Criteria for considering studies for this review
Types of studies
Randomised controlled trials or quasi-randomised controlled trials where there was sufficient evidence that intervention and control groups were similar at baseline.
Types of participants
People with CF, diagnosed according to Rosenstein (Rosenstein 1998), including all ages with any degree of disease severity. Studies with participants enrolled during a period of stability or during a respiratory exacerbation and those studies where aerosolised medications were used as a single-dose, acutely or as long-term maintenance therapy were all considered.
Types of interventions
Nebuliser systems were compared for delivery of each of the following:
hypertonic sodium chloride;
other aerosolised medications.
For each medication conventional systems were compared with any other identified aerosol delivery system listed below, or with another type of conventional system (post hoc change). For this purpose we considered conventional systems to be a compressor combined with a jet nebuliser, including open vent jet systems and breath-assisted open vent systems.
Systems for comparison were:
different types of conventional system (for example open-vent jet system versus breath-assisted open-vent jet system) (post hoc change);
adaptive aerosol delivery (AAD) nebuliser systems (post hoc change);
AAD incorporating vibrating mesh technology (VMT) nebuliser systems;
vibrating mesh technology systems;
ultrasonic nebuliser systems.
Conventional nebuliser systems consist of a compressor coupled with a nebuliser chamber. The compressor entrains room air, compresses it to a higher pressure and emits the air at a given flow rate. The air enters the nebuliser chamber and passes through a small hole, a venturi, beyond which the air expands rapidly creating a negative pressure; this draws the medication up a feeding tube where it is atomised into particles. The particle sizes are variable, larger particles will impact on the baffle within the nebuliser chamber and onto the walls of the chamber and be returned back to the well of the chamber to be re-nebulised. The smaller particles will be continuously released from the nebuliser chamber during both inspiration and expiration of the person using the nebuliser system.
There are three main types of conventional nebuliser system: the jet nebuliser; the open-vent jet nebuliser; and the breath-assisted open-vent jet nebuliser. The jet nebuliser works continuously as described above. Open-vent jet nebulisers incorporate an open vent to allow extra air to be sucked into the chamber during inspiration. This results in greater air flow through the chamber and so greater densities of smaller respirable particles over a shorter period of time. Breath-assisted open-vent jet systems use a valve system to allow air to be drawn in during inspiration as per the open-vent design. During expiration the valve closes and the flow of air through the chamber is decreased to that coming from the compressor only. This decreases the amount of particles released during expiration and therefore decreases medication wastage (O'Callaghan 1997). One last adaptation of compressor and nebuliser systems is holding chambers. This is a chamber which is attached to the nebuliser and aerosol generated continuously by the nebuliser is held within the chamber. A negative pressure is created within the chamber during inspiration causing a valve to open and air to be entrained. This air picks up aerosol and delivers it to the person breathing in. An expiratory valve diverts expired air away from the chamber and the chamber continues to fill with aerosol. Holding chambers are designed to reduce medication wastage (O'Callaghan 1997).
A large number of conventional compressor and nebuliser combinations are available and these combinations have differing characteristics in terms of aerosol particle size, nebulisation time and mass of medication delivered (Higgenbottam 1997). Conventional nebulisation systems tend to be cheaper than the alternatives and are less prone to reliability or delivery problems (or both) due to poor cleaning and maintenance. They are, however, noisy and bulky and therefore less portable; they also produce variable particle sizes and have a larger residual volume as compared to alternative systems, so leading to more wastage of medication.
Two nebuliser systems, the Halolite® and Prodose®, were the first and second generation of nebuliser systems to utilise AAD. These systems are no longer available as they have been superseded by an AAD nebuliser system incorporating vibrating mesh technology (VMT); the I-Neb AAD system®. With AAD, pressure changes relating to airflow are continuously analysed and timed pulses of aerosol (during the first 50% to 80% of inspiration only) are delivered based on the prior three breaths until the preset dose; an actuation, is delivered. This eliminates wastage of medication during exhalation which occurs with continuously delivering nebulisers and optimises deposition. These systems were designed to give optimal efficiency and therefore require an alteration in the priming dose of medication used as compared to conventional nebuliser systems.
AAD incorporating VMT
One nebuliser system, the I-Neb AAD system®, utilises VMT and AAD in combination in order to optimise deposition and treatment times. As detailed above, AAD occurs along with the use of VMT, as detailed below. Inhalation technique is assessed; the nebuliser system will not operate unless correctly set up and used at the appropriate angle. The system also stores adherence and delivery data such as treatment date, time, duration and completeness of dose which can be downloaded by the clinician or the person using the I-Neb using software supplied by Philips (Insight®). These nebuliser systems were designed to give optimal efficiency and therefore require an alteration in the priming dose of medication used as compared to conventional nebuliser systems.
This technology aerosolises medication utilising a vibrating, perforated mesh to generate particles. This is achieved by using a piezoelectric element which either vibrates a transducer horn or which is annular and encircles the mesh causing it to vibrate. Both methods result in medication being pumped through the perforated mesh creating homogenous particles. Some meshes are created with an electroplating technique which forms tapered holes and others by precision laser-drilling (Kesser 2009). Vibrating mesh systems are silent, portable (being small and battery powered), fast and produce more homogenously-sized particles as compared to conventional systems. There are a number of systems available. The Omron MicroAir®, the Aerogen Aeroneb Go®, and the Pari eFlow Rapid® were designed to be similar in efficiency to conventional breath-enhanced nebulisers by using larger particle sizes, a system housing which causes a high residual dose within the nebuliser system, or a medication reservoir with a larger residual volume. Other nebuliser systems were designed to give optimal efficiency and may therefore require an alteration in the priming dose of medication used. The Aerogen OnQ®, Aerodose®, Aeroneb Pro® and Solo®, Pari eFlow® and Philips I-Neb® aim to deliver medication more efficiently and quicker. Some VMT systems are currently available for clinical use while others have only been utilised in research. A number of VMT systems use the piezoelectric crystal technology associated with ultrasonic nebulisers (see below) to create the vibration necessary to pump medication through a mesh.
Ultrasonic nebulisers utilise a piezoelectric crystal which vibrates creating standing waves within the surface of the medication, droplets move away from the crests of these waves becoming an aerosol. Large particles impact on a baffle to be re-nebulised in the same way as jet nebulisers. Ultrasonic nebulisers may be smaller and are quieter and quicker than conventional systems. There is controversy, however, as to whether they are suitable to nebulise certain medications.
Types of outcome measures
Treatment time (for single nebulised treatment)
Quality of life (as measured by e.g. the Cystic Fibrosis Questionnaire (CF-Q) (Henry 2003) or Cystic Fibrosis Quality of Life (CFQoL) (Gee 2000) both validated measures of quality of life in people with CF)
Deposition (as measured by radio labelling or by serum, sputum or urine levels of the studied medication) (post hoc change to consider sputum and urine levels)
Adherence (percentage of prescribed treatment taken)
Burden of care (as measured by a validated tool)
Respiratory function tests
force expiratory volume at one second (FEV1)
forced vital capacity (FVC)
forced mid-expiratory flow (FEF25-75)
time to next exacerbation (measured in days and as defined by Rosenfeld 2001)
total exacerbations within study period
Need for additional antibiotic treatment during study period
Cost (including nebuliser system, consumables and medication costs where a particular brand is required in order to use the nebuliser system)
Patient satisfaction and preference with nebuliser system (e.g. weight, dimensions, time taken to clean equipment, noise levels, power supply, average number of doses per fully charged batteries, cleaning regimen, nebuliser system and consumables availability, customer support, etc.)
Nebuliser system reliability
Search methods for identification of studies
The authors identified relevant studies from the Cochrane CF and Genetic Disorders Group's CF Trials Register using the terms: aerosol delivery AND nebuliser.
The CF Trials Register is compiled from electronic searches of the Cochrane Central Register of Controlled Trials (CENTRAL) (updated each new issue of The Cochrane Library), quarterly 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 is identified by searching the abstract books of 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 Cystic Fibrosis and Genetic Disorders Group Module.
Date of the most recent search of the CF Trials Register: 15 Oct 2012.
Searching other resources
The authors searched the reference lists of each included study for additional publications. They contacted the authors of identified studies for further study details or for information on other published or unpublished studies. They approached the manufacturers of each type of nebuliser system assessed and also manufacturers of the medications identified to provide details of any studies or any data relevant to this review. They emphasised that they wished to access both published and unpublished work, with results which were either positive or negative for the manufacturer.
Data collection and analysis
Selection of studies
Two authors (TD and PW) independently reviewed all citations and abstracts identified by the search to determine which of the papers assessed they would include. Where the two authors disagreed, they resolved this by consensus; a third author (NM) was available to review the paper if consensus was not possible. They recorded any areas of disagreement. The authors excluded non-RCTs, although they did include randomised cross-over studies. The authors also excluded studies comparing nebuliser versus inhaler systems.
Data extraction and management
Two authors (TD and PW) independently performed data extraction and recorded data using Review Manager (RevMan 2011). The authors recorded any areas of disagreement which occurred. They noted details of randomisation, allocation concealment, degree of blinding, inclusion and exclusion criteria, participant type (for example adult or paediatric), and dropouts and withdrawals and how these were accounted for. For short-term studies (up to and including four weeks), the authors reported outcomes for single-intervention studies separately; then reported outcomes at one week, between one and two weeks, more than two weeks to three weeks, more than three weeks to four weeks. For long-term studies (over four weeks), they reported outcomes at three months, six months, twelve months and annually thereafter. They presented other time points reported in the included studies as appropriate.
Some included studies compared more than two nebulisers of the same type (e.g. conventional) for administering the same drug. In these cases, for parallel studies, the authors presented the data for each combination of comparisons on the same graph. For cross-over studies, this is not possible and we have presented the data in an additional table.
For studies that did not present data in an appropriate form to enter into the meta-analysis, where possible the authors calculated the standard deviations (SDs) from the standard errors (SEs) or the confidence intervals (CIs) presented in the published papers.
Assessment of risk of bias in included studies
The authors assessed the risk of bias using the tool described within the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).
Generation of allocation sequence
The authors considered this as having a low risk of bias if a computer algorithm or a similar process based on chance was used to randomise participants to treatment groups. They identified this as having a high risk of bias for sequences which could be attributed to prognosis, degree of disease severity, age etc. They considered studies as having an unclear risk of bias where the generation of allocation sequence was not identified.
Concealment of allocation
The authors considered concealment of allocation to have a low risk of bias where it was not possible for the investigators to foresee the allocation of participants to a particular treatment group, for example centralised or pharmacy-controlled randomisation, pre-numbered or coded identical containers administered serially to participants, on-site locked computer system, or sequentially numbered, sealed, opaque envelopes. They considered the concealment of allocation to have a high risk of bias if the investigator was able to predict the allocation, for example, alternation; the use of case record numbers, dates of birth or day of the week. They graded the risk of bias as unclear if the concealment of allocation was not described.
The authors reported on the degree of blinding employed in each study. Given the specific systems for nebulisation considered within this review, the blinding of the investigator and participants was generally not possible; however, blinding of the person analysing the data was possible. The risk of bias generally increases when few people are blinded to an intervention, thus the risk of bias was higher if the data analyst was not blinded.
Incomplete outcome data
The authors judged a study to be at low risk of bias from incomplete outcome data if there were either no missing outcome data (all participants included in the analysis are exactly those who were randomized into the study) or any missing data or withdrawals were unlikely to be directly related to the intervention, for example if the participant moved away. They judged a study to have an unclear risk of bias if the number of participants randomized into each intervention group was not clearly reported, or numbers completing study not clearly reported. They judged the study to have a high risk of bias if the proportion of incomplete outcome data across groups was not balanced across intervention groups or if reasons for withdrawal or dropout were not given.
The authors judged there to be a low risk of bias from the selective reporting of data if all pre-specified outcomes were reported adequately. They judged there to be an unclear risk of bias if insufficient information was available to make a judgement of either low or high risk. They judged there to be a high risk of bias if not all pre-specified outcomes were reported at all, or non-significant results were not reported; also, if there was non-reporting of outcomes that would very likely have been recorded (e.g. reporting of FVC but not FEV1).
Other sources of bias
Fifteen studies identified as suitable for inclusion were randomised cross-over studies. The authors assessed the risk of bias for cross-over studies as suggested in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) and considered:
whether the cross-over design was suitable;
whether there was a carry-over effect;
whether only first-period data were available;
incorrect analysis; and
comparability of results with those from parallel-group studies.
Measures of treatment effect
For binary outcome measures, the authors assessed data on the number of participants with each outcome event, by allocated treated group, irrespective of compliance and whether or not the individual was later thought to be ineligible or otherwise excluded from treatment or follow-up. They calculated a pooled estimate of the treatment effect for each outcome across studies using the risk ratio (RR) and 95% CIs where appropriate.
For continuous outcomes, they recorded either mean relative change from baseline for each group or mean post-treatment or intervention values and their SDs (presented separately). Where SEs were reported, the authors calculated the SDs where possible. They calculated a pooled estimate of treatment effect by calculating the mean difference (MD) and 95% CIs. If outcomes were reported using different units of measurement, the authors calculated the standardised mean difference (SMD) and 95% CIs.
They considered studies identifying interventions of varying duration separately; those of up to four weeks being short term and those over four weeks long term. They included single-dose interventions as most deposition data were of this data type.
For time-to-event outcomes included in the review, the authors planned to obtain a mixture of logrank and Cox model estimates from the studies. They planned to combine results using the generic inverse variance method as they would have converted the logrank estimates into log hazard ratios and SEs as detailed in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). This was not required during analysis of the data obtained from the included papers.
The authors reported results for both individual studies and meta-analyses with a point estimate together with an associated CI. The point estimate gave magnitude and direction of effect as compared with control and the CI indicated the certainty or uncertainty of this estimate.
Unit of analysis issues
Cross-over studies are those in which each individual receives both treatments in random order. The advantage of this design is that the effect of treatments can be compared within each participant. This is appropriate when within-participant variation is small compared to that between participants, and can allow studies with a smaller number of participants to be conducted. The design is not suitable when the condition of participants is not stable over time, or when the effect of one treatment can 'carry over' from one period to the next. This carry over may be minimised with an adequate wash-out period. The issue of participant stability is obviously pertinent to the CF patient group. A meta-analysis can be conducted if, in the original data, the relevant estimates of treatment effect with SEs are provided. For future updates of the review, where estimates are not available and the authors cannot retrieve the information from the authors, they will consult a statistician for the best way of reporting these data.
The authors identified 11 of the 15 cross-over studies were suitable for inclusion in the analysis, considered the methods recommended by Elbourne to combine results from these cross-over studies (Elbourne 2002) and treated the data as we would for parallel studies.
Dealing with missing data
The authors have reported on whether the original investigators employed an intention-to-treat analysis (analysis based on the initial treatment allocation, not on the treatment eventually administered). They have assessed whether the numbers and reasons for dropouts and withdrawals in all intervention groups are described or whether it is specified that there were no dropouts or withdrawals. They contacted the primary investigators of any identified study where data they required for analysis was not published in the published paper.
Assessment of heterogeneity
The greater the consistency between the primary studies in a meta-analysis, the more generalisable are the results. Heterogeneity refers to the genuine differences between studies rather than those that occur by chance. When the authors are able to include and combine more studies in future updates, they will test for heterogeneity using the I2 statistic (Higgins 2003). They will use a simplified categorization of heterogeneity such that they consider heterogeneity to be low if the I2value is up to 25%, moderate up to 50%, high up to 75% and very high over 75%.
Assessment of reporting biases
The authors attempted to minimise publication bias by directly contacting manufacturers of nebuliser systems for data from all studies, regardless of positive or negative outcome, carried out with the specified medications and using the identified nebuliser systems. Had they been able to combine at least 10 studies, they planned to assess publication bias using a funnel plot analysis, bearing in mind that there are other reasons for funnel plot asymmetry, which would require caution in interpretation. The authors assessed selective reporting by comparing study protocols to final publications where possible, in order to make sure that all outcomes measured were reported. Where this was not possible, they compared the measurements identified within the methods section of the paper with the measurements reported on within the results section.
The authors analysed the included data using a fixed-effect model. They planned to employ a random-effects model if they identified moderate or high degrees of heterogeneity (as defined above).
Subgroup analysis and investigation of heterogeneity
As there were insufficient studies, the authors were unable to perform the planned subgroup analysis for children (up to 16 years) compared to adult (16 and older) and by disease severity. There are a number of ways of measuring disease severity and new methods are being developed as the decline in respiratory function slows over time in the CF population. However, spirometry has been an accepted standard in disease monitoring (Davies 2009) and provides a simple classification system for the purposes of this review. In future revisions, and when sufficient studies are available for subgroup analysis, the authors intend to describe disease severity as severe (FEV1 below 30% predicted), moderate (FEV1 between 31% and 60% predicted) and mild (FEV1 over 60% predicted) (Davies 2009).
The authors will perform sensitivity analyses when they are able to include a sufficient number of studies in the future. Where they identify moderate or high degrees of heterogeneity, they will test the robustness of their findings by performing a sensitivity analysis excluding studies with higher risk of bias (i.e. quasi-randomised). They will consider all sources of bias.