Validation and refinement of PROSASH model using the neutrophil‐to‐lymphocyte ratio in patients with HCC receiving sorafenib

The recently developed PROSASH model is proving to be a useful tool in risk‐group discrimination in hepatocellular carcinoma (HCC) patients treated with sorafenib. Several studies highlighted that the neutrophil‐to‐lymphocyte ratio (NLR) is one of the most important predictors of survival in HCC patients treated with sorafenib. The aims of the present study were to validate the PROSASH model and determine whether the incorporation of inflammatory markers can improve risk stratification. This study included 438 patients. According to the four categories of the PROSASH model, median overall survival (OS) was 20.0, 14.9, 8.5 and 3.0 months respectively (P < .001). The Harrell's c for this categorized model was 0.621. NLR (cut‐off 3) stratified OS in each of the PROSASH categories. After reclassification, median OS was 21.0, 15.1, 8.2 and 4.1 months (P < .001). The Harrell's c increased from 0.621 to 0.673 (P = .001). Integrating NLR into the PROSASH model allowed a more accurate classification of the patients in the risk groups.


| INTRODUC TI ON
The authors used data collected by two phase III randomized clinical trials (RCT) that compared sorafenib with brivanib 9 and sunitinib. 10 PROSASH model, composed by vascular invasion, age, ECOG score, AFP, albumin, creatinine, AST, extra-hepatic spread and aetiology, highlighted that it is possible to predict the survival of patients before to start sorafenib. PROSASH model identified four risk categories with a range of survival between 4 months (high-risk category) to 20 months (low-risk category).
Nevertheless, the PROSASH model still lacks an external validation in routine clinical practice. In addition, two studies highlighted that neutrophil-to-lymphocyte ratio (NLR) with the cut-off of 3 is one of the most important predictors of survival in HCC patients treated with sorafenib, 11,12 but it remains unknown whether this prognostic factor could help furtherly refining the PROSASH model.
The aims of the present study were to validate the PROSASH model in a cohort of HCC patients treated with sorafenib in daily clinical practice, and to verify whether the incorporation of inflammatory markers can determine an improvement of its risk stratification ability.

| MATERIAL S AND ME THODS
The present multicentric retrospective study included 438 ad- Several lines of information on neutrophil, lymphocyte and platelet counts were retrieved from blood tests carried out at baseline (the day before the start of treatment). Complete blood counts have been carried out with XE-2100 (Sysmex).
NLR was obtained by dividing the absolute neutrophil count by the absolute lymphocyte count and the cut-off was set at 3. 11,12

Key points
• The recently developed PROSASH model is proving to be a useful tool in risk-group discrimination in hepatocellular carcinoma (HCC) patients treated with sorafenib.
• The aims of the present study were to validate the PROSASH model and determine whether the incorporation of inflammatory markers can improve risk stratification.
• Integrating NLR into the PROSASH model allowed a more accurate classification of the patients in the risk groups.

| S TATIS TIC AL ANALYS IS
Patients forming the study population were consecutively treated from 2008 to 2018 at four tertiary referral hospitals. Since this was a retrospective observational study, without any intervention comparison, no sample size was required. Overall survival was the main outcome measure and was estimated using Kaplan-Meier methods; differences between subgroups were assessed through log-rank test.
The first aim of the study was to validate the PROSASH model. To accomplish this task, the Harrell's c-statistic was calculated for both the categorization into four different risk classes and the linear predictor, by fitting a Cox regression on overall survival, thus obtaining also the 95% confidence intervals. 13 C-indexes were compared when necessary

| RE SULTS
Characteristics of the cohort (n = 438) are reported in Table 1     The main limitation of this study was its retrospective nature.

| D ISCUSS I ON
However, cases were consecutively selected and this may have reduced potential bias. Furthermore, we do not have data on sequence treatment because this cohort included patients treated before the results of all second-line studies.

| CON CLUS IONS
To conclude, PROSASH model did not show good accuracy in prognostication of HCC patients treated with sorafenib in routine clinical practice. Present results suggest that the integration of NLR index in PROSASH model might increase its accuracy, but its final role in a refined prognostic model would be assessed more adequately in dedicated larger studies.

ACK N OWLED G EM ENT
We don't have acknowledgements.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.