Methods
Data sources and searches
We aimed to identify all randomized controlled trials (RCTs) comparing HMEs and HHs in adult critically ill patients. We applied standard filters for the identification of RCTs using the MEDLINE and PUBMED search engines (from inception to June 2014), using English language restrictions. Our search included the following keywords: heat and moisture exchangers, heated humidifiers, airway humidification, artificial humidification, artificial airway occlusion, mortality, pneumonia and humans and randomized clinical trial.
Selection of studies
Trials comparing any type of HH, including systems with heated and non-heated wire, with HMEs in adult critically ill patients were included. We restricted the analysis to RCTs to guarantee control of selection bias. We included only published full papers and excluded abstracts. Study designs containing inadequately adjusted planned co-interventions and crossover trials were excluded. The intervention of interest was the use of HH and HME in reducing artificial airway occlusion, pneumonia and mortality. Studies were further divided according to the use of HH with heated and HH with non-heated wire to perform the subgroup analysis.
Outcome measures
The primary outcome was the incidence of artificial airway occlusion; the secondary outcomes were the incidences of pneumonia and mortality.
Data extraction and quality assessment
Initial selection was performed by two pairs of independent reviewers (MV and DC, PP and YS) screening titles and abstracts. For detailed evaluation, a full-text copy of all studies of possible relevance was retrieved. Data from each study were extracted independently by paired and independent reviewers (LB and DC, PP and YS) using a standardized data abstraction form. Data extracted from the publications were independently checked for accuracy by two other reviewers (GS and AE). Quality assessment of these studies included: (1) use of randomization sequence generation, (2) reporting and type of allocation concealment, (3) blinding, (4) reporting of incomplete outcome data and (5) comparability of the groups at baseline. Quality assessment is reported in Additional file
1. Two reviewers (MV and LB) independently used these criteria to evaluate trial quality. We solved any possible disagreement by consensus in consultation with two other reviewers (GS and AE) if needed.
Qualitative analysis
A narrative summary approach was used to explore study characteristics and quality indicators in describing variation among studies and to consider possible implications for this in our understanding of the outcomes of the RCTs included in the Cochrane review [
5,
6].
Quantitative analysis
The meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines [
7]. Meta-analysis was performed with mixed random effect using the DerSimonian and Laird method. Results were graphically represented using forest plot graphs. The relative risk (RR) and 95% CI for each outcome were separately calculated for each trial, pooling data when needed, according to an intention-to-treat principle. The choice to use RRs was driven by the design of meta-analysis based on RCTs. Tau
2 was used to define the variance between studies. The difference in the estimates of treatment effect between the treatment groups for each hypothesis was tested using the two-sided
z test with statistical significance considered at
p value <0.05. The homogeneity assumption was checked with the
Q test with a degree of freedom (df) equal to the number of analyzed studies minus 1. The heterogeneity was measured by the
I
2 metric, which describes the percentage of total variation across studies that is due to heterogeneity rather than chance.
I
2 was calculated as:
I
2 = 100% Å ~ (Q − df)/Q
where Q is Cochran’s heterogeneity statistic and df is degrees of freedom. A value of 0% indicates no observed heterogeneity, and larger values show increasing heterogeneity. We decided a priori to analyze all the outcomes according to the following categories when possible: HME vs HH, HME vs HH with heated wire and HME vs HH with non-heated wire.
Weighted random-effect meta-regression analysis was performed to test the relationship between the effect size on each considered outcome and the following covariates, with each one analysed separately: duration of MV, pneumonia incidence, intensive care unit (ICU) length of stay, percentage of respiratory diagnoses at ICU admission, simplified acute physiology score (SAPS), age and acute physiology and chronic health evaluation II (APACHE II) score.
Analyses were conducted with OpenMetaAnalyst (version 6) and SPSS version 20 (IBM SPSS). Weighted linear regression was used to evaluate potential publication bias, with the natural log of the RR as the dependent variable and the inverse of the total sample size as the independent variable. This is a modified Macaskill test that gives more balanced type I error rates in the tail probability areas in comparison to other publication bias tests [
8].
Discussion
In this systematic review and meta-analysis, we found: (1) no significant difference in artificial airway occlusion, pneumonia or mortality between HMEs and HHs, (2) no effect of HHs with and without heated wire compared to HMEs; however HHs with non-heated wire had the lower RR for artificial airway occlusion compared with HME, and (3) independently from the HH type, an advantage of HMEs in airway occlusion incidence was observed in studies with high incidence of pneumonia, and a trend toward favoring HMEs was observed for studies with prolonged MV. To our knowledge, this is the first systematic review performed (1) by dividing RCTs according to HHs with heated and non-heated wire and (2) including a meta-regression analysis on the potential effects of clinical variables on the efficacy of the two devices.
According to the American Association for Respiratory Care (AARC) guidelines, HHs should provide an absolute humidity level of between 33 and 44 mgH2O/L, whereas HMEs should provide a minimum humidity level of 30 mgH2O/L [
1]. HHs may produce insufficient heat and humidification when the temperature is improperly selected or pre-set at a non-adjustable level rather than at the clinical setting [
1]. However, insufficient heat and humidification may occur with HMEs too [
1]. Only 37% of HMEs have been found to meet the standard criteria advocated by the AARC guidelines [
1]. Insufficient airway humidification may lead to an increase in tracheal tube occlusion, a serious adverse event that may occur in mechanically ventilated patients and requires timely intervention. In an RCT comparing HMEs and HHs with increasing minute ventilation, the authors found that after 72 hours the inner diameter of the endotracheal tube decreased by 2.5–6.5 mm when gas conditioning was performed using HMEs and by 1.5 mm with HHs [
72]. A systematic review showed that in patients ventilated more than 48 hours, there is no difference in tracheal tube occlusion when comparing HMEs and HHs [
84]. In this meta-analysis, we found no differences in the incidence of artificial airway occlusion, but stratifying the comparison according to the type of HHs we found less risk for airway occlusion in HHs without heated wire compared with HME. However, these data were not confirmed by the sub-analysis comparing hydrophobic and hygroscopic HMEs with HHs. Probably, the main determinants of artificial airway occlusion are the duration of mechanical ventilation and pneumonia, rather than humidifier type per se
, even if a prolonged use of HME (<72 hours) may increase this risk. Long-term invasive MV and the presence of ventilator-associated pneumonia (VAP) increased the risk of artificial airway occlusion threefold in one study [
85] and twofold in another [
86]. This is the first meta-analysis reporting a meta-regression of included studies in this field. Our meta-regression showed that in studies with high incidence of pneumonia and prolonged MV, the HMEs had a slight advantage in terms of the artificial airway occlusion.
Earlier models of HME were associated with an increased incidence of airway occlusion, which led to the exclusion of patients at high risk from the studies [
86]. In contrast, trials using HMEs with enhanced intrinsic humidifying performance showed no difference in the incidence of airway occlusion [
86]. A Cochrane review states that hydrophobic HMEs may reduce the risk of pneumonia and the use of an HME may increase artificial airway occlusion in certain subgroups of patients [
4]. Our analysis includes more recently published studies. According to our meta-regression, the HME may reduce the risk of airway occlusion in selected patients affected by pneumonia.
Hess et al. concluded in their clinical practice guidelines that HMEs are associated with lower incidence of pneumonia compared with HHs [
85]. However, there are concerns about the increased airway resistance and care of HME filters [
85]. Kola et al. found a significant reduction in pneumonia using HMEs during MV, particularly when patients are ventilated for 7 days or more [
86]. Hess et al. included studies published between 1990 and 1998 in their analysis of pneumonia [
85]. Kola et al. reported the same results as Hess et al. but they only included one more study in their meta-analysis. Accordingly, the underlying mechanism of reduction in pneumonia may be due to the dryness of the ventilator circuit when using HMEs [
85,
86]. Therefore, HMEs minimized the need for septic manipulations or aspirations of the airway/circuit and the circuit condensate [
85,
86]. Furthermore, the inclusion of more recent studies may have changed the results. Indeed, Siempos et al. did not find any superiority of HMEs compared to HHs in reducing pneumonia, mortality or morbidity [
87]. The results of Siempos et al. were groundbreaking and in line with the RCTs published at that time. The inclusion of three further RCTs [
80‐
82] with 870 patients dramatically changed the previous results. The Cochrane review by Kelly et al. included adult and pediatric patients treated with HMEs and HHs [
4]. There was no overall effect on artificial airway occlusion, mortality, pneumonia or respiratory complications; however, the arterial partial pressure of carbon dioxide (PaCO
2) and minute ventilation were increased while body temperature was lower when HMEs were compared to HHs [
4] in a meta-analysis including 18 RCTs and 2442 adult critically ill patients. In line with the available literature, we did not find any difference in artificial airway occlusion, pneumonia or mortality between HMEs and HHs, even if according to their nature, they have different characteristics. However, HMEs were found to increase the PaCO
2 and work of breathing probably due to higher dead space, and to reduce the inner diameter of the endotracheal tube during prolonged MV [
4]. Indeed, HME may negatively impact on ventilator function while increasing the dead space [
11]. In spontaneously and assisted breathing patients, this requires increased minute ventilation and then the work of breathing, to maintain constant alveolar ventilation and PaCO
2. In controlled MV, the additional dead space of HMEs may reduce alveolar ventilation and increase PaCO
2 [
11]. This effect of HME dead space may be further exacerbated by protective ventilation at low tidal volume (VT) and by using HMEs with a larger dead space [
11].
Humidification is mandatory during MV. Nowadays, the airway humidification is appropriate in the absence of any contraindications listed by the AARC guidelines, such as altered body temperature, airway thermal injury, under hydrated secretions, increased work of breathing, hypoventilation, condensation and airway dehydration [
1]. Clear advantages in terms of clinical outcomes for different humidification devices are far from being demonstrated. The present meta-analysis reported no superiority of HMEs over HHs in term of clinical outcomes, with similar results even when stratifying the studies according to the type of HH, while some advantage of HMEs might be possible in patients with pneumonia or those with a long MV time. The choice of humidifiers should be made according to the clinical context trying to avoid possible complications and reaching the appropriate performance at lower costs. However, to help clinicians make the correct choice between HHs and HMEs, further high-quality RCTs are needed to evaluate the incidence of respiratory complications other than pneumonia, gas exchange and work of breathing when comparing the HH and HME devices.
This systematic review and meta-analysis has several limitations that must be addressed. First, the quality of the included RCTs was relatively low and our conclusions may be limited by this point. Second, the diagnosis of pneumonia was differently defined across the studies and often mixed with VAP. Third, the definition of mortality varied between the studies: three studies reported the ICU mortality, two studies reported overall mortality, two studies reported hospital mortality, one studies reported mortality during MV and in four studies mortality was not reported. Fourth, we performed the meta-analyses of outcomes if reported by more than three RCTs. Fifth, as we found only one additional RCT published between 2006 and 2013, the present results may depend on the studies published before 2006. However, in contrast to previous reports we included a meta-regression analyzing and interpreting data from a new point of view. Sixth, we found few studies reporting the effective tracheal tube lumen and most of them with provided a poor definition of pneumonia diagnosed at ICU admission. Seventh, we are not able to stratify the meta-analysis according to the baseline respiratory condition or inclusion criteria. This did not allow us to suggest the use of HMEs or HHs in different respiratory diseases.
New, prospective RCTs are needed in terms of assessing the effect of HHs vs HMEs in patients with respiratory failure due to pneumonia, on airway diameter, the amount of secretion and the occurrence of artificial airway obstruction and VAP.