Response categorization and outcomes in extremely premature infants born at 22–26 weeks gestation that received inhaled nitric oxide for hypoxic respiratory failure

To evaluate the outcomes of extremely premature infants who received inhaled nitric oxide(iNO) for hypoxic respiratory failure(HRF). Retrospective analysis of 107 infants born 22–26 weeks gestation who received iNO for HRF at a single institution. Infants were categorized as positive, negative, or no responders based on change in FiO2 or OI. Underlying physiology was determined using Echocardiography/Radiography/Biochemistry. 63% of infants had a positive response; they received iNO earlier and were more likely to have acute pulmonary hypertension(PH). Positive response correlated with decreased incidence of death or grade 3 BPD at 36 weeks postmenstrual age, as compared to a negative response. Extremely premature infants have a positive response rate to iNO comparable to term infants when used for PH in the transitional period. Infants with a negative response to iNO had worse outcomes, necessitating the determination of the underlying physiology of HRF prior to iNO initiation.


INTRODUCTION
The incidence of hypoxemic neonatal respiratory failure (HRF) is high among neonates born ≤26 weeks gestation and increases with decreasing gestational age (GA) reaching nearly 100% by 22 weeks [1]. Oxygenation in the transitional period is influenced by right ventricular (RV) performance, pulmonary vascular resistance (PVR), degree of fetal lung maturity and lung recruitment. Therefore, the etiology of HRF in extremely premature infants is diverse and may include both respiratory [surfactant deficiency, retained fetal lung fluid, pneumonia, pulmonary hypoplasia, fetal lung immaturity] and cardiovascular [RV dysfunction, hemodynamically significant patent ductus arteriosus (PDA), acute pulmonary hypertension (PH)] phenotypes [2,3] which are likely to benefit from a physiology-specific approach. Among sick extremely preterm infants, acute PH may occur due to an incomplete transition from fetal to neonatal circulation with persistence of elevated pulmonary arterial vascular tone resulting in high pulmonary arterial pressures, impaired pulmonary blood flow, persistent right-to-left intra/ extracardiac shunt, and the development of hypoxemia out of proportion to parenchymal lung disease [4]. Inhaled nitric oxide (iNO), a selective pulmonary vasodilator, is a standard therapy for term and late preterm infants with HRF complicated by PH. The incidence of acute PH in the extremely preterm neonate born ≤26 weeks with HRF may be as high as 9.5% overall and 18.5% at 22 weeks [5].
Studies, including randomized trials, examining the use of iNO in the preterm population are inconsistent and often contradictory. Several studies have suggested that preterm infants with HRF at lower GA are less likely to respond to iNO, and outcomes, for example, risk of bronchopulmonary dysplasia (BPD) or death, after receiving iNO were similar; however, the incidence of acute PH in these studies was unknown [6][7][8][9]. Several recent studies, however, have demonstrated the opposite. Baczynski et al. demonstrated that preterm infants, with an average GA of 27 weeks, who received iNO for HRF had a similar response rate to iNO (46%) as term infants [10]. In addition, positive responders were more likely to survive without disability at 18 months (51%) vs non-responders (15%). Ahmed et al. reported that infants who had evidence of acute PH on targeted neonatal echocardiography (TnECHO) evaluation had positive response rate of 82%, independent of gestational age [11].
To date, no studies have examined the response to iNO and clinical outcomes exclusively in infants born less than 26 weeks GA. The aim of this study was to evaluate the response rates to iNO when used for HRF in infants born between 22 to 26 weeks GA, and to examine the clinical outcomes of these infants.

METHODS
This was a retrospective study of premature infants, born at 22 to 26 +6 weeks GA, who received iNO between July 2010 and April 2017 at a single tertiary center (University of Iowa Hospitals and Clinics, Iowa City, Iowa). The study was approved by the Institutional Review Board. Patients were identified from a clinically maintained database and the patient's electronic medical record was used to abstract data.

Criteria for eligibility
Neonates were included if they had clinical evidence of HRF, defined as fraction of inspired oxygen (FiO2) of ≥0.5 or oxygenation index (OI) ≥10, and received iNO for ≥12 h [adapted from [12,13]. During the study period, it was standard unit practice to initiate iNO based on the above HRF criteria. Although there was minor variation in practice, the above "symptomatic management for hypoxemia" criteria were agreed upon and followed for most extremely preterm patients. Neonates with congenital anomalies, known or suspected genetic or chromosomal syndromes, or diagnosis of major congenital heart disease, were excluded. Patent foramen ovale/atrial septal defect, small (<1 mm) ventricular septal defect and a diagnosis of PDA did not disqualify a patient from the study.

Clinical characteristics
Patient demographic and antenatal information were collected, including gender, GA, birth weight, maternal antenatal steroids, diagnosis of chorioamnionitis, gestational diabetes, maternal hypertension, multiple gestation, history of intrauterine fetal demise, duration of rupture of membranes, mode of delivery, cord arterial blood gas, APGAR score at 1 and 5 min, chest compressions during resuscitation, mode of ventilation following resuscitation, and total doses of surfactant received prior to iNO administration. Immediately prior to and 2 h after iNO administration, markers of clinical illness severity were collected, including mode of ventilation, mean airway pressure (MAP), fraction of inspired oxygen (FiO 2 ), respiratory severity score (RSS) [calculated as MAP × FiO 2 ], OI [calculated as (MAP × FiO 2 × 100)/PaO 2 )], blood gas, lactate, preductal and postductal oxygen saturations, heart rate, systolic and diastolic blood pressures, and mean arterial pressure at 2 and 1 h prior, time of initiation, and 1 and 2 h post administration. In addition, inotrope score [dopamine dose (mcg/kg/ min) + dobutamine dose (mcg/kg/min) + epinephrine dose*100 (mcg/kg/ min) + norepinephrine dose*100 (mcg/kg/min) + vasopressin dose *10 (mu/kg/min) + milrinone dose * 100 (mcg/kg/min)] was calculated for infants on cardiovascular support [14]. The infants' age in hours and weight at the time of initiation were collected, as well as the total dose of iNO received (summation of the dose of iNO multiplied by the time received, in parts per million).

Response to iNO
The response to iNO was determined 2 h after initiation of therapy by consensus of three neonatologists (JMD, PJM, REG) who were blinded to the clinical outcome. An infant was determined to have a positive response if FiO 2 decreased by ≥0.2 or OI decreased by ≥20%. Conversely, an infant was determined to have a negative response if FiO 2 increased by ≥0.2 or OI increased by ≥20%. An infant was determined to be a non-responder if the criteria above were not fulfilled.

Physiology and echocardiographic evaluation
For further assessment of the underlying physiology, data was collected to evaluate for possible etiologies requiring the initiation of iNO, including acute PH, PDA, sepsis, or lung disease. The information collected included diastolic hypotension (less than the third percentile for gestational age) at time of initiation, C-reactive protein (CRP) levels, white blood cell count, left shift, positive culture from a sterile site (blood, urine, cerebrospinal fluid), a diagnosis of necrotizing enterocolitis (NEC) at the time of initiation, and a chest x-ray score at the time of initiation [15]. In addition, when available, TnECHOs performed within 24 h of iNO administration were reviewed by a neonatologist (DRR) trained in Hemodynamics who was blinded to the clinical outcome. Echocardiography information was obtained, including RV systolic pressure (as measured by tricuspid regurgitation velocity), PDA diameter, PDA shunt direction, left atrial to aortic root ratio, diastolic flow indices in the aorta, and septal wall motion.
The underlying physiology was categorized as acute PH, hemodynamically significant PDA, post-ligation cardiac syndrome, sepsis, severe lung disease, or unknown/other by two investigators blinded (JMD, PJM) to the response to iNO and the clinical outcomes; disagreement between the investigators was resolved by a 3rd blinded party (JMK). A diagnosis of PH was determined by echocardiography data (elevated RVSp > 35, right to left shunting across the PDA for at least 10% of systole, or septal wall flattening or paradoxical motion) or by clinical data, which was defined as HRF at less than 4 h of life, with evidence of abnormal systemic blood flow (hypotension or lactic acidosis), without evidence of lung disease on chest x-ray. If lung disease was present, as defined below, this was considered the predominant physiology. The presence of a hemodynamically significant PDA was determined by Iowa PDA score [16]. Post-ligation cardiac syndrome was defined as presence of hypotension and/or need for inotrope support with associated oxygenation and ventilation failure in the 24 h following PDA ligation [17]. Sepsis or systemic inflammatory response syndrome (SIRS) were defined as a positive culture from a sterile site or significant elevation of CRP or immature WBC forms. Severe lung disease was determined by chest x-ray score of 3 or greater [15].

Outcomes
The primary outcome was the composite of death or grade 3 BPD, defined as need for mechanical ventilation at 36 weeks postmenstrual age [18]. Secondary outcomes included death, grade 3 BPD, duration of invasive ventilation after initiation of iNO, duration of non-invasive positive pressure ventilation after iNO initiation, retinopathy of prematurity (ROP) requiring treatment, necrotizing enterocolitis, intraventricular hemorrhage, periventricular leukomalacia, need for PDA ligation and age at discharge.

Statistical analysis
Univariate analysis of variance was used to compare the demographics, resuscitation, markers of clinical illness prior to iNO administration, and outcome characteristics between the three response groups. Mean with standard deviation and median with interquartile range were calculated for continuous variables with normal and non-normal distribution, respectively. Logistic regression was used to evaluate the relationship between response, gestational age, and postnatal age with outcome. Our dataset was largely complete, therefore, no adjustment for missing data was performed. A post-hoc analysis was performed to compare the outcome variables for infants treated with iNO at less than 96 h of life. A power analysis was not performed for this study. All records meeting the inclusion criteria in the study time period were included. SPSS v27 (IBM Corp, Armonk, NY, USA) was used for analysis.

RESULTS
During the study period, 423 patients between the GA 22-26 +6 weeks were admitted to NICU, out of which 174 (41%) received iNO and 107 infants satisfied the eligibility criteria ( Fig. 1). In total, 53 infants were not eligible for inclusion as their GA was greater than 27 weeks (n = 28), FiO 2 was less than 0.5 or OI was less than 10 at the time of initiation (n = 21), or information was not available (n = 4). A further 14 infants were excluded as the duration of iNO administration was less than 12 h (n = 12) or structural heart disease was identified (n = 2). In the included cohort, 67 infants (63%) demonstrated a positive response, 27 infants (25%) had no response, and 13 infants (12%) had a negative response. There were no differences in demographic or antenatal characteristics between the three groups, except that the "positive response" group received iNO therapy at a younger age than both the "no response" and "negative response" groups (10 h vs 204 and 144 h, p < 0.001) ( Table 1). Figure 2 demonstrates no difference in the proportion of positive, negative, or nonresponders according to GA. Using logistic regression, postnatal age [OR 1.1 (1.0, 1.10, p = 0.01),] but not GA [OR 0.9 (0.7, 1.3)] was associated with a positive response. There were no intergroup differences for all markers of clinical illness severity, including RSS, OI, FiO 2 , pH, receipt of cardiotropic support, and inotrope score, Table 1. Demographic characteristics, markers of clinical illness severity, and physiology data of positive responders, non-responders, and negative responders of extremely premature infants less than 26 6/7 weeks gestation at birth who received iNO for at least 12 consecutive hours. immediately prior to iNO initiation. Following iNO administration, there was a significant decrease in the FiO 2 and RSS in the positive responder group, with no change in either in the non-response group, and an increase in both measures in the negative responder group (Supplementary Fig. 1).

Physiology classification
For the infants included in this study, assignment of relevant underlying physiology at time of iNO initiation was possible for 93 infants (87%); specifically, 40 infants were determined to have acute PH, 11 infants had a hemodynamically significant PDA, 12 infants had sepsis, SIRS, or NEC, and 30 infants had severe lung disease. For the "positive response" group, 67% of infants had acute PH, 9% had sepsis or SIRS, and 24% had lung disease (p < 0.001). For the "no response" group, only 15% had acute PH, 26% had a hemodynamically significant PDA, 7% had sepsis or SIRS, and 52% had severe lung disease. For the "negative response" group, 33% had a hemodynamically significant PDA, 42% had sepsis or SIRS, and 25% had severe lung disease. The proportion of infants with PH was highest among responders (p < 0.001), while PDA was higher in both non-responder and negative responders (p = 0.02) and lung parenchymal disease was higher (p = 0.003) among non-responders.

Neonatal morbidity and mortality
The incidence of the combined outcome of death or grade 3 BPD (invasive ventilation at 36 weeks postmenstrual age) was lower in the "positive response" group as compared to the "no response" and "negative response" groups [67% vs 85% and 100%, p = 0.01; ( Table 2)]. Infants in the negative response group had an increased incidence of multiple neonatal morbidities, including grade 3 BPD (100% vs 57% and 79%, p = 0.04), duration of non-invasive positive pressure ventilation after iNO (235 days vs 143 and 166 days, p = 0.01), and ROP requiring treatment among survivors (67% vs 18% and 20%, p = 0.02). There were no intergroup differences in the incidence of duration of invasive ventilation after iNO, NEC, intraventricular hemorrhage, periventricular leukomalacia, and age at discharge. A post-hoc analysis was performed on all included infants that received iNO prior to 96 h of life. There was a significantly higher incidence of the combined outcome of death or grade 3 BPD in the no response and negative response groups compared to the positive response group [61% vs 89% vs 100%, p = 0.044; (Table 3)]. Infants in the negative response group had increased incidence of death (26% vs 56% vs 71%, p = 0.026). The effect on grade 3 BPD and ROP was not significant in this cohort.

DISCUSSION
Active resuscitation of extremely preterm neonates, who represent a population of increased developmental vulnerability, is conducted with increasing frequency and success [19]. Hypoxemia and acute PH are common sequelae; therefore, it is crucial to understand the role of contemporary pulmonary vasodilator therapies. In this study, we demonstrated, in a cohort of infants born at the limits of viability [22 to 26 weeks GA], a positive response rate to iNO therapy comparable with term infants. Moreover, infants that do have a positive response similar to a more mature cohort with a mean GA of 27.7 weeks [10] have a decreased likelihood of the composite outcome of death or grade 3 BPD as compared to those patients that either do not respond or have a negative response. To date, randomized trials have failed to identify a clear benefit of prophylactic administration of iNO to preterm infants to prevent BPD. Multiple bodies including the National Institutes of Health [20], American Academy of Pediatrics [21], and the American Heart Association/American Thoracic Society [22], advocate against its use except under very specific rescue circumstances such as history of preterm, prolonged rupture of membranes with pulmonary hypoplasia [22], or documented PH [20]. In spite of this, off-label use is common and neonates < 26 weeks GA represent the subpopulation in which iNO use has increased most dramatically [21,23]. The literature suggests that a perception of safety and clinicians' previous experience are the most important drivers for those who chose to prescribe iNO for HRF among preterm infants, despite a lack of large-scale studies demonstrating benefit [24]. Both older GA [12,25] and heavier birthweight [26] have been associated with positive iNO responses in randomized trials and observational studies [11,27,28] both of which focused on patients 25-32 weeks'; however, no studies to date have examined the 22-26 week population [12,29,30]. Animal models suggest biological plausibility for nitric responsiveness even at 22 weeks GA. In fetal lambs, by the equivalent of 24 weeks gestation, the arterial smooth muscle that controls vascular tone is mature. Endothelial nitric oxide synthase (eNOS) is present throughout gestation and the production of endogenous NO modulates vascular tone in late gestation lambs [31]. Resistance arteries, isolated from preterm lambs, relax with administration of an NO donor [32]. In vivo, preterm lambs (0.6 gestation) exposed to doses from 5-20 ppm of iNO demonstrate a dose-dependent reduction in pulmonary artery pressure, increase in ventilation-perfusion matching, and increased PaO 2 [33]. Consistent with these findings in fetal lambs, in our study population, extremely preterm infants were capable of responding to iNO and had a positive response rate comparable to term infants [34,35].
Greater responsiveness in the transitional period is likely related to the specific etiology of HRF and ambient physiology. Most term infants are treated with iNO in the early postnatal period [35] based on the hypothesis that acute PH is the most likely etiology for HRF based either on echocardiography evidence or clinical findings. Given the early postnatal age of the positive responders in our extremely preterm cohort, we speculate that responders had acute PH similar to term infants, which was related to the transition from fetal to neonatal physiology, especially given the relatively high incidence of acute PH at the extremes of prematurity [5]. The mechanism by which iNO improves oxygenation among acute PH patients is related to a decline in PVR which may improve pulmonary blood flow thereby improving the delivery of oxygenated blood to the systemic circulation. In addition, iNO may help to improve RV function by decreasing downstream afterload [36]. The timing of response in our study is consistent with the literature; in that preterm infants were more likely to have a positive response to iNO when it was initiated within the first 72 h of life [10,11]. Also a plausible contributor is evolution of disease. Acute PH is often transitory and cardiopulmonary disease may transition into inflammatory lung disease, shunt physiology or infection. Simultaneously, the neonate is exposed to noxious stimuli, lung damage, oxidative stress and potentially organ injury. Both absence of the physiology which iNO is designed to treat (acute PH), and presence of mediators which result in suboptimal medication response may be a part of the mechanism of non-response later in life. Further study into this would be valuable because it is certainly possible that both mechanisms are simultaneously at play. The reduced risk of death or grade 3 BPD in iNO responders, independent of GA, is a clinically relevant finding which adds to the already controversial literature. Several randomized control trials and a meta-analysis have not found that the prophylactic use of iNO impacts survival or development of BPD in premature infants [7,9,37], though these studies have included a heterogenous population of infants without a clear indication for iNO. Importantly, the trials failed to consider the possibility of negative responders. The inclusion criteria used in the clinical trials of prophylactic iNO for prevention of BPD were based on specific metrics of lung parenchymal disease such as requirement for surfactant or continuous positive airway pressure within the first 24 postnatal hours [38], requirement for mechanical ventilation for at least 2 days after the first 7 postnatal days [39] and requirement for ventilation support between 7 and 21 postnatal days [40]. It is likely that the enrolled population include a diverse spectrum of cardiorespiratory diseases including physiology which has either a low likelihood of benefit or even a risk of harm (e.g., ductal shunt physiology). Because we separated our cohort into responders, who had a high likelihood of favorable physiology, and non-responders and negative responders who had a high likelihood of unfavorable physiology, our ability to detect improvement in death/BPD was more precise than that of most trials, which did not stratify for disease physiology as a component of the trial design. Our results are in keeping with the literature that iNO responders have lower risk of BPD, or the composite of BPD and death, when used for its anti-remodeling properties [40,41] or pulmonary vasodilator for HRF [10,11]. The mechanism by which iNO modulates the risk of BPD is not known, however, several possibilities exist; first, animal experimental studies have shown that iNO prevents fibrin deposition and structural lung changes associated with the development of BPD; [42,43] second, iNO has been shown to decrease the presence of inflammatory cells in animal lung tissue and to reduce the concentration of several cytokines, known to contribute to BPD, in tracheal aspirates from premature infants [42,44]. Third, by improving the oxygenation status of the preterm infant at an earlier age, less mechanical ventilation and oxygen may be required subjecting the lung to less barotrauma and less oxidative injury thus contributing to less disease progression. The latter is plausible in our cohort as positive responders required fewer days of positive pressure ventilation compared to the negative response group. Interestingly, infants in the positive response group had a decreased rate of ROP requiring treatment as compared to those in the negative response group. We hypothesize that this may, similarly, be related to limitation of oxygen exposure early in life.
Understanding the reasons for a lack of a positive response to iNO is important for refining the care and designing more precision-based trials for the use of this therapy. In our cohort, negative responders were more likely to have a diagnosis of PDA or sepsis as compared to the other groups. Although the specific mechanism remains to be elucidated, the following factors may be contributory; first, by administering iNO to an infant with a PDA, a decline in PVR may promote increased left to right shunting of blood across the PDA, compromise lung compliance and precipitate hypoxemia; second, patients with sepsis may have a high burden of inflammatory cytokines and free radicals, to which peroxynitrite and other nitric-oxide mediated reactive oxygen species may contribute; finally, lung parenchymal disease was the primary underlying physiology of the non-response group [42][43][44]. With increasing atelectasis or consolidation of the lungs, iNO cannot diffuse to the peripheral pulmonary vasculature, thus its effect may be attenuated. Similarly, if neither pulmonary vascular dysregulation nor ventilation to perfusion matching are components of the pathophysiology for a particular patient, a therapy targeted towards this avenue is unlikely to demonstrate benefit. Our study is the first to describe a negative response to iNO in premature infants and suggest that this be a component of future studies in this field. It is likely that previous trials of iNO in preterm infants, where patients were randomized based exclusively on HRF (and not echocardiography confirmed PH), may have included both iNO negative responders, as well as nonresponders. These data suggest vigilance in the appraisal of iNO response in extremely preterm infants, especially in patients without echocardiography confirmed PH; specifically, clinicians should consider that a clinical deterioration after iNO initiation may reflect a negative response (rather than assuming refractory PH) prompting timely echocardiography evaluation and/or discontinuation of iNO treatment, where relevant.
Our study had several limitations. First, this was a retrospective study, and collection of data was dependent on the details available from the electronic medical record. In addition, due to the retrospective nature of the study, there were not clearly defined indications for iNO use in this population, but initiation was determined based on physician discretion. As the underlying physiology was determined after iNO was initiated, it can be difficult to compare these infants as similar groups as the underlying physiology was starkly different in the responder groups. Second, echocardiography was not consistently performed before/after iNO administration; thus, some patients lacked data regarding physiologic characterization at the time of iNO initiation. Serial echocardiography, particularly in the nonresponder and negative responder groups would have been helpful to determine the physiological contributors to deterioration. Third, although the overall sample size was not small, there were a small number of negative responders upon which to make conclusions and we lack long-term data for the cohort. Fourth, our data, which only includes infants from a single institution with unique site-specific guidelines and team approach, may limit its applicability [45]. Finally, our study did not examine healthy matched controls to compare the outcome data.

CONCLUSION
Extremely preterm neonates born at 22-26 weeks gestation demonstrate improved oxygenation in response to inhaled nitric oxide at a rate comparable to term infants, particularly during the transitional period when acute PH is the prevailing physiology. Responders may have a lower burden of severe respiratory morbidity and mortality. Importantly, some extremely preterm neonates may develop progressive hypoxemia following initiation of nitric oxide, thus earlier recognition of hypoxic respiratory failure requiring escalation of care and earlier use of echocardiography, prior to initiation of therapy, can help to determine the population of greatest likelihood to benefit as well as to help with stratification for future clinical trials.

DATA AVAILABILITY
Requests to view the included data can be made to the corresponding author.