Predicting risk of upper gastrointestinal bleed and intracranial bleed with anticoagulants: cohort study to derive and validate the QBleed scores

Hippisley-Cox et al now have a well-oiled risk engine, generating QBleed to add to their product range.

The average GP practice has hundreds of patients in Atrial Fibrillation. I struggle to decide who to anticoagulate and who not, I have relied on the RCTs and the expert committees, which tell me that a CHADS2 score of 2 or more should always favour anticoagulation, as the HAS-BLED bleeding risk score will rarely outweigh the benefit gained. Des Spence and I have argued that the CHADS2 ( and the CHA2DSVASC ) clearly overestimates stroke risk in today's patients, and needs recalibration. The HAS-BLED score is somewhat crude, but demonstrates that many of the same factors which increase stroke risk will also increase bleeding risk. In my practice many of the factors are weighed on clinical instinct ( compliance, confusibility, etc. ), so I have a keen interest in an up-to-date bleeding risk estimate, and an accurate Stroke risk score.

Hippisley-Cox et al offer me QBleed and QStroke. It should be possible to combine the two into an anticoagulation decision tool - whereby the clinical parameters are entered, and the risk engine computes the number of strokes avoided ( 60% of QStroke ) and the number of serious bleeds in the same timeframe.

I ran a test for a woman aged 70, 75, and 80 years, without other factors. QRisk effectively recalibrates CHADS scores downward, and so it came as no surprise to me to note that the tipping point was somewhere between 75 and 80 years, and not younger as the expert committees claim. But a real problem for GPs ( as demonstrated in Hippisley-Cox et al's fig.2.) is that intracranial and GI bleeds increase exponentially after age 80, underlining the reluctance of many GPs to start Warfarin in the very elderly patient with numerous other problems. It is a pity that QStroke does not extend beyond 84 years. Another problem is that many of the risks occur in the induction phase, or in time out of therapeutic range. It is still unclear whether these are inevitable penalties of Warfarin, as oppsed to the newer anticoagulants.

Given the new NICE Atrial Fibrillation guideline, I look forward to a Qdecision tool combining the Qstroke estimate of benefit with the QBleed estimate of harm.