Associations between tricuspid annular plane systolic excursion to reflect right ventricular function and acute kidney injury in critically ill patients: a SICS-I sub-study

This is a sub-study of the Simple Intensive Care Studies I (SICS-I), a single-center, prospective observational study designed to evaluate the diagnostic and prognostic value of combinations of clinical and hemodynamic variables in critically ill patients (NCT02912624) [20, 21]. This sub-study, which we will from now on refer to as the current study, was implemented in the SICS-I on February 10, 2016. The local institutional review board (Medisch Ethische Toetsingscommissie of the University Medical Center Groningen; UMCG) approved the study (M15.168207). We report our study in adherence to STROBE guideline (Additional file 1: Table S1) [22].

All acutely admitted patients of 18 years and older with an expected ICU stay of at least 24 h were eligible for inclusion. Exclusion criteria were discharge within 24 h and absence of informed consent. In the present study, we further excluded patients in which we were unable to obtain measurements of right ventricular function and patients with history of chronic kidney disease or dialysis prior to admission. Study size was dependent on patients included within SICS-I and amount of available data.

We registered patient characteristics at admission, including the APACHE IV scores and use of vasopressors and/or inotropes according to our protocol [23]. Variables necessary for the diagnosis (and severity) of AKI were recorded up until 72 h after admission. The hemodynamic variables were measured once (within first 24 h of ICU admission) at the earliest possibility. We prospectively recorded data from physical examination, including heart rate, systolic blood pressures (SBP), diastolic blood pressures (DBP), central venous pressure (CVP), mean arterial pressure (MAP) and central temperature. We recorded several CCUS variables (Vivid-S6, GE Healthcare, London, UK) including cardiac index, tricuspid annular plane systolic excursion (TAPSE) and right ventricular systolic excursion (RV S’). Cardiac index was estimated by the velocity time integral of left ventricular outflow tract times heart rate times left ventricular outflow tract area (cardiac output = VTI × HR × LVOT area), divided by body surface area. All measurements were conducted by research-students, who were not involved in patient care. All researchers were trained by experienced cardiologist-intensivists to conduct a focused CCUS before contributing to the study. Training included studying of theory, practice on healthy individuals and lastly supervised CCUS in patients. The first 20 examinations were supervised by a senior researcher. The images and measurements of cardiac index were validated and remeasured by an independent core laboratory (Groningen Image Core Lab, UMCG, Groningen, the Netherlands) for purposes of the main study.

We defined AKI and its severity according to the KDIGO criteria based on serum creatinine, urinary output recorded cumulative every 24 h and use of RRT [24]. The modification of diet in renal disease (MDRD) formula was used for estimation of the ideal serum creatinine for each individual as baseline assuming a creatinine clearance of 75 ml/min/m², as suggested by [25, 26]. Chronic kidney disease was defined by serum creatinine above 177 µmol/L (following the definition used by the ‘Nationale Intensive Care Evaluatie’) [27]. Cutoff values for cardiac function were below 2.2 L/min/m² for low cardiac index, below 17 mm for low TAPSE and below 9.5 cm/s as low RV S’ [28]. Patient outcome variables were the occurrence and maximum severity of AKI within the first 72 h of ICU stay, number of days alive outside ICU and 90-day mortality.

To assess selection bias we compared baseline characteristics between included and excluded patients. We aimed to minimize possible misclassification bias by providing comprehensive protocols and training for involved researchers.

The overall statistical methods were described in the pre-defined statistical analysis plan of SICS-I (NCT02912624). Continuous variables were reported as means [with standard deviations (SD)] or medians [with interquartile ranges (IQR)] depending on distribution. Categorical data were presented in proportions. Associations were calculated as odds ratios (OR) with 95% confidence intervals (CI). Student’s T test, Mann–Whitney U test or the Chi-square tests were used as appropriate. A two-sided p value of ˂0.05 was considered statistically significant. Associations of hemodynamic variables with the occurrence of AKI were first explored by univariate analysis. We used univariate associations with p < 0.1 for entrance of a variable into the multivariable model, while the predisposing variables age and sepsis were included in the model based on the literature, irrespective of their p values in the univariate analyses. No missing data were imputed, and a multivariate analysis was performed based on available original data, except for variables with more than 50% missing values, which were excluded from the model [23]. Variables were assessed for collinearity before entrance in the multivariable model. If two variables were strongly correlated, only one variable was included in the multivariate model based on the strongest univariate association. The final model was based on logistic regression analysis to identify the variables independently associated with the development of AKI. Discrimination (to distinguish AKI patients from non-AKI patients) of the final model was evaluated with receiver operating characteristic (ROC) curves. Calibration of the multivariable model (i.e., the number of AKI cases that the model predicts correctly across different risk groups) was checked with the Hosmer–Lemeshow test and by plotting observed AKI proportions against predicted risks of 10 equally sized groups. Analyses were performed using Stata, version 15 (StataCorp, College Station, TX, USA).

All hemodynamic variables were statistically significantly associated with AKI in the univariate analysis (Additional file 1: Table S3). CVP was missing in over 50% of cases (measured in only 112 patients) and thus was not included in the final multivariate model. Due to collinearity between MAP and systolic and diastolic blood pressures, we only included MAP in the model. No further collinearity was observed. All remaining variables with p < 0.1 in univariate analysis were entered into the adjusted model.

In multivariate analysis, positive end-expiratory pressure (PEEP), cardiac index, RV S’, vasopressors and central temperature were shown to have a p > 0.1 and were removed from the model (p = 0.95, p = 0.71, p = 0.19, p = 0.17 and p = 0.10, respectively). In the final model heart rate was not significantly associated with AKI. Higher age, higher weight, increased APACHE IV scores, lower MAP and lower TAPSE were all statistically significantly associated with AKI (Table 2). The receiver operating characteristic (ROC) curve of the final model yielded an area under the curve of 0.75 (95% CI 0.73–0.80), pseudo-R² = 0.15. Hosmer–Lemeshow goodness-of-fit test showed a χ² 10.04; p = 0.2623.

A sensitivity analysis was performed with the final model and CVP in patients in which this was available (n = 94), which showed no statistically significant association for CVP (OR 1.08; 95% CI 0.99–1.20, p = 0.098) or lower TAPSE (OR 1.04; 95% CI 0.96–1.13, p = 0.252) with AKI (Table 3).

Singular proxies of venous hemodynamic function have previously been associated with the development of AKI [18, 29‐33]. Chen et al. found that peripheral edema was associated with a 30% higher risk of AKI in 2238 critically ill patients. In the same article the authors showed that each cm H₂O higher CVP was associated with a 2% higher adjusted risk of AKI in 4761 critically ill patients [17]. We assessed right ventricular function measured by TAPSE and confirmed these observations of the potential role of central venous hemodynamic function, assuming that venous congestion, with CVP as a proxy, leads to impaired right ventricular function, and vice versa. In the sensitivity analysis including CVP the point estimates were similar, but in a relatively small sample size these effects did not reach statistical significance. Right ventricular function has also been suggested to be associated with the development of AKI by data from patients with heart failure by Mullens et al. [34] and pulmonary hypertension by Haddad et al. [35]. Guven et al. [36] described that preoperative right ventricular function was associated with an increased risk and severity of AKI in 595 heart transplant patients. To our best knowledge, in no studies these measurements were performed prospectively in an unselected population of the critically ill. We evaluated two measures of right ventricular function: TAPSE and RV S’. Huang et al. [37] elegantly described that most studies use these as indices of RV function, but evaluation of isolated TAPSE is potentially misleading. Other indices of RV function or for example inferior vena cava measurements may also reflect venous congestion and be associated with AKI [38]. Fluid balance may influence right ventricular function and thus interact with the association with AKI [39]. The optimal variable representing venous hemodynamic function for evaluation of association with AKI in the critically ill is to be elucidated. A combination of previously mentioned variables might show a higher predictive value for venous congestion and AKI in critically ill patients compared to TAPSE alone.

Only limited data specifically address the association between cardiac index and AKI. Many studies focus on the association between (septic) shock and AKI [40, 41], while data on the detailed association between cardiac index, MAP and renal blood flow in critically ill patients in general and AKI are sparse [42, 43]. In a retrospective cohort study of 1879 critically ill patients, both impaired left and right ventricular function were associated with AKI and increased mortality [44]. One study included 10 patients with sepsis and AKI and suggested that the renal perfusion fraction of cardiac output measured by phase contrast magnetic resonance imaging was very low (7.1%) [45]. One hypothesis is that AKI develops after ischemic and hypoxic damage due to restricted blood flow, possibly caused or exaggerated by a reduced cardiac index. A higher cardiac index may be protective for AKI, or reverse, that a low cardiac index would be associated with increased incidences of AKI, although other mechanisms may be involved. We found no association between cardiac index and AKI, while a lower MAP appeared to be associated with AKI. Findings from a previous prospective observational study evaluating associations between hemodynamic variables and the progression of AKI in septic patients suggested that a lower MAP is associated with AKI [40].

Similar to other studies, higher PEEP showed association with AKI; however, in the multivariate model PEEP did not remain significant [46, 47]. The fact that PEEP in this study did not contribute enough to the final model might be explained by the relatively low PEEP levels and the regularly used lower tidal volumes (for the purposes of lung-protective ventilation) leading to substantial lower intrathoracic pressures compared with the mentioned studies. Moreover, only 57% of our population was mechanically ventilated, potentially leading to a bias in our multivariate analysis in terms of this variable, and we did not evaluate other ventilation settings such as plateau pressure.

We conducted a single-center study. Collaboration with other centers and other ICUs will increase generalizability. We included a broad population of acutely admitted critically ill patients, while investigation of hemodynamic variables in subgroups might yield more specific clues of the role of hemodynamic variables in the pathophysiology of AKI according to diagnosis, e.g., sepsis or shock patients. Our observations encourage to elaborate specifically on the role of venous hemodynamic function in the development of AKI and to search for a practical proxy, or perhaps a combination of multiple variables, for assessment of venous congestion. This is also in line with the recently proposed research agenda for AKI [48].

Several limitations of our study must be acknowledged. First, our study is a prospective observational study, which hampers causal inferences. As it is a data-driven analysis, results need to be confirmed in other studies. Second, we collected kidney function up to 72 h, i.e., early AKI. Therefore, the incidence and severity of AKI may both have deteriorated or improved after 72 h. It has been suggested that the incidence of AKI is highest during the initial days of ICU stay [49], although continued fluid administration during or after early AKI might prevent recovery of kidney function [50]. Variables identified as associated with AKI might vary with the timing of assessment of variables and diagnosing AKI [51], so that longer follow-up will probably lead to identification of another set of variables associated with AKI. Third, CVP was measured in fewer patients and had to be excluded from multivariate analysis in this cohort. This is an important limitation, as CVP may serve as one of the proxies for venous congestion and right ventricular function. This, however, does not influence the conclusion that TAPSE was associated with AKI, whereas cardiac index was not. Fourth, we used the MDRD formula to estimate baseline serum creatinine since these data were not collected in the main study. Even though this method is widely accepted, this might have led to either an over- or underestimation of AKI incidence [52, 53]. Fifth, we used serum creatinine as a marker of kidney function. Despite recent observations on the accuracy of serum creatinine [26], serum creatinine is still a surrogate for kidney function. Also, creatinine might be biased by interventions during ICU stay, such as hemodilution due to a positive fluid balance. Last, 21% of the patients were excluded from analyses due to inadequate quality of CCUS images which may be difficult to obtain in ICU patients [54]. Weight and mechanical ventilation were significantly different in this group, which are factors that often hamper transthoracic CCUS. The use of vasopressors was also more prevalent in the excluded group, but as APACHE IV score was not significantly different, we assumed that the groups were similar. We thus do not expect that this has influenced the findings of this study.