The Glasgow prognostic score at diagnosis is an independent predictor of survival in advanced stage classical Hodgkin lymphoma

Abstract

Positron emission tomography (PET)-guided strategies emerge as standard-of-care for advanced-stage classical Hodgkin lymphoma (cHL) while conflict remains regarding early stage (I – IIa) disease and the necessity for treatment-intensification among patients with a positive PET after two chemotherapy cycles (PET-2) (Borchmann et al, 2017). Risk-stratification remains crucial, increasing the requirement for predictive biomarkers. Immunity and inflammatory response impact the tumour microenvironment and malignancy progression (Mantovani et al, 2008). Metabolic and inflammatory parameters of the peripheral blood, and ratios of the latter, correlate with outcome in lymphoma patients (Tadmor et al, 2015). There is evidence outlining the prognostic significance of inflammation-based scores including the Glasgow prognostic score (GPS; integrating serum C-reactive protein and albumin), platelet-to-lymphocyte ratio (PLR) and neutrophil-to-lymphocyte ratio (NLR). From a clinicopathological perspective this is especially apparent in cHL, with scarce malignant cells embedded in a multifaceted inflammatory environment (Skinnider & Mak, 2002). GPS was recently shown to predict outcome in B cell lymphoma (Hao et al, 2017), whereas only NLR and PLR were established as prognosticators in cHL (Marcheselli et al, 2017; Reddy et al, 2017). In this retrospective study, approved by the local ethics committee, we investigated the prognostic capabilities of baseline GPS in cHL as a complementary resource for risk stratification. Patients with cHL admitted to our institutions between January 2000 and July 2017 were screened. The initial assessment (Figure S1) identified 171 eligible patients. Centralized review of pathology reports was conducted and central haematopathology assessment was performed in 140/171 (81·9%) cases. Clinical information was collected from original files and data were recorded (Table S1). Scores/ratios were calculated as described (Hao et al, 2017). Patients were treated with a chemotherapy regimen of the treating physician's choice with respective current German Hodgkin Study Group study protocols serving as institutional standard. PET scans were not employed on a routine basis. Progression-free survival (PFS) and overall survival (OS) were estimated by means of the Kaplan–Meier method and the univariate log-rank test. Optimal cut-offs for NLR (6·267) and PLR (262·8) were determined utilizing an algorithm proposed by Budczies et al (2012) (Figure S2). Characteristics associated with OS or PFS with at least a trend towards statistical significance (P < 0·07) were included in a proportional-hazard-model. Differences among subgroups were assessed using the chi-square and Mann-Whitney-U-test. All investigations were conducted using GraphPad PRISM6 (GraphPad Software Inc., San Diego, CA, USA)and SPSS 24 (IBM, Armonk, NY, USA). Clinicopathological characteristics are summarized in Table SI. Median age at diagnosis was 40·0 years (range 18–82 years) with a median follow-up of 94·0 months. Median NLR was 3·1 (range 0·23–53·9), with a median PLR of 161·5 (range 12·9–920). Baseline distribution into GPS groups and pre-treatment characteristics stratified on GPS and NLR groups are depicted in Table SII. Multivariate analysis revealed Eastern cooperative Oncology group performance status > 1 (Hazard ratio [HR] = 5·151, 95% confidence interval [CI] = 1·268–20·930, P = 0·022) and GPS to be the only independent predictors of OS, and GPS was one of two (alongside elevated lactate dehydrogenase) independent predictors of PFS (OS: HR = 3·271; 95% CI = 1·327–8·059, P = 0·010; PFS: HR = 2·206; 95% CI = 1·294–3·761, P = 0·004). The impact of GPS on OS and PFS is presented in Fig 1 and Table 1. Dichotomizing NLR and PLR identified no impact of PLR on outcome; although NLR did impact on outcome, this effect was lost upon multivariate analysis. Subcategorizing results by early (I – IIa) and advanced (IIb – IV) stage we found GPS to maintain predictive properties regarding PFS (P = 0·0001) and OS (P < 0·0001) in advanced stage disease whereas this was less striking in early stage disease (PFS, P = 0·3258; OS, P = 0·03) (Figure 1). Two-year freedom from progression was 100%, 90% and 78% according to a GPS of 1, 2 or 3 and 96·6%, 81·0% and 61·7% at 5 years (P < 0·0001). Overall survival according to GPS at 2 and 5 years was 100%, 100%, 91·2% and 98·1%, 97·1%, 75·5%. Moreover, we assessed the association of GPS/NLR groups with baseline clinical characteristics (Table SII). Choice of therapy resulted in a trend towards improved OS and PFS favouring BEACOPP (bleomycin, etoposide, doxorubicin, cyclophosphamide, vincristine, procarbazine, prednisolone)-type chemotherapy (OS: HR = 1·919; 95% CI = 0·7125–5·169, P = 0·1972; PFS: HR = 1·877; 95% CI = (0·9089–3·876), P = 0·088). Meanwhile GPS maintained its prognostic implications upon correction for choice of therapy (P < 0·001 for both OS and PFS). The need for refined risk-stratification complementing PET-guided subsequent treatment is evident. This is the first study illustrating the prognostic impact of GPS in cHL. Its predictive capabilities, especially amongst advanced stage patients, may contribute to treatment guidance in PET-2-positive patients (Zinzani et al, 2016; Borchmann et al, 2017). An investigation of GPS in prospective trails incorporating PET-guided treatment is however required in order to validate our current observations. There is growing evidence of a drastic impact of GPS on outcome in various malignancies, including lymphoma and solid tumours (Hao et al, 2017). A most recent study by Reddy et al (2017) assessed the prognostic capabilities of NLR and PLR in early stage cHL revealing independent adverse prognostic implications of an elevated PLR. Other studies made similar observations for elevated NLR (Marcheselli et al, 2017). Our study expands on previous efforts in that cHL of all stages were assessed for baseline GPS alongside both PLR and NLR. Deviations from previous results probably result from the inclusion of patients with all disease stages and the size of specific subgroups may have prevented the detection of independent effects on clinical outcome. Limitations of the study include its sample size and retrospective design resulting in the lack of centralized pathology and radiology review in a subset of patients and the potential for fragmentary data. Our results require validation in larger and independent data-sets, derived from PET-guided prospective trails before suggestions regarding personalized risk-assessment and ultimately treatment intensifications should be made. Our data show that baseline GPS correlates with rates of relapse and refractory disease across all primary stages of cHL, which was preserved in advanced stage disease upon multivariate analysis. GPS constitutes a promising means of risk-stratification in cHL requiring further validation. The authors would like to thank Birte Mergelkuhl and Rudina Marx for their dedicated administrative assistance. Funding: The authors declare no funding relevant to the current manuscript. Study concept: NG, HW, SF. Data collection: NG, HW, SF, HL, HB, VB, HM, SK. Data analysis and creation of figures and tables: NG, HB, SF, DR. Initial Draft of manuscript: NG. Critical revision and approval of final version: all authors. The authors declare no conflicts of interest. Fig S2. Definition of optimal cut-off values for PLR (262·8 - A) and NLR (6·267 - B) utilizing a mixture model of two Gaussian distributions fitted to the histogram of the respective biomarker employing the R-based web-application by Budczies et al (Budczies et al, 2012). Table S1. Clinicopathological baseline characteristics of the study group Table S2. Baseline clinical characteristics stratified on Glasgow prognostic score - and Neutrophil-to-lymphocyte ratio –group Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.