Surgical Management of Pancreatic Cancer With Liver Metastasis (Oligometastasis)

Study Design and Registry Framework This investigation represents a nationwide retrospective cohort study based on the German Diagnosis-Related Group (DRG) hospital discharge database covering the period 2010-2023. The DRG system constitutes a mandatory administrative registry capturing all inpatient hospitalizations across Germany, thereby providing full national coverage without center-level voluntary participation bias.

The registry includes patient demographics, ICD-10 diagnosis codes, OPS procedure codes, discharge status, and institutional identifiers. For this study, all adult patients (≥18 years) with a principal diagnosis coded as malignant neoplasm of the pancreas (ICD-10 C25) who underwent pancreatic resection were included.

Quality Assurance Plan The German DRG registry operates under a legally mandated standardized coding framework. Coding is performed by trained hospital coders and subject to institutional internal audits as well as external audits by health insurance funds and the Medical Service of Health Insurance (MD). Financial reimbursement is directly linked to coding accuracy, creating systematic incentives for data validity.

For this study:

Only validated national DRG datasets were used. Data were obtained from the Federal Statistical Office. All data were anonymized prior to analysis. Logical consistency checks were performed before statistical analysis. No site-level monitoring was required, as this is a population-based administrative dataset.

Data Validation and Registry Procedures

Data validation in the DRG system includes:

Automated plausibility checks at the hospital level. Cross-validation between diagnosis and procedure codes. Reimbursement-driven coding audits. Periodic national validation procedures.

Before analysis, additional study-specific validation steps were performed:

Removal of patients with missing age or sex. Hierarchical ranking of pancreatic procedures to avoid double counting. Exclusion of implausible combinations (e.g., total pancreatectomy following partial resection within same admission).

Verification of coding consistency between C78.7 (secondary liver malignancy) and liver resection codes.

Data Checks and Internal Consistency

Data were checked against predefined logical rules:

Age ≥18 years. Valid combinations of pancreatic resection types. Liver resection categorized as non-anatomic or anatomic based on OPS coding. Cross-check of metastatic liver diagnosis (C78.7) with liver resection procedures.

Range checks for continuous variables (e.g., age). Inconsistencies were excluded or corrected according to predefined rules.

Source Data Verification

As this is a nationwide administrative dataset, direct medical record review was not possible. However, source data integrity is supported by:

Mandatory hospital coding standards. Financial audit procedures. External audits by statutory health insurance authorities. National validation studies demonstrating reliability of DRG coding for major surgical procedures.

Therefore, indirect source data verification is inherent to the registry's structure, although patient-level chart verification was not performed.

Data Dictionary

Key variables included:

Demographics:

• Age (years)
• Sex (male/female)

Diagnoses:

• ICD-10 C25 (malignant neoplasm of pancreas)
• ICD-10 C78.7 (secondary malignant neoplasm of liver) Comorbidities defined via Charlson algorithm (Quan et al.)

Procedures: OPS codes for:

• Distal pancreatectomy
• Pancreatic head resection
• Total pancreatectomy
• Liver wedge resection
• Segmentectomy
• Hemihepatectomy
• Thermal ablation

Outcomes:

In-hospital mortality (binary) Failure to rescue (death after major complication) Institutional Variable

Annual hospital caseload categorized as:

<10, 10-19, 20-49, ≥50 pancreatic resections/year

Coding systems used:

ICD-10-GM (German modification) OPS (German procedure coding system) Charlson Comorbidity Index (Quan adaptation)

Data Management:

• Secure data storage.
• Anonymized dataset.
• Reproducible R scripts.

Analysis Workflow:

• Descriptive statistics.
• Group comparisons.
• Multivariable logistic regression.
• Volume-outcome marginal effect estimation.

Change Management Any analytic modification required documentation and version tracking in the statistical analysis script.

Sample Size Assessment

This study is based on complete nationwide capture (N = 68,303 resections). Given the large cohort size:

Statistical power was sufficient to detect small effect sizes. Mortality differences ≥1% were detectable with high precision. No formal prospective power calculation was required due to exhaustive national sampling.

The subgroup with simultaneous liver resection (n=2,447) provided adequate power for multivariable modeling.

Plan for Missing Data Administrative datasets typically have minimal missingness for key variables (age, sex, discharge status).

Approach:

Exclusion of patients with missing age or sex. No imputation performed. Comorbidity coding assumed absent if not coded. Logical inconsistencies were excluded prior to analysis. No multiple imputation was required.

Statistical Analysis Plan

Primary Objective:

Assess association between simultaneous liver resection and in-hospital mortality.

Secondary Objective:

Assess failure to rescue and volume-outcome relationships.

Analytical Principles

• Two-sided hypothesis testing.
• Significance threshold p < 0.05.
• Reporting of odds ratios with 95% confidence intervals.

Methods:

Descriptive analysis (means, medians, proportions). Chi-square and Wilcoxon tests. Multivariable logistic regression using Firth's correction to reduce small-sample bias.

Adjustment variables:

• Age group
• Sex
• Pancreatic procedure type
• Individual Charlson comorbidities
• Hospital case volume
• Year of operation

Subgroup analyses:

• Anatomic vs non-anatomic liver resection.
• Volume-stratified outcome modeling.
• Marginal standardized predicted probabilities for hospital volume categories.

All analyses were performed in R (version 4.3).

Limitations of Registry-Based Design No histopathological differentiation beyond ICD-10 C25. No systemic therapy data. No long-term survival. No 90-day mortality. No tumor burden quantification. Heterogeneous patient selection likely.

Summary This registry-based nationwide analysis applies structured validation procedures, predefined analytic rules, and standardized coding systems to evaluate the safety of simultaneous pancreatic and liver resections. While administrative data limit oncologic interpretation, the methodological framework ensures robust short-term safety assessment across a complete national cohort.