Adverse prognostic significance of CD20 positive Reed–Sternberg cells in classical Hodgkin's disease*


  • *

    Presented in part at the American Society of Hematology, December 1999.

Carol S. Portlock, Memorial Sloan–Kettering Cancer Center, 1275 York Avenue, New York, NY 10021, USA.


The prognostic significance of CD20 positive classical Hodgkin's disease (cHD) is uncertain. All cHD cases referred to the Memorial Sloan–Kettering Cancer Center (MSKCC) were retrospectively identified (5/92–11/00); the samples were immunostained, and clinical data ascertained. Cases were re-reviewed without knowledge of clinical outcome. Univariate and multivariate analyses were performed 248 patients had cHD: 28 CD20+ (11%); 220 CD20. All clinical characteristics were comparable except haemoglobin level at presentation. With a median follow-up of 29·2 months, significant prognostic factors in multivariate analysis were: CD20 positivity, elevated white blood cell count (WBC) and low absolute lymphocyte count for time-to treatment failure (TTF); and for overall survival (OS), CD20 positivity, elevated WBC count, bone marrow involvement and age ≥45 years. TTF was significantly poorer for ABVD-treated patients with CD20+ cHD as compared with CD20 cHD. Among 167 patients treated at MSKCC, both TTF (P < 0·0001) and OS (P = 0·017) were significantly decreased in CD20+ patients as compared with CD20 cHD. CD20+ cHD is a poor prognostic factor for TTF and OS. All cHD cases should be immunophenotyped for CD20. A large prospective trial is needed to confirm these findings.

Hodgkin's disease (HD) is now a highly curable lymphoma in which the thrust of clinical research has shifted to the reduction of therapy and its side effects in good risk patients, and the identification of those at high risk who require novel approaches.

The World Health Organization (WHO) pathology classification (Harris et al, 2000) recognizes two HD subsets: nodular lymphocyte predominant HD and classical HD (cHD). Although both are B-cell derived lymphomas, their histopathology, immunophenotyping, molecular genetics, clinical presentation and outcomes are different. Consequently, management recommendations now reflect this distinction and the histological subset at diagnosis has become an increasingly important prognostic factor.

Within cHD, however, the histological subsets of nodular sclerosis (NSHD), mixed cellularity (MCHD), lymphocyte-rich, and lymphocyte-depleted, have some differences of clinical presentation but management recommendations are not significantly different based upon the histological subset. Rather, this large group of patients (approximately 95% of all HD) is managed according to clinical presentation parameters such as stage, disease bulk and systemic symptoms. Attempts to identify poor prognostic risk groups in which initial therapy might be substantially altered to improve outcome (such as upfront autologous transplantation), have not proven successful to date (Kewalramani & Moskowitz, 2001; Andrieu et al, 2002; Federico et al, 2003).

In our experience at Memorial Sloan–Kettering Cancer Center (MSKCC), one histologic group that appeared to have a particularly poor outcome with conventional HD initial therapy was cHD with Reed–Sternberg (HRS) cell positivity for the CD20 antigen. Our initial preliminary retrospective report showed that 9% of cHD cases had detectable CD20 positive HRS cells (Donnelly et al, 1999). Based on these initial observations, a comprehensive retrospective study was undertaken, asking whether the presence of Reed–Sternberg cell CD20 positivity in cHD has prognostic significance when conventional therapy is utilized.

This analysis reports the data of 248 previously untreated patients (28, CD20 positive; 220, CD20 negative) diagnosed at MSKCC from May 1992 to November 2000.


Study population

The reports of all the cHD cases reviewed by the Department of Pathology at the MSKCC between May 1992 and November 2000 were retrospectively examined. For the purposes of this study, cHD included the nodular sclerosing, mixed cellularity and lymphocyte-depleted Rye classification subtypes. The initial diagnostic biopsy of this consecutive cohort of patients was stained by immunohistochemistry at our institution and clinical data was ascertained. Cases with evidence of composite histology were excluded. Patients with a known history of human immunodeficiency virus infection were excluded. All cases were re-reviewed and confirmed by a single pathologist without knowledge of clinical outcome.

Pathology and immunochemical procedures

Tissue samples were fixed in formalin and embedded in paraffin, cut in to 4 μm sections and mounted in super frost/Plus (Fisher) slides. The sections were deparaffinized in xylene and rehydrated in graded ethanol. For epitope retrieval, the tissue sections were heated in 0·01 mol/l citric acid buffer (pH 6·00) in a microwave oven for 30 min at high power, to bring the temperature of the solution to 90–95°C, and cooled at room temperature for 30 min.

Endogenous peroxidase was blocked by incubation in 3% hydrogen peroxide in phosphate-buffered saline (PBS) (pH 7·4) at room temperature for 10 min. After washing in PBS, the slides were placed in a humid chamber and incubated with non-immune suppressor serum for 10 min to block non-specific protein-finding sites. The tissue sections were then incubated overnight with L26 monoclonal antibody specific for the CD20 epitope (Dako Cytomation California Inc., Carpinteria, CA, USA) at a dilution of 1:1000 (immunoglobulin in a concentration of 0·03 μg/l) in antibody diluent (Dako). The slides were washed in PBS and 0·001% Triton X PBS three times and then incubated in biotinylated horse anti-mouse serum [1:500 in 0·5% bovine serum albumin (BSA)/PBS], washed in PBS four times and incubated in peroxidase-conjugated streptavidin (1:1000 0·5% BSA/PBS). After 15 min incubation the slides were again washed in PBS four times and kept in 0·01% Triton ×/PBS. Peroxidase colour was developed with 3,3′-diamino-benzidine (DAB) and hydrogen peroxide (Dako), and counter stained with haematoxylin. External controls were run with each section, (lymph node with follicular hyperplasia). Other cells in the sections serve also as internal controls.

Criteria for CD20 positivity

Typically, since most cases were of nodular sclerosis type, Reed–Sternberg cells and variants formed ill-defined clusters in the centre of the nodules when immunostained for CD30. In each case, we took the cytologic features and frequency of CD30 positive cells in these clusters and compared them with similar cells seen in the subsequent sections stained for CD20. Most cells in the CD30 stained sections showed the characteristic membrane, Golgi zone positive staining, or both (Fig 1). In those cases that were CD30 positive, we found only membrane with occasional Golgi zone staining in the corresponding sections stained for CD20 (Fig 2). We did not attempt to count the percentage of CD30 positive cells in each field that were also positive for CD20, since these were variable in different areas and we could not exclude contaminating immunoblasts, which can be positive for CD30, CD20 or both. These counts can be highly subjective in biopsies that ranged from small fragments to entire lymph nodes. By these criteria, most cells (more than 50%) were positive for CD20 in 25 cases. In three biopsies, there were only one or two positive clusters of CD20 positive abnormal cells, and these were classified as cases with focal positivity for this marker.

Figure 1.

Cluster of Reed–Sternberg cells and variants, positive for CD30. Original magnification 400×; inset, 600×.

Figure 2.

Cluster of Reed–Sternberg cells and variants positive for CD20. Note the membrane and occasional Golgi zone staining only. Original magnification 400×; inset, 600×.

Clinical data

The medical records of all patients were retrospectively reviewed and relevant clinical data, including extent of disease, pretreatment laboratory values, treatment and outcomes, were obtained. Advanced stage disease was defined as Ann Arbor stage IIX, IIIB, or IV. Wherever possible, risk groups were assigned to patients with advanced stage disease according to the Hasenclever (Hasenclever & Diehl, 1998) and Straus (Straus et al, 1990) models. Since MSKCC is a referral centre for relapsed or refractory cHD, the distinction was made whether initial treatment was given at MSKCC or at outside institutions to minimize selection bias.

The characteristics of the patient database are outlined in Table I.

Table I.  Patient characteristics.
Total number of patients248 
 Male 17106
 Female 11114
Age (years)
 Median 33 
 Range 12--79 
Age stratified (years)
 <45 25168
 ≥45  352
 Normal 11118
 >Normal  847
Extranodal sites involved
 ≤1 sites 16144
 >1 site 1160
 I/II/III 21176
 IV  744
Bone marrow
 Uninvolved 23208
 Involved  511
 >2  848
 ≤2  854
B symptoms
 Presence 1693
 Absence 12120
 NSHD 24178
 MCHD  337
 >10·5 g/dl 13137
 ≤10·5 g/dl 1045
 <40 g/l 1363
 ≥40 g/l  8107
 <0·6 × 109/l  622
 ≥0·6 × 109/l 14155
 >15 × 109/l  332
 ≤15 × 109/l 19149
 Involved  218
 Uninvolved 26196
Bulky disease
 Yes 1472
 No 14144
First treated at MSKCC
 Yes 15153
 No 1367
 Yes 12123
Follow up time (months)


All patients had a routine initial evaluation including history and physical examination, routine blood work, torso imaging with chest X-ray and computed tomography, and selected bone marrow biopsy. Nuclear imaging was performed as indicated. Staging designations are those of the Cotswolds modification of the Ann Arbor stage.


Standard regimens with or without adjuvant radiation therapy (according to MSKCC guidelines) were utilized in the majority of cases: adriamycin, bleomycin, vincristine, dacarbazine (ABVD) in 137, Stanford V (mechlorethamine, adriamycin, vinblastine, vincristine, bleomycin, etoposide, prednisone) in 57, MOPP (methotrexate, Oncovin, procarbazine, prednisone)/ABVD in 10, MOPP/adriamycin, bleomycin, vinblastine (ABV) or cyclophosphamide, oncovin, procarbazine, prednisone (COPP)/ABV in 12, MOPP or C-MOPP (MOPP + cyclophosphamide) in 3, bleomycin, etoposide, adriamycin, cyclophosphamide, vincristine, procarbazine, prednisone (BEACOPP) in 3, cyclophosphamide, doxorubicin, vincristine, prednisone (CHOP) or CHOPE (CHOP + etoposide) in 8, other chemotherapy in 13, and radiotherapy alone in 4. There were two in whom the details of the initial therapy were not available, and one who refused therapy.

Statistical analysis

Time to treatment failure (TTF) was measured according to the method of Kaplan and Meier, from the beginning of treatment to primary treatment failure, relapse, or last follow-up. The overall survival time (OS) was measured from the beginning of treatment to death or last follow-up. Actuarial Kaplan–Meier OS and TTF survival curves were calculated using the SAS statistical software program (SAS/STAT Software, 1997).

In the univariate analysis, the log-rank test was used to test the association between clinical or laboratory variables and TTF or OS. Factors that were potentially predictive of TTF or OS (P < 0·1) were entered into a multivariate analysis using the Cox proportion hazards model. The association between CD20 expression and clinical or laboratory parameters were examined by using Fisher's exact test. The non-parametric Mann–Whitney test was used to evaluate the correlation between patient age and CD20 expression. P-values <0·05 were used to define statistical significance. All statistical analyses were performed utilizing Splus version 2000 and the SAS statistical software program.


There were 248 patients identified in this retrospective review: 28 were CD20+ (11% of all patients); 220 were CD20. All characteristics evaluated, except haemoglobin (Hb) concentration ≤10·5/dl (which was significantly more common among CD20+ cases), were comparable in the two groups. The factors that were analysed are listed in Table I: sex, age, lactate dehydrogenase (LDH; normal versus abnormal), number of extranodal sites (0–1 vs. >1), stage (early: I, II, IIIA versus advanced: IIX, IIIB, IV), bone marrow status (uninvolved versus involved), Hasenclever score (>2 vs. ≤2), B symptoms (absence versus presence), histopathologic subtype (NSHD versus MCHD), Hb concentration (>10·5 vs. ≤10·5 g/dl), albumin level (≥40 vs. <40 g/l), absolute lymphocyte count (ALC) (≥0·6 vs. <0·6 × 109/l), white blood cell count (WBC) (≤15 vs. >15 × 109/l), inguinal adenopathy (uninvolved versus involved), bulky disease, as defined in the Cotswolds staging (absent versus present), institution of first therapy (MSKCC versus other), and first chemotherapy (ABVD versus other regimen).

Actuarial analysis

The median follow-up for all patients was 29·2 months, with a range of 0·47–110·9 months. For the CD20 positive group, the median follow-up was 25·1 months; and for the CD20 negative group, 29·7 months. The follow-up was not significantly different for the two groups (P = 0·112).

TTF and OS were evaluated for each of the characteristics listed in Tables II and III, respectively. For all evaluable patients, negative factors that were significantly different in univariate analysis of TTF included CD20 positivity (P < 0·0001), presence of B symptoms (P < 0·001), more than one extranodal site (P = 0·036), advanced stage (P = 0·063), elevated WBC (P = 0·012), reduced Hb (P < 0·001), low ALC (P = 0·016), low albumin (P = 0·007), and elevated erythrocyte sedimentation rate (ESR) (P = 0·033).

Table II.  Univariate analysis of time to treatment failure.*
  1. *Variables with P > 0·1 are not listed.

B symptom0·0003
Bone marrow0·074
Extranodal site0·036
Advanced stage0·063
Elevated WBC0·012
Reduced Hb0·001
Low ALC0·016
Low albumin0·007
Elevated ESR0·033
Table III.  Univariate analysis of overall survival.*
  1. *Variables with P > 0·1 are not listed.

Age ≥45 years vs. <45 years0·001
NSHD histology0·005
Inguinal involvement0·099
Bone marrow0·048
Extranodal site0·031
Stage IV0·044
Advanced stage0·10
Elevated WBC0·053
Reduced Hb0·006
Low ALC (0·6 × 109/l)0·01
Low albumin (40 g/l)0·06

Significant negative factors for OS in univariate analysis included: age ≥45 years (P = 0·001), NSHD histology (P = 0·005), CD20 positivity (P = 0·035), positive bone marrow (P = 0·048), number of extranodal sites (P = 0·031), stage IV disease (P = 0·044), elevated WBC (P = 0·054), reduced Hb (P = 0·006), and low ALC (P = 0·010).

A step-wise regression analysis was performed to determine the most significant variables among the patient characteristics listed in Table I. For TTF, the final model included CD20 positivity, elevated WBC and low ALC. By using CD20 negativity as a reference, the hazard ratio (HR) for CD20 positivity was 4·393 [95% confidence interval (CI): 2·255–8·557]. The HR for WBC is 2·235 (elevated WBC as the reference) (95% CI: 1·204–4·150). The HR for low versus high ALC was 1·869 (95% CI: 0·971–3·599). For OS, the final model also included CD20 positivity, elevated WBC, bone marrow and age ≥45 years. Again, by holding CD20 negativity as a reference, the HR for CD20 positivity was 3·331 (95% CI: 1·029–10·788); the HR for elevated versus no elevated WBC was 3·299 (95% CI: 1·296–8·393); the HR for positive versus negative bone marrow was 3·373 (95% CI: 1·062–10·715) and the HR for older age versus younger age was 6·015 (95% CI: 2·383–15·186) (Tables IV and V).

Table IV.  Multivariate analysis of time to treatment failure.
VariableHazard ratio95% CI
Low ALC1·8690·971–3·599
Table V.  Multivariate analysis of overall survival.
VariableHazard ratio95% CI
Elevated WBC3·2991·296–8·393
Age ≥45 years vs. <45 years6·0152·383–15·186
Bone marrow3·3731·062–10·715

In addition to patient characteristics, the effect of treatment regimen on outcome was also analysed according to CD20 status (negative versus positive). For those patients receiving ABVD and those receiving other regimens, the TTF was significantly poorer for patients with CD20 positive cHD, compared with those with CD20 negative cHD (P < 0·0001 for patients treated with ABVD and P < 0·003 for patients treated with other regimens). OS was also significantly poorer for ABVD-treated patients (P = 0·020), but not for those receiving other regimens (P = 0·410). On the other hand, there were marginally significant differences in TTF or OS among patients receiving salvage autologous transplantation according to CD20 status (P = 0·060 for TTF and P = 0·070 for OS).

Among the 167 patients that were initially treated at MSKCC, CD20 positivity resulted in a significantly decreased TTF as compared with CD20 negative cHD (P < 0·0001), and OS was also decreased (P = 0·017). Similar findings were determined for patients receiving initial treatment elsewhere (P = 0·002 for TTF and P = 0·791 for OS) (Figs 3–5).

Figure 3.

Time to treatment failure for all patients (pts) initially treated at MSKCC, according to CD20 status.

Figure 4.

Overall survival for all patients (pts) initially treated at MSKCC, according to CD20 status.

Figure 5.

Time to treatment failure for all patients (pts) initially treated at MSKCC and initially receiving ABVD, according to CD20 status.


This retrospective comparison of newly diagnosed patients with cHD found a significant poor risk feature associated with the immunophenotypic presence of Reed–Sternberg cell CD20 antigen positivity. This CD20 positive patient group represented only 9·8% of all cHD (first treated at MSKCC), yet these patients experienced 25% of all relapses after conventional therapy.

The CD20 antigen is a 35 kD transmembrane nonglycosylated phosphoprotein, which functions as a calcium channel (Polyak & Deans, 2002). It is involved in the regulation of B-cell activation, proliferation, and differentiation. Despite recent evidence suggesting a B-cell origin for HRS cells (Thomas et al, 2002), most do not stain for CD20. The German Hodgkin's Study Group examined 1286 cases of cHD and found 4·9% to have positively-stained HRS cells (von Wasielewski et al, 1997). Others have shown varying frequencies of CD20 expression, ranging from 14% to 58% (Hall et al, 1988; Schmid et al, 1991; Rassidakis et al, 2002; Tzankov et al, 2003).

In most of our CD20 positive cases (25/28), at least 50% of the Reed–Sternberg cells and variants were positive for CD20, with a distinct membrane staining. The frequency of Reed–Sternberg cells that stained with CD20 were variable within the ill-defined clusters in those cases with a nodular sclerosis pattern. In three cases, the positive cells were present in only one or two small clusters, and classified as focally positive. The intensity of staining was variable within each case, often within the same cluster of cells. There was no diffuse cytoplasmic or nucleolar staining of Reed–Sternberg cells or variants; some cells showed a dense spot of the Golgi area (see Fig 2).

Three groups (von Wasielewski et al, 1997; Rassidakis et al, 2002; Tzankov et al, 2003) have also retrospectively analysed their HD database with regard to the prognostic significance of CD20 in cHD. Rassidakis et al (2002) reported a frequency of 22% in cHD. This group defined CD20 positivity to be the presence of any membrane and/or cytoplasmic CD20. They also assessed the degree of stain intensity as absent versus weak to moderate versus bright. In 598 patients analysed, Rassidakis et al (2002) could not demonstrate any subset in which CD20 positivity conferred a significantly poorer failure free survival. The site of CD20 antigen expression (membrane versus cellular) was not evaluated in this prognostic analysis, however.

The German Lymphoma Study Group (von Wasielewski et al, 1997) reported a frequency of 4·8% CD20 positivity among 1751 patients with cHD. It was not specified in their report whether CD20 positivity included cytoplasmic staining or was limited to membrane staining, as in our study. On univariate analysis, these authors found a significantly worse prognosis for CD20 positive cases (which were also CD15 and CD30 negative) as compared with the most common immunophenotype of CD15, CD30 positive and CD20 negative (83% of cases in this common subtype) with P = 0·0005 for Freedom from Treatment Failure and P = 0·01 for overall survival. CD20 positive cases with CD15 and/or 30 positive immunophenotypes did not differ significantly in outcome from those that were CD15, CD30 positive and CD20 negative. On multivariate analysis, however, these authors reported that only the absence of CD15 was a negative prognostic factor.

Because of the small numbers in our retrospective series, we were unable to analyse the effect of CD20 positivity in the presence or absence of the other markers of cHD. The frequency of these three variables in our patient database were: CD20+, CD15+, CD30+ = 22; CD20+, CD15, CD30+ = 4; CD20+, CD15+, CD30 = 1; and CD20+, CD15, CD30 = 0. The CD30 status of one patient was unknown. Thus, the majority of patients had all three markers present, representing 76% of all CD20+ cases in our series.

A recent report by Tzankov et al (2003) retrospectively analysed the clinico-pathologic features of 24 cases of CD20 positive cHD. These authors reviewed a 25-year database, in which the patients received radiotherapy alone in early stage and combined modality regimens in advanced stage presentations. Their criteria for CD20 positivity was limited to HRS membrane staining (20% of all cHD). In a multivariate analysis comparing a control group of 95 patients to 24 CD20 positive patients, CD20 was not an independent prognostic factor for failure-free survival (FFS) or OS. Treatment regimen and stage were prognostic factors for FFS; and treatment regimen, B symptoms, and age for OS. In the most recently treated cohort (1981–1999), there was no significant difference in FFS according to CD20 status. This might be due to the effect of small sample size and more effective treatments.

In order to remove selection bias from our retrospective series analysis, we evaluated the impact of institutional location of initial therapy on TTF and OS outcomes. When limited to previously untreated patients diagnosed and treated at MSKCC, the adverse effect of CD20 positive histology on TTF and Overall Survival was maintained (Figs 3 and 4). These findings were also confirmed in the cohort of patients initially treated elsewhere.

The standard combination chemotherapy regimen utilized in cHD is ABVD. CD20 positive status was also an adverse prognostic factor for TTF among those receiving this regimen (see Fig 5) as well as other cHD regimens, but not for OS. Of interest are eight patients who received CHOP or CHOPE (Lester et al, 2001) chemotherapy in an effort to treat this poor risk lymphoma with more aggressive therapy. Five of eight patients have remained disease-free since their initial therapy, suggesting that these higher dose doxorubicin-based regimens may be promising alternatives to ABVD. Further research will be needed to confirm this initial impression, however.

Among the eligible patients who had relapsed or refractory CD20 positive cHD, autologous stem cell transplantation (ASCT) therapy resulted in outcomes which were poorer, but not significantly different from those with CD20 negative histology (TTF, P = 0·06; OS, P = 0·07). Neither TTF nor OS following ASCT were adversely affected by CD20 status.

Although based on a limited retrospective review at MSKCC, we conclude that the presence of CD20 positive cells in cHD is a poor risk prognostic factor with initial therapy, for TTF and OS. A prospective study is needed to confirm our findings. Salvage autologous transplantation may be successful and appears to eradicate much of the adverse significance of CD20 status. On the other hand, ABVD chemotherapy for initial therapy appears to be unsuccessful in this poor risk group. All cases of cHD should undergo immunophenotypic analysis for CD20, in addition to the CD15 and CD30 antigens. More intensive doxorubicin-containing regimens and rituximab (anti-CD20 antibody) (Rehwald et al, 2003) should be considered for prospective study in this setting.