The determinants of venous return from the peripheral circulation are, from Guyton’s experiments: 1.) the mean circulatory filling pressure (P_MCF), 2.) right atrial pressure (P_RA) and 3.) the resistance to venous return (R_VR) (Guyton, 1955; Sylvester et al., 1983; Jacobsohn et al., 1997; Feihl and Broccard, 2009a). Together the P_MCF and the P_RA define the pressure gradient for venous return.

In addition to detailing the peripheral vascular determinants of blood returning to the heart, Guyton expanded our understanding of hemodynamics by adding to his analysis the cardiac determinants of blood flow from the heart. He did so by superimposing the venous return and Starling-Sarnoff curves (Guyton, 1955); this depiction is commonly referred to as the ‘Guyton diagram.’ These curves can be placed over each other because they both have P_RA on the x-axis and blood flow on the y-axis (Figure 6). Though it will be developed in more detail below, the Guyton diagram introduces an important distinction between intravascular and transmural pressures. The P_RA and P_MSF measured on the Guyton diagram are intravascular pressures. Thus, the pressure gradient for venous return is directly related to the difference between the intravascular P_MSF and P_RA (Equation 1). The Starling mechanism, however, is related to right atrial transmural pressure which is the pressure within the right atrium less its ambient pressure (i.e., the pericardial pressure). The transmural right atrial pressure is a static pressure that determines cardiac myocyte stretch which servo-controls the ejected stroke volume to match the venous return inflow.

Like any model, the value of the Guyton diagram is that it makes explicit the system’s independent and dependent variables. Independent variables are those things the clinician can change or control (e.g., vascular volume and capacitance, airway pressure), whereas dependent variables are what the clinician wants to predict or study (e.g., cardiac output) by manipulating the independent variables. These distinctions are critical when considering the effects of any intervention in the ICU (e.g., prone position).

Nestled within the venous return curve are some of the independent variables of the circulatory system touched upon above: 1.) vascular capacitance, 2.) total vascular volume and 3.) the R_VR. Thus, increasing total vascular volume via intravenous fluids and/or decreasing vascular capacitance via alpha-agonists both augment the V_S and, therefore, P_MSF. On the Guyton diagram, raising P_MSF right-shifts the venous return curve such that there is increased blood flow to the heart for any given P_RA. Similarly, beta-agonists (Green, 1977) and/or shunting blood from long to short time-constant vascular beds decreases the R_VR (Caldini et al., 1974); this also enhances venous return for any given P_RA. On the Guyton diagram, diminished R_VR is manifested by an increased slope of the venous return curve (Figure 6). The converse also holds, diminished blood volume, increased capacitance and/or increased R_VR all reduce venous return for any given P_RA. One clinically-important scenario wherein vascular capacitance rises (i.e., which decreases P_MSF) is reduced adrenergic tone (e.g., sedation, anesthesia, relief of hypoxemia) (Bressack and Raffin, 1987).

Found within the cardiac function curve are additional independent variables: heart rate, rhythm, valve function, afterload, inotropic and lusitropic states (Feihl and Broccard, 2009a; Feihl and Broccard, 2009b). Consequently, rate and rhythm control (e.g., cardioversion), afterload reduction (e.g., vasodilator therapy, pulmonary vascular recruitment), enhanced contractility and improved relaxation (e.g., epinephrine infusion) all increase the slope of the Starling-Sarnoff curve. With this, blood flow from the heart is enhanced for any given P_RA. The converse also holds, for example, rapid atrial dysrhythmia coupled with torrential tricuspid regurgitation and severe pulmonary arterial hypertension decreases the slope of the cardiac function curve, that is to say, reduce cardiac output for any given P_RA (Figure 6).

But what about the P_RA itself? Is it an independent variable? In Guyton’s experimental work on venous return, P_RA was studied as an independent variable. However, on the Guyton diagram, which considers both venous return and cardiac function simultaneously, P_RA is not independent. This was clearly stated by Guyton in his initial proposal: “right atrial pressure is not one of the primary determinants of cardiac output but, instead, is itself determined simultaneously with cardiac output” (Guyton, 1955). Later, Fiehl and Broccard expanded upon P_RA as a dependent variable in their excellent review (Feihl and Broccard, 2009a). Accordingly, when analyzing venous return and cardiac function simultaneously, the dependent variable is the equilibrium formed at their intersection–the operating point. Thus, both the x- (i.e., P_RA) and y- (i.e., cardiac output) Cartesian coordinates are equally dependent upon the system. This may be counterintuitive given the convention of placing the independent variable on the x-axis, however, with the Guyton diagram this is a vestige of his initial work on venous return. When it is understood that the operating point is the dependent variable, the circular and specious reasoning that the concept of venous return is incorrect because ‘raising P_RA reduces venous return per Guyton but augments cardiac output by Starling’ becomes moot. Rather, at any given time (or in response to an intervention, such as the prone position) there are characteristics of the peripheral circulation and heart that, in tandem, produce a unique cardiac output and P_RA (Guyton, 1955). To clarify this, a modified Guyton model is proposed below to disclose the clinically-relevant independent variables.

While the proposed model is meant to illuminate the clinically-relevant independent variables determining Q_circulatory, equation 11 has important caveats (Magder, 2012). The most important is that it is predicated upon the intersection of two hypotenuses; in vivo, both the venous return and cardiac function curves have portions that flatten out. When the operating point falls upon the flat portion of the cardiac function curve, Q_circulatory depends only upon the independent variables of cardiac function: P_PC, the right atrial pressure at which the cardiac function curve begins to plateau, P_RAplat and the R_cardiac (Figure 9). Q c i r c u l a t o r y = P R A p l a t − P P C R c a r d i a c (12)

Fundamentally, this equation relays that Q_circulatory is no longer determined by peripheral factors when the operating point is above the P_RAplat. Changing P_MSF or R_VR only alter P_RA with fixed Q_circulatory.

In a manner similar to cardiac function, the venous return curve also flattens when the P_RA falls below venous collapse pressure, P_CRIT (Magder, 2012). This is the formation of a Starling resistor when the great veins enter the thorax and is observed with ultrasound as collapse of the great veins. This phenomenon is also termed ‘waterfall’ physiology because the pressure below P_CRIT has no bearing on flow, just as the height of a waterfall does not mediate its flow (Permutt and Riley, 1963). Consequently, when the operating point lies to the left of P_CRIT (i.e., on the “flat portion” of the venous return curve) Q_circulatory becomes independent of cardiac function or P_PC; the independent variables are P_CRIT, P_MSF and R_VR (Figure 9). Q c i r c u l a t o r y = P M S F − P C R I T R V R (13)

In other words, when venous limited, reducing R_cardiac (i.e., improving cardiac function) or changing P_PC has no bearing on Q_circulatory; only changing P_MSF, R_VR or P_CRIT might alter total flow.

Recently, Lai and colleagues studied the effect of prone position on the determinants of venous return (Lai et al., 2021). They measured P_MSF by extrapolating to zero flow a series of P_RA–cardiac output pairings in response to increasing airway pressure. Though this method overestimated P_MSF in a porcine model (Berger et al., 2016), this observation was restricted to euvolemic conditions which are less likely in ARDS patients in the ICU. Nevertheless, considering the discussion on P_MSF measurement above, a retrospective calculation of P_MSA would be of great interest given that the average P_MSF measured by Lai et al. was clinically quite high, especially in the prone position. Irrespective of absolute values, Lai and colleagues observed that P_MSF increased significantly from the semi-recumbent to prone position; they hypothesized that this was due to increased intra-abdominal pressure (IAP). However, the baseline value and change in IAP had no bearing on P_MSF behavior. This is unsurprising given what is known about the mechanisms by which PEEP increase P_MSF. Initially, it was also hypothesized that IAP mediated P_MSF augmentation with PEEP application and/or stiffening of the chest wall (i.e., akin to prone position) in early canine models (Scharf et al., 1977). However, IAP had no role in raising P_MSF, instead, adrenergic reflexes (i.e., changing vascular capacitance) and redistribution of blood volume from the central to peripheral circulation were the main drivers of P_MSF rise (Scharf and Ingram, 1977; Fessler, 1995; Fessler, 1997). Accordingly, central blood volume, adrenergic reserve and exogenous vasoactive agents all undoubtedly mediate the change in P_MSF upon pronation, rather than IAP. Parenthetically, this could also explain hemodynamic differences noted between elective surgical and critically-ill ARDS patients when prone position is employed (Edgcombe et al., 2008). The latter are more likely to be on vasoactive agents and volume-loaded, while the former more likely euvolemic; as well, anesthetic agents may blunt reflexive changes in vascular capacitance which would limit P_MSF rise in the operating room. As described above, P_MSF is directly related to Q_circulatory when the patient is not cardiac limited and without concurrent changes in P_PC, R_cardiac or R_VR.

There are no known direct measurements of P_PC in humans with ARDS placed in the prone position. Yet, inferences can be made given the mathematical approximation of P_PC presented above. The prone position increases the elastance (i.e., stiffness) of the chest wall (E_CW) (Pelosi et al., 1998). To the extent that pronation also decreases the elastance (i.e., improves compliance) of the lungs by alveolar recruitment, the E_CW relative to the elastance of the respiratory system (i.e., the lungs and the chest wall together, E_RS) rises. Multiplying the mean airway pressure generated by mechanical ventilation by the E_CW/E_RS ratio approximates P_PC augmentation when a patient is passive with the ventilator. For example, if the mean airway pressure is 10 mmHg with an E_CW/E_RS ratio of 0.3 in the supine position, then 3 mmHg is added to the P_PC. If mean airway pressure remains constant and prone position increases the E_CW/E_RS ratio to 0.5, then 5 mmHg is added to the P_PC.

Additionally, pericardial restraint could play an important role determining P_PC, especially if there is comorbid acute cor pulmonale (ACP). Typically, when right atrial volume is low (i.e., estimated by a transmural pressure below 5 mmHg (Hamilton et al., 1994)), there is little recoil pressure generated by the pericardium around it. As atrial volume increases beyond this, the pericardial sac is engaged and moves up its volume-pressure relationship. This leads to an increasingly large elastic recoil pressure from the pericardium, which raises the P_PC. Elevated P_PC, therefore, restricts right ventricular filling and ‘protects’ from overdistention; however, this blunts Q_circulatory by narrowing the P_MSF–P_PC gradient.

In the setting of ACP, often seen in moderate-to-severe ARDS (Guérin and Matthay, 2016; Mekontso Dessap et al., 2016), pericardial recoil may play an important role upon prone position. With ACP, co-existent right atrial distension elevates P_PC by pericardial recoil; this is especially true with P_RA above 10–12 mmHg (Hamilton et al., 1994). While prone position is expected to further increase P_PC (i.e., by increasing P_PL), to the extent that the elevated P_PL shrinks cardiac volume, P_PC may remain constant, or even fall, as the elastic recoil pressure imparted by the pericardium is reduced. More simply, the rising P_PL experienced by the pericardial space is offset by falling recoil pressure of the pericardium. This was originally observed in models of continuous positive airway pressure in heart failure (Huberfeld et al., 1995). Were this to occur upon prone position in a patient with ACP, P_PC would remain constant or fall. Taken with the effect of prone position on P_MSF noted above, the P_MSF–P_PC gradient would be maintained (or enhanced) and so too would Q_circulatory if R_cardiac and R_VR remain constant.

Finally, some have argued for the execution of prone position with thoracoabdominal supports that allow the abdomen to hang freely (Chiumello et al., 2006). These supports are typically placed mid-sternum and below the pelvis. Chiumello and colleagues compared these supports to the abdomen flush with the bed in prone ARDS patients (Chiumello et al., 2006). They found that the supports accentuated local pressure without any benefit to gas exchange while diminishing stroke volume. Given support placement directly at the sternum, it is possible that P_PC is accentuated, reducing the P_MSF–P_PC gradient and Q_circulatory barring a concomitant decrease in R_cardiac or R_VR.

While not a commonly-employed term within the sphere of clinical hemodynamics, ‘cardiac resistance’ (R_cardiac) is analogous to R_VR. Graphically and mathematically, R_cardiac is simply the inverse slope of the cardiac function curve. A decrease in R_cardiac (i.e., a steeper slope of the cardiac function curve) represents improved cardiac function and raises the operating point (i.e., Q_circulatory) unless the system is venous limited. In an elegant ultrasonographic study, Vieillard-Baron and colleagues illuminated the salubrious effects on the RV prompted by prone position (Vieillard-Baron et al., 2007). In 21 patients with P_aO₂/F_iO₂ ratio of less than 100 mmHg and ACP defined as RV enlargement and septal dyskinesia, 18 h of prone position led to a significant reduction in heart rate and increase in cardiac output. Furthermore, RV end-diastolic area fell while LV end-diastolic area increased and tricuspid regurgitation was reduced. Taken together, the rise in cardiac output with diminished RV size strongly implies reduced R_cardiac as a mechanism of improved Q_circulatory, at least in patients with ACP. The mechanism for this improvement (detailed at the outset of this review and by others (Repessé et al., 2016)) was reduced pulmonary vascular impedance to flow facilitating RV ejection (Jardin and Vieillard-Baron, 2003; Vieillard-Baron et al., 2007) which improves stroke volume and cardiac output for any given P_RA.

Recent studies also imply reduced R_cardiac. Ruste and colleagues investigated the hemodynamic effects of prone position in over 100 patients (Ruste et al., 2018). 25% of prone sessions led to significantly increased cardiac output, while 23% had a significant decrease; the remainder showed no change. Importantly, of those sessions where cardiac output rose, 56% had no change or a decrease in global end-diastolic volume (GEDV) measured by transpulmonary thermodilution. Rising cardiac output without an increase in end-diastolic volume infers reduced R_cardiac. Importantly, static GEDV with prone position could signify a shrinking RV end diastolic volume with enlarging LV end diastolic volume consistent with the reduced RV-to-LV end-diastolic area ratio observed with echocardiography by Vieillard-Baron et al. (Vieillard-Baron et al., 2007). Finally, Boesing and colleagues recently published on different PEEP titration strategies and their interaction with prone position (Boesing et al., 2022). In this study, esophageal pressure (P_ES) was used as a surrogate for P_PL. Curiously, the PEEP titration strategy that led to the greatest increase in cardiac output from supine to prone was associated with the smallest rise in transmural P_RA (i.e., P_RA less P_ES), in other words, the least preload augmentation. Similar to the observations by Ruste and colleagues, this finding suggests, but does not prove, enhanced cardiac function (i.e., reduced R_cardiac).

In the study of Lai and colleagues (Lai et al., 2021), the R_VR was calculated from semi-recumbent to prone position in ARDS patients. In total, R_VR increased in the vast majority, though there were a few with stable or slightly diminished R_VR. Like P_MSF, the change in R_VR was not related to IAP and like P_MSF, this is unsurprising given the foundational work of Takata and Robotham (Takata et al., 1990). In their original model, Takata and Robotham proposed that the relationship between great vein pressure and IAP would behave analogously to West zones in the lung. That is, if the IAP is much greater than inferior vena cava (IVC) pressure (i.e., zone 2), then venous return is impaired when the abdomen is pressurized by diaphragmatic descent and, in theory, prone position. However, if IAP is much less than IVC pressure (i.e., zone 3), then increased IAP generated by diaphragmatic descent (or prone position) enhances venous return. Their initial work confirmed this model, however, they later found that the model held even with an open abdomen and evisceration, that is, constant IAP (Takata and Robotham, 1992). Thus, the ambient pressure of import was more likely focal subcostal, crural, or intra-hepatic pressure, rather than general IAP. This was observed by Decramer and colleagues (Decramer et al., 1984) and explored further by Brienza et al. in a porcine model (Brienza et al., 1995) and Jellinek et al. in humans (Jellinek et al., 2000). Accordingly, diaphragmatic shape-matching between the liver and upper abdomen, active versus passive diaphragm displacement, intra-hepatic compliance (e.g., intrinsic liver disease) and the use of focal thoracoabdominal supports, among other factors might affect hepatic pressure (P_hepatic) upon pronation. Diminished venous pressure (e.g., hypovolemia, venodilation) relative to P_hepatic might increase R_VR. By contrast, elevated venous pressure (e.g., high blood volume, low venous capacitance) relative to P_hepatic might blunt a rise in R_VR with prone positioning.

Another possible mechanism for increased R_VR with prone position follows that of P_MSF. As described above, reflex sympathetic tone is a key mediator of increased P_MSF. However, when alpha agonists act upon veins to increase the V_S, resistance necessarily rises. This is because change in volume is proportional to the second power of vessel diameter but resistance is related to the fourth power. More concretely, if the diameter of a vein falls by 20% from its baseline, its volume is diminished by 36% (i.e., this reduces its capacitance, increases P_MSF) but its resistance rises by 244% (Rothe, 1993). Because the splanchnic circulation is a crucial reservoir for venous blood, the rise in resistance in response to V_S recruitment can be offset by beta-agonism (Green, 1977) in the hepatic veins, or redistribution of blood flow to short time constant vascular beds, as noted above (Magder, 2016) (Figure 4). Nevertheless, hepatosplanchnic blood flow during prone position in ARDS changes little (Hering et al., 2002; Matejovic et al., 2002). Interestingly, one study found decreased renal blood flow (Hering et al., 2001)—a fast time-constant bed; diversion of blood in this manner contributes to increased R_VR. A final, potential mechanism for R_VR augmentation with prone position lies in the superior vena cava (SVC). Fessler found that the rise in total R_VR following PEEP application was predominantly due to the veins draining into the SVC rather than the IVC (Fessler et al., 1992). Because P_PL is the pressure that surrounds SVC and prone tends to raise P_PL for any given P_AW (see equation 11 above), it is possible that mechanical compression of the SVC contributes to R_VR (Lansdorp et al., 2014; Berger et al., 2016). Regardless of the mechanism, R_VR is a critical determinant of Q_circulatory (Pinsky, 2021).

Taking the above into consideration, a key factor when predicting the hemodynamic response to prone position is the location of the operating point whilst semi-recumbent; is the operating point ‘cardiac limited’, ‘venous limited’ or ‘unlimited’ (Figure 6) (Magder, 2012)? Knowing this focuses the clinician on the independent variables most likely affecting Q_circulatory. For instance, if the operating point is cardiac limited (Figure 6) we see that changes in P_MSF and R_VR play no role, while changes in cardiac characteristics (e.g., R_cardiac) mediate Q_circulatory. Of course, this depends on how close the operating point is to the P_RA at which the cardiac function curve flattens out, but this is, nevertheless, a reasonable clinical heuristic. Jozwiak and colleagues studied 18 ARDS patients with elevated right ventricular-to-left ventricular end-diastolic areas (RVEDA/LVEDA), but without ACP (Jozwiak et al., 2013). Prior to prone position, the change in cardiac output in response to a passive leg raise was evaluated. By the model above, ‘cardiac limitation’ is detected when a patient is preload unresponsive. In this state, only improved cardiac function during pronation (i.e., reduced R_cardiac) increases Q_circulatory; changes in P_MSF and R_VR shift the operating point along the x-axis, but not the y-axis. In other words, P_RA changes but not blood flow. Jozwiak and colleagues found that in ‘cardiac limited’ patients, prone position significantly reduced pulmonary vascular resistance and the RVEDA/LVEDA which should diminish R_cardiac and improve Q_circulatory. However, these patients were also found to have depressed left ventricular ejection fraction. Furthermore, in the face of prone position, systemic afterload increased; total R_cardiac, therefore, did not improve.

When patients are not ‘cardiac limited,’ the operating point may be either ‘unlimited’ or ‘venous limited.’ In the study of Jozwiak and colleagues, imaging of the great veins was not reported, but those patients who were preload responsive were unlikely to have great vein collapse (i.e., venous ‘waterfall’) given that their average, baseline P_RA was relatively high (i.e., 15 mmHg) with increased RVEDA/LVEDA ratios. Thus, based on equation 11 above, the change in Q_circulatory was probably subject to all of: P_MSF, P_PC, R_cardiac and R_VR. Given what we know from Lai and colleagues, prone position likely increased P_MSF; P_PC may have increased less than the rise in P_PL because of reduced pericardial restraint and R_cardiac fell due to diminished pulmonary vascular resistance. Each of these effects raise Q_circulatory, presumably offsetting heightened R_VR with prone (Figure 10).

It is also possible for preload responsive patients to be ‘venous limited’ as described by equation 13 above. When the operating point lies on the flat portion of the venous return curve (i.e., below P_CRIT) then R_cardiac ceases to affect Q_circulatory. Said another way, blood flow is determined solely by peripheral venous factors. When ‘venous limited’, volume status is likely a crucial determinant of the hemodynamic response to prone position based on the model of Takata and Robotham described above (Takata et al., 1990). A zone 3 abdomen might have a stable or enhanced P_MSF relative to P_CRIT and blunt any increase in R_VR (i.e., stable or increased Q_circulatory), while a zone 2 abdomen would diminish P_MSF relative to P_CRIT and favour elevated R_VR (i.e., stable or reduced Q_circulatory). There is little data on ‘venous limited’ ARDS patients being placed in prone. In the study by Lai and colleagues, there were four ‘preload responsive’ patients who had no change (n = 3) or a decrease (n = 1) in Q_ciculatory when placed in prone position. These patients may have been venous limited, but this data was not collected. Given that at low trans-mural pressure, the great veins are very compliant (Bodson and Vieillard-Baron, 2012), generation of a hemodynamically-significant Starling resistor, i.e., ‘venous limitation,’ should lead to great vein collapse throughout most of the respiratory cycle. In a patient passive with the ventilator, collapse is an inspiratory event for the SVC and expiatory event for the IVC. Collecting this data with ultrasound before and after pronation could help delineate this hemodynamic phenotype.

Finally, it is possible to be both venous and cardiac limited simultaneously, in other words, the operating point is on both the flat portion of the venous return and cardiac function curves concurrently. This might happen in states of high P_CRIT (e.g., high PEEP, high subdiaphragmatic pressure) coupled with depressed cardiac function. In the setting of ARDS, this could be a syndrome of alveolar over-distension (Jardin and Vieillard-Baron, 2003). Prone position in such a patient might reduce Q_circulatory, especially if the patient is hypovolemic. Managing this hemodynamic phenotype might involve PEEP titration to reduce P_CRIT and enhance cardiac function as this could move the operating point onto steep portions of the venous return and cardiac function curves.