As tension pneumothorax is frequently a difficult clinical diagnosis encountered during emergent situations, the primary objective of this systematic review and meta-analysis is to utilize the existing world literature (including that from case reports and series) on tension pneumothorax in order to describe and contrast the clinical manifestations of the disorder among patients receiving positive pressure ventilation versus those that are breathing unassisted (i.e. breathing spontaneously and not receiving positive pressure ventilation).
A secondary objective is to determine if a difference exists in the time to cardiac or respiratory arrest or requirement for thoracic decompression (i.e. needle or tube thoracostomy) among patients receiving positive pressure ventilation as compared to those who are breathing unassisted.
Using Ovid, we will search MEDLINE and MEDLINE In Process & Other Non-indexed Citations, EMBASE, the Cochrane Database of Systematic Reviews, and the Cochrane Central Register of Controlled Trials (CENTRAL) from their first available date without restrictions. We will also query PubMed. To identify unpublished or ongoing/recently concluded studies, we will write colleagues and content experts, search personal files, and investigate two clinical trials registries (ClinicalTrials.gov and Current Controlled Trials). In addition, we will use the PubMed “related articles” and Google Scholar “cited by” features and hand-survey reference lists of included articles and tension pneumothorax review papers identified during the conduct of the search. Authors of selected articles will be contacted for additional information as necessary. Searches will be updated to within three months of submission of the systematic review for peer review.
Types of study to be included
Cohort, case-control, and cross-sectional studies as well as case reports and case series will be eligible for inclusion in the systematic review.
Condition or domain being studied
Tension pneumothorax, often defined as hemodynamic compromise in a patient with an expanding intrapleural air mass, is an uncommon yet potentially catastrophic clinical diagnosis most frequently encountered in the pre-hospital, Emergency Department, and Intensive Care Unit (ICU) settings. Although a valid estimate of the incidence of tension pneumothorax remains to be determined, this condition has been suggested to occur among 5% of major trauma patients managed in the pre-hospital environment and 1% to 3% of adult ICU patients. In one retrospective cohort study, the adjusted risk of death among mechanically ventilated patients was reported to be approximately 13 times higher among those that developed a tension pneumothorax as compared to those that did not.
The population of interest will include adults/adolescents (>=12 years old) or children (<12 years old) with a tension pneumothorax. We will define tension pneumothorax according to a previously suggested working definition as “a pneumothorax that results in significant respiratory or hemodynamic compromise that reverses on thoracic decompression alone”. Patients aged <12 years will be analyzed separately from adults as their vital signs vary significantly below this age, and their clinical manifestations likely differ from adults as their mediastinum and thoracic wall are more flexible.
Studies of any design, including case reports and series, which report original data on clinical manifestations of tension pneumothorax will be considered for inclusion in the systematic review. Although all observational studies, case reports, and case series of tension pneumothorax will be included, articles will be stratified according to whether two investigators independently agreed that the clinical condition of the patient met the tension pneumothorax working definition. Investigations of fatal cases will be included if the condition causing death was thought by the study authors to be suggestive of a tension pneumothorax and associated with expulsion of air following thoracic decompression or confirmed by a pathologist on autopsy. We will exclude studies that did not describe ventilation status, as well as reports of patients with diving-related pulmonary barotrauma; a previous contralateral pneumopneumectomy, traumatic diaphragmatic hernia, or tension pneumopericardium or pneumoperitoneum as well as chronic (as defined by the authors) or loculated pneumothoraces and investigations of patients undergoing thoracic surgery or laparoscopy at the time of onset of their tension pneumothorax clinical manifestations. Although reports of patients with bilateral pneumothoraces will also be excluded, those that describe the clinical manifestations of a tension pneumothorax in a patient with a previously treated contralateral pneumothorax will be included. All of the above excluded conditions were selected as they represent special, uncommon, or less relevant associated or principal patient conditions, which have the potential to misrepresent the more common clinical manifestations of tension pneumothorax.
The exposure of interest will include positive pressure ventilation. We will define positive pressure as either invasive (e.g., via an endotracheal tube and mechanical ventilator) or non-invasive (e.g., bag-valve-mask ventilation).
The clinical manifestations of patients who are receiving positive pressure ventilation will be compared to those who are breathing unassisted (i.e. breathing spontaneously and not receiving positive pressure ventilation) as the principal comparison exposure.
The principal outcome of the systematic review will include a comparison of the occurrence rate of clinical manifestations of tension pneumothorax among patients receiving positive pressure ventilation as compared to those that are breathing unassisted (i.e. spontaneously breathing and not receiving positive pressure ventilation). We will define clinical manifestations as patient-level findings/data, which may be gathered by clinicians during a medical interview or through physical examination, invasive monitoring or treatment equipment (e.g., intravascular catheters or mechanical ventilators), or diagnostic studies (e.g., echo- or electrocardiograms). As experimental evidence suggests that the hemodynamic status of the patient may be the principal difference between patients with varying respiratory status, we will specifically examine if adverse hemodynamic effects (decreases in arterial blood pressures or lower presenting arterial blood pressure values, hypotension, or cardiac arrest) are more common among patients receiving positive pressure ventilation as compared to those who are breathing unassisted.
We will also examine whether the time to cardiac or respiratory arrest (or requirement for thoracic decompression) differs among patients with a tension pneumothorax according to their presenting respiratory status.
Data extraction, (selection and coding)
Independently and in duplicate, two investigators will screen citation titles and abstracts, review potentially relevant articles in full, and extract data using a pre-designed electronic spreadsheet. These data collection instruments will be piloted on a random selection of five observational studies and 30 case reports/series until reliable data collection can be demonstrated (kappa statistic >=0.75). All non-English citations will be read in full by an interpreter, and, among those that satisfy selection criteria for inclusion in the systematic review, their data will be extracted only once by this individual. Data extracted from individual articles will include:
• 1) Study characteristics, including year of publication, Country of conduct, and design.
• 2) Characteristics of the included study participants, including the number enrolled as well as their age and gender, pneumothorax etiology (as suggested by the authors or through consensus between investigators), and ventilation status (i.e. whether they were receiving positive pressure ventilation or were breathing unassisted), as well as whether a definition of tension pneumothorax was provided and how this condition was specifically defined. We will define positive pressure ventilation as either invasive (e.g., via an endotracheal tube and mechanical ventilator) or non-invasive (e.g., bag-valve-mask ventilation).
• 3) Whether the following clinical manifestations were present before and after thoracic decompression: chest pain, dyspnea or shortness of breath, respiratory distress, subcutaneous emphysema, hypoxia [either arterial oxygen saturation (SpO2) <92% or partial pressure of arterial oxygen (PaO2) <8kPa/60 mmHg on room air or requiring supplemental oxygen], tachypnea (respiratory rate >=20), tachycardia (heart rate >=100), hypotension (systolic blood pressure<= 90 and/or mean arterial pressure <= 60 mmHg), jugular venous distention, increased peak inspiratory or airway pressure and/or whether resistance to assisted ventilation was noted by the clinicians managing the patient, any reported relevant invasive hemodynamic (e.g., cardiac output or central venous or pulmonary arterial pressures) or respiratory (mechanical ventilatory) measurements, and whether the patient developed a respiratory and/or cardiac arrest (and which occurred first). We will also record the first reported type of cardiac arrest rhythm and the approximate time to cardiac or respiratory arrest (or requirement for thoracic decompression) where available. Finally, we will extract values for respiratory and heart rates, systolic, diastolic, and mean arterial blood pressures, and SpO2 or PaO2 [as well as the accompanying fraction of inspired oxygen (FIO2) that the patient was receiving] at baseline and before and after thoracic decompression where available. Through consensus between the two extracting investigators, the clinical manifestations of tension pneumothorax will be extracted as proximal to the author’s description of its diagnosis and/or treatment as possible. We will accept authors’ definitions for subjective clinical manifestations such as respiratory distress and for the presence or absence of dichotomous clinical manifestations (e.g., tachycardia or hypotension) where absolute numbers are not afforded.
• 4) Whether the following ipsilateral and contralateral chest signs were present before and after thoracic decompression: percussion hyperressonance, decreased air entry, thoracic hyper- or hypo-expansion, chest wall hyper- or hypo-mobility, and contralateral tracheal deviation.
• 5) Whether the following X-ray or computed tomography signs of tension pneumothorax were present before and after thoracic decompression: a large (>50% total lung volume) pneumothorax, tracheal and/or mediastinal shift, increased rib spacing, and/or a flattened ipsilateral hemidiaphragm .
• 6) The initial and subsequent method of thoracic decompression for treatment of tension pneumothorax (needle or tube thoracostomy or another method) and whether these were effective.
• 7) Any confounding treatments or pathologies that could alter the clinical manifestations of patients with a tension pneumothorax, including sedative and cardiorespiratory depressant drugs, rib fractures and flail chest, pulmonary contusions, hemothoraces, high spinal cord injury, any reported type of pulmonary disease, and decreased level of consciousness [as defined by the authors or reported using a commonly used clinical cutoff on a validated scale [e.g., Glasgow Coma Scale (GCS) score =13].
Risk of bias (quality) assessment
Two investigators will also independently determine the risk of bias among the included studies. Risk of bias for cohort and case-control studies will be evaluated using the Newcastle Ottawa Scale, which evaluates selection and attrition bias, between-group comparability, and ascertainment of exposure bias using a starring system. We will also describe the recruitment (prospective versus retrospective) and enrollment (consecutive versus non-consecutive) method of each observational study, as well as whether a definition of tension pneumothorax was explicitly provided and whether this aligned with the working definition. Case reports and series will be evaluated as to whether the authors provided absolute numbers rather than narrative or subjective descriptions for clinical manifestations such as tachycardia or hypotension. We will also quantify the extent of missing data among the outcome variables used in statistical analyses. Disagreements in any of the above methodological assessments will be resolved by consensus or arbitration by a third investigator.
Strategy for data synthesis
Overview of the Data Synthesis Strategy:
In order to provide an evidence-informed description of the clinical manifestations of tension pneumothorax, we will first conduct a narrative synthesis of the systematic review results. Where appropriate, this will be followed by an observational and individual patient data meta-analysis using case reports and series. For the reasons outlined above under selection criteria, the below described data synthesis strategy will be performed separately for adults/adolescents and children.
All clinical manifestations reported among observational studies of patients with a tension pneumothorax who were receiving positive pressure ventilation will be compared to those reported for patients breathing unassisted either directly (if patients with both types of ventilation status existed within studies) or indirectly (if only studies of patients with one versus another type of ventilation status are available). This will be done qualitatively and, where possible, quantitatively through use of meta-analysis. Subsequently, summary statistics will be calculated describing the occurrence rate of clinical manifestations reported among case reports and series of patients with a tension pneumothorax stratified by patient ventilatory status. These results will then be compared to those obtained from the observational studies in order to examine for similarities and/or differences. Finally, an individual patient data meta-analysis will be conducted in order to examine whether differences may exist in clinical manifestations, including decreases in arterial blood pressures or lower presenting arterial blood pressure values, hypotension, and cardiac arrest, among patients receiving positive pressure ventilation versus those breathing unassisted.
Narrative Synthesis of Results:
Articles will first be grouped according to the ages of the included study patients (adults/adolescents versus children) and their design (observational versus case reports and series). After studies have been appropriately grouped, the principal characteristics of the observational studies (including their design, year of publication, setting, and a description of the clinical manifestations of the included patients stratified by ventilation status) will be presented in one or more summary tables before any statistical combination of their results is contemplated. For case reports and series, the details of the reported patients, including their baseline characteristics (e.g., age, gender, and the etiology of their pneumothorax); potential confounding conditions or treatments; and clinical, radiological, and invasive hemodynamic or respiratory clinical manifestations will first be presented in summary tables stratified by ventilatory status before any individual patient data meta-analyses are conducted. We will also present the reported times to cardiac or respiratory arrest (or, where these are unavailable, the time to requirement for thoracic decompression) among patients according to their ventilatory status.
Observational Study Data Meta-Analysis:
We will begin our observational study data meta-analysis by calculating individual study estimates of the occurrence rate of clinical manifestations among patients receiving positive pressure ventilation versus those that are breathing unassisted. The standard error and 95% confidence interval of these estimates will then be determined using the Clopper-Pearson exact binomial method.
Among studies involving patient populations with similar baseline characteristics and only one type of ventilatory status (e.g., positive pressure ventilation), individual study estimates of the occurrence rate of clinical manifestations will be pooled separately by patient ventilatory status. If studies provide the percentage of patients that were receiving positive pressure ventilation, but do not stratify their results by ventilatory status, then those in which <50% of patients were breathing unassisted will be included in the positive pressure ventilation category. Where possible, we will also determine a pooled estimate of the weighted or standardized mean difference in continuously measured clinical manifestations such as systolic or mean arterial pressures between these two patient populations (or the associated standardized mean difference in these values between baseline and presentation, where available). As variability in our pooled estimates beyond chance is expected across studies, these estimates will be calculated using random-effects models according to the method proposed by DerSimonian and Laird.
The pooled estimates obtained from the above calculations will then be compared qualitatively in order to determine if differences in the frequency of occurrence of clinical manifestations appear to exist according to the ventilatory status of the patients. Although we believe that it would be unlikely that any of the available observational studies will report adjusted odds ratios relating the frequency of occurrence of clinical manifestations between patients receiving positive pressure ventilation versus those breathing unassisted, where available these will also be pooled using random-effects models.
We will examine for evidence of between-study statistical heterogeneity by calculating I-squared inconsistency and Cochrane’s Q statistics (as part of a hypothesis test of heterogeneity). As suggested by Higgins and coworkers, we will consider I-squared statistics =25%, 50%, and 75% to represent low, moderate, and high degrees of inter-study heterogeneity, respectively. In the presence of greater than a low degree of between-study heterogeneity, we will conduct subgroup analyses and univariate meta-regression (p-value <0.10 considered significant given the low power of these tests) in order to explore the influence of sources of clinical and methodological study variation on the meta-analysis results. A priori study covariates of interest will include: 1) study temporality (prospective versus retrospective), 2) study subject enrolment (consecutive versus non-consecutive), 3) patient level of consciousness (decreased versus not decreased as reported by authors or measured the GCS  or another validated scale), 4) use of antihypertensive medications or the presence of pulmonary disease among =50% of patients who are breathing unassisted, and 5) proportion of patients receiving positive pressure ventilation (>=50% versus <50%).
Individual Patient Data Meta-Analysis Using Case Reports and Series:
After stratifying the results of case reports and series by the presenting age of the included patient(s) into adults/adolescents versus children, we will begin by examining the distribution of all continuous variables using histograms and measures of central tendency. Normally distributed data will be summarized using means (with standard deviations) and compared using t-tests (with an unequal variance option where appropriate) while skewed data will be summarized using medians (with interquartile ranges) and compared using the Mann-Whitney U-test. Dichotomous data will be summarized using proportions and compared using Fisher’s exact test.
As stepwise regression procedures may result in biased estimated coefficients and optimistic fits (especially with small sample sizes), our analyses will test the a priori hypothesis that patients who are receiving positive pressure ventilation have a higher risk of adverse hemodynamic complications. We will also test examine whether these complications may be confounded or modified by patient age and/or the presence of decreased level of consciousness or pulmonary disease. Although we will also test whether differences exist in the risk of other clinical manifestations between patients of varying respiratory status, these analyses will be largely exploratory and therefore more susceptible to type I error.
As the reported clinical manifestations of patients with a tension pneumothorax may be clustered or correlated within hospitals or treatment locations and are expected to occur relatively commonly, all associations will be determined using generalized estimating equations with a log or identity link, a binomial distributional family, and an exchangeable within-group correlation structure. These equations are a valid method for the analysis of common outcomes among correlated data, and can be used to adjust for the influence of data clustering when fitting multivariable binomial, logistic, and linear regression models. Potential modifying or confounding variables included in the model will include patient age (as a continuous variable) and the presence or absence of decreased level of consciousness and/or pulmonary disease (as dichotomous variables). If these models are unable to converge (as may occur when modeling the risk of an outcome), we will attempt to model the log using clustered logistic regression. Finally, any potential differences in presenting systolic, diastolic, or mean arterial blood pressure (or differences in presenting systolic, diastolic, mean arterial blood pressures from baseline) will then be examined using generalized estimating equations with an identity link, a Gaussian distributional family, and a similar modeling strategy.
In order to examine the robustness of our findings, we will also conduct a series of sensitivity analyses, which will be reported alongside our principal results upon submission of the study findings for peer review. First, we will assess whether our observed outcomes were sensitive to the definition of tension pneumothorax. After conducting analyses using the data extracted from all reported cases, this will be done by recalculating all outcomes using only those cases that both authors agreed met the tension pneumothorax working definition.
Second, as we anticipate that case reports and series may sometimes fail to report data on potentially important clinical manifestations (e.g., the presence or absence of hypotension), we will also perform multiple imputation and imputation. After performing these imputations, our regression analyses will then re-conducted on each of the imputed data sets. The estimated association obtained from each analysis will then be averaged to obtain an overall point estimate of the estimated association. Imputation will then be used in order to provide an estimate of the extremes of influence of the missing values on the estimated outcomes between groups. This will be performed by first assigning all the missing values in the positive pressure ventilation group a value of “0” and all the missing values for these variables in the breathing unassisted group a value of “1”. After recalculating the model with the inclusion of these imputed values, we will then reverse the assignment of the “0” and “1” values and then again recalculate the model in order to provide the opposite extreme estimate.
Analysis of subgroups or subsets
None planned. ***References for the above protocol available from the authors upon request.
Results of the systematic review will be disseminated through an end-of-grant approach (presentation at major regional and international meetings) and through an integrated knowledge translation intervention (where we have involved key stakeholders in several major groups who have an interest in the clinical manifestations of tension pneumothorax).
Contact details for further information
Departments of Surgery and Community Health Sciences
University of Calgary
Intensive Care Unit Administration
Ground Floor McCaig Tower
Foothills Medical Center
3134 Hospital Drive Northwest
Calgary, Alberta, Canada
Organisational affiliation of the review
University of Calgary
Dr Derek Roberts,
Dr Simon Leigh-Smith, Defence Medical Services, Royal Navy and Royal Infirmary Edinburgh, Edinburgh, United Kingdom Dr Peter Faris, Department of Community Health Sciences, University of Calgary and Alberta Health Services - Research Excellence Support Team, Calgary, Alberta, Canada Dr Chad Ball, Departments of Surgery and Oncology as well as Regional Trauma Services, University of Calgary and the Foothills Medical Centre, Calgary, Alberta, Canada Ms Helen Lee Robertson, Health Sciences Library, Health Sciences Centre, University of Calgary, Calgary, Alberta, Canada Dr Christopher Blackmore, Department of Surgery, University of Calgary and the Foothills Medical Centre, Calgary, Alberta, Canada Dr Elijah Dixon, Departments of Surgery, Oncology, and Community Health Sciences, University of Calgary and the Foothills Medical Centre, Calgary, Alberta, Canada Dr Andrew Kirkpatrick, Departments of Surgery and Critical Care Medicine, University of Calgary and the Foothills Medical Centre, Calgary, Alberta, Canada Dr John Kortbeek, Departments of Surgery and Critical Care Medicine, University of Calgary and the Foothills Medical Centre, Calgary, Alberta, Canada Dr H. Thomas Stelfox, Departments of Critical Care Medicine, Medicine, and Community Health Sciences, University of Calgary and the Foothills Medical Centre, Calgary, Alberta, Canada
Anticipated or actual start date
01 June 2013
Anticipated completion date
01 March 2014
Conflicts of interest
Subject index terms status
Subject indexing assigned by CRD
Subject index terms
Humans; Pneumothorax; Thoracic Injuries
Reference and/or URL for protocol
It is currently being submitted to Systematic Reviews for peer review.
Stage of review
Date of registration in PROSPERO
27 September 2013
Date of publication of this revision
27 September 2013
Stage of review at time of this submission
Piloting of the study selection process
Formal screening of search results against eligibility criteria
Risk of bias (quality) assessment
PROSPERO This information has been provided by the named contact for this review. CRD has accepted this information in good faith and registered the review in PROSPERO. CRD bears no responsibility or liability for the content of this registration record, any associated files or external websites.