Prediction of Postoperative Hypokalemia in Patients with Severe Carotid Artery Stenosis undergoing Standard Carotid Endarterectomy: A Retrospective Cohort Study

Article information

Ann Geriatr Med Res. 2026;30(1):18-27

Publication date (electronic) : 2025 December 29

doi : https://doi.org/10.4235/agmr.25.0164

Suiyuan Hu^1,*, Xuan Lai^1,*, Yunfeng Han^,2

¹Department of Geriatrics, Peking University Third Hospital, Beijing, China

²Department of Neurosurgery, Peking University Third Hospital, Beijing, China

Corresponding authors: Yunfeng Han, MD Department of Neurosurgery, Peking University Third Hospital, No. 49, Huayuan North Road, Haidian District, Beijing 100191, China E-mail: hyfln@126.com

*These authors contributed equally to this study.

Received 2025 October 9; Revised 2025 November 27; Accepted 2025 December 22.

Abstract

INTRODUCTION

Hypokalemia, characterized by serum potassium concentration less than 3.5 mmol/L, is a pervasive clinical concern that becomes particularly critical in the perioperative setting.1-3) Although mild hypokalemia (serum K^⁺ 3.0–3.5 mmol/L) is often asymptomatic, epidemiological evidence recognized it as an increased risk of stroke and mortality.4)

Progressive declines in serum potassium levels can lead to significant systemic complications, including muscle weakness, flaccid paralysis, metabolic alkalosis, renal dysfunction, and paralytic ileus. Severe cases may provoke life-threatening cardiac arrhythmias, which can rapidly progress to respiratory failure and cardiac arrest.1,3,5) Its prevalence and impact are magnified in surgical populations,6) especially those with underlying cardiovascular comorbidities,7,8) such as patients undergoing carotid endarterectomy (CEA), a condition that primarily affects older patient populations.

In the surgical patient, even mild hypokalemia can lower the threshold for cardiac arrhythmias, particularly in the context of anesthesia, intraoperative stress, blood loss, and the use of vasoactive drugs.9-11) It can also potentiate the effects of neuromuscular blocking agents and delay recovery from anesthesia.12) For vascular surgery patients, this electrical instability poses a direct threat to cerebral and cardiac perfusion.13-15) Postoperative hypokalemia further exacerbates arrhythmogenic susceptibility, impairs hemodynamic stability, and is associated with an increased incidence of adverse neurological and cardiovascular events.13-15) Moreover, preoperative hypokalemia has been established as an independent predictor of 30-day major adverse cardiovascular events after noncardiac surgery.2)

Consequently, prompt identification and correction of hypokalemia are imperative to optimizing postoperative recovery and outcomes, particularly in older patients who tolerate electrolyte imbalances poorly. Nonetheless, a clear understanding of predictive factors and a validated risk assessment tool for postoperative hypokalemia specifically following CEA, with consideration of geriatric-specific risk profiles, remains lacking.

Therefore, the primary aim of this study was to identify perioperative factors associated with potassium dysregulation that correlate with the incidence of postoperative hypokalemia in patients undergoing standard CEA, with emphasis on variables relevant to aging populations. Based on these factors, we sought to develop and validate an individualized predictive nomogram to facilitate early risk stratification and tailored perioperative management.

MATERIALS AND METHODS

Study Population

This retrospective cohort study recruited 1,189 patients with severe carotid artery stenosis who underwent standard CEA at Peking University Third Hospital from October 2021 to May 2023.

The inclusion criteria were as follows: (1) diagnosis of severe carotid artery stenosis (NASCET standard) and undergoing standard CEA at Peking University Third Hospital, and (2) availability of standardized electrolyte assessments both preoperatively and postoperatively.

The exclusion criteria were as follows: (1) nonatherosclerotic causes of carotid artery stenosis, such as arteritis or arterial dissection; (2) incomplete laboratory test results; (3) coexisting primary disorders that cause hypokalemia, including primary aldosteronism, renal tubular acidosis, Cushing's syndrome, and hereditary conditions such as hypokalemic periodic paralysis; and (4) long-term use of medications that significantly affect serum potassium levels, such as oral potassium supplements or diuretics.

A total of 1,076 patients with complete datasets were ultimately enrolled (Fig. 1). The study protocol was approved by the Ethics Committee of Peking University Third Hospital (ethics number: S2018206). Written informed consent was acquired from all participants prior to their enrollment in the study. All procedures were conducted in accordance with the Declaration of Helsinki.

Fig. 1.

Predictive nomogram for postoperative hypokalemia risk. The model incorporates preoperative and postoperative variables, including potassium (K), hemoglobin (HGB), body mass index (BMI), postoperative visual analog scale (VAS) score, and frailty. To use the nomogram, locate each patient's value on the corresponding variable axis, draw a line upward to the points axis to determine the points for each variable, sum all points to obtain the total points, and finally draw a line downward to the hypokalemia risk axis to read the predicted probability of hypokalemia.

RESULTS

Baseline Characteristics

A total of 1,076 patients were ultimately included in the analysis. The enrollment flow diagram is shown in Fig. S1.

The vast majority of patients (82.1%, n=883) had a normal serum potassium level (K≥3.5 mmol/L). Hypokalemia was present in 17.9% (n=193) of cases, with most (17.4%, n=187) falling into the mild range (3.0≤K<3.5 mmol/L). No cases of severe hypokalemia (K<2.5 mmol/L) were observed.

Demographic characteristics, medical history, and laboratory parameters are summarized in Table 1. The entire cohort comprised 889 (82.6%) males and 187 (17.4%) females, with a median age of 65 years (range, 35–88 years). The median preoperative potassium level was 4.10 mmol/L. Patients in the postoperative hypokalemia group had significantly lower body mass index (BMI), hemoglobin, preoperative potassium, and low-density lipoprotein cholesterol (LDL-C) levels, along with higher pain scores (VAS), and a greater proportion of frail individuals (all p<0.05). No other significant differences were observed in demographics, comorbidities, or remaining laboratory parameters.

Table 1.

Baseline characteristics of the overall study population

In the subgroup of patients aged ≥70 years (n=346; hypokalemia n=56 and normokalemia n=290), those with postoperative hypokalemia showed lower BMI, hemoglobin, creatinine, albumin and preoperative potassium levels, and higher CK-MB levels, a higher proportion of females and frailty (all p<0.05) (Table S1).

When analyzing patients within the normokalemic range, significant differences emerged between those maintaining postoperative K^⁺ >4.0 mmol/L versus 3.5–4.0 mmol/L. The high-normal potassium group was older and exhibited elevated preoperative potassium, blood urea nitrogen (BUN), and creatinine, alongside lower estimated glomerular filtration rate (eGFR), reduced LVEF, and higher diabetes prevalence (all p<0.05). Complete results are available in Table S2.

Univariate Analyses for Factors Associated with Postoperative Hypokalemia

Results from univariate analysis indicated that BMI, hemoglobin (HGB), K, frailty, postoperative VAS, creatinine, and LDL-C exhibited statistically significant differences between the hypokalemia group and the normokalemia group (Table 2). Variables with p<0.05 were included in the subsequent binary logistic regression analysis (Fig. S2).

Table 2.

Univariate analysis of predictors in the overall cohort (p<0.05)

Among patients aged ≥70 years, BMI, HGB, K, ALB, Cr, sex, frailty, CK-MB, and postoperative VAS were significantly associated with postoperative hypokalemia (all p<0.05) (Table S3).

Multivariate Analysis and Nomogram Construction

Multivariate analyses identified preoperative potassium (OR=0.60, 95% CI 0.50–0.72), HGB (OR=0.74, 95% CI 0.63–0.88), BMI (OR=0.74, 95% CI 0.63–0.88) and postoperative VAS (OR=1.28, 95% CI 1.09–1.51) as independent predictors of postoperative hypokalemia in patients with severe carotid artery stenosis following standard CEA (all p<0.05). Frailty showed a borderline association with postoperative hypokalemia (OR=1.56, 95% CI 1.00–2.44, p=0.05). ORs with 95% CI are presented in Table 3.

Table 3.

Multivariate regression analysis of the overall cohort

A nomogram was developed based on these independent predictors to estimate the individual risk of postoperative hypokalemia (Fig. 1).

Internal Validation of the Nomogram

Internal validation was performed using 500 bootstrap resamples. The AUC for the nomogram was 0.710, indicating good discriminatory ability (Fig. 2A). Calibration plots demonstrated strong agreement between predicted and observed probabilities of hypokalemia (Fig. 2B), confirming the model’s robustness and clinical applicability. In summary, the nomogram exhibits considerable discriminative power and calibration accuracy.

Fig. 2.

Internal validation of the nomogram. (A) The receiver operating characteristic curves of the nomogram model. The area under the curve (AUC) of 0.710 quantifies overall discriminative ability (0.5 = random chance, 1.0 = perfect discrimination). (B) Calibration plot of the nomogram model. The 45° diagonal represents the Ideal line (perfect calibration). The Apparent line shows observed performance in this dataset; the Bias-corrected line (from bootstrap resampling) adjusts for overfitting, providing a more realistic estimate for new patients.

DISCUSSION

This cohort study establishes that postoperative hypokalemia is a common complication, occurring in 17.9% of patients following standard CEA for high-grade stenosis—a finding consistent with the broader hospital-based literature reporting rates of 14%–40%.1) Such potassium (K^⁺) imbalances are frequently observed in the perioperative setting, with an incidence of 0.2%–16.0% in the general surgical population, escalating markedly to 2.9%–71.0% among high-risk patient cohorts.16)

We focused specifically on CEA because this standardized vascular procedure is performed in a patient population particularly susceptible to hypokalemia due to prevalent cardiovascular comorbidities and age-related physiological changes.6-8) Furthermore, in CEA patients, even mild hypokalemia carries heightened clinical significance, as electrolyte disturbances can precipitate arrhythmias or hemodynamic instability that may compromise cerebral perfusion critically.13-15) This procedural homogeneity helps control for surgical variability and strengthens internal validity. However, we acknowledge that this focus inherently limits generalizability to other surgical contexts. Future research should validate and adapt this predictive model in diverse surgical cohorts to assess broader clinical applicability.

In the present study, preoperative potassium, HGB, BMI, frailty and postoperative VAS were identified as independent risk factors for postoperative hypokalemia. Among these, preoperative serum potassium, HGB, and BMI were protective factors, while frailty and postoperative VAS score were risk factors. This suggested that the occurrence of hypokalemia after CEA arose from the synergistic interplay of multiple pathophysiological mechanisms, primarily involving pre-existing potassium depletion, frailty and compromised nutritional status, and pain-induced stress responses.

The strong inverse correlation between preoperative potassium levels and the risk of postoperative hypokalemia highlighted the critical role of pre-existing potassium depletion. This observation was consistent with previous studies across various surgical contexts,17-21) which collectively underscored that preexisting hypokalemia or even physiologically low-normal potassium reserves, constituted the strongest predictor of postoperative hypokalemia. For instance, a retrospective cohort study on patients undergoing en bloc resection for oral cancer identified a preoperative serum potassium cutoff of 3.98 mmol/L as a significant predictor of postoperative hypokalemia (relative risk=1.76).17) Similarly, another study in the same population reported that preoperative levels below 3.87 mmol/L were independently associated with hypokalemia (OR=2.484, p=0.008).21) These thresholds lied above the conventional lower limit of normal (3.5 mmol/L), supporting the notion that even within the normal range, relatively lower levels may reflect ongoing potassium depletion and reduced buffering capacity, thereby increasing perioperative risk.21) Surgical stress exacerbates potassium loss and transcellular shifts. In patients with preoperative negative potassium balance, sympathetic activation22) and insulin secretion23) further promoted intracellular shifting. Consequently, there were growing evidence to suggest that the potassium target should be raised to at least 4.0 mmol/L, with proactive supplementation considered when levels fall below.1,17,24) The aforementioned findings were supported by studies in elderly patients and those undergoing total joint arthroplasty, which also identified preoperative hypokalemia as an independent risk factor for postoperative potassium disturbances.18-20)

Although neither group in our study met the criteria for anemia, a modest peri-operative decline in hemoglobin was still associated with a higher risk of postoperative hypokalemia. This observation was consistent with the findings of Pan et al.,19) who identified preoperative red blood cell count as an independent predictor of hypokalemia after total joint arthroplasty (OR=0.417, p=0.027) and suggested that the effect might be attributable to intra- and postoperative loss of red blood cells and fluids. From a pathophysiological perspective, the following mechanisms may explain this link: (1) acute loss of red-cell mass reduces total-body potassium buffering capacity25); (2) consequent blood-volume contraction triggers sympathetic activation and subsequent β₂-adrenergic stimulation of the Na^⁺/K^⁺-ATPase, thereby promoting a shift of potassium into cells.26,27)

Large-scale observational studies have consistently demonstrated a strong association between nutritional vulnerability and hypokalemia. In the multinational PDOPPS cohort, lower serum potassium was associated with poorer nutritional status, reflected through reductions in BMI, body weight, serum albumin, phosphorus, urea, and muscle mass, along with elevated inflammatory markers and impaired residual kidney function.28) Similarly, in patients with anorexia nervosa during refeeding, low BMI, together with hypoalbuminemia and binge-purge behavior, served as a critical predictor for hypokalemia risk, which could lead to life-threatening conditions.29) Low BMI and frailty represented overlapping clinical manifestations of nutritional depletion that converge on potassium homeostasis disruption through shared mechanisms. Compared with non-frail older adults, those with severe frailty exhibited a stepwise increase in electrolyte disturbances, including hypokalemia.30) Central to this relationship is the reduction in lean body mass, as skeletal muscle contain approximately 80% of the intracellular potassium and was the single largest pool of body potassium.31) Lower BMI is often indicative of reduced muscle mass, directly limiting potassium storage capacity.32,33) This was especially consequential in older adults, in whom sarcopenia and frailty frequently coexist.34,35) Moreover, frailty often accompanied and exacerbated chronic malnutrition and inadequate intake36); it was also associated with comorbidities and polypharmacy that promote potassium wasting.37-39) Under physiological stress such as surgery, these patients lacked metabolic reserve to compensate for potassium losses from bleeding, drainage, or internal shifts, markedly increasing their susceptibility to hypokalemia. Empirical evidence supported this elevated risk. Frailty has been identified as an independent predictor of hypokalemia in patients with acute ischemic stroke,40) and significantly increased the risk of postoperative complications across in non-cardiac and cardiac surgery, with higher morbidity, longer hospital stay, and greater mortality compared with non-frail patients.41,42) The consensus for orthopedic patients advocated systematic peri-operative management of frailty, including regular electrolyte monitoring to prevent frailty-related complications.43) In our study, although underlying medical histories were similar between groups, the higher prevalence of frailty in the hypokalemia group suggested that frailty captured a dimension of vulnerability not reflected by binary comorbidity data. This association was biologically plausible: frail individuals likely exhibited reduced total body potassium stores due to sarcopenia (muscle loss), chronic subclinical malnutrition with inadequate dietary potassium intake, and impaired neurohormonal responses to surgical stress. These mechanisms reflected decreased physiological reserve rather than specific disease diagnoses, explaining why frailty was elevated in the hypokalemia group despite comparable comorbidity profiles. In summary, low BMI and frailty were interrelated indicators of poor nutritional reserve that predisposed to hypokalemia through shared mechanisms—reduced potassium storage, impaired intake, and diminished adaptive capacity. Their co-occurrence identified a high-risk phenotype prone to postoperative hypokalemia requiring intensified monitoring and prevention in clinical practice.

Our study identified postoperative pain as an independent determinant of hypokalemia. This finding was consistent with previous research demonstrating that intravenous patient-controlled analgesia significantly reduced the severity and incidence of postoperative hypokalemia following laparoscopic cholecystectomy.44) The association was likely mediated through several mechanisms: First, pain-induced catecholamine release stimulated β₂-adrenergic receptors and activated Na^⁺/K^⁺-ATPase, promoting intracellular potassium translocation.9) Second, pain-related discomfort may suppress oral intake, thereby exacerbating negative potassium balance. Although patients in our study resumed oral intake within 6 hours post-surgery, inadequate consumption due to pain may have contributed to potassium depletion. Third, although less commonly observed, pain-associated anxiety could provoke hyperventilation, leading to respiratory alkalosis,45) which further facilitated potassium shifts into cells.46)

Furthermore, our analysis revealed a paradoxical association: patients who maintained higher postoperative potassium levels presented with baseline characteristics traditionally predictive of adverse outcomes. This counterintuitive observation invites mechanistic interpretation. The recently published POTCAST trial in the New England Journal of Medicine (NEJM) provides a critical framework for understanding this phenomenon.47) That multicenter, randomized study demonstrated that actively maintaining serum potassium in the high-normal range (4.5–5.0 mmol/L) reduced the composite risk of sustained ventricular tachycardia, appropriate ICD therapy, arrhythmic or heart failure hospitalization, and mortality by 24% (hazard ratio=0.76, 95% CI 0.61–0.95, p=0.01) in high-risk patients with cardiovascular disease.47) Building on this foundation, we hypothesize that the mild renal impairment frequently observed in older, comorbid patients may trigger compensatory potassium conservation mechanisms that inadvertently confer arrhythmic protection. This adaptive renal response could stabilize perioperative potassium homeostasis, effectively “pre-conditioning” these patients against ventricular arrhythmias despite their otherwise unfavorable risk profile.

The developed nomogram offers a practical tool for individualized risk assessment, and the predictive stability of these variables was further confirmed by machine learning algorithms. The nomogram achieved an AUC of 0.710, indicating moderate discriminative ability. While not exceptional, this performance is generally considered acceptable for initial model development, particularly for complex multifactorial outcomes like perioperative hypokalemia. The model's utility lies in its practical application as a low-cost screening tool using routine preoperative variables to identify high-risk patients who may benefit from increased vigilance and preemptive measures, rather than serving as a definitive diagnostic instrument. Its clinical value should be evaluated through prospective implementation studies focused on improving patient outcomes. For clinical implementation, this tool could be integrated into electronic health records during preoperative assessment to flag potentially high-risk patients. As an illustrative example, one might consider a predicted risk threshold, such as approximately 30%, which could represent a substantially elevated risk, to trigger a bundled care pathway:

1) Proactive potassium supplementation: For instance, patients with preoperative K^⁺ below an illustrative cutoff like 4.0 mmol/L might be candidates for supplementation.

2) Enhanced nutritional support: For example, patients with BMI below 24 kg/m² or HGB under 130 g/L might receive dietitian consultation. These values could serve as potential markers of compromised metabolic reserve, particularly relevant in certain populations.

3) Frailty-specific management: For frail patients, implement comprehensive geriatric assessment, medication review, and early mobilization.

4) Aggressive multimodal pain control: Prioritize regional anesthesia and scheduled non-opioid analgesia to maintain postoperative VAS scores at a low level (e.g., ≤3/10), which might help minimize catecholamine-driven potassium shifts.

5) Postoperative monitoring: For illustration, patients above a certain risk threshold, such as roughly 20%, could undergo serum potassium measurements at intervals like 6, 12, and 24 hours postoperatively.48) These stratified interventions could facilitate resource allocation and preemptive management in vulnerable older adults, though all proposed thresholds require prospective validation to establish optimal cutoffs and improve clinical efficacy.

Several limitations should be acknowledged. First, medication data that can influence potassium levels, such as diuretics, ACE inhibitors/ARBs, corticosteroids, and insulin, were not systematically collected.49) In this surgical cohort, loop diuretic use was uncommon, and thiazide exposure was likely limited to low-dose combination formulations. Nevertheless, this remains an important limitation, as uneven distribution of these medications between groups could have confounded our findings. Future studies should incorporate detailed medication records to better control for these factors and refine the model. Second, our study relied on a single postoperative potassium measurement within 24 hours. Serial measurements at key timepoints (e.g., immediately post-surgery in the post-anesthesia care unit, at 6, 24, and 48 hours) would provide a more comprehensive and physiologically relevant profile, enabling detection of transient hypokalemia, capturing trends over time, identifying distinct temporal risk patterns, and assessing the effectiveness of corrective interventions. Future studies should implement structured potassium monitoring protocols with time-series analysis to better guide intervention timing. Third, the single-center design limits external generalizability, and our internal bootstrapping validation may not capture population-level heterogeneity. While internal validation suggests robustness, the model currently lacks external validation. Future multicenter, prospective studies involving diverse patient populations, healthcare settings, and surgical teams are required to externally validate the nomogram before widespread clinical implementation. Finally, observational design precludes causal inference—randomized trials are needed to test whether correcting identified deficits improves outcomes.

In conclusion, we developed a validated tool to predict postoperative hypokalemia after CEA. The model highlights that in addition to biochemical and surgical factors (preoperative hypokalemia and pain), geriatric syndromes like frailty and nutritional status (reflected by low BMI and hemoglobin) are pivotal risk determinants. This facilitates early, individualized management, including tailored potassium supplementation, nutritional support, and aggressive pain control, especially for vulnerable older adults, to mitigate complications and promote recovery.

SUPPLEMENTARY MATERIALS

Article information Continued