Figure 2. Therapeutic Misalignments in the ARDSNet Low Tidal Volume Trial ⁵,¹⁰.

The effects of changing tidal volume on mortality in the ARMA trial were significantly different based on pre-randomization compliance levels (Breslow-Day test; p = 0.003 for interaction).¹⁰ In patients with lower pulmonary compliance pre-randomization (compliance < 0.6 mL/cm H₂O/kg PBW), increasing tidal volume (n=156) increased mortality compared to decreasing tidal volume (n=217) (42 vs. 29%; p=0.02). In patients with higher pre-randomization pulmonary compliance (≥ 0.6 mL/cm H₂O/kg PBW), increasing tidal volume (n=57) decreased mortality compared to decreasing tidal volume (n=90) (21 vs. 37%; p=0.07). This interaction remained significant (p < 0.05) in univariable and multivariable analyses controlling for pre-randomization differences in APACHE II score, age, and PaO₂/FiO₂ ratio. Furthermore, this interaction was qualitatively unchanged in multiple sensitivity analyses, including dividing the compliance subgroups at the median and removing “miscast patients” (patients with tidal volumes <6 ml kg pre-randomization that were increased to 6 ml/kg and patients with tidal volumes >12 ml/kg pre-randomization that were decreased to 12 ml/) in both the spline and median divided analyses.¹⁰ (For further details of these analyses, please visit: http://www.cc.nih.gov/ccmd/htmlpg/ccmsupplemental.html)

Figure reproduced from Crit Care Med 2007, 35(6):1509-1516.¹⁰