Types of studies

Prospective randomised and quasi‐randomised trials. We excluded short‐term cross‐over studies as not being informative for our primary outcome measure.

Types of participants

Preterm infants (< 34 weeks) with early or established BPD. Established BPD was defined as oxygen need and evidence of respiratory insufficiency at 28 days of age. Early BPD was defined as a persistent oxygen need, respiratory insufficiency and high predicted likelihood of development of established BPD after a minimum of 21 days of age. Maximum age of enrolment into the study was 36 weeks' postmenstrual age.

Types of interventions

We searched for studies comparing two different levels of fluid intake. Because there is a wide range of possible fluid intakes in such infants, in order to group the studies we assumed that normal fluid intake is 140 to 150 ml/kg/day, and we intended to group the subjects into unrestricted (≥ 150 ml/kg/day) and fluid restricted (< 150 ml/kg/day).

Types of outcome measures

Electronic searches

We conducted a comprehensive search including: Cochrane Central Register of Controlled Trials (CENTRAL; 2016, Issue 1) in the Cochrane Library (searched 16 February 2016); MEDLINE via PubMed (1966 to 16 February 2016); Embase (1980 to 16 February 2016); and CINAHL (1982 to 16 February 2016) using the following search terms: (fluid OR food, formulated) AND (bronchopulmonary dysplasia OR BPD OR chronic lung disease OR CLD), plus database‐specific limiters for RCTs and neonates (see Appendix 1 for the full search strategies for each database). We did not apply language restrictions.

We searched clinical trials' registries for ongoing or recently completed trials (ClinicalTrials.gov; the World Health Organization’s International Clinical Trials Registry Platform www.who.int/ictrp/search/en/; and the ISRCTN Registry).

Searching other resources

We examined the references in all studies identified as potentially relevant.

Selection of studies

All three review authors screened the title and abstract of all studies identified by the above search strategy. We assessed the full text of any potentially eligible reports and excluded those studies that did not meet all of the inclusion criteria. We discussed any disagreements until consensus was achieved.

Data extraction and management

All three review authors extracted the data separately. Any disagreements were discussed until consensus was achieved. We contacted trial authors for further information if data from the trial reports was insufficient.

Assessment of risk of bias in included studies

We used the criteria and standard methods of Cochrane Neonatal. The methodological quality of the studies was assessed using the following key criteria: allocation concealment (blinding of randomisation), blinding of intervention, completeness of follow‐up, and blinding of outcome measurement/assessment. For each criterion, assessment was 'yes', 'no' or 'unclear'. Additional information from the trial authors was requested to clarify methodology and results as necessary. This information was included in the table 'Characteristics of included studies'.

In addition, the review authors independently assessed the risk of bias for each study using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

The methodological quality of the studies was assessed using the following criteria.

1. Sequence generation (evaluating possible selection bias). For each included study, we described the method used to generate the allocation sequence as: adequate (any truly random process e.g. random number table; computer random number generator); inadequate (any non‐random process e.g. odd or even date of birth; hospital or clinic record number); or unclear.

2. Allocation concealment (evaluating possible selection bias). For each included study, we described the method used to conceal the allocation sequence as: adequate (e.g. telephone or central randomisation; consecutively numbered sealed opaque envelopes); inadequate (open random allocation; unsealed or non‐opaque envelopes; alternation; date of birth); or unclear.

3. Blinding (evaluating possible performance bias). For each included study, we described the methods used to blind study participants and personnel from knowledge of which intervention a participant received. Blinding was assessed separately for different outcomes or classes of outcomes. We assessed the methods as: adequate, inadequate, or unclear for participants; adequate, inadequate, or unclear for study personnel; and adequate, inadequate, or unclear for outcome assessors.

4. Incomplete outcome data (evaluating possible attrition bias through withdrawals, dropouts, protocol deviations). For each included study and for each outcome, we described the completeness of data including attrition and exclusions from the analysis. We stated whether attrition and exclusions were reported, the numbers included in the analysis at each stage (compared with total randomised participants), reasons for attrition or exclusion where reported, and whether missing data were balanced across groups or were related to outcomes. We assessed methods as: adequate (< 20% missing data); inadequate (> 20% missing data); or unclear;

5. Selective reporting bias. For each included study where the protocol is available (through trials' registers), we described how we investigated the possibility of selective outcome reporting bias and what we found. We assessed the methods as: adequate (where it was clear that all of the study's prespecified outcomes and all expected outcomes of interest to the review had been reported); inadequate (where not all the study's prespecified outcomes had been reported; one or more reported primary outcomes were not prespecified; outcomes of interest were reported incompletely and so could not be used; study failed to include results of a key outcome that would have been expected to have been reported); or unclear.

6. Other sources of bias. We noted other possible sources of bias (for example, whether there was a potential source of bias related to the specific study design or whether the trial was stopped early owing to some data‐dependent process). We assessed whether each study was free of other problems that could put it at risk of bias as: yes; no; or unclear.

7. Funding sources: when reported we divided funding sources into funding from local funds (hospital, research centre, or university), charitable foundations, government agencies, or from industry. If funding was from industry, and industry also had substantial input into study design or conduct or analysis, then that was determined to be a high risk of bias. Funding from sources that had no potential financial benefit from the results was considered to be a low risk of bias. Unclear or partial funding was considered unclear risk of bias.

Measures of treatment effect

We calculated risk ratio (RR) and risk difference (RD) for dichotomous data and mean difference (MD) for continuous data, with respective 95% confidence intervals (CIs). When it was deemed appropriate to combine two or more study arms, we obtained the treatment effects from the combined data using the methods described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We determined the number needed to treat for an additional beneficial outcome (NNTB) or number needed to treat for an additional harmful outcome (NNTH) for a statistically significant difference in the RD.

Unit of analysis issues

The only trial included was an individually randomised parallel group trial.

Dealing with missing data

We requested additional data from the trial investigators for data on important outcomes which were missing.

Assessment of heterogeneity

If more than one trial was included in a meta‐analysis, we planned to examine the treatment effects of individual trials and heterogeneity between trial results by inspecting the forest plots. We planned to calculate the I² statistic for each analysis to quantify inconsistency across studies and describe the percentage of variability in effect estimates that may be due to heterogeneity rather than sampling error. The I² statistic was to be analysed as follows: less than 25% = no heterogeneity; 25% to 49% = low heterogeneity; 50% to 74% = moderate heterogeneity; and 75%+ = high heterogeneity. If moderate or high heterogeneity was detected (I² statistic > 50%), we would have explored the possible causes (for example, differences in study design, participants, interventions, or completeness of outcome assessments) in sensitivity analyses.

Assessment of reporting biases

We planned to investigate publication bias by using funnel plots if at least 10 clinical trials were included in the systematic review (Higgins 2011).

Data synthesis

We used the standard methods of the Cochrane Neonatal Review Group. For categorical outcomes, typical estimates for relative risk (RR) and risk difference (RD) were calculated along with their 95% confidence intervals. For significant results, we planned on reporting the number needed to treat for an additional beneficial outcome or number needed to treat for an additional harmful outcome (NNTB/NNTH). A fixed‐effect model was assumed. Continuous outcomes were analysed using weighted mean difference, assuming a fixed‐effect model.

Subgroup analysis and investigation of heterogeneity

Subgroup analysis for the severity of the disease, the different levels of fluid restriction and the duration of the intervention were planned but could not be performed as only one study was found and data for these subgroups was unavailable.

Sensitivity analysis

Sensitivity analysis was planned to eliminate trials considered of low quality, but only one trial was found.