Reassessing anaemia in gastrectomy: Diagnostic definitions and predictive value in the POWER 4 cohort

Preoperative anaemia is a common finding in patients undergoing gastrectomy, and is frequently targeted in optimisation strategies. However, its independent association with postoperative complications remains unclear, and the prognostic value of redefining diagnostic thresholds has not been formally assessed.

This prospective substudy of the POWER 4 cohort included 742 patients undergoing elective gastrectomy for cancer. Anaemia was defined by WHO sex-specific thresholds (Hb <13 g/dL in men, <12 g/dL in women) and a universal threshold of <13 g/dL. Four subcohorts were analysed: A1 (WHO, full cohort), A2 (universal, full), B1 (WHO, untreated), B2 (universal, untreated). Multivariable logistic regression was used to assess associations with postoperative complications. Discrimination was evaluated using AUC, AIC/BIC, and Net Reclassification Improvement (NRI). Generalised additive models (GAMs) were used to explore continuous relationships between haemoglobin and risk.

Anaemia prevalence was 38.0% with WHO thresholds and 52.4% with the universal threshold. In full-cohort models, anaemia was not independently associated with complications (A1 OR 0.92 [95% CI 0.60-1.41]; A2 OR 0.89 [0.59-1.36]). Model performance was similar (AUC 0.686). In untreated patients (n = 584), discrimination improved (B1 AUC 0.750; B2 AUC 0.743), but anaemia remained non-significant. NRI modestly favoured the universal threshold (A1/A2: 0.048; B1/B2: 0.072), with most reclassification occuring in intermediate risk groups. GAMs showed a smooth monotonic relationship without a clear inflection point.

Anaemia was not independently associated with postoperative complications. Redefining thresholds altered prevalence but did not improve prediction. Haemoglobin may be better used as part of an integrated perioperative risk assessment rather than a binary stratifier.