Calculating mixed venous saturation during veno-venous extracorporeal membrane oxygenation

Abstract

Introduction: Recirculation (R), the shunting of arterial blood back into to the venous lumen, commonly occurs during veno-venous extracorporeal membrane oxygenation (VV-ECMO) and renders the monitoring of the venous line oxygen saturation no longer reflective of patient mixed venous oxygen saturation (S(V)O(2)). Previously, we failed to prove the hypothesis that, once R is known, it is possible to calculate the S(V)O(2) of a patient on VV-ECMO. We hypothesize that we can calculate S(V)O(2) during VV-ECMO if we account for and add an additional correction factor to our model for dissolved oxygen content. Therefore, the purpose of this study is to derive a more accurate model that will allow clinicians to determine S(V)O(2) during VV-ECMO when ultrasound dilution is being used to quantify R.

Methods: Using an extracorporeal circuit primed with fresh porcine blood, two stocks of blood were produced; (1) arterial blood (AB), and (2) venous blood (VB). To mimic recirculation, the AB and VB were mixed together in precise ratios using syringes and a stopcock manifold. Six paired stock AB/VB sets were prepared. Two sets were mixed at 20% R increments and 4 sets were mixed at 10% R increments. The partial pressure of oxygen (pO(2) ) and oxygen (O(2)) saturation of the stock blood and resultant mixed blood was determined. The original model was modified by modeling the residual errors with linear regression.

Results: When using the original model, as the partial pressure of arterial oxygen (P(a)O( 2)) of the stock AB increased, the calculated S(V)O(2) was higher than actual, especially at higher R levels. An iteration of the original model incorporating the P(a)O(2) level (low, medium, high) and R was derived to fit the data.

Conclusions: The original model using R and circuit saturations for the calculation of S(V)O( 2) in VV-ECMO patients is an oversimplification that fails to consider the influence of the high pO(2) of arterial blood during therapy. In the future, further improvements in this model will allow clinicians accurately to calculate S(V)O(2) in conjunction with recirculation measurements.

Figure 1

This is a published formula that is thought to be a way to calculate R. Where R is recirculation (%), SO_2preox is the O₂ saturation of the blood entering the oxygenator, SO_2postox is the O₂ saturation of the blood exiting the oxygenator and S_VO₂ is the mixed venous O₂ saturation in the patient. * (See reference for source)

Figure 2

Model 1 is a published formula. Where R is recirculation (%), SO_2preox is the O₂ saturation of the blood entering the oxygenator, SO_2postox is the O₂ saturation of the blood exiting the oxygenator and S_VO₂ is the mixed venous O₂ saturation in the patient. * (See reference for source)

Figure 3

This is a picture of the mixture apparatus. This allowed precise mixing of the arterial and venous sources.

Figure 4

Model 2 is the first attempt to modify Model 1. Where R is recirculation (%), SO_2preox is the O₂ saturation of the blood entering the oxygenator, SO_2postox is the O₂ saturation of the blood exiting the oxygenator and S_VO₂ is the mixed venous O₂ saturation in the patient. Notice the modification to correct for the error introduced by recirculation (0.168*R).

Figure 5

Model 3 is the final modification of Model 1. Where R is recirculation (%), SO_2preox is the O₂ saturation of the blood entering the oxygenator, SO_2postox is the O₂ saturation of the blood exiting the oxygenator and S_VO₂ is the mixed venous O₂ saturation in the patient. Notice the modification to correct for the error introduced by recirculation (0.005*R) as well as the addition of a variable to account for the relationship of R and P_aO₂ (0.163*R*P_aO₂).

Graph 1. Model 1 S_vO₂ Residuals versus Recirculation

The residuals on this graph show the lack of precision that is present when the data is applied to the original math model (Model 1). The RMSE is 12.96 and the bias is 8.35 showing that the previous model overestimated the predicted values.

Graph 2. Model 2 S_vO₂ Residuals versus Recirculation

The residuals on this graph show the improvement in precision when Model 2 is compared to Model 1. The RMSE is 8.63 and the bias is -1.47 showing that this model is an improvement over the previous model and has some degree of under estimation.

Graph 3. Model 3 S_vO₂ Residuals versus Recirculation

The residuals on this graph show the dramatic improvement of Model 3 when compared to Model 1. The RMSE is 4.03 and the bias is -0.90 showing that the final model has an improved accuracy and precision when compared to the original model.

Graph 4. Evaluation of residuals on a 45 degree scatter plot for Model 1

This scatter plot shows that Model 1 had a larger variance in the predicted SvO₂ and the gross degree of overestimation.

Graph 5. Evaluation of residuals on a 45 degree scatter plot for Model 2

This scatter plot shows that the variance and accuracy of Model 2 was an improvement when compared to Model 1.

Graph 6. Evaluation of residuals on a 45 degree scatter plot for Model 3

This scatter plot shows that the variance and accuracy of Model 3 is dramatically improved when compared to Model 1.