3.1 The second follow-up inquiry of patients and their baseline characteristics
The retrospective study initially selected 979 consecutive patients enrolled in the SYSUCC study between January 2007 and May 2012, with 749 patients were finally selected for analysis, nearly half of whom had a history of prior hepatitis B virus (HBV) infection (343/749, 45.8%). As of May 2012, the death toll was 404 (58.8%). The median follow-up time was approximately 16.6 months (range: 0.1–112.4), and the median follow-up interval was 2.0 months (range: 0.4–8.0). Overall, 120 patients had new progress after TACE. Among them, 36 had lymph node metastasis, 66 had distant metastasis, 59 had vascular invasion, and 74 had ascites.
Performance status (PS) score, major tumor diameter, white blood cell (WBC) count, alpha-fetoprotein (AFP) level, location of pathological changes, Child-Pugh class, and the number of intrahepatic pathological changes were significantly different among the three groups of patients, (P<0.05) (Table 1).
3.2 Correlation Between OS and ΔLDH After TACE Treatment
In the correlation study, we first used the Kaplan–Meier method and assessed the median OS of the three groups. Then, the log-rank test was used for statistical analysis. The median OS was 34.5 months (95%CI: 28.4–42.3) in the ΔLDH witnin ±80 U/L group; 23.9 months (95%CI: 19.8–34.0) in the ΔLDH≤-80 U/L group; and 11.5 months (95%CI: 6.7–18.2) in the ΔLDH≥80 U/L group (Figure 2). This difference among groups was statistically significant (P<0.0001).
Univariate analysis showed that PS score (1: HR=3.07, 95%CI=2.29–4.11); main tumor diameter (cm) (5 cm: HR=3.32, 95%CI=2.60–4.23); AFP (ng/mL) (≥25: HR=2.27, 95%CI=1.77–2.90); number of intrahepatic pathological changes (≤3: HR=2.10, 95%CI=1.53–2.90 and >3: HR=4.07, 95%CI=3.05–5.44); location of pathological changes (left or right: HR=2.12; 95%CI=1.54–2.91 and both sides: HR=4.11, 95%CI=3.08–5.48); WBC (109/L) (≥11: HR=1.80, 95%CI=1.20–2.70); and Child–Pugh class (B: HR=1.84, 95%CI=1.43–2.36) were related to higher risk of death in the ΔLDH≥80 U/L group than the ΔLDH within ±80 U/L group and were considered risk factors. However, age (years) (≥55: HR=0.87, 95%CI=0.70–1.68) and hemoglobin (g/L) (≥120: HR=0.72, 95%CI=0.56–0.93) were related to a less serious risk and were protective factors (Table 2). Similar results were found in the ΔLDH≤-80 U/L group.
Confounding factors according to clinical and statistical reasons included age, sex, Hgb, Child–Pugh class, PS score, location of pathological changes, AFP, number of intrahepatic pathological changes, WBC count, and main tumor diameter.
Table 3 shows that the risk of death increased by 24% when ΔLDH≤-80U/L (unadjusted HR: 1.24, 95%CI: 0.99–1.55) and by 131% when ΔLDH≥80U/L (unadjusted HR: 2.31, 95%CI: 1.74–3.06). After adjustment for confounders, the positive association was more consistent than before (ΔLDH≤-80U/L: HR=1.11, 95%CI=0.88–1.39; : HR=1.40, 95%CI=1.04–1.87). Thus, it is evident that an increase (ΔLDH≥80U/L, P<0.05) in LDH levels after TACE treatment increases the risk of death in patients.
3.3 The nonlinear relationship and threshold effect of ΔLDH on OS
As shown in Figure 3, there is a non-linear U-shaped relation between ΔLDH and OS. After the lowest point (ΔLDH=0 U/L), HR gradually increased with decreasing or increasing LDH, and the risk increased significantly with increasing LDH. ΔLDH had a significant effect on the threshold for OS (P<0.001) (Table 4). ΔLDH<0 (100 U/L) had an HR of 0.96 (95%CI: 0.93–1.00) and ΔLDH>0 (100 U/L) had an HR of 1.11 (95%CI: 1.07–1.15). The significance of ΔLDH threshold effect on OS was also considerable (P=0.021) after adjusting for potential influencing factors including age, sex, WBC, Hgb, PS score, main tumor diameter, Child–Pugh class, location of pathological changes, AFP, and the of intrahepatic pathological changes.
In order to verify the data distribution and correlation between ΔLDH and OS, we drew a scatter diagram (Figure. 4), which further proved that the increase and decrease of LDH were related to the decreased OS.
3.4 Sensitivity Analysis
We first assessed the relationship between ΔLDH and patient mortality at 6 months, and then at 12, 24, and finally 36 months using a univariate logistic regression model to analyze the relationship between LDH and OS after TACE treatment. In order to more intuitively show the differences between groups, we drew a histogram. Further detailed information of the methods is presented as supplementary information (Table S1, Figure S2) (P<0.05).