3.1 General characteristics of the patients for1HNMR analysis
The information regarding patients with LF is provided in Table 1. The group treated by ALSSs comprised 54 males and three females, and the control group comprised 46 males and four females. All the subjects in the ALSSs group satisfied the diagnostic criteria for LF as described in 2012 in the Clinical Practice Guideline for Liver Failure promulgated by the Liver Failure and Artificial Liver Group, Chinese Society of Infectious Diseases, Chinese Medical Association. The exclusion criteria comprised: presence of mental or neurological disorders; allergic constitution; inability to tolerate blood products during ALSSs; or an incomplete record of liver failure due to various factors. The ALSSs-treated and control groups were comparable as they exhibited no statistically significant differences in age or gender (P>0. 05).
The PTA of the serum samples taken after ALSSs treatment were observed to have significantly increased (28.87 ± 0.98 before treatment vs. 38.21± 2.44% after treatment, P<0.001). In contrast, the MELD score and levels of total bilirubin (TBIL), ALT and AST in serum after ALSSs treatment were markedly decreased. No significant differences were observed in serum creatinine levels or INRs between samples taken before and after ALSSs treatment.
3.2 PCA and PLS-DA analysis on 1H NMR data
In this paper, 1HNMR data was processed by total normalization and creatinine normalization before multivariate analysis. The PCA and OPLS-DA scores plots are shown in Fig.1. The results showed that the health group, pre-treatment and post-treatment of ALSSs groups were overlapped in the PCA scores plots by total normalization (Fig.1a). However, these groups could be partly separated by group in the OPLS-DA scores plot generated by creatinine normalization (Fig.1d). The established OPLS-DA model evaluated its accuracy and predictability with three parameters (R2X, R2Y and Q2Y), when R2Y(cum) and Q2(cum) are >0.5, the model is accurate[27]. The values of R2Xcum, R2Ycum, and Q2 in the OPLS-DA model based on total normalization are 0.302, 0.408, and 0.378, respectively, whereas those in the model based on creatinine calibration are 0.512, 0.428, and 0.414. Thus, the OPLS-DA model based on creatinine normalization has good predictive power. The results of the sevenfold cross-validation and permutation test (200 times) show an R2 of 0.488 and Q2 of -0.13, indicating that the model is not over-fitted and the model is effective[28].
3.3 Changes of metabolites in LF patients after ALSSs treatment
A total of 21 serum metabolites in 1HNMR spectra were identified and confirmed (Table 2) by comparing their chemical shifts and coupling patterns with the corresponding values according to the method described in a previous paper [26]. Our results show significant differences between samples taken before and after ALSSs treatment in the ratios of serum leucine to creatinine, isoleucine to creatinine, acetate to creatinine, alanine to creatinine, creatine to creatinine, taurine to creatinine, and lactate to creatinine (Independent-sample t-test, P < 0.05). Figure 2 shows that variables 15 (taurine:creatinine), 16 (creatine:creatinine), and 21 (lactate:creatinine) were closely correlated with ALSSs treatment. In addition, it can be clearly seen that these three ratios returned to normal levels after ALSSs treatment (Figure 3).
3.4 ROC analysis of metabolites in patients before and after ALSSs treatment
Analysis of receiver-operating characteristics (ROC) was conducted before and after ALSSs treatment, and the identified metabolites were used as variables. The results are provided in Table 2. The AUROC values of the creatine:creatinine ratio, taurine: creatinine ratio, and lactate:creatinine ratio were 0.633, 0.644, and 0.650, which were higher than the MELD scores (AUROC=0.593) and prothrombin activity (PTA, AUROC=0.562, Table 1). Therefore the lactate:creatinine ratio was found to show the highest diagnostic efficiency regarding ALSSs treatment.
3.5 Correlation of lactate:creatinine ratio with survival of LF patients
Of 105 inpatients with LF, 48 survived for more than six months after ALSSs treatment (the survival group), 24 died within three months of treatment, and 33 died within one month of treatment. These data were employed for retrospective analysis. The clinical data are presented in Table 3. The results (Figure 4a) show that the survival group exhibited a lactate:creatinine ratio before treatment of 0.038±0.002 (n=48), the group who died within three months had a lactate:creatinine ratio of 0.048±0.005 (n=24) and the group who died within one month had a corresponding ratio of 0.052±0.005 (n=33). These data indicate that the groups of patients who died showed higher lactate:creatinine ratios than the group of patients who survived for more than six months. The correlation analysis shows that the lactate:creatinine ratio is negatively correlated with survival period of liver-failure patients (r = - 0.26, p = 0.001).
In addition, the result shows that the correlation coefficient between the serum lactate:creatinine ratio and PTA was - 0.186 (p = 0.025, Figure 4b).
3.6 Use of lactate:creatinine ratio and PTA to predict one-month survival of LF patients
A total of 81 inpatients with LF who received ALSSs treatment, were included for retrospective cohort analysis in this study. Of these, 49 survived for more than six months, while 32 died within one month after ALSSs treatment. Our results showed that the lactate-creatinine ratio in the death group remained unchanged after ALSSs treatment (without statistical differences). However, the lactate-creatinine ratio in the survival group fell markedly (0.052 ± 0.005 vs. 0.025 ± 0.002) after ALSSs treatment (p < 0.05). In comparison, the serum PTA levels in both the death group and survival group obviously increased, but no statistical differences of those were observed after ALSSs treatment (Fig.5).
The AUROC (95% confidence intervals) of the lactate-creatinine ratio after ALSSs treatment for diagnosis of survival group from death group was 0.682 (0.502-0.862), the sensitivity was 44.8%, and the specificity was 82.1%. Whereas, The AUROC of PTA after ALSSs treatment for diagnosis of those was 0.591 (0.40-0.783), the sensitivity was 40.7%, and the specificity was 76.8% (Fig.6). These results indicate that the lactate-creatinine ratio may be more reliable than measures of PTA to evaluate the therapeutic effect of ALSSs treatment in LF patients.