A new working formulation of non-Hodgkin's lymphomas. A retrospective study of the new NCI classification proposal in comparison to the Rappaport and Kiel classifications

Two hundred thirty cases of malignant non-Hodgkin's lymphomas were reclassified in a retrospective study according to the New Working Formulation for Clinical Usage of the NCI as compared to the Rappaport and Kiel classifications. The reproducibility for the individual schemes this study was 81% (Rappaport), 79% (Kiel), and 85% (New Working Formulation). In keeping with the results of the NCI international study, all lymphomas were subdivided into 3 prognostic groups: (1) low-grade malignancy (6.0 years median survival); (2) intermediate-grade malignancy (3.5 years median survival); and (3) high-grade malignancy (1.4 years median survival). The NCI-proposed New Working Formulation for Clinical Usage is thus recommended as practical and unprejudicing classification scheme for general application; however, its usefulness as tool for translating one classification scheme into another appears limited.     

A critical analysis of the classifications of non-Hodgkin's lymphomas.

The Rappaport classification of non-Hodgkin's lymphomas was proposed almost a quarter century ago, before the advent of modern immunology. This classification, which is based entirely on morphologic features, has proved its clinical usefulness. In light of recent scientific advances, however, its terminology is not appropriate. Five new classifications have been proposed recently, each claiming to have more merit than the others. The purpose of this study is to critically analyze and evaluate these newly proposed classifications to determine which classification is conceptually and scientifically acceptable as well as clinically useful. The results of the study show that there are more similarities than differences among the Rappaport. Lukes and Collins, Dorfman, British, and WHO classifications; the Kiel classification, however, is fundamentally different (Tables 8, 9, 11). None of these classifications can be used in its proposed form. Based on the analysis of these classifications, a compromise working classification is proposed which incorporates the relevant concepts and terminology from the Rappaport, Berard, Dorfman, WHO, and Lukes and Collins classifications (Tables 15, 16). The proposed compromise classification is an attempt to reconcile the various classifications, and to stimulate others to offer modifications which may bring about a final solution to the problem of classification of non-Hodgkin's lymphomas.     

Histologic subtypes and survival of Chinese patients with non-Hodgkin's lymphomas

The histologic subtypes and survival of 840 Chinese patients with non-Hodgkin's lymphoma (NHL) were reviewed. All cases were classified according to the Rappaport and Kiel systems and the Working Formulation (WF). A low incidence of nodular/follicular lymphomas (12%) was observed. The most common histologic subtypes were diffuse histiocytic, diffuse centroblastic, and diffuse large cell lymphomas, according to Rappaport, Kiel, and the WF, respectively. A high proportion (24%) of the cases were unclassifiable, according to Kiel, because of the precise terms of the classification. The "favorable"-prognosis NHL, according to Rappaport, or the low-grade NHL, according to Kiel and the WF, had a more indolent clinical course. However, except for the nodular mixed (Rappaport) or follicular mixed (WF) lymphomas which appeared to behave like the more aggressive NHL, a plateau was not seen in the survival curves of our patients with other kinds of favorable-prognosis or low-grade NHL, indicating the lack of curative potential of these tumors. The lymphoplasmacytoid lymphoma, according to Kiel, also appeared to have a more aggressive clinical course. A plateau was seen in most of the other survival curves of patients with the more aggressive tumors, indicating the prospect of cure. However, the prognosis of the very aggressive tumors such as the diffuse lymphoblastic and diffuse small noncleaved cell lymphomas, according to the WF, remains very poor with a median survival of less than 10 months.     

Non-Hodgkin's lymphomas: a simplified classification scheme

1,292 cases of malignant Non-Hodgkin's lymphomas were classified according to the Rappaport and Kiel classifications and to the New Working Formulation for Clinical Usage (NWF). Subsequently, all lymphoma entities were grouped together following the simple criteria of cell size and growth pattern (small cell, large cell, and mixed cellularity; nodular and diffuse growth). Supplementing these morphological criteria, were data from immunological and cytochemical cell marker studies. The accumulated information was correlated to clinical-prognostic parameters. The results show that using above simplified criteria, all Non-Hodgkin's lymphomas can be classified according to a simple scheme which is of clinical relevance. It can further be deduced from all currently used lymphoma classifications, avoids confusion, and offers abbreviated data useful for computer storage.     

Primary gastrointestinal non-Hodgkin's lymphomas. A retrospective clinicopathologic study of 150 cases

The records of 150 primary gastrointestinal (GI) lymphomas in adults collected from 1974 to 1982 at the Department of Pathology, University of Vienna, were reviewed. One hundred thirty-three cases of malignant lymphomas (ML) were analyzed with respect to histologic type, presenting tumor stage, and clinical course, as well as for factors influencing prognosis. The histologic type of ML as assessed by the Working Formulation and the Kiel, the Lukes and Collins, and the Rappaport classifications showed only a minor influence on prognosis. MLs of follicular center cell origin prevailed in the stomach and large cell, immunoblastic MLs prevailed in the bowel. Immunoperoxidase studies indicated a B-cell nature of GI MLs and demonstrated intracytoplasmic IgM kappa or lambda in most of the MLs of the small lymphocytic, plasmacytoid, and immunoblastic type, respectively. The 105 cases of gastric MLs represented 3.6% of all malignancies of the stomach collected during the study period. Clinical symptoms preceded the diagnosis by 4.4 months on average, and endoscopic biopsy specimens indicated malignancy in 78%. Presenting tumor stages of gastric MLs according to the Ann Arbor staging system were Stage I in 20%, Stage II in 76.2%, and Stage IV in 3.8%. The 28 cases of intestinal ML localized in the small and large bowel without any site prevalence presented with Stage I in 14%, Stage II in 82%, and Stage III in 4%. Tumor resection was performed in 90% of all cases and was followed by multiagent therapy in 53%. Radical tumor resection was obtained in 58% of the gastric MLs and only 28.6% of the intestinal MLs and was closely related to tumor stage. Statistical analysis demonstrated a significant influence of the presenting tumor stage on prognosis as expressed by the overall 2-year survival rate of 70% for Stage I versus 39% for Stage II ML. In addition, Stage II1 according to Musshoff et al. run a better course than II2 as shown by the disease-free 2-year survival rate of 49% versus 15%, respectively. Radical tumor resection was a major determinant of survival and cure of disease as exhibited by the disease-free 2-year survival rate of 57% after radical resection versus 8% after nonradical resection of ML. Finally, diffuse tumor growth and tumor penetration of the gastric wall beyond serosa decreased the survival rates.     

Correlation of surface receptors with histological appearance in 29 cases of non-Hodgkin lymphoma.

The receptor patterns of cell suspensions from 29 cases of non-Hodgkin lymphoma were correlated with the histology of the nodes from which the cells were taken. Twenty-two were judged to be predominantly or largely B-cell, and because of this preponderance these were divided by a method based on the distribution of surface immunoglobulin and the expression of Fc and C3 receptors. "Mature" B-cell and B-mixed tumours showing capping surface Ig with Fc and/or C2 receptors correlated well with a nodular growth pattern, and consisted of what Rappaport (1966) calls "poorly differentiated" lymphocytes equivalent to the "small cleaved" cells as defined by Lukes and Collins (1975). Ten of the 14 patients in this receptor category are alive between 12 and 30 months after diagnosis. Receptor-silent and "immature" B-cell tumours with non-capping surface Ig correlated predominantly with the Rappaport histiocytic lymphoma and Lukes and Collins' large cleaved and large non-cleaved lymphomas, though these histological categories also included a wide variety of other receptor types such as T-cell, Receptor-overlap and the single true Macrophage tumour.Five of the 11 patients with receptor-silent or immature B-cell tumours are alive between 7 and 15 months after the diagnosis. Diffuse mixed and diffuse poorly differentiated lymphocytic lymphomas in Rappaport's classification correlated poorly with receptors, mature and immature B-cell tumours being equally represented.     

Primary gastrointestinal lymphoma. A clinicopathologic study of 102 patients

Clinicopathologic findings in 102 patients with primary gastrointestinal lymphomas were reviewed. Abdominal pain was the common presenting symptom. The primary sites of the tumors were: 67 in the stomach, 24 in the small intestine, 7 in the ileocecal region, 3 in the large intestine, and 1 in the esophagus. The disease more frequently affected males than females and showed peak incidence in the 5th decade of life. Gastric lymphomas usually presented with a single lesion, but multiple lesions were frequent in the small intestine. The body and/or antrum of the stomach were the commonest sites of the lymphomas. Gastric lymphomas were diagnosed at an earlier stage than intestinal lymphomas. There was 1 case with Hodgkin's disease. The remaining 101 patients with non-Hodgkin's lymphomas were classified according to the Rappaport and the Kiel classifications. The proportion of nodular lymphomas in the present series was 7%. The frequencies of diffuse histiocytic type ard germinal center cell tumors were 62 and 74%, respectively. Cox's multivariate analysis for prognostic factors revealed that the stage of the tumor, sex, and age were prognostically significant.     

Epstein-Barr virus-associated antibody pattern in untreated non-Hodgkin lymphoma patients. Relationship to clinical variables and lymphocyte functions

Sera from 296 unselected and untreated patients with non-Hodgkin lymphoma (NHL) classified according to the Rappaport and the Kiel systems were analzyed for antibodies to Epstein-Barr virus (EBV). The aim of the study was to determine whether antibody spectra and liters to EBV-coded antigens correlated to clinical and immunological variables and whether the liters were of any prognostic significance. Increased antibody titers to EB viral capsid antigen (VCA) and slightly raised titers to early antigens (EA) of the diffuse (D) and restricted (R) types were noted frequently. Anti-VCA antibody titers correlated to clinical stage and age of the patients but not to histological. subgroups according to the Rappaport or the Kiel classification systems. However, and-VCA titers ? 1:2560 were seen only in diffuse lymphomas according to the Rappaport and in non-follicle cell-derived lymphomas according to the Kiel classifications. Patients with complement-receptor-positive diffuse lymphomas had higher anti-VGA titers than complement-receptor-positive nodular cases. Anti-VCA titers also correlated positively to serum IgG levels (p <0.01). Total number of lymphocytes separated from peripheral blood and mitogen induced (ConA, PWM) DNA synthesis were recorded before treatment in 54 of the patients. The patients exhibited a significant lymphocytopenia as well as a significantly reduced lymphocyte response to mitogens (p <0.001) compared to healthy controls. Elevated anti-VCA titers and anti-EA titers correlated to a good mitogen-induced lymphocyte response (p <0.05). Only anti-D 1:40 at diagnosis predicted a poor prognosis.     

Morphometric study of primary cutaneous germinal center cell lymphomas.

BACKGROUND. Primary cutaneous germinal center cell lymphomas (PCGCCL) are B-cell neoplasias; most are mixed cell type, composed of small and large cleaved cells and large noncleaved cells (centrocytes and centroblasts). Because of the significant pleomorphism of the cells, which ranged in size from small elements with indented nuclei to larger ones with multilobulated vesiculous nuclei, histopathologic examination of PCGCCL of mixed cell type has the worst degree of histologic interobserver reproducibility. METHODS. In this study, the authors determined whether the light microscopic morphologic classification criteria applied in quantifying cleaved and noncleaved cells in these lymphomas are substantiated by morphometric analysis and have good reproducibility in differentiating the various cell subsets. They used an image analyzer (IBAS 2000, Zeiss Kontron, Munich, Germany) to study 17 centroblastic/centrocytic and centroblastic follicular and/or diffuse primary cutaneous lymphomas. RESULTS. The data obtained showed morphometric differences between the follicular and diffuse patterns of neoplastic cells. In follicular neoplasms, the cells tend to be smaller, more cleaved, and more monomorphic than those observed in the diffuse forms. In all the follicular and diffuse cases, the authors observed a unimodal population with a wide peak composed of cells with intermediate morphometric features that cannot be labeled properly. CONCLUSIONS. These data underline the considerable overlap of nuclear parameters of lymphocytes between the subtypes of cells in cutaneous mixed small and large cell lymphoma, making the Kiel classification and Working Formulation unsuitable for this kind of neoplasia and supporting revision of classification criteria in cutaneous germinal center cell lymphomas.     

Prognostic value of the histologic classification of peripheral T-cell lymphoma: a clinico-pathologic study of 71 HTLV-1 negative cases.

The histologic and clinical features of 71 cases of peripheral T-cell lymphoma (PTCL) have been studied. All patients were HTLV-1 negative. The T-cell phenotype was demonstrated by immunohistochemistry on cryostat sections (41 cases) and paraffin-embedded sections (30 cases). All cases were histologically classified according to the updated Kiel classification of non-Hodgkin's lymphoma (low and high-grade) and according to a Working Formulation (WF)-based classification (predominantly small cells, mixed small and large cells, and predominantly large cells). Most cases were in the high-grade group according to both classifications. The prognostic value of these two classifications was comparatively assessed. The analysis of the actuarial survival curves showed that, by using the updated Kiel classification, low-grade PTCL had a survival probability higher than high-grade PTCL although the difference was not statistically significant. Similar results were obtained when the WF-based classification was applied: furthermore, actuarial survival curves of mixed small and large cell PTCL, and of large cell PTCL were rather similar, thus indicating that differentiating these two categories has a limited prognostic value.     

Comparison of S-phase fraction, working formulation, and Kiel classification in non-Hodgkin's lymphoma

The prognostic value of S-phase fraction (SPF), determined by flow cytometric study from paraffin-embedded tissue, and grading by Working Formulation (WF) and Kiel classification were compared among 245 patients with non-Hodgkin's lymphoma followed for the median of 89 months or until death. Histologic reclassification and SPF determinations were done without knowledge on clinical data. SPF (P equals 0.0001), WF (P equals 0.0003), and Kiel classification (P equals 0.0008) were associated with mortality in lymphoma in a univariate analysis, and WF and SPF were independent prognostic factors in Cox's multivariate analysis. Although SPF correlated strongly both with WF and Kiel grades (P less than 0.0001), low-grade and high-grade malignant lymphomas according to Kiel classification, and high-grade lymphomas according to WF could be divided into groups with significantly different outcome by SPF. The results suggests a role for SPF in therapeutic decision-making.     

Prognostic Value of the Histologic Classification of Peripheral T-cell Lymphoma: A Clinico-Pathologic Study of 71 HTLV-1 Negative Cases

The histologic and clinical features of 71 cases of peripheral T-cell lymphoma (PTCL) have been studied. All patients were HTLV-1 negative. The T-cell phenotype was demonstrated by immunohistochemistry on cryostat sections (41 cases) and paraffin-embedded sections (30 cases). All cases were histologically classified according to the updated Kiel classification of non-Hodgkin's lymphoma (low and high-grade) and according to a Working Formulation (WF)-based classification (predominantly small cells, mixed small and large cells, and predominantly large cells). Most cases were in the high-grade group according to both classifications. The prognostic value of these two classifications was comparatively assessed. The analysis of the actuarial survival curves showed that, by using the updated Kiel classification, low-grade PTCL had a survival probability higher than high-grade PTCL although the difference was not statistically significant. Similar results were obtained when the WF-based classification was applied: furthermore, actuarial survival curves of mixed small and large cell PTCL, and of large cell PTCL were rather similar, thus indicating that differentiating these two categories has a limited prognostic value.     

Lymphomas in dogs. A morphologic, immunologic, and clinical study

One hundred seventy-six canine lymphomas were classified morphologically using four of the major human lymphoma classification schemes (Rappaport, Lukes-Collins, Kiel, and the Working Formulation). All 176 dogs received the same chemotherapeutic protocol. Sixty-two of these lymphomas had their immunophenotypes established by examination of cell surface markers by automated cytofluorography. Several different morphologic types of canine lymphoma were identified and these were comparable to morphologic categories in human classification schemes. Follicular and low grade lymphomas were rare. The two most common morphologic types were diffuse large cell (centroblastic) and immunoblastic. The Kiel classification appeared to be the most useful human scheme for classifying the canine lymphomas. Cytofluorographic analysis was generally straightforward, and 60 of the 62 lymphomas were placed into one of three immunophenotypic categories: 27 pan-T(LQ1)+SIg+, 21 pan-T(LQ1)-SIg+, and 12 pan-T(LQ1)+SIg-. Two of the lymphomas could not be characterized immunologically because a pre-existing or reactive non-neoplastic population of lymphocytes made interpretation of single cell suspension analysis difficult. The authors identified correlations between morphology and survival and disease-free remission; dogs with high-grade tumors generally survived the longest and had the longest remissions. No correlations were identified between high concentrations of serum lactate dehydrogenase, age, sex, or stage of disease, and morphology, immunophenotype, remission, or survival times. A significant correlation between clinical illness and survival time was documented. The median age of the dogs was nine years, no significant effect of sex on prevalence was observed, and some breeds were significantly overrepresented. Significant morphologic-immunophenotypic correlations included shorter remission and survival times for T-cell tumors than B-cell tumors, and a highly significant correlation between the pan-T(LQ1)+SIg-"T cell" phenotype and hypercalcemia.     

DNA-grading of malignant lymphomas. I. Prognostic significance, reproducibility and comparison with other classifications

The prognostic significance of the DNA Malignancy Grade (DNA-MG) was tested for 83 malignant non Hodgkin lymphoma patients in comparison with three different subjective morphological classification systems (New Working Formulation for Clinical Usage, Rappaport and Kiel Classification). Monolayer smears prepared from paraffin embedded tissue and imprint smears from fresh cut lymph-nodes were investigated. Feulgen staining was performed automatically. The scalar DNA-MG was determined by rapid interactive DNA-cytometry, using an automatic microscope and a TV-image analysis system. A strong influence of the DNA-MG on length of survival was found. Compared with morphological classification systems, the DNA-MG was of greater prognostic validity as revealed by different statistical tests. Significant differences of survival probabilities between some groups with differences of 0.5 DNA-MG only were found. The interobserver reproducibility of this new prognostic index was found to be 95% within a range of +/-0.4 DNA-MG.     

Comparison of three pathologic classifications (Rappaport, LSG and Working Formulation) for the treatment of non-Hodgkin's lymphoma

Comparison of three pathologic classifications (Rappaport, LSG and Working Formulation) was made for 292 cases of non-Hodgkin's lymphoma who were treated at the Department of Radiology, Tokyo Medical and Dental University, from 1952 to 1984. The Working Formulation was found to be the most convenient classification for the treatment of non-Hodgkin's lymphoma in Japan. There were two remarkable characteristics in the cases of non-Hodgkin's lymphoma classified by the Working Formulation. They were better case distribution having a strong relation to clinical prognosis and fewer high-grade cases in Waldeyer's ring lymphoma.     

Lymphoplasmacytic lymphoma. A clinicopathologic study of a previously unrecognized composite variant

Six cases are reported of a previously undescribed unusual composite variant of lymphoplasmacytic lymphoma that is not readily classifiable by either the Rappaport or the Lukes and Collins classifications, or by the recently proposed Working Formulation. These cases are characterized by a partly nodular, partly diffuse proliferation of lymphoid cells surrounded by a separate, zonally distinct proliferation of large plasmacytoid cells. The latter cells are located in medullary and paracortical areas in lymph nodes and in the marginal zone of the white pulp and in the red pulp in the spleen. This distinct zonal characteristic was noted in the small bowel as well in one case. Five of our six patients were male, and their ages ranged from 46 to 68 years. Three had a monoclonal serum IgM and one had hyperglobulinemia that was not further characterized. Three had evidence of an altered immune state. In all cases, monoclonal IgM was demonstrated in involved tissues by an immunoperoxidase technique. These cases are unusual because of the unique topographic segregation of the varying types of tumor cells in all cases, and because of the association of paraproteinemia with nodular lymphoma in three. The resemblance of this tumor to the plasma cell variant of giant lymph node hyperplasia with which two cases were originally confused, and to lymph nodes in patients with autoimmune diseases, may be responsible for its lack of recognition.     

Clinico-pathological features of malignant lymphomas in 294 Hong Kong Chinese patients, retrospective study covering an eight-year period

The clinical records and histological material from 294 adult Chinese patients with malignant lymphoma were examined. These patients were first seen at the Queen Mary Hospital, Hong Kong, during the 8-year period 1975-82. There were 27 patients (9.2%) with Hodgkin's disease (HD) and 267 with non-Hodgkin's lymphoma (NHL). The median age at presentation was younger for HD (45 years) and the male: female ratio was higher (2:1) than the corresponding figures for NHL of 51 years and 1.4:1. In 76 patients (28.5% of NHL), the disease was thought to have originated in an extra-nodal site, 48 of these cases being gastrointestinal lymphomas. It was possible to reclassify 234 NHL according to the Rappaport and Kiel classifications, and the Working Formulation (WF) proposed by the US National Cancer Institute Study; for HD, the Rye classification was used in 26 cases where suitable material was available. Nodular/follicular lymphomas made up 17.1% of nodal NHL and 5.3% of extra-nodal NHL. The "histiocytic" (Rappaport) or large-cell (WF) subtype was the commonest amongst diffuse NHL. There were only four cases of Burkitt's lymphoma. For HD, the nodular sclerosing subtype was commonest in females (5 out of 8 cases) and for males, the commonest was mixed cellularity (10 out of 18 cases). Of patients with nodal NHL 64.7%, presented with Stage IV disease. For HD, there were about equal numbers of patients presenting with Stage II and Stage IV disease (10 and 9 respectively). The low incidence of Hodgkin's disease and of follicular lymphomas is comparable to figures from other "oriental" countries such as Japan.     

Transferrin receptor expression by non-Hodgkin's lymphomas. Correlation with morphologic grade and survival

The monoclonal antibody OKT9 was applied to cryostat sections of 267 non-Hodgkin's lymphomas and related neoplasms. It was found that the transferrin receptor is expressed by a wide variety of B- and T-lineage non-Hodgkin's lymphomas. The OKT9 staining also was loosely correlated with the three morphologic grades of non-Hodgkin's lymphomas identified by the International Working Formulation. In general, higher grade lymphomas more often and more intensely expressed the T9 antigen. However, transferrin receptor expression by certain histologic subtypes of lymphoma did not correlate with their morphologic grade: low-grade follicular lymphomas expressed the T9 antigen more frequently than diffuse low-grade lymphomas; diffuse small cleaved cell lymphomas were stained by OKT9 less often than other histologic subtypes of intermediate-grade lymphomas; and diffuse immunoblastic lymphomas expressed transferrin receptors less often than the other high-grade histologic subtypes of non-Hodgkin's lymphoma. Intermediate lymphocytic lymphomas, not recognized in the International Working Formulation, were infrequently and weakly stained by OKT9 in a manner similar to diffuse low-grade lymphomas. We obtained clinical follow-up data on 43 individuals with chronic lymphocytic leukemia/small lymphocytic lymphoma and 64 individuals with diffuse large cell and immunoblastic lymphoma. Transferrin receptor expression in these two groups did not correlate significantly with survival.     

Classification of malignant lymphoma

Over the past 30 years different lymphoma classifications were used world-wide. In Europe, the Kiel classification of K. Lennert dominated the scene, it was updated 1988 and 1992. In North America the Working Formulation, primarily designed to be an instrument for translation of one classification into the other, was predominantly used. Both classifications, however, showed little correspondence with each other, which made it difficult to compare the results of randomized clinical trials on both sides of the Atlantic Ocean. In 1994 the Revised European American Lymphoma (R.E.A.L.) classification was proposed by the International Lymphomas Study Group (ILSG). In contrast to the existing classifications it abandoned the grading of malignancies. The ILSG focused on diagnostic reproducibility and on the definition of distinct clinical-pathological lymphoma entities. Similar to the Kiel classification the new classification scheme was based on cell lineage (T- and B-cell origin) and cell differentiation (precursor and mature lymphomas) and included a number of distinct extranodal lymphomas. The new WHO classification is basically identical with the R.E.A.L. scheme and exhibits only minor changes. The WHO classification includes all hematopoietic and lymphoid neoplasms. It represents the first generally accepted classification, providing hematologists and oncologists with a solid diagnostic basis for therapeutic decisions.     

Radiation therapy of Waldeyer's ring lymphoma

Forty-nine patients with biopsy-proven Waldeyer's ring lymphoma were treated with radiation therapy between 1968 and 1979; 45 to 50 Gy were given to Waldeyer's ring structures with a 5- to 10-Gy boost to the primary site. Uninvolved low cervical nodes received 40 to 50 Gy. Chemotherapy was reserved for treatment failures. Actuarial 5-year survival was 53%, disease-free survival was 48%, and local control was 98%. Patients with Stage I disease fared better than patients with Stage II disease, and unilateral adenopathy conferred a better prognosis than bilateral adenopathy. When classified by the Rappaport system, nodular lymphomas had a better prognosis than diffuse lymphomas, and of the diffuse lymphomas, histiocytic lymphoma was a more lethal disease than lymphocytic lymphoma. Patients with diffuse undifferentiated lymphomas in the Rappaport system, or high-grade lymphomas in the Working Formulation, fared poorly. Most relapses were systemic, and actuarial 5-year survival after salvage was only 20%. A logical approach to the treatment of this disease can be based on these prognostic features.