Predicting Clinical Outcome in Patients With Hodgkin Lymphoma

Until recently, the systems for predicting outcome in patients with Hodgkin lymphoma have been based upon relatively straightforward clinical criteria such as age, gender, anatomical stage, and routine measurements, including the sedimentation rate, blood count, and serum albumin concentration. These parameters reflect indirectly on the biologic heterogeneity of the disease, but only at one remove. Such baseline prognostic indices, however, have proven inferior to predictors based on functional imaging with FDG-PET performed as a measurement of response during chemotherapy. Despite this shift toward using in-treatment imaging as the prognostic benchmark, the recent description of the clinical implications of macrophage infiltration has revived interest in pathobiology both as an indicator of outcome and as a potential target for therapy. But enumeration of macrophages is difficult to reproduce using immunohistochemistry, especially on tissue microarrays that may sample only a small region of each biopsy. Now, a collaboration between the Department of Pathology and Experimental Therapeutics of the British Columbia Cancer Agency and four North American cooperative groups has tested a method of multigene expression analysis to derive a prognostic score from routine diagnostic biopsy material. This approach holds the prospect of a new means of predicting outcome based upon molecular phenotype.

In this study, requiring as little as 200 ng of RNA extracted in most cases from a single 10 mm section of formalin-fixed, paraffin-embedded tissue, the authors used a new technique called NanoString technology to examine the pattern of expression of 259 genes. Cases for a training set were drawn from the recently reported intergroup E2496 trial that showed equivalent outcomes in patients with advanced Hodgkin lymphoma treated with either ABVD or the Stanford V regimen. From a total of 794 available biopsies, 290 were studied, and based on overall survival of the patients from whom the samples were derived, a 52 gene prognostic set was developed. Next, this prognostic set was tested on a validation cohort consisting of 78 patients treated with ABVD. This validation cohort was enriched for treatment failure but was otherwise similar to patients treated with ABVD in a population-based registry. From the original 52 gene set, analysis of expression of 23 genes was found to generate a robust prognostic index. Of those 23 genes, 20 were overexpressed and three were underexpressed in the group at highest risk of death. The genes overexpressed reflected the presence of increased macrophage numbers, such as CD68, IL15RA, and STAT1; genes indicative of a Th1 response such as IFN-γ; the targets of IFN-γ such as CXCL11, IRF1, and TNFSF10; genes of HLA class I; and genes expressed by cytotoxic T cells or NK cells. The high-risk group was found to have an excess of patients with a high international prognostic score, positive Epstein-Barr virus-encoded RNA (EBER) expression, and a histology other than nodular sclerosis, although the molecular signature remained independently predictive in multivariate analysis.