Gas exchange calculation may estimate changes in pulmonary blood flow during veno-arterial extracorporeal membrane oxygenation in a porcine model.

Veno-arterial extracorporeal membrane oxygenation (V-A ECMO) is used as rescue therapy for severe cardiopulmonary failure. We tested whether the ratio of CO2 elimination at the lung and the V-A ECMO (V˙co2ECMO/V˙co2Lung) would reflect the ratio of respective blood flows and could be used to estimate changes in pulmonary blood flow (Q˙Lung), i.e., native cardiac output. Four healthy pigs were centrally cannulated for V-A ECMO. We measured blood flows with an ultrasonic flow probe. V˙co2ECMO and V˙co2Lung were calculated from sidestream capnographs under constant pulmonary ventilation during V-A ECMO weaning with changing sweep gas and/or V-A ECMO blood flow. If ventilation-to-perfusion ratio (V˙/Q˙) of V-A ECMO was not 1, the V˙co2ECMO was normalized to V˙/Q˙ = 1 (V˙co2ECMONorm). Changes in pulmonary blood flow were calculated using the relationship between changes in CO2 elimination and V-A ECMO blood flow (Q˙ECMO). Q˙ECMO correlated strongly with V˙co2ECMONorm (r2 0.95-0.99). Q˙Lung correlated well with V˙co2Lung (r2 0.65-0.89, P < = 0.002). Absolute Q˙Lung could not be calculated in a nonsteady state. Calculated pulmonary blood flow changes had a bias of 76 (-266 to 418) mL/min and correlated with measured Q˙Lung (r2 0.974-1.000, P = 0.1 to 0.006) for cumulative ECMO flow reductions. In conclusion, V˙co2 of the lung correlated strongly with pulmonary blood flow. Our model could predict pulmonary blood flow changes within clinically acceptable margins of error. The prediction is made possible with normalization to a V˙/Q˙ of 1 for ECMO. This approach depends on measurements readily available and may allow immediate assessment of the cardiac output response.

INTRODUCTION

Extracorporeal membrane oxygenation (ECMO) is increasingly used as rescue therapy for severe cardiopulmonary failure (2). In veno-arterial (V-A) ECMO treatment, the native heart and lung work in parallel with the extracorporeal circuit and the assessment of native cardiac output [i.e., blood flow through the lungs (Q̇Lung)] is difficult. The ongoing unloading of the right ventricle even at low V-A ECMO blood flow (Q̇ECMO) makes assessment of cardiac function during V-A ECMO treatment challenging. Monitoring of the cardiac function and the evolution of native cardiac output during V-A ECMO treatment is not well standardized. Echocardiography is often used, but it requires specific knowledge (1) and routine echocardiographic parameters may not be useful in this context because of altered circulatory physiology and changing cardiac loading conditions (12). Monitoring of the evolution of native cardiac output based on simple, noninvasive, and readily available measurements would therefore be helpful in clinical practice, particularly during weaning, since early weaning success is associated with a favorable prognosis (9).

Gas exchange during V-A ECMO should reflect the combined effect of ventilation and perfusion of the native lung and those of the V-A ECMO circuit (21). We hypothesize that during V-A ECMO weaning the ratio between changes in CO2 elimination at the lung and the V-A ECMO (V̇co2ECMO and V̇co2Lung) is the same as the ratio between changes in the respective flows (Q̇ECMO and Q̇Lung). We tested this hypothesis in this preliminary, hypothesis-generating study by measuring the elimination of CO2 over the native lung and the V-A ECMO and the respective blood flows and compared the calculated flow changes with those directly measured from the pulmonary artery and V-A ECMO circuit.

METHODS

Animal care, surgery, and anesthesia.

This study was performed as a preliminary, independent substudy of a yet-unpublished project evaluating regional abdominal circulation during V-A ECMO and systemic inflammation, where measurements were done before the main study protocol was started. The study complied with the Guide for the Care and Use of Laboratory Animals (National Academy of Sciences, 1996) and Swiss National Guidelines and was approved, including an amendment for this substudy, by the Commission of Animal Experimentation of Canton Bern, Switzerland (BE119/17).

We studied a convenience sample of four animals (2 male and 2 female, 51.5 ± 1.3 kg) before the main study protocol was started. The pigs fasted for 12 h with free access to water. After anesthesia induction with intravenous midazolam and atropine and oral intubation, anesthesia was maintained with propofol and fentanyl, and the depth was controlled by repeatedly testing the response to nose pinch in additional to bispectral index target < 60 (BIS Quatro; Covidien, Mansfield, MA). Additional injections of fentanyl (50 µg) or midazolam (5 mg) were given as needed. Muscle relaxation was induced with rocuronium (0.5 mg/kg). Mechanical ventilation [volume control mode, positive end-expiratory pressure (PEEP) 5 cmH2O, fraction of inspired O2 () 0.3] was initiated with a tidal volume (Vt) of 7 mL/kg and a respiratory rate aiming at an end-tidal Pco2 () of 45 mmHg. A 5-Fr introducer sheath was placed in the right carotid artery for arterial blood pressure measurement and arterial blood gas sampling. Two three-lumen central venous lines were placed in the right and left jugular veins for right atrial pressure measurement and continuous administration of sedatives and vasopressors. V-A ECMO with right atrial-aortic cannulation and a left atrial vent (Maquet Cardiohelp, Quadrox MECC oxygenator, Rastatt, Germany; Medtronic cannula and vent, Minneapolis, MN) were installed via a sternotomy, and a bolus of 2,500 IE of unfractioned heparin was given. An appropriately sized ultrasonic flow probe was placed on the pulmonary artery (16- or 18-mm internal diameter, Transonic PAU series, Ithaca, NY). During surgery, fluid was supplemented with Ringer lactate at an initial rate of 5 mL·kg−1·min−1 and increased to 10 mL·kg−1·min−1. Any visible blood loss was replaced by hydroxyethyl starch (HES; 6% Voluven; Fresenius Kabi, Bad Homburg, Germany), and V-A ECMO pump speed was adjusted to achieve a mixed or central venous saturation > 50%.

Measurements and data recording.

Pulmonary blood flow, i.e., cardiac output (Q̇Lung) and V-A ECMO blood flow (Q̇ECMO), was measured on the pulmonary artery main trunk and arterial ECMO tubing (Transonic PAU series, Ithaca, NY). Pulmonary end-tidal Pco2 () and Pco2 at the membrane lung (peCO2ECMO) were measured with a sidestream capnograph (GE Medical, Module E-COVX with automated correction to BTPS conditions). The carbon dioxide production (V̇co2) was calculated individually for native and membrane lungs from the tidal Pco2 tracing as described below. We recorded sweep gas flow (V̇ECMO) manually. Arterial blood gases were taken before and after the study period. Pulmonary ventilation (V̇Lung) was kept constant. In the first animal, ventilator settings were kept identical to those before V-A ECMO [tidal volume (Vt) 0.465 L, 12 breaths/min], whereas in the subsequent animals V̇Lung was reduced to 2 L/min (Vt 0.25 L, 8 breaths/min) as V-A ECMO was started and kept constant thereafter. In all animals 5 cmH2O PEEP and volume control mode were used (Servo-i; Maquet, Solna, Sweden). The fraction of inspired oxygen was set at 0.30. Measurements were performed in healthy animals 30 min after surgery was completed. Eventually, the pigs were euthanized by injection of 40 mmol of potassium chloride, and V-A ECMO stopped in deep anesthesia. Data were recorded with LabVIEW (National Instruments Corp., Austin, TX) for off-line analysis with SOLEASY (ALEA Solutions, Zürich, Switzerland) and MATLAB R2019a (MathWorks, Natick, MA).

Experimental protocol.

The experiment consisted of three phases with varying sweep gas-to-blood flow ratios (i.e., the V̇/Q̇ of the membrane lung) to determine how the sweep gas-blood flow relationship at the V-A ECMO influences extracorporeal CO2 elimination (V̇co2ECMO). First, we reduced Q̇ECMO and V̇ECMO in parallel (stable V̇/Q̇ = 1; “reduction of V̇&Q̇” phase, rV̇&Q̇ECMO). Then we lowered V̇ECMO with a constant Q̇ECMO (V̇/Q̇ toward shunt; “reduction of V̇” phase, rV̇ECMO). Finally, we tested a V-A ECMO weaning trial, where Q̇ECMO was reduced but V̇ECMO was kept constant (V̇/Q̇ toward dead space; “reduction of Q̇” phase, rQ̇ECMO).

Q̇ECMO and V̇ECMO were set at 4 L/min each at baseline and afterward reduced, depending on the respective phase, to 75%, 50%, and 25% of baseline with an interval of 1 min for each condition (Fig. 1). The left atrial vent was clamped during these procedures, and the stepwise reduction of blood flow was not supported by vasopressors or inotropes.

Fig. 1.

Experimental protocol with stepwise reduction of veno-arterial extracorporeal membrane oxygenation (V-A ECMO) sweep gas flow (V̇ECMO) and/or blood flow (Q̇ECMO). Steps 1–4 of each phase [reduction of V̇(V), reduction of Q̇(Q), and reduction of both (VQ)] indicated at top.

Calculation of V̇co2 for V-A ECMO.

Expiratory concentration of CO2 at the V-A ECMO exhaust was calculated from the expiratory partial pressure of CO2 at the V-A ECMO exhaust and used to calculate V̇co2 (16, 23), using actual barometric pressures (on average 722 mmHg). The experiments were performed at 540 m above sea level.

Calculation of V̇co2 for the lung.

Mean pulmonary expired carbon dioxide () was calculated by averaging the end-tidal carbon dioxide () curve over the respiratory cycle with correction for the inspiratory-to-expiratory (I:E) ratio:

This was verified by integration of the expiratory Pco2 curve, which delivers the same result.

We then calculate V̇co2Lung:

Blood flow calculations.

Figure 2 depicts the situation during V-A ECMO schematically. We define the following relationships, whereby Q̇ is flow and Δv-aCO2 is the inflow-outflow difference in blood CO2 content in a given segment (Δv-aoCO2 is the difference between venous and aortal CO2 content, Δv-LACO2 is the difference between venous and left atrial CO2 content, and Δv-pmCO2 is the difference between venous and postmembrane CO2 content):

Fig. 2.

Schematics for veno-arterial extracorporeal membrane oxygenation (V-A ECMO). Q̇ECMO, V-A ECMO blood flow; Q̇Lung, lung blood flow; RA, right atrium; V̇co2ECMO, elimination of CO2 (V̇co2) at V-A ECMO; V̇co2Lung, V̇co2 at lung; V̇o2, oxygen intake; cvCO2, mixed venous CO2 content; cpmCO2, post oxygenator CO2 content; claCO2, left atrial CO2 content; caoCO2, aortal CO2 content.

We then implement and into :

We now solve for Q̇Lung:

As we aim to calculate Q̇Lung with expired gas phase measurements only rather than calculating blood gas content from multiple blood gas samples, we modify with the following assumptions. As carbon dioxide production and carbon dioxide elimination are mathematical opposites, we use the absolute value function, thus eliminating negative values:

We now implement – into :

simplifies to

There is a fixed relationship of Q̇Lung and Q̇ECMO with the respective eliminated CO2. This expresses our hypothesis that the ratio between the differences in V̇co2ECMO and V̇co2Lung is the same as the ratio between the differences in the respective flows (Q̇ECMO and Q̇Lung). In our experimental setup, we cannot expect to reach a steady state, as step changes were set at 1 min. Therefore, we calculate pulmonary blood flow using the differences in V̇co2 and Q̇ECMO during V-A ECMO weaning rather than applying it to steady-state conditions.

Normalization of uneven V̇/Q̇ at the V-A ECMO.

During phase rV̇&Q̇ECMO with a constant V̇/Q̇ECMO of 1, we expect the relationship in to work. However, ∆V̇co2ECMO is influenced by V̇ECMO and Q̇ECMO. Q̇ECMO determines the amount of CO2 transported toward the membrane lung, and V̇ECMO determines the amount of CO2 eliminated over the membrane lung with a major impact on ∆V̇co2ECMO (10, 13, 17). ∆V̇co2ECMO therefore does not necessarily represent ∆Q̇ECMO, when V̇/Q̇ECMO differs from 1. During phase rQ̇ECMO, V̇co2 may decouple from Q̇ECMO. Accordingly, the ratio ∆V̇co2ECMO/∆V̇co2Lung is affected by V̇ECMO despite unchanged blood flows.

To correct for uneven V̇/Q̇, we normalized ∆V̇co2ECMO into a new variable, ∆V̇co2ECMONorm, only dependent on Q̇ECMO and independent of V̇ECMO, with . The correction factor f is expressed in .

A formal deduction of this normalization is found in the .

Statistical analysis.

For statistical, mathematical, and graphical analysis, we used MATLAB R2019a (MathWorks, Natick, MA), including an extension pack under a creative commons license for the creation of Bland–Altman plots (15). Data are presented either individually or as a range. Correlation coefficients were calculated with Pearson’s square (r2). Agreement between methods (calculated and measured Q̇Lung) was assessed with Bland–Altman analysis.

RESULTS

Baseline.

At baseline V̇ECMO and Q̇ECMO of 4 L/min, V̇co2ECMO was between 202 and 243 mL/min, whereas V̇co2Lung was between 13 and 193 mL/min, corresponding to a measured Q̇Lung of 10–964 mL/min and representing a normal V̇co2 production for swine (6) (step 1 for V̇, Q̇, and V̇Q̇ in Table 1).

Table 1.

Individual data sets

	ECMO				Lung
Step	V̇	Q̇	V̇co2	V̇co2Norm	V̇	Q̇	V̇co2	Q̇calc
	Animal 1
V̇
1	4,000	4,105	214	217	5,600	964	189	3,572
2	3,000	4,092	177	212	5,600	917	189	3,647
3	2,000	4,049	135	209	5,600	1,125	195	3,778
4	1,000	4,071	77	203	5,600	980	196	3,934
Q̇
1	4,000	4,113	226	229	5,600	920	197	3,529
2	4,000	3,147	202	179	5,600	1,035	200	3,520
3	4,000	2,058	173	128	5,600	1,458	215	3,463
4	4,000	1,207	140	88	5,600	1,915	244	3,348
V̇Q̇
1	4,000	4,068	211	213	5,600	843	202	3,859
2	3,000	3,231	168	174	5,600	1,157	199	3,686
3	2,000	2,191	120	126	5,600	1,376	227	3,945
4	1,000	1,178	66	73	5,600	1,550	242	3,932
	Animal 2
V̇
1	4,000	4,013	222	223	1,800	389	52	935
2	3,000	4,010	194	229	1,800	387	61	1,077
3	2,000	3,982	150	229	1,800	370	66	1,141
4	1,000	3,994	86	223	1,800	358	60	1,065
Q̇
1	4,000	4,079	259	261	1,800	105	64	1,006
2	4,000	3,016	236	205	1,800	503	73	1,073
3	4,000	1,995	205	150	1,800	968	72	966
4	4,000	1,048	160	97	1,800	1,349	87	944
V̇Q̇
1	4,000	4,016	245	245	1,800	126	75	1,236
2	3,000	3,008	195	195	1,800	560	76	1,170
3	2,000	2,019	142	143	1,800	991	88	1,245
4	1,000	1,094	67	71	1,800	1,472	105	1,626
	Animal 3
V̇
1	4,000	4,062	263	266	1,800	10	14	217
2	3,000	4,043	228	270	1,800	10	18	264
3	2,000	4,025	177	274	1,800	5	19	283
4	1,000	3,989	102	266	1,800	2	22	333
Q̇
1	4,000	4,031	287	288	1,800	4	23	324
2	4,000	2,966	261	225	1,800	8	22	296
3	4,000	1,994	228	167	1,800	328	42	502
4	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A
V̇Q̇
1	4,000	4,074	260	262	1,800	9	42	651
2	3,000	3,008	211	211	1,800	36	37	529
3	2,000	1,973	168	167	1,800	399	57	677
4	1,000	641	79	64	1,800	1327	103	1,035
	Animal 4
V̇
1	4,000	4,170	239	244	2,000	59	22	378
2	3,000	4,216	197	240	2,000	72	30	528
3	2,000	4,188	154	244	2,000	39	34	592
4	1,000	4,186	89	241	2,000	21	32	561
Q̇
1	4,000	4,177	248	254	2,000	9	31	515
2	4,000	3,031	225	195	2,000	294	36	556
3	4,000	2,064	192	142	2,000	616	58	845
4	4,000	1,060	149	91	2,000	909	68	801
V̇Q̇
1	4,000	4,031	231	232	2,000	14	49	848
2	3,000	3,098	188	191	2,000	259	55	899
3	2,000	2,108	138	142	2,000	602	65	964
4	1,000	1,051	75	78	2,000	928	78	1,060

Values (in mL/min) are individual data for all animals at baseline [step 1 at reduction of ventilation (V̇), reduction of blood flow (Q̇), and reduction of both (V̇Q̇)] and every step of blood flow reduction. Extracorporeal membrane oxygenation (ECMO) Q̇ and Lung Q̇ denote readings from the respective flow probes. CO2 elimination (V̇co2) values were calculated according to – in methods with the reported barometric pressure for each day (728, 726, 711, and 721 mmHg). Note that 1) in animal 1 ventilation is high because baseline settings at respirator were 5.6 L/min [tidal volume (Vt) 465 mL, frequency 12 times/min] and that 2) during reduction of Q̇ phase the cardiovascular system of animal 3 did not support the ECMO reduction to 25% of baseline, and therefore no measurement is available (N/A). V̇co2Norm refers to a calculated V̇co2 for a sweep gas-to-blood flow ratio normalized toward 1 (for details see ). Q̇calc, calculated lung Q̇.

Measurements at the V-A ECMO.

Per protocol, Q̇ECMO remained unchanged from baseline during phase rV̇ECMO (98–100% of baseline or 3,989–4,186 mL/min) and was reduced to a quarter of baseline in phase rV̇&Q̇ECMO (641–1,178 mL/min, 16–29% of baseline). In phase “rQ̇ECMO,” Q̇ECMO was reduced to approximately a quarter in all animals except animal 3 because of hemodynamic instability (25.4–49.5% of baseline or 1,048–1,994 mL/min) (Table 1).

The normalization function was calculated by fitting our data points into and retrieving the constant c = 1.157 (r2 = 0.995, P < 0.001). V̇co2ECMONorm correlated highly with Q̇ECMO, and the normalization improved correlation significantly (Fig. 3, A and B, respectively). In phase rV̇ECMO, reducing V̇ECMO without any change in Q̇ECMO, V̇co2ECMONorm was 194–249 mL/min or 93.3–100.1% of baseline. Without normalization, V̇co2ECMO decoupled from Q̇ECMO, with a decrease from 205–246 mL/min to 73–96 mL/min in this phase (Table 1, Fig. 3). V̇co2ECMO values for phase rV̇&Q̇ECMO dropped to roughly a quarter of baseline (64–74 mL/min, 25–33% of baseline) in parallel with reduced Q̇ECMO. During phase rQ̇ECMO, V̇co2ECMONorm was 84–156 mL/min or 38–58% of baseline.

Fig. 3.

Effect of the normalization of the sweep gas flow-to-blood flow ratio on the veno-arterial extracorporeal membrane oxygenation (V-A ECMO). A: scatterplot for V-A ECMO blood flow (Q̇ECMO) vs. elimination of CO2 at V-A ECMO (V̇co2ECMO). Smallest points represent phase rV̇ECMO; medium-sized points represent phase rQ̇ECMO; large points represent phase rV̇&Q̇ECMO. No correlations reached significant levels (P < 0.05). B: scatterplot for Q̇ECMO vs. V̇co2ECMO normalized to ventilation-to-perfusion ratio (V̇/Q̇) = 1 (V̇co2ECMONorm), all data points considered. Smallest points represent phase rV̇ECMO; medium-sized points represent phase rQ̇ECMO; large points represent phase rV̇&Q̇ECMO. In phase rQ̇ECMO, animal 3 did not tolerate the last reduction in V-A ECMO flow.

Measurements at the lung.

During unchanged Q̇ECMO (phase rV̇ECMO), Q̇Lung remained close to baseline (2–980 mL/min) and did not change much within one animal, and V̇co2 stayed constant, accordingly.

During reduction of Q̇ECMO in phase rV̇&Q̇ECMO and phase rQ̇ECMO, Q̇Lung increased from its low baseline values to 928–1,550 mL/min, and 328–1,914 mL/min, respectively (Table 1). V̇co2Lung followed the changes in Q̇Lung to 74–232 mL/min (rise of 28–57 mL/min from baseline, with stepwise increases in every animal) for phase rV̇&Q̇ECMO and 39–233 mL/min for phase rQ̇ECMO (rise of 18–45 mL/min from baseline) and remained steady at full Q̇ECMO (phase rV̇ECMO, 21–188 mL/min; change of 7–8 mL/min from baseline) (Table 2). Q̇Lung and V̇co2Lung showed a high correlation (Fig. 4).

Table 2.

Calculation of stepwise reductions

	ECMO				Lung
Step	ΔV̇	ΔQ̇	ΔV̇co2	ΔV̇co2Norm	ΔV̇	ΔQ̇	ΔV̇co2	ΔQ̇calc
	Animal 1
V̇
1→2	−1,000	−13	−37	−6	0	−47	0	−1
2→3	−1,000	−43	−42	−2	0	208	7	122
3→4	−1,000	22	−59	−7	0	−145	1	−2
Summed up	−3,000	−34	−138	−15	0	16	7	119
Q̇
1→2	0	−966	−24	−50	0	115	3	66
2→3	0	−1,089	−29	−51	0	423	15	313
3→4	0	−851	−32	−40	0	457	28	605
Summed up	0	−2,906	−86	−142	0	995	47	984
V̇Q̇
1→2	−1,000	−837	−43	−38	0	314	−3	−59
2→3	−1,000	−1,040	−47	−48	0	219	28	612
3→4	−1,000	−1,013	−54	−54	0	174	15	286
Summed up	−3,000	−2,890	−144	−140	0	707	41	838
	Animal 2
V̇
1→2	−1,000	−3	−27	6	0	−2	9	−5
2→3	−1,000	−28	−42	1	0	−17	4	−189
3→4	−1,000	12	−61	−6	0	−13	−6	12
Summed up	−3,000	−19	−130	1	0	−32	7	−181
Q̇
1→2	0	−1,063	−22	−56	0	398	9	160
2→3	0	−1,021	−31	−55	0	465	0	−9
3→4	0	−947	−44	−53	0	381	15	268
Summed up	0	−3,031	−98	−164	0	1,244	23	419
V̇Q̇
1→2	−1,000	−1,008	−50	−50	0	434	0	7
2→3	−1,000	−989	−52	−52	0	431	12	236
3→4	−1,000	−925	−75	−72	0	481	17	217
Summed up	−3,000	−2,922	−177	−174	0	1,346	30	460
	Animal 3
V̇
1→2	−1,000	−19	−36	4	0	0	3	−15
2→3	−1,000	−18	−50	4	0	−5	2	−8
3→4	−1,000	−36	−75	−8	0	−3	3	13
Summed up	−3,000	−73	−161	0	0	−8	8	−9
Q̇
1→2	0	−1,065	−25	−63	0	4	−1	−11
2→3	0	−972	−33	−58	0	320	20	326
3→4	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A
Summed up	0	−2,037	−59	−121	0	324	19	315
V̇Q̇
1→2	−1,000	−1,066	−49	−51	0	27	−5	−99
2→3	−1,000	−1,035	−43	−45	0	363	20	459
3→4	−1,000	−1,332	−89	−103	0	928	45	588
Summed up	−3,000	−3,433	−181	−199	0	1,318	61	948
	Animal 4
V̇
1→2	−1,000	46	−42	−4	0	13	8	−84
2→3	−1,000	−28	−43	4	0	−33	4	−31
3→4	−1,000	−2	−65	−2	0	−18	−2	−2
Summed up	−3,000	16	−150	−3	0	−38	10	−117
Q̇
1→2	0	−1,146	−24	−59	0	285	4	88
2→3	0	−967	−32	−53	0	322	22	410
3→4	0	−1,004	−43	−52	0	293	10	196
Summed up	0	−3,117	−99	−164	0	900	37	693
V̇Q̇
1→2	−1,000	−933	−43	−40	0	245	7	158
2→3	−1,000	−990	−50	−49	0	343	10	195
3→4	−1,000	−1,057	−63	−65	0	326	13	213
Summed up	−3,000	−2,980	−155	−154	0	914	29	565

Calculations of stepwise reductions [reduction of ventilation (V̇), reduction of blood flow (Q̇), and reduction of both (V̇Q̇)] for all lung and extracorporeal membrane oxygenation (ECMO) data. Values (in mL/min) are individual data for all animals for measurements performed at the lung. The summed up category refers to the cumulative change from step 1 to step 4. Note that during reduction of Q̇ phase the cardiovascular system of animal 3 did not support the ECMO reduction to 25% of baseline, and therefore no measurement is available (N/A). Q̇calc, calculated lung Q̇.

Fig. 4.

Correlation between lung blood flow (Q̇Lung) and CO2 elimination at the lung (V̇co2Lung) absolute values: scatterplot for Q̇Lung vs. V̇co2Lung, all data points considered. Smallest points represent phase rV̇ECMO; medium-sized points represent phase rQ̇ECMO; large points represent phase rV̇&Q̇ECMO. Note that in animal 1 ventilation and thus V̇co2Lung is high, because baseline settings at respirator were 5.6 L/min [tidal volume (Vt) 465 mL, 12 times/min]. This was the first animal, and the ventilator settings were not adjusted from previous settings. In phase rQ̇ECMO, animal 3 did not tolerate the last reduction in veno-arterial extracorporeal membrane oxygenation (V-A ECMO) flow.

Calculation of Q̇Lung.

The calculation of pulmonary blood flow from absolute V̇co2 values is imprecise and leads to a consistent overestimation (Table 1). This overestimation increases with increasing V̇/Q̇ at the lung, which is shown in animal 1, where we had increased ventilation compared with the other animals. In phase rV̇ECMO, we observe no change in measured Q̇Lung as well as calculated changes in Q̇Lung. When differences between the short stepwise flow reductions are considered (Table 2), correlations are reestablished (Fig. 5) and the respective Bland–Altman plot (Fig. 5) shows a small bias with acceptable limits of agreement. True blood flow changes are underestimated since bias is positive. Bias stays constant over the measured range (R2 = −0.16, P = 0.5). When phase rV̇ECMO is excluded due to no expected change in blood flow, out of 23 blood flow change calculations, an opposite direction of the flow change is calculated in 4 instances. In all of these instances, the value of the change is below the least significant change, which is 113 mL/min. When the entire reduction steps are summarized (Table 2 and Fig. 5), the relationship becomes overt.

Fig. 5.

A: Bland–Altman plot for all data points during veno-arterial extracorporeal membrane oxygenation (V-A ECMO) weaning. Bias is positive but close to zero, with wide limits of agreement. Bias stayed constant over increasing changes in lung blood flow (Q̇Lung) (R2 = 0.014). LoA, limits of agreement. B: scatterplot for the real change in Q̇Lung vs. the calculated change in Q̇Lung during V-A ECMO weaning. Smallest points represent phase rV̇ECMO; medium-sized points represent phase rQ̇ECMO; large points represent phase rV̇&Q̇ECMO. Linear regressions yield animal 1: y = 0.75 × x + 73.34; animal 2: y = 0.44 × x – 47.85; animal 3: y = 0.73 × x + 7.17; animal 4: y = 0.8 × x – 30.17. C: scatterplot for subsumed weaning steps for each animal. Linear regressions yield animal 1: y = 0.91 × x + 125.05; animal 2: y = 0.47 × x – 166.98; animal 3: y = 0.70 × x + 34.8; animal 4: y = 0.79 × x – 84.95.

DISCUSSION

We show in a preliminary analysis that measurements of V̇co2 at both lung and V-A ECMO are possible with simple sidestream technology. Our model for the estimation of changes in Q̇Lung predicts the directional change of pulmonary blood flow, i.e., cardiac output with acceptable accuracy in this small sample size (3). The measurements needed for our calculations (Q̇ECMO, V̇ECMO, V̇Lung, peCO2ECMO, ) are easily performed with the use of standard sidestream capnographs, all of which are readily available in an ICU setting or an operating theater and require no specific training.

As expected from the ventilation-perfusion concept and the gas content equations in Fig. 2 (14), we found that decrease in Q̇ECMO and the consecutive increase in Q̇Lung leads to a respective change in V̇co2Lung and V̇co2ECMONorm. A closer look at as the background of our hypothesis shows an adaptation of the classic Berggren shunt equation (11). This seems intuitive, as the V-A ECMO is in concept an anatomical right-to-left shunt, where the ability to ventilate and oxygenate the shunted blood will clearly affect its functional influence (Fig. 2). Changing the sweep gas-to-blood flow ratio on the ECMO will vary the function of this anatomical shunt from true shunt (V̇ECMO = 0 at any Q̇ECMO) to dead space (Q̇ECMO = 0 at any V̇ECMO). V̇co2ECMO only represents the shunt correctly as long as sweep gas/blood flow on the V-A ECMO are kept at a ratio of 1 (in phase rV̇&Q̇ECMO). For sweep gas-to-blood flow ratios differing from 1, sweep gas flow (V̇ECMO) will drastically change the amount of the eliminated CO2 (10, 17) independently of blood flow, a known phenomenon in states of shock or multiorgan failure (13). We could simulate this in the derivation of our normalization procedure (see , Figs. A2 and A3) and reproduce it in the experiment during the steps in phase rV̇ECMO (Table 1).

Fig. A2.

Three-dimensional mesh plot showing postmembrane Pco2 (, mmHg) as a function of ventilation (L/min) and blood flow (L/min).

Fig. A3.

Three-dimensional mesh plot showing elimination of CO2 at veno-arterial extracorporeal membrane oxygenation (V̇co2ECMO, mL/min) as a function of ventilation (L/min) and blood flow (L/min).

The normalization of V̇co2ECMO reestablishes a sweep gas-to-blood flow ratio of 1, and therefore restores the correlation between V̇co2ECMONorm and Q̇ECMO. This newly calculated V̇co2ECMONorm now is only dependent on blood flow and independent from ventilation and thus eliminates the influence of V̇/Q̇ECMO mismatch on blood flow calculations. We used our data to calculate the constant c with a curve fitting function, in order to stay independent from blood gas measurements, although individual calculations would be possible from premembrane pH. We see the high goodness of fit of this normalization procedure as an indirect proof of the normalization function (see , Fig. A6). During V-A ECMO weaning with a sweep gas-to-blood flow ratio of 1, it seems of little practical importance. Normalization might be particularly helpful to wean a low blood flow system with the primary intention to eliminate CO2, where the effect of increased ventilation is most relevant (5) (see , Fig. A3). Whether this might be applicable to a veno-venous configuration would need to be investigated. In a veno-arterial configuration, normalization might allow accurate estimations of postmembrane CO2 pressures in blood, enabling a continuous gaseous oxygenator measurement to derive blood gas tensions (see , Fig. A2).

Fig. A6.

Curve fitting for correction factor f as a function of ventilation-to-perfusion ratio (V̇/Q̇).

A high V̇/Q̇Lung will significantly increase the overall amount of CO2 eliminated and thus lead to an overestimation of pulmonary blood flow, whereas a reduction in V̇ECMO will lead to a decrease in eliminated CO2 and thus to a rise in venous CO2 content. This in turn increases V̇co2Lung, to achieve a new steady state. However, as the V-A ECMO and the lung both drain venous blood from the right atrium, V̇co2ECMO should increase simultaneously with the new steady state to fulfill . Our short measurement periods did preclude a steady state for CO2 elimination. Calculations of total blood flow for any given moment may therefore be impossible, because the lack of a steady state does not allow for sufficient accuracy. As we calculated Q̇Lung through a deliberate step change in V̇co2, a steady state is not necessary, as there is no need for an absolute reference point. This also allows calculations for different settings of V̇Lung (as shown with animal 1), as long as V̇Lung remains constant.

The ratio of ventilation to perfusion in the lung will vary with hypoxic vasoconstriction, shunt, alveolar collapse, and dead space. Our V̇co2Lung, estimated from end-tidal Pco2 in healthy lungs, showed an acceptable relationship with Q̇Lung, but stable minute ventilation on the lung was mandatory. As Q̇Lung is the quantity to be calculated, a normalization procedure is not possible. As V̇co2Lung can only represent blood flow that participates in gas exchange, shunt due to supine positioning of the animals could explain the bias of underestimation of changes in pulmonary blood flow with our method.

There are several possible limitations to our method: First, a V̇/Q̇Lung mismatch (e.g., high shunt and/or high dead space) might result in a decrease of Q̇Lung-V̇co2Lung correlation and might thus increase the bias significantly. Second, we did not document every V-A ECMO flow change with blood gas samples, because our aim was to calculate Q̇Lung with gaseous measurements. Nevertheless, a meticulous documentation of blood gas status would strengthen our hypothesis and allow for alternative calculations of gas content and direct calculations of the normalization function. Third, V̇co2 was calculated with sidestream capnography, which is of limited accuracy. Signal shifts in the Pco2-time tracing may introduce an error here. We did not rely on a breath-by-breath measurement, but averaged values over 1 min may help to minimize this possible influence. Mainstream calorimetric modules are available and used in assessing cardiac output, alveolar and dead space ventilation (7, 18–20). Mainstream capnography at the V-A ECMO gas outlet is feasible and may deliver accurate results for oxygen intake (V̇o2) and V̇co2 (4, 22). This might improve our results and overall accuracy compared with our calculations from sidestream end-tidal carbon dioxide. Fourth, this study was conducted in a small, clearly preliminary set of healthy animals and without any cardiovascular support.

The large scatter in pulmonary flow reflects the individual variability of native cardiac output during V-A ECMO treatment. In conclusion, we show that measurements of V̇co2 at the V-A ECMO are easily performed. A normalization procedure allows estimation of V̇co2 only dependent on blood flow without the influence of a V̇/Q̇ mismatch. This in turn lays the basis of blood flow calculations using V̇co2 values. Calculations of pulmonary blood flow using absolute values of carbon dioxide elimination are not possible in a nonsteady state with our method. The concept can be derived from basic physiological equations. Whether our method may result in a clinically useful approach and support V-A ECMO weaning, where assessment of cardiac output may help to evaluate weanability, has to be further evaluated. These preliminary findings need further confirmation in a larger study, also investigating low and high V̇/Q̇ states at the lung before exploring clinical applications.

GRANTS

The Department of Intensive Care Medicine has received unrestricted educational grants from the following organizations for organizing a quarterly postgraduate educational symposium, the Berner Forum for Intensive Care (until 2015): Fresenius Kabi, GSK, MSD, Lilly, Baxter, Astellas, AstraZeneca, B. Braun, CSL Behring, Maquet, Novartis, Covidien, Nycomed, Pierre Fabre Pharma AG (formerly known as RobaPharm), Pfizer, Orion Pharma, Bard Medica S.A., Abbott AG, and Anandic Medical Systems.

The Department of Intensive Care Medicine has received unrestricted educational grants from the following organizations for organizing biannual postgraduate courses in the fields of critical care ultrasound, management of ECMO and mechanical ventilation: Pierre Fabre Pharma AG (formerly known as RobaPharm), Pfizer AG, Bard Medica S.A., Abbott AG, Anandic Medical Systems, PanGas AG Healthcare, Orion Pharma, Bracco, Edwards Lifesciences AG, Hamilton Medical AG, Fresenius Kabi (Schweiz) AG, Getinge Group Maquet AG, Dräger Schweiz AG, and Teleflex Medical GmbH.

DISCLOSURES

The Department of Intensive Care Medicine, University Hospital Bern, has, or has had in the past, research contracts with Orion Corporation, Abbott Nutrition International, B. Braun Medical AG, CSEM SA, Edwards Lifesciences Services GmbH, Kenta Biotech Ltd, Maquet Critical Care AB, and Omnicare Clinical Research AG and research and development/consulting contracts with Edwards Lifesciences SA, Maquet Critical Care AB, and Nestlé. The money was paid into a departmental fund; no author received personal financial gain. K. F. Bachmann, D. Berger, and L. Gattinoni filed a patent for the method described. None of the other authors has any conflicts of interest, financial or otherwise, to disclose.

AUTHOR CONTRIBUTIONS

K.F.B. and D.B. conceived and designed research; K.F.B., M.H., and D.B. performed experiments; K.F.B. analyzed data; K.F.B., M.H., L.G., and D.B. interpreted results of experiments; K.F.B. prepared figures; K.F.B. and D.B. drafted manuscript; K.F.B., S.M.J., J.T., L.G., and D.B. edited and revised manuscript; K.F.B., M.H., S.M.J., J.T., L.G., and D.B. approved final version of manuscript.