Baseline patient characteristics
A total of 251 patients were included in the analysis; their characteristics are listed in Table 1. Low SMD was related to high total fat areas. Female was associated with low SMI. The correlation between SMI and SMD was assessed (Fig 1). A positive correlation between SMI and SMD was found (R2=0.058, P<0.001). The low-SMD group exhibited a lower SMI than the high-SMD group (42.31±10.18 vs. 48.77±11.67 cm2/m2, P<0.001). Among the 166 patients with high SMD before chemotherapy, 53 (31.9%) had low SMI, whereas among the 85 patients with low SMD before chemotherapy, 49 (57.6%) had low SMI. Compared with high SMD, the risks of low SMI were 1.516 (95% confidence interval [CI]: 1.164-1.973) among patients with low SMD. Moreover, 49 of all 251 patients (19.5%) had both low SMI and low SMD.
Survival according to SMI and SMD
OS and PFS were assessed according to SMI and SMD (Fig. 2). With the univariate analysis, low SMI was not related to OS (median, 6.0 versus 8.0 months; P=0.076) or PFS (P=0.752). Patients with low SMD had poorer OS than those with high SMD (6.1 versus 7.9 months, P=0.010). However, there were no differences in PFS (P=0.116) with respect to SMD. Furthermore, patients with both low SMI and low SMD showed poorer OS than others (4.1 vs 7.8 months, P=0.004) (Fig. 3). We also performed a multivariate Cox proportional hazard regression for SMI, SMD, and baseline characteristics (Table 2). Low SMI, low SMD, and co-presence of low SMI and low SMD were statistically significant prognostic factors for OS but not for PFS (Low SMI, hazard ratio [HR]: 1.35, 95% confidence interval [CI]: 1.03–1.78, P=0.032; low SMD, HR: 1.45, 95% CI: 1.09–1.93, P=0.011; and Co-presence of low SMI and low SMD, HR: 1.58, 95% CI: 1.12–2.23, P=0.010). Based on the multivariate analysis results, the C index of low SMI, low SMD, and co-presence of low SMI and low SMD were 0.64, 0.65 and 0.66, respectively. We evaluated the relation between change pattern (increase or decrease) of SMI or SMD and survival during chemotherapy. There is no OS and PFS difference according to change pattern of SMI or SMD. Moreover, Eastern Cooperative Oncology Group performance status and type of first line chemotherapy regimen were prognostic factors for OS, while the number of metastatic sites and baseline CA19-9 levels were related to both OS and PFS (Supplementary Material 1).
Chemotherapy response and toxicities according to SMI and SMD
Chemotherapy response was assessed according to SMI, SMD, and their rate of change (Table 3). Objective responses were not related to low SMI or low SMD (mean SMI: CR/PR vs. SD/progressive disease [PD], 46.58 vs. 46.62 cm2/m2, P=0.981; mean SMD: CR/PR vs. SD/PD, 43.27 vs. 43.48, respectively, HU, P=0.856). Disease control was also not related to low SMI or low SMD (mean SMI: CR/PR/SD vs. PD, 47.21 vs. 45.44 cm2/m2, P=0.249; mean SMD: CR/PR/SD vs. PD,43.58 vs. 42.85, respectively, HU, P=0.484). We also assessed the correlation between chemotherapy response and the rate of change of SMI and SMD and found that objective response was not related to change in SMI or SMD (SMI change: CR/PR vs. SD/PD, -1.99 vs. -4.15%, respectively, P=0.157; SMD change: CR/PR vs. SD/PD, -3.95 vs. -3.05%, respectively, P=0.501). Disease control was also not related to change (SMI change: CR/PR/SD vs. PD, -1.10 vs. -2.85%, respectively, P=0.298; SMD change (%), CR/PR/SD vs, PD, -1.67 vs. -1.26, respectively, P=0.784).
We also investigated the relationship between chemotherapy-related toxicities and SMI or SMD (Table 4). Low SMI and low SMD were not related to grade 3 or higher neutropenia, anaemia, thrombocytopenia, fatigue, or diarrhea, separately. However, all grade 3 or higher adverse events were more frequently reported by patients with low SMI (43% by high-SMI patients vs. 59% by low-SMI patients, P=0.019) as well as by patients with low SMD (44% by high-SMD patients vs. 60% by low-SMD patients, P=0.023).
Survival rates among chemotherapy responders and non-responders
We assessed survival according to the presence of low SMI and low SMD in both chemotherapy responders and non-responders (Fig. 4). OS was not associated with SMD status among responders (CR/PR); however, among non-responder patients (SD/PD), the low SMD group showed poorer OS than the high SMD group (median: 5.6 vs 7.4 months, P=0.006). Thus, we analysed survival after progression at 8 weeks of initiation of chemotherapy according to SMD status among non-responders. In this case, the low SMD group showed poorer survival after progression at 8 weeks than the high SMD group (median: 2.2 vs. 3.4 months, P=0.004). Moreover, OS was not associated with SMI in either responder (P=0.489) or non-responder patients (P=0.061).