Combined test of third lumbar skeletal muscle index and prognostic nutrition index improve prognosis prediction power in resected colorectal cancer liver metastasis

Patients’ common characteristics and survival

539 patients with CRLM were recruited in this study. Mean (SD) age was 60.6 (±11.7) years old, ranging from 24 to 92, and 355 (65.9%) patients were male. Univariate and multivariable Cox regression analysis for OS are outlined in Table 1. Higher T stage, higher N stage, lower differentiation grade and microvascular involvement were significantly associated with worse outcomes (all P-value<0.05). Besides, patients received simultaneous resection of liver metastasis demonstrated better survival outcomes (P-value<0.01). Interestingly, there were no statistically significant differences between study participants in regard to gender, age, primary tumor location and peri-neural invasion in our series (P-value>0.05). From multivariable analysis, we found that T stage (P-value=0.02), tumor differentiation (P-value=0.01), Baseline SMI level (P-value<0.01) and simultaneous metastasectomy status (P-value<0.01) were regarded as independent risk indicators for OS.

Correlation of baseline physiological compositions and clinical characteristics

Median follow-up duration was 24 months, ranging from 3 months to 77 months. Median BMI was 24.1 kg/m² (ranging from 14.9 to 36.8 kg/m²), including males of 26.0 kg/m² and females of 21.7 kg/m² (P-value<0.001), while median SMI value of males has no statistical difference compared to females (40.5 cm²/m² versus 39.0 cm²/m², P-value=0.064). Linear regression relationship between SMI and PNI was determined in Figure 1 (P<0.05) and the R square for relation was 0.01.

When cut-off values for SMI were applied (43 cm²/m² for males and 41 cm²/m² for females), the cohort was divided into 309 (57.3%) and 230 (42.7%) patients with low and high SMI. Baseline low SMI was observed in 116 of 184 females (63.0%) and 193 of 355 males (54.4%) (P=0.05). Comparisons of clinical features of CRLM patients with low or high-SMI were shown in Table 2. Patients with low SMI had significantly smaller pre-albumin (P-value=0.01), smaller PNI (P-value=0.04) and shorter hospital stays (P-value=0.03) compared with those with high SMI. There was no difference in the value of transferrin, CRP, lymphocyte count, CEA and CA19-9. To identify the prognostic role of baseline SMI, ROC curves and Kaplan-meier analysis were performed. As is shown in Figure 2, time-dependent ROC curves were separately constructed to compare clinical value of baseline SMI and PNI. Area under curve (AUC) of baseline SMI and PNI were 0.69, 0.57 in Figure 2A and 2B, respectively. CRLM patients with low SMI was demonstrated to have shorter duration of PFS (Figure 2C, P-value=0.02, HR: 1.31; 95%CI: 1.00-1.71) and OS (Figure 2D, P-value<0.001, HR:1.728; 95%CI: 1.186-2.517) than their high SMI counterparts.

SMI as an indicator on treatment response and tumor progression

CRLM patients was divided into two different cohorts based on whether simultaneous liver metastasectomy was performed: simultaneous liver resection (SLR) cohort and no simultaneous liver resection (N-SLR) cohort. Totally, SLR cohort patients had longer duration of PFS compared with N-SLR patients (Figure 3A, P-value<0.01, HR: 3.19; 95%CI: 2.42-4.21). Specifically, SLR cohort included 181 patients, occurrence or metastasis was observed in 50 patients (27.6%), while N-SLR cohort included 358 patients and occurrence or metastasis was observed in 194 patients (54.2%).

To compare indicative efficacy of SMI on therapeutic response, patients at the status of PR (namely during the treatment process had increased SMI, whereas patients regarded as having PD/SD had decreased SMI (loss of muscle volume) (Figure 3B, Fisher’s exact test, P-value<0.01). Table 3 included SMI value in initial diagnosis, PR, SD and PD status. Furthermore, a subgroup analysis was also performed to compared indicative role of SMI. In SLR cohort, median value of SMI at status of initial diagnosis, PR, SD and PD was separately 41.87cm²/m², 46.89cm²/m², 34.36 cm²/m² and 28.55 cm²/m² (P-value<0.01), whereas value of SMI at status of initial diagnosis, PR, SD and PD was 39.17 cm²/m², 45.51 cm²/m², 34.75 cm²/m² and 30.47 cm²/m² (P-value<0.001), which is also shown in Figure 3C and 3D.

6 months’ loss of SMI is indicative for worse prognosis in CRLM patients

To explore the indicative clinical value of SMI changes in CRLM patients, further analysis was performed. Table 4 confers all changes of SMI and SMD in each period after the operation. Compared with the significant changes in muscle area (MA), our analysis showed that muscle density was not statistically changed after operation. For the change of MA, at the first postoperative evaluation of SMI based on CT scan (3 months after operation), median change of MA was -3.7% (Figure 4A) (skeletal muscle mass decreased by 3.7 percent), ranging from -25.6% to 19.4% (skeletal muscle mass increased by 19.4 percent). On the following 6 and 12 months postoperative CT scan, median change of MA was separately -8.8% (95% CI: -11.7% to -2.3%) and -10.1% (95% CI: -15.2% to -6.5%) (Figure 4C and 4E). It is reported that 8% decreased SMI was regarded as the cut-off value of major SMI loss [33]. Figure 4B demonstrated that CRLM patients with loss of SMI at the 3 months CT scan had not statistically different survival outcome with another SMI-stable group (P=0.61). As are shown in Figure 4D and 4F, 8% cut-off loss of SMI recorded from more than 6 months postoperative CT scan was regarded as a significant risk factor indicating worse prognosis in CRLM patients (P-value<0.001). Subgroup analysis was also performed for SLR cohort patients and N-SLR cohort patients, respectively (Supplementary Figure 2A–2F).

Prognostic efficacy of PNI in CRLM patients

Since 3, 6 and 12 month PNI were also recorded, prognostic efficacy of PNI was also calculated. Figure 5 revealed that, for all patients, AUC of 3 months, 6 months and 12 months were separately 0.54, 0.73 and 0.67. Further analysis was also performed in both SLR and N-SLR group patients. As are shown in Supplementary Figure 2G–2L, 6 months PNI demonstrated more accurate prognostic value with a higher AUC in SLR (0.72) and N-SLR cohorts (0.73).

Consistency analysis of SMI and PNI in resected CRLM patients

Totally, there is no statistical difference of median SMI between baseline and 3 months postoperative CT scan (Table 4), whereas 6 months and 12 months CT scan showed totally decreased SMI (P-value<0.001). As status of SMI and PNI in 6 and 12 months conferred prognostic significance, Kappa analysis of was made to compare the consistency among 3, 6 and 12 months. As is shown in Table 5, Kappa value of SMI and PNI between 6-month’s and 12-month’s cohort were 0.62 and 0.53 (P-value<0.001). In contrast, 3 months’ SMI and PNI were not statistically significantly consistent with 6 months’ and 12 months’ result.

Combined test of post-operative SMI and PNI improve the prognosis prediction power in CRLM patients

We further constructed the binary logistic regression model combining the tests of SMI and PNI. Figure 6 depicted prognostic potency of different postoperative period (3-, 6- and 12-month) with ROC curves. Furthermore, when the two biomarkers were combined using the binary logistic modeling, all three AUCs were significantly better than single SMI or PNI test. AUC of 3-month, 6-month and 12-month were separately 0.65, 0.85 and 0.80 (Figure 6).

Patients and characteristics

The study was approved by the Ethical Committee of Zhongshan Hospital, Fudan University. All included patients met the following criteria: 1) primary tumor resection with or without simultaneous liver metastasectomy of CRLM at Fudan University Zhongshan Hospital between January 2013 and December 2016; 2) available blood test records; 3) available follow-up information. The exclusion criteria were as follows: patients with infections, hematological disease, hyperpyrexia, and intestinal obstruction at test were excluded; patients with no available CT images and incomplete clinical and pathologic data were excluded. The type of surgical resection and the extent of lymph node dissection were selected according to Chinese colorectal cancer treatment guidelines [28, 29]. Tumor response including partial response (PR), stable disease (SD) and progressive disease (PD) were based on response evaluation criteria in solid tumors (RECIST 1.1) [30]. All patients provided written informed consent. Details of flow gram was demonstrated in Supplementary Figure 1.

Preoperative anthropometric and blood-chemistry measurements

Physical status and preoperative laboratory values were obtained within 1 week prior to initiation of surgery. Body mass index (BMI) was calculated by dividing the body weight in kilograms by the square of the height in meters. CT was performed on an average of 4.2 days (range, 1–12 days) before surgical operation using a 320-slice multi-detector CT scanner (Aquilion ONE; Toshiba Medical Systems Corporation, Otawara, Japan) [31]. Skeletal muscle areas, including the psoas, erector spinae, quadratus lumborum, transversus abdominis, external and internal obliques, and rectus abdominis, were identified and quantified using -29 to 150 Hounsfield units. Specifically, CT images of less than 12-month’s post-operation were recorded as three cohorts (3-month cohort, 6-month cohort and 12-month cohort). SMI was calculated by normalizing skeletal muscle areas for height (cm²/m²) [32]. As is reported in 2013 [33], patients with SMI<41 cm²/m² (women) and <43cm²/m² (men) were significantly associated with low survival, thus we choose this value as a baseline cut-off value. PNI was calculated based on the serum albumin and total lymphocyte count, using the following equation: PNI=serum albumin (g/dL) + 5*total lymphocyte count (/mL) [34].

Follow-up and statistics

The follow-up period was as follows: first follow-up was performed within 2–3 months after the baseline and subsequent follow-up cycles usually ranged from 3 to 6 months or even shorter which depended on the clinical situations and (or) tumor relapse or metastasis being suspected. Data of overall survival (OS) and relapse/metastasis time were also documented. Duration of OS was calculated from the date of baseline to the last follow-up or tumor-specific death. Progression free survival (PFS) was determined from the date of diagnosis to the date of progression under the regular follow-up.

Statistical analyses were performed using the SPSS statistical package (22.0; SPSS) and prism 6 (GraphPad Software, Inc., 2015). Mean changes in SMI were analyzed with paired t-tests. The correlations between continuous valuables were analyzed using Spearman rank correlation test. Predictive accuracy of tests was assessed by calculating AUC. Besides, Time-dependent cut-off values were determined when positive likelihood ratio (PLR) were the largest one [35, 36]. PLR was calculated as follows: PLR=sensitivity/(1–specificity). Binary logistic regression model of SMI and PNI were estimated based on SPSS 22.0. PFS and OS analyses were carried out using the Kaplan–Meier method and results were compared using a log-rank test. A multivariable Cox proportional hazards model predicting OS was performed using backward stepwise selection. Risk factors were expressed as the hazard ratio [HR, 95% confidence interval (CI)]. Statistical significance was defined as P-value less than 0.05.