Introduction

Serum phosphorus plays a pivotal role in various biologic processes including energy production, membrane transport, and cellular signaling [1]. Serum phosphorus levels are tightly regulated via dietary absorption, bone formation, and renal excretion, and thus directly reflect renal dysfunction and vascular calcification [2,3]; therefore, patients with renal dysfunction and cardiovascular disease typically show a graded association between serum phosphorus and poor prognosis [4–7].

In addition to the well-known adverse effects of serum phosphorus abnormalities, an increase in morbidity after cardiovascular surgeries has also been reported [8–10]. However, in contrast to the results of percutaneous coronary intervention [11,12], previous studies have shown an inconsistent relationship between serum phosphorus abnormalities and coronary artery bypass grafting (CABG) outcomes [8,13]. In addition, to the best of our knowledge, no prior studies have used the clinical criteria definition of phosphorus abnormalities to perform a direct three-group comparison of the outcomes of CABG among normal, hypo- and hyperphosphatemia groups.

Considering that serum phosphorus is a well-known marker of vascular calcification and an assay is readily available, the effect of preoperative serum phosphorus abnormalities defined based on clinical criteria on the outcomes of CABG need to be re-evaluated. Therefore, in the present study, we evaluated whether preoperative serum phosphorus abnormalities are related to outcomes of CABG using the definition of serum phosphorus abnormalities from clinical criteria and a relatively new statistical technique, inverse probability weighting (IPW) adjustment.

Materials and methods

Results

Patient characteristics

The baseline characteristics of the three groups are summarized in Table 1. Of the enrolled 4,989 patients, 4,544 (91.1%) 238 (4.8%), and 207 (4.1%) were stratified into the normal, hypophosphatemia, and hyperphosphatemia groups, respectively (Table 1). Mean serum phosphorus levels were 3.5 (±0.5) mg/dl, 2.1 (±0.3) mg/dl and 5.1 (±0.6) mg/dl in each group. The three groups had significantly different baseline characteristics including male sex, age, diabetes, ejection fraction < 40%, chronic kidney disease, old MI, emergent operation, off-pump technique, valve-combined operation, and the use of the right gastroepiploic artery. After IPW adjustment, the balance between the three groups was improved and presented as the change in the average standardized absolute mean difference in S1 Fig.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Baseline characteristics.

https://doi.org/10.1371/journal.pone.0225720.t001

Serum phosphorus abnormalities and clinical outcomes

The overall incidence of all-cause death was 4.1% (206/4989), with a 3.7% incidence in the normal group, 9.7% in the hypophosphatemia group, and 6.3% in the hyperphosphatemia group (P < 0.6). In multivariate Cox regression analysis only hypophosphatemia was significantly associated with the incidence of all-cause death (HR 1.76; 95% CI 1.13–2.76; P = 0.01) (Table 2). However, on further analysis using IPW adjustment, the association between hypophosphatemia and all-cause death was not significant (HR 1.52; 95% CI 0.89–2.61; P = 0.12). The incidence of graft failure was significantly higher in the hypophosphatemia group after adjustment using multivariate Cox regression analysis and IPW analysis (HR 2.14; 95% CI 1.22–3.75; P = 0.01, HR 2.51; 95% CI 1.37–4.61; P = 0.003, respectively).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Clinical outcomes.

https://doi.org/10.1371/journal.pone.0225720.t002

To determine whether the incidence of all-cause death and graft failure varied based on other covariates, we calculated the odds ratios (ORs) of various complex subgroups (Figs 1 and 2). Preoperative hypophosphatemia was associated with an increased incidence of all-cause death only in patients with chronic kidney disease or those undergoing non-emergent operation (P for interaction = 0.05 and 0.03, respectively). Other variables showed no interaction with all-cause death or graft failure.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Subgroup analysis of all-cause death according to the following variables: male sex, chronic kidney disease, ejection fraction<40%, valve combined operation, emergent operation, and the off-pump technique.

https://doi.org/10.1371/journal.pone.0225720.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Subgroup analysis of graft failure according to the following variables: male sex, chronic kidney disease, ejection fraction<40%, valve combined operation, emergent operation, and the off-pump technique.

https://doi.org/10.1371/journal.pone.0225720.g002

Discussion

In the present study, preoperative hypophosphatemia defined based on clinical criteria appeared to be associated with all-cause death and graft failure on univariate analysis. However, further analyses using IPW adjustment revealed that the association with all-cause death was not significant, and subgroup analysis indicated that this association varied based on chronic kidney disease and emergent operation. Interestingly, preoperative hypophosphatemia was independently associated with graft failure after CABG.

The relationship between serum phosphorus level and vascular calcification has been established based on a large body of evidence [2], and serum phosphorus abnormalities have been implicated in cardiovascular mortality and calcification of coronary arteries [4,5]. However, previous studies of patients undergoing coronary revascularization have shown inconsistent results [11–13]. Interestingly, all of these studies stratified patients into multiple groups in a graded manner by applying discrete criteria within normal range. Such methods might be convenient for showing statistical relationships but, in our opinion, does not reflect the clinical significance of phosphorus abnormalities. Moreover, the accuracy of the statistics applied in those studies to adjust covariates for comparing more than two groups is also questionable. Therefore, we stratified patients into normal, hypo-, and hyperphosphatemia groups using phosphorus levels defined based on clinical criteria and analyzed data using IPW adjustment.

Our study demonstrated that hyperphosphatemia is not associated with all-cause mortality. This result corresponds with the findings of a recent study on CABG [13]. The suggested explanation was that diseased arteries are fully replaced by endothelialized arteries or vein grafts in CABG, in contrast to other cardiovascular disease or coronary interventions, and so vascular calcification biomarkers are relatively useless in predicting outcomes [13]. On the other hand, in this study, hypophosphatemia appeared to be associated with all-cause death in multivariate analysis. Although this relationship has been previously suggested in open heart surgey [8], our subgroup analysis revealed that the association varied significantly based on other covariates and was limited to those with chronic kidney disease or those undergoing non-emergent operation. Considering that phosphorus-relevant biomarkers such as bone minerals, vitamin D, and parathyroid hormone are closely related to increases in cardiovascular risk, especially in the patients with kidney disease [7,17,18], these results might be related to the three main regulation mechanisms (i.e., diet intake, bone formation, and renal excretion), which interact dynamically. Poor oral intake is frequently observed in patients undergoing emergent procedures [19].

For adjustments among the three groups, we used a relatively new statistical method in this study. The propensity scores of relevant covariates are generally used to allow an observational study to mimic some of the particular characteristics of randomized trials [20] and there are four broad ways to use calculated propensity scores: matching, IPW, stratification, and covariate adjustment [21]. Although matching is the most frequently used method, using a logistic regression model to estimate the propensity score may lead to biased results in misspecified models [22]. Moreover, adequate adjustment with these methods could be challenging in comparisons among more than two groups. Therefore, using a generalized boosted regression has been proposed as an alternative approach for estimating propensity scores in IPW, and has also been suggested to be a suitable substitute matching method in some models [20]. For this study, institutional statistical experts participated in the design, and chose IPW as the most suitable statistical method to directly compare both hypo- and hyperphosphatemia to the normal group.

Interestingly, even after adjustment, hypophosphatemia was independently associated with graft failure. We suspect that this was related to endothelial dysfunction in hypophosphatemia [23]. In endothelial cells, nitric oxide is produced from L-arginine and molecular oxygen, and acts as a potent vasodilator that determines vascular remodeling [24]. However, in an experimental study, simulated hypophosphatemia decrease endothelial nitric oxide synthase activity and nitric oxide production via reduced intracellular calcium and increased protein kinase C β2 [23]. In addition, previous studies suggested that hypophosphatemia can cause a wide variety of cardiovascular diseases such as heart failure after cardiac surgery [8,25]. However, considering that graft failure after CABG can be caused not only by stenosis or thrombosis but also by native vessel flow or subclavian steal [26], further investigation of the relationship between hypophosphatemia and graft failure in CABG is needed.

There were several limitations to this study. First, since our study had a retrospective observational design, the possibility that hidden bias or unmeasured confounding factors might have influenced the results even after adjustment cannot be completely ruled out. Second, serum phosphorus abnormalities were determined based on a single preoperative measurement, and the duration of hypo- or hyperphosphatemia could not be considered in our analysis. Rapid turnover and tight regulation might have led to variation according to the timing of measurement. So, whether the correction of serum phosphorus level would improve clinical outcomes remains uncertain in this study and needs further investigation. Third, our study could not explain the underlying pathophysiologic mechanisms or determine whether phosphorus supplementation could improve outcomes. Finally, since we analyzed 14 years of data, our institutional protocols for surgical and perioperative management would have changed during the study period.

Conclusions

In conclusion, preoperative serum phosphorus abnormalities defined based on clinical criteria were not associated with clinical outcomes of CABG. However, further studies will be needed to reveal the relationship between hypophosphatemia and graft failure.

Supporting information

Acknowledgments