Bedside assessment of left atrial pressure in critical care: a multifaceted gem

As demonstrated in Fig. 1, there is variation throughout the cardiac cycle and the pressure at a specific time point has consequences for both incoming flow from the PV (downstream) into the LA and ongoing flow from the LA into the left ventricle (LV). It is quite difficult to express LV filling pressure (LVFP) as a single value on the LV and LA pressure tracing because the pressures fluctuate and LV filling is a complex process.

Mean LAP and LVEDP are not telling us the same thing yet are often used interchangeably. The LVEDP provides information about the LV operating compliance and is the closest estimate of LV preload as a surrogate for LVEDV. Patients with similar LVEDP can have markedly different LAP, which is determined by the operating compliance of the LA [9]. This concept is perhaps most relevant to critical care as changes to compliance can occur with fluid challenges and mechanical ventilation for example. The mean LAP integrates the atrial pressure tracing throughout systole and diastole providing a measure of the hemodynamic load determined by the LA operating compliance (and indirectly left ventricular operating compliance through atrioventricular coupling). It is the mean LAP that is reflected back to the pulmonary venous circulation impacting right ventricular performance [9, 10].

The ‘mid A wave pressure’ (mean value of the A‐wave amplitude) is recommended in consensus statements to estimate end-diastolic LAP that correlates most closely with LVEDP [11], whereas the mean LAP is obtained by temporal integration of the instantaneous PAOP over the entire cardiac cycle (Fig. 3). Mean LAP and end-diastolic LAP can differ significantly in the presence of large ‘V’ waves that occur in severe mitral regurgitation and with reduced LA compliance [12] (Fig. 3). Some suggest that the mean LAP as opposed to the end-diastolic LAP makes more sense when wanting to differentiate pre- from post-capillary pulmonary hypertension (PH) [9, 10]. Certainly, in the critically ill patient with hypoxic respiratory failure and RV dysfunction the more crucial question must be what the cumulative haemodynamic load on the pulmonary vascular system is. The answer to this lies with measurement of the mean LAP.

The impact of different PH haemodynamic subgroups on RV function is increasingly recognised [13]. A higher incidence of RV dysfunction and RV–pulmonary arterial uncoupling (measured by tricuspid annular planar systolic excursion (TAPSE)/systolic pulmonary artery systolic pressure (sPAP) ratio) was found in those with pre-capillary and combined pre- and post-capillary PH than in isolated post-capillary PH [14]. ePLAR (echocardiographic pulmonary-to-left atrial ratio using tricuspid regurgitant velocity and E/e′) appears to be a simple, non-invasive ratio in differentiating pre- and post-capillary PH with reasonable accuracy, albeit in non-critically ill cohorts [15] (Fig. 4). Patients with RV dysfunction coupled with a low/normal mean LAP and high pulmonary pressures may benefit from pulmonary vasodilators, e.g. nitric oxide. In contrast, those with a high mean LAP and isolated post-capillary PH may derive benefit from diuretics, and pulmonary vasodilators in this group may worsen pulmonary oedema [16]. These diverging treatment strategies emphasise the potential benefit of amalgamating LAP measurement into categorising RV–pulmonary circuit dysfunction. Further investigation of the feasibility and utility of ePLAR in critically ill patients with RV dysfunction would be of interest.

The challenges in correlating PAOP, LAP and LVEDP when using a PA catheter have been subject to intense evaluation in previously published works [17, 18] and are summarised in Table 1. Table 2 summarises non-critical care studies investigating the correlation between the PAOP and LVEDP during left heart catheterisation (LHC) showing varying results [19‐22]. Data comparing PAOP and LVEDP in critical care populations are scare and conflicting, and a tabulated summary is provided in Table 3 [23‐25]. In 1974, Lozman et al. evaluated five ventilated post-operative cardiac surgical patients without ARDS and showed that the relationship between PAOP and directly measured LAP was lost at PEEP levels above 15 cm H20 [23]. Jardin et al. demonstrated that below a PEEP of 10cmH20, PAOP correlated with invasively measured LVEDP; however, this correlation was diminished at PEEP values > 10 [24]. Teboul et al. have shown that PAOP correlated strongly with invasively measured post-A wave LVEDP in patients with ARDS with PEEPs up to 20 cm H20. They suggested this observed correlation of PAOP and LVEDP is due to surrounding diseased lung preventing alveolar vessel compression [25].

Investigation of LAP non-invasively using Doppler has been studied for over 30 years [26]. The most recent 2016 American Society of Echocardiography and the European Association of Cardiovascular Imaging (ASE/EACVI) guidelines estimate mean LAP through Doppler assessment of diastolic blood flow between the left atrium and left ventricle (mitral E to A wave ratio), tissue Doppler imaging (TDI) of the mitral annulus, the tricuspid regurgitant flow velocity and LA volumes as shown in Fig. 5 [27]. Importantly for the critical care physician who is interested in presence of raised LAP for treatment decisions, these guidelines began to differentiate between the two major objectives—LV diastolic dysfunction and LAP (Fig. 5).

The Euro-Filling study enrolled 159 patients in 9 centres, comparing non-invasive LAP measurements using ASE/ESCVI guidelines with invasive measurements of LVEDP. Of those with a normal non-invasive LAP, only 65% had a normal invasive LVEDP. Of those with an elevated non-invasive LAP, 79% had elevated invasive LVEDP. Overall, the sensitivity was 75% and specificity 74% giving a PPV of 39% and NPV 93% with an AUC of 0.78 [28]. In a similar study, 90 patients undergoing invasive cardiac catheterisation underwent TTE immediately prior to the procedure. The non-invasive LAP was accurate, when compared to the pre-A invasive measurement in 75% and inaccurate in 25%. In the latter group, the non-invasive LAP was overestimated in 8/20 and underestimated in 12/20 [29]. In these and other studies, the use of a single parameter as opposed to the guidelines use of 4 parameters was found to be highly inaccurate. Even the favoured one of E/e′ delivered only a moderate correlation to invasive measurements. Once again, these studies do not include the type of patients commonly found in the critical care setting who often have cardiac pathologies other than coronary artery disease. A tabulated summary is provided in Table 2 [28‐30].

It is of little surprise that pathophysiology unique to critical illness (mechanical ventilation, vasoactive agents, fluid shifts) can make application of these algorithms more challenging [31]. Brault et al. [32] who compared ASE/EACVI echo Doppler LAP algorithms to PAOP (measured at end expiration and averaged over five non-consecutive cardiac cycles) in 98 mechanically ventilated patients found a sensitivity and specificity of 74% for ASE/EACVI algorithms to predict elevated PAOP ≥ 18 mmHg. Agreement between echocardiography and PAOP was moderate (Cohen’s Kappa, 0.48; 95% CI, 0.39–0.70). Overall, the guidelines show better discriminatory performance in the critically ill than previous iterations as shown by Clancy et al. [33] and offer an unrivalled framework in our patient group. Table 3 provides a tabulated summary of critical care studies that have compared echo Doppler LAP to PAOP values in critically ill patients [32, 34‐40].

The ratio of early diastolic mitral inflow to average mitral annular tissue velocity (E/e′) has been most extensively studied in the cardiology population [28, 30] and has gained some interest in the critical care literature [32, 34, 36‐39]. E/e′ is less load dependent and can be used to assess for raised LAP in those with atrial fibrillation (AF), making it a favoured choice in critical care. A septal E/e′ of > 11, as well as lack of mitral E velocity beat to beat variation, are suggestive of raised LAP in AF [27].

As with any haemodynamic measurement, the use of a single parameter to evaluate LAP should be avoided, and E/e′ is no exception [19]. Although a normal E/e′ does not rule out high LAP, an E/e′ > 15 does have a high specificity in identifying a high LAP [41]. This is perhaps of greatest pragmatic benefit when decisions on further fluid resuscitation are needed at the bedside: an E/e′ > 15 in this scenario would strongly favour a patient with ‘fluid intolerance’. At the other extreme, a low lateral E/e′ of < 8 has shown good diagnostic accuracy to predict PAOP < 18 mmHg [34].

A further challenge of the algorithm to identify patients with high LAP in critical care is the inability of the LA to dilate acutely (in comparison to the right atrium) [42]. Critically ill patients can have acutely high LAP despite a normal LA size, for example, those with volume overload or sepsis and acute diastolic dysfunction [3]. In summary, a dilated LA (LA volume index (LAVI) ≥ 34mls/m²) should raise suspicion for raised LAP, but a normal LA size should not exclude raised LAP. Echocardiographic evaluation of the interatrial septal (IAS) kinetics throughout the respiratory cycle may add pertinent information. Patients with fixed bowing of the IAS to the right are more likely to have raised LAP [43]. Considering right atrial pressure is important however given it is the relative pressure difference between the atria that determines position of the interatrial septum. Additional parameters, including a reduced E wave deceleration time (< 160 ms), alterations in the pulmonary venous Doppler waveform such as a reduced contribution to left atrial filling during systole (S/D ratio < 1), reduced isovolumetric relaxation time (IVRT) of < 60 ms and a mitral ‘L’ wave of > 20 cm/sec, may be additive in identifying raised LAP. An appraisal of their merits and disadvantages is discussed by Nagueh et al. [27].

Overall, when it comes to LAP measurement there exists a lack of uniformity in methods and ‘what’ is being measured. These issues are further compounded by the heterogeneity of the populations included. We shouldn’t be too hasty however in abandoning LAP measurement at the bedside altogether. A non-invasive, rapid beside screening tool to identify patients with possible raised LAP could be ‘the rule of 8’s’: lateral E/e′ > 8 [34] and a lateral e’ ≤ 8 cm/s [32]. This tool could serve as a trigger to temporarily halt further fluid resuscitation and instigate multimodal assessment of cardiopulmonary performance as proposed in Fig. 6.

LA strain uses angle independent speckle tracking imaging to assess LA function and stiffness [44]. The increased participation of LA contraction to end-diastolic LV filling is increased in the presence of LV diastolic dysfunction, up to the point when LA is failing because of excessive LVEDP. Studies have demonstrated an inverse relationship between LA global strain and LV end-diastolic pressures [45]. LA strain should be measured using a non-foreshortened apical-4-chamber (A4C) view of the LA where values of LA strain for reservoir, conduit and pump functions are measured [46] (Fig. 7). Inoue et al. evaluated 322 patients referred for left or right heart catheterisation in a multicentre study [47]. Cut off values for LA reservoir strain of < 18% and LA pump strain of < 8% had an AUC of 0.76 and AUC 0.77, respectively, for detecting increased LVFP (defined as PAOP > 12 mmHg or LVEDP > 16 mmHg). LA strain didn’t perform well in predicting LVFP in those with AF. These values have been proposed to serve as substitute parameters for those with missing criteria in ASE/EACVI algorithms that would otherwise be classified as ‘indeterminate’ (Fig. 5), providing there are no exclusion criteria (AF, mitral valve disease, and left bundle branch block amongst others) [48].

There has been increasing interest in the utility of the relative left atrial volume change over the cardiac cycle to predict filling pressure. The hypothesis being that a smaller volume expansion of the LA between systole and diastole predicts higher LAP. The value, expressed as a percentage, is known as the left atrial expansion index (LAEI) and is calculated by the formula: (Volmax − Volmin) × 100%/Volmin, where Volmax = maximal LA volume and Volmin = minimal LA volume. Genovese et al. investigated its use in over six hundred patients with chronic cardiac disease [49]. A reasonable linear correlation was found between logarithmically transformed LAEI and PAOP (r = 0.73, p < 0.001). Whilst LA strain and LAEI have shown promise in the cardiology setting, prospective data are needed to assess their role in the critical care arena.

Patients may have pre-existing diastolic dysfunction or develop de novo diastolic dysfunction because of critical illness such as sepsis [4]. An important concept to appreciate is that the ‘diastolic stress test’ of critical illness can shift patients from a normal ‘resting’ LAP to a high LAP state, with corresponding increases in E/e′ ratio. This is because the mitral annular velocity (e’) of the stiff left ventricle cannot increase to match the increased mitral E velocity as occurs with increased cardiac output demand [27]. This can be particularly problematic during ventilatory weaning leading to Weaning-induced Pulmonary Oedema (WiPO). There is no validated cut off E/e′ value to predict WiPO, though higher values are associated with increased risk of weaning failure. The reader is directed to detailed review of this topic elsewhere [50, 51]. Repeated echocardiographic assessment with an approach as outlined in Fig. 6 could help identify those at risk of WiPO and help guide treatment strategies. For example, patients with elevated LAP may benefit from more aggressive diuresis, higher levels of PEEP and a planned extubation to non-invasive ventilation.

A PAOP ≥ 18 mmHg was a commonly accepted criterion to define cardiogenic oedema in ARDS [52]. However, it was not a ‘hard’ value, was seldom measured and it was increasingly appreciated that raised LVFP could coexist with ARDS, hence it was removed from revised diagnostic criteria [53]. Authors of the guidelines highlight the ongoing complexities in differentiating cardiogenic from non-cardiogenic pulmonary oedema and describe scenarios [53].

An elderly patient with chronic obstructive lung disease and congestive cardiac failure, with a central venous pressure of 15 mmHg and fulfilling ARDS criteria, is described as probably having an overlap of cardiogenic and non-cardiogenic pulmonary oedema [53]. In contrast, a multi-trauma patient fulfilling ARDS criteria with a small, hyperdynamic LV without pericardial effusion is likely to have non-cardiogenic pulmonary oedema ARDS. The latter scenario highlights the benefit of incorporating echocardiography, however, it describes the extreme ends of both echocardiographic and clinical spectrums where treatment decisions are often easier.

Unfortunately, many of our patients, like the elderly patient described with pre-existing cardiorespiratory comorbidity, fall into a ‘grey zone’ and there is frequent overlap of cardiogenic and non-cardiogenic aetiologies [54]. Ray et al. studied over 500 elderly patients presenting to the emergency department with acute respiratory failure and showed that those with cardiogenic pulmonary oedema had the highest mortality at 21%. Importantly, around one third of the total cohort was deemed to have inappropriate early treatment and this was associated with a doubling of in hospital mortality [54]. Early detection of raised LAP and lung oedema to prevent inappropriate therapy therefore is a key goal.

Lung ultrasound (US) assessment of B lines is relatively quick and can be used to identify pulmonary oedema [55, 56]. B lines can be seen in other non-cardiogenic lung oedema states particularly relevant to our population (interstitial syndrome of ARDS, pulmonary fibrosis) [56], hence the need to contextualise echocardiographic and clinical findings [57]. B line quantification methods have shown good diagnostic accuracy against extra vascular lung water impedance techniques in critically ill patients [58, 59]. A simplified 4-sector method described by Mayr et al. only just underperformed against the more laborious 28-sector method, and a cut off value of ≥ 15 B lines resulted in a sensitivity of 91.7% and specificity of 92.1% to identify patients with increased extravascular lung water (AUC 0.978) [59]. Counting the number of B lines can be difficult in those with coalesced lines and lung US scores evaluating the percentage of B lines occupying the pleural line may be better, however, the time required for post processing and offline analysis limits its application at the bedside for most users at this time [58]. Furthermore, in the shocked patient with echocardiographic LAP parameters falling into the ‘grey zone’, the finding of a predominant A line pattern, that is highly specific for a low/normal PAOP, can increase confidence that repeating a fluid challenge is unlikely to result in pulmonary oedema [60].

Time critical decisions on haemodynamic resuscitation often centre on ‘fluid tolerance versus intolerance’, as opposed to definitive diagnosis, and we often need rapid, yet rich haemodynamic information. Tempered with an awareness of the caveats, trends in LAP coupled to the upstream hydrostatic consequence using lung US could provide this information (proposed in Fig. 6).