Prognostic value of perioperative changes in serum primary metabolites in patients after major surgery under general anesthesia: an exploratory secondary analysis of the TAPIR trial

Abstract

Purpose

Major surgery under general anesthesia substantially alters physiologic homeostasis. Nevertheless, the intricate effects on the metabolome are poorly studied. Metabolic fingerprints may allow the identification of patients at risk for unfavourable outcomes.

Methods

We conducted a secondary, exploratory, targeted metabolomic analysis of 177 high-risk patients undergoing major abdominal surgery under general anesthesia enrolled in the Targeting preoperatively Assessed Personal cardiac Index in major abdominal suRgery patients (TAPIR) randomized controlled trial. We analyzed primary serum metabolites using liquid chromatography coupled with triple quadrupole mass spectrometry before surgery (on preoperative day 0) and on postoperative day 3 (POD3). Our primary aim was to investigate postoperative alterations in primary serum metabolites. Secondary objectives included analyses in different subgroups, including patients with postoperative complications (composite of complication, delirium, acute kidney injury, and infection) up to day 30. We applied regression analyses and calculated false discovery rate-adjusted P values to address multiplicity.

Results

Of the 37 metabolites analyzed, 20 were different on POD3 after comparison with before surgery (lower: 4-hydroxyproline, alanine, asparagine, citrulline, cystine, dimethylglycine, glutamine, glutamic acid, glycine, guanosine, histidine, niacinamide, serine, uric acid, and xanthine; higher: isoleucine, leucine, methionine, methionine sulfoxide, pyruvic acid; adjusted P values < 0.05). We found no statistically significant preoperative differences between patients with and without postoperative complications (all adjusted P values ≥ 0.05). Postoperatively, patients with (vs without) delirium within thirty days after surgery (n = 13/177) showed lower levels of alanine, asparagine, citrulline, cystine, glutamine, glutamic acid, serine, threonine, and tyrosine after adjusting for preoperative metabolite levels.

Conclusion

Major abdominal surgery under general anesthesia was associated with complex changes in primary metabolites. We identified alterations in certain metabolites that were associated with postoperative delirium. Future research may establish metabolic patterns allowing the identification of patients at risk for unfavourable postoperative outcomes.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12630-025-02984-6.

Medical advances and the changing demographic profile of society are driving the increasing complexity and frequency of surgery.¹ The rising number of postoperative complications places a significant burden on the health and wellbeing of patients and health care systems.² There is therefore an urgent need to optimize the allocation of resources to meet the increased demand.² Existing and validated scoring systems and risk calculators for expected outcomes are limited in their ability to predict individual risk.³ Nevertheless, a comprehensive preoperative assessment of anticipated postoperative outcomes, the potential need for postoperative monitoring, and individualized therapy are essential to ensure efficient use of resources. It is already known that major surgery under general anesthesia significantly alters physiologic homeostasis by affecting vital functions—such as gas exchange, body temperature, cardiovascular dynamics, plasma volume, and electrolyte concentrations.⁴ Furthermore, general anesthesia decreases metabolic rate,⁵ modulates oxidative stress,⁶ and alters mitochondrial function and individual metabolic pathways (glucose, protein, and lipid metabolism).^5,7,8

To meet the demand for improving surgical quality, predicting complications, and optimizing patient outcomes, metabolomics—a low-molecular-weight metabolites profiling technique—can help identify pathologically or surgically altered metabolic pathways. Metabolomics is an evolving omics technology that provides insight into physiologic function and enables the identification of individual metabolic fingerprints in surgical patients resulting from external and internal factors.⁹ Nevertheless, the complex effects of major surgery under general anesthesia on the metabolome and the prognostic value of perioperative changes in the metabolome for patient outcome remain largely unknown. As part of the metabolome, primary metabolites are essential for maintaining the basic life processes of the organism, provide a wide range of well-defined compounds, and appear to be particularly affected in the sparse literature to date on perioperative metabolic changes.^10–13

Therefore, the primary objective of this study was to investigate perioperative changes in primary serum metabolites in high-risk patients undergoing major abdominal surgery under general anesthesia. Secondary objectives included analysis of changes in primary serum metabolites in different subgroups, including patients with postoperative complications, such as delirium, acute kidney injury, and infection.

Methods

Study design and clinical data

We conducted a secondary exploratory targeted metabolomics analysis of serum samples of patients included in the Targeting preoperatively Assessed Personal cardiac Index in major abdominal suRgery patients (TAPIR) randomized controlled trial.¹⁴ The Ethics Committee of the Hamburg Chamber of Physicians (Hamburg, Germany) approved the trial, and it was registered at ClinicalTrials.gov (NCT02834377). All patients gave written informed consent. For detailed information, we refer to the original publication.¹⁴ Briefly, 188 high-risk adult patients undergoing elective major abdominal surgery under general anesthesia were enrolled (Electronic Supplementary Material [ESM] eAppendix 1) and randomized to receive either routine or personalized hemodynamic management on the basis of their individual baseline cardiac index (TAPIR treatment arms).¹⁴ In all patients, general anesthesia was maintained either with inhalational agents or via total intravenous anesthesia, with or without a neuraxial regional technique. The TAPIR trial collected baseline and perioperative clinical data, including age, sex, height, body weight, baseline risk factors (renal, pulmonary, cardiac, liver impairment, and immunosuppression), American Society of Anesthesiologists (ASA) Physical Status classification, Patient Health Questionnaire score, duration of surgery, blood loss, fluid balance, and use of vasopressors. Postoperative outcomes were assessed according to the European Perioperative Clinical Outcome definitions.^14,15

Primary endpoint

We primarily investigated differences between preoperative (on preoperative day 0 [D0]; day of surgery) and postoperative (on postoperative day 3 [POD3]) levels of primary serum metabolites in all patients.

Subgroup analyses

In the first subgroup analysis, we analyzed the preoperative and postoperative differences between patients receiving routine or personalized hemodynamic management (TAPIR treatment arms: routine vs personalized). We conducted a second set of subgroup analyses to investigate associations between metabolic alterations and the most common postoperative complications observed in the TAPIR trial. Therefore, we analyzed preoperative and postoperative differences between patients with and without occurrence of the primary TAPIR composite outcome of major postoperative complications (ESM eAppendix 2),¹⁴ acute kidney injury (AKI), delirium, and infection (collapsed composite of postoperative infections) within thirty days after surgery. We defined AKI based on the Kidney Disease Improving Global Outcomes (KDIGO) guidelines.¹⁶ We identified delirium on the basis of the European Perioperative Clinical Outcome (EPCO) definitions.¹⁵ The collapsed composite (any event vs none) of postoperative infections included surgical site infection, urinary tract infections, confirmed bloodstream infections, pneumonia, and infections of unknown etiology.¹⁵

Analysis of primary metabolites

We analyzed primary metabolites (listed in Table 1) using liquid chromatography coupled with triple quadrupole mass spectrometry (LC-MS/MS) from 20-µL serum samples of the patients which were obtained immediately before surgery (D0) and on POD3. Electronic Supplementary Material eAppendix 3 contains further information on preanalytical preparation, separation technique, mass spectrometric analysis (LCMS-8050 equipment set-up), and mass transitions.

Table 1.

List of primary metabolites analyzed

Compound	Category	HMDB ID
4-hydroxyproline	Amino acids: nonproteinogenic	HMDB0000725
Adenosine	Nucleosides and nucleotides	HMDB0000050
Alanine	Amino acids	HMDB0000161
Aspargine	Amino acids	HMDB0000168
Asymmetric/symmetric dimethylarginine	Amino acids: nonproteinogenic	HMDB0003334/ HMDB0001539
Cholesterol	Others	HMDB0000067
Citrulline	Amino acids: nonproteinogenic	HMDB0000904
Creatinine	Others	HMDB0000562
Cystine	Amino acids	HMDB0000192
Desoxycholic acid/chenodesoxycholic acid	Organic acids	HMDB0000626/ HMDB0000518
Dimethylglycine	Amino acids: nonproteinogenic	HMDB0000092
Glutamic acid	Amino acids: nonproteinogenic	HMDB0000148
Glutamine	Amino acids	HMDB0000641
Glycine	Amino acids	HMDB0000123
Glycodeoxycholic acid	Organic acids	HMDB0000631
Guanosine	Nucleosides and nucleotides	HMDB0000133
Histidine	Amino acids	HMDB0000177
Hypoxanthine	Nucleosides and nucleotides	HMDB0000157
Inosine	Nucleosides and nucleotides	HMDB0000195
Isoleucine	Amino acids	HMDB0000172
Leucine	Amino acids	HMDB0000687
Methionine	Amino acids	HMDB0000696
Methionine sulfoxide	Amino acids: nonproteinogenic	HMDB0002005
Niacinamide	Others	HMDB0001406
Pyruvic acid	Glycolytic system	HMDB0000243
Serine	Amino acids	HMDB0000187
Succinic acid	Tricarboxylic acid cycle	HMDB0000254
Taurochenodeoxycholic acid	Organic acids	HMDB0000951
Taurocholic acid	Organic acids	HMDB0000036
Threonine	Amino acids	HMDB0000167
Thymidine monophosphate	Nucleosides and nucleotides	HMDB0001227
Tryptophan	Amino acids	HMDB0000929
Tyrosine	Amino acids	HMDB0000158
Uric acid	Organic acids	HMDB0000289
Uridine	Nucleosides and nucleotides	HMDB0000296
Valine	Amino acids	HMDB0000883
Xanthine	Nucleosides and nucleotides	HMDB0000292

HMDB ID = Human Metabolome Database Identifier (freely available electronic database containing detailed information about small molecule metabolites found in the human body)

Statistical analysis

We wrote a data analysis and statistical plan after accessing the data. In descriptive analyses for clinical variables, we report the median [interquartile range] for continuous variables. For categorical variables, we report absolute and relative frequencies (percentages). We imputed missing values for serum metabolites with the half of the minimum observed value. There were few missing data points in clinical variables, and we did not impute these.

To evaluate differences between the preoperative and postoperative levels of primary serum metabolites, we used linear mixed models with time (D0 vs POD3) as the fixed effect and patient as the random effect (intercept) to account for repeated measurements. Before modelling, we log₁₀-transformed (original values + 1 to avoid missing values for 0 values) and then z-standardized (mean, 0; standard deviation [SD], 1) all metabolite levels. The regression coefficients (β_z) of time including 95% confidence intervals (CIs) are an estimate of the difference in metabolite levels between D0 and POD3.

To analyze preoperative and postoperative group differences (e.g., patients with postoperative delirium vs without postoperative delirium) in metabolite levels, we used linear mixed models with time, group, and the interaction of both (time × group) as fixed effects and patient as the random effect (intercept) to account for repeated measurements. In detail, we analyzed group differences using post hoc contrasts of the estimated marginal means. In addition, to obtain an estimate of postoperative group differences in metabolite levels adjusted for preoperative levels, we applied additional linear regression models with POD3 metabolite levels (e.g., postoperative asparagine) as a response variable and D0 levels (e.g., preoperative values of asparagine) as well as group as covariates (analysis of covariance [ANCOVA] approach). We interpreted absolute values of the mean differences (β_z) > 0.1 as small effects, > 0.3 as medium effects, and > 0.5 as large effects.¹⁷

To check the consistency and robustness of the results, we recalculated all models after excluding patients with influential data points (model-specific difference in fits [DFFITS] > 0.2) and performed all analyses nonparametrically (preoperative vs postoperative differences: Wilcoxon signed-rank tests; group differences: Mann–Whitney U tests).

To address the issue of multiple comparisons (multiplicity), we calculated both the raw P values and the Benjamini–Hochberg adjusted P values.¹⁸ This methodology enables the control of the rate of incorrectly rejected null hypotheses (i.e., the falsediscovery rate [FDR]). This approach is less conservative and provides higher power than classical methods for controlling the family-wise error rate (e.g., the Bonferroni procedure). Although this is associated with a higher probability of type 1 errors, it is well suited for primarily exploratory analyses. In the Results section, we primarily focus on findings with FDR-adjusted P values < 0.05.

We preformed all analyses with R version 4.1.0 (The R Foundation for Statistical Computing, Vienna, Austria) and RStudio version 1.4.1717 (RStudio Inc., Boston, MA, USA).^19,20

Sample size calculation

This secondary analysis of the TAPIR trial was to be exploratory in nature; thus, the results should primarily be regarded as hypothesis-generating. Nevertheless, on the basis of a simplified sample size calculation, preoperative and postoperative differences (primary endpoint) with effect sizes of Hedges’ |g| ≥ 0.21 (small effect)—or converted to standardized regression coefficients |β_z| = 0.11—are detectable with P values < 0.05 when paired samples t tests are applied (n = 177; power of 0.8; two-sided α = 0.05).

Results

Sample characteristics

The TAPIR trial included 188 patients.¹⁴ Preoperative blood samples were not available in 2 patients for medical reasons. Additionally, postoperative blood samples were not available for 9 patients (two deceased, seven due to medical or organizational reasons). As a result, preoperative and postoperative serum samples were available for 177 patients to determine metabolite levels. Table 2 presents their baseline demographic and clinical characteristics.

Table 2.

Baseline demographic and clinical patient characteristics

Variable
Age (yr), median [IQR]	65 [55–74], N = 177
Body height (cm), median [IQR]	174 [167–180], N = 177
Body weight (kg), median [IQR]	75.5 [63.8–86.0], N = 176
BMI (kg·m−2), median [IQR]	24.7 [22.0–27.7], N = 176
LOS: ICU (days), median [IQR]	1 [1–4], N = 177
LOS: hospital (days), median [IQR]	16 [11–24], N = 177
Sex (male), n/total N (%)	108/177 (61%)
Obesity (BMI ≥ 30 kg·m−2), n/total N (%)	28/176 (16%)
CKD* (yes), n/total N (%)	45/177 (25%)
Risk factor for cardiac/respiratory complications† (yes), n/total N (%)	87/177 (49%)
Immunosuppressive therapy (yes), n/total N (%)	81/177 (46%)
ASA Physical Status, n/total N (%)
II	23/176 (13%)
III	131/176 (74%)
IV	22/176 (13%)
Major surgery, n/total N (%)
General	109/177(62%)
Aortic surgery	25/177 (14%)
Gynecologic	23/177 (13%)
Urologic	20/177 (11%)
Postoperative complications, n/total N (%)
TAPIR primary composite outcome up to POD 30 (cumulative, yes)	75/177 (42%)
Delirium‡ up to POD 30 (cumulative, yes)	13/177 (7%)
AKI‡ up to POD 30 (cumulative, yes)	15/177 (89%)
Infection‡,§ up to POD 30 (cumulative, yes)	66/177 (37%)

^*Serum creatinine at admission ≥ 1.3 mg·dL⁻¹

^†See ESM eAppendix 1

^‡European Perioperative Clinical Outcome (EPCO) definitions¹⁵

^§The collapsed composite (any event vs none) of postoperative infections included surgical site infection (≤ POD30: n = 36), urinary tract infections (≤ POD30: n = 8), confirmed bloodstream infections (≤ POD30: n = 11), pneumonia (≤ POD30: n = 18), and infections of unknown etiology (≤ POD30: n = 5)

AKI = acute kidney injury; ASA = American Society of Anesthesiologists; BMI = body mass index; CKD = chronic kidney disease; ICU = intensive care unit; IQR = interquartile range; LOS = length of stay; POD = postoperative day; TAPIR = Targeting preoperatively Assessed Personal cardiac Index in major abdominal suRgery patients randomized controlled trial¹⁴

Primary endpoint: postoperative changes in serum metabolome

Overall, 54% (n = 20/37) of the metabolites were significantly different after surgery compared with before surgery (FDR-adjusted P values < 0.05; Fig. 1). Fifteen metabolites exhibited significantly lower levels at the postoperative (vs preoperative) time point, with large effect sizes (β_z ≤  −0.5) observed for citrulline, 4-hydroxyproline, cystine, glutamine, niacinamide, and histidine, and moderate effect sizes (β_z: −0.3 to −0.49) for guanosine, dimethylglycine, alanine, glycine, glutamic acid, uric acid, serine, and asparagine. Five metabolites exhibited significantly higher levels at the postoperative (vs preoperative) time point, with moderate (methionine, leucine, and pyruvic acid) to large (methionine sulfoxide and isoleucine) effect sizes. Electronic Supplementary Material eTable 1 provides details on the number of missing values, descriptive statistics, and results of the regression models.

Fig. 1.

Summary of serum metabolites with significant changes between the preoperative time point and three days after surgery. (A) Model-based mean differences (unit: z-score) including 95% CIs (i.e., the regression coefficients for Time). Serum metabolites (in light blue) are significantly decreased. Metabolites in light green are significantly increased (all FDR-adjusted P values < 0.05). (B) Violin plots of individual changes (in SDs) including means (dots, i.e., the regression coefficients for Time from panel A).

CI = confidence interval; FDR = false discovery rate; SD = standard deviation

We found no significant preoperative or postoperative differences between patients in the two treatment arms of the TAPIR trial (all FDR-adjusted P values > 0.05; ESM eTable 2; eAppendix 4).

Subgroup analyses: serum metabolome in patients with versus without postoperative complications within thirty days after surgery

Preoperative differences

We found no significant differences in the preoperative metabolite levels for patients with vs without occurrence of the primary TAPIR outcome (composite of postoperative complications) or infection (all FDR-adjusted P values > 0.05; Fig. 2). Patients with (vs without) postoperative delirium and patients with (vs without) AKI had significantly higher preoperative succinic acid levels. In both cases, this was owing to influential data from one patient identified in regression diagnostics. After exclusion in sensitivity analyses, we found no significant differences (FDR-adjusted P values > 0.05).

Fig. 2.

Heat map for the main findings of the explorative analyses of postoperative complications. We present the preoperative (D0) and postoperative (POD3) model-based mean differences (unit: z-score). In addition, we present postoperative mean differences adjusted for preoperative metabolite levels (ANCOVA POD3). Negative/positive values indicate lower/higher serum metabolite levels in patients with (vs without) the postoperative complication up to postoperative day 30 indicated in the overarching columns. We also report the regression weight of the interaction term of group and time (group × time), i.e., difference in change between D0 and POD3 in patients with (vs without) the postoperative complication. Per complication (overarching columns), we adjusted all P values (D0, POD3, ANCOVA POD3, and group × time) using the FDR approach. We grouped the serum metabolites according to hierarchical cluster analysis (Euclidean distances of the correlation matrix of metabolite levels at D0, agglomerative clustering with complete linkage).

^*P value of model based mean difference/regression weight < 0.05

^**P value < 0.05 after FDR adjustment for multiple comparisons; interpretation of the absolute values of the mean differences/regression weights: > 0.1 small effect size, > 0.3 medium effect size, and > 0.5 large effect size

ANCOVA = analysis of covariance; DCA/CDCA = deoxycholic acid/chenodeoxycholic acid; FDR = false discovery rate; POD = postoperative day

Postoperative differences

On postoperative day 3, patients with (vs without) delirium within thirty days after surgery (n = 13/177) had significantly lower levels of alanine, asparagine, cholesterol, citrulline, cystine, dimethylglycine, glutamine, glutamic acid, serine, threonine, tyrosine, and uric acid (Fig. 2). Except for cholesterol, dimethylglycine, and uric acid, these metabolites differed significantly between the two groups after adjustment for preoperative metabolite levels. The interaction terms (group × time) for serine and tyrosine showed FDR-adjusted P values < 0.05, indicating significant differences in the preoperative and postoperative changes between patients with (vs without) delirium. For alanine, asparagine, citrulline, cystine, glutamic acid, and threonine, the raw P values of the interaction terms were < 0.05. For all of the above metabolites, the models indicate greater decreases between the preoperative and postoperative time points in patients with (vs without) delirium (ESM eTable 2).

We observed no significant differences in metabolite levels on postoperative day 3 between patients in whom the TAPIR composite outcome occurred by postoperative day 30 (n = 75/177) and those in whom it did not (all FDR-adjusted P values > 0.05). Similarly, we observed no differences between patients with (n = 66/177) and those without infections by postoperative day 30 (all FDR-adjusted P values > 0.05). On postoperative day 3, we found lower serine levels in patients with AKI within thirty days after surgery (n = 15/177) than in patients without. Nevertheless, after adjusting for preoperative metabolite levels, the FDR-adjusted P value was > 0.05. Electronic Supplementary Material eTable 2 summarizes the results of the above analyses.

Discussion

In this secondary, exploratory, targeted metabolomic analysis of high-risk patients undergoing major abdominal surgery under general anesthesia enrolled in the TAPIR trial,¹⁴ we observed complex changes in metabolites analyzed before and after major surgery. In secondary subgroup analyses, there were no statistically significant preoperative differences between patients with and without specific postoperative complications. Nevertheless, we observed significant early postoperative changes in patients who developed postoperative delirium.

Perioperative metabolome changes

Previous research suggests an initial postoperative breakdown of protein metabolism, followed by an catabolic postoperative stress response.²¹ Nevertheless, our findings suggest more nuanced metabolic changes. While some of the changed metabolites were increased postoperatively (methionine, methionine sulfoxide, leucine, and isoleucine), others were decreased (citrulline, 4-hydroxyproline, cystine, glutamine, histidine, dimethylglycine, alanine, glycine, glutamic acid, serine, asparagine, and xanthine). These varied responses imply that postoperative metabolic processes are more complex than previously thought, possibly involving distinct anabolic and catabolic pathways that are selectively activated depending on specific physiologic needs and stressors.

There is increasing evidence that oxidative stress, redox regulation, and antioxidant defence are pivotal in perioperative metabolism.⁶ In line with this, we found a number of changes in metabolites involved in both the formation of reactive oxygen species and the regulation of antioxidant pathways. Postoperatively, glutamine levels were decreased, while methionine and methionine sulfoxide levels were increased. These changes may reflect a defence mechanism against oxidative stress induced by surgery.^22–24 Glutamine is a conditionally essential amino acid in stress and injury; e.g., it plays a key role in acute-phase protein synthesis, wound healing, regulation of tissue protection, and immune response.^23,24 It is also critical for glutathione preservation, maintenance of antioxidant capacity, and cellular metabolism support.^23,24 In critically ill surgical patients, reduced plasma glutamine levels have been associated with the duration of treatment.²⁵ Despite its potential benefits, the clinical application of glutamine remains controversial, particularly in patients with shock or renal failure, where its use may pose risks.²⁶

Consistent with our findings, it is well established that the amino acid derivative uric acid decreases after surgery, likely owing to its strong antioxidant properties.^10,11 Uric acid is primarily synthesized from purines, such as guanosine monophosphate, which is metabolized into uric acid via guanosine and xanthine.²⁷ The observed reductions in plasma concentrations of both xanthine and guanosine in our study may suggest increased use of this metabolic pathway, possibly as part of the body’s antioxidant defence mechanism during the postoperative period.

Consistent with our findings, citrulline depletion has been reported after surgery.^28,29 Citrulline, as a source of arginine and subsequently nitric oxide, has been shown to exhibit beneficial effects in patients with arteriosclerosis, pulmonary and systemic hypertension, and heart failure by contributing to the functionality of endothelial and immune cells.³⁰ While nitric oxide is known to have a vasodilatory effect that is beneficial to the cardiovascular system under certain circumstances, it also causes oxidative stress.⁶ In line with our results, 4-hydroxyproline, a marker of collagen turnover,³¹ has been reported to decrease after major abdominal surgery for colorectal and pancreatic cancer.^12,13

Similar to our results, a previous study found elevated levels of the branched-chain amino acids (BCAA) leucine and isoleucine postoperatively in children with congenital heart disease.³² In critically ill patients with sepsis, lower levels of BCAAs were associated with higher intensive care unit mortality and higher twenty-eight-day mortality.³³ Given their anabolic/anticatabolic properties, BCAA supplementation is being tested in clinical trials in patients with sepsis, burn, and trauma.³⁴ Taken together, these findings suggest that elevated BCAAs may represent a protective, homeostatic response and contribute to recovery by supporting anabolic processes.

Whether the metabolic changes described above represent compensatory processes, homeostatic mechanisms, an adaptive stress response, or the first steps in recovery and remission remains speculative at this time. Further research is needed to clarify the precise role of these metabolic changes in the postoperative period and to determine their potential contribution to patient outcomes.

Subgroup analyses

We found no preoperative or postoperative differences between patients who received routine or personalized hemodynamic management (TAPIR treatment arms). This suggests that the type of hemodynamic management may not have influenced the metabolic changes observed in our study, although further research is needed to confirm this finding and assess its clinical implications.

In our analysis, patients who developed postoperative delirium had lower levels of alanine, asparagine, citrulline, cystine, glutamine, glutamic acid, serine, threonine, and tyrosine in trend. In addition, our metabolomics analysis suggests that lower serum serine and tyrosine levels may be associated with postoperative delirium, which is a novel finding. Our findings may pave the way for future therapeutic interventions aimed at correcting these metabolic disturbances to prevent or attenuate postoperative delirium. We discuss these findings in detail in ESM eAppendix 5.

Notably, we did not observe any specific postoperative metabolomic alterations in patients who later developed a postoperative complication (composite outcome), AKI, or infection. This may suggest that these complications may be more closely related to surgical trauma than to the metabolic background of the patients. Alternatively, future studies investigating such associations may need to focus on other metabolites and at other time points. Furthermore, the underlying (patho-)physiologic mechanisms of postoperative complications are heterogeneous and may lead to different metabolic alterations. In addition, statistical power constraints, especially in smaller subgroups (e.g., 15 patients with AKI), may have limited our ability to detect significant differences. Nevertheless, we identified certain metabolic alterations (before FDR adjustment) for all postoperative complications that may be of interest for future research.

Limitations

Our study successfully identified several metabolic changes that occur after major surgery under general anesthesia. Nevertheless, we cannot exclude interindividual metabolomic variations or influences of metabolic processes from both exogenous and endogenous stimuli, such as diet, preoperative medication, menstrual cycle, or circadian rhythm.^35–39 Factors such as age, sex, and pre-existing comorbidities (e.g., obesity or chronic kidney disease) as well as the duration and type of surgery, fluid therapy, hemodynamic status, and perioperative medications may also significantly influence the metabolome (ESM eAppendix 4; eTable 3). Additionally, variations in LC-MS/MS measurements per se (i.e., measurement error) may have introduced variability.⁴⁰ The number of metabolites assessed could be regarded as modest, considering the potential of MS-targeted metabolomics.⁴¹ Owing to the samples collected, we were only able to measure serum metabolites. It could be of clinical interest to correlate such measurements with those from other matrices (urine and feces) for a more integrative assessment of the patients’ trajectory. The timing of the postoperative blood sampling should also be considered when interpreting the results. Lastly, the single-centre design and the focus on high-risk patients limit the generalizability of our findings.

Conclusions

We observed complex perioperative changes in the metabolome after major surgery under general anesthesia. Certain metabolites were differentially regulated in patients who developed postoperative complications, particularly delirium. In the future, metabolomics may allow the definition of metabolic fingerprints of specific risk groups, leading to improved and individualized patient risk assessment and postoperative care. Our study provides important information to guide further research.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

Open Access funding enabled and organized by Projekt DEAL.