Introduction
Over the past few decades the mortality after heart surgery has decreased significantly [1]. However, despite this progress, there are still multiple problems in heart surgery that need to be addressed. Cardiac surgery-associated acute kidney injury (CSA-AKI) is one of the most serious and common complications, with an incidence of up to 43% [2]. Patients with acute kidney injury (AKI) show significantly lower short- and long-term survival rates [2]. Therefore, it should be a major goal to prevent the occurrence of AKI after heart surgery.
Ranucci et al. have shown, that a low oxygen delivery index (DO2i) during surgery is an independent risk factor for AKI in patients undergoing heart surgery with the use of cardiopulmonary bypass (CPB) [3]. Due to hemodilution, hemoglobin levels may decrease, which could result in a low DO2i. To prevent the DO2i from falling below a critical level, it is recommended by the EACTS/EACTA/EBCP guidelines that the pump flow rate of the heart-lung machine (HLM) should be adjusted according to the arterial oxygen content [4]. Compensating a low level of hemoglobin by raising the pump flow and getting the DO2i to a noncritical value is one of the main targets of a so-called GDP (goal-directed perfusion) management [5].
To date, most studies investigated the influence of DO2i and the incidence of CSA-AKI in conventional cardiopulmonary bypass cases with total sternotomy. In minimally invasive cardiac surgery (MICS), the arterial pump flow is sometimes limited due to smaller venous cannulas [6]. Since we had no GDP monitoring available and did not follow a GDP concept in the study period, we wanted to know if we reached the recommended DO2i threshold during CPB in our MICS cases. Further to this, we explored whether there is a relationship between low DO2 and postoperative AKI. Hence, we focused on patients undergoing minimally invasive reconstruction or replacement of the mitral valve via partial right-sided anterior thoracotomy.
Methods
Patient Population
We retrospectively analyzed records of all patients (>=18 years) who underwent either minimally invasive mitral valve repair or replacement via right-sided anterolateral minithoracotomy at our institution between July 2017 and May 2020. The presence of renal replacement therapy (RRT) before surgery was the only exclusion criterion. In the study period we identified 294 patients who underwent mitral valve surgery. One patient met the exclusion criterion and the perfusion data for two other patients were missing. Thus, a total of 291 patients were included in the final analysis.
Anesthesiological Management
Induction of anesthesia was done with sufentanil and propofol. Cisatracurium was commonly used for muscle relaxation. Maintenance of anesthesia was continued with sevoflurane and sufentanil boli. A central venous catheter was placed via the internal jugular vein for central venous pressure measurement, and as vascular access. Anesthesia during perfusion was continued via propofol perfusor and intermittent boli of sufentanil.
CPB Management and Calculation of Oxygen Delivery- Related Parameters
The procedures were conducted with three different types of HLMs with roller pumps (HL20; Maquet, Rastatt, Germany, S3 or S5; Sorin Group, Munich, Germany). Electronic patient data and CPB data were obtained by JOCAP XL (Maquet, Rastatt, Germany) or ECC-Online (Sorin Group, Mirandola, Italy). During perfusion, data were recorded every 15 to 20 seconds. Data- Master (Sorin Group, Mirandola, Italy) for continuous online arterial oxygen partial pressure, venous saturation and venous hematocrit measurement, as well as vacuum assisted venous drainage (Maquet, Rastatt, Germany) were used in all cases. GDP monitoring was not available. Two CBP sets were used: oxygenator Quadroxi Adult with venous Reservoir VHK 71000 (Maquet, Rastatt, Germany) or CAPIOX FX-25 with Capiox Advance Hardshell Venous Reservoir (Terumo, Shibuya, Japan) both with integrated arterial filter. Priming volume in both sets consisted of 1500 ml Sterofundin ISO (B.Braun, Melsungen, Germany) with 10000 I.E. Heparin. Target cardiac index (CI) was 2.4 l/min/m2 for calculation of pump flow rate with a mean arterial pressure of 50 – 70 mmHg and a bladder temperature of 34 – 35° Celsius during CPB. Before initiation of CPB, retrograde autologous priming (RAP) was performed as a standard of care. Since cardioplegic solution was not suctioned from the right atrium, removal of the crystalloid volume was done with a hemoconcentrator. Transfusion trigger on CBP is 8g/dl at our institution, but also with the possibility of an individual team decision, based on venous saturation, NIRS value and volume status of the patient on bypass.
To calculate oxygen delivery (DO2) during CPB, several steps were needed. Each perfusion record was edited manually and unnecessary data were removed, since data recordings did not only contain the pump flow during, but also the pump flow be- fore and after termination of CPB. Initiation and termination of bypass time was also excluded from analysis by removing the pump flow at the beginning and at the end, which was below 70% of the target pump flow, to exclude times where the heart ejects. Furthermore, hemoglobin values, which were needed to calculate DO2 during CPB, were not saved in the perfusion records. During CPB, arterial and venous blood gas and an ethylenediaminetetraacetic acid (EDTA) sample were taken every 30 minutes by the perfusionist and sent to the main lab. An ABL 800 (Radiometer GmbH, Krefeld, Germany) was used for blood gas analysis and a Sysmex XN-2000 (Sysmex Corporation, Kōbe, Japan) was used for haemoglobin determination from the EDTA samples. Out of these three hemoglobin values, the lowest one was used for DO2 calculation until a new value was available. Automatic data processing of edited perfusion records and calculation of the required variables was done using MATLAB 2019b (MathWorks, Natrick, Massachusetts, USA). DO2 was calculated using the following formula:
- DO2= Pump Flow (l/min)*10*(Hb (mg/dl)* 1.34*Hb saturation +0.003*arterial oxygen tension) [5]
Arterial oxygen tension on CPB is usually kept between 150 and 200 mmHg in our institution, so we assumed a saturation of 99% and an arterial oxygen tension of 150 mmHg to simplify the calculation of the DO2-values. DO2i was then calculated as shown in formula 2:
- DO i=(DO )/BSA
Body surface area (BSA) was calculated according to Du Bois‘ formula [7]. The lowest DO2i on CPB was determined with the help of MATLAB by calculating moving averages of the DO2i over 5, 10 and 20 minutes and subsequently determining the lowest value. We also calculated the cumulative and the largest area under the curve (AUC) below the DO2 threshold as first described by Mukaida et al. and Oshita et al. (see figure 1) [8,9]. As threshold of the DO2i we choose 272 ml/min/m2 [3].
Surgical Management
Cannulation for extracorporeal circulation was performed using the Seldinger technique via femoral artery and vein. In accordance with consultation between the surgeon and perfusionist, an arterial cannula (Fem-Flex II; Edwards-Lifesciences, Irvine, California, USA) with a size of 16 Fr., 18 Fr., or 20 Fr. was used, depending on the calculated blood flow. For venous cannulation, VFEM femoral cannulas (Edwards-Lifesciences, Irvine, California, USA) with a size of 22 Fr. or 24 Fr. were used. If patients had been diagnosed with a persistent foramen ovale or atrial septal defect, venous cannulation was performed with a bicaval 22/22 Fr. or 23/25 Fr. RAP cannula (Sorin Group Italy, Mirandola, Italy). The surgical approach to the heart was through a 4-5 cm right anterolateral incision in the 4th or 5th intercostal space. An optic and the aortic clamp were also introduced into the thoracic cavity through two small separate incisions. Cardioplegia was applied via a 9 Fr. aortic root cannula (MIAR; Medtronic, Dublin, Ireland) in antegrade direction. As a standard of care in our institution, 2000 ml of 4 °C cold Custodiol® solution (Dr. Franz Köhler Chemie GmbH, Bensheim, Germany) were administered via gravity to induce cardioplegic arrest for at least six minutes. Repeated application of 1000 ml Custodiol® solution was performed when longer cross clamp times were necessary according to the surgeon’s preference.
Statistical Methods
All metric variables are expressed as mean and standard deviation if normally distributed or median with 25th/75th percentiles if not. Categorical data are expressed as absolute numbers and percentages. The influence of the different variables on AKI was examined in a univariate analysis. For normally distributed data, the Welch-t-test, and for non-normally distributed data, the Mann-Whitney U test, were used. Non-normally distributed data was identified using quantile-quantile-plots. For categorical variables, Fisher’s exact test was used. Due to the retrospective and explorative nature of our study, we did not apply any method for adjusting p-values. Diagnosis of AKI was based on creatinine and urine output according to Kidney Disease: Improving Global Outcomes (KDIGO) criteria [10].
All factors that were associated with AKI in univariate analysis were entered into a multivariate logistic regression analysis to define the independent risk factors for AKI. For all tests, a p-value < 0.05 was considered statistically significant. Statis- tics were calculated using R 4.1.2 (R Core Team, R Foundation for Statistical Computing, Vienna, Austria.)
Results
Baseline Variables
The baseline characteristics of the whole population, as well as baseline characteristics of patients with and without AKI, are shown in table 1. Patients with AKI were significantly older and had a higher European System for Cardiac Operative Risk Evaluation (EuroSCORE) II.
Intraoperative Variables
All oxygen delivery-related parameters tended to be lower in patients with AKI than compared with those who did not develop AKI in the first 48 hours after surgery (see table 2). The difference in cumulative and largest AUC almost reached statistical significance. Mean and nadir hemoglobin levels during CPB were both significantly lower in patients with AKI, while peak free hemoglobin was higher. In the AKI group significantly more patients were transfused with packed red blood cells (PRBC).
Postoperative Variables
Postoperative data are presented in table 3. Overall, 37 patients (12.7%) developed AKI ac- cording to KDIGO-criteria. A majority of patients reached AKI stage 1, while RRT was necessary in five patients. Ventilation time, as well as intensive care unit (ICU) and hospital stay, were significantly prolonged in patients with AKI. Additionally, the amount of PRBC transfused during ICU stay was higher in patients with AKI. Patients with AKI also had a lower urinary output and required more frequent reoperation for bleeding.
Comparison of Mitral Valve Replacement and Repair
Since univariate analysis revealed that patients with mitral valve replacement developed AKI (22 vs. 15 cases) more often than patients with mitral valve repair, we decided to do a subgroup analysis between these two interventions. As shown in table 4, patients with mitral valve replacement were older, had a higher EuroSCORE II, longer CBP time and had a lower hemoglobin and nadir DO2i on CPB than patients with mitral valve repair. The subgroup analysis further revealed that patients who developed AKI in the mitral valve replacement group had a higher EuroSCORE II and they were transfused with significantly more PRBCs intraoperatively than patients without AKI. In patients with AKI after mitral valve repair only the EuroSCORE II was significantly higher compared with patients without AKI. In both groups, patients with AKI tended to be older. Patients with mitral valve replacement who developed AKI after surgery had a longer CPB-time, which was almost statistically significant.
Logistic Regression Analysis
To determine the independent risk factors for AKI, we performed a multivariate logistic regression analysis including the risk factors identified in the univariate analysis and all stages of AKI as dependent variables.
Because there is a correlation between nadir hemoglobin during CPB and the amount of PRBCs administered, we examined these variables separately. As shown in table 5, only EuroSCORE II, reoperation for bleeding and the amount of PRBC perioperative were independently associated with AKI.
Discussion
AKI after cardiac surgery remains a major problem leading to higher morbidity and mortality in patients. In our study group, patients with AKI had longer ventilation time, longer ICU and hospital stay and a higher 30-day mortality rate than patients without AKI. These findings are well known and have already been reported and therefore emphasize the importance of reducing the incidence of AKI in order to improve outcomes after cardiac surgery [11,12]. Moreover, treatment costs for patients with AKI are higher and increased time in the ICU puts additional strain on the availability of ICU beds, especially in light of the recently worsening shortage of hospital staff. The overall incidence rate of AKI was relatively low in our study group, compared with reported incidences of CSA-AKI of up to 40% [2]. Since we also included urine output and not solely a rise in serum creatinine for calculation of AKI according to KDIGO-criteria, unrecorded cases of AKI are probably lower than in other studies, where serum creatinine alone was used for the diagnosis of AKI.
Logistic regression analysis revealed that a higher EuroSCORE II was an independent risk factor for developing postoperative AKI. The calculation of the EuroSCORE II takes variables into account that are known to be a risk factor for AKI on their own, such as age and preoperative renal impairment [13,14]. In the literature a higher EuroSCORE II has been reported to be a risk factor for AKI [15].
Packed Red Blood Cells
Other independent risk factors were reoperation for bleeding, as well as intraoperative administration of PRBCs. Since reoperation for bleeding is often accompanied by low cardiac output and hypotonia, both of which factors are known to be a risk factor for AKI alone [16,17], the result of the logistic regression analysis is not surprising and agrees with the results of other studies [18,19]. Although administration of PRBCs is described by numerous retrospective studies as a risk factor for AKI there is still an ongoing debate regarding this topic [20–22]. On the one hand, there are retrospective observational studies that showed a better outcome and a reduction of AKI with a liberal transfusion regime [23,24]. In contrast, several randomized controlled trials comparing a liberal versus a restrictive transfusion strategy failed to show superiority of a restrictive strategy [25–27]. In a review paper, Patel and Murphy pointed out that a major bias in the results of observational studies is the fact that transfused patients tended to be sicker [28]. Indeed, in our study population transfused patients had a higher EuroSCORE II (3.8 ± 3.8 vs. 1.9 ± 1.8; p < 0.001), they were older (71.2 ± 8.7 vs. 62.5 ± 11.0; p < 0.001) and had a lower eGFR preoperatively (65.1 ± 19.4 vs. 74.3 ± 16.5; p = 0.001) than non-transfused patients. Regarding this, it is possible that there is only a correlation between transfusion and AKI rather than a causal relationship. Therefore, the result of our logistic regression analysis regarding transfusion must be viewed with caution and further studies are needed to investigate the influence of transfusion on AKI.
In contrast to several other retrospective studies, the difference in DO2 during CPB between patients with and without AKI was not statistically significant and could not be identified as an independent risk factor for AKI in our group [3,5,29,30]. However, these studies investigated conventional procedures. Vandewiele et al. performed a propensity matched comparison between conventional cardiac surgery and MICS and reached a similar result, since they could not identify DO2 as a risk factor for AKI in MICS [6]. With respect to the subgroup analysis, one cannot conclude from this study whether a GDP management in MICS might prevent AKI or not. On the one hand, there is a strong trend across our study population towards reduced DO2 during CPB and AKI. On the other, if we look at the results of the subgroup analysis from patients with mitral valve replacement, DO2i in patients with and without AKI was almost the same. Since nadir DO2i was below 272 ml/min/m2, the threshold which was deter- mined by Ranucci et al. in 2005, it is therefore conceivable that the incidence of AKI could be reduced with GDP management [3]. This is supported by the fact that, to date, two randomized controlled trials have demonstrated a reduction in AKI with GDP management, with a threshold of the DO2i of 280 and 300 ml/min/m2, respectively [31,32]. However, Magruder et al. determined a lower threshold of the DO2i in their study (225 ml/min/m2) than Ranucci et al. [29]. Since nadir DO2i in our patients with mitral valve replacement and AKI was higher (236 ml/min/m2), a further increase in DO2i might not have the desired effect if the determined threshold from Magruder et al. would be sufficient for the prevention of AKI. The average CI in our study was relatively low (2.1 l/min/m2). Nevertheless, mean DO2i was 297 ml/min/m2 across all patients. This was due to the fact, that nadir hemoglobin during CPB was 9.8 g/dl. In the setting of MICS an increase in pump flow to enhance the DO2 is often not possible, due to the limited venous return of the smaller cannulas compared to conventional cardiac surgery. Maintaining a higher hemoglobin level during CPB in MICS is therefore crucial to maintain DO2 at a sufficient level. We mainly achieved this by lowering priming volume with RAP and filtration of the extra fluid of the cardioplegia with a hemoconcentrator as standard of care.
Retrospective analysis of DO2 has several biases, which makes interpretation of the results difficult, especially when we take temperature into account. Since the Hb-level of the patients usually decreases with ongoing bypass time, it is likely that the lowest DO2i occurs often at the end of CPB. At this point, rewarming of the patient is normally taking place or has already been completed. Since the temperature during the time of the lowest DO2i is not published in the studies that investigated the critical threshold of the DO2i, we do not really know at what temperature the critical threshold is valid. This may explain the different thresholds, determined by the various study groups. Another potential bias is the fact, that calculation of DO2 during CPB relies only on pump flow. In times when the heart is ejecting blood, calculation of DO2 only with pump flow leads to a miscalculation of the actual DO2. We tried to address this problem by excluding calculation of DO2 at the initiation and termination of CPB.
Limitations
Limitations of our work are the retrospective design, as well as the lack of inline monitoring for continuous measurement of DO2 during CPB. Another limitation is the absence of a flow probe distally placed, after possible shunts (recirculation line for blood samples, shunt during hemoconcentration). For these reasons, the real DO2 values were most likely lower than the calculated ones.
Conclusion
Although growing evidence is supporting the use of GDP management during CPB, we could not identify oxygen delivery in general or an oxygen delivery index threshold in particular as independent risk factor [31,32]. However, since our average of the DO2i was below the published threshold, prospective randomized trials are needed to define the effect of GDP management in MICS. We were able to show that patients with a higher predicted mortality risk were more likely to develop AKI and, due to lower preoperative hemoglobin levels, also experienced more pronounced low oxygen delivery during CPB. Since we could clearly demonstrate that there were big differences in the occurrence of AKI with respect to the performed procedure, it may be helpful for future studies to include only high-risk patients for AKI. This would result in fewer patients being needed for reliable statistical power to meet the study goals. For a sufficient DO2 in MICS it is important to limit hemodilution and preserve the hemoglobin of the patient as much as possible to compensate for the potential lower pump flow compared to conventional procedures. In patients with poor baseline conditions, it may be helpful to use a second venous cannula to achieve better venous return to compensate for the lower hemoglobin levels with a higher pump flow.
Ethics
This study was approved by the responsible local ethics committee (Medical Faculty of the Eberhard-Karls-University and the University Hospital Tübingen) and complies with the Declaration of Helsinki (number of ethics registration: 729/2020BO2). Written informed patient consent was waived due to the retrospective nature of the study and analysis of routine inpatient data.
Conflict of Interest
The authors declare no conflict of interest.