Incidence,risk factors,and prediction model of piperacillin/tazobactam-associated hypokalemia in neonates

Abstract

Background:

Maintaining optimal potassium levels is critical in neonates. Piperacillin/tazobactam (TZP) is associated with hypokalemia in adults, but neonatal data are limited.

Objectives:

This study aimed to investigate the incidence, severity, and determinants of TZP-associated hypokalemia (TAH) in neonates and develop a predictive model for early detection.

Design:

Retrospective cohort study.

Methods:

Neonates treated with TZP between January 2019 and December 2024 at Xiamen Women and Children’s Hospital were included. Demographic, laboratory, and medication data were extracted. Naranjo probability scores assessed TAH causality. Multivariate logistic regression identified predictors, and a nomogram predicted TAH occurrence. Nomogram performance was evaluated.

Results:

Among 1027 neonates initially screened, 358 met the inclusion criteria and were ultimately included in the analysis. The incidence of TAH in this cohort was 20.4% (73/358). Independent predictors included higher baseline serum creatinine (p = 0.012, OR 1.02, 95% CI 1.00–1.03), dopamine use (p = 0.002, OR 3.37, 95% CI 1.54–7.36), and lower serum calcium (p = 0.049, OR 0.31, 95% CI 0.10–0.99) was a protective factor. The nomogram showed good predictive accuracy (AUC = 0.771) and net benefit. High-risk neonates had a higher hypokalemia incidence (log-rank test p < 0.001).

Conclusion:

Higher baseline serum creatinine, dopamine use, and lower serum calcium are independent predictors of neonatal TAH. The nomogram offers a user-friendly tool for TAH prediction, facilitating early detection and management in clinical practice. This study supports early detection, diagnosis, and intervention for neonatal TAH.

Plain language summary

Predicting potassium problems in newborns receiving piperacillin/tazobactam: a tool for early warning

What is this study about? Newborns in the hospital sometimes need strong antibiotics like piperacillin/tazobactam (TZP) to treat serious infections. However, this medicine can lower potassium levels in the blood (a condition called hypokalemia), which can be dangerous for a baby’s heart and muscles. This study looked at how often this happens in newborns and identified which babies are most at risk, so doctors can monitor them more closely. What did the researchers do? The researchers reviewed the medical records of 358 newborns who received TZP at a children’s hospital between 2019 and 2024. They checked which babies developed low potassium during treatment and used a special scoring system (the Naranjo scale) to confirm the link to the antibiotic. They then analyzed the babies’ characteristics, lab results, and other medications to find clues that might predict this side effect. What did the study find? 1. About 1 in 5 newborns (20.4%) developed low potassium while on TZP. 2. Newborns were at higher risk if they: Had higher levels of creatinine (a marker of kidney function) at the start of treatment. Were also receiving a medication called dopamine. Had lower levels of calcium in their blood. 3. Using these three factors, the researchers created an easy-to-use chart (nomogram) that helps doctors estimate a baby’s risk of low potassium before it happens. Why is this important? Low potassium can cause serious problems, especially in fragile newborns. This study gives doctors a practical tool to identify which babies need extra monitoring and early intervention while they are on TZP. By catching the problem early, doctors can prevent complications and keep newborns safer during treatment. What are the limitations? This study was done at one hospital and used past medical records, so further research in larger, more diverse groups of patients is needed to confirm these findings.

Keywords

hypokalemia incidence neonate piperacillin/tazobactam prediction nomogram risk factor

Introduction

Maintaining optimal blood potassium concentration is crucial for neonatal growth, development, and metabolic homeostasis.^1,2 Hypokalemia, defined as serum potassium <3.5 mmol/L, can result from three primary mechanisms: renal/enteric potassium wasting, intracellular shifting, or insufficient intake. This electrolyte imbalance manifests through neuromuscular symptoms (muscle weakness, ileus), cardiovascular complications (arrhythmias, ECG changes), and potentially progresses to respiratory paralysis or cardiac arrest in severe cases.^3,4 From an epidemiological perspective, multiple studies reveal hypokalemia affects over 20% of hospitalized patients (K⁺ < 3.5 mmol/L), with 3.5%–5% developing severe deficiency (<3.0 mmol/L).^5–7 Timely identification and correction of hypokalemia determinants, therefore, constitute critical care priorities.

Pharmaceutical agents constitute a common contributor to potassium depletion.^2,5,8,9 Besides conventional diuretics¹⁰ and corticosteroids,¹¹ certain medications can induce hypokalemia. Piperacillin-tazobactam (TZP), a β-lactam/β-lactamase inhibitor combination extensively used for neonatal sepsis and multidrug-resistant infection,^12–15 has drawn regulatory attention. Following a 2020 Japanese pharmaceutical alert,¹⁶ TZP’s safety profile now explicitly lists hypokalemia as a clinically significant adverse reaction.

Emerging evidence suggests a notable incidence of TAH. A randomized controlled trial comparing TZP with fosfomycin in urinary infections documented 12.6% TAH incidence.¹⁷ Risk stratification studies identify multiple predisposing factors: Seo et al.¹⁸ reported independent associations with advanced age, female sex, prolonged therapy (⩾7 days), and high-dose regimens (>300 mg/kg/day). By contrast, Kuramoto et al.⁷ identified age as the sole independent predictor, revealing significant knowledge gaps in neonatal TAH risk stratification. This evidence paucity underscores the urgent need for systematic investigation of TAH epidemiology in neonates. Elucidating incidence rates, severity patterns, and modifiable risk factors will inform clinical guidelines for safer TZP utilization in this vulnerable population.

Methods

This study was reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement.¹⁹ A completed STROBE checklist is provided as Supplemental File 1.

Study design and population

The retrospective study was conducted at Xiamen Women and Children’s Hospital in Fujian, China. We screened all neonates (aged ⩽ 28 days) who received intravenous TZP between January 1, 2019 and December 31, 2024. To be included, neonates had to meet all of the following criteria: (1) Age ⩽ 28 days; (2) received intravenous TZP therapy; and (3) serum potassium levels were measured during treatment. Participants were excluded based on the following criteria: (1) TZP treatment duration <3 days; (2) missing baseline/during/post-treatment serum potassium records; (3) pre-existing hypokalemia; and (4) dialysis/continuous kidney replacement therapy.

Data collection from electronic medical records comprised three domains: (1) Demographic characteristics: gender, gestational age (GA), postnatal age (PNA), postmenstrual age (PMA), birth weight (bWT), current weight (cWT), Apgar scores, and TZP administration details (dose, frequency, and duration); (2) laboratory parameters: baseline potassium (nearest pre-treatment measurement), treatment-phase potassium, recovery-phase potassium, renal/liver function; and (3) clinical data: concurrent medications and infection sites.

Concomitant medication analysis

All concomitant medications administered during TZP therapy were systematically recorded. To ensure analytical robustness and clinical relevance, medications were included in the regression analysis if they met either of the following criteria: (1) frequency of use >5% in the cohort and (2) known association with electrolyte disturbances based on prior literature (e.g., diuretics, vasoactive agents, corticosteroids, insulin). Drugs that could independently cause hypokalemia (e.g., furosemide, hydrocortisone) were retained in the univariate analysis. However, to avoid multicollinearity and overfitting, only variables with p < 0.01 in univariate analysis were entered into the multivariate model.

Definition and application of TAH cases using the Naranjo scale

TAH assessment employed the Naranjo probability scale, a validated 10-item questionnaire categorizing causality into four levels: definite (>9), probable (5–8), possible (1–4), and doubtful (<1). This algorithm systematically evaluates the probability that TZP induces hypokalemia by weighted scoring of temporal relationships, dose-response correlations, and alternative etiological factors. To ensure that the subsequent risk factor analysis specifically addressed hypokalemia attributable to TZP, a neonate was classified as a TAH case only upon fulfilling two criteria: (1) development of hypokalemia during TZP treatment and (2) a Naranjo score of ⩾1, indicating at least a “Possible” causal relationship.

Given the retrospective study design, certain Naranjo items—specifically Item 5 (rechallenge) and Item 7 (toxic drug concentrations)—were often unavailable and were therefore scored as “0” (unknown) in accordance with standard practice for retrospective adverse drug reaction (ADR) assessments. This approach preserved methodological consistency while acknowledging the inherent limitations of retrospective data in fully capturing all elements of the Naranjo algorithm.

Outcomes and definitions

The incidence, severity, and risk factors of TAH were evaluated. Multivariate logistic regression was performed to identify independent predictors and establish a nomogram to predict the occurrence of TAH.

The severity of hypokalemia is classified according to the Common Terminology Criteria for Adverse Events 5.0 (CTCAE 5.0). It is graded as follows: grade 0 for normal levels, grade 1 for levels between the lower limit of normal (LLN) and 3.0 mmol/L, grade 2 for levels between LLN and 3.0 mmol/L that require intervention, grade 3 for levels below 3.0–2.5 mmol/L that require hospitalization treatment, grade 4 for levels below 2.5 mmol/L that are life-threatening, and grade 5 for fatal cases. Grades 1 and 2 are classified as mild hypokalemia, while grade 3 is considered moderate hypokalemia. Grades 4 and 5 are categorized as severe hypokalemia.

Statistical analysis

Statistical analysis followed a structured approach with these four consecutive steps: First, categorical variables were summarized using percentages, while continuous variables were expressed as either mean with standard deviation (SD) for parametric data or median with interquartile range (IQR) for non-parametric data. Second, the Kolmogorov–Smirnov test was systematically applied to assess the normality assumption for all continuous variables. Third, based on these distributional assessments, we implemented appropriate statistical tests: the Student’s t-test for variables with normal or approximately normal distributions, and the Mann–Whitney U test for variables violating the normality assumption. Finally, categorical data comparisons were performed using either the Pearson Chi-square test or Fisher’s exact test, the latter being employed when expected cell counts fell below 5.

Statistical analyses proceeded in sequential phases. First, intergroup comparisons between hypokalemia and normokalemia cohorts were systematically performed for all study variables. The investigation then employed a two-stage regression approach: (1) comprehensive univariate logistic regression screening for potential predictors, followed by (2) refined multivariate logistic regression analysis (significance threshold p < 0.01) to identify independent determinants. This threshold was applied to minimize overfitting and prioritize strong predictors. The regression outcomes delivered critical effect estimates through odds ratios (ORs) with corresponding 95% confidence intervals (CIs). These validated predictors subsequently informed the development of a clinical prediction nomogram for neonatal TAH. The scoring system operates through three distinct steps: (1) cumulative score calculation by aggregating individual predictor values, (2) projection of total scores on the nomogram axis, and (3) vertical axis alignment to determine probabilistic TAH risk estimates.

The evaluation of predictive performance involved three complementary analytical approaches: First, discrimination capacity was quantified using the receiver operating characteristic (ROC) curve analysis with calculation of the area under the curve (AUC). Second, calibration accuracy was visualized through calibration curves plotting predicted probabilities against observed frequencies. Third, clinical utility was assessed via decision curve analysis (DCA) to estimate the model’s net benefit across different threshold probabilities. Internal validation was conducted through bootstrap resampling with 1000 iterations to evaluate the stability of both discrimination and calibration metrics.

Threshold determination for clinical stratification employed ROC-derived optimal cutoff values using Youden’s index, which maximizes the sum of sensitivity and specificity for hypokalemia prediction. This threshold enabled risk stratification of patients into distinct prognostic categories (low risk vs high risk) within the development cohort. Subsequent survival analyses incorporated Kaplan–Meier curves to visualize group-wise differences in tachyarrhythmia (TAH) incidence, with statistical comparisons performed using the log-rank test.

All analyses adopted a two-tailed testing framework with statistical significance defined as p < 0.05. Analytical processes were implemented using IBM SPSS Statistics (v25.0) and R Statistical Computing Environment (v3.6.3).

Results

General characteristics

Electronic medical records were screened to identify all neonates who received intravenous TZP therapy between January 2019 and December 2024. Application of the inclusion criteria (age ⩽ 28 days, receipt of intravenous TZP, and having serum potassium levels measured during treatment) yielded 1027 potentially eligible neonates. These neonates were then assessed against the exclusion criteria. A total of 669 neonates were excluded: 32 due to TZP treatment duration <3 days, 537 due to missing baseline, during treatment, or post-treatment serum potassium records, 84 due to pre-existing hypokalemia, and 16 due to dialysis or continuous kidney replacement therapy. Consequently, the final analytical cohort comprised 358 neonates (Figure 1).

Figure 1.

Flowchart of cohort identification and selection process.

The study population consisted predominantly of male infants (60.9%) with a median GA of 35.6 weeks (IQR 33.29–38.86). Neonatal assessment revealed Apgar scores of 9 (IQR 7–10) at 1 min, 9 (IQR 8–10) at 5 minutes, and 9 (IQR 9–10) at 10 min. Clinical parameters showed the baseline TZP therapy duration was 6 days (IQR 5–7) with corresponding serum potassium levels averaging 4.33 mmol/L (IQR 3.95–4.76). Lower respiratory tract infection constituted the primary therapeutic indication (81.5%), while adjuvant medications included levocarnitine (52.5%) and dopamine (50.8%). Comparative analyses between groups are detailed in Table 1.

Table 1.

Baseline demographic and clinical characteristics of neonates treated with TZP therapy (N = 358).

Variables	Total	Hypokalemia	Normokalemia	Statistics	p Value
Variables	(n = 358 )	(n = 73 )	(n = 285 )	Statistics	p Value
Sex (male/female), n	218/140	49/24	169/116	1.494	0.222
Gestational age (weeks)	35.6 (33.3–38.9)	34.6 (31.4–38.4)	36.0 (32.3–38.9)	−1.139	0.255
Postnatal age (days)	4 (1–11)	0 (0–1)	6 (1–12)	−6.521	<0.001
Postmenstrual age (weeks)	36.9 (33.3–39.9)	35.1 (31.7–39.0)	37.1 (33.6–39.9)	−2.412	0.016
Birth weight (kg)	2.4 (1.6–3.1)	2.3 (1.7–2.9)	2.4 (1.6–3.1)	−0.540	0.589
Current weight (kg)	2.4 (1.7–3.1)	2.3 (1.7–2.9)	2.4 (1.7–3.1)	−0.873	0.382
Duration of TZP therapy (days)	6 (5–7)	6 (5–8)	6 (4–7)	−1.848	0.065
Apgar score
1 min	9 (7–10)	8 (7–9)	9 (8–10)	−2.988	0.003
5 min	9 (8–10)	9 (8–9)	9 (8–10)	−1.958	0.050
10 min	9 (9–10)	9 (9–10)	9 (9–10)	−1.271	0.204
Serum potassium (mmol/L)	4.3 (4.0–4.8)	4.4 (4.0–5.0)	4.3 (3.9–4.7)	−1.401	0.161
Serum sodium (mmol/L)	139.8 (137.2–142.3)	138.8 (137.3–141.7)	140 (137.1–142.4)	−0.838	0.402
Serum calcium (mmol/L)	2.1 (1.9–2.3)	1.9 (1.7–2.1)	2.2 (2.0–2.4)	−5.484	<0.001
Serum magnesium (mmol/L)	0.8 (0.7–0.9)	0.8 (0.7–0.8)	0.8 (0.7–0.9)	−1.381	0.167
Total bilirubin (μmol/L)	87.1 ± 51.3	72.9 ± 46.4	90.8 ± 51.9	−2.677	0.008
Direct bilirubin (μmol/L)	6.9 (5.4–10.1)	6.5(5.3–8.2)	7.3 (5.4–10.6)	−2.135	0.033
Alkaline phosphatase (U/L)	170 (126–246)	161(124–209)	175 (126–258)	−1.816	0.069
Alanine transaminase (U/L)	9.0 (6.0–13.0)	7.5(5.3–12.0)	9.0 (6.3–13.0)	−1.957	0.050
Aspartate transaminase (U/L)	37.2 (23.9–58.0)	48.0(33.0–62.0)	34.0 (22.9–55.0)	−2.836	0.005
Total protein (g/L)	45.3 (40.9–50.1)	41.8(38.7–48.8)	46.1 (41.7–50.2)	−3.283	0.001
Plasma albumin (g/L)	27.7 (24.5–31.3)	25.7(22.2–29.4)	28.1 (25.2–32.1)	−3.816	<0.001
Blood urea nitrogen (mmol/L)	4.67 (3.04–8.12)	5.83(3.89–9.09)	4.50(2.86–7.51)	−2.858	0.004
Serum creatinine (μmol/L)	58.9 (38.3–76.7)	74.2(59.1–91.5)	55.2 (36.0–71.9)	−5.769	<0.001
Infectious source, n (%)
Respiratory	292 (81.5)	62 (84.9)	230 (80.7)	0.691	0.406
Central nervous	124 (34.6)	23 (6.4)	101 (28.2)	0.529	0.397
Digestive	57 (15.9)	13 (17.8)	44 (15.4)	0.244	0.622
Bacteremia	38 (10.6)	3 (4.1)	35 (12.3)	4.090	0.043
Coadministered drugs, n (%)
Caffeine	119 (33.2)	34 (46.6)	85 (29.8)	7.348	0.007
Dopamine	182 (50.8)	58 (79.5)	124 (43.5)	30.04	<0.001
Dobutamine	71 (19.8)	24 (32.9)	47 (16.5)	9.814	0.002
Fluconazole	128 (35.8)	40 (54.8)	88 (30.9)	14.472	<0.001
Calf pulmonary surfactant	37 (10.3)	16 (21.9)	21 (7.4)	13.275	<0.001
Insulin	25 (7.0)	11 (15.1)	14 (4.9)	9.228	0.002

Incidence of hypokalemia and Naranjo criteria

Among the 358 patients treated with piperacillin-tazobactam (TZP), 73 (20.4%) developed TAH, which was investigated as an ADR in this study. According to the Naranjo causality assessment scale, 57 TAH cases (78.1%) were categorized as “probable” ADRs (mean score 5.12), and 16 cases (21.9%) as “possible” ADRs (mean score 4.00). No cases met the criteria for “definite” or “doubtful” ratings. The overall mean Naranjo score for TAH events was 4.88. As all cases met the predefined criterion (Naranjo score ⩾1), they were included in subsequent risk-factor analyses.

In terms of severity, the distribution among TAH cases was as follows: Among the 73 neonates with TAH, 43 (58.9% of cases) were classified as mild, 23 (31.5% of cases) as moderate, and 7 (9.6% of cases) as severe. Following clinical interventions—including potassium supplementation or drug discontinuation—normal potassium levels were achieved in 69 patients (94.5%). Dynamic monitoring of serum potassium levels demonstrated the following trends: a pre-treatment median level of 4.41 mmol/L (IQR 3.99–5.03), a post-treatment nadir of 3.07 mmol/L (IQR 2.72–3.36), and a recovery level of 4.38 mmol/L (IQR 4.04–4.79; Figure 2).

Figure 2.

Change in serum potassium values.

Screening for predictive factors

Univariate logistic regression analysis (p < 0.01) was used to screen candidate variables for inclusion in the multivariate model. Significant predictors included: serum calcium (p < 0.001), total bilirubin (p = 0.008), total protein (p < 0.001), plasma albumin (P < 0.001), blood urea nitrogen (p = 0.004), serum creatinine (p < 0.001), co-administered dopamine (p < 0.001), dobutamine (p = 0.002), insulin (p = 0.004), fluconazole (p < 0.001), and caffeine citrate (p = 0.007) (Table 2).

Table 2.

The univariate analyses of the potential predictive factors associated with hypokalemia.

Variables	p Value	OR	OR 95% CI
Sex (male/female), n	0.223	1.401	0.815	2.411
Gestational age (weeks)	0.311	0.970	0.916	1.028
Postnatal age (days)	0.002	0.663	0.498	0.857
Postmenstrual age (weeks)	0.024	0.934	0.880	0.991
Birth wt (g)	0.639	0.932	0.695	1.250
Wt (g)	0.428	0.886	0.657	1.195
Duration of TZP therapy (days)	0.304	1.045	0.961	1.135
Apgar score
1 min	0.101	0.920	0.832	1.016
5 min	0.470	0.952	0.832	1.088
10 min	0.885	0.989	0.847	1.154
Serum potassium (mmol/L)	0.098	1.421	0.938	2.152
Serum sodium (mmol/L)	0.729	1.007	0.969	1.046
Serum calcium (mmol/L)	<0.001	0.089	0.037	0.213
Serum magnesium (mmol/L)	0.412	0.485	0.087	2.724
Total bilirubin (μmol/L)	0.008	0.993	0.987	0.998
Direct bilirubin (μmol/L)	0.216	0.980	0.949	1.012
Alkaline phosphatase (U/L)	0.554	1.000	0.998	1.001
Alanine transaminase (U/L)	0.665	0.998	0.992	1.005
Aspartate transaminase (U/L)	0.882	1.000	0.999	1.001
Total protein (g/L)	<0.001	0.959	0.939	0.98
Plasma albumin (g/L)	<0.001	0.936	0.906	0.967
Blood urea nitrogen (mmol/L)	0.004	1.088	1.027	1.152
Serum creatinine (μmol/L)	<0.001	1.027	1.017	1.037
Infectious source, n (%)
Respiratory	0.407	1.348	0.666	2.729
Central nervous	0.529	0.838	0.483	1.453
Digestive	0.622	1.187	0.601	2.343
Bacteremia	0.055	0.306	0.091	1.025
Coadministered drugs, n (%)
Dopamine	<0.001	5.020	2.717	9.278
Fluconazole	<0.001	2.713	1.605	4.587
Calf pulmonary surfactant	0.001	3.529	1.734	7.182
Dobutamine	0.002	2.480	1.389	4.430
Insulin	0.004	3.434	1.488	7.927
Caffeine citrate injection	0.007	2.051	1.213	3.468

Notably, furosemide—a known hypokalemia-inducing diuretic—was used in 37.2% of patients but did not meet our pre-specified threshold of p < 0.01 (univariate p = 0.034). Other medications with potential hypokalemic effects (e.g., hydrocortisone, insulin) were included in the initial screening, but only those meeting the significance threshold proceeded to multivariate analysis.

Multivariate logistic regression revealed three independent predictors of hypokalemia (Figure 3): lower serum calcium (OR 0.31, 95% CI 0.10–0.99; p = 0.049), higher baseline serum creatinine (OR 1.02, 95% CI 1.00–1.03; p = 0.012), and dopamine use (OR 3.37, 95% CI 1.54–7.36; p = 0.002). Variables with p < 0.01 in univariate analysis were included in the multivariate model to enhance model stability.

Figure 3.

Forest plot summarizing the risk factors for TAH in neonates.

ROC-derived thresholds and binary predictors

ROC analysis identified optimal cut-off values for predicting TAH: serum calcium ⩽2.06 mmol/L (AUC = 0.708, sensitivity 67.1%, specificity 67.0%), and serum creatinine > 62.3 μmol/L (AUC = 0.719, sensitivity 72.6%, specificity 61.8%) (Supplemental Table 3 and Supplemental Figures 1 and 2). These were used to define binary predictors: hypocalcemia and elevated creatinine.

In the multivariable logistic regression model incorporating hypocalcemia, elevated creatinine, and dopamine use, hypocalcemia was associated with a 2.33-fold increased risk of TAH (OR 2.33, 95% CI 1.26–4.30; p = 0.007), elevated creatinine with a 2.38-fold increased risk (OR 2.38, 95% CI 1.26–4.48; p = 0.008), and dopamine use with a 3.61-fold increased risk (OR 3.61, 95% CI 1.91–6.84; p < 0.001; Supplemental Table 4).

Construction of the predictive model and nomogram

A logistic regression model was employed to construct a nomogram based on the three aforementioned factors, namely serum calcium, baseline serum creatinine, and dopamine use. Higher total points on the nomogram indicate a greater risk of neonatal hypokalemia (Figure 4(a)). For instance, if a neonate uses dopamine and has serum calcium of 2.8 mmol/L and baseline serum creatinine of 180 μmol/L, the total score is approximately 135, indicating an estimated hypokalemia risk of 50% for this case.

Figure 4.

(a) Nomogram for the prediction of neonatal TAH. (b) ROC curve for the neonatal TAH prediction model. (c) Calibration curve for predicting the probability of neonatal TAH. (d) Decision curve analysis in the prediction of neonatal TAH.

Predictive accuracy and net benefit

In the training cohort, the predictive model exhibited excellent discriminative ability as evidenced by an AUC of 0.771 (95% CI 0.724–0.814; Figure 4(b)). This was complemented by calibration analysis which showed that the predicted probabilities were closely aligned with observed outcomes along the ideal diagonal (Figure 4(c)). More importantly, DCA confirmed the clinical utility of the model, demonstrating substantially greater net benefit across a wide threshold probability range compared to default strategies (Figure 4(d)).

Comparison of hypokalemia incidence

The optimal cutoff value for determining TAH incidence was identified as 84 points through Youden’s index analysis. This threshold demonstrated 80.8% sensitivity and 66.3% specificity, with corresponding positive and negative predictive values of 38.1% and 93.1%, respectively. Based on this calculated cutoff, we established distinct risk stratification cohorts: high-risk (⩾84 points) and low-risk (<84 points) groups. Subsequent Kaplan–Meier analysis revealed significantly different clinical trajectories between groups. The high-risk cohort exhibited substantially reduced hypokalemia-free rates compared to the low-risk group (p < 0.001), with an adjusted hazard ratio of 4.78 (95% CI 3.72–7.58; Figure 5).

Figure 5.

Kaplan–Meier curves for risk groups.

Discussion

As documented in previous literature, drug-induced hypokalemia manifests commonly among hospitalized patients.^20,21 Approximately 20% of hospitalized patients are affected by hypokalemia, with 40% of cases originating pharmacologically.¹⁷ Controlling disease progression necessitates assessing hypokalemia risk in patients, enhancing potassium monitoring during drug administration, and appropriately managing symptoms. Hypokalemia can be induced by penicillin antibiotics through their “non-reabsorbable anion effect.”³ Piperacillin, a potent organic acid, dissociates into non-reabsorbable anions within the distal renal tubules, generating a negative luminal voltage gradient favoring potassium secretion, thus increasing potassium excretion and ultimately leading to hypokalemia.^22,23 In addition, the high sodium content in intravenous piperacillin infusion can expand extracellular fluid volume, promoting greater transportation of sodium ions to distal renal units, further stimulating potassium ion excretion.²⁴

Previous research has primarily focused on adults in exploring the effects of risk factors on TAH, neglecting younger age groups. This study aimed to develop an easy-to-understand visual model for assessing TAH risk factors in neonates, constituting a novel and significant contribution. Findings revealed a 20.4% incidence of TAH in neonates, falling between rates reported by Kuramoto et al. (24.8%)⁷ and Seo (13.9%).¹⁸ More notably, the incidence of moderate-to-severe hypokalemia among all TZP-treated neonates was 8.4% (30/358) in our study, which is substantially higher than the corresponding figures reported by Kuramoto et al (6.4%),⁷ Seo et al (4.5%),¹⁸ and Kaye et al (1.3%).²⁵ This result requires clinical attention due to its severity.

Our analysis revealed that higher baseline serum creatinine and dopamine use were independent risk factors associated with TAH. Lower serum calcium levels, which are influenced by maternal factors, tend to be higher at birth and gradually decline over time. Due to immaturity of the nephron and normal developmental changes, premature infants often have higher baseline creatinine levels at birth compared to full-term infants, reflecting their renal immaturity.²⁶ This study involved a predominantly preterm population, with a median GA of 35.6 weeks (IQR, 33.29–38.86 weeks). Elevated creatinine values suggest immature renal function, which can make neonates more susceptible to electrolyte disturbances that may lead to hypokalemia in the diseased state. In addition, it affects the renal excretion of penicillin antibiotics and enhances the “non-reabsorbable anionic effect” of penicillin antibiotics. The ROC-derived thresholds identified in this study (creatinine > 62.3 μmol/L, calcium ⩽2.06 mmol/L) provide practical tools for rapid risk stratification. Neonates meeting these criteria, especially those concurrently receiving dopamine, should be prioritized for frequent potassium monitoring during TZP therapy. These thresholds can be integrated into clinical decision support systems to facilitate early intervention.

Comorbid medications can increase the risk of hypokalemia. Univariate analysis showed that diuretics, vasoactive analogues, glucocorticoid analogues, insulin, and numerous other drugs were associated with increased hypokalemia risk.²⁷ However, multivariate logistic regression analyses identified only dopamine use as an independent risk factor for TAH. Dopamine is a catecholamine that can cause hypokalemia at moderate doses (3–5 ug/kg.min). This is due to its agonism of α and β-adrenergic receptors, stimulation of insulin secretion, activation of Na⁺-K⁺-ATPase, and transfer of extracellular potassium into the cell.²⁸ In addition, dopamine’s impact on renal blood flow and glomerular filtration rate can also contribute to electrolyte imbalances, including potassium disturbances.²⁹ Therefore, although dopamine is an essential medication in neonatal care, it is crucial to monitor electrolyte levels.

It is noteworthy that furosemide, a well-known diuretic associated with hypokalemia, showed an association in univariate analysis (p = 0.034) but was not included in the final multivariate model, as it did not reach our pre-specified threshold of p < 0.01. Therefore, although furosemide use was prevalent (37.2%) and associated with hypokalemia at a univariate level, it did not meet our pre-specified statistical threshold for further multivariable adjustment. This methodological choice underscores that our model identifies independent predictors with the strongest statistical evidence, rather than all potential clinical associations.

Median neonatal calcium was significantly lower in the hypokalemia group (1.90 mmol/L) than in the normokalemia group (2.19 mmol/L). Multivariate logistic regression analyses highlighted serum calcium (p = 0.049, OR 0.31, 95% CI 0.10–0.99) as a protective factor against TAH occurrence, implying hypocalcemia as a potential predictor of hypokalemia in neonatal electrolyte disorders. Calcium is known to mitigate hyperkalemia-induced cardiac effects,³⁰ but its relationship with hypocalcemia during hypokalemia warrants study, especially in neonates with immature electrolyte regulation.

Several studies have reported a higher prevalence of hypokalemia in women compared to men, potentially attributed to differences in body weight composition.^5,17,31,32 In addition, low BMI has been identified as a hypokalemia risk factor due to its correlation with potassium levels and muscle mass.^7,31,33 Although no gender differences were detected in our study, neonates exhibited lower muscle mass percentage of body weight compared to adults, potentially explaining the higher incidence of severe hypokalemia observed in neonates.

One study reported that extremely preterm infants born small for gestational age, who were managed with currently recommended early parenteral nutrition, had a high risk of early hypokalemia and hypophosphatemia.³⁴ Other studies have suggested a common association between magnesium deficiency and hypokalemia and proposed that concomitant magnesium deficiency can exacerbate hypokalemia, thereby complicating potassium therapy administration.³⁵ In our univariate analysis, we observed lower PMA and serum magnesium levels in the hypokalemia group compared to the normokalemia group. This suggests that age and serum magnesium may serve as risk factors for neonatal TAH. However, our multivariate logistic regression did not reveal any association with hypokalemia. This lack of association may be attributed to various factors, such as the small sample size, the complexity of neonatal physiology, or the multifaceted nature of hypokalemia etiology.

Limitations

The International Neonatal Consortium has noted the absence of established actionable reference values for routine laboratory tests in neonates, in contrast to initiatives like the CALIPER project that provide pediatric reference intervals. Therefore, we used adult threshold values for certain diagnostic criteria.³⁶ Although this approach is practical in the absence of neonatal-specific reference values, it may not accurately reflect neonatal physiology. This limitation highlights the need for future research to establish neonatal-specific reference values for improved diagnostic accuracy and clinical care.

Additional limitations of our study include its retrospective, single-center design with a small sample size, which may limit generalizability. The predictive model also lacks external validation, necessitating further validation with larger, more diverse samples. Furthermore, while the provided binary thresholds enhance clinical interpretability, they require external validation and may oversimplify continuous physiological relationships. Although we systematically recorded concomitant medications, the relatively low usage of certain drugs (e.g., Omeprazole) may have limited our ability to detect their independent effects. Furthermore, unaccounted factors such as parenteral nutrition, underlying diseases, and infection severity might introduce bias. Thus, prospective multicenter studies are needed to ensure unbiased and comprehensive results. It is also worth noting that other structured approaches for ADR prediction, including machine learning–based or multidimensional scoring tools, have been proposed in pharmacovigilance research and could allow finer-grained analysis in future prospective designs.³⁷

Conclusion

In summary, this study identified higher baseline serum creatinine, dopamine use, and lower serum calcium as predictors of neonatal TAH, presenting the first user-friendly and relatively personalized model for predicting neonatal TAH. Visualization of predictors and personalized models can offer clinicians a simple and intuitive tool for early neonatal TAH detection and management, thereby enhancing patient safety. Regular updates to the model in alignment with emerging clinical trial data can reinforce its utility and relevance.

Supplemental Material

sj-docx-3-taw-10.1177_20420986261434697 – Supplemental material for Incidence, risk factors, and prediction model of piperacillin/tazobactam-associated hypokalemia in neonates

Supplemental material, sj-docx-3-taw-10.1177_20420986261434697 for Incidence, risk factors, and prediction model of piperacillin/tazobactam-associated hypokalemia in neonates by Jun Wan, Shiteng Lin, Wei Zhuang, Weida Chen, Yanting Huang, Zhiyi Huang, Wanlong Lin and Yao Chen in Therapeutic Advances in Drug Safety

Supplemental Material

sj-docx-4-taw-10.1177_20420986261434697 – Supplemental material for Incidence, risk factors, and prediction model of piperacillin/tazobactam-associated hypokalemia in neonates

Supplemental material, sj-docx-4-taw-10.1177_20420986261434697 for Incidence, risk factors, and prediction model of piperacillin/tazobactam-associated hypokalemia in neonates by Jun Wan, Shiteng Lin, Wei Zhuang, Weida Chen, Yanting Huang, Zhiyi Huang, Wanlong Lin and Yao Chen in Therapeutic Advances in Drug Safety

Supplemental Material

sj-docx-5-taw-10.1177_20420986261434697 – Supplemental material for Incidence, risk factors, and prediction model of piperacillin/tazobactam-associated hypokalemia in neonates

Supplemental material, sj-docx-5-taw-10.1177_20420986261434697 for Incidence, risk factors, and prediction model of piperacillin/tazobactam-associated hypokalemia in neonates by Jun Wan, Shiteng Lin, Wei Zhuang, Weida Chen, Yanting Huang, Zhiyi Huang, Wanlong Lin and Yao Chen in Therapeutic Advances in Drug Safety

Supplemental Material

sj-docx-6-taw-10.1177_20420986261434697 – Supplemental material for Incidence, risk factors, and prediction model of piperacillin/tazobactam-associated hypokalemia in neonates

Supplemental material, sj-docx-6-taw-10.1177_20420986261434697 for Incidence, risk factors, and prediction model of piperacillin/tazobactam-associated hypokalemia in neonates by Jun Wan, Shiteng Lin, Wei Zhuang, Weida Chen, Yanting Huang, Zhiyi Huang, Wanlong Lin and Yao Chen in Therapeutic Advances in Drug Safety

Supplemental Material

sj-docx-7-taw-10.1177_20420986261434697 – Supplemental material for Incidence, risk factors, and prediction model of piperacillin/tazobactam-associated hypokalemia in neonates

Supplemental material, sj-docx-7-taw-10.1177_20420986261434697 for Incidence, risk factors, and prediction model of piperacillin/tazobactam-associated hypokalemia in neonates by Jun Wan, Shiteng Lin, Wei Zhuang, Weida Chen, Yanting Huang, Zhiyi Huang, Wanlong Lin and Yao Chen in Therapeutic Advances in Drug Safety

Supplemental Material

sj-jpg-1-taw-10.1177_20420986261434697 – Supplemental material for Incidence, risk factors, and prediction model of piperacillin/tazobactam-associated hypokalemia in neonates

Supplemental material, sj-jpg-1-taw-10.1177_20420986261434697 for Incidence, risk factors, and prediction model of piperacillin/tazobactam-associated hypokalemia in neonates by Jun Wan, Shiteng Lin, Wei Zhuang, Weida Chen, Yanting Huang, Zhiyi Huang, Wanlong Lin and Yao Chen in Therapeutic Advances in Drug Safety

Supplemental Material

sj-jpg-2-taw-10.1177_20420986261434697 – Supplemental material for Incidence, risk factors, and prediction model of piperacillin/tazobactam-associated hypokalemia in neonates

Supplemental material, sj-jpg-2-taw-10.1177_20420986261434697 for Incidence, risk factors, and prediction model of piperacillin/tazobactam-associated hypokalemia in neonates by Jun Wan, Shiteng Lin, Wei Zhuang, Weida Chen, Yanting Huang, Zhiyi Huang, Wanlong Lin and Yao Chen in Therapeutic Advances in Drug Safety

Footnotes

Acknowledgements

None.

Declarations

ORCID iDs

Jun Wan

Wei Zhuang

Supplemental material

Supplemental material for this article is available online.

References

Lorenz

JM.

Assessing fluid and electrolyte status in the newborn. National Academy of Clinical Biochemistry. Clin Chem 1997; 43(1): 205–210.

Bonilla-Félix

Potassium regulation in the neonate. Pediatr Nephrol 2017; 32(11): 2037–2049.

Gennari

FJ.

Hypokalemia. N Engl J Med 1998; 339(7): 451–458.

Kardalas

Paschou

Anagnostis

, et al. Hypokalemia: a clinical update. Endocr Connect 2018; 7(4): R135–R146.

Paice

Paterson

Onyanga-Omara

, et al. Record linkage study of hypokalaemia in hospitalized patients. Postgrad Med J 1986; 62(725): 187–191.

Sørensen

Matzen

LE.

[Serum electrolytes and drug therapy of patients admitted to a geriatric department]. Ugeskr Laeger 1993; 155(48): 3921–3924.

Kuramoto

Masago

Kashiwagi

, et al. [Incidence and risk factors of hypokalemia in tazobactam/piperacillin-administered patients]. Yakugaku Zasshi 2019; 139(12): 1591–1600.

Kokot

Hyla-Klekot

Drug-induced abnormalities of potassium metabolism. Pol Arch Med Wewn 2008; 118(7–8): 431–434.

Ben Salem

Hmouda

Bouraoui

. Drug-induced hypokalaemia. Curr Drug Saf 2009; 4(1): 55–61.

10.

Lin

Wong

LYF

Cheung

BMY

. Diuretic-induced hypokalaemia: an updated review. Postgrad Med J 2022; 98(1160): 477–482.

11.

Tsai

Hsu

, et al. Life-threatening hypokalemia in an asthmatic patient treated with high-dose hydrocortisone. Am J Med Sci 2004; 327(3): 152–155.

12.

Perry

Markham

Piperacillin/tazobactam: an updated review of its use in the treatment of bacterial infections. Drugs 1999; 57(5): 805–843.

13.

Drug Approval Package: Zosyn (Piperacillin Tazobactam) NDA #050684s045 050750s012. https://www.accessdata.fda.gov/drugsatfda_docs/nda/2005/050684_S045_050750_s012_ZosynTOC.cfm (accessed 17 February 2024).

14.

Korang

Safi

Nava

, et al. Antibiotic regimens for early-onset neonatal sepsis. Cochrane Database Syst Rev 2021; 5(5): CD013837.

15.

Wolf

Simon

The use of piperacillin-tazobactam in neonatal and paediatric patients. Expert Opin Drug Metab Toxicol 2009; 5(1): 57–69.

16.

Pharmaceuticals and Medical Devices Agency (PMDA). Revisions of PRECAUTIONS (FY2020), https://www.pmda.go.jp/english/safety/info-services/drugs/revision-of-precautions/0008.html (accessed 17 February 2024).

17.

van der Heijden

CDCC

Duizer

Fleuren

HWHA

, et al. Intravenous flucloxacillin treatment is associated with a high incidence of hypokalaemia. Br J Clin Pharmacol 2019; 85(12): 2886–2890.

18.

Seo

Kim

Incidence and determinants of piperacillin/tazobactam-associated hypokalemia: a retrospective study. Antibiotics (Basel) 2022; 11(8): 1138.

19.

von Elm

Altman

Egger

, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet 2007; 370(9596): 1453–1457.

20.

Seo

Kim

Electrolyte disorders associated with piperacillin/tazobactam: a pharmacovigilance study using the FAERS database. Antibiotics (Basel) 2023; 12(2): 240.

21.

Morton

Hypokalaemia in pregnancy – Prevalence, underlying causes, and an approach to investigation. Obstet Med 2024; 17(4): 213–220.

22.

Zaki

Lad

Piperacillin-tazobactam-induced hypokalemia and metabolic alkalosis. Indian J Pharmacol 2011; 43(5): 609–610.

23.

Pandya

Gupta

Malhotra

, et al. Piperacillin-tazobactam induced hypokalaemia. Int J Basic Clin Pharmacol 2018; 7(12): 2459.

24.

Brunner

Frick

. Hypokalaemia, metabolic alkalosis, and hypernatraemiadue to “Massive”sodium penicillin Therapy. Br Med J 1968; 4(5630): 550–552.

25.

Kaye

Rice

Dane

, et al. Fosfomycin for injection (ZTI-01) versus piperacillin-tazobactam for the treatment of complicated urinary tract infection including acute pyelonephritis: ZEUS, a phase 2/3 randomized trial. Clin Infect Dis 2019; 69(12): 2045–2056.

26.

Drukker

Guignard

JP.

Renal aspects of the term and preterm infant: a selective update. Curr Opin Pediatr 2002; 14(2): 175–182.

27.

Salem

Hmouda

Bouraoui

Drug-induced hypokalaemia. Curr Drug Saf 2009; 4(1): 55–61.

28.

Clausen

Flatman

JA.

The effect of catecholamines on Na-K transport and membrane potential in rat soleus muscle. J Physiol 1977; 270(2): 383–414, https://pubmed.ncbi.nlm.nih.gov/198530/ (accessed 4 March 2024).

29.

Somlo

Doroshow

Lev-Ran

, et al. Effect of low-dose prophylactic dopamine on high-dose cisplatin-induced electrolyte wasting, ototoxicity, and epidermal growth factor excretion: a randomized, placebo-controlled, double-blind trial. https://pubmed.ncbi.nlm.nih.gov/7738626/ (accessed 4 March 2024).

30.

Celebi Yamanoglu

Yamanoglu

. The effect of calcium gluconate in the treatment of hyperkalemia. Turk J Emerg Med 2022; 22(2): 75–82.

31.

Kleinfeld

Borra

Gavani

, et al. Hypokalemia: are elderly females more vulnerable? J Natl Med Assoc 1993; 85(11): 861–864.

32.

Fukui

Otani

Katoh

, et al. Female gender as a risk factor for hypokalemia and QT prolongation after subarachnoid hemorrhage. Neurology 2002; 59(1): 134–136.

33.

Rundo

Sagild

Total and exchangeable potassium in humans. Nature 1955; 175(4461): 774.

34.

Boubred

Herlenius

Bartocci

, et al. Extremely preterm infants who are small for gestational age have a high risk of early hypophosphatemia and hypokalemia. Acta Paediatr 2015; 104(11): 1077–1083.

35.

Huang

Kuo

Mechanism of hypokalemia in magnesium deficiency. J Am Soc Nephrol 2007; 18(10): 2649–2652.

36.

Allegaert

Hildebrand

Singh

, et al. The publication quality of laboratory values in clinical studies in neonates. Pediatr Res 2023; 94(1): 96–98.

37.

Yadesa

Kitutu

Tamukong

, et al. Development and validation of “prediction of adverse drug reactions in older inpatients (PADROI)” risk assessment tool. Clin Interv Aging 2022; 17: 195–210.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.02 MB

0.01 MB

0.02 MB

0.43 MB

0.44 MB