This was a multicenter, randomized, controlled trial of IV SB vs. IV normal saline in high risk adult patients requiring cardiovascular surgery to test the hypothesis that SB would be protective against AKI post-operatively. We developed our inclusion criteria based on studies describing patients at risk for developing severe renal dysfunction after requiring cardiovascular surgery (Higgins et al., 1992; Andersson et al., 1993; Chertow et al., 1997, 1998; Fortescue et al., 2000; Thakar et al., 2005; Brown et al., 2007). Inclusion criteria included planned receipt of cardiovascular surgery as well as increased risk factors for AKI post-operatively, including either an estimated GFR of <60 ml/min/1.73 meters squared or the combination of age greater than 70 years and complex surgery (as defined in Table 1). Exclusion criteria included eGFR of <15 ml/min, emergency surgery, distal aortic surgery, and pre-operative bicarbonate level >30 mEq/l.

Patients undergoing cardiovascular surgery who met the inclusion and exclusion criteria were identified pre-operatively by study personnel who explained the study protocol and obtained written informed consent for the trial. Patients eligible for the study were identified more than 12 h prior to elective or urgent surgery to allow enough time to make an informed decision regarding the trial. Qualifying patients were randomized to receive either SB or normal saline (controls) as perioperative infusions. Subjects enrolled in the trial received a weight based pre-operative bolus of the designated fluid followed post-operatively by a weight based infusion of the designated fluid for 22 h. Serum creatinine levels were obtained pre-operatively and then on a daily basis for at least 72 h post-operatively. The serum creatinine level was measured daily using the Jaffe method. Other data collected during the trial period included length of stay in the hospital and need for dialytic therapy for treatment of AKI during hospitalization. The protocol for this prospective randomized controlled multicenter trial was approved by the Wake Forest and Ohio State School of Medicine's Institutional Review Boards and registered with Clinical trials.gov (NCT0048435).

All surgical procedures were performed through a median sternotomy. Equipment used during cardiopulmonary bypass included: a roller pump system (Terumo system I) with trillium coated tubing (Medtronic), a membrane oxygenator (Affinity trillium coated, Medtronic's), and an arterial filter (351 20 micron by Medtronic). The circuit was primed with Plasmalyte A, as well as 10,000 units of unfractionated heparin, 25 mg of Mannitol, and 50 mEq SB with a final pump volume of ~700 ml. The patients were resuscitated with crystalloids (lactated ringer's or normal saline solutions), blood products or colloids (albumin 5%). The patient's maximum blood glucose levels were maintained at <190 mg/dl using insulin infusions ± boluses as deemed necessary. Throughout the procedure, the patients' anesthetic was maintained using volatile anesthetics, benzodiazepines, and narcotics (fentanyl, morphine, and methadone) as per anesthesiologists' preference. Perioperative hemodynamic management targeted either the baseline cardiac output (CO) prior to induction or a CO >2.1 l/min/m2 and a mean arterial blood pressure >60 mmHg. The intraoperative fluid management was directed by the treating anesthesiologist and documented. All clinical data was collected prospectively and entered into a computerized database by the research team.

Those patients allocated to the normal saline arm (controls) underwent a protocol consisting of an initial bolus consisting of 0.154 M NaCl (NS) at 5.0 ml/kg (or a maximum bolus dose of 500 ml to avoid excessive fluid overload) given over 15 min prior to initiation of cardio-pulmonary bypass. Following the bolus, these patients were infused with NS (0.154 M) at a rate of 1.0 ml/kg/h (or a maximum rate of infusion of 125 mL/min to avoid excessive fluid overload) for a total of 10 h, and decreased to 0.4 ml/kg/h for a remaining 12 h; essentially until the following day. Those patients allocated to the SB (NaHCO₃) arm received a 0.150 M NaHCO₃ solution prepared in a sterile water base. Bicarbonate solution was prepared by a central pharmacy at each site. Blinding of clinical staff to treatment arm was not performed as the data collected in the study was objective data. This protocol consisted of an initial bolus of 0.150 M NaHCO₃ at 5.0 ml/kg (or a maximum bolus dose of 500 ml) given prior to initiation of cardio-pulmonary bypass, followed by a 0.150 M NaHCO₃ infusion at a rate of 1.0 ml/kg/h (or a maximum rate of infusion of 125 mL/min) for a total of 10 h followed by a NaHCO₃ infusion of 0.4 ml/kg/h for another 12 h, as was done with the NS treated patients. The goal bicarbonate dose in this study was ~3.0 mEq/kg. This dose was based on what our group found to be efficacious in our previous studies (Fisher et al., 2004, 2005). The study protocol included mechanisms to adjust the bicarbonate infusion rates based on post-operative lab values. The bicarbonate infusion rate was decreased by one half (1/2) for all patients whose serum bicarbonate level was greater than or equal to 30 mmol/L post-operatively. This half dose infusion was continued until the serum bicarbonate returned to a level less than 30 mmol/L at which time the rate was increased back to the previous level. For any patient with a post-operative serum bicarbonate level greater than or equal to 35 mmol/ml the infusion rate was decreased to 10 mL per hour (KVO) until the bicarbonate level dropped to less than 30 mmol/ml at which time the infusion rate was increased back to previous levels.

This protocol consisted of ~22 h of IV study infusion. Any other IV fluid that the respective surgeons or anesthesiologists deemed necessary was allowed as part of routine patient management. For those patients felt to be volume overloaded, the randomized fluid was reduced to 10 mL per hour until the clinical conditions allow the higher rate to be resumed.

The primary outcome of the trial was to determine the rate of AKI in the study population. The study sample size was determined based on previous information in a high risk population. Historic data reported rates of ARF, defined arbitrarily as a change in SCr ≥ 1.0 mg/dL, to be ~11.7% in patients treated with sodium chloride (SC) and 2.7% in patients treated with the bicarbonate solution (Fisher et al., 2004). To observe these differences between treatment groups in ARF with 90% power and a 5% type one error rate, ~360 patients (180 in each randomized group) would need to be enrolled in the trial.

After the study was initiated, a consensus definition of AKI was introduced in 2004 (RIFLE) and subsequently modified based on the AKIN staging system (Mehta et al., 2007). This system defines stage 1 of AKI as a 1.5 fold or ≥0.3 mg/dl increase in the serum creatinine level from baseline. After the initiation of the trial the AKIN staging system had become a clearly defined reference for AKI and in consultation with our DSMB, but prior to data analysis, the AKIN definition of stage I AKI was chosen as the primary means to determine the presence of AKI. For analytic purposes, we evaluated our results based on three definitions of ARF: ARF defined as an increase in serum creatinine (SCr) by 1 mg/dl in 72 h, ARF (AKI) defined as an increase of 25% in SCr, and ARF (AKI) defined as an increase in SCr of 0.3 mg/dl.

Secondary endpoints of interest include the maximum change in estimated serum creatinine and the glomerular filtration rate (GFR) during the first 72 h post-operatively, LOS, incidence of dialysis (HD) and mortality.

At the interim time, we tested the hypothesis of a difference in the proportion of patients with any AKI between treatment groups, as described by the AKIN staging system (AKI as a 1.5 fold or ≥0.3 mg/dl increase in serum creatinine from baseline). Data from 120 patients was available for analysis (2 patients did not have post-surgery creatinine measurements recorded at OSU—one in each treatment arm). AKI was observed in 31 (26%) of patients, with nearly equal numbers in each randomized group [14 in the normal saline (24%), 17 in the SB group (27%)], resulting in an interim test p-value (chi-square test) of 0.95. Given this information, we could clearly state that early efficacy was not met. In fact based on this information, conditional power based on the effect size determined with the 120 patients accrued, was estimated to be only 12%. Given this very low estimated conditional power, we, along with our data and safety monitoring committee, recommended closing the trial and reporting on the current study.

The primary analysis of AKI rates between treatment groups was evaluated by the chi-squared test and included estimates of the relative risk and associated 95% confidence intervals. Secondary outcomes included length of stay and maximum change in GFR which were compared by a two-sample t-test (without assuming equal variances) or by the chi-square test for categorical outcomes such as mortality and incidence of dialysis. All tests were two-sided and considered significant at the 0.05 level. Analyses and data management were performed using SAS software, Version 9.2 of the SAS System. Copywrite, SAS Institute Inc., Cary, NC, USA.

Patients were recruited to the study between August 2006 and February 2011. Patient characteristics of the study population are shown in Table 2. The average patient age was ~70 years and the majority of the participants were Caucasians. The type of surgery performed included coronary artery bypass grafting (alone or in combination with valvular surgery), valvular surgery, aortic (ascending and aortic arch) as well as placement of ventricular assist devices.

Due to slow enrollment, the study team and independent clinical ethicist agreed to perform an interim analysis once 120 subjects were enrolled to determine if treatment efficacy had been met. A total of 123 patients were enrolled between the two centers to either receive intravenous SB (n = 64) or SC (n = 59) infusions. As described earlier, accrual to this study was stopped early due to no evidence of efficacy (See Interim analysis). Data from 120 patients were available for analysis (2 patients did not have post-surgery creatinine measurements recorded at OSU;–one in each treatment arm). One patient suffered from infiltration of SB around the peripheral intravenous access site. The infiltration was treated, and the patient dropped out of the study.

A similar number of patients in the two treatment groups developed AKI, defined as an increase in serum creatinine the first 48 h after surgery of 0.3 mg/dl or more. Specifically 14 patients (24%) who received SC and 17 patients (27%) who received SB were observed to develop AKI post-surgery, resulting in a relative risk of AKI of 1.1 (95% CI: 0.6–2.1, chi-square p-value = 0.68) for patients receiving SB compared to those who received SC. AKI in the first 72 h occurred in 42 patients [22 patients (38%) in the SC group and 20 patients (32%) in the SB group] resulting in an estimated relative risk of 0.9 (95% CI: 0.5–1.4). AKI defined as an increase of 25% or more in serum creatinine (AKI25) in the first 24 or 48 h was also not substantially different between groups (Table 3). Moreover, ARF defined in the original study design, was not improved post-surgery by the use of SB. ARF was not frequently observed; only 3 patients (5%) in the SC group and 7 patients (11%) in the SB group had ARF (RR = 2.2; 95% CI: 0.6–8.0). A sub analysis of patients who had contrast during surgery, did not provided a substantial change to these conclusions concerning differences in AKI or ARF post-surgery between these groups.

We observed a similar number of deaths (10 in the SC group and 9 in the SB group) and patients in need for post-operative dialysis (5 in the SC group and 4 in the SB group) in both groups. Further, the number of days spend in the hospital did not differ substantially between the treatment groups with an average of 14.2 days [standard deviation (sd) = 15.4] in SB patients and 14.6 days (sd = 14.2) in SC patients (t-test assuming unequal variances p-value = 0.88).

This study was funded, in part, by a grant for the Southeastern Kidney Council.