Introduction

Acute kidney injury (AKI) is a global clinical problem among hospitalized patients associated with high morbidity, both short- and long-term mortality, and increased healthcare expenditure [1, 2]. AKI-associated mortality has been reported to be as high as 23% or > 1.7 million deaths per year [1,2,3]. Previous studies have attempted to identify novel AKI biomarkers, effective interventions and pharmacological treatments [1, 4,5,6,7,8,9]. Unfortunately, most treatments are unsuccessful [6, 10]. Thus, risk stratification and the identification of patients who are at high risk of developing AKI in order to prevent AKI events is crucial.

Phosphate is the most abundant intracellular anion, and it plays an important role in several aspects of cellular metabolism [11, 12]. Serum phosphate derangements with hyperphosphatemia and hypophosphatemia are common clinical problems with a prevalence as high as 45% of hospitalized patients [13, 14]. Renal phosphorus excretion is the key mechanism for maintaining phosphate balance [13]. Thus, reduced glomerular filtration rate (GFR) in patients with AKI or chronic kidney disease (CKD) can result in hyperphosphatemia [15, 16]. On the other hand, acute hyperphosphatemia itself can lead to AKI via acute phosphate nephropathy [17,18,19,20]. However, it is still unclear if elevated admission serum phosphate levels in general hospitalized patients carry a higher risk of in-hospital AKI. The objective of this study was to evaluate the risk of developing AKI in all hospitalized patients across a spectrum of phosphate levels.

Materials and methods

Study population

This was a single-center retrospective study conducted at a tertiary referral hospital. All hospitalized adult patients who had admission phosphate measurement available between January and December 2013 at Mayo Clinic, Rochester, MN, USA were enrolled in the study. Exclusion criteria were: patients without serum phosphate measurement within 24 h of admission, patients with end-stage renal disease (ESRD), patients who lacked serum creatinine (SCr) measurement during hospitalization, and patients who presented with AKI at the time of admission. For patients with multiple admissions during this period, only the first hospital admission was analyzed. ESRD was identified based on the 9th International Classification of Diseases (ICD-9) code assignment (Supplementary Table 1) or an estimated (e)GFR of less than 15 ml/min/1.73 m2.

Data collection

Clinical characteristics, demographic information, and laboratory data were collected using manual and automated retrieval from the institutional electronic medical record system. The admission serum phosphate level, defined as the first serum phosphate level within 24 h of hospital admission, was collected. eGFR was derived using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation [21]. CKD was defined as a calculated eGFR < 60 ml/min/1.73 m2. The Charlson Comorbidity score [22] was computed for comorbidities at the time of admission. Principal diagnoses were grouped based on ICD-9 codes at admission (Supplementary Table 2).

Clinical outcomes

The primary outcome was AKI, based on the Kidney Disease Improving Global Outcomes (KDIGO) SCr criterion [23]. AKI was defined as an increase in SCr ≥ 0.3 mg/dl within 48 h or ≥ 1.5 times baseline within 7 days after the admission date. Baseline SCr was defined as the minimum SCr measured within 1 year before admission since this method not only can detect more AKI cases, but also provides a better predictive ability for 60-day mortality than using the most recent value of pre-admission SCr, in our institution [24]. Outpatient SCr was used since studies have shown that it represents a steady state and provides a more robust assessment of baseline renal function than inpatient SCr [24,25,26]. If outpatient SCr was not available, the Modification of Diet in Renal Disease equation [27] was used to estimate baseline SCr level, assuming a normal baseline GFR of 75 ml/min/1.73 m2, in accordance with the guidelines [23]. The accuracy of this method was also previously validated in our institution [25]. AKI was staged for severity [23] as shown in Supplementary Table 3.

Statistical analysis

All continuous variables are reported as mean ± SD. All categorical variables are reported as count with percentage. Baseline demographics and clinical characteristics were compared among admission phosphate groups, using ANOVA for continuous variables and the Chi square test for categorical variables. We categorized admission serum phosphate levels, based on its distribution into six groups (< 2.4, 2.4 to < 2.9, 2.9 to < 3.4, 3.4 to < 3.9, 3.9 to < 4.4, and ≥ 4.4 mg/dl). Serum phosphate level 2.4 to < 2.9 mg/dl (lowest incidence of AKI) was selected as the reference group for outcome comparison. We performed univariate analysis and then multivariate logistic regression analysis to evaluate the independent association between admission phosphate levels and AKI. Odds ratio (OR) with 95% confidence interval (CI) is reported. OR was adjusted for priori-defined variables. The adjusting variables were age, sex, ethnic group, baseline GFR, principal diagnosis, Charlson score, comorbidities, medications, vasopressor use and need for mechanical ventilation at hospital admission. Comorbidities were coronary artery disease (CAD), hypertension (HTN), diabetes mellitus (DM), congestive heart failure (CHF), peripheral vascular disease (PVD), and stroke. Medications were angiotensin-converting-enzyme inhibitors (ACEIs), angiotensin II receptor blockers (ARBs), nonsteroidal anti-inflammatory drugs (NSAIDs) and diuretics. A stepwise backward logistic regression analysis was used to derive the final multivariate model. The calibration of the model was assessed by the Hosmer–Lemeshow Goodness of Fit test. The discrimination of the model was assessed by C-statistic. A two-tailed p value of < 0.05 was considered statistically significant. All analyses were performed using JMP statistical software (version 10, SAS Institute, Cary, NC, USA).

Results

A total of 7529 patients with available serum phosphate measurement within 24 h were identified. After excluding 24 patients who lacked SCr measurement during hospitalization, 363 patients with ESRD and 2106 patients with AKI at presentation, 5036 patients were enrolled (Supplementary Fig. 1).

Baseline characteristics

The baseline characteristics of the 5036 patients according to the various serum phosphate levels are summarized in Table 1. The distribution of phosphate levels was as follows: < 2.4 mg/dl, 458 patients (9%); 2.4–2.9 mg/dl, 650 patients (13%); 2.9–3.4 mg/dl, 1168 patients (23%); 3.4–3.9 mg/dl, 1308 patients (26%); 3.9–4.4 mg/dl, 867 patients (17%); and ≥ 4.4 mg/dl, 585 patients (12%). Of the 5036 patients, 4,625 (92%) were Caucasian and 2744 (54%) were male. Mean age was 60 ± 18 years. Patient comorbidities included HTN (46%), DM (19%), CAD (15%), and CHF (5%). Prior to admission, 29% of the patients were taking diuretics, 32% were taking ACEIs or ARBs, and 23% were taking NSAIDs.

Table 1 Baseline clinical characteristics of the study population
Full size table

Analysis of the principle admission diagnosis showed that patients with a diagnosis of hematology/oncology presented with high admission serum phosphate levels, while patients with infectious disease and endocrine/metabolic diagnoses presented with low admission serum phosphate levels (Table 1).

Admission serum phosphate levels and risk of AKI

The incidence of AKI was 10.5, 9.5, 11.8, 10.0, 12.8, and 17.9% in patients with admission phosphate < 2.4, 2.4–2.9, 2.9–3.4, 3.4–3.9, 3.9–4.4, and ≥ 4.4 mg/dl, respectively (Fig. 1). The highest AKI incidence was observed in patients with admission phosphate ≥ 4.4 mg/dl, with the association preserved for all stages of AKI (Table 2).

Fig. 1
figure 1

In-hospital acute kidney injury between the various admission serum phosphate levels

Full size image
Table 2 Patient outcomes according to the serum phosphate level at hospital admission
Full size table

To assess whether admission serum phosphate levels contributed to AKI development, logistic regression models were built, using 2.4–2.9 mg/dl (lowest incidence of AKI) as a reference range. Both unadjusted and adjusted analyses demonstrated that an admission phosphate level > 4.4 mg/dl was associated with an increased risk of AKI with ORs of 2.07 (95% confidence interval, CI 1.49–2.92) and 1.72 [1.20–2.47], respectively (Table 3). Conversely, admission serum phosphate levels < 4.4 mg/dl were not predictive for the development of AKI during hospitalization.

Table 3 Odds ratios for the association between admission serum phosphate levels and in-hospital acute kidney injury occurrence
Full size table

A sensitivity analysis using the most recent SCr measured within 1 year before hospital admission as the baseline SCr was performed. The results of the sensitivity analysis in Supplementary Table 4 were consistent with the primary analysis. A backward stepwise logistic regression was also performed to build the best model. The results of the analysis by backward stepwise logistic regression were consistent with our planned covariate-adjusted analysis as shown in Table 4.

Table 4 Multivariate model using backward stepwise logistic regression
Full size table

Subgroup analysis stratified by CKD status

Of 5,036 patients, 699 (14%) had CKD. Subgroup analysis to assess the risk of AKI stratified by CKD status was performed (Supplementary Table 5). In multivariate analysis, the risks of AKI with an admission phosphate level > 4.4 mg/dl were 1.47 [0.72–2.99] and 1.89 [1.24–2.91] in patients with CKD and without CKD, respectively.

Discussion

This study demonstrates that an elevated admission phosphate level is associated with AKI development during hospitalization. The highest incidence of AKI was observed among patients with admission phosphate ≥ 4.4 mg/dl at all stages of AKI, while the lowest incidence of AKI was observed when admission phosphate was 2.4–2.9 mg/dl. An elevated admission phosphate > 4.4 mg/dl was associated with a 1.72-fold increased risk of AKI during hospitalization.

There are several plausible explanations for the association between increased AKI risk and elevated phosphate levels. Previous reports of phosphate administration during bowel preparation have demonstrated that elevated phosphate levels can lead to AKI due to phosphate nephropathy [28,29,30]. Elevated phosphate levels in the plasma or tubular compartments can result in deposition of calcium phosphate crystals in renal tubules causing AKI [17,18,19,20, 28,29,30]. Hyperphosphatemia is also seen in rhabdomyolysis [31], when phosphate is released from damaged muscle cells [32]. The release of myoglobin and phosphate can subsequently cause AKI via tubular injuries [31, 32]. Elevated phosphate levels can also be found in the setting of tumor lysis syndrome (TLS), a significant cause of AKI [33]. Expectedly, our study demonstrated that patients with a primary diagnosis of hematology/oncology presented with higher admission serum phosphate levels. In patients with TLS, although hyperuricemia plays an important role in AKI via pro-inflammatory pathways and AKI-related crystal-dependent pathways [34], elevated phosphate levels with calcium phosphate deposition in the renal tubules can also cause AKI [35].

Although hyperphosphatemia has been associated with poor clinical outcomes including cardiovascular diseases and mortality among patients with CKD/ESRD [36], the risk of elevated admission serum phosphate levels for in-hospital AKI is still unclear. In this study, we are the first to demonstrate that hyperphosphatemia in non-AKI patients at hospital admission is associated with subsequent AKI development during hospitalization. However, the effects of normalizing phosphate levels and the risk of AKI were not the focus of our present study. Although the beneficial effect of phosphate-binder treatment on mortality in patients with CKD is debatable [37], its use is generally recommended for prevention and treatment of CKD-mineral and bone disorder (CKD-MBD) [38]. The role of phosphate-binders to prevent AKI in hyperphosphatemic patients is still unclear. Theoretically, current phosphate binders would not reduce serum phosphate levels quickly [39,40,41] and would have little role in patients who are unable to consume much food such as hospitalized patients with AKI.

This study has several limitations. (1) It is a single-center, retrospective study. The patient population in this study is relatively homogeneous consisting predominantly of Caucasians. Thus, further studies with a more heterogeneous population are desirable to ascertain the clinical effects of admission phosphate on AKI. (2) Although subgroup analysis in the CKD population showed the relatively comparable effect size of a 1.47-increased AKI risk with an admission phosphate level > 4.4 mg/dl, it did not reach statistical significance, likely due to the limited number of CKD patients (n = 699), especially those with eGFR < 30 ml/min/1.73 m2 in our study resulting in a lack of power or the collinearity between phosphate levels and SCr in the CKD subgroup. (3) Only 58% of patients had available outpatient SCr. Although the method to estimate baseline SCr adopted in this study was suggested by the guidelines [23], the surrogates for baseline SCr are known to be inaccurate, particularly in CKD patients, and result in bi-directional misclassification of the incidence and prognosis of AKI in a hospital setting [25]. (4) Due to the retrospective design and limited data, this study did not investigate some variables that might have modulated or confounded the association between serum phosphate and AKI, such as nutritional intake, body composition, and proteinuria. A multi-center, prospective study is ultimately required to address these limitations.

In summary, this study demonstrates that elevated admission phosphate levels are associated with an increased risk for in-hospital AKI.