Identifying patients with early-stage classic Hodgkin lymphoma (cHL) at higher risk of treatment failure is important to tailor therapy and limit long-term toxicity. Clinical risk factors and early metabolic response with [18F]FDG-PET are used in practice, yet approximately 15% of patients still fail to achieve durable remission with standard first-line regimens. Over the years, many tissue-based biomarkers have been proposed to add prognostic discrimination. Among these, CD68, c-MET, and HLA class II have been reported in prior studies to correlate with outcome, but results have been inconsistent and have not translated into routine clinical use. This study evaluated whether IHC detection of these three markers provides prognostic information in a rigorously controlled, randomized-trial population from the EORTC-GELA H9 intergroup study.
The parent EORTC-GELA H9 randomized trial enrolled 1,591 clinical stage I–II (supradiaphragmatic) cHL patients between 1998 and 2002. Diagnostic unstained whole tissue sections were available for 580 patients (36% of the trial). Samples were collected from 50 centers across ten European countries. For the biomarker validation study, a case-cohort sampling strategy was applied among the 580 patients with available tissue to form the analysis cohort.
IHC was performed on formalin- or Bouin-fixed paraffin sections using antibodies and staining protocols consistent with the original studies that reported prognostic associations. The applied reagents included a rabbit polyclonal anti-c-MET antibody, a mouse monoclonal anti–HLA class II (clone CR3/43), and a mouse monoclonal anti-CD68 (clone KP1). Appropriate positive and negative controls were included, and EBER in situ hybridization was used to determine tumor cell EBV status where applicable.
Scoring criteria replicated prior work: c-MET was scored positive when cytoplasmic staining was present in at least 30% of Hodgkin Reed–Sternberg (HRS) cells, with slides considered evaluable only if ≥10 HRS cells were present. Membranous HLA class II staining was scored positive when expression on HRS cell membranes exceeded surrounding lymphocytes and at least 50 HRS cells were present; heterogeneous patterns were scored by the predominant pattern or deemed indecisive if no predominant pattern existed. CD68 was quantified by automated digital morphometry in tumor-rich areas, expressed as the percentage of positive pixels of total tissue pixels; the cut-off for high versus low CD68 was set at the median percentage, consistent with the original automated-scoring approach.
A case-cohort design was used. From the 580 patients with tissue, a random sub-cohort comprising 33% (n = 193), stratified by favorable versus unfavorable risk group and treatment arm, was selected. The sub-cohort included 41 patients who experienced a PFS event; all additional cases (patients with PFS events) were added to yield a total case-cohort of 257 patients containing 105 PFS events.
The primary endpoint was progression-free survival (PFS), defined as time from registration to progression/relapse or death from any cause. Overall survival was a secondary endpoint. Cox regression models with Barlow weighting were used to estimate hazard ratios (HRs) and robust 95% confidence intervals, accounting for the case-cohort sampling. Analyses were pre-specified and performed for the whole cohort and stratified by risk group and treatment arm as appropriate. Sensitivity analyses applied alternative weighting schemes (Prentice and Self & Prentice).
Of evaluable cases, c-MET expression in tumor cells was observed in 67% (108/162). Membranous HLA class II expression on HRS cells was present in 51% (70/137), consistent with previous reports. CD68 expression was scored across cases by digital analysis; the median percentage of CD68-positive tissue pixels was approximately 9.4%, and values above the median were defined as high.
In univariate weighted Cox regression models for PFS, none of the three biomarkers showed a statistically significant association with outcome. The estimated HRs (with 95% confidence intervals) were: c-MET HR 0.752 (95% CI: 0.456–1.241); HLA class II HR 0.854 (95% CI: 0.503–1.448); CD68 HR 0.864 (95% CI: 0.518–1.440). Because no biomarker reached significance in univariate analysis, no combined biomarker classifier was constructed or tested further in multivariable combinations.
Using standardized IHC methods and a pre-specified, statistically rigorous case-cohort approach within a randomized trial population, the study found that c-MET, HLA class II, and CD68 assessed by the described scoring systems did not provide prognostic information for PFS in early-stage cHL in this cohort. The observed prevalence of marker expression was in line with earlier reports for c-MET and HLA class II, suggesting successful technical replication of staining and scoring methods.
The lack of validation contrasts with some prior studies that reported adverse prognosis associated with high CD68 or absence of c-MET or HLA class II. The authors note that trial-specific factors may influence findings, including risk-adapted treatment regimens, the exclusion of advanced-stage patients, and potential challenges in scoring reproducibility across multi-center archival tissue. These aspects could reduce the ability to detect prognostic effects observed in other cohorts.
Key limitations noted in the source include potential influences from the trial population (early-stage, risk-adapted treatments) and technical or scoring reproducibility challenges. Individual participant-level data are not publicly available due to GDPR-related restrictions; data access requests can be submitted to EORTC through their controlled data sharing process. The study also reports that staining was not evaluable for all collected tissues because of minimum HRS cell thresholds and other technical exclusions.
In this validation study within the EORTC-GELA H9 intergroup randomized trial, IHC assessment of c-MET, HLA class II, and CD68 did not show prognostic value for progression-free survival in early-stage classic Hodgkin lymphoma patients. Given these findings, the three candidate biomarkers—using the applied staining and scoring methods—do not provide sufficient evidence to support implementation for routine prognostic stratification in this clinical setting.