Association of Preoperative Hyponatremia With Surgical Outcomes: A Systematic Review and Meta-analysis of 32 Observational Studies

Abstract

Background

Preoperative hyponatremia is prevalent in patients undergoing surgical procedures, but it is uncertain if hyponatremia will lead to increased risk of surgical mortality and morbidity.

Methods

A systematic search of Medline (PubMed), Embase, and Cochrane Library from inception through July 2, 2021, was performed. Full-length articles that reported on the association between surgical outcomes among adults aged ≥18 years with documented preoperative hyponatremia were included.

Findings

We identified 32 observational studies comprising 1 301 346 participants. All studies had low risk of bias. When adjusted for covariates, patients with hyponatremia had significantly higher odds of developing major complications (defined as a composite measure of 9 major complications) compared with patients with normal sodium concentrations (adjusted odds ratio = 1.37; 95% CI, 1.23-1.53; I² = 78%; N = 10). Additionally, patients with preoperative hyponatremia also significantly higher hazards of early mortality (<90 days) compared with patients with normonatremia (adjusted hazard ratio = 1.27; 95% CI, 1.13-1.43; I² = 97%; N = 10) after adjustment for covariates. Preoperative hyponatremia also had significant associations with respiratory, renal, and septic complications. In terms of prognostic performance, preoperative hyponatremia performed adequately in predicting major complications in surgical patients (area under the curve = 0.70; negative likelihood ratio, 0.90) with a specificity of 88% and a sensitivity of 25%.

Interpretation

Our meta-analysis suggests that preoperative hyponatremia is associated with poorer early mortality and major morbidity outcomes in surgical patients. Hyponatremia is also a specific prognosticator for major complications in surgical patients, reiterating its potential use as a clinical indicator of poor outcomes.

The homeostasis of serum sodium concentration is maintained by an interplay of mechanisms including the renin-angiotensin-aldosterone axis, sympathetic neural activity, release of myocardial natriuretic peptides, and antidiuretic hormone (1, 2). Disorders of sodium concentrations include hypernatremia and hyponatremia, with hyponatremia being far more prevalent (3). Hyponatremia is frequently encountered in hospitalized patients (4) and is a known predictor of adverse outcomes in patients admitted for medical conditions including congestive cardiac failure, chronic kidney disease, and liver disease (5-11), even in mild cases (12, 13).

Preoperative hyponatremia is a prevalent occurrence seen in approximately 1 in every 13 patients who are planned for surgeries (14). In geriatric patients undergoing orthopedic surgeries, it has been reported that mortality increases with serum sodium levels below 135 mmol/L (15). Although there is other emerging evidence that suggests preoperative hyponatremia may be associated with poor perioperative outcomes in other surgical populations (16-18), some studies found no significant association (5, 9). In a previous large-scale US surgical database study conducted in 2012, preoperative hyponatremia was associated with higher risk of 30-day mortality and morbidities including perioperative major coronary events, wound infections, and pneumonia (14). In the intervening years, other large-scale observational studies in various regions have been conducted (9). However, to our knowledge, there has not been any systematic, evidence-based review of the association between preoperative hyponatremia and surgical outcomes to date.

Materials and Methods

Search Strategy and Selection Criteria

We searched PubMed, Embase, and Cochrane Library from inception through July 2, 2021, using free text and controlled vocabulary terms. We also identified 2 additional studies via a “snowball” search of reference lists of included articles. We searched the databases with keywords including “hyponatremia,” “outcome,” and “surgical” or relevant synonyms such as “dysnatremia,” “complication,” “morbidity,” “mortality,” and “postoperative.” The full search strategy is included in Supplementary Methods (19).

Three authors (M.Y.G., C.B.T., and R.Y.K.T.) retrieved the records and screened all titles and abstracts to determine eligibility independently. The authors then independently retrieved full texts of the articles to evaluate eligibility for inclusion and extracted data manually into a standardized extraction template. In the case of disagreement, consensus on the eligibility for inclusion was reached by discussion. A fourth author (N.H.W.L.) reviewed the included studies and verified extracted data. The study selection process is outlined in Fig. 1.

Figure 1.

PRISMA flow diagram of the study selection process.

In this systematic review and meta-analysis, we included any observational studies published as full-length articles in peer-reviewed journals that reported on the association of postoperative outcomes among adult patients with documented preoperative hyponatremia. The exposure group was patients with preoperative hyponatremia, as defined by the original studies, where serum sodium levels were measured before surgery. The comparator was patients with normonatremia. Outcomes of interest included postoperative mortality and morbidity outcomes. These include major complications (surgical site infection, wound dehiscence, pulmonary complications, unplanned reintubation, renal complications, neurological complications, cardiac complications, deep venous thromboembolism, and sepsis) as well as outcomes of early and late mortality and length of hospital stay.

We excluded studies that lacked a comparator group, studies without adequate quantifiable data on outcomes (eg, only P values provided without quantifiable effect measure), non-English studies, and articles that were case reports, case series, reviews, letters, conference abstracts, or non-full-length articles.

We extracted unadjusted and maximally adjusted measures of associations for all perioperative and postoperative mortality and morbidity-related outcomes reported in each study. When neither measure was reported, we recorded the raw frequency count to calculate the unadjusted measures of association. We also collected key data on the study characteristics, outcome measures, and hyponatremia-associated information from each included study. See Supplementary Methods (19) for the full list.

Three authors (M.Y.G., C.B.T., and R.Y.K.T.) independently applied the Newcastle-Ottawa Scale tool to evaluate for risk of bias (Supplementary Table S2) (19, 20).

Data analysis

We performed separate meta-analyses for individual outcomes using the inverse variance-weighted mixed-effects model, pooling unadjusted and maximally adjusted effect measure separately. We assessed and considered between-study heterogeneity as significant if the Q-test P value was <0.10 or if the I² statistic was ≥50%. To assess for potential sources of heterogeneity, we performed further subgroup analyses and univariate random-effects meta regression for the study-level characteristics. See Supplementary Methods (19) for details.

Additionally, we performed meta-analyses of the prognostic value and performance of hyponatremia for the outcome of mortality and the composite outcome measure of major complications using a hierarchical model for meta-analytical integration. We generated a summary receiver operator characteristic curve, Fagan nomogram, coupled funnel plots and calculated area under the curve, sensitivity, specificity, positive and negative likelihood ratios, and positive and negative predictive values to evaluate the prognostic accuracy of hyponatremia in predicting the unadjusted odds of major complications and mortality. We used the O:E ratio as a measure of calibration. The O:E ratio is the ratio of the number of observed outcomes divided by the number of outcomes expected by the prediction model. We pooled O:E ratios from the included studies using the restricted maximum likelihood estimation and the HartungKnapp-Sidik-Jonkman method to derive all confidence intervals and summary statistics (21, 22).

We conducted all analyses on Stata (version 17), Review Manager (version 5.4.1), and R (version 4.0.3) using the meta, metafor, and metamisc packages.

We conducted this review in accordance with the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines (23); the checklist can be found in Supplementary Table S1 (19). We registered our study on PROSPERO (CRD42021285685). Institutional review board approval was not required for this study because it involved data extracted from literature available in the public domain. The requirement for informed consent was waived and all research adhered to the tenets of the Declaration of Helsinki.

Results

We retrieved 11 051 abstracts from the initial database search and subsequently screened 11 011 records after removing duplicates. Eventually, 32 studies were included (5, 7, 9, 14, 16-18, 24-48), and 29 studies were eligible for meta-analyses (Table 1) (5, 7, 9, 14, 16-18, 24, 26-33, 35-44, 46, 48). To prevent double representation of the same dataset, the study with the longest period of enrollment was selected when there were multiple studies using the same dataset to analyze the same outcomes (17, 28).

Table 1.

Summary of included studies

First author (last name), year country	Study design	Total sample size % Male Average age, y	Patient characteristics	Type of surgery	Definition of hyponatremia	Timepoint of preoperative serum sodium measurement	Outcomes reported	Covariates adjusted for	NOS score (of 9)
Pennington, 2021 (43) USA	Retrospective population database study	10 654 45.60% NR	Patients who require lumbar interbody fusion patients	Lumbar interbody spinal fusion	<135 mEq/L	NR	Length of hospitalization stay Major morbidity 30-d readmission 30-d reoperation	NR	7
McCauslands, 2014 (16) USA	Retrospective study	16 206 44.80% 62.5	Adult patients admitted for major orthopedic procedures	Major orthopedic surgery	<135 mEq/L	Serum sodium during hospitalization most proximal to the day of surgery	30-d mortality	Age, race, sex, Deyo-Charlson Comorbidity Index, and individual codes for congestive heart failure, diabetes, cancer, liver disease, and fracture	8
Bokhari, 2019 (27) USA	Retrospective cohort study	58049 NR NR	Patients admitted for elective degenerative spine surgery	Elective degenerative spine surgery	NR	NR	Length of hospitalization stay (prolonged hospital stay) Blood transfusion requirement	NR	7
Tinning, 2015 (45) UK	Retrospective cohort study	3897 20.90% 83	Patients with hip fracture	Operative treatment for hip fracture	<135 mEq/L	On admission	Overall mortality	NR	7
Crestanello, 2013 (17) USA	Retrospective cohort study	2247 66.60% 62	Patients going for cardiac surgery	Cardiac surgery	<135 mEq/L	Latest sodium measurement within 30 d before surgery	Mortality Neurologic complication Renal complication	Refer to appendix E2 of article	8
Hackworth, 2008 (7) USA	Retrospective cohort study	213 78.00% 51	Patients with liver cirrhosis	Liver transplant	≤130 mEq/L	Measured immediately before transplantation	Delirium Acute cellular rejection Mortality at 3 mo Pneumonia Acute renal failure Transplant rejection	NR	7
Leise, 2014 (39) USA	Retrospective cohort study	19 537 66.72% 53	Liver transplant patients with serum NA levels immediately before LT available	Liver transplant	≤130 mEq/L	NR	30-d mortality 90-d mortality 1-y mortality In-hospital mortality	Age, sex, race, presence of ascites, etiology of liver disease, MELD score, dialysis, and life support	8
Madsen, 2015 (41) Denmark	Retrospective cohort study	7317 23.80% 83	Surgically treated hip fracture patients	Hip fracture surgery	<135 mEq/L	At admission	30-d mortality	Refer to index table in article	8
Feinstein, 2015 (30) USA	Retrospective cohort study	214 64.50% 67	Patients with squamous cell carcinoma	General surgery	<135 mEq/L	24 h before surgery	Length of hospitalization Mortality Cardiac complication Pulmonary complication Infection complication Renal complication Blood transfusion required	NR	7
Khan, 2020 (37) USA	Retrospective cohort study	16 238 69.36% 70 in non hyponatremia, 71.6 in hyponatremia	Patients who had CABG, valve, or CABG and valve procedures from 2000 to 2016 and available preoperative serum sodium within 30 days of the index procedure were included in the study.	CABG, valve or CABG, and valve procedures	<135 mEq/L	Preoperative period	Long-term mortality (10 y) Operative mortality Length of hospitalization Stroke or TIA Prolonged ventilation Atrial fibrillation Renal failure	Age, sex, body mass index, smoking status, diabetes, renal failure, hypertension, cerebrovascular accident, endocarditis, chronic lung disease, peripheral vascular disease, cerebrovascular disease, prior CABG or valve surgery, previous myocardial infarction, NYHA functional class, surgery status, type (CABG, valve, or CABG and valve surgery) and year of index surgery, glucose, creatinine, and LVEF	8
Hagino, 2013 (33) Japan	Retrospective cohort study	512 21.50% 82.6	Patients who had fractures of femoral neck and trochanteric fracture	Femoral head replacement and open reduction and fixation	<135 mEq/L	NR	Overall mortality	NR	8
Londono, 2006 (40) Spain	Retrospective cohort study	241 63.00% 55	Patients with cirrhosis who were going for transplantation	Liver transplantation	<130 mEq/L	Time of transplantation	Mortality at 1 y Infection complication Neurologic complication Renal complication Vascular complications Acute rejection Biliary complications	NR	7
Shavit, 2018 (44) Israel	Retrospective observational study	1008, 60.00% 70	Patients with nondialysis-dependent chronic kidney disease, stage 3-4 CKD	Cardiac surgery	<135 mEq/L	Preoperative	Late mortality (survival at 13 y) Renal failure requiring dialysis Ventilation Perioperative AMI, low cardiac output Atrial fibrillation Infection complication Stroke Dialysis	NR	7
Fukuhara, 2009 (31) Japan	Retrospective analysis	134 54.50% 53	Patients who went for LDLT for end-stage liver diseases	Living donor liver transplantation	≤130 mEq/L	Pretransplant: lowest preoperative sodium score was used	Acute graft loss (rejection) Sepsis Neurologic complication Acute rejection Biliary complications	NR	7
Yun, 2009 (5) USA	Retrospective cohort	2175 55.70% 50.1	Patients who have end-stage liver disease going for orthotopic liver transplantation	Orthotic liver transplantation	<135 mEq/L (mild hyponatremia: 125-134 mEq/L, severe hyponatremia: <125 mEq/L)	Immediately before OLT	90-d mortality	Covariates at the time of OLT such as age, sex, total bilirubin, creatinine, and diagnostic categories were also taken into account.	8
Wang, 2016 (46) China	Retrospective study	2733 84.45% 47.5	HBV cirrhosis patient	Liver transplantation	<135 mmol/L (mild hyponatremia: 130-135 mmol/L, moderate hyponatremia: 125-130 mmol/L, severe hyponatremia ≤125 mmol/L)	Pretransplantation	Overall mortality	Adjusted for transplant year, recipient sex, recipient age, and graft type	8
Yang, 2018 (48) Korea	Retrospective study	1164 72.00% 53.5	Patients who need liver transplantation	Liver transplantation	<130 mEq/L	Preoperative measured within 1 mo of transplant Intraoperative measurements calculated as an average of the following	1-y mortality	Living versus deceased donor, recipient age, sex, body mass index, history of hypertension, diabetes mellitus, preoperative diuretics administration, hepatorenal syndrome, cold ischemic time, warm ischemic time, MELD score, graft-recipient body weight ratio, ABO-incompatibility transplantation, preoperative left ventricular ejection fraction, operation time, red blood cell transfusion during surgery, intraoperative mean blood glucose, preoperative hemoglobin, and preoperative albumin	8
Xu, 2019 (47) China	Retrospective study	842 74.50% 64	Patients with gastric adenocarcinoma scheduled for curative gastrectomy	Radical gastrectomy	<130 mmol/L	After admission	NR	NR	7
Leung, 2012 (14) USA	Prospective Cohort	964 263 43.2% (overall) NR	Adult patients (≥18 years old) from all the participating sites undergoing any major surgery	Any major surgery	<135 mEq/L	Preoperative bloodwork performed within 1 mo of surgery	30-d mortality Cardiac complication Pneumonia Infectious complications Stroke Mortality Pneumonia Cardiac complication Infection complication Neurologic complication	Date of birth, sex, surgical profile (eg, principal procedure, inpatient vs outpatient status, emergency vs nonemergency surgery, surgical specialty), preoperative characteristics (eg, height, weight, smoking history, alcohol consumption, functional health status, classification, 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, creatinine levels)	8
Pennington, 2020 (42) USA	Retrospective	20 817 50.29% NR	Patients identified in NSQIP from 2012 to 2014	Cervical spine surgery	<135 mEq/L	Preoperative (varied)	30-d mortality Major morbidity Length of hospitalization 30-d readmission 30-d reoperation	19 covariates for our major morbidity model, 14 for mortality, 26 for hospitalization duration, 21 for 30-d readmission model, and 18 for 30-d reoperation	8
Lee, 2020 (38) USA	Retrospective	10 623 43.60% 69.4	NSQIP data from 2012 to 2016	Anatomic or reverse total shoulder arthroplasty	<135 mEq/L	Last known sodium record within 90 d	Major morbidity Length of hospitalization Non-home discharge 30-d readmission 30-d reoperation	Age, BMI, sex, race, smoking status, diabetic treatment, dyspnea, functional status, history of COPD, history of CHF, dialysis treatment, steroid use, bleeding disorder, and sepsis	8
Abola, 2019 (18) USA	Retrospective database review	88 103 37.62% NR	Patients aged ≥18 y in the NSQIP database from 2012 to 2014	Total knee arthroplasty	<135 mEq/L	NR	Major morbidity Length of hospitalization 30-d readmission 30-d reoperation	age, sex, race, BMI, history of diabetes, smoking, dyspnea, functional status, history of CHF, history of COPD, dialysis, hypertension requiring medication, disseminated cancer, chronic steroid use, ASA class, bleeding disorders, weight loss, evidence of wound infection, preoperative blood transfusion, and sepsis	8
Gu, 2020 (32) USA	Retrospective database review	25 517 42.1% NR	American College of Surgeons NSQIP database for all revision hip and knee arthroplasties between 2007 and 2016	Revision hip and knee arthroplasties	<135 mEq/L	Before revision joint replacement	Deep wound infection Organ space infection Infectious complications (major; ie, combined deep wound and organ/space, self-calculated) 30-d mortality Length of hospitalisation (extended LOS >7 days) Sepsis Myocardial infarction Pneumonia Pulmonary complication Required transfusion Urinary tract infection Mortality Sepsis Septic shock Pneumonia Cardiac complication Pulmonary complication Infection complication Renal complication Renal failure	smoking status, diabetic status, BMI, dyspnea status, ASA class, type of anesthesia used, and preoperative functional status. Medical comorbidity data collected included CHF, COPD, hypertension, renal failure, dialysis dependence, steroid use, weight loss, bleeding disorders, and preoperative transfusion	8
Crestanello (2), 2013 (28) USA	Prospective	4370 66.2% Hyponatremia: 61 No hyponatremia: 62	2002 to 2008	Cardiac surgery (CABG, valve, CABG and valve, others)	<135 mEq/L	any of the 3 most recent daily sodium determinations performed within 30 d before surgery. For patients who had multiple sodium determination in a single day, the mean of those determinations was used as the sodium level for that day	Pulmonary complication 30-d mortality Late mortality Overall mortality Length of hospitalization Operative complication Infectious complications Neurologic complications Renal failure Dialysis (renal failure requiring dialysis) Mortality Operative complication (perioperative MI; reoperation for bleeding, for cardiac tamponade, for native or prosthetic valve dysfunction, and for graft occlusion; other cardiac reoperations; and noncardiac reoperations) Pulmonary complication Infection complication Renal failure Blood transfusion required	Varies for each outcome measure. Refer to Table 3 of original article (28)	8
Boin, 2010 (26) Brazil	Retrospective data review	318 71.10% 46.45	OLT using the piggyback techniques	Liver transplant	≤ 130 mEq/L	NR	Mortality	NR	7
Hennrikus, 2015 (35) USA	Retrospective data review	748 Normonatremia = 41.6% Hyponatremia = 39.4% Normonatremia = 60.5 ± 14.5 Hyponatremia = 67.2 ± 15.5	All consecutive hospitalized orthopedic surgery patients	Orthopedic surgeries (TKR, THR, spine fusion, others)	<135 mEq/L	Within 30 d preoperatively	Non-home discharge	Model, adjusted for age, sex, BMI, operating time, blood loss, type of procedure, and Charlson comorbidity index as a composite of comorbidities was used to determine the significance of the association between the type of hyponatremia and discharge to an extended-care facility	8
Hefler-Frischmuth, 2018 (34) Austria	Retrospective cohort study	498 0.00% 58.7	Patients with epithelial ovarian cancer	Cytoreductive surgery for epithelial ovarian cancer	≤134 mmol/L	24-72 h before surgery	Overall mortality	NR	8
Karapanagiotou, 2012 (36) Greece	Retrospective data review	80 65.21% 51.68	Patients younger than 18 years and patients with retransplantations, synchronous hepatorenal transplantations, or acute hepatic failure were excluded from the study	Cadaveric donor liver transplant	<130 mEq/L	NR	30-d mortality 1-y mortality Acute renal failure	Renal failure, and neurological disorders as categorical variables, as well as MELD, APACHE II SOFA scores, and age as continuous variables	8
Dawwas, 2007 (29) UK, Ireland	Prospective	5152 59.60% 50.4	Adults with chronic liver disease underwent a first single-organ liver transplant	First single-organ liver transplant	<135 mEq/L Mild hyponatremia: 130-134 mEq/L, severe hyponatremia: <130 mEq/L)	Immediately before transplantation	90-d mortality Overall mortality Renal complication Renal failure requiring dialysis Vascular complications Acute rejection Biliary complications	NR	8
Berardi, 2020 (25) Italy	Retrospective study	89 54.00% 66	Patients with stage I-III pancreatic ductal adenocarcinoma	Radical surgery for pancreatic ductal adenocarcinoma	<135 mEq/L	NR	Morbidity, postoperative stay	NR	7
Cecconi, 2016 (9) United Kingdom	Retrospective study	35 816 49% 59	All patients undergoing noncardiac surgery in participating hospitals	Inpatient noncardiac surgery	≤137 mmol/L (mild hyponatremia)	Preoperative serum sodium measurement was the most recent measurement performed within the 28 d before surgery	In-hospital mortality, censored at 30 d	NR	8
Ayus, 2020 (24) Argentina	Retrospective cohort study	1571 19.6% 82 (normonatremia), 86 (chronic prolonged hyponatremia), 83 (recent hyponatremia)	Patients with hip fracture	Hip fracture repair	<135 mmol/L	Within 30 d before admission with previously normal plasma sodium (recent hyponatremia)	Outcomes of interest included both in-hospital and postdischarge outcomes. Postoperative complications (within 30 d after hip surgery) that were evaluated included the development of acute MI, heart failure, significant atrial or ventricular arrhythmias, stroke, venous thromboembolism (deep venous thrombosis or pulmonary embolism), and sepsis. In-hospital mortality, LOS (in days), hospital readmission for any cause within 30 d after discharge, and death from any cause during follow-up	Adjusted for the propensity score for presenting with hyponatremia as a continuous variable in addition to directly adjusting for age, sex, BMI, preadmission diagnosed dementia, prior heart failure, preexisting CKD, and risk factors for hip fracture Adjusted for propensity to present with hyponatremia, age, sex, and Charleston comorbidity score	8

Abbreviations: AMI, acute myocardial infarction; ASA, American Society of Anaethesiologists; BMI, body mass index; MELD, Model for End-Stage Liver Disease; CABG, coronary artery bypass graft; CHF, congestive heart failure; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; HBV, hepatitis B virus; LDLT, living donor liver transplantation; LOS, length of hospital stay; LT, liver transplantation; LVEF, left ventricular ejection fraction; MI, myocardial infarction; NA, not applicable; NR, not reported; NSQUIP, National Surgery Quality Improvement Program; NYHA, New York Heart Association; OLT, orthotopic liver transplantation; THR, total hip replacement; TKR, total knee replacement; TIA, transient ischemic accident.

Study Characteristics

A total of 11 051 abstracts were identified, and 32 studies met our inclusion criteria (Fig. 1). These consisted of 29 retrospective studies (n = 327 561 patients) and 3 prospective studies (n = 973 785 patients) (14, 17, 28). Twelve studies included patients who underwent orthopedic surgeries (16, 18, 24, 27, 32, 33, 35, 38, 41-43, 45); 10 studies involved patients who underwent liver transplantation (5, 7, 26, 29, 31, 36, 39, 40, 46, 48); 4 studies for cardiac surgical procedures (17, 28, 37, 44); 3 studies on surgeries of multiple disciplines (9, 14, 30); and 1 study each on radical gastrectomy (47), surgical procedures for epithelial ovarian cancer (34), and radical surgical procedures for pancreatic ductal adenocarcinoma (25). Sixteen studies included patients from North America (5, 7, 14, 16-18, 27, 28, 30, 32, 35, 37-39, 42, 43), 8 studies from Europe (9, 25, 29, 34, 36, 40, 41, 45), 5 from Asia (31, 33, 46-48), 2 from South America (24, 26), and 1 from Israel (44).

Eighteen studies included early mortality (5, 7, 9, 14, 16, 24, 26, 28, 29, 32, 33, 36, 37, 39-42, 44) and 9 studies included late mortality (7, 24, 28, 29, 36, 37, 39, 44, 48) as part of mortality outcomes. Sixteen studies included at least 1 outcome that qualified as a major complication (7, 14, 18, 24, 28-32, 36-38, 40, 42-44), 4 reported on respiratory complications (7, 14, 28, 32), 6 on cardiac complications (14, 24, 30, 32, 37, 44), 10 on renal complications (7, 28-32, 36, 37, 40, 44), 7 on neurological complications (7, 14, 24, 28, 31, 40, 44), and 5 on postoperative sepsis (7, 24, 29, 31, 32). Definitions of the outcome measures can be found in the following sections.

Risk of Bias

On assessment with Newcastle-Ottawa Scale, all the studies were allocated a score of 7 or higher, indicating that the studies were largely assessed to be high quality with a low risk of bias. We assessed the risk of bias using the Newcastle-Ottawa Scale, which grades studies as having a high (<5 stars, in total), moderate (5-7 stars), or low (≥8 stars) risk of bias (20). The full assessment can be found in Supplementary Table S2 (19).

Definition of Preoperative Hyponatremia

Twenty-two studies defined hyponatremia as serum sodium levels of either < or ≤135 mEq/L (5, 14, 16-18, 24, 25, 28-30, 32-35, 37, 38, 41-46), 8 defined hyponatremia as either < or ≤130 mEq/L (7, 26, 31, 36, 39, 40, 47, 48), 1 study used the cut off of ≤137 mEq/L to define hyponatremia (9), and 1 study did not specify the numerical definition of hyponatremia (27). Three of the studies reported outcomes stratified by the degree of severity of hyponatremia of patients (9, 16, 29).

Definition of Major Complications

We defined major complications as a composite outcome measure comprising the following 9 morbidities: surgical site infection, wound dehiscence, pulmonary complications, unplanned reintubation, renal complications, neurological complications, cardiac complications, deep venous thromboembolism, and sepsis. In line with previous large-scale clinical studies in the American College of Surgeons National Surgical Quality Improvement Program, this definition ensures the comprehensive inclusion of clinically relevant outcomes (38, 42). We excluded all measures of mortality in this composite measure because we conducted a separate analysis for mortality outcomes.

If a study reported 2 or more morbidity outcomes that qualified as major complications (eg, renal insufficiency, pneumonia), we selected the outcome with the larger raw frequency counts to prevent double representation of study participants (Table 2).

Table 2.

Summary of specific complications included in the composite indicator of major complications

	Major morbidity (NOS)	Surgical site infection	Wound dehiscence	Pulmonary complications	Unplanned reintubation	Renal complications	Neurological complications	Cardiac complications	Deep Venous thromboembolism	Sepsis
Abola 2019	1.05 (0.93-1.19)
Ayus 2020										1.84 (1.01-3.35)
Cresanello 2 2013				1.73 (1.32-2.27)		Renal failure: 1.34 (0.96-1.87) Renal failure requiring dialysis: 1.64 (1.11-2.44)	0.9 (0.54-1.50)
Gu 2020		Deep wound infection: 1.09 (06.78-1.77) Organ space infection: 2.32 (1.68-3.19)		2.05 (1.314-3.199)				1.197 (0.629-2.279)		2.533 (1.776-3.614)
Khan 2020						1.52 (1.20-1.93)	1.48 (1.09-2.02)
Lee 2020	2.00 (1.43-2.73)
Leung 2012				1.17 (1.12-1.22)			1.07 (0.96-1.20)	1.21 (1.14-1.29)
Pennington 2020	1.24 (1.10-1.40)
Pennington 2021	1.22 (1.03-1.44)
Shavit 2018						1.37 (1.10-1.70)

Bolded values represent the selected outcomes, with larger raw frequency counts, that were chosen to prevent double representation of study participants.

Definition of Specific Systemic Morbidities

Respiratory complications

Of the 5 included studies (7, 14, 28, 30, 32), all but 1 specified a timeframe 30 days from date of surgery as the inclusion period for postoperative respiratory complications (7, 14, 30, 32). Specific respiratory morbidities reported included pneumonia (all studies), unplanned intubation or reintubation (2 studies) (28, 30), ventilatory support for >24 hours after surgery (28), and pulmonary embolism (28). Hackworth et al and Leung et al reported figures specifically for pneumonia (7, 14), whereas the remaining studies reported respiratory/pulmonary complications as a composite figure comprising a combination of the aforementioned morbidities.

Renal complications

All 10 included studies reported on the renal complications of acute renal failure or acute kidney injury (7, 28-32, 36, 37, 40, 44). Crestenallo et al and Londono et al specified renal failure as either a postoperative serum creatinine concentration greater than 2 mg/dL and/or 2-fold increase from preoperative values (28, 40). Five of the included studies specified the need for renal replacement therapy/dialysis in their definition of renal impairment (7, 28, 29, 31, 44). Feinstein et al included urinary tract infections in the reported figures for renal complications (30).

Sepsis

All 5 included studies defined sepsis as the detection of pathogenic organisms in blood cultures (24, 29, 31, 32, 40).

Neurological complications

Of the 7 included studies (7, 14, 28, 31, 37, 40, 44), all but 2 included postoperative stroke and/or transient ischemic attack in their outcome measures (7, 14, 28, 37, 44). Other neurological complications assessed included coma, paralysis or paraplegia, and seizures. Londono et al and Fukuhara et al reported on neurological disorders defined as the development of abnormal neurological functions (31, 40). Hackworth et al and Fukuhara et al included central pontine myelinolysis as part of the outcome measures (7, 31). Notably, Hackworth et al reported on delirium, which included cases unresponsiveness resulting from stroke or central pontine myelinolysis (7).

Cardiac complications

Ayus et al, Feinstein et al, Shavit et al, Gu et al, and Leung et al included perioperative myocardial infarction as part of cardiac complications (14, 24, 30, 32, 44), with Gu et al and Leung et al also including cardiac arrest as part of the definition (14, 32). Shavit et al also included congestive cardiac failure with New York Heart Association functional class III-IV and low cardiac output under cardiac complications (44). Khan et al defined cardiac complications solely by the presence of atrial fibrillation (37).

Definition of Mortality

Early mortality

We defined early mortality as postoperative mortality at first time point of measurement reported in the study, with duration of <90 days. Twelve studies defined early mortality as 30-day all-cause mortality (death occurring within 30 days of surgical procedures regardless of cause, in or out of the hospital, including intraoperative and postoperative deaths) (7, 9, 14, 16, 24, 28, 32, 36, 39, 41, 42, 44). Four studies used 90-day all-cause mortality as outcome (5, 26, 29, 40). Khan et al reported on operative mortality (37), whereas Hagino et al reported on inpatient mortality without a specified duration (33).

Late mortality

Late mortality was defined as any death that occurred both after discharge from the hospital and more than 90 days from the operative procedure.

Definition of extended length of hospitalization

Regarding outcomes of use of hospital resources, 9 studies reported on length of hospitalization (18, 27, 28, 30, 32, 37, 38, 42, 43), 2 on non-home discharge (35, 38), 6 on 30-day readmission (18, 24, 32, 38, 42, 43), and 4 on 30-day reoperation (18, 38, 42, 43).

Abola et al, Lee et al, Pennington et al (2020), and Pennington et al (2021) defined extended length of stay as one that was greater than the 75th percentile of the entire cohort (18, 38, 42, 43). Gu et al defined extended length of stay as an inpatient hospital stay >7 days (32). The rest of the studies defined length of hospitalization as the number of days postoperative to discharge (27, 28, 30, 37).

Association of Preoperative Hyponatremia With Surgical Outcomes

Major complications

Meta-analysis

Compared with the control group (patients with normonatremia), patients with hyponatremia had significantly higher odds of developing major postsurgical procedures complications, defined as a composite measure of 9 major complications (adjusted odds ratio [aOR] = 1.37; 95% CI, 1.23-1.53; I² = 78%; N = 10) (Figure 2A) when adjusted for covariates. Table 2 summarizes the specific major complications identified in each study.

Figure 2.

Meta-analysis of the adjusted associations of preoperative hyponatremia with (A) major complications and (B) early mortality.

Subgroup analyses and meta-regression

From subgroup analyses of the adjusted association by type of surgical procedures, this association of hyponatremia and increased risk of major complications was preserved in all subgroups (Supplementary Fig. S1a (19)). In the orthopedics subgroup, the association remained significant with reduced between-study heterogeneity (aOR = 1.40; 95% CI,= 1.16-1.68; I² = 81%; N = 6). The association also remained significant in the cardiac subgroup with elimination of heterogeneity (aOR = 1.51; 95% CI, 1.31-1.73; I² = 0%; N = 3).

To further account for the observed between-study heterogeneity and to identify potential confounders, we also performed univariate meta-regression based on 2 variables: the mean age of study participants and sex proportions (given by percentage of males). Notably, sex was identified as a significant contributor of heterogeneity accounting for 23.34% between-study heterogeneity (Supplementary Table S3 (19)). Meta-analysis based on serum sodium cutoff value was not performed given that the same cutoff of <135 mEq/L were used for all included studies.

Publication bias and influence analysis

By visual inspection of funnel plot, we observed asymmetry (Supplementary Fig. S1b (19)). This was confirmed on the Egger linear regression test in which funnel plot asymmetry was detected (P = 0.0095).

Specific systemic complications

We also conducted further exploratory meta-analyses looking at the association of preoperative hyponatremia with the specific systemic complications including respiratory, renal, neurological, cardiac, biliary complications, and sepsis. Of note, only meta-analyses of respiratory complications (aOR = 1.47; 95% CI,= 1.07-2.01; I² = 82%; N = 3), renal complications (aOR = 1.42; 95% CI, 1.23-1.64; I² = 0%; N = 3), and sepsis (aOR = 2.33; 95% CI,= 1.72-3.17; I² = 0%; N = 2) yielded statistically significant adjusted associations (Fig. 3A-3C).

Figure 3.

Significant associations of preoperative hyponatremia with specific systemic morbidities: (A) respiratory complications, (B) renal complications, and (C) sepsis.

There were no significant adjusted associations between preoperative hyponatremia and neurological complications (OR = 1.14; 95% CI, 0.93-1.4; I² = 33%; N = 4) (Supplementary Fig. S2a (19)) and cardiac complications (aOR = 1.05; 95% CI, 0.80-1.40; I² = 87%; N = 4) (Supplementary Fig. S2b (19)).

Early mortality

Meta-analysis

Preoperative hyponatremia was associated with higher adjusted hazard of early mortality. Patients with hyponatremia had a significantly increased hazard of early mortality compared with patients with normonatremia (adjusted hazard ratio [aHR] = 1.27; 95% CI; 1.13-1.43; I² = 97%; N = 10) after adjusting for covariates (Fig. 2B).

Subgroup analyses and meta-regression

Subgroup analysis by degree of hyponatremia showed that more severe hyponatremia was associated with a higher hazard of early mortality (serum sodium ≤130 mEq/L: aHR = 1.59; 95% CI,= 1.26-2.02; I² = 19%; N = 3 vs serum sodium 131-135 mEq/L: aHR = 1.32; 95% CI, 1.04-1.67; I² = 52%; N = 3) (Supplementary Fig. S3a (19)).

Further prespecified subgroup analysis of adjusted association by type of surgery (Supplementary Fig. S3b (19)) demonstrated that the association in the orthopedics surgery subgroup remained significant with eliminated between-study heterogeneity (aHR = 1.47; 95% CI,= 1.25-1.72; I² = 22%; N = 4). The association also remained significant in the others/general subgroup, which comprised patients undergoing noncardiac (9), cardiac (28), and any major surgeries (14) (aHR = 1.34; 95% CI, 1.12-1.60; I² = 80%; N = 3). This association was also near statistical significance in the liver transplant subgroup (aHR = 1.02; 95% CI, 0.98-1.06; I² = 73%; N = 3).

To further account for the observed between-study heterogeneity and to identify potential confounders, we performed univariate meta-regression based on 3 variables: the mean age of study participants, and sex proportions (given by percentage of males), and serum sodium cutoff value for determination of preoperative hyponatremia (Supplementary Fig. S3cS3d (19)). Data and variables used for meta-regression are found in Table 1. Notably, mean age of study participants was identified as a significant effect moderator and accounted for 45.92% of between-study heterogeneity (Supplementary Table S4 (19)).

Publication bias and influence analysis

By visual inspection of the funnel plot, we observed minimal asymmetry (Supplementary Fig. S3e (19)). This was reinforced with the Egger linear regression test in which funnel plot asymmetry was not detected (P = 0.1332).

Late mortality

We also performed meta-analysis on the association of preoperative hyponatremia with late mortality. Most studies reported higher adjusted odds of late mortality for patients with hyponatremia undergoing surgery when compared with the control group (aOR = 1.20; 95% CI, 1.05-1.36; I² = 93%; N = 7) (Supplementary Fig. S4 (19)).

Length of hospitalization stay

We also explored the association of preoperative hyponatremia with length of hospital stay, as an indicator of its effect on the utility of hospital resources. There was also a significant adjusted association (aOR = 1.37; 95% CI, 1.22-1.53; I² = 90%; N = 9) noted for preoperative hyponatremia and length of hospital stay (Supplementary Fig. S5 (19))

Performance of Preoperative Hyponatremia in Predicting for Major Complications

Preoperative hyponatremia performed adequately in predicting for major complications based on the summary receiver operator characteristic curve (area under the curve = 0.70; 95% CI, 0.66-0.74) (Fig. 4). In particular, it was 88% specific (95% CI, 84-92) for a poor outcome albeit with low sensitivity of 25% (95% CI, 17-34). Visual inspection of coupled funnel plots did not indicate a clear threshold effect (Supplementary Fig. S6 (19)).

Figure 4.

Summary receiver operator characteristic curve of prognostic performance of preoperative hyponatremia and major complications.

Assuming a 10% pretest probability of developing major morbidity among patients based on figures published in a previous large-scale surgical cohort study (42), the presence of hyponatremia (positive likelihood ratio, 2) would be associated with a 19% posttest probability (positive predictive value = 19%) of developing major complication as illustrated by the Fagan nomogram (Supplementary Fig. S7 (19)).

As a measure of calibration, we performed a meta-analysis of the O:E ratio. The meta-analysis summary estimate for O:E ratio was 1.80 (95% CI, 1.49-2.16; 95% prediction interval, 1.04-3.11; Supplementary Fig. S8 (19)), indicating on average that 80% fewer deaths were observed than predicted by the presence of preoperative hyponatremia.

The meta-analysis of concordance (C) statistic (as a measure of discrimination) however was not possible because the included studies, being observational in nature, did not report C statistic nor other quantities (eg, linear predictor, Somer D statistics) that may be used to estimate the C statistics (49).

Discussion

In our study, we observed that preoperative hyponatremia was associated with a 2-fold increased risk of early mortality and 2.5-fold increased risk of major complications after surgical procedures across 1 301 346 participants included in 32 studies. This analysis was comprehensive and included all types of surgeries, patient populations, and patient outcomes. Meta-analyses were conducted on maximally adjusted estimates to minimize potential confounding effects of covariates, whereas subgroup and sensitivity analyses were also performed to account for between-study heterogeneity.

Our study also included secondary analyses on the performance of preoperative hyponatremia in predicting for major complications in surgical patients. We observed that it was 88% specific and 25% sensitive in predicting for major complications. The low sensitivity may limit the clinical use of hyponatremia in ruling out postoperative complications. This may be attributable to the wide range of preoperative causes of postsurgical complications, some of which may not be associated with hyponatremia. Nonetheless, its high specificity provides impetus for future surgical outcome prognostic models to include hyponatremia as one of the factors considered.

To the best of our knowledge, this is the first systematic review and meta-analysis, and the most comprehensive evidence-based clarification to date, on the associations of preoperative hyponatremia with surgical outcomes. Previous epidemiological studies established that hyponatremia, regardless of its etiology, entails poorer prognoses for hospitalized patients (50, 51). The results of our study therefore add to these findings, demonstrating specifically that preoperative hyponatremia was associated with poorer mortality and morbidity outcomes in surgical patients (52).

An interesting finding of our study is that even in preoperative patients with only a mild decrease in serum sodium (131-135 mEq/L), there is a statistically higher adjusted hazard of early mortality (aHR = 1.32; 95% CI, 1.04-1.67; I² = 52%; N = 3). The subgroup analysis we performed also revealed that the hazards of early mortality increase with the severity of hyponatremia. This finding is also in agreement with the finding of a previous meta-analysis conducted on all hospitalized patients with multiple disease types that reported that even a moderate reduction of serum sodium was associated with increased mortality (51).

Symptoms and consequences of hyponatremia can vary from mild to life-threatening. Hyponatremia, a disorder of water and sodium balance, and thus plasma osmolality, causes increased risk of adverse events including cerebral edema, falls, fractures, and osteoporosis (53). Studies in animals showed that hyponatremia exacerbates senescence manifestations in tissue organs (54) and increased expression of apoptotic proteins leading to accelerated cell death (55). Although hyponatremia may be associated with increased risks of mortality and morbidity, hyponatremia may also be a marker of poor outcome attributable by the patient’s underlying comorbidities, whether diagnosed or undiagnosed. Among hospitalized patients, euvolemic hyponatremia is the most common cause of hyponatremia, with syndrome of inappropriate antidiuretic hormone secretion being the most common underlying etiology (56). This is particularly common in patients with diseases related to the central nervous system, tumors, or secondary to medications (57). As such, clinical judgment of physicians is paramount in correctly assessing possible causes for patients’ hyponatremia before surgical procedures, and to treat them accordingly. This is in particularly important in patients with underlying comorbidities in which refractory hyponatremia may manifest, such as in patients with severe liver cirrhosis and congestive cardiac failure (58, 59). For instance, hyponatremia has been recognized as an independent predictor of mortality in end-stage liver disease, leading to the incorporation of serum sodium levels in the calculation of the Model for End-Stage Liver Disease (60).

The increased mortality observed may also be a result of the increased risk of specific systemic major complications postoperatively, as demonstrated in our meta-analysis. A previous study reported that myocardial injury after noncardiac surgical procedures and sepsis were 2 of the most important postoperative complications independently associated with early mortality (61). Both of these morbidity outcomes were included as part of the composite measure of major complications in our study. Separately, another study noted that preoperative hyponatremia showed a strong association with postoperative infections, with patients with preoperative hyponatremia more than 3 times more likely to develop postoperative infections (62). The study reported that this was possibly attributable to the impaired function of helper T cells, which are quintessential in regulating host immune response and preventing breakdown of microbial target function that occurs because of cellular edema of mucosal membranes (63).

Relating to the utility of hospital resources, we have also observed that hyponatremia is associated with increased odds of longer hospital stays, perhaps because hyponatremia impairs patient rehabilitation and slows recovery (56). At the pathophysiological level, it has been hypothesized that a longer hospital stay may be attributable to neurological dysfunction that results from cerebral oedema and ischemia associated with hyponatremia (64).

Preoperative hyponatremia can be multifactorial, and determination of underlying etiology is crucial for treatment. There are various proposed established algorithms to do so, and most of the etiologies are identifiable with simple screening tests and are treatable (56). This is especially important for causes, which may have unpredictable perioperative implications if undiagnosed, such as renal impairment, cerebral salt-wasting syndrome, heart failure, and liver disease, among others (1).

Some limitations to our study must be acknowledged. First, because of a lack of sufficient studies on the association of preoperative hyponatremia with specific systemic outcomes and late mortality, meaningful sensitivity analyses to account for the observed between-study heterogeneity could not be performed. Second, we were unable to conduct meta-analyses based on the urgency of the surgery because this was not explicitly reported by most of the studies. Third, we were also unable to evaluate for possible differences in surgical outcomes based on chronicity of preoperative hyponatremia because this was not stratified or reported by most of the studies. In addition, some of the studies did not specify the exact time of sodium measurement. However, Ayus et al (24) and Verghese et al (15) showed that chronic prolonged hyponatremia was associated with increased risk of postoperative death and prolonged hospital stay, respectively. Our study also did not address the clinical question of whether correction of preoperative hyponatremia will change outcomes because this was not investigated in most studies (65). Finally, this meta-analysis may be limited by the type of primary studies available, which were largely observational studies. This analysis thus could not achieve a conclusive analysis of prognostic performance. Such an analysis may be better answered by future prognostic studies, with data on calibration and discrimination available.

Conclusion

In conclusion, our study sheds light on the increased risk of poorer early mortality and morbidity surgical outcomes associated with preoperative hyponatremia, suggesting that clinicians should consider preoperative hyponatremia as a potential factor to risk-stratify patients planned for surgical procedures. This is particularly important for patients with more severe hyponatremia. Further studies are needed to elucidate the role of preoperative hyponatremia in accounting for poorer surgical outcomes, and if its correction will improve surgical outcomes.

Abbreviations

aHR

adjusted hazard ratio

aOR

adjusted odds ratio

C

concordance

O:E

observed to expected ratio

Funding

C.B.T. gratefully acknowledges the National University of Singapore Yong Loo Lin School of Medicine Dean’s Research Development Award and Singapore Public Service Commission Medicine Scholarship for supporting this work. The funders had no role in the design and conduct of the study, nor the decision to prepare and submit the manuscript for publication.

Author Contributions

All authors approved the final version of the manuscript. Conceptualization: C.B.T., M.Y.G., M.Y.G., and N.H.W.L. Data curation: C.B.T., M.Y.G., and R.Y.K.T. Formal analysis: C.B.T., M.Y.G., and R.Y.K.T. Investigation: C.B.T., M.Y.G., and R.Y.K.T. Methodology: C.B.T., M.Y.G., R.Y.K.T., W.J.L., and N.H.W.L. Software: C.B.T., M.Y.G., and R.Y.K.T. Supervision: W.J.L. and N.H.W.L. Writing – original draft: C.B.T., M.Y.G., and R.Y.K.T. Writing – review & editing: W.J.L. and N.H.W.L. M.Y.G., M.Y.G., and C.B.T. had access to the dataset and N.H.W.L. had final responsibility for the decision to submit the manuscript for publication.

Disclosures

The authors have nothing to disclose.

Availability of Data and Analysis

All data generated or analyzed are included in this article and supplementary files.