Doppler trans‐thoracic echocardiography for detection of pulmonary hypertension in adults
Abstract
This is a protocol for a Cochrane Review (Diagnostic test accuracy). The objectives are as follows:
To determine the diagnostic accuracy of trans‐thoracic Doppler echocardiography for detecting pulmonary hypertension.
Background
Target condition being diagnosed
Pulmonary hypertension (PH) is an important cause of morbidity and mortality. Although the exact prevalence of PH is not known due to the various etiology, a previous study reported that PH affects up to 100 million people worldwide (Schermuly 2011). Pulmonary hypertension leads to a substantial loss of exercise capacity, as well as causing right ventricular overload, resulting in heart failure and early mortality. A recently published study reported that three‐year survival ranges from 58.2% to 73.3% (Ling 2012).
Pulmonary hypertension is diagnosed when resting mean pulmonary arterial pressure is 25 mmHg or more at right‐heart catheterization. It is a progressive disease that if left untreated can be fatal, although the rate of progression is highly variable.
Pulmonary hypertension is divided into five etiological categories according to World Health Organization (WHO) criteria (Simonneau 2013). Group 1 is pulmonary arterial hypertension (PAH). This group consists of sporadic idiopathic pulmonary hypertension (IPAH), heritable IPAH, and PAH due to diseases that localize to small pulmonary muscular arterioles. These diseases include connective tissue disease, HIV infection, portal hypertension, congenital heart disease, schistosomiasis, chronic hemolytic anemia, persistent pulmonary hypertension of the newborn, pulmonary veno‐occlusive disease, and pulmonary capillary hemangiomatosis. Drug‐induced PAH (including anorexigenic) and toxin‐induced PAH are also considered to belong to group 1 PAH.
Group 2 PH is pulmonary hypertension due to left heart disease, and is the most common form of PH worldwide. Pulmonary hypertension due to systolic dysfunction, diastolic dysfunction, or valvular heart disease is included in this group.
Group 3 PH is pulmonary hypertension due to lung disease or hypoxaemia. This includes PH caused by chronic obstructive pulmonary disease (COPD), interstitial lung disease, pulmonary disease with a mixed restrictive and obstructive pattern, sleep‐disordered breathing, and alveolar hypoventilation disorders.
Group 4 PH is chronic thromboembolic pulmonary hypertension due to chronic thromboembolic occlusion of the proximal or distal pulmonary vasculature.
Group 5 PH is pulmonary hypertension with unclear, multifactorial mechanisms.
Although the epidemiology of PH varies among the five groups, the majority of available evidence relates to group 1 PAH. The prevalence of group 1 PAH in the general population is estimated to be 5 to 15 cases per 1 million adults (Humbert 2006; Ling 2012). The prevalence of PH in groups 2 to 5 is unknown due to the broad classification and multiple etiologies. One cohort study reported that the proportion of each group among people with pulmonary hypertension is 0.18, 0.35, 0.17, 0.09, and 0.21 (Meredith 2014).
The prognosis of PH depends on various factors. The Registry to Evaluate Early and Long‐term PAH Disease Management (REVEAL) risk score is used to predict disease progression (Benza 2010). Poor prognostic indicators include age at initial presentation > 50 years (Peacock 2007), male sex and ≥ 60 years (Marcus 2008), persistent (WHO) functional class III or IV (Appendix 1), pericardial effusion and elevated right atrial pressure (Paulus 2007; Torbicki 2007). The natural history and prognosis of group 1 PAH is better studied than that of groups 2 through 5. In general, in the absence of therapy, those with group 1 PAH have worse survival than groups 2 through 5. Symptomatic patients with IPAH who do not receive treatment have a median survival of approximately three years. Data from the REVEAL registry reported that, from the time of diagnostic right‐heart catheterization, people with PAH had one‐, three‐, five‐, and seven‐year survival rates of 85%, 68%, 57%, and 49%, respectively (Benza 2010).
Early detection and treatment of PH is suggested because advanced disease may be less responsive to therapy (Galie 2013). Primary therapy of PH is directed at the underlying cause of the PH. People with persistent PH with WHO functional class II, III, or IV despite treatment of the underlying cause of the PH should be referred to a specialized center to be evaluated for advanced therapy. Advanced therapy is directed at the PH itself, rather than the underlying cause of the PH. It includes treatment with prostacyclin pathway agonists, endothelin receptor antagonists, phosphodiesterase 5 inhibitors, and soluble guanylate cyclase stimulant.
Index test(s)
Estimated systolic pulmonary artery pressure (PAP) by Doppler trans‐thoracic echocardiography will be an Index test. Systolic PAP can be estimated from the maximum tricuspid regurgitation (TR) jet velocity by using the modified Bernoulli equation and adding right atrial pressure (RAP) (systolic PAP = 4 (v)2 + RAP, where v is the peak velocity of the TR jet) (Berger 1985). The majority of people have some degree of TR, and the utilization of TR to estimate systolic PAP is the most common practice in echocardiography (Kaplan 2010). There are several methods to estimate RAP (e.g. clinical estimation from jugular venous pressure, using a fixed value from 5 mmHg to 10 mmHg, using the diameter and collapse of the inferior vena cava during spontaneous respiration); however, using the diameter and collapse of the inferior vena cava during spontaneous respiration is the most common method to estimate RAP (Rudski 2010). Several studies demonstrated adequate correlation between the estimated systolic PAP by Doppler trans‐thoracic echocardiography and the direct measurement of mean PAP with right‐heart catheterization (Zhang 2010), and if the estimated systolic PAP is higher than 35 to 40 mmHg, further evaluation is recommended to determine if PH is present (Rudski 2010). Although several studies demonstrated adequate correlation between the estimated systolic PAP and by echocardiography and the direct measurement of right‐heart catheterization, several studies reported that overinflated lung lowered the diagnostic accuracy of echocardiography (Fisher 2007).
Clinical pathway
Symptoms and signs of PH are non‐specific, which frequently results in a delay in diagnosis (Brown 2011). The initial symptoms of PH are exertional dyspnoea, fatigue, chest pain, syncope, palpitations, and peripheral edema. Many people with PH visit hospital complaining of these non‐specific symptoms.
Physicians suspect the presence of PH by these nonspecific symptoms or initial evaluation by chest radiograph or electrocardiogram. The classic chest radiograph of a person with PH shows enlargement of the central pulmonary arteries, right ventricular enlargement, and right atrial dilatation. Electrocardiogram of a person with PH may show signs of right ventricular disease, which include right axis deviation, an R wave/S wave ratio greater than 1 in lead V1, incomplete or complete right bundle branch block, or increased P wave amplitude in lead II.
When PH is suspected by these signs or initial screening tests, diagnostic evaluations are performed to confirm that PH exists and to determine its severity and identify its cause (Figure 1) (Rubin 2016).
Clinical pathway of diagnostic evaluation for pulmonary hypertension among adults, adolescents, and children.
PH: pulmonary hypertension; PAH: pulmonary arterial hypertension (Rubin 2016)
The first step of diagnostic testing is evaluation with echocardiogram. Pulmonary artery pressure can be estimated by this non‐invasive method. When echocardiogram does not suggest PH, further evaluation depends on clinical suspicion. If clinical suspicion for PH is still high, right‐heart catheterization should be considered. When echocardiogram is suggestive of PH, clinicians should evaluate if left heart disease exists to adequately explain the degree of estimated PH. Patients with enough left heart disease on the echocardiogram to explain the degree of estimated PH do not require further evaluation to determine the etiology of PH.
Patients who have no left heart disease, which seems insufficient to explain the degree of estimated PH, should undergo additional diagnostic testing to determine the etiology of PH and appropriate treatment, which may include pulmonary function test, ventilation‐perfusion scanning, overnight oximetry, polysomnography, or laboratory testing (e.g. autoimmune serologies, HIV serology, and liver function tests) according to the history and physical examination.
Right‐heart catheterization is indicated to confirm the PH and its severity. Direct pressure measurement of pulmonary artery pressure with right‐heart catheterization is the clinical reference standard to confirm PH (Lewis 2016).
Alternative test(s)
Some additional tests (e.g. chest radiograph, magnetic resonance, computed tomography) are considered to confirm the etiology of PH, however Doppler echocardiography plays a central role in the diagnosis and management of PH. Doppler echocardiography or right‐heart catheterization is essential to confirm the elevation of PAP (Freed 2016).
Rationale
Many patients are diagnosed at a late stage of the disease because at the beginning of the disease symptoms and signs of PH are non‐specific (Brown 2011). Early and accurate detection of PH would therefore be of great benefit. While direct pressure measurement with right‐heart catheterization is the clinical reference standard for PH, it is not routinely used due to its invasiveness and complications (Connors 1996).
Trans‐thoracic Doppler echocardiography is less invasive, less expensive, and widely available compared with right‐heart catheterization; it is therefore recommended that echocardiography should be used as an initial screening method and as a method of monitoring disease progression (Rudski 2010).
However, several studies have questioned the accuracy of non‐invasively measured PAP (Arcasoy 2003; Fisher 2009; Rich 2011). To avoid late detection, accurate triage of PH is necessary. We will therefore evaluate the diagnostic accuracy of echocardiography to triage patients suspected PH.
We hypothesize that Doppler echocardiography could be a beneficial test to triage PH since ultrasound devices have become common in a variety of settings.
Objectives
To determine the diagnostic accuracy of trans‐thoracic Doppler echocardiography for detecting pulmonary hypertension.
Secondary objectives
We will aim to study several possible source of heterogeneity as below.
-
Types of PH defined by WHO classification (group 3 or not). We expect that it will be more difficult to measure tricuspid regurgitation in patients with lung disease, especially COPD, than in patients without lung disease, which could decrease the sensitivity and specificity of echocardiography to detect PH.
-
Mechanical ventilation (including non‐invasive positive pressure ventilation) or not.
-
Estimation of right atrial pressure (whether estimated by using the diameter and collapse of the inferior vena cava during spontaneous respiration or by any other methods).
Methods
Criteria for considering studies for this review
Types of studies
We will include reports on the diagnostic accuracy of trans‐thoracic Doppler echocardiography for detecting pulmonary hypertension. We will consider diagnostic test accuracy studies (consecutive series or random sample) of Doppler echocardiography against right‐heart catheterization (Bossuyt 2008). We will exclude diagnostic case‐control studies (two‐gate design). Case‐control designs are known to overestimate the sensitivity and specificity that a diagnostic test has in clinical practice (Rutjes 2005). We will exclude studies if right‐heart catheterization was not the reference standard, and the reference standard threshold is different to 25 mmHg. We will exclude case studies that did not provide sufficient diagnostic test accuracy data (true positive (TP), false positive (FP), true negative (TN), and false negative (FN) values, based on the reference standard). We will include studies that provided data from which we could extract TP, FP, TN, and FN values, based on the reference standard. We will contact study authors for missing data.
Participants
We will include all adults (16 years of age or older) with suspected PH. We will exclude any participant already diagnosed as having PH.
We will not exclude participants based on sex or cause of PH.
Index tests
Measurement by Doppler trans‐thoracic echocardiography will be an index test. Systolic PAP will be calculated from the maximum tricuspid regurgitation jet velocity by using the modified Bernoulli equation and adding RAP (Berger 1985). We will include all methods of measuring RAP (e.g. clinical estimation from jugular venous pressure, using a fixed value from 5 mmHg to 10 mmHg, using the diameter and collapse of the inferior vena cava during spontaneous respiration).
Target conditions
The target condition will be pulmonary hypertension (PH) regardless of WHO classification group 1 to 5 (Simonneau 2013).
Reference standards
We will define pulmonary hypertension as when mean of PAP assessed by right‐heart catheterization is ≥ 25 mmHg (Hoeper 2013).
Search methods for identification of studies
Electronic searches
We will search the following databases:
-
MEDLINE Ovid SP (1946 to date);
-
Embase Ovid SP (1974 to date);
-
Web of Science Core Collection (1970 to date).
We will search the following trials registries:
-
US National Institutes of Health Ongoing Trials Register ClinicalTrials.gov (clinicaltrials.gov/);
-
World Health Organization International Clinical Trials Registry Platform (apps.who.int/trialsearch/).
A proposed MEDLINE search strategy is detailed in Appendix 2, which we will adapt for use in the other databases. We have combined search terms describing the target condition and the index text. We have not used search terms to describe diagnostic study designs as this is not currently recommended in the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy (De Vet 2008). We will not apply any restrictions on language or type of publication.
Searching other resources
To identify additional published, unpublished, and ongoing studies, we will enter relevant studies identified from the above sources into Web of Science and use the ‘Related Articles’ feature. We will check the reference lists of all primary studies and relevant systematic reviews and will contact authors of all included studies to identify other published, unpublished, or ongoing searches. We will search for conference proceedings through Embase and the Web of Science.
Data collection and analysis
Selection of studies
We will undertake the systematic review using the methods outlined in the Cochrane Handbook for Reviews of Diagnostic Test Accuracy (Deeks 2013). Two review authors (JK and SS) will independently review titles and abstracts identified by the search strategy. JK and SS will retrieve the full text of potentially relevant studies and will independently assess the full text against the eligibility criteria outlined in Criteria for considering studies for this review. Differences will be resolved by discussion between the review authors (JK and SS), with a further review author (YK) acting as arbiter. We will provide details of both included and excluded studies in the respective tables of the review.
Data extraction and management
Two review authors (JK and SS) will independently extract data on study characteristics, participant demographics, sample size, test methods, methodological quality, sensitivity, and specificity (Appendix 3; Appendix 4; Appendix 5). Both review authors will then extract data to construct a 2x2 contingency table. Disagreements will be resolved by consensus and with the further review author (YK) acting as arbiter.
Assessment of methodological quality
We will use the QUADAS‐2 tool to assess the quality of studies (Whiting 2011). We will record the assessment on a study quality assessment form (Appendix 6). The qualities to be assessed are described in detail in (Appendix 7). For each item in the quality assessment form, we will include a description of how the study addressed the issue and enter a judgement of 'low,' 'high,' or 'unclear' for overall risk of bias for each of the four domains. In addition, we will add a judgement of 'low,' 'high,' or 'unclear' for the overall concern of applicability to the review question for domains 1, 2, and 3. We will present an 'Assessment of methodological quality' table that will show all judgements made for all included studies. Two review authors (JK and SS) will independently assess methodological quality. Disagreements will be resolved by discussion between the review authors, with a further review author acting as an arbiter (YK).
Statistical analysis and data synthesis
We will perform data synthesis using the methods recommended by the working group of the Cochrane Collaboration on systematic reviews of diagnostic test accuracy (Deeks 2013). We will extract accuracy data for all thresholds used in the primary studies. We will represent individual studies' sensitivities and specificities in forest plots, sorted by sensitivity, in order to inspect between‐study variability. These individual studies' accuracy estimates will be also represented in a Receiver Operating Characteristic (ROC) plot of sensitivity versus 1‐specificity to visually assess the correlation between both indices. Given the lack of validated cut‐offs of the index tests (Galie 2009), we expect variability in cut‐off points chosen in the included studies. Therefore, if the included primary studies report accuracy data using different cut‐off values, we will meta‐analyze pairs of sensitivity and specificity using the hierarchical summary ROC (HSROC) model (Rutter 2001), which allows for the possibility of variation in threshold between studies, while also accounting for variation within and between studies and any potential correlation between sensitivity and specificity. If studies report data at multiple thresholds, we will use the threshold as primary analysis that was prespecified as primary endpoint in each study or was the nearest to 40 mmHg if not prespecified. If a sufficient number of primary studies report data using common cut‐off values, we will perform meta‐analyses using the bivariate model (Reitsma 2005). This model accounts for intrastudy accuracy variability and interstudy variations in test performance with the inclusion of random effects. We will analyze all studies sharing the same threshold at the same time and will obtain summary accuracy estimates. We will present these estimates with 95% confidence ellipses in the ROC space. We will use pooled estimates of sensitivity and specificity to calculate the positive and negative likelihood ratios and diagnostic odds ratio (Glas 2003). We will estimate explanatory variables for the bivariate model or HSROC model in order to analyze how these variables individually affect sensitivity and specificity. We will undertake all analyses using Review Manager 5 (Review Manager 2014), STATA software, version 13.0 (Stata 2013), or SAS software (SAS 2011).
Investigations of heterogeneity
To test whether either sensitivity or specificity, or both, differ in subgroups of studies we will use a bivariate model or HSROC model in which we will add explanatory variables representing the following subgroups.
-
Types of PH defined by WHO classification (group 3 or not). We expect that to measure tricuspid regurgitation will be more difficult in patients with lung disease, especially COPD, than in patients without lung disease, which could decrease the sensitivity and specificity of echocardiography to detect PH.
-
Mechanical ventilation (including non‐invasive positive pressure ventilation) or not.
-
Estimation of right atrial pressure (whether estimated by using the diameter and collapse of the inferior vena cava during spontaneous respiration or by any other methods).
Sensitivity analyses
We will examine the robustness of the meta‐analysis by conducting sensitivity analysis. We will check the impact of excluding from the meta‐analysis studies according to domains of the QUADAS‐2 assessment. We anticipate that studies at high risk of bias in domains 1 and 4 would have a great impact on meta‐analysis because high risk of bias in these domain would cause selection bias. Due to the invasiveness of the reference standard, less severe cases would not be verified by right‐heart catheterization, which would lead to partial verification bias. We will therefore perform additional sensitivity analysis by using prevalence as surrogate of partial verification, high prevalence or low prevalence, for which the median prevalence among included studies would be the cut‐off point.
Although uninterpretable results would be likely to occur among patients with overinflated lung, which is associated with the severity of pulmonary hypertension, it is not always considered test positive. We will therefore perform sensitivity analysis by including uninterpretable results as both test negative and positive.
Assessment of reporting bias
We do not plan to explore reporting bias due to a lack of suitable statistical methods (Deeks 2013).
Clinical pathway of diagnostic evaluation for pulmonary hypertension among adults, adolescents, and children.
PH: pulmonary hypertension; PAH: pulmonary arterial hypertension (Rubin 2016)
