Types of studies

Randomised trials (published and unpublished). Both random allocation and quasi‐random allocation (e.g. where there is alternate allocation to groups) were included. Parallel and cross‐over trials were eligible.

Types of participants

People of all ages and of both sexes with CF diagnosed by genetic testing or evidence on sweat chloride or nasal potential difference, including all degrees of disease severity.

Types of interventions

Nebulised hypertonic saline, where timing of inhalation was the randomised element in the study protocol:

1. hypertonic saline inhalation up to six hours before airway clearance techniques, compared to inhalation during airway clearance techniques;

2. hypertonic saline inhalation up to six hours before airway clearance techniques, compared to up to six hours after airway clearance techniques;

3. hypertonic saline inhalation during airway clearance techniques, compared to up to six hours after airway clearance techniques;

4. morning compared to evening inhalation with any definition provided by the author. If not defined, we accepted midnight to midday as morning and midday to midnight as evening.

Note: many individuals perform two treatments with physical airway clearance techniques each day. Even if these treatments are performed as far as possible from each other (i.e. 12 hours apart), any threshold greater than six hours would mean that the hypertonic saline labelled ‘before’ would actually be closer to ‘after’ the previous treatment with physical airway clearance techniques. Therefore, six hours is the broadest threshold possible to capture all potentially relevant trials. However, some of the mechanisms by which timing may affect the outcome are short‐lived. Therefore, a sensitivity analysis is performed that only considers trials where the hypertonic saline is within 30 min of the techniques (seeSensitivity analysis).

The timing regimen could be a single treatment or could be maintained for any duration.

Hypertonic saline treatment had to be a minimum of a single dose of at least 3% concentration. If studies mentioned co‐interventions (such as bronchodilators and other inhaled medications), these were permitted provided they were the same on all trial days and taken either before or after the period during which hypertonic saline and airway clearance techniques were used.

Airway clearance techniques (ACT) had to be a minimum of 10 minutes in duration and include at least one of the following:

1. postural drainage with percussion and vibration (PDPV). In other reviews this has been described as conventional chest physiotherapy (CCPT) (Elkins 2006c; Main 2009; van der Schans 2009);

2. active cycle of breathing techniques (ACBT).This comprises relaxation or breathing control, forced expiration technique (FET), thoracic expansion exercises and may include postural drainage or percussion (Robinson 2010);

3. autogenic drainage (AD). This breathing technique uses high expiratory flow rates at varying lung volumes to enhance mucus clearance while avoiding airway closure;

4. positive expiratory pressure (PEP) (Elkins 2006a);

5. oral oscillatory devices that provide oscillating PEP (such as the Flutter, Cornet and Acapella) or intrapulmonary percussive ventilation, which provides continuous oscillation of the air pressure in the airways via the mouth (Morrison 2009);

6. thoracic oscillating devices applied via a vest to provide external chest wall oscillation (Morrison 2009);

7. exercise prescribed for the purpose of airway clearance either independently or as an adjunct to other techniques.

Types of outcome measures

Electronic searches

We identified relevant trials from the Cochrane Cystic Fibrosis and Genetic Disorders Group's Cystic Fibrosis Trials Register using the term 'hypertonic saline'.

The Cystic Fibrosis Trials Register is compiled from electronic searches of the Cochrane Central Register of Controlled Trials (CENTRAL) (updated each new issue of the Cochrane Library), weekly searches of MEDLINE, a search of Embase to 1995 and the prospective handsearching of two journals ‐ Pediatric Pulmonology and the Journal of Cystic Fibrosis. Unpublished work was identified by searching the abstract books of relevant conferences, including the three major cystic fibrosis conferences: the International Cystic Fibrosis Conference, the European Cystic Fibrosis Conference and the North American Cystic Fibrosis Conference. For full details of all searching activities for the register, please see the relevant sections of the Cochrane Cystic Fibrosis and Genetic Disorders Group's website.

Date of last search of the Cochrane Cystic Fibrosis and Genetic Disorders Group's Cystic Fibrosis Trials Register: 28 February 2019.

In addition, we searched the Physiotherapy Evidence Database (PEDro) (Michaleff 2011) and the WHO International Clinical Trials Registry Platform for all years available (Appendix 1; Appendix 2). Date of last search: 21 February 2019.

Searching other resources

We hand‐searched the biennial Australia and New Zealand CF Conference Proceedings from 1995 to 2017.

We contacted trial authors, experts and manufacturers of hypertonic saline to identify additional trials.

We searched the reference lists of the included studies.

Selection of studies

Both authors (RD, ME) independently selected the trials to be included in the review. We performed initial screening by title and abstract, and reviewed the full text for studies remaining after the initial screening phase. We resolved disagreements by discussion. We excluded irrelevant records and have presented the details of these excluded trials and the reasons for exclusion in tabular form (Characteristics of excluded studies).

Data extraction and management

Each author independently extracted data from the included studies using a standardised data extraction form. We resolved any disagreements by discussion. Where data were absent or difficult to interpret in the presented form, the authors contacted the trial investigators to gain the information required to evaluate risk of bias in the trial and to facilitate data analysis. Data extraction was also carried out at the editorial base for the Dentice trial given we (the Cochrane Review authors) are also two of the trial authors (Dentice 2012).

Assessment of risk of bias in included studies

We independently determined the risk of bias for each trial using published and unpublished data from the trial investigators, following the domain‐based evaluation as described in the Cochrane Handbook for Systematic Reviews of Interventions 5.1 (Higgins 2011). The risk of bias was also carried out at the editorial base for the Dentice trial given we (the Cochrane Review authors) are also two of the trial authors (Dentice 2012). We assessed the following domains as either low risk, unclear risk, or high risk of bias:

1. sequence generation;

2. allocation concealment;

3. blinding (of participants, personnel and outcome assessors);

4. incomplete outcome data;

5. selective outcome reporting;

6. other sources of bias.

We resolved any disagreements by discussion. In addition, each author independently rated each trial on the PEDro Scale (Maher 2003; de Morton 2009), using published and unpublished data from the trial investigators. The PEDro Scale was chosen because it has acceptably high reliability for individual ratings and for consensus ratings (Maher 2003; Shiwa 2011). It has undergone extensive validation including convergent and construct validity and the appropriateness of the scoring system has been justified with Rasch analysis (de Morton 2009). The PEDro Scale is also widely used and understood by physiotherapists (Elkins 2013). The Cochrane Risk of Bias tool has reliability that ranges from poor to slight (Armijo‐Olivo 2012; Hartling 2009; Hartling 2013).

Measures of treatment effect

For dichotomous data we planned to use the risk ratio (RR) with 95% confidence intervals (95% CIs) as a measure of treatment effect with an intention‐to‐treat analysis. An intention‐to‐treat analysis means that, where participants did not receive treatment as allocated, and were measures of outcomes were available, the analysis was performed as if participants received the treatment to which they were allocated. We considered a trial to have analysed by intention to treat if the report explicitly states that all participants received treatment or control conditions as allocated. Mortality and adverse events were the only dichotomous outcomes.

For continuous data we planned to record the difference in mean change from baseline and standard deviation (SD) (or standard errors (SE)) for each group, or the final group means and SDs if change data are unavailable. We planned to calculate a pooled estimate of treatment effect using the mean difference (MD) with 95% CIs.

However, given that both of the included trials were cross‐over in design, we used the generic inverse variance method. Where the MD and SE of the MD were available, we entered these directly into the meta‐analysis. Where only group means and SDs were available, we calculated the MD by subtraction of one group mean from the other. We imputed the SD of the differences from that obtained from the data in a similar trial, taking into account the paired nature of the data. We then calculated SEs using the formula: imputed SD / √n.

Unit of analysis issues

We incorporated data from cross‐over trials into meta‐analysis using the generic inverse variance method, involving expression of data in terms of the paired mean differences between treatments and their SE. We calculated these values either from paired individual patient data provided by authors, or by calculation of mean differences between interventions and their SE from means, SDs and P values reported in the manuscript (Elbourne 2002). We intended to combine data from parallel‐designed trials with those from cross‐over trials in the meta‐analyses, but only cross‐over trials were obtained. If necessary, we intended to calculate the SEs in parallel trials from the MDs between treatments and their CIs, but only cross‐over trials were obtained.

Dealing with missing data

Where data were absent or difficult to interpret in the presented form, we contacted the trial investigators in an attempt to obtain the data in a form that would facilitate data analysis. Had we not been able to obtain the data, our plan was to analyse only the available data (available‐case analysis). Additional data were obtained from the authors of the van Ginderdeuren trial (Van Ginderdeuren 2011). Additional data were sought from the authors of the O'Neill trial (O'Neill 2016) and data were received for all requested outcomes.

Assessment of heterogeneity

We visually inspected forest plots for overlap of the CIs and estimated statistical heterogeneity using the I² value (Higgins 2003). Given that thresholds for the interpretation of I² can be misleading (since the importance of inconsistency depends on several factors) we chose a rough guide to interpretation as follows: moderate heterogeneity was defined as an I² value of 30% to 60%, substantial heterogeneity was defined as an I² value of 50% to 90%, and considerable heterogeneity was defined as an I² value of 75% to 100%. The P value from the chi‐squared test was also used to interpret the importance of the observed value of I². Clinical heterogeneity was also assessed by considering differences in trial designs and participants characteristics.

Assessment of reporting biases

If we had included sufficient trials in the review, we planned to assess publication bias using a funnel plot. If we suspected outcome reporting bias, we planned to contact trial investigators to clarify whether they measured and analysed certain outcomes and obtained the data. However, only two eligible trials were obtained; this was fewer than the recommended number of studies to construct a funnel plot so we were unable to assess whether outcome reporting bias was likely.

Data synthesis

We analysed the following between‐group comparisons where possible:

1. hypertonic saline inhalation up to six hours before airway clearance techniques, compared to inhalation during airway clearance techniques;

2. hypertonic saline inhalation up to six hours before airway clearance techniques, compared to up to six hours after airway clearance techniques;

3. hypertonic saline inhalation during airway clearance techniques, compared to up to six hours after airway clearance techniques;

4. morning compared to evening inhalation with any definition provided by the author. If not defined, we accepted midnight to midday as morning and midday to midnight as evening.

We performed meta‐analyses using a fixed‐effect model. However, if we had observed substantial heterogeneity, we planned to use a random‐effects model. We planned to further explore heterogeneity in the planned subgroup analyses.

Subgroup analysis and investigation of heterogeneity

We planned a subgroup analysis for different concentrations of hypertonic saline inhalation: low hypertonic concentrations (3% to 5%); medium (6% to 7%); and high (8% to 10%). We also planned to perform a subgroup analysis for intervention duration: less than two weeks; two to eight weeks; and greater than eight weeks. However, the included studies did not permit these analyses.

Sensitivity analysis

We planned to conduct sensitivity analyses to assess the robustness of the review results by repeating the analyses with the following adjustments:

• exclusion of trials with unclear or inadequate sequence generation;

• exclusion of trials with unclear or inadequate allocation concealment;

• exclusion of trials that scored less than 5 out of 10 on the PEDro Scale (Maher 2003);

• modifying the definition of 'before' and 'after' airway clearance techniques to within 30 minutes of the techniques;

• with and without cross‐over trials.

However, the included studies did not permit these analyses.