Figure 1. Crude risk of 30-day postoperative mortality according to preoperative sodium level. To convert sodium to millimoles per liter, multiply by 1.0.
Figure 2. Association of 30-day postoperative mortality with hyponatremia in the full cohort and subgroups. Body mass index is calculated as weight in kilograms divided by height in meters squared. ASA indicates American Society of Anesthesiologists; COPD, chronic obstructive pulmonary disease.
Leung AA, McAlister FA, Rogers SO Jr, et al. Preoperative hyponatremia and perioperative complications. Arch Intern Med. Published online September 10, 2012. doi: 10.1001/archinternmed.2012.3992.
eTable 1. Baseline Characteristics of the 233 048 Cases Without Record of Preoperative Serum Sodium Collection Included Into Sensitivity Analysis
eTable 2. Incidences of 30-Day Postoperative Outcomes for the 233 048 Cases Without Record of Preoperative Serum Sodium Collection Included Into Sensitivity Analysis
eTable 3. American Society of Anesthesiologists (ASA) Physical Status Classification
Leung AA, McAlister FA, Rogers SO, Pazo V, Wright A, Bates DW. Preoperative Hyponatremia and Perioperative Complications. Arch Intern Med. 2012;172(19):1474-1481. doi:10.1001/archinternmed.2012.3992
Author Affiliations: Division of General Medicine and Primary Care (Drs Leung, Pazo, Wright, and Bates) and Department of Surgery, Center for Surgery and Public Health (Dr Rogers), Brigham and Women's Hospital, Boston, Massachusetts; and Division of General Internal Medicine, Department of Medicine, University of Alberta, Edmonton, Alberta, Canada (Dr McAlister).
Background Although hyponatremia has been linked to increased morbidity and mortality in a variety of medical conditions, its association with perioperative outcomes remains uncertain.
Methods To determine whether preoperative hyponatremia is a predictor of 30-day perioperative morbidity and mortality, we conducted a cohort study using the American College of Surgeons National Surgical Quality Improvement Program database to identify 964 263 adults undergoing major surgery from more than 200 hospitals (from January 1, 2005, to December 31, 2010) and observed them for 30-day perioperative outcomes. We used multivariable logistic regression to estimate relative risks for death, major coronary events, wound infections, and pneumonia occurring within 30 days of surgery and quantile regression to estimate differences in average length of hospital stay.
Results A total of 75 423 patients with preoperative hyponatremia (sodium level <135 mEq/L [to convert to millimoles per liter, multiply by 1.0]) were compared with 888 840 patients with normal baseline sodium levels (135-144 mEq/L). Preoperative hyponatremia was associated with a higher risk of 30-day mortality (5.2% vs 1.3%; adjusted odds ratio [aOR], 1.44; 95% CI, 1.38-1.50), and this finding was consistent in all the subgroups. This association was particularly marked in patients undergoing nonemergency surgery (aOR, 1.59; 95% CI, 1.50-1.69; P < .001 for interaction) and American Society of Anesthesiologists class 1 and 2 patients (aOR, 1.93; 95% CI, 1.57-2.36; P < .001 for interaction). Furthermore, hyponatremia was associated with a greater risk of perioperative major coronary events (1.8% vs 0.7%; aOR, 1.21; 95% CI, 1.14-1.29), wound infections (7.4% vs 4.6%; 1.24; 1.20-1.28), and pneumonia (3.7% vs 1.5%; 1.17; 1.12-1.22) and prolonged median lengths of stay by approximately 1 day.
Conclusion Preoperative hyponatremia is a prognostic marker for perioperative 30-day morbidity and mortality.
Hyponatremia is common, is potentially reversible, and has important consequences in a variety of clinical conditions.1- 3 Furthermore, the concept of “asymptomatic” hyponatremia has been questioned, as even those with mild hyponatremia are recognized to be at increased risk for serious adverse outcomes.4- 6 Among hospitalized patients, the presence of hyponatremia has been associated with increased mortality,7- 13 prolonged lengths of stay,8,10,12,13 and greater utilization costs.12
However, most existing studies have focused on patients admitted to medical services.7 As such, the association between preoperative hyponatremia and perioperative outcomes remains largely unexplored. Although the preoperative period offers unique opportunities for physicians to perform risk stratification,14,15 anticipate complications, improve perioperative care,16,17 and, in some cases, intervene on modifiable risks,18 the implications of hyponatremia detected preoperatively are unclear. Therefore, we designed this study to evaluate the prognostic implications of preoperative hyponatremia on selected perioperative outcomes across a broad range of surgical conditions.
This study was approved by the institutional review board at Partners HealthCare.
We assembled a cohort through the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) Participant Use Data Files. The ACS NSQIP is a nationally validated, outcome-based, risk-adjusted program developed to improve the quality of surgical care for adults in the United States.19 Presently, there are nearly 300 participating academic and community hospitals. This database provides patient-level information for major surgical procedures, in the inpatient and outpatient setting, across different surgical specialties. One of the strengths of this database is that preoperative, intraoperative, and postoperative data are prospectively collected for each surgical case through trained nurse reviewers using standardized methods, definitions, and end points. Data are monitored weekly to ensure appropriate case selection, and data validation is performed through audits to ensure a rich source of reliable clinical information.19,20 For cases collected between January 1, 2005, and December 31, 2010, the ACS NSQIP database excluded all the surgical procedures directly related to acute trauma, transplantation cases, surgical procedures for brain-death organ donors, and concurrent cases. Furthermore, cases of inguinal herniorrhaphies exceeding 3 in 8 days and breast lumpectomies exceeding 3 in 8 days were excluded from each participating site to ensure a diverse case mix of procedures.20
The study cohort consisted of adult patients (≥18 years old) from all the participating sites undergoing any major surgery between January 1, 2005, and December 31, 2010.21 We excluded patients who had undergone any other major surgical procedure in the previous 30 days and patients for whom we could not verify the surgical history. We defined hyponatremia as a sodium measurement less than 135 mEq/L (to convert to millimoles per liter, multiply by 1.0), and we further classified hyponatremia according to mild (130-134 mEq/L) and moderate to severe (<130 mEq/L) categories. To explore the association between preoperative hyponatremia and perioperative outcomes, we excluded patients who did not have a preoperative serum sodium level recorded and those with hypernatremia (sodium level ≥145 mEq/L) from the primary analysis. The preoperative laboratory value was defined as the most recent sodium level measured within 90 days of surgery.
The following baseline data were retrieved for each patient: date of birth, sex, surgical profile (eg, principal procedure, inpatient vs outpatient status, emergency vs nonemergency surgery, and surgical specialty), preoperative characteristics (eg, height, weight, smoking history, alcohol consumption, functional health status, American Society of Anesthesiologists [ASA] classification,22 and history of comorbidities, such as diabetes mellitus, pulmonary disease, hepatobiliary disease, cardiac disease, renal disease, and cerebrovascular disease), and preoperative laboratory data (eg, serum sodium and creatinine levels). These data were collected according to strict definitions from medical records, operating room logs, anesthesia records, telephone interviews, and letters.
The primary outcome was 30-day mortality (defined as any death occurring within 30 days of surgery regardless of cause, in or out of the hospital, and included intraoperative and postoperative deaths). Secondary outcomes, which were events occurring within 30 days of surgery (in the intraoperative and postoperative periods), included major coronary event (a composite of myocardial infarction and cardiac arrest), stroke, postoperative wound infection (a composite of superficial incisional, deep incisional, and organ-space surgical site infections), pneumonia, and length of hospital stay. Detailed definitions are online.20
Descriptive statistics were reported for baseline characteristics between patients with normal sodium measurements and those with hyponatremia. All study outcomes were dichotomized, except length of stay, which remained as a continuous variable. For the discrete outcomes, crude comparisons were made using the χ2 test, and odds ratios (ORs) were reported. We used multivariable logistic regression, accounting for all measured patient- and procedure-related factors, to calculate adjusted ORs (aORs) with 95% CIs. The data were then fit according to a nonparametric quantile regression model, and adjusted estimates for the median change in length of stay were determined according to baseline sodium category. (In contrast to multiple linear regression, where there is an assumption that residuals are normally distributed, this model assumption is not required for quantile regression.) Data were mostly complete, but missing values were handled in the analysis by dummy coding.23 We subsequently conducted a secondary analysis for the primary outcome according to the number of days from the preoperative sodium measurement to the time of operation. Furthermore, sensitivity analyses for potential selection bias were performed to account for the patients who were excluded from the initial cohort assembly because they did not have a preoperative sodium measurement recorded. Finally, we performed stratified analyses to screen for effect modification. All the statistical analyses were performed using a commercially available software program (SAS, version 9.3; SAS Institute, Inc).
A total of 1 334 886 surgical cases were identified from the ACS NSQIP database during the study. We excluded 137 575 patients because they were younger than 18 years, had undergone surgery in the preceding 30 days, or had preoperative hypernatremia. Of the remaining 1 197 311 patients, 233 048 did not have any documented preoperative sodium measurements and were, accordingly, excluded from the primary analysis. A final cohort of 964 263 patients was assembled.
Preoperative hyponatremia was present in 75 423 surgical patients (7.8%), and most cases were mild in severity (Table 1). The greatest prevalence of preoperative hyponatremia was in those undergoing cardiac surgery (11.8%) and vascular surgery (11.2%), followed by general (7.5%), orthopedic (7.1%), and other (6.1%) procedures. Patients with hyponatremia were typically older, were more likely to be male, had greater comorbidity, and were more likely to be inpatients or those undergoing emergency surgery. Nearly all the patients had their preoperative blood work performed within 1 month of surgery. Sodium measurements were performed within 1 day of surgery in 95.0% of all emergency patients and in 30.6% of all nonemergency patients.
For the binary end points, the outcome incidences were rare (<10%). Thus, the aORs approximate adjusted relative risks.24 As such, we refer to relative ratios between odds and risks interchangeably throughout the presentation of the results.
During the study, 15 630 deaths were observed within 30 days of surgery. The risk of death was related to the magnitude of preoperative hyponatremia (Figure 1). Compared with patients with normal preoperative sodium levels (1.3% deaths), those with mild hyponatremia (4.6% deaths; aOR, 1.38; 95% CI, 1.32-1.45) and moderate to severe hyponatremia (9.6% deaths; aOR, 1.72; 95% CI, 1.58-1.88) exhibited higher mortality rates, even after adjustment for baseline imbalances in patient and procedural characteristics (P < .001 for pairwise comparison between mortality risk in mild and moderate to severe categories) (Table 2).
Compared with patients with normal sodium levels, the presence of preoperative hyponatremia was likewise associated with a higher risk of major coronary events (1.8% vs 0.7%; aOR, 1.21; 95% CI, 1.14-1.29), wound infections (7.4% vs 4.6%; 1.24; 1.20-1.28), and pneumonia (3.7% vs 1.5%; 1.17; 1.12-1.22). As with all-cause mortality, patients with more severe levels of hyponatremia exhibited higher event rates than those with milder degrees of hyponatremia, even after adjustment (Table 2). Only the rate of stroke was not significantly higher in patients with hyponatremia (0.5% vs 0.3%; P = .23), although this may be related to inadequate power since this outcome was rare.
We subsequently examined lengths of stay (Table 3). The unadjusted median lengths of stay were 2.0 to 4.0 days longer in patients with hyponatremia vs those with normal sodium levels, varying according to surgical specialty and year. After controlling for all the covariates to account for differences in case mix, hyponatremia was typically associated with longer median lengths of stay by approximately 1 day for most surgical procedures during the study, and it was never associated with decreased length of stay.
Sodium levels are subject to change over time. Therefore, we restricted the analysis to patients with recent (ie, ≤2 weeks before surgery), very recent (ie, ≤1 week before surgery), and same-day (ie, ≤1 day before surgery) sodium measurements, and the overall risks remained broadly similar. The odds of death within 30 days of surgery was consistently higher in patients with hyponatremia regardless of time of blood sample collection (aOR, 1.42; 95% CI, 1.36-1.49 for recent; 1.40, 1.34-1.46 for very recent; and 1.35, 1.29-1.41 for same day).
During initial cohort assembly, 233 048 patients were excluded because their sodium levels were not measured preoperatively (eTable 1 and eTable 2). Thus, we performed sensitivity analyses with a total cohort of 1 197 311 individuals, assuming that the initially excluded patients all had normal sodium levels, and the odds of death was still higher in patients with hyponatremia (aOR, 1.43; 95% CI, 1.37-1.50). We again repeated the analyses with the assumption that all these patients had hyponatremia at baseline. The risk of death in patients with hyponatremia remained elevated (aOR, 1.39; 95% CI, 1.33-1.45). These 2 estimates reflect the range of possible outcomes given the 2 extremes of selection bias.
Finally, we conducted subgroup analyses for the primary outcome (Figure 2). None of the baseline patient characteristics or procedural factors changed the overall association between preoperative hyponatremia and the increased risk of 30-day mortality. As was consistent with the main study estimates, there was an increased risk of death in all the subgroups. The association was particularly marked in patients undergoing nonemergency surgery (P < .001 for interaction) and in patients classified as ASA class 1 or 2 (P < .001 for interaction) (for definitions of the ASA classification, see eTable 3).
In this observational study of nearly 1 million patients undergoing major surgery, we found that preoperative hyponatremia was present in approximately 1 in 13 patients, and this group had a 44% increased risk of 30-day perioperative mortality, even after adjustment for all other potential risk factors. Furthermore, preoperative hyponatremia was common and was an independent negative prognostic factor in patients undergoing nonemergency surgery (with an incidence of 1 in 16 patients and an associated 59% increased risk of death) and in those classified as ASA class 1 or 2 (with an incidence of 1 in 21 patients and an associated 93% increased risk of death). Moreover, the excess risk was present even for patients with mild hyponatremia. Preoperative hyponatremia was also associated with an increased risk of perioperative major coronary events, surgical site wound infections, pneumonia, and prolonged hospital stays.
Previous studies7- 12 examining the prognosis related to hyponatremia have mostly been in the inpatient setting; these studies have focused on medical services or have examined hospitalwide admissions to single centers. Hyponatremia has been documented to be a negative prognostic factor in congestive heart failure,25 liver disease,26- 28 chronic kidney disease,29,30 pneumonia,13 and hospitalized10,11 populations. Until now, however, there has been little evidence to link preoperative hyponatremia with adverse perioperative outcomes, as previous research has been limited to exploratory subgroup analyses. In a previous single-center study,10 community-acquired hyponatremia was associated with a higher risk of in-hospital mortality in a subgroup of 6393 patients admitted to surgical services (aOR, 2.21; 95% CI, 1.49-3.29). Similarly, a more recent study11 reported that in-hospital mortality was increased with hyponatremia in a subgroup of 11 079 patients admitted for orthopedic procedures (aOR, 2.31; 95% CI, 1.25-4.27), but it did not differentiate between preoperative and postoperative hyponatremia. The magnitude of risk that we report is smaller compared with that of previous work (although the present crude estimates were larger) likely because (1) we adjusted for confounders that were not measured in previous studies (such as functional status and procedural factors), (2) we restricted the exposure to preoperative laboratory work, and (3) we examined 30-day outcomes rather than index hospitalization events only. In favor of this reasoning, Greenblatt and colleagues31 observed that preoperative hyponatremia was independently associated with surgical site infection, controlling for patient demographic characteristics, preoperative comorbidities, and operative variables; their reported risk (aOR, 1.20; 95% CI, 1.02-1.42) was nearly identical to that of the present study (aOR, 1.24; 95% CI, 1.20-1.28). Altogether, the present study is consistent with previous work and provides a more refined estimate of the true risk of perioperative complications associated with preoperative hyponatremia.
Controversy exists over whether hyponatremia is a marker or a mediator of mortality and other adverse events.30,32 Physiologically, hyponatremia results from disturbances in water balance, which is normally regulated by the actions of antidiuretic hormone. Accordingly, some possible causes of preoperative hyponatremia include volume depletion, pain and nausea, and predisposing medical conditions (eg, congestive heart failure), which are all marked by higher levels of circulating antidiuretic hormone.1- 3 Addressing this, Waikar and colleagues30 recently demonstrated that the association between hyponatremia and mortality was independent of any antidiuretic hormone–mediated mechanism. However, it remains to be proved whether hyponatremia itself is a causal determinant of adverse events or whether it solely indicates the presence of other serious underlying conditions (overt or subclinical) that, in turn, mediate morbidity and mortality. Although it is clear that sodium concentrations are critical to various cellular functions in the body,1,30 there is still no known pathogenic mechanism to explain how hyponatremia could directly cause the observed increased risk of perioperative morbidity and mortality. Given the present state of information, hyponatremia should be considered as a sensitive surrogate marker for comorbidity and disease severity.
Although this study provides evidence that preoperative hyponatremia is associated with perioperative morbidity and mortality, further research is needed to establish whether correcting preoperative hyponatremia will mitigate risks. Legitimate concern should be raised about the safety of intervention as overly rapid or large changes to sodium levels over a short time can be potentially disastrous.33 Conversely, if monitored correction of hyponatremia is found to be safe and beneficial, it would strengthen causal inference and would be transformative to routine care since serum sodium is not presently recognized as an independent and reversible risk factor for perioperative complications.18
These findings must be interpreted in the context of the study design. As with all nonexperimental designs, this study is subject to potential confounding. Although information on many clinically relevant variables was collected and we performed careful statistical adjustments in the analyses to account for differences between exposure groups, unmeasured confounders may still be present. Second, we reported 30-day perioperative outcomes because these were routinely reported, checked, and validated by the ACS NSQIP. As such, complications that occurred after that period were not considered. However, more distant events are also less likely to be direct sequelae of surgery itself or the preoperative state. Third, although the data we used are rich and mostly complete, there are some inherent limitations to the data set, some of which were deliberately introduced to safeguard the privacy of patients, such as removal of unique patient identifiers. We also lacked medication data and, therefore, could not determine how risk may vary according to various drug exposures. Moreover, it was impossible to completely ensure that some individuals did not contribute more than once to the analysis during the study interval, although it would be improbable based on the sampling methods used and the exclusion of patients with recent surgical histories.20 Furthermore, the results of the sensitivity analyses suggest that any selection bias resulting from missing laboratory data was unlikely to be significant. Finally, hyponatremia is a heterogeneous disorder with a variety of potential etiologies, yet a limitation of this data set is that we did not classify study participants according to serum osmolality or clinical volume status. As such, although these findings may not be specific to any particular subgroup, it can also be said that the broad inclusion of all types of hyponatremia potentially enhances the generalizability of the study.
In conclusion, we found that preoperative hyponatremia (even to a mild degree) is common and predicts postoperative morbidity and mortality, even in relatively healthy patients (ie, classified as ASA class 1 or 2) and those undergoing nonemergency surgery. These findings give rise to several key implications. First, these results illustrate that even mild perturbations of serum sodium are not inconsequential and should not be ignored. Hyponatremia, when detected preoperatively, should be considered a prognostic marker for perioperative complications, and its presence should alert physicians to a situation of increased risk necessitating closer surveillance in the perioperative period. Second, whenever possible, the underlying cause of hyponatremia should be determined. Although the effectiveness and safety of intervening on preoperative hyponatremia have not yet been established, one reasonable approach is to monitor for perioperative complications in all patients at risk and to selectively treat hyponatremia before nonemergency surgical procedures when a reversible cause is found. Accordingly, further studies on how to best address hyponatremia in the perioperative setting are needed.
Correspondence: David W. Bates, MD, MSc, Division of General Internal Medicine and Primary Care, Brigham and Women's Hospital, Brigham Circle, 1620 Tremont St, Third Floor, Boston, MA 02120 (firstname.lastname@example.org).
Accepted for Publication: June 7, 2012.
Published Online: September 10, 2012. doi:10.1001/archinternmed.2012.3992
Author Contributions:Study concept and design: Leung, McAlister, Wright, and Bates. Acquisition of data: Rogers. Analysis and interpretation of data: Leung, McAlister, Rogers, Pazo, Wright, and Bates. Drafting of the manuscript: Leung. Critical revision of the manuscript for important intellectual content: Leung, McAlister, Rogers, Pazo, Wright, and Bates. Statistical analysis: Leung. Obtained funding: Bates. Administrative, technical, and material support: Rogers and Bates. Study supervision: Rogers and Bates.
Financial Disclosure: None reported.
Funding/Support: Dr Leung is supported by a Clinical Fellowship Award from Alberta Innovates–Health Solutions and by a Fellowship Award from the Canadian Institutes for Health Research. Dr McAlister is supported by a career salary award from Alberta Innovates–Health Solutions as a senior health scholar.
Disclaimer: The American College of Surgeons National Surgical Quality Improvement Program and the hospitals participating in the ACS NSQIP are the source of the data used herein; they have not verified and are not responsible for the statistical validity of the data analysis or the conclusions derived by the authors.