Table 3.

Change in VACS index between baseline and 18 months by randomization group and cumulative adherence (high versus low) under different methods assumptions. ZINC trial 2013–2015

	Adjusted mean difference in change from baseline to 18 months (95% CI)
Per-protocol effect—main analysisa	−8.01 (− 16.42, 0.01)
Per-protocol effect—high adherence ≥70%a	−4.07 (− 11.50, 2.75)
Per-protocol effect—high adherence ≥90%a	− 12.34 (− 12.14, − 4.14)
Per-protocol effect—high adherence ≥80%—riboflavin testa	−3.48 (− 10.44, 4.27)
Per-protocol—change in the temporal order of post-baseline covariatesa	− 7.63 (− 15.21, 0.39)
Intention-to-treat with inverse probability of censoring weightsa	−7.08 (− 13.98, 0.04)
Intention-to-treat—original analysis with multiple [4] imputation	− 4.68 (− 9.62, 0.25)
Per-protocol—original “standard” analysis [4]	−7.49 (− 13.74, − 1.23)

^aEstimated using inverse probability weighting