Patient characteristics
A total of 3015 patients were retrospectively screened at Shanghai Chest Hospital and 126 patients were enrolled in this study according to the inclusion criteria. 62 patients and 64 patients were grouped into BM group and BMM group. In this study, the characteristics of the two groups were well balanced (Table 1).
The metastatic organs of all patients, coexisting with initial bone metastases in this study, were lung (66/126,52.4%), brain (40/126,31.7%), pleura (30/126,23.8%), liver (16/126,12.7%), adrenal gland (7/126,5.6%), distant lymph nodes (7/126,5.6%), pericardium (6/126,4.8%) and nodules of subcutaneous metastasis (2/126,1.6%) (Fig. 2).
Comparison of survival outcome of different bone metastases patterns
The median follow-up time for all patients was 44.5 months (95%CI 37.9–51.1 months). 59.5% patients (n = 75) died during the follow-up period. The overall median 1LrwPFS was 13.9 months (95% CI 12.653–15.147), 2LrwPFS was 14.5 months (95% CI 11.665–17.335) and rwOS was 40.1 months (95% CI 35.996–44.204).
Compared with patients in BM group, patients in BMM group had worse rwOS (35.2 months vs. 42.9 months, HR = 0.512,95% CI 0.320–0.821, P = 0.005) (Fig. 3A) and shorter 2LrwPFS (12.8months vs. 17.0 months, HR = 0.575,95% CI 0.374–0.881, P = 0.011) (Fig. 3B). There was no statistically significant difference in 1LrwPFS (12.7months vs. 14.0months, HR = 0.838,95% CI 0.587–1.198, P = 0.333) and PPS (10.6 months vs. 6.2months, HR = 0.731,95% CI 0.475–1.123, P = 0.152) between BM group and BMM group.
No CR was observed in BM group and BMM group during third-generation EGFR-TKI treatment. The disease control rate (DCR) was 93.5% (58/62) and objective response rate (ORR) was 46.8% (29/62) in BM group. The BMM group achieved a similar DCR (96.9%,62/64) and ORR (48.4%,31/64) (Fig. 3E). Detailed data for the best response outcome between BM and BMM group was shown in Fig. 3C and 3D.
Multivariate regression analysis on rwOS and 2LrwPFS
Based on the results of the univariate analysis, multivariate analysis used the Cox model to examine the impact of important clinical characteristics on rwOS and 2LrwPFS. In multivariate analysis, patients with BMM (P = 0.002), performance status (PS) score ≥ 2 (P = 0.000) and TP53 alteration positive (P = 0.003) were independent prognostic factors of worse rwOS (Table 2). 2LrwPFS was independently related to patients with BMM (P = 0.014) and PS score (P = 0.000) (Table 3).
In addition, log-rank tests demonstrated that rwOS differed for patients according to TP53 alteration status and performance status. 2LrwPFS was different for patients according to performance status. Patients with TP53 alteration negative had a median rwOS of 42.9months compared to 33.4 months for patients with TP53 alteration positive (log-rank test, P = 0.038, Fig. 4A). Patients with PS0-1 had a median rwOS of 41.8months compared to 18.1 months for patients with PS ≥ 2 (log-rank test, P = 0.000, Fig. 4B). Patients with PS0-1 had a median 2LrwPFS of 16.4months compared to 8.0 months for patients with PS ≥ 2 (log-rank test, P = 0.000, Fig. 4C).
Correlations of 1LrwPFS, 2LrwPFS and PPS to rwOS
In this study, we assessed the associations of 1LrwPFS, 2LrwPFS and PPS to rwOS using patient-level data. Linear regression and Spearman rank correlation analysis demonstrated that 2LrwPFS was strongly correlated with rwOS (r = 0.621, R2 = 0.568, P =0.000) (Figure 4E), while 1LrwPFS (r = 0.421, R2 = 0.210, P =0.000) and PPS (r = 0.406, R2 = 0.154, P =0.000) was moderately correlated with rwOS (Figure 4D, 4F). 2LrwPFS had a high impact on rwOS.
Subgroup analysis of patients in BM and BMM group
Survival analysis of rwOS and 2LrwPFS showed the superiority of BM group over BMM group. Subsequently, we performed subgroup analysis to determine the survival effect of different bone metastases patterns (BM group and BMM group) in various subgroups. All subgroup showed that BMM group was inferior to BM group in rwOS (Fig. 5A) and 2LrwPFS (Fig. 5B).