A relapse calculator for severe ANCA vasculitis, built from PEXIVAS

Design

Post hoc risk-prediction model derivation from a randomised trial

Population

649 PEXIVAS participants in remission after severe ANCA-associated vasculitis

Primary outcome

First relapse within 2 years of remission

Effect

100 relapses; optimism-corrected C statistic 0.709; calibration slope 0.99

Researchers used 649 participants from the PEXIVAS trial who reached remission after presenting with severe ANCA-associated vasculitis (eGFR below 50 or diffuse alveolar haemorrhage). Of these, 100 relapsed within two years.

The final model estimates two-year relapse risk from five things: sex, whether induction used oral cyclophosphamide or rituximab, ANCA subtype, non-haemorrhagic respiratory or mucosal and eye involvement at presentation, and eGFR at remission. Discrimination was moderate (optimism-corrected C statistic 0.71) and calibration good.

Relapse risk drives decisions about how long to continue maintenance and how often to monitor. A structured estimate is better than intuition, but this model was built in a trial population with severe disease and has not been tested in other cohorts. It is a tool to inform discussion rather than to set maintenance duration on its own.

• Record the factors linked to relapse: sex, induction agent, ANCA subtype, respiratory or mucosal and eye involvement, and eGFR at remission.
• Use relapse risk to inform the intensity of monitoring after remission.
• Do not shorten maintenance therapy on the basis of this model alone until it is externally validated.
• Expect about one in six patients with severe disease to relapse within two years of remission, as in this cohort.

Why it matters

It moves relapse prediction from impression towards a structured estimate using routinely available information.

Don't overread it

The model has not been externally validated and comes from a trial population with severe kidney or lung involvement.

The statistics, in plain English

A C statistic of 0.71 means that in about 71% of pairs of patients, the model gives the one who relapses a higher risk score. That is moderate: useful at the group level, less reliable for an individual.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.