Albumin as a predictor in a nomogram for severe complications following gastrointestinal surgery and how to manage it

Abstract

BACKGROUND

Perioperative serum albumin (ALB) levels affect the outcomes of patients undergoing gastrointestinal surgery. ALB supplementation in patients with low protein levels may therefore be beneficial. However, the specific effects of and optimal levels for serum ALB supplementation remain to be elucidated.

AIM

To develop and validate a nomogram for predicting severe complications on the basis of the serum ALB concentration and to establish optimal postoperative serum ALB concentration ranges.

METHODS

A total of 409 patients who underwent gastrointestinal surgery were randomly divided into a training cohort and a validation cohort. Patients with Clavien-Dindo grade II or higher complications were classified as having severe complications. A nomogram was developed on the basis of the serum ALB concentration to predict the risk of severe complications. Receiver operating characteristic curve, calibration curve and decision curve analysis were performed to validate the nomogram. The χ² test was used to compare the incidence of severe complications across different ALB ranges.

RESULTS

Logistic regression analysis revealed that the American Society of Anesthesiologists grade, surgery duration, preoperative serum ALB concentration, and age were significant predictors of severe complications. The nomogram developed using these predictors yielded an area under the area under the curve of 0.728 in the training cohort and 0.700 in the validation cohort; the calibration curves demonstrated good consistency between the predicted and observed probabilities in the two cohorts. Decision curve analysis demonstrated that the nomogram provided net benefits in predicting severe complications within threshold probability ranges of 19%-81% and 9%-66% in the two cohorts. Compared to those with relatively low peak postoperative serum ALB concentrations, patients with relatively high concentrations had a significantly lower incidence of severe complications (60.5% vs 40.7%; P = 0.041).

CONCLUSION

A nomogram that incorporates four clinical characteristics could be valuable for predicting the risk of severe complications. An appropriate range of serum ALB concentrations may be associated with a lower incidence of severe complications.

Key Words: Gastrointestinal surgery; Serum albumin; Nomogram; Complications; Clavien-Dindo

Core Tip: Serum albumin (ALB) is valuable for reducing the occurrence of severe complications during the perioperative period and for postoperative recovery in patients who undergo gastrointestinal surgery. This study developed a nomogram based on four clinical characteristics, which may be valuable for predicting the risk of severe postoperative complications and could conveniently provide surgeons with early decision-making and timely interventions for high-risk patients. Additionally, compared to those with relatively low peak postoperative serum ALB concentrations, patients with relatively high postoperative serum ALB concentrations had a significantly lower incidence of severe complications. This study provides a valuable reference for perioperative ALB usage.

INTRODUCTION

Serum albumin (ALB), a plasma protein synthesized by the liver, maintains plasma colloid osmotic pressure, increases the circulating blood volume, and exerts immunomodulatory effects[1]. It also plays important roles in metabolism, anticoagulation, acid-base regulation, and antioxidation[2]. Hypoalbuminemia (defined as a serum ALB concentration less than 35 g/L)[3] is associated with postoperative complications in gastrointestinal surgery, including infections[4], anastomotic fistulas[5], respiratory failure, and deep vein thrombosis[6-10]. After surgery, the body needs ALB to repair tissue damage and mitigate inflammation[11]; as a critical carrier of endogenous and exogenous substances, ALB plays a vital role in promoting recovery. Therefore, clinicians usually empirically administer supplemental ALB to patients with hypoproteinemia to increase serum levels, and doing so can provide sufficient nutritional support and antioxidant substances to protect cells from the damage caused by oxidative stress[12].

Studies have shown that hypoproteinemia may be an independent risk factor for postoperative complications[13]. Preoperative hypoalbuminemia can be used to predict hospital mortality, reoperation risk and urinary tract infection risks[14], and, as a retrospective study revealed, the length of hospital stay after surgery for primary colorectal cancer[15]. The Expert Consensus Statement on the use of ALB in critically ill patients recommends that, for those who undergo abdominal surgery, the serum ALB concentration should be maintained at ≥ 30 g/L during the perioperative period[16]. However, as ALB exerts hyperosmotic effects, excessive infusion may cause dehydration, increased circulatory overload, congestive heart failure, and pulmonary edema. Therefore, patient serum ALB concentrations should be maintained within an appropriate range.

Although studies have shown that hypoproteinemia can be used to predict the risks of various complications of gastrointestinal surgery, research based on the Clavien-Dindo grading system has been limited, and effective risk scoring tools based on serum ALB levels are lacking. In addition, the optimal postoperative ALB level remains controversial. Nomograms have been widely used as predictive tools in recent years[17].

The purpose of this study was to utilize perioperative factors to develop a nomogram model for obtaining risk scores to predict the probability of severe complications and simultaneously to establish optimal postoperative serum ALB levels to provide decision-support tools for optimizing clinical management.

MATERIALS AND METHODS

Study design and ethical approval

In this retrospective study, 444 inpatients who underwent gastrointestinal surgery at our hospital from January 2019 to May 2023 were identified from the hospital information management system. Of these patients, 35 patients who also underwent resection of other organs were excluded, and 409 patients were ultimately included. The study was approved by the ethics committees of our hospital (No. IIT2025-055-002).

Inclusion criteria: (1) Age older than 18 years old and having undergone gastrointestinal surgery; (2) Available preoperative/postoperative serum ALB test results; (3) No serious impairment of hepatic or renal function; and (4) Complete medical records.

Exclusion criteria: (1) Severe preoperative malnutrition (cachexia or body mass index < 18); (2) Previous abdominal surgery; and (3) Congenital disorders that affect the serum ALB concentration. In accordance with the Clavien-Dindo grading system, a total of 189 patients were classified as having grade II or higher complications, which we defined as severe complications, while the remaining 220 patients were defined as having general complications.

Data collection

Medical records, including demographics (age and sex), American Society of Anesthesiologists (ASA) physical status, surgical operation, surgical duration, and preoperative total bilirubin, preoperative alanine aminotransferase, preoperative aspartate aminotransferase, preoperative gamma-glutamyl transferase, preoperative ALB, lowest postoperative ALB and highest postoperative ALB levels within 5 days after surgery, were reviewed by clinicians. An ASA grade of II or less was defined as a low ASA; otherwise, it was defined as a high ASA. If conversion to open surgery occurred, the procedure was recorded as laparotomy. Preoperative total bilirubin, preoperative alanine aminotransferase, preoperative aspartate aminotransferase, preoperative gamma-glutamyl transferase and preoperative ALB levels were defined according to the level measured closest to the start of surgery. ALB levels were measured by bromocresol green assay. The Clavien-Dindo grading system was used to classify postoperative complications into grades I-V. Grade I includes minor risk events not requiring therapy (with exceptions of analgesic, antipyretic, antiemetic, and antidiarrheal drugs or drugs required for lower urinary tract infection)[18]. Grade II and III complications require either pharmacological or procedural interventions, respectively. Grade IV complications are life-threatening complications, whereas patient death is classified as grade V[19].

Statistical analyses

The first author (Feng F) received systematic training in biostatistics, and the statistical methods for this study were reviewed by Peng JR from Beijing Shijitan Hospital.

Following assessments of distribution normality with the normality test, the means ± standard deviations were used for continuous variables with a normal distribution, and the medians (interquartile ranges) were used for continuous variables that were not normally distributed; comparisons between groups were analyzed with the nonparametric Mann-Whitney U test, and severe complication rates between groups were compared using the χ² test.

The dataset was divided into a training cohort (70%) and a validation cohort (30%) for model construction and internal validation using generalized linear models; variables with P < 0.05 after univariable logistic regression analysis were examined for multicollinearity by applying the variance inflation factor, and they were included in the multivariable logistic regression analysis if no correlations were identified in the multicollinearity test. Variables with P < 0.05 or with known clinical significance were defined as independent predictive factors and subsequently used to construct the nomogram.

The odds ratios (ORs) and 95% confidence interval (CI) were calculated for each variable associated with the occurrence of severe complications. The area under the curve (AUC), as well as calibration curve analysis and decision curve analysis (DCA), were used to evaluate the nomogram.

The association between the serum ALB concentration and hospitalization duration was analyzed using Spearman’s correlation analysis. To eliminate the interference of confounding factors, stratified logistic regression analysis was used to explore the importance of different subgroups of patients with different serum ALB levels to the incidence of severe complications. We converted the categorical variables into dummy variables for analysis.

All statistical analyses were performed with the R statistical software package, version 4.4.1; SPSS software, version 25 (IBM, Armonk, NY, United States); and Graph Pad Prism software, version 8.0 (GraphPad Software, San Diego, CA, United States). All tests were two-tailed, and P < 0.05 was considered to indicate statistical significance.

RESULTS

Characteristics of patients in both cohorts

In this study, we included only patients with gastrointestinal malignancies who required partial resection; no resections involving the hepatopancreatic region or other anatomical areas were performed. In the entire patient group, the training cohort and the validation cohort, totals of 46.2% (189/409), 46.2% (132/286) and 46.3% (57/123), respectively, had severe complications. Among the patients with severe complications in the overall cohort, 142 patients were classified as grade II, 35 as grade III, and 12 as grade IV, including surgical site infections requiring antibiotic therapy, pulmonary infections, blood transfusions, and total parenteral nutrition. A total of 409 patients were randomly divided into a training cohort and a validation cohort at a ratio of 7:3. The demographic and clinical characteristics of all patients are summarized in Table 1. The training and validation cohorts were generally comparable in terms of demographic and clinical characteristics (P > 0.05). A flow diagram for the construction and validation of the nomogram is shown in Figure 1.

Figure 1 Flow diagram of the predictive model. DCA: Decision curve analysis; ROC: Receiver operating characteristic.

Table 1 Comparison of characteristics between training cohort and validation cohort, n (%)/median (interquartile range)/mean± SD.

Characteristics	Training cohort (n = 286)	Validation cohort (n = 123)	P value
Age (years)	65.00 (57.00-72.00)	65.00 (59.00-71.00)	0.784
Sex			0.672
Male	183 (63.99)	76 (61.79)
Female	103 (36.01)	47 (38.21)
American Society of Anesthesiologists grade			0.430
Low	203 (70.98)	92 (74.80)
High	83 (29.02)	31 (25.20)
Operation			0.823
Laparoscopy	164 (57.34)	72 (58.54)
Laparotomy	122 (42.66)	51 (41.46)
Surgical duration (minutes)	264.00 (210.00-315.00)	252.00 (193.00-317.00)	0.925
Preoperative total bilirubin (μmol/L)	11.50 (8.70-14.60)	11.90 (8.80-15.50)	0.497
Preoperative alanine aminotransferase (U/L)	14.00 (10.00-20.00)	13.00 (9.00-21.00)	0.869
Preoperative aspartate aminotransferase (U/L)	16.00 (14.00-20.00)	17.00 (14.00-21.00)	0.480
Preoperative gamma-glutamyl transferase (U/L)	19.00 (14.00-30.00)	18.00 (13.00-25.00)	0.224
Preoperative albumin (g/L)	37.16 ± 4.22	37.88 ± 4.35	0.117

Screening of the characteristics of the nomogram

Univariable logistic regression analysis revealed that four characteristics (age, ASA grade, surgical duration, and preoperative serum ALB concentration) were significantly associated with severe complications. Multivariate logistic regression analysis revealed that ASA grade, surgical duration, and preoperative serum ALB concentration were independent predictive factors for severe complications (P < 0.05). Although the P value for age was 0.098 (P > 0.05), on the basis of clinical experience and previous studies[20], age was also considered a significant risk factor and was therefore incorporated into the nomogram. The variance inflation factor values of the four characteristics were all < 1.25, suggesting the absence of significant multicollinearity (Figure 2A).

Figure 2 Construction of nomogram through variables selection. A: The forest plot of univariable logistic regression for the training cohort, and the variance inflation factor values of characteristics incorporated into the nomogram; B: Nomogram for predicting severe complications after gastrointestinal surgery. ALB: Albumin; ALT: Alanine aminotransferase; ASA: American society of Anesthesiologists; AST: Aspartate aminotransferase; GGT: Gamma-glutamyl transferase; OR: Odds ratio; TBil: Total bilirubin; VIF: Variance inflation factor.

Nomogram construction and validation

We constructed a nomogram for severe complications according to the four characteristics identified above. An example of the use of the nomogram for predicting the risk probability for severe complications in a given patient is shown in Figure 2B. The total score was determined on the basis of the individual scores calculated using the nomogram; most patients in the present study had total risk points ranging from 40 to 160. The AUC of the nomogram was 0.728 (95%CI: 0.670-0.786) in the training cohort and 0.700 (95%CI: 0.608-0.792) in the validation cohort (Figure 3A and B), indicating good discrimination and suggesting that the nomogram could serve as a convenient tool for clinicians to predict severe complications. The calibration curves of the nomogram revealed the predicted and observed probabilities in both the training and validation cohorts (Figure 3C and D). The Hosmer-Lemeshow goodness of fit test yielded nonsignificant P values of 0.903 and 0.482 in the training and validation cohorts, respectively, suggesting that the nomogram has good calibration fitness. To evaluate the clinical practicality of the nomogram, a decision curve was constructed, which revealed that the threshold probabilities of the nomogram in the training cohort and validation cohort were 19%-81% and 9%-66%, respectively, indicating that the application of this nomogram to predict severe complications would add net benefits beyond either a treat-all scheme or a treat-none scheme (Figure 3E and F).

Figure 3 Discrimination, calibration, and decision curve of nomogram for predicting severe complications after gastrointestinal surgery. A: Training cohort receiver operating characteristic curve; B: Validation cohort receiver operating characteristic (ROC) curve; C: Training cohort calibration curve; D: Validation cohort calibration curve; E: Training cohort decision curve; F: Validation cohort decision curve. AUC: Area under the curve.

Correlation analysis of ALB level and length of hospitalization

We collected preoperative ALB levels, the lowest ALB levels within 5 days postsurgery, and the length of hospitalization from all patients. Spearman’s correlation analysis revealed that the P values were all < 0.001 (Figure 4A). Negative correlations were observed between the length of hospitalization and both preoperative ALB levels (r = -0.18) and the lowest serum ALB levels within 5 days postsurgery (r = -0.24).

Figure 4 Association between albumin levels and severe complications. A: Correlation analysis albumin (ALB) level and length of hospitalization; B: Comparison of ALB levels between general complication and severe complication group; C: Distribution of incidence of severe postoperative complications across different ALB level ranges.

Risk of severe complications associated with low serum ALB levels

We first compared the preoperative serum ALB concentration and lowest serum ALB concentration within 5 days after surgery between the general complication group and severe complication group and confirmed the presence of significant between-group differences (P < 0.001; Figure 4B). Preoperative serum ALB concentrations were stratified into three subgroups with values ranging from 30 g/L to 35 g/L (n = 109), from 35 g/L to 40 g/L (n = 177), and from 40 g/L to 45 g/L (n = 91) (Table 2). The lowest serum ALB concentrations within 5 days postsurgery were also stratified into three subgroups with values ranging from 25 g/L to 30 g/L (n = 96), from 30 g/L to 35 g/L (n = 216), and from 35 g/L to 40 g/L (n = 88) (Table 3).

Table 2 Stratified multivariable logistic regression analysis of preoperative albumin on severe complications.

Characteristics	Univariable logistic regression		Multivariable logistic regression
	OR (95%CI)	P value	OR (95%CI)	P value
Age (years)	1.003 (1.014-1.053)	0.001	1.019 (0.997-1.041)	0.091
Sex
Male	Reference		-
Female	1.074 (0.718-1.607)	0.729	-
American Society of Anesthesiologists grade
Low	Reference		Reference
High	2.608 (1.669-4.075)	< 0.001	1.715 (1.010-2.912)	0.046
Operation
Laparoscopy	Reference		Reference
Laparotomy	1.695 (1.141-2.519)	0.009	1.356 (0.863-2.130)	0.187
Surgical duration (minutes)	1.005 (1.003-1.007)	< 0.001	1.005 (1.003-1.008)	< 0.001
Preoperative total bilirubin (μmol/L)	1.012 (0.987-1.038)	0.341	-
Preoperative alanine aminotransferase (U/L)	1.001 (0.990-1.012)	0.893	-
Preoperative aspartate aminotransferase (U/L)	1.012 (0.995-1.029)	0.182	-
Preoperative gamma-glutamyl transferase (U/L)	1.002 (0.998-1.005)	0.355	-
Preoperative albumin (g/L)
(30, 35)	3.302 (1.701-5.405)	< 0.001	2.223 (1.192-4.415)	0.012
(35, 40)	1.234 (0.732-2.079)	0.430	1.021 (0.589-1.770)	0.942
(40, 45)	Reference		Reference

OR: Odds ratio.

Table 3 Stratified multivariable logistic regression analysis of lowest postoperative albumin levels within 5 days on severe complications.

Characteristics	Univariable logistic regression		Multivariable logistic regression
	OR (95%CI)	P value	OR (95%CI)	P value
Age (years)	1.003 (1.014-1.053)	0.001	1.022 (1.000-1.043)	0.047
Sex
Male	Reference		-
Female	1.074 (0.718-1.607)	0.729	-
American Society of Anesthesiologists grade
Low	Reference		Reference
High	2.608 (1.669-4.075)	< 0.001	2.085 (1.253-3.467)	0.005
Operation
Laparoscopy	Reference		Reference
Laparotomy	1.695 (1.141-2.519)	0.009	1.203 (0.775-1.886)	0.410
Surgical duration (minutes)	1.005 (1.003-1.007)	< 0.001	1.004 (1.002-1.007)	< 0.001
Preoperative total bilirubin (μmol/L)	1.012 (0.987-1.038)	0.341	-
Preoperative alanine aminotransferase (U/L)	1.001 (0.990-1.012)	0.893	-
Preoperative aspartate aminotransferase (U/L)	1.012 (0.995-1.029)	0.182	-
Preoperative gamma-glutamyl transferase (U/L)	1.002 (0.998-1.005)	0.355	-
Lowest postoperative albumin within 5 days (g/L)
(25, 30)	3.066 (1.666-5.644)	< 0.001	2.331 (1.214-4.476)	0.011
(30, 35)	2.256 (1.328-3.832)	0.003	1.725 (0.978-3.043)	0.060
(35, 40)	Reference		Reference

OR: Odds ratio.

Multivariable logistic regression analysis revealed that patients with preoperative ALB concentrations ranging from 30 g/L to 35 g/L had a 1.223-fold increased risk of developing severe complications compared to those with preoperative serum ALB concentrations ranging from 40 g/L to 45 g/L (95%CI: 1.192-4.415; P = 0.012), suggesting that preoperative hypoproteinemia can lead to severe complications. With respect to postoperative serum ALB concentrations, with the reference range set from 35 g/L to 40 g/L, patients whose lowest concentrations ranged from 25 g/L to 30 g/L had a 1.331-fold increased risk of severe complications (95%CI: 1.214-4.476; P = 0.011). These results indicated that low preoperative or postoperative serum ALB concentrations are associated with a greater risk of severe complications.

Establishing the optimal range of postoperative ALB levels

We confirmed that low ALB concentrations significantly increased the risk of severe complications. Therefore, what range of postoperative ALB levels should be maintained to reduce the risk of severe complications? Given the variability in postoperative ALB supplementation among patients, we performed a stratified analysis on the basis of peak serum ALB levels within 5 days after surgery to evaluate the effects of this supplementation. Considering ALB levels greater than 30.0 g/L, we found that the peak incidence of severe complications occurred at 33.6 g/L; therefore, we adopted a 3.6 g/L interval for stratification and conducted the χ² test to assess differences in the incidence of severe complications among the groups. The incidence of severe complications was significantly greater in the 30.0-33.6 g/L group (60.5%) than in the 33.6-37.2 g/L group (40.7%), and a significant intergroup difference in rates was observed according to Pearson’s χ² test (χ² = 4.162; P = 0.041; Figure 4C). However, similar differences were not found between the other groups. These findings suggested that 33.6-37.2 g/L may be the optimal range for postoperative ALB levels, and this range may be associated with a reduction in the incidence of severe complications, providing actionable guidance for rational ALB supplementation.

DISCUSSION

ALB promotes anastomotic healing following gastrointestinal surgery and reduces the incidence of complications, such as infections, hemorrhage, and anastomotic leakage. Therefore, maintaining a certain level of ALB is highly important for reducing severe complications. However, few studies have investigated the development of ALB-based risk scoring tools for complications[21-23], and the optimal range for ALB supplementation remains controversial. In this study, we developed a nomogram model based on the serum ALB concentration, which can be used to predict the risk of severe complications. Moreover, we clarified the relationship between the appropriate range of serum ALB levels and the reduction in the incidence of severe complications, which could provide reference data for the rational use of serum ALB levels.

Previous studies have demonstrated that the preoperative serum ALB concentration is a significant predictor of postoperative complications, including intra-abdominal infections and prolonged ileus[24]. However, studies of the risk of complications based on the Clavien-Dindo grading system after gastrointestinal surgery have been limited[25,26]. A large-scale retrospective study revealed that age, preoperative serum ALB concentration, and the use of open surgery were independent risk factors predicting severe complications of grades III and IV after radical surgery for rectal cancer, despite an AUC of 0.7191 (95%CI: 0.6182-0.8199) in the validation set, and DCA demonstrated suboptimal clinical utility. In contrast, while demonstrating a similar AUC for predicting severe complications (grade II or higher), our nomogram exhibited superior clinical utility with DCA[20].

The nomogram includes four clinical characteristics: (1) Age; (2) ASA grade; (3) Surgical duration; and (4) Preoperative ALB concentration. The physiological stress of anesthesia and long surgical durations place significant demands on older adults, often resulting in longer recovery periods and a greater risk of perioperative complications. Surgery induces a significant stress response in patients, activating both the hypothalamic-pituitary-adrenocortical axis and the sympathetic-adrenomedullary axis to release substantial amounts of glucocorticoids, promoting protein catabolism while suppressing synthesis and resulting in a catabolic-dominant metabolic state that leads to a negative nitrogen balance[27]. Hypoalbuminemia is often considered a common risk factor in surgical and hospitalized patients, and it can cause edema of the anastomosis and ultimately lead to diffuse peritonitis[28], further exacerbating the patient’s physical and psychological burdens.

Studies have revealed that low preoperative serum ALB concentrations significantly increase the incidence of postoperative complications[29,30]. We conducted a stratified logistic regression analysis and reported that both preoperative and postoperative low serum ALB concentrations were risk factors for severe complications, highlighting the critical importance of maintaining perioperative serum ALB levels, consistent with previous findings[31-34].

Although the serum ALB concentration plays an important role in reducing postoperative complications, its use should not be unlimited because excessive infusion may cause dehydration, pulmonary edema and a significant economic burden. Therefore, maintaining levels within an optimal range is essential[35].

We analyzed the highest level of ALB within 5 days after surgery to determine the optimal range. Five days is a common time point for the occurrence of complications, and selecting the highest value can reduce the intervention effects of different doses of supplemental ALB. The optimal range reduces the incidence of severe complications while avoiding excessive supplementation of ALB without increasing health or economic costs. This range is also one of the highlights of this study.

Previous studies of ALB-based predictive models for postoperative complications have remained limited, with most studies focusing solely on preoperative ALB as a risk factor for severe complications[36-38]. The innovation of our study is the development of a risk scoring model based on the integration of preoperative characteristics, and a nomogram can more conveniently facilitate early decision-making and timely interventions for high-risk patients. Additionally, this study provides new ideas for how to supplement ALB reasonably.

The present study had several limitations. First, this research was limited by the availability of clinical data, and variables such as intraoperative blood loss, renal function parameters, and transfusion volumes were not obtained. Second, due to the limitations of single-center studies, the nomogram was only internally validated. Therefore, further multicenter studies and external validation are necessary to improve the reliability of the nomogram. Third, differences in both the surgical approach and the operating surgeon could have introduced baseline imbalances. Finally, prospective, controlled experiments must be designed to validate the rationality of the optimal range of ALB levels.

CONCLUSION

This study developed and internally validated a convenient and useful nomogram incorporating four predictors to identify patients at risk of severe complications after gastrointestinal surgery. We also established the optimal range of serum ALB levels to provide decision-making support for optimized clinical management to reduce complications and length of hospitalization.