Types of studies

We sought randomised controlled trials.

Types of participants

We planned to include all adults with a positive anterior chamber tap for CMV DNA. We planned to exclude the following groups of participants:

1. Pregnant women

2. Children

3. People who are allergic to the systemic drugs

4. People who have neutropenia (reduction in white blood cells) or are predisposed to it due to a systemic medical condition

Types of interventions

We planned to include studies that compare:

• oral valganciclovir to oral ganciclovir;

• topical ganciclovir to systemic valganciclovir/ganciclovir therapy;

• oral/topical antiviral therapy to placebo treatment;

• intravitreal ganciclovir injections to ganciclovir implant;

• different doses of the same drug;

• different duration of the same drug.

Types of outcome measures

Electronic searches

The Cochrane Eyes and Vision Information Specialist conducted systematic searches in the following databases for randomised controlled trials and controlled clinical trials. There were no language or publication year restrictions. The date of the search was 21 March 2017.

• Cochrane Central Register of Controlled Trials (CENTRAL; 2017, Issue 2) (which contains the Cochrane Eyes and Vision Trials Register) in the Cochrane Library (searched 21 March 2017) (Appendix 1);

• MEDLINE Ovid (1946 to 21 March 2017) (Appendix 2);

• Embase Ovid (1980 to 21 March 2017) (Appendix 3);

• ISRCTN registry (www.isrctn.com/editAdvancedSearch; searched 21 March 2017) (Appendix 4);

• US National Institutes of Health Ongoing Trials Register ClinicalTrials.gov (www.clinicaltrials.gov; searched 21 March 2017) (Appendix 5);

• World Health Organization International Clinical Trials Registry Platform (www.who.int/ictrp; searched 21 March 2017) (Appendix 6).

Searching other resources

We planned to screen the reference lists of potentially relevant studies to identify further trials for this review. We planned also to contact investigators and experts in the field to help identify unpublished data or ongoing studies. We did not handsearch conference proceedings or journals specifically for the purposes of this review.

Selection of studies

Two review authors (AA, JM) independently assessed the titles and abstracts obtained from the electronic and manual searching. Once we had screened the titles and abstracts, we had 'yes', 'no', and 'unclear' lists. The two review authors planned also to independently conduct all relevant full‐copy assessments for all 'yes' and 'unclear' studies, to decide which studies we would include and to document excluded studies in a 'Characteristics of excluded studies' table. We planned to contact authors for further clarification where studies lacked the required information and when we could not agree on them.

Data extraction and management

We planned to extract the following study characteristics: methods, participants, interventions, and outcomes, including type of chart used for measuring vision. To measure our primary outcome, clinical resolution of inflammation would include one or more of the following: absence of cells in the anterior chamber, resolution of corneal oedema or keratic precipitates or both. We planned to analyse logMAR (log of the minimum angle of resolution) score, converting from letters as appropriate, following guidance given in Chaper 18 of the Cochrane Handbook for Systematic Reviews of Interventions (Stewart 2011). Following identification of relevant studies, two review authors (AA, JM) planned to independently extract the relevant data using a simple form that would be piloted.

Assessment of risk of bias in included studies

We planned to use the Cochrane tool for assessing risk of bias as described in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a). We will assess the following six parameters for each included study:

• sequence generation

• allocation concealment

• blinding (masking)

• incomplete outcome data

• selective reporting

• other potential sources of bias

Measures of treatment effect

We planned to use risk ratios with 95% confidence intervals as the measure of effect for dichotomous variables (resolution of inflammation, glaucoma, corneal decompensation, and adverse events). We planned to use the mean difference (with a 95% confidence interval) as the measure of effect for continuous variables (visual acuity, quality of life).

We planned to analyse time to recurrence as a hazard ratio, as outlined in Chapter 7 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b). We also planned to investigate adjustments for repeated measures (summary effect over all time points).

We planned to measure visual acuity in logMAR scale. Where visual acuity is measured on different scales, we planned to convert it into logMAR before analysis.

Unit of analysis issues

Trials may randomise one or both eyes to the intervention or comparator. If people are randomly allocated to treatment, but only one eye per person is included in the trial, then there will not be a unit‐of‐analysis issue. In these cases, we planned to document how the eye was selected. If people are randomly allocated to treatment, but both eyes are included and reported, we planned to analyse as "clustered data", that is to adjust for within‐person correlation as summarised in Chapter 9 of the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2011). We planned to contact the trial investigators for further information in order to do this.

We excluded studies that had allocated different eyes to different treatments, as there may be a confounding cross‐over effect due to systemic absorption.

Dealing with missing data

We planned to follow the general recommendations for dealing with missing data as outlined in Chapter 16 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011c).

• We planned to report follow‐up for each treatment group for each individual study, and collect data on reasons for missingness, where available.

• Whenever possible, we planned to contact the original investigators to request missing data, in particular where review outcomes have not been reported.

• We planned to make explicit the assumptions of any methods used to address missing data, for example that the data are assumed missing at random, or that missing values were assumed to have a particular value such as a poor outcome.

• We planned to address the potential impact of missing data on the findings of the review in the 'Discussion' section.

We planned not to use missing summary data (for example no usable data) as a reason to exclude a study from the review. We planned to include the study in the review, and discuss the potential implications of its absence from a meta‐analysis.

Assessment of heterogeneity

We planned to check for heterogeneity by examining the:

• characteristics of the study;

• forest plot of results of the study;

• results of the Chi² test for statistical heterogeneity;

• results of the I² statistic to quantify inconsistencies between study results.

We planned to use the following guide to interpret the I² statistic, following guidance given in Chapter 9 of the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2011):

• 0% to 40%: may not be important;

• 30% to 60%: may represent moderate heterogeneity;

• 50% to 90%: may represent substantial heterogeneity;

• 75% to 100%: considerable heterogeneity.

Assessment of reporting biases

If there are 10 or more studies contributing to a meta‐analysis, we planned to prepare a funnel plot. An asymmetric funnel plot is possible evidence of publication bias, although there may be other explanations for the asymmetry (Sterne 2011).

We planned to check for selective outcome reporting bias by checking that the individual trials have reported all the outcomes that they collected data for; where possible, we planned to look at the trial protocol to check this.

Data synthesis

We planned to pool data using a random‐effects model, unless there are three or fewer trials, in which case we planned to use a fixed‐effect model.

There is likely to be some clinical or methodical diversity in a meta‐analysis, therefore heterogeneity is inevitable. If we identify heterogeneity, we planned to check that the data are correct. If there is evidence of substantial heterogeneity, for example, I² greater than 50%, Chi² less than 0.1 and different direction of effects, we planned not to pool data, as the pooled effect may not be a good summary of the individual trial results.

Subgroup analysis and investigation of heterogeneity

We did not plan any subgroup analyses.

Sensitivity analysis

We planned to compare all analyses with different models, that is random‐effects compared to fixed‐effect.

For the analysis of the primary outcome, we planned to examine the effect of excluding the following types of trials:

1. Trials graded as high risk of bias on any parameter

2. Unpublished trials

3. Industry‐funded studies