Prone positioning in conscious patients on medical wards: A review of the evidence and its relevance to patients with COVID-19 infection

COVID-19, respiratory failure, ARDS and CARDS

COVID-19 is a novel human infection and investigation of its pathophysiology is a rapidly evolving area of science. It causes a spectrum of disease with most hospitalised patients requiring admission for oxygen therapy (82% in one cohort).³ An acute respiratory failure may develop in those with pneumonia. This has been observed as marked hypoxaemia with a normal work of breathing in some.⁴ The mechanism for this is yet to be elucidated but is likely multifactorial. Lung compliance can remain normal in early disease. This may be due to oedema distributed to the interstitium rather than alveoli initially.⁵

In addition, there may be a failure of the normal hypoxic vasoconstriction mechanism secondary to endothelial dysfunction in areas of lung consolidation.⁶ This may also result in complement-mediated coagulopathy: formation of microvascular pulmonary thrombi which further contributes to ventilation/perfusion (V_a/Q) mismatch and worsening respiratory failure.⁷ Pulmonary infiltrates tended to be bilateral (79%), peripheral (54%) or in the right lower lobe (27%).⁸

Many patients meet the ‘Berlin criteria’ for ARDS (summarised in Table 1).⁹ The ratio of partial pressure of arterial oxygen (PaO₂) to fraction of inspired oxygen (FiO₂) is used as a marker for the degree of hypoxaemia. Compared to COVID-19 patients who did not develop ARDS, those who did were older, had a higher fever on admission and were more likely to report dyspnoea. Bilirubin, urea and d-dimer were significantly higher and lymphocyte counts significantly lower.¹⁰ However, the Berlin definition of ARDS is extremely broad and does not reflect any single underlying pathology.

Some have suggested labelling ARDS in COVID-19 as ‘atypical’ ARDS or CARDS (COVID-19 with ARDS).⁵ Gattinoni et al have hypothesised a model of either ‘L’ or ‘H’ phenotypes. ‘L’ is typified by low lung elastance (high compliance), low V_a/Q ratio, low lung weight and low recruitability seen in early disease. Some patients progress to ‘H’: high elastance (low compliance), high right to left shunt, high lung weight and high recruitability, reflecting more ‘typical’ ARDS.¹¹

This is still debated, given that progression from ‘L’ to ‘H’ phenotypes has not been documented clinically to date, and other data suggest that many ‘atypical’ features of CARDS (including preserved compliance) simply reflect existing heterogeneity in existing cohorts of ARDS patients.¹² In addition, many have cautioned against diverting from established ventilation strategies in ventilated patients.

Intubation and invasive mechanical ventilation (IMV) may be an option in those critically ill with COVID-19; however, it may not be deemed appropriate for many patients. This may be due to a less severe respiratory failure or inappropriateness of escalation to ICU-level care. As the pandemic continues to develop worldwide, IMV may simply not be an option for many in resource-poor settings. Other strategies that can be employed before critical deterioration are therefore attractive.

Oxygen therapy

The mainstay of therapy in COVID-19 infection is optimising oxygen delivery. In response to anticipated high oxygen demand within hospitals, National Institute for Health and Care Excellence (NICE) guidance has recommended an adjustment to normal target SpO₂ targets in all hospitalised patients (92–96% in first instance, 90–94% if high overall oxygen flow demands).¹³ On medical wards oxygen is delivered via ‘fixed’ devices (Venturi or similar), achieving an FiO₂ of 0.24–0.6, or ‘variable’ devices (facemask ± reservoir bag), achieving an FiO₂ of 0.6–1.0. Continuous positive airway pressure (CPAP) and high flow nasal cannula (HFNC) devices can be used to enhance oxygen delivery.

CPAP delivers a constant positive pressure to the airways via an airtight mask or hood. It aims to deliver flow above that of a patient's peak inspiratory flow, maintaining the end expiratory airway pressure at a prescribed level (positive end expiratory pressure or PEEP, measured in cmH₂O). This increases mean airway pressure, prevents collapse of alveoli at end expiration and improves oxygenation. NICE has recently stated that ‘CPAP is the preferred form of non-invasive ventilatory support in the management of the hypoxaemic COVID-19 patient. Its use does not replace IMV, but early application may provide a bridge to IMV’.¹⁴

HFNC circuits are becoming a more common sight on medical wards and are often initiated by critical care outreach teams. These deliver up to 60 L/min of humidified, warmed and oxygenated flow into the upper respiratory tract via modified nasal cannula. Humidification and warming increases tolerability, augments secretion clearance and prevents airway dehydration. An FiO₂ of 1.0 may be achieved via oxygen flushing of the upper airways. This decreases dead space ventilation and acts as an ‘anatomical oxygen reservoir’.¹⁵ PEEP delivery via HFNC is variable (3–5mmHg) and depends on factors such as cannula fit into the nostrils and degree of mouth versus nose breathing. Currently ‘the use of HFNO is not advocated in COVID-19 patients based on lack of efficacy, oxygen use and infection spread’.¹⁴

Given the above, CPAP is one potential therapy in those in whom simple oxygen therapy is not enough. However, CPAP (and HFNC) require equipment and specialist nursing and are categorised as aerosol generating procedures (AGPs),¹⁶ mandating use of appropriate personal protective equipment to minimise risk of airborne transmission to healthcare workers.¹⁷ In addition, application of PEEP to spontaneously breathing patients who exhibit respiratory distress may result in a phenomenon known as patient self-induced lung injury (P-SILI). Forceful inspiratory effort results in large transpulmonary pressure fluctuations (where the transpulmonary pressure is the difference between alveolar and pleural pressures). This exerts large shear forces in vulnerable lung tissue, increasing tissue stress and vessel permeability.^5,18

The prone position

Simply, this is when a patient is repositioned from the supine to lie on their front. An early report into using this technique to improve oxygenation in patients with acute respiratory failure included a single awake, non-intubated patient, where its use apparently ‘deferred intubation’.¹⁹ Since then the technique is largely utilised as a ‘rescue therapy’ in mechanically ventilated patients with severe ARDS. The PROSEVA (Prone Positioning in Severe Acute Respiratory Distress Syndrome) randomised controlled trial reported an impressive mortality benefit in IMV patients with severe ARDS when this technique was used alongside lung protective ventilation (23.6% versus 41.0% 90-day mortality, hazard ratio [HR] 0.44, 95% confidence interval [CI] 0.29–0.67).²⁰

A later Cochrane systematic review concluded there were mortality benefits in adult ventilated patients with ARDS, with pooled mortality from eight trials indicating a reduction from 47% to 42% (relative risk [RR] 0.90, 95% CI 0.82–0.98, quality of evidence ‘very low’). Treatment with prone positioning for >16 hours/day resulted in a difference in mortality versus non-prone of 47 versus 36% (five trials, RR 0.77 95% CI 0.61–0.99, quality of evidence moderate).²¹

Given the established benefits of prone position in these patients, some have suggested that the benefit could extend to those with early COVID-19 disease on medical wards.

The physiological viewpoint

When patients lie supine, ventilation of the lung is not even. The weight of the heart and mediastinum causes compression atelectasis of underlying lung (left lower lobe). Also, ventral (non-dependent) alveoli distend more readily versus the relatively compressed dorsal alveoli. This is due to both gravity and the ‘shape mismatch’ of the triangular lung expanding within a spherical chest cavity on inspiration.²² This results in heterogenous distribution of transpulmonary pressures (thus ventilation) across a ventral-dorsal axis.²³ The graduated non-cardiogenic pulmonary oedema in dependent regions seen in typical ARDS may exacerbate this.

Pulmonary perfusion across lung regions in the supine position is also heterogenous. It is greatest in dependent dorsal regions, which is partly due to gravity and partly to the effects of the vascular architecture in these regions.²⁴ PEEP exaggerates this gradient. Together these contribute to a V_a/Q mismatch which impacts on arterial oxygenation in the presence of ARF and ARDS.

When a patient is placed prone, this therefore has several effects. It reverses the compression atelectasis of the heart and mediastinum. It leads to more homogenous distribution of ventilation across the lung and minimises distending forces in ventral alveoli.²⁴ Perfusion persists to the dorsal region, with some evidence suggesting a greater capacity for hypoxic vasoconstriction in ventro-cranial regions. This may be due to higher production of nitric oxide (a potent vasodilator) in the endothelium of the dorsal lung which may modify the vasoconstrictive mechanism to hypoxia.²⁴ Thus prone positioning reduces the relative shunt fraction significantly (around 30%).²³ Improvements in PaO₂/FiO₂ of 34–62% have been documented in a review of observational data, with a variable temporal response (a trend towards an immediate response within 30 minutes followed by a continued response up to 24 hours).²⁴

In addition, drainage of secretions is also aided when prone, where material in the dorsal lung travels to open airways. This may explain a reduced incidence of ventilator-associated pneumonia (VAP) in the prone position.²⁵ Improvements in thoraco-abdominal compliance, particularly in patients with higher BMIs, are also observed by removing the detrimental effect weight of the chest wall.²⁴

Evidence for the prone position in IMV patients with COVID-19

In one single centre observational study from Wuhan, Pan et al utilised prone position in seven of 12 mechanically ventilated patients when the PaO₂/FiO₂ was persistently <20.0 kPa. The PaO₂/FiO₂ increased from 16±8.1 to 24.3±18.7 kPa in prone (p=0.065 but deemed ‘clinically relevant’). The authors used a novel index for measuring lung recruitability and reported significantly increased recruitability among those where prone position was used (p=0.02). However, three of these seven patients received extra corporeal membrane oxygenation (ECMO) support and outcome data were not provided.²⁶

More recently a cohort study from America has described the management of 66 IMV patients with COVID-19. In this study, 31 patients underwent prone positioning (in response to PaO₂/FiO₂ <26.6 kPa), with a mean of two ‘sessions’ over 72 hours. This improved PaO₂/FiO₂ and compliance from median values of 20 kPa and 33 ml/cmH₂O (interquartile ranges [IQR] of 125–183 and 24–46) to 31.1 kPa and 35ml/cmH₂O (IQR 167–265 and 34–47), which was similar to the response seen in non-COVID-19 ARDS patients.¹² Interestingly only 85% (n=56) met the Berlin criteria for ARDS despite intubation.

The Surviving Sepsis campaign advocates for a trial of prone positioning in mechanically ventilated patients who meet the moderate to severe ARDS definition.²⁷ They recommend periods of 12–16 hours, reflecting existing evidence for non-COVID ARDS.

The emergence of prone teams has been witnessed during this pandemic, with a heightened interest in utilising this technique for COVID-19 with ARDS. NHS England suggesting use of proning teams to ‘improve efficiency’ of instigating the technique.²⁸

Evidence for the prone position in conscious, non-ventilated COVID-19 patients

Given the above, it is plausible that the underlying mechanism leading to improvement in oxygenation is analogous between invasively ventilated ICU patients with severe ARDS and non-ventilated ward patients. A small number of reports have emerged and are summarised in Table 2.

Caputo et al, 2020.²⁹ Consecutive adult patients presenting to a single centre with hypoxia (SpO₂ <90%) who failed to respond to supplemental oxygen (SpO₂ to >93%) and who were able to self-prone were enrolled. 18 patients required intubation within 48 hours of admission (36% total). 12 participants were kept on nasal cannula pre-prone positioning, despite failure to achieve SpO₂ >90%.

Elharrar et al, 2020.³⁰ Inclusion criteria included COVID-19-positive status, presence of posterior lesions on chest CT and an oxygen requirement, and exclusion criteria included requirement of immediate intubation or reduced consciousness. Only 28% of 88 consecutive patients met these criteria. It is unclear if patients that responded were treated with HFNC and/or tolerated prone positioning for longest. After 10 day follow up five participants were intubated.

Sartini et al, 2020.³¹ Here an opportunistic cross-sectional survey was conducted on a single day. Patients with ARDS secondary to COVID-19 who were already undergoing prone positioning in combination with CPAP outside of the ICU were analysed. Improvement was sustained after returning to supine and stopping CPAP in 12 participants when measured at the 1-hour mark. However, a selection bias is implicit in the study design. Interestingly, around half of participants had a PaO₂/FiO₂ ≤70 mmHg prior to prone positioning, which appears critically low in spontaneously breathing patients a median of 5 days into treatment with prone + CPAP.

The Intensive Care Society have recently advocated for a trial of prone positioning in conscious patients with suspected/confirmed COVID-19 and FiO₂ of ≥28% to maintain SpO₂ 92–96% (or 88–92% in those at risk of hypercapnic respiratory failure) despite basic respiratory support. They suggest utilising cycles of 30 minutes to 2 hours, rotating through prone to right sided to supine to left sided positioning.³² Pre-pandemic data were referenced here.

Pre-pandemic evidence for conscious prone positioning

Several key trials conducted before the emergence of COVID-19 are summarised in Table 3.

Scaravilli et al, 2015.³³ 13 had a diagnosis of pneumonia and 5 had a pre-existing diagnosis of COPD. Any improvements in those who were maintained on facemasks only is not reported separately. Oxygen mask use fell on prone positioning from 56% pre prone positioning to 37% during prone positioning across all 43 prone episodes; the effect of HFNC/CPAP likely contributed to the difference seen.

Ding et al, 2020.³⁴ 20 patients with pneumonia (influenza in 45%) were diagnosed with ARDS on a trial of PEEP 5cmH₂O via CPAP/BiPAP (10 moderate, 10 severe). Patients then moved from HFNC, to HFNC + prone, to CPAP/BiPAP, to CPAP/BiPAP + prone to IMV if the target SpO₂ of >90% could not be maintained. 11 avoided intubation (success group), where the largest increase in PaO₂/FiO₂ was seen between HFNC to HFNC + prone (Table 3). Only one patient in the success group did not progress to requiring NIV based on their protocol, and a significant improvement in PaO₂/FiO₂ was also seen between HFNC + prone to NIV (131±38, 156±36, p=0.046). The retrospective separation of the success and failure overstates these benefits, although the success group did have significantly higher PaO₂/FiO₂ at baseline possibly indicating benefit in less severe disease.

Valter et al, 2003.³⁵ Four cases of hypoxaemic respiratory failure secondary to pneumonia are reported. BiPAP/CPAP were used pre-prone positioning in three patients, which was de-escalated in all three post prone positioning. One patient died in this group (of a cerebral infarct 18 days later).

Bellone and Basile, 2018.³⁶ Three cases of acute respiratory failure secondary to unspecified pneumonia were followed over 9 days. Prone positioning was started due to lack of response to a combination of HFNC and BiPAP over the first 48 hours of admission. A continued improvement to repeated episodes of prone positioning with HFNC was seen over the following 7 days and all patients were discharged from hospital.

Safety

Most safety data are of low quality, and almost exclusively derived from intubated patients. Here the pooled adverse event profile (seven studies, >7000 patients) indicated an RR of 1.10 (1.01 to 1.2), (quality of evidence deemed ‘very low’). Without the concern for dislodgement of tracheal tube in conscious, non-intubated patients, the main safety issues are development of pressure sores (face, sternum in particular) and loss of venous access.³⁷

There was a low rate of intolerance to the prone position (2/43 total prone procedures), with no other adverse events noted previously.³³ However, the safety profile of this technique on the general medical ward is unknown. The risk of aspiration in obtunded patients, for example, would have more severe consequences in those without a protected airway. Table 4 outlines contraindications for use of this technique in ward patients as outlined by the Intensive Care Society UK.³²