Increased serum bicarbonate in critically ill patients: a retrospective analysis

Transcription

1 Intensive Care Med (2015) 41: DOI /s ORIGINAL Alexandre Braga Libório Danilo Teixeira Noritomi Tacyano Tavares Leite Candice Torres de Melo Bezerra Evandro Rodrigues de Faria John A. Kellum Increased serum bicarbonate in critically ill patients: a retrospective analysis Received: 28 October 2014 Accepted: 8 January 2015 Published online: 20 January 2015 Ó Springer-Verlag Berlin Heidelberg and ESICM 2015 Take-home message: In critically ill patients, a high serum bicarbonate level and its duration are associated with a longer duration of mechanical ventilation, longer ICU length of stay and higher hospital mortality. Electronic supplementary material The online version of this article (doi: /s ) contains supplementary material, which is available to authorized users. A. B. Libório ()) T. T. Leite C. T. de Melo Bezerra Department of Clinical Medicine, Universidade Federal do Ceará, Av. Abolição, 4043 Ap 1203, Ed. Jangada, Fortaleza, CE CEP , Brazil alexandreliborio@yahoo.com.br Tel.: D. T. Noritomi Unidade de Terapia Intensiva, Hospital Paulistano, São Paulo, SP, Brazil E. R. de Faria Pronefron Section, Fresenius Medical Group, Fortaleza, CE, Brazil J. A. Kellum Department of Critical Care Medicine, The Center for Critical Care Nephology, University of Pittsburgh School of Medicine, Pittsburgh, PA, USA Abstract Purpose: Although metabolic alkalosis is a common occurrence, no study has evaluated its prevalence, associated factors or outcomes in critically ill patients. Methods: This is a retrospective study from the Multiparameter Intelligent Monitoring in Intensive Care II database. From 23,529 adult patient records, 18,982 patients met the inclusion criteria. Results: Serum bicarbonate levels demonstrated a U-shaped association with mortality with knots at 25 and 30 meq/l. Of the total included patients, 5,565 (29.3 %) had at least one serum bicarbonate level measurement [30 meq/l. The majority were exposed to multiple factors that are classically associated with metabolic alkalosis (mainly diuretic use, hypernatremia, hypokalemia and high gastric output). Patients with increased serum bicarbonate exhibited increased ICU LOS, more days on mechanical ventilation and higher hospital mortality. After multivariate adjustment, each 5-mEq/l increment in the serum bicarbonate level above 30 meq/l was associated with an odds ratio of 1.21 for hospital mortality. The association between increased serum bicarbonate levels and mortality occurs independently of its possible etiologies. Conclusion: An increased serum bicarbonate level is common in critically ill patients; this can be attributed to multiple factors in the majority of cases, and its presence and duration negatively influence patient outcomes. Keywords Metabolic alkalosis Mortality Critically ill care Acid-base disorders Electronic database Epidemiology Introduction Metabolic alkalosis can be accompanied by several physiological effects, including hemodynamic consequences as a result of direct effects or by means of ionized calcium concentration reduction, which causes potential cardiac arrhythmias and respiratory suppression [1]; hypokalemia, which causes muscle weakness [2]; and difficult mechanical ventilation weaning [3, 4]. Considering the main etiologies [dehydration, diuretic use, renal replacement therapy (RRT), hypokalemia and gastric suctioning], critically ill patients are especially prone to

2 480 developing metabolic alkalosis. While metabolic acidosis has long been associated with adverse outcomes and has been a matter of several laboratory and clinical studies [5, 6], relatively little attention has been devoted to metabolic alkalosis. The impact of alkalosis on mortality in general patients has only been evaluated in one unadjusted study of critically and non-critically ill patients, in which the majority of patients had respiratory alkalosis [7]. Given that metabolic alkalosis represents up to half of all acidbase disturbances in hospitalized patients [8] and is often caused by medical interventions, it is concerning that no large studies have been performed in critically ill patients evaluating the prevalence, etiology and clinical consequences of increased serum bicarbonate levels. In the present report, we reported the prevalence rate of increased serum bicarbonate levels and the associated factors during ICU stays. Moreover, we hypothesized that increased serum bicarbonate is associated with the main outcomes of critically ill patients. Methods Study design and setting The multiparameter intelligent monitoring in intensive care II (MIMIC-II) project is maintained by the Laboratory for Computational Physiology at the Massachusetts Institute of Technology (MIT) and contains de-identified data on patients hospitalized at an ICU at Beth Israel Deaconess Medical Center from 2001 to 2008 [9]. The database is freely available, so any researcher who accepts the data-use agreement and has attended protecting human subjects training can apply for permission to access the data. The Institutional Review Boards of the Massachusetts Institute of Technology and Beth Israel Deaconess Medical Center approved the use of the database and granted a waiver of informed consent. A description of MIMIC-II database is detailed in the Supplemental Appendix. Participants We included all adult patients with an ICU length of stay (LOS) [24 h who had at least one serum bicarbonate measurement within the first 36 h after ICU admission. Patients with serum bicarbonate [28 meq/l associated with acidemia (serum ph\7.35) and CO 2 partial pressure [45 mmhg were considered to have pure respiratory acidosis and were excluded from the study. Additionally, patients with extremely high serum bicarbonate levels ([70 meq/l) were excluded because a measurement/ archival error was likely. Definitions and outcomes Acute kidney injury (AKI) was defined according to the creatinine-based Kidney Disease Improving Global Outcome (KDIGO) criteria [10]. Briefly, only serum creatinine levels were used to classify AKI stage. Given that the MIMIC-II database does not contain previous serum creatinine measurements, the lowest value during the hospital stay was used to determine baseline renal function. We calculated the strong ion difference (SID = Na?? K?? Ca 2?? Mg 2? - Cl - ; all in meq/l) with electrolytes measured on the same day maximum serum bicarbonate levels were achieved. Recorded outcomes included the number of days on mechanical ventilation per 1,000 patient days of ICU stay, hospital LOS, hospital mortality and post-discharge mortality. Post-discharge mortality was assessed through data from the Social Security Death Index (SSDI) during a 9-month period after hospital discharge. The SSDI is not perfect and may contain incorrect and/or missing data. Only the first ICU admission was considered from each patient. Blood gas measurement Venous and arterial blood samples were considered, and in patients with abnormal CO 2 partial pressure values, the standard serum bicarbonate level was calculated. Each patient had serum bicarbonate levels measured at admission (within 36 h of admission to the ICU). Additionally, patients were classified according to the maximum serum bicarbonate level measured during their ICU stay. Because our objective was to evaluate only the metabolic component of alkalosis, we preferred using the standard bicarbonate level when the pco 2 level was out of the normal range. By using this approach, we could guarantee that variations in ph were due to pco 2 variations only in the normal range. To verify the association between the duration of increased serum bicarbonate levels with the outcome in each subject, we calculated the proportion of days with a high serum bicarbonate level as the number of days with serum bicarbonate \30 meq/l. Statistical analysis First, we used descriptive statistics, including the mean with standard deviation, median and interquartile range or frequencies, to describe the population as appropriate. Next, we investigated the association between the maximum serum bicarbonate level as a continuous variable and in-hospital mortality. All tests were two sided, and p \ 0.05 was considered significant.

3 481 Based on the initial exploration of maximum serum bicarbonate levels for hospital mortality, which revealed a U-shaped association, we assessed a cubic spline regression model with serum bicarbonate as a continuous variable and identified two knots at 24 and 31 meq/l. The values were then categorized as \25 or [30 versus meq/l. This same approach was made using SBE with one knot at -1 and 5 meq/l. After testing for collinearity, multiple covariate analyses were applied. We adjusted for age, gender, simplified acute physiology score (SAPS-I), sequential organ failure assessment (SOFA), main comorbidities, type of admission (clinical or surgical), diuretic use, either previous to or during the ICU stay, alkali administration, hypo-/hyperkalemia, hypo-/hypernatremia, RRT and gastric output, minimum serum bicarbonate level, partial CO 2 pressure [45 mmhg, AKI severity, and mechanical ventilation during ICU stay. For interpretation, we provided odds ratio (OR) values for each 5-mEq/l increase or decrease. Because ICU discharge and mortality represent competing risks, we used the cumulative incidence function to analyze time to hospital discharge and mortality over 28 days. Statistical analyses were performed using SPSS 19.0 for Windows and R-project. A complete statistical analysis description is shown in the supplementary material. Results Population The MIMIC-II database contains the records of 32,425 patients, of which 24,581 were adults aged C15 years at the time of admission. In total, 5,403 patients were excluded because their ICU LOS was less than 24 h, and another 68 patients were excluded because of insufficient data on the serum bicarbonate levels. In addition, 124 patients with serum bicarbonate levels greater than 28 meq/l were considered to have pure respiratory acidosis, and 4 other patients were excluded because they had serum bicarbonate levels [70 meq/l (Fig. 1). The main data regarding demographics, illness severity, acid-base status, and main outcomes of the excluded patients are displayed in the supplementary Table 1. Therefore, the final analytic cohort contained 18,982 patients (80 % of all adult patients admitted to the ICU during the period). The mean age upon admission was 63.8 ± 17.5 years, and 8,159 were females (43.0 %). The mean SOFA and SAPS-I scores on admission were 5.9 ± 4.0 and 13.5 ± 5.9, respectively. The overall ICU and in-hospital mortality were 7.7 and 11.5 %, respectively (Table 1). Admission age<15 7,844 ICU stay LOS<24h 5,403 Pure respiratory acidoses 124 MIMIC-II 32,425 Adults 24,581 Final cohort 18,982 No serum bicarbonate record 68 Increased serum bicarbonate levels during ICU stay and in-hospital mortality As shown in Fig. 2, the association between in-hospital mortality and maximum serum bicarbonate levels during ICU stays results in a U shape. The lowest mortality was noted in the group of patients with serum bicarbonate levels between 25 and 30 meq/l. This increase in mortality was significant when the serum bicarbonate level was less than 25 meq/l, and the highest mortality was achieved when the serum bicarbonate levels were less than 20 meq/l. In addition, if a patient had at least 1 day with serum bicarbonate [30 meq/l, a stepwise and continuous increase in mortality was evident. Patient characteristics according to the maximum serum bicarbonate level during ICU stay Serum bicarbonate > 70mEq/L (likely lab error) 4 Fig. 1 Patient distribution from the Multiparameter Intelligent Monitoring in Intensive Care II (MIMIC-II) database and exclusion criteria Based on the above analysis indicating that mortality increased if the patient either experienced at least 1 day of serum bicarbonate levels [30 meq/l or never achieved 25 meq/l, we divided the patients into groups using this cutoff (25 30 meq/l) rather than the classic normal range of meq/l. During their ICU stay, 5,565 patients (29.3 %) had increased serum bicarbonate levels for at least 1 day. The majority of patients (86.6 %) developed increased serum bicarbonate within the first 72 h of the ICU stay, and all except 59 developed it in the first 7 days. Another 3,400 patients (17.9 %) had persistent metabolic acidosis during their ICU stay, and 10,017 patients (52.8 %) had maximum serum bicarbonate levels within the reference range (25 30 meq/l) during their ICU stay. The majority of patients (n = 3,351) with increased serum bicarbonate levels during their ICU stay also had metabolic acidosis at another period of the ICU stay, and

4 482 Table 1 General patient characteristics according to the maximum serum bicarbonate level during ICU stay Total patients (n = 18,982) No increase in serum bicarbonate (n = 13,417) Increased serum bicarbonate (n = 5,565) p Age (years), mean ± SD 63.8 ± ± ± Male [n (%)] 8,159 (43.0) 5,773 (43.0) 2,386 (42.9) Hypertension [n (%)] 6,121 (32.2) 4,463 (33.3) 1,658 (29.8) \ Uncomplicated diabetes [n (%)] 3,687 (19.3) 2,576 (19.2) 1,111 (19.6) Complicated diabetes [n (%)] 937 (4.9) 668 (5.0) 269 (4.8) Obesity [n (%)] 330 (1.7) 190 (1.4) 140 (2.5) \ Congestive heart failure [n (%)] 3,782 (19.9) 2,318 (17.3) 1,464 (26.3) \ COPD [n (%)] 3,128 (16.5) 1,846 (13.8) 1,282 (23.0) \ Cardiac arrhythmia [n (%)] 3,747 (19.7) 2,455 (18.3) 1,292 (23.2) \ Liver disease [n (%)] 948 (5) 694 (5.1) 264 (4.7) Metastatic cancer [n (%)] 848 (4.5) 615 (4.6) 233 (4.2) Previous diuretic use a [n (%)] 7,985 (42.1) 5,301 (39.5) 2,684 (48.2) \ Surgical patients [n (%)] 8,343 (43.9) 5,948 (44.3) 2,395 (43.0) \ SAPS on ICU admission, median (IQR) 14 (10 18) 13 (9 17) 16 (12 19) \ SOFA on ICU admission, median (IQR) 6 (2 9) 5 (2 8) 7 (4 10) \ AKI \ No-AKI 12,372 (65.2) 9,899 (73.8) 2,473 (44.4) Stage 1 3,860 (20.3) 2,210 (16.5) 1,650 (29.6) Stage 2 1,219 (6.4) 537 (4.0) 682 (12.2) Stage 3 1,535 (8.1) 771 (5.7) 764 (13.7) SID (meq/l), mean ± SD (n = 18,352) 41.4 ± ± ± 4.3 \ Serum albumin (g/dl), mean ± SD (n = 8,226) 2.8 ± ± ± 0.7 \ Maximal ph during ICU stay, median (IQR) ( ) 7.49 ( ) \ ( ) Serum bicarbonate on ICU admission (meq/l), mean ± SD 24.3 ± ± ± 5.9 \ Minimal serum bicarbonate during ICU stay (meq/l), 21.2 ± ± ± 5.5 \ mean ± SD Maximal serum bicarbonate during ICU stay (meq/l), 28.6 ± ± ± 4.2 \ mean ± SD Days on mechanical ventilation (per 1,000 patient days \ of ICU stay) ICU LOS, days median (IQR) 2.7 ( ) 2.2 ( ) 5.6 ( ) \ Hospital LOS, days median (IQR) 8 (5 14) 7 (4 11) 12 (7 20) \ ICU mortality [n (%)] 1,468 (7.7) 920 (6.9) 548 (9.8) \ In-hospital mortality [n (%)] 2,185 (11.5) 1,423 (10.6) 762 (13.7) \ SID = Na?? K?? Ca 2?? Mg 2? - Cl - COPD chronic pulmonary obstructive disease, SAPS simplified acute physiology score, SOFA sequential organ failure assessment, ICU intensive care unit, SID strong ion difference, LOS length of stay, AKI acute kidney injury a Previous diuretic use refers to its use previous to ICU admission only 2,163 patients (11.4 %) maintained serum bicarbonate levels within the meq/l range during the entire ICU stay. Patient characteristics according the maximum serum bicarbonate level are presented in Table 1. Factors associated with increased serum bicarbonate levels We analyzed the occurrence of factors commonly described in association with metabolic alkalosis. Patients with increased serum bicarbonate were more likely to have been exposed to diuretics prior to or during their ICU stay and were also administered more alkali treatment. Moreover, COPD and hypokalemia were more common in these patients. High gastric output ([750 ml) during ICU stay was four times more common in patients with increased serum bicarbonate than in other patients (supplementary Table 2). RRT was performed in 926 patients (4.9 %) and was more common in patients with increased serum bicarbonate concentrations. In fact, more than one-third of patients (37.1 %) undergoing RRT had high serum bicarbonate levels. As hypernatremia has been recently associated with metabolic alkalosis [11], we evaluated its occurrence and found that hypernatremia was more frequent in patients with increased serum bicarbonate. Additionally, hypoalbuminemia was more frequent in these patients, supporting the notion that hypoalbuminemia potentially causes metabolic alkalosis, as indicated by the physicochemical approach (Table 1) [12]. Finally, urine chloride measurements were requested for 923 patients, of which 397 had increased serum bicarbonate

5 483 Fig. 2 Hospital mortality according to the maximum serum bicarbonate concentration measured during ICU stays. Linear regression: R 2 = 0.024; cubic polynomial regression: R 2 = levels. Of these 397 patients, only 49 (12.3 %) had a urinary chloride value less than 20 meq/l, suggesting the presence of chloride-responsive metabolic alkalosis in these patients. After investigating all of these possible putative causes, the majority of patients with increased serum bicarbonate (58.4 %) had three or more potentially associated risk factors, whereas the conditions of 408 patients (7.3 %) could not be attributed to the possible explanations we included for the high serum bicarbonate levels. Increased serum bicarbonate and outcomes Patients with increased serum bicarbonate levels had prolonged hospital LOS and more days on mechanical ventilation (Table 1). In addition, these patients exhibited increased in-hospital mortality rates (13.7 vs %, p \ ) compared with other patients. The same trend was observed regarding ICU mortality rates (9.8 vs. 6.8 %, p \ 0.001). The univariate in-hospital mortality OR was 1.34 (95 % CI ) for patients with increased serum bicarbonate levels compared with all other patients. An increased serum bicarbonate value was associated with an increased risk of in-hospital death in unadjusted and fully adjusted models (Table 2). In the fully adjusted model, each 5-mEq/l increment was associated with an OR of 1.21 for hospital mortality. In accordance with the physicochemical approach, we also disclosed an association between a larger SID ([40 meq/l) and mortality (OR 1.19; 95 % CI , for each 5-mEq/l increment. Considering only patients discharged alive from the hospital (n = 16,797), we still detected an adjusted increment in post-discharge mortality (hazard ratio % CI ) for each 5-mEq/l increment. We also performed uni- and multivariate analyses evaluating the proportion of days with increased serum bicarbonate levels, and patients experiencing high serum bicarbonate levels for more than half of their ICU stay exhibited a further increase in the in-hospital mortality rate (see supplementary Table 3). Because metabolic acidosis is highly associated with mortality, we also performed an analysis excluding patients with serum bicarbonate levels B24 meq/l during their ICU stay; the 28-day cumulative incidence plots revealed better mortality rates in patients who maintained serum bicarbonate levels within the meq/l range during their ICU stay than in those with increased serum bicarbonate levels (see supplementary Fig. 1). Metabolic alkalosis at ICU admission and subgroup analysis We also investigated patients with increased serum bicarbonate levels at admission. Among 5,565 patients with increased serum bicarbonate levels during the ICU stay, 1,352 (24.3 %) already had increased serum bicarbonate levels on ICU admission. When taking into consideration only records of ICU admission, an increased serum bicarbonate level remained associated

6 484 Table 2 Association of the maximum serum bicarbonate level during ICU stay with in-hospital mortality Univariate analysis Model 1 Model 2 Model 3 OR (95 % CI) b OR (95 % CI) b OR (95 % CI) b OR (95 % CI) Maximum serum bicarbonate meq/l Ref. Ref. Ref. Ref. Ref. Ref. Ref. Maximum serum bicarbonate under 25 meq/l, per each 5-mEq/l ( ) ( ) ( ) ( ) Maximum serum bicarbonate above 30 meq/l, per each 5-mEq/l ( ) ( ) ( ) ( ) Model 1 adjusts for age and gender. Model 2 adjusts for the covariate in model 1 and simplified acute physiology score (SAPS- I), sequential organ failure assessment (SOFA), main comorbidities (hypertension, congestive heart failure, cardiac arrhythmias, chronic pulmonary obstructive disease, diabetes mellitus, lymphoma, metastatic cancer, liver disease, obesity) and type of admission (clinical or surgical). Model 3 adjusts for covariates in model 2 and the main causes of metabolic alkalosis (diuretic before or during ICU stay, alkali administration, hypokalemia, hypernatremia, renal replacement therapy and gastric output), minimum serum bicarbonate level, partial CO 2 pressure [45 mmhg, acute kidney injury severity, mechanical ventilation and hyponatremia during ICU stay. The bias and 95 % CI were estimated by bootstrap re-sampling. The odds ratio for mortality is displayed for each 5 meq/l in the serum bicarbonate level All p values \ with worse outcome (see supplementary Fig. 2). Interestingly, the adjusted OR for mortality in these patients was almost similar to that in patients with metabolic acidosis on ICU admission when compared to patients with normal serum bicarbonate levels of ( ) and ( ) for each 5-mEq/of serum bicarbonate above 30 meq/l or below 25 meq/l, respectively. We also performed a stratified analysis according to factors classically associated with metabolic alkalosis. In all subgroups except those with high gastric output, at least a trend for a significant association between an increased serum bicarbonate level and in-hospital mortality was noted. These data are presented in Fig. 3. Sensitivity analysis evaluating only arterial blood gas measurement is shown in the supplementary text and supplementary Table 4. In these patients, pure metabolic alkalosis (ph [7.45) was present in 85.1 %, and the remaining had concomitant respiratory acidosis. Adjusted mortality for increased ph was 1.09 (95 % CI ) for each 0.1 increment above Discussion Our analysis of a cohort of almost 19,000 critically ill patients from a single large academic center revealed that increased serum bicarbonate levels ([30 meq/l) were common and associated with longer ICU LOS, prolonged mechanical ventilation, and increased mortality. Moreover, our analysis suggests that factors leading to metabolic alkalosis can be identified in the majority of cases and that multiple factors (more than three) are often observed. To the best of our knowledge, this is the first study evaluating increased serum bicarbonate levels in critically ill patients. The prevalence of metabolic alkalosis in ICU patients has rarely been studied [13]. In the largest study to date, using a mixed population of critically and non-critically ill patients, metabolic alkalosis was the main acid-base disorder observed in hospitalized patients [8]. Even using a cutoff of 30 meq/l, we verified a prevalence rate similar to the alkalosis prevalence reported in other studies [8, 14]. Although metabolic alkalosis has been classically associated with several comorbidities [congestive heart failure and malnutrition (COPD)], the majority of patients developed increased serum bicarbonate levels during their ICU stay. Several therapies or complications during critical care potentially lead to the development of high serum bicarbonate levels. In fact, our analysis disclosed that patients with high serum bicarbonate levels had been exposed to multiple factors; it was difficult to determine only one possible etiology. Some conditions not traditionally associated with increased serum bicarbonate levels have been confirmed by our data. Hypernatremia and low serum albumin levels were more frequently observed in patients with increased serum bicarbonate concentrations. Because many patients developed such a disturbance during the ICU stay, it is difficult to assert that an increased serum bicarbonate level itself is responsible for a prolonged ICU stay. We must take into consideration

7 485 Fig. 3 Fully adjusted odds ratio values for in-hospital mortality in patients experiencing at least 1 day of increased serum bicarbonate levels in subgroups defined by the main associated factors. Adjusted for age, gender, simplified acute physiology score (SAPS-I), sequential organ failure assessment (SOFA), main comorbidities (hypertension, congestive heart failure, cardiac arrhythmias, chronic pulmonary obstructive disease, diabetes mellitus, lymphoma, metastatic cancer, liver disease and obesity), that patients with longer ICU stays are more exposed to situations leading to metabolic alkalosis and also have more serum bicarbonate records, making it difficult to affirm there is a causality association. However, the great majority of patients (86.6 %) developed increased serum bicarbonate levels within the first 72 h of the ICU stay, thus making ICU length of stay an unlikely significant bias in our analysis. We demonstrated that increased an serum bicarbonate level is associated with prolonged mechanical ventilation, in accordance with the physiological knowledge that higher ph is associated with respiratory depression [15]. Also, our data clearly demonstrate that the degree of increase in serum bicarbonate levels and duration of elevation are associated with a higher risk of death. The association between alkalosis and higher mortality has only been reported in one previous study [7]; in this study, the majority of patients had respiratory alkalosis, and only 156 patients had elevated bicarbonate levels. Although an increased serum bicarbonate level was associated with mortality, it is important to state that metabolic acidosis is more detrimental, having a greater impact on mortality. Given that serum bicarbonate levels resulted in a reverse U-shaped association with mortality in critically ill patients and many patients with increased serum bicarbonate levels also developed metabolic acidosis type of admission (clinical or surgical), other main causes of metabolic alkalosis (diuretics prior to or during ICU stay, alkali administration, hypokalemia, hypernatremia, renal replacement therapy and gastric output), minimum serum bicarbonate levels, partial CO 2 pressure [45 mmhg, acute kidney injury severity, mechanical ventilation, and hyponatremia during ICU stay. p = for alkali administration during their ICU stay, two approaches were used to analyze the data. First, we included both the maximum and minimum serum bicarbonate levels during the ICU stay for each patient in the multivariate analysis; second, we compared the effect of high serum bicarbonate levels exclusively in patients who did not develop metabolic acidosis during their ICU stay. With both approaches, we identified an increase in the mortality rate related to high serum bicarbonate levels. To further investigate whether increased serum bicarbonate levels alone impacted mortality, we performed a subgroup analysis according to exposure to each factor classically associated with metabolic alkalosis. In all patient subgroups, except in those with high gastric output, we identified an association between increased serum bicarbonate levels and mortality rates. We hypothesize that high gastric output-associated alkalosis was not associated with mortality because it is the most easily remedied with standard treatment (chlorideresponsive alkalosis). The main strength of our study is the high number of general ICU patients included. This large cohort made it possible to employ a customized methodology, in which we did not assume the traditional reference range for normal serum bicarbonate values (the normal range could be quite misleading in the ICU context). Finally, we

8 486 analyzed the magnitude and time of exposure to high serum bicarbonate levels and the association of these parameters with the mortality rate, demonstrating that this response is truly a biologically independent phenomenon. Several limitations to our study should be noted. First, it was a retrospective study, and it was not possible to affirm there is a causal association between the serum bicarbonate level and outcomes. Although we adjusted for the presence of important comorbidities (such as COPD, heart failure), we could not adjust for the severity of these comorbidities. Second, the database contains data spanning almost a decade, during which changes in the management of critically ill patients and, therefore, changes in patient outcomes have potentially occurred. Because the MIMIC-II database is completely de-identified, we were unable to divide the patients into groups that corresponded to different treatment periods. Also, we did not adopt a physicochemical approach throughout the entire study because many patients had at least one missing laboratory measurement (mainly albumin and phosphate). Moreover, although all data used in the SID calculation were measured on the same day, the dynamic of critically ill patients electrolytes limits our conclusion regarding the physicochemical approach. In conclusion, the risk associated with increased serum bicarbonate levels has been quantified. Future studies focusing on the mechanisms leading to a high serum bicarbonate level, its prevention and correction are warranted. Acknowledgments Results from the paper were presented at the 2014 Annual Meeting of the American Society of Nephrology, held in November 2014 in San Diego, CA. Conflicts of interest A.B.L. is the recipient of a grant from the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq). The funders had no role in the study design, data collection and analysis, decision to publish or preparation of the manuscript. References 1. Chernow B, Zaloga G, McFadden E et al (1982) Hypocalcemia in critically ill patients. Crit Care Med 10: Marti G, Schwarz C, Leichtle AB et al (2014) Etiology and symptoms of severe hypokalemia in emergency department patients. Eur J Emerg Med 21: Berthelsen P, Gøthgen I, Husum B, Jacobsen E (1985) Oxygen uptake and carbon dioxide elimination after acetazolamide in the critically ill. Intensive Care Med 11: Krintel JJ, Haxholdt OS, Berthelsen P, Brøckner J (1983) Carbon dioxide elimination after acetazolamide in patients with chronic obstructive pulmonary disease and metabolic alkalosis. Acta Anaesthesiol Scand 27: Morris CG, Low J (2008) Metabolic acidosis in the critically ill: part 1. Classification and pathophysiology. Anaesthesia 63: Maciel AT, Noritomi DT, Park M (2010) Metabolic acidosis in sepsis. Endocr Metab Immune Disord Drug Targets 10: Anderson LE, Henrich WL (1987) Alkalemia-associated morbidity and mortality in medical and surgical patients. South Med J 80: Hodgkin JE, Soeprono FF, Chan DM (1980) Incidence of metabolic alkalemia in hospitalized patients. Crit Care Med 8: Saeed M, Villarroel M, Reisner AT et al (2011) Multiparameter intelligent monitoring in intensive care II (MIMIC-II): a public-access intensive care unit database. Crit Care Med 39: Molitoris BA, Levin A, Warnock DG et al (2007) Improving outcomes of acute kidney injury: report of an initiative. Nat Clin Pract Nephrol 3: Lindner G, Schwarz C, Grüssing H, Kneidinger N, Fazekas A, Funk GC (2013) Rising serum sodium levels are associated with a concurrent development of metabolic alkalosis in critically ill patients. Intensive Care Med 39: Kaplan LJ, Kellum JA (2010) Fluids, ph, ions and electrolytes. Curr Opin Crit Care 16: Galla JH (2000) Metabolic alkalosis. J Am Soc Nephrol 11: Yunos NM, Kim IB, Bellomo R et al (2011) The biochemical effects of restricting chloride-rich fluids in intensive care. Crit Care Med 39: Webster NR, Kulkarni V (1999) Metabolic alkalosis in the critically ill. Crit Rev Clin Lab Sci 36: