Predictive value of cytokines for developing complications after polytrauma

Table 1 Overview of included studies, the studied cytokines and the outcome parameters (acute respiratory distress syndrome, sepsis, muli-organ dysfunction syndrome, multi-organ failure, mortality)

No.	Ref.	Year	Design	No pts. (control)	Cytokines	ARDS (%)	Sepsis (%)	MODS (%)	MOF (%)	Mortality (%)
1	Billeter et al[35]	2009	P-coh	1032	IL-6					10%
2	Bogner et al[36]	2009	P-coh	58	IL-6, -8, -10				74%	19%
3	Cook et al[58]	2013	P-cc	83 (18)	G-CSF		7%			7%
4	Cuschieri et al[34]	2010	P-coh	152	IL-6			37%		5%
5	Donnelly et al[37]	1994	P-coh	15	IL-6, -8, -1β; TNF-	49%				33%
6	Dresing et al[26]	2004	P-coh	30	IL-6; TNF-			13%		19%
7	Egger et al[38]	2004	P-coh	26	IL-6, -8		35%
8	Flores et al[39]	2001	P-coh	43	IL-6		49%			16%
9	Frangen et al[59]	2008	P-cc	71 (25)	IL-17, -6					22%
10	Frank et al[11]	2002	P-cc	77 (15)	IL-6, -8					9%
11	Frink et al[3]	2009	P-coh	143	IL-1β, -6, -8, -10; TNF-		29%	17%		15%
12	Gebhard et al[40]	2000	P-coh	94	IL-6					19%
13	Giamarellos-Bourboulis et al[55]	2008	P-cc	69 (10)	IL-6, -8; TNF-, IFN-γ		62%			35%
14	Gouel-Chéron et al[53]	2012	P-cc	100 (18)	IL-6, -10		37%			5%
15	Haasper et al[28]	2010	P-coh	94	IL-6		16%	22%		13%
16	Hayakawa et al[31]	2011	P-coh	45	TNF-			53%		25%
17	Heizmann et al[52]	2008	R-cc	195 (10)	IL-2, -4, -10, -11, -12, -18; IFN-γ					19%
18	Jastrow et al[32]	2009	P-coh	48	IL-6, -8, -10, -1β, -2, -4, -12; TNF-				23%	17%
19	Keel et al[41]	2009	P-coh	83	IL-6		40%			12%
20	Lausevic et al[33]	2008	P-coh	65	IL-6, -10		62%		55%	51%
21	Lausevic et al[29]	2010	P-coh	65	IL-6, -10		63%			51%
22	Law et al[42]	1994	P-coh	13	IL-6, -8; TNF-				46%	23%
23	Lendemans et al[13]	2004	P-coh	16	IL-6, -10; TNF-				56%
24	Liener et al[43]	2002	P-coh	94	IL-8	0%	0%		0%	19%
25	Livingston et al[44]	1988	P-coh	20	IFN-γ		30%			15%
26	Maier et al[27]	2007	P-coh	251	IL-6, -8, -10				34%	12%
27	Meade et al[45]	1994	P-coh	25	IL-6, -8; TNF-	36%
28	Menges et al[50]	1999	P-coh	68	IL-10, -1; TNF-		25%		25%	1%
29	Mommsen et al[30]	2009	P-coh	55	IL-18		42%	13%		13%
30	Neidhardt et al[54]	1997	P-cc	417 (137)	IL-10	5%	11%	22%		22%
31	Oberholzer et al[46]	2000	P-coh	1276	IL-6, IL-10		14%	40%		7%
32	Partrick et al[56]	1996	P-cc	27 (6)	IL-6, -8				33%	7%
33	Paunel-Görgülü et al[47]	2011	P-coh	47 (17)	IL-6		38%			11%
34	Raymondos et al[48]	2012	P-coh	24	IL-6, -8, -1β, TNF-	29%				4%
35	Roetman et al[60]	2008	P-cc	229 (110)	IL-18, -4; IFN-γ					16%
36	Schinkel et al[61]	2005	P-cc	216 (110)	IL-11				4%	16%
37	Sherry et al[14]	1996	R-cc	66 (10)	IL-10	8%	39%			2%
38	Sousa et al[51]	2015	P-coh	99	IL-6, -10; TNF-	19%		34%		28%
38	Spielmann et al[57]	2001	P-cc	47 (15)	TNF-	11%	30%	51%		23%
39	Svoboda et al[62]	1994	P-cc	42 (12)	IL-1β, -2, -6; TNF-				33%	26%
40	Wick et al[49]	2000	P-coh	37	IL-12			11%		16%
41	Yagmur et al[63]	2005	P-cc	99 (10)	IL-1, -2, -6, -8; TNF-					17%

P-coh: Prospective cohort study; P-cc: Prospective case-control study; R-cc: Retrospective case-control study; IL: Interleukin; TNF: Tumor necrosis factor; IFN: Interferon; Pts: Patients; Y: Yes; N: No.

Table 2 Value of cytokine concentrations for predicting acute respiratory distress syndrome

Ref.	Year	Design	No pts.	ARDS n (%)	Predicts ARDS	Results
IL-6
Donnelly et al[37]	1994	P-coh	15	7 (49%)	N	[IL-6] is not significantly different in ARDS
Meade et al[45]	1994	P-coh	25	9 (36%)	N	[IL-6] is higher in patients with ARDS after onset of symptoms; does not predict development of ARDS
Raymondos et al[48]	2012	P-coh	24	7 (29%)	Y	[IL-6] is significantly higher in patients at high risk for ARDS
Sousa et al[51]	2015	P-coh	99	19 (19%)	Y	[IL-6] is significantly higher at 72 h post injury
IL-8
Donnelly et al[37]	1994	P-coh	15	7 (49%)	Y	[IL-8] is significantly higher in patients with ARDS, starting at 16 h post injury
Meade et al[45]	1994	P-coh	25	9 (36%)	N	[IL-8] is higher in patients with ARDS after onset of symptoms; does not predict development of ARDS
Raymondos et al[48]	2012	P-coh	24	7 (29%)	Y	[IL-8] is significantly higher in patients at high risk for ARDS
IL-10
Neidhardt et al[54]	1997	P-cc	417	19 (5%)	N	[IL-10] is not related to the development of ARDS
Sherry et al[14]	1996	R-cc	66	5 (8%)	N	[IL-10] is not related to the development of ARDS
Sousa et al[51]	2015	P-coh	99	19 (19%)	Y	[IL-10] is significantly higher in patients with ARDS upon admission, at 24 + 48 + 72 h post injury
Spielmann et al[57]	2001	P-cc	47	5 (11%)	N	[IL-10] is not related to the development of ARDS
TNF-α
Donnelly et al[37]	1994	P-coh	15	7 (49%)	N	[TNF-α] below detection limit
Meade et al[45]	1994	P-coh	25	9 (36%)	N	[TNF-α] below detection limit
Sousa et al[51]	2015	P-coh	99	19 (19%)	N	[TNF-α] is not related to the development of ARDS
IL-1β
Donnelly et al[37]	1994	P-coh	15	7 (49%)	N	[IL-1β] below detection limit
Meade et al[45]	1994	P-coh	25	9 (36%)	N	[IL-1β] below detection limit

P-coh: Prospective cohort study; P-cc: Prospective case-control study; R-cc: Retrospective case-control study; IL: Interleukin; TNF: Tumor necrosis factor; Pts: Patients; Y: Yes; N: No.

Table 3 Value of cytokine concentrations for predicting sepsis

Ref.	Year	Design	No pts.	Sepsis n (%)	Diagnostic tests	Predicts sepsis	Results
IL-6
Billeter et al[35]	2009	P-coh	1032			Y	[IL-6] is significantly higher in sepsis between days 3-7
Egger et al[38]	2004	P-coh	26	9 (35%)		N	[IL-6] is significantly higher in sepsis before clinical manifestations; does not predict sepsis
Flores et al[39]	2001	P-coh	43	21 (49%)		N	[IL-6] is not significantly altered in sepsis
Giamarellos-Bourboulis et al[55]	2008	P-cc	69	43 (62%)	ROC AUC 0.500 (95%CI: 0.304-0.696, P > 0.05)	N	[IL-6] is not related to the development of sepsis
Gouel-Chéron et al[53]	2012	P-cc	100	37 (37%)	> 67.1 pg/mL: Sensitivity 85%; specificity 73%	Y	[IL-6] > 67.1 pg/mL is predictive for sepsis on days 1 + 2 (OR = 10.9)
Haasper et al[28]	2010	P-coh	94	15 (16%)		N	[IL-6] is not significantly different in sepsis
Keel et al[41]	2009	P-coh	83	33 (40%)		Y	[IL-6] is significantly higher in sepsis on days 5 + 14
Lausevic et al[33]	2010	P-coh	65	41 (63%)		N	[IL-6] is not predictive for sepsis
Oberholzer et al[46]	2000	P-coh	1276	179 (14%)		Y	[IL-6] is significantly higher in septic patients
Paunel-Görgülü et al[47]	2011	P-coh	47	18 (38%)	AUC ROC 0.79 (day 5 post injury)	Y	[IL-6] is significantly elevated on days 5 + 9 in sepsis
IL-8
Egger et al[38]	2004	P-coh	26	9 (35%)		N	[IL-8] is not significantly altered in sepsis
Giamarellos-Bourboulis et al[55]	2008	P-cc	69	43 (62%)	AUC ROC 0.453 (95%CI: 0.254-0.652, P > 0.05)	N	[IL-8] is not predictive for sepsis
IL-10
Gouel-Chéron et al[53]	2012	P-cc	100	37 (37%)		N	[IL-10] is not related to the development of sepsis
Lausevic et al[33]	2010	P-coh	65	41 (63%)		Y	[IL-10] is significantly lower in sepsis on days 1 + 2
Menges et al[50]	1999	P-coh	68	17 (25%)		Y	[IL-10] is significantly higher in sepsis and MOF after 6 d
Neidhardt et al[54]	1997	P-cc	417	45 (11%)		Y	[IL-10] is significantly higher in sepsis on days 1 + 3 + 5 + 7 + 10 + 14 + 21
Sherry et al[14]	1996	R-cc	66	26 (39%)		Y	[IL-10] is significantly higher in sepsis
TNF-α
Giamarellos-Bourboulis et al[55]	2008	P-cc	69	43 (62%)	AUC ROC 0.466 (95%CI: 0.274-0.657, P > 0.05)	N	[TNF-α] is not related to the development of sepsis
Menges et al[50]	1999	P-coh	68	17 (25%)		Y	[TNF-α] is significantly higher in sepsis and MOF after 8 d
IFN-γ
Giamarellos-Bourboulis et al[55]	2008	P-cc	69	43 (62%)		N	[IFN-γ] below detection limit
Livingston et al[44]	1988	P-coh	20	6 (30%)		Y	[IFN-γ] is markedly lower in sepsis after 14 d
G-CSF
Cook et al[58]	2013	P-cc	83	6 (7%)		Y	[G-CSF] > 500 pg/mL is significantly associated with sepsis
IL-18
Mommsen et al[30]	2009	P-coh	55	23 (42%)		Y	[IL-18] is significantly higher in sepsis on days 3-6 post injury
IL-1
Menges et al[50]	1999	P-coh	68	17 (25%)		Y	[IL-1] is significantly higher in sepsis and MOF on days 3 + 5 + 6 + 9 - 13

P-coh: Prospective cohort study; P-cc: Prospective case-control study; R-cc: Retrospective case-control study; IL: Interleukin; TNF: Tumor necrosis factor; IFN: Interferon; ROC: Receiver operating characteristic; AUC: Area under curve; Pts: Patients; Y: Yes; N: No.

Table 4 Value of cytokine concentrations for predicting muli-organ dysfunction syndrome

Study	Year	Design	No pts.	MODSn (%)	Diagnostic tests	Predicts MODS	Results
IL-6
Cuschieri et al[34]	2010	P-coh	152	29 (37%)	> 350 pg/mL: Sensitivity 79%, specificity 76%; OR = 3.87 (95%CI: 1.13-11.19)	Y	[IL-6] > 350 pg/mL is highly associated with MODS
Frink et al[3]	2009	P-coh	143	24 (17%)	r = 0.35; > 761.7 pg/μL: Sensitivity 16.7%, specificity 98.3%	Y	[IL-6] > 76.6 pg/μL is associated with MODS with accuracy of 84.7%
Haasper et al[28]	2010	P-coh	94	21 (22%)		Y	[IL-6] is significantly higher in MODS on days 1 + 7
Oberholzer et al[46]	2000	P-coh	1276	516 (40%)		Y	[IL-6] is significantly higher in (severe) MODS
Sousa et al[51]	2015	P-coh	99	34 (34%)	> 294 pg/mL: AUC ROC 0.769 (95%CI: 0.414-0.736)	Y	[IL-6] > 294 pg/mL is associated with MODS at 48 + 72 h post injury
IL-8
Frink et al[3]	2009	P-coh	143	24 (17%)	r = 0.53; sensitivity 0%	N	[IL-8] is significantly higher in MODS; does not predict development of MODS
IL-10
Frink et al[3]	2009	P-coh	143	24 (17%)	r = 0.31; sensitivity 0%	N	[IL-10] is significantly higher in MODS; does not predict development of MODS
Neidhardt et al[54]	1997	P-cc	417	92 (22%)		Y	[IL-10] is significantly higher in MODS on days 1 + 3 + 5 + 7 + 10 + 14 + 21 post injury
Spielmann et al[57]	2001	P-cc	47	24 (51%)		N	[IL-10] is not related to the development of MODS
Sousa et al[51]	2015	P-coh	99	34 (34%)	> 4.93 pg/mL: AUC ROC 0.700 (95%CI: 0.506-0.841)	Y	[IL-10] > 4.93 pg/mL is associated with MODS at 24 + 72 h post injury
TNF-α
Frink et al[3]	2009	P-coh	143	24 (17%)	r = 0.32; sensitivity 0%	N	[TNF-α] is significantly higher in MODS; does not predict development of MODS
Hayakawa et al[31]	2010	P-coh	45	24 (53%)		Y	[TNF-α] is significantly higher in MODS on days 3 + 5
Sousa et al[51]	2015	P-coh	99	34 (34%)		Y	[TNF-α] is associated with MODS at 48 h post injury
Spielmann et al[57]	2001	P-cc	47	24 (51%)		N	[TNF-α) is not associated with MODS
IL-1β
Frink et al[3]	2009	P-coh	143	24 (17%)	r = 0.00; sensitivity 0%	N	[IL-1β] is not related to development of MODS
IL-12
Wick et al[49]	2000	P-coh	37	4 (11%)		Y	[IL-12] is significantly lower in patients with MODS
IL-18
Mommsen et al[30]	2009	P-coh	55	7 (13%)		Y	[IL-18] is significantly higher in MODS on days 2 + 3 + 6 + 7 + 9 + 10 + 13 + 14
MIF
Hayakawa et al[31]	2010	P-coh	45	24 (53%)		Y	[MIF] is significantly higher in MODS

P-coh: Prospective cohort study; P-cc: Prospective case-control study; R-cc: Retrospective case-control study; IL: Interleukin; TNF: Tumor necrosis factor; IFN: Interferon; ROC: Receiver operating characteristic; AUC: Area under curve; r: Correlation coefficient between cytokine and development of MODS; MODS: Muli-organ dysfunction syndrome; Pts: Patients; Y: Yes; N: No.

Table 5 Value of cytokine concentrations for predicting multi-organ failure

Ref.	Year	Design	No pts.	MOFn (%)	Diagnostic tests	PredictsMOF	Results
IL-6
Bogner et al[36]	2009	P-coh	58	43 (74%)		Y	[IL-6] is significantly higher in MOF at 0 - 24 + 72 h
Frank et al[11]	2002	P-cc	77		r = 0.25 on day 2	N	[IL-6] is significantly higher in MOF; no reliable predictor due to low r
Jastrow et al[32]	2009	P-coh	48	11 (23%)	AUC ROC 0.816; (IL-6) > 0.861 pg/mL: sensitivity 57%, PPV 100%	Y	[IL-6] > 0.861 pg/mL is highly predictive for MOF
Lausevic et al[33]	2008	P-coh	65	36 (55%)		Y	[IL-6] is significantly higher in MOF on all days of hospitalization
Lendemans et al[13]	2004	P-coh	16	9 (56%)		Y	[IL-6] is significantly higher in MOF after two weeks
Law et al[42]	1994	P-coh	13	6 (46%)		N	[IL-6] is elevated in MOF, does not predict MOF
Maier et al[27]	2007	P-coh	251	85 (34%)	AUC ROC 0.70 for late-onset MOF	Y	[IL-6] is predictive for (late) MOF
Partrick et al[56]	1996	P-cc	27	9 (33%)		Y	[IL-6] is significantly higher in MOF at 12 + 36 h
Svoboda et al[62]	1994	P-cc	42	14 (33%)		N	[IL-6] is higher in MOF at day 1, does not predict MOF
IL-8
Bogner et al[36]	2009	P-coh	58	43 (74%)		Y	[IL-8] is significantly higher in MOF from 0-72 h
Frank et al[11]	2002	P-cc	77		r = 0.32 on day 2	N	[IL-8] is significantly higher in MOF; not reliable due to low r
Jastrow et al[32]	2009	P-coh	48	11 (23%)		Y	[IL-8] is significantly higher in MOF from 0-24 h
Law et al[42]	1994	P-coh	13	6 (46%)		N	[IL-8] is elevated in MOF, does not predict MOF
Maier et al[27]	2007	P-coh	251	85 (34%)	AUC ROC 0.69 for late-onset MOF	Y	[IL-8] is predictive for (late) MOF
Partrick et al[56]	1996	P-cc	27	9 (33%)		Y	[IL-8] is significantly higher in MOF at 12 + 36 + 84 h
IL-10
Bogner et al[36]	2009	P-coh	58	43 (74%)		Y	[IL-10] is significantly higher in MOF in early post-injury phase (< 12 h)
Jastrow et al[32]	2009	P-coh	48	11 (23%)	AUC ROC 0.776; (IL-10) > 38.6 pg/mL: Sensitivity 71%, PPV 77%	Y	[IL-10] > 38.6 pg/mL is predictive for MOF
Lausevic et al[33]	2008	P-coh	65	36 (55%)		Y	[IL-10] is significantly higher in MOF in very early post injury phase
Lendemans et al[13]	2004	P-coh	16	9 (56%)		Y	[IL-10] is significantly higher in MOF on days 3 + 4
Maier et al[27]	2007	P-coh	251	85 (34%)	AUC ROC 0.60 for late-onset MOF	N	[IL-10) is not predictive for MOF
Menges et al[50]	1999	P-coh	68	17 (25%)		Y	[IL-10] is significantly higher in sepsis and MOF after 6 d
TNF-α
Jastrow et al[32]	2009	P-coh	48	11 (23%)		Y	[TNF-α] is significantly higher in MOF from 2 – 6 + 10 – 24 h
Lendemans et al[13]	2004	P-coh	16	9 (56%)		Y	[TNF-α] is significantly higher in MOF on days 7 + 8 + 10 + 11
Menges et al[50]	1999	P-coh	68	17 (25%)		Y	[TNF-α] is significantly higher in sepsis and MOF after 8 d
Svoboda et al[62]	1993	P-cc	42	14 (33%)		Y	[TNF-α] is higher in MOF, but only after onset of symptoms
IL-1(β)
Menges et al[50]	1999	P-coh	68	17 (25%)		Y	[IL-1] is significantly higher in sepsis and MOF on days 3 + 5 + 6 + 9 - 13
Svoboda et al[62]	1994	P-xx	42	14 (33%)		N	[IL-1β] is not related to MOF
IL-2
Svoboda et al[62]	1994	P-cc	42	14 (33%)		N	[IL-2] is not related to MOF
IP-10
Jastrow et al[32]	2009	P-coh	48	11 (23%)	> 889.9 pg/mL has a sensitivity of 71% and PPV of 100%	Y	[IP-10] is highly predictive for MOF (AUC ROC 0.939)
Eotaxin
Jastrow et al[32]	2009	P-coh	48	11 (23%)	> 193.8 pg/mL has a sensitivity of 71% and PPV of 62%	Y	[Eotaxin] is highly predictive for MOF (AUC ROC 0.810)
MIP-1β
Jastrow et al[32]	2009	P-coh	48	11 (23%)	> 248.6 pg/mL has a sensitivity of 71% and PPV of 77%	Y	[MIP-1β] is highly predictive for MOF (AUC ROC 0.871)
IL-11
Schinkel et al[61]	2005	P-cc	216	9 (4%)		N	[IL-11[ is not significantly different in MOF

P-coh: Prospective cohort study; P-cc: Prospective case-control study; R-cc: Retrospective case-control study; IL: Interleukin; TNF: Tumor necrosis factor; IFN: Interferon; ROC: Receiver operating characteristic; AUC: Area under curve; r: Correlation coefficient between cytokine and development of MOF; PPV: Positive predictive value; MOF: Multi-organ failure; Pts: Patients; Y: Yes; N: No.

Table 6 Value of cytokine concentrations for predicting mortality

Ref.	Design	No pts.	Mortalityn (%)	Follow-up	Diagnostic tests	Predicts mortality	Results
IL-6
Bogner et al[36]	P-coh	58	11 (19%)	90 d		Y	[IL-6] is significantly higher in non-survivors at 0 + 6 h
Cuschieri et al[34]	P-coh	152	4 (5%)	In-hospital		N	[IL-6] is not significantly higher in non-survivors
Dresing et al[26]	P-coh	30	6 (19%)	29 d		Y	[IL-6] is significantly higher in non-survivors on days 3 + 5
Frink et al[3]	P-coh	143	21 (15%)	In-hospital	> 2176.0 pg/mL: Sensitivity 28.6%, specificity 100% on day 1	Y	[IL-6] is highly predictive for non-survival (AUC ROC 0.858)
Frangen et al[59]	P-cc	71	16 (22%)	In-hospital		Y	[IL-6] is significantly higher in non-survivors
Gebhard et al[40]	P-coh	94	18 (19%)	In-hospital		Y	[IL-6] is significantly higher in non-survivors at 4 + 6 + 12 h post injury
Maier et al[27]	P-coh	251	29 (12%)	In-hospital	AUC ROC 0.60	N	[IL-6] is not predictive for non-survival
Sousa et al[51]	P-coh	99	28 (28%)	72 h	> 276 pg/mL: AUC ROC2 0,775 (95%CI: 0.591-0.960)	Y	[IL-6] > 276 pg/mL is significantly correlated with non-survival
Svoboda et al[62]	P-cc	42	11 (26%)	In-hospital	> 400 pg/mL has a sensitivity of 100%	Y	[IL-6] > 400 pg/mL is significantly correlated with non-survival
Yagmur et al[63]	P-cc	99	17 (17%)	60 d		Y	[IL-6] is significantly elevated in non-survivors
IL-8
Bogner et al[36]	P-coh	58	11 (19%)	90 d		Y	[IL-8] is significantly higher in non-survivors at 6 + 24 h
Liener et al[43]	P-coh	94	18 (19%)	15 d		Y	[IL-8] is significantly higher in non-survivors from 30 min-24 h
Maier et al[27]	P-coh	251	29 (12%)	In-hospital	AUC ROC 0.45	N	[IL-8] is not predictive for non-survival
Yagmur et al[63]	P-cc	99	17 (17%)	60 d		Y	[IL-8] is significantly elevated in non-survivors
IL-10
Bogner et al[36]	P-coh	58	11 (19%)	90 d		Y	[IL-10] is significantly higher in non-survivors at 72 h post injury
Gouel-Chéron et al[53]	P-cc	100	5 (5%)	14 d		Y	[IL-10] is significantly higher in non-survivors when detectable on days 1 + 2
Heizmann et al[52]	R-cc	195	37 (19%)	42 d		N	[IL-10] tends towards lower levels in non-survivors; not significant
Maier et al[27]	P-coh	251	29 (12%)	In-hospital	AUC ROC 0.51	N	[IL-10] is not predictive for non-survival
Neidhardt et al[54]	P-cc	417	92 (22%)	21 d		Y	[IL-10] is significantly increased in non-survivors on days 1 + 3
Sherry et al[14]	R-cc	66	1 (2%)	50 d		N	[IL-10] is not related to non-survival
Sousa et al[51]	P-coh	99	28 (28%)	72 h	> 8.24 pg/mL: AUC ROC 0.871 (95%CI: 0.715-1.000)	Y	[IL-10] > 8.24 pg/mL is associated with non-survival at 48 + 72 h post injury
TNF-α
Dresing et al[26]	P-coh	30	6 (19%)	29 d		N	[TNF-α] is not significantly elevated in non-survivors
Sousa et al[51]	P-coh	99	28 (28%)	72 h		N	[TNF-α] is not significantly elevated in non-survivors
Spielmann et al[57]	P-cc	47	11 (23%)	6 d		N	[TNF-α] is not significantly elevated in non-survivors
Svoboda et al[62]	P-cc	42	11 (26%)	In-hospital		Y	[TNF-α] is significantly elevated in non-survivors
Yagmur et al[63]	P-cc	99	17 (17%)	60 d		N	[TNF-α] is not significantly elevated in non-survivors
IL-18
Heizmann et al[52]	R-cc	195	37 (19%)	42 d		N	[IL-18] tends towards lower levels in non-survivors; not significant
Mommsen et al[30]	P-coh	55	7 (13%)	14 d		Y	[IL-18] is significantly increased in non-survivors on days 2-7
Roetman et al[60]	P-cc	229	36 (16%)	30 d		N	[IL-18] median value is significantly lower in non-survivors
IL-2
Heizmann et al[52]	R-cc	195	37 (19%)	42 d		N	[IL-2] tends towards lower levels in non-survivors; not significant
Svoboda et al[62]	P-cc	42	11 (26%)	In-hospital		N	[IL-2] is not related to non-survival
Yagmur et al[63]	P-cc	99	17 (17%)	60 d		Y	[IL-2] is significantly increased in non-survivors
IL-1
Svoboda et al[62]	P-cc	42	11 (26%)	In-hospital		N	[IL-1] is not related to non-survival
Yagmur et al[63]	P-cc	99	17 (17%)	60 d		N	[IL-1] is not related to non-survival
IL-12
Heizmann et al[52]	R-cc	195	37 (19%)	42 d		N	[IL-12] tends towards lower levels in non-survivors; not significant
Wick et al[49]	P-coh	37	6 (16%)	In-hospital		Y	[IL-12] is significantly lower in non-survivors
IL-11
Schinkel et al[61]	P-cc	216	34 (16%)	In-hospital		N	[IL-11] is lower in non-survivors, only reaching significance after week 4
Heizmann et al[52]	R-cc	195	37 (19%)	42 d		N	[IL-11] tends towards lower levels in non-survivors; not significant
IL-17
Frangen et al[59]	P-cc	71	16 (22%)	In-hospital		N	[IL-17] is not related to non-survival
IL-4
Heizmann et al[52]	R-cc	195	37 (19%)	42 d		N	[IL-4] tends towards lower levels in non-survivors; not significant
Roetman et al[60]	P-cc	229	36 (16%)	30 d		N	[IL-4] is not related to mortality
IFN-γ
Heizmann et al[52]	R-cc	195	37 (19%)	42 d		N	[IFN-γ] tends towards lower levels in non-survivors; not significant
Roetman et al[60]	P-cc	229	36 (16%)	30 d		N	[IFN-γ] inconsistently detectable

P-coh: Prospective cohort study; P-cc: Prospective case-control study; R-cc: Retrospective case-control study; IL: Interleukin; TNF: Tumor necrosis factor; IFN: Interferon; AUC: Area under the receiver operating characteristic (ROC) curve; Pts: Patients; Y: Yes; N: No.