Meta-Analysis of Preclinical Studies of Fibrinolytic Therapy for Acute Lung Injury

Background Acute lung injury (ALI) is characterized by suppressed fibrinolytic activity in bronchoalveolar lavage fluid (BALF) attributed to elevated plasminogen activator inhibitor-1 (PAI-1). Restoring pulmonary fibrinolysis by delivering tissue-type plasminogen activator (tPA), urokinase plasminogen activator (uPA), and plasmin could be a promising approach. Objectives To systematically analyze the overall benefit of fibrinolytic therapy for ALI reported in preclinical studies. Methods We searched PubMed, Embase, Web of Science, and CNKI Chinese databases, and analyzed data retrieved from 22 studies for the beneficial effects of fibrinolytics on animal models of ALI. Results Both large and small animals were used with five routes for delivering tPA, uPA, and plasmin. Fibrinolytics significantly increased the fibrinolytic activity both in the plasma and BALF. Fibrin degradation products in BALF had a net increase of 408.41 ng/ml vs controls (P < 0.00001). In addition, plasma thrombin–antithrombin complexes increased 1.59 ng/ml over controls (P = 0.0001). In sharp contrast, PAI-1 level in BALF decreased 21.44 ng/ml compared with controls (P < 0.00001). Arterial oxygen tension was improved by a net increase of 15.16 mmHg, while carbon dioxide pressure was significantly reduced (11.66 mmHg, P = 0.0001 vs controls). Additionally, fibrinolytics improved lung function and alleviated inflammation response: the lung wet/dry ratio was decreased 1.49 (P < 0.0001 vs controls), lung injury score was reduced 1.83 (P < 0.00001 vs controls), and BALF neutrophils were lesser (3 × 104/ml, P < 0.00001 vs controls). The mortality decreased significantly within defined study periods (6 h to 30 days for mortality), as the risk ratio of death was 0.2-fold of controls (P = 0.0008). Conclusion We conclude that fibrinolytic therapy may be effective pharmaceutic strategy for ALI in animal models.

iNtRODUctiON Suppressed fibrinolysis is a pathological hallmark of acute lung injury (ALI) in addition to pulmonary edema and cytokine/ chemokine "storm" (1,2). Over the last two decades, the mortality of acute respiratory distress syndrome (ARDS), the late stage of ALI remains unacceptably high. There are approximately 200,000 new cases annually in the United States (3,4). ALI could be caused by pulmonary (e.g., pneumonia and smoke inhalation) or systemic disorders (e.g., sepsis, hemorrhagic shock, and trauma) (5). The heterogeneity of ALI apparently makes it difficult to identify the etiology precisely and promptly for designing casespecific salutary interventions. To date, supportive strategies have been shown to be beneficial (5). In addition, some promising therapeutic strategies are being evaluated by registered clinical trials, including stem cell therapy (6,7), corticosteroid, interferon beta, and tumor necrosis factor-alpha (8).
In injured lungs, alveolar fibrinolytic activity is depressed markedly and intravascular and extracellular fibrin is deposited in the air spaces (9,10). The eliminated fibrinolytic activity was predominately attributed to elevated plasminogen activator inhibitor-1 (PAI-1) in both the plasma and bronchoalveolar lavage fluid (BALF). The fibrinolytic system is composed of proteases and anti-proteases, including plasminogen, plasminogen activators (tissue-type plasminogen activator, tPA; urokinase plasminogen activator, uPA), plasmin, PAI-1, and plasmin catalytic antagonists (α2-antiplasmin and α2-macroglobulin). Both uPA and tPA proteolytically cleave zymogen plasminogen to generate plasmin with catalytic activity, which degrades fibrin. To date, the fibrinolytic therapy has clinically been applied to pleural effusion/empyema as fibrinolysins (11), cardiovascular diseases as thrombolytics (12), and obstructive airway diseases as mucolytics (13,14). The benefit of fibrinolytic therapy for ALI, however, is still at the earlier stage of preclinical studies. Intravenous delivery of either uPA or tPA might be protective for traumatic lung injury, as improved survival and gas exchange were observed in treated pigs (15). Moreover, tPA attenuated pulmonary abnormalities in smoke inhalation injured sheep (16). Interestingly, an earlier pilot study reported a potential improvement in lung function following administration of either uPA or tPA in 20 ALI patients (17). Because of inconclusive preclinical studies to address optimized dose, routes, and benefit, clinical trials have not been conducted to date.
The main purpose of this meta-analysis is, therefore, to assess preclinical studies of ALI for the potential effects of fibrinolytic therapy. Additionally, we quantified the differences in outcomes between large and small animal models, variable fibrinolytic regimes, and routes. Our analysis suggests that three fibrinolytic regimens may benefit ALI by improving gas exchange, inflammation, and lung injury.

mateRiaLS aND metHODS
The study was conducted in accordance with the methods recommended in the PRISMA guidelines. Please see Datasheet S1 in Supplementary Material for detail search strategies.

Data Sources
Three investigators (CL, ZS, and YM) independently searched the published studies indexed by the PubMed, Embase, Web of Science, and CNKI on March 2017, using the strategy: (fibrinolytics OR plasmin OR fibrinase OR fibrinolysin OR alfimeprase OR uPA OR urokinase OR abbokinase OR breokinase OR winkase OR kinlytic OR tPA OR activase OR reteplase OR rctPA OR retavase OR actilyse OR repilysin OR alteplase OR tenecteplase OR TNKase OR TNK tPA OR metalyse OR SK OR streptokinase OR streptase OR kabikinase OR actase OR thrombolysin OR eminase OR desmoteplase OR pro-urokinase OR staphylokinase OR plasminogen activator) AND (lung injury OR ALI OR ARDS OR respiratory distress syndrome OR septic OR sepsis OR bacteremia OR endotoximia OR multi-organ failure OR respiratory failure OR pneumonia OR shock OR pulmonary edema OR lung edema OR edematous OR pulmonary edema). The fourth investigator (H-LJ) was consulted in case of no consensus on inclusion. There was no language restriction in the searching and non-English literature was translated into English via a professional service. We also checked the references of included studies for additional publications that were not hit by the searching strategy.

inclusion and exclusion criteria
Studies were included in the current meta-analysis if: (1) the species of studies were animals; (2) animal models were ALI, including Pseudomonas aeruginosa pneumonia, Klebsiella pneumonia, fibrotic lung injury, trauma, septic shock, burn and smoke inhalation injury, and disseminated intravascular coagulation; (3) the animals were treated with three fibrinolytics, genetically engineered for overexpressing of plasminogen activators or knocking out PAI-1; and (4) results were expressed or could be digitized or converted to mean and SD.
The following studies were excluded: (1) review articles, letters, case reports, posters, or without objective data to be evaluated; (2) data were from PAI-1 transgenic, PAI-1 knock-in, and tPA-and uPA-deficient animals; (3) insufficient publications existed to perform a meta-analysis; (4) the number of control group or experimental group was less than three animals; (5) the animals were treated with combined heparin, antithrombin, activated protein C, and other medicines with potential effects on fibrinolysis; and (6) combined regimens and fibrinolytic therapy for lung injury.

Data extraction
Data extraction was carried out by three authors (CL, ZS, and YM). The following items from the eligible studies were extracted: article information (first author name, publication date, and country of origin), animal species, gender, method of ALI induction, type and dose of fibrinolytics, duration of treatment, delivery approach, and number of animals for both control and treated groups. Data were extracted as mean and SD if available. When only graphic presentations were available, values for mean and SD were obtained via calibrating images using GetData Graph Digitizer software (version 2.26.0.20) (18)(19)(20). If raw data were represented by median and interquartile Fibrinolytics Improve ALI Frontiers in Immunology | www.frontiersin.org August 2018 | Volume 9 | Article 1898 range (20,21) or range (22), these values were converted to SD by the formulas: 2 ], and mean = (a + 2 m + b)/4, where m represents median, a and b are lower and upper range (23). If the SE was reported (20), it was converted to SD using the function, SD SE = ×√n, where n is the sample size. For one study (24), we combined individual data using the formulas, X= , where x represent variance, X is the mean of the pooled individual data. If neither SD nor SEM were found (15,25), we borrowed SD value of similar studies for the same parameter. For parameters reported with divergent units, for example, PaO2, PaCO2, and neutrophils, the units were converted to the same one. For lung water content that shown as weight gain or lung leak index, we converted them to W/D ratio following the formulas: W/D ratio = (W0 + W1)/ D0 = lung leak index × 100, where W0 represents normal animal lung wet weight, W1 represents weigh gain, and D0 represents normal animal lung dry weight. Lung injury scores were analyzed as occurrence of at least one of following indices: (1) leukocyte and red blood cell infiltration, (2) alveolar epithelium damage,

Quality assessment
Risk of biases was assessed using the Cochrane Handbook for Systematic Reviews of Interventions (26). All included studies were assessed on seven fronts: randomization (selection bias), blinding of personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), allocation concealment (selection bias), selective outcome reporting (reporting bias), and other biases. These assessments are included in Table 6. In addition, we examined the quality of the included 22 studies with the ARRIVE guidelines specifically for animal studies ( Table 2).

Statistical analysis
We performed statistical analysis in the preclinical studies using Review Manager (RevMan), version 5.3 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014) and STATA V.12 (StataCorp. College Station, TX, USA). The mean differences were considered to be statistically significant when P ≤ 0.05. If P-value was less than 0.05, or the I 2 -value was greater than 50%, the overall estimate was analyzed in a random effects model. Groups within a parameter were included if there were ≥3 comparisons. Eleven readouts were selected: PaO2, PaCO2, plasma thrombin-antithrombin complexes (TATc), plasma PAA, fibrin degradation products (FDP) in BALF, PAA in BALF, PAI-1 in BALF, neutrophils in BALF, lung water content, lung injury score, and mortality. The data were combined using inverse variance method and shown as weighted mean differences (WMD) with 95% confidence intervals (95% CI) in the forest plots in addition to mortality. The measures for the parameters (PaO2 and PaCO2) at the same time points were continuously monitored. If unavailable, we used the data collected at an adjacent time point. We pooled dichotomous variables (i.e., mortality) using the Mantel-Haenszel method. If the mortality was measured within a period, the end time point was used.
The RR was computed with the random effects model for a high heterogeneity. The potential publication bias was assessed with funnel plots and the Egger's regression test (40) (Stata, version 12). Heterogeneity among studies was defined with the I 2 -statistic function as an unimportant (I 2 < 25%), a moderate (25% < I 2 < 75%), or a high degree of heterogeneity (I 2 > 75%).
To eliminate heterogeneity, the meta-analyses were further performed for data grouped by animal size (small or large animals), individual fibrinolytics (tPA, uPA, or plasmin), and routes (i.v., i.p., i.t., nebulization, or transgenic). Small animals included mice, rats, and rabbits. Large animals were comprised of sheep, pigs, and dogs. Subgroup analysis was not performed when there was only one sample. In addition, the robustness of the results was confirmed by sensitivity analysis. If multiple studies with the same first author and publication year, they were distinguished with an asterisk (*) (19,20). The same control group was used for several comparisons and denoted with superscript letters a, b, c, and d (15,18,29,32).

mortality within Defined Follow-Up Periods
The mortality associated with fibrinolytic therapy was evaluated (Figure 2). Compared with controls, fibrinolytics significantly reduced the deaths of treated animals, as shown by the overall risk ratio (RR) (0.21, 95% CI: 0.08 to 0.52, P = 0.0008). Furthermore, we analyzed mortality of small and large animals separately ( Table 4) Table 3.
Overall arterial carbon dioxide pressure (PaCO2), another physiological parameter for gas exchange capacity of the lung, showed a significant reduction of 12 mmHg (95% CI: −18 to −5 mmHg, P = 0.0001) in treated animals ( Figure 3B). Moreover, five of seven tPA studies (42 controls, 36 treated animals) favored fibrinolytic therapy with a decrease of 21 mmHg (95% CI: −34 to −8 mmHg, P = 0.002) in PaCO2, whereas the other two studies did not. In contrast, uPA did not significantly alter PaCO2 (0.3 mmHg, 95% CI: −2 to 2 mmHg, P = 0.79) in three studies ( Table 4). Regarding routes of administration, all of the five studies investigating i.t. treatment (29 controls, 23 treated animals) demonstrated a favor, as there was a 31 mmHg reduction in PaCO2 (95% CI: −45 to −18 mmHg, P < 0.00001). In contrast, i.v. (3 studies) and nebulization (2 studies) did not exhibit significant effects on PaCO2 with a change of 1 mmHg ( Table 4). Nine studies in small animals (66 controls, 60 treated animals) were evaluated for the efficacy of fibrinolytic therapy in the reduction of PaCO2. Five studies favored fibrinolytic therapy, and four studies did not. Taken together, the overall PaCO2 showed a reduction of 15 mmHg in small animals (95% CI: −23 to −7 mmHg, P = 0.0005). However, fibrinolytics did not reduce PaCO2 significantly in large animals ( Table 4). In addition, the overall benefit of fibrinolytics on PO2 and PaCO2 levels at the same time point were confirmed by analyzing data at the endpoint of each study ( Table 4, data with *).

assessment of Bias and Sensitivity analysis
To determine if potential threats to internal validity influenced our findings, we evaluated the quality of included 22 studies in addition to the checklists of the ARRIVE guidelines ( Table 2). Incomplete outcome data and selective outcome reporting in all studies were low risk. Randomization, blinding of personnel, allocation concealment, and blinding outcome assessment in 11 projects were low risk, whereas the risk in remaining projects was unclear ( Table 6). Funnel plot found no asymmetrical distribution for PaO2 (Figure 7A), which was confirmed by the Egger regression ( Figure 7B; P = 0.199 for PaO2). Without imputing any missing studies for PaO2, the Trim and Fill analysis exhibited a symmetrical funnel plot too (data not shown).
To test the stability and dependability of the results, we omitted one study (34), which had a relative large sample size. The combined WMD of PaO2 for remaining 17 studies was estimated by the sensitivity plot again, yielding a value of 15 (95% CI: 7 to 24) (Figure 8), which is same as the estimate of overall effect. These results indicated that the reliability of our meta-analysis was considerably strong.

DiScUSSiON
Fibrinolytic therapy for ALI has been emerging during the last decade (41). Comparing with systematic evaluation (including systematic review and meta-analysis) of the efficacy of anticoagulants in clinical and preclinical studies of ALI (42), the potential benefit of fibrinolytics for ALI patients has not been tested by well-designed clinical trials yet. In this analysis, we  have performed meta-analysis and systematic review, focusing on efficacy in 22 published preclinical studies of ALI. Our results suggest that fibrinolytic therapy significantly improves gas exchange, reduces alveolar neutrophils, increases fibrinolytic activity, reduces the quantity of pulmonary edema fluid, and suppresses the histologic severity of lung injury in a fibrinolysin-, route-, and species-dependent manners.
The overall effects of fibrinolytics on death toll, oxygenation, fibrinolysis, and lung function were corroborated by analyzing subgrouped data and sensitivity. To our knowledge, this is the first meta-analysis to summarize previous preclinical studies aiming to provide evidence for further animal studies and clinical trials.

Fibrinolytic therapy for aLi is Feasible and tolerant
All of three fibrinolytic reagents, tPA, uPA, and plasmin were administered for 14 animal models of ALI in the included studies. The large range of applied doses demonstrates feasibility and tolerance. In addition, tolerability of tPA was confirmed in mice by a classic study (43). Airway bleeding was observed only where treatment was given locally with a large dose of  tPA (≥1 mg/kg/d). Although the scarcity of gross hemorrhage and other severe adverse events applied via five routes were reported in 6 of 22 studies, considering the non-adherence to the ARRIVE guidelines (44) and informal systematic measurements of local and systemic hemorrhage quantitatively, the potential risk for bleeding associated with fibrinolytic therapy could not be ruled out.

mortality improved by Fibrinolytic therapy
Overall mortality was improved during the defined follow-up period from 1 h to 28 days. The most effective route to reduce mortality is intratracheal administration, and plasmin is the least effective lytic to improve survival rate. It is note worthy that only two of 10 studies followed mortality 48 h to 30 days. Other eight studies determined death toll within 2 days. Given the inconsistent follow-up time that differs from clinical studies, the overall improvement of mortality by fibrinolytic therapy shall be confirmed by further studies designed per the ARRIVE guidelines (44) and with extended follow-up periods identical to that for ALI/ARDS clinical studies.

Fibrinolytic therapy improves Gas exchange
Intriguingly, fibrinolytics increase gas exchange based on data at the same time-point, and to the most extent when delivered via intratracheal route. Fibrinolytics benefited gas exchange differently between small and large animals. It is probably due to divergent dose applied between species. Furthermore, the structure of the respiratory tract, lung function, inner surface of the lung for gas exchange, and oxygen consumption are body size dependent. Additionally, inconclusive results of oxygenation are seen for plasmin due to insufficient samples for meta-analysis. Hypoxia is correlated with the mortality of ALI (45). This notion is supported by the benefit of both gas exchange and survival rate by intratracheal delivery of fibrinolytic regimens.

Fibrinolytic therapy alleviates Lung injury
Our meta-analysis demonstrates that fibrinolytic therapy restores the dysfunctional fibrinolysis and coagulation in ALI. This benefit does not depend on routes and fibrinolytic regimens. Alveolar fibrin deposition attracted neutrophils and fibroblasts, and decreased lung edema fluid clearance (46). We identify that fibrinolysin tPA and i.v. route are the most potent for reducing neutrophil infiltration. We also find that fibrinolytics facilitate edema fluid resolution in small animals to a greater extent than in large animals. Our previous studies proved that uPA and plasmin but not tPA could activate apically located epithelial sodium channels, a predominate pathway to remove edema fluid (47,48). Considering that tPA competitively binds to PAI-1 with a greater affinity than uPA, delivered tPA may form complexes with elevated PAI-1 and serve as a "cage" to separate endogenous uPA from PAI-1 (49). Under this situation, freed uPA is able to activate epithelial sodium channels and resolve edema fluid. It is unclear why intraperitoneal administration eliminates lung edema to the most extent. At least, this is the only way to get rid of direct addition of extra fluid to the flooded air spaces intratracheally or indirectly leaking to the alveoli through the impaired blood-gas barrier post intravenously fluid infusion. In addition, we find that plasmin may be the best fibrinolysin to improve lung injury score. Plasmin may defragment deposited fibrin. However, this is not supported by FDP content in BALF. It is probably due to the cleavage of epithelial sodium channels by plasmin proteolytically to expedite transalveolar fluid re-absorption (unpublished data). Exogenously applied tPA displayed anti-inflammatory effects (50). Therefore, anti-inflammatory properties of three fibrinolytic agents could be a mechanism for alleviating ALI. Taken together, fibrinolytic regimens may suppress the mortality in preclinical models of ALI through multifaceted mechanisms in a route-and model-dependent manner. ALI animal models may not be representative of human ALI, because of the timing and severity of ALI induction, the dose and timing of the treatment in relation to ALI induction, the use of small/young animals without comorbid illnesses, and lack of administration of standard of care co-interventions, such as fluids and antibiotics during the study period. How well animal models of ALI mimic the pathophysiology of human ALI has also been a contentious issue. Thus, the effect of construct validity on fibrinolytic therapy of ALI remains to be determined.

Limitations of Our analysis
Our review has some limitations. First, the duration of delivery of fibrinolytic agents and the period of observation were not consistent. More than half of the included studies do not adhere to the ARRIVE guidelines. Second, although subgroups and sensitivity analyses were conducted, we could not completely explain the substantial heterogeneity and some diverse effects in experimental ALI. Third, uncertainty of the potential benefits of fibrinolytics in ALI may be improved by future preclinical studies with larger numbers of animal studies. For example, we could not perform dose-effect analysis because of a paucity of sufficient data. Finally, the effects of fibrinolytics on ALI animal models provide basic information for further preclinical studies. Considering the difference between relative homogeneous animals (e.g., healthy young or adult animals without comorbidity and extremely heterogeneous patients), it shall be cautious to link the findings in preclinical studies to ALI patients.

cONcLUSiON
Our results identify that the best route is intratracheal delivery, and the most efficacious regimen is tPA. FDA approved fibrinolytic therapies for cardiovascular and pleural diseases, including hypertensive intraventricular hemorrhage (51), loculated pleural effusions, parapneumonic effusions, pleural empyema, malignant effusions, hemothorax, and myocardial infarction (52)(53)(54)(55), demonstrate the tolerance and feasibility of this treatment. This meta-analysis, however, did not provide data on the optimized dose of each therapy, the optimal time