combination chemotherapy (CHT) with high-dose meth-otrexate (HD-MTX) and radiotherapy (RT) during thetime covered by this study. The protocol was ﬁrst pub-lished in 1992 , proving to be superior to RT alonein several smaller phase II clinical trials, and eventuallybecame widely accepted as clinical standard. The en-couraging results of the initial clinical studies were con-ﬁrmed by a larger multicenter trial showing superioroutcome of combination therapy, when compared to pre-viously published results on RT alone . Even withouta larger randomized study available, these results ledto a widely accepted consensus to include HD-MTX inPCNSL therapy . However, the beneﬁt of the newtherapy modality for a general population of PCNSLpatients was less striking in large retrospective multi-center studies [4, 7, 13, 30] than that expected basedon the smaller Phase II trials. Subgroups of PCNSLpatients clearly had prolonged survival after an aggres-sive therapy regimen , but for a majority of cases theprognosis remained poor [14, 17, 29].In contrast to other PCNSL series in the past thatwere based on clinical records, the present work relieson the neuropathological archives, reﬂecting the ex-perience of a single referral center for stereotacticneurosurgery. The PCNSL patients referred to our in-stitution include a group of patients with histologicallyproven CNS lymphoma in relatively poor clinical con-dition who received only supportive treatment. Ourapproach to identifying PCNSL patients based on thepathology archive rather than on clinical records of large specialized haemato-oncological centres enabledus to reﬂect this aspect of PCNSL management duringthe 1990s that may have been missed by other retro-spective studies. Encompassing a time of signiﬁcantchanges in PCNSL treatment, it also aimed at investi-gating the general acceptance of such an important newtreatment modality as combined RT
CHT, and its ac-tual impact on clinical outcome.
Materials and methods
Search methods and case deﬁnition
Patient identiﬁcation was based on the neuropathological archive,encompassing the years from 1989 to 2001. The series was restrictedto cases diagnosed by stereotactic biopsy, an approach that representsthe state-of-the art diagnostic procedure in PCNSL and allowed us torelate the annual patient numbers to the overall frequency of braintumours diagnosed by the same method. 210 cases of cerebral lympho-ma were identiﬁed. 34 were secondary lymphoma and
or had evidenceof immunosuppression (history of HIV-related disease or tested positivefor HIV, long-term immunosuppressive medication due to organ trans-plantation or auto-immune disease, pre-existing immune defects or pre-vious chemotherapy for solid tumours). 12 cases were excluded becausethe clinical material was not sufﬁcient to exclude immunosuppression, orto determine survival. 164 PCNSL cases met the following inclusioncriteria: (1) high-grade lymphoma diagnosed by routine histology andimmunohistochemistry, (2) clinical records sufﬁcient to exclude immu-nosuppression and
or systemic lymphoma at the time of diagnosis (3)survival data available until July, 2002.
Stereotactic biopsy technique and processing of specimens
Stereotactic serial biopsy was performed under computer guidance,with a neuropathologist attending the surgical procedure and evaluatingintra-operative tissue preparations. Consecutive specimens were eitherﬁxed in formaline, or used for touch
smear preparations stained withmethylene blue in an alternating manner. The clinical records fromthe Department of Stereotactic Neurosurgery were retrieved to obtain adescription of the pre-operative cranial computed tomography (cCT), asummary of the presenting complaints, the pre-operative clinical course,and the approximate date of onset of the symptoms. The neuropatholo-gical work-up of older cases, particularly the immunohistochemical de-tection of leucocyte antigens, was completed with current methods of immunohistochemical signal detection and antigen retrieval if necessary.In addition to the available routinely performed immunohistochemicalstainings, new 5
m thick parafﬁn sections were cut from selected casesand stained following routine protocols using the following antibodies:polyclonal anti-CD3, monoclonal anti-CD45 (clone LCA), monoclonalanti-CD20, clone L26, and monoclonal mouse anti-CD79a, clone JCB(DAKO, Glostrup, Denmark). Antibody binding was detected using theHistostain Plus Kit (Zymed, distributed by Zytomed, Berlin, Germany)following the manufacturer’s instructions. All histopathological reportsavailable from the neuropathological archive were reviewed for clinicaldata provided by the referring physician, radiographic ﬁndings, intra-operative, and histopathological diagnoses. Histologically, all but twocases were consistent with diffuse large B-cell lymphoma according tothe current WHO classiﬁcation including the expression of the B-celllineage markers CD20 and CD79a; the two exceptions showed surfaceimmunoreactivity for the T-lineage marker CD3.
Corticosteroid therapy and intra-operative diagnosis
Patients were considered to have received corticosteroid therapy if they either were receiving treatment at the time of the surgical procedureor had discontinued treatment less than 5 days prior to the stereotacticbiopsy. The intra-operative diagnosis was based on methylene blue-stained smear or touch preparations, which were evaluated by the attend-ing neuropathologist during each biopsy procedure in the operatingtheater. The diagnostic categories considered for statistical analysis were‘‘high-grade lymphoma,’’ ‘‘probable high-grade lymphoma, other differ-ential diagnoses not excluded,’’ or ‘‘other diagnosis’’ such as glioma ormetastasis. The accuracy of the intra-operative diagnosis was graded as‘‘complete correlation’’ if intra-operative diagnosis of lymphoma wasconﬁrmed by histological diagnosis based on parafﬁn-embedded speci-mens, ‘‘partial correlation’’ if lymphoma had been considered during thesurgical procedure but other diagnoses could not be ruled out, and ‘‘nocorrelation’’ if intra-operative diagnosis did not include lymphoma andthe diagnosis of PCNSL was based solely on the parafﬁn-embeddedmaterial) .
Clinical follow-up was recorded until July, 2002. All clinical recordsavailable from the archive of the Department of Stereotactic Neuro-surgery were reviewed. In the majority of cases, the patients had beenreferred from other medical institutions, which limited the availableinformation on the preclinical patient history to data provided by the