Cardiorespiratory Dynamics During Prone Positioning for Severe Ards: A Retrospective Cohort Study

INTRODUCTION

Acute respiratory distress syndrome (ARDS) is a grave respiratory ailment characterized by a high mortality rate and significant comorbidity in survivors. It primarily involves increased permeability of the alveolar-capillary barrier, resulting in diminished oxygen exchange, reduced lung compliance, and pulmonary edema. The hospital mortality rate for ARDS varies between 35% and 45%, depending on disease severity, with more severe cases having higher mortality rates¹. Consequently, the timely recognition and appropriate treatment of ARDS are paramount in reducing mortality and enhancing clinical outcomes.

While patients with mild ARDS and adequate oxygenation can be managed conservatively, those with moderate to severe ARDS typically necessitate mechanical ventilation. In cases of severe ARDS where conventional treatments prove ineffective, clinicians may consider prone positioning as an alternative strategy. Prone positioning, initially shown to improve oxygenation in ARDS patients in the 1970s^2,3 and reduce mortality in a randomized trial⁴, is now frequently employed as a rescue therapy for severe ARDS patients unresponsive to traditional treatments. The use of this maneuver witnessed a surge during the COVID-19 pandemic⁵, as a substantial portion of COVID-19 patients hospitalized for acute respiratory failure progressed to moderate to severe ARDS, requiring mechanical ventilation. The rapid progression of the disease, coupled with its high prevalence, placed unprecedented strain on healthcare systems, leading to shortages of intensive care unit (ICU) capacity and mechanical ventilators. This scarcity of resources escalated the demand for alternative respiratory management strategies, such as prone positioning, to alleviate the burden.

In addition to its well-documented benefits in oxygenation^6,7, there is growing evidence supporting the mortality-reducing effects of prone positioning. A meta-analysis involving 1,867 patients reported a 16% reduction in mortality in patients with severe hypoxemia (PaO2/FiO2 < 100 mmHg) (risk ratio [RR] 0.84, 95% confidence interval 0.74 - 0.96, p = 0.01)⁸, and similar results have been reported in other meta-analyses^9–11. Furthermore, the PROSEVA trial demonstrated a reduction in 90-day mortality (hazard ratio 0.44, 95% CI 0.29 to 0.67) with early, prolonged prone positioning in mechanically ventilated patients with severe ARDS⁴.

While studies have extensively documented the physiological effects of prone positioning on oxygenation, such as improved lung perfusion¹² and reduced differences in ventral-dorsal transpulmonary pressure¹³, it is essential to recognize that the effects of this maneuver extend beyond the respiratory system. Prone positioning has been found to exert extrapulmonary hemodynamic effects, including an increase in cardiac output¹⁴ and improvements in right ventricular volume-pressure overload¹⁵. Paired with the observations that right ventricular dysfunction is both common in ARDS¹⁶ and associated with increased mortality¹⁷ the observed benefits of prone positioning may derive from a composite of both respiratory and non-respiratory effects.

In this study, we aim to investigate the cardiopulmonary characteristics of a cohort of patients who underwent prone positioning for severe ARDS. Specifically, our objectives are to 1) elucidate the changes in cardiorespiratory dynamics during neuromuscular blockade and prone positioning and 2) determine the associations between changes in cardiorespiratory dynamics following prone positioning and mortality.

METHODS

RESULTS

We initially included 190 patients but subsequently excluded 54: 19 did not have recorded cardiorespiratory dynamics measures, 5 received neuromuscular blockade prior to admission, and 30 had less than one hour between neuromuscular blockade and prone positioning (eFigure 1). Our final analysis cohort consisted of 136 patients (Table 1).

Figure 1 shows the unadjusted distribution of features for both neuromuscular blockade and prone positioning. In this figure, a plausible physiologic effect would be displayed as a smooth gradient of color from blue to red and greater odds (red) is associated with a greater chance than a value in that range was observed after the event of interest. The largest effect is in the oxygenation parameters (A-a gradient, S:F ratio, and P:F ratio) after prone positioning (Figure 1c) with little effect seen on circulatory parameters (Total norepinephrine dose, mean arterial pressure). This is consistent with a larger effect on respiration than circulation. For the whole cohort, prone position was associated with an improvement in A-a gradient of 113 mmHg (95% CrI 78 – 149) between the pre-proning values and 10 hours post proning (Figure 2). Norepinephrine dose, on the other hand, did not significantly change before and after prone positioning (Estimated difference: 0.04 mcg/min 95% CrI -1.00 – 1.07).

• Download figure
• Open in new tab

Figure 1

Changes in cardiorespiratory parameters after neuromuscular blockade and prone positioning. Odds ratio (color) of the post neuromuscular blockade (panels a and b) or prone positioning (panels c and d) state as a function of the quartile (panels b and d) or decile (panels a and c) for the parameters listed along the left. For example, the red color in the upper left box of panel C signifies that the lowest decile of A-a gradient was four times more likely to be after prone positioning than before prone positioning.

Note the top row in panel (c) shows monotonic relationship of A-a gradient with lower values having a high likelihood of being observed after proning (red colors) and high values have a low likelihood of being observed after proning (blue colors). In contrast, A-a gradient in panel (a) has an inconsistent trend.

• Download figure
• Open in new tab

Figure 2

The distributions of observed values for (a) A-a gradient and (b) norepinephrine equivalent dose for four hours before and 24 hours after prone positioning. The black line represents the median value while the blue and red shaded regions represent the 5^th to 95^th and 25^th to 75^th percentile values respectively. Observations were aligned to 15-minute boundaries and carried forward until a change was observed.

When we examine the combined effects (Figure 3), neuromuscular blockade was estimated to have a moderate effect on the coefficient of sample entropy, average instantaneous heart rate, and PCO2 while A-a gradient, detrended fluctuation analysis of the heart rate, PCO2, total norepinephrine dose, mean arterial blood pressure, coefficient of sample entropy, local dynamics score, and heart rate were estimated to have small effects. This is consistent with broad cardiovascular and respiratory effects of neuromuscular blockade. For prone positioning, only A-a gradient and PCO2 were estimated to change after prone positioning. This is consistent with the primary effect of prone positioning primarily on the respiratory system.

• Download figure
• Open in new tab

Figure 3

Estimated effects for cardiorespiratory predictors before and after (a) neuromuscular blockade and (b) prone positioning. The red line represents the MAP estimate from the posterior predictive and is omitted when the 95% highest density posterior interval contains the null effect. The grey band represents the 95% highest density posterior interval. The vertical lines represent the percentiles of observed values. Panel A shows that the effects of neuromuscular blockade are diverse with observed impact on ventilation (A-a gradient and pCO2), hemodynamics (norepinephrine equivalent dose and mean arterial pressure), and heart rate dynamics (coefficient of sample entropy and instantaneous heart rate). In contrast, Panel B shows that the dominant effect of prone positioning is on ventilation with modest effects on hemodynamics and nearly no effect on heart rate dynamics.

For the outcome of 7-d mortality, there was a high probability that the baseline factors of increasing age, male sex, and higher baseline A-a gradient were associated with increased risk of death (Table 2). Worsening shock as demonstrated by increased total vasopressor requirement was associated with worse prognosis after the first hour. In contrast, an increase in PCO2 was associated with worse prognosis at the first hour with decreasing effect seen over time. A decrease in instantaneous heart rate, corresponding to an increase in mean RR interval, and a decrease in heart rate variability, corresponding to a decrease in standard deviation of the RR interval, were associated with improved prognosis. A comparison of selected predictors and their association with the events of neuromuscular blockage and prone positioning are shown in Figure 4.

• Download figure
• Open in new tab

Figure 4

This graph shows that the effect of prone positioning and neuromuscular blockade (Panel A) for a subset of predictors and the association of those predictors with 7-day mortality (Panel B). In panel A, the mean value from the posterior predictive is shown in red and the line is omitted when the 95% credible interval includes no effect. The grey band represents the 95% credible interval for the posterior predictive. Note that many of the estimated effects are small with little variation over the range of commonly observed values (Z -2 to 2). However, the estimated A-a gradient and the instantaneous heart rate (Average RR interval) have large estimated effects for the events of prone positioning and neuromuscular blockade respectively.

In panel B, the adjusted association of selected predictors after prone positioning with 7-day mortality. The closed circle is the mean posterior parameter estimate and bars represent the 95% credible interval. The probability that the parameter is greater or less than zero is above the line. The first row shows the estimated effect of the baseline A-a gradient in a joint model with the changes at each time point. The estimated effect is consistent with a high probability of direction. In contrast, the second line shows the effect of the change in pCO2 with immediate increases associated with increased mortality (and immediate decreases associated with decreased mortality). The effect is smaller and less certain as time progresses after prone positioning.

DISCUSSION

In summary, the phenotype of improved prognosis after proning was associated with decreased vasopressor requirement, improved ventilation, decreased heart rate, and decreased heart rate variability. One explanation of these findings is improved right ventricular function resulting from improving oxygenation and ventilation due to prone positioning.

Our analysis supports previous evidence regarding the beneficial effects of prone positioning on oxygenation and ventilation. These improvements are consistent with the primary physiological effect of prone positioning on the respiratory system, which includes enhanced lung perfusion and decreased ventilation-perfusion mismatch. Additionally, our analysis did not show large immediate changes in cardiovascular parameters after prone positioning. The findings of this study shed light on the complex interplay of physiologic factors resulting from the use of prone positioning as a therapeutic strategy for severe ARDS.

Interestingly, neuromuscular blockade was associated with changes in several cardiorespiratory predictors, including A-a gradient, heart rate variability, and PCO2. This observation underscores the broader physiological effects of neuromuscular blockade beyond muscle relaxation. However, it is worth noting that only a few of these changes had a significant clinical impact, emphasizing the need for careful consideration of the risks and benefits of neuromuscular blockade in ARDS management.

Furthermore, our analysis of predictors of short-term mortality following prone positioning revealed several key insights. Increasing age and higher baseline A-a gradient were associated with an elevated risk of mortality consistent with previously published work. Worsening shock, as evidenced by increased total vasopressor requirements, was predictive of worse prognosis after the first hour post-proning. Conversely, an increase in PCO2 was initially associated with a worse prognosis, but its effect diminished over time. This suggests that the dynamics of carbon dioxide levels may have complex implications for ARDS outcomes.

The observation that a decrease in instantaneous heart rate and heart rate variability was associated with improved prognosis raises intriguing questions about the role of cardiac function in ARDS management. It is plausible that improved right ventricular function, resulting from enhanced oxygenation and ventilation due to prone positioning, contributes to the observed changes in heart rate parameters.

CONCLUSION

In conclusion, this study contributes valuable insights into the effects of prone positioning and neuromuscular blockade in severe ARDS management. The immediate changes in prone positioning primarily impact respiratory physiology, with limited influence on circulatory parameters. In contrast, neuromuscular blockade exhibited a more nuanced effect with broad effects on respiratory and cardiovascular physiology. Predictors of short-term mortality after prone positioning include both respiratory and cardiovascular parameters suggesting that extrapulmonary effects, such as improvement in right ventricular heart function, might also contribute to the benefit of prone positioning.

DATA AVAILABILITY

All data produced in the present study are available upon reasonable request to the authors.

CONTRIBUTIONS

Conceptualization: AJB, JRM

Methodology: AJB, SWL, JRM

Software: AJB, SWL

Validation: AJB, JRM

Formal analysis: AJB, SWL

Investigation: AJB, SWL

Resources: JRM

Data Curation: AJB, SWL

Writing - Original Draft: AJB, SWL

Writing - Review & Editing: AJB, SWL, JRM

Visualization: AJB

Supervision: AJB, JRM

Project administration: AJB

Funding acquisition: JRM

FUNDING

The work of AB was conducted with the support of the iTHRIV Scholars Program. The iTHRIV Scholars Program is supported in part by the National Center for Advancing Translational Sciences of the National Institutes of Health under Award Numbers UL1TR003015 and KL2TR003016 as well as by The University of Virginia. This content is solely the responsibility of the authors and does not necessarily represent the official views of NIH or UVA.

FIGURES

• Download figure
• Open in new tab

eFigure 1