The Effect of Ibuprofen Exposure and Patient Characteristics on the Closure of the Patent Ductus Arteriosus in Preterm Infants

Abstract

Spontaneous closure of the ductus arteriosus depends on gestational age (GA) and might be delayed in preterm infants, resulting in patent ductus arteriosus (PDA). Ibuprofen can be administered to enhance closure, but the exposure‐response relationship between ibuprofen and the closure of PDA remains uncertain. We investigated the influence of patient characteristics and ibuprofen exposure on ductus closure. A cohort of preterm infants with PDA and treated with ibuprofen was analyzed. Ibuprofen exposure was based on a previously developed population pharmacokinetic study that was in part based on the same study population. Logistic regression analyses were performed with ductus closure (yes/no) as outcome, to analyze the contribution of ibuprofen exposure and patient characteristics. In our cohort of 263 preterm infants (median GA 26.1 (range: 23.7–30.0) weeks, birthweight 840 (365–1,470) g) receiving ibuprofen treatment consisting of 3 doses that was initiated at a median postnatal age (PNA_start) of 5 (1–32) days, PDA was closed in 55 (21%) patients. Exposure to ibuprofen strongly decreased with PNA_start. Overall, the probability of ductus closure decreased with PNA_start (odds ratio (OR): 0.7, 95% CI: 0.6–0.8) and Z‐score for birthweight (Z _{Birthweight‐for‐GA}; OR: 0.8, 95% CI: 0.6‐1.0), and increased with GA (OR: 1.5, 95% CI: 1.1–1.9). For patients with PNA_start < 1 week, concentrations of ibuprofen, GA, and Z _{Birthweight‐for‐GA} predicted probability of ductus closure. During a window of opportunity for ductus closure within the first days of life, probability of closure depends on GA, Z _{Birthweight‐for‐GA}, and ibuprofen exposure. Increased, yet unstudied dosages might increase the effectivity of ibuprofen beyond the first week of life.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

☑ Ibuprofen is the most frequently used drug for the treatment of patent ductus arteriosus in preterm neonates, but the exposure‐response relationship between ibuprofen and closure of the ductus remains uncertain.

• WHAT QUESTION DID THIS STUDY ADDRESS?

☑ This study examined the effects of patient characteristics and ibuprofen exposure on the closure of the ductus arteriosus in a clinical, observational cohort.

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

☑ This study provides relevant insights into the timing of ibuprofen treatment initiation and highlights the importance of timely and adequate dosing.

• HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

☑ This study provides relevant insights regarding important patient characteristics, such as postnatal‐ and gestational age and exposure to ibuprofen, that enhance the potential of future studies to define an exposure‐response relationship.

The ductus arteriosus usually closes within hours after birth of a term born infant, but in prematurely born neonates closure is often delayed or does not occur spontaneously, resulting in a patent ductus arteriosus (PDA). The consequent ductal left‐to‐right shunt can result in increased pulmonary blood flow and systemic hypoperfusion, which in turn can lead to comorbidities, such as bronchopulmonary dysplasia intraventricular hemorrhage, necrotizing enterocolitis, and renal failure. ¹

Spontaneous closure of the ductus is an effect of multiple complex physiological mechanisms. A drop in prostaglandin E₂ (PGE₂) levels due to lost supply by the placenta after birth is one of these mechanisms. Because of higher sensitivity to PGE₂ earlier in pregnancy, this mechanism of ductus closure may be less effective in preterm infants. ² Cyclo‐oxygenase (COX) inhibitors can be used to pharmacologically stimulate closure of the ductus by decreasing PGE₂ production. Other suggested mechanisms that make preterm infants more prone for PDA are an increased nitric oxide sensitivity in combination with decreased expression of ion channels in the ductus that induce vasodilation, and an immature anatomy of the ductus that postpones anatomic closure. ²

Besides other drugs, such as acetaminophen and indomethacin, the COX‐inhibitor ibuprofen is the most frequently used for the treatment of PDA. ³ , ⁴ The effects of ibuprofen have been compared with placebo or to alternative treatments, such as indomethacin. In a meta‐analysis of all currently studied treatment strategies, oral ibuprofen treatment with 20 mg/kg on the first day of treatment, followed by 10 mg/kg daily on days 2 and 3, was shown to be the most effective for PDA closure. ⁵ This analysis did not identify patient characteristics, such as gestational age (GA) or postnatal age (PNA) as confounding factors, but might have been limited by the fact that these characteristics were not always reported in the examined studies. ⁶ The probability of spontaneous closure increases with GA, and as a result the effectiveness of ibuprofen may appear higher in infants with a lower GA compared with studies in infants with a higher GA. ⁷ Moreover, most studies reported on scenarios in which ibuprofen treatment is started within the first few days of life. The clearance of ibuprofen increases with PNA, resulting in a lower exposure to ibuprofen. ⁸ , ⁹ These factors should be considered simultaneously for the evaluation of the efficacy of ibuprofen.

The aim of this study was to quantify the contribution of patient characteristics and ibuprofen exposure to the closure of PDA in a large population of preterm infants with varying patient characteristics and ibuprofen dosing regimens. The overall population was evaluated as well as a subset of the population in which ibuprofen was initiated during the first week of life.

METHODS

Patients and data

Data on a clinical cohort of preterm infants (n = 298) who all received ibuprofen for the treatment of PDA were available for analysis. Data from part of this cohort (n = 43) were also used for the development of a recently published population pharmacokinetic (PK) model that is described in more detail in the next section. ⁸ All patients had confirmed PDA based on an echocardiographic ultrasound prior to treatment initiation. Treatment with ibuprofen was initiated if the PDA was judged as hemodynamically significant by the clinical team (pediatric cardiologist and neonatologist) according to the local protocol and there was no contraindication for ibuprofen (i.e., hepatic or renal failure, severe thrombocytopenia or other known clotting disorders, recent intraventricular hemorrhage or other bleeding, sepsis, suspected or confirmed necrotizing enterocolitis, or severe hyperbilirubinemia). For infants born before June 2015, the protocol for ibuprofen treatment consisted of a loading dose of 10 mg/kg bodyweight on day 1 followed by 2 maintenance doses of 5 mg/kg on days 2 and 3 (also written as 10‐5‐5 mg/kg). After June 2015, a new dosing regimen was applied that accounts for the maturation of ibuprofen clearance with PNA. ⁹ Loading and maintenance doses were adapted to PNA, resulting in 10‐5‐5 mg/kg for PNA < 70 hours, gradually increasing per day up to 20‐10‐10 mg/kg if PNA was above 196 hours. Ibuprofen was administered intravenously or orally using the same dosages, and an ibuprofen cycle consisting of three doses could be repeated as judged to be clinically deemed necessary. Typically, an ultrasound was made after the third dose to assess the status of the PDA. To prevent unnecessary burden for the patient, the attending physician could postpone or cancel an ultrasound after three doses if the outcome was expected to be clinically irrelevant (apparent signs for open ductus such as a persistent murmur or low diastolic blood pressure or no signs of a clinically relevant ductus). If necessary, ibuprofen treatment could be continued for an additional 3 to 6 days. For the purpose of this analysis, all ultrasounds were re‐assessed by one experienced clinician (author J.d.K.) who determined whether PDA was closed or not. Additionally, the diameter of the ductus, the diameter of the left pulmonary artery, the ratio between the diameter of the ductus and the left pulmonary artery, the maximal diastolic velocity, ratio between atrium to aortic valve ratio, flow through the left pulmonary artery, and flow through the aorta were documented for each ultrasound. For each patient, the Z‐score for birthweight (Z _{Birthweight‐for‐GA}) was calculated based on birthweight, GA, and Dutch perinatal growth curves. ¹⁰ The cohort has been previously described by de Klerk et al. ¹¹

Ibuprofen exposure and concentrations

A previously developed population PK model for ibuprofen in preterm infants ⁸ was used to calculate ibuprofen exposure upon the applied dosing strategies based on relevant patient characteristics. For oral administrations, 100% bioavailability was assumed and the absorption rate constant was set at 0.6562 hour⁻¹. ¹² Based on this PK model clearance keeps increasing with PNA, whereas physiologically a plateau‐function in clearance is expected at higher PNA. ¹³ Maximum PNA in the PK model dataset was 18 days. To prevent extrapolation of the maturation function outside the range in PNA of the patients that were included in the PK study, for infants in the current study with a PNA above 18 days (n = 18), clearance was calculated as if their age was 18 days. ⁸ If one or more plasma samples of ibuprofen were available for an individual patient (n = 150 plasma samples from 43 patients), the model was used to best fit the plasma‐concentration time curve of that individual patient, also referred to as individual predicted exposure. If no concentration was available, the infants’ characteristics (i.e., GA, PNA, and small for GA (SGA)) were used to calculate the typical value for the population, also called population predicted exposure. Comparison of the demographics between the group with PK samples and the group without PK samples did not reveal any differences.

Ibuprofen exposure was examined using different metrics. First, area under the curve between the first dose and 24 hours after the third dose (AUC_0‐72hours) was tested based on the results of Hirt et al. ⁹ Second, the lowest ibuprofen concentration during the first 3 days of ibuprofen treatment of the patient (C _{trough72hours}) was selected based on ibuprofen’s mechanism of action that concerns COX‐2 inhibition via competitive and reversible COX‐2 binding. ¹⁴ Therefore, a certain minimal concentration would be needed to maintain COX‐2 inhibition and achieve closure of the ductus. ¹⁵ C _{trough72hours} could be the trough concentration of the first, second, or third dose, dependent on which was the lowest. Additionally, average C _{trough72hours} (C _{trough72hours_average}) was defined as the average trough concentration after the first, second, and third doses.

Logistic regression analysis

Only results upon the first ibuprofen treatment cycle that consisted of three doses were used. To account for variable dosing and ultrasound timing, patients who only received 1 or 2 ibuprofen doses, patients with more than 72 hours between 2 of the first 3 doses, and patients with an ultrasound made more than 1 week after the first dose were excluded to assure that closure (yes/no) happened within the interval where the effect of ibuprofen would still be relevant. If no ultrasound was made after 3 doses and ibuprofen treatment was continued, the ductus was assumed to be open 24 hours after the third dose. Patients without at least one ultrasound after diagnosis were excluded.

Closure of the ductus (yes/no) ~ 24 hours after the first ibuprofen treatment episode (3 ibuprofen doses) was the primary outcome for logistic regression. The influence of the patient characteristics GA, PNA at the start of treatment (PNA_start), postmenstrual age, sex, birthweight, and Z _{Birthweight‐for‐GA} on the primary outcome were analyzed, as well as ibuprofen exposure using the following different measures: the cumulative first 3 dose amounts, route of administration, AUC_0‐72hours, C _{trough72hours}, and C _{trough72hour_average}. First, a univariate analysis was performed to examine and compare predictive performance of all covariates. Second, a multivariate analysis was performed for which all relevant covariates were included as predictors. Insignificant predictors (P > 0.05) were removed one‐by‐one to obtain the final model in which each predictor was significant (P < 0.05).

Subgroup analysis

Finally, the multivariate analysis was repeated on a subgroup of the dataset in which the maximal PNA_start was 7 days because in clinical practice ibuprofen treatment is usually initiated during the first week of life. ¹⁶

For both multivariate analyses interaction terms were tested and collinearity was assessed by the variance inflation factor. Potential overparameterization of the final models was assessed by using a Hosmer‐Lemeshow test. ¹⁷

Simulation of dosing strategies

Coefficients of the final model were used to calculate the probability of closure for hypothetical preterm infants’ representative for the study population. GA was set at 24, 25, 26, 27, or 28 weeks with median birthweight for each GA for Dutch male infants. ¹⁰ In simulations, these infants were treated with ibuprofen intravenously 10‐5‐5 mg/kg, 20‐10‐10 mg/kg. Population predicted C _{trough72hours} and AUC_0‐72hours were obtained for each dosing strategy based on a previously developed population PK model in preterm infants. ⁸

RESULTS

Patients and data

Of the total of 299 infants, at least one follow‐up ultrasound was made in 295 patients. Upon diagnosis, the median diameter of the PDA was 2.2 (range: 1.2–4.1) mm in this overall cohort. During the first follow‐up ultrasound, performed at a median of 78 hours (range: 60–71 hours) after diagnosis and 71 hours (range: 56–165 hours) after the first dose, PDA was closed in 78 patients (26%). In the remaining patients, the median diameter of the PDA was 2.0 mm (range: 0.5–4.0 mm). In 46 patients, a second follow‐up ultrasound was made after repeated ibuprofen treatment cycles at 112 hours (range: 72–3,662 hours) after the first ultrasound. In 7 of these patients, PDA was reported to be closed, whereas the median diameter in the other patients was 1.8 mm (range: 0.6–3.0 mm).

Following the exclusion criteria, 263 preterm infants were included in the data analysis (Figure  1 ). In 55 infants (21%) the ductus arteriosus was closed after 3 ibuprofen doses, of which in 52 patients’ treatment was initiated during the first week of life. For 114 patients, the first follow‐up ultrasound was postponed to beyond 3 doses and the ductus was assumed to have not closed after 3 doses. Ultrasounds made after more than 3 doses showed that in 103 (90%) of these patients, the ductus was still open which justified the assumption. Patient characteristics of the analyzed infants are presented in Table  1 . Figure  2 shows the C _{trough72hours} and AUC_0‐72hours vs. PNA_start.

Figure 1.

Data selection flowchart. PDA, patent ductus arteriosus.

Table 1.

Summary of the patient characteristics of the patients in median and ranges

	Outcome after 3 doses ibuprofen		% Closed	Total
	Open	Closed
Number of patients	208	55	21	263
Male (n)	118	26	18	144
Female (n)	90	29	24	119
GA (weeks)	26.1 (23.7–29.4)	26.6 (24.0–30.0)*		26.1 (23.7–30.0)
Birthweight (g)	840 (365–1,320)	840 (440–1,470)		840 (365–1,470)
Small for GA (n)	52	19	27	71 (27%)
Postmenstrual age (weeks)	27.6 (24.7–32.8)	27.6 (25.0–31.1)		27.6 (24.7–32.8)
PNA at treatment initiation (days)	6 (1–33)	4 (2–17)**		5 (1–33)
PNA at treatment initiation < 8 days (n)	128	52	29	180
PNA at treatment initiation ≥ 8 days (n)	80	3	4	87
Z‐score of birthweight for GA	−0.3 (−5.2 to 3.3)	−0.7 (−4.7 to 1.5)*		−0.4 (−5.2 to 3.3)
Diameter PDA at diagnosis (mm)	2.3 (1.2–4.1)	2.0 (1.0–3.5)		2.2 (1.0–4.1)
Diameter PDA at first follow‐up ultrasound (mm)	2.0 (0.5–4.0)	0.0 (0.0–0.0)*		1.7 (0.0–4.0)
Time between diagnosis and first follow‐up ultrasound (hour)	78 (60–171)	87 (68–158)		85 (60–171)
Dosing regimenc
10‐5‐5 mg/kg	154	42	21	196
20‐10‐10 mg/kg	31	3	9	34
Other	23	10	30	33
Route of ibuprofen administration (n)
i.v.	189	55	23	244
Oral	19	—	0	19
AUC0‐72hours (mg*hour/L)	895 (51–2,567)	1,363 (232–2,605)**		992 (51–2,605)
C trough72hours (mg/L)	3.3 (0.0–25.9)	12.8 (0.0–22.2)**		4.6 (0.0–25.9)

*P value < 0.05, **P value <0.01, based on Mann‐Whitney U test.

AUC_0‐72hours, area under the curve between the first dose and 24 hours after the third dose; C _{trough72hours}, minimal trough concentration of the first 3 doses; GA, gestational age; PDA, patent ductus arteriosus; PNA, postnatal age.

Figure 2.

Lowest ibuprofen trough concentration in the 72 hours after start of treatment (left) and ibuprofen area under the curve (right) for each patient included in the logistic regression analysis. Circles represent appropriate for GA infants, and squares represent small for GA infants. AGA, appropriate for gestational age; GA, gestational age; SGA, small for gestational age.

Univariate analysis

In the univariate analysis GA (P = 0.02), PNA_start (P = 0.00006), (P = 0.02), C _{trough72hours} (P = 0.006), C _{trough_average} (P = 0.002), and AUC_0‐72hours (P = 0.001) significantly predicted closure of the ductus. PNA_start provided the biggest drop (30 points) in Akaike Information Criterion and was therefore the most predictive covariate. For every week increase in GA, the odds of ductus closure were found to increase with 30% (odds ratio (OR): 1.30, 95% CI: 1.04–1.64). For every day increase in PNA_start, the odds of closure decrease with 23% (OR: 0.77, 95% CI: 0.67–0.86). For every unit increase in Z _{Birthweight‐for‐GA} the odds of closure decrease with 21% (OR: 0.79, 95% CI: 0.65–0.97). For every mg/L increase in C _{trough72hours} the odds of closure increase with 6% (OR: 1.06, 95% CI: 1.02–1.1). For every 100 mg*hour/L in AUC_0‐72hours, the odds of closure increase with 8% (1.0008, 95% CI: 1.0003–1.001). Postmenstrual age, birthweight, sex, route of administration, and cumulative dose were not predictive for closure of the ductus.

Multivariate analysis complete dataset

GA, Z _{Birthweight‐for‐GA}, PNA_start, and AUC_0‐72hours were included in the multivariate analysis. With both PNA_start and AUC_0‐72hours in the model, PNA_start was more predictive for ductus closure than AUC_0‐72hours, which no longer met the significance criterion (P = 0.12) and was therefore removed from the model. Based on the final model, the odds of ductus closure increase with increasing GA (OR: 1.47, 95% CI: 1.12–1.94) and decreases with increasing PNA_start (OR: 0.74, 95% CI: 0.64–0.84) and Z _{Birthweight‐for‐GA} (OR: 0.79, 95% CI: 0.63–1.00). A Hosmer‐Lemeshow test did not suggest a misfit of the model (P = 0.06). Based on these results, the probability of closure based on these patient characteristics can be calculated with Eq. 1, and parameter estimates and corresponding ORs are presented in Table  2 . In Figure  3 the probability of closure with PNA_start is visualized for infants with varying Z _{Birthweight‐for‐GA}. AUC_0‐72hours was initially selected because it was more predictive than C _{trough72hours} in the univariate analysis. Repetition of the multivariate with C _{trough72hours} instead of AUC_0‐72hours resulted in a similar conclusion regarding the impact of GA, PNA_start, and ibuprofen exposure.

$$\text{Probability}\text{of}\text{closure}=\frac{\exp (‐9.91‐0.30\ast {\text{PNA}}_{\text{Start}}+0.38\ast \text{GA}‐0.23\ast Z_{\text{Birthweight ‐ for ‐ GA}})}{1+\exp (‐9.91‐0.30\ast {\text{PNA}}_{\text{Start}}+0.38\ast \text{GA}‐0.23\ast Z_{\text{Birthweight ‐ for ‐ GA}})}$$ (1)

Table 2.

Estimated odds by the logistic regression analysis with closure of the ductus after 3 ibuprofen doses as outcome measure

	Estimated coefficient (95% CI)*	Estimated OR (95% CI)	P value	Variance‐inflation factor
Complete study population
PNA of treatment initiation (days)	−0.30 (−0.45 to 0.18)	0.74 (0.64–0.84)	0.15 × 10−4	1.07
GA (weeks)	0.38 (0.12–0.66)	1.47 (1.12–1.94)	0.0058	1.12
Z‐score for birthweight	−0.23 (−0.46 to 0.002)	0.79 (0.63–1.00)	0.048	1.06
Subgroup: maximum PNA at treatment initiation 7 days
GA (weeks)	0.57 (0.27–0.90)	1.77 (1.31–2.46)	0.0004	1.1
C trough72hours (mg/L)	0.11 (0.04–0.19)	1.12 (1.04–1.21)	0.004	2.5
Z‐score for birthweight	−0.59 (−0.98 to −0.23)	0.55 (0.38–0.79)	0.002	2.4

First, the logistic regression based on the complete study population is presented, and below the logistic regression based on the subset of the study population with a maximum PNA at treatment initiation of 7 days.

CI, confidence interval; C _{trough72hours}, minimal trough concentration of the first 3 doses; GA, gestational age; OR, odds ratio; PNA, postnatal age.

Figure 3.

Predicted probability of closure of the patent ductus arteriosus vs. PNA at treatment initiation based on the logistic regression analysis on the complete study population, for infants with different GAs. GA, gestational age; PNA, postnatal age.

Subgroup analysis – First week of life

In a multivariate analysis on a subset of the dataset in which the maximal PNA_start was 7 days, C _{trough72hours} was found to be more predictive for closure than PNA_start. With GA, PNA_start, Z _{Birthweight‐for‐GA}, and C _{trough72hours} included as predictors, PNA_start was the least significant predictor (P = 0.16). Upon removal of PNA_start, all remaining predictors (GA, Z _{Birthweight‐for‐GA}, and C _{trough72hours}) were significant predictors for closure, without a suggestion of misfit (Hosmer‐Lemeshow test P value of 0.6). Based on these parameter estimates, the odds of closure are expected to increase with GA (OR: 1.77, 95% CI: 1.31–2.46) and increased C _{trough72hours} (OR: 1.12, 95% CI: 1.04–1.21), and to decrease with Z _{Birthweight‐for‐GA} (OR: 0.55, 95% CI: 0.38–0.79). Parameter estimates are presented in Table  2 . The probability of closure during the first week of life can be calculated with Eq. 2.

$$\begin{array}{cc}& \text{Probability}\text{of}\text{closure}\text{during}\text{first}\text{week}\text{of}\text{life}\hfill \\ \hfill & =\frac{\exp (‐17.53+0.11\ast C_{\text{trough72h}}+0.57\ast \text{GA}‐0.59\ast Z_{\text{Birthweight ‐ for ‐ GA}})}{1+\exp (‐17.53+0.11\ast C_{\text{trough72h}}+0.57\ast \text{GA}‐0.59\ast Z_{\text{Birthweight ‐ for ‐ GA}})}\hfill \end{array}$$ (2)

Based on this model, the typical infant in our population with a GA of 26.1 weeks, a Z _{Birthweight‐for‐GA} of −0.4 and a C _{trough72hours} of 4.6 mg/mL would have a probability of ductus closure of 13%. Increasing C _{trough72hours} with 1–5.6 mg/L increases the probability of closure to 15% and the maximum C _{trough72hours} of 25.9 mg/mL results in a probability of closure of 63%. The estimated probability of closure upon different treatment strategies based on C _{trough72hours} when ibuprofen is given in the first week of life is presented in Figure  4 . The figure illustrates that the regimen of 20‐10‐10 mg/kg resulted in the highest values for C _trough72hour (Figure  S1 ) with the highest probability of closure compared to 10‐5‐5 mg/kg and PNA adjusted dosing. ⁸ Because the minimum PNA_start at which 20‐10‐10 mg/kg was administered in our population was 4 days, probabilities of closure below this age are extrapolated (dashed lines). Interindividual variability in PKs and the resulting variability in C _{trough72hours} and predicted probability are is presented in Figure  S2 , and variability in predicted exposure arising from the logistic regression model coefficient variation is presented in Figure  S3 .

Figure 4.

Probability of closure of the ductus arteriosus vs. PNA at treatment initiation during the first week of life, predicted by the logistic regression model based on the subset of the dataset with a maximal PNA at treatment initiation of 7 days. Dashed lines represent extrapolations of the study population, since the 20‐10‐10 mg/kg regimen was not administered to infants below a PNA of 4 days. PNA, postnatal age.

Sensitivity analyses on the assumptions on PDA status after 3 doses for infants without an ultrasound, and on the availability of PK samples for a subset of the population are presented in Tables  S1, S2, and S3 . Estimated probability and observed closure vs. identified predictors by the subgroup analysis are presented in Figure  S4 . Weighted residuals of the subgroup model are presented in Figure  S5 . These analyses show little sensitivity to the assumptions and adequate fit of the data.

DISCUSSION

In this study, we included a large cohort of preterm infants with a PDA that was treated with ibuprofen initiated at a wide range of PNAs. In the overall population, the odds of closure decrease with PNA_start and increase with GA, and lower birthweight for GA, and with these predictors no additional influence of ibuprofen exposure could be identified in this population that all received ibuprofen in varying dosing regimens. If treatment is initiated during the first week of life, however, concentrations of ibuprofen were predictive for closure where PNA_start was not. Importantly, a very low closure rate was observed in the study population, especially after the first week of life. Spontaneous closure is most likely to occur in the first days after birth, ⁷ , ¹⁸ which might explain these results. In our study population, very low exposure to ibuprofen was observed when treatment was initiated beyond the first week of life (Figure  2 ), even if high ibuprofen dosages were used as suggested by the meta‐analysis of Mitra et al. ⁵ Therefore, we cannot exclude that the low closure rates are explained by the low ibuprofen concentrations in these neonates with a PNA beyond the first week of life. Lower effectiveness of pharmacological treatment upon later treatment initiation was also found by Relangi et al., ¹⁹ however, low‐quality evidence suggests no important differences in mortality or PDA ligation upon treatment initiation during or beyond the first week of life. ⁶

In the first week of life, the odds of ductus closure were found to increase with C _{trough72hours}, and not by PNA_start. To our knowledge, only one exposure‐response study on ibuprofen for the treatment of PDA was performed before, namely by Hirt et al. ⁹ In this study, with a maximal PNA_start of 11 days, a target AUC_0‐72hours of 900 mg*hour/L was identified. In our population, this target was achieved in more than 50% of the population, whereas the closure rate was much lower (21% vs. 86% observed by Hirt et al. ⁹ ). Based on the PK model used to determine ibuprofen exposure, C _{trough72hours} was more variable than AUC_0‐72hours, especially at lower C _{trough72hours}, which might explain why C _{trough72hours} was found to be more predictive than AUC_0‐72hours in the subgroup analysis. Better prediction of closure by C _{trough72hours} was suggested before ¹⁵ and can be explained by a required minimal inhibition of COX. However, whereas a dosing regimen of 20‐10‐10 mg/kg may result in higher levels of C _{trough72hours}, higher peak levels and the higher risk for toxicity should be considered as well. ¹⁵ Especially during the first days of life where the risk of intraventricular hemorrhage is high due to hyporeactivity of platelets, ²⁰ additional risks of high ibuprofen doses should be carefully outweighed against expected benefits. In order to achieve higher C _{trough72hours} values, twice‐daily dosing, as previously suggested by Flint et al., ¹⁵ would be a good alternative. Continuous administration of ibuprofen has also been shown to result in higher efficacy with fewer complications. ²¹ PNA adapted dosing is necessary to obtain comparable C _{trough72hours} for all preterm infants.

As expected, GA was predictive for closure. PDA is most common in the most extreme preterm infants, who unfortunately also might be most vulnerable for the side effects of ibuprofen. It remains a challenge to balance the expected benefit of ibuprofen treatment early in life with increased risk for adverse events. Even though clearance is estimated to increase with GA, similar exposure across the range in GA was observed. At lower GAs the ductus is most sensitive for vasodilatory signals and less to vasoconstricting signals, ² which might be translated into a need for higher ibuprofen exposure if GA is low. Based on the available dataset, it was not possible to identify a target C _{trough72hours}, or to identify whether such a target C _{trough72hours} is dependent on GA or PNA_start. In order to examine the need for an individualized target, further studies are necessary in which doses are adapted such that C _{trough72hours} across the range in PNA_start. Safety of higher exposure with potentially increased risks for side effects, including kidney failure and intraventricular hemorrhages, should also be carefully evaluated.

Infants with a low Z _{Birthweight‐for‐GA} had an increased probability of ductus closure in our analyses. This is an interesting finding that is in line with a previous study that found a similar increased effect in small for GA infants. ²² According to the population PK model used to calculate exposure, SGA infants (Z _{Birthweight‐for‐GA} < −2) had an increased clearance of ibuprofen, ⁸ resulting in lower C _{trough72hours} compared with their weight appropriate counterparts (Figure  2 ). Therefore, counterintuitively, the relatively smallest infants are expected to have a higher probability of closure with lower C _{trough72hours}. Neither the effect of being SGA on ibuprofen clearance, nor the effect of Z _{Birthweight‐for‐GA} on the probability of closure can be explained physiologically and would need further validation to adjust clinical practice to, but both findings can be interpreted as a sign that these infants require further studies, and should be examined as a separate group of patients that differ from appropriate for GA neonates.

A strength of this study is the combination of PK and pharmacodynamics analyses in a cohort of patients with a wide range in PNA at ibuprofen treatment initiation. PKs are rapidly changing in the neonatal population, resulting in up to two‐fold differences in minimum plasma concentration (C _min)_72hours and AUC_0‐72hours between treatment initiation upon the same dose in mg/kg on PNA days 1 and 7. To account for these changes, it is essential to examine exposure, and not just the dose per kg bodyweight. This is illustrated by the fact that the cumulative first three dose amounts were not a significant predictor for closure of the PDA in the univariate analysis, but C _{trough72hours} was.

A limitation of the available dataset was the observational nature. In our cohort, a selection bias applies as all included patients had a PDA at a relatively late time point in life. The closure rate of the overall population was relatively low compared to other ibuprofen cohorts, ⁹ but also compared with recorded spontaneous closure. ⁷ Physicians had freedom to deviate from treatment protocols, which make these results a good reflection of real practice, but also dependent on the neonatal intensive care unit and corresponding treatment practices, and increased the number of assumptions that had to be made. For example, the status of the PDA was assumed to be open after three doses if the ultrasound was postponed. This assumption was deemed appropriate for the majority of affected infants and was therefore preferred over excluding all patients with a postponed ultrasound because these were not selected randomly but based on clinical sings, which therefore would introduce more bias. Additionally, because this study was not placebo‐controlled the results should be interpreted with caution. For further optimization of the treatment of PDA, in future studies with ibuprofen, both spontaneous closure and changing PKs should be considered during the study design. Another limitation was that plasma samples were not available for all patients. Despite the inclusion of covariates on ibuprofen clearance, interindividual variability remained high (40%) in the PK study meaning that true exposure might deviate from the population predicted exposure. ⁸ Plasma samples for all patients could therefore have resulted in more precise exposure calculations, combined with more knowledge on bioavailability of ibuprofen which was now assumed to be 100%. The assumption that clearance did not further increase beyond a PNA of 18 days might have resulted in slightly overestimated exposure for infants above PNA_start of 18 days. However, due to the limited number of infants and the already very low predicted exposure for these infants, we expect little impact of this assumption.

In conclusion, in our cohort of relatively late treated preterm born infants, we showed a very low closure rate, especially after the first week of life. There is a window of opportunity for ductus closure within the first days of life, with a probability of closure that depends on GA, birthweight‐for‐GA, and ibuprofen exposure. Increased, yet unstudied dosages might be needed to increase the exposure and effect of ibuprofen beyond the first week.

FUNDING

No funding was received for this work.

CONFLICT OF INTEREST

The authors declared no competing interests for this work.

AUTHOR CONTRIBUTIONS

A.G.J.E., S.V., R.B.F., and S.H.P.S. wrote the manuscript. J.d.K., I.K.M.R., and S.H.P.S. designed the research. A.G.J.E., S.V., B.G., J.d.K., E.H.J.K., M.v.D., S.W., I.K.M.R., C.A.J.K., and S.H.P.S. performed the research. A.G.J.E. and S.V. analyzed the data.

Supporting information

Supinfo S1