We retrieved 800 potentially relevant documents from PubMed (n = 81), Embase (n = 244), Web of Science (n = 324), Cochrane Library (n = 88), or manual search (n = 3). After combining and deleting duplicate documents, 609 studies were retained. According to the title and abstract, preliminary screening was conducted to further exclude 549 articles. Subsequently, we obtained the full text of the remaining 60 articles and screened them by reading. Four of these studies were excluded because they did not provide access to the full text; 9 studies were not relevant to our theme. The inclusion and exclusion criteria of the research groups of 22 studies did not meet the conditions we defined, 5 studies did not establish a control group, 5 studies did not report on TRALI-related risk factors, and 2 studies were conference summaries. Finally, we included 13 articles in the meta-analysis. The details of the screening are shown in Figure 1.
The 13 articles included in this study included a total of 11,380,242 participants: 3179 cases and 11377063 controls. These studies were published from 2004 to 2015. The research participants originated from four countries: the United States, the Netherlands, Croatia, and Italy. The sample size varied from 26 to 11,378,264. The 13 studies were all case-control studies. Three articles were multi-center studies. We used the NOS to assess the quality of each study. Three studies scored seven points, eight studies scored eight points, and two studies scored nine points. Quality assessments indicated that all included studies had high quality. Table 1 and Supplementary Table E2 show the basic characteristics of all included studies and the NOS-based scoring results.
Table 1
Characteristics of all the included studies
Author
|
Country
|
Year
|
Intervention
|
Study period
|
Study type
|
Number of patients (TRALI / transfused controls)
|
Male
|
Age
|
RBC units
|
FFP units
|
PLT units
|
Identified risk factors
|
NOS score
|
Toy et al
|
America
|
2015
|
TRALI patients versus transfused controls
|
2006-2009
|
case-control study
|
308(145/163)
|
75(52)/90(55)
|
58±19/56±20
|
-
|
-
|
-
|
chronic alcohol abuse, current smoker, shock before transfusion, positive fluid balance before transfusion
|
9
|
Mulder et al
|
the Netherlands
|
2014
|
2009-2012
|
304(21/283)
|
13(61.9)/172(60.8)
|
4.6±5.2/5±5.8
|
0.8±0.7/1.3±3.5
|
0.5±0.6/0.5±1.6
|
0.3±0.5/0.5±6
|
mechanical ventilation, sepsis, high Pediatric Risk of Mortality III score
|
8
|
Menis et al
|
America
|
2014
|
2007-2011
|
11378264(2556/11375708)
|
1125(44)/4891555(43)
|
-
|
-
|
-
|
-
|
platelet (PLT)- and plasma-containing transfusions, persons ages 65 to 79 years, females, white people, with 6-month histories of post inflammatory pulmonary fibrosis, tobacco use
|
8
|
Zah-Bogović et al
|
Croatia
|
2014
|
2009-2010
|
252(32/220)
|
24(75)/138(62.7)
|
62.6±10.85/66.3±10.28
|
4(2.5-6)/2(1-3)
|
5(0-6)/0(0-2)
|
0(0-8)/0(0-0)
|
total number of transfused blood units, duration of cardiopulmonary bypass
|
7
|
Teofili et al
|
Italy
|
2014
|
2005-2011
|
71(14/57)
|
|
34.4±5.8/34.4±4.8
|
9(3.75-13.25)/5(3-6.5)
|
8(2.25-15.5)/4(0-8)
|
0(0-11)/0(0-5)
|
Pregnancy-related hypertensive disorders
|
8
|
Toy et al
|
America
|
2012
|
2006-2009
|
253(89/164)
|
44(49)/91(56)
|
54±20/56±20.2
|
2(0-25)/2(0-12)
|
2(0-40)/0(0-21)
|
0(0-7)/0(0-3)
|
higher IL-8 levels, liver surgery, chronic alcohol abuse, shock, higher peak airway pressure while being mechanically ventilated, current smoking, positive fluid balance, receipt of plasma or whole blood from female donors, volume of HLA class II antibody with normalized background ratio more than 27.5, volume of anti–human neutrophil antigen positive by granulocyte immunofluoresence test
|
9
|
Vlaar et al
|
the Netherlands
|
2011
|
2006-2009
|
48(16/32)
|
12(75)/20(63)
|
74±7.4/67±9.3
|
3(2-5)/2(1-2)
|
2(0-4)/0(0-2)
|
1(0-2)/0(0-1)
|
Age, time on the cardiopulmonary bypass, total amount of blood products, number of red blood cells stored more than 14 days, total amount of plasma, presence of antibodies in donor plasma, total amount of transfused bioactive lipids
|
8
|
Vlaar et al
|
the Netherlands
|
2010
|
2004-2007
|
218(109/109)
|
70(64)/66(61)
|
59±17/57±16
|
1.8±0.3/1.9±0.2
|
1.3±0.2/0.9±0.2
|
0.4±0.1/0.2±0.1
|
emergency cardiac surgery, hematologic malignancy, massive transfusion, sepsis, mechanical ventilation, and high Acute Physiology and Chronic Health Evaluation II score
|
8
|
Edens et al
|
America
|
2010
|
2008-2009
|
66(22/44)
|
-
|
28.3±7.5/28.2±8.6
|
22.3±16.2/13.2±6.9
|
21.6±14.6/11.7±6.6
|
4.7±11.2/1.1±1.5
|
pulmonary injury, volume of FFP transfused
|
7
|
Sanchez et al
|
America
|
2007
|
2002-2004
|
26(6/20)
|
3(50)/14(70)
|
-
|
3/16 people
|
6/14 people
|
2/11 people
|
Whole blood units, Leukocyte-reduced units
|
8
|
Gajic et al
|
America
|
2007
|
2 years
|
148(74/74)
|
37(50)/37(50)
|
64(52-78)/61(53-73)
|
1(0-2)/2(0-2)
|
1(0-4)/0(0-1)
|
0(0-0)/0(0-0)
|
transfusion of plasma from female donors, number of pregnancies among the donors, number of donor units positive for anti-granulocyte antibodies, anti–HLA class II antibodies, andconcentration of lysophosphatidylcholinein the donor product
|
7
|
Silverboard et al
|
America
|
2005
|
2000-2002
|
102(35/67)
|
22(63)/56(84)
|
38.7±14.4/31.2±16.7
|
15±17/8±13
|
-
|
-
|
the amount of transfused blood
|
8
|
Gajic et al
|
America
|
2004
|
2001-2002
|
182(60/121)
|
33(55)/71(59)
|
65(50-75)/68(53-77)
|
58/115 people
|
43/60 people
|
32/43 people
|
Thrombocytopenia, transfusion of fresh frozen plasma
|
8
|
TRALI, transfusion-related acute lung injury; RBC, red blood cell; FFP, fresh frozen plasma; PLT, blood platelet; NOS, Newcastle-Ottawa Scale.
|
Risk factors
The current meta-analysis summarizes 20 potential risk factors from four aspects, which have been reported in at least two studies: host-related, blood transfusion-related, device-related, and surgery-related factors (Supplementary Table E3). Figure 2 shows the forest plots of potential risk factors predisposing patients to TRALI.
Host-related factors
Age. Two studies, including 11378312 patients, showed that the age of blood transfusion patients was significantly related to the occurrence of TRALI. The greater the age, the greater the risk. There was no obvious heterogeneity in the studies (fixed effect model, i2 = 15.9%; OR = 1.16, 95% CI = 1.08–1.24, P < 0.01; Figure 2A).
Sex. Two studies including 11378290 patients indicated a significant correlation between sex and the occurrence of TRALI. TRALI is more likely to occur in women than in men undergoing blood transfusion treatment. There was no obvious heterogeneity in the studies (fixed effect model, i2 = 0.0%; OR = 1.26, 95% CI = 1.16–1.38, P < 0.01; Figure 2B).
Tobacco use. Three studies including 11378825 patients indicated a significant correlation between tobacco use and the occurrence of TRALI. There was no obvious heterogeneity in the studies (fixed effect model, I2 = 0.0%; OR = 3.82, 95% CI = 1.91–7.65, P < 0.01; Figure 2C).
Chronic alcohol abuse. Two studies including 561 patients indicated a significant correlation between chronic alcohol abuse and the occurrence of TRALI. There was no obvious heterogeneity in the studies (fixed effect model, I2 = 0.0%; OR = 3.82, 95% CI = 2.97–26.83, P < 0.01; Figure 2D).
Tidal volume. When the two studies that reported tidal volume were aggregated, tidal volume was not associated with the occurrence of TRALI. There was some heterogeneity in the studies (random effect model, I2 = 52.2%; OR = 1.37, 95% CI = 0.89–2.11, P = 0.155 > 0.05; Figure 2E).
Fluid balance. Two studies including 561 patients indicated a significant correlation between fluid balance and the occurrence of TRALI. There was a high degree of heterogeneity in the studies (random effect model, I2 = 71.3%; OR = 1.24, 95% CI = 1.08–1.42, P < 0.01; Figure 2F).
Liver disease. Two studies including 505 patients indicated that liver disease had no significant effect on the occurrence and development of TRALI. There was a high degree of heterogeneity in the studies (random effect model, I2 = 85.5%; OR = 1.51, 95% CI = 0.48–4.78, P = 0.482 > 0.05; Figure 2G).
Sepsis. Three studies including 11378786 patients indicated that sepsis had no significant effect on the occurrence and development of TRALI. There was a high degree of heterogeneity in the studies (random effect model, I2 = 93.9%; OR = 2.08, 95% CI = 0.46–9.28, P = 0.339 > 0.05; Figure 2H).
Hematology-oncology. Three studies including 11378543 patients indicated that hematology-oncology had no significant effect on the occurrence and development of TRALI. There was a high degree of heterogeneity in the studies (random effect model, I2 = 87.9%; OR = 1.40, 95% CI = 0.25–7.80, P = 0.701 > 0.05; Figure 2I).
Shock before transfusion. Two studies including 561 patients indicated that patients with shock before blood transfusion were more likely to experience TRALI after blood transfusion. There was no obvious heterogeneity in the studies (fixed effect model, I2 = 0.0%; OR = 4.41, 95% CI = 2.38–8.20, P < 0.01; Figure 2J).
ASA score. Two studies including 300 patients indicated that transfusion patients with a high ASA score were likely to develop TRALI. There was no obvious heterogeneity in the studies (fixed effect model, I2 = 0.0%; OR = 2.72, 95% CI = 1.43–5.16, P < 0.01; Figure 2K).
Transfusion-related factors
Amount of transfusions. Seven studies including 976 patients indicated a significant correlation between large-scale blood transfusion and the occurrence of TRALI. There was a high degree of heterogeneity in the studies (random effect model, I2 = 77.4%; OR = 1.40, 95% CI = 1.14–1.72, P < 0.01; Figure 2L).
Blood products from female donors. Four studies including 553 patients indicated that blood products from female donors had no significant effect on the occurrence and development of TRALI. There was no obvious heterogeneity in the studies (fixed effect model, I2 = 20.0%; OR = 1.09, 95% CI = 0.94–1.27, P = 0.319 > 0.05; Figure 2M).
RBC units. Five studies including 519 patients indicated that the number of RBC units had no significant effect on the occurrence and development of TRALI. There was some heterogeneity in the studies (random effect model, I2 = 58.8%; OR = 0.97, 95% CI = 0.72–1.31, P = 0.859 > 0.05; Figure 2N).
PLT units. Five studies including 11378475 patients indicated that the number of PLT units had no significant effect on the occurrence and development of TRALI. There was some heterogeneity in the studies (random effect model, I2 = 69.6%; OR = 1.38, 95% CI = 0.27–1.26, P = 0.160 > 0.05; Figure 2O).
FFP units. Five studies including 393 patients indicated a significant correlation between the number of FFP units and the occurrence of TRALI. There was some heterogeneity in the studies (random effect model, I2 = 71.1%; OR = 1.21, 95% CI = 1.01–1.46, P = 0.043 < 0.05; Figure 2P).
Device-related factors
Mechanical ventilation. Three studies including 775 patients indicated that patients on mechanical ventilation during blood transfusion therapy were more likely to develop TRALI. There was no obvious heterogeneity in the studies (fixed effect model, I2 = 0.0%; OR = 4.13, 95% CI = 2.20–7.76, P < 0.01; Figure 2Q).
Pump time during cardiopulmonary bypass. Two studies including 300 patients indicated that pump time during cardiopulmonary bypass had no significant effect on the occurrence and development of TRALI. There was a high degree of heterogeneity in the studies (random effect model, I2 = 95.6%; OR = 1.04, 95% CI = 0.96–1.12, P = 0.329 > 0.05; Figure 2R).
Procedure-related factors
Surgical intervention. Two studies including 97 patients indicated that surgical intervention had no significant effect on the occurrence and development of TRALI. There was no obvious heterogeneity in the studies (fixed effect model, I2 = 0.0%; OR = 1.21, 95% CI = 0.17–8.63, P = 0.848 > 0.05; Figure 2S).
Liver transplantation. Three studies including 587 patients indicated no significant correlation between liver transplantation and the occurrence of TRALI. There was no obvious heterogeneity in the studies (fixed effect model, I2 = 15.1%; OR = 2.95, 95% CI = 0.80–10.84, P = 0.188 > 0.05; Figure 2T).
Publication Bias and sensitivity analysis
We used the Begg funnel plot and Egger linear regression test to evaluate the publication bias in these studies. Because each factor was generally included in a few studies, we only conducted a bias test for risk factors that were included in more than three studies. The assessment of publication bias for potential risk factors of TRALI is shown in Supplementary Table E4. Among the factors that had a significant correlation with the occurrence of TRALI, the number of transfusions had some publication bias, and the number of FFP units did not. Among the factors that were not significantly related to TRALI, there was no publication bias for blood products from female donors and number of RBC units and PLT units (Figure 3). Because the number of studies included was small, we did not conduct a sensitivity analysis.