Hyperchloremia and postoperative acute kidney injury: a retrospective analysis of data from the surgical intensive care unit

This retrospective observational study was approved by the Institutional Review Board (IRB) of the Seoul National University Bundang Hospital (SNUBH) (approval number B-1806/474-105). The requirement for written informed consent was waived by the IRB. The study included all adult patients (≥ 19 years old) admitted to SNUBH between 1 January 2011 and 30 June 2016, who were admitted to the surgical ICU after undergoing surgical procedures. If a patient was admitted to the surgical ICU more than once during the study period, only the last postoperative ICU admission was considered. The exclusion criteria were as follows: (1) lack of accurate records of main perioperative laboratory investigations; (2) death within 72 h of postoperative ICU admission; (3) preoperative AKI; or (4) receiving chronic renal replacement therapy before surgery.

As of August 2017, SNUBH is a 1360-bed tertiary care hospital with five ICUs (medical, surgical, neurologic, and emergency I and II). During the study period, surgical ICU admission was indicated according to the complexity of the surgery and the severity of the patient’s condition. The decision on ICU admission was made by the intensivist, who was the main researcher (IAS) in the present study.

For the diagnosis of postoperative AKI, we employed the criteria and grading system laid out in the Kidney Disease: Improving Global Outcomes (KDIGO) guidelines [10] because the KDIGO classification is recognized as the most adequate tool for this purpose [11, 12]. Specifically, AKI diagnosis was based only on creatinine levels (Additional file 1). Creatinine levels were obtained by venous measurement; the most recent measurement obtained within 1 month before surgery was considered as the baseline value. Postoperative AKI was defined as AKI diagnosed within 3 days postoperatively. During the study period, postoperative AKI was diagnosed by certified nephrologists; for the purpose of this study, two certified intensivists (TKO and IAS) reviewed all AKI diagnoses to confirm postoperative AKI. If there were disagreements between the two certified intensivists, the final decision was made through consultation with the nephrologists.

Demographic, clinical characteristics, and laboratory test results were obtained by retrospective review of medical charts. At SNUBH, samples for measuring serum creatinine and electrolyte levels are collected during routine preoperative laboratory testing. For the purpose of this study, the baseline creatinine and electrolyte (chloride) levels were defined as the levels measured most recently in blood sampled from the vein within 1 month before surgery. For the diagnosis of postoperative AKI, creatinine levels were measured in samples collected on postoperative days (PODs) 0–3. To identify patients with CKD, we obtained the preoperative estimated glomerular filtration rate (eGFR, mL·min^− 1·1.73·m^− 2) using the Modification of Diet in Renal Disease formula [13]. We recorded information on perioperative fluid management, including the type and dosage of fluid infused: NaCl 0.9% (mL·kg^− 1), NaCl 0.45% (mL·kg^− 1), balanced electrolyte solution (Ringer’s lactate or Plasmalyte; mL·kg^− 1), colloid (hydroxyethyl starch; mL·kg^− 1), red blood cell infusion (packs), and free water-containing dextrose (mL·kg^− 1). The use of relevant drugs during PODs 0–3 was also recorded: inotropes or vasopressors (norepinephrine, vasopressin, dopamine, dobutamine, epinephrine), diuretics (mannitol, furosemide), radiocontrast agents, nephrotoxic antibiotics (aminoglycoside, cephalosporin, vancomycin, and sulfonamide), and nonsteroidal anti-inflammatory drugs. Data on invasive and non-invasive measurement of intraoperative blood pressure were collected, and intraoperative hypotension was defined as intraoperative mean blood pressure < 60 mmHg for > 1 min. Intraoperative fluid balance (percent) was calculated using the following formula: [total input fluid (L) – total output fluid (L)] × 100 × [weight on admission (kg)]^− 1.

As described previously [9], hyperchloremia was defined as postoperative Cl⁻ > 110 mmol·L^− 1 measured at least once during PODs 0–3. Exposure to hyperchloremia on the day of surgery (POD 0) was recorded as an additional independent variable. The largest (maximum) value of Cl⁻ measured during PODs 0–3 was also retained. The increase in Cl⁻ was computed as follows the maximum Cl⁻ or sodium levels during PODs 0–3 minus the preoperative Cl⁻. Metabolic acidosis was diagnosed in patients with simultaneous presence of pH < 7.35 and HCO₃⁻ < 24 mEq·L^− 1 on arterial blood gas analysis during PODs 0–3.

SNUBH medical record technicians blinded to the study goals collected all electronic medical record data. During data collection (i.e., until the statistical analysis was performed), the main researchers were blinded to data on Cl⁻, which represent the most important data from the perspective of this study.

Patients were stratified into four groups according to baseline eGFR (≥ 90, 60–89, 30–60, and < 30 mL·min^− 1·1.73·m^− 2), incidence of hyperchloremia during PODs 0–3, increase in serum Cl⁻ (quartiles Q1–Q4) during PODs 0–3, and incidence of AKI (yes/no) during PODs 0–3, as explained in detail subsequently. Continuous variables were analyzed using the t test, while categorical variables were analyzed using the chi-square test.

We tested the potential association between postoperative AKI and exposure to perioperative hyperchloremia and increase in Cl⁻. Restricted cubic splines (RCS) were used to analyze the relationship between the probability of postoperative AKI and either maximum Cl⁻ or increase in Cl⁻ perioperatively. Since the relationship between the probability of occurrence of postoperative AKI and the increase in Cl⁻ was not linear and the third quartile (Q3) of the distribution of values for the increase in Cl⁻ was close to an inflection point where the slope of the RCS graph changed, the distribution of values of increase in Cl⁻ was divided into quartiles, and the increase in Cl⁻ was analyzed as a categorical variable.

Next, we performed univariable logistic regression analysis for occurrence of postoperative AKI during PODs 0–3. Only covariates with P < 0.2 on univariable analysis were included in the multivariable model, with two types of main independent variables (perioperative hyperchloremia as binary exposure; perioperative increase in Cl⁻ as continuous exposure). We generated separate multivariable models for exposure to hyperchloremia and for increase in Cl⁻ during PODs 0–3 to avoid multi-collinearity between the two models.

Because CKD is a well-known risk factor for AKI [14], we investigated the interaction between preoperative eGFR (normal kidney function, > 90 mL·min^− 1·1.73·m^− 2; CKD stage 2, 60–89 mL·min^− 1·1.73·m^− 2; CKD stage 3, 30–60 mL·min^− 1·1.73·m^− 2; CKD stage 4 or 5, < 30 mL·min^− 1·1.73·m^− 2) and the two types of main independent variables (hyperchloremia as binary exposure; increase in Cl⁻ as continuous exposure). When interaction between preoperative eGFR and the main independent variables was noted, we performed three additional subgroup analyses. In these subgroup analyses, we performed Bonferroni correction to control type 1 error, and a Bonferroni-corrected P value <0.013 was considered to indicate statistical significance. The same method of analysis as described above was applied for postoperative AKI stage ≥ 2 as a dependent variable. For all multivariable models, the goodness of fit was confirmed using the Hosmer-Lemeshow test.

Additionally, Pearson correlation analysis was performed to evaluate the simple relationship between Cl⁻ and infusion of various fluids during PODs 0–3. All analyses were performed using IBM SPSS (version 24.0; IBM Corp., Armonk, NY, USA) and R (version 3.3.3; various R packages; http://​www.​r-project.​org), with statistical significance set at P < 0.05 for group analyses and at a Bonferroni-corrected P value <0.013 for subgroup analyses.

Between 1 January 2011 and 30 June 2016, 12,746 patients were admitted to the ICU following surgery. Of these, 3854 postoperative ICU admissions were excluded because they represented multiple admissions; only the last ICU admission was considered for each patient. Additionally, 182 patients were excluded due to incomplete medical records, 358 patients due to preoperative AKI, 195 patients due to death within 72 h of ICU admission, and 166 due to receiving chronic renal replacement therapy before surgery. In total, 7991 patients were included in the final analysis. Of these, 1187 patients (14.9%) were diagnosed with postoperative AKI, of whom 186 (2.3%) exhibited AKI of stage ≥ 2 (Fig. 1). Table 1 summarizes the differences between patients with hyperchloremia and those with non-hyperchloremia during PODs 0–3. The two groups did not differ significantly in the incidence of postoperative AKI (289/1876, 15.4% vs 898/6115, 14.7%; P = 0.443) or AKI of stage ≥ 2 (51/1876, 2.7% vs 135/6276, 2.2%; P = 0.199).

The RCS describing the relationship between postoperative AKI and maximum Cl⁻ are provided in Additional file 2 (panel A), while the results of the univariable logistic regression analyses are presented in Additional files 3 and 4. The covariates with P < 0.2 on the univariable analysis were entered into the multivariable logistic regression model, and the results of multivariable analysis based on model 1 are provided in Table 2. In the overall sample (7991 patients), exposure to hyperchloremia was not associated with postoperative AKI (odds ratio (OR), 1.09; 95% confidence interval (CI), 0.80–1.49, P = 0.571; model 1) or with AKI of stage ≥ 2 (OR, 0.77; 95% CI 0.40–1.49; P = 0.437; model 2).

The RCS describing the relationship between postoperative AKI and increase in Cl⁻ are provided in Additional file 1 (panel B). Based on the shape of the RCS curve, the distribution of values for the increase in Cl⁻ was divided into quartiles: Q1, increase in Cl⁻ ≤ 1 mmol·L^− 1, 2535 patients (31.7%); Q2, increase in Cl⁻ of 1–3 mmol·L^− 1, 1593 patients (20.0%); Q3, increase in Cl⁻ of 3–6 mmol·L^− 1, 2075 patients (26.0%); and Q4, increase in Cl⁻ > 6 mmol·L^− 1, 1788 patients (22.4%).

The results of the multivariable regression analysis of increase in Cl⁻ are provided in Table 3. The overall P value for the 4-level factor (Q1–Q4) was 0.756. However, there was interaction between perioperative Cl⁻ increase and preoperative kidney function (Q1 * eGFR ≥ 90 mL·min^− 1·1.73·m^− 2 and Q4 * eGFR < 30 mL·min^− 1·1.73·m^− 2), and thus subgroup analyses were performed for each predefined eGFR threshold (≥ 90, < 90, < 60, and < 30 mL·min^− 1·1.73·m^− 2). On subgroup analysis, a Cl⁻ increase > 6 mmol·L^− 1 was associated with postoperative AKI in patients with CKD stage ≥ 3 (eGFR < 60 mL·min^− 1·1.73·m^− 2) (OR, 1.42; 95% CI, 1.09–1.84; P = 0.009 vs a Cl⁻ increase ≤ 1 mmol·L^− 1). In addition, there was no significant association between the increase in Cl⁻ and occurrence of postoperative AKI of stage ≥ 2, and there was no significant interaction for AKI stage ≥ 2 between perioperative Cl⁻ increase and preoperative kidney function (Table 4).