MDS-type abnormalities within myeloma signature karyotype (MM-MDS): only 13% 1-year survival despite tandem transplants

Authors


Joth L. Jacobson, MS, Cancer Research and Biostatistics, 1730 Minor Ave STE 1900, Seattle, WA 98101–1468, USA. E-mail: jothj@crab.org

Abstract

Summary. Cytogenetic abnormalities (CA), especially of chromosome 13, have been used to identify a subgroup of previously untreated multiple myeloma (MM) patients with very poor prognosis despite high-dose therapy (HDT). We examined the prognostic implications of CA in 1000 MM patients receiving melphalan-based tandem autotransplants (median follow-up, 5 years). Negative consequences for both overall survival (OS) and event-free survival (EFS) in the presence of any CA were confirmed, especially when detected within 3 months of HDT. In the context of standard prognostic factors (SPF), ‘MM-MDS’ (MM karyotype that contains, in addition, CA typical of MDS) imparted a poor OS and EFS, after adjusting for any CA and all individual CA. One-year mortality was also high, especially for the MM-MDS subgroup with trisomy 8 within a MM signature karyotype (87%vs 34% in its absence, P < 0·001). No patient remained event free 5 years post transplant in the presence of these baseline high-risk CA. However, certain trisomies (e.g. chromosomes 7 and 9) and del 20 had favourable clinical consequences. The higher risk that is associated with CA compared with SPF justifies routine cytogenetic studies in all patients with MM at diagnosis and whenever additional treatment decisions are considered, such as in planning HDT either for initial response consolidation, at the time of primary unresponsiveness to induction therapy, or at relapse.

The clinical course of patients with multiple myeloma (MM) is quite variable and can be partially predicted by a number of laboratory parameters such as beta-2-microglobulin (β2m), C-reactive protein (CRP) and the plasma cell labelling index (PCLI) (Latreille et al, 1982; Boccadorro et al, 1987; Greipp et al, 1988, 1993). Recently, abnormalities of chromosome 13 (mainly complete and partial deletions of the long arm), found to be present in 16% of untreated patients, were recognized as the major adverse disease feature with high-dose therapy (HDT) (Tricot et al, 1995) and, subsequently, also with standard-dose therapy (SDT) (Seong et al, 1998). While proliferation-independent genetic measurements such as DNA flow cytometry (Barlogie et al, 1985) and, in recent years, interphase fluorescence hybridization (FISH) have revealed aneuploidy in virtually all patients with active MM (Drach et al, 1995; Tabernero et al, 1996; Zandecki et al, 1996), abnormal metaphases are obtained only in about one-third of newly diagnosed patients (Dewald et al, 1985; Gould et al, 1988; Sawyer et al, 1995). Cytogenetic abnormalities (CA) are typically complex in MM and represent a hallmark of this disease, involving many chromosomes that are altered both numerically and structurally (Rao et al, 1998; Sawyer et al, 1998a, 2001). The mechanisms underlying such gross chromosome instability (CIN) are not understood but may be related to telomere shortening (Klausner, 2002). Although MM-associated recurrent CA have been recognized (Chesi et al, 1996, 1997, 1998a,b), the clinical implications have not been established, except for the consistently poor prognosis associated with the deletion 13 syndrome and, more recently, hypodiploidy (Seong et al, 1998; Avet-Loiseau et al, 2002; Fassas et al, 2002). As part of a comprehensive cytogenetic analysis in all patients referred to the Arkansas myeloma program for HDT, we have reassessed the clinical implications of commonly altered chromosomes along with those of standard prognostic factors (SPF). Unlike previous studies, the current investigation focuses on CA present during the work-up of patients for HDT (within 3 months of HDT) with different prior treatment histories.

Patients and methods

Between September 1989 and June 1998, 1000 eligible patients were enrolled in melphalan-based HDT trials at the Myeloma Institute for Research and Therapy (MIRT) at the University of Arkansas for Medical Sciences, the details of which, along with specifics of laboratory follow-up, have been reported previously (Desikan et al, 2000). Abnormalities of chromosome 13 (henceforth referred to as CA13) were recognized as the dominant adverse CA, while certain recurrent structural and numeric abnormalities were also considered, such as deletions and translocations involving chromosomes 1, 6, 11, 14, 15, 16, 17 and 22 (Desikan et al, 2000).

In the current study, all of the chromosomes considered had regions showing consistently recurring clonal aberrations by both conventional cytogenetics and by spectral karyotyping (Table I) (Dewald et al, 1985; Sawyer et al, 1995, 1998b, 1998c, 2001; Zandecki et al, 1996). Specific CA were labelled using a descriptor string that contains three parts: (1) type of abnormality, (2) chromosome involved, (3) amount of chromosome involved. For the type of abnormality, the options were: additions/trisomies (add), deletions/monosomies (del), duplications (dup), translocations (t) and isochromes (iso). The chromosomes involved were 1–22. The amount of chromosome involved could be the whole chromosome, p-arm, or q-arm. For example, ‘del01p’ denotes a chromosome 1 p-arm deletion, ‘del13’ denotes a whole deletion (or monosomy) of chromosome 13 and ‘add08’ denotes a whole addition (or trisomy) of chromosome 8. MM-MDS denotes a MM karyotype that contains, in addition, cytogenetic abnormalities typical of MDS [del05, del05q, del07, del07q, add08, t(1;7), del20q]. Hyperdiploidy was defined as a modal chromosome number in the range 47–80, or > 92; hypodiploidy was defined as a modal chromosome number in the range 1–45 or 81–91.

Table I.  (A) CA considered in the analysis and the percentage of patients with specific CA (among patients with any CA) present < 3 months prior to HDT (n = 141). : (B) CA groups (among patients with any CA) present < 3 months prior to HDT.
Chromosome numberWhole additionsWhole deletionsDeletionsTranslocationsDuplicationsIsochromesAny*
p-armq-armp-armq-armp-armq-armp-armq-arm
1  20% 11%35% 4% 0% 52%
2     6%      6%
343%          43%
4     1%      1%
532% 6%  8%  6%     47%
6  4% 16%  7%  0%  24%
733% 5%  3% 2% 2%     43%
811%15% 2%  10%   0% 35%
949%    1%      50%
1145% 4%  3% 16%     62%
12   6%        6%
13 36% 19%  6%     52%
14 19%  1% 12%     30%
1554%          54%
16 4%16%  4%  9%     26%
17  9% 0%  1%    1% 11%
1819%     4%     23%
1950%54% 0%  7% 5%     60%
2016%11%  7% 1% 1%     34%
2135%    1%      35%
22 20%  4%  5%     27%
X
Y
Any*77%66%26%45%26%56% 4%0%1%100%† 
 NumberPercentage
  • *

    ‘Any’ represents the number of patients with an abnormality of any chromosome as presented in row/column. For example, 77% had a whole addition of at least one chromosome.

  • CA present in 141 of 732 patients within 3 months of HDT.

Jumping 1 q-arm1813%
Any translocation8560%
Any numeric CA13495%
Any structural CA7956%
Hyperdiploidy9064%
Hypodiploidy5142%
MM-MDS4733%
 t(1;7)2/47 4%
 Del059/4719%
 Del05q11/4723%
 Del077/4715%
 Del07q4/47 9%
 Add0815/4732%
 Del20q10/4721%

The availability of cytogenetic data depended on the nature of a patient's initial referral. In the case of patients on our Total Therapy program (Barlogie et al, 1997, 1999), cytogenetic data were available prior to therapy and serially until HDT was initiated. Other patients had SDT initiated elsewhere and were presented to our centre in various stages: partially responsive, primary refractory, resistant or untested relapse. In these patients, cytogenetic data were available only prior to peripheral blood stem cell mobilization strategies or even HDT itself.

Among all 1000 patients, 7738 cytogenetic results were available in 989 patients, including 3144 cytogenetic results obtained at some time prior to HDT in 965 patients. As a complete cytogenetic history is still rarely available at most centres prior to transplant, this analysis considers only the 732 patients in whom 3033 cytogenetic results were available within 3 months (< 3 months) of HDT. The median number of cytogenetic studies in all 965 patients prior to HDT was two (range 1–15). In the 732 with samples taken < 3 months prior to HDT, the median number of cytogenetic studies was one (range 1–4). The incidence of CA at any time pre-HDT was 37%, including 19% with CA present within 3 months of HDT.

Statistical considerations

The large number of CA under investigation and the low incidence rate of many of the CA made it desirable to reduce the list of CA considered for analysis with outcome variables, to avoid ending up with results that might be spurious and irreproducible. To subset the list of specific CA, we first calculated the minimum incidence rate necessary for a given CA to achieve adequate statistical power (≥ 80%) to distinguish a meaningful effect size in a univariate test for prognostic effect. For this purpose, we chose a hazard ratio of at least 3·5 when compared with patients without a given abnormality as an effect size. As the hazard ratio for the presence of any CA was slightly less than 3·0, an increase to 3·5 seemed appropriate to identify specific CA with a comparatively higher prognostic importance. Using these assumptions, the minimum incidence rate was calculated to be 2% (15 or more) in the overall population (n = 732).

Outcome definitions.  Overall survival (OS) was calculated as the time from initiation of HDT to death from any cause or last contact. Event-free survival (EFS) was calculated as the time from initiation of HDT to either progression of disease or death from any cause or last contact. Secondary endpoints of interest were also explored. One-year mortality was calculated as death within 1 year of HDT initiation.

Analysis methods.  Survival curves were estimated by the product-limit method (Kaplan & Meier, 1958) and compared using the log-rank test (Mantel, 1966). Cox proportional hazards regression (Cox, 1972) was used to assess the influence of cytogenetics and standard prognostic factors on survival outcomes. Logistic regression was used to relate cytogenetic and prognostic factors to dichotomous outcomes. Multivariate models were constructed using stepwise regression methods. All covariates were included in the stepwise selection.

A chromosome cluster tree was created using hierarchical clustering methods (Everitt, 1980). The distance measured between chromosomes, the per cent disagreement, was calculated as follows:

image

Results

Median estimates of OS and EFS in the 732 patients with cytogenetics available within 3 months of HDT were 42 months (95% CI 39–47 months) and 20 months (95% CI 18–22 months) respectively (Fig 1). Outcomes in this group of 732 patients with CA available within 3 months of HDT were similar to outcomes in the overall population of 1000 patients (data not shown).

Figure 1.

OS and EFS for the 732 HDT patients. Numbers in parentheses are the 95% confidence interval for the median estimate. The percentage of surviving patients is shown on the y axis. N = number of patients.

The clinical characteristics of this population are representative of patient referrals over the past decade to the MIRT. In patients with cytogenetic samples taken within 3 months of HDT, 141 of 732 patients (19%) exhibited one or more CA. Patients with these pre-HDT CA were significantly more likely to have an elevated β2m, lactate dehydrogenase (LDH) and CRP, and decreased albumin and haemoglobin. Patients with pre-HDT CA were less often responsive to prior SDT, had received more SDT and had a higher incidence of IgA isotype disease (Table II).

Table II.  Patient and laboratory characteristics.
Laboratory characteristicAll patients
n = 732
< 3 month pre-HDT samples
CA
n = 141
No CA
n = 591
  1. All characteristics presented differed significantly (P < 0·05) between patients with a CA and those without. Variables not statistically different between CA and no CA included: age, Durie–Salmon stage, creatinine and IL-6. BM, bone marrow.

IgA isotype21%33%20%
Responsive to prior Rx62%44%66%
Prior SDT ≥ 12 months41%47%36%
Albumin ≤ 0·0035 g/l41%52%32%
C-reactive protein ≥ 4·0 mg/l52%59%49%
Haemoglobin ≤ 10 g/dl26%42%22%
LDH (UI/l) ≥ 200 UI/l14%25%17%
Plasmacytosis BM biopsy ≥ 25%26%58%21%
Serum β2m ≥ 4·0 mg/l19%39%16%

The incidence of specific CA in the 732 patients with cytogenetics studies available < 3 months prior to HDT are presented in Table I. The most common CA were whole chromosome additions, specifically of chromosomes 3, 5, 7, 9, 11, 15, 19 and 21, that occurred in 32–54% of patients with CA. Approximately half of the patients were found to have CA13, most of which were whole or q-arm deletions. MM-MDS was found in 47 patients (33% of patients with CA); hypodiploid karyotype was found in 59 (42%) and hyperdiploidy in 90 (64%) of the CA patients. MM-MDS, comprising t(1;7), del 5q, − 5, del 7q, − 7, + 8 and del 20q, was sometimes associated with similar MDS lesions in separate metaphases (i.e. outside the context of a MM signature karyotype).

CA in MM typically involve a large number of altered chromosomes. In this group of 141 patients with pre-HDT CA, the number of chromosomes exhibiting CA ranged between one and 17, with a median of seven altered chromosomes. A clustering tree of the CA present in these 141 patients was developed to examine associations between CA (among the specific CA that were found in at least five patients; Fig 2). This tree shows the most apparent cluster to include the trisomies listed earlier (3, 5, 7, 9, 11, 15, 19, 21), except trisomy 8, which was associated with translocations involving chromosome 6q. Other apparent clusters included monosomy 5 and 7 (MM-MDS), translocations on 11q and 14q [11 of 22 patients had a t(11;14) translocation], and translocation 1q, monosomy 13 and jumping 1q. The high coincidence rate among trisomies (3, 5, 7, 9, 11, 15, 19, 21) could imply that one could use the presence of any of these trisomies as a prognostic marker, a possibility that is investigated later.

Figure 2.

CA clustering tree in 141 patients with pre-HDT CA.

Prognostic implications

Pre-HDT CA impart a poor prognosis for both OS and EFS (Fig 3). The median OS in the 141 patients with pre-HDT CA was 16 months, compared with 52 months in patients with no CA < 3 months prior to HDT (P < 0·001). The median EFS in the 141 patients with pre-HDT CA was 9 months, compared with 26 months in patients with no CA ≤ 3 months prior to HDT (P < 0·001). As data were available on the specific chromosome alterations that comprised a patient's abnormal karyotype, the prognostic implications of individual CA were then analysed among the 141 patients with CA (Fig 4).

Figure 3.

OS and EFS by presence of CA < 3 months pre-HDT. Numbers in parentheses are the 95% confidence interval for the median estimate. The percentage of surviving patients is shown on the y axis. N = number of patients.

Figure 4.

Kaplan–Meier median OS and EFS for individual CA and CA groups considered in the analysis. *indicates P < 0·05 in the univariate Cox model for OS compared with patients with other CA. **indicates P < 0·05 in univariate Cox model for EFS compared with patients with other CA. ***indicates P < 0·05 of both OS and EFS. The percentage of surviving patients is shown on the y axis. N = number of patients.

OS.Figure 4 presents the Kaplan–Meier calculated median OS according to both individual CA and CA groups of interest, presented in order from the longest to shortest OS. In univariate models, Del01p, T01q, Add08, MM-MDS, p-arm deletions, or q-arm translocations resulted in a worse prognosis when compared with patients with other CA at a P < 0·05 level. Only patients with a Add09 CA had a comparatively better prognosis in univariate models (P < 0·05) (Table III). Although the power to test outcomes among these CA is very low, considering CA occurring in less than 15 patients, the following had a worse OS in univariate models (P < 0·005): T06q (10 patients), del11q (four patients), T13q (nine patients), Del16q (five patients), T19q (seven patients) and T22q (seven patients). Thus, almost every q-arm translocation we considered for analysis was a marker for poor OS.

Table III.  Univariate and multivariate regression models in patients with abnormal karyotypes (n = 141).
 OSEFS
HR (95% CI)PHR (95% CI)P
  1. HR, hazard ratio; 95% CI, 95% confidence interval; P, regression model P-value.

Univariate
 Del01p1·7 (1·1, 2·6)0·015  
 T01q1·5 (1·1, 2·2)0·025  
 Add082·9 (1·7, 5·1)< 0·0012·0 (1·2, 3·5)0·011
 Add090·62 (0·4, 0·9)0·008  
 MM-MDS2·0 (1·4, 3·0)< 0·0011·4 (1·0, 2·0)0·051
 q-arm translocation1·5 (1·1, 2·1)0·0261·5 (1·1, 2·1)0·029
 p-arm deletion1·6 (1·1, 2·5)0·021  
 Any structural CA1·5 (1·1, 2·1)0·027  
Multivariate #1
 Add084·0 (2·2, 7·2)< 0·0012·1 (1·2, 3·7)0·009
 Add090·45 (0·3, 0·7)< 0·001  
 Del200·54 (0·3, 0·9)0·046  
Multivariate #2
 Add082·9 (1·5, 5·4)< 0·0012·1 (1·2, 3·7)0·009
 Add090·44 (0·3, 0·7)< 0·001  
 Del200·53 (0·3, 0·9)0·038  
 MM-MDS1·7 (1·1, 2·6)0·009  

First-year mortality in patients with CA was 39%vs 12% in patients with no CA (P < 0·001). Patients with a poor-risk CA for OS in univariate models also had higher first-year mortality rates when compared with patients with other CA: Del01p (61%vs 35%, P = 0·01), T01q (53%vs 33%, P = 0·02), Add08 (87%vs 34%, P < 0·001), MM-MDS (51%vs 34%, P = 0·05), p-arm deletions (57%vs 35%, P = 0·03) or q-arm translocations (48%vs 29%, P = 0·02).

EFS. Patients with Add08, Add18, T01q, Del13 or T11q had an estimated median EFS of < 8 months, while Add11, Add19 or Del16 CA had an estimated median EFS of > 11 months (Fig 4). Considering CA groups, patients with p-arm deletions, MM-MDS or q-arm translocations had an estimated median EFS of < 8 months. Only Add08, MM-MDS and q-arm translocations had a worse prognosis in univariate models (P < 0·05) (Table III). Similarly to OS, almost all q-arm translocations were markers for poor EFS in univariate models when also considering CA that occurred in less than 15 patients.

A 5-year EFS was rare in patients with CA present < 3 months prior to HDT; only two patients were alive and event free 5 years post the initiation of HDT. No patients with a poor-risk CA identified earlier for OS and EFS (i.e. Add08, MM-MDS, Del01p, Del13) remained event free at 5 years.

A multivariate analysis was first performed that examined only individual CA (Table III, multivariate #1). Add08 was identified as the common marker of poor prognosis for both OS and EFS. Add09 and Del20 were markers for a comparatively better OS. Inclusion of MM-MDS, ploidy status (hypo- versus hyperdiploid) and indicators of CA groups across chromosomes (p-arm deletions, q-arm deletions, p-arm translocations, q-arm translocations, any structural CA, any numeric CA) identified only MM-MDS as adding prognostic information to the OS model (multivariate #2).

Standard prognostic factors

We next examined whether associations existed between SPF and the specific CA identified earlier as prognostic markers for OS and/or EFS. All CA were associated with an increased incidence of disease that was resistant to standard therapy (P < 0·05). MM-MDS was associated with the IgA isotype and poor prognosis MM lab markers [elevated β2m, CRP, interleukin 6 (IL-6), and decreased albumin, P < 0·05].

Multivariate models combining CA and SPF

To examine the prognostic importance of the CA identified earlier in the complete group of patients, multivariate Cox models were developed in the overall population (n = 732) that combined SPF and CA (Table IV). For OS, any CA, MM-MDS and Add08 all had estimated hazard ratios greater than 2·0 even after adjusting for SPF. The presence of Add07 conferred an improved hazard (HR = 0·45) that could probably be expected of any of the trisomies that clustered together (see Fig 2). In fact, if an indicator of any trisomy from the cluster was used in the model instead of Add07, the estimated hazard ratio for any chromosome abnormality marker was 0·67 with a P-value of 0·077: an indication that most of the trisomies present in the cluster would be an independent predictor of improved survival. For EFS, any CA had an estimated hazard ratio of 2·0 and, within that group, MM-MDS again conferred a worse prognosis (HR = 1·5). The poor prognostic impact of MM-MDS could not be explained simply by the presence of Add08 prior to HDT, as MM-MDS remained a marker for poor prognosis even after adjusting for Add08 (Table IV). β2m ≥ 4 mg/l and prior SDT of a length of 12 months or more were independent predictors of poor prognosis post transplant for both OS and EFS. The Kaplan–Meier survival by CA group is presented in Fig 5.

Table IV.  Multivariate regression models combining CA and SPF in all patients (n = 732).
 OSEFS
Step*HR (95% CI)PStep*HR (95% CI)P
  • *

    The step at which that variable was entered into the stepwise model.

  • HR, hazard ratio; 95% CI, 95% confidence interval; P, regression model P-value.

Cytogenetics
 MM-MDS12·2 (1·4, 3·3)< 0·00161·5 (1·1, 2·2)0·033
 Any CA32·5 (1·9, 3·4)< 0·00112·0 (1·6, 2·6)< 0·001
 Add0760·45 (0·3, 0·7)< 0·001   
 Add0872·7 (1·4, 5·4)0·003   
SPF
 Age ≥ 65 years81·5 (1·1, 2·0)0·012   
 Albumin < 0·0035 g/l91·3 (1·1, 1·6)0·018   
 β2m ≥ 4 mg/l21·8 (1·4, 2·3)< 0·00121·5 (1·2, 1·9)< 0·001
 CRP ≥ 4 mg/l   41·3 (1·1, 1·5)0·003
 IgA myeloma   51·3 (1·1, 1·6)0·018
 IL-6 ≥ 3·5 pg/ml51·4 (1·1, 1·7)0·003   
 Prior SDT ≥ 12months41·6 (1·3, 1·9)< 0·00131·4 (1·1, 1·6)< 0·001
Figure 5.

OS and EFS by presence of the ‘poor-risk’ CA Add08 or other MM-MDS ≤ 3 months pre-HDT. Numbers in parentheses are the 95% confidence interval for the median estimate. The percentage of surviving patients is shown on the y axis. N = number of patients.

Prognostic importance of CA present ≥ 3 months prior to HDT

CA present ≥ 3 months prior to HDT imparted a worse prognosis than CA present at any time prior to HDT (median OS 16 months vs 27 months, median EFS 9 months vs 13 months; data not shown). As data were available in the group of patients with ≥ 3 months pre-HDT history of cytogenetics, we examined whether CA present prior to HDT, but not present < 3 months prior to HDT, were of a prognostic value in this group of patients. The group of patients with CA present < 3 months included those with and without CA present > 3 months prior to HDT. We identified 604 patients who had at least one cytogenetic study in both time intervals of interest (< 3 months prior to HDT, ≥ 3 months prior to HDT). In this group, 104 patients (18%) had CA present only < 3 months prior to HDT and 121 patients (21%) had CA present only ≥ 3 months prior to HDT. Incidence rates of specific CA in the group of 104 patients were similar to the 141 used earlier, and were similar in comparison to the incidence rates in the 121 patients with CA only ≥ 3 months prior to HDT.

Kaplan–Meier estimates of OS and EFS for each of the three groups (CA < 3 months prior to HDT, CA only ≥ 3 months prior to HDT, no CA) are presented in Fig 6. The median OS and EFS for patients with CA present only ≥ 3 months prior to HDT were 40 and 20 months respectively. Both medians were more than double those in the CA < 3 months median estimates, but were still significantly worse (P < 0·001) than patients with no CA at any time prior to HDT. Similar to the analysis of CA present < 3 months pre-HDT, trisomies of chromosomes 3, 5, 7, 9 and 11 were markers for a comparatively better OS when present only ≥ 3 months prior to HDT. In fact, the median OS for patients with these trisomies was very close to the median OS for patients with no CA at any time pre-HDT (data not shown). Deletions of chromosomes 6 and 13 were markers for a worse prognosis (P < 0·05) when present only ≥ 3 months prior to HDT, with a median OS of approximately 26 months. Thus, even though a patient may present with no CA < 3 months prior to HDT, knowledge of the cytogenetic history at earlier time points can still be of prognostic value. However, it is clear that CA present < 3 months prior to HDT impart a far worse prognosis than any CA that may have been present at earlier times.

Figure 6.

Comparison of CA present ≤ 3 months pre-HDT (including patients with and without CA > 3 months prior to HDT) to CA present only > 3 months pre-HDT in 604 patients. Numbers in parentheses are the 95% confidence interval for the median estimate. The percentage of surviving patients is shown on the y axis. N = number of patients.

Discussion

Among the haematological malignancies, MM is associated with the most profound genetic instability, resembling the findings in solid tumours. Although a few recurring, mainly 14q, abnormalities can be recognized by interphase FISH, which have adverse prognostic implications in cases of t(4;14) and superior outcome in patients with t(11;14) (Moreau et al, 2002), the mere detection of metaphase CA is a well-known adverse laboratory feature in MM.

We and others have previously reported on the particularly poor risk associated with CA13 and have confirmed the adverse consequences of hypodiploidy (Smadja et al, 2001; Fassas et al, 2002), when detected in metaphase spreads at diagnosis. The presence of CA13 or hypodiploid CA at relapse was associated with a median survival of only 6 months compared with 24 months in the presence of other CA, and 60 months when CA were absent at the time of relapse (Barlogie et al, 2002; Shaughnessy et al, 2003a). The adverse consequences of CA13 also pertained when the entire cytogenetic history of 1000 consecutive patients enrolled in tandem transplants was considered (Desikan et al, 2000). The current examination differs from our previous work (Tricot et al, 1995) in that the pre-HDT observation window was limited to 3 months in order to facilitate the comparison with other trials enrolling patients, at various stages of their disease, in autograft-supported HDT programs. In this study, we have also undertaken one of the first attempts to account for the entire complexity of individual CA. Not surprisingly, CA present within 3 months of HDT carried a greater risk compared with CA that were present at an earlier time point but absent within 3 months of HDT. We interpreted this observation to indicate a rapidly fatal course in those patients whose CA survive either pre-HDT induction or salvage therapies.

The consideration of individual CA, and adjusting for SPF, revealed that MM-MDS (and particularly trisomy 8 within a MM signature karyotype) imparted a poor EFS and OS. We coined the term ‘MM-MDS’ to reflect the presence of a CA typical of MDS within the context of a MM signature karyotype, especially as some patients, in addition to normal diploid metaphases, also exhibited MDS CA only in separate metaphases that usually concurred with those present within the MM-MDS lesions (Sawyer et al, 1998a). The observation of MDS lesions in normal haematopoietic and MM cells, which are at least morphologically identical, is consistent with the theory that this terminally differentiated B-cell neoplasm is susceptible to similar genetic lesions to those occurring in normal haematopoiesis.

Examination of the potential clinical associations revealed the MM-MDS patient group to be the only one associated with six adverse SPF. The apparently higher frequency of MM-MDS than MDS prior to HDT [47 patients (6%) vs six patients (1%), P < 0·001] suggests a greater sensitivity of the malignant MM population than of normal haematopoietic cells to a common injury. Alternatively, especially in patients with a limited prior drug exposure, genetic damage to a common precursor cell can be hypothesized, which is in keeping with occasional case reports. Scrutiny of normal and MM cells by interphase FISH will be critical to determine the relative frequencies of specific MDS lesions within both normal haematopoietic cells and MM cells, and thus clarify their underlying biology.

Although still conferring a high risk, CA13 (del 13 and del 13q) detected within 3 months of HDT was not an independent risk factor once MM-MDS was accounted for (see Fig 4 and Table III). However, among patients with CA only detectable at an earlier time point but not within 3 months, del 13, and del 6, were markers of a worse prognosis.

The notion that most trisomies (except at 08) clustered together and imparted superior survival when examined in the context of SPF (trisomy 7) was surprising (see Table IV). When examined only in the context of the 141 patients with CA, both add 09 and del 20 conferred a favourable clinical outcome (see Table III).

In light of universal aneuploidy recognized by interphase FISH, we previously postulated that the presence of CA in approximately one-third of untreated MM characterized a tumour that could proliferate and divide in vitro in the absence of stromal cell support, and thus reflect autonomously proliferating MM cells imposing a poor survival in vivo (Shaughnessy et al, 2003b). However, the detection of the beneficial implications of certain CA such as trisomies, which also had to survive HDT-preceding therapy and perform successful mitotic division in vitro without stromal support, suggests that such lesions containing such CA are exquisitely sensitive to high-dose melphalan, which was the HDT regimen employed, in vivo. By contrast, MDS lesions, typically induced by a prolonged low-dose alkylating agent therapy such as melphalan, would be expected to survive even high-dose melphalan and thus account for the poor prognosis associated with MM-MDS.

The routine application of gene array analysis (Zhan et al, 2002) over the past 2 years, both prior to initiation of therapy and at relapse, along with in vivo follow-up studies within days of single-agent therapy (Shaughnessy et al, 2002), should help to reveal genes that distinguish MM-MDS from other CA, and help clarify the molecular mechanisms associated with good-risk CA.

Compared with SPF, most CA pose a much greater hazard that exceeds the relative risk associated with interphase-FISH-defined abnormalities. We hence recommend that cytogenetic studies should be an integral part of the evaluation of patients with MM, both at diagnosis and at critical therapeutic decision points. When performed in the context of complimentary genetic investigations (interphase and metaphase FISH, gene expression profiling), the molecular correlates of high-risk CA can hopefully be revealed (akin to the BCR-ABL gene fusion in Philadelphia-chromosome-positive acute lymphoblastic and acute myeloblastic leukaemia) and appropriate targets identified for more effective therapeutic intervention.

Acknowledgments

This work was supported in part by research grant CA55819 from the National Cancer Institute in Bethesda, MD, USA.

Ancillary