Types of studies

We will include randomised controlled trials (RCTs) only in the review.

Types of participants

We will include studies where participants are aged 50 years or older and have non‐exudative AMD early, intermediate or late as defined by the Age Related Eye Disease Study (AREDS 2001). Late stages can feature geographic atrophy; a well‐demarcated area of retinal pigmented epithelium atrophy with or without central foveal involvement.

Types of interventions

We will include trials where PBM is compared to standard care with or without sham treatment. We will include studies using visible to NIR light (500 nm to 1000 nm) delivered by low‐level lasers or light‐emitting diodes. Trials that have adequately described the PBM therapy parameters in terms of wavelength(s), dose, frequency, duration and coverage will be included.

Standard care consists of modifying risk factors; smoking cessation, nutritional advice and supportive measures; referral to low‐vision services and visual rehabilitation.

Types of outcome measures

Electronic searches

The Cochrane Eyes and Vision Information Specialist will search the following electronic databases for randomised controlled trials and controlled clinical trials. There will be no language or publication year restrictions.

• Cochrane Central Register of Controlled Trials (CENTRAL) (which contains the Cochrane Eyes and Vision Trials Register) in the Cochrane Library (latest issue) (Appendix 1);

• MEDLINE Ovid (1946 to present) (Appendix 2);

• Embase Ovid (1980 to present) (Appendix 3);

• ISRCTN registry (www.isrctn.com/editAdvancedSearch (Appendix 4);

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

• World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) (www.who.int/ictrp) (Appendix 6).

Searching other resources

We will review the reference lists of included trial reports and related systematic reviews to identify additional potentially relevant trials. We will contact medical device companies conducting studies on PBM for information about any ongoing or completed but not published clinical trials. We will search abstracts from the annual meetings of the European VitreoRetinal Society, the Macula Society, the Retina Society, subspecialty meetings from the American Academy of Ophthalmology, and the American Society of Retinal Surgeons for ongoing trials.

Selection of studies

Two review authors will screen independently the titles and abstracts resulting from the searches using web‐based software (Covidence 2015). We will obtain full‐text copies of potentially relevant trials and contact trial investigators for further information if required. Discrepancies between authors as to whether or not studies meet inclusion criteria will be resolved by discussion. We will document the excluded studies and reasons for exclusion.

For potentially eligible studies identified on trials registers we will do the following.

• If the study has a completion date more than two years previously, we will look for publications of this trial and contact the investigators if necessary to obtain published or unpublished data from the trial. If eligible, we will include the study in the review irrespective of whether we can identify a publication.

• If the study has a completion date within two years, or in the future, we will document the study in the ongoing studies section.

Data extraction and management

Two review authors will extract data independently using an online form developed by Cochrane Eyes and Vision and using Covidence. We will pre‐pilot the data extraction template. We will resolve discrepancies by discussion. Two attempts will be made to contact trial investigators for missing data. Data will be directly imported into Review Manager 5 (RevMan 5) (Review Manager 2014); and the accuracy of the data import will be checked by one author.

Assessment of risk of bias in included studies

Two review authors will assess independently the risk of bias using Cochrane's 'Risk of bias' tool for assessing risk of bias in each included study (Higgins 2011). We will resolve disagreements by discussion. We will specifically consider and report on the following sources of bias.

• Selection bias (random sequence generation, allocation concealment): was the sequence of allocation generated using a random procedure and was the allocation concealed to people recruiting/enrolling participants and to participants?

• Performance bias (masking of participants and researchers): were the recipients of care unaware of their assigned intervention? Were persons providing care unaware of the assigned intervention?

• Detection bias (masking of outcome assessors). Were persons evaluating outcomes unaware of the assigned intervention?

• Attrition bias: were the rates of follow‐up and compliance similar in the groups? Was the analysis by intention to treat and were there any post‐randomisation exclusions?

• Selective outcome reporting bias: is there any evidence that the outcomes that were measured have not been reported?

We will grade each domain as low risk of bias, high risk of bias or unclear (lack of information or uncertainty of potential for bias). We will contact trial investigators for clarification of parameters graded as 'unclear'.

Measures of treatment effect

We will calculate the mean difference for the following continuous outcomes: visual acuity, contrast sensitivity, reading speed and LLD score. Where possible, we will check for the skewness of continuous data (Altman 1996).

We will calculate the risk ratio for adverse events.

Unit of analysis issues

PBM can be applied unilaterally or bilaterally. Usually it is applied to the affected eye(s). A pilot interventional study showed bilateral drusen size reduction when an ultra low energy nanosecond laser was applied unilaterally in AMD‐affected eyes (Guymer 2014; Jobling 2015). Application to only one eye may have an effect in the fellow eye, although the evidence is unclear and beyond the scope of this review.

Dealing with missing data

If possible, we will conduct an intention‐to‐treat (ITT) analysis. We will use imputed data if computed by the trial investigators using an appropriate method, but will not impute missing data ourselves. If ITT data are not available, we will do an available case analysis. This assumes that data are missing at random. We will assess whether this assumption is reasonable by collecting data from each included trial on the number of participants excluded or lost to follow‐up and reasons for loss to follow‐up by treatment group, if reported.

Assessment of heterogeneity

We will examine the overall characteristics of the studies, in particular the type of participants and types of interventions, to assess the extent to which the studies are similar enough to make pooling study results sensible. We will look at the forest plots of study results to see how consistent the results of the studies are, in particular looking at the size and direction of effects. We will calculate I² which is the percentage of the variability in effect estimates that is due to heterogeneity rather than sampling error (chance) (Higgins 2002). We will consider I² values over 50% to indicate substantial inconsistency but will also consider Chi² P value. As this may have low power when the number of studies are few we will consider P < 0.1 to indicate statistical significance of the Chi² test.

Assessment of reporting biases

We will use the 'Risk of bias' assessment tool to look for selective or incomplete reporting. See Assessment of risk of bias in included studies.

If there are 10 trials or more included in a meta‐analysis, we will construct funnel plots and consider tests for asymmetry for assessment of publication bias, according to Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

Data synthesis

We will pool data using a random‐effects model in RevMan 5. If there are fewer than three trials in a comparison we will use a fixed‐effect model.

If there is inconsistency between individual study results such that a pooled result may not be a good summary of the individual trial results — for example, the effects are in different directions or I² > 50% and P < 0.1 — we will not pool the data but will describe the pattern of the individual study results.

If there is statistical heterogeneity we may pool the data if all the effect estimates are in the same direction, such that a pooled estimate would seem to provide a good summary of the individual trial results.

Subgroup analysis and investigation of heterogeneity

If there are sufficient trials we will compare the effect of treatment in the following subgroups.

• early, intermediate and late dry‐AMD.

Sensitivity analysis

We will examine the impact of excluding studies at high risk of bias in one or more domains.