Dynamic Postoperative Trends for Early Detection of Complications After Gastrointestinal Surgery

1. INTRODUCTION 1.1 Background and Rationale Postoperative complications occur in approximately 10-35% of patients undergoing gastrointestinal (GI) surgery and are associated with increased morbidity, prolonged hospital stay, readmissions, and mortality. Early recognition of complications such as surgical site infection, intra-abdominal collections, or anastomotic leakage is essential to prevent progression to sepsis, multi-organ failure, and adverse long-term outcomes. Postoperative surveillance in routine clinical practice relies on clinical assessment, laboratory inflammatory markers, and imaging studies, with imaging typically reserved as a second-level investigation. Inflammatory biomarkers such as white blood cell count, C-reactive protein (CRP), and procalcitonin (PCT) are widely used as surrogate indicators of postoperative complications. However, postoperative recovery is a dynamic, time-dependent process, characterised by an initial inflammatory response followed by gradual resolution. Deviations from this expected trajectory, such as persistence, secondary rises, or discordant trends across different parameters, may occur in patients with self-limiting postoperative courses but may also precede clinically apparent complications by several days. As a result, isolated abnormal values or transient increases may either represent benign postoperative physiology or early signals of pathological evolution, making interpretation in the immediate postoperative phase inherently challenging. In addition, uncertainty in interpreting postoperative inflammatory markers frequently influences antimicrobial prescribing. Empirical initiation or prolonged continuation of antibiotics is common in the presence of non-specific inflammatory responses, even in the absence of confirmed infection. This practice contributes to antimicrobial overuse, increased costs, and antimicrobial resistance. A trajectory-based interpretation of postoperative inflammatory markers may therefore not only improve early complication detection but also support more rational antimicrobial stewardship by identifying patients in whom antibiotic therapy could be safely de-escalated. Most existing studies focus on static cut-off values at predefined postoperative time points, rather than on the longitudinal evolution of combined physiological and inflammatory markers. Consequently, there is limited evidence to guide clinicians in interpreting the overall trajectory of postoperative parameters and in distinguishing physiological postoperative inflammation from early evolving complications. This uncertainty directly affects decisions regarding the appropriate timing of escalation to second-level diagnostic investigations. Therefore, a comprehensive trajectory-based evaluation of postoperative recovery, integrating routinely collected physiological and laboratory markers over time, is needed to improve early risk stratification and clinical decision-making. The DYNAMIC-GI study was designed to address this unmet need by systematically analysing the temporal evolution of postoperative physiological and inflammatory parameters and their association with postoperative complications after GI surgery. 1.2 Study Hypothesis The study hypothesises that postoperative complications after GI surgery are associated with abnormal temporal trajectories of physiological and inflammatory markers. Specifically:

• Patients with an uncomplicated postoperative course exhibit coherent, self-limiting recovery trajectories, consistent with physiological postoperative inflammation.
• Patients who develop postoperative complications demonstrate persistent, non-resolving, or secondary rising trajectories of these parameters. Assessment of the magnitude of these deviations and the coherence between different parameters may provide earlier and more clinically meaningful signals of postoperative complications, improving discrimination between physiological postoperative responses and early evolving complications. 1.3 Objectives Primary Objective: to identify specific trajectory patterns that may enable early prediction of postoperative infective complications after gastrointestinal (GI) surgery requiring interventional procedures or reoperation. Secondary Objectives
• To assess the relative reliability of individual markers and the added value of consistent or discordant elevations across different combinations of markers.
• To explore whether postoperative trajectory patterns of inflammatory markers are associated with the initiation, continuation, or de-escalation of postoperative antibiotic therapy.
• To develop predictive models based on trajectory patterns for the early detection of postoperative infective complications.
• To explore whether trajectory patterns differ according to:
• Different surgical interventions
• Elective versus urgent surgery
• Open versus minimally invasive surgical approaches

1. STUDY DESIGN The DYNAMIC-GI study is a multicentre, prospective, snapshot observational cohort study. No experimental intervention will be applied. All data derive from routine perioperative clinical care. No changes to postoperative management will occur. This design ensures high feasibility, minimal ethical burden, and strong external validity.
2. METHODS: PARTICIPANTS, INTERVENTIONS, AND OUTCOMES

3.1. Study setting

The study will include a broad range of participating centres, such as:

• Academic or community referral hospitals
• Regional and general hospitals Participating centres may differ in surgical volume, case mix, and postoperative care pathways, intentionally reflecting real-world clinical heterogeneity.

Eligible procedures will include major elective or urgent GI surgery under general anaesthesia, at higher risk of early postoperative complications, involving:

• Upper GI tract (oesophagus, stomach, duodenum)
• Hepato-pancreato-biliary surgery
• Small bowel and colorectal surgery Surgical procedures under local anaesthesia, including soft tissue procedures, proctological interventions, abdominal wall surgery, hiatal repair, exploratory laparotomy, or adhesiolysis in the absence of bowel resection, will not be considered for inclusion. See Appendix A for additional information.

3.2 Eligibility Criteria

Inclusion Criteria:

• Age ≥ 18 years
• Elective or urgent major GI surgery
• Availability of postoperative clinical data and at least three postoperative measurements of inflammatory markers to allow identification of trajectory patterns

Exclusion Criteria:

• Procedures without postoperative monitoring
• Death within 24 hours of surgery
• Absence of key postoperative clinical or laboratory data

3.3 Interventions None. This is a purely observational study.

3.4 Outcomes Primary Outcome: Occurrence of postoperative infective complications within 90 days, defined as any infective complication Clavien-Dindo grade > III.

Secondary Outcomes:

• Deep or superficial surgical site infection
• Reoperation
• ICU re-admission
• 30- and 90-day mortality
• Length of hospital stay
• Hospital re-admission

3.5. Variable of interest Clinical Parameters (Daily Preoperative - POD 0 to discharge when available)

• Blood pressure (systolic, diastolic, mean)
• Heart rate
• Body temperature
• Urinary output Laboratory Parameters (Daily Preoperative - POD 0 to discharge when available)
• White blood cell count
• Neutrophil count
• Lymphocyte count
• Neutrophil-to-lymphocyte ratio
• Platelets count
• INR
• C-reactive protein
• Procalcitonin
• Serum creatinine
• Serum total bilirubin
• Serum amylase
• Serum lipase
• Serum lactate

3.6. Participant timeline Participants will be recruited over a 3-month period. Complications will be recorded up to 90 days after the intervention.

3.7. Sample size No formal hypothesis-driven sample size calculation is required. However, because the primary outcome is major postoperative complications (Clavien-Dindo > II), the cohort size is planned to ensure an adequate number of outcome events for trajectory-based modelling. Assuming a major complication rate of approximately 10-15%, a sample of 1,000-1,500 patients is expected to yield 100-225 major complications, which should be sufficient for multivariable and longitudinal analyses.

The number of outcome events is considered adequate to support trajectory-based analyses, including mixed-effects models and multivariable regression frameworks, while limiting the risk of model overfitting and ensuring stable estimation of associations between postoperative marker trajectories and clinical outcomes. The planned sample size is also expected to allow exploratory subgroup analyses by surgical approach, urgency of surgery, and procedure type, while maintaining sufficient statistical precision.

3.8 Statistical analysis Baseline demographic, clinical, and surgical characteristics will be summarised using descriptive statistics. Continuous variables will be reported as mean and standard deviation or median and interquartile range, as appropriate, based on their distribution. Categorical variables will be presented as absolute frequencies and percentages. For estimates, 95% confidence intervals will be provided to quantify statistical precision.

The primary outcome is the occurrence of infective postoperative complications within 90 days, defined as Clavien-Dindo grade ≥ II. Secondary outcomes include specific postoperative complications, reinterventions, ICU readmission, mortality, and length of stay.

Postoperative physiological and laboratory parameters collected daily from postoperative day 0 to day 7 will be analysed as longitudinal measures. Temporal trajectories of these parameters will be modelled using mixed-effects regression models, incorporating patient-level random effects to account for within-subject correlation over time. Time will be modelled flexibly, using either categorical postoperative day indicators or continuous functions, depending on model fit and interpretability. Both linear and non-linear trends will be explored where appropriate.

Associations between marker trajectories and the primary outcome will be assessed by including interaction terms between time and outcome status, allowing comparison of temporal patterns between patients with and without postoperative complications. Multivariable models will adjust for clinically relevant baseline covariates, including age, sex, ASA score, type of surgery, urgency (elective versus urgent), and surgical approach, selected a priori based on clinical relevance rather than automated selection procedures.

For secondary analyses, similar modelling strategies will be applied to specific complications such as surgical site infection, intra-abdominal complications, and need for reoperation. Time-to-event outcomes, where relevant, will be analysed using survival methods, including Cox proportional hazards models, with time-varying covariates considered when appropriate.

Exploratory analyses will assess concordance and discordance between different inflammatory markers by jointly modelling multiple trajectories and evaluating whether combined patterns provide improved discrimination compared with single markers. Additional exploratory analyses will investigate the association between marker trajectories and postoperative antibiotic initiation, continuation, or de-escalation.

The extent and patterns of missing data will be described. Analyses will primarily rely on likelihood-based mixed-effects models, which provide valid inference under the missing-at-random assumption. Sensitivity analyses using multiple imputation may be conducted if the extent of missing data is substantial. In cases of significant missingness in covariates or other variables not directly related to the primary study outcomes, multiple imputation will also be considered.

All statistical tests will be two-sided, with a nominal significance level of 0.05. Given the exploratory nature of the study, no formal multiple-testing adjustment will be applied, and results will be interpreted with appropriate caution. Statistical analyses will be performed using Stata (version 16 or later) and R (version 4.2.0 or later).

3.9 Recruitment Recruitment for the DYNAMIC-GI project will be continuous throughout its duration, targeting surgeons, academic communities, and potential stakeholders. The study is currently endorsed by the Collaborative Alliance for Research and Education in Surgery (CARES).

Strategies to achieve adequate participant enrolment will include regular newsletters, updates, and promotion through social media, as well as national and international conferences.

4 METHODS: DATA COLLECTION, MANAGEMENT, AND ANALYSIS

4.1 Data collection methods The duration of enrolment will be from September 1, 2026, to November 30, 2026. The scheme of the evaluations will follow the standard methods of the centre participating in the study. Clinical information collected at baseline will be recorded in accordance with routine practice at participating centres for patients undergoing GI surgery. The variables collected at baseline will include those listed in Appendix B.

4.2 Data Management Data for the DYNAMIC-GI study will be collected by investigators using an electronic data capture form (eCRF) built within the REDCap environment. REDCap is designed to comply with Good Clinical Practice (GCP) guidelines and the International Council for Harmonisation (ICH), ensuring adherence to regulatory standards. Each local investigator will be responsible for entering data into a REDCap online form for each patient included in the study.

Personal patient-identifying data (name, date of birth, address, phone number, etc.) will not be recorded. Instead, a progressive identification code will be used. Epidemiological, clinical, and treatment-related data will be collected in the eCRF, which can be accessed through a secure system. A link to the eCRF will be sent to a designated contact person (Local Lead) at each participating centre, ensuring that no patient-identifying data are transmitted.

1. ETHICS The study is non-interventional and based exclusively on routinely collected clinical data. It will be conducted in accordance with the General Data Protection Regulation (GDPR), International Council for Harmonisation - Good Clinical Practice (ICH-GCP) principles, the Declaration of Helsinki and National regulations Ethics Committee approval will be obtained at the coordinating centre, with local approval or notification as required at participating centres.

Written informed consent will be obtained from all participating patients prior to data collection, unless a waiver is granted by the local Ethics Committee or Institutional Review Board (IRB) in accordance with national regulations for observational studies. Patients will receive clear information regarding the purpose of the study, the nature of the data collected, data protection measures, and their right to withdraw consent at any time without affecting their clinical care.

Authorship will follow ICMJE recommendations , with recognition of all participating centres.