Model-based PEEP optimisation in mechanical ventilation.

BACKGROUND: Acute Respiratory Distress Syndrome (ARDS) patients require mechanical ventilation (MV) for breathing support. Patient-specific PEEP is encouraged for treating different patients but there is no well established method in optimal PEEP selection.
METHODS: A study of 10 patients diagnosed with ALI/ARDS whom underwent recruitment manoeuvre is carried out. Airway pressure and flow data are used to identify patient-specific constant lung elastance (E lung) and time-variant dynamic lung elastance (E drs) at each PEEP level (increments of 5 cm H2O), for a single compartment linear lung model using integral-based methods. Optimal PEEP is estimated using E lung versus PEEP, Edrs-Pressure curve and E drs Area at minimum elastance (maximum compliance) and the inflection of the curves (diminishing return). Results are compared to clinically selected PEEP values. The trials and use of the data were approved by the New Zealand South Island Regional Ethics Committee.
RESULTS: Median absolute percentage fitting error to the data when estimating time-variant E drs is 0.9% (IQR = 0.5-2.4) and 5.6% [IQR: 1.8-11.3] when estimating constant E lung. Both E lung and E drs decrease with PEEP to a minimum, before rising, and indicating potential over-inflation. Median E drs over all patients across all PEEP values was 32.2 cmH2O/l [IQR: 26.1-46.6], reflecting the heterogeneity of ALI/ARDS patients, and their response to PEEP, that complicates standard approaches to PEEP selection. All E drs-Pressure curves have a clear inflection point before minimum E drs, making PEEP selection straightforward. Model-based selected PEEP using the proposed metrics were higher than clinically selected values in 7/10 cases.
CONCLUSION: Continuous monitoring of the patient-specific E lung and E drs and minimally invasive PEEP titration provide a unique, patient-specific and physiologically relevant metric to optimize PEEP selection with minimal disruption of MV therapy.

1 Introductions

Acute respiratory distress syndrome (ARDS) and acute lung injury (ALI), occurs due to severe inflammatory response of the lung, resulting in direct alveolar injury, pulmonary oedema and alveolar collapse [1,2]. The lung injury greatly impairs the patients breathing, reducing alveolar gas exchange, resulting in possible mortality and morbidity if not given a proper treatment. ALI/ARDS patients are associated with high morbidity, mortality up to 60% [3] and significant medical cost [4].

Patients diagnosed with ALI/ARDS are mechanically ventilated for breathing support [5,6]. Various mechanical ventilation (MV) modes have been introduced to clinicians for the support of patients with ALI/ARDS [7]. However, the fundamentals of MV remains in selecting an optimal positive end-expiratory pressure (PEEP) to maximise patients' lung recruitment, prevent alveoli collapse, and avoid ventilator induced lung injury (VILI) [8]. The heterogeneity of the disease and patients' variable response to MV, encourages PEEP treatment to be patient-specific and individualised. However, there is no gold standard method in PEEP selection; consequently, optimising patient-specific PEEP in MV remains a challenge for clinicians [9-11].

Model-based and patient-specific approaches offer the ability to identify intra- and inter-patients variability and thus, potential to guide MV therapy based on patient's condition and needs [12,13]. This approach provides the opportunity to balance risk of lung injury and lung function support and reduce work of breathing [14] during MV. However, to date, only a few have been tested [15-17] and their potential in critical care is not yet validated.

This research presents several model-based approaches to identify patient-specific disease state and patient-specific response to MV therapy using patient-specific, constant lung elastance (E) [16,18] with comparison of dynamic lung elastance (E) in ALI/ARDS. Dynamic lung elastance (E) is a time-variant lung elastance during each breath in MV. Eand Eare thus proposed for guiding PEEP selection. By monitoring both the identified parameters (Elastance = 1/Compliance) through limited PEEP titration, it is possible to identify PEEP settings that maximize recruitment, minimize work of breathing without inducing lung injury.

2 Methods

2.1 Study Design

Ten patients in the Intensive Care Unit (ICU), Christchurch Hospital, New Zealand, diagnosed with ALI or ARDS (PaO(PF ratio) between 150-300 mmHg), underwent a modified protocol-based recruitment manoeuvre (RM) [17]. PEEP is increased with increments of 5cmHfrom zero PEEP (ZEEP) until peak airway pressure reaches a limit of 45cmH[19]. Patients were sedated and paralyzed with muscle relaxants to prevent spontaneous breathing efforts. All patients were ventilated using Puritan Bennett PB840 ventilators (Covidien, Boulder, CO, USA) with volume control (tidal volume, V= 400~600ml), synchronized intermittent mandatory ventilation (SIMV) mode, throughout the trial. The clinical trials and the use of the data were approved by the New Zealand, South Island Regional Ethics Committee.

A heated-pneumotachometer with Hamilton Medical flow sensor (Hamilton Medical, Switzerland) connected to the ventilator circuit Y-piece is used to record patient's airway pressure and flow data. A Dell™ (Dell, Austin, TX, USA) laptop was used in conjunction with National Instruments USB6009 and Labview Signal Express (National Instruments, Austin, TX, USA) to obtain measurements at a sampling rate of 100 Hz. Analysis was performed using MATLAB (The Mathworks, Natick, Massachusetts, USA).

2.2 Model-based Analysis

The model-based approach incorporates a physiologically relevant and validated recruitment model [17,20] with the use of a single compartment linear lung model that captures fundamental lung mechanics and properties in real-time to identify patient-specific constant lung elastance (E) and dynamic lung elastance (E) during MV. The model uses transpulmonary pressure (P), volume (V) and flow (Q) and offset pressure (P), to identify lung elastance (E) and resistance (R). Patient-specific lung elastance, Ereflects the lung stiffness (1/Compliance). Therefore, a lower Eis a more compliant lung. Eis identified from measured data using an integral-based method [21]. The model is defined:

Airway pressure is related to transpulmonary pressure (P) and pleural pressure (P) by:

When the patient is sedated and fully dependant on the ventilator to breathe, it can be assumed that there is no chest wall activity, allowing Pto be omitted in this case. Equation 1 is then further modified to eliminate P, yielding:

Patient-specific dynamic lung elastance, E, is identified as a time-variant lung elastance and Equation (3) is defined:

To ensure that the identified parameters of constant Eand time-variant E(E) are valid, the absolute percentage error between the identified model and measured clinical pressure data is reported.

2.3 Model-Based PEEP Selection

During each breathing cycle, as PEEP rises, lung elastance (E) falls as new lung volume is recruited faster than the pressure build-ups in the lung. If little or no recruitment occurs, Erises with PEEP indicating that pressure above that PEEP level was unable to recruit significant new lung volume and is, instead, beginning to stretch already recruited lung [22]. Hence, recruitment and potential lung injury can be balanced by selecting PEEP at minimum E.

Compared to a single, constant Evalue at each PEEP, identifying time-variant Eallows this change to be seen dynamically within each breath as pressure increases thus allowing a more detail view of patient's lung physiological condition. Three model-based approaches based on patient-specific Eand Etrajectory in a patient's breath at different PEEP levels are used to optimize PEEP selection.

Minimum : locates the point where minimum Eoccurs over all PEEP values (and pressure for E) during the recruitment manoeuvre.

Minimum : EArea is obtained by integrating Eover time during the patient's breathing cycle at each PEEP. EArea is more clinically relevant than median or mean Ethroughout each breath and can be shown to be proportional to patient-specific work of breathing.

Inflection Method: This method detects the inflection in the EArea-PEEP and E-PEEP curves. Inflection is defined here at the PEEP value with Evalue 5-10% above (before) minimum EArea or E(105~110% of minimum EArea or E). PEEP is selected where inflection occurs, as a point of diminishing returns.

The overall approach implies that as long as Efalls during each breath, as PEEP level increases, that recruitment of new volume outweighs lung stretching as flow and volume follow a path of lesser or least resistance. These methods are thus attempts to maximize recruitment (Minimum Eand Minimum EArea) and also ensure safety from excessive pressure (Inflection Method). These metrics are three of many possibilities to demonstrate the concept.

2.4 Edrs Area and Work of Breathing

These approaches were also compared with selecting PEEP using the identified minimum or inflection of constant E, for comparison to other similar work [23]. Patient-specific Eand Eare only analyzed during inspiration and not during the expiratory cycle. This choice was made because increases in pressure induce lung damage as it passes a limit and thus expiration (decreasing pressure) should not be used to guide PEEP selection.

A higher resolution of the trend changes in Ecan be observed using EArea. EArea is obtained through integration of Ewith time. It is also known that the work of breathing (WOB) [24,25] for a patient is proportional to lung elastance. In general, more work is required to fill a given lung volume with higher elastance. WOB is defined:

Substituting Pfrom Equation (3) into Equation (5) and using P= 0, (atmospheric).

From Equation (6), work of breathing can be divided into work to overcome lung elastance (WOB= E) and work to overcome airway resistance (WOB= R). Substitution of dynamic lung elastance, E, for constant Eenables a derivation for WOB:

EArea in Equation (8) is the integral of Equation (7), yielding the relation of Eto the work of breathing required to overcome lung elastance at a given level of PEEP and mode of MV.

2.5 Analysis and Comparisons

In this study, Eand median Eare compared using Pearson's linear correlation coefficients to relate these metrics. Eand EArea are also compared to median Eand WOBto ensure there was no loss of information for each patient at different PEEP values, and to show the validity of Equation (7) and using EArea. Finally, clinically selected PEEP is compared to the value determined by proposed model-based metrics.

3 Results

Table 1 shows the clinical details of the 10 patients recruited with their clinical diagnostics, and PF ratios. Table 2 shows the median [Inter-quartile Range (IQR)] Efor each patient and PEEP, and absolute percentage fitting error. Median absolute percentage fitting error (APEEdrs(t)) across all patients and PEEP is 0.9% [IQR: 0.5-2.4]. Median Eat each PEEP is 32.2cmH[IQR: 26.1-46.6]. Median [IQR] Edecreases with increasing PEEP until the minimum E. Patients who suffer from COPD (Patients 1, 4, 5, 9 and 10) have significantly higher Ethan others (P < 0.0001), as expected clinically. Table 3 shows the constant lung elastance (E) at each PEEP with median = 32.2cmH[IQR: 25.0-45.9], and absolute percentage fitting (APEElung) at 5.6% [IQR: 1.8-11.3]. Table 4 shows the EArea at each PEEP with median [IQR] of 34.0cmH[IQR: 24.7-48.5].

Table 1

Patient demography.

Patients	Sex	Age (year)	Clinical Diagnostic	PF Ratio
1	F	61	Peritonitis, COPD	214
2	M	22	Trauma	180
3	M	55	Aspiration	222
4	M	88	Pneumonia, COPD	165
5	M	59	Pneumonia, COPD	285
6	M	69	Trauma	280
7	M	56	Legionnaires	265
8	F	54	Aspiration	302
9	M	37	H1N1, COPD*	182
10	M	56	Legionnaires, COPD	237

*Chronic Obstructive Pulmonary Disease

Table 2

Patient-specific dynamic lung elastance (E) at each PEEP level.

Patient	Dynamic Lung Elastance, Edrs (cmH2O/l) Median [IQR]							Edrs (cmH2O/l) Median [IQR]	APE* (%) Median [IQR]
	PEEP (cmH2O)
	0	5	10	15	20	25	30
1	63.1[46.9-114.9]	53.8[43.0-80.2]	43.6[38.4-54.5]	35.0[33.3-39.4]	33.4[32.0-34.2]	31.1[32.0-32.4]	PEEP 2732.2[31.9-32.6]	35.0[32.5-51.2]	1.1[0.5-4.1]
2	30.8[26.3-45.1]	26.4[23.7-31.4]	23.1[22.0-24.3]	22.1[22.0-22.6]	22.5[22.4-22.6]	PEEP 2223.1[22.9-23.2]		23.1[22.5-26.4]	0.7[0.6-2.4]
3	26.9[22.6-36.9]	22.1[20.2-25.6]	18.3[18.0-19.0]	17.3[17.2-17.4]	17.5[17.1-17.5]	17.8[17.4-18.7]	PEEP 2819.2[17.9-19.7]	18.3[17.6-21.4]	0.6[0.5-1.3]
4	73.2[50.4-144.4]	70.4[49.9-126.9]	54.5[41.7-82.3]	36.8[30.6-43.9]	28.5[25.6-31.4]	25.9[21.6-28.4]	23.1[19.4-25.5]	36.8[26.6-66.4]	3.4[0.9-5.4]
5	105.7[80.6-199.8]	97.8[77.5-166.8]	89.3[74.3-143.4]	79.4[68.6-107.3]	67.3[61.4-79.4]	52.3[52.0-55.8]		84.4[67.3-97.8]	3.2[0.9-6.0]
6	30.4[25.9-39.1]	26.2[25.5-27.2]	23.3[22.4-23.5]	21.6[21.5-21.8]	21.8[21.3-22.5]	23.3[22.6-23.9]		23.3[21.8-26.2]	0.8[0.6-1.2]
7	49.3[46.1-62.4]	42.2[41.5-43.1]	44.3[41.8-47.7]	53.6[48.8-59.7]	PEEP 1652.4[50.3-57.6]			49.3[43.8-52.7]	1.6[1.3-2.0]
8	45.7[37.9-67.8]	37.2[32.9-43.0]	31.8[29.9-33.5]	28.8[28.0-29.8]	27.4[27.1-27.9]	26.8[26.3-27.0]	27.0[26.8-27.5]	28.8[27.1-35.9]	0.8[0.5-2.2]
9	58.1[47.1-100.8]	40.5[36.4-52.8]	39.9[35.8-48.7]	31.2[30.2-33.6]	28.3[27.9-29.0]	26.3[26.3-26.5]	26.2[25.8-26.5]	31.2[26.8-40.4]	0.8[0.4-2.1]
10	54.4[48.1-76.2]	45.2[41.9-51.8]	39.4[38.4-41.7]	35.9[35.7-36.0]	33.9[33.7-34.1]	33.9[33.4-34.6]	PEEP 2733.9[33.2-34.8]	35.9[33.9-43.8]	0.4[0.4-0.9]
Median[IQR]	51.9[30.8-63.1]	41.4[26.4-53.8]	39.7[23.3-44.3]	33.1[22.1-36.8]	28.4*[22.5-33.9]	26.3*[23.1-32.2]	26.6*[23.1-32.2]	32.2[26.1-46.6]	0.9[0.5-2.4]

*APE - Absolute Percentage Fitting Error (%)

*Values presented include value from different PEEP.

Table 3

Patient-specific constant lung elastance (E) at different PEEP.

Patient	Constant Lung Elastance, Elung (cmH2O/l)							Elung (cmH2O/l) Median [IQR]	APE (%) Median [IQR]
	PEEP (cmH2O)
	0	5	10	15	20	25	30
1	53.8	47.0	41.2	32.8	32.8	32.1	PEEP 2732.2	34.7[32.4-45.5]	7.2[1.7-19.0]
2	27.7	25.3	22.8	22.3	22.6	PEEP 2223.1		23.0[22.6-25.3]	2.5[1.1-7.7]
3	24.0	21.6	18.3	17.3	17.4	18.1	PEEP 2819.1	18.3[17.6-20.9]	4.2[1.6-6.6]
4	60.2	59.7	50.1	35.1	27.8	25.3	22.5	35.1[25.9-57.3]	17.7[15.4-32.1]
5	87.4	84.0	81.2	74.3	65.7	53.1		77.8[65.7-84.0]	15.7[9.2-19.8]
6	27.1	25.5	22.8	21.6	21.8	23.4		23.1[21.8-25.5]	2.7[2.2-4.2]
7	47.7	42.5	45.5	55.7	PEEP 1655.3			47.7[44.8-55.4]	6.2[5.0-7.7]
8	41.7	35.5	31.2	28.7	27.5	26.6	27.0	28.7[27.2-34.4]	2.9[1.3-8.7]
9	51.3	39.1	38.2	31.1	28.2	26.2	26.1	29.7[26.2-38.7]	3.1[1.0-10.8]
10	51.0	44.1	39.2	35.8	33.9	34.0	PEEP 2734.2	35.8[34.1-42.9]	2.0[1.0-5.6]
Median [IQR]	49.4[27.7-53.8]	40.8[25.5-47.0]	38.7[22.8-45.5]	31.9[22.3-35.5]	28.0*[22.6-33.9]	26.2*[23.3-32.6]	26.6*[22.5-32.2]	32.2[25.0-45.9]	5.6[1.8-11.3]

Eg. PEEP 16 is included in PEEP 20 Median [IQR]

*Values presented include value from different PEEP. Eg. PEEP 16 is included in PEEP 20 Median [IQR]

Table 4

Patient-specific EArea at different PEEP.

Patient	Edrs Area (mH2Os/l)							Edrs Area (cmH2Os/l)Median [IQR]
	PEEP (cmH2O)
	0	5	10	15	20	25	30
1	84.6	49.5	37.1	28.9	26.6	25.7	PEEP 2725.7	28.9[25.9-46.4]
2	34.0	24.8	21.0	20.2	20.3	PEEP 2220.7		20.9[20.3-24.8]
3	37.7	27.6	22.2	20.8	19.1	19.7	PEEP 2818.9	20.8[19.3-26.3]
4	102.2	91.2	61.7	37.9	31.7	48.1	47.5	48.1[40.3-83.8]
5	118.7	99.9	89.1	70.6	75.7	42.9		82.4[70.6-99.9]
6	29.4	23.8	20.8	21.6	19.5	20.8		21.2[20.8-23.8]
7	37.6	33.8	31.3	37.9	PEEP 1632.1			33.8[31.9-37.7]
8	55.1	38.5	32.0	29.0	27.5	24.1	24.3	29.0[25.1-36.9]
9	106.5	55.2	51.3	38.3	34.1	31.6	31.3	38.4[32.2-54.2]
10	74.7	52.6	44.0	39.5	37.3	37.2	PEEP 2737.3	39.5[37.3-50.5]
Median [IQR]	64.9[37.6-102.2]	44.0[27.6-55.2]	34.6[22.2-51.3]	33.5[21.6-38.4]	29.6*[20.3-34.1]	25.7*[20.8-38.6]	28.5*[24.3-37.3]	34.0[24.7-48.5]

*Values presented include value from different PEEP. Eg. PEEP 16 is included in PEEP 20 Median [IQR]

Figure 1 shows patient-specific time-varying Eat each PEEP level for Patients 2, 6, 8 and 10. Edecreases as pressure increases at each PEEP. However, at higher PEEP, this trend can reverse indicating stretching exceeding recruitment of new lung volume. The optimal PEEP derived by minimum Eis indicated.

Figure 1

Dynamic lung elastance (. Top Left Panel: Patient 2, Top Right Panel: Patient 6. Both patients show significant Edrop from lower zero PEEP to PEEP 15cmH. Further increase of PEEP to 20cmHshows increase of overall E. Bottom Left Panel: Patient 8, Bottom Right Panel: Patient 10. Both patients show a consistent drop in overall Ewith increasing of PEEP and overall Edid not rise with PEEP for the entire ranged considered.

Figure 2 shows patient-specific EArea for Patients 2, 6, 8 and 10 with PEEP. The optimal PEEP is derived using minimum EArea and Inflection method with the band of 5-10% above minimum EArea shown by the dashed-lines.

Figure 2

. Top Left Panel: Patient 2, Top Right Panel: Patient 6. Bottom Left Panel: Patient 8, Bottom Right Panel: Patient 10. Severe COPD or patients with similar clinical features (e.g. Patient 10) showed significantly higher EArea compared to other patients. PEEP selection is based on minimum E-Area and the inflection method with PEEP increase.

Figure 3 shows patient-specific constant lung elastance (E) with increasing PEEP for Patients 2, 6, 8 and 10. Edecreases with PEEP and the trend is similar to the EArea-PEEP plot of Figure 2, as expected from the high correlation. The optimal PEEP using minimum Eand Inflection E(Dashed-lines) are also indicated.

Figure 3

. Top Left Panel: Patient 2, Top Right Panel: Patient 6. Bottom Left Panel: Patient 8, Bottom Right Panel: Patient 10. PEEP derived from Minimum Eand Inflection method are as indicated.

Across all 10 patients, patient-specific constant lung elastance (E) can be represented by the median of dynamic lung elastance (E) with correlation R = 0.987. Correlation of Eand WOBis R = 0.815. EArea and median Eare also closely correlated with R = 0.896. Hence, Ecan be represented with EArea, where EArea captures all Evalues in a given breath and thus, is a more physiologically representative metric. Finally, validating Equation (2), EArea is correlated to the work to overcome lung elastance, WOB, as expected, with R = 0.936. The correlations are shown in Figure 4.

Figure 4

Pearson's Correlation. Top Left Panel: E-Median E, R = 0.987. Top Right Panel: E-WOB, R = 0.815. Bottom Left Panel: EArea-Median E, R = 0.896. Bottom Right Panel: EArea-WOB, R = 0.936.

Table 5 compares clinically selected PEEP during MV therapy with PEEP selected using Minimum E, and Minimum EArea and the Inflection method. The clinical values are set over a much narrower range, both higher and lower than those selected using E. Minimum EArea always selects a higher PEEP, by definition, than the Inflection method. However, Minimum EArea selects PEEP similar to or higher than Minimum E, where it also thus adds consideration of the reduction in overall WOBin selecting PEEP. PEEP derived from minimum Eand Inflection Eare also indicated.

Table 5

PEEP (cmH) selection in clinical and model-based approach.

	Patients
Selection Methods	1	2	3	4	5	6	7	8	9	10
Clinical	10	12	10	10	12	11	7.5	12	10	10
Minimum Edrs	20	15	15	25	25	15	5	20	15	20
Minimum Edrs Area	25	15	20	20	25	20	10	25	25	20
InflectionEdrs Area	14~16	6~9	15~17	16~18	22~24	7.5~12	5~7.5	21~23	20~23	12~16
Minimum Elung	25	15	15	30	25	15	5	25	30	20
Inflection Elung	13~17	6~9	8~10	26~27	21~24	7.5~10	5	12~18	19~22	12 ~15

4 Discussion

4.1 Model-based PEEP Selection

Median fitting error for time-variant Ein Table 2 is less than 1%, showing that a single compartment lung model can be used for time-varying Eestimation. The wide range of patient-specific Eacross all patients and PEEP shown in Table 2 reflects the heterogeneity of ALI/ARDS patient condition and response to PEEP that makes standardising and PEEP selection difficult [26]. Compared to the estimation of Ein Table 3, median fitting error is 5.6% and in specific cases, fitting error can be as high as 15.7-17.7% (Patients 4 and 5). This latter result indicates that a first order model can be used to estimate most patient-specific constant E, but, in several cases, the model may not accurately represent patients' physiological condition. Time-varying Eprovides a better model fit across all patients and also provides a clearer insight into the patient's physiological condition, and is thus the better model-based metric.

Figures 1 and Figure 2 shows E-Pressure-PEEP curves and EArea decrease with increasing PEEP, lung pressure, and volume over each breath. In the beginning of the recruitment manoeuvre, at zero end-expiratory pressure (ZEEP), Eis relatively very high for all patients with median 51.9cmH[IQR: 30.8-63.1]. In particular, chronic obstructive pulmonary disease (COPD) patients or patients with similar clinical features [27] (Patients 1, 4, 5, 9 and 10) have initially the highest Emedian, as expected, from 63.1cmH[IQR: 57.2-81.3] versus 30.8cmH[IQR: 29.5-46.6] for the other patients (p = 0.0079). As PEEP rises, it is observed that Ecurves drop at patient-specific rates. High constant lung elastance, Eat ZEEP and decreasing elastance as PEEP increments are also observed in Figure 3 for Patient 10.

In all cases, patient-specific Eand Edecrease to a patient-specific minimum before increasing at higher PEEP. Minimum Eand Esuggest the point where the lung is most compliant, if ventilated at that PEEP level. Further increases in PEEP and pressure thus lead to increased Eor E, and thus increase detrimental effects. In particular, increases in Eor Ecan be associated with overstretching of the patient's lung [16,28]. However, the heterogeneity of ALI/ARDS means there is a possibility of overstretching of healthy lung units even at low PEEP and airway pressures [10]. Thus, Minimum or, perhaps preferably, Inflection Eand Ecan provide a potentially higher resolution metric.

Patients 2 and 6 (Figure 1, 2, 3: Top panels) are examples where patient-specific E, EArea and Eincrease after descending to a minimum. Results suggest that further increases of PEEP and inflation pressures will stretch lung units causing possible damage, as seen by increasing Eat higher PEEP. The rise of Eoccurs at relatively low PEEP and pressure 15-20cmHin these two patients.

In contrast, Patients 8 and 10 (Figure 1, 2, 3: Bottom panels) never see Eor Erising even at the maximum PEEP used in this study. However, the Erange at higher PEEP for Patients 8 and 10 (PEEP 15~30cmH) is relatively small with median E= 31.3cmH, [IQR = 27.2-33.9]. This outcome indicates that further increases of PEEP from 15 to 30cmHhas no added advantage in reducing E, suggesting PEEP selection should be made at using the Inflection method.

Table 2 shows median [IQR] Efor every patient and PEEP. The IQR range drops significantly for every patient as PEEP increases. This range also indicates lung status or condition with the influence of pressure. A small IQR range indicates that the lung is ventilated at a PEEP level where maximal lung recruitment occurs over a narrow pressure range as tidal volume, Vis fixed in the MV mode used. A high IQR range shows the opposite. Hence, the lengths along pressure in Figure 1 also indicate how readily the patient was recruited and that easiest recruitment occurs at minimum E[29].

Table 4 shows the patient-specific EArea at each PEEP. It is found that EArea is closely related to median E, as shown in Figure 4. EArea at lower PEEP with median 64.9 cmH2Os/l [IQR: 37.6-102.2] is observed and as PEEP increases, EArea decreases. Upon reaching minimum EArea, patient-specific EArea increase with PEEP (Patients 2, 4, 6, 7 and 10). This trend is similar to the trend observed in patient-specific dynamic E(Table 2) and constant E(Table 3). Optimal PEEP derived using minimum or inflection method in EArea is similar to minimum patient-specific Ebut different as EArea considers the whole inspiration and the effect of WOB. It is also found that EArea is closely correlated to work in overcoming the lung elastic properties (WOB). This means that EArea provides combined information of patients-specific lung physiological conditions as well as work of breathing.

Table 5 shows the model-based approaches to PEEP selection compared to clinically selected PEEP. For 9 of 10 patients, the PEEP value selected using Minimum Eand EArea results in a value higher than the clinically selected PEEP. This latter result suggests that these patients could be treated at PEEP levels higher than clinically selected PEEP. When Minimum Eor EArea metrics are compared with Minimum E[16], they result in selecting similar PEEP. However, selecting PEEP is a trade off in minimizing lung pressure and potential damage, versus maximizing recruitment. Hence, the Inflection method offers similar recruitment at a lower PEEP and may be a safer choice, although its selected values are still higher than clinically selected in 7 of 10 cases. Overall, these results reflect the heterogeneity of the ALI/ARDS lung and the need for patient-specific approaches to select PEEP.

Patient 9 is an interesting case which illustrates the model's potential to capture unique patient-specific lung recruitment and condition as it occurs in a clinically and physiologically relevant manner. When the patient is ventilated from PEEP of 5 to 10cmH, median Eonly decreases by less than 1.0cmH. However, when PEEP is increased to 15cmH, the median Edrops significantly, as shown in Figure 5. This smaller Edrop suggests that only minimal lung volume is recruited from PEEP of 5 to 10 cmH. The significant drop in Eat PEEP 15cmHindicates that PEEP 15 cmHhas overcome recruitment resistance and additional new lung volume is recruited. Patient 9 was diagnosed with H1N1 and high PEEP for lung recruitment has proven to be beneficial for these patients [30]. Similar trends can be observed in Figure 5 bottom panels with the EArea-PEEP plot and E-PEEP-plot.

Figure 5

Patient 9 . Top Panel: Box-and-whisker diagram for Patient 9 Ewhen PEEP increase from 5 to 10cmH. The Edrops significantly when PEEP is increase from 10 to 15cmH. Bottom Left Panel: EArea-PEEP plot for Patient 9. Bottom Right Panel: E-PEEP for Patient 9.

4.2 Limitations

In this research, the lung model used to identify patient-specific Ecomprised a single compartment lung model. It was initially proposed for simple computational analysis and neglects the effect of nonlinear flow [31]. However, this analysis is based predominantly on trend comparisons, where the patient is their own reference. In addition, the model is simple and capable of capturing the fundamental lung mechanics, which varies intra- and inter- patients. Hence, this limitation should be minimal in this case, but should be confirmed with direct prospective clinical studies.

During the clinical trials, the patients were sedated and paralyzed using muscle relaxants. It is assumed that after sedation, the patient will be fully dependant on mechanical ventilation and not have spontaneous breathing effort. This assumption thus assumes the patient's pleural pressure (P) after sedation is zero and allows Pin Equation (3) to be omitted, which may not be entirely valid [32]. However, this assumption is made for the first step study to prove the concept within a simpler situation. Otherwise, the terms Eand Ewould represent a respiratory system elastance [31] and time-variant dynamic respiratory system elastance. However, given the low fitting errors observed, this issue should have little impact in this research.

During the course of estimating patient-specific Eor E, respiratory system resistance, R, is assumed overall constant within a physiological range [33] as PEEP increases. This assumption may not be entirely valid in some cases [33,34]. However, continuous measurements of respiratory resistance are not typically available and the effect of this resistive term is limited mathematically in its impact. Equally, trend comparison, as used here, across PEEP values will reduce the impact.

The identification of E, Eand EArea during MV is presented as a method to select PEEP, but there is currently no conclusive, optimum overall Eor EArea in patients. Erange varies depending on patient disease state and thus will also change over time. However, this trial includes only 10 patients, and there is not yet enough clinical data to indicate an optimum E, Eor EArea value for a specific patient or group. On-going, prospective trials with more specific patient groups should develop more conclusive outcomes, relating specific set values of Emetrics to effective patient-specific treatments and clinical outcome.

In particular, the time-varying Evalue and its change over a given breathing cycle, provides additional insight to guide ventilation that is not investigated here. For example, changes in ventilator pattern or mode to modify the Etrajectory could also be used with this data to guide therapy choice. However, this study does not have the numbers or design to provide that advice, or specific Evalues associated with specific decrease state or lung damage.

5 Conclusions

The model-based approach presented provides patient-specific, physiological insight not directly measurable without additional invasive, disruptive and clinically intensive test manoeuvres. This method can be directly implemented using modern ventilators with minimal, limited PEEP titrations, and thus without significant interruption to ongoing therapy. In particular, the full manoeuvres used here would not be required for clinical use, and only modest PEEP changes (3-8cmH) would be required to determine if Ewas decreasing at a different PEEP. Eoffers higher resolution in patients' response to change of pressure and PEEP, which is potentially, a better metric compared to existing constant lung elastance estimation. Thus, the overall method is readily generalisable and clinical practicable. It is able to capture patient-specific condition and responsiveness to PEEP and recruitment accurately, and as clinically expected. Hence, the approach presented offers significant potential to improve clinical insight and delivery of mechanical ventilation, and should be prospectively tested.

6 List of Abbreviations

ALI: Acute lung injury; APE: Absolute percentage error; ARDS: Acute respiratory distress syndrome; COPD: Chronic Obstructive Pulmonary Disease; E: Patient-specific constant lung elastance; E: Patient-specific dynamic lung elastance; FiO: Fraction of Inspired Oxygen; ICU: Intensive care unit; IQR: Interquartile Range; MV: Mechanical ventilation; PaOPartial pressure of oxygen in arterial blood; P: Airway pressure; PPleural pressure; P: Transpulmonary pressure; PEEP: Positive end expiratory pressure; PF Ratio: PaO; P: Offset pressure; Q: Flow; RM: Recruitment manoeuvre; R: Resistance; SIMV: Synchronized intermittent mandatory ventilation; t: Time; V: Volume; VILI: Ventilation induced lung injury; V: Tidal volume; WOB: Work of Breathing; WOB: Work to overcome respiratory system elastance; WOB: Work to overcome airway resistance; ZEEP: Zero PEEP

7 Competing Interests

The authors declare that they have no competing interests.

8 Authors Contribution

YSC, JGC, GMS created and defined the model. YSC, JGC and TD had input to analysis of results. GMS, AS implemented trials clinically with input from all others. All authors had input in writing and revising the manuscript. All authors have read and approved the final manuscript.

9 Consent

Written informed consent was obtained from the participant and or relative/friends/family of this study. A copy of written consent is available for review by the Editor-in-Chief of this journal.