This prospective single-center study was conducted according to the STROBE statement and recruited consecutive patients undergoing colorectal surgery from November 2019 to May 2023 in an academic tertiary hospital in Greece. Patients were enrolled in the study after providing informed consent. The study protocol was approved by the hospital's Ethical and Scientific Review Board (Approval No. 42687/0519) and was registered in an open-access database available on the Internet (www.clinicaltrials.gov: NCT04748744).

In this study, data were evaluated from 403 consecutive patients who underwent colorectal surgery at the Surgical Clinic of the General University Hospital of Patras (GUHP). Patients were included in the study, provided they had completed the necessary informed consent documents after a counseling session with members of the research team. In these sessions, the purposes of the study, the research perspective, the interventions to which they were to be subjected, as well as the fact that their participation is voluntary and they retain the right to withdraw from it at any time, even after its end, until the publication of its results, were analyzed to the patients. Patients were included in this study if they met the following inclusion criteria: (1) age 18 years and older, (2) ability to undergo surgical intervention on an urgent or an elective basis, and (3) requiring colorectal surgery for any surgical pathology. We included both elective and urgent/emergency operations, as they will be further analyzed as separate subgroups in the analysis. Due to the effects of systematic inflammatory states and neurological degenerative disorders in BChE levels, we decided to exclude patients who exhibited signs of metastatic disease, either localized or with extended metastatic disease burden.

In total, 489 patients were screened for inclusion in the present study. Of them, 67 were deemed unfit for surgery, or their management plan was switched to conservative management before undergoing surgery, and 19 patients were not able to provide consent at the time of operation or in the early postoperative period.

During hospitalization, patient serum samples were obtained at four time points: (1) on the day of surgery, preoperatively, and a second sample immediately postoperatively, (2) first postoperative day, (3) on the third postoperative day, and (4) on the fifth postoperative day.

The reasoning behind this measurement protocol was to incorporate preoperative (baseline) measurements of BChE within our modeling process. For the measurement of BChE, materials and resources of the GUHP were used, which had been approved at the time of study submission. The quantitative in vitro determination of BchE in the serum was done using a colorimetric method. For the determination of BChE levels, a spectrophotometric method with a Randox RX Imola Autoanalyzer was used. The values were expressed in IU/L (international units per liter). The values of BChE levels, as well as all the patient data concerning their hospitalization, were recorded in a special postoperative monitoring software of the clinic, transferred to separate databases, and anonymized before the analysis phase. As part of the project's protocol, the biological patient samples were kept until the end of the analysis phase (September 2023), after which they were appropriately disposed of and destroyed without storing any tissue samples.

Our primary endpoint was the development of surgical site infections (SSIs), as a subcategory of septic complications. SSIs were defined using the CDC criteria for diagnosis and classified in accordance with the CDC/NSQIP classifications of SSIs. SSIs were classified as superficial incisional SSIs, deep incisional SSIs, and organ/space SSIs (20, 21).

As a secondary endpoint of our study, we set the development of any postoperative septic complication, with a subset analysis on patients whose septic profile was secondary to an anastomotic leak.

For the definition of septic syndromes, the latest definition according to Sepsis-3 was used: sepsis should be defined as life-threatening organ dysfunction caused by a dysregulated host response to infection. For clinical operability, organ dysfunction can be represented by an increase in the Sequential Organ Failure Assessment (SOFA) score of 2 or more points, which is associated with in-hospital mortality greater than 10%. For this composite measure, all septic complications were considered, including but not limited to anastomotic leaks, surgical site infections (of all grades), hospital-associated pneumonia, UTIs, and more.

As a special patient subpopulation of interest, patients with septic syndrome secondary to anastomotic leaks were studied separately. The diagnosis of the leak was made using clinical indicators and was radiologically confirmed in all cases as per local workup protocol. These three outcomes were studied in relation to age, gender, preoperative diagnosis of the patient, the degree of urgency of the surgery, the presence of malignancy, the duration of the operation, and the number of pRBC units transfused.

As a final secondary outcome of the study, we set the development of any postoperative complication, as defined by the ESA-ESICM joint task force on perioperative outcome measures. The existence or not of a postoperative complication, and its categorization, was made after the evaluation of the patient by at least two doctors at the time of diagnosis, and any disagreements as to the definition and identification of it were resolved by a senior, third specialized colorectal surgeon.

All patients received a bundle of SSI prophylaxis based on the 2019 NICE guidelines for the prevention of SSIs. Patients in our institution received preoperative antibiotics within 30–60 min from the first incision. Appropriate warming, hair removal, and glycemic control were ensured throughout the procedure. Skin preparation included alcohol-based chlorhexidine solution for all patients. Elective colorectal patients received intravenous cefuroxime and metronidazole. Emergency cases routinely received beta-lactamase plus metronidazole; however, ciprofloxacin was also utilized in some cases. In our institution, mechanical bowel preparation is used routinely in elective cases. Oral antibiotics in elective colorectal cases are not part of the institution's protocol and therefore were not administered (20).

For the statistical analysis of the results of the work, the statistical data processing packages SPSS (IBM SPSS Statistics for Windows, Version 28.0. Armonk, NY, USA: IBM Corp), jamovi [The jamovi project (2021). jamovi (Version 1.6)], and the R programming language were used. The variables of interest were expressed as binomial (binomial variables) and concerned the development or not of SSI, the development or not of postoperative sepsis, and the presence of anastomotic leakage, as a cause of postoperative sepsis. The presence or absence of malignancy, gender, operative approach, urgency on which the patient was operated, smoking, and the ASA score greater than 2 were also expressed as binomial. Preoperative diagnosis, type of surgery, TNM staging, and the number of RBC units administered were expressed as categorical variables. The days of hospitalization, age, BChE levels in the blood, BMI, and surgical time in minutes were the continuous variables. At the same time, some continuous variables were deemed necessary to be transformed either logarithmically or to the root of the variable to meet the requirements of the regression models.

The initial approach to the data was made using univariate analysis techniques for pairs of variables to establish the existence of any associations between them. For the comparison of continuous variables, the means for each category and the standard deviations were checked for statistical significance using either t-tests (Student's) or Mann–Whitney U-test when the use of parametric tests was not feasible. The normality of data was evaluated by using visualization of the variable distribution (histograms) and with the utilization of the Shapiro–Wilk test. For the comparison of categorical variable values, Fischer's exact test and the Chi-squared test of statistical significance were used as appropriate. For all tests of statistical significance, p < 0.05 was considered the threshold of significance, while all tests were two-tailed.

As the final step in the statistical analysis of the study data, the construction of a predictive model for the outcomes of interest was defined. For the creation of the predictive models, we relied on the principles of logistic regression modeling. The stepwise selection regression technique was used for the selection of parameters and manual parameter selection to achieve optimal model fit.

To assess the weighed and independent utility of BChE in assessing SSI patients, multivariate logistic regression analysis using backward variable selection techniques was used. The optimal logistic regression model resulted from both manual extraction and insertion of certain variables, determined by the reported AIC. After constructing the original model, we utilized the omnibus likelihood ratio test to assess for each model whether the variance explained by the model in the observed data is statistically more significant than the unexplained variance. We selected the variables that proved to be independently associated with SSI, to construct the final predictive model. To assess the final model fit, McFadden's pseudo-R² test was employed.

For the evaluation of the goodness of fit of the predictive models, on our data, the Akaike information criterion and the R² indices according to McFadden's pseudo-R² test were employed.

Finally, after construction of the optimal model, we evaluated its ability to predict the occurrence of SSIs by using bootstrapping sample drawing and plotting the corresponding ROC curve.

Among the remaining patients, 15.2% (61 patients) experienced surgical site infection (SSI) as the predominant postoperative complication. In addition, 6.5% (26 patients) developed any form of postoperative infection, while 4.2% (17 patients) developed septic complications, primarily due to postoperative leakage from a newly formed anastomosis. Most of the enrolled patients (69.7%) had a complication-free early postoperative period. The development of any complication affected a total of 122 patients, which corresponds to 30.3% of the participants. SSI as a complication affected 61 patients (15.2%).

Operative approach, urgency of operation, ASA grading, preoperative BMI, and diabetes mellitus were all found to be significantly associated with the occurrence of SSI in the univariate analysis (Table 1). These are the parameters that multivariable analysis and model building will be based on; however, crucial parameters such as sex and age will also be included based on previous larger studies that have documented a strong relationship between them and SSI.

Looking at Table 2, we can observe that patients with surgical site infection (SSI) tend to differ from uncomplicated patients in terms of the mean BChE levels in their blood serum. On the day of surgery and the first postoperative day, patients with SSI had higher BChE levels on average (5.41 KU/L vs. 5.16 KU/L, p = 0.164 and 4.69 KU/L, vs. 4.61 KU/L, p = 0.658, respectively). This trend reverses on the third and fifth postoperative days, where patients who developed SSI had significantly lower serum BChE levels on average (3.90 KU/L vs. 4.54 KU/L, p < .001 and 4.14 KU/L vs. 4.73 KU/L, p < .001, respectively). Preoperative (baseline) serum BChe levels did not differ significantly between patients.

Multivariate analysis revealed that when adjusted for other factors, lower BChE levels on the first postoperative day were associated with 2.6 times higher odds of developing SSI (OR: 2.6, 95%CI: 1.3–3.9, p-value < 0.05). Similar results were found for low BChE levels on the third postoperative day as they were associated with 2.53 times higher odds for developing SSI (OR: 2.5, 95%CI: 1.27–3.87, p-value < 0.05) when adjusted for other factors. Lastly, when adjusting for other factors, BChE levels on the fifth day were not independent risk factors for SSI development (OR: 0.38, 95%CI: 0.02–1.23, p-value > 0.05). All the above results were obtained, using baseline BChE measurements on the operative day, as the reference level (Table 3).

An ASA score of less than 2 was an independent negative predictor of SSI occurrence (Table 3, with an OR of 0.138). Malignancy status was also an independent predictor of SSI, with an OR of 0.190 (no malignancy vs. malignancy).

Gender, age higher than 65 years, and length of hospitalization were not independently, significantly associated with the development of postoperative SSI. Operative approach, urgency of operation, smoking, diabetes mellitus, BMI, and TNM levels did not prove to have statistically significant association in the initial, cumulative model, and therefore were not included in the final model. The final ROC curve and AUC can be seen in Supplementary Figure S1. The multivariable logistic regression model was able to predict the occurrence of postoperative surgical site infection with satisfying accuracy following the bootstrapping process. The pseudo-R² (McFadden's) confirmed a relatively good fit of our model to the data, and the final model was the one with the optimal AIC value (Table 3), further confirming that this is the optimal parameter selection (from the available) to predict SSI utilizing BChE levels. The overall accuracy of the final model is 0.952 with a sensitivity of 0.978 and a specificity of 0.852. The model's AUC (area under the curve) was calculated as 0.981 (Table 3), indicating that it has good predictive capability for the outcome of interest.

When examining the relationship between BChE and the development of any type of infectious/septic complication, the differences do not appear to be statistically significant in favor of any patient group (see Table 4). Therefore, based on the univariate analysis, BChE does not seem to correlate with septic complications when they are grouped together.

The same pattern is observed when one examines the mean BChE levels in the serum of patients with postoperative anastomotic leakage (see Table 5). We observe that the BChE levels in patients with leakage did not significantly differ from those of uncomplicated patients.

The association between BChE levels and the development of any form of sepsis on the first, third,^, and fifth day after surgery were also evaluated. BChE levels in patients with sepsis did not show statistically significant differences compared to uncomplicated patients during any time point postsurgery between patients with any form of sepsis and the cohort that did not (4.94 vs. 4.6, p = 0.903 4.81 vs. 4.54, p = 0.818 and 4.88 vs. 4.73, p = 0.433, respectively; tested with Mann–Whitney U-test and Welch's t-test).

On the day of surgery and the first postoperative day, patients with any complication exhibited higher mean BChE levels although this difference was not statistically significant (5.31 KU/L vs. 5.16 KU/L, p-value = 0.438; and 4.68 KU/L vs. 4.61 KU/L; p-value = 0.640, respectively). During the third and fifth day postsurgery, a statistically significant difference was found in the mean serum BChE levels between patients with any complication and those without. Patients had significantly lower mean BChE levels (4.22 KU/L vs. 4.54 KU/L p-value = 0.015, and 4.45 KU/L vs. 4.73 KU/L, p-value = 0.029, respectively).