Discussion

Subgroup analysis and clinical outcomes

Baseline demographic characteristics and clinical comorbidities demonstrated no statistically significant association with the composite short-term outcome. Specifically, there was no significant relationship based on patient gender,​​​​​​ X2(1, N=196)=0.30, p=0.583, with a negligible effect size (Cramer’s V=0.039). Similarly, the presence of major medical comorbidities, including diabetes mellitus or hypertension or both, failed to demonstrate a significant association with clinical outcomes, X2(1, N=196)=0.49, p=0.483, Cramer’s V=0.050.

Conversely, the operative approach was a highly significant predictor of short-term patient outcomes, X2(1, N =196)=11.26, p<0.001. This association demonstrated a moderate-to-strong practical effect size, Cramer’s V=0.240, highlighting that the invasiveness of the surgical approach (open versus minimally invasive) is a critical factor in prognostic clinical modelling for this cohort.

A shift in anaemia distribution

This study identifies a notable shift in anaemia distribution, characterised by a migration toward more severe deficiency levels during the perioperative period in vascular surgery patients. Specifically, while 79 (40.3%) of the population presented with mild preoperative anaemia, 99 (50.5%) progressed to moderate postoperative severity and 19 (9.7%) reached severe level. As highlighted by Crispell et al. (2023) and Shah et al. (2023), surgical trauma triggers a significant interleukin-6 (IL-6)-mediated rise in hepcidin [1,7]. This induces a functional iron deficiency by sequestering iron within the reticuloendothelial system, effectively neutralising the patient's ability to mobilise iron stores to compensate for intraoperative losses, thereby explaining the observed migration [1,7].

Comparative model performance and predictive utility

This study demonstrates that postoperative anaemia severity is a potent, independent predictor of 30-day adverse outcomes in vascular surgery patients, with superior discriminative accuracy compared to both preoperative and persistent perioperative anaemia models.

When evaluating predictive accuracy, the postoperative model outperformed both preoperative and persistent anaemia models (AIC=184.10; R2=0.13; AUC=0.69; p=0.001). In contrast, we found weaker evidence supporting the relationship between preoperative anaemia (AIC=188.10; R2=0.10; AUC=0.67; p=0.005) or persistent perioperative anaemia (AIC=191.74; R2=0.04; AUC=0.60; p=0.025) and 30-day outcomes. Our finding aligns with Crispell et al. (2023) and Shah et al. (2023), who posit that the postoperative state serves as a more dynamic indicator of clinical risk [1,7]. This is because it incorporates the cumulative impact of surgical blood loss, haemodilution, and the inflammatory hepcidin blockade, providing a more comprehensive reflection of the patient's physiological status than a static preoperative measure [1,7].

Dose-response relationship between anaemia severity and postoperative outcomes

Our findings demonstrate that while preoperative anaemia is a recognised risk factor in traditional vascular surgery literature, postoperative anaemia severity serves as a significantly more potent indicator of 30-day adverse outcomes. Analysis of the postoperative model identified a clear dose-response relationship, where the risk of the composite 30-day outcome (prolonged hospitalisation seven or more days, unplanned readmission, or mortality) escalated predictably alongside the degree of anaemia severity. Specifically, while mild anaemia did not significantly alter the risk profile (OR: 1.18; 95% CI: 0.45-3.00; p=0.749), the risk escalated sharply for patients with moderate (OR: 4.31; 95% CI: 1.75-10.63; p=0.002) and severe postoperative anaemia (OR: 9.69; 95% CI: 1.17 -80.41; p=0.035), though the wide confidence interval reflects the limited number of patients in the severe anaemia stratum (19, 9.7%) and should be interpreted with caution. Although the present study focuses on vascular surgery, comparable relationships between perioperative anaemia and adverse postoperative outcomes have been reported across other high-risk surgical specialities, including cardiac surgery, suggesting that the adverse impact of perioperative anaemia may represent a broader perioperative phenomenon observed across high-risk surgical populations rather than a finding unique to vascular surgery [4,8,9].

Threshold effects

Our analysis identified a clear severity-response relationship with high odds ratios (ORs) observed in our moderate (OR 4.31) and severe (OR 9.69) cohorts, mirroring the findings of Makar et al. (2024) and Musallam et al. (2011), who demonstrated that the OR of unplanned readmission escalate sharply once discharge Hb falls below 100 g/L [8,9]. This threshold effect is supported by the foundational work of Warner et al. (2023), whose large-scale observational evidence suggests that for every 10 g/L (1 g/dL) decrease in postoperative Hb, the risk of 30-day readmission increases by 8-9% [8,10].

While our study used WHO severity categories rather than continuous Hb values, our "Moderate" and "Severe" groups, both of which fall below the 100 g/L tipping point, captured a similarly significant escalation in risk (OR: 4.31 and 9.69, respectively). This suggests that the WHO Moderate classification serves as a reliable clinical proxy for the high-risk thresholds identified in these large-scale surgical cohorts, confirming that even incremental drops in Hb have a compounded impact on patient recovery.

It should be noted that while Makar et al. (2024) and Warner et al. (2024) focused primarily on readmission rates, our analysis used a composite 30-day outcome. This broader outcome captures the total burden of suboptimal recovery, including both in-hospital delays (prolonged length of stay) and post-discharge complications (unplanned readmission and mortality). This choice of endpoint likely accounts for the high predictive utility of our postoperative model, as it reflects the reality that severe anaemia often results in extended primary hospitalisations, which may statistically mask the risk of subsequent readmission if evaluated in isolation.

The high composite adverse outcome rate (158, 80.6%) warrants careful interpretation, as prolonged hospitalisation (seven or more days) accounted for most events (140, 71.4%). In the context of major vascular surgery, where extended postoperative recovery is frequently required due to procedural complexity and patient comorbidity, a prolonged length of stay may not necessarily indicate significant postoperative morbidity. Consequently, the overall composite adverse outcome rate is likely to be influenced by the inclusion of length of stay as an outcome component and should not be interpreted as indicating that the majority of patients experienced severe postoperative morbidity.

Physiological thresholds in vascular patients

The threshold effect identified in our analysis suggests that while mild anaemia appears relatively well-tolerated, reaching moderate or severe grade represents a critical clinical turning point. The pronounced risk escalation observed in moderate (p=0.002; OR: 4.31; 95% CI: 1.75-10.63) and severe (p=0.035; OR: 9.69; 95% CI: 1.17-80.41) cohorts may be explained by the oxygen supply-demand mismatch described by Leiner et al. (2020) in high-risk surgical patients, whereby haemodynamic instability and impaired oxygen delivery contribute to cellular oxygen debt and postoperative complications [11]. In surgical patients broadly, cardiovascular comorbidity significantly amplifies anaemia-related operative risk; in vascular surgery specifically, coexisting coronary artery disease (CAD) may further reduce tolerance to perioperative haemodynamic stress, with major adverse cardiac event rates reported at approximately 16% following open aortic repair and substantially higher among patients with concomitant anaemia and CAD [12,13]. Once Hb falls below a critical oxygen delivery threshold, tissue oxygen supply dependency ensues, and anaerobic metabolism is initiated, converting moderate haematological deficiency into a primary driver of major clinical morbidity [11,14].

Clinical implications

These findings highlight the postoperative period as a vital window for intervention, particularly for the vascular surgery population. For a vascular patient, moderate anaemia is not merely a laboratory abnormality but a state of haemodynamic fragility that can precipitate myocardial ischaemia or surgical site complications [11]. Therefore, consideration of intravenous iron supplementation in anaemic vascular surgery patients may be warranted in line with current perioperative anaemia guidelines [2]; however, the evidence base for postoperative intravenous iron specifically remains limited, and optimal timing, dosing, and patient selection require further prospective investigation [2,3].

Limitations and future research

Several limitations of this study warrant consideration. Primarily, the single-centre retrospective design and a relatively modest cohort may limit the generalisability of our findings to the extent to which these results can be reliably applied to the broader population of vascular surgery patients in different hospital settings. Within the severe postoperative anaemia stratum, the small sample size (19, 9.7%) contributed to diminished precision and wider confidence intervals for our highest risk estimation. Nevertheless, the monotonic dose-response relationship maintained across all models suggests that the underlying biological association remains statistically coherent, aligning with the Makar et al. (2024) and Warner et al. (2024) studies cited herein [8,15].

Secondly, the absence of serial Hb measurements across the perioperative period prevented the study from distinguishing between acute and chronic anaemia, which may have distinct pathophysiological implications for vascular recovery.

It should be noted that the “Persistent Perioperative Anaemia model” was operationalised as a binary variable (Yes/No) rather than an ordinal severity scale, as the combination of two timepoints did not permit reliable stratification into four severity levels without substantially reducing cell counts. While this approach limits direct comparability of AUC values across models, it retains clinical relevance by capturing the presence or absence of sustained haematological compromise.

Furthermore, the retrospective nature of the data collection meant that data attrition was an unavoidable limitation, as we could not retrospectively recapture lost clinical entries or influence patient selection, which resulted in a high exclusion rate (639, 76.5%), potentially affecting the sample representativeness and introducing selection bias.

This design also introduces the risk of unmeasured confounding, including objective frailty indices, baseline nutritional status, procedural complexity, precise intraoperative blood loss, and the transfusion volumes of whole blood or specific cell components, which fell outside the scope of our data extraction. These factors can independently drive both the depth of perioperative anaemia and the 30-day composite outcome. In particular, the lack of granular data regarding allogeneic red blood cell transfusions is a notable constraint; such interventions not only acutely alter postoperative Hb measurements but also carry independent prognostic weight that can skew 30-day clinical trajectories.

Consequently, while our baseline findings offer valuable foundational insights, they underscore the critical necessity for future well-powered, prospective, multi-centre longitudinal studies. Moving toward a prospective methodology will not only eliminate retrospective data attrition but will also allow for the construction of robust multi-variable risk adjustment models that seamlessly integrate the unmeasured confounders identified in this cohort. Specifically, future protocols should systematically track intraoperative blood loss, blood transfusions, standardised objective frailty indices, and complete baseline iron panels. Furthermore, scaling this research across multiple tertiary centres is essential to increase the sample size within the high-risk severe postoperative anaemia stratum, thereby narrowing our risk estimation confidence intervals. Ultimately, these expanded prospective datasets will be vital to establish definitive, risk-stratified Hb triggers and refine targeted postoperative PBM pathways tailored specifically to the unique haemodynamic requirements of vascular surgery patients.