Patient selection
Prospective follow-up data of 1297 patients with colon cancer treated at the Affiliated Hospital of Jiangnan University between April 1, 2015 to May 31, 2019 were retrospectively collected. The patients were randomly allocated in a 2:1 ratio to the primary and validation cohorts using R version 4.0.3. The inclusion criteria for both cohorts were as follows: (1) age of 18-60 years, (2) postoperative pathological diagnosis of colon cancer, (3) undergoing radical resection, and (4) TNM stage I to III. The exclusion criteria were as follows: (1) incomplete clinical data, (2) death in the hospital or within 30 days postoperatively, (3) preoperative neoadjuvant therapy, (4) concurrent secondary tumor or rectal cancer, (5) previous history of other malignancies, and (6) pregnancy. This study was approved by the Ethics Committee of the Affiliated Hospital of Jiangnan University (no. 20220025), and the review board waived the requirement for informed consent from all subjects.
Data collection
Through the prospectively maintained colon cancer database, the following clinical data were carefully collected for each patient: demographics and clinical features [sex, age, body mass index (BMI), age-adjusted Charlson Comorbidity Index (aCCI), carcinoembryonic antigen (CEA), albumin, and bowel obstruction], nutritional status [nutrition risk screening 2002 (NRS 2002), patient-generated subjective global assessment (PG-SGA), and third lumbar spine skeletal muscle mass index (L3MI)], surgical characteristics [operation type and postoperative comprehensive complication index (CCI)], and pathological information (tumor location, size, differentiation, histology, and stage). These candidate variables of interest were derived from previous studies and clinical experience. Postoperative complications were assessed using CCI scores 34, where a CCI score of 26.2 was considered the threshold for serious complications 35. Tumor staging was performed based on pathological reports using the American Joint Committee on Cancer 8th edition.
Nutritional status assessments
NRS 2002. The NRS 2002 was used to evaluate nutritional risk, which accounts for three parts: disease severity (stress metabolism due to the degree of illness), impaired nutritional status (based on weight loss, BMI, and food intake), and 70-year age adjustment (Supplementary Table 1). Patients are scored using the first two components, a score of 0-3 according to whether they are absent, mild, moderate, or severe, and a score of 1 is added if the patient is aged ≥70 years. The NRS 2002 score, ranging from 0 to 7, was categorized into three groups: no nutritional risk (0-2), nutritional risk (3-4), and severe nutritional risk (≥5) 36.
PG-SGA. The PG-SGA, specially designed for nutritional status assessment of patients with cancer, comprises four patient self-report components (weight loss, food intake, symptoms, activities and functions) and three professional assessment components (disease and its relationship to nutritional needs, metabolic demand, and physical examination of fat, muscle, and fluid status) 37. The original PG-SGA and its rating methods are available on the PG-SGA/Pt-Global Platform (https://pt-global.org/pt-global/). We used only the global PG-SGA categories of well nourished (A), moderately malnourished (B), and severely malnourished (C).
L3MI. The skeletal muscle index of the third lumbar vertebra (L3MI) is widely used to assess the nutritional status and reflects protein-energy malnutrition 38. The OsiriX open-source software (version 8.5.2: Pixmeo Sarl, Geneva, Switzerland) was employed to analyze computed tomography images using muscle-specific Hounsfield Unit thresholds of -29 to 150 39. The cross-sectional areas of the skeletal muscles at L3, including the psoas, erector spinae, quadratus lumborum, transversus abdominis, external and internal obliques, and rectus abdominis muscles (Supplementary Figure 2) were calculated (cm2) and subsequently normalized by the square of the height (m2) to obtain the L3MI. According to the previous study, a priori definition of low L3MI (LL3MI) was used (male: L3MI < 40.8 cm2/m2 and female: L3MI < 34.9 cm2/m2) 40.
Study outcome
The primary outcome of our study was OS, defined as the time from the diagnosis of colon cancer to death from any cause. Information on treatment and living status was obtained through telephone, outpatient, or inpatient follow-ups after completion of the primary therapy. Follow-up visits were scheduled every three months for two years postoperatively and every six months for three years and beyond. The end date of the follow-up was November 30, 2022.
Statistical analysis
Statistical analyses were performed using R (version 4.0.3; http://www.r-project.org/). The measurement data included in this study were converted to categorical variables, presented as frequencies and corresponding percentages. Chi-squared tests were used to compare the baseline features between the primary and validation cohorts. A univariate Cox regression analysis was performed for each variable, and the significant factors were included in the multivariate Cox regression analysis to identify independent risk predictors associated with OS using the enter method. A nomogram was accordingly formulated based on the significant predictors using R version 4.0.3. For user convenience, an interactive dynamic nomogram was constructed using the package of DynNom. Five-fold cross-validation was performed 1000 times for internal validation of the prediction model. The discrimination performance and predictive accuracy were assessed by the concordance index (C-index), receiver operating characteristic curve (ROC), and calibration curve. A decision curve analysis (DCA) was conducted to assess the clinical benefits of the model. The same assessments were performed in the validation cohort. The total points of each patient were calculated based on the established nomogram model, and three groups of patients with different prognostic risks were depicted using X-tile (version 3.6.1) 41. Survival curves were delineated by the Kaplan-Meier analysis according to the dichotomized risk groups and compared using the log-rank test. All tests were two-sided, and statistical significance was set at P < 0.05.