Abstract
Purpose
While the relationship between serum phosphorus concentrations and the prognosis of chronic obstructive pulmonary disease (COPD) has been previously investigated, the prognostic significance of dynamic serum phosphorus fluctuations in patients with COPD admitted to intensive care units (ICUs) remains poorly understood. This study aimed to evaluate the association between longitudinal serum phosphorus trajectories during the early ICU stay and 28-day all-cause mortality, and to explore their potential utility as a prognostic indicator.
Methods
This retrospective cohort analysis utilized data from COPD patients registered in the Medical Information Mart for Intensive Care IV database (MIMIC-IV, v3.0). Group-based trajectory modeling (GBTM) was applied to delineate distinct serum phosphorus trajectories over the first seven days following ICU admission. The relationship between these trajectory groups and all-cause mortality was examined using Kaplan–Meier survival curves and multivariable Cox proportional hazards regression. Subgroup analyses were performed to verify the robustness of this association across different patient subgroups.
Results
A total of 3627 ICU-admitted patients with COPD were enrolled. GBTM distinguished two distinct longitudinal patterns: a high-phosphate trajectory (mean phosphate range: approximately 5.0–6.5 mg/dL; n = 320, 8.8%) and a low/normal-phosphate trajectory (mean phosphate range: approximately 3.2–3.8 mg/dL; n = 3307, 91.2%). Patients in the high-phosphate group exhibited markedly lower 28-day cumulative survival. Multivariable Cox regression confirmed that this trajectory was independently associated with elevated 28-day mortality risk. Furthermore, the trajectory-based model demonstrated superior predictive performance compared with single-time-point phosphate measurements.
Conclusion
Longitudinal trends in serum phosphate levels serve as a significant and independent risk factor for short-term mortality in critically ill patients with COPD.
Keywords: chronic obstructive pulmonary disease, serum phosphate, trajectory analysis, group-based trajectory modeling, intensive care unit, mortality
Introduction
Chronic obstructive pulmonary disease (COPD) is a prevalent respiratory disorder arising from the interplay between genetic predisposition and environmental exposures. Its central pathological hallmark is irreversible structural damage to the airways and/or alveolar spaces, resulting in persistent respiratory symptoms and progressive airflow limitation.1,2 Currently, the global prevalence of COPD among adults is estimated at approximately 10.3%, and the disease now constitutes the third leading cause of death worldwide.3 The acute exacerbation phase is characterized by a rapid and sustained worsening of respiratory symptoms that substantially exceeds day-to-day variability; this deterioration not only accelerates the decline in pulmonary function but is also strongly linked to an increased requirement for mechanical ventilation, prolonged hospitalization, and elevated mortality.4 Despite continuous advances in clinical care, more than half of all COPD cases globally remain undiagnosed. Concurrently, rates of intensive care unit (ICU) admission and mortality among affected individuals persist at high levels; compounded by progressive population aging, this imposes a substantial economic and societal burden on healthcare systems worldwide.5 These challenges underscore the pressing need to identify reliable prognostic biomarkers capable of facilitating early risk stratification and informing the development of individualized therapeutic strategies.
Serum phosphate participates in numerous essential physiological processes, including membrane transport, energy metabolism, bone mineralization, and muscle contraction.6–8 Phosphorus plays a critical role in diaphragmatic energy metabolism, as adequate intracellular phosphate is essential for maintaining adenosine triphosphate (ATP) production and optimal muscle contractility. Both hypophosphatemia and hyperphosphatemia have been implicated in respiratory muscle dysfunction. Hypophosphatemia reduces the ATP/adenosine diphosphate (ADP) ratio, thereby impairing diaphragmatic contractility and increasing the risk of weaning failure from mechanical ventilation.9,10 Conversely, hyperphosphatemia may compromise respiratory muscle efficiency through intracellular calcium-phosphate precipitation, which interferes with mitochondrial function.11 In addition, mechanical ventilation itself can influence serum phosphate homeostasis via ventilator-induced alterations in acid-base balance and systemic inflammation. Patients with COPD are particularly vulnerable to these phosphate-related respiratory disturbances, given the high prevalence of pre-existing diaphragmatic weakness in this population. Disturbances in serum phosphate homeostasis are frequently encountered in critically ill populations; the reported prevalence of hypophosphatemia and hyperphosphatemia among ICU patients is approximately 20% and 45%, respectively.12,13 A growing body of evidence has documented an association between abnormal serum phosphate concentrations and mortality risk in patients with COPD. Specifically, elevated serum phosphate levels have been positively correlated with increased in-hospital mortality among patients with acute exacerbation of COPD (AECOPD), and hyperphosphatemia has been proposed as a potential risk factor for adverse outcomes in this population.14 Conversely, hypophosphatemia has likewise been identified as a prognostic risk factor in AECOPD patients.15 Compared with single-time-point or variability-based measures, trajectory analysis more accurately captures longitudinal temporal patterns of serum phosphate fluctuation. To date, however, existing investigations have predominantly focused on general critically ill or septic cohorts, and the prognostic implications of dynamic serum phosphate changes specifically in COPD patients remain inadequately defined. Accordingly, this study was designed to evaluate the impact of serum phosphate trajectories on mortality risk among COPD patients admitted to the ICU, with the ultimate goal of guiding phosphate management and improving clinical outcomes.
Materials and Methods
Data Source and Study Population
Patients diagnosed with COPD were identified and enrolled from the Medical Information Mart for Intensive Care (MIMIC-IV, version 3.0) database. This publicly accessible critical care repository encompasses comprehensive patient-level data, including demographics, laboratory findings, vital signs, comorbidities, and therapeutic interventions (https://mimic.mit.edu/). One of the authors, Pei Zhou, was granted access to the MIMIC-IV database (certificate number: 13521 212). The study protocol was approved by the Institutional Review Board of Beth Israel Deaconess Medical Center, which waived the requirement for informed consent; accordingly, written informed consent was not mandated.
The eligible cohort consisted of patients admitted to the intensive care unit for the first time who carried a diagnosis of COPD, defined according to the International Classification of Diseases, 10th Revision (ICD-10). Individuals meeting any of the following exclusion criteria were removed from the analysis: (i) age under 18 years; (ii) ICU length of stay less than 24 hours; (iii) fewer than three serum phosphate measurements obtained within the first 7 days following ICU admission, or fewer than two measurements recorded within the initial 72 hours; (iv) a documented diagnosis of end-stage renal disease (ESRD); (v) coexisting severe hepatic disease; and (vi) patients with multiple hospital admissions, for whom only the first hospitalization and the corresponding first ICU record were retained. The detailed screening and selection process is depicted in Figure 1.
Figure 1.
Flow chart of patient selection.
Variable Extraction
This study included the following variables: (1) Demographic characteristics: age, sex, ethnicity and weight; (2) Disease severity on ICU admission: sequential organ failure assessment (SOFA) score, glasgow coma scale (GCS) score; (3) Comorbidities: myocardial infarction, congestive heart failure, diabetes, chronic kidney disease (CKD), cerebrovascular disease, paraplegia, malignant tumors, acute kidney injury (AKI) and Charlson Comorbidity Index (CCI) score; (4) Laboratory parameters: white blood cell count(WBC), platelet count, hemoglobin, creatinine, blood urea nitrogen (BUN), glucose, calcium, sodium and potassium; (5) Vital signs: heart rate, respiratory rate, mean arterial pressure, and oxygen saturation (SpO2); (6) Treatment: mechanical ventilation during hospitalization (yes/no), use of vasopressors (yes/no), and use of glucocorticoids (yes/no); (7) For variables measured repeatedly, the first value recorded on the ICU admission day was used; (8) Primary endpoints: 28-day mortality after ICU admission.
Statistical Analysis
Group-based trajectory modeling (GBTM) was applied to delineate distinct longitudinal patterns of serum phosphate concentrations during the initial 7 days following ICU admission. Model estimation was performed using the hlme() function from the lcmm R package, specifying a quadratic temporal function and omitting random effects. The optimal two-trajectory solution was determined by evaluating models with two to four classes, guided by the Bayesian Information Criterion (BIC), mean posterior probability (≥0.70), and minimum class size (≥5%). This process yielded a high-phosphate trajectory (8.8%) and a low/normal-phosphate trajectory (91.2%). Variables with >20% missing values were excluded from the analysis. For the remaining covariates, missing values were handled via multiple imputation using chained equations (MICE, m = 5, maxit = 50), with the trajectory group variable excluded from the imputation procedures. Between-group comparisons were conducted using analysis of variance (ANOVA) or Kruskal–Wallis tests for continuous variables, and chi-squared (χ2) or Fisher’s exact tests for categorical variables, as appropriate. Survival outcomes were examined using Kaplan–Meier curves with Log rank testing, and the independent association between phosphate trajectory group and 28-day all-cause mortality was evaluated through multivariable Cox proportional hazards regression with a stepwise, progressively adjusted approach. Subgroup analyses were performed across strata defined by sex, age, SOFA score, mechanical ventilation, glucocorticoid therapy, AKI status and CKD status, potential interaction effects were formally assessed using likelihood ratio tests. To compare the prognostic utility of trajectory-based classification versus single time-point phosphate measurements, six nested models were constructed and evaluated based on the concordance index (C-index) and likelihood ratio test. All statistical analyses were executed in R (version 4.6.0), and a two-sided P value < 0.05 was considered statistically significant.
Results
Characteristics of Serum Phosphate Trajectories
Application of GBTM identified two distinct longitudinal patterns of serum phosphate over the initial 7 days of ICU stay (Figure 2). The predominant trajectory, designated Group 2 and comprising 91.2% of the cohort (n = 3307), was characterized by consistently low to normal phosphate concentrations, with mean values ranging from approximately 3.2 to 3.8 mg/dL throughout the observation period. In contrast, a smaller subset of patients, Group 1 (8.8%, n = 320), exhibited a persistently elevated phosphate trajectory, with mean levels between approximately 5.0 and 6.5 mg/dL. Although serum phosphate in this group displayed a gradual downward trend over the course of the week, concentrations remained above the upper limit of the normal reference range (4.5 mg/dL) for the majority of the first 7 days.
Figure 2.
Phosphate trajectory groups in COPD patients during ICU stay.
Baseline Characteristics
A total of 3627 patients with COPD were included, of whom 3307 (91.2%) were classified as the low-phosphate/normal trajectory group (Group 2) and 320 (8.8%) as the high-phosphate trajectory group (Group 1). A comparison of baseline characteristics between the two groups is shown in Table 1. There were no significant differences in age or gender distribution between the high-phosphate trajectory group and the low-phosphate/normal trajectory group; however, the clinical presentation in the high-phosphate trajectory group was more severe: mean arterial pressure was lower, respiratory rate was higher, and haemoglobin, haematocrit, and serum calcium levels were reduced, whilst urea, creatinine, and serum potassium levels were significantly elevated. In terms of disease severity, the high-phosphate group had significantly higher SOFA scores and a significantly heavier burden of complications, with higher prevalence rates of myocardial infarction, congestive heart failure, diabetes, and kidney disease. In terms of treatment, the high-phosphate group had higher rates of vasopressor and glucocorticoid use, as well as longer ICU stays. Most importantly, the 28-day mortality rate was significantly higher in the high-phosphate group.
Table 1.
Baseline Characteristics of ICU-Admitted COPD Patients by Phosphate Trajectory Groups
| Variable | Group 1 (n=320) | Group 2 (n=3307) | P-value |
|---|---|---|---|
| Demographics | |||
| Age, years | 73.00 [65.51, 80.55] | 72.67 [65.34, 80.20] | 0.841 |
| Gender | 0.639 | ||
| Female, % | 158 (49.4) | 1584 (47.9) | |
| Male, % | 162 (50.6) | 1723 (52.1) | |
| Weight | 84.10 [68.88, 100.00] | 78.20 [63.70, 96.00] | <0.001 |
| Ethnicity | 0.246 | ||
| Asian, % | 1 (0.3) | 45 (1.4) | |
| Black, % | 24 (7.5) | 279 (8.4) | |
| Hispanic, % | 9 (2.8) | 53 (1.6) | |
| White, % | 230 (71.9) | 2380 (72.0) | |
| Other, % | 56 (17.5) | 550 (16.6) | |
| Vital signs | |||
| MBP, mmHg | 73.64 [68.03, 79.57] | 77.77 [71.91, 85.26] | <0.001 |
| HR, beats/min | 84.64 [71.96, 97.40] | 83.00 [73.26, 95.14] | 0.324 |
| RR, beats/min | 20.19 [17.59, 23.33] | 19.50 [17.27, 22.06] | 0.004 |
| SpO2, % | 95.96 [94.51, 97.45] | 96.17 [94.52, 97.54] | 0.21 |
| Laboratory parameters | |||
| WBC, 10^9/L | 11.40 [8.10, 15.20] | 10.50 [7.70, 14.50] | 0.024 |
| Platelet, 10^9/L | 201.00 [142.00, 274.00] | 209.00 [158.00, 272.00] | 0.072 |
| Hemoglobin, g/dL | 9.50 [8.15, 11.12] | 10.90 [9.10, 12.40] | <0.001 |
| Hematocrit, % | 30.65 [26.10, 35.32] | 34.20 [29.35, 38.80] | <0.001 |
| Creatinine, mg/dL | 2.00 [1.40, 3.20] | 1.00 [0.70, 1.40] | <0.001 |
| BUN, mg/dL | 46.00 [27.00, 76.25] | 21.00 [15.00, 32.00] | <0.001 |
| Glucose, mg/dL | 136.00 [106.75, 183.00] | 130.00 [104.00, 169.00] | 0.029 |
| Sodium, mmol/L | 138.00 [135.00, 141.25] | 139.00 [136.00, 142.00] | 0.053 |
| Potassium, mmol/L | 4.60 [4.10, 5.30] | 4.20 [3.90, 4.70] | <0.001 |
| Calcium, mmol/L | 8.50 [7.90, 8.90] | 8.70 [8.20, 9.10] | <0.001 |
| Severity of illness | |||
| SOFA score | 8.00 [5.00, 10.00] | 4.00 [2.00, 6.00] | <0.001 |
| GCS score | 15.00 [14.00, 15.00] | 15.00 [14.00, 15.00] | 0.951 |
| Comorbidities | |||
| Myocardial infarct, % | 108 (33.8) | 774 (23.4) | <0.001 |
| Congestive heart failure, % | 235 (73.4) | 1642 (49.7) | <0.001 |
| Diabetes, % | 145 (45.3) | 1169 (35.3) | 0.001 |
| Cerebrovascular disease, % | 37 (11.6) | 533 (16.1) | 0.036 |
| Paraplegia, % | 10 (3.1) | 173 (5.2) | 0.109 |
| CKD, % | 184 (57.5) | 841 (25.4) | <0.001 |
| AKI, % | 302 (94.4) | 2650 (80.1) | <0.001 |
| Malignant cancer, % | 48 (15.0) | 589 (17.8) | 0.219 |
| CCI score | 8.00 [6.00, 10.00] | 6.00 [5.00, 8.00] | <0.001 |
| Intervention | |||
| Mechanical ventilation, % | 200 (54.2) | 211 (57.8) | 0.596 |
| Vasopressor, % | 180 (56.2) | 1034 (31.3) | <0.001 |
| Steroid, % | 163 (50.9) | 1345 (40.7) | <0.001 |
| Clinical outcomes | |||
| ICU LOS, days | 4.00 [2.20, 8.13] | 2.91 [1.82, 5.29] | <0.001 |
| 28-day mortality, % | 147 (45.9) | 511 (15.5) | <0.001 |
Abbreviations: MBP, mean arterial blood pressure; HR, heart rate; RR, respiratory rate; SpO2, oxygen saturation; WBC, White blood cell; BUN, blood urea nitrogen; SOFA, sequential organ failure assessment; GCS, glasgow coma scale; CKD, chronic kidney disease; AKI, acute kidney injury; CCI, charlson comorbidity index; ICU, intensive care unit; LOS, length of stay.
Association Between Serum Phosphate Trajectory and 28-Day Mortality
Univariable and multivariable Cox proportional hazards regression analyses were performed to evaluate the association between serum phosphate trajectory groups and 28-day all-cause mortality. Covariates demonstrating a statistically significant association with mortality in the univariable analysis (P < 0.05), together with clinically established confounders, were entered into the multivariable models. The results of the sequential Cox regression analyses are presented in Table 2. In the unadjusted model (Model 1), patients in the high-phosphate trajectory group exhibited a 2.77-fold greater hazard of 28-day mortality relative to the low/normal-phosphate group (HR = 2.77, 95% CI: 2.31–3.33, P < 0.001). Following adjustment for demographic characteristics including age, sex, ethnicity, and body weight (Model 2), the mortality risk associated with the high-phosphate trajectory intensified (HR = 3.03, 95% CI: 2.51–3.65, P < 0.001). In the fully adjusted model (Model 3), which additionally accounted for SOFA score, comorbidities (myocardial infarction, cerebrovascular disease, paraplegia, diabetes mellitus, AKI), Charlson Comorbidity Index score, vital signs (heart rate, respiratory rate, oxygen saturation), laboratory parameters (white blood cell count, platelet count, BUN, creatinine, calcium), and therapeutic interventions (mechanical ventilation, vasoactive agents), the high-phosphate trajectory persisted as an independent predictor of 28-day mortality (HR = 2.29, 95% CI: 1.85–2.84, P < 0.001).
Table 2.
Cox Proportional Hazards Regression Analysis of Phosphate Trajectory Groups and 28-Day Mortality
| Variable | Model I HR (95% CI, p-value) | Model II HR (95% CI, p-value) | Model III HR (95% CI, p-value) |
|---|---|---|---|
| Trajectory group | |||
| Group 2 | 1.00 (Reference) | 1.00 (Reference) | 1.00 (Reference) |
| Group 1 | 2.77 (2.31–3.33, p<0.001) | 3.03 (2.51–3.65, p<0.001) | 2.29 (1.85–2.84, p<0.001) |
Notes: Model I (Crude): Unadjusted. Model II (Demographics-adjusted): Adjusted for age, gender, weight and ethnicity. Model III (Fully adjusted): Adjusted for age, gender, ethnicity, weight, SOFA score, major comorbidities (myocardial infarction, cerebrovascular disease, paraplegia, diabetes, acute kidney injury), vital signs (heart rate, respiratory rate, mean blood pressure, oxygen saturation), laboratory parameters (white blood cell count, platelet count, blood urea nitrogen, creatinine, calcium), and treatment measures (mechanical ventilation, vasopressor use).
The 28-day survival curves stratified by trajectory group are depicted in Figure 3. The cumulative survival probability was significantly lower in the high-phosphate trajectory group (red curve) than in the low/normal-phosphate trajectory group (blue curve), a difference confirmed by the Log rank test (P < 0.001). Collectively, these findings indicate that a dynamic trajectory of persistently elevated serum phosphate during the ICU stay—independent of illness severity and baseline health status—carries substantial predictive value for 28-day mortality risk among patients with COPD.
Figure 3.
Kaplan-Meier survival curves for 28-day mortality by phosphate trajectory groups.
Comparative Predictive Value of Serum Phosphate Trajectories versus a Single Admission Measurement
To evaluate the incremental prognostic utility of longitudinal serum phosphate trajectories over a single admission measurement, we sequentially incorporated the initial serum phosphate level (modeled as both a continuous and a categorical variable), the phosphate trajectory group assignment, and their combination into a baseline covariate model (C-index = 0.7171) (Table 3). The addition of the initial serum phosphate level as a continuous variable yielded only a marginal improvement in model discrimination (ΔC-index = +0.0016; AIC decreased from 9246.36 to 9241.31); a similarly modest gain was observed when initial phosphate was entered as a categorical variable (ΔC-index = +0.0023; AIC = 9239.19). In marked contrast, incorporating the serum phosphate trajectory grouping substantially enhanced discriminatory performance (ΔC-index = +0.0119; AIC reduced to 9191.90), with the magnitude of improvement in the C-index exceeding that of the single continuous phosphate measurement by approximately 7.4-fold (0.0119 vs 0.0016). In the combined model incorporating both variables, the discriminatory power remained largely unchanged compared to the trajectory model alone (ΔC-index = +0.0119 for continuous; +0.0123 for categorical), and the AIC actually rose. This suggests substantial information overlap between static admission measurements and dynamic trajectories, indicating that the trajectory grouping captures the majority of the predictive information. Collectively, these findings indicate that dynamic serum phosphorus trajectories during the ICU stay possess greater prognostic value than a single admission assessment, suggesting that cumulative exposure to persistent hyperphosphataemia may better reflect the risk of mortality than phosphorus levels at a single point in time.
Table 3.
Comparison of Predictive Value Between Single Phosphate Measurement and Dynamic Trajectory for 28-Day Mortality
| Model | AIC | C-Index | ΔC-Index | P-value |
|---|---|---|---|---|
| Base | 9246.36 | 0.7171 | Reference | - |
| Base + Single (continuous) | 9241.31 | 0.7187 | +0.0016 | 0.008 |
| Base + Single (categorical) | 9239.19 | 0.7194 | +0.0023 | 0.009 |
| Base + Trajectory | 9191.90 | 0.7290 | +0.0119 | <0.001 |
| Base + Both (continuous) | 9193.89 | 0.7290 | +0.0119 | <0.001 |
| Base + Both (categorical) | 9193.15 | 0.7294 | +0.0123 | <0.001 |
Notes: Base model adjusted for age, gender, ethnicity, SOFA score, comorbidities (congestive heart failure, renal disease, malignancy, acute kidney injury), laboratory parameters (white blood cell count, platelet count, creatinine), vital signs (heart rate, respiratory rate, mean arterial pressure, oxygen saturation), and mechanical ventilation. Single (continuous) = first phosphate measurement as a continuous variable; Single (categorical) = first phosphate categorized as hypophosphatemia/normal/hyperphosphatemia; Trajectory = phosphate trajectory group; Both = single phosphate plus trajectory group simultaneously.
Subgroup Analysis
To evaluate the consistency of the association between high phosphate trajectories and 28-day mortality across different clinical characteristics, we performed stratified analyses by sex, age, SOFA score, AKI, mechanical ventilation, glucocorticoid use, and CKD (Figure 4). In the overall population, high phosphate trajectories were associated with a significantly increased risk of 28-day mortality (HR 2.77, 95% CI 2.31–3.33). The direction and magnitude of this association remained largely consistent across most predefined subgroups, with no significant interactions observed for sex (P for interaction = 0.548), age (P = 0.183), SOFA score (P = 0.427), AKI (P = 0.850), mechanical ventilation (P = 0.514), or CKD (P = 0.351). Notably, a significant interaction was found for glucocorticoid use (P for interaction = 0.005). The HR was substantially higher in patients not receiving glucocorticoids (HR 2.62, 95% CI 1.97–3.48) compared with those who did (HR 1.57, 95% CI 1.20–2.04). This finding suggests that glucocorticoid therapy may partially attenuate the association between high phosphate trajectories and mortality risk. Collectively, these results demonstrate that the adverse prognostic impact of high phosphate trajectories is generally robust across diverse patient subgroups, while glucocorticoid use may serve as a potential effect modifier in this association.
Figure 4.
Subgroup analysis of the impact of high-phosphorus trajectories on 28-day mortality across different subgroups.
Discussion
Leveraging the MIMIC-IV database, this study is the first to apply GBTM to delineate the longitudinal patterns of serum phosphate levels among patients with COPD during their ICU stay. Two distinct trajectory patterns were identified: a low/normal-phosphate trajectory, encompassing 91.2% of the cohort, and a high-phosphate trajectory, observed in 8.8% of patients. The high-phosphate trajectory was independently and robustly associated with 28-day mortality, an association that persisted following comprehensive adjustment for potential confounders. Moreover, the dynamic trajectory-based model demonstrated superior predictive performance compared with a single admission serum phosphate measurement, indicating that the temporal evolution of serum phosphate concentrations conveys prognostic information beyond that captured by conventional cross-sectional assessments.
A substantial body of clinical evidence indicates that electrolyte disturbances are highly prevalent among patients with COPD,16 a phenomenon that is particularly pronounced in those requiring ICU admission.17 These imbalances typically arise from a confluence of factors, encompassing both the underlying disease process and associated therapeutic interventions. Despite this, the clinical significance of serum phosphate derangements is frequently underestimated in routine practice. Phosphate imbalance represents a common electrolyte disorder and has been closely linked to an elevated risk of mortality.18 Preserving serum phosphate concentrations within the normal homeostatic range is essential for averting disease progression and adverse clinical outcomes. To date, the preponderance of research has concentrated on the relationship between a single admission serum phosphate measurement and adverse outcomes in critically ill populations.19 Raymond Farah et al demonstrated that hypophosphataemia carries prognostic significance in acute exacerbations of COPD.15 Li et al reported that elevated admission serum phosphate levels were positively associated with increased in-hospital mortality among patients with AECOPD, proposing that hyperphosphataemia may constitute a potential risk factor for in-hospital death.14 Conversely, Du et al identified a non-linear relationship between serum phosphate concentrations and both in-hospital and 90-day mortality in patients with severe COPD.20 Critically, all of the aforementioned investigations relied upon single cross-sectional assessments, thereby limiting their capacity to capture the dynamic fluctuations of serum phosphate over time and their sustained influence on prognosis. The present study is the first to employ a latent class mixture model to delineate longitudinal serum phosphate trajectories in COPD patients admitted to the ICU. Our findings revealed that the 28-day mortality risk in the high-phosphate trajectory group was substantially greater than that observed in the low/normal-phosphate trajectory group, and that the predictive performance of the dynamic trajectory model surpassed that of a single admission phosphate measurement. These results indicate that the temporal evolution of serum phosphate concentrations encodes prognostic information that conventional cross-sectional assessments fail to capture, thereby offering a novel foundation for early risk stratification in patients with COPD. The aetiology of elevated serum phosphate concentrations in critically ill patients is multifactorial and complex, encompassing reduced renal phosphate excretion due to impaired kidney function, intracellular phosphate release secondary to tissue injury, extracellular phosphate shifts driven by acidaemia, and increased exogenous phosphate loading, among other contributors.21–23 Nevertheless, the precise mechanisms through which hyperphosphataemia confers an elevated mortality risk in patients with COPD remain incompletely understood. Several interconnected pathophysiological pathways may be implicated. First, elevated serum phosphate levels can induce the osteogenic differentiation of vascular smooth muscle cells into osteoblast-like phenotypes, thereby accelerating vascular calcification, compromising microcirculatory perfusion, and ultimately precipitating organ dysfunction23 Moreover, the systemic inflammation and oxidative stress inherently present in patients with COPD may further potentiate the vascular endothelial injury attributable to hyperphosphataemia. Second, disturbances in phosphate homeostasis can impair tissue energy provision while simultaneously elevating energy expenditure, thereby promoting the release of mitochondrial reactive oxygen species and triggering inflammatory cascades, cellular senescence, and apoptosis.24–26 Against the background of already heightened inflammation during acute exacerbations of COPD, this establishes a deleterious “inflammation–hyperphosphataemia” vicious cycle. Furthermore, serum phosphate concentrations are independently associated with coronary artery calcification scores.27 Given that cardiovascular disease represents a frequent comorbidity in the COPD population, persistently elevated phosphate levels may exacerbate myocardial ischaemia and heart failure. These mechanisms likely act in concert, collectively contributing to the unfavourable prognosis observed among patients following a high-phosphate trajectory.
The trajectory patterns identified in this study may have practical implications for clinical monitoring. We propose that at least three phosphate measurements during the first 7 days of ICU admission may be sufficient to identify patients with a persistently high trajectory, who could be considered for closer surveillance. However, the optimal monitoring strategy requires prospective validation before formal recommendations can be made. Whether our findings extend to non-COPD ICU populations remains uncertain, as the specific trajectory patterns may differ by underlying disease; future studies are needed to validate our results in broader ICU cohorts. Additionally, a key question is whether hyperphosphatemia is merely a prognostic marker or a causal determinant of poor outcomes. Although we observed a robust association, the observational design precludes causal inference, and whether phosphate-lowering interventions confer clinical benefit remains unknown, requiring investigation in randomized controlled trials. Our findings should therefore be considered hypothesis-generating.
The present study leveraged serial serum phosphate measurements from COPD patients admitted to the ICU within the MIMIC-IV database to construct longitudinal trajectories of serum phosphate concentrations over the course of their ICU stay, and subsequently evaluated the association between distinct trajectory groups and 28-day mortality risk. The characterisation of dynamic trajectories facilitates the early identification of COPD patients at elevated mortality risk, thereby informing personalised monitoring strategies and timely therapeutic interventions. Nonetheless, several limitations of this study warrant consideration. First, given its retrospective cohort design, and although GBTM was employed to comprehensively capture the temporal dynamics of serum phosphate levels with adjustment for multiple known confounders, residual confounding from unmeasured variables—such as nutritional support regimens, phosphate binder administration, and ICU sedation protocols—may still influence the findings. The results should therefore be regarded as hypothesis-generating and require corroboration through prospective investigations. Second, this study examined serum phosphate trajectories confined to the initial 7 days of ICU admission; the trajectory patterns and their prognostic utility over extended observation windows—such as 14 days or the entire ICU length of stay—remain to be elucidated in future research. Nevertheless, the dynamic fluctuations in serum phosphate concentrations during this early 7-day period are sufficient to identify high-risk patients, providing a timely foundation for risk stratification in early clinical decision-making.
Conclusion
In summary, this study is the first to delineate longitudinal phosphate trajectories in ICU-admitted COPD patients. Our findings suggest that a high-phosphate trajectory may be an independent risk factor for 28-day mortality and may offer improved prognostic value over single phosphate measurements. These findings provide a preliminary basis for early risk stratification in COPD patients. However, given the retrospective design, prospective studies are warranted before clinical implementation.
Funding Statement
No funding was received.
Data Sharing Statement
Publicly available datasets were analyzed in this study. This data can be found here: https://mimic.mit.edu/.
Ethics Statement
The study was a retrospective analysis of publicly available, de-identified data, involved no direct contact with participants, no participant recruitment, no clinical intervention, and no additional collection of biological samples or personal identifiable information. According to Article 32 (Items 1 and 2) of the Measures for Ethical Review of Life Science and Medical Research Involving Human Subjects (China, issued February 18, 2023), use of de-identified public data may be exempt from additional ethical review when data use remains within original consent scope. Therefore, no additional institutional ethics approval was required for this analysis.
Author Contributions
All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.
Disclosure
The authors declare no competing interests.
References
- 1.Agustí A, Celli BR, Criner GJ, et al. Global initiative for chronic obstructive lung disease 2023 report: GOLD executive summary. Europ Resp J. 2023;61:2300239. doi: 10.1183/13993003.00239-2023 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Brightling C, Greening N. Airway inflammation in COPD: progress to precision medicine. Europ Resp J. 2019;54:1900651. doi: 10.1183/13993003.00651-2019 [DOI] [PubMed] [Google Scholar]
- 3.Kahnert K, Jörres RA, Behr J, Welte T. The diagnosis and treatment of COPD and its comorbidities. Deutsches Arzteblatt Int. 2023;120(25):434–11. doi: 10.3238/arztebl.m2023.027 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Prediletto I, Giancotti G, Nava S. COPD exacerbation: why it is important to avoid ICU admission. J Clin Med. 2023;12:3369. doi: 10.3390/jcm12103369 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Hartl S, Lopez-Campos JL, Pozo-Rodriguez F. Risk of death and readmission of hospital-admitted COPD exacerbations: European COPD audit. Europ Resp J. 2016;47(1):113–121. doi: 10.1183/13993003.01391-2014 [DOI] [PubMed] [Google Scholar]
- 6.Kalantar-Zadeh K, Ganz T, Trumbo H, Seid MH, Goodnough LT, Levine MA. Parenteral iron therapy and phosphorus homeostasis: a review. Am J Hematol. 2021;96(5):606–616. doi: 10.1002/ajh.26100 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Couce ML, Saenz de Pipaon M. Bone mineralization and calcium phosphorus metabolism. Nutrients. 2021;13:3692. doi: 10.3390/nu13113692 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Peacock M. Phosphate METABOLISM IN HEALTH AND DISease. Calcified Tissue Int. 2021;108(1):3–15. doi: 10.1007/s00223-020-00686-3 [DOI] [PubMed] [Google Scholar]
- 9.Sinatra N, Cuttone G, Geraci G, et al. Correlation between hypophosphatemia and hyperventilation in critically ill patients: causes, clinical manifestations, and management strategies. Biomedicines. 2025;13:2382. doi: 10.3390/biomedicines13102382 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Supinski GS, Morris PE, Dhar S, Callahan LA. Diaphragm dysfunction in critical illness. Chest. 2018;153(4):1040–1051. doi: 10.1016/j.chest.2017.08.1157 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Pongsuwan K, Kusirisin P, Narongkiattikhun P, Chattipakorn SC, Chattipakorn N. Mitochondria and vascular calcification in chronic kidney disease: lessons learned from the past to improve future therapy. J Cell Physiol. 2022;237(12):4369–4396. doi: 10.1002/jcp.30891 [DOI] [PubMed] [Google Scholar]
- 12.Suzuki S, Egi M, Schneider AG, Bellomo R, Hart GK, Hegarty C. Hypophosphatemia in critically ill patients. J Crit Care. 2013;28:536.e539–519. doi: 10.1016/j.jcrc.2012.10.011 [DOI] [PubMed] [Google Scholar]
- 13.Chaba A, Doola R, Laupland KB. Feeding-associated hypophosphatemia in critically ill patients. A multicentre, observational study. J Crit Care. 2026;92:155400. doi: 10.1016/j.jcrc.2025.155400 [DOI] [PubMed] [Google Scholar]
- 14.Li S, Huang Q, Nan W, He B. Association between serum phosphate and in-hospital mortality of patients with AECOPD: a retrospective analysis on eICU database. Heliyon. 2023;9(9):e19748. doi: 10.1016/j.heliyon.2023.e19748 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Farah R, Khamisy-Farah R, Arraf Z, Jacobson L, Makhoul N. Hypophosphatemia as a prognostic value in acute exacerbation of COPD. Clin Respir J. 2013;7:407–415. doi: 10.1111/crj.12027 [DOI] [PubMed] [Google Scholar]
- 16.Deep A, Behera PR, Subhankar S, Rajendran A, Rao CM. Serum electrolytes in patients presenting with acute exacerbation of Chronic Obstructive Pulmonary Disease (COPD) and their comparison with stable COPD patients. Cureus. 2023;15(4):e38080. doi: 10.7759/cureus.38080 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Menteş O, Çelik D, Yildiz M, et al. Electrolyte imbalance and its prognostic impact on all-cause mortality in ICU patients with respiratory failure. Medicina. 2025;61:642. doi: 10.3390/medicina61040642 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Chen Y, Luo M, Xu H, Zhao W, He Q. Association between serum phosphate and mortality in critically ill patients: a large retrospective cohort study. BMJ Open. 2021;11(9):e044473. doi: 10.1136/bmjopen-2020-044473 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Yang J, Cheng Y, Wang R, Wang B. Association between early elevated phosphate and mortality among critically ill elderly patients: a retrospective cohort study. BMC Geriatr. 2022;22(1):208. doi: 10.1186/s12877-022-02920-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Du S, Lin K, Li J. Association between the serum phosphate levels and hospital mortality as well as 90-day mortality among critically ill patients with chronic obstructive pulmonary disease: a retrospective cohort study. Int J Chronic Obstr. 2024;19:1681–1693. doi: 10.2147/copd.S465752 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Thongprayoon C, Cheungpasitporn W, Chewcharat A. Hospital-acquired serum phosphate derangements and their associated in-hospital mortality. Postgrad Med J. 2022;98(1155):43–47. doi: 10.1136/postgradmedj-2020-138872 [DOI] [PubMed] [Google Scholar]
- 22.Shuto E, Taketani Y, Tanaka R. Dietary phosphorus acutely impairs endothelial function. J Am Soc Nephrol. 2009;20(7):1504–1512. doi: 10.1681/asn.2008101106 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Raikou VD. Serum phosphate and chronic kidney and cardiovascular disease: phosphorus potential implications in general population. World J Nephrol. 2021;10(5):76–87. doi: 10.5527/wjn.v10.i5.76 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Hu MC, Moe OW. Phosphate and cellular senescence. Adv Exp Med Biol. 2022;55–72. doi: 10.1007/978-3-030-91623-7_7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Maique J, Flores B, Shi M, et al. High phosphate induces and klotho attenuates kidney epithelial senescence and fibrosis. Front Pharmacol. 2020;11:1273. doi: 10.3389/fphar.2020.01273 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Michigami T, Yamazaki M, Razzaque MS. Extracellular phosphate, inflammation and cytotoxicity. Adv Exp Med Biol. 2022;15–25. doi: 10.1007/978-3-030-91623-7_3 [DOI] [PubMed] [Google Scholar]
- 27.Voelkl J, Egli-Spichtig D, Alesutan I, Wagner CA. Inflammation: a putative link between phosphate metabolism and cardiovascular disease. Clin Sci. 2021;135:201–227. doi: 10.1042/cs20190895 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Publicly available datasets were analyzed in this study. This data can be found here: https://mimic.mit.edu/.