Use of ibuprofen for the closure of patent ductus arteriosus in preterm infants: a systematic review of meta-analyses

MAs of randomized controlled trials (RCTs) or quasi-randomized controlled trials (QRCTs) and observational studies, investigating the use of ibuprofen, including versus paracetamol or indomethacin, for the treatment of PDA in preterm infants. Excluded publications included expert opinions, narrative and qualitative reviews, previous versions of current/updated MAs, and ibuprofen's use as prevention against PDA. Also excluded were MAs of studies that were all included in more recent MAs that are included in our analysis. Such older MAs, however, were included in the current analysis if reported pairwise comparisons between interventions that were not included in the more recent MAs.

Preterm infants (≤37 weeks) with HSPDA. There were no inclusion or exclusion criteria on birth weight or postnatal age.

Ibuprofen versus indomethacin or paracetamol, via any formulation and dosing regimen, or comparing different formulations or doses of ibuprofen, for PDA treatment.

Outcomes of interest were efficacy and safety outcomes of ibuprofen as reported in pairwise comparisons included in the MAs. Successful PDA closure was the outcome of interest, and there were no inclusion or exclusion criteria on the reporting of other outcomes. The analysis in this review only included outcomes that are reported in two or more MAs/NMAs. As reported in included studies, the effect measure of ibuprofen as a risk ratio (RR), odds ratio (OR) or mean difference (MD). These were provided at 95% confidence intervals (CI). For NMAs, the SUCRA (surface under the cumulative ranking curve) measures could be included, presented by a mean SUCRA score and a median ranking for the treatment modality.

We searched Embase, PubMed and the Cochrane Database of Systematic Reviews (CDSR). Search terms used were based on variations of the key terms ‘ibuprofen’ and ‘patent ductus arteriosus’ in literature, from inception to June 2020. Detailed search strategies are in Supplementary Material 2. Google Scholar and references of the included studies were manually searched for additional relevant articles in the gray literature over the same search duration. An English language restriction was imposed on the search.

Two reviewers independently screened the identified studies based on the above criteria, including duplicate removal, first by title and abstract for an initial eligibility assessment, before a final full-text screening. Any disagreement was resolved by consensus and referral to a third reviewer. With a similar process, two reviewers independently extracted the main study characteristics and the outcomes of interest.

As generally performed in the literature [9], the quality of methods and risk of bias are evaluated for included MAs to enable the interpretation of conclusions identified from the MAs. Two reviewers independently assessed MAs' methodological quality based on AMSTAR-2 (A Measurement Tool to Assess Systematic Reviews-2, https://amstar.ca/Amstar_Checklist.php) [10]. No identified tool has been proposed to especially evaluate the methodological quality of NMAs. Because of its good validity, reliability and responsibility, AMSTAR-2 is currently the most commonly used tool for the assessment of methods quality in MAs and, in the current study, consistent with relevant similar studies in the literature, will also be used to assess the quality in NMAs [11]. The quality of methods was rated as high, moderate, low and critically low. Disagreements were resolved by consensus and referral to a third reviewer.

The ROBIS (Risk of Bias in Systematic Reviews) tool [12] is most proposed for assessing the risk of bias in MAs, including NMAs. Two reviewers independently evaluated the risk of bias in the studies. The risk of bias was rated ‘high risk of bias’, ‘unclear risk of bias’, or ‘low risk of bias’. Disagreements were resolved by consensus and referral to a third reviewer.

Based on the AMSTAR-2 assessment of the four studies, the quality of methods was assessed to be low in three MAs [23–25], critically low in one MA [26], moderate in one study [22] and high in the two Cochrane studies [13,21]. Assessment items were studies underperformed can be seen in Supplementary Material 6.

Based on the ROBIS assessment, the risk of bias was determined to be high in two studies [25,26] and unclear in two studies [23,24] and low in three studies [13,21,22]. Domains of ROBIS where studies underperformed were identifying and selecting studies, data collection and study appraisal, and the synthesis and findings (Supplementary Material 7).

The use of mixed oral and IV of ibuprofen was assessed against mixed oral and IV formulations of indomethacin or paracetamol in several studies. The rate of PDA closure was not statistically different using ibuprofen against indomethacin [13,23]. A no-difference trend was also observed against paracetamol [21,23]. For the need for surgical ligation efficacy outcome, ibuprofen was compared with indomethacin and paracetamol, where no difference was measured [13,21,23].

Regarding adverse events, ibuprofen was inferior to indomethacin when it comes to neonatal mortality [23]. Based on the same MA, however, analyzing ibuprofen versus indomethacin based on direct and indirect comparisons of RCTs and observational studies, ibuprofen did not differ in mortality from indomethacin. Against paracetamol, mortality did not differ with ibuprofen [21,23,25].

Against indomethacin, ibuprofen use was significantly associated with a lower incidence of NEC, IVH, oliguria, increased creatinine level, intestinal perforation, platelet count and days on ventilator support, at high AMSTAR and low ROBIS [13]. As per a low AMSTAR with unclear ROBIS NMA, by Marconi et al. [23], however, the NEC advantage with ibuprofen was cancelled, as no difference in GI bleeding was reported, and lower rates of BPD and significantly reduced serum bilirubin levels were associated with indomethacin. Compared with paracetamol, ibuprofen was not differently associated with neonatal mortality, NEC, BPD, IVH and oliguria events [23,25]. However, paracetamol was associated with a lower incidence of GI bleeding [21,23,25], increased serum bilirubin [21], and decreased platelet count post-treatment [21].

Three MAs [22,24,26] reported no difference between IV ibuprofen and IV indomethacin for successful PDA closure. However, against oral paracetamol, IV ibuprofen was reported to be inferior for PDA closure based on one MA, with moderate AMSTAR and low ROBIS [22], including when IV ibuprofen is via continuous infusion. Based on the SUCRA analysis, by Mitra et al. [22], the IV ibuprofen had a low median rank of 8 among ten treatment modalities for PDA closure (Table 5). Regarding the need for surgical ligation, in two studies, the standard dose IV ibuprofen did not differ from the IV indomethacin [22,26].

The neonatal mortality, NEC, IVH, GI bleeding, and intestinal perforation did not differ between IV ibuprofen and IV indomethacin based on three studies [22,24,26]. This included the IV continuous infusion of ibuprofen, for mortality and NEC [22]. No difference in the mortality, NEC, and IVH events rates was also reported against paracetamol in one study [22], including the IV continuous infusion of ibuprofen for NEC [22]. Based on the Mitra et al. SUCRA analysis [22], the IV ibuprofen ranked sixth among ten treatment modalities for association with NEC events (Table 7).

One MA reported IV ibuprofen to be inferior to IV indomethacin concerning the BPD events, with low AMSTAR and unclear ROBIS [24]. However, the rate of the BPD event between IV ibuprofen and indomethacin was not found different in two other MAs, at moderate AMSTAR and low ROBIS [22] and critically low AMSTAR and high ROBIS [26], including in the case when the IV ibuprofen was a continuous infusion [22]. The BPD with IV ibuprofen was also not different compared with oral paracetamol [22].

IV ibuprofen had a significantly lesser association with elevated creatinine levels and oliguria than IV indomethacin [22,26]. However, oliguria was not different between IV ibuprofen and oral paracetamol [22].

For PDA closure, oral ibuprofen was not statistically different from IV indomethacin based on two MAs [22,26], and was also not different compared with both oral and IV indomethacin in a third study [13]. Against oral paracetamol, oral ibuprofen was not different for the PDA closure in two MAs [22,25]. Out of four pooled pairwise analyses that evaluated the need for surgical ligation with oral ibuprofen versus indomethacin, three reported no difference compared with IV indomethacin [13,22,26]. In the remaining study, oral ibuprofen did not perform differently from mixed oral and IV indomethacin administration [13].

Against IV indomethacin, oral ibuprofen was not statistically different in neonatal mortality [22,24], and also not significantly different in BPD [22,26]. Mortality and BPD were not different between oral ibuprofen and oral paracetamol [22]. Based on a SUCRA analysis [22], oral ibuprofen has third and second lowest-risks for mortality and BPD, respectively, among 10 treatment modalities (Table 6 & 7). IVH was not different between oral ibuprofen and IV indomethacin or oral paracetamol [22]. The only GI bleeding study with oral ibuprofen reported a not statistical difference from IV indomethacin [26]. Oral ibuprofen was significantly associated with less NEC compared with mixed oral and IV indomethacin or IV indomethacin [13,22]. In contrast, oral ibuprofen was reported as not different from IV indomethacin for NEC events in one study with critically low AMSTAR and high ROBIS [26]. Oral ibuprofen was also compared against oral paracetamol with no significant difference in the incidence of NEC [22]. Oliguria was statistically less with oral ibuprofen compared with IV indomethacin [22] but was not different compared with oral paracetamol [22]. Oral ibuprofen was ranked fourth in risk for oliguria among ten treatment modalities based on one SUCRA analysis (Table 11) [22]. In one MA, oral ibuprofen was not different in creatinine levels compared with IV indomethacin, but was associated with a significantly reduced increase in creatinine levels compared with oral indomethacin [26]. In another study, against mixed oral and IV formulations use of indomethacin, no difference was reported with oral ibuprofen [13].

Only one study evaluated the effect of ibuprofen formulation on PDA closure [13]; oral ibuprofen had a lower risk of failure to close the PDA than IV ibuprofen. In the same study, however, for the need for surgical ligation, no difference was observed.

For adverse events, BPD and IVH were not different between the two formulations in one study [22]. With a similar trend, but at a higher AMSTAR, oliguria did not differ between the two formulations in another study [13]. Only the serum creatinine level differed between the formulations, where oral ibuprofen was associated with lower increase in levels [13]. One MA reported the need for retreatment with oral versus IV ibuprofen to close the PDA, which was statistically significant with oral ibuprofen [22]. Other efficacy and safety outcomes were not different between the oral and IV formulations.

One MA, by Mitra et al., reported the difference between HD ibuprofen and IV indomethacin [22], where HD oral ibuprofen was associated with a significant advantage in relation to PDA closure and the need for surgical ligation. Still, it was not different with neonatal mortality, NEC, and BPD. HD IV ibuprofen performed differently from HD oral ibuprofen against IV indomethacin, where there was no difference for both PDA closure and the need for surgical ligation. HD IV ibuprofen was not different from IV indomethacin concerning neonatal mortality, NEC, BPD, IVH, and oliguria [22].

According to Mitra et al., HD oral and IV ibuprofen performed similarly against oral paracetamol [22]. No difference with HD ibuprofen against the oral paracetamol was observed for PDA closure and the need for surgical ligation. A similar trend was observed for neonatal mortality and NEC.

Regarding the effect with HD versus standard dose of mixed oral and IV ibuprofen use, by Ohlsson et al. [13], PDA closure was significantly better achieved when at HD, but not for the need for surgical ligation, based on one RCT. A similar trend of no difference was observed with oliguria and creatinine levels, based on one RCT [13].

Looking at the ibuprofen formulations separately, one MA found no difference between HD oral ibuprofen versus standard-dose oral ibuprofen for PDA closure [22], but reported that HD of either oral or IV ibuprofen was more effective than standard-dose IV ibuprofen for PDA closure. For the need for surgery, this was significantly lesser with HD oral ibuprofen than standard doses of either oral or IV ibuprofen [22]. In contrast, the HD IV ibuprofen did not perform differently for the same outcome from the standard doses of either oral or IV ibuprofen [22]. Based on the same MA, HD IV ibuprofen did not perform differently from either oral or IV ibuprofen for BPD and IVH events.

This is consistent with the SUCRA analysis performed by Mitra et al. in their NMA [22]; whereby, for PDA closure, out of 9 treatment modalities, the two highest-ranked were HD oral ibuprofen and HD IV ibuprofen. Based on the 95% confidence intervals reported, no significant difference could be seen within at least the first five ranked treatments, including oral ibuprofen and oral paracetamol (Table 4).

In the current study, general observations are consistent with recent trends of having ibuprofen as first-line therapy for PDA in neonates, particularly against the older and previous standard treatment, indomethacin, due to reported enhanced efficacy and reduced toxicity in the literature [16,31–34]. Paracetamol is a more recently suggested alternative for PDA, where it demonstrated efficacy and safety, particularly advantageous in patients with clinical contraindications to ibuprofen [6]. Our observation is that paracetamol, overall, performed better than indomethacin compared with ibuprofen. Also consistent with our observations, is the recent trends of increased interest in the off-label oral ibuprofen compared with IV ibuprofen, particularly as IV ibuprofen may not be available in all practices, especially in low-income countries. Oral ibuprofen has the advantage of a lower acquisition cost compared with IV ibuprofen, added to ease of availability and administration [35–41].

The current study's applicability is limited by the lack of information about a consistent administration timing of ibuprofen medication. The medication timing highly depends on the clinical diagnostic criteria used to define the HSPDA. A study by Koehne et al. found, however, that it is normal for HSPDA to be clinically silent for the first 2–3 days of life and, hence, accurate and early diagnosis of HSPDA best depends on early ECHO assessment [42]. There is, nonetheless, no evidence in included MAs of overall consistency in how early and how often the ECHO was performed for PDA diagnosis in individual studies. In the Cochrane review, by Ohlsson et al. [13], for example, the diagnosis among included RCTs varied from prophylactic (within 24 h) to 2, 3, 7 or 21 days of life, to selective, to not stated.

Several limitations exist in the current systematic review. The analysis only included outcomes reported in two or more MAs. Hence, this did not include all reported efficacy and safety outcomes in the literature, excluding, for example, neurodevelopment impairment. This does not undermine how important such outcomes are. However, given the objective of the current review, which is to simplify access to summaries of outcomes where an information overload exists, a decision was made whereby outcomes that are reported in only one MA are easily accessible and interpretable, noting that, for the interest of care providers, we identified the literature studies were these were reported (Supplementary Material 5). Besides, some studies were included in more than one included MAs, which may create double counting of data, which is inherent in systematic reviews of MAs and is challenging to eliminate. However, as already discussed, we looked to minimize this limitation's impact by excluding older versions of more recent MAs' updates. The language-restricted search is another limitation and might have missed relevant non-English published studies. No resources were available to authors, however, to translate non-English studies if found. Furthermore, while we believe that our literature search was comprehensive and included all relevant studies, it is always possible that additional search terms and combinations of them may identify other studies.