Criteria for considering studies for this review
Types of studies
All randomised clinical trials, regardless of publication status and year, language, or number of participants. We will consider quasi-randomised studies and other studies that may be retrieved with the searches only for their report on harms.
Types of participants
Participants regardless of age, sex, and ethnic origin, diagnosed as any types of non-alcoholic fatty liver disease including non-alcoholic fatty liver, non-alcoholic steatohepatitis, or cirrhosis fulfilling the following criteria:
Histological evidence of hepatic steatosis, or steatoﬁbrosis, or cirrhosis.
Minimal alcohol intake: a daily alcohol intake less than 20 g in women and 40 g in men (Becker 1996; Neuschwander 2003); or ongoing or recent alcohol consumption less than 21 drinks on average per week in men and less than 14 drinks on average per week in women (Chalasani 2012).
There will be no competing aetiologies for hepatic steatosis (eg, significant alcohol consumption, hepatitis B, hepatitis C, medications, parenteral nutrition, Wilson's disease, and severe malnutrition). There will be no co-existing causes for chronic liver disease (eg, haemochromatosis, autoimmune liver disease, chronic viral hepatitis, and Wilson's disease) (Chalasani 2012).
Types of interventions
All types of omega-3 polyunsaturated fatty acid at any dose, duration, and administration methods versus no intervention, placebo, or other interventions. Co-interventions will also be considered if used equally in all groups.
Types of outcome measures
Liver histopathology: number of people with histological improvement/deterioration in the degree of fatty liver infiltration, inflammation, and fibrosis.
Adverse events: defined as any untoward medical occurrence not necessarily having a causal relationship with the treatment but resulting in the discontinuation of treatment. Severe adverse events will be defined according to the International Committee on Harmonization guidelines (ICH-GCP 1997), as any event that would increase mortality; is life-threatening; requires inpatient hospitalisation; results in a persistent or significant disability; or any important medical event that may jeopardise the participant or require intervention to prevent it.
NAFLD fibrosis score based on six readily available variables (age, body mass index, hyperglycaemia, platelet count, albumin, aspartate aminotransferase/alanine aminotransferase ratio), calculated using the published formula (naﬂdscore.com) (Chalasani 2012).
Laboratory tests: number of people with any reported improvement/deterioration in serum activities of aspartate aminotransferase, alanine aminotransferase, gamma glutamyl transpeptidase, and ferritin, or mean decrease in the level of serum activities of aspartate aminotransferase, alanine aminotransferase, gamma glutamyl transpeptidase, and ferritin.
Search methods for identification of studies
We will search the Cochrane Hepato-Biliary Group Controlled Trials Register (Gluud 2013), the Cochrane Central Register of Controlled Trials (CENTRAL) in The Cochrane Library, MEDLINE (PubMed), EMBASE, Science Citation Index Expanded (Royle 2003), and registered clinical trials on www.clinicaltrails.gov. In addition, we will also search The Chinese Biomedical Database (CBM). The preliminary search strategies are given in Appendix 1 with the planned time spans of the searches. If needed, we will improve the searches at the review stage.
Searching other resources
We will scan the bibliographic references of identified randomised clinical trials, review articles, and meta-analyses in order to find other relevant randomised clinical trials. We will also search conference databases and ongoing trials for relevant trials.
Data collection and analysis
We will carry out the review according to the recommendations of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), and the Cochrane Hepato-Biliary Module (Gluud 2013). We will use the software package Review Manager 5 provided by The Cochrane Collaboration (RevMan 2012).
Selection of studies
We will use the search strategies described in Appendix 1 to obtain titles and abstracts of studies that may be relevant for the review. Two review authors (YY Liu and PZ Su) will examine the titles and abstracts to retrieve potentially relevant reports independently. We will examine the full text to decide on the relevance of the publication. We will resolve disagreements by consensus. We will list the fulfilment of the inclusion criteria of the included trials and reasons for exclusion of the excluded studies during the selection process. We will contact authors of the trials if information about methodology or data are unclear or missing.
Data extraction and management
Two review authors (SH L and K X) will extract data. We will translate all publications published in non-English language journals (except for Chinese) before assessment. For trials reported in more than one publication, we will extract data from each report separately and combine information across multiple data collection forms. We will contact the study authors to acquire accurate data if there are discrepancies between published versions or if any further information is required. We will resolve disagreements by consultation with all review authors. From each trial, we will extract the following information: date, location; study design, length of follow-up, and use of intention-to-treat analysis; total number, number of people randomised, number of participants allocated to each intervention group; diagnostic criteria; mean (or median) age, sex ratio, and country; total number of intervention groups, dose and duration of administration of omega-3 polyunsaturated fatty acid and of additional intervention(s), and also of other traditional agents or placebo in the control groups; number of events in the intervention group and in the control group for each of the outcome measures mentioned above. If information is not available in the published trial, we will contact authors of the publications in order to assess the trials correctly.
Assessment of risk of bias in included studies
Two review authors (SH L and K X) will follow the instructions given in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), and the Cochrane Hepato-Biliary Group Module (Gluud 2013), to assess the risk of bias of the included trials.
Two review authors will independently assess the risk of bias of each included trial according to the recommendations in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), the Cochrane Hepato-Biliary Group Module (Gluud 2013), and methodological studies (Schulz 1995; Moher 1998; Kjaergard 2001; Wood 2008; Lundh 2012; Savović 2012 Savović 2012a). We will use the following definitions in the assessment of risk of bias.
Allocation sequence generation
Low risk of bias: sequence generation was achieved using computer random number generation or a random number table. Drawing lots, tossing a coin, shuffling cards, and throwing dice are adequate if performed by an independent person not otherwise involved in the trial.
Uncertain risk of bias: the method of sequence generation was not specified.
High risk of bias: the sequence generation method was not random.
Low risk of bias: the participant allocations could not have been foreseen in advance of, or during, enrolment. Allocation was controlled by a central and independent randomisation unit. The allocation sequence was unknown to the investigators (eg, if the allocation sequence was hidden in sequentially numbered, opaque, and sealed envelopes).
Uncertain risk of bias: the method used to conceal the allocation was not described so that intervention allocations may have been foreseen in advance of, or during, enrolment.
High risk of bias: the allocation sequence was likely to be known to the investigators who assigned the participants.
Blinding of participants, personnel, and outcome assessors
Low risk of bias: blinding was performed adequately, or the assessment of outcomes was not likely to be influenced by lack of blinding.
Uncertain risk of bias: there was insufficient information to assess whether blinding was likely to induce bias on the results.
High risk of bias: no blinding or incomplete blinding, and the assessment of outcomes were likely to be influenced by lack of blinding.
Incomplete outcome data
Low risk of bias: missing data were unlikely to make treatment effects depart from plausible values. Sufficient methods, such as multiple imputation, were employed to handle missing data.
Uncertain risk of bias: there was insufficient information to assess whether missing data in combination with the method used to handle missing data were likely to induce bias on the results.
High risk of bias: the results were likely to be biased due to missing data.
Selective outcome reporting
Low risk of bias: all outcomes were pre-defined and reported, or all clinically relevant and reasonably expected outcomes were reported.
Uncertain risk of bias: it is unclear whether all pre-defined and clinically relevant and reasonably expected outcomes were reported.
High risk of bias: one or more clinically relevant and reasonably expected outcomes were not reported, and data on these outcomes were likely to have been recorded.
For a trial to be assessed with low risk of bias in the selective outcome reporting domain, the trial should have been registered either on the www.clinicaltrials.gov web site or a similar register, or there should be a protocol (eg, published in a paper journal). In the case when the trial was run and published in the years when trial registration was not required, we will carefully scrutinize all publications reporting on the trial to identify the trial objectives and outcomes. If usable data on all outcomes specified in the trial objectives are provided in the publications results section, then the trial can be considered a low risk of bias trial in the 'Selective outcome reporting' domain.
Low risk of bias: the trial appears to be free of industry sponsorship or other type of for-profit support that may manipulate the trial design, conductance, or results of the trial.
Uncertain risk of bias: the trial may or may not be free of for-profit bias as no information on clinical trial support or sponsorship was provided.
High risk of bias: the trial was sponsored by industry or received other type of for-profit support.
Low risk of bias: the trial appears to be free of other sources of bias.
Uncertain risk of bias: there is insufficient information to assess whether other sources of bias are present.
High risk of bias: it is likely that potential sources of bias related to the specific trial design used, or other bias risks, are present.
We will consider trials with low risk of bias if assessed with 'low risk of bias' in all the above domains; in all remaining cases, the trials will be considered as high risk of bias trials.
Measures of treatment effect
We will analyse dichotomous data by calculating the odds ratios (OR) with 95% confidence interval (CI). For continuous data, we will use mean difference (MD) as the effect measure of choice. We will report the 95% CI for all effect sizes.
Unit of analysis issues
We will include individual randomised clinical trials, clustered randomised clinical trials, and cross-over trials. However, we will analyse data extracted from the three types of clinical trials in three groups (group of individual randomised clinical trials, group of clustered randomised clinical trials, and group of cross-over trials) instead of analysing all data as a whole. Cluster-randomised trials will be meta-analysed using the generic inverse-variance method. We are uncertain about whether omega-3 polyunsaturated fatty acids have a long-lasting effect on the human body after a period of intake, so we will only analyse the first period of the data retrieved from all included cross-over trials.
Dealing with missing data
We will analyse all the trials in two ways: analysing only the available data and using intention-to-treat analyses. When analysing only the available data, we will include all the data extracted from all the trials as well as the data we get from authors of the trials (ignoring the remaining missing data). In the case of intention-to-treat analyses, we will include data on the total number of randomised participants, no matter how the original trialists analysed the data. This will involve imputing outcomes for the missing participants.
Regarding the primary outcomes, we will include participants with incomplete or missing data in sensitivity analyses by imputing them according to the following scenarios (Hollis 1999).
Poor outcome analysis: assuming that drop-outs/participants lost from both the experimental and the control groups experienced the outcome, including all randomised participants in the denominator.
Good outcome analysis: assuming that none of the drop-outs/participants lost from the experimental and the control groups experienced the outcome, including all randomised participants in the denominator.
Extreme case analysis favouring the experimental intervention ('best-worse' case scenario: none of the drop-outs/participants lost from the experimental group, but all of the drop-outs/participants lost from the control group experienced the outcome, including all randomised participants in the denominator.
Extreme case analysis favouring the control ('worst-best' case scenario): all drop-outs/participants lost from the experimental group, but none from the control group experienced the outcome, including all randomised participants in the denominator.
Assessment of heterogeneity
We will analyse heterogeneity using the Chi2 test, with an alpha of 0.05 used for statistical significance and with the I2 test (Higgins 2011). We define I2 values of 25%, 50%, and 75% as corresponding to low, medium, and high levels of heterogeneity.
Assessment of reporting biases
We will use funnel plot asymmetry to assess the existence of publication bias and other biases (Egger 1997). We will need at least 10 trials before we can draw a funnel plot.
We will use both a random-effects model and a fixed-effect model to analyse the data. If we find significant differences in the results between the two models, we will provide both results. If the difference in the results is not significant, then we will present the results from using the random-effects model (Higgins 2011).
Trial sequential analysis
Trial sequential analysis is a tool for quantifying the statistical reliability of the data in a cumulative meta-analysis (CTU 2011; Thorlund 2011), adjusting alpha and beta values for sparse data and repetitive testing on accumulating data (Brok 2008; Wetterslev 2008; Brok 2009; Thorlund 2009; Wetterslev 2009; Thorlund 2010; Thorlund 2011). Trial sequential analysis is a methodology that combines a required information size calculation (cumulated sample sizes of included trials) with the threshold of statistical significance. In order to control for the risks of random errors due to sparse data and multiplicity, we will perform trial sequential analyses for the dichotomous outcomes as well as for the continuous outcomes (Brok 2008; Wetterslev 2008; Brok 2009; Thorlund 2009; Wetterslev 2009; Thorlund 2010; Thorlund 2011). We will base our calculations on the diversity-adjusted required information size on the proportion of participants with the outcome in the conventional group, a relative risk reduction of 20%, an alpha (type I error) of 5%, a beta (type II error) of 20%, and the diversity of the meta-analysis (Wetterslev 2009).
Subgroup analysis and investigation of heterogeneity
Heterogeneity may exist among the trials due to clinical diversity or methodological diversity (Higgins 2011). Variability in risk of bias and in the participants, interventions, and outcomes may be potential sources of heterogeneity in this review.
Subgroup analyses will be performed on subsets of participants as follows:
Trials with low risk of bias compared to trials with high risk of bias.
Adults compared to children (under 16 years old).
Large dose of omega-3 polyunsaturated fatty acid compared to small dose of omega-3 polyunsaturated fatty acid.
Long-time intake of omega-3 polyunsaturated fatty acid compared to short-time intake of omega-3 polyunsaturated fatty acid (we will decide the specific quantity of dose and length of time during the review preparation).
We plan to carry out the following sensitivity analyses:
'Summary of findings' tables
We will construct 'Summary of findings' tables on results of primary outcomes using GRADEpro software (version 3.2) (GRADE 2011). We will resolve disagreements by discussion.