Asymmetric risk score architecture in atrial fibrillation: renal impairment inflates bleeding but not thrombotic risk estimation
I Alevroudis, A Evangeliou, S Vergopoulos, M Petridou, D Oikonomou, E A Pavlou, D O Pavlou, S Mitsiadis, S Tzikas, C Papadopoulos, G Giannopoulos, V VassilikosAbstract
Background/Introduction
In atrial fibrillation (AF), anticoagulation decisions balance thromboembolic risk (CHA2DS2-VASc) against bleeding risk (HAS-BLED). Renal impairment is incorporated into HAS-BLED but is absent from CHA2DS2-VASc. This asymmetry may bias decision-making: renal patients are flagged as high bleeding risk while their concomitantly elevated thrombotic burden—driven by comorbidity clustering—remains unrecognised by the thrombotic score. We hypothesised that this creates a net bias toward anticoagulation undertreatment.
Purpose
To quantify the risk estimation discordance generated by the asymmetric incorporation of renal impairment into HAS-BLED but not CHA2DS2-VASc, and to evaluate whether this translates into a risk–treatment imbalance.
Methods
We analysed 424 consecutive patients with non-valvular AF and available renal data from a single-centre registry. Renal impairment was defined as elevated creatinine or documented chronic kidney disease. CHA2DS2-VASc and HAS-BLED scores were compared between groups. A corrected HAS-BLED was calculated by subtracting the renal component (1 point) to isolate the structural bias. The proportion simultaneously classified as high thrombotic (CHA2DS2-VASc ≥4) and high bleeding risk (HAS- BLED ≥3) was compared. Observed bleeding events and anticoagulation rates were assessed.
Results
Of 424 patients, 49 (11.6%) had renal impairment. Despite renal disease being absent from CHA2DS2-VASc, renal patients had significantly higher thrombotic scores (4.6±1.8 versus 3.8±1.7;p=0.005) driven by comorbidity clustering: higher diabetes (49.0% versus 30.5%; p=0.015), trends toward more vascular disease (20.5% versus 9.4%) and hypertension (87.8% versus 75.7%). HAS-BLED was markedly higher (3.3±0.9 versus 2.1±1.0; p<0.001), with 87.8% classified as high bleeding risk versus 28.8%. After removing the renal point, corrected HAS-BLED narrowed substantially (2.3 versus 2.1), indicating the gap was largely driven by the structural inclusion of renal disease itself. Critically, 71.4% of renal patients had both high thrombotic and bleeding risk versus 24.0% without renal disease. Despite this dual burden, observed bleeding was lower in renal patients (4.1% versus 6.9%; p=0.76) and anticoagulation rates were similar (63.3% versus 54.8%).
Conclusions
Current risk stratification creates a structural asymmetry: renal impairment inflates predicted bleeding via HAS-BLED while its associated thrombotic burden remains invisible to CHA2DS2- VASc. This disproportionately flags renal patients as high bleeding risk without recognising their elevated stroke risk, potentially discouraging anticoagulation in those who may benefit most. The low observed bleeding despite high predicted risk suggests HAS-BLED overestimates haemorrhagic risk here. These findings support renal impairment as a thrombotic risk modifier and call for score recalibration to restore balanced net clinical benefit assessment in AF with kidney disease.Rena asymmetric risk