Prognostic analysis and a new risk model for Hodgkin lymphoma in Japan
- First Online:
- Cite this article as:
- Itoh, K., Kinoshita, T., Watanabe, T. et al. Int J Hematol (2010) 91: 446. doi:10.1007/s12185-010-0533-9
- 136 Views
The Japan Clinical Oncology Group conducted two multicenter phase II trials in 200 patients with advanced Hodgkin lymphoma (HL) in the 1990s. Among 181 patients whose histopathological specimens were available and reviewed by 6 hematopathologists, 167 (92.3%) were diagnosed with HL. Five-year overall survival (OS) among these 167 patients was 88.3%, including 89.2% among nodular sclerosis and 82.2% among mixed cellularity cases. International prognostic score was not closely associated with OS. Seven unfavorable prognostic factors for OS on univariate analysis were male, B symptoms, clinical stage of III or IV, elevated serum LDH, elevated alkaline phosphatase, elevated β2-microglobulin, and pathological subtype (mixed cellularity and lymphocyte depletion). On multivariate analysis, male [HR 3.30 (95% CI 1.15–9.52, p = 0.027)] and elevated serum LDH [HR 2.41 (95% CI 1.07–5.43, p = 0.034)] were independent factors for OS. Based on these prognostic factors, the 5-year OS was 95.7% in the low-risk group (no adverse factor), 87.9% in the intermediate-risk group (1 adverse factor) and 73.3% in the high-risk group (2 adverse factors). This simple prognostic model for HL warrants further validation studies.
KeywordsInternational prognostic scoreMulticenter phase II trialPrognostic factorOverall survivalMale genderLDH
Most of the patients with advanced Hodgkin lymphoma (HL) could be induced into complete remission (CR) with state-of-the-art combination chemotherapy or chemo-radiotherapy, and in patients with advanced HL who relapsed after achieving CR, there are some therapeutic options for curing the disease, including conventional salvage chemotherapy and high-dose chemotherapy followed by autologous stem-cell transplantation . However, the excellent outcomes in the initial treatments for HL do not necessarily result in excellent survival, because 20–30% of patients with advanced HL are not cured of their disease, and moreover, the treatments are associated with increased risks of late toxicities such as secondary malignancies, cardiopulmonary toxicities, and cerebrovascular diseases [2–5]. It still seems to be necessary to identify the high-risk group of the minority of patients with fatal outcome.
Many prognostic factors for failure-free survival have been described in patients with advanced HL. These included age, sex, clinical stage, B symptoms, number of nodal sites, laboratory data such as serum albumin, hemoglobin, white cell count, lymphocyte count, etc. . The international prognostic score (IPS)  was widely accepted as the prognostic index in advanced HL. However, only 7% of the patients had the worst adverse score of 5 or higher of IPS which represents a very high risk, and was associated with 56% of the overall survival (OS) at 5 years. Thus, it was concluded that a distinct group of patients at very high risk could not be identified by the IPS .
Considering the various effective treatment options and their late toxicities, it is important to identify the prognostic factors for OS in patients with advanced HL. In particular, this is relevant to the question of whether early high-dose chemotherapy with autologous stem-cell transplantation should be used as a consolidation therapy in patients with responses to induction therapy, who are nevertheless considered to remain at high risk for relapse. To address the ability to predict the prognosis of patients with advanced HL, we analyzed patients with advanced HL enrolled in the Japan Clinical Oncology Group-Lymphoma Study Group (JCOG–LSG) trials. The aims of this study were to validate the IPS in terms of OS, to evaluate the OS according to several prognostic factors including histological subtypes of HL, and to find a better prognostic model for patients with advanced HL, who were enrolled in JCOG–LSG trials with state-of-the-art combination chemotherapy or chemo-radiotherapy.
2 Patients and methods
2.1 Patients and treatments
The JCOG–LSG conducted two multicenter phase II trials for advanced HL in the 1990s that tested the efficacy of the ABVd regimen (JCOG9305)  and ABV regimen followed by involved-field radiotherapy (IF-RT) (JCOG9705) . Major eligibility criteria were age between 15 and 69 years, and Eastern Cooperative Oncology Group (ECOG) performance status of 0–3 in the two trials, and clinical stage of II, III or IV in JCOG9305 and clinical stage of IB, IIB, III, or IV or any stage with bulky lesion in JCOG9705. Bulky lesion was defined as a mass of at least 10 cm (largest diameter) and a bulky mediastinum (ratio of the mediastinum to the thorax of at least one-third at the level of the largest diameter while the patient was standing). A total of 128 patients from 35 participating institutes were enrolled in JCOG9305 between 1993 and 1997 to assess the efficacy of the ABVd regimen, which consisted of doxorubicin, bleomycin, vinblastine and a reduced dose of dacarbazine of two-thirds (250 mg/m2) of that in the original ABVD regimen. The reasons for modification of the original ABVD regimen in both JCOG studies were that dacarbazine was highly emetic and it was not approved for the treatment of HL in Japan at that time. In JCOG9705, a total of 72 patients from 25 participating institutes were enrolled between 1998 and 2000 to assess the efficacy of the ABV regimen, in which the dose of doxorubicin was increased to 120% of that in the original ABVD regimen and dacarbazine was not utilized. Patients were evaluated for response after 4 cycles of chemotherapy. All patients received 2 additional cycles of chemotherapy. For those with CR after 4 cycles, chemotherapy was finished after a total of 6 cycles. Patients who were in CR or uncertain CR (CRu) after 6 cycles were given 2 additional cycles of chemotherapy. In patients with bulky lesions, IF-RT with 30–40 Gy was added if patients entered into CR or CRu after 4 or 6 cycles. Regardless of whether the lesion was bulky or non-bulky, IF-RT was added if patients entered into partial remission (PR) in JCOG9705.
CR was defined as the disappearance of all measurable or assessable diseases and all signs and symptoms of the disease lasting for at least 4 weeks. PR was defined as a reduction of 50% or greater in the sum of the perpendicular diameters of all measurable lesions and the appearance of no new lesions for at least 4 weeks. CRu was defined as the maintenance of PR for at least 3 months without any treatment. Progressive disease was defined as an increase of 25% in the size of any lesion or development of any new lesions. Relapse was defined as an increase of 25% in the size of any lesion or development of any new lesions in CR or CRu patients. The details of the results of each clinical study will be published elsewhere.
All of the protocols described above including the informed consent document were approved by both the JCOG Protocol Review Committee and the institutional review board of each institution. The protocol of JCOG0108A, an ancillary study with secondary use of the data acquired by the above-mentioned JCOG studies, was also approved by the JCOG Protocol Review Committee.
2.2 Consensus diagnosis
The procedure of reaching a consensus diagnosis of HL according to the WHO classification has been described . Briefly, 6 hematopathologists consisting of 4 panelist pathologists and 2 consulting pathologists reviewed the histopathological specimens independently. Immunohistochemical studies were conducted on paraffin sections by means of the avidin–biotin–peroxidase complex technique and a panel of monoclonal antibodies including antibodies against CD20 (L26; DakoCytomation, Glostrup, Denmark), CD3 (PS-1; Novocastra, Newcastle, UK), CD15 (MMA; Becton Dickinson, San Jose, CA, USA) and CD30 (BerH2; DakoCytomation, Glostrup, Denmark). All 6 hematopathologists and 1 hematologist performed the central pathologic review, in which the case report forms of the patients were available for reference of clinical information. A consensus diagnosis was established when agreement was reached by three-fourths or greater majority of the 4 panelist pathologists with no opposition from the 2 consulting pathologists and the hematologist. The cases with discordant pathological diagnosis were re-evaluated until agreement by two-thirds or greater majority was reached among the 6 pathologists by means of reconciliation. Then, a consensus diagnosis was made. The present study included patients in two multicenter phase II trials for advanced HL who were diagnosed with HL by central pathological review.
2.3 Statistical analysis
All statistical analyses were performed by a statistician (K.Y.) at the JCOG Data Center. Patients with lymphocyte depletion had been reported as having a worse prognosis than those with other subtypes , but this subgroup contained only 7 patients in this study. Therefore, patients with lymphocyte depletion were grouped together with patients with mixed cellularity who had also been shown to have a worse prognosis . OS was the endpoint of all statistical analyses. OS was calculated from the date of enrollment in respective study to the date of death from any cause or to the date of last follow-up in living patients. OS was estimated by the Kaplan–Meier method. The log-rank test was used to assess the significance of unadjusted differences in OS for each prognostic factor. Multivariate analysis was performed by the Cox proportional hazards model to identify subsets of prognostic factors for OS. All p values were two-sided and p values less than 0.05 were considered significant. There is no widely agreed approach to building a multivariate prognostic model from a set of candidate predictors [13, 14] and, in consideration of the limitation of events in our study, the data were analyzed from points of significancy and parsimony. A prognostic model was established by fitting all variables that significantly influenced OS in multivariate analysis, and the risk groups were identified according to the established model. For comparing OS between the risk groups, the overfitting-corrected p values were derived by fivefold cross-validation. All statistical analyses were performed using SAS release 9.1.3 (SAS Institute, Inc., Cary, NC).
3.1 Histopathological distribution
Histopathological distribution of advanced HL among 167 patients
Number of patients (%)
Nodular lymphocytic predominance
3.2 Clinical characteristics
Patient characteristics (n = 167)
Number of patients (%)
Performance status (0/1/2/3)
Extranodal sites (0/1/≥2)
Sites of organ involvement
Bone marrow (yes)
Baseline hematological data
Hemoglobin (<10.5 g/dl)
White blood cells (≥15000/μl)
Lymphocytes (<600/μl or <8%)
3.3 OS according to histology and IPS
3.4 Unfavorable prognostic factors by multivariate analysis
Univariate survival analysis
White blood cells
<600/μl or <8%
Multivariate survival analysis
3.5 Risk group model
It is recognized that there is an uneven geographical distribution of malignant lymphomas throughout the world. Namely, the incidence of T cell lymphoma is relatively high in Asia compared with Western countries. On the contrary, the incidence of HL in Japan was reported to be 4.4% of malignant lymphomas and this is relatively low compared with those in Western countries [11, 15, 16]. The low incidence of HL in Japan limited the evaluation of the applicability of IPS or other prognostic factors to Japanese patients with advanced HL treated with an established protocol considered to be state-of-the-art combination chemotherapy or chemo-radiotherapy. To our knowledge, this is the first report to validate the IPS comprehensively and to analyze the conventional prognostic factors for OS in a large number of Japanese patients with advanced HL treated with established protocols of state-of-the-art combination chemotherapy or chemo-radiotherapy and diagnosed by central pathological review. The histopathological distributions of advanced HL according to the WHO classification in 167 patients in Japan were determined, and showed that the proportion of patients with nodular sclerosis in Japan (68.9%) was higher, while the proportion with mixed cellularity (20%) was similar to those in Western countries [12, 16]. The survival of each HL subtype in Japan was similar to those in Western countries .
In our study, there were only 5 patients with IPS score of 5 or higher, accounting for only 3% of the entire study population. Their 6-year OS was 60%, indicating that a distinct group of patients at very high risk could not be determined on the basis of IPS. Even in the original IPS paper, only 7% of the patients had a score of 5 or higher representing a very high risk and had a 59% of OS at 5 years. The results were very similar to our study. Therefore, in our study, it can be concluded that a distinct group of patients at a very high risk could not be identified by IPS, as same as stated in the original IPS paper . Furthermore, the survival curves of patients with an IPS score of 1, 2 or 3 were not clearly separated from each other (Fig. 2). Therefore, IPS was not closely associated with OS in our study. It has been reported that the IPS score 0–4 versus 5 or 6 was found to have prognostic significance for disease-specific survival in a report of large number of Japanese patients with HL treated variously, in which the presence of T cell and/or cytotoxic antigen in Hodgkin’s and Reed-Sternberg cells also showed a significant poor prognosis . However, in that study, neither patient number nor survival rate of patients in each IPS score was shown at all, thus it may be said that IPS was not adequately validated in Japanese patients with advanced HL treated with state-of-the-art combination chemotherapy or chemo-radiotherapy. As it had been concluded that a distinct group of patients at very high risk could not be identified by the IPS , attempts have been made to determine more suitable factors that could detect the poor-risk population among patients with HL [17–19]. Namely, it was reported that the number of involved anatomic sites combined with the IPS  or interleukin-10 (IL-10) level added to the IPS  could detect the subgroup of HL patients with poor prognosis.
As the initial treatments for HL led to excellent outcomes and the rescue treatments could improve the clinical outcomes, the prognostic factors for OS might be more important than the prognostic factors for progression-free survival (PFS) in patients with advanced HL. In our study in which the diagnosis of HL was based on central pathological review, OS was independently affected by male and elevated serum LDH on multivariate analysis. Only male and clinical stage of III or IV among the 7 factors in the IPS were significantly associated with poor OS in the univariate analysis, and male remained significant in the multivariate analysis. The German Hodgkin Study Group suggested that hematotoxicities were more pronounced in females although this did not translate into increased infection, and female patients had similar response rates as males but fewer relapses and deaths, leading to a significantly better freedom from treatment failure in a large retrospective analysis . Sex might be associated with the metabolism of anticancer drugs . Elevated serum LDH was previously reported to be prognostically unfavorable in advanced HL [22, 23] and is also one of the most important factors in the international prognostic index of non-HL . Therefore, elevated serum LDH might reflect the total status of HL, including both constitutional and disease-related elements.
As a post hoc sensitivity analysis, we also performed the analysis using stepwise variable selection methods, and the results were shown that male [HR 6.18 (95% CI 2.28–16.70, p < 0.001)] and elevated serum LDH [HR 2.87 (95% CI 1.32–6.24, p = 0.008)] remained significant and serum albumin level of less than 4 g/dl [HR 3.38 (95% CI 1.35–8.51, p = 0.01)] was also significant. Although serum albumin was significantly correlated with B symptom (p < 0.001) and serum alkaline phosphatase (p = 0.001), serum albumin did not show the significancy over male and serum LDH as the prognostic factor and did not show the prognostic relevance for OS in univariate analyses. Although stepwise variable selection method was widely used, model by stepwise method is not necessarily considered the best with regard to the statistical issues, which were often discussed and criticized [25–28]. In this study, analysis was performed following the prospectively planned method, and the final model was evaluated by cross-validation, one of the internal validation methods to resolve these statistical issues. The OS curves between risk groups derived from our final model were significantly different, and the results were validated by cross-validation. Based on a combination of model fit and parsimony considerations, our final model incorporated two prognostic factors: male and elevated serum LDH.
Generally, complex models with a large number of prognostic predictors are not practical and simple models are easier to evaluate and are preferable in routine clinical practice. The prognostic model for advanced HL in our study, including sex and serum LDH, was considered to be very simple. However, it is discussed that prognostication with the prospective studies has a limitation by the exclusion of patients with poor condition and prognostic models using data of prospective study might be difficult for generalization. On this concern, adequate consideration should be necessary.
Prognosis of the patients with advanced HL is improved in advance of treatment, and then prognostic factors may differ according to the state of the treatment. In the original IPS paper, eligible patients with advanced HL for the original IPS study were limited to those who were 15–65 years old and were treated with an established protocol still considered to be state-of-the-art, with at least four planned cycles of combination chemotherapy (preferably containing doxorubicin) with or without radiotherapy. This means that IPS was established in the patients who could be safely treated with state-of-the-art therapy, excluding both elderly patients of more than 65 years and those who were poorly treated probably because of poor condition. Nonetheless, IPS has been used widely, because everybody wants to know the prognostic state of patients with advanced HL treated with state-of-the-art therapy. In our study, eligible patients for the prognostic analysis are almost same as patients for the original IPS study. Then, our prognostic model could be accepted for general use, although further studies should be warranted to validate our prognostic model.
Unfortunately, data on the β2-microglobulin level of 56 patients were missing. In univariate analysis, β2-microglobulin was found to be highly significant in 111 patients. However, multivariate analysis including β2-microglobulin revealed that there was no significant factor detected. Then, β2-microglobulin was excluded from the final multivariate analysis for OS. Serum β2-microglobulin levels are known to reflect renal function and membrane turnover, the latter of which is associated with tumor mass and growth rate. Elevated β2-microglobulin level was reported to predict poor survival in several hematological malignancies including low-grade lymphoma , large cell lymphoma  and HL [31–33]. Interestingly, Vassilakopoulos et al.  reported that the β2-microglobulin level was a powerful independent prognostic factor for OS, but not for failure-free survival in optimally treated patients with HL. The prognostic impact of β2-microglobulin on OS should be re-evaluated in future.
In conclusion, despite the limitation of a small number of patients, our prognostic model was considered to be a simple method of predicting OS in Japanese patients with advanced HL. Further studies to validate our prognostic model and to re-evaluate the prognostic impact on OS of sex and serum LDH combined with β2-microglobulin are warranted.
We thank Kiyoshi Mukai (Tokyo Medical University), Shigeo Nakamura (Nagoya University), and Kouichi Ohshima (Kurume University) for pathological review as members of an expert panel. This study was supported by Grants-in-Aid for Cancer Research from the Ministry of Health, Labour and Welfare of Japan (2S-1, 5S-1, 8S-1, 11S-1, 11S-4, 14S-1, 14S-4, 17S-1, 17S-5).