Failure of Ductus Arteriosus Closure Is Associated With Increased Mortality in Preterm Infants

METHODS

The objective of this retrospective study was to investigate the possible impact of the failure of ductal closure on mortality in very low birth weight (VLBW) infants who were born at ≤29 weeks' gestational age (GA). The institutional review board at the University of Oklahoma Health Sciences Center (OUHSC) approved the study, and informed consent was waived.

VLBW infants who were born at ≤29 weeks' GA, had a birth weight of ≤1500 g, and were admitted to the NICU of the OUHSC Children's Hospital from January 2005 to August 2007 were included in the study. Electronic medical charts were reviewed. Figure 1 depicts the study population. Because shunting across the PDA is a normal phenomenon during postnatal transition, we excluded patients who died during the first 3 postnatal days. Patients with major congenital anomalies were also excluded. In addition, because the status of ductal patency cannot be objectively verified without the use of echocardiography, we excluded patients who did not have at least 1 echocardiographic study performed. Finally, because surgical ligation of the ductus in our patient population took place over a wide range of postnatal ages and because the impact of ligation on mortality and morbidity is complex and difficult to assess, we excluded from the primary analysis patients who had undergone ligation (n = 28); however, to assess the potential impact of ligation on mortality, we included these neonates in our secondary analyses (Fig 1). Of the 423 patients screened, 301 were included in the primary analysis. These patients were divided into 2 groups: closed ductus and persistent ductus. The closed ductus group (n = 260) comprised patients whose ductus closed either spontaneously or in response to exposure to COX inhibitors. The persistent ductus group (n = 41) comprised patients whose ductus was open (ie, blood flow detected by color Doppler) on the last echocardiogram obtained before discharge or death regardless of treatment status. Patency of the ductus on the last echocardiogram in the patients who were enrolled in this group was not attributed to reopening of the PDA as a result of clinical deterioration caused by sepsis or other pathology. The size of the PDA on the last echocardiogram varied from small (≤1.5 mm; n = 22), to moderate (1.6–2.5 mm; n = 13) to large (>2.5 mm; n = 6). Patients in the persistent ductus group had not been exposed to COX inhibitors because of the presence of a clinically “nonsignificant” PDA (n = 12) or had received either a partial (n = 3) or a complete (n = 15) single course (3 doses of a COX inhibitor) or prolonged or multiple courses (n = 11) of COX inhibitors but failed to achieve complete ductal closure at the time of discharge or death. Among the 29 patients in the persistent ductus group who received treatment with a COX inhibitor, the primary reasons for not proceeding with surgical ligation were a ductal diameter of <2 mm without hemodynamic instability (n = 13), having received a COX inhibitor at the time of death (n = 7), change in the direction of ductal shunting from left-to-right to right-to-left (n = 7), thought to be unstable to tolerate the procedure (n = 3), and documented sepsis (n = 2).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1

Study population. The shaded areas represent the secondary analyses.

As per the clinical protocol in the NICU at OUHSC, the diagnosis of a PDA in very preterm infants needs to be confirmed by echocardiography. Accordingly, a screening echocardiogram is usually obtained for these patients on postnatal days 2 to 4. Occasionally, however, when the patient receives no or only minimal respiratory support and has no evidence of hemodynamic instability, the neonatologist may forego obtaining an echocardiogram. As mentioned previously, in this study, we excluded all patients without at least 1 echocardiogram documenting the status of the PDA. A PDA is treated when it is considered hemodynamically significant on the basis of the ductal size and/or clinical symptoms attributed to ductal shunting. When a PDA is considered not significant, the patient is followed by serial echocardiograms until discharge and then followed by a cardiologist as an outpatient. A significant PDA is treated with either indomethacin or ibuprofen lysine according to the neonatologist's preference. Ductal response to COX inhibitor is evaluated by echocardiography after each treatment course. If the patient fails to respond to 2 to 3 courses of COX inhibitor treatment with complete closure of the ductus, a significant PDA is ligated, whereas a nonsignificant PDA is followed as described. For the purposes of this study, significant PDA is defined as a PDA with a ductal diameter of ≥1.5 to 2.0 mm and/or clinical signs and symptoms of significant left-to-right ductal shunting such as an active precordium, bounding pulses, wide pulse pressure, and systemic hypotension.

To adjust for the effects of disease severity on outcome, we used the clinical risk index for babies (CRIB) score.^19,20 We also collected data on patient characteristics and diseases that have been shown to affect mortality. These data include GA, birth weight, antenatal steroid exposure, Apgar score at 5 minutes, gender, IVH grades III to IV, NEC (Bell stages II–III), and culture-proven sepsis.

GA and birth weight between the groups were compared using 2-tailed Student's t test. Nonparametric test (Mann-Whitney) was used to compare data with non-Gaussian distribution. Binomial parameters between the groups were compared with χ² or Fisher's exact test as appropriate. Univariate logistic regression was used to predict the effect of selected variables on mortality. These variables included GA, antenatal steroid exposure, Apgar score at 5 minutes, gender, CRIB score, IVH III to IV, NEC II to III, sepsis, and exposure to COX inhibitors. Variables with P < .1 in the univariate logistic regression analyses were used simultaneously as potential confounders in multivariate logistic regression models to define the odds of mortality in case of a persistent PDA. Finally, Cox proportional hazard regression was used to take into account time until death and assess the independent effect of each risk factor on mortality. The time to event was defined as birth to the identification of death.

In addition, we performed a number of secondary data analyses. First, we excluded patients who died during the first week and then those who died during the first 2 weeks of postnatal life. In addition, we included patients who underwent ductal ligation in the analysis by using 3 different scenarios. First, we included the ligated patients in the closed ductus group and then in the persistent ductus group. Finally, using the median age of death in patients in the persistent ductus group as the cutoff point (14 days), a patient with a ligated PDA was considered to belong to the closed or the persistent ductus group when ligation was performed before or after the cutoff point, respectively.

RESULTS

The study population that underwent the primary analysis comprised 301 VLBW infants who were born at ≤29 weeks' GA (Fig 1). There were 260 patients in the closed ductus group and 41 patients in the persistent ductus group (Table 1). Patients with a persistent ductus had a lower birth weight, were less mature, had a higher CRIB score and higher incidence of IVH grades III to IV, and were more likely to receive a COX inhibitor than those in the closed ductus group (Table 1). Mortality rate was higher among patients with a persistent ductus (29 of 41 [70.7%]) compared with those with a closed ductus (29 of 260 [11.2%]; P < .0001; Tables 1 and 2). On additional analysis of the persistent ductus group, mortality rates were found to be 55% (12 of 22) in patients with a small PDA (ductal diameter ≤1.5 mm) compared with 89% (17 of 19) in those with a moderate or large PDA (ductal diameter ≥1.6 mm) on the last echocardiogram (P = .014). Table 3 lists the causes of death in the closed and persistent ductus groups as well as in patients who underwent ligation. It is of note that more patients died from multiorgan failure in the persistent ductus group compared with the other 2 groups.

Univariate logistic regressions of potential confounders that affect mortality in the primary analysis between the closed and persistent ductus groups are shown in Table 4. After adjusting for GA, Apgar score at 5 minutes, CRIB score, IVH III to IV, NEC stages II to III, sepsis, and exposure to COX inhibitors in multiple logistic regression models, the odds for mortality were significantly higher when the PDA persisted (odds ratio: 16.79 [95% confidence interval [CI]: 6.05–46.61]; P < .0001). By using the Cox proportional hazard regression model to incorporate the effect of time until death while controlling for other significant covariates, the hazard for death for the patients in the persistent ductus group remained significantly higher than in the closed ductus group (hazard ratio: 8.34 [95% CI: 4.30–16.06]; P < .0001; Table 5).

Patients who were treated with COX inhibitors in the closed and persistent ductus groups (n = 159) received indomethacin (n = 120), ibuprofen (n = 35), or both (n = 4). Exposure to indomethacin or ibuprofen was similarly associated with an increased mortality in univariate regression analysis (data not shown).

On additional analysis, excluding patients who died during the first postnatal week, the persistent ductus group had a higher mortality rate (22 of 34 [64.7%]) compared with the closed ductus group (25 of 256 [9.8%]; P < .0001; Table 2). Similarly, excluding patients who died during the first 2 weeks, patients in the persistent ductus group had a higher mortality rate (13 of 25 [52%]) than those in the closed ductus group (20 of 251 [8%]; P < .0001). Adjusting for the confounders, the odds for mortality and the hazard for death remained higher when PDA persisted (Table 5).

Among the 28 patients who had undergone PDA ligation and thus were excluded from the primary analysis, 2 died during their initial hospitalization. To investigate whether exclusion of these patients affected the observed difference in mortality between the closed and persistent ductus groups, we included them in the secondary analyses by using 3 different scenarios (see “Methods”). Regardless of the group assignment of the patients with a surgically ligated ductus arteriosus, the confounder-adjusted odds for mortality and hazard for death remained higher for patients who were assigned to the persistent ductus group (Table 5).

Survival data at 9 months of age were available for 227 of the 301 patients who were enrolled in the primary analysis. The Kaplan-Meier curve for this subset showed significant difference in mortality patterns between the 2 groups (log-rank test P < .0001; Fig 2). Data on the fate of the PDA were available only on 7 of the 12 patients who were discharged from the hospital with a persistent ductus. Five of the 7 patients had their PDA still open on follow-up visit in the clinic (ages of 66, 113, 121, 158, and 365 days), whereas the remaining 2 patients had a closed ductus on the follow-up visit at 88 and 251 days of postnatal life, respectively.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2

Kaplan-Meier survival curves of closed versus persistent ductus. Log-rank test P < .0001.

DISCUSSION

In this retrospective study, we found that failure of ductal closure in VLBW neonates who were born at ≤29 weeks' GA was associated with an increase in mortality. This association remained significant after adjustment for the degree of immaturity; initial disease severity; and morbid pathologies such as severe IVH, NEC, and sepsis. The hazard for death for patients in the persistent ductus group was eightfold higher compared with the closed ductus group. Furthermore, this association persisted even when patients who died in the first week or first 2 weeks of postnatal life were excluded and when ligated patients were included in the analysis. The reasons for the increased mortality are unknown but may be, at least in part, related to the pathophysiology of the PDA. Left-to-right shunting across the PDA significantly affects both the pulmonary and the systemic circulation, resulting in pulmonary overcirculation and decreased systemic perfusion, respectively.^21–26 Pulmonary overflow can lead to pulmonary edema, increased need for ventilatory support, maldevelopment of the alveolar structure, and ultimately BPD.^21–23 The associated compromised systemic blood flow and hypotension can affect tissue oxygen delivery and result in abnormal organ function and tissue injury.^24–26 From this standpoint, it is an interesting finding that the most common cause of death in persistent ductus group was multiorgan failure. An alternative explanation for the association between increased mortality and persistent PDA is that persistent ductus may be only a marker of disease severity; however, a possible contributory role of ductal shunting in the observed increased mortality might be supported by our finding that patients with a moderate to large PDA had higher mortality than those with a small ductus arteriosus.

Because there is a lack of randomized clinical trials comparing outcomes in neonates who are treated for a PDA with those who do not receive pharmacologic treatment or surgical closure, some authors have questioned the justification for administering COX inhibitors or performing surgical ligations to close the PDA.^12–14 Because of the growing concerns about the potential harmful effects of surgical ligation of the ductus, the criticism of the present approach to the treatment of the PDA has especially targeted ligation. Indeed, surgical closure of the PDA leads to an impairment of cardiovascular function.^27,28 Although transient, the documented deterioration in the cardiovascular status and the suspected deterioration in pulmonary function after surgical ligation may have long-term implications, especially because there is an association between PDA ligation and BPD and poor neurodevelopmental outcome.^29,30 Although PDA ligation may be only a marker rather than the cause of these complications, because of the lack of randomized trials, a cause–effect relationship cannot be ruled out. As a consequence, some investigators have suggested a conservative approach to the treatment of PDA whereby surgical closure is reserved for patients in intractable heart failure¹³; however, it is unclear whether the prolonged exposure to pulmonary overcirculation and compromised systemic perfusion is less harmful than surgical ligation. Along these lines, we found that inclusion of patients who had undergone surgical ligation into the closed or persistent ductus group or choosing their group assignment according to the median time of death in the persistent ductus group did not affect the finding of the primary analysis, because a persistent PDA remained associated with a significantly higher mortality in these secondary analyses; therefore, surgical ligation in this patient population does not seem to have played an appreciable role in influencing mortality.

Because the common practice has been to close the PDA pharmacologically or surgically, data on the outcome of preterm infants with a persistent ductus are scarce. To our knowledge, only 1 other study addressed this issue and reported a fourfold increase in the mortality of preterm infants with a persistent PDA.³¹ The finding of an eightfold increase in the mortality of VLBW neonates with a persistent PDA in our study strongly supports and extends the observations of the previous study.³¹

Patients who were discharged from the hospital with a persistent ductus remained asymptomatic at the follow-up outpatient cardiology clinic visits during the first postnatal year. This observation is consistent with the findings of a recent study that demonstrated that none of the 21 VLBW infants who were included in this study and discharged from the hospital with a PDA developed congestive heart failure by the time of the follow-up visit.³²

Our study suffers from several limitations. Given its retrospective nature, the 2 groups were not similar in many of their clinical characteristics. We adjusted for the identified differences by using multivariate logistic and proportional hazard regression models, yet certain differences may have remained unidentified and thus not adjusted for by our analysis. Another limitation is that the persistent ductus group included patients with and without exposure to COX inhibitors; however, because we were interested in the effect of failure of closure of the ductus as an independent variable regardless of the cause, this heterogeneity is likely of less importance. In addition, we corrected for the effect of COX inhibitor exposure in the regression models. Finally, it is of note that the CIs for the odds for mortality and, to a lesser extent, the hazard for death were relatively wide; however, the finding that the lower limits of these CIs are 6.05 and 4.30, respectively, is reassuring and lends support for the validity and potential clinical significance of our observations.

CONCLUSIONS

We found a significant increase in mortality in very preterm infants when the ductus arteriosus remained open. It is important to emphasize that our findings do not establish the persistent ductus arteriosus as the cause of the observed increased mortality; nevertheless, it does raise this possibility. Although the controversy regarding medical and/or surgical closure of PDA can be settled only by appropriately designed and powered randomized studies, the finding of an increased mortality in patients with a persistent PDA in this study provides additional information and can be of help in the design of future prospective clinical trials.