Usefulness of Inflammation-Based Prognostic Scores for Predicting the Risk of Complications After Radical Resection of Colorectal Carcinoma.

PURPOSE: We aimed to investigate the value of inflammation-based prognostic scores for predicting early complications after radical surgery for colorectal carcinoma.
METHODS: We retrospectively analyzed data of 154 patients who underwent elective resection of colorectal carcinoma between January 2017 and December 2018 at Beijing Friendship Hospital. Univariate, multivariate, and receiver operating characteristic curve analyses were conducted. As inflammation indices, we evaluated the preoperative modified Glasgow Prognostic Score (GPS), as well as the C-reactive protein/albumin ratio (CAR), postoperative GPS, and C-reactive protein levels on postoperative day 3 (POD3).
RESULTS: Within 30 days postoperatively, complications occurred in 80 patients (51.9%). And high levels of preoperative mGPS (P=0.002), preoperative CAR (P=0.019), POD3 CAR (P<0.001) and POD3 poGPS (P<0.001) can significantly affect postoperative complications after surgery for colorectal cancer, with CRP on POD3 (odds ratio, 1.015; 95% confidence interval, 1.006-1.024; P=0.001) as independent risk factors. Among all inflammation-based indicators, POD3 CAR had the highest area under the curve (0.711) and positive predictive value (83.2%). Higher CAR (≥2.6) on POD3 was associated with a higher rate of complications (92.9% vs 36.6%, P<0.001), especially of infectious nature (54.8% vs 16.1%, P<0.001).
CONCLUSION: CAR≥2.6 on POD3 reflects sustained systemic inflammation and represents a useful predictor of complications after surgery for colorectal carcinoma, facilitating early detection, timely intervention, and enhanced recovery.

Introduction

Colorectal carcinoma is the third most common malignant tumor worldwide, with surgical resection being the only effective treatment currently available.1,2 Despite important advances in surgical techniques, the overall rate of complications remains high, at around 30%.3 Postoperative complications are associated with increased treatment costs, prolonged hospital stay, delayed adjuvant chemotherapy, increased risk of recurrence, and a detrimental impact on survival.4–7 Therefore, timely and precise detection and treatment of any complication are critical to improving prognosis after colorectal carcinoma surgery.

In recent years, inflammatory response-based prognostic scoring systems have been developed as tools for clinical evaluation to aid in decision-making. The most common predictors include the C-reactive protein/albumin ratio (CAR), modified Glasgow Prognostic Score (mGPS), postoperative Glasgow Prognostic Score (poGPS), and neutrophil-to-lymphocyte ratio. Most studies in patients with colorectal carcinoma focused primarily on the oncologic prognosis and rarely reported on the performance of predictors for postoperative complications. Tissue trauma caused by surgery affects the body’s metabolic, neuroendocrine, and immune response. An immune response is reflected as an increase in the expression of pro-inflammatory cytokines and the subsequent increase in the levels of C-reactive protein (CRP) and albumin in the acute phase protein.8 Warschkow et al conducted a meta-analysis of studies covering 1832 patients and found that increased CRP levels on postoperative days (PODs) 3 and 4 were good predictors of postoperative complications.9 Labgaa et al found that early postoperative reduction in albumin levels was a predictor of postoperative complications.10 Therefore, we hypothesized that indices that combine CRP and albumin information would be useful for patient stratification according to the risk of postoperative complications. In this study, we examined the value of CAR, mGPS, and poGPS as predictors of early postoperative complications in patients who underwent surgery for colorectal carcinoma.

Patients and Methods

Study Design and Selection of Participants

The retrospective study was approved by the Ethics Committee of Beijing Friendship Hospital. Between January 2017 to December 2018, 154 patients without preoperative infection of the respiratory, digestive, or urinary tract underwent radical surgery for colorectal cancer and were enrolled in the study. And the operations were performed by the same group of surgeons. All the patients have signed written informed consent forms for their data to be used in the study and the data were kept confidential. This study was conducted in accordance with the Declaration of Helsinki.

Data Collection

We retrospectively examined the records maintained in our hospital’s database and extracted data regarding age, sex, body mass index, presence of chronic diseases, tumor location, neoadjuvant therapy, and pathological stage. We also collect data regarding the surgery, including the operation time, operation method, and intraoperative blood loss. Laboratory data collected in this study included preoperative albumin levels, preoperative mGPS, CAR on POD3, and poGPS on POD3.

Preoperative and postoperative inflammatory responses, respectively, were evaluated based on the mGPS and poGPS (Table 1), which represent widely validated and independent systemic inflammation-based prognostic scores.11 Many studies have reported that poGPS on PODs 3 and 4 is a predictor of infectious complications.12

Table 1

Overview of the Glasgow Prognostic Score Systems Based on Biomarkers of Systemic Inflammation

Scoring System	Points Allocated
Modified Glasgow Prognostic Score (mGPS)
C-reactive protein ≤10 mg/L and albumin ≥35 g/L	0
C-reactive protein >10 mg/L	1
C-reactive protein >10 mg/L and albumin <35 g/L	2
Postoperative Glasgow Prognostic Score (poGPS)
C-reactive protein <150 mg/L	0
C-reactive protein >150 mg/L and albumin >25 g/L	1
C-reactive protein >150 mg/L and albumin <25 g/L	2

Definition of Postoperative Complications

The severity of complications occurring within 30 days after surgery was graded based on the Clavien-Dindo classification.13 The main types of infectious complications included wound infection, urinary tract infection, pulmonary infection, abdominal pelvic infection, and sepsis. Surgical site infection was classified as superficial incision infection, deep incision infection, and organ infection, according to the guidelines issued by the Centers for Disease Control and Prevention in 2017.14

Data Analysis

Statistical analyses were conducted using SPSS version 22.0 (IBM Corp., Armonk, NY, USA) and MedCalc version 18.2 (MedCalc Software bvba, Ostend, Belgium). Continuous data were expressed as mean ± standard deviation and compared between groups using the independent sample t-test. Categorical data were compared using the chi-square test or Fisher’s exact probability test. To identify independent risk factors for postoperative complications, variables showing significant association (P<0.05) with the outcome on univariate analysis were entered into the multiple logistic regression analysis, and the results were expressed as odds ratios (ORs) with 95% confidence intervals (95% CIs). Receiver operating characteristic curve analysis was used to estimate the initial predictive value of POD3 CAR, preoperative mGPS, and POD3 poGPS. The area under the curve (AUC) was calculated, and the sensitivity, selectivity, positive predictive value (PPV), and negative predictive value (NPV) at the optimal cutoff value (corresponding to the maximum value of the Jordan index) were obtained for each predictor. Relationships with P<0.05 were considered statistically significant.

Results

Study Population and Baseline Characteristics

A total of 154 patients (95 male, 59 female) were included in the study, with an average age of 64.9±11.4 years. Within days postoperatively, complications occurred in 80 patients (51.9%), and two patients died.

Analysis of Possible Risk Factors for Postoperative Complications

Univariate analysis revealed that sex, preoperative albumin levels, preoperative CRP, use of neoadjuvant therapy, operative time, preoperative mGPS, preoperative CAR, POD3 CAR, POD3 poGPS, POD3 CRP and POD3 albumin were significantly associated with the 30-day incidence of postoperative complications (P<0.05) (Table 2). Both POD3 CAR and poGPS were derived from the same CRP and albumin in the same patients and highly interrelated, and the same for preoperative CAR and mGPS. We do not put them in the same multivariate analysis. On multivariate analysis, POD3 CRP (OR=1.015; 95% CI=1.006–1.024;P=0.001) and operative time (OR=1.006; 95% CI=1.001–1.010; P=0.018) were independent risk factors for postoperative complications (Table 3). While preoperative CRP appeared to have similar predictive value as that of POD3 CRP (OR=1.045; 95% CI=0.983–1.112), the association of preoperative CRP with the 30-day incidence of postoperative complications did not reach statistical significance (P=0.160). As POD3 CRP was an independent risk factor for postoperative complications, we speculated that the diagnostic value of POD3 CAR and poGPS was higher than that of preoperative indicators.

Table 2

Univariate Analysis of Risk Factors Associated with Postoperative Complications

Characteristics	All (n=154)	Postoperative Complications		P-value
		Yes (n=80)	No (n=74)
Age,year	64.9±11.4	64.4±12.2	65.4±10.6	0.607
Sex,n(%)				0.004
Male	95(61.7)	58(72.5)	37(50)
Female	59(38.3)	22(27.5)	37(50)
Body mass index, kg/m2	23.7±3.5	23.4±3.6	23.9±3.5	0.382
Chronic disease,n(%)				0.907
Hypertension	59(38.3)	28(35.0)	31(41.9)
Diabetes mellitus	25(16.2)	15(18.8)	10(13.5)
Dyslipidemia	10(6.5)	5(6.3)	5(6.8)
Hepatocirrhosis	1(0.6)	1(1.3)	0(0.0)
Preoperative CRP,mg/L	6.9±10.5	8.7±13.4	4.9±5.2	0.018
Preoperative albumin, g/L	37.1±3.7	36.5±4.2	37.8±3.1	0.032
Preoperative CAR	0.2±0.4	0.3±0.5	0.1±0.1	0.019
Preoperative mGPS≥1,n(%)	26(16.9)	21(26.3)	5(6.8)	0.002
Surgical approach, n(%)				0.782
Laparoscopy	132(85.7)	70(87.5)	62(83.8)
Converted laparoscopy	8(5.2)	4(5.0)	4(5.4)
Laparotomy	14(9.1)	6(7.5)	8(10.8)
Operative time, minutes	216.2±94.2	235.1±99.5	195.7±84.0	0.009
Estimated blood loss, mL	164.4±647.7	239.4±891.5	83.3±79.0	0.123
Location of the tumor,n(%)				0.152
Colon	74(48.1)	34(42.5)	40(54.1)
Rectum	80(51.9)	46(57.5)	34(45.9)
Neoadjuvant treatment,n(%)	9(5.8)	9(11.3)	0(0.0)	0.003
Pathological stage, n(%)				0.534
pCR	2(1.3)	1(1.3)	1(1.4)
0	1(0.6)	0(0.0)	1(1.4)
I	42(27.3)	20(25.0)	22(29.7)
II	49(31.8)	27(33.7)	22(29.7)
III	46(29.9)	22(27.5)	24(32.4)
IV	14(9.1)	10(12.5)	4(5.4)
CRP on POD3, mg/L	57.1±49.8	73.9±54.0	38.9±4.4	<0.001
Albumin on POD3, g/L	33.2±3.7	32.4±3.8	34.0±3.6	0.009
CAR on POD3	1.8±1.7	2.4±1.8	1.2±1.1	<0.001
poGPS≥1 on POD3,n(%)	20(13.0)	18(22.5)	2(2.7)	<0.001

Notes: Data represent mean ± standard deviation or frequency, as appropriate.

Abbreviations: CRP, C-reactive protein; CAR, C-reactive protein/albumin ratio; mGPS, modified Glasgow Prognostic Score; pCR, pathologic complete response; POD, postoperative day; poGPS, postoperative Glasgow Prognostic Score.

Table 3

Multivariate Analysis of Factors Associated with Postoperative Complications

Risk Factor	OR	95% CI	P-value
Sex, male vs female	1.786	0.849–3.758	0.127
Preoperative albumina	0.989	0.881–1.110	0.845
Preoperative CRPb	1.045	0.983–1.112	0.160
Operative timec	1.006	1.001–1.010	0.018
Neoadjuvant treatment	0.496	0.155–1.583	0.237
POD3 CRPb	1.015	1.006–1.024	0.001
POD3 albumina	0.980	0.877–1.095	0.720

Notes: For continuous variables, the ORs were calculated per unit increase (in ag/L, bmg/L,cminutes).

Abbreviations: 95% CI, 95% confidence interval; CRP, C-reactive protein; CAR, C-reactive protein/albumin ratio; OR, odds ratio; POD, postoperative day.

Usefulness of CAR, mGPS, and poGPS as Predictors of Postoperative Complications

Because POD3 CRP was previously reported to serve as a predictor of postoperative complications,9 we conducted receiver operating characteristic curve analysis for all five indicators of inflammatory response (preoperative mGPS, preoperative CAR, as well as CRP, CAR, and poGPS on POD3). For POD3 CAR, the AUC was 0.711 and an optimal cutoff of 2.6 was identified (sensitivity, 51.3%; specificity, 87.8%; PPV, 83.2%; NPV, 61.3%). For preoperative mGPS (Figure 1, Table 4), the AUC was 0.613 and an optimal cutoff of 0.1 was identified (sensitivity, 66.3%; specificity, 50.0%; PPV, 58.9%; NPV, 57.8%). For preoperative mGPS, the AUC was 0.600, with the cutoff value at 0 (sensitivity, 73.8%; specificity, 6.9%; PPV, 46.1%; NPV, 19.2%), 1 (sensitivity, 18.8%; specificity, 93.2%; PPV, 75.0%; NPV, 51.5%) and 2 (sensitivity, 7.5%; specificity, 0.0%; PPV, 100.0%; NPV, 50.0%). For poGPS on POD3, the AUC was 0.599, with cutoff value at 0 (sensitivity, 77.5%; specificity, 2.7%; PPV, 46.3%; NPV, 10.0%), 1 (sensitivity, 21.3%; specificity, 97.3%; PPV, 89.5%; NPV, 53.3%) and 2 (sensitivity, 1.3%; specificity, 0.0%; PPV, 100.0%; NPV, 48.4%). For CRP on POD3, the AUC was 0.706 and an optimal cutoff of 65.5 mg/L was identified (sensitivity, 48.8%; specificity, 79.7%; PPV, 73.2%; NPV, 60.2%).

Figure 1

Receiver operating characteristic (ROC) curve of risk factors associated with postoperative complications.

Abbreviations: CAR, C-reactive protein/albumin ratio; CRP, C-reactive protein; mGPS, modified Glasgow Prognostic Score; POD, postoperative day; poGPS, postoperative Glasgow Prognostic Score; pre, preoperative.

Table 4

Predictive Power Analysis of Risk Factors Associated with Postoperative Complications

Predictor	AUC	P-value	Optimal Cutoff	Sensitivity	Specificity	Positive Predictive Value	Negative Predictive Value
Preoperative CAR	0.613	0.045	0.1	66.3%	50.0%	58.9%	57.8%
Preoperative mGPS	0.600	0.032
			0	73.8%	6.8%	46.1%	19.2%
			1	18.8%	93.2%	75.0%	51.5%
			2	7.5%	0.0%	100.0%	50.0%
POD3 CRP	0.706	<0.001	65.5	48.8%	79.7%	73.2%	60.2%
POD3 CAR	0.711	<0.001	2.6	51.3%	87.8%	83.2%	61.3%
POD3 poGPS	0.599	0.046
			0	77.5%	2.7%	46.3%	10.0%
			1	21.3%	97.3%	89.5%	53.3%
			2	1.3%	0.0%	100.0%	48.4%

Abbreviations: AUC, area under the curve; CAR, C-reactive protein/albumin ratio; CRP, C-reactive protein; mGPS, modified Glasgow Prognostic Score; POD, postoperative day; poGPS, postoperative Glasgow Prognostic Score.

The AUC was above 0.5 for all four variables analyzed, indicating good predictive value for postoperative complications (Figure 1, Table 4). From the perspective of clinical practice, PPV is the most impactful parameter because it reflects the likelihood that the screening will correctly identify patients with the outcome (high risk of postoperative complications). The present findings suggest that there is an 83.2% probability of correctly identifying the high risk of early complications among patients with CAR≥2.6 on POD3. Compared to preoperative CAR, preoperative mGPS, and POD3 poGPS, POD3 CAR provided significantly superior predictive value. Moreover, it reflected in the increased chance to correctly identify patients at risk. Therefore, we further examined the usefulness of POD3 CAR as an early predictor of specific postoperative complications.

CAR as a Predictor of Specific Postoperative Complications

For this analysis, the patients were stratified according to the CAR cutoff: high CAR (≥2.6) versus low CAR (<2.6). Complications occurred in 39 (92.9%) of the 42 patients with CAR≥2.6 and in 41 (36.6%) of the 112 patients with CAR<2.6 (Table 5). The difference between groups was statistically significant (P<0.001).

Table 5

Incidence of Postoperative Complications According to Postoperative C-Reactive Protein/Albumin Ratio (CAR)

Characteristics	CAR<2.6 (n=112)	CAR≥2.6 (n=42)	P-value
Overall	41 (36.6%)	39 (92.9%)	<0.001
Grade I	9 (8.0%)	3 (7.1%)	0.362
Temperature >38°C after surgery	4	1
Superficial incision infection	0	1
Urinary retention	4	1
Transient confusion	1	0
Grade II	21 (18.8%)	18 (42.9%)	<0.001
Postoperative blood transfusion >2 units	3	2
Total parenteral nutrition >2 weeks	1	0
Deep incision infection	2	2
Mild ileus	3	1
Chylous leakage	1	0
Urinary tract infection	3	1
Pneumonia	1	3
Mild abdominal and pelvic infection	7	9
Grade IIIa	7 (6.3%)	9 (21.4%)	<0.001
Acute hepatic injury	1	0
Mild-to-moderate anastomotic leakage	1	1
Organ infection	4	2
Moderate abdominal and pelvic infection	0	3
Mild-to-moderate anastomotic bleeding	0	1
Pleural effusion	1	1
Severe ileus	0	1
Grade IIIb	4 (3.6%)	3 (7.1%)	0.063
Severe abdominal and pelvic infection	1	1
Severe anastomotic leakage	2	1
Severe anastomotic bleeding	1	0
Rectovaginal fistula	0	1
Grade IV	0	4 (9.5%)	0.001
Septic shock	0	4
Grade V	0	2 (4.8%)	0.019
Grade III or greater	11 (9.8%)	18 (42.9%)	<0.001
Postoperative infective complications	18 (16.1%)	23 (54.8%)	<0.001

The incidence of specific complications was examined in each group. Compared to patients with high CAR, those with low CAR on POD3 had a lower incidence of mild complications (grade I or II; 50.0% vs 26.8%, P<0.001), as well as a lower incidence of severe complications (grade III or higher; 42.9% vs 9.8%, P<0.001) (Table 5). In particular, infectious complications occur significantly more frequently in patients with high CAR on POD3 (54.8% vs 16.1%, P<0.001) (Table 5).

Usefulness of CAR as a Predictor of Surgical Site Infection, Infective Complications, and Anastomotic Leakage

For predicting surgical site infection (Figure 2), POD3 CAR exhibited an AUC of 0.760 (95% CI=0.685–0.825), the sensitivity of 75.00%, and specificity of 68.85%. For predicting infectious complications (Figure 3), POD3 CAR exhibited an AUC of 0.702 (95% CI=0.623–0.773), the sensitivity of 59.09%, and specificity of 74.55%. For predicting anastomotic leakage (Figure 4), POD3 CAR exhibited an AUC of 0.798 (95% CI=0.726–0.859), the sensitivity of 100%, and specificity of 58.11%.

Figure 2

(A, B) Receiver operating characteristic curve analysis of C-reactive protein/albumin ratio (CAR) as a predictor of surgical site infection (SSI).

Abbreviations: AUC, area under the curve; CI, confidence interval.

Figure 3

(A, B) Receiver operating characteristic curve analysis of C-reactive protein/albumin ratio (CAR) as a predictor of infective complications.

Abbreviations: AUC, area under the curve; CI, confidence interval.

Figure 4

(A, B) Receiver operating characteristic curve analysis of C-reactive protein/albumin ratio (CAR) as a predictor of anastomotic leakage.

Abbreviations: AUC, area under the curve; CI, confidence interval.

Discussion

In this single-center retrospective analysis of 154 patients who underwent radical surgery for colorectal carcinoma, we found evidence suggesting that high levels of preoperative mGPS (P=0.002), preoperative CAR (P=0.019), POD3 CAR (P<0.001) and POD3 poGPS (P<0.001) can significantly affect postoperative complications after surgery for colorectal cancer. On receiver operating characteristic curve analysis, we found good predictive value for all three inflammation-based scores evaluated. However, CAR on POD3 provided the highest PPV, which is of particular importance in the clinical setting.

Surgery induces local tissue damage, physical barrier damage, and potential exposure to environmental and commensal microorganisms, all of which can lead to localized inflammation. Moreover, patients with surgery often undergo other invasive procedures such as venous catheterization, general anesthesia, endotracheal intubation, and catheterization. These procedures damaged the skin and epithelial defenses, causing inflammation at distant anatomical sites. Therefore, monitoring inflammation perioperatively is essential, as well as to understand the relationship between inflammation and the risk of postoperative complications. Indeed, an increasing number of studies are being designed to examine the role of the systemic inflammatory response as a contributing factor to postoperative complications.

Both preoperative and postoperative CRP levels have been reported as important predictors of postoperative survival in patients with colorectal cancer.9,15,16 CRP is an important marker of inflammation, and its elevation in cancer patients is mainly due to the inflammatory response to the tumor and surgery. Nason et al17 reported that CRP on POD3 is a useful predictor of infectious complications in patients who undergo colorectal surgery. However, in the clinical setting, there are several limitations to the use of postoperative CRP as a predictor of postoperative complications. The main disadvantages are low predictive accuracy and high time lag.18,19 CRP and white blood cell levels increase non-specifically in response to surgical stress.20 Moreover, as suggested by a recent study, postoperative changes in CRP levels are noted later than the changes in other inflammatory markers such as interleukin-6.19 Therefore, CRP alone is not a sufficiently sensitive descriptor of the inflammatory state of patients in the early stages after surgery.

In order to improve the accuracy of the prediction of postoperative complications, CAR considers CRP levels concomitantly with albumin levels. Hepatocytes synthesize albumin. And albumin serves as a multifunctional protein with antioxidant, immunomodulatory, and detoxifying action.21 Among patients with malignant tumors, albumin levels differ significantly between survivors and non-survivors.22 Some scholars have pointed out that decreased albumin levels reflect negative nitrogen balance and a decreased rate of toxic metabolite clearance. Abundant expression of inflammatory factors in the plasma causes damage to the endothelial cells of capillaries, which allows albumin to leak through the damaged capillary endothelium into the interstitial space, resulting in hypoproteinemia,21,23 which is an important prognostic index of surgical outcomes.24 In our study, the combined indices of CRP and albumin (CAR, mGPS, poGPS) provided superior prognostic efficacy compared to that noted for CRP. CAR on POD3 exhibited the highest probability to detect patients at risk (PPV=83.2%), and the overall predictive value of preoperative CAR was poor. Preoperative mGPS (PPV=100.0%) and poGPS on POD3 (PPV=100.0%) exhibited a higher probability to adequately identify patients with high risk when the cutoff value was 2.

Alazawi et al reported that most patients who underwent surgery had a systemic inflammatory response on POD3,25 which is consistent with our present findings. Specifically, we found high predictive value for all inflammation-based prognosis scores obtained on POD3, and the incidence of postoperative complications was significantly higher in patients with CAR≥2.6 than in those with CAR<2.6 on POD3 (overall incidence, as well as the incidence of severe complications and that of infectious complications).

In our study, subgroup analysis confirmed the good prognostic value of CAR on POD3 for predicting the 30-day incidence of surgical site infection (AUC=0.760; 95% CI=0.685–0.835; P=0.016), infectious complications (AUC=0.702; 95% CI=0.623–0.773), and anastomotic leakage (AUC=0.798; 95% CI=0.726–0.859). Surgical site infection is the most common complication after surgery, with a high incidence and strong impact on the duration of hospitalization.26 In a prospective study by Goulart et al.27 CAR was superior to white blood cell count and procalcitonin as an independent predictor of surgical site infection. On the other hand, CRP was reported as the best postoperative inflammatory response marker for predicting anastomotic leakage, providing high sensitivity.28 Our present findings confirm the usefulness of CRP as an inflammatory marker but highlight the benefit of considering albumin concomitantly with CRP (especially as CAR) to improve the accuracy of predicting the risk of complications after radical surgery for colorectal carcinoma.

In recent years, the concept of enhanced recovery after surgery has become increasingly adopted in the field of colorectal surgery. Within this framework, minimally invasive surgery is preferred, as it helps reduce surgical trauma and ensuing inflammatory response to surgical stress as much as possible. However, various complications may still occur. Therefore, further studies are warranted to elucidate the extent to which such principles can minimize inflammatory response to surgical stress.

Several limitations of our study should be acknowledged. First, this was a retrospective study with a relatively small sample. Second, because of the small sample size, our analysis did not differentiate between laparoscopic and open surgery. Future studies with prospective design and large sample size are warranted. In particular, it is desirable to develop a comprehensive score based on mGPS, CAR, and poGPS. And the score should be easy to evaluate in clinical practice. It would help clinicians detect early postoperative complications and intervene promptly, thereby reducing the duration and costs of hospitalization, and promoting rapid recovery.