Professional Documents
Culture Documents
Goasguen 1999
Goasguen 1999
www.elsevier.com/locate/leukres
Abstract
The relationship between plasmablastic cells and outcome in multiple myeloma (MM) has been established for nearly 15 years.
But the assessment of these cells is not easy to perform and it allows the identification of only a small proportion of patients. We
investigated the plasma cell morphology using a progressive evaluation of consecutive criteria: nucleolus, chromatin and
nuclear-cellular ratio (N/C). The combination of these three items produces a subclassification where four cellular subtypes
identify 93% of the plasma cells, and these subtypes are related to the outcome. The interest of this methodology is to be based
on the mature plasma cells that are easier to identify than the plasmablastic cells. These new cell subtypes introduce a new
classification for patients: Group 1 includes patients with at least 66% mature plasma cells (P000). Both Group 2 and 3 have less
than 66% P000 and are separated by their degree of maturation (Proplasma I ] Proplasma II + plasmablastic). The distinction of
these three groups of patients is highly related to the prognosis (PB10 − 4). These results have been confirmed on a second group
of patients coming from a different institution. In conclusion, we propose a new methodology for the plasma cell evaluation in
MM, that is based on the morphological criteria and that has the advantage of identifying an intermediate (30%) subgroup of
patients with a prognostic significance. © 1999 Elsevier Science Ltd. All rights reserved.
0145-2126/99/$ - see front matter © 1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 5 - 2 1 2 6 ( 9 9 ) 0 0 1 3 2 - 0
1134 J.E. Goasguen et al. / Leukemia Research 23 (1999) 1133–1140
Fig. 1. Construction of algorithm for automatic subclassification of plasma cells in MM. The three questions are consecutives, and the answers
are yes (1) or no (0). Four subtypes represented 93% of the cells. These more frequent subtypes can be easily recognised by direct observation.
Since the three questions are consecutive, one cell can and compared by the logrank test. The confident inter-
be defined as ‘P111’ when the plasma cell (P) is defined val was defined as 95% and the lower limit for significa-
by ‘presence of nucleolus’, ‘presence of blastic chromatin’ tion is 0.05.
and ‘N/C ratio\ 0.6 ’. This cell is very similar to the
description of a plasmablastic cell [3,8]. On the con-
trary, a cell defined by P000 is a mature plasma cell 3. Results
(Marschalko cell) since there is no nucleolus (0), no
blastic chromatin (0), and no N/C \ 0.6 (0). This al- Among the 92 patients included in Study 1, two early
gorithm is represented on Fig. 1 where finally eight deaths were observed and excluded for statistical
subtypes are defined. These subtypes are represented in comparison.
a synoptic table (Fig. 2) and illustrated in Fig. 3 to
explain their significance. 3.1. Response to treatment
2.3.2. Method for bone marrow plasma cell e6aluation Every patient was classified as ‘responder’ (R) and
The morphological evaluation was performed on ‘non responder’ (NR). Forty-two patients had a regres-
bone marrow smear, 100 consecutive plasma cells were sion of the disease (responder group) but 34 were
analysed for these three criteria. But it is clear that the evaluated as non-responders. Twelve patients had to
morphologists did not identify the height subtypes by stop their treatment for medullar toxicity and two cases
using their pseudonyms (P111, P110 and so on). They were lost for this information.
were working using worksheets where they just had to Response to treatment is demonstrated to be inde-
put 1 or 0 for each of the three questions, cell by cell. pendent of the morphologic subtypes. The percentage
Some morphologists were working using a short pro- of cells is somewhat different when considering P110
gram on a computer. The calculator produced the (R, 11.8%; NR, 5.4%) and P000 (R, 54.2%; NR,
proportion of each subtypes as soon as 100 cells were 64.9%), but is not statistically significant. It is also
counted. observed that most of the time, only three subtypes are
generally identified: P000, P100 and P110. The overall
2.4. Statistical methods survivals (R, 43.3 months; NR, 42.4 months), survival
curves and risk of death were very similar for both
The percentage of cells with one criteria (for exam- groups.
ple, presence of nucleolus) can be sorted by increasing
value, demonstrating two waves clearly separated by a 3.2. Relationship between delineated subtypes and
cut-off point. This cut-off was at 25% for the nucleolus classical morphology
and 66% for P000 in our two series.
All data were recorded using Dbase Software and The mature plasma cell is included in the P000
then transferred to PCSM Software (DeltaSoft, France) subtype: nucleolus is absent, chromatin is clumped (in
for statistical calculations and survival studies. Survival blocks), and the cytoplasm is abundant. The P100
curves were established by the Kaplan-Meier method subtype is similar to P000 but shows one nucleolus
1136 J.E. Goasguen et al. / Leukemia Research 23 (1999) 1133–1140
Fig. 2. Synoptic representation of plasma cell subtypes as defined by algorithm system. P, plasma cell; first number represents presence (1) or
absence (0) of nucleolus; second number represents presence (1) or absence (0) of blastic chromatin; last number represents presence (1) or absence
(0) of N/C ratio\ 0.6.
surrounded by blocks of chromatin (P100 was com- 3.3.3. Percentage of P000 (mature plasma cells)
monly called Proplasma cell I in our group). P110 has The presence of P000 is discussed for more or less
a nucleolus and a blastic chromatin, but cytoplasm is two third of P000 cells. Forty-eight cases had less than
large and we called it, Proplasma cell II. PB are usually 66% P000 and a survival duration of 31 months com-
defined as having a nucleolus but some of them may be pared to 42 cases with more than 66% P000 and 52
considered as PB, even without displaying a nucleolus, months survival time (P= 0.0050). When the group is
when the chromatin is blastic and the cytoplasm re- restricted to only responders (39 cases) and non-respon-
duced. These cells are called P011 according to the ders (37 cases) the median survivals are 31 and 60
present algorithm. months respectively (P=0.0050) Fig. 5.
Fig. 3. Synoptic representation of plasma cells as defined by new algorithm system. Cells with nucleolus are represented by P111, P110, P101,
P100. Cells with blastic chromatin are defined as P111, P110, P011, P010 and cells with very high N/C ratio (\ 0.6) are P111, P101, P011, P001.
ments evaluation we defined three groups of patients: vivals are 46, 26 and 10 months for 36, 19 and 7 cases
Group 1: P000] 66%. Then for patients with P000B respectively for Group 1, Group 2 and Group 3 (P B
66%, two groups are distinguished: Group 2: P100] 0.0001) Fig. 8.
P110+P111, and Group 3: P100B P110 + P111.
Results are displayed in Table 3 and survival curves are
shown in Fig. 7.
Table 1
Discrimination of the survival duration according to the percentage
of plasmablastic cells as evaluated by P111 (B2 or ]2%) for all cases
(A, 90 cases) and cases restricted to responders and non responders
(B, 76 cases)
P111B2% P111]2% P
Fig. 4. Survival curves for Series 1 patients restricted to responders
A n 59 31 and non-responders (76 cases). Fifty-two cases with P111 B 2% (me-
Median survival (months) 54 20 B10−4 dian 56 months) are compared to 24 patients with P111 \ 2% (me-
B n 52 24 dian 21 months). PB0.001. When all patients from Series 1 (90 cases)
Median survival (months) 56 21 B10−4 are included, medians are 54 months and 20 months, respectively
(PB0.0001).
1138 J.E. Goasguen et al. / Leukemia Research 23 (1999) 1133–1140
Fig. 5. Survival curves for Series 1 patients restricted to responders Fig. 7. Survival curves for Series 1 patients restricted to responders
and non-responders (76 cases). Thirty-nine cases with P000 B66% and non-responders (76 cases). Group 1: 39 patients P000 \66%
(median 31 months) are compared to 37 patients with P000 \66% (median 61 months); Group 2: 29 patients P000 B66% and P100\
(median 60 months). P=0.0050. Any change is observed when all P111 + P110 (median 32 months); Group 3: 8 patients P000 B 66%
patients from Series 1 (90 cases) are included. and P100B P111 + P110 (median 19 months). P value for these three
groups is B 0.0001 and is unchanged when all patients from Series 1
are included (90 cases: Group 1, 44 cases, median 52 months; Group
4. Discussion
2, 35 cases, median 32 months; Group 3, 11 cases median 20 months).
Fig. 6. Survival curves for Series 1 patients including responders and Fig. 8. Survival curves for Series 2 patients including 62 cases from a
non-responders and comparing patients associating P111 B 2% and single institution. As upper defined: Group 1, 36 cases, median 46
P000 \ 66% (39 cases, median 60 months) with patients having months; Group 2, 19 cases, median 26 months; Group 3, 7 cases,
P111 \ 2% and P000B 66% (22 cases, median 21 months) PB0.0001. median 10 months; PB0.0001.
J.E. Goasguen et al. / Leukemia Research 23 (1999) 1133–1140 1139
Table 2
Comparison of survival curves for multiple myeloma patients in two different studies (1 and 2) according to the proposed reclassification (Group
1, P000]66%; Group 2, P000B66% and P100\P110+P111; Group 3, P000B66% and P100BP110+P111)
P100\P110+P111 P100BP110+P111
Group 1 Group 2 Group 3
Study 1 Group Aa n 44 35 11
Median survival (months) 52 32 20 B10−4
b
Group B n 39 29 8
Median survival (months) 61 32 19 B10−4
Study 2 n 36 19 7
Median survival (months) 46 26 10 B10−4
a
All 90 patients from Study 1.
b
Restricted to responder and non responder from Study 1.
through all the countries in the world. Moreover this the algorithm explains why the myeloma cell morphology
method allows the identification of a group including a is so difficult to assess.
reduced number of patients, which represents only Our study reproduces some of the results already
around 10% [3,8] of the MM. Finally any significant demonstrated [3,8] because P111 and PB cells are very
difference between the survival of the mature, intermedi- close. As shown in Table 1 or Fig. 3, we found a high
ate and immature subgroups had also been demonstrated correlation (P B 0.0001) between survival and P111 per-
meaning that this classification is restricted to worse centage meaning that this morphological evaluation is at
patients. least as consistent as those already published.
For these major reasons we decided to build up a new To compare the impact of each criteria on the al-
approach for the evaluation of plasma cell morphology gorithm we defined a critical value for them, keeping in
based on criteria which would be easily evalu-ated and mind that the definition of a ‘cut-off’ had to be simple,
would produce a positive (not by exclusion) evaluation and useful. A cut-off at more or less than 25% cells with
for all cells. Taking into account all possible morpholog- nucleolus, is definitely significant, but the P value is only
ical criteria (all those already described plus: position of 0.0330. Each criteria by itself has a significant value but
the nucleus; existence of golgi-zone; size of nucleus; the inclusion of three criteria for the construction of the
aspect of chromatin) we could not obtain a valuable algorithm produces the best result, and the algorithmic
classification is drawn by the nucleolus and the chro-
reproducibility between the five observers. We excluded
matin aspect.
some of them to keep only those that produced a valuable
The key point of the preceding reports [3,8] is the
concordance (\85% of cells for five morphologists).
recognition of PB based on the notion that it is an index
To avoid the situation where cells are defined by
for immaturity. In our system, the immaturity can be
exclusion, we decided to evaluate only the criteria and not
assessed by P111 or P111+ P110 and so on. We tested
the cell, producing this algorithm. The advantage of the
different combinations based on the immaturity notion,
algorithm is that all cells are identified (because at the but the most sensitive was firstly to consider the mature
first step they do not have a name). However four compartment (which can be most easily evaluated). Since
subtypes represent 93% of the cell population (P111+ their is a clear cut-off for patients with more or less than
P110+P100+ P000). These cells are currently and easily two third of P000, we firstly defined the Group 1 for all
recognised since P111 is somewhat similar to PB (PB cases with P000 ] 66%. Then, (for P000 B 66%) the index
should be P111 +P011). P110 is very close to ‘immature of immaturity can be evaluated for two different situa-
myeloma cells’ [3] that we called Proplasma II, and P100 tions: Group 2, P100\ P110+ P111 versus Group 3,
is a mature plasma cell but with a nucleolus (so-called P110+ P111\ P100. These two groups can be easily
Proplasma I). The second advantage of the algorithm is distinguished because this is the situation where Pro-
the recognition of 7% of cells that do not have any name. plasma I (mature with nucleolus) are more or less
These cells (P101, P011, P010, P001) originate the frequent than all plasma cells having a blastic chromatin
discordance between observers because they do not fit and a nucleolus (Plasmablastic+ Proplasma II).
with any classification. In some cases, these ‘aberrant’ The definition of these three subgroups is very sensitive
plasma cells may represent more than 20% of the for the survival (P B 0.0001). The advantage of this
proliferating population (personal observation). Of proposal is that if the PB are underestimated or missed,
course, it was not in our goal to give them a name and they will be fused with P110 but the notion of immaturity
this was not necessary with the proposed procedure. But will be preserved.
1140 J.E. Goasguen et al. / Leukemia Research 23 (1999) 1133–1140
In most of the cases, the percentage of P101, P011, project, assisted in data interpretation, drafting, revis-
P010 and P001 is very low. Since we demonstrated the ing and gave final approval to the paper. C. Mathiot
impact of the immaturity, we reclassified P101 and contributed to the concept and design, assisted in the
P011 with the immature compartment (P111+P110) analysis of data, helped in the drafting of the paper,
and we gathered P010 and P001 with P000. The num- critical revision and gave final approval. J.M. Scheiff,
ber of these cells being generally so small, this gathering M. Bizet, and B. Ly-Sunnaram contributed to the
did not produce any change in survival curves, median concept, design, data analysis, provided study materials
survival or statistical tests. and gave final approval to the article. B. Grosbois
To be sure that the five observers did not deviate in helped to assemble the data, contributed to the concept
their evaluation, a control study was tested by a single and design and gave fund approval to the paper. M.
reviewer (JG) using the same methodology. Even Monconduit contributed to the concept and design,
though the median survivals are somewhat different, provided study materials and gave final approval. J.-L.
(Fig. 8), the sub-classification in three different groups Michaux contributed to the concept and design and
is highly significant (P B0.0001), providing also the gave final approval.
advantage to identify an intermediate subgroup repre-
senting 30% of patients (Group 1, 58%; Group 2, 30%;
Group 3, 11%). References