Patient characteristics
Overall, 287 patients were evaluated. Of them, 214 (74.6%) patients died, and 73 patients were right-censored. Table 1 shows the baseline characteristics of the SBRT group (n=133) and the IMRT group (n=154). Before propensity score matching, the proportion of patients with unfavorable baseline characteristics was higher in the SBRT group than that in the IMRT group, including TACE refractory disease, older age, PT, albumin, ALBI score, and ALBI grade (all P<0.05). After matching, 102 paired patients from the SBRT and IMRT groups were selected. The patient characteristics were well balanced between the two groups in the matched cohort.
Overall survival, progression-free survival, intrahepatic control, and local control
The median follow-up time was 31 months (range, 3–84 months). Before propensity score matching, the median OS in the IMRT and SBRT groups was 10.0 vs. 10.0 (HR=0.893; 95% CI: 0.677–1.178; P = 0.422) months; PFS, 6.0 vs. 6.0 (HR=1.103; 95% CI: 0.855–1.422; P =0.450) months; IC, 9.0 vs. 11.0 (HR=1.269; 95% CI: 0.792–2.035; P =0.322) months. There was no significant difference between the IMRT group and the SBRT group with respect to the 6-, 12-, 24-, and 60-month cumulative OS (73.2, 48.1, 25.1, 10.7% vs. 70.3, 46.5, 29.3, 11.3%, P=0.407; Figure 1a), PFS (52.3, 30.6, 20.7, 8.7% vs. 51.3, 21.6, 10.0, 8.3%, P=0.424; Figure 1b), IC (61.8, 44.3, 35.6, 26.0% vs. 75.4, 46.3, 33.0, 26.4%, P=0.094; Figure1c), and LC (84.5, 62.6, 58.1, 52.7% vs.87.0, 64.8, 54.6, 54.6%, P=0.498; Figure1d).
Consistent results of no statistically significant difference in OS, PFS, IC, and LC were obtained after propensity score matching (Figure 2a, 2b, 2c, 2d). The median OS, PFS, and IC was 8 vs. 10 (HR=0.795; CI: 0.574–1.099; P=0.165), 6 vs. 6 (HR=1.048; CI: 0.776–1.415; P=0.758), and 8 vs. 11 (HR=0.756; CI: 0.512–1.116; P =0.159) months in the IMRT and SBRT groups. The 6-, 12,- 24-, and 60-month cumulative OS (73.5, 42.9, 23.6, 7.6% vs. 72.4, 45.1, 29.8, 13.2%, P=0.151; Figure2a), PFS (53.9, 29.3, 21.8, 7.5% vs. 54.5, 19.3, 12.0, 9.6%, P=0.744; Figure2b), IC (61.4, 45.7, 39.0, 26.8% vs. 75.1, 45.8, 35.9, 28.7%, P=0.144; Figure2c), and LC (85.2, 56.5, 52.1, 47.4% vs. 87.4, 65.2, 62.1, 62.1%, P=0.191; Figure 2d) rates of the IMRT group were similar to those of the SBRT group.
Multivariate survival analysis in the matched cohort (Online Resource: Table S2) showed that the AFP level (HR=1.433; 95% CI: 0.928–2.214; P=0.023) and ALBI score (HR=1.860; 95% CI: 1.298–2.665; P=0.004) were independent predictors of OS. Further, ALBI score (HR, 1.609; 95% CI, 1.154–2.244; P= 0.005) was an independent predictor of PFS.
Subgroup analysis bybiologically effective dose
Univariable survival analysis showed that BED10 was a prognostic factor for OS, PFS, IC, and LC. BED10 was excluded from multivariate analysis to avoid collinearity). Therefore, we further performed subgroup analyses by BED10. No optimal cutoff value was found in ROC analyses for the IMRT group. In ROC analyses of the matched cohort, the optimal cutoff for predicting OS, PFS , IC and LC of the SBRT group was 100 Gy (AUC=0.601, Online Resource: Figure S1a), 100 Gy (AUC=0.670; Online Resource: Figure S1b), 100 Gy (AUC=0.601; Online Resource: Figure S1c), and 100 Gy (AUC=0.598; Online Resource: Figure S1d), respectively. Therefore, the prognostic effect of BED10 ≥100 Gy for the OS, PFS, IC, and LC in the SBRT group was further investigated.
In Kaplan-Meier analyses of the unmatched cohort, the 6-, 12-, 24-, and 60- month OS (87.5%, 61.5%, 46.2%, and 46.2% vs. 63%, 36.5%,22.6%, and 8.5%; P=0.0078; Figure 3a), PFS (66.9%, 46%, 30.7%, and 30.7% vs. 38.8%, 15.6%, 5.4%, and 3.6%; P=0.0011; Figure 3b), IC (90.9%, 69.2%, 60.6%, and 60.6% vs. 62.8%, 40.1%, 25.5%, and 17%; P=0.0059; Figure 3c), and LC (95.5%, 86.8%, 77.1%, 77.1% vs. 84.7%, 58.0%, 47.0%, 47.0%, 17%; P=0.04; Figure 3d) were significantly better in patients treated with a high BED10 level (BED10 ≥100 Gy) than in patients with a low BED10 level (BED10 <100 Gy). The results of ROC analyses and Kaplan-Meier analyses with a BED10 cutoff value of 100 Gy are shown in Online Resource:Table S3.
Toxicity
In the matched IMRT group, 15 (14.7%) patients experienced RILD (c-RILD, n=2; nc-RILD with elevation of CP-score ≥ 2, n =13). Of them, 6 (6/89 with CP-A, 6.7%) and 9 (9/13 with CP-B, 69.2%) had CP-A and CP-B disease, respectively. Five patients died within 2 months and 2 died within 5-6 months due to RILD after IMRT. There were 4 patients (3.9%) with tumor adjacent to the gastrointestinal tract (<1 cm) who experienced radiation-induced gastrointestinal (GI) bleeding within 5–6 months.
In the matched SBRT group, 12 (11.8%) patients developed RILD (c-RILD, n=2; nc-RILD with elevation of CP-score ≥2, n=10). Of them, 5 (5/91 with CP-A, 5.4%) and 7 (7/11 with CP-B, 63.6%) had CP-A and CP-B disease, respectively. Among the patients who received BED10 ≥100 Gy, 4 patients (1 with CP-A5, 3 with CP-A6) treated with 42–45 Gy in 3 fractions experienced nc-RILD with elevation of CP-score ≥2 after SBRT. Among the patients who received BED10 <100 Gy group, 2 patients with CP-B7 died within 3 and 5 months due to c-RILD after SBRT, and 3 patients (2.9%) with tumors adjacent to the gastrointestinal tract (<1 cm) experienced radiation-induced GI bleeding within 4–7 months.