Potential treatment EMs for SRI-4: characteristics that need to be balanced across trials
| Baseline characteristic | Expected effect-modifying relationship (relative to placebo) with outcome | Anticipated bias (direction and magnitude) for indirect comparison* | Rationale for anticipated bias (values are belimumab trials vs anifrolumab trials) | Adjustment in model | |
| SLEDAI-2K | Treatment effect is larger in populations with higher SLEDAI-2K values |  | Moderate/large bias in favour of anifrolumab | Strong treatment EM with moderate difference (mean of 10.4 vs 11.3) | Balance mean SLEDAI-2K across trials |
| Race | Treatment effect is smaller in Black African ancestry race than other races |  | Negligible bias in favour of anifrolumab | Moderate treatment EM with small difference (20% vs 14% Black African Ancestry) | Not adjusted in the base-case analysis† |
| C3 | Treatment effect is larger among patients with low C3 concentration |  | Moderate bias in favour of belimumab | Moderate treatment EM with moderate difference (49% vs 36% low C3) | Balance proportion with low C3 across trials |
| C4 | Treatment effect is larger among patients with low C4 concentration |  | Moderate bias in favour of belimumab | Moderate treatment EM with moderate difference (37% vs 23% low C4) | Balance proportion with low C4 across trials |
| Anti-dsDNA | Treatment effect is larger among anti-dsDNA positive patients |  | Small bias in favour of belimumab | Small treatment EM with large difference (71% vs 40% positive) | Balance proportion anti-dsDNA positive‡ |
| Glucocorticoid use | Treatment effect is larger among patients with any glucocorticoid use |  | Negligible bias in favour of belimumab | Moderate treatment EM with small difference (88% vs 82% with any use) | Not adjusted in the base-case analysis† |
| Smoking status | Treatment effect is smaller in smokers | ? | Bias for characteristic is unknown | Smoking status is not reported for any trial in the evidence base | None |
| BMI | Treatment effect is smaller in patients with high BMI | ? | Bias for characteristic is unknown | BMI is only available for some trials (but appears similar across trials that report it)§ | None |
*Arrow thickness indicates the strength of the effect.
†Characteristic was adjusted for in sensitivity analyses.
‡Positivity in belimumab trials defined based on 30 IU/mL threshold, while positivity in anifrolumab trials was defined based on 15 IU/mL threshold. While it was possible to alter the definition in the belimumab trials to match the definition in the anifrolumab trials, a 15 IU/mL threshold was not clinically meaningful for the belimumab trials.
§BMI of 27.6 kg/m2 for pooled TULIP trials (reported in Bruce et al18) and BMI of 25.4 kg/m2 for pooled belimumab trials.
Anti-dsDNA, anti–double-stranded DNA antibody; BMI, body mass index; C3/C4, complement component 3/4; EM, effect modifier; SLEDAI-2K, SLE Disease Activity Index 2000; SRI-4, SLE Responder Index-4.