‘Are Routine Post-discharge Diuretics Necessary After Pediatric Cardiac Surgery?’

A prospective, one-armed, safety non-inferiority trial with historical controls was performed at a single-center, quaternary, children’s hospital. Inclusion criteria were children aged 3 months–18 years after pediatric cardiac surgery resulting in a two-ventricle repair between 7/2020 and 7/2021. Eligible patients were compared with patients from a 5-year historical period (selected using a database search). The intervention was that “regular risk” patients received no diuretics and pre-specified “high risk” patients received 5 days of twice per day furosemide at discharge. 61 Subjects received the intervention. None were readmitted for pleural effusions, though 1 subject was treated for a symptomatic pleural effusion with outpatient furosemide. The study was halted after an interim analysis demonstrated that 4 subjects were readmitted with pericardial effusion during the study period versus 2 during the historical control (2.9% versus 0.2%, P = 0.003). We found no evidence that limited post-discharge diuretics results in an increase in readmissions for pleural effusions. This conclusion is limited as not enough subjects were enrolled to definitively show that this strategy is not inferior to the historical practice. There was a statistically significant increase in readmissions for pericardial effusions after implementation of this study protocol which can lead to serious complications and requires further study before conclusions can be drawn regarding optimal diuretic regimens.


Introduction
Diuretics are routinely prescribed post-discharge after twoventricle pediatric cardiac surgical repairs despite a near absence of evidence to support this practice. Inflammation and fluid administration in the peri-operative period result in fluid overload which can manifest as pleural effusions post-operatively [1]. However, advances in bypass and surgical techniques have reduced the degree of fluid overload. Furthermore, in-patient post-operative diuresis can quickly resolve fluid overload. Once fluid overload is relieved, in the absence of residual lesions or univentricular anatomy, further diuretic therapy may not be required to prevent the accumulation of pleural effusions.
Pleural effusions are a known complication after pediatric cardiac surgery. Pleural effusions may accumulate due to increased capillary permeability, increased hydrostatic pressure, lower oncotic pressure, fluid overload in excess of what the lymphatic system can accommodate and trauma to the lymphatic system. Described risk factors pertinent to twoventricle repairs include systemic venous hypertension, small pulmonary arteries, severe tricuspid regurgitation, and surgical trauma [1]. In 2002, Bosci et al. found 39% of patients in a prospective study developed pleural effusions while in-patient [2]. In this cohort, the chest tubes were usually removed on postoperative day one and a diuretic regimen was not described. These data are from an era when more inflammation from All work performed at Ann & Robert H. Lurie Children's Hospital of Chicago.

3
bypass could be expected and did not describe pleural effusion re-accumulation after discharge.
A systematic review of pleural effusion after discharge post-congenital cardiac surgery in 2020 found only ten papers to include and only two that dealt with patients with two ventricles [3]. This review found a 3% adjusted readmission rate for pleural effusions. This may be higher than expected in the current era as the two-ventricle data were derived from a 1997 study by Vaynblat et al. of Tetralogy of Fallot repaired at a median age of 9.4 years [4]. The Vaynblat study described an association between increased pleural drainage and higher post-operative right atrial pressures supporting the notion that systemic venous hypertension increases pleural effusions. The second two-ventricle study in the systematic review found a high rate of pleural effusions in children re-presenting with pericardial effusions, but did not describe the rate of isolated pleural effusions in the setting of a randomized controlled trial of peri-operative steroids versus placebo [5].
To investigate our own experience with pleural effusion readmissions, we queried a surgical database to examine children over 1 month of age post two-ventricle repair with bypass and found only one readmission for pleural effusions among 476 consecutive cases over a 30-month period. During this period, standard practice was to prescribe furosemide as an outpatient that would eventually be stopped at cardiology follow-up which was typically 4-6 weeks after surgery. We considered this in conjunction with informal discussions with adult providers who do not routinely prescribe diuretics after discharge. We hypothesized that diuretics may not be necessary to prevent pleural effusion re-accumulation once fluid overload is resolved and when there is an absence of risk factors such as systemic venous hypertension and residual anatomic lesions.
In order to evaluate this hypothesis, we designed a prospective, one-armed, safety trial in two phases. Our intervention was to discharge "regular risk" patients after twoventricle repair home on no diuretics and to send those with pre-specified "high risk" factors home with 5 days of furosemide twice per day. Our specific hypothesis was that the study protocol would generate a non-inferior rate of readmissions due to pleural effusions during the study period compared to a historical control. The secondary hypothesis was that readmission for pericardial effusion would be noninferior to the historical control.

Study Design
This study was approved by the Institutional Review Board at Ann & Robert H. Lurie Children's Hospital. Consent from parents, and assent from subjects over age 11, was obtained post-operatively.

Study Population
Inclusion criteria were children aged 3 months-18 years with two-ventricle physiology after cardiac surgical repair. Exclusion criteria included (a) ongoing need for diuretics after surgery (i.e., heart failure, greater than mild residual shunt, greater than mild mitral or aortic valve regurgitation, greater than mild systolic dysfunction, lung disease, or systemic hypertension), (b) single ventricle palliation, (c) chylous effusion, and (d) heart transplant (due to heart failure team preference).

Measurements
Demographic, surgical, in-patient, and outpatient follow-up data were obtained prospectively by study personnel and logged in a shared excel spreadsheet. Data on historical controls were abstracted from the electronic health record (EHR). Readmission data were collected for any readmission, but specifically for readmission for pleural or pericardial effusion.

Intervention Protocol
The treatment protocol involved discharging subjects on one of the three following regimens (1) no diuretics for "regular risk" subjects (2) furosemide 1 mg/kg/dose (maximum 20 mg) twice per day for 5 days for "high risk" subjects (3) ad lib diuretic regimen if the discharging physician in conjunction with the study team determined a unique situation required diuretic adjustment. No protocolized adjustment was made for in-patient treatment. Physicians were aware of the study and were encouraged to attempt to maintain diuresis to remove fluid overload while in-patient, but also encouraged not to keep subjects in the hospital longer for ongoing diuresis.
"High risk" criteria were generated a posteriori by study team consensus. These criteria included (a) chest tube drainage > 4 ml/kg/day in the 24 h prior to removal and removed on the day of discharge (b) subject > 10% fluid overloaded on day of discharge based on discharge weight/pre-operative weight (c) Pleural effusion evident on discharge chest radiograph or echocardiogram (> trace on echocardiogram) or re-development of pleural effusion while in-patient after chest tube removal (d) greater than trace pericardial effusion on discharge echocardiogram (e) post-Tetralogy of Fallot repair with a transannular patch (due to presumed venous hypertension from elevated right ventricular end diastolic pressure) and (f) post-atrial septal defect (ASD) repair based on surgical team concern for higher risk of pleural effusion.
Subjects post complete atrioventricular canal (AVC) repair were added mid-study to the "high risk" group after a pericardial effusion developed outpatient in a single patient per the surgical team's preference.
The study was separated into two phases with safety and feasibility established during a pre-planned, limited, phase one that included the first fifteen subjects. During phase one, the study team and discharging physician met to determine the appropriate treatment regimen based on study guidelines. Each subject was seen within 1 week of discharge and had a routine chest X-ray at that visit.
After no pleural or pericardial effusions were noted in phase one, we proceeded with phase two. During phase two, the consenting study member and discharging physician assigned the treatment regimen based on the pre-specified risk category. During phase two, routine chest X-rays were not ordered and follow-up was per our standard within 10-14 days. Routine echocardiograms are not performed at surgical follow-up in our institution. Follow-up was obtained by the surgical clinic team which tracked addition of outpatient diuretics, readmissions, and any other parent concerns for 1 month after discharge.

Power Considerations
Data on 476 two-ventricle patients who underwent an eligible procedure were pulled from the EHR 1/1/2017-6/30/2019-prior to study baseline. One of these patients had a pleural effusion, so we opted to use a non-inferiority design to determine the number of patients recruited into this study. To determine that the experimental protocol was not inferior to these historical controls, assuming a 0.2% historical event rate, 5% non-inferiority margin (corresponding to 4 pleural effusion readmissions), 80% power, and a 10% type-I error rate, 152 patients would be required.

Statistical Analysis Plan
We planned to describe the characteristics of intervention and historic patients, using median and interquartile ranges for continuous variables and counts for categorical variables. Then, should pleural effusion cases have developed, we would use a non-inferiority margin approach assuming a 5% non-inferiority margin (corresponding to 4 or 5 events) to indicate inferiority of the proposed protocol with the institutional historical control.

Modification Due to Safety Concerns
A number of subjects met the secondary endpoint of readmission due to pericardial effusions. This led to concern for an increase in pericardial effusions due to the study protocol so the study was paused after enrolling patients for 1 year. At a pre-planned safety review, an interim analysis was performed to determine whether the rate of pericardial effusions in a 5-year historical period was different than that observed in the study period. A longer historical control was chosen over the original due to the rarity of pericardial effusions. To identify historical controls, the institutional surgical database was searched for patients meeting study inclusion/exclusion criteria in a 5-year period just prior to the study during which diuretics were routinely prescribed for approximately 1 month after discharge. Using the same search methodology, the institutional surgical database was searched in the study period to retrieve data on all eligible patients from that period (not just those screened for the study prospectively so more subjects were found for this analysis). We did not exclude patients from the study period for residual lesions, chylous effusions, or longer length of stay as we did in the prospective study, as doing so would have made the selection criteria different from the historical control. We confirmed that we identified all of the pericardial effusions in the study period that were prospectively recorded in the study log, which suggests the search criteria used were adequate to identify pericardial effusions. The rate of pericardial effusions in the historical period was compared to the study period using Fisher's exact test due to the small numbers.

Results
There were 124 subjects screened and 61 included in the analysis (see Fig. 1, CONSORT diagram). Sixty were consented and one additional subject was included because she developed a symptomatic pericardial effusion after receiving the study regimen. This patient was not enrolled in the study as it has just been paused. However, the patient had an ASD repair and received the five days 'high risk' regimen exactly as she would have if she had been enrolled as this had become the hospital norm. She developed a symptomatic pericardial effusion and these data were deemed important enough to add her to study analysis as we had just paused the study to consider the incidence of pericardial effusions. The study was subsequently terminated due to an increased incidence of readmissions for pericardial effusions.
Demographics and description of surgical repairs for included patients are given in Table 1. There were 23 females (37.7%) with a median age of 41 months [inner quartile range (IQR) 9-75 months] and median weight of 13.6 kg (IQR 8.0-20.8 kg) at the time of surgery. The median fluid overload (pre-operative weight/discharge weight) was − 1.9% (IQR − 3.7 to 0.7%). No patients were discharged with a fluid overload > 10% and three patients were discharged with a fluid overload > 5%, none of whom were readmitted.
Out of 61 patients, 36 were discharged with no diuretics. 24 were discharged with 5 days of furosemide due to "high risk" designation. Seventeen of these were due to the type of surgical repair (ASD or AVC). One patient was discharged with the 'ad lib' regimen which was 10 days of furosemide due to mild plus atrioventricular valve regurgitation after AVC repair per surgeon request.
No patients were readmitted for pleural effusion. Five pleural effusions were detected at outpatient follow-up. Three were trace to small and untreated. One was diagnosed due to tachypnea at follow-up and was treated outpatient with furosemide. One patient had a small pleural effusion in the presence of tamponade physiology from a large pericardial effusion for which the patient was readmitted. Two additional patients received short courses of outpatient furosemide for lower extremity edema.
The study was halted for an interim analysis after four patients were readmitted for pericardial effusions. When compared to the historical baseline during which two patients had been readmitted for pericardial effusion, we found this was a significant increase in readmissions during the study period [ Table 3. Based on these findings, the study was terminated for safety concerns.
We investigated how fragile our study results were by calculating a fragility index, i.e., the minimum number of patients whose readmission for a pericardial effusion would have needed to be prevented during the study period to render the risks not statistically different, assuming a 5% type-I error rate. This index was 2 for pericardial effusion readmissions, and 1 for symptomatic pericardial effusions, indicating that our findings were quite fragile.
Lastly, there were two asymptomatic pericardial effusions treated with diuretics as an outpatient and one asymptomatic small pericardial effusion that was not treated.

Discussion
No subjects were readmitted for pleural effusions while using a regimen of no diuretics in "regular risk" or a 5-day course in "high-risk" patients after discharge. Four subjects were readmitted for pericardial effusions which was an increase compared to a historical control. The number of subjects readmitted for pericardial effusion was quite low with a small fragility index which constrains the conclusions that can be drawn from this. Despite the small number of subjects affected by pericardial effusion, one presented in tamponade which highlights the importance of further exploring this finding. It is especially important to explore these new findings regarding the incidence of pericardial effusions when considering that other studies have also found that diuretics may not be necessary to prevent pleural effusions [6] and programs may be considering reducing outpatient diuretic use.
Trivedi et al. published a quality improvement project that reduced their post-discharge diuretic regimen from 32 to 10 days [6]. They found no change in the small number of pleural effusions discovered after discharge. Only three pleural effusions were found out of 115 patients post-protocol implementation and it was unclear if any were readmitted. They found no change in readmission rates pre-and post-protocol implementation. Interestingly, the frequency of pericardial effusions decreased (p = 0.03) in their postintervention group. This could be related to a protocolized practice change shortly before beginning their quality improvement project that involved the use of ibuprofen as an outpatient when pericardial inflammation was suspected. This may explain the discrepant pericardial effusion findings between these two studies.
Trivedi's quality improvement project and this prospective study both found extremely low rates of pleural effusion and no increase in readmissions for pleural effusions. This study was halted prior to reaching a goal of 154 subjects. However, the absence of any readmissions for pleural effusion through 61 subjects suggests that if routine diuretic therapy has any utility in preventing readmissions for pleural effusions, then the number needed to treat is likely to be high. That number needed to treat would only consider readmission for pleural effusion. Diuretics may not be needed to prevent pleural effusion once the initial fluid overload has been ameliorated as the lymphatic system may be able to increase clearance when needed, especially in the absence of the systemic venous hypertension seen in single ventricle physiology.
In contrast, the lymphatic system may not clear pericardial effusions and we found a higher frequency of this complication while using our diuretic regimen. These results are driven by very small numbers and the fragility index shows us that a change in 1 or 2 patient outcomes would have made the results non-significant statistically. We also examined these re-presentations by symptoms since the decision to readmit may vary between providers and found that this was also increased. Given the statistical significance of both of these outcomes combined with the plausibility that longer courses of diuretics suppress pericardial effusions suggest that diuretics may decrease the frequency of these effusions becoming clinically relevant. The frequency of readmissions for pericardial effusion during the historical control was 0.2%, similar to the rate of 1.1% found in retrospective review of 142,633 cases reported in the pediatric health information database [7]. We suspect many of those patients in the database would have been treated with diuretics as we believe this is a widespread practice. We saw a higher rate during our study era (2.9%) which supports the hypothesis that diuretics, a treatment for pericardial effusions, may decrease the symptoms.
If diuretics suppress the incidence of symptomatic pericardial effusions, this still may not justify routine use of diuretics after all two-ventricle repairs given the very low numbers. A more targeted approach may be sufficient to prevent an increase of pericardial effusions in the absence of routine, prolonged, diuretic courses. This targeted approach could be used for patients with a rub, those with an ASD repair and those with diffuse ST segment changes, all risk factors for pericardial effusion. Non-steroidal antiinflammatory drugs (not used routinely on the subjects in our study) may more appropriately address the cause of pericardial effusion and thus be more effective than furosemide. Further study of when or if furosemide is necessary post-discharge and how to address pericardial effusions in that context warrants further study. This could be pursued most efficiently in a quality improvement format that allows for flexibility in treatment regimen.
The limitations of this study include the unexpected increase in pericardial effusions that led to stopping the study prior to full enrollment. Additionally, we did not have standardized criteria for pericardial effusion detection, treatment, or readmission in the historical or study periods. This may have led to readmissions that could have been treated outpatient, one reason 'symptomatic' effusions were included in the analysis as this is not provider

Conclusion
We found no evidence that limited post-discharge diuretics results in an increase in readmissions for pleural effusions. This conclusion is limited as not enough subjects were enrolled to definitively show that this strategy is not inferior to the historical practice. There was a statistically significant increase in readmissions for pericardial effusions after implementation of this study protocol which can lead to serious complications and requires further study before conclusions can be drawn regarding optimal diuretic regimens.
Acknowledgments Alexa Harris PA-C and Grace Golan PA-C for their invaluable efforts in the surgical clinic follow up of these subjects. Dr. Ricardo Briceño for sharing his post-discharge diuretic regimen and follow up with us which informed our study design. We also appreciate consultations with Dr. Evonne Morrell regarding her experience with pericardial effusions during the quality improvement project she published.
Author contributions All authors contributed to the study conception and design. Material preparation, data collection, and analysis were performed in part by all authors. The first draft of the manuscript was written by Jamie Penk and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.