The study was approved by the French ethical committee for human subjects (2017-A01324-49). All patients have received a written document for research purposes and gave their consent prior to participation. The study was performed in accordance with the ethics standards laid down in the 1964 Declaration of Helsinki. The manuscript was drafted according to the STROBE guidelines.

We performed a post hoc analysis of a multicentric study comprising four tertiary university hospitals (Dijon, Amiens, Rouen, and Strasbourg, France), including adult patients who underwent cardiac surgery with CPB between January 2017 and December 2021 (Guinot et al., 2023b).

The inclusion criteria were as follows: patient age ≥ 18 years, following angiotensin-converting enzyme inhibitor treatment, cardiac surgery with the use of CPB: coronary artery bypass graft (CABG), the surgical correction of valve disease (aortic, mitral, tricuspid, and pulmonary), combined surgery, ascending aortic disease, and others.

The non-inclusion criterion was off-pump cardiac surgery. The exclusion criteria were left ventricular assist devices and heart transplantation.

All data were anonymously extracted from each center’s institutional database. The following baseline patient characteristics were prospectively collected: demography (age, gender, and BMI) and comorbidities (arteriopathy, stroke, chronic obstructive pulmonary disease, and atrial fibrillation). Cardiovascular risk factors include the type of surgery (valvular surgery, CABG, combined surgery, aortic disease, and other), the presence of a critical preoperative condition, the American Society of Anesthesiology (ASA) score, and special operative status (emergency, redux, and endocarditis). Intraoperative data were also collected. EuroSCORE II (European System for Cardiac Operative Risk Evaluation) was calculated for each patient.

AKI was defined as an increase in serum creatinine greater than 27 μmol/L compared with preoperative baseline or urine output less than 0.5 mL kg^-1∙h^-1 according to the KDIGO (Kidney disease improving global outcomes) definition (Kellum et al., 2013). Cardiogenic shock was defined as any inotropic drug requirement to obtain hemodynamic stability with left and/or right ventricular dysfunction. Postoperative atrial fibrillation was defined as a new onset of irregularly irregular heart rate in the absence of P-waves lasting at least 30 s or for the duration of the ECG recording (if < 30 s) during the postoperative period (Campbell et al., 2022). Patients were continuously monitored for the first 48–72 postoperative hours. A new episode was defined as any episode of atrial fibrillation in patients naïve to preoperative atrial fibrillation.

The patients were managed based on local protocols, according to national and international guidelines. Following surgery, all patients were monitored in the intensive care unit and managed by a specialized team, which included a cardiologist trained in the care of postoperative cardiac surgery. Hemodynamic support was guided by institutional protocols to achieve mean arterial pressure > 65 mmHg, cardiac index > 2.2 L min^-1∙m^-2, and urine output > 0.5 mL kg^-1∙h^-1. They also benefited from permanent monitoring of the electrocardiogram, oxygen saturation, and central venous pressure. Biological monitoring was performed several times a day at a frequency determined by the attending physician.

A sample size of 564 patients can demonstrate an absolute difference of 5% of AKI between the two groups (for an incidence of AKI of 25%) with a risk alpha of 0.05 and a power of 0.8. Normality was assessed visually using histograms and with the Shapiro–Wilk test. The quantitative variables were presented as mean ± standard deviation (SD) or as median [25%–75% interquartile range], as appropriate. The qualitative variables were presented as numbers (percentages).

We used a combined statistical approach to assess the association between preoperative ACE inhibitor discontinuation and postoperative AKI. First, we performed propensity score matching using a nearest-neighbor algorithm with 1:1 matching without replacement and a caliper distance of less than 0.2 of the standard deviation of the logit of the propensity score, to reduce the influence of potential confounders between the two groups. Only complete datasets were used. Seventeen covariates were selected and entered: center, age, dyslipidemia, coronary artery disease, smoking, renal function, preoperative atrial fibrillation, statin, metformin, number of procedures, valvular surgery, length of aortic clamping, glucose–insulin–potassium, hypothermia, del Nido cardioplegia, custodial cardioplegia, and perioperative blood transfusion. The choice of these variables was performed considering the already validated prognostic scores in perioperative setting (EuroSCORE II) and the clinically pertinent and non-redundant status with a higher baseline disequilibrium between the two groups. The baseline disequilibrium and the adequacy of covariate balance in the matched sample were assessed using absolute standardized mean differences (mean difference expressed in units of SD). In the matched cohort, the comparison regarding the primary endpoint between the two groups of treatment was performed using the McNemar test. The comparisons regarding the secondary endpoints between the two groups were performed using the McNemar test or the Wilcoxon signed-rank test, as appropriate. Because we tested multiple outcomes, the threshold for statistical significance was corrected by using a Bonferroni correction and set at p < 0.006 (0.05 divided by 8, the number of assessed outcomes). Second, we performed a sensitivity analysis using a logistical model approach and an unmatched cohort. We performed a univariable logistical model with the apparition of postoperative AKI as a dependent variable and each preoperative factor as an independent variable. To take in account the effect of the center, we include the center in the multivariate logistic regression model. Non-redundant and clinically pertinent variables with a p-value <0.2 in univariate analyses were introduced in a multivariable logistic regression model. The association between risk factors and AKI was evaluated by the odds ratio (OR) and its 95% confidence interval. The calibration of the final logistic model was assessed using the Hosmer–Lemeshow statistic. The validity conditions for the logistic regression were checked in order to have at least 5–10 events for each independent variable in the multivariable model. All analyses were performed using R software version 3.4.4 (R Foundation for Statistical Computing, Wien, Austria). We used the “MatchIt” package for propensity score matching.

In total, 3,000 patients underwent cardiac surgery with CPB during the study period, of whom 1,072 were observing ACE inhibitor therapy. One-to-one propensity score matching has led to two matched cohorts of 283 patients (Figure 1).

The mean age was 68.6 years (SD 9.4) in the ACE inhibitor continuation group and 68.7 years (SD 9.42) in the ACE inhibitor withholding group. The mean left ventricular ejection function (LVEF) was 55% (SD 12.5 and 11.9) in both groups. The mean estimated eGFRs were 78 mL∙min^-1∙1.73m^-2 (SD 25.8) in the ACE inhibitor continuation group and 79 mL∙min^-1∙1.73m^-2 (SD 25.7) in the ACE inhibitor withholding group. The groups were considered well-balanced after matching for several variables (Table 1).

Seventy-one (25.1%) patients developed postoperative AKI in the withholding ACE inhibitor group and sixty-eight (24%) in the ACE inhibitor maintenance group, which was not a statistically significant difference (p = 0.843). The postoperative outcomes are presented in Tables 2, 3 and Figure 2, according to ACE inhibitor group management.

We did not observe a statistically significant difference in terms of mortality (3.18% vs 4.24%, p = 0.646). In addition, there was no statistically significant difference in postoperative atrial fibrillation (24.4% vs 24.0%, p = 1) and postoperative cardiogenic shock (15.9% vs 20.5%, p = 0.184), despite the group. The incidence of severe hypotension requiring vasopressor use was similar between the two groups (72,4% vs. 67,1%, p = 0.214). Finally, the ICU and hospital length of stays did not differ between the two groups (3 [2; 5] vs. 3 [2; 5] days, p = 0.963 and 9.5 [8; 12] vs. 10 [8; 14] days, p = 0.638).

The continuation of the ACE inhibitor was not associated with postoperative AKI in univariable logistic regression (OR: 1.2 (0.9–1.5), p = 0.246). The logistic regression results regarding the factors associated with postoperative AKI are presented in Figure 3. These factors comprise age, high blood pressure, atrial fibrillation, diabetes on insulin, critical state, LVEF, estimated eGFR, P2Y12 inhibitor use, endocarditis, number of procedures, only valvular surgery, only CABG surgery, length of CPB time, norepinephrine use, and intraoperative blood transfusion.

Because of its retrospective nature, we do not know the dosage and the type of molecule used. The use of the KDIGO definition, which gives a central place to serum creatinine in the definition of AKI, is not necessarily a reflection of structural renal damage (Coca et al., 2014). Cola et al. had demonstrated that despite an increase in postoperative creatinine levels, urinary biomarkers of tubular damage did not increase in patients who continued ACE inhibitors. It seems legitimate to ask whether the follow-up period is sufficient. Finally, it might be interesting to study the rate of long-term ACE inhibitor treatment in cases of temporary withdrawal for cardiac surgery. Indeed, analysis of the data from Antoniak’s study on American veterans shows a lower rate of long-term ACE inhibitor treatment in cases of temporary withdrawal, which could have a negative impact on the prognosis of patients (Antoniak et al., 2021).