Introduction

Pericardial complications following cardiac surgeries are among the most common and debilitating complications that can have a significant impact on long-term survival. Common pericardial complications following cardiac surgeries are postpericardiotomy syndrome (PPS) (incidence rate: 10%–40%), pericarditis (25%–50%) and pericardial effusion (PE) (10%–60%).1–4 Also, as pericardial inflammation may trigger arrhythmias, the occurrence of postoperative atrial fibrillation (POAF) is a life-threatening complication in acute pericarditis, with a high risk of recurrence among predisposed individuals.5

There is limited knowledge about the pathogenesis of these complications; however, PPS is thought to be an immune-mediated process caused by pericardial and pleural damage and other potential acquired factors that are not well understood.6

Postoperative mortality, morbidity and healthcare costs can be reduced by appropriate prophylaxis.7 The effectiveness of pharmacological strategies proposed for treating or preventing these complications, such as colchicine, non-steroidal anti-inflammatory drugs (NSAIDs) or corticosteroids, is uncertain and requires further evaluation.8

Previous systematic reviews have focused on the effect of colchicine in reducing pericardial complications by pooling results of the few existing primary studies.2–4 9 10 These reviews have sometimes pooled studies with different designs, each of which belongs to a different level of evidence.2 11 Also, none of these reviews have assessed the overall certainty of evidence. Relying purely on the statistical significance of a meta-analysis can be misleading in clinical decision-making.12 Additionally, other treatment choices have not been investigated, and their effect has not been compared with colchicine and each other.2–4 10 11 13 Therefore, we performed a systematic review using network meta-analysis (NMA) of randomised control trials (RCTs) to identify the most effective pharmacological treatments for pericardial complications following cardiac surgery.

Methods

We registered our protocol with the International Prospective Register of Systematic Reviews on 24 June 2023 (CRD42023434974) and followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension statement for NMA to report our findings.14 15

Searches

We performed the systematic search in PubMed/MEDLINE, EMBASE and the Cochrane Central Register of Controlled Trials (CENTRAL) on 22 February 2023, without any language or publication date restriction. Additionally, we searched for more eligible trials by reviewing reference lists of related guidelines, reviews and all included studies. On 22 January 2024, we updated the search. See online supplemental etables 2 and 3 for more details regarding search terms, date of searches and an exported strategy. The references were added to an EndNote 20 reference manager, which removed duplicate records.16

Eligibility criteria and study selection

We included peer-reviewed RCTs that enrolled adults (18 years or older) undergoing cardiac surgery (surgeries via sternotomy and pericardiotomy, including coronary artery bypass grafting (CABG), valve surgery and other major cardiac surgeries). We included trials that reported on at least one of the selected pericardial complications (PPS, PE or pericarditis) as primary or secondary outcomes. We excluded trials that included populations undergoing minor cardiac procedures or with a prior history of major pericardial disease. Online supplemental etable 1 presents more details on eligibility criteria and our PICOTS.

We used the Rayyan online software to facilitate literature screening.17 Potentially eligible studies were screened by pairs of reviewers who independently reviewed titles, abstracts and full-text articles based on eligibility criteria. Disagreements were resolved through consensus or adjudication with a third reviewer.

Data extraction

We collected data independently and duplicated on the following: study characteristics (authors, publication year, country of recruitment and trial registration), characteristics of populations (sample size, age, sex, types of cardiac surgery and underlying comorbidities), type of surgical procedure, characteristics of interventions and comparators (formulation and drug classification, description, dosage and time of intervention), risk of bias and measures of association with incidence of the outcomes of interest. A third-party adjudicator intervened to settle disputes if reviewers could not reach a consensus.

Risk of bias assessment

We used the Cochrane risk of bias tool for randomised trials (ROB V.2.0) to assess the risk of bias of the included RCTs.18 We rated the risk of bias on five main domains: randomisation and allocation concealment, blinding and deviation from intended intervention, loss to follow-up and missing outcome data, outcome measurement and deviations from the registered protocol.18 19

Data synthesis and statistical methods

We summarised the effects of interventions using relative risks (RR) and corresponding 95% CI. For continuous outcomes, we reported mean differences (MD) with associated 95% CI. We performed a frequentist random-effects NMA using the restricted maximum likelihood estimator (REML) with the Netmeta package in R (V.4.03, R Foundation for Statistical Computing) for each outcome. We generated treatment nodes based on the mechanism of action.20 21

The transitivity of each group of trials included in network models was considered based on clinical and methodological knowledge. In the case of intervention intransitivity (due to drugs’ different mechanisms of action) that was incomparable in one network, we decided to run separate network meta-analyses for each group of interventions. This ensures that all treatments were similar and comparable in terms of effect modifiers.22

We performed a pairwise random effects meta-analysis to determine the direct estimates and evaluate any inconsistencies. Node-splitting models were used to assess local incoherence and obtain indirect estimates (the difference between direct and indirect evidence in closed loops when possible).23 To evaluate heterogeneity, we examined forest plots visually and calculated the I2 statistical test. We classified heterogeneity levels as follows: 0%–40% as potentially unimportant, 30%–60% as moderate, 50%–90% as considerable and 75%–100% as critical based on Cochrane recommendations.24 We assessed publication bias by visually inspecting the Funnel plots and conducting Egger’s statistical test for comparisons with 10 or more trials.

For the visual presentation of the network geometry and network summary of the risk of bias, we used the Confidence in Network Meta-Analysis online application.25

Subgroup analyses

We investigated three predefined hypotheses to explain the variability between studies using meta-regression analysis to determine whether the risk of bias, timing of receiving interventions and type of cardiac surgery affect the treatment effects. We conducted subgroup analyses only if each subgroup contained two or more studies and explored subgroup effects with a test of interaction.

Assessment of the certainty of evidence

We used the Grading of Recommendations, Assessment, Development and Evaluations (GRADE) approach for NMA to evaluate the certainty of evidence. We assigned a high, moderate, low or very low certainty rating for each comparison and outcome, considering the risk of bias, inconsistency, indirectness, publication bias, intransitivity, incoherence and imprecision.12 26–28

The minimally contextualised approach used for making judgments of imprecision considers only whether CIs include a minimally important effect. It does not consider the magnitude of plausible effects captured by CIs. The results were reported using guidance from the GRADE working group, which involves using different adjectives based on the certainty of evidence.12

ResultsSearch results

Our search found a total of 2286 articles and included 39 RCTs with a total of 6419 patients. Out of the included RCTs, four examined the therapeutic effects,7 29–31 and the rest of the studies examined prophylactic effects. Prophylactic studies were RCTs that randomised participants to intervention groups before surgery, with the aim of investigating the occurrence of one of the outcomes of interest. For therapeutic studies, all patients who developed an outcome of interest after cardiac surgery were randomly assigned to intervention groups with the aim of recovery.

10 trials were not included in the final NMA due to intervention intransitivity or overlapped population (online supplemental efigure 1).

Study characteristics

We included 6419 patients, with an average age of 61.18 years, and 70% men. The mean prevalence of hypertension, diabetes mellitus, smoking and chronic obstructive pulmonary disease (COPD) was 60.5%, 26.9%, 33.1% and 8.7%, respectively, among the included participants. An average of 6.8% of patients had a prior history of cardiac surgery. Online supplemental etables 4 and 5 present more details of patients’ characteristics and interventions. Online supplemental efigure 4 shows the worldwide distribution of the included studies and populations.

Risk of bias

We judged 14 studies as low, 9 as having some concerns and 16 as high risk of bias. Most bias concerns were raised from the randomisation process and selection of the reported results. 17 trials (43.6%) adequately generated their randomisation process and allocation sequence concealment. 19 trials (48.7%) had a credible registered protocol without any deviations from it (Online supplemental efigures 3 and 4).

Network meta-analysis

The network is mainly connected via a control node. There were a few head-to-head RCTs forming closed loops to provide indirect evidence (for network diagrams for each outcome, see figure 1 and online supplemental efigures 35, 44, 51, 58 and 70). In networks with closed loops, our node-splitting models showed no evidence of incoherence (online supplemental efigures 26, 33 and 63). Online supplemental efigures 6, 14, 32, 30, 37, 46, 53, 60 and 72 present the pairwise and network forest plots for each outcome separately.

Figure 1

Network maps for the pericardial complications. The size of the nodes (circle) corresponds to the number of patients randomised to that intervention. The thickness of the lines corresponds to the number of studies for each comparison. ARBs, angiotensin receptor blockers; MgSO4, magnesium sulfate; NSAIDs, non-steroidal anti-inflammatory drugs; TXA, tranexamic acid.

Network heterogeneity was unimportant for the incidence of PPS (I2=8.7%, 95% CI 0.0% to 81.0%), pericarditis (I2=0.0%, 95% CI 0.0% to 89.6%), POAF (I2=0.0%, 95% CI 0.0% to 64.8%), mortality (I2=0.0%), adverse events (AEs) (I2=0.0%), incidence of PE (I2=0.0%, 95% CI 0.0% to 33.5%) and considerable for the incidence of pleural effusion (I2=71.0%, 95% CI 26.5% to 88.6%) and duration of ICU stay (I2=62.1%, 95% CI 0.0% to 91.2%).

We decided to conduct two separate network meta-analyses for PE outcomes. This decision was based on consulting with expert clinicians and the need to prevent intransitivity. Considering the different mechanisms of action and contexts of clinical use, we performed the first NMA (NMA #1) using anti-inflammatory and cardiovascular-affecting drugs and the second one (NMA #2) with coagulation-affecting drugs.

In investigating networks that included 10 or more studies to assess publication bias, all Funnel plots were symmetric, and Eggers’ tests were non-significant (online supplemental efigures 19 and 65).

We summarised the contributions of each direct comparison in the network, along with both the network and indirect estimates, using the network contribution matrices for each outcome. Please see online supplemental etables 10 to 17.

NMA results for the prophylactic interventionsIncidence of PPS

Seven trials reported on the incidence of PPS, including 2061 patients and 310 events. We found that colchicine reduced the risk of PPS compared with control (RR 0.53, 95% CI 0.38 to 0.73; high certainty); corticosteroids may have no significant effect on PPS (RR 0.88, 95% CI 0.56 to 1.36; low certainty), and statin results were very uncertain (very low certainty of evidence). See table 1 and online supplemental efigures 5 to 8.

Table 1

Evidence profile for the effectiveness of medications on the prevention of pericardial complications

Incidence of postoperative PE

18 studies reported a total of 437 PE events among 2971 included participants. None of the suggested interventions showed a significant effect on preventing PE. However, prophylactic heparin may reduce the risk of PE (RR 0.35, 95% CI 0.13 to 0.98; low certainty). These two NMAs’ summary of findings is presented in table 1 and online supplemental efigures 13 to 16 and 22 to 25.

Incidence of postoperative pericarditis

Nine studies reported the incidence of postoperative pericarditis, including 897 participants and 189 events. Beta-blockers may have a mild effect on preventing postoperative pericarditis (RR 0.66, 95% CI 0.45 to 0.97; low certainty). Colchicine, non-steroidal anti-inflammatory drugs (NSAIDs), and digoxin had no statistically significant effects (table 1 and online supplemental efigures 29 to 32).

Incidence of postoperative pleural effusion

For this outcome, eight trials reported 1125 events for 2104 patients. Colchicine was ineffective (RR 0.69, 95% CI 0.46 to 1.05; low certainty) in reducing the risk of pleural effusion compared with control. Results for other interventions were uncertain (table 1, and online supplemental efigures 35 to 39).

Incidence of postoperative all-cause mortality

The reviewed trials reported 22 (1.5%) deaths. 10 trials included in the NMA with 1503 patients. None of the interventions significantly affected the mortality rate (table 1 and online supplemental efigures 44 to 48).

Duration of ICU stay

Six RCTs reported hospitalisation duration with an average stay of 2.2 days. Our findings showed none of the interventions had a significant effect on the overall duration of ICU stay compared with control (colchicine and NSAIDs had moderate certainty, and vitamin D, corticosteroids and beta-blockers had low certainty of evidence). Online supplemental efigures 51 to 55 show the network geometries, the forest plot and rankings.

Incidence of POAF

There were 489 atrial fibrillation cases reported after surgery in 1975 patients. Results showed that beta-blockers reduce the risk of POAF with a larger effect than the control group (RR 0.4, 95% CI 0.20 to 0.81; moderate certainty). NSAIDs, then fish oil and colchicine with an inferior certainty, may reduce the risk of POAF (NSAIDs: RR 0.37, 95% CI 0.23 to 0.59; fish oil: RR 0.28, 95% CI 0.09 to 0.90; colchicine: RR 0.37, 95% CI 0.23 to 0.59, all low certainty). See table 1 and online supplemental efigures 58 to 62 for network geometries, the forest plot and rankings.

Heterogeneity and subgroups

For each outcome, we conducted subgroup analyses for types of surgery (CABG, valve surgery and mixed), time of interventions (preoperative, postoperative, and intraoperative) and risk of bias (low, high and some concerns). We discovered only one subgroup analysis that demonstrated statistical significance. Colchicine was found to be beneficial in reducing pleural effusion when administered during mixed types of surgery, but not in the case of only CABG and valve surgeries (p<0.01 for interaction; see online supplemental efigures 11, 12, 20, 21, 28, 42, 43 and 66 to 69. See online supplemental etable 6 for meta-regression for potential sources of heterogeneity.

Meta-analysis results for the therapeutic interventions

As we found few studies that investigated the treatment effect of colchicine only, we used meta-analysis to pool the effect values. We ran a meta-analysis for outcomes with at least two or more studies. The analysis results presented in table 2 and online supplemental etable 7 indicate that none of them was statistically significant.

Table 2

Evidence profile for the effectiveness of medications on the treatment of pericardial complications

DiscussionMain findings

This NMA of 39 RCTs with a total population of 6419 patients evaluated the comparative effectiveness of prophylactic and therapeutic agents for PPS, PE and pericarditis after open heart surgeries.

Our study demonstrates that colchicine use is associated with reduced risk of PPS. Beta-blockers may prevent postoperative pericarditis and reduce POAF incidence with a large magnitude of effect compared with control. Fish oil may have a significant impact on reducing the risk of POAF. NSAIDs and colchicine may reduce the risk of POAF but with an inferior effect to beta-blockers. Also, prophylactic heparin may reduce the risk of PE (low certainty of evidence). We found no evidence of a difference in the risk of pleural effusion, all-cause mortality or postoperative ICU stay. Although there was no difference in the incidence of serious AEs for any of the interventions, some RCTs reported a higher incidence of gastrointestinal discomfort for colchicine compared with the control group.

In relation to previous findings

We have compared our NMA results with prior large RCTs and meta-analyses in online supplemental etable 8. The results of this NMA suggest that colchicine has a favourable effect on preventing the occurrence of PPS. This finding is in agreement with other previous large trials and meta-analyses.2 3 8 10 32–34 Corticosteroids and statins, consistent with other studies, had no significant effect on PPS. Since our NMAs’ results for corticosteroids and statins were based on a limited number of RCTs, further investigation into these interventions is strongly recommended.2 13 35 36 In line with previous reports, none of the proposed interventions reported a meaningful effect for the treatment or prevention of PE.4 7 30 33 37 38 Our results showed that prophylactic heparin may effectively prevent PE. This result was raised from a single high risk of bias RCT with 60 participants.39 This study showed that heparin use can significantly (p value=0.02) decrease the incidence of mild PE.39 However, considering the low certainty and the wide CI, it is uncertain whether there is a true causative relationship. As for pericarditis, our results showed that the evidence for the effect of beta-blockers on the incidence of postoperative pericarditis is scarce. Similarly, other interventions did not show a significant effect, which was consistent with previous trial results.40–42

The NMA demonstrated that fish oil may have a significant impact on the prevention of POAF. This result came from a high risk of bias RCT that enrolled 73 patients. Although some prior trials showed that fish oil is not effective in reducing POAF, our NMA estimate aligns with the effect observed in a recent NMA and some prior meta-analyses.43–45 Some trials investigating fish oil’s effectiveness in preventing POAF excluded patients with a history of supraventricular arrhythmia, while others included them. This heterogeneity between the participants of these two groups of studies might explain the difference in the final results. However, our NMA results were assessed as low certainty of evidence, andthe mentioned recent NMA results for the efficacy of fish oil were reported as very low certainty. This shows the results are very uncertain for clinical application.

Strengths and limitations

To our knowledge, this is the first NMA to study and compare the effectiveness of pharmacological preventions and treatments for postoperative pericardial complications. Our analysis is up-to-date and includes all suggested interventions for the most common pericardial outcomes. We applied a rigorous methodological approach and considered all NMA preassumptions to minimise bias in results. The quality of included trials was assessed using the Cochrane risk of bias tool, and the certainty of evidence was assessed based on the GRADE approach. We used a minimally contextualised approach, drawing conclusions from the NMA to enhance clinical usefulness.27 46

Our study has several limitations. First, there were only a limited number of trials assessing the effect of NSAIDs, corticosteroids, statins, and other agents on pericardial complications after cardiac surgeries, and some of them were assessed as studies with a high risk of bias, which leads to low certainty in results. Also, moderate network heterogeneity was reported for the incidence of PE, and considerable heterogeneity was noted for the incidence of pleural effusion and duration of ICU stay.

We were unable to assess network incoherence for some outcomes because the network provided only a few closed loops and was mainly connected via a control node.

Future directions and clinical implementation

There has been a shift towards the use of minimally invasive procedures to prevent major complications and improve patient outcomes across medical specialties. Nevertheless, open cardiac surgeries are still inevitable in certain situations despite serious complications. This underscores the importance of taking measures to prevent severe complications.

The rationale for preventing postoperative pericardial complications lies in the significant morbidity and mortality associated with these conditions. Early intervention through pharmacological measures can mitigate the risks, enhance recovery and improve overall patient-important outcomes.2

Colchicine, as the key prophylactic intervention, is generally initiated at a loading dose of 0.5–1 mg, followed by 1–2 mg one or two times a day, depending on renal function. The timing for initiation seems better when it begins before the surgery or within the first few days after surgery.3 10

Our study has shown that there are limited proposed interventions with the potential to prevent or treat these complications. Furthermore, the number of trials investigating these interventions, such as statins, NSAIDs, corticosteroids and digoxin, is limited, making it challenging to draw a definitive conclusion.

On the other hand, most of the available studies have focused on colchicine administration. Although colchicine has shown a good effect, it has reported side effects and an increasing risk of gastrointestinal events and myalgias in some previous analyses.47 48

In conclusion, we strongly recommend more investigations on other interventions and suggest proposing and examining new interventions in large, high-quality trials due to the critical clinical importance of pericardial complications and the weak body of evidence for some interventions.