Table 3.

Causal Mediation Analysis of [NET+Ns] as mediator of multi-organ failure in severe COVID-19

Exposure (causal variable): t1-SOFA, or t1-SF Mediator: [NET+Ns], or [ANC] Outcome: t2-SOFA, or ICUFD, or t2-SF-ratio	Natural direct effecta	Natural indirect effectb	Total Effectc	Percent mediatedd	Percent eliminatede	Indirect effect of eliminating [NET+Ns]f-h
Causal pathway hypotheses	Effect (95% CI)	Effect (95% CI)	Effect (95% CI)	Effect (95% CI)	Effect (95% CI)	Effect (95% CI)
A] Direct effect: t1-SOFA→ t2-SOFA Indirect effect: t1-SOFA→[NET+Ns]→t2-SOFA	3.3 (2.0, 4.7)	2.6 (-1.4, 5.9)	5.9 (1.7, 9.9)	44.1 (16.5, 110.6)	46.9 (15.8, 124.6)	0.98 f (0.29, 2.06)
B] Direct effect: t1-SOFA → ICUFD Indirect effect: t1-SOFA →[NET+Ns]→ICUFD	-12.3 (-17.9, -6.2)	-11.0 (-18.5, 4.3)	-23.3 (-33.0, -6.3)	47.1 (22.0, 72.3)	51.1 (22.8, 80.4)	3.0 g (0.85, 7.09)
C] Direct effect: t1-SF → t2-SF Indirect effect: t1-SF →[NET+Ns]→t2-SF	124 (-263, 364)	46 (-224, 481)	170 (103, 235)	27.2 (-107, 372)	15.6 (-9.7, 64.2)	16.8 h (-4.6, 42.8)
D] Direct effect: t1-SF → ICUFD Indirect effect: t1-SF →[NET+Ns]→ICUFD	1.7 (-18.1, 43.9)	13.8 (-36.0, 34.3)	15.4 (5.1, 20.7)	89.3 (-131, 225.5)	22.6 (-15.5, 43.4)
E] Direct effect: t1-SOFA → t2-SOFA Indirect effect: t1-SOFA →[ANC]→t2-SOFA	2.5 (1.0, 4.3)	1.1 (-0.7, 2.2)	3.6 (1.9, 4.8)	31.1 (-18.0, 63.0)	36.5 (-28.2, 70.8)
F] Direct effect: t1-SOFA → ICUFD Indirect effect: t1-SOFA →[ANC]→ICUFD	-13.6 (-22.0, -6.8)	-1.1 (-8.7, 6.2)	-14.6 (-21.0, -9.5)	7.4 (-45.5, 77.8)	0.1 (-54.7, 67.0)

Legend: Causal effect estimates are for t1-SOFA from Q1-Q3 (1 to 6.25), and for t1-SF ratio from Q1-Q3 (135.8 to 346.2) in this prospective pilot cohort. [ANC], absolute neutrophil counts from CBC-differential; ICUFD, ICU-free days at day 28 with death as [-1]; NET, neutrophil extracellular traps; [NET+Ns], immunotyped (DEspR+CD11b+)NET-forming neutrophil counts (10³[K]/μL); SF, SF ratio SpO2/FiO2; SOFA, sequential organ failure assessment score (points) without neurological component; t1, time of consented blood sample draw for analysis of NET+N levels; t2, time before ICU/hospital-discharge or ICU-death < 28 days; 95% CI (n, m), were calculated using bootstrapping with 10,000 replicates

^aNatural direct effect: the direct effect of exposure on outcome if the mediator [NET+Ns] is set at the value it would naturally take when SOFA-1 score is at its reference value

^bNatural indirect effect: the effect of exposure on outcome mediated by DEspR+[NET+Ns], accounting for exposure and mediator interaction

^cTotal Effect: the total effect of exposure on outcome through direct and indirect pathways

^dThe percent of the effect of the exposure on the outcome mediated by DEspR+[NET+N] counts

^eThe percent of the effect of exposure on the outcome that would be eliminated by reducing the DEspR+[NET+N] count to 0 (total effect - controlled direct effect under a hypothetical situation where DEspR+[NET+N] is 0) / total effect)

^fIn patients with t1-SOFA ≥ 2, the indirect effect of eliminating [NET+Ns] decreased t2-SOFA by 0.98 points (95% CI: 0.27,2.13), ~15% absolute risk reduction (arr) (37). In patients with t1-SOFA ≥ 4, the indirect effect of eliminating [NET+Ns] is 1.41 95% CI (0.47,3.05), ~22% arr

^gIn patients with t1-SOFA ≥ 2, outcome model fit in those with t1-SOFA ≥ 2; 50% rrr, relative risk reduction compared to corticosteroid therapy (38-39)

^hIn patients with t1-SF ratio <300, outcome model fit in those with t1-SF ratio <300, not including interaction (p>0.77)