Three‐dimensional body composition parameters using automatic volumetric segmentation allow accurate prediction of colorectal cancer outcomes

Abstract

Background

Parameters obtained from two‐dimensional (2D) cross‐sectional images have been used to determine body composition. However, data from three‐dimensional (3D) volumetric body images reflect real body composition more accurately and may be better predictors of patient outcomes in cancer. This study aimed to assess the 3D parameters and determine the best predictive factors for patient prognosis.

Methods

Patients who underwent surgery for colorectal cancer (CRC) between 2010 and 2016 were included in this study. Preoperative computed tomography images were analysed using an automatic segmentation program. Body composition parameters for muscle, muscle adiposity, subcutaneous fat (SF) and abdominal visceral fat (AVF) were assessed using 2D images at the third lumbar (L3) level and 3D images of the abdominal waist (L1–L5). The cut‐off points for each parameter were determined using X‐tile software. A Cox proportional hazards regression model was used to identify the association between the parameters and the treatment outcomes, and the relative influence of each parameter was compared using a gradient boosting model.

Results

Overall, 499 patients were included in the study. At a median follow‐up of 59 months, higher 3D parameters of the abdominal muscles and SF from the abdominal waist were found to be associated with longer overall survival (OS) and disease‐free survival (all P < 0.001). Although the 3D parameters of AVF were not related to survival outcomes, patients with a high AVF volume and mass experienced higher rate of postoperative complications than those with low AVF volume (27.4% vs. 18.7%, P = 0.021, for mass; 27.1% vs. 19.0%, P = 0.028, for volume). Low muscle mass and volume (hazard ratio [HR] 1.959, P = 0.016; HR 2.093, P = 0.036, respectively) and low SF mass and volume (HR 1.968, P = 0.008; HR 2.561, P = 0.003, respectively), both in the abdominal waist, were identified as independent prognostic factors for worse OS. Along with muscle mass and volume, SF mass and volume in the abdominal waist were negatively correlated with mortality (all P < 0.001). Both AVF mass and volume in the abdominal waist were positively correlated with postoperative complications (P < 0.05); 3D muscle volume and SF at the abdominal waist were the most influential factors for OS.

Conclusions

3D volumetric parameters generated using an automatic segmentation program showed higher correlations with the short‐ and long‐term outcomes of patients with CRC than conventional 2D parameters.

Introduction

Recently, focus on the body composition of patients with cancer has increased owing to widespread computed tomography (CT) scan usage, leading to more studies predicting outcomes based on the body composition parameters. ¹ , ² , ³ , ⁴ Several representative studies have confirmed the negative effect of abnormal body composition on patient survival ³ , ⁴ , ⁵ and correlation of the same with adverse surgical and medical treatment outcomes. ¹ , ² , ³ , ⁴ , ⁵ However, lack of standardized criteria has led to heterogeneity across studies, and the time‐consuming and labour‐intensive nature of body composition measurements has limited the clinical applicability of these parameters.

Most patients with cancer have an abnormal body composition at the time of treatment initiation, influenced by factors, such as nutrition, lifestyle and metabolic changes caused by tumour‐related inflammation. ⁶ Maintaining or increasing skeletal muscle can significantly improve overall health during treatment, ⁷ whereas muscle loss can lead to decreased function and increased disability. ⁸ Weight loss in patients with cancer results in the depletion of adipose tissue and skeletal muscle, which is different from malnutrition, and cannot be reversed by conventional nutritional support. ⁹ , ¹⁰ , ¹¹

Body surface area (BSA) and body mass index (BMI) are commonly used body indices. However, they do not reflect body composition, which is defined by the proportion and distribution of bones, muscles and adipose tissues. Increased fat content in skeletal muscles may negatively affect muscle function by reducing muscle density, ¹² , ¹³ thereby leading to compromised survival in patients with cancer. ¹ , ¹⁴

Colorectal cancer (CRC) is the third most frequently diagnosed cancer and second in terms of cancer‐related mortality worldwide. ¹⁵ In addition to the tumour stage, the patient's body composition and functional status can affect his or her long‐term survival and may be associated with an increase in postoperative morbidity. ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ Moreover, some authors have pointed out that muscle loss can independently elevate the risk of postoperative complications and delayed recovery. ²⁰ Even patients with CRC, without metastasis, have a high prevalence of muscle abnormalities, and pre‐existing comorbidities in these patients were found to be associated with low skeletal muscle density (SMD). ²¹

Most studies have evaluated muscle and adipose tissue parameters using two‐dimensional (2D) images of the third lumbar (L3) vertebra. ¹ , ³ , ¹⁴ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ Given the high prevalence of CRC and its associated unfavourable prognosis, the clinical application of three‐dimensional (3D) images instead, for assessing treatment outcomes, is being considered. In the present study, we aimed to analyse skeletal muscle, abdominal visceral fat (AVF) and subcutaneous fat (SF), in terms of volume (cubic centimetres) and mass (grams), from 3D images of the abdominal waist (L1–L5 levels) and compare them with conventional parameters from 2D images of the L3 level. We were the first who conducted an analysis of parameters at L3, abdominal waist and whole abdomen levels, aiming to compare them in order to ascertain body composition parameters with high predictive accuracy.

Methods

Study population

Clinical, pathological and survival outcomes were retrieved from prospectively collected cancer databases at the Seoul National University Hospital, Seoul, Korea. Patients who met the following criteria were included in the study: (1) underwent curative‐intent surgery for primary stage II–III CRC between January 2010 and December 2016 and (2) had available preoperative and postoperative abdominal and pelvic CT images. Patients with hereditary CRC syndrome, metastatic disease (M1), synchronous or metachronous cancer, CT without contrast enhancement or a follow‐up period of <5 years were excluded. Patients who received preoperative neoadjuvant chemotherapy or chemoradiotherapy were also excluded.

Clinical data

Patient demographics and clinical data, including age, sex, BMI, American Society of Anesthesiologists grade, comorbidities and preoperative symptoms, such as obstruction, perforation, bleeding and tumour stage, according to the 8th edition of the American Joint Committee on Cancer (AJCC), were recorded. The postoperative complications and pathological data were analysed.

Overall survival (OS) was defined as the time from the date of surgery to death from any cause. Disease‐free survival (DFS) was defined as the time from the date of surgery to CRC recurrence or death from any cause. CRC recurrence included local and distant metastases. Local recurrence was defined as pathological, radiological or endoscopic evidence of recurrent CRC at a previous anastomosis or near the operative fields. Distant metastasis was defined as pathological or radiological evidence of recurrent CRC in organs other than the colorectum, such as the liver, lungs and peritoneum. We collected data until the patient's death or till the end of June 2021, when the follow‐up period exceeded 5 years.

Segmentation of body composition using DeepCatch

Automatic segmentation of body composition was performed using preoperative abdominal and pelvic CT scans with DeepCatch (https://medicalip.com/DeepCatch; DeepCatch, Version 1.X.X.X; MEDICAL IP Co., Ltd., Seoul, Korea). Preoperative CT was performed within 3 months of surgery. The preoperative CT was performed within 3 months before surgery.

Images of abdominal and pelvic CT with contrast enhancement were uploaded as Digital Imaging and Communications in Medicine (DICOM) files to the DeepCatch program, and patients' body composition was segmented into seven areas, namely, the skin, bone, muscle, AVF, SF, internal organs and central nervous system, with different colours using the program. ²⁴ Convolutional neural network, a deep learning model, was used to learn the metadata from CT images. ²⁵ After measuring the linear attenuation coefficient of the Hounsfield unit (HU) and pixels from the CT image, the volume, area and linear attenuation of body composition were calculated in the boundary line of the body area and presented in a report.

Body composition parameters were measured at the L3 vertebral level and abdominal waist (L1–L5).

Definition of two‐dimensional body composition parameters

To evaluate skeletal muscle area, the conventional 2D parameters included the skeletal muscle index (SMI), SMD, skeletal muscle gauge (SMG) and normal attenuation muscle area (NAMA)/total abdominal muscle area (TAMA). SMI was calculated by dividing the total muscle surface area at the L3 level by the square of the patient's height (cm²/m²). ⁷ SMD was defined as the mean muscle attenuation in HU of the muscle voxels. ²⁶ SMG (cm² × HU/m²) was calculated by multiplying SMI and SMD. ⁴

To assess the skeletal muscle quality, amount of fat in the muscle (myosteatosis) was evaluated. Skeletal muscle was sub‐segmented into three groups, based on the amount of intramuscular fat measured by the HU of CT images, namely, (1) intramuscular adipose tissue (−190 to −30 HU), (2) low‐attenuation muscle area (−29 to +29 HU) and (3) NAMA (+30 to +150 HU). ¹³ , ²⁷ The NAMA/TAMA ratio was used as an indicator of muscle quality.

To assess abdominal fat, the AVF area (square centimetres) and the SF area (square centimetres) were calculated and normalized by height (square metres).

Definition of three‐dimensional body composition parameters

The volume of muscle and fat was determined from the tissue areas and distances between the scans of the skeletal muscle, AVF and SF. The skeletal muscle mass (grams) was estimated by multiplying the muscle volume with muscle density (1.06 g/cm³), and the fat mass (grams) was estimated by multiplying fat volume with adipose tissue density (−0.92 g/cm³). ²⁸

Determination of cut‐off points

As the body compositions of men and women are different, the cut‐off points were determined separately according to sex. We used X‐tile software (Rimm Laboratory, Yale School of Medicine, New Haven, CT, USA) to determine the cut‐off points. OS was used as an outcome measure with a minimal P‐value approach for cut‐off optimization. ²⁹

Statistical analysis

Demographic data are presented as numbers, percentages or means with standard deviations or ranges, as appropriate. The χ ² test or Student's t‐test was used to compare categorical or continuous variables according to the body composition. The parameters were grouped as dichotomous variables based on the cut‐off points.

OS and DFS were analysed using the Kaplan–Meier method. The log‐rank test and univariate Cox regression model were used to evaluate the prognostic factors for survival among the high and low muscle or fat groups. In multivariate analysis, age, sex, comorbidities and AJCC stage were used as fixed adjustment factors, and other significant factors in univariate analysis were included in the multivariate analysis using a backward stepwise process. Correlation analysis was performed between body composition parameters and treatment outcomes, including mortality, relapse and complications. A gradient boosting model (GBM) was used to compare the relative influence of body composition parameters on OS.

Statistical significance was set at P < 0.05, and confidence interval (CI) was set at 95%. Analyses were conducted using the SPSS software Version 25 (SPSS, Chicago, IL, USA) and R (R Core Team, 2021).

Results

Baseline characteristics

A total of 2127 patients, who underwent surgery for CRC between January 2010 and December 2016, were screened for eligibility, and 499 patients were finally analysed (Figure  1 ). The mean time between preoperative CT and surgery day was 10.6 days (range, 1–90 days).

Figure 1.

Flowchart. AI, artificial intelligence; CRT, chemoradiotherapy; CT, computed tomography.

Patient characteristics according to 3D body composition parameters in the abdominal waist are shown in Table S1 . Patients with low muscle mass and volume were older and had a lower BMI (all P < 0.001). Male patients had higher AVF masses and volumes (all P < 0.001) and lower SF masses (P = 0.004). More patients with low SF volume had obstructive symptoms before surgery than those with high SF volume (P = 0.003). The mean tumour size was larger in patients with low muscle mass and low SF mass and volume; however, the AJCC stage, lymphatic invasion, venous invasion and perineural invasion were not significantly different between the groups. Postoperative complications were more frequent in the high AVF mass and volume groups (27.4% vs. 18.7%, P = 0.021, and 27.1% vs. 19.0%, P = 0.028, respectively); however, the Clavien–Dindo grade was not significantly different between the groups.

Survival analysis using three‐dimensional parameters

Cut‐off points for each parameter are listed in Table 1 .

Table 1.

Definitions of cut‐off points for body composition parameters

Variable	2D				3D
	L3		Cut‐off		Abdominal waist	Cut‐off
			Male	Female		Male	Female
Muscle	Muscle index	SMI, cm2/m2	39.0	40.1	Muscle mass, g	316.8	283.5
		SMD, HU	34.8	33.9
		SMG, cm2 × HU/m2	1666.5	1433.3
	Muscle adiposity	NAMA/TAMA	0.791	0.707	Muscle volume, cm3	665.9	607.7
Abdominal visceral fat (AVF)		AVF area, cm2/m2	23.8–89.8	84.1	AVF mass, g	132.6	516.1
					AVF volume, cm3	381.9	1235.4
Subcutaneous fat (SF)		SF area, cm2/m2	40.2	104.3	SF mass, g	200.2	309.7
					SF volume, cm3	393.2	723.2

Abbreviations: 2D, two‐dimensional; 3D, three‐dimensional; HU, Hounsfield unit; NAMA, normal attenuation muscle area; SMD, skeletal muscle density; SMG, skeletal muscle gauge; SMI, skeletal muscle index; TAMA, total abdominal muscle area.

The median follow‐up time was 59 (range, 12–122) months. Kaplan–Meier survival analysis revealed significantly worse survival in patients with low muscle mass in the abdominal waist than in those with high muscle mass (5‐year OS: 70.2% vs. 86.9%, P < 0.001; 5‐year DFS: 61.7% vs. 81.0%, P < 0.001; Figure 2 A,B ). Similarly, the low muscle volume group showed significantly worse survival rates than the high muscle volume group (5‐year OS: 62.5% vs. 86.0%, P < 0.001; 5‐year DFS: 54.2% vs. 78.7%, P < 0.001; Figure 2 C,D ).

Figure 2.

Kaplan–Meier curves for overall survival (OS) and disease‐free survival (DFS) according to muscle mass and volume in the abdominal waist. (A) OS in high versus low muscle mass, (B) DFS in high versus low muscle mass, (C) OS in high versus low muscle volume and (D) DFS in high versus low muscle volume.

In the analysis of abdominal fat composition, patients with a low SF mass in the abdominal waist showed significantly lower OS and DFS than those with a high SF mass (5‐year OS: 70.8% vs. 87.9%, P < 0.001; 5‐year DFS: 61.7% vs. 81.0%, P < 0.001; Figure 3 A,B ). The results of SF volume were similar (5‐year OS: 63.5% vs. 86.1%, P < 0.001; 5‐year DFS: 57.7% vs. 78.5%, P < 0.001; Figure 3 C,D ). However, there was no significant difference in OS or DFS between the low and high AVF mass and volume groups (Figure  4 ).

Figure 3.

Kaplan–Meier curves for overall survival (OS) and disease‐free survival (DFS) according to subcutaneous fat (SF) mass and volume in the abdominal waist. (A) OS in high versus low SF mass, (B) DFS in high versus low SF mass, (C) OS in high versus low SF volume and (D) DFS in high versus low SF volume.

Figure 4.

Kaplan–Meier curves for overall survival (OS) and disease‐free survival (DFS) according to abdominal visceral fat (AVF) mass and volume in the abdominal waist. (A) OS in high versus low AVF mass, (B) DFS in high versus low AVF mass, (C) OS in high versus low AVF volume and (D) DFS in high versus low AVF volume.

Cox regression for multivariate analysis

In univariate analysis of clinical factors, age, obstruction, AJCC stage, lymphatic invasion and venous invasion were significant risk factors for OS (Table  S2 ). Obstruction, AJCC stage, lymphatic invasion, venous invasion and perineural invasion were significant prognostic factors for DFS.

In multivariate analysis of the 2D parameters, low SMI, SMD, SMG and AVF area were independent risk factors for OS (hazard ratios [HRs] 2.008, 2.430, 2.328 and 2.568, respectively; all P < 0.05; Table 2 ), whereas a low SMD was an independent risk factor for DFS (HR 2.049; P = 0.018).

Table 2.

Multivariate analysis of body composition parameters for overall and disease‐free survival

Variables	Overall survival			Disease‐free survival
	HR	95% CI	P	HR	95% CI	P
2D parameters from L3 level
SMI, cm2/m2
High	1.0
Low	2.008	1.139–3.540	0.016	1.316	0.756–2.292	0.332
SMD, HU
High	1.0
Low	2.430	1.327–4.449	0.004	2.049	1.129–3.718	0.018
SMG, cm2 × HU/m2
High	1.0			1.0
Low	2.328	1.353–4.006	0.002	1.629	0.954–2.782	0.074
AVF area, cm2
High	1.0			1.0
Low	2.568	1.373–4.802	0.003	1.382	0.698–2.736	0.353
SF area, cm2
High	1.0			1.0
Low	1.770	0.944–3.320	0.075	0.860	0.494–1.499	0.596
NAMA/TAMA
High	1.0			1.0
Low	1.478	0.861–2.536	0.156	1.113	0.682–1.816	0.668
3D parameters from abdominal waist
Skeletal muscle mass, g
High	1.0
Low	1.959	1.132–3.391	0.016	1.507	0.871–2.607	0.142
Skeletal muscle volume, cm3
High	1.0
Low	2.093	1.049–4.175	0.036	2.644	1.395–5.012	0.003
AVF mass, g
High	1.0			1.0
Low	1.127	0.522–2.431	0.761	0.794	0.355–1.778	0.575
AVF volume, cm3
High	1.0			1.0
Low	1.272	0.610–2.653	0.521	0.704	0.316–1.569	0.390
SF mass, g
High	1.0			1.0
Low	1.968	1.191–3.251	0.008	1.385	0.836–2.296	0.207
SF volume, cm3
High	1.0			1.0
Low	2.561	1.390–4.720	0.003	2.123	1.144–3.937	0.017

Abbreviations: 2D, two‐dimensional; 3D, three‐dimensional; AVF, abdominal visceral fat; CI, confidence interval; HR, hazard ratio; HU, Hounsfield unit; NAMA, normal muscle attenuation area; SF, subcutaneous fat; SMD, skeletal muscle density; SMG, skeletal muscle gauge; SMI, skeletal muscle index; TAMA, total abdominal muscle area.

For the 3D parameters, low skeletal muscle mass and volume and low SF mass and volume in the abdominal waist (HRs 1.959, 2.093, 1.968 and 2.561, respectively; all P < 0.05) were significant independent risk factors for OS (Table  2 ). Low skeletal muscle volume (HR 2.644; P < 0.05) and low SF volume in the abdominal waist (HR 2.123; P = 0.017) were significant risk factors for DFS.

Correlation analysis for death, relapse and complications

All 2D and 3D skeletal muscle parameters, except for NAMA/TAMA, were negatively correlated with mortality (Table  3 ). Relapse was negatively correlated with 2D and 3D muscle parameters, except for SMI and NAMA/TAMA; 3D SF parameters were also negatively correlated with death and relapse. Postoperative complications were positively correlated with AVF parameters.

Table 3.

Correlation coefficients between body composition parameters and cancer treatment outcomes

Variables		Death		Relapse		Complications
		Correlation coefficients	P	Correlation coefficients	P	Correlation coefficients	P
Muscle parameters
2D	SMI	−0.084	0.060	−0.053	0.240	0.097	0.031
	SMD	−0.127	0.004	−0.104	0.020	−0.053	0.240
	SMG	−0.157	<0.001	−0.112	0.012	−0.028	0.537
	NAMA/TAMA	−0.070	0.078	−0.038	0.396	−0.030	0.503
3D	Muscle mass	−0.177	<0.001	−0.135	0.003	0.032	0.470
	Muscle volume	−0.188	<0.001	−0.190	<0.001	0.033	0.458
Abdominal visceral fat parameters
2D	AVF area	−0.087	0.053	−0.047	0.299	0.091	0.043
3D	AVF mass	−0.010	0.821	−0.003	0.938	0.098	0.028
	AVF volume	−0.014	0.761	0.003	0.949	0.103	0.021
Subcutaneous fat parameters
2D	SF area	−0.044	0.325	−0.001	0.989	0.025	0.573
3D	SF mass	−0.197	<0.001	−0.137	0.002	−0.015	0.739
	SF volume	−0.188	<0.001	−0.154	0.001	−0.031	0.484

Note: Bold data indicate significant correlations (P < 0.05). Abbreviations: 2D, two‐dimensional; 3D, three‐dimensional; AVF, abdominal visceral fat; NAMA, normal muscle attenuation area; SF, subcutaneous fat; SMD, skeletal muscle density; SMG, skeletal muscle gauge; SMI, skeletal muscle index; TAMA, total abdominal muscle area.

Gradient boosting model for comparison of parameters

GBM demonstrated that muscle and fat volumes in the abdominal waist had the greatest influence on OS (Figure  5 ).

Figure 5.

Gradient boosting model for relative influences of parameters on overall survival: (A) muscle parameters and (B) subcutaneous fat (SF) parameters. 2D, two‐dimensional; 3D, three‐dimensional; NAMA, normal muscle attenuation area; SMD, skeletal muscle density; SMG, skeletal muscle gauge; SMI, skeletal muscle index; TAMA, total abdominal muscle area.

Discussion

The present study demonstrated that the 3D volumetric parameters measured in the abdominal waist (L1–L5 levels) are highly correlated with death, relapse and postoperative complications compared with the well‐known conventional 2D parameters at the L3 level. The cut‐off points for the 3D parameters set in this study separated the survival curves well. High volumes of skeletal muscle and SF were correlated with a low risk of death and relapse, and a high volume of AVF was correlated with an increased risk of postoperative complications.

Sarcopenia measured at the L3 level, using abdominal CT, has been studied comprehensively till date and proven to be an independent predictor of survival in patients with CRC. ³ , ³⁰ , ³¹ , ³² SMI, SMD, SMG and NAMA/TAMA have been reported to be the independent prognostic factors for survival of patients with CRC, and these parameters showed negative correlations with death and recurrence in the present study as well. Nevertheless, the 3D parameters of muscle mass and volume showed even higher correlations compared with muscle parameters at the L3 level with death and relapse in the present study, which could be due to the fact that wider 3D parameters reflect the patient's whole‐body composition more accurately than narrow 2D parameters.

The DeepCatch program automatically determines 3D body compositions retrieved from numerous 2D images within a short period of time, leading to high clinical applicability. The manual segmentation process requires long time, intensive labour, adequate training and two or more researchers for correction. The DeepCatch program produces body composition data in <5 min per patient, thereby providing clinicians with sufficient time to use the parameters before the initiation of treatment to improve the outcomes. The program provides extensive information in a simple and understandable form that can be quickly processed by a specialist with any amount of work experience.

The present study included patients with rectal and colon cancers. The patients were analysed in previous studies using 2D parameters, ³ , ²⁰ , ²¹ , ³⁰ , ³¹ , ³² , ³³ , ³⁴ and our subgroup analysis was only for patients with colon cancer (n = 408); in all cases, the results were consistent with those of patients with CRC. For muscle mass of patients with colon cancer, the 5‐year OS was 91.6% versus 77.6% (P < 0.001) and the 5‐year DFS was 87.0% versus 77.6% (P = 0.021) (data not shown). For muscle volume, the 5‐year OS rate was 90.4% versus 73.5% (P < 0.001) and the 5‐year DFS rate was 86.9% versus 67.6% (P < 0.001). AVF did not have a significant effect on patient survival in terms of fat mass and volume. SF mass and volume had significant effects on OS (91.0% vs. 81.4%, P = 0.011, and 89.9% vs. 78.1%, P = 0.014, respectively) but not on DFS. To increase applicability of the study results, we included both colon and rectal cancers in the present study.

When considering the higher correlations of 3D parameters, measured at L1–L5, over the 2D parameters, in L3 single‐cut images, we may assume the whole abdominal images to possibly have even higher correlations with treatment outcomes, as they cover more areas of the body. Further analysis showed that the survival curves, according to the groups, by whole abdominal parameters closely resembled those observed in the abdominal waist circumference groups (Table S3 and Figures S1 – S3 ). Using Cox regression analysis, we identified low muscle mass, muscle volume and SF volume in the whole abdomen as independent predictors of unfavourable OS (all P < 0.05; Table S4 ). For DFS, low muscle volume was an independent prognostic factor for negative outcomes (P = 0.030). Notably, the correlation analysis showed similar results for the abdominal waist parameters (Table  S5 ). However, we chose the 3D parameters of the abdominal waist for the study, as the images used for whole abdominal analysis in this study covered different regions of the abdomen between individual cases.

AVF mass and volume showed strong correlations with postoperative complications in our study. Complication rate in the low muscle volume group was 20.2%, which was not significantly different from that in the high muscle volume group (23.7%, P = 0.470; Table S1 ). The Clavien–Dindo grades of complications were not significantly different either (5.3% vs. 13.5% for grade ≥3, P = 0.324; Table S1 ). In the present study, complication rate was related to AVF volume and mass.

These findings underscore the significance of AVF in assessing the risk of postoperative complications in patients with CRC. Van Vugt et al. had reported that low skeletal muscle mass and density are associated with impaired postoperative recovery and increased complications after CRC surgery. ³³ Ding et al. reported that patients with visceral obesity had longer surgery time, greater blood loss, greater length of the resected bowel and higher intraoperative and postoperative complication rates than those without visceral obesity. ³⁵ This could possibly be due to the technical difficulties associated with surgery in patients with obesity. In addition, this could be attributed to an abnormal systemic metabolism profile, as adipose tissue is an active endocrine organ involved in the development of metabolic syndrome and can affect the regulation of inflammation. ³⁴

Emerging opportunities to assess and modify muscle and fat tissues before and after cancer treatment will stimulate the development of novel treatment protocols. Treating patients with chronic diseases, including cancer, is well known to be expensive for both patients and the government. ³⁶ Therefore, preoperative body composition analysis could offer a new opportunity not only for identifying high‐risk patients before surgery but also for identifying those who have the potential to incur high costs. Development of the concept and application of a tailored approach throughout the entire treatment period for each patient, according to their body composition parameters and the principles of personalized medicine, can have a clinically significant perspective.

This study, however, had some limitations. First, this was a retrospective study, with an inevitable selection bias. The patients were randomly selected, which might have produced a significant bias in the results despite the adjusted multivariate analysis. Second, new cut‐offs were set in the present study population, and those were not validated in other study cohorts. As Asians have different body compositions from Western populations and the 3D body compositions of patients with CRC have not been reported, new cut‐off points were considered for this study. The cut‐off points should be validated in another cohort of Asian patients with CRC to confirm our study results. Third, data on the patients' social level, sports activities, habits and diet, which can also affect body composition, were not included. In addition, the interactions between changes in body composition, immunologic responses, chemotherapy and cancer outcomes were not analysed. Further research exploring the relationships across environmental factors, body composition and cancer outcomes would be required in future.

In conclusion, the 3D volumes of skeletal muscle and SF in the abdominal waist could serve as more clinically valuable predictors of survival of patients with CRC than 2D single‐cut composition parameters. AVF volume and mass correlated with the risk of postoperative complications. The broad scope of assessment allowed for a more comprehensive evaluation of body composition and more accurate prediction of short‐ and long‐term outcomes in patients with CRC.

Conflict of interest statement

Sang Joon Park is the founder and CEO of MEDICAL IP Co., Ltd. The authors declare no conflicts of interest.

Supporting information

Figure S1. Kaplan–Meier curves for overall survival (OS) and disease‐free survival (DFS) according to muscle mass and volume in the whole abdomen. (A) OS in high vs. low muscle mass, (B) DFS in high vs. low muscle mass, (C) OS in high vs. low muscle volume, (D) DFS in high vs. low muscle volume.

Figure S2. Kaplan–Meier curves for overall survival (OS) and disease‐free survival (DFS) according to subcutaneous fat (SF) mass and volume in the whole abdomen. (A) OS in high vs. low SF mass, (B) DFS in high vs. low SF mass, (C) OS in high vs. low SF volume, (D) DFS in high vs. low SF volume.

Figure S3. Kaplan–Meier curves for overall survival (OS) and disease‐free survival (DFS) according to abdominal visceral subcutaneous fat (AVF) mass and volume in the whole abdomen. (A) OS in high vs. low AVF mass, (B) DFS in high vs. low AVF mass, (C) OS in high vs. low AVF volume, (D) DFS in high vs. low AVF volume.

Table S1. Clinical characteristics according to the 3D body composition parameters of the abdominal waist.

Table S2. Univariate analysis of clinical factors associated with overall survival and disease‐free survival.

Table S3. Definition of cut‐off points for whole‐abdomen (3D) body composition parameters.

Table S4. Multivariate analysis of whole‐abdomen (3D) body composition parameters for overall and disease‐free survival.

Table S5. Correlation coefficients between whole‐abdomen (3D) body composition parameters and cancer treatment outcomes.

Bimurzayeva A., Kim M. J., Ahn J.‐S., Ku G. Y., Moon D., Choi J., et al (2023) Three‐dimensional body composition parameters using automatic volumetric segmentation allow accurate prediction of colorectal cancer outcomes, Journal of Cachexia, Sarcopenia and Muscle, doi: 10.1002/jcsm.13404