In lupus, the infection risk followed the steroid, not the immunosuppressant

Design

target trial emulation from claims data with inverse probability weighting and marginal structural Cox models

Population

6,168 patients with non-renal systemic lupus erythematosus starting azathioprine, belimumab, methotrexate or mycophenolic acid analogues, 2011 to 2023

Primary outcome

time to first infection-related hospitalisation

Effect

24-month incidence 13% vs 8% (mycophenolic acid analogues vs belimumab); HR 1.55 (95% CI 1.07 to 2.25), null after adjusting for time-varying glucocorticoid dose

A target trial emulation used US claims data to compare infection-related hospitalisation among 6,168 patients with non-renal systemic lupus erythematosus starting azathioprine, belimumab, methotrexate or a mycophenolic acid analogue between 2011 and 2023. Patients with prior lupus nephritis, transplantation, or rituximab or cyclophosphamide exposure were excluded. Treatment weighting balanced baseline characteristics, and marginal structural Cox models then incorporated monthly glucocorticoid dose as it changed during follow-up.

Two-year cumulative incidence of infection-related hospitalisation was 13% (95% CI 10 to 15) with mycophenolic acid analogues, 10% with azathioprine (8 to 12), 10% with methotrexate (8 to 11) and 8% with belimumab (5 to 10). In intention-to-treat analysis, mycophenolic acid analogues carried higher risk than belimumab (HR 1.55, 1.07 to 2.25) and methotrexate (HR 1.32, 1.04 to 1.68). Per-protocol analysis widened the gap against belimumab for both azathioprine (HR 2.04, 1.01 to 4.16) and mycophenolic acid analogues (HR 2.27, 1.11 to 4.62).

Then came the finding that reframes all of it: once time-varying monthly glucocorticoid dose was modelled, no between-treatment difference remained significant. Read carefully, that has two readings the authors are honest about - the steroid may be the cause of the infections, or it may be the marker of the disease activity that made a particular drug necessary, and it may also be the pathway by which an ineffective drug causes infection. What it does not support is ranking these four agents by infection risk and choosing on that basis. What it does support is treating the steroid dose as the modifiable exposure: record it monthly, taper it deliberately, and judge a steroid-sparing agent by whether it actually spared steroid.

• Do not choose between these four agents on infection risk alone - the differences did not survive adjustment
• Record monthly glucocorticoid dose as a measured outcome of treatment, not a background detail
• Judge a steroid-sparing agent by the dose it actually spared at 6 and 12 months
• Escalate immunosuppression rather than tolerating a persistent prednisolone requirement
• Vaccinate and counsel about infection according to steroid exposure, not just the immunosuppressant

Why it matters

The infection risk attributed to choice of immunosuppressant in lupus may mostly be the steroid that went with it.

Don't overread it

An observational emulation, not a randomised trial; adjusting for glucocorticoid dose can mask a genuine drug effect as well as remove confounding.

The statistics, in plain English

Hazard ratios with intervals like 1.01 to 4.16 are technically significant and practically uninformative about size - the lower bound is no effect at all. That such differences vanished when glucocorticoid dose was added is not proof the drugs are equivalent; adjusting for a variable that lies on the causal pathway can hide a real drug effect as easily as reveal a confounded one. This is claims data emulating a trial, so unmeasured disease activity is the standing limitation, and disease activity is precisely what determines both the drug choice and the steroid dose.

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.