Types of studies

We included controlled trials using random or quasi‐random patient allocation, including cluster‐randomised controlled trials. We did not include cross‐over trials. Studies published as abstracts were eligible for inclusion only if assessment of study quality was possible, and if other criteria for inclusion were fulfilled.

Types of participants

We included preterm infants fed formula (exclusively or as a supplement to human breast milk) after discharge from hospital. The intervention may have commenced up to one week before planned discharge from hospital. We did not include in this review trials that randomly assigned infants to nutrient‐enriched formula versus standard term formula more than one week before hospital discharge (and then continued the intervention after hospital discharge).

Types of interventions

• Standard term formula: energy content ≤ 72 kcal/100 mL and protein content ≤ 1.7 g/100 mL

versus

• Postdischarge formula: energy content > 72 kcal/100 mL (but ≤ 75 kcal/100 mL) and protein content > 1.7 g /100 mL; or

• Preterm formula: energy content > 75 kcal/100 mL and protein content > 2.0 g/100 mL.

The formula could be fed as the sole diet or as a supplement to human breast milk. Infants in trial groups should have received similar care other than the type of formula. Target levels prescribed for volume of intake and advice or support for demand feeding should have been no different among groups.

Types of outcome measures

Electronic searches

We updated the search on 8 September 2016 using a combination of the following terms: (infant nutrition OR infant formula OR milk OR formula OR nutrient OR fortif* OR supplemt*) AND (postdischarge OR post‐discharge OR discharge) plus the following database‐specific terms limited by relevant search filters for clinical trials, as recommended in the Cochrane Handbook for Systematic Reviews of Interventions.

• PubMed: ((infant, newborn[MeSH] OR newborn OR neonate OR neonatal OR premature OR low birth weight OR VLBW OR LBW or infan* or neonat*) AND (randomized controlled trial [pt] OR controlled clinical trial [pt] OR randomized [tiab] OR placebo [tiab] OR drug therapy [sh] OR randomly [tiab] OR trial [tiab] OR groups [tiab]) NOT (animals [mh] NOT humans [mh]))

• Embase: (infant, newborn or newborn or neonate or neonatal or premature or very low birth weight or low birth weight or VLBW or LBW or Newborn or infan* or neonat*) AND (human not animal) AND (randomized controlled trial or controlled clinical trial or randomized or placebo or clinical trials as topic or randomly or trial or clinical trial)

• Cumulative Index to Nursing and Allied Health Literature (CINAHL): (infant, newborn OR newborn OR neonate OR neonatal OR premature OR low birth weight OR VLBW OR LBW or Newborn or infan* or neonat*) AND (randomized controlled trial OR controlled clinical trial OR randomized OR placebo OR clinical trials as topic OR randomly OR trial OR PT clinical trial)

• Cochrane Library: (infant or newborn or neonate or neonatal or premature or very low birth weight or low birth weight or VLBW or LBW)

We applied no language restrictions.

We searched ClinicalTrials.gov and Current Controlled Trials for completed and ongoing trials.

Searching other resources

We examined the references provided in studies identified as potentially relevant. We searched abstracts from annual meetings of the Pediatric Academic Societies (1993 to 2016), the European Society for Pediatric Research (1995 to 2016), the UK Royal College of Paediatrics and Child Health (2000 to 2016) and the Perinatal Society of Australia and New Zealand (2000 to 2016). We considered trials reported only as abstracts to be eligible if sufficient information was available from the report, or from contact with study authors, to fulfil the inclusion criteria.

Selection of studies

Two review authors screened the titles and abstracts of all studies identified by the above search strategy. We reassessed the full texts of potentially eligible reports and excluded studies that did not meet all of the inclusion criteria. We discussed disagreements until consensus was achieved.

Data extraction and management

We used a data collection form to aid extraction of relevant information from each included study. Two review authors extracted study data separately. We discussed disagreements until consensus was achieved and asked investigators for further information if data provided in the trial reports were insufficient.

Assessment of risk of bias in included studies

We used criteria and standard methods of the Cochrane Neonatal Review Group to assess the methodological quality of included trials. Two review authors conducted assessment of risk of bias and resolved disagreements by discussion. We requested additional information from trial authors to clarify methods and results if necessary.

We made explicit judgements about whether studies were at high risk of bias across four domains, according to the criteria suggested in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

• Random sequence generation: We categorised the method used to generate the allocation sequence as:

  • low risk ‐ any truly random process (e.g. random number table, computer random number generator);

  • high risk ‐ any non‐random process (e.g. odd or even date of birth, hospital or clinic record number); or

  • unclear risk ‐ no or unclear information provided.

• Allocation concealment: We categorised the method used to conceal the allocation sequence as:

  • low risk (e.g. telephone or central randomisation, consecutively numbered sealed opaque envelopes);

  • high risk ‐ open random allocation (e.g. unsealed or non‐opaque envelopes, alternation; date of birth); or

  • unclear ‐ no or unclear information provided.

• Blinding: We assessed blinding of participants, clinicians and caregivers and outcome assessors separately for different outcomes and categorised the methods as:

  • low risk;

  • high risk; or

  • unclear risk.

• Incomplete outcome data: We described completeness of data, including attrition and exclusions from analysis for each outcome and reasons for attrition or exclusion, when reported. We assessed whether missing data were balanced across groups or were related to outcomes. When sufficient information was reported or was supplied by trial authors, we planned to reinstate missing data to the analyses. We categorised completeness as:

• • low risk: ≤ 10% missing data;

  • high risk: > 10% missing data; or

  • unclear risk: no or unclear information provided.

We assessed the likely magnitude and direction of bias and whether we considered it likely for bias to impact study findings. We planned to explore the impact of the level of bias by performing sensitivity analyses.

Measures of treatment effect

We analysed treatment effects in individual trials by using Review Manager 5.3 and reported risk ratios (RRs) and risk differences (RDs) for dichotomous data and mean differences (MDs) for continuous data, with respective 95% confidence intervals (CIs). We determined the number needed to treat for an additional beneficial outcome (NNTB) or an additional harmful outcome (NNTH) for analyses with statistically significant differences among RDs.

Unit of analysis issues

The unit of analysis was the participating infant in individually randomised trials. For cluster‐randomised trials (had we identified any for inclusion), we planned to undertake analyses at the level of the individual while accounting for clustering of data by using methods recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

Dealing with missing data

We requested additional data from trial investigators if data on important outcomes were missing or were reported unclearly. When data were still missing, we examined the impact of this on effect size estimates by performing sensitivity analyses.

Assessment of heterogeneity

We examined treatment effects in individual trials and heterogeneity between trial results by inspecting forest plots if more than one trial was included in a meta‐analysis. We calculated the I² statistic for each analysis to quantify inconsistency across studies and to describe the percentage of variability in effect estimates that may be due to heterogeneity rather than to sampling error. If we detected moderate or high (I² > 50%) heterogeneity, we explored possible causes (e.g. differences in study design, participants or interventions; completeness of outcome assessments) by performing sensitivity analyses.

Assessment of reporting biases

If more than five trials were included in a meta‐analysis, we conducted a funnel plot analysis.

Data synthesis

We used the fixed‐effect model in RevMan 5.1 to perform meta‐analysis.

Subgroup analysis and investigation of heterogeneity

We prespecified the following subgroup analyses.

• Very preterm (< 32 weeks' gestation) or VLBW (< 1500 g) infants (vs infants at 32 to 36 weeks' gestation or birth weight 1500 to 2499 g).

• Infants who were small for gestational age (< 10th percentile for weight) at hospital discharge (vs infants ≥ 10th percentile).

• Infants with chronic lung disease receiving supplemental oxygen therapy at hospital discharge (vs infants without chronic lung disease).