Types of studies

Types of participants

Types of interventions

Types of outcome measures

We used the standard search strategy for the Cochrane Neonatal Review Group (see: Cochrane Neonatal Review Group, search strategy for Specialised Register in The Cochrane Library). The review authors undertook a comprehensive search which included searches of the following electronic databases: The Cochrane Central Register of Controlled Trials (CENTRAL, The Cochrane Library, Issue 2, 2007), MEDLINE (1966 to June 2007), EMBASE (1988 to June 2007), CINAHL (1982 to June 2007), PEDro (1929 to June 2007) and Web of Science. The review authors cross-referenced relevant literature including identified trials and review articles.

Searches of the electronic databases were based on the following search strategy:
1. physiotherapy OR physical therapy, AND
2. infant OR neonate OR newborn, AND
3. chest OR lung OR respiratory
4. NOT cystic fibrosis

The search also included searches of previous reviews including cross-references, abstracts, conference and symposia proceedings, expert informants and journal hand searching restricted to the English language. We handsearched conference Proceedings of the Society for Pediatric Research (SPR) (1967 to 2007), the European Society for Pediatric Research (ESPR) (2005), Pediatric Academic Societies Abstracts 2006, Australasian Physiotherapy conference abstracts (2003 to 2006), theses and dissertation lists, The Chartered Society of Physiotherapists (1988 to 2005) and trial registries (Australia, UK, USA) for unpublished work. No other language restrictions were applied.

We examined the title and abstract of each retrieved study to assess eligibility. If there was uncertainty, we examined the full paper.

We used the standard methods of the Neonatal Review Group of the Cochrane Collaboration.Two review authors searched for and considered trials for inclusion, evaluated methodological quality and extracted data independently. Differences in interpretation were resolved by discussion with a third review author.
Where necessary, we contacted investigators of identified trials for additional information regarding trial methodology or outcome data (Crane 1978; Etches 1978; Finer 1978; Curran 1979; Tudehope 1980; Hartrick 1982; Peters 1983; Fiser 1985; Leung 1998; Main 2001). Since many of these studies were performed quite a few years ago some of the authors either no longer had the data available (Etches 1978; Finer 1978; Curran 1979; Tudehope 1980) or could not be contacted (Crane 1978; Hartrick 1982). At the time of this review, additional data had been received from Peters 1983, and was included in this review.

Quality assessment
We conducted quality assessment according to the methods described in section six of the Cochrane Handbook for Systematic Reviews of Interventions (Handbook 2007). We considered four major sources of potential bias and methods of avoidance when assessing the trial quality:

(1) selection bias - blinding of randomisation;
(2) performance bias - blinding of intervention;
(3) attrition bias - complete follow-up;
(4) detection bias - blinding of outcome assessment.

A quality rating was assigned to each trial for the criterion of blinding of randomisation as follows: (A) adequate, (B) unclear, (C) inadequate, or (D) not used. A quality rating of (A) yes, (B) can't tell, or (C) no, was assigned to the other quality components (completeness of follow-up and blinding of outcome assessment). High quality trials were defined as those receiving an A rating for blinding of randomisation (central computerised randomisation service or sealed opaque envelopes). The quality assessment rating included in the Table of 'Characteristics of Included Studies' refers to blinding of randomisation (allocation concealment) only.

Data analysis
Two review authors independently extracted data using prepared data extraction forms. We resolved any discrepancies by discussion with a third review author.

We used the Review Manager software (RevMan 4.2.7) (RevMan 2003) for statistical analyses. In the cross-over trial (Peters 1983) investigators provided individual patient data for each intervention on oxygenation. From this data the authors derived the group means and standard deviations for per cent change in oxygenation. The numbers of infants who experienced one or more hypoxaemic episode during each intervention were included for the outcome of hypoxaemia.

Due to insufficient data it was not possible to conduct a meta-analysis. For individual trials, where possible, mean differences and 95% confidence intervals (CI) are reported for data measured on a continuous scale. For categorical outcomes, relative risk and 95% confidence intervals are reported.

Methods for future updates of this review
We did not implement all of the methods outlined in the protocol as there were insufficient data.
For future updates of the review, we will analyse categorical data using relative risk (RR), risk difference (RD) and number needed to treat (NNT) where appropriate. We will use a fixed-effect model to pool results. We will use weighted mean differences (WMD) for data measured on a continuous scale. We will report the 95% confidence intervals (CI) for all estimates. We will perform sensitivity analyses to evaluate the effect of the trial quality (excluding quasi-randomised trials and considering trials with minimal bias). We will assess heterogeneity by visual inspection of the outcomes tables and by using an I-squared test of heterogeneity (Higgins 2002). Where statistical heterogeneity is found, the review authors will look for an explanation using prespecified subgroup analyses.

The following subgroup analyses will be performed:

1. Population

• Gestational age: < 30 weeks; < 37 weeks; 37 weeks and over

• Birth weight : < 1500 grams, < 2500 grams, 2500 grams and over

• Underlying pulmonary disorder: respiratory distress syndrome (RDS), aspiration, infection or chronic lung disease (CLD)

2. Intervention - techniques

• Type of technique: percussion (including cupping with a face mask, contact heel percussion and finger percussion); vibration (with fingers or mechanical vibrator)

• Frequency: four hours or less; > four hours

The search strategy initially retrieved 2036 references of which 1993 publications were excluded based upon title and abstract. We identified the remaining 43 publications as potentially eligible for inclusion in this review and requiring more detailed examination. After examination of the publications, we excluded 40 publications from the review. Thirty-one publications were excluded as they did not fulfil the inclusion criteria. Of these 31, we excluded 23 trials as they were not randomised or quasi-randomised controlled trials, six trials because the participants were not newborn infants with respiratory disease in the first four weeks of life, one because the participants were not mechanically ventilated, and a further one trial because the intervention was not a physiotherapy technique. We excluded a further four trials (Crane 1978; Hartrick 1982; Fiser 1985; Leung 1998) as the publications were in the form of conference abstracts and despite attempts to do so no further data could be obtained from the authors.
We excluded another five trials (Etches 1978; Finer 1978; Curran 1979; Tudehope 1980; Main 2001) as no useable data were reported or could be obtained from the authors.

(For further details on excluded studies see the table of 'Characteristics of Excluded Studies').

One ongoing trial was identified (Hough 2007) (See table of 'Characteristics of Ongoing Studies').

Three trials were included in the review.

The study by Raval 1987 compared the effects of CPT in 20 premature neonates with a gestational age of 28 to 32 weeks. Infants who were intubated and ventilated for respiratory distress syndrome (RDS) were randomly assigned to two groups. The intervention in the CPT group (n = 10) consisted of postural drainage in the head up and head down position, percussion, vibration and suction. The control group (n = 10) procedure consisted of suctioning only. Each of these procedures was performed every two to three hours in the first 24 hours of life. Outcomes measured used were continuous TcPO2 monitoring, arterial blood gas (ABG) sampling, weight of sputum cleared, mortality rate and rates of intracranial lesions.

Peters 1983 evaluated the responses of critically ill neonates to two different techniques of CPT administration. Thirty neonates (25 to 42 weeks gestational age) who were 28 days old or younger were randomised in a two-period cross-over design to one of two groups. Group 1 received postural drainage, endotracheal instillation of saline, suction and percussion consisting of cupping with a Bennett face mask (Bennett percussion) followed by Boyd percussion (contact heel percussion) an hour later. Group 2 received the same interventions in the reverse order. The neonates' responses for heart rate, mean arterial pressure (MAP), intracranial pressure (ICP), TcPO2 and TcPCO2 were recorded during the pre-CPT, postural drainage, percussion, ETT instillation and suctioning, and immediately post-CPT phases.

Wong 2003 enrolled 56 neonates who were aged less than 37 weeks gestation and required mechanical ventilation if they met the inclusion criteria of the presence of a segmental or lobar collapse confirmed on chest x-ray. Lung squeezing technique (LST), a form of manual chest wall compression, was compared in a group of 26 neonates with conventional treatment of postural drainage, percussion and vibration (PDPV). Therapy sessions were performed twice daily at specified times until resolution of atelectasis occurred. Re-expansion of atelectasis, recurrence of atelectasis, duration of ventilation, duration of oxygen dependency, occurrence of bronchopulmonary dysplasia (BPD), rates of intracranial lesions and mortality rate were the outcomes measured in this study.

Types of participants
A total of 106 infants were randomised in the three included trials investigating CPT in neonates.
Participant recruitment and inclusion criteria varied between the three trials; Raval 1987 included preterm infants with a birth weight less than 2000 grams, Wong 2003 recruited mechanically ventilated infants with a gestational age of less than 37 weeks, and Peters 1983 included any newborn infant requiring intubation. Gestational ages varied from between 28 to 32 weeks (Raval 1987), 22 to 36 weeks (Wong 2003) and 25 to 42 weeks (Peters 1983). Birth weights varied correspondingly with one study not stating birth weight (Raval 1987), the second ranging from 540 to 2900 grams (Wong 2003), and in third the study (Peters 1983) the infants tending to be heavier (ranging in weight from 750 to 4030 grams). Postnatal age at the time of the study varied from 1 to 12 hours of age (Raval 1987) to 1 to 10 days (Peters 1983). In the trial by Wong 2003, the postnatal age was not stated. The primary diagnosis for enrolled infants in Raval 1987 was RDS, whereas for the study by Wong 2003 the main diagnosis was atelectasis. In the third study, the diagnoses included hyaline membrane disease (HMD), pneumonia and transient tachypnoea of the newborn (Peters 1983). All three studies included only infants who were mechanically ventilated. Only two of the studies stated exclusion criteria. Peters 1983 included infants requiring sedation for irritability and/or restlessness, while Wong 2003 stated their exclusion criteria as persistent pulmonary hypertension, meconium aspiration syndrome, congenital heart defects, pneumonia presenting with generalised patchy consolidation, post-cardiothoracic surgery, pleural effusion and pneumothorax.

Types of interventions
The types of interventions assessed differed across the included studies. Raval 1987 compared CPT comprising postural drainage in the head up and in the head down positions, percussion, vibration and suctioning (PDPV) with a control group receiving standard care. Each of these procedures was performed every two to three hours throughout the first 24 hours of life. Wong 2003 also used the PDPV technique combination but compared this with another CPT technique called the lung squeezing technique (LST) until resolution of atelectasis. The remaining study compared individual techniques of percussion with a face mask (cupping) and contact heel percussion using a cross-over design (Peters 1983). Each infant received each technique once in a randomised order over approximately a three hour period.

Types of outcome measures
Short-term physiological outcomes of oxygenation could be included from one cross-over trial (Peters 1983). In this trial, transcutaneous oxygenation, changes in oxygen requirements and hypoxic episodes were measured 10 minutes pre and post-intervention. Although Peters 1983 measured PaCO2, we could not include this data in the analyses as the standard deviation data could not be obtained from the author. In the trial by Raval 1987, oxygenation was reported in figure format and as the authors could not be contacted, we could not include the data on oxygenation in the review. Secretion clearance was also used as an outcome measure in two of the studies (Raval 1987; Wong 2003). Wong 2003 defined the secretions as obvious if the amount collected during a single intervention was greater than 0.2 ml. Raval 1987 measured the amounts of secretions collected over a 24-hour period and reported average secretion per procedure. Rates of intracranial lesions diagnosed by ultrasound scan and mortality rate were measured in the two parallel group studies. Ultrasound scans were done at three to five days of postnatal life in one trial (Raval 1987) and for three consecutive days after the first intervention in the other (Wong 2003). For the purposes of this review the highest grade of haemorrhage was included. Although the study by Peters 1983 did not include ultrasound findings, they did report changes in intracranial pressure and mean arterial pressure; however, these measurements were not included in our pre-specified outcomes. Wong 2003 was the only study to document duration of ventilation, duration of oxygen dependency and bronchopulmonary dysplasia, and the presence of atelectasis on chest radiographs.

(For further details on included studies see the table, 'Characteristics of Included Studies').

Concealment of allocation
Of the three included studies, only one reported an adequate method of allocation concealment (Wong 2003). This study used a computerised random number allocation placed in sealed opaque envelopes, and stratified by mode of ventilation [high frequency oscillatory ventilation (HFOV) or conventional mechanical ventilation (CMV)]. The other two reported randomly assigning participants to groups but did not report on the method of allocation concealment (Peters 1983; Raval 1987).

Blinding of intervention
Due to the nature of the intervention of CPT, blinding of intervention was not employed in any of the included studies.

Completeness of follow-up
There was no loss to follow-up reported in any of the included studies. Only one study reported withdrawals (Wong 2003); however, outcomes for these infants were included in the analysis.

Blinding of outcome measures
In summary, only one study was of a high methodological quality (Wong 2003). It is difficult to determine adequately the quality of the remaining two trials (Peters 1983; Raval 1987) due to lack of information in the reported studies and the inability to obtain clarification from the authors.

The results of three trials are included in this review (Peters 1983; Raval 1987; Wong 2003).

Active CPT versus nonactive techniques/standard care (Comparison 1)

One trial (Raval 1987) with a total of 20 infants compared active CPT in the form of postural drainage, percussion and vibration (PDPV) with the nonactive technique of suction alone.

Primary outcome measures
(Duration of mechanical ventilation, duration of supplemental oxygen, duration of hospital stay):
Raval 1987 did not address any of the preplanned primary outcome measures.

Secondary outcomes

Weight of secretions cleared per procedure per day (Outcome 1.1)

There was no statistical difference in the amount of sputum cleared with CPT as compared to the amount cleared with suction alone (MD 0.01 g; 95% CI -0.10 to 0.12).

Rates of IVH (Outcomes 1.2 - 1.3)

No statistically significant difference was found for either the rate of any intraventricular haemorrhage (RR 2.33; 95% CI 0.83 to 6.54) or for the rate of a grade 3-4 IVH (RR 11.00; 95% CI 0.69 to 175.86).

Comparison of active CPT techniques (Comparison 2)

One trial (Peters 1983) (n = 30) reported change in oxygenation comparing the techniques of percussion versus cupping.

Primary outcome measures
(Duration of mechanical ventilation, duration of supplemental oxygen, duration of hospital stay):
Peters 1983 did not address any of the preplanned primary outcome measures.

Secondary outcomes measures

Change in oxygenation per procedure per baby (Outcome 2.1 - 2.2)

An increase in the incidence of hypoxaemia was shown with cupping when compared to contact heel percussion (RR 0.53; 95% CI 0.28 to 0.99) and an associated increase in oxygen requirement (MD -9.68; 95% CI -14.16 to -5.20).

Comparison of LST and PDPV (Comparison 3)

One trial (Wong 2003) (n = 56) compared the techniques of LST and PDPV.

Primary outcome measures
(Duration of mechanical ventilation, duration of supplemental oxygen, duration of hospital stay):
Wong 2003 did not address any of the preplanned primary outcome measures.

Secondary outcomes

Non-resolution of atelectasis after initial intervention (Outcome 3.1)

When compared with PDPV, LST resulted in a reduction in the numbers of infants with non resolution of atelectasis (RR 0.25; 95% CI 0.11 to 0.57).

Secretion clearance per procedure per baby (Outcome 3.2)

There was no statistically significant difference between the LST and PDPV groups in the numbers of infants who had less than 0.2 ml of secretions removed post-treatment (RR 0.96; 95% CI 0.62 to 1.49).

Rates of IVH grade 3-4 (Outcome 3.3)

There was no difference in the rate of progression of IVH to grade 3-4 when LST was compared to PDPV (RR 0.87; 95% CI 0.35 to 2.17).

Rates of any PVL (Outcome 3.4)

There was no difference in the rate of PVL (RR 0.87; 95% CI 0.21 to 3.52) between the techniques of LST and PDPV.

Due to small numbers, meta-analyses and planned subgroup analyses could not be undertaken.

The results of this review do not provide sufficient evidence to guide clinical practice on the use of CPT techniques in infants on ventilatory support in today's neonatal intensive care settings. Although concern for the safety of active CPT in preterm infants has been reported, possible adverse effects of CPT could not be evaluated due to insufficient data. In view of this and the lack of clear evidence for benefit, it would seem wise to use this intervention cautiously.

Further well-designed trials are required to assess the risks and benefits of CPT in the treatment of respiratory diseases in ventilated neonates. Future trials should be adequately powered to address clinically important outcomes, particularly for the high risk population of infants < 30 weeks gestation. Clinically important outcomes which should be assessed include duration of ventilation, duration of oxygen therapy, length of hospital stay, and presence of intracranial lesions. Shorter-term outcomes such as resolution of atelectasis, oxygenation, and other lung function variables such as ventilation distribution should also be included. Costs also need to be considered.

Some of the important clinical and economic outcomes that remain unmeasured by current research are likely to require a very large sample size, therefore, a large multicentre trial would be recommended.

Risk of bias table

Risk of bias table

Risk of bias table