Association between blood phosphorus level and clinical outcomes in critically ill patients: A systematic review and meta‐analysis

Abstract

Introduction

Phosphorus is essential for biological functions and disturbances in its levels can cause phosphorus imbalance, which has ben linked to important clinical outcomes. The study aims to investigate the relationship between blood phosphorus levels and key outcomes, including mortality, intensive care unit stay, and mechanical ventilation duration.

Method

This systematic review and meta‐analysis was conducted in accordance with Preferred Reporting Items for Systematic reviews and Meta‐Analyses guidelines. A comprehensive search was performed in PubMed, Web of Science, Scopus, and Embase up to October 2024. Studies involving adult intensive care unit patients with reported blood phosphorus levels and associated clinical outcomes were included. Risk of bias was assessed using the ROBINS‐I tool.

Result

A total of 43 observational studies were included. Hypophosphatemia was not significantly linked to intensive care unit or hospital mortality but was associated with prolonged intensive care unit stay (+1.53 days), and mechanical ventilation duration (+1.54 days). Hyperphosphatemia was significantly associated with increased intensive care unit mortality (RR: 1.21, 95% CI: 1.13, 1.29; P < 0.001; n = 13) and hospital mortality (RR: 2.10, 95% CI: 1.49, 2.95; P < 0.001; n = 8), longer intensive care unit stay (+0.80 days), and extended mechanical ventilation (+4.23 days). Potential sources of heterogeneity included the use of renal replacement therapy among patients in the intensive care unit length of stay analysis and the timing of phosphorus measurement in the intensive care unit mortality analysis in hypophosphatemia. Also, patients receiving renal replacement therapy emerged as a potential source of heterogeneity in the analysis of hospital mortality in hyperphosphatemia. The evidence quality was low to very low, with high heterogeneity across studies.

Conclusion

Pooled heterogeneous data from available low‐quality evidence suggests an association of phosphate abnormalities with longer duration of mechanical ventilation, increased intensive care unit and hospital stays, and hyperphosphatemia with mortality. These findings could underline the need for regular phosphorus monitoring and targeted interventions in the intensive care unit.

INTRODUCTION

Phosphorus is a vital mineral for physiological processes like energy metabolism, ¹ cellular signaling, and acid‐base balance. ² In critical illness, metabolic stress and organ dysfunction often disturb blood phosphorus levels, resulting in hypophosphatemia or hyperphosphatemia. ³ Intensive Care Unit (ICU) patients—especially those with sepsis, trauma, or multiorgan failure—are vulnerable to such imbalances due to malnutrition, renal impairment, and aggressive medical treatments. ⁴ These imbalances are increasingly recognized as clinically significant, with growing evidence linking them to patient prognosis and outcomes in critical care settings. ⁵

Hypophosphatemia (serum phosphorus <2.5 mg/dL or <0.8 mmol/L) is common in ICU patients, ⁶ , ⁷ with prevalence ranging from 10.8% ⁴ to 30.9%, ⁸ depending on timing and population. It is linked to worse outcomes, such as respiratory muscle weakness, prolonged mechanical ventilation, and increased mortality. ⁹ , ¹⁰ A systematic review showed it also leads to longer ICU and hospital stays. ⁹ Common causes of hypophosphatemia include intracellular shifts, malnutrition, and renal losses. ⁶ It may also result from refeeding syndrome or diabetic ketoacidosis, where insulin drives phosphorus into cells.

Hyperphosphatemia (serum phosphorus >4.5 mg/dL or >1.45 mmol/L) ¹¹ is also prevalent in critical care settings, often reflecting underlying renal impairment or tissue injury. ³ Its prevalence can reach up to 45% ³ and is associated with increased mortality and heightened requirements for continuous renal replacement therapy, ⁵ indicating it may be both a marker of disease severity and a contributor to adverse clinical trajectories. Hyperphosphatemia results from renal dysfunction, excessive phosphate intake, or tissue injury. ³ , ¹¹ Despite these insights, the literature presents inconsistencies. Suzuki et al. reported no independent association between hypophosphatemia and mortality in a cohort of 2730 critically ill patients, ¹² suggesting it may merely reflect illness severity rather than directly influence outcomes. Similarly, hypophosphatemia was linked to a longer hospital stay, but it was not a reliable indicator of mortality in the ICU population, according to Haider et al. ¹³ and Suzuki et al. ¹² These conflicting results may stem from methodological differences, including retrospective designs, heterogeneous patient populations, and unadjusted confounders, such as nutritional status or therapeutic interventions. Furthermore, the dynamic nature of phosphorus levels during critical illness—often fluctuating due to treatment or disease progression—complicates efforts to establish a clear causal link with outcomes. ²

Beyond prevalence and prognosis, understanding whether phosphorus levels act as modifiable risk factors or passive biomarkers is crucial. This could inform ICU management strategies, like phosphorus repletion for hypophosphatemia or phosphate‐lowering interventions for hyperphosphatemia. To date, systematic reviews have lacked precise and comprehensive synthesis of evidence on hypophosphatemia and hyperphosphatemia in critically ill patients, with incomplete data on prevalence, prognostic associations, and clinical outcomes across various settings. This systematic review and meta‐analysis aims to address this gap by evaluating the association between blood phosphorus levels—analyzed both continuously and categorically—and key clinical endpoints, including mortality, ICU length of stay, duration of mechanical ventilation, and complication rates.

METHODS

Study design and registration

This systematic review and meta‐analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) guidelines to ensure transparency and reproducibility. ¹⁴ The protocol was prospectively registered with the International Prospective Register of Systematic Reviews (PROSPERO) under registration number [CRD42025644415] prior to data extraction, with any amendments to the protocol documented and justified in the final report.

Eligibility criteria

The population consisted of adult patients (aged ≥18 years) admitted to an ICU with critical illness, including conditions such as sepsis, septic shock, acute kidney injury, trauma, burns, or postoperative states (e.g., cardiac or gastrointestinal surgery).

The exposure involved the measurement of blood phosphorus levels (serum or plasma phosphate) at ICU admission or during the ICU stay, according to the primary papers categorization or reported as either hypophosphatemia (<2.5 mg/dL or <0.8 mmol/L) or hyperphosphatemia (>4.5 mg/dL or >1.45 mmol/L). However, the definition of hypophosphatemia was not uniform across all included studies. While several studies used a threshold of <0.8 mmol/L, some studies applied lower study‑specific cut‑off values (e.g., <0.6, <0.65, or <0.48 mmol/L). In the present meta‑analysis, hypophosphatemia and hyperphosphatemia were analyzed according to the definitions reported in the original studies. Whenever available, patients with hypophosphatemia or hyperphosphatemia were compared with patients with normal serum phosphate levels (normophosphatemia) as the reference group, based on the comparisons reported in the original studies. The outcomes of interest included at least one of the following; Primary outcomes include: (a) ICU mortality, (b) hospital mortality, and secondary outcomes include: (c) ICU length of stay, (d) hospital length of stay, (e) duration of mechanical ventilation.

Search strategy

A comprehensive literature search was conducted across PubMed, Web of Science, Scopus, and Embase from inception to October 27, 2024. The search strategy combined Medical Subject Headings (MeSH) and free‐text terms related to phosphorus levels (e.g., “phosphate,” “hypophosphatemia,” “hyperphosphatemia,” “serum phosphorus”), and outcomes (e.g., “mortality,”). A sample PubMed search included: (“phosphate*”[All Fields] OR “hypophosphatemia”[All Fields] OR “hyperphosphatemia”[All Fields] OR “phosphorus”[MeSH Terms]) AND (“mortality”[MeSH Terms] OR “hospital mortality”[MeSH Terms]). The full search strategies for each database are provided in Table S1. Reference lists of included studies and relevant reviews were manually searched.

Study selection

Observational studies with controls group conducted on human participants, published in English, and with accessible full texts were included. Animal studies, reviews, editorials, conference abstracts, and case reports were excluded.

Data extraction

Data extraction was performed independently by two reviewers using a standardized, piloted form. The extracted information included several key study characteristics, such as the author, year, country, design (cohort or cross‐sectional), sample size, and the specific ICU population (e.g., sepsis, trauma). Patient characteristics, including age range (18–75 years), sex distribution, primary diagnosis, and follow‐up duration, were also documented. For phosphorus measurements, the timing (either at admission or during ICU stay), the definition of hypo‐ and hyperphosphatemia, and the assay methods used were recorded. Outcomes of interest included ICU and hospital mortality (expressed as risk ratios [RRs]), ICU and hospital length of stay (in days), mechanical ventilation duration (in days) with corresponding effect estimates (e.g., RRs, weighted mean differences [WMDs]), and 95% confidence intervals (CIs). Additionally, the analysis accounted for potential confounders, including age, sex, and illness severity. In the event of disagreements, a third reviewer acted as an adjudicator. Discrepancies were resolved through discussion and consensus. Additionally, to address missing data, corresponding authors were contacted and given a 2‐week period to respond.

Risk of bias assessment

The Risk of Bias in Non‐randomized Studies of Interventions (ROBINS‐I) tool was applied to assess study quality across domains: confounding bias, participant selection, exposure assessment, misclassification during follow‐up, missing data, outcome measurement, and selective reporting. Studies were rated as having moderate or serious overall bias. Two reviewers independently evaluated each study, with disagreements resolved through discussion.

Statistical analysis

Meta‐analyses were conducted using a random‐effects model (DerSimonian‐Laird method) in STATA (version 17.0). For dichotomous outcomes (ICU and hospital mortality), pooled RRs with 95% CIs were calculated. For continuous outcomes (ICU length of stay, hospital length of stay, mechanical ventilation duration), WMDs with 95% CIs were computed. Analyses were stratified by hypophosphatemia (n = 41 studies) and hyperphosphatemia (n = 20 studies), reflecting the reported study counts. For each study, we extracted the effect estimates reported by the authors, including odds ratios (ORs) or hazard ratios (HRs) with their corresponding 95% CIs. Effect estimates reported by the original studies, including ORs with their 95% CIs, were extracted. To harmonize effect measures across studies, ORs were either treated as equivalent to RRs or converted to RRs depending on the outcome incidence. When the incidence of the outcome was <10% or when the OR ranged between 0.5 and 2.5, the OR was considered an acceptable approximation of the RR. If the OR was <0.5 or >2.5 and the incidence in the non‐exposed group exceeded 10%, ORs were converted to RRs using the Zhang and Kai correction method. ¹⁵ All effect measures were transformed to their natural logarithms and pooled using random‐effects models. The final summary estimate for each outcome was expressed as a pooled RR with its 95% CI. ¹⁶

Heterogeneity was assessed using the I ² statistic (<25%: low, 25%–50%: moderate, >50%: high) and Cochran's Q test (p < 0.001 indicating significance). Subgroup analyses explored heterogeneity by study design (cohort vs. cross‐sectional), patients receiving renal replacement therapy, timing of phosphorus measurement, adjustment for cofounders, type of comparison group, and number of patients, with findings reported in Tables 2 and 3. Sensitivity analyses tested robustness by excluding influential studies identified via influence diagnostics. Publication bias was evaluated using Egger's and Begg's tests, with trim‐and‐fill analysis applied where significant (p < 0.05).

Table 2.

The result of the subgroup analysis of hypophosphatemia and clinical outcomes in critically ill patients.

Sub‐grouped by	No. of trials	Effect sizea	95% CI, P value	I 2 (%)	P for heterogeneity	P for between subgroup heterogeneity
ICU mortality
Type of Study						0.401
Cohort	27	1.00	(0.90, 1.12), 0.966	66.3	<0.001
Cross‐sectional	3	1.44	(0.62, 3.32), 0.394	92.6	<0.001
Patients receiving renal replacement therapy						0.650
Yes	7	0.98	(0.70, 1.36), 0.887	76.0	<0.001
No	23	1.06	(0.91, 1.24), 0.440	82.1	<0.001
Timing of phosphorus measurement						0.050
ICU admission	14	1.13	(0.91, 1.39), 0.262	85.1	<0.001
During ICU stay	14	0.91	(0.78, 1.05), 0.202	58.0	0.003
Not report	2	1.22	(0.99, 1.50), 0.532	0.0	0.776
Adjustment for confounders						0.795
Yes	16	1.02	(0.90, 1.17), 0.730	73.6	<0.001
No	14	1.07	(0.79, 1.45), 0.663	84.2	<0.001
Number of patients						0.591
<1000	19	1.08	(0.86, 1.37), 0.669	81.7	<0.001
≥1000	11	1.00	(0.87, 1.16), 0.066	75.6	<0.001
Hospital mortality
Type of Study						0.867
Cohort	16	1.12	(0.90, 1.39), 0.305	66.0	<0.001
Cross‐sectional	1	1.24	(0.66, 2.31), 0.499	–	–
Patients receiving renal replacement therapy						0.529
Yes	2	1.41	(0.66, 3.02), 0.337	65.7	0.088
No	15	1.09	(0.88, 1.36), 0.423	65.0	<0.001
Timing of phosphorus measurement						0.507
ICU admission	8	1.03	(0.85, 1.26), 0.758	11.4	0.341
During ICU stay	9	1.18	(0.84, 1.64), 0.341	78.0	<0.001
Adjustment for confounders						0.707
Yes	12	1.11	(0.86, 1.44), 0.425	70.4	<0.001
No	4	1.03	(0.79, 1.32), 0.803	0.0	0.989
Number of patients						0.287
<1000	10	1.29	(0.84, 1.97), 0.240	75.4	<0.001
≥1000	7	1.01	(0.86, 1.19), 0.930	25.1	0.237
ICU LOS
Type of Study						0.192
Cohort	20	1.18	(0.65, 1.70), <0.001	96.3	<0.001
Cross‐sectional	2	3.94	(−0.18, 8.05), 0.061	96.0	<0.001
Patients receiving renal replacement therapy						0.014
Yes	2	5.74	(2.34, 9.15), 0.001	0.0	0.705
No	21	1.43	(0.88, 1.99), <0.001	97.1	<0.001
Timing of phosphorus measurement						0.180
ICU admission	11	1.23	(0.55, 1.90), <0.001	97.0	<0.001
During ICU stay	9	1.98	(1.26, 2.69), <0.001	87.4	<0.001
Not report	2	−0.38	(−3.87, 3.12), 0.833	0.0	0.800
Adjustment for confounders						0.797
Yes	11	1.44	(0.88, 2.01), <0.001	94.4	<.001
No	11	1.65	(0.21, 3.08), 0.024	94.9	<0.001
Number of patients						0.148
<1000	15	2.14	(0.69, 3.58), 0.004	92.4	<0.001
≥1000	7	0.96	(0.28, 1.64), 0.006	98.6	<0.001
Hospital LOS
Type of Study						0.001
Cohort	18	0.62	(−0.65, 1.90), 0.337	89.7	<0.001
Cross‐sectional	1	8.90	(7.20, 10.60), <0.001	–	–
Patients receiving renal replacement therapy						0.992
Yes	2	1.38	(−13.61, 16.36), 0.857	86.1	0.007
No	17	1.30	(−0.17, 2.77), 0.083	93.5	<0.001
Timing of phosphorus measurement						0.444
ICU admission	9	1.64	(−0.91, 4.19), 0.208	95.7	<0.001
During ICU stay	8	0.88	(−0.35, 2.10), 0.160	65.2	0.005
Not report	1	−2.70	(−8.89, 3.49), 0.392	–	–
Adjustment for confounders						0.357
Yes	10	0.62	(−0.88, 2.12), 0.417	91.6	<0.001
No	8	2.66	(−1.41, 6.73), 0.200	90.5	<0.001
Number of patients						0.131
<1000	18	1.33	(−0.29, 2.96), 0.108	93.0	<0.001
≥1000	1	−0.04	(−0.77, 0.69), 0.914	0.0	<0.001
MV
Type of Study						0.086
Cohort	17	1.32	(0.67, 1.97), <0.001	96.1	<0.001
Cross‐sectional	2	2.77	(1.24, 4.30), <0.001	76.1	0.041
Patients receiving renal replacement therapy						0.330
Yes	2	4.14	(−1.44, 9.73), 0.146	87.6	0.005
No	16	1.34	(0.63, 2.06), <0.001	97.5	<0.001
Timing of phosphorus measurement						0.376
ICU admission	8	1.07	(−0.13, 2.27), 0.081	98.5	<0.001
During ICU stay	8	2.04	(1.12, 2.96), <0.001	93.8	<0.001
Not report	2	0.88	(−1.54, 3.29), 0.477	0.0	0.987
Adjustment for confounders						0.823
Yes	8	1.49	(0.59, 2.38), <0.001	96.2	<0.001
No	10	1.66	(0.46, 2.86), 0.007	97.5	<0.001
Number of patients						0.044
<1000	13	2.03	(0.99, 3.06), <0.001	94.8	<0.001
≥1000	5	0.67	(−0.15, 1.49), 0.107	97.7	<0.001

^a Calculated by Random‐effects model.

Abbreviation: CI, confidence interval.

Table 3.

The result of the subgroup analysis of hyperphosphatemia and clinical outcomes in critically ill patients.

Sub‐grouped by	No. of trials	Effect sizea	95% CI, P value	I 2 (%)	P for heterogeneity	P for between subgroup heterogeneity
ICU mortality
Type of Study						–
Cohort	13	1.21	(1.13, 1.29), <0.001	80.5	<0.001
Cross‐sectional	0	–	–	–	–
Patients receiving renal replacement therapy						0.749
Yes	3	1.21	(0.98, 1.50), 0.078	73.8	0.022
No	10	1.26	(1.14, 1.39), <0.001	82.6	<0.001
Timing of phosphorus measurement						0.442
ICU admission	9	1.23	(1.11, 1.36), <0.001	76.4	<0.001
During ICU stay	3	1.20	(0.98, 1.47), 0.080	89.8	<0.001
Not report	1	1.40	(1.16, 1.69), <0.001	0.0	<0.001
Adjustment for confounders						0.104
Yes	12	1.23	(1.14, 1.32), <0.001	81.8	<0.001
No	1	1.12	(1.03, 1.21), 0.006	0.0	<0.001
Number of patients						0.072
<1000	2	1.46	(1.17, 1.82), 0.001	16.4	0.274
≥1000	11	1.18	(1.11, 1.26), <0.001	80.5	<0.001
Hospital mortality
Type of Study						0.549
Cohort	7	1.96	(1.38, 2.78), <0.001	78.8	<0.001
Cross‐sectional	1	3.29	(1.79, 2.78), <0.001	–	–
Patients receiving renal replacement therapy						0.001
Yes	1	1.13	(0.87, 1.47), 0.361	0.0	<0.001
No	7	2.37	(1.64, 3.44), <0.001	74.7	<0.001
Timing of phosphorus measurement						0.318
ICU admission	6	1.95	(1.32, 2.88), 0.001	81.4	<0.001
During ICU stay	2	2.61	(1.71, 3.94), <0.001	2.6	0.311
Adjustment for confounders						–
Yes	8	2.10	(1.49, 2.95), <0.001	79.2	<0.001
No	–	–	–	–	–
Number of patients						0.387
<1000	3	2.98	(1.16, 7.67), 0.024	73.5	0.023
≥1000	5	1.90	(1.32, 2.75), 0.001	82.5	<0.001
ICU LOS
Type of Study						–
Cohort	10	0.80	(0.57, 1.04), 0.001	89.2	<0.001
Cross‐sectional	–	–	–	–	–
Patients receiving renal replacement therapy						0.972
Yes	1	0.80	(0.57, 1.03), <0.001	0.0	<0.001
No	9	0.79	(0.53, 1.06), <0.001	90.3	<0.001
Timing of phosphorus measurement						0.439
ICU admission	7	0.84	(0.60, 1.08), <0.001	85.3	<0.001
During ICU stay	3	0.48	(−0.41, 1.37), 0.290	93.0	<0.001
Adjustment for confounders						–
Yes	10	0.80	(0.57, 1.04), 0.001	89.2	<0.001
No	–	–	–	–	–
Number of patients						0.331
<1000	2	−0.99	(−4.71, 2.72), 0.600	79.9	0.026
≥1000	8	0.85	(0.63, 1.08), <0.001	90.1	<0.001
Hospital LOS
Type of Study						0.003
Cohort	7	0.25	(−1.14, 1.64), 0.726	88.8	<0.001
Cross‐sectional	1	3.00	(1.71, 4.29), <0.001	–	–
Patients receiving renal replacement therapy						0.781
Yes	1	0.80	(0.25, 1.35), 0.004	0.0	<0.001
No	7	0.54	(−1.19, 2.27), 0.540	89.9	<0.001
Timing of phosphorus measurement						0.971
ICU admission	4	0.54	(−2.24, 3.32), 0.705	90.7	<0.001
During ICU stay	3	0.55	(−2.25, 3.36), 0.699	90.9	<0.001
Not report	1	0.80	(0.25, 1.35), 0.004	0.0	<0.001
Adjustment for confounders						–
Yes	8	0.65	(−0.64, 1.94), 0.323	88.8	<0.001
No
Number of patients						0.932
<1000	2	1.57	(−12.63, 15.78),0.828	95.0	<0.01
≥1000	6	0.95	(−0.17, 2.07), 0.096	86.8	<0.001

^a Calculated by Random‐effects model.

Abbreviation: CI, confidence interval.

Grading of evidence

The certainty of evidence for each outcome was evaluated using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework, considering risk of bias, inconsistency (I ² > 60% for all outcomes), indirectness, imprecision, and publication bias.

RESULTS

Literature search

The systematic review and meta‐analysis findings are summarized in Figure 1, which outlines the study selection process. Initially, 13,904 articles were identified through database searches. Following the removal of 843 duplicates and the exclusion of 259 ineligible studies (98 animal studies and 161 non‐original articles), 12,802 records underwent title and abstract screening. Two independent reviewers screened these records, excluding 12,709 irrelevant articles. Next, 93 full‐text articles were assessed for eligibility, with 50 excluded due to irrelevant exposures or outcomes, insufficient data, lack of control groups, or study populations of children. Ultimately, 43 observational studies were included in the quantitative analysis, ⁴ , ⁵ , ⁸ , ¹² , ¹³ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² , ³³ , ³⁴ , ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵² , ⁵³ , ⁵⁴ with 38 addressing hypophosphatemia ⁴ , ⁵ , ⁸ , ¹² , ¹³ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³³ , ³⁴ , ³⁵ , ³⁶ , ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵² , ⁵³ , ⁵⁴ and 20 focusing on hyperphosphatemia. ⁴ , ⁵ , ⁸ , ¹³ , ¹⁷ , ²¹ , ²⁶ , ²⁷ , ²⁹ , ³² , ³⁶ , ³⁷ , ⁴¹ , ⁴³ , ⁴⁵ , ⁴⁷ , ⁴⁸ , ⁵¹ , ⁵⁴ Of these 43 observational studies, 3 were cross‐sectional studies ¹³ , ¹⁸ , ⁵⁰ and the remaining were cohort studies. The process is detailed in a PRISMA flow diagram (Figure 1).

Figure 1.

Literature search and review flow diagram for selection of studies.

Study characteristics

The observational studies included in Tables S3 and S4 evaluated the impact of hypophosphatemia and hyperphosphatemia in critically ill patients. Clinical outcomes assessed across the studies included ICU and hospital mortality, mechanical ventilation, length of ICU, and hospital stay. The study populations focused on critically ill patients, including those with sepsis, septic shock, acute kidney injury, trauma, burns, gastrointestinal surgery, cardiac surgery, and patients requiring mechanical ventilation or continuous renal replacement therapy. The studies were conducted between 1998 and 2024, with patient ages ranging from 18 to 75 years. Follow‐up durations varied across studies, spanning from a few days to several years. All studies included both male and female participants. These comprehensive analyses provide valuable insights into the association between phosphate abnormalities and clinical outcomes in critically ill patients. The timing of phosphate measurement varied across the included studies. Specifically, phosphate levels were assessed at ICU admission in 22 studies, whereas 18 studies evaluated hypophosphatemia or hyperphosphatemia occurring during the ICU stay. In three studies, the timing of phosphate measurement was not clearly reported.

Risk of bias assessment

The ROBINS tool classified 18 studies as moderately biased, ⁵ , ⁸ , ¹⁹ , ²³ , ²⁷ , ³² , ³⁴ , ³⁵ , ³⁶ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁸ , ⁵¹ , ⁵³ , ⁵⁴ while other studies had a serious bias ⁴ , ¹² , ¹³ , ¹⁷ , ¹⁸ , ²⁰ , ²¹ , ²² , ²⁴ , ²⁵ , ²⁶ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³³ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ , ⁴⁶ , ⁴⁷ , ⁴⁹ , ⁵⁰ , ⁵² due to residual confounding or insufficient information regarding participant selection (Table S2).

Findings from the meta‐analysis

Association between hypophosphatemia and clinical outcomes

The meta‐analysis revealed no significant association between hypophosphatemia and risk of ICU mortality (RR: 1.04, 95% CI: 0.91, 1.19; P = 0.553; I ² = 80.4%, P _{heterogeneity} = <0.001; n = 30; Figure S1) OR risk of hospital mortality (RR: 1.12, 95% CI: 0.91, 1.38; P = 0.272; I ² = 64.2%, P _{heterogeneity} = <0.001; n = 17; Figure S2). In patients with hypophosphatemia, ICU length of stay was 1.53 days longer compared with those with normal phosphorus levels (WMD: 1.53, 95% CI: 0.98, 2.08; P < 0.001; I ² = 96.8%, P _{heterogeneity} = <0.001; n = 22). Similarly, the duration of mechanical ventilation was 1.54 days longer in hypophosphatemic patients (WMD: 1.54, 95% CI: 0.84, 2.23; P < 0.001; I ² = 97.2%, P _{heterogeneity} = <0.001; n = 18). Hospital length of stay showed a trend toward being longer by 1.19 days, although this was not statistically significant (WMD: 1.24, 95% CI: −0.20, 2.68; P = 0.091; I ² = 92.9%, P _{heterogeneity} = <0.001; n = 18). (Table 1 ). The certainty of the evidence was rated as very low according to the GRADE framework (Table S5).

Table 1.

Association between blood phosphorus level and clinical outcomes in critically ill patients.

	Pairwise meta‐analysis
	Studies, n	Effect size (95% CI)	P value	I 2, %	P heterogeneity	GRADE
Hypophosphatemia
ICU mortality	30	RR: 1.04 (0.91, 1.19)	0.533	80.4	<0.001	⨁◯◯◯ Very low
Hospital mortality	17	RR: 1.12 (0.91, 1.38)	0.272	64.2	<0.001	⨁◯◯◯ Very low
ICU LOS (days)	22	WMD: 1.53 (0.98, 2.08)	<0.001	96.8	<0.001	⨁◯◯◯ Very low
Hospital LOS (days)	18	WMD: 1.24 (−0.20, 2.68)	0.091	92.9	<0.001	⨁◯◯◯ Very low
MV (days)	18	WMD: 1.54 (0.84, 2.23)	<0.001	97.2	<0.001	⨁◯◯◯ Very low
Hyperphosphatemia
ICU mortality	13	RR: 1.21 (1.13, 1.29)	<0.001	80.5	<0.001	⨁◯◯◯ Very low
Hospital mortality	8	RR: 2.10 (1.49, 2.95)	<0.001	79.2	<0.001	⨁◯◯◯ Very low
ICU LOS (days)	10	WMD: 0.80 (0.57, 1.04)	<0.001	89.2	<0.001	⨁⨁◯◯ Low
Hospital LOS (days)	8	WMD: 0.65 (−0.64, 1.94)	0.323	88.8	<0.001	⨁◯◯◯ Very low
MV (days)	4	WMD: 4.23 (1.63, 6.82)	0.001	98.3	<0.001	⨁◯◯◯ Very low

Abbreviations: CI, confidence interval; GRADE, Grading of Recommendations Assessment, Development, and Evaluation; LOS, length of stay; MD, mean difference; MV, mechanical ventilation; RR; relative risk.

Moreover, subgroup analyses for hypophosphatemia are summarized in Table 2. Overall, the associations between hypophosphatemia and ICU or hospital mortality were consistent across study design, adjustment for confounders, renal replacement therapy status, timing of phosphorus measurement, type of comparison group, and study sample size, with no significant between‐subgroup heterogeneity observed. In contrast, for non‐mortality outcomes, hypophosphatemia was associated with longer ICU length of stay and increased duration of mechanical ventilation, particularly in studies measuring phosphorus levels during ICU stay and in smaller studies (<1000 patients). Although substantial heterogeneity remained across most subgroups, the direction of effect was generally consistent, suggesting robustness of the main findings. Patients receiving renal replacement therapy and the timing of phosphorus measurement could explain the observed heterogeneity for ICU length of stay and ICU mortality, respectively.

Potential sources of heterogeneity included the use of renal replacement therapy among patients in the ICU length of stay analysis and the timing of phosphorus measurement in the ICU mortality analysis.

The subgroup analysis based on the timing of phosphorus measurement did not show a statistically significant difference between studies measuring phosphate levels at ICU admission and those assessing hypophosphatemia during the ICU stay across most outcomes (Table 2). Although the effect estimates varied slightly between these categories, the test for between‑subgroup heterogeneity was not significant for ICU mortality, hospital mortality, length of stay, and mechanical ventilation, suggesting that the timing of phosphate measurement had limited influence on the pooled results.

A sensitivity analysis was carried out for the hospital length of stay and revealed that the pooled effect size was affected when Olthof et al. ³⁸ , Miller et al. ³⁶ , and Padelli et al. ³⁹ were excluded. Moreover, it was found that the effect size for mechanical ventilation changed when the studies by Alharbi et al. ¹⁷ and Miller et al. ³⁶ were omitted. Furthermore, by excluding Alharbi et al. ¹⁷ , the overall effect was altered. Despite this, other clinical outcomes were not affected by one study.

In addition, hypophosphatemia and hospital mortality risk were found to have statistically significant publication bias (P = 0.012, Egger's test; P = 0.036, Begg's test). We employed the trim‐and‐fill method to find sources of bias, and the results were comparable. For the other clinical outcomes, there was no indication of publication bias.

Association between hyperphosphatemia and clinical outcomes

The meta‐analysis found significant associations between hyperphosphatemia and increased risk of mortality, both in the ICU (RR: 1.21, 95% CI: 1.13, 1.29; P < 0.001; I ² = 80.5%, P _{heterogeneity} < 0.001; n = 13; Figure S3) and in the hospital (RR: 2.10, 95% CI: 1.49, 2.95; P < 0.001; I ² = 79.2%, P _{heterogeneity} < 0.001; n = 8; Figure S4). Additionally, hyperphosphatemic patients had longer ICU stays by 0.80 days (WMD: 0.80, 95% CI: 0.57, 1.04; P < 0.001; I ² = 89.2%, P _{heterogeneity} < 0.001; n = 10). The duration of mechanical ventilation was also extended by 4.23 days (WMD: 4.23, 95% CI: 1.63, 6.82; P = 0.001; I ² = 98.3%, P _{heterogeneity} < 0.001; n = 4). However, no significant difference was found in hospital length of stay (WMD: 0.65, 95% CI: −0.64, 1.94; P = 0.323; I ² = 88.8%, P _{heterogeneity} < 0.001; n = 8) (Table 1 ). The overall certainty of the evidence was rated as very low or low for all outcomes using the GRADE framework (Table S6). Due to the high heterogeneity, we conducted subgroup analyses based on study types and age groups. The details of these subgroups are provided in Table 3.

Subgroup analyses (Table 3) demonstrated that hyperphosphatemia was consistently associated with increased ICU and hospital mortality across most subgroups, including study design, renal replacement therapy status, timing of phosphorus measurement, adjustment for confounders, study sample size, and type of comparison group. In contrast, no significant associations were observed for ICU or hospital length of stay. Although substantial heterogeneity persisted in several subgroups, the direction and magnitude of the mortality associations remained largely consistent. Patients receiving renal replacement therapy emerged as a potential source of heterogeneity in the analysis of hospital mortality.

The subgroup analysis according to the timing of phosphorus measurement showed generally consistent results across most outcomes (Table 3). Hyperphosphatemia measured at ICU admission and during the ICU stay was associated with a comparable increase in ICU and hospital mortality, and the test for between‑subgroup heterogeneity was not statistically significant. Similarly, no significant differences were observed between these categories for ICU and hospital length of stay, indicating that the timing of phosphate assessment did not substantially modify the overall associations.

In the sensitivity analysis of ICU length of stay, Broman et al. ⁸ was excluded from the pooled effect size. Also, the exclusion of Miller et al. ³⁶ from the analysis reformed the overall effect size of the hospital's length of stay. Furthermore, the sensitivity analysis for mechanical ventilation showed that removing three studies from the analysis altered the overall effect.

Moreover, statistically significant publication bias was observed for hyperphosphatemia and ICU mortality risk (P < 0.001, Egger's test) and hospital mortality risk (P = 0.020, Egger's test). To identify potential sources of bias, we used the trim‐and‐fill method, and the results remained consistent. No evidence of publication bias was found for the other clinical outcomes.

DISCUSSION

To our knowledge, the present study is the most comprehensive systematic review and meta‐analysis of observational studies in which 43 original research articles investigating the relationship between phosphate abnormalities (hypophosphatemia in 38 studies and hyperphosphatemia in 20 studies) were included. The pooled results revealed that experiencing at least one episode of hyperphosphatemia in ICU patients was associated with increased risk of mortality (both ICU and hospital mortality) and poor clinical outcomes, including prolonged ICU length of stay and longer mechanical ventilation duration. Hypophosphatemia, although not a significant predictor of mortality, was associated with increased ICU length of stay and mechanical ventilation duration.

Previous meta‐analyses have investigated the impact of phosphate abnormalities on outcomes in critically ill patients, and their findings closely align with ours. Regarding hyperphosphatemia, the most comparable study to ours is a meta‐analysis by Zheng et al., which evaluated nine studies encompassing 47,570 patients. ³ They similarly found that hyperphosphatemia was significantly associated with increased all‐cause mortality in critically ill patients (OR = 2.85). However, some relevant studies were not included in their meta‐analysis ²⁶ , ²⁹ and several cohort studies have been published since then. ³⁷ , ⁴⁵ Other meta‐analyses focusing on specific conditions among critically ill patients, such as sepsis, ⁵⁵ or on non‐critically ill patients, including those with end‐stage renal disease ⁵⁶ and coronary heart disease, ⁵⁷ have also reported consistent results.

Hyperphosphatemia often reflects underlying renal dysfunction, which itself is an independent risk factor for mortality in critically ill patients. Thus, this raises the possibility that hyperphosphatemia may not act as an independent risk factor, but serve as a marker of illness severity. However, in critical illness, the causes and consequences of hyperphosphatemia may be more complex. Ischemic tissue injury, cardiac arrest, lactic acidosis, and diabetic ketoacidosis can lead to hyperphosphatemia in ICU patients. ⁵⁸ , ⁵⁹ , ⁶⁰ , ⁶¹ , ⁶² Also, several pathophysiological mechanisms may contribute to the increased mortality rate in hyperphosphatemia. Elevated serum phosphate levels have been shown to impair endothelial function. Studies indicate that high phosphate concentrations can reduce endothelial nitric oxide synthase activity, leading to decreased nitric oxide production and subsequent endothelial dysfunction. ⁶³ Furthermore, phosphate overload has been linked to systemic inflammation. Research demonstrates that excessive phosphate intake can directly induce inflammatory responses, contributing to vascular calcification and increased cardiovascular mortality, particularly in patients with chronic kidney disease. ⁶⁴ Hyperphosphatemia may also impair oxygen delivery. While the direct impact of hyperphosphatemia on oxygen transport is less well‐established, the resultant endothelial dysfunction and systemic inflammation can exacerbate organ dysfunction, potentially impairing oxygen utilization at the tissue level.

As mentioned above, hypophosphatemia was not significantly associated with mortality, but it was a significant predictor of poor clinical outcome, i.e., increased duration of mechanical ventilation and ICU length of stay. This suggests that, although hypophosphatemia may not directly impact survival, it could contribute to prolonged critical illness and delayed recovery. Other meta‐analyses also indicated similar results. A meta‐analysis by Sin et al. encompassing 12 studies found that hypophosphatemia was not significantly associated with increased all‐cause mortality (risk ratio: 1.13; 95% CI: 0.98–1.3) but it was related to extended hospital and ICU stays. ⁹ Another meta‐analysis by Liu et al. published in Chinese ⁶⁵ similarly indicated that critically ill patients with hypophosphatemia had increased hospital and ICU length of stay and higher mechanical ventilation duration. In this study, inconsistent with Sin et al. and our study, hypophosphatemia was identified as a significant predictor for higher mortality rate. It should be noted that the Liu et al. meta‐analysis included 8 Chinese studies and 1 English study, so the different populations examined in their study could explain the different results reported for mortality. These findings emphasize the need for careful phosphate management, not only to prevent severe depletion but also to mitigate its potential complications.

Phosphate is essential for production of adenosine triphosphate (ATP), which is the primary energy source for cellular functions. A deficiency in phosphate can lead to decreased ATP levels, and this can impair the function of the respiratory muscles, which may contribute to prolonged mechanical ventilation. ⁷ Also, hypophosphatemia can potentially result in hemodynamic instability through decreased myocardial contractility and arrhythmias. ⁷ Furthermore, hypophosphatemia can impair leukocyte function and thus weaken the body's ability to combat infections. ⁷ All these conditions can prolong ICU stays and worsen patient outcomes.

In order to identify the potential sources of substantial heterogeneity found in our results and to explore the effect of the other variables on the observed associations, we performed subgroup analyses. The results showed that hypophosphatemia was associated with an increased duration of mechanical ventilation only in patients who developed hypophosphatemia during ICU stay, not in those who had hypophosphatemia at ICU admission. One possible explanation for this finding is that hypophosphatemia developing during the ICU stay may reflect the suboptimal quality of care patients received in the ICU. Such a condition could lead to prolonged mechanical ventilation in these patients. In contrast, hypophosphatemia present at ICU admission may represent a pre‐existing or transient condition during the acute phase of critical illness that has been appropriately managed. The other results remained unchanged across the tested subgroups, and the altered results could not be considered reliable findings due to the insufficient number of studies in the corresponding categories (n = 1–3). Regarding heterogeneity, the timing of phosphorus measurement and adjustment for confounders may explain the statistical heterogeneity in the association between hypophosphatemia and hospital mortality. Similarly, for the association between hypophosphatemia and ICU mortality, heterogeneity could be explained by the timing of phosphorus measurement. The sources of heterogeneity were not identified through subgroup analyses for other outcomes.

In the present meta‐analysis, we perform a comprehensive database search to ensure all eligible studies are included. Also, we applied standard tools to evaluate the quality of the included studies and the certainty of the evidence. However, this meta‐analysis had some limitations that should be kept in mind while interpreting the findings. First, as all included studies had observational designs, drawing a definitive causal relation between phosphorus status and clinical outcomes is not possible. Also, we do not know to what extent the observed relation between phosphate disturbance and mortality and/or other clinical outcomes is explained by the underlying cause of hyper/hypophosphatemia. In addition, the cut offs used for the definition of hypophosphatemia were different across included studies, which should be considered while interpreting the findings. Furthermore, we observed substantial statistical heterogeneity in the pooled results. Although the subgroup analyses were performed to identify potential sources of heterogeneity, for most outcomes, the source of heterogeneity was not identified through subgroup analysis. Also, due to insufficient data in primary studies, we could not perform further subgroup analyses to examine the effect of other factors that may influence outcomes and/or contribute to statistical heterogeneity, such as the treatment patients received, the duration and severity of phosphate disturbance, and the serum levels of vitamin D and parathyroid hormone (PTH).

In addition, we detected evidence of publication bias using Egger's test for hospital mortality (in relation to hypophosphatemia) as well as ICU and hospital mortality (in relation to hyperphosphatemia). This bias may have resulted in overstimation of the pooled effect size for these associations and should be considered when interpreting these finding. Restricting the included studies to English language publications may also have influenced the interpretation and generalizability of the findings.

Finally, the findings of the present meta‐analysis highlight the prognostic importance of phosphate abnormalities in critically ill patients. Our study revealed that hyperphosphatemia was significantly associated with mortality, prolonged hospital and intensive care unit length of stay, and increased mechanical ventilation duration. Patients with hypophosphatemia also had longer mechanical ventilation duration, and prolonged hospital and intensive care unit length of stay. These results emphasize the potential value of routine phosphate monitoring and targeted interventions to improve patient outcomes in the intensive care unit setting.

AUTHOR CONTRIBUTIONS

Pardis Irandoost: conceptualization; investigation; methodology; validation; visualization; writing—review and editing; project administration; supervision; data curation. Mohammad Mostafa Ansari: conceptualization; investigation; writing—original draft; methodology; visualization. Mohammad Niakan Lahiji: investigation; writing—review and editing; visualization; methodology. Sepide Talebi: conceptualization; investigation; validation; visualization; writing—original draft; formal analysis. Kimia Torabinasab: conceptualization; investigation; methodology. Samira Pourmoradian: conceptualization; methodology; validation; writing—original draft; writing—review and editing. Marziyeh Ashoori: conceptualization; investigation; writing—original draft; writing—review and editing. Maryam Milanifard: writing—original draft; investigation; visualization. Omid Moradi Moghaddam: conceptualization; investigation; funding acquisition; writing—original draft; methodology; visualization; writing—review and editing; formal analysis; supervision; project administration.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

Supporting information

Supporting File 1