Mortality and bleeding associated with the management of sub-massive pulmonary embolism: a systematic review and Bayesian network meta-analysis

Current guidelines recommend anticoagulation (AC) for low and intermediate-risk pulmonary embolism (PE) and systemic thrombolysis (tPA) for high risk (massive) PE. How these treatment options compare with other modalities of treatment such as catheter directed thrombolysis (CDT), ultrasound assisted catheter thrombolysis (USAT), and administering lower dose of thrombolytics (LDT) is unclear. There is no study that has compared all these treatment options. We conducted a systematic review and Bayesian network meta-analysis of randomized controlled trials in patients with submassive (intermediate risk) PE. Fourteen randomized controlled trials were included, comprising 2132 patients. On Bayesian network meta-analysis, a significant decrease in mortality was noted in tPA versus AC. There was no significant difference between USAT versus CDT. For risk of major bleeding, there was no significant difference in relative risk of major bleeding between tPA versus AC and USAT versus CDT. tPA was found to have a significantly higher risk of minor bleeding and a lower risk of recurrent PE compared to AC. Systemic thrombolysis is associated with a significant reduction in mortality and recurrent PE compared to anticoagulation but an increased risk of minor bleeding. There was no difference in risk of major bleeding. Our study also shows that while the newer modalities of treatment for pulmonary embolism are promising, there is lack of data to comment on the purported advantages.

Venous thromboembolism (VTE) is the third most frequent acute cardiovascular syndrome globally behind myocardial infarction and stroke. The annual incidence of pulmonary embolism (PE) is between 39 and 115 per 100,000 population [1][2][3][4] . The current guidelines recommend anticoagulation for low and intermediate risk PE and systemic thrombolysis for high risk (massive) PE 1 . The PEITHO trial that studied systemic thrombolysis(tPA) versus anticoagulation in patients with submassive (intermediate risk) PE, found that systemic thrombolysis with Tenecteplase prevented death or hemodynamic deterioration but at an increased risk of bleeding 5 . This spurred an interest in catheter directed thrombolysis where lower doses of thrombolytics can be administered thus theoretically reaping the benefits of thrombolysis while mitigating the bleeding risks. There are two methods for catheter directed thrombolysis, standard catheter directed thrombolysis (CDT) and ultrasound assisted catheter thrombolysis (USAT). The standard catheter directed thrombolysis delivers thrombolytic agents locally through a multiple side hole catheter directly into the pulmonary artery thrombus. Ultrasound assisted catheter device (USAT) utilizes an ultrasonic core to deliver acoustic energy that helps in disrupting the thrombus thereby www.nature.com/scientificreports/ improving penetration of thrombolytics into the thrombus. This technique was purported to achieve similar goals of CDTs (catheter directed thrombolysis) with lower bleeding risks [6][7][8][9][10] . However, the purported advantages of USAT over CDT have not been demonstrated in clinical practice 11 . Recently, a few studies have suggested that administering lower dose of thrombolytics systemically (LDT) is effective while preventing bleeding complications. This modality is potentially more cost effective than catheter directed methods as it does not involve using complex equipment or the Cath lab/IR suite and saves operator involvement and time.
As there is no study to date that has compared the above treatment modalities, clinicians are relying mainly on personal preferences and institutional protocols. We hope to bridge this knowledge gap by conducting a systematic review and Bayesian network meta-analysis for treatment of submassive pulmonary embolism. The treatment modalities compared are anticoagulation (AC), systemic thrombolysis with full dose thrombolytics (tPA), low dose thrombolytics (LDT), catheter directed thrombolysis (CDT), and ultrasound assisted catheter thrombolysis (USAT). The primary outcomes studied were in-hospital mortality risk and risk of major bleeding. The secondary outcomes were risk of minor bleeding and recurrent PE. The advantages of using this method are manifold: it allows for indirect comparison of two or more treatments that have never been directly compared provided they are linked via a common comparator; it allows greater statistical precision through incorporation of indirect evidence that is not taken into account with pairwise meta-analysis; it can be used to rank treatment modalities with respect to clinical efficacy/harm, which will be useful when determining policies, guidelines surrounding the choice of treatment 12 .
In assessment of mortality risk and risk of major bleeding, fourteen studies were included comprising 2132 patients. Network plot is given in Fig. 2. Node size was based on sample size and edge width by number of studies. There were 13 studies with AC as a comparator group, 10 studies with tPA, 2 studies with USAT and LDT and one study with CDT.
Mortality risk. Compared with AC, only tPA was found to have significantly lower risk of mortality (RR: 0.35; 95% CrI: 0.09-0.81). When compared to AC, we could not draw any inference with LDT, USAT and CDT as their 95% credible intervals included zero. There was no significant difference between USAT versus CDT (RR: 1.01; 95% CrI: 0.01-72. 8). No inferences could be drawn between any other comparisons as their 95% credible intervals included either zero as the lower limit or an exceedingly large number as the upper limit due to having zero events in one of the treatment arms. The Heat Plot is given in Fig. 3. On meta-regression, publication year was not found to have a significant influence (p = 0.35). The SUCRA (surface under the cumulative ranking curve) ranking for lowest risk of in-hospital mortality was for USAT (SUCRA: 85.15) followed by CDT (SUCRA: 84.56), LDT (SUCRA: 45.41), tPA (SUCRA: 32.41) and lastly anticoagulation (SUCRA: 2.46). Rankogram for mortality risk is given in Fig. 4.
Major bleeding. Regarding risk of major bleeding, there was no significant difference between tPA and AC (RR: 0.95; 95% CrI: 0.31-2.42) and CDT versus USAT (RR:0.41; 95% CrI: 0.01-12.32). No inferences could be drawn from any other comparisons as their 95% credible intervals included either zero as the lower limit or an exceedingly large number as the upper limit due to having zero events in one of the treatment arms. The Heat Plot is given in Fig. 5. On meta-regression, publication year was not found to have a significant influence (p = 0.36). The SUCRA (surface under the cumulative ranking curve) ranking for lowest risk of major bleeding was for LDT (SUCRA: 80.04), followed by CDT (SUCRA: 76.08), USAT (SUCRA: 65.17), tPA (SUCRA: 15.51) and AC (SUCRA: 13.21). Rankogram is given in Fig. 6. Table 1. Summary of included studies.  Risk of bias. Within-study risk of bias was assessed using the Cochrane risk of bias assessment tool for RCTs. All included studies were of good quality and with low risk of bias. We did not find any bias for selective report-  www.nature.com/scientificreports/ ing within studies. To assess the risk of bias due to missing evidence we used the ROB-MEN tool 24 . It is a novel tool that incorporates qualitative and quantitative methods, developed by authors involved with Cochrane reviews and methods groups. With ROB-MEN, we initially assess within study bias between pairwise comparisons followed by qualitative assessment of publication bias and quantitative assessment of publication bias. Thereafter, it runs a network meta-regression using the smallest observed variance as a covariate to estimate small study effects and finally calculate the overall risk of bias. For across study assessment of bias, for qualitative assessment of publication bias, we made the presumption that procedures and newer techniques are more favored. In quantitative assessment of bias, we considered the contribution from evidence to be substantially biased if the difference between treatment groups was at least 15%. On analysis, the overall risk of bias for comparisons were either "low risk" or had "some concerns". None of the comparisons were found to have high risk of bias. The details of risk of bias assessment is given in the supplemental file.

Discussion
While anticoagulation has been the main treatment option for patients with submassive (intermediate risk) pulmonary embolism, the PEITHO trial demonstrated that systemic thrombolysis prevented death or hemodynamic deterioration but at the expense of increased risk of bleeding 5 . A meta-analysis that looked at outcomes with thrombolytic therapy versus anticoagulation had shown lower mortality risk with systemic thrombolysis but at the expense of major bleeding risk 25 . Catheter directed thrombolytic methods using either a pigtail catheter or ultrasound assisted thrombolysis were developed to improve mortality rates while decreasing risk of bleeding. This is also the theoretical reasoning behind administering lower dose thrombolytics systemically. The efficacy of catheter directed thrombolytic methods in improving cardiovascular hemodynamics was demonstrated in the  www.nature.com/scientificreports/ ULTIMA trial, SEATTLE II trial and from initial results from PERFECT registry 8,9,26 . But, despite having data on the safety and efficacy of catheter directed methods, randomized controlled trial data on these techniques is scarce. It is also not clear if ultrasound assistance provides greater benefit. The OPTALYSE-PE trial found lower doses of tPA to be effective with USAT, but this was a single arm study 27 . The only randomized controlled trial comparing USAT and CDT to date is the SUNSET sPE trial, which was performed in patients with submassive PE; There was no difference in safety outcomes between the two groups and standard CDT showed better hemodynamic outcomes 11 . The MOPETT trial that compared LDT versus AC found patients in LDT group to have significant reduction in pulmonary artery pressure 17 .
To the best of our knowledge, ours is the first study comparing full dose systemic thrombolysis versus low dose thrombolysis versus standard catheter directed thrombolysis versus ultrasound assisted catheter thrombolysis versus anticoagulation alone in the treatment of acute submassive (intermediate-high risk) pulmonary embolism. In our Bayesian network meta-analysis, significant reduction in mortality was noted in tPA versus  www.nature.com/scientificreports/ AC. There was no significant difference between USAT versus CDT. No inferences could be drawn between any other comparisons. Regarding risk of major bleeding, there was no significant difference in relative risk of major bleeding between tPA versus AC and USAT versus CDT. No inferences could be drawn from any other comparisons. We also found that systemic thrombolysis has a significantly lower risk for recurrent PE compared to anticoagulation but there is a higher risk of minor bleeding complications. Our study differs from the previously published meta-analysis in that we found no significant difference in risk of major bleeding between systemic thrombolysis and anticoagulation.
Our study also shows that while the newer modalities of treatment for pulmonary embolism are promising, there is lack of data to comment on the purported advantages. The available studies are small sized and thereby lack power to demonstrate any statistical significance. There are two ongoing trials that seek to clarify this: HI-PEITHO and PEITHO-3. These studies are estimated to be completed by 2025 and 2027 respectively. The Higher-Risk Pulmonary Embolism Thrombolysis (HI-PEITHO) study is a multinational multicenter randomized controlled trial comparing ultrasound assisted thrombolysis plus anticoagulation (USAT) versus anticoagulation (AC) alone in intermediate-high risk pulmonary embolism 28 . PEITHO-3 trial is a multinational multicenter randomized controlled trial that compare the efficacy and safety of reduced dose alteplase (LDT) with standard heparin anticoagulation 29 (AC). The results of these highly anticipated trials might shed light on this important topic.

Methods
This systematic review and meta-analysis is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) Extension statement for systematic reviews incorporating network metaanalyses for health care interventions 30  Search was restricted to adult patients using limit/age filters available on the databases. The articles were downloaded to EndNote and reviewed by two independent reviewers (J.K and B.K), who are both attending physicians in Internal Medicine. Duplicates were identified on EndNote and removed manually. Disagreements between reviewers were resolved with discussion and achieving consensus. www.nature.com/scientificreports/ Study selection. We included randomized controlled trials on adult population that reported at least one relevant clinical outcome of interest for our study. Only studies that specified sub massive (intermediate risk) cases of pulmonary embolism were considered. Sub massive PE was defined as cases with acute PE with objective evidence of RV dysfunction but hemodynamically stable. Evidence of RV dysfunction included positive CT or echo findings and/or elevated cardiac biomarkers such as troponin or BNP or NT-pro BNP. Cases without RV dysfunction were not included in this study. Studies where PE grade was not assessed were not considered. Disagreements between reviewers were resolved by discussion and achieving mutual consensus and after consultation with the lead author (D.M).
Data extraction. From selected articles, we extracted the number of participants in each treatment arm, total patient sample size, in-hospital mortality, major bleeding, minor bleeding, and recurrent PE events in each group. Major bleeding events were any significant bleeding events as defined by the International Society of Thrombosis and Hemostasis (ISTH) 31 . Bleeding events not fulfilling the definition for major bleeding were counted as minor bleeding events. If clarification were needed regarding the study or variable of interest, corresponding authors were contacted to ensure the study/variables met our case definitions. Quality of studies were assessed using the Cochrane bias risk assessment tool 32 . Risk of bias due to missing evidence was conducted using the ROB-MEN tool 24 .
Data synthesis and analysis. Data analysis was conducted in R, version 4.1.2, using BUGSnet and R2WinBUGS packages 33 . Outcomes were combined using random effects model using uninformative priors.
We specified a burn-in of 50,000 iterations followed by 100,000 iterations with 10,000 adaptations and convergence was checked and confirmed. We estimated the risk ratio (RR) which was reported as 95% credible intervals (CrI). Markov chain Monte Carlo (MCMC) modeling was used to estimate the relative ranking probability of the clinical outcome of interest with each treatment modality and a rankogram was created based on surface under the cumulative ranking curve (SUCRA). Beta-binomial regression was used to handle double-zero studies. Inconsistency was assessed based on the inconsistency model described in the NICE-DSU TSD 4 34 . Here, inconsistency model is weighted against how well they fit the data compared to the consistency model to determine if there is evidence of inconsistency. Correlation plots were created using the posterior mean deviances of each data point between the consistency and inconsistency model (Supplemental File). As we included studies spanning over three decades, we conducted a meta-regression to assess if publication year was an influencing factor.

Data availability
All data generated or analyzed during this study are included in this published article and its supplementary information files.