Patent Ductus Arteriosus Ligation in Extremely Preterm Infants and Death or Neurodevelopmental Impairment. Dany Weisz

Transcription

1 Patent Ductus Arteriosus Ligation in Extremely Preterm Infants and Death or Neurodevelopmental Impairment by Dany Weisz A thesis submitted in conformity with the requirements for the degree of Masters of Science. Institute of Health Policy, Management and Evaluation University of Toronto Copyright by Dany Weisz 2016

2 Patent Ductus Arteriosus Ligation in Extremely Preterm Infants and Death or Neurodevelopmental Impairment Dany Weisz Masters of Science Institute of Health Policy, Management and Evaluation University of Toronto 2016 Abstract Objective: Evaluate the association between patent ductus arteriosus (PDA) ligation vs. medical management and neonatal and neurodevelopmental outcomes. Methods: Retrospective cohort study of extremely preterm infants with PDA born between The primary outcome was death or neurodevelopmental impairment (NDI) at months. Secondary outcomes included death, chronic lung disease (CLD), and NDI. Multivariable logistic regression (MLR) analysis and marginal structural models (MSM) were used to adjust for perinatal and postnatal confounders. Results: Of 754 infants with PDA, 184(24%) underwent ligation. Compared with medically treated infants, ligated infants had similar odds of death/ndi (aor 0.83, 95%CI: ), NDI (aor 1.27, 95%CI: ), and CLD (aor 1.36, 95%CI: ), but lower mortality (aor 0.09, 95%CI: ). Conclusions: PDA ligation is not associated with adverse outcomes and may reduce mortality. Previously reported associations of ligation with increased morbidity are likely due to bias from confounding by indication, rather than a detrimental causal effect of ligation. ii

3 Acknowledgements Thank you to Dr. Prakesh Shah, my supervisor, for outstanding guidance and mentorship during all phases of this project. I also sincerely thank my thesis committee members: Dr. Lucia Mirea, Dr. Patrick McNamara, Dr. Joseph Kim and Dr. Linh Ly, for their constructive review, feedback and direction. Thank you to Dr. William Benitz (Stanford University) and Dr. Catherine Birken (University of Toronto) for their excellent reviews, to Dr. Paige Church and Dr. Ed Kelly for their collaboration, and to my colleagues in the Department of Newborn and Developmental Paediatrics, Sunnybrook Health Sciences Centre for their support. Most importantly, a loving thank you and appreciation to my wife, Sharon, and our children for their patience and enthusiasm. I am so grateful and this would not have been possible without your strong support. iii

4 Table of Contents Preamble Background and Rationale Patent Ductus Arteriosus: Anatomy and Closure Physiology and Continuum of the Ductal Shunt In Neonates Natural History of PDA in Preterm Infants Clinical Importance of the PDA in Preterm Infants Cerebral Intraventricular Hemorrhage Necrotizing Enterocolitis Chronic lung disease Neurodevelopmental Impairment Diagnosis of Haemodynamically Significant PDA Therapeutic Approaches to the PDA: Rationale and Options for Treatment Pharmacotherapeutic Agents for Ductal Closure Universal Medical Prophylaxis for Patent Ductus Arteriosus Early Medical Treatment of Asymptomatic PDA Diagnosed by Echocardiographic Screening Medical Treatment of Symptomatic PDA Surgical PDA Ligation Association of PDA Treatment and Adverse Outcomes Surgical Ligation and Health Outcomes: Methodological Issues in Previous Studies Confounding by Indication Selection Bias The PDA Ligation Decision: Uncertainty, Practice Variability and the Urgent Need for Clarity Methods Design and Setting Management of PDA Outcomes and Assessment Data Sources and Collection Potential Confounders Perinatal Covariates Postnatal Morbidities Occurring Prior to Ductal Closure Statistical Analyses Multivariable Logistic Regression Marginal Structural Models Marginal Structural Models: Background Assumptions of Marginal Structural Models Marginal Structural Models, Stage 1 Analyses: Estimation of Stabilized Inverse Probability of Treatment Weights Marginal Structural Models, Stage 2 Analyses: Weighted Estimation of the Effect of Ligation on Outcomes Subcohort Analyses to Reduce Selection and Information Bias Infants Lost To Follow-up Sample Size iv

5 2.7 Research Ethics and Data Sharing Results Multivariable Logistic Regression Marginal Structural Models Infants Lost to Neurodevelopmental Follow-up Discussion Main Findings and Comparison to Previous Literature Strengths Limitations Implications for Practice Implications for Future Research Conclusions Appendix 1: Data Collection Form Appendix 2: Covariate Definitions Appendix 3: Multivariable Logistic Regression Model Results (full cohort, n=754) Appendix 4: Missing Data Appendix 5: Marginal Structural Model Stage 1 Results Pooled Logistic Regression Analyses (full cohort, n=754) MSM Numerator Term: No covariates MSM Denominator Term: Antenatal and Perinatal Covariates Only MSM Denominator Term: Antenatal, Perinatal and Postnatal Covariates Arising Prior to Ductal Closure MSM Weighted Logistic Regression Analysis: Ligation vs. Medical Treatment Distribution and Q-Q plot of the stabilized inverse probability of treatment weights (sweight) for each outcome adjusted for antenatal, perinatal and postnatal covariates v

6 . List of Tables Table 1: Neonatal and neurodevelopmental outcomes from previous studies for infants with a patent ductus arteriosus treated with surgical ligation compared with medical management only. Table 2: Classification of the severity of neurodevelopmental impairment in infants. Table 3: Echocardiographic classification of PDA haemodynamic significance. Table 4: Antenatal and perinatal characteristics of cohort of extremely preterm infants with clinically and echocardiographically significant patent ductus arteriosus. Table 5: Morbidity arising during the NICU course, period of ductal patency (prior to surgical ligation or medical closure), and during the at-risk period for surgical ligation in the full cohort of infants (n=754). Table 6: Neonatal and neurodevelopmental outcomes of ligated vs. medically treated infants (full cohort, n=754) estimated using multivariable logistic regression. Table 7: Antenatal and perinatal characteristics of ligated infants who had previously failed cyclooxygenase inhibitor therapy and medically treated infants with persistent haemodynamically significant PDA after cyclooxygenase inhibitor therapy (n=308). Table 8: Morbidity arising during the NICU course, period of ductal patency (prior to surgical ligation or medical closure), and during the at-risk period for surgical ligation for ligated infants who had previously failed cyclooxygenase inhibitor therapy (COXI) and the subgroup of medically treated infants with a haemodynamically significant PDA after cyclooxygenase inhibitor therapy. vi

7 Table 9: Neonatal and neurodevelopmental outcomes of ligated infants who had previously failed cyclooxygenase inhibitor treatment vs. medically treated infants with haemodynamically significant patent ductus arteriosus after cyclooxygenase inhibitor therapy (n=308), estimated using multivariable logistic regression. Table 10: Associations of ligation vs. medical treatment and the primary composite outcome of death or moderate-severe neurodevelopmental impairment, and the secondary outcome of death before discharge from NICU among cohorts restricted to survivors by postnatal age. Table 11: Association between surgical ligation and neonatal and neurodevelopmental outcomes for the entire cohort of infants (n=754) with a clinical and echo diagnosis of PDA, estimated using stabilized inverse probability of treatment weights and marginal structural models. Table 12: Association between surgical ligation and neonatal and neurodevelopmental outcomes for the subgroup of infants (n=308) comprising ligated infants who had previously failed cyclooxygenase inhibitor (COXI) treatment and medically treated infants with an echocardiography confirmed HSPDA after COXI treatment, estimated using stabilized inverse probability of treatment weights and marginal structural models.* Table 13: Mean, standard deviation and Kolmogorov-Smirnov test results for stabilized weights for each study outcome. Table 14: Antenatal and perinatal characteristics, postnatal morbidity, ductal characteristics and neonatal outcomes of surviving infants based on completion of neurodevelopmental evaluation at follow-up. Table 15: Comparison of pooled adjusted odds ratios from previous meta-analysis where studies adjusted for antenatal and perinatal covariates only vs. adjusted odds ratios for primary and secondary outcomes from the current study. vii

8 List of Figures Figure 1: Directed acyclic graph depicting the relationship between physiological instability and systemic inflammation, PDA ligation and the outcomes of death, severe neonatal morbidities and neurodevelopmental impairment. Figure 2: Flow diagram of infants included in the study. Figure 3: Kaplan-Meier analysis of time to PDA closure of medically (red) versus surgically (blue) treated infants over the first 10 weeks of life. Figure 4: Average daily mean airway pressure (cmh 2 O) with 68% and 95% confidence intervals (based on the standard error) over the first 40 days of life for medically treated (blue solid line) vs. ligated infants (red dashed line) prior to ductal closure. Infants no longer contributed data after the date of ductal closure, leading to the widening of confidence intervals over time as the number of infants with persistent haemodynamically significant PDA diminished with time. The median date of ligation was day of life (DOL) 29 (vertical interrupted line), with interquartile range [DOL 22, DOL 38] (solid grey box). The earliest date of ligation was on DOL 7. Figure 5: Average daily mean airway pressure (cmh 2 O) with 68% and 95% confidence intervals (based on the standard error) over the first 40 days of life and prior to ductal closure for the subcohort of medically treated infants with echo-proven significant PDA after cyclooxygenase inhibitor treatment (blue solid line) vs. ligated infants who had previously failed cyclooxygenase inhibitor treatment (red dashed line). Infants no longer contributed data after the date of ductal closure, leading to the widening of confidence intervals over time as the number of infants with persistent haemodynamically significant PDA diminished with time. The median date of ligation was day of life (DOL) 29 (vertical interrupted line), with interquartile range [DOL 22, DOL 38] (solid grey box). The earliest date of ligation was on DOL 7. Figure 6: Kaplan-Meier analysis of time to PDA closure for ligated infants who had previously failed cyclooxygenase inhibitor therapy vs. medically treated babies who had viii

9 echocardiographically significant PDA after cyclooxygenase inhibitor treatment but who were not treated with ligation. Figure 7: Kaplan Meier analysis of survival in ligated vs. medically treated infants in the subcohort with echocardiography proven HSPDA after COXI therapy that were alive at day 20 of life. Infants were censored if discharged home or transferred to a community NICU. Ligated infants (top red line) had significantly greater survival than medically treated infants (bottom blue line) (Log-Rank test, p <0.0001). ix

10 List of Appendices Appendix 1: Data Collection Form Appendix 2: Covariate Definitions Appendix 3: Multivariable Logistic Regression Model Results Appendix 4: Missing Data Appendix 5: Marginal Structural Model Results x

11 List of Abbreviations ASQ BSID BW CGA COXI CLD CP GA GMFCS HSPDA HUS IPTW IUGR IVH MLR MSM NDI NEC NICU OR PDA RCT ROP SD SGA SNAP II VCP VILI Ages and Stages Questionnaire Bayley Scales of Infant Development birthweight corrected gestational age cyclooxygenase inhibitors chronic lung disease cerebral palsy gestational age Gross Motor Functional Classification System haemodynamically significant patent ductus arteriosus head ultrasound inverse probability of treatment weight intra-uterine growth restriction intraventricular hemorrhage multivariable logistic regression marginal structural model neurodevelopmental impairment necrotizing enterocolitis neonatal intensive care unit odds ratio patent ductus arteriosus randomized control trial retinopathy of prematurity standard deviation small for gestational age Score for Neonatal Acute Physiology II vocal cord paresis ventilator induced lung injury xi

12 Objective: Does patent ductus arteriosus ligation in preterm infants increase death or neurodevelopmental impairment compared with medical management? Preamble At the end of 20 th century, the clinical management of patent ductus arteriosus (PDA) in extremely preterm infants was commonly characterized by aggressive treatment aimed at achieving rapid postnatal closure, often within the first two weeks of life. The strong association between PDA and adverse outcomes had fostered the ideology that an infants' duration of exposure to ductal shunting should be minimized. Pharmacotherapeutic prophylaxis was frequently administered at birth, and early echocardiographic screening and pharmacological treatment aimed at ductal closure was provided. Infants with respiratory failure who demonstrated clinical and echocardiographic signs of PDA uniformly received medical treatment, and surgical ligation was promptly performed if medical closure failed or was contraindicated. Over the past decade, several large retrospective cohort studies have associated PDA ligation with increased neonatal and neurodevelopmental morbidity, including chronic lung disease, retinopathy of prematurity, cerebral palsy, cognitive deficits and hearing and visual impairments. The publication of these studies has been associated with a secular trend toward a reduction in treatment with surgical ligation for persistent symptomatic PDA. While the collective impact of these studies has been to prompt concern about causing harm by treating with surgical ligation, significant methodological shortcomings limit the validity of these studies, especially residual bias due to confounding by indication. As a result, neonatologists and paediatric cardiac surgeons frequently face significant controversy with the PDA ligation decision; clinicians must navigate high impact literature associating PDA surgery with increased morbidity, yet fraught with biases that make the true risks, and any benefits, of ligation uncertain. 1

13 1 Background and Rationale 1.1 Patent Ductus Arteriosus: Anatomy and Closure The ductus arteriosus (DA) is the fetal vascular connection between the main pulmonary artery and the descending aorta. The DA is one of several normal developmental mechanisms that divert oxygen-replete blood away from the high resistance pulmonary circuit to the systemic circulation. After birth, an abrupt decrease in circulating prostaglandins combined with increased arterial oxygen tension leads to ductal vasoconstriction and DA closure in nearly all term infants within the first week of life. In contrast, delayed postnatal closure of the ductus arteriosus, termed patent ductus arteriosus (PDA), occurs in up to 60% of infants born at less than 29 weeks gestational age. 1, 2 In preterm infants, the anatomical characteristics of the DA and physiological pathways that facilitate ductal closure are immature. Relative to the muscular DA that is groomed for rapid vasoconstriction in term infants; the preterm DA is comparatively less muscular and has reduced sensitivity of the metabolic pathways of ductal vasoconstriction and closure. 1.2 Physiology and Continuum of the Ductal Shunt in Neonates During fetal life, low systemic vascular resistance (SVR) due to the low resistance placenta, combined with elevated pulmonary vascular resistance (PVR) result in pulmonary artery to aorta ('right to left') flow across the ductus arteriosus (DA). During normal neonatal transition, increased SVR associated with umbilical cord clamping occurs alongside a decrease in PVR precipitated by ventilation in air and increased arterial pressure of oxygen (PaO 2 ). The previously right-to-left ductal shunt becomes balanced (symmetrically bidirectional) at 5 minutes after birth, mostly left to right by minutes, and entirely left-to-right by 24 hours of age. 3, 4 2

14 In preterm neonates, the ductal shunt is variable in direction, reflecting the effects of underlying disease states on pulmonary and systemic haemodynamics. The shunt may be conceptualized as residing on a continuum between life-saving conduit, neutral bystander and pathological entity. In infants with critical congenital heart disease or myocardial dysfunction, patency of the ductus arteriosus (DA) may be necessary for adequate pulmonary blood flow (e.g., tricuspid atresia) or systemic blood flow (e.g., critical aortic stenosis). In severe persistent pulmonary hypertension of the newborn (PPHN), the postnatal failure of optimal vasorelaxation of pulmonary arterioles (e.g., due to asphyxia, respiratory distress syndrome, etc) results in persistently high PVR and persistence of the right-to-left ductal shunt. The right-to-left shunt may reduce right ventricular afterload and support post-ductal systemic blood flow, albeit with deoxygenated blood. A bidirectional shunt in milder cases of PPHN may play a neutral role, merely permitting the noninvasive estimation of the systemic-pulmonary pressure gradient. If the ductus arteriosus remains patent after birth, preterm infants who experience the expected fall in PVR may be susceptible to the effects of a large systemic-to-pulmonary (left-to-right) shunt. Blood flows across the PDA continuously in systole and diastole, resulting in volume overload of the pulmonary artery, pulmonary veins, and left heart. Shunt volume (Q) is directly proportional to 4 th power of the ductal radius (r), and the aorto-pulmonary pressure gradient (, and is inversely proportional to the ductal length (L) and blood viscosity ( ). Increased pulmonary blood flow may lead to alveolar edema, reduced pulmonary compliance, increased need for respiratory support, and pulmonary hemorrhage. Increased blood flow to the left heart results in dilatation and increased end-diastolic pressures in the left ventricle and atrium. Ductal diastolic 'steal' from the descending aorta, shorter diastolic (and coronary perfusion) times due to tachycardia, and increased myocardial oxygen demand may result in subendocardial ischemia and mesenteric hypoperfusion. 3

15 1.3 Natural History of PDA in Preterm Infants The natural history of PDA in preterm infants has been described in small cohorts of infants for who echocardiographic assessment of PDA was performed but no pharmacological or surgical treatment was administered. While the merits of exclusively conservative management of the PDA cannot be elucidated without an adequate comparator group (of infants treated for PDA), these studies provide insight into the short and long-term likelihood of spontaneous closure. In a retrospective cohort study of 103 extremely preterm infants who did not receive PDA treatment, 91 survived beyond 72 hours, of whom 70 had an echocardiographic diagnosis of PDA on day 3 of life. 5 Of these, 51 (73%) experienced spontaneous ductal closure prior to discharge home. In very low birthweight infants discharged home from the NICU with PDA, most (87 93%) infants undergo spontaneous closure in infancy 6 or early childhood Clinical Importance of the PDA in Preterm Infants Although the natural history of the PDA is toward spontaneous closure, the clinical importance of the PDA in preterm infants is underscored by its incidence and consistent association with adverse neonatal outcomes. PDA is common in preterm infants, occurring on the third day of life in up to 65% of preterm infants born at gestational age < 29 weeks. 8 Persistent patency on day 3 of life identifies infants at higher risk of all major complications and morbidities of prematurity, including death, intraventricular hemorrhage (IVH), chronic lung disease (CLD), necrotizing enterocolitis (NEC) and retinopathy of prematurity (ROP). 8 While the association of PDA with adverse outcomes is strong, causation remains unestablished. Clinical trials aimed at facilitating closure of the symptomatic PDA have failed to demonstrate a reduction in severe morbidities of prematurity. However, inconsistency across studies in the definition of PDA haemodynamic significance has led to concerns of reduced validity Cerebral Intraventricular Hemorrhage Early neonatal cardiorespiratory instability is associated with germinal matrix bleeding and potential extension into the ventricular system (intraventricular hemorrhage (IVH)) and/or 4

16 periventricular hemorrhagic infarction (PVHI). Mild and severe IVH are associated with progressively higher odds of moderate-severe NDI compared with no IVH. 10 Cerebral ischemiareperfusion injury is a possible pathophysiology. Most (90%) IVH occurs in the first week of life, corresponding with the emergence of the left-to-right PDA shunt and associated increased left heart volume loading and cerebral (pre-ductal) perfusion. The coincidental timing of PDA and IVH suggests that the PDA may contribute to reperfusion injury in preterm infants. A potential causal role for PDA is supported by trials that have repeatedly and conclusively demonstrated a reduction in IVH and symptomatic PDA with the administration of prophylactic indomethacin, possibly by mitigating the emergence of a significant ductal shunt. However, prophylactic ibuprofen has been demonstrated to facilitate early ductal closure without reducing IVH. The disparity between indomethacin and ibuprofen in IVH prevention suggests that either early PDA closure is not the causal mechanism of IVH prevention or that differences in the administration of the COXI (eg. timing, dose) may result in similar early PDA closure but a differential impact on IVH. Taken together, both the mechanism of action of indomethacin in IVH prevention and a causal relationship between PDA and IVH remain speculative. 11, Necrotizing Enterocolitis Necrotizing enterocolitis (NEC) is an inflammatory intestinal condition of preterm infants with a multifactorial pathophysiology influenced by genetic predisposition, intestinal immaturity and ischemia, imbalance in microvascular tone, abnormal intestinal microbial colonization, and mucosal immunoreactivity. 13 Infants with severe NEC develop a profound systemic inflammatory response. Mortality is high and survivors have a high incidence of neurodevelopmental impairment. 14 A causal role for PDA in the development of NEC is supported by results of a randomized controlled trial of early prophylactic surgical PDA ligation vs. conservative management which demonstrated reduced NEC in the interventional arm, though the incidence of NEC in the noninterventional arm was atypically high. 15 Subsequent observational studies have strongly associated PDA with the development of NEC in extremely preterm infants. 8, 16, 17 However, 5

17 placebo-controlled trials of PDA treatment have failed to demonstrate a reduction in NEC despite achieving ductal closure. 18 The pathophysiological role of PDA in the development of NEC is not well understood, but is thought to be mediated by intestinal hypoperfusion. Patent ductus arteriosus is associated with diastolic flow reversal in the abdominal aorta, celiac artery and superior mesenteric artery, and with reduced mesenteric tissue oxygenation Chronic lung disease Preterm infants born before 32 weeks gestational age who require positive pressure ventilation or supplemental oxygen at 36 weeks corrected gestational age are assigned a diagnosis of moderateto-severe chronic lung disease. A large ductal shunt leads to pulmonary overcirculation, alveolar edema, decreased pulmonary compliance and an increased need for invasive mechanical ventilation. Longer exposure to the ductal shunt and larger PDA shunt volumes, as assessed by echocardiography, have been associated with increased mortality and CLD, supporting a pathological role of the PDA. 20, 21 However, placebo-controlled trials of PDA treatment have failed to demonstrate a reduction in CLD despite achieving ductal closure. Although trials have been criticized for suboptimal patient selection and open-label treatment, these results suggest that ductal closure may not modify the increased risk of CLD associated with PDA Neurodevelopmental Impairment Among infants born extremely preterm, death or neurodevelopmental impairment (NDI) is estimated to occur in up to 50% NDI is commonly defined as a composite outcome which includes neuromotor impairment (e.g., cerebral palsy), neurocognitive impairment (cognitive or 25, 26 language delay) and neurosensory impairment (hearing, vision or both). Cerebral injury/dysmaturation and subsequent NDI is likely the final common pathway after cardiorespiratory instability leads to hypoxia-ischemia-reperfusion injury, inflammation, or arrested development of sensitive, immature white and grey matter (Figure 1) Isolated 6

18 cerebral specific injury (e.g., isolated arterial thrombo-embolic stroke, meningitis without associated sepsis) is uncommon. Predictive models have identified risk factors for death and NDI at different time points during NICU care. At birth, GA, birthweight (BW), multiple gestation, antenatal corticosteroids, intrauterine growth restriction (IUGR) and gender are the most important prognostic perinatal risk 23, 36 factors for death and NDI. Postnatally, large PDA and sepsis are significantly associated with NDI. 37, 38 At the time of NICU discharge, neonatal morbidities of prematurity such as CLD, retinopathy of prematurity (ROP), sepsis and major brain injury can reliably predict NDI at 18-23, months corrected age. Figure 1: Directed acyclic graph depicting the relationship between physiological instability and systemic inflammation, PDA ligation and the outcomes of death, severe neonatal morbidities and neurodevelopmental impairment (From Weisz and McNamara, J Clin Neo ). BPD, bronchopulmonary dysplasia; ROP, retinopathy of prematurity 7

19 1.5 Diagnosis of Haemodynamically Significant PDA Clinical examination is a common mechanism for the initial suspicion of PDA in preterm infants. The clinical signs of PDA in preterm infants are related to the physiological effects of left heart volume loading and diastolic 'steal' from the aorta to the pulmonary artery. The precordium is active and a holosystolic murmur is present, often loudest at the upper left sternal border. This may be accompanied by a wide pulse pressure, bounding peripheral pulses and diastolic hypotension. Increased pulmonary blood flow reduces pulmonary compliance leading to a need for more supplemental oxygen, increased work of breathing and ventilator support. Echocardiography (echo) is the primary method for the definitive diagnosis of haemodynamically significant PDA (HSPDA). The PDA can be easily and reliably imaged and PDA severity may be classified as haemodynamically significant or non-significant. In general, the haemodynamic significance of the PDA can be considered the interplay between the ductal shunt volume and the compensatory capacity of the systemic and pulmonary circulations. Studies have demonstrated that PDA size 1.5mm on the first day of life predicts a subsequent symptomatic PDA and correlates well with Doppler flow pattern in assessments of haemodynamic significance. 44 The narrowest PDA diameter is typically recorded on all echocardiograms and ductal diameter < 1.5mm is generally considered small and not haemodynamically significant. Echo indices of left heart volume loading, such as left ventricle and left atrium dilatation, and ratio of the left atrium to the aortic root (LA:Ao) correlate with the need for treatment. LA:Ao ratio > 1.4 has a high sensitivity for HSPDA. With each echocardiogram performed, a summary classification for the PDA haemodynamic significance (large, moderate, or small) is provided that incorporates echo data such as ductal size and flow pattern, indices of left heart volume loading and pulmonary overcirculation and systemic arterial diastolic flow reversal. The echocardiographic significance of the PDA, and 20, 21, 45 longitudinal exposure to a HSPDA correlates with key neonatal outcomes such as CLD. 8

20 1.6 Therapeutic Approaches to the PDA: Rationale and Options for Treatment The consistent association of PDA with adverse outcomes has been the impetus for treatment aimed at ductal closure despite a lack of evidence from clinical trials to demonstrate that pharmacological or surgical ductal closure mitigates these outcomes. Methods to close or minimize the effects of a PDA include conservative management (e.g. fluid restriction, diuretics, ventilation strategies), cyclooxygenase inhibitors (COI) such as indomethacin or ibuprofen, or surgical ligation Pharmacotherapeutic Agents for Ductal Closure Failure or delay of ductal closure in preterm infants occurs, in part, due to elevated early postnatal levels of circulating prostaglandin E 2 (PGE 2 ), which is produced from membrane phospholipids by enzymes such as prostaglandin H synthase (PGHS) complex, and prevents ductal vasoconstriction. Pharmacotherapeutic strategies aimed at ductal closure have targeted the cyclooxygenase (COX) and peroxidase (POX) moieties of PGHS to effect a reduction in circulating prostaglandins and PDA closure. Indomethacin and ibuprofen are the most studied and commonly used cyclooxygenase inhibitors (COXI) to facilitate ductal closure. Commonly used treatment regimens include indomethacin 0.2mg/kg intravenously (IV) every 12 hours for three doses and ibuprofen 10mg/kg followed by two additional 5mg/kg doses 24 hours apart. Ibuprofen (administered via intravenous or oral routes) has similar efficacy in ductal closure as indomethacin, but with fewer adverse events such as renal insufficiency or necrotizing enterocolitis, though these results may have been influenced by heterogeneity in GA, dosages and route of administration. 47 Prolonged courses of indomethacin, or the administration of a total indomethacin dose > 0.6mg/kg has been associated with increased necrotizing enterocolitis, without improved rates of ductal closure. 48 Oral and 47, 49, 50 intravenous ibuprofen have similar efficacy for ductal closure and adverse events. COXI administration for PDA closure has been associated with a transient reduction in cerebral and splanchnic blood flow, oliguria, weight gain, hyperbilirubinemia and gastrointestinal injury. 9

21 Therefore, renal failure, severe jaundice, intestinal perforation and necrotizing enterocolitis are contraindications to COXI administration. Since 2012, two randomized controlled trials of acetaminophen have demonstrated similar efficacy to ibuprofen in the early treatment of PDA in preterm infants, though trials have 51, 52 enrolled relatively mature infants Universal Medical Prophylaxis for Patent Ductus Arteriosus Patent ductus arteriosus prophylaxis is defined as the routine administration of a COXI within the first 6-12 hours of life without the use of screening echocardiography. Prophylactic indomethacin (0.1mg/kg daily for three days) reduces the incidence of severe intraventricular hemorrhage, periventricular leukomalacia, pulmonary hemorrhage, symptomatic PDA and the need for surgical PDA ligation. 11, 53 However trials evaluating prophylactic indomethacin employed 'back-up' treatments for subsequent symptomatic PDA, and therefore only provide a comparison between prophylactic and symptomatic treatment vs. symptomatic treatment alone. The broader use of prophylactic indomethacin decreased after publication of the Trial of Indomethacin Prophylaxis in Preterm infants (TIPP) which reported no difference in the primary outcome of death or neurodevelopmental impairment at 18 months corrected postnatal age in extremely low birthweight infants. 43 However, the clear short-term benefits, coupled with a lack of demonstrable harm at long-term follow-up and the known suboptimal reliability of neurodevelopmental testing at 18 months, has led to its continued use in some centres. Studies that have followed infants to school age have shown an improvement in neurodevelopment in boys who received indomethacin prophylaxis. 54 A recent small randomized placebo-controlled trial reported that prophylactic acetaminophen accelerated PDA closure in moderately preterm infants, but did not affect neonatal outcomes

22 1.6.3 Early Medical Treatment of Asymptomatic PDA Diagnosed by Echocardiographic Screening Given the high rate of spontaneous ductal closure (30-40% in preterm infants born at less than 29 weeks gestational age), indiscriminate administration of indomethacin prophylaxis has been criticized for subjecting a large minority (up to 46%) of infants to treatment for whom there is no benefit. 53 An alternative strategy is to employ early echocardiographic screening for PDA with the targeted administration of treatment to infants at risk of failure of spontaneous early ductal constriction. Recent trials have reported that early asymptomatic treatment reduces the risk of pulmonary hemorrhage and subsequent symptomatic PDA, but without improvement in neonatal outcomes. Van Overmeire et al. randomized 127 preterm infants (GA weeks) with PDA diagnosed by echocardiography to early (Day 3) or late (Day 7) treatment with indomethacin. The infants treated early had greater PDA closure but incurred more short-term side effects (lower urine output, higher serum creatinine, higher FiO 2 and mean airway pressure) and adverse events (composite of mortality, NEC, or cystic periventricular leukomalacia (PVL)). 56 Another placebo-controlled trial of targeted early PDA treatment with indomethacin after echocardiographic screening in the first 12 hours of life demonstrated a reduction in early pulmonary hemorrhage in the treatment group, but there was no difference in the rate of all pulmonary hemorrhage and the study was underpowered to detect differences in neonatal outcomes Medical Treatment of Symptomatic PDA For infants with clinical signs of PDA (murmur, wide pulse pressure, active precordium, cardiomegaly, etc), contemporary trials have primarily compared the timing of therapy (early [Day 2-5] vs. moderately late [day 7-14]). This comparison was performed by randomizing infants to early COXI treatment vs. placebo, with later treatment as 'backup' for persistent PDA. Unfortunately, this methodology does not permit the evaluation of PDA treatment vs. no treatment, and thus the benefits/sequelae of treatment remain uncertain. Recent trials have evaluated whether infants should be treated at the first clinical signs of PDA vs. a delayed approach with later treatment. Overall, the early treatment of clinically symptomatic PDA (at Day 3) increases PDA closure rates but may also increase adverse effects 11

23 without improving neonatal outcomes (e.g., mortality, CLD), compared with moderately late treatment at Day 7 to In addition, treatment in the 'late' group was more selective, with only 45% of control infants receiving open-label treatment. The efficacy of PDA treatment after the first 2 weeks of life, and its impact on neonatal outcomes, has not been evaluated Surgical PDA Ligation Surgical PDA ligation provides definitive ductal closure, and is usually only considered when medical treatments have either failed or were contraindicated. 59 Factors associated with the decision to treat a PDA with surgical ligation, include clinical signs of respiratory failure and/or systemic hypoperfusion, and echocardiographic criteria. 60, 61 Failure to wean an infant with PDA from invasive mechanical ventilation is the sine qua non of considering referral for surgical ligation. Ligation is performed under general anaesthesia with the infant supported by invasive mechanical ventilation. The surgical approach comprises a left lateral thoracotomy, retraction of the left lung, and the application of a ligature or clip to the PDA. Surgical mortality is low. 62 Immediate surgical complications include bleeding, pneumothorax, chylothorax, inadvertent ligation of a primary bronchus or branch pulmonary artery and vocal cord paresis (VCP). Left VCP occurs due to intra-operative injury to the left recurrent laryngeal nerve in 5-50% of infants. 186,187 Post-operatively, preterm infants are at risk of a low cardiac output state known as 'post ligation cardiac syndrome' (PLCS) which likely occurs due to increased LV afterload. 1.7 Association of PDA Treatment and Adverse Outcomes Over the past decade, several large cohort studies have reported increased neonatal morbidity and NDI in early childhood among infants treated with PDA ligation compared to medical management alone. 25, 26, A recent systematic review and meta-analysis of randomized trials and controlled observational studies demonstrated higher CLD, ROP, and NDI in ligated compared to medically-treated infants. 68 In contrast, mortality was lower in ligated compared to medically-treated infants

24 In a large retrospective cohort study of preterm infants born < 32 weeks GA with a symptomatic PDA, Mirea et al compared neonatal outcomes according to PDA treatment assignment. 63 After adjustment for antenatal and perinatal confounders, infants treated with surgical ligation had lower mortality but higher odds of CLD and ROP, compared with infants treated with medical management alone (Table 1). Similarly, in a retrospective review of 426 extremely low (<1000g) birthweight (ELBW) infants, Kabra et al detected higher CLD and ROP in 110 infants who underwent surgical ligation compared to 316 infants who received medical management only (Table 1). 25 Higher CLD among infants with PDA ligation was also detected by Madan et al 26 in a review of 2,838 extremely low birthweight infants with treatment for a symptomatic PDA. In a post-hoc secondary analysis of a small randomized controlled trial 15 (RCT) that compared early prophylactic surgical ligation (95% ligated) with delayed selective ligation (56% ligated) in ELBW infants requiring supplemental oxygen, Clyman et al. 65 found higher risk of moderate to severe CLD among infants treated with prophylactic compared to selective ligation (48% vs. 21%, p<0.05). 65 An association of PDA ligation with increased NDI has been detected by Madan et al 26, Kabra et al 25, Bourgoin et al 66 and Janz-Robinson et al 67. (Table 1). In these large retrospective cohort studies, infants were categorized by treatment assignment (conservative management, COXI only, or surgical ligation) and outcomes compared between treatment groups. Surgically treated infants had more neurodevelopmental impairment compared with medically or conservatively treated infants. Retrospective epoch studies have similarly associated ligation and NDI by demonstrating improvement in neurodevelopmental outcome after units reduced the proportion of infants referred for surgical ligation. Wickremasinghe et al. evaluated the months neurodevelopmental outcomes after moving from an early, aggressive surgical approach (infants with echocardiographic PDA after medical therapy were immediately referred for surgery) to a delayed selection ligation approach (infants with a persistent PDA were referred for ligation if the PDA was clinically and echocardiographically significant). Infants in the selective ligation 13

25 epoch were less likely to be treated with surgery (66% vs. 100%) and had less NDI (aor 0.07, 95% CI ). 69 Aspects of PDA ligation that have been postulated to contribute to the risk of NDI include surgical and anaesthesia effects, and post-operative haemodynamic compromise. Vocal cord paresis is a common surgical complication and is associated with an increased risk of death, extubation failure and chronic lung disease, need for gastrostomy tube, and gastroesophageal reflux disease. 62, Recent studies have associated use of halothane gases for anaesthesia in young children with NDI. 73, 74 Preterm infants are at risk of post-operative hypotension and 60, 61, cardiogenic shock due to PLCS, which may result in cerebral hypoperfusion and injury. In light of concerns regarding NDI and neonatal morbidities, the safety of PDA ligation has recently been questioned. 69, Concerns regarding the association of ligation and increased CLD, severe ROP and NDI have driven a secular trend toward a reduction in infants being treated with surgical ligation in North American centres. 84, 85 However, some studies have estimated lower mortality among infants with a PDA treated with ligation compared to medical management alone (Table 1). 14

26 Table 1: Neonatal and neurodevelopmental outcomes reported for infants with a patent ductus arteriosus treated with surgical ligation compared with medical management only. Study Kabra 2007* Madan 2009 Characteristics ELBW infants with symptomatic PDA PDA ligation (n=110) Medical only (n=316) ELBW infants with PDA Primary ligation (n=135), Indo only (n=1525), Indo and ligation (n=775) No treatment (n=403) Clyman Post-hoc analysis of RCT comparing 2009 early prophylactic ligation (n=40) vs. delayed selective ligation (n=44) in ELBW infants Mirea 2012 Janz- Robinson 2015ǁ Bourgoin 2016 Infants with GA 32 weeks with a PDA. Conservative (n=577), Indo only (n=2026), Indo+ligation (n=626), Primary ligation (n=327) Infants with GA 28 weeks. No PDA or clinically insignificant PDA (n=826), Pharmacological treatment only (n=569), Ligation (n=78) Infants with GA 28 weeks with PDA. Conservative (n=505), Ibuprofen only (n=248), Ligation (n=104) Odds Ratios (95% Confidence Intervals) Death or NDI Death NDI Severe ROP CLD 1.55 ( ) 1.03 ( ) ( ) 0.46 ( ) 1.15 ( ) 0.41 ( ) ( ) 1.53 ( ) 2.20 ( ) ( ) 1.9 ( ) 1.91 ( ) 10.7 ( ) 1.81 ( ) 3.10 ( ) 3.79 ( ) 2.30 ( ) CLD, chronic lung disease; GA, gestational age; Indo, indomethacin; NDI, neurodevelopmental impairment; RCT, randomized controlled trial; ROP, retinopathy of prematurity; * Adjusted for antenatal steroids, gestational age at birth, sex, multiple births, mother's education, and total dose of indomethacin received per kg of bodyweight between birth and discharge from the study center 15

27 Data shown for Indomethacin and Ligation vs. Indomethacin only. Adjusted for center, gestational age at birth, birthweight, gender, prophylactic indomethacin, Apgar score, severe RDS, growth restriction, antenatal steroids, antenatal/postnatal infection, maternal marital status and maternal age. Unadjusted odds ratios computed from randomized controlled trial. Data shown for any ligation vs. no ligation. Adjusted for gestational age, antenatal steroids, multiple births, gender, small for gestational age, Score for Neonatal Acute Physiology II. This data was provided by the primary author and is unpublished. ǁ Compared ligation vs. no treatment (either no PDA or clinically insignificant PDA) for the outcome of bilateral blindness and cognitive impairment > 2 standard deviations below mean. Adjusted for SGA, antenatal corticosteroids, multiple gestation, gender and other perinatal confounders (not specified). Compared ligation vs. conservative management. Adjusted for gender, GA, birthweight Z-score, antenatal corticosteroids, gestational hypertension, clinical chorioamnionitis, Apgar score, place of hospitalization, place of birth, year of birth, delivery characteristics 16

28 1.8 Surgical Ligation and Health Outcomes: Methodological Issues in Previous Studies Observational studies to date have associated PDA ligation with lower mortality but increased neurodevelopmental impairment compared with medical management alone. The divergence of these competing outcomes (Table 1) may be explained by several possible situations: First, surgical ligation may improve the survival of infants with PDA but may be simultaneously neurologically detrimental. Second, ligation may improve the survival of infants with PDA, but the infants referred for ligation are at higher pre-ligation risk of NDI (confounding by indication and increased pre-ligation illness severity). Finally, the decreased mortality may be a spurious finding influenced by survival bias (where moribund infants with a PDA die before becoming eligible for ligation), and the increase in NDI may be either a true detrimental effect of ligation or the effect of confounding by indication Confounding by Indication A serious concern in observational studies is bias arising when treatment assignment is not independent of baseline prognostic factors. The methodology used in the observational studies described above suggests the authors did not adequately address sources of treatment selection bias. Multivariable analyses were conducted controlling for GA and other antenatal or perinatal covariates (Table 1). This set of covariates, if complete, would be sufficient to balance baseline prognostic factors for interventions that occur shortly after birth. However, PDA ligation typically occurs several weeks after birth, and the interval accumulation of PDA related and unrelated postnatal comorbidities influences both treatment assignment and outcomes. No studies to date have addressed this time-dependent confounding by indication that infants referred for ligation may be more 'ill' and have larger ductal shunts at the time of the decision to treat with surgery, compared with infants who are treated with medical management alone. Illness severity, characterized by postnatal morbidities such as IVH and sepsis, and parameters of 27, 29, 39, physiological instability such as hypotension predict both neonatal morbidities and NDI. 86, 87 Severe IVH is a true confounder as it is associated with both PDA ligation and NDI 88-92, and 17

29 is not on the causal pathway. Although IVH has been found to occur after treatment with ligation and indomethacin 93, most (90%) IVH occurs in the first week of life 94, preceding the timing of surgical ligation reported by most studies. 93, Other studies have demonstrated that IVH does not worsen after PDA treatment with indomethacin or ligation. 12, 99, 100 Additional potential postnatal confounders include hypotension, postnatal sepsis and NEC, which increase illness severity and are associated with both PDA ligation and death or NDI. 64, 87, These factors, however, can occur both before, during, or after PDA treatment and thus may be confounders in some infants, and intermediates in others. Therefore, it is necessary to obtain data on the timing of hypotension, IVH, NEC and sepsis, relative to surgical ligation to correct for possible bias due to these potential confounders in multivariable analyses examining the impact of surgical PDA ligation. Taking these time-dependent covariates into account would enable a more reliable estimation of the causal effects of ligation, compared with medical management. The intensity and duration of invasive mechanical ventilation is a particularly important confounder in the association between ligation and adverse outcomes. Prolonged ventilator dependence is a commonly used clinical criterion in the decision to treat with ligation (vs. medical management) and also strongly predicts neonatal morbidities such as CLD, ROP and NDI The need for invasive mechanical ventilation reflects a combination of the effects of pulmonary overcirculation and alveolar edema from the PDA and underlying respiratory insufficiency due to severe respiratory distress syndrome, extreme prematurity and ventilator induced lung injury (VILI). The relative independent contribution of the PDA to ongoing respiratory insufficiency is, at present, difficult to quantify. Infants with similar echocardiographic indices of PDA haemodynamic significance may have varying degrees of respiratory failure (severe to none) owing to differences in nascent lung disease and tolerance of the increased pulmonary blood flow from the ductal shunt. Infants with greater respiratory failure are more likely to be referred for ligation, and the association of pulmonary insufficiency with adverse outcomes renders this a key confounder. However, no studies to date have adequately accounted for the intensity and duration of invasive mechanical ventilation support required prior to the administration of PDA treatment. 18

30 1.8.2 Selection Bias In previous studies 26, 63, 96, 108, survival bias may have influenced the reported lower mortality among ligated infants. Ligation was generally undertaken later in life after failure of medical therapy, meaning that ligated infants were more likely to have already survived the critical period of high early neonatal mortality. This PDA treatment paradigm implies that some of the sickest infants, treated initially with conservative management and/or indomethacin, may have died prior to becoming 'eligible' for ligation, resulting in selection bias in assembling the cohort of ligated infants. The possibility of survival bias is supported by studies which found no difference in mortality between medically and surgically treated groups of infants who both 26, 63, 64, received treatments at a similar postnatal age. If ligation truly improved survival (and survival bias was not present), then a delay or reduction in rates of surgical treatment would be expected to increase mortality. Some studies have demonstrated consistent mortality rates after moving to a delayed selective ligation strategy from an early routine ligation strategy following indomethacin failure. 69, 81 Several studies have reported higher mortality rates in infants with persistent PDA refractory to medical therapy, 113, 114 though the effect of ligation in potentially mitigating this risk was not assessed. Taken together, survival bias may be present and any beneficial effect of ligation on mortality remains untested. An additional form of selection bias may exist due to the common treatment paradigm where medically treated infants must have a demonstrated persistent significant PDA after failure of medical treatment in order to be considered eligible for ligation. Previous studies have uniformly categorized infants by treatment assignment irrespective of ductal patency following medical therapy. If infants whose PDA closed with medical therapy (and who were therefore never considered for ligation) were healthier (e.g., lower illness severity and thus at lower risk of adverse neonatal outcomes) than medically treated infants whose PDAs remained patent, then the inclusion of the former may introduce bias in favour of the medically treated group. 19