Early initiation of peritoneal dialysis improves postoperative recovery in children with right ventricular outflow tract obstructive lesions at high risk of fluid overload: a propensity score-matched analysis

Abstract

OBJECTIVES

Postoperative fluid overload is common in children after cardiac surgery, especially for those with right ventricular outflow tract obstruction, which is associated with poor outcomes. This study was conducted to investigate whether early peritoneal dialysis (PD) was associated with improved outcomes in these children at high risk of fluid overload.

METHODS

Between January 2010 and January 2015, a total of 2555 consecutive patients with right ventricular outflow tract obstruction underwent anatomical repair. Using empirical risk evaluation, 219 patients at high risk of fluid overload were identified. A propensity score matching was performed to correct the selection bias and identify the comparable patient groups: the early PD group, in whom PD was initiated within 6 h of admission in paediatric ICU, and the control group, without early PD. The mechanical ventilation time, vasoactive-inotropic score and time to negative fluid balance were compared in 45 matched patient pairs (totally 90).

RESULTS

After propensity matching, there were no statistically significant differences between the 2 groups in terms of demographics and preoperative characteristics. The early PD group had shorter mechanical ventilation time [median 49 h, interquartile range (IQR) 31–97 h vs median 76 h, IQR 55–166 h; P < 0.01]; lower vasoactive-inotropic score (median 17, IQR 16–21 vs median 22, IQR 18–26; P < 0.01); shorter duration of inotrope requirement (median 7 days, IQR 6–9 days vs median 8 days, IQR 7–13 days; P < 0.01); shorter time to negative fluid balance (median 20 h, IQR 13–34 h, vs median 48 h, IQR 40–74 h; P < 0.01) and a higher rate of negative fluid balance at 24 h (69% vs 29%, P < 0.01).

CONCLUSIONS

When compared with the control group, the early PD group showed shorter mechanical ventilation time, less inotropic requirement and lower time to attain negative fluid balance. On the basis of our empirical risk-evaluation practice, early PD could improve immediate postoperative recovery in children with right ventricular outflow tract obstruction.

INTRODUCTION

Postoperative fluid overload is common in children after cardiac surgery, especially for those with right ventricular outflow tract (RVOT) obstructive lesions [1]. The causes of fluid overload are multiple, including haemodynamic instability, massive fluid or blood/products administration, capillary leak, low cardiac output, longer cardiopulmonary bypass (CPB) time and/or acute kidney injury [2, 3]. All of these could lead to systemic and pulmonary tissue oedema, prolonging postoperative recovery time and increasing morbidity and mortality [4, 5].

Fluid therapy for the management in the immediate postoperative period is crucial, and many therapies have been developed to maintain postoperative negative fluid balance. Conventional fluid overload avoidance strategies include fluid restriction and diuretic therapy as well as inotropic support. When these fail, renal replacement therapy may be considered [6]. Placement and use of the peritoneal dialysis (PD) catheter is a safe, simple, well-established and low-cost procedure. It remains controversial that PD is associated with decreased mortality. However, recent evidence supports that PD may shorten the paediatric intensive unit stay rather than decreased early mortality. Therefore, it might be cost-effective in high-risk patients [7–12].

Data on the optimal timing of PD in paediatric cardiac surgery patients are still not well known. Patients undergoing reconstruction of RVOT are at risk for postoperative ventricular dysfunction, which contribute to capillary leak and interstitial fluid accumulation. On the basis of our empirical practice, CPB time, the vasoactive-inotropic score (VIS) and right ventricular pressure to left ventricular pressure (RVP/LVP) ratio at the end of CPB may be used to identify the patients at high risk for fluid overload. However, the optimal timing of PD is not clear for these high-risk patients in our centre. We, therefore, hypothesized that early PD is associated with improved immediate postoperative recovery in children with RVOT obstruction at high risk of fluid overload.

MATERIALS AND METHODS

Study population

This was a retrospective observational study including all patients with RVOT obstruction who underwent anatomical repair between January 2010 and January 2015. It was approved by the Ethics Committee of Fuwai Hospital.

The inclusion and exclusion criteria are as follows: all patients at high-risk of fluid overload, with RVOT obstructive lesions [tetralogy of Fallot (TOF), pulmonary artery atresia and ventricular septal defect (VSD) and the double-outlet right ventricle with pulmonary stenosis] and younger than 3 years were included in the study. The ‘high-risk category’ for fluid overload was based on an empirical practice developed since 2010 in our centre, in which patients met all 3 haemodynamic parameters at the end of CPB: postoperative right ventricular dysfunction (RVP/LVP ratio ≥ 0.65), requiring high inotropic support (VIS ≥ 20) and longer CPB time (CPB ≥ 120 min). Patients were excluded if born with a gestational age of less than 36 weeks, documented chromosomal abnormality, extracorporeal membrane oxygenation initiation, delayed sternal closure and a history of abdominal operation.

Timing of peritoneal dialysis

There were 2 different PD-initiation strategies to deal with these high-risk patients in our centre. Some patients preferred to adopt an aggressive PD strategy, who met the criteria of high-risk fluid overload based on the empirical practice. Dialysis was implemented directly in the operating room or in paediatric intensive care unit (PICU) within 6 h of admission. A relatively conservative PD placement based on clinical triggers such as positive fluid balance, oliguria or evidence of acute kidney injury was implemented in other patients. Given the different timing of PD, we, therefore, compared outcomes to determine whether early PD, based on present empirical practice, was associated with improved clinical outcomes.

Fluid management strategies

In the early PD group, the high-risk patients were placed with a PD catheter in the operating room or PICU within 6 h of admission, and everyone initiated dialysis in the immediate postoperative period. The paediatric Tenckhoff PD catheter (37 cm; Baxter Healthcare Co. Ltd, Guangzhou, China) was inserted into the abdominal cavity either intraoperatively and/or postoperatively via subumbilical and/or subxiphoid approach. The dialysate solutions of 2.5% and 4.25% dextrose, warmed prior to infusion, were connected to an enclosed system. Initially, PD was started with 10 ml/kg fill volume of 2.5% dextrose dialysate and unfractionated heparin 200 units/l. Started with 1 h standard cycle (5-min fill, 15-min dwell and 40-min drain), and then the intensivist would adjust the cycle from 1 to 4 h according to quantitative changes in urine output. Potassium chloride was infused by central venous catheter to attain a serum potassium concentration of 3.5–4.0 mmol/l. Dextrose concentration was increased to achieve net negative fluid balance with tolerable haemodynamics. Diuresis is simultaneously initiated with a low dose of furosemide at 0.05–0.1 mg/kg/h by continuous infusion. PD will be terminated when urine output >2 ml/kg/h with a net negative fluid balance for a minimum of consecutive 2 days and resolution of anasarca.

In the control group, the 1st-line therapy for these children is aggressive diuresis beginning with furosemide between 0.1 and 0.5 mg/kg/h, targeting net negative fluid balance. Peritoneal drainage was routinely implemented when patients presented with massive ascites by trocars (20 GA, BD Angiocath, Brazil). PD was only considered when the aggressive diuresis therapy failed.

Perioperative management

All the operations were performed by 4 comparable qualified senior surgeons in our centre, and surgical techniques were substantially same for all cases. For TOF, selection of the RVOT reconstruction technique was based on preoperative angiocardiographic and echocardiographic data, as well as the assessment of pulmonary valve structure and body surface area-specific pulmonary annulus diameter during the operation. If the pulmonary annulus and main pulmonary artery were more than 2–3 SDs below normal, a transannular patch would be indicated. Otherwise, the patient was accepted as a candidate for valve-sparing surgery. For pulmonary artery atresia VSD patients, a valved conduit would be used, either the valved bovine jugular vein or pulmonary homograft with valve. Modified ultrafiltration was performed routinely for all participants. Pressures were recorded in both right and left ventricular cavities by needle puncture after weaning CPB. Intraoperative transoesophageal echocardiography was commonly used to assess for the residual defect and the need to return to bypass. When the residual VSD ≥ 3 mm or RVOT gradient ≥ 40 mmHg, it often predicted the need for immediate reoperation. Under this condition, these cases would be excluded from the study.

All patients were transferred to PICU immediately after operation, and there was no treatment bias between these 2 groups. Fentanyl (5–10 μg/kg/h) and midazolam (100–200 μg/kg/h) were commonly used to achieve sedation and analgesia. Mechanical ventilation was triggered in a synchronized intermittent mandatory ventilation model, with PB840 ventilators (NellcorPuritanBennett Ireland, Galway, Ireland) and DrägerSavina ventilators (Drägerwerk AG & Co. KGaA, Lübeck, Germany). Blood pressure (BP), heart rate, central venous pressure and left atrial pressure were routinely monitored to evaluate cardiac function and circulatory blood volume. Dopamine (2–6 μg/kg/min) and milrinone (0.4–0.8 μg/kg/min) were the 1st-line agents to treat either mild (10–20% decrease in mean arterial BP for age) or moderate (20–30% decrease in mean arterial BP for age) hypertension. Epinephrine (0.03–0.1 μg/kg/min) should be considered in patients with severe ventricular dysfunction and hypertension (>30% decrease in mean arterial BP for age).

Total fluid output was calculated as the sum of urine output, PD ultrafiltrate, stool mass, nasogastric tube losses and chest tube output. The sum of crystalloids and colloids was defined as total fluid intake. The VIS is determined as follows: dopamine dose (µg/kg/min) + dobutamine dose (µg/kg/min) + [100× epinephrine dose (µg/kg/min)] + [10 × milrinone dose (µg/kg/min)] + [10 000 × vasopressin dose (U/kg/min)] + [100 × norepinephrine dose (µg/kg/min)] [13].

The primary outcome is mechanical ventilation time. Secondary outcomes are the VIS and time to negative fluid balance in our study.

Statistical analysis

We performed a matched propensity score (PS) analysis to assess treatment effects and reduce the impact of selective bias and potential confounding. PS 1-to-1 matching is utilized with the nearest-neighbour algorithm without replacement and a 0.02 caliper setting. Age, weight, surgeons, CPB, the VIS and RVP/LVP ratio at the end of the surgery are put into a logistic regression model to estimate the PS. We then derived, for each patient, the probability PS of initiating early PD and compared when exposed to patients in the control group. Following matching, absolute standardized differences are used to assess prematch imbalances and postmatch balance in baseline covariates. The best balance is reflected by a standardized difference below 10%. Analysis of matched pair outcomes is performed after the best balance among groups is obtained. The appropriate statistical methods are tested: categorical outcomes with the exact McNemar’s test, and continuous outcomes with either a paired t-test or the Wilcoxon sign-rank test depending on the distribution. Descriptive statistics is used to compare variables between the unmatched groups, and generally, continuous variables are expressed as mean ± standard deviation or median, interquartile ranges as appropriate and discrete variables as frequencies (n). Because of the long observation period, the Mann–Whitney U non-parametric test statistics is applied for comparisons as more robust. The 2-tailed χ² analysis or the Fisher’s exact test is used for categorical variables. A 2-sided P-value <0.05 is regarded as significant. All the statistical analyses are performed using the SPSS statistical software (IBM SPSS Statistics for Windows, version 22, IBM Corporation, Armonk, NY, USA).

RESULTS

Patients and data

During the period between January 2010 and January 2015, a total of 2555 patients with RVOT obstructive lesions (TOF 2092, pulmonary artery atresia/VSD 298 and the double-outlet right ventricle/PS 165) were hospitalized for anatomical repair, whereas 219 patients met the criteria for high-risk inclusion. Of 219 patients, 90 patients were included in this propensity score matching (PSM) analyses. One hundred and twenty-nine patients were excluded: 6 patients were of gestational age <36 weeks, 2 were with chromosomal abnormality, 15 had delayed sternal closure or extracorporeal membrane oxygenation, only 1 had abdominal operation before and remaining 105 patients were not included for the PSM analysis. Details were depicted in Fig. 1.

Clinical variables and outcomes

All 219 patients underwent RVOT reconstruction, in which transannular patch repair was performed in 77% of the patients. Fifty-three patients met the criteria for high-risk fluid overload, for whom dialysis was implemented directly during the immediate postoperative period, and a relatively conservative PD placement was implemented in 166 patients. Although we failed to set up the negative fluid balance in spite of aggressive diuretics in 73 (44%) children in the control group, we had to implement PD again over the next several days.

Table 1 lists the demographic data and baseline characteristics of patients prior and after matching. Before matching, there was no significant difference between the 2 groups except age, weight and surgeons. The control group showed a younger age and lower weight than those of the early PD group (P = 0.04 and P < 0.01, respectively). After PSM, 45 cases from each group were well matched by a 1:1 PSM algorithm. There was no significant difference among propensity-matched groups with regard to baseline characteristics. Six variables were also well balanced with a standardized difference of ≤10% (Table 2).

Of the 45 propensity matching pairs, we found that the decreased time of mechanical ventilation was observed in the early PD group (P < 0.01). The significantly lower maximum VIS and shorter duration of inotrope requirement were also observed in the early PD group during the 1st postoperative day. With respect to the time to achieve postoperative negative fluid balance, it was also shorter (P < 0.01). In contrast to the control group, the early PD group not only achieved negative fluid balance more rapidly but also maintained a higher rate of negative fluid balance throughout the 1st postoperative day (P < 0.01). The results showed a lower arterial lactate level in the early PD group on postoperative Day (POD) 1 (P < 0.01). Detailed postoperative outcomes are given in Table 3.

No serious complications such as intra-abdominal haemorrhage, visceral injury and peritonitis due to PD in study patients were noted. Only 1 patient in the early PD group along with 2 in the control group had moderate abdominal distention. Overall, 3 in-hospital deaths were recorded: 2 patients in the control group and 1 patient in the early PD group. In the control group, 1 died on POD11 due to multiple organ dysfunction syndromes and another patient died on POD5 because of acute respiratory distress syndrome. The patient in the early PD group died on POD6 because of low cardiac out syndrome.

DISCUSSION

To our knowledge, this is the 1st study to assess the benefits of the early PD following cardiac surgery in children with RVOT obstructive disorders based on risk factors assessment. This retrospective study showed an association with early PD with shorter duration of mechanical ventilation and decreased time to achieve a negative fluid balance along with less inotropic requirement when compared with the control group. Emerging evidence has suggested that persistent positive fluid balance after paediatric cardiac surgery is associated with poor outcomes, including longer mechanical ventilation time and inotropic support, as well as the requirement of renal replacement therapy and extracorporeal membrane oxygenation support [14–16]. Reportedly, fluid restriction, diuretics and PD comprise the mainstay of fluid balance management, and increasing literatures are suggesting that PD is a safe and effective alternative [7–12]. Many centres have adopted the use of PD for fluid management, but the indications and timing have not been well studied in different populations.

The timing of initiation of PD, as previous reported, was quite different. Alkan et al. [8] and Chan et al. [10] reported that PD should be performed if oliguria persists for more than 4 h despite medication, and in the absence of the established oliguria, PD should be performed with increased creatinine levels in association with one of the following: clinical signs of fluid overload, hyperkalaemia (5.5 mmol/l), persistent metabolic acidosis or low cardiac output syndrome. Sorof et al. [9] presented that indications for starting PD included physical evidence of total body fluid overload/anasarca, urine output, acid base or electrolyte disturbances and low cardiac output with renal insufficiency. Kwiatkowski et al. [17] proposed a much aggressive recommendation that prophylactic PD was initiated if age <3 months undergoing CPB, heart transplant <6 months, patients undergoing TOF repair <4 months and CPB >120 min. All these variables of indications or clinical practices may not be accurate to determine the timings of PD in CHD patients with RVOT obstructive lesions. Considering both systolic and diastolic right ventricular dysfunction following reconstruction of the RVOT surgery, they were prone to fluid overload due to capillary leak and interstitial fluid accumulation. To identify the high-risk fluid overload patients in the early stage, we developed an empirical risk assessment practice since 2010. We found that early PD based on this empirical risk-evaluation method could improve immediate postoperative outcomes in these patients.

In addition, our PD protocol was quite different from conventional PD prescriptions for a shorter dwelling time and a much longer draining time. The main reason is that we deem 15-min dwelling as able to adequately remove excessive fluid from the body. However, it may be not enough for the metabolite excretion in these patients. Also, longer draining time effectively reduces intra-abdominal fluid retention, which may increase intra-abdominal pressure. As shown, even though the early PD group had lower urine output on POD1, it had much more total output. Most of the negative fluid balance was mainly achieved by the PD catheter output from dialysis rather than urine output. As systemic oedema was alleviated, we began to prolong the draining time. Thus, urine output increased simultaneously on POD2. In other words, early PD had achieved a good renal-replacement effect or alleviated kidney burden during the oedema period.

Limitations

The major limitation of the study is its single-centre retrospective design. Because of the lack of randomization for observational design, selection bias inevitably arises vis-a-vis prospective studies. We addressed this by utilizing PS matching to avoid confounding; however, there is a possibility of bias due to unmeasured or unknown confounders, which can be addressed only by a randomized controlled trial. Further study is needed to replicate these findings in independent cohorts, thereby promoting the validity and generalizability of the observation. In addition, the fluid removal benefits from prophylactic PD in RVOT obstruction patients may not be eligible for other types of CHD [18, 19]. Ryerson et al. [20] just reported that prophylactic PD catheter placement did not decrease the time to achieve a negative fluid balance after the Norwood procedure. We thought that our empirical risk-evaluated PD practice might show much better results on children with right-side heart obstructive lesions. In addition, it merits further elaboration as to why these patients with RVOT obstructive disorders particularly benefit from early PD, given that many other patients are also quite prone to fluid overload.

Despite the limitations of a retrospective study, our experience showed that when compared with the control group, the early PD group was associated with lower time to attain negative fluid balance, less inotropic requirement and shorter mechanical ventilation time. Based on the empirical risk-evaluation practice, early PD improves immediate postoperative recovery in children with RVOT obstructive lesions.

Funding

This work was supported by the Special Scientific Research Fund of Public Welfare Profession of China [2016-F01].

Conflict of interest: none declared.

REFERENCES

Author notes