Summary of findings 1. Summary of findings.

Outcome	Effects and confidence in the effect estimates						Comments**
	Indomethacin		Ibuprofen		Acetaminophen
Severe Intraventricular Haemorrhage
Placebo comparator 127 per 1000 (12.7%)	Network RR 0.66 (0.49, 0.87)	Network absolute risk difference* 43 fewer per 1000 (from 65 fewer to 16 fewer)	Network RR 0.69 (0.41, 1.14)	Network absolute risk difference 39 fewer per 1000 (from 75 fewer to 18 more)	Network RR 1.17 (0.04, 55.2)	Network absolute risk difference 22 more per 1000 (from 122 fewer to 1000 more)	· Prophylactic indomethacin probably results in a small reduction in severe IVH · Prophylactic ibuprofen probably results in a small reduction in severe IVH · The evidence is very uncertain about the effect of prophylactic acetaminophen on severe IVH
	Moderate ⊕⊕⊕◯ Confidence in estimate due to imprecision1		Moderate ⊕⊕⊕◯ Confidence in estimate due to imprecision2		Very Low ⊕◯◯◯ Confidence in estimate due to imprecision3
Rank [Median (95% CrIs)] 3 (2‐4)	Rank 2 (1‐3)		Rank 2 (1‐4)		Rank 4 (1‐4)
	Based on 2629 infants (16 RCTs)		Based on 863 infants (6 RCTs)		Based on 48 infants (1 RCT)
Mortality
Placebo comparator 161 per 1000 (16.1%)	Network RR 0.85 (0.64 to 1.1)	Network absolute risk difference 24 fewer per 1000 (from 58 fewer to 16 more)	Network RR 0.83 (0.57 to 1.2)	Network absolute risk difference 27 fewer per 1000 (from 69 fewer to 32 more)	Network RR 0.49 (0.16 to 1.4)	Network absolute risk difference 82 fewer per 1000 (from 135 fewer to 64 more)	· Prophylactic indomethacin probably results in a moderate reduction in mortality · Prophylactic ibuprofen may result in a moderate reduction in mortality · The evidence is very uncertain about the effect of prophylactic acetaminophen on mortality
	Moderate ⊕⊕⊕◯ Confidence in estimate due to imprecision4		Low ⊕⊕◯◯ Confidence in estimate due to imprecision5		Very Low ⊕◯◯◯ Confidence in estimate due to risk of bias and imprecision6
Rank [Median (95% CrIs)] 4 (3‐4)	Rank 2 (1‐4)		Rank 2 (1‐4)		Rank 1 (1‐4)
	Based on 2877 infants (19 RCTs)		Based on 914 infants (7 RCTs)		Based on 208 infants (2 RCTs)
Surgical PDA closure
Placebo comparator 87 per 1000 (8.7%)	Network RR 0.40 (0.14 to 0.66)	Network absolute risk difference 52 fewer per 1000 (from 75 fewer to 30 fewer)	Network RR 0.24 (0.06 to 0.64)	Network absolute risk difference 66 fewer per 1000 (from 82 fewer to 31 fewer)	_______	_______	· Prophylactic indomethacin probably results in a moderate reduction in need for surgical PDA closure · Prophylactic ibuprofen probably results in a moderate reduction in need for surgical PDA closure · There is no evidence on the effect of prophylactic acetaminophen on need for surgical PDA closure
	Moderate ⊕⊕⊕◯ Confidence in estimate due to imprecision7		Moderate ⊕⊕⊕◯ Confidence in estimate due to imprecision8		_______
Rank [Median (95% CrIs)] 3 (3‐3)	Rank 2 (1‐2)		Rank 1 (1‐2)		_______
	Based on 1800 infants (11 RCTs)		Based on 873 infants (6 RCTs)		_______
Necrotizing Enterocolitis
Placebo comparator 65 per 1000 (6.5%)	Network RR 0.76 (0.35 to 1.2)	Network absolute risk difference 16 fewer per 1000 (from 42 fewer to 13 more)	Network RR 0.73 (0.31 to 1.4)	Network absolute risk difference 18 fewer per 1000 (from 45 fewer to 26 more)	_______	_______	· Prophylactic indomethacin results in trivial difference in NEC · Prophylactic ibuprofen results in trivial difference in NEC · There is no evidence on the effect of prophylactic acetaminophen on NEC
	High ⊕⊕⊕⊕ Confidence in estimate		High ⊕⊕⊕⊕ Confidence in estimate		_______
Rank [Median (95% CrIs)] 3 (3‐3)	Rank 2 (1‐3)		Rank 1 (1‐3)		_______
	Based on 2543 infants (14 RCTs)		Based on 905 infants (7 RCTs)		_______
Gastrointestinal perforation
Placebo comparator 47 per 1000 (4.7%)	Network RR 0.92 (0.11 to 3.9)	Network absolute risk difference 4 fewer per 1000 (from 42 fewer to 137 more)	Network RR 2.6 (0.42 to 20.0)	Network absolute risk difference 76 more per 1000 (from 27 fewer to 897 more)	_______	_______	· Prophylactic indomethacin probably results in trivial difference in gastrointestinal perforation · The evidence is very uncertain about the effect of prophylactic ibuprofen on gastrointestinal perforation · There is no evidence on the effect of prophylactic acetaminophen on gastrointestinal perforation
	Moderate ⊕⊕⊕◯ Confidence in estimate due to imprecision9		Very Low ⊕◯◯◯ Confidence in estimate due to imprecision10		_______
Rank [Median (95% CrIs)] 2 (1‐3)	Rank 1 (1‐3)		Rank 3 (1‐3)		_______
	Based on 1221 infants (2 RCTs)		Based on 177 infants (2 RCTs)		_______
Chronic Lung Disease
Placebo comparator 359 per 1000 (35.9%)	Network RR 1.10 (0.93 to 1.3)	Network absolute risk difference 36 more per 1000 (from 25 fewer to 108 more)	Network RR 1.00 (0.83 to 1.3)	Network absolute risk difference 0 fewer per 1000 (from 61 fewer to 108 more)	_______	_______	· Prophylactic indomethacin may result in a small increase in chronic lung disease · Prophylactic ibuprofen may result in trivial difference in chronic lung disease · There is no evidence on the effect of prophylactic acetaminophen on chronic lung disease
	Low ⊕⊕◯◯ Confidence in estimate due to inconsistency and imprecision11		Low ⊕⊕◯◯ Confidence in estimate due to imprecision12		_______
Rank [Median (95% CrIs)] 1 (1‐3)	Rank 3 (1‐3)		Rank 2 (1‐3)		_______
	Based on 2106 infants (10 RCTs)		Based on 904 infants (7 RCTs)		_______
Cerebral Palsy
Placebo comparator 110 per 1000 (11%)	Network RR 0.97 (0.44 to 2.1)	Network absolute risk difference 3 fewer per 1000 (from 62 fewer to 121 more)	_______	_______	Network RR 0.36 (0.01 to 6.3)	Network absolute risk difference 70 fewer per 1000 (from 109 fewer to 583 more)	· Prophylactic indomethacin may result in trivial difference in cerebral palsy · There is no evidence on the effect of prophylactic ibuprofen on cerebral palsy · The evidence is very uncertain about the effect of prophylactic acetaminophen on cerebral palsy
	Low ⊕⊕◯◯ Confidence in estimate due to imprecision13		_______		Very Low ⊕◯◯◯ Confidence in estimate due to imprecision14
Rank [Median (95% CrIs)] 2 (1‐3)	Rank 2 (1‐3)		_______		Rank 1 (1‐3)
	Based on 1367 infants (4 RCTs)		_______		Based on 35 infants (1 RCT)

1. In the direct comparison, the credible intervals include moderate benefit (73 fewer per 1000) to small benefit (27 fewer per 1000). Therefore, the certainty of evidence was rated down by one level for imprecision. No further change was made based on the network estimates

2. In the direct comparison, the credible intervals include moderate benefit (82 fewer per 1000) to small harm (33 more per 1000). Therefore, the certainty of evidence was rated down by one level for imprecision. No further change was made based on the network estimates

3. 95% CrIs include appreciable benefit and very large harm. In the direct comparison, the certainty of evidence was rated down by one‐level for serious imprecision. Based on the network estimates, the certainty was rated down by two more levels due to very serious imprecision (implausible effect sizes) in the network estimates

4. In the direct comparison, the credible intervals include moderate benefit (61 fewer per 1000) to small harm (17 more per 1000). Therefore, the certainty of evidence was rated down by one level for imprecision. No further change was made based on the network estimates

5. In the direct comparison, the credible intervals include appreciable benefit (72 fewer per 1000) and harm (48 more per 1000). Therefore, the certainty of evidence was rated down by two levels for very serious imprecision. No further change was made based on the network estimates.

6. In the direct comparison, the certainty of evidence was rated down due to substantial risk of bias in the included studies; the certainty was further rated down two levels for very serious imprecision as the credible intervals include appreciable benefit (85 fewer per 1000) and harm (76 more per 1000). Therefore, the overall certainty of evidence for the direct estimate was rated as very low. No further change was made based on the network estimates.

7. In the direct comparison, the credible intervals include moderate benefit (88 fewer per 1000) to small benefit (25 fewer per 1000). Therefore, the certainty of evidence was rated down by one level for imprecision. No further change was made based on the network estimates

8. The certainty of evidence for the direct comparison was high. However, the 95% credible intervals in the network estimates include appreciable benefit (82 fewer) to small benefit (31 fewer). Hence, the certainty of evidence was rated down by one level due to imprecision

9. 95% CrIs of the network estimates include small benefit (42 fewer) to appreciable harm (137 more). Hence, the certainty of evidence was rated down by one level due to imprecision

10. In the direct comparison, the credible intervals included trivial benefit (7 fewer per 1000) to appreciable harm (191 fewer per 1000). Therefore, the certainty of evidence was rated down by one level for imprecision. 95% CrIs of the network estimates include small benefit (27 fewer) to very large harm (897 more). Hence, the certainty was rated down by two more levels due to very serious imprecision (implausible effect sizes) in the network estimates.

11. In the direct comparison, the certainty of evidence was rated down one level due to serious inconsistency; the certainty was further rated down one level for imprecision as the credible intervals include small benefit (33 fewer per 1000) to appreciable harm (111 more per 1000). Therefore, the overall certainty of evidence for the direct estimate was rated as low. No further change was made based on the network estimates.

12. In the direct comparison, the credible intervals include moderate benefit (86 fewer per 1000) to large harm (132 more per 1000). Therefore, the certainty of evidence was rated down by two levels for imprecision (as the confidence limits include appreciable benefit or harm). No further change was made based on the network estimates

13. In the direct comparison, the credible intervals include moderate benefit (60 fewer per 1000) to large harm (111 more per 1000). Therefore, the certainty of evidence was rated down by two levels for imprecision (as the credible intervals include appreciable benefit and harm). No further change was made based on the network estimates

14. In the direct comparison, the credible intervals include moderate benefit (59 fewer per 1000) to very large harm (797 more per 1000). Therefore, the certainty of evidence was rated down by two levels for imprecision (as the credible intervals include appreciable benefit and harm). The 95% CrIs of the network estimates include large benefit (109 fewer) to very large harm (583 more). Hence the certainty of evidence was rated down by one more level due to very serious imprecision (implausible effect sizes) in the network estimates

* A network absolute risk difference was calculated from the network RR estimates using an assumed control risk that was derived by dividing the total event number by the total infant number in the control groups in the network

**Comments on interpretation of effect sizes are based on a priori defined thresholds as follows: (a) For the outcome of mortality: Small benefit/harm was defined as <20 fewer or more per 1000, respectively. Moderate benefit/harm was defined as 20 to 50 fewer or more per 1000, respectively. Large benefit/harm was defined as >50 fewer or more per 1000 respectively; (b) For all other outcomes listed in the summary of findings table: Any effect <20 fewer or more per 1000 was defined as a trivial benefit or harm. No direction of effect was specified for trivial effects. Small benefit/harm was defined as 20‐50 fewer or more per 1000 respectively. Moderate benefit/harm was defined as 50‐100 fewer or more per 1000 respectively. Large benefit/harm was defined as >100 fewer or more per 1000, respectively. Language for interpretation used in this column is based on the GRADE informative statements to communicate the findings of systematic reviews of interventions by Santesso 2020.

GRADE Working Group grades of evidence
High certainty: we are very confident that the true effect lies close to that of the estimate of the effect.
Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.
Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect.
Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.