| Objective: The prognostic value of serum liver function tests（LFTs）for gastric cancer patients with liver metastasis are not fully understood.This study aimed to develop a liver chemistry score（LCS）and incorporate it into prognosis determination.Materials and Methods: Data of patients with gastric cancer liver metastasis（GCLM）who received palliative chemotherapy in two general hospitals in Anhui Province were retrospective collected and set as development and validation cohort.In development cohort,LCS was generated by Lasso regression dimensionality reduction based on LFT.Cut-off draw calibration curve to reflect the calibration degree of the model.Values of the score were determined by restricted cubic spline.Survival differences between patients with low and high score were estimated by Kaplan-Meier method,propensity score matching was used to balance clinical characteristics between the two subgroups.Then the independent prognostic factors of patients were selected by univariate and multivariate Cox regression analysis,respectively,and the nomogram predictive model was constructed.The model was verified internally and externally by resampling.The difference of the model was reflected by drawing receiver operating characteristic curve（ROC）and calculating the area under the curve,and calibration curve was drew to reflect the calibration degree of the model.Results: 336 and 72 patients with GCLM were included in development and validation cohort,and the median survival time was 9.9 and 10.2 months,respectively.LASSO regression finally reached a two-parametric LCS based on levels of aspartate aminotransferase（AST）and alkaline phosphatase（ALP）at the time of liver metastasis,which calculated as 0.03343515 × ln（AST,U/L）+ 0.02687997 × ln（ALP,U/L）.The optimal cut-off value of 0.232 were used to separate patients as low or high score group.Patients with low or high LCS experienced different survival times in development（m OS: 13.5 vs.7.3 months）and validation cohort（m OS: 14.5 vs.8.5 months）,in addition,death risk was higher in patients with high score than low score（HR=1.70,95%CI:1.37～2.12,P<0.001）.Multivariate Cox regression analysis revealed that shown poor performance status（ECOG 2～）,existence of extrahepatic metastasis,multiple liver metastasis,high level of CA19-9（>35 U/ml）,as well as high LCS（≥0.232）were independent prognostic factors for GCLM patients.A nomogram including above five parameters was constructed and showed superior performance than model not including LCS.AUC of 6-months ROC improved from 0.647（95%CI: 0.584～0.711）to 0.699（0.638～0.759）in development cohort,and 0.837（0.734～0.940）to 0.875（0.784～0.966）in validation cohort.In addition,the nomogram predicted survival fit well with actual survival both in the internal and external validation as shown in calibration curves.Conclusions: A Liver chemistry score based on AST and ALP was constructed which showed ability in separating survival of GCLM patients in this study,and confirmed that the score is an independent prognostic factor in patients.In addition,a nomogram which involving ECOG performance,existence of extrahepatic metastasis,multiple liver metastasis,high level of CA19-9 and the LCS can realize the individual prediction of survival probability of GCLM patients.It has higher prediction ability than traditional model and externally validated,and useful in clinical practice. |
PDF Full Text Request