)
RARS705 (7.1)0.98 (0.87–0.99)0.93 (0.79–0.97)0.3 (0.1–0.7)0.011703 (4.3)0.97 (0.87–0.99)0.4 (0.1–1.6)NS
RCMDS2512 (48.0)0.73 (0.50–0.87)0.57 (0.33–0.75)2.2 (1.1–4.5)0.021254 (16.0)0.95 (0.68–0.99)1.8 (0.5–5.6)NS
RCMD14852 (35.1)0.76 (0.67–0.83)0.58 (0.47–0.67)1.8 (1.1–2.9)0.01414816 (10.8)0.90 (0.82–0.94)1.4 (0.6–3.1)NS
5q- Syndrome389 (23.7)0.88 (0.72–0.95)0.80 (0.61–0.91)0.8 (0.4–1.8)NS385 (13.2)0.92 (0.77–0.97)1.2 (0.4–3.7)NS
MDS-u92 (22.2)0.73 (0.28–0.92)0.73 (0.28–0.92)1.5 (0.3–6.3)NS91 (11.1)0.87 (0.39–0.98)1.9 (0.2–14.7)NS
RAEB-110452 (50.0)0.58 (0.46–0.69)0.24 (0.14–0.36)3.7 (2.3–6.0)0.000110431 (29.8)0.62 (0.50–0.72)5.2 (2.5–14.7)0.0001
RAEB-211576 (66.1)0.46 (0.35–0.56)0.19 (0.11–0.29)5.7 (3.6–9.0)0.000111570 (60.9)0.36 (0.26–0.46)13.9 (7.1–27.1)0.0001
IPSS score:     0.00001    0.00001
Low-risk17728 (15.8)0.95 (0.89–0.97)0.81 (0.70–0.88)1778 (4.5)0.96 (0.91–0.98)
Intermed -125681 (31.6)0.81 (0.75–0.86)0.64 (0.56–0.70)2.1 (1.4–3.3)0.000125633 (12.9)0.89 (0.84–0.93)3.1 (1.4–6.7)0.004
Intermed -213784 (61.3)0.50 (0.40–0.59)0.21 (0.13–0.31)6.3 (4.1–9.7)0.000113758 (42.3)0.53 (0.42–0.62)15.7 (7.4–33.1)0.0001
High-risk6040 (66.7)0.36 (0.22–0.50)0.17 (0.07–0.30)9.3 (5.7–15.1)0.00016041 (68.3)0.21 (0.11–0.34)37.4 (17.3–80.5)0.0001
IPSS cytogenetic categ.     0.00001    0.00001
Good374103 (27.5)0.85 (0.81–0.89)0.65 (0.58–0.72)37444 (11.8)0.89 (0.84–0.92)
Intermediate17474 (42.5)0.69 (0.61–0.76)0.48 (0.38–0.57)1.7 (1.3–2.3)0.000117451 (29.3)0.71 (0.63–0.78)2.7 (1.8–4.1)0.0001
Poor8256 (68.3)0.33 (0.22–0.45)0.19 (0.07–0.30)4.4 (3.2–6.1)0.00018245 (54.9)0.38 (0.26–0.50)7.8 (5.1–11.9)0.0001
NCCSa     0.00001    0.00001
Very good133 (23.1)NR0.91 (0.51–0.99)13
Good405118 (29.1)0.84 (0.79–0.87)0.64 (0.57–0.70)3.0 (0.9–9.4)0.0340549 (12.1)0.88 (0.84–0.91)0.3 (0.2–0.4)0.0001
Intermediate13860 (43.4)0.68 (0.59–0.76)0.41 (0.30–0.52)5.5 (1.7–17.6)0.00413848 (34.8)0.67 (0.57–0.75)
Poor4225 (59.5)0.45 (0.29–0.61)0.31 (0.16–0.47)7.5 (2.2–24.9)0.0014221 (50.0)0.50 (0.33–0.64)1.5 (0.9–2.6)NS
Very poor3227 (84.4)0.05 (0.03–0.19)NA48.8 (14.4–165.3)0.00013222 (68.7)NA5.6 (3.3–9.6)0.0001
Karyotype     0.00001    0.00001
Normal29176 (26.1)0.85 (0.79–0.89)0.66 (0.58–0.73)29127 (9.3)0.90 (0.85–0.93)
Abnormal339157 (46.3)0.66 (0.60–0.71)0.45 (0.38–0.51)1.8 (1.3–2.3)0.00001339113 (33.5)0.67 (0.62–0.73)3.7 (2.4–5.6)0.0001
Number of abnormalities     0.00001    0.00001
None29176 (26.1)0.85 (0.79–0.89)0.66 (0.58–0.73)29127 (9.3)0.90 (0.85–0.93)
One24798 (39.7)0.76 (0.70–0.82)0.50 (0.42–0.57)1.3 (1.0–1.8)0.0424769 (27.9)0.75 (0.69–0.80)2.8 (1.8–4.4)0.0001
Two4623 (50.0)0.53 (0.35–0.67)0.50 (0.32–0.64)2.1 (1.3–3.4)0.0014615 (32.6)0.67 (0.49–0.80)4.1 (2.1–7.7)0.0001
Three138 (61.5)0.47 (0.18–0.72)0.35 (0.10–0.63)2.3 (1.1–4.7)0.028136 (46.1)0.56 (0.24–0.79)5.7 (2.3–13.8)0.0001
Four118 (72.7)0.17 (0.01–0.50)NA13.7 (6.5–29.0)0.0001117 (63.6)NA20.9 (9.0–48.6)0.0001
Five76 (85.7)0.21 (0.01–0.58)NA7.1 (3.1–16.4)0.000175 (71.4)NA16.5 (6.3–43.4)0.0001
Sixb1514 (93.3)NANA17.1 (9.5–31.9)0.00011511(73.3)NA25.6 (12.3–52.9)0.0001
Single defectsc     0.00001    0.00001
3q defects106 (60.0)0.41 (0.11–0.70)0.28 (0.04–0.59)3.2 (1.4–7.4)0.007104 (40.0)0.60 (0.20–0.80)12.8 (6.9–23.6)0.0001
Del(5q)5517 (31.9)0.85 (0.71–0.92)0.68 (0.50–0.80)0.9 (0.3–1.9)NS5511 (20.0)0.84 (0.70–0.91)2.3 (0.3–2.6)NS
Del(7q)2312 (52.1)0.61 (0.35–0.79)0.35 (0.12–0.60)2.2 (1.2–4.0)0.013238 (34.8)0.66 (0.38–0.83)4.3 (1.9–9.6)0.0001
−7149 (64.3)0.40 (0.14–0.66)0.15 (0.01–0.45)3.3 (1.4–7.6)0.0001149 (64.3)0.35 (0.11–0.61)10.7 (5.0–22.9)0.0001
Der(1;7)53 (60.0)NR0.50 (0.05–0.84)4.1 (2.0–8.1)NS51 (20.0)0.75 (0.13–0.96)1.9 (0.3–14.0)NS
+83516 (45.7)0.68 (0.48–0.82)0.32 (0.14–0.52)1.9 (1.1–3.3)0.0173510 (28.6)0.76 (0.57–0.88)3.2 (1.6–6.7)0.002
Del(11q)53 (60.0)NR0.80 (0.20–0.97)0.9 (0.3–2.8)NS5NANS
Del(12p)174 (23.5)0.82 (0.54–0.94)0.73 (0.41–0.89)0.8 (0.3–2.1)NS171 (5.9)NR0.5 (0.1–4.2)NS
Del(20q)2010 (50.0)0.89 (0.64–0.97)0.38 (0.14–0.63)1.3 (0.7–2.6)NS206 (30.0)0.84 (0.59–0.95)2.5 (1.0–6.1)0.041
-Y8NANANANS8NS
Double incl. del(5q)c178 (47.0)0.53 (0.23–0.75)0.53 (0.23–0.75)1.9 (0.9–3.9)NS173 (17.6)0.75 (0.42–0.91)2.2 (0.7–7.4)NS
Double incl. any otherc2513 (52.0)0.52 (0.28–0.71)0.44 (0.21–0.65)2.4 (1.3–4.3)0.0032510 (40.0)0.66 (0.40–0.83)4.8 (2.3–9.9)0.0001

Considering biological parameters, the IPSS and NCCSS cytogenetic categories as well as the number and type of cytogenetic abnormalities had a significant influence on OSand LFS when censoring for specific treatments was applied (Table 3, Figs. 1 and 2). When single defects were considered, del(5q), der(1;7), del(11q), del(12p), del(20q), and –Y presented an OS and a risk of AML evolution similar to those of normal karyotypes, whereas 3q abnormalities, −7, del(7q), +8 presented a significantly worse OS and a significantly higher risk of AML progression (Table 3). The OS of del(20q) was similar to that of normal karyotypes, but the LFS was significantly worse (P = 0.025 with censoring and P = 0.04 without censoring). When −7 patients were compared to del(7q) patients no difference in OS and a significant difference in LFS (P = 0.02) became apparent (Table 5). In comparison with normal karyotypes, −7 patients presented a death rate of 24.3% (95% CI = 10.9–54.1) with a HR of 3.3 (95% CI: 1.4–7.6), whereas del(7q) patients presented a death rate of 21.9% (95% CI = 12.1–39.7) with a HR of 2.8 (95% CI: 1.4−5.3). In addition, AML evolution rates were 41.0% (95% CI = 20.5–82.1) with a HR of 11.2 (95% CI: 5.0–25.2) for −7 patients versus 10.9% (95% CI = 4.5–26.1) with a HR of 3.0 (95% CI: 1.1–7.9) for del(7q) patients (P = 0.02). When −7 and del(7q) patients were compared without censoring for specific treatments no significant difference in OS and a trend toward a significant difference in LFS (P = 0.06) were noted (Table 5).

Figure 1.

OS according to the NCCSS with censoring. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Figure 2.

LFS according to the NCCSS with censoring. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Table 5. Influence of the cytogenetic pattern on OS and PFI by univariate analysis
 Overall survivalAML progression
CategoryaTot.Events (%)At 2-year (95% CI)At 5-years (95% CI)HR (95% CI)PTot.Events (%)At 2-years (95% CI)HR (95% CI)P
  1. a

    Patients with “del(7q)” were compared to those with “−7” and patients with “double abnormalities including del(5q)” to those with “double abnormalities including any other.”

  2. b

    When censoring was not applied, no statistical difference in OS and a trend toward a statistical significant difference in LFS (P = 0.06).

  3. c

    When censoring was not applied, no statistical difference was noted in OS and LFS was noted.

Single defects
Del(7q)2311 (47.3)0.64 (0.36–0.82)0.33 (0.09–0.59)235 (21.7)0.74 (0.43–0.89)
−7b146 (42.8)0.58 (0.21–0.82)0.21 (0.01–0.58)0.9 (0.3–2.3)0.7148 (57.1)0.43 (0.16–0.68)0.3 (0.09–0.82)0.02
Double abnormalities
Including del(5q)178 (47.0)0.53 (0.23–0.75)0.53 (0.23–0.75)172 (11.7)0.84 (0.49–0.96)
Includ. any otherc2511 (44.0)0.49 (0.24–0.70)0.41 (0.17–0.63)1.2 (0.4–3.0)0.7c257 (28.0)0.71 (0.45–0.86)2.5 (0.52–12.70)0.25c

Considering double defects, four patients only presented double defects including −7/del(7q), two died 4 and 7 months after clinical diagnosis and all experienced disease evolution. Thus, only double defects including del(5q) and other double defects were statistically analyzed. These cytogenetic categories presented similar OS and LFS with and without censoring (Table 5). However, when censoring for specific treatments was applied, in comparison with normal karyotypes, the OS of double defects including 5q- entailed a HR of 2.1 (95% CI: 1.4–4.5) with a P = 0.04, whereas the OS of other double defects had a HR of 2.8 (95% CI: 1.4–5.4) with a P < 0.001. In addition, the AML evolution rate was 5.1% (95% CI = 1.3–20.5) with a HR of 1.7 (95% CI: 0.4–7.4) for patients with double defects including 5q- versus 15.4% (95% CI = 7.3–32.2) with a HR of 4.1 (95% CI = 1.8–9.7) for those with other double defects.

A progressive increase in the number of chromosomal defects was significantly correlated with a progressive worsening of OS and a progressive increase for risk of AML evolution (Table 3).

When cytogenetic defects were grouped within the five different NCCSS subgroups, a statistically significant influence of each new cytogenetic category on both OS and LFS (P < 0.0001; P < 0.0001) was noted (Figs. 1 and 2). In particular, a significantly different OS was observed when the very good NCCSS category was compared with the good (P = 0.04), the good with the intermediate (P = 0.007), the intermediate with the poor (P = 0.005), and the poor with the very poor (P < 0.0001). No patient of the very good category progressed to AML. When compared to the intermediate category, the good NCCS category presented a significantly better LFS (P = 0.0001), whereas the poor and the very poor showed a significantly worse LFS (P < 0.009 and P < 0.0001) (Table 3). In addition, a significantly different LFS was noted when the poor and the very poor categories were compared (P = 0.03).

When censoring was not performed a significant difference in OS was observed when the very good category was compared to the good (P = 0.03), the good to the intermediate (P = 0.0001), the poor to the very poor (P < 0.00001). Instead, only a trend toward a significantly different OS was noted when the intermediate and poor categories were compared (P = 0.09) (Table 4, Fig. 3). This same result was obtained when the LFS of the intermediate and poor categories were compared (P = 0.09). Instead, the LFS of the good category was significantly better than that of the intermediate category (P < 0.0001) and the LFS of the very poor category was significantly worse than that of the intermediate category (P < 0.0001) (Table 4, Fig. 4).

Figure 3.

OS according to the NCCSS without censoring. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Figure 4.

LFS according to the NCCSS without censoring. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Multivariate analysis

To estimate whether the NCCSS grouping was superior to the standard IPSS cytogenetic grouping for predicting OS and LFS, six different multivariate models including variables which revealed a significant prognostic influence on univariate analyzes were developed. As in a previous study [14], these models were compared by means of the AIC (Tables 6 and 7). Independently of censoring, all multivariate models had a significant influence on both OS and LFS (P < 0.0001 and P < 0.0001 respectively). According to the AIC, the model that included the number of cytopenias, the WHO classification and the NCCSS cytogenetic grouping was the best (AIC = 1896) for predicting OS, followed by the model that included the number of cytopenias, the blast cell percentage and the NCCSS cytogenetic grouping (AIC = 1907) (Table 6). When no censoring was applied these findings were maintained: AIC values were 2417 and 2429, respectively. Considering LFS, the model that included the number of cytopenias, the blast cell percentage and the NCCSS cytogenetic grouping was the best (AIC = 1130), although it was not much better than the model that included the number of cytopenias, the WHO classification and the NCCSS cytogenetic grouping (AIC = 1137) (Table 7). When no censoring was applied, these findings were maintained: AIC values were 1423 and 1430, respectively.

Table 6. Comparison of Six Multivariate Models Significantly Effecting OS by Means of the Akaike Information Criterion (AIC) with Censoringa
First model AIC = 1927Second model AIC = 1907Third model AIC = 1925
  1. a

    With no censoring in univariate analyses: AIC = 2460 for the first model; AIC = 2429 for the second model; AIC = 2443 for the third model; AIC = 2454 for the fourth model; AIC = 2417 for the fifth model and AIC = 2430 for the sixth model.

VariablesCat.HR (95% CI)PVariablesCat.HR (95% CI)PVariablesCat.HR (95% CI)P
Age<65Age<65Age<65
≥651.6 (1.2–2.1)0.001≥651.6 (1.2–2.2)0.001≥651.6 (1.2–2.2)0.001
Cytopenia0Cytopenia0Cytopenia0
 11.4 (0.8–2.4)NS 11.4 (0.8–2.4)NS 11.4 (0.8–2.4)NS
 21.9 (1.1–3.3)0.01 21.9 (1.2–3.4)0.01 22.1 (1.2–3.5)0.008
 32.8 (1.6–5.2)0.001 32.8 (1.6–5.2)0.001 32.9 (1.6–5.4)0.0001
Blast cell perc.<5Blast cell perc.<5Blast cell perc.<5
 5–103.0 (2.0–4.5)0.0001 5–102.8 (1.9–4.2)0.0001 5–103.0 (2.0–4.5)0.0001
 11–204.3 (2.9–6.1)0.0001 11–203.9 (2.7–5.6)0.0001 11–204.2 (2.9–6.1)0.0001
IPSS Cytog. Cat.GoodNCCSS Categ.Very GoodNumber of Chrom. Abnor.None
 Iterm.1.5 (1.1–2.2)0.01 Good3.5 (0.8–14.4)NS One1.3 (0.9–1.9)NS
 Poor3.5 (2.3–5.1)0.0001 Intermediate5.2 (1.2–21.9)0.02 Two2.1 (1.2–3.6)0.003
     Poor9.7 (2.2–42.6)0.002 Three2.1 (0.9–5.1)0.08
     Very Poor29.7 (6.7–132.4)0.0001 Four6.0 (2.5–14.0)0.0001
         Five9.3 (3.2–26.6)0.0001
         Six8.6 (4.2–17.8)0.0001
Fourth model AIC = 1921Fifth model AIC = 1896Sixth model AIC = 1909
VariablesCat.HR (95% CI)PVariablesCat.HR (95% CI)PVariablesCat.HR (95% CI)P
Age<65Age<65Age<65
≥651.6 (1.2–2.2)0.001≥651.6 (1.2–2.1)0.001≥651.6 (1.2–2.1)0.003
Cytopenia0Cytopenia0Cytopenia0
 11.3 (0.7–2.2)NS 11.3 (0.7–2.2)NS 11.3 (0.7–2.1)NS
 21.5 (0.9–2.6)NS 21.5 (0.9–2.6)NS 21.6 (0.9–2.8)NS
 32.3 (1.2-4.3)0.008 32.1 (1.2–4.1)0.01 32.3 (1.2–4.4)0.09
WHO class.RAWHO class.RAWHO class.RA
 RARS0.3 (0.1–1.5)0.06 RARS0.3 (0.1–0.9)0.03 RARS0.3 (0.1–0.9)0.03
 RCMD1.6 (0.9–2.8)NS RCMD1.7 (0.9–3.2)0.06 RCMD1.7 (0.9–3.0)NS
 RCMDS2.4 (1.1–5.3)0.02 RCMDS2.6 (1.2–5.6)0.01 RCMDS2.3 (1.1–5.1)0.03
 5q- Syn.1.3 (0.5–2.9)NS 5q- Syn.1.1 (0.5–2.6)NS 5q- Syn.0.9 (0.4–2.3)NS
 U-MDS2.1 (0.5–9.3)NS U-MDS1.8 (0.4–7.8)NS U-MDS1.9 (0.4–8.5)NS
 RAEB-13.9 (2.2–7.1)0.0001 RAEB-13.6 (2.0–6.4)0.0001 RAEB-13.7 (2.0–6.5)0.0001
 RAEB-25.6 (3.2–10.0)0.0001 RAEB-25.2 (2.9–9.4)0.0001 RAEB-25.4 (3.0–9.7)0.0001
IPSS Cytog. Cat.GoodNCCSS Categ.Very goodNumber of Chrom. Abnor.None
 Iterm.1.5 (1.0–2.1)0.02 Good4.1 (0.9–16.9)0.05 One1.3 (0.9–1.8)NS
 Poor3.2 (2.2–4.8)0.0001 Intermediate5.6 (1.3–23.7)0.01 Two1.7 (1.2–3.5)0.006
     Poor10.2 (2.3–44.8)0.002 Three1.9 (0.8–4.5)NS
     Very poor36.4 (8.1–162.9)0.0001 Four5.8 (2.5–13.7)0.0001
         Five17.7 (6.1–50.9)0.0001
         Six10.4 (4.9–21.7)0.0001
Table 7. Comparison of Six Multivariate Models Significantly Effecting LFS by Means of the Akaike Information Criterion (AIC) with Censoringa
First model AIC = 1146Second modelb AIC = 1130Third model AIC = 1147
  1. a

    With no censoring in univariate analyses: AIC = 1442 for the first model; AIC = 1423 for the second model; AIC = 1439 for the third model; AIC = 1450 for the fourth model; AIC = 1430 for the fifth model and AIC = 1447 for the sixth model.

  2. b

    As no patients of the very good category progressed to AML, the intermediate cytogenetic category was used for comparisons.

VariablesCat.HR (95% CI)PVariablesCat.HR (95% CI)PVariablesCat.HR (95% CI)P
Age<65Age<65Age<65
≥651.0 (0.7–1.5)NS≥651.0 (0.7–1.5)NS≥651.0 (0.7–1.5)NS
Cytopenia0Cytopenia0Cytopenia0
 12.4 (1.0–5.8)0.04 12.4 (1.0–5.7)0.04 12.2 (0.9–5.2)NS
 22.6 (1.1–6.2)0.03 22.5 (1.1–6.1)0.03 22.7 (1.1–6.6)0.02
 34.2 (1.7–10.4)0.002 33.8 (1.5–9.6)0.004 34.4 (1.8–11.0)0.001
Blast cell perc.<5Blast cell perc.<5Blast cell perc.<5
 5–105.2 (3.0–9.1)0.0001 5–105.1 (2.9–8.9)0.0001 5–105.7 (3.3–9.9)0.0001
 11–2010.3 (6.3–16.8)0.0001 11–209.4 (5.7–15.5)0.0001 11–2010.6 (6.5–17.3)0.0001
IPSS Cytog. Cat.GoodNCCSS Categ.Very GoodNumber of Chrom. Abnor.None
 Iterm.2.2 (1.4–3.5)0.001 Good0.4 (0.3–0.7)0.0001 One2.3 (1.4–3.8)0.001
 Poor3.8 (2.3–6.3)0.0001 Intermediate Two2.6 (1.2–5.5)0.01
     Poor1.9 (1.1–3.5)0.002 Three3.5 (1.4–8.8)0.009
     Very Poor2.9 (1.5–5.7)0.001 Four2.9 (0.8–10.1)NS
         Five15.6 (4.5–53.9)0.0001
         Six9.2 (3.8–22.1)0.0001
Fourth model AIC = 1921Fifth model AIC = 1896Sixth model AIC = 1909
VariablesCat.HR (95% CI)PVariablesCat.HR (95% CI)PVariablesCat.HR (95% CI)P
Age<65Age<65Age<65
≥651.0 (0.7–1.5)NS≥651.0 (0.7–1.5)NS≥651.0 (0.7–1.5)NS
Cytopenia0Cytopenia0Cytopenia0
 12.3 (0.9–5.5)0.05 12.1 (0.9–5.1)NS 11.9 (0.8–4.7)NS
 22.4 (0.9–5.8)0.05 22.3 (0.9–5.4)NS 22.4 (0.9–5.9)0.05
 33.9 (1.5–9.9)0.004 33.3 (1.3–8.5)0.01 34.0 (1.6–10.2)0.004
WHO class.RAWHO class.RAWHO class.RA
 RARS0.5 (0.01–2.2)NS RARS0.4 (0.1–1.9)NS RARS0.4 (0.1–1.7)NS
 RCMD0.9 (0.4–2.5)NS RCMD1.1 (0.4–2.8)NS RCMD0.9 (0.4–2.3)NS
 RCMDS0.8 (0.2–3.7)NS RCMDS0.7 (0.2–3.7)NS RCMDS0.7 (0.2–3.7)NS
 5q- Syn.1.6 (0.5–5.5)NS 5q- Syn.1.4 (0.4–4.9)NS 5q- Syn.0.8 (0.2–2.7)NS
 U-MDS U-MDS U-MDS
 RAEB-14.9 (2.2–11.2)0.0001 RAEB-14.6 (2.0–10.5)0.0001 RAEB-14.5 (2.0–10.2)0.0001
 RAEB-29.6 (4.3–21.3)0.0001 RAEB-28.7 (3.9–19.6)0.0001 RAEB-28.4 (3.8–18.8)0.0001
IPSS Cytog. Cat.GoodNCCSS Categ.Very GoodNumber of Chrom. Abnor.None
 Iterm.2.3 (1.4–3.6)0.001 Good0.4 (0.3–0.7)0.0001 One2.2 (1.4–3.7)0.002
 Poor3.8 (2.3–6.4)0.0001 Intermediate Two2.6 (1.2–5.6)0.014
     Poor1.9 (1.1–3.5)0.002 Three3.3 (1.3–8.4)0.011
     Very Poor2.9 (1.5–5.7)0.001 Four2.9 (0.8–10.0)NS
         Five23.4 (6.6–82.5)0.0001
         Six10.6 (4.1–24.3)0.0001

Discussion

The principal aim of this study was to evaluate the efficacy of the new NCCSS to predict the clinical outcome of 630 consecutive MDS patients from a single Institution. In our series, clonal chromosomal abnormalities were revealed in 53.8% of the 630 de novo consecutive MDS patients. This frequency is similar to that reported by other studies [3-13, 16, 17], even if it should be underlined that our study did not include patients diagnosed as RAEB in transformation (RAEB-t), now considered AML by the WHO classification. Instead, the NCCSS study [17] also analyzed RAEB-t patients, as the exclusion of this MDS FAB subtype led to a statistically lower risk in the entire sample without affecting the relative position of the five prognostic subgroups. In addition, it has been reported that the upcoming IPSS revision will include RAEB-t patients as well.

Many studies based on univariate analyzes, have debated whether del(7q) and −7 should be included in different prognostic cytogenetic subgroups. A fundamental contribution to resolve this issue will surely be provided by new molecular technologies, for example, a SNP array study recently demonstrated that haploinsufficiency of defined 7q regions may translate to different clinical outcomes for MDS and AML patients carrying a 7q deletion and dysplastic features [36]. The IPSS [8], the Spanish MDS Cytogenetic Working Group [10] and another study [16] have reported that del(7q) and −7 present a similar poor OS and a similar high risk of MDS/AML evolution. In contrast, the GA data set [13] and other reports [11, 12, 37] have revealed that isolated del(7q) and −7 were associated with an intermediate and a poor clinical outcome respectively. In particular, two previous studies by our own group [11, 12] revealed that in univariate and multivariate analyzes del(7q) was associated with a survival probability better than that of the poor IPSS cytogenetic category and similar to that of the intermediate category, whereas the risk of MDS/AML evolution was similar to those of both IPSS cytogenetic categories. However, the AIC revealed that the IPSS prognostic power was not ameliorated by the introduction of del(7)(q31q35) as a new entity. In this study, univariate analysis revealed that after censoring for specific treatments patients with del(7q) and −7 presented a similar OS and a significantly different LFS (P = 0.02), a datum that was not confirmed by multivariate analysis. In addition, when no censoring was applied −7 and del(7q) patients showed similar OS and a trend toward a significantly different LFS. However, once again this datum was not confirmed by multivariate analysis. Thus, our data partially agree with a recent study that analyzed the prognostic implications of sole chromosome seven abnormalities in myeloid malignancies [38]. That study reported no significant difference in OS and LFS comparisons, adjusted for the presence of either excess blasts or multilineage dysplasia among patients with der(1;7)(p10;p10), del(7q), and −7. Instead, in our series the five patients with der(1;7)(q10;q10) experienced a clinical outcome similar to that of normal karyotypes. Thus, in agreement with Sanada et al. [39] and in contrast with Slovak et al [40], our result confirms that der(1;7)(q10;p10) truly defines a distinct MDS risk group with an OS and a LFS significantly better than those associated with −7 and del(7q).

Despite the efforts of many researchers [6-13, 16, 20], the prognostic significance of rare single chromosomal defects, double defects and unrelated clones, which flag the extreme heterogeneity of the MDS cytogenetic pattern [2, 21, 28-31, 41], has remained ill defined. In fact, most of the aforementioned studies provided conflicting results due to the small number of patients studied. Recently, the availability of 2,801 karyotypes allowed the NCCSS to overcome these problems and to define the clinical outcome of thirteen rare single defects [17], a result confirmed by the IPSS-R study [18]. In our series, except for der(1;7), the other most common rare defects were i(17q), del(17p), and trisomy 13. However, they were present in only four patients each, so no statistical survival analysis could be performed. Death occurred in three patients with i(17q), in two with del(17p) and in three with +13, whereas MDS/AML evolution occurred in two patients with i(17q), in three with del(17p) and in three with +13. So, these data confirm the poor prognostic significance of 17p abnormalities [10, 13, 17] and for the first time suggest that also +13 might be associated with an unfavourable clinical outcome. Trisomy 13 has been rarely observed in MDS [42] and in AML it is strongly associated with AML1/RUNX mutations and increased FLT3 expression [43].

Considering double abnormalities, the only possible comparison was between double defects including del(5q) and other double defects. A recent cooperative study revealed that patients with one defect in addition to del(5q) presented an OS similar and a risk of AML transformation higher than those of patients with del(5q) alone [44]. Intriguingly, a current SNP-A report showed that deletions involving the centromeric or telomeric extremes of 5q are associated with a more aggressive disease phenotype and additional chromosomal lesions [45]. This finding appears to be confirmed by our study. After censoring, patients with an isolated del(5q) presented an OS and a LFS similar to those of chromosomally normal patients and thus significantly better than those of patients with double defects including del(5q) (P = 0.001 and P = 0.001). However, without censoring OS and LFS were similar among patients with normal karyotypes, del(5q) and double defects including del(5q) (Table 3). These last patients presented an OS and a LFS similar to those of patients with other double defects independently of censoring (Table 5). When the OS of patients with double defects including del(5q) was compared to that of chromosomally normal patients and the OS of patients with other double defects was compared to that of chromosomally normal patients a similar HR was observed: 2.1 (95% CI: 1.4–4.5) versus 2.8 (95% CI: 1.4–5.4) with censoring and 1.9 (95% CI: 1.3–2.3) versus 2.4 (95% CI: 1.3–4.3) with no censoring. When these same comparisons were made for LFS the HR was 1.7 (95% CI: 0.4–7.4) versus 4.1 (95% CI: 1.8–9.7) with censoring and 2.2 (95% CI: 0.7–7.4) versus 4.7 (95% CI: 2.3–9.9) with no censoring. Thus, our data do not confirm the observation made by the NCCSS that the OS of patients with double defects including del(5q) is significantly better than that of patients with other double defects [17, 18].

Despite these differences, when censoring was performed and our patients were grouped according to the NCCSS a significant difference in OS and in LFS (Table 3, Figs. 1 and 2) among all the newly defined cytogenetic categories was noted. However, when no censoring was performed only a trend toward a statistical significant difference in OS and LFS was observed when the intermediate and the poor NCCSS categories were compared (Table 4, Figs. 3 and 4). This datum might suggest that despite recent NCCSS efforts the prognostic relevance of a significant proportion of single cytogenetic abnormalities is still ill defined [46]. In fact, the NCCSS considers these truly rare chromosomal defects as a single group and assigns them to the intermediate category which is no different than the current IPSS. In addition, the prognostic relevance of the seven new NCCSS categories is based on a relatively small number of patients and thus their prognostic relevance could vary from one study to another and in relation to the different treatments applied [46]. In the present series der(1;7) determined a favourable outcome, but in another an unfavourable outcome [40]. Despite these drawbacks, it has been observed that the NCCSS is practically more useful to consolidate the very good and good categories as well as the very poor and poor categories [46]. This fact is confirmed by our study as a clear cut difference in OS and LFS was noted when the very good was compared to the good category and the poor to the very poor.

Further proof of the NCCSS effectiveness in improving the prognostic stratification of cytogenetic abnormalities came from the comparison of six multivariate models. The most effective model to predict OS included the WHO classification and the NCCSS cytogenetic grouping (AIC = 1,896 with censoring and AIC = 2,417 with no censoring) (Table 5), whereas the model that included the blast cell percentage and the NCCSS cytogenetic grouping and the model that included the WHO classification and the NCCSS cytogenetic grouping were almost equally effective for predicting LFS (AIC = 1,130 versus 1,137 with censoring; AIC = 1,423 versus 1,430 with no censoring) (Table 6).

In conclusion, the NCCS (i) improves the prognostic stratification of the good and poor IPSS cytogenetic categories by introducing the very good and the very poor categories; (ii) is still incomplete in establishing the prognostic relevance of rare/double defects, ii) applied to patients who receive supportive treatment only identifies five different prognostic subgroups, but applied to patients treated with specific therapies reveals only a trend toward a significantly different OS and LFS when patients of the poor and intermediate cytogenetic categories are compared, (iii) combined with the WHO classification is much more effective than the IPSS in predicting MDS clinical outcome.

Author Contributions

Marina Boni, Paola Maria Cavigliano, Irene Dambruoso, Rita Zappatore performed reasearch and contributed essential reagents.

Acknowledgments

Paolo Bernasconi performed research, designed the research study, wrote the paper. Catherine Klersy analyzed the data.

Ancillary