ANALYSIS OF THE RELATIONSHIP BETWEEN EARLY SERUM PHOSPHATE LEVELS AND SHORT-TERM MORTALITY IN SEPTIC PATIENTS: A RETROSPECTIVE STUDY BASED ON MIMIC-IV

Abstract

Objective: The aim of the study is to explore the impact of early serum phosphate levels on the prognosis of critically ill patients with sepsis. Methods: In this retrospective large cohort study, data of patients with sepsis were obtained from the Medical Information Mart for Intensive Care IV database. Patients were retrospectively divided into a control group and three study groups according to their daily serum phosphate levels within 2 days of intensive care unit (ICU) admission. A Cox regression model was used to evaluate the association between serum phosphate levels and 28-day morbidity. Results: This study included 9,691 patients diagnosed with sepsis. During the first 2 days of ICU admission, patients with hyperphosphatemia in either of the 2 days had higher 28-day mortality, while patients in the hypophosphatemia group had lower 28-day mortality (first day, 32.9% vs. 16.3%; second day, 36.3% vs. 14.7%). After adjusting for potential confounders, hyperphosphatemia was significantly associated with 28-day mortality; however, only hypophosphatemia on the second day was independently associated with reduced 28-day mortality. After stratification in the hypophosphatemia group, subgroup analysis showed that only the association between the mild hypophosphatemia group and 28-day mortality reached statistical significance (hazard ratio = 0.76, 95% CI = 0.65–0.89, P = 0.001). Conclusions: Mild hypophosphatemia might improve the short-term prognosis of patients with sepsis, and hyperphosphatemia is an independent risk factor for the outcomes of septic patients. After ICU admission, the serum phosphate levels on the second day had a better independent correlation with 28-day mortality, which prompted us to reconsider the optimal timing of phosphate evaluation.

INTRODUCTION

Phosphorus is an essential component in living organisms and is involved in many pathophysiological processes in the body, mainly in the form of phosphate in humans. Phosphate homeostasis is vital for skeletal development, energy metabolism, membrane structural integrity, and cell signaling (^1–3). Most phosphate bound to calcium in the body (approximately 85%) is stored in bones and teeth, and the rest (approximately 14% of the total) is found primarily in tissue cells. Phosphate in the serum and extracellular fluid represents only 1% of systemic phosphate. Only 30% is inorganic phosphate, representing the regulated variable and the component that can be measured and regulated in clinical practice (^1,2). Phosphate is also a necessary anion in the extracellular fluid. Normal serum phosphate levels range from 2.5–4.5 mg/dL (0.80–1.45 mmol/L). Homeostatic regulation of serum phosphate levels involves several major organ systems, including the cardiovascular, respiratory, immune, urinary, and neuromuscular (⁴). Serum phosphate depends on dietary phosphate intake, extracellular phosphate transfer, glomerular function, and tubular phosphate reabsorption and is influenced by various regulatory factors (⁵). Moreover, a transient redistribution of serum phosphate is also stimulated by insulin and respiratory alkalosis (⁶). Abnormalities in serum phosphate can be associated with various diseases, especially in critically ill patients (^2,7,8).

Sepsis is a multifaceted host response to an infecting pathogen that endogenous factors may significantly amplify and is also one of the most common diseases in the intensive care unit (ICU) (⁹). Although many advances have been made in treating sepsis, sepsis remains a significant cause of morbidity and mortality worldwide. Given that the overall incidence of sepsis appears to be increasing rapidly, overall mortality is not significantly improving, demonstrating the challenge’s continuing magnitude (¹⁰). The global burden of sepsis may be much higher than previously reported (¹¹). Many septic patients present with abnormal serum phosphate levels, and the predictive value of serum phosphate has been studied extensively in other specific diseases. However, its prognostic value for patients with sepsis remains poorly investigated (^12–15).

In previous studies of critically ill patients with abnormal serum phosphate levels, most concluded that hyperphosphatemia was associated with a worse clinical outcome. In contrast, the relationship between hypophosphatemia and clinical outcome was inconsistent (^14,16–22). Possible reasons may be the different definitions of hypophosphatemia, the study population, and whether the results were adjusted for confounding factors. The relationship between serum phosphate and patient prognosis is complex and is related not only to serum phosphate concentration levels but also to phosphate fluctuations (²³). In dangerous acute critical illnesses such as acute sepsis, the clinical outcome of patients may be susceptible to both serum phosphate concentration and fluctuating levels. The discrepancy in the results of numerous studies suggests that the effect of abnormal serum phosphate on patient prognosis is complex and may differ in the prognosis of patients with different etiologies. These conclusions were equally controversial in studies on the relationship between sepsis and serum phosphate levels. For instance, in the study by Shor et al. (²²), severe hypophosphatemia in sepsis increased the risk of death by nearly 8-fold. In contrast, the Jang et al. (²⁰) found that patients in the hypophosphatemia group had relatively lower mortality. Still, hypophosphatemia was not independently associated with 28-day mortality in patients with sepsis, and hyperphosphatemia was an independent risk factor for increased mortality.

Disease severity and etiology vary considerably in intensive care units (ICUs). Therefore, it is necessary to re-evaluate the relationship between serum phosphate levels and clinical outcomes in patients with sepsis. Patients with sepsis often have changes in homeostasis of the internal environment in the early stages, leading to fluctuations in serum ion concentrations; therefore, the initial serum phosphate concentration may not adequately reflect the relationship between serum phosphate abnormalities and patient clinical outcomes. We conducted a single-center retrospective cohort study using patient data from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database to investigate the association between serum phosphate abnormalities in the first 2 days and 28-day mortality in patients with sepsis.

MATERIALS AND METHODS

Sources of data

The data for our study were obtained from a public database named the MIMC-IV version 2.0 (https://physionet.org/content/mimiciv/2.0) (²⁴). The database provides critical care data for more than 53,000 patients admitted to ICUs at the Beth Israel Deaconess Medical Center from 2008 to 2019. We completed the National Institutes of Health web-based training course and the Protecting Human Research Participants examination (no. 47648550) to gain access to the database. Hospitalization information was deidentified, and informed consent was not required.

Patients

Patients meeting the diagnostic criteria of Sepsis-3 within 24 hours of ICU admission were eligible for inclusion in our study; namely, all patients with suspected infection and an acute change in total Sequential Organ Failure Assessment (SOFA) score of 2 points or higher were considered to have sepsis (⁹). The exclusion criteria were as follows: (1) patients younger than 18 years; (2) patients with multiple records of ICU admission, excluding records other than the first ICU admission; (3) patients who did not have complete records within 2 days after admission to the ICU, and patients with missing values were excluded from our research; and (4) patients whose length of stay in the ICU were less than 2 days.

Data collection

Structured Query Language was used to extract data. We extracted or calculated the following variables, including (1) baseline characteristics: age, sex, and weight; (2) pre-ICU comorbidities: hypertension, diabetes, chronic kidney disease, myocardial infarction, congestive heart failure, chronic pulmonary disease, liver disease, and malignant cancer; (3) the severity of organ dysfunction: Sequential Organ Failure Assessment, SAPS II (Simplified Acute Physiology Score), OASIS (Oxford Acute Severity of Illness Score); (4) laboratory parameters (within 2 days after ICU admission): white blood cell count, hemoglobin, platelet, serum creatinine, blood lactate (maximum value within 2 days), and daily laboratory serum electrolyte records (phosphate, sodium, potassium, calcium, bicarbonate, chloride); (5) treatment received: vasoactive agent use, mechanical ventilation received, and renal replacement therapy; (6) hospital and ICU stay data were extracted and calculated as follows: all-cause 28-day mortality, hospital and ICU length of stay (HLOS and ILOS). Vasoactive agents were defined as any use of norepinephrine, epinephrine, dopamine, phenylephrine, milrinone, and dobutamine within the first 2 days of ICU admission. It is well known that elevated lactate levels are directly related to the prognosis of patients with sepsis, so we chose the maximum lactate value over 2 days. Because electrolyte levels are more likely to fluctuate because of homeostatic and therapeutic factors, we calculated the mean values of daily laboratory serum electrolyte results and mean values of others laboratory parameters in 2 days.

We extracted serum phosphate values within the first and second days after ICU admission and divided the patients into four groups according to their daily serum phosphate concentrations. Because normal serum phosphate levels range from 2.5 to 4.5 mg/dL (0.80–1.45 mmol/L) in adults, hypophosphatemia is defined as serum phosphate concentrations less than 2.5 mg/dL (0.80 mmol/L), and hyperphosphatemia is defined as serum phosphate concentrations greater than 4.5 mg/dL (1.45 mmol/L) (⁴). The control group was patients with explicitly normal phosphate values (2.5 ≤ P ≤ 4.5 mg/dL). Patients with all phosphate values less than 2.5 mg/dL composed the hypophosphatemia group, and those with all phosphate values greater than 4.5 mg/dL composed the hyperphosphatemia group. The mixed group consisted of patients with at least one phosphate value (P < 2.5 mg/dL or P > 4.5 mg/dL) and the remaining phosphate values in the other groups.

In the subgroup analysis, patients in the hypophosphatemia group were divided into three groups according to their serum phosphate concentrations. In a retrospective study by Li et al. (²⁵), the authors found that severe hypophosphatemia (<1.5 mg/dL in this study) had higher in-hospital mortality, but this was not statistically significant in the overall septic population. The defined thresholds in our subgroup analysis also referred to the findings of the study mentioned previously. Patients with all phosphate values of 1.5 mg/dL or less composed the severe hypophosphatemia group, and those with all phosphate values greater than 1.5 mg/dL composed the mild hypophosphatemia group. The mixed hypophosphatemia group consisted of the remaining patients in the hypophosphatemia group with phosphate values in different subgroup ranges.

Outcome variables

The primary endpoint was the all-cause 28-day mortality. The ICU length of stay and length of hospital stay were secondary endpoints. Moreover, secondary endpoints were extracted only for descriptive purposes. The 28-day mortality data were confirmed by inspecting the death records in the database.

Statistical analysis

Continuous variables in baseline characteristics are presented as mean ± SD or median with interquartile ranges. Differences in continuous data were compared by the Student t test and the Mann-Whitney test as appropriate. Categorical variables are presented as percentages. Analysis of variance (or Kruskal-Wallis test) and χ² (or Fisher exact) tests were used to compare the different groups. The cumulative risk of death within the different groups of daily serum phosphate levels is presented using Kaplan-Meier curves. Kaplan-Meier survival curves were compared using the log-rank test. A Cox regression model was used to estimate the hazard ratio (HR) for mortality for each study and subgroup compared with the control group while adjusting for the confounding variables. A univariable Cox regression model was used to identify risk factors affecting survival. Two different models were used to adjust for potential confounders. Model 1 included the confounding variables that were statistically significant in the prior univariable Cox regression analysis for mortality; model 2 included all the confounding variables: age, sex, weight, all ICU comorbidities, SOFA score, SAPS II, OASIS score, white blood cell count, hemoglobin, platelet, creatinine, blood lactate, and daily laboratory serum electrolyte results within 2 days after ICU admission (phosphate, sodium, potassium, calcium, bicarbonate, chloride), and treatment received (vasoactive agent use, ventilation received, vasopressor therapy, renal replacement therapy). The software R (http://www.R-project.org; The R Foundation; version 4.1.1) was used for statistical analyses.

RESULTS

Characteristics of patients

A total of 9,691 patients with sepsis were enrolled in this study, which included 7,503 survivors and 2,188 nonsurvivors who were stratified according to 28-day mortality. A flowchart of the study cohort selection process is presented in Figure 1. The characteristics of the participants are listed in Table 1.

Fig. 1: Flowchart of patient selection for the study. Visual representation of how the 9,691 ICU stays that we analyzed were selected from the 76,943 admissions in MIMIC-IV.

Table 1 - Baseline characteristics of the total cohort, 28-day survivors, and 28-day nonsurvivors

	All patients	Survivors	Nonsurvivors	P
Variables	(N = 9,691)	(n = 7,503)	(n = 2,188)
Age, y	65.21 (16.61)	63.81 (16.66)	70.01 (15.53)	<0.001
Sex, n (%)	5,623 (58.0)	4,390 (58.5)	1,233 (56.4)	0.076
Weight, kg	85.87 (25.04)	86.86 (25.12)	82.48 (24.47)	<0.001
SOFA	3.00 (2.00, 5.00)	3.00 (2.00–5.00)	4.00 (2.00–5.00)	<0.001
OASIS	38.00 (32.00–45.00)	37.00 (31.00–43.00)	43.00 (37.00–49.00)	<0.001
SAPS II	42.00 (33.00–52.00)	40.00 (31.00–49.00)	50.00 (40.00–59.00)	<0.001
WBC, K/μL	13.35 (6.53)	13.01 (6.20)	14.53 (7.43)	<0.001
Hemoglobin, g/dL	10.30 (1.86)	10.35 (1.82)	10.16 (1.96)	<0.001
PLT, K/μL	186.80 (99.01)	187.44 (96.21)	184.62 (108.08)	0.241
Lactate, mmol/L	3.13 (2.69)	2.89 (2.35)	3.94 (3.50)	<0.001
Creatinine, mg/dL	1.62 (1.51)	1.56 (1.53)	1.83 (1.41)	<0.001
Renal replacement therapy, n (%)	957 (9.9)	559 (7.5)	398 (18.2)	<0.001
Mechanical ventilation, n (%)	6,375 (65.8)	4,844 (64.6)	1,531 (70.0)	<0.001
Vasoactive drugs, n (%)	5,730 (59.1)	4,271 (56.9)	1,459 (66.7)	<0.001
Myocardial infarction, n (%)	1819 (18.8)	1,334 (17.8)	485 (22.2)	<0.001
Congestive heart failure, n (%)	3,106 (32.1)	2,292 (30.5)	814 (37.2)	<0.001
Chronic pulmonary disease, n (%)	2,763 (28.5)	2,102 (28.0)	661 (30.2)	0.048
Diabetes, n (%)	2,967 (30.6)	2,311 (30.8)	656 (30.0)	0.481
Chronic kidney disease, n (%)	2,142 (22.1)	1,568 (20.9)	574 (26.2)	<0.001
Malignant cancer, n (%)	1,350 (13.9)	913 (12.2)	437 (20.0)	<0.001
Hypertension, n (%)	3,897 (40.2)	3,076 (41.0)	821 (37.5)	0.004
Liver disease, n (%)	1820 (18.8)	1,239 (16.5)	581 (26.6)	<0.001
Laboratory serum electrolyte (day 1)
Sodium, mmol/L	138.63 (5.23)	138.61 (4.92)	138.70 (6.19)	0.475
Potassium, mmol/L	4.24 (0.60)	4.22 (0.58)	4.30 (0.65)	<0.001
Calcium, mmol/L	8.15 (0.78)	8.15 (0.77)	8.17 (0.83)	0.274
Chloride, mmol/L	104.65 (6.57)	104.82 (6.27)	104.08 (7.46)	<0.001
Bicarbonate, mmol/L	22.15 (4.73)	22.41 (4.56)	21.28 (5.17)	<0.001
Phosphate (mg/dl)	3.83 (1.39)	3.74 (1.33)	4.14 (1.54)	<0.001
Laboratory serum electrolyte (day 2)
Sodium, mmol/L	138.74 (5.13)	138.70 (4.78)	138.87 (6.17)	0.185
Potassium, mmol/L	4.11 (0.54)	4.08 (0.51)	4.21 (0.61)	<0.001
Calcium, mmol/L	8.16 (0.69)	8.16 (0.66)	8.17 (0.79)	0.329
Chloride, mmol/L	104.70 (6.47)	104.70 (6.12)	104.71 (7.53)	0.965
Bicarbonate, mmol/L	23.22 (4.71)	23.70 (4.48)	21.55 (5.08)	<0.001
Phosphate, mg/dL	3.55 (1.45)	3.41 (1.36)	4.02 (1.64)	<0.001

OASIS, oxford acute severity of illness score; PLT platelet; SAPS II, simplified acute physiology score; SOFA, sequential organ failure assessment; WBC, white blood cell.

The mean age of study participants was 65.21 ± 16.61 years, with 58.0% male. The overall 28-day mortality rate was 22.5%. Nonsurvivors were lower in weight and older (P < 0.001), and a higher proportion had comorbidities. The SAPS II, OASIS, and SOFA scores were significantly higher in the nonsurvivors than in the survivors. We also found that nonsurvivors had higher lactate levels, white blood cell counts, serum phosphate concentrations, and creatinine levels. There were some differences in the other biochemical parameters, but the overall difference was minor. Nonsurvivors were more likely to require renal replacement, mechanical ventilation, or supportive treatment with vasoactive agents (Table 1).

Clinical outcomes and characteristics of the study groups

Table 2 shows septic patients’ baseline characteristics and outcomes according to phosphate levels in 2 days. The number of patients in the hypophosphatemia and mixed phosphate groups was more balanced on the second day compared with the first day. A much smaller proportion of patients in the hypophosphatemia and control groups required vasoactive agents and renal replacement therapy. The SAPS II, OASIS, and SOFA scores of patients in the hyperphosphatemia and mixed groups were significantly higher than other groups. The hyperphosphatemia and mixed groups had relatively higher blood lactate concentrations, and their levels were similar in these two groups. In the first 2 days after ICU admission, the 28-day mortality rate was lowest in the hypophosphatemia group and highest in the hyperphosphatemia group (Table 2). The proportion of patients with comorbidities was lower in the hypophosphatemia group and highest in the hyperphosphatemia group (Supplementary Table 1, https://links.lww.com/SHK/B668).

Table 2 - Baseline characteristics and outcomes of septic patients according to phosphate levels in 2 days

Variables	Time	Group 1	Group 2	Group 3	Group 4	P
	Day 1	(n = 4,260)	(n = 631)	(n = 1,420)	(n = 3,380)
	Day 2	(n = 4,761)	(n = 1,692)	(n = 1,557)	(n = 1,681)
SOFA	Day 1	3.00 (2.00–4.00)	3.00 (2.00–4.00)	4.00 (3.00–6.00)	3.00 (2.00–5.00)	<0.001
	Day 2	3.00 (2.00–4.00)	3.00 (2.00–4.00)	4.00 (3.00–6.00)	4.00 (2.00–5.00)	<0.001
OASIS	Day 1	37.00 (31.00–43.00)	36.00 (31.50–42.00)	41.00 (34.00–48.00)	39.00 (33.00–46.00)	<0.001
	Day 2	37.00 (32.00–43.00)	36.00 (31.00–42.00)	41.00 (35.00–48.00)	40.00 (34.00–46.00)	<0.001
SAPS II	Day 1	39.00 (31.00–48.00)	38.00 (30.00–45.00)	50.00 (41.00–61.00)	43.00 (34.00–54.00)	<0.001
	Day 2	40.00 (32.00–50.00)	37.00 (29.00–46.00)	51.00 (42.00–61.00)	44.00 (35.00–55.00)	<0.001
Lactate	Day 1	2.56 (1.84)	2.49 (1.91)	3.69 (3.39)	3.73 (3.16)	<0.001
	Day 2	2.72 (2.06)	2.60 (1.89)	3.96 (3.58)	4.04 (3.45)	<0.001
Vasoactive agents, n (%)	Day 1	2,320 (54.5)	316 (50.1)	934 (65.8)	2,160 (63.9)	<0.001
	Day 2	2,641 (55.5)	883 (52.2)	1,082 (69.5)	1,124 (66.9)	<0.001
Outcomes
Hospital LOS, d	Day 1	14.61 (12.90)	14.41 (15.45)	16.99 (15.22)	16.43 (15.03)	<0.001
	Day 2	14.82 (13.39)	14.78 (13.08)	16.93 (15.48)	17.28 (16.06)	<0.001
ICU LOS, d	Day 1	7.05 (7.00)	7.16 (6.47)	8.05 (7.23)	8.06 (7.75)	<0.001
	Day 2	7.08 (6.77)	7.42 (7.41)	8.08 (7.55)	8.56 (8.13)	<0.001
28-Day mortality, n (%)	Day 1	835 (19.6)	103 (16.3)	467 (32.9)	783 (23.2)	<0.001
	Day 2	983 (20.6)	248 (14.7)	565 (36.3)	392 (23.3)	<0.001

Group 1: control group; Group 2: hypophosphatemia group; Group 3: hyperphosphatemia group; Group 4: mixed phosphate group; LOS, length of stay; OASIS, oxford acute severity of illness score; SAPS II, simplified acute physiology score; SOFA, sequential organ failure assessment.

In Figure 2, Kaplan-Meier survival curves of 28-day mortality are plotted for different groups and compared using the log-rank test. Compared with the other groups, the hypophosphatemia group had the lowest 28-day mortality each day. The 28-day mortality was in the order of hyperphosphatemia group > mixed group > control group > hypophosphatemia (Table 2, Fig. 2). The hypophosphatemia and control groups still had relatively better lactate levels and less time to stay in the ICU and the hospital. The details are presented in Table 2.

Fig. 2: Kaplan-Meier curve for the cumulative hazard of 28-day mortality A visual summary of the cumulative risk of 28-day mortality in the first 2 days. The “first day” represents the Kaplan-Meier curves of four groups (control group, hypophosphatemia group, hyperphosphatemia group, mixed phosphate group) on the first day of ICU stay. The “second day” represents the Kaplan-Meier curves of four groups (control group, hypophosphatemia group, hyperphosphatemia group, mixed phosphate group) on the second day of ICU stay.

Associations between serum phosphate and mortality

To eliminate the influence of possible confounding factors, univariate and multivariate Cox regression analyses were used to determine the relationship between serum phosphate levels and all-cause 28-day mortality. The results are presented in Table 3 and Supplementary Table 2, https://links.lww.com/SHK/B668. Compared with patients in the control group, patients with hyperphosphatemia and those in the mixed group had increased 28-day ICU mortality with a nonadjusted HR (hyperphosphatemia group: HR = 1.85, 95% CI = 1.65–2.07, P < 0.001; mixed group: HR = 1.22, 95% CI = 1.11–1.34, P < 0.001, respectively). The association between hypophosphatemia and 28-day mortality was not statistically significant (hypophosphatemia group: HR = 0.82, 95% CI = 0.67–1.01, P = 0.058). However, after adjusting for model 1 or model 2, the association between the mixed group and 28-day mortality was no longer statistically significant (model 1: HR = 0.97, 95% CI = 0.88–1.08, P = 0.608; model 2: HR = 0.98, 95% CI = 0.88–1.08, P = 0.642, respectively). These patients in the hyperphosphatemia group still had increased 28-day ICU mortality, and the result was consistent after being adjusted by model 1 and model 2 (model 1: HR = 1.20, 95% CI = 1.05–1.37, P = 0.008; model 2: HR = 1.17, 95% CI = 1.02–1.33, P = 0.025, respectively).

Table 3 - Univariable and multivariable Cox regression analyses for 28-day mortality of septic patients according to phosphate levels in 2 days

Variables	Unadjusted, HR (95% CI), P	Model 1, HR (95% CI), P	Model 2, HR (95% CI), P
First day
Normal phosphate	1.0 (Ref)	1.0 (Ref)	1.0 (Ref)
Hypophosphatemia	0.82 (0.67–1.01), 0.058	0.91 (0.74–1.12), 0.374	0.93 (0.76–1.15),0.503
Hyperphosphatemia	1.85 (1.65–2.07), <0.001	1.20 (1.05–1.37),0.008	1.17 (1.02–1.33),0.025
Mixed phosphate	1.22 (1.11–1.34), <0.001	0.97 (0.88–1.08),0.608	0.98 (0.88–1.08),0.642
Second day
Normal phosphate	1.0 (Ref)	1.0 (Ref)	1.0 (Ref)
Hypophosphatemia	0.69 (0.6–0.79), <0.001	0.78 (0.68–0.90), 0.001	0.79 (0.68–0.91), 0.001
Hyperphosphatemia	1.99 (1.8–2.21), <0.001	1.34 (1.19, 1.52), <0.001	1.39 (1.23, 1.58), <0.001
Mixed phosphate	1.17 (1.04–1.31), 0.01	0.92 (0.82, 1.04), 0.197	0.95 (0.84, 1.07), 0.378

After regrouping all patients according to the second-day serum phosphate values, the relationships between serum phosphate levels and 28-day mortality were significantly different from those of the first-day results. Patients in the hyperphosphatemia and mixed groups with a nonadjusted HR had an increased risk for events, but patients with hypophosphatemia had a reduced risk (hyperphosphatemia group: HR = 1.99, 95% CI = 1.8–2.21, P < 0.001; mixed group: HR = 1.17, 95% CI = 1.04–1.31, P = 0.01; hypophosphatemia group: HR = 0.69, 95% CI = 0.6–0.79, P < 0.001, respectively). After adjusting for the two models, the association between the hypophosphatemia group and mortality remained statistically significant (model 1: HR = 0.78, 95% CI = 0.68–0.90, P = 0.001; model 2: HR = 0.79, 95% CI = 0.68–0.91, P = 0.001, respectively). Patients in the hyperphosphatemia group still had an increased risk of death. All the results are shown in Table 3.

Subgroup analysis was conducted to further clarify the association between different degrees of hypophosphatemia and mortality in patients with sepsis. The above results showed that the association between hypophosphatemia on the first-day and 28-day mortality was not statistically significant, and we found that mortality was close across subgroups of hypophosphatemia on the first day. Hence, we only performed a stratified analysis for patients in the hypophosphatemia group according to the second-day serum phosphate values. The normal serum phosphate group was used as the reference group. Patients with mild hypophosphatemia had the lowest mortality rate compared with other patients with hypophosphatemia. We found that only mild hypophosphatemia benefited the 28-day mortality of patients independently after adjusting for confounders by model 2 (HR = 0.76, 95% CI = 0.65–0.89, P = 0.001) (Fig. 3).

Fig. 3: The association between two-day serum phosphate levels and different time mortalities in septic patients. The HR was calculated after adjusting all variables by model 2. Subgroup analyses only involved patients with hypophosphatemia and normal phosphate on the second day.

DISCUSSION

In this retrospective cohort study, our primary research objective was to investigate the association between early serum phosphate levels and 28-day mortality in patients with sepsis. We found that the serum phosphate level was an independent predictor of mortality in patients with sepsis. However, the timing of the assessment was crucial, where serum phosphate levels on the second day revealed this association more accurately than the initial value. Hyperphosphatemia on either of the 2 days was an independent risk factor of 28-day mortality in patients with sepsis. In contrast, hypophosphatemia with stable serum phosphate levels on the second day had a beneficial effect on 28-day mortality; mild hypophosphatemia was independently associated with reduced 28-day mortality. To the best of our knowledge, this is the first study to determine the protective effect of hypophosphatemia on the outcomes of patients with sepsis after adjusting for confounding factors. It revealed that mild hypophosphatemia has the best beneficial impact on 28-day mortality.

Although some studies have explored the relationship between serum phosphate abnormalities and the prognosis of critically ill patients, their findings are varied and controversial. This may be due to the small sample size, different definitions of serum phosphate thresholds, and different initial etiologies of the study population. Most findings suggest that hyperphosphatemia is associated with adverse clinical outcomes. In the study by Haider et al. (¹⁷), which involved 2,390 critically ill patients, they reported that hyperphosphatemia was an independent risk factor for mortality, and the length of hospital stay of patients with hyperphosphatemia was longer than patients with hypophosphatemia or normal phosphate; however, hypophosphatemia was not associated with mortality. Other studies also found that hyperphosphatemia was associated with worse outcomes in critically ill patients (^14,18,20,26). In all of these studies, the normal serum phosphate group was the control group for the higher risk of hyperphosphatemia. Nevertheless, studies focusing on the association between serum phosphate levels and the clinical outcomes of patients with sepsis remain relatively short. In our study with septic patients as the study population, the results regarding the effect of hyperphosphatemia were similar to those of the studies mentioned previously, supporting that hyperphosphatemia is an independent risk factor for poor prognosis. While many findings were consistent with the association between hyperphosphatemia and poor prognosis, most studies on the prognostic impact of hypophosphatemia have reached different conclusions. In a small sample of 55 patients, the results showed that severe hypophosphatemia was associated with an eight-fold increased mortality risk (²²). In a retrospective study including patients with bloodstream infection in different ICUs, it was concluded that hypophosphatemia was independently associated with a two-fold increase in mortality in patients with bloodstream infection (²⁷). However, this study only included patients from a single hospital and assessed a single measurement of serum phosphate during ICU admission. The findings of two multicenter retrospective studies regarding the relationship between hypophosphatemia at admission and prognosis were inconsistent (^16,28). None of the above studies considered the effect of possible fluctuations in serum phosphate levels on outcomes, and most studies only evaluated initial serum phosphate levels. In a study by Kim et al. (²³), higher concentrations of serum phosphate and fluctuations in levels were associated with higher hospital mortality, suggesting that fluctuations in serum phosphate levels were also an essential factor in prognosis. In addition, patients with serum phosphate levels between 2 and 3 mg/dL had the lowest mortality, indicating that people with mild hypophosphatemia might have a better outcome (²³). In our study, we also found relatively higher mortality in the mixed group, which included more patients with fluctuating phosphate. In several studies investigating the relationship between phosphate disturbances and the prognosis of patients with sepsis, researchers have confirmed that hyperphosphatemia is an independent risk factor for increased mortality. However, they failed to demonstrate that the relationship between hypophosphatemia and clinical outcomes was statistically significant after adjusting for other confounders. Interestingly, the patients with hypophosphatemia in these studies had lower mortality rates than those with normal serum phosphate levels (^20,25,29). For instance, in a retrospective study by Li et al. (²⁵), the authors found that increased serum phosphate levels within the normal range were also significantly associated with a higher mortality risk in patients with sepsis. After adjusting for several confounders, patients with severe hypophosphatemia (<1.5 mg/dL in this study) had higher in-hospital mortality, but this was not statistically significant in the overall septic population (²⁵). It can be inferred that mild hypophosphatemia may have some potentially beneficial effects in patients with sepsis. In summary, many studies have reported that elevated serum phosphate levels are associated with increased mortality. Conversely, whether decreased serum phosphate levels are also associated with reduced mortality has not been confirmed. Although some findings suggest that patients with hypophosphatemia have lower mortality, no study has verified that this relationship is statistically significant after adjusting for confounders.

In our study, we grouped patients with fluctuating serum phosphate levels separately from those with stable phosphate levels to attenuate the effect of fluctuating phosphate levels on the relationship between serum phosphate levels and patient prognosis. The results showed that hyperphosphatemia is an independent risk factor for increased mortality in patients with sepsis, which is in line with many previous studies. However, in contrast to previous findings, we confirmed the protective effect of hypophosphatemia on the prognosis of patients with sepsis. In addition, our study found that the timing of serum phosphate assessment was also an essential factor and that initial hypophosphatemia after ICU admission was not independently associated with prognosis. Mild hypophosphatemia on the second day was independently associated with a better prognosis, possibly due to more significant changes in body homeostasis in early sepsis patients, resulting in initial serum phosphate instability. The number of people in each group was more evenly distributed in the serum phosphate group on the second day than on the first day. Our study suggests that optimal serum phosphate levels and the timing of evaluation may need to be reconsidered in patients with sepsis. We also noted that the number of patients with hypophosphatemia was unevenly distributed among the groups in our study, which might reduce the statistical significance of the association between some hypophosphatemia levels and mortality, and more studies are needed to investigate the effect of different periods and levels of hypophosphatemia on the outcome of critically ill patients.

The mechanism by which elevated serum phosphate levels lead to poor prognosis is unclear. Possible explanations include low muscle strength, phosphate-induced cytotoxicity, vascular calcification, cardiovascular disease, etc. (^30–34) Increased serum phosphate levels can increase the risk of cardiovascular disease and microvascular dysfunction in patients through various mechanisms, such as inflammation, oxidative stress, and vascular calcification. Impairment of cardiovascular function is an essential pathological process in sepsis. These findings suggest that patients with sepsis may be more sensitive to the risk of increased serum phosphate. In our study, patients with hypophosphatemia required a lower vasoactive agent infusion rate and had a lower proportion of cardiovascular comorbidities. This finding implies that the beneficial effects of lower serum phosphate levels may be associated with reduced cardiovascular dysfunction. It remains unclear whether lower serum phosphate levels directly reduce mortality. Further studies are needed to explore the underlying mechanisms and assess the potential prognostic impact of the trend and magnitude of serum phosphate fluctuations.

There are some limitations to our investigation. First, this was a single-center retrospective cohort study; therefore, the results should be interpreted cautiously when generalizing our findings to other populations and areas. Second, the results may be affected by the number of phosphate measurements that vary between patients, which may not accurately reflect the accuracy of serum phosphate fluctuations. Third, because this was a retrospective study, there was a possibility of selection bias.

In conclusion, mild hypophosphatemia was associated with a better short-term prognosis in patients with sepsis, and hyperphosphatemia was an independent risk factor for prognosis. It may be necessary to re-evaluate the definition of normal reference ranges for serum phosphate levels in patients with sepsis. The reduced impact of phosphate fluctuations and serum phosphate levels on the second day after ICU admission had a better correlation with prognosis, which prompted us to reconsider the timing of serum phosphate monitoring and treatment of phosphate abnormalities.