Effect Modifying Role of Serum Calcium on Mortality-Predictability of PTH and Alkaline Phosphatase in Hemodialysis Patients: An Investigation Using Data from the Taiwan Renal Registry Data System from 2005 to 2012

Predicting mortality in dialysis patients based on low intact parathyroid hormone levels is difficult, because aluminum intoxication, malnutrition, older age, race, diabetes, or peritoneal dialysis may influence these levels. We investigated the clinical implications of low parathyroid hormone levels in relation to the mortality of dialysis patients using sensitive, stratified, and adjusted models and a nationwide dialysis database. We analyzed data from 2005 to 2012 that were held on the Taiwan Renal Registry Data System, and 94,983 hemodialysis patients with valid data regarding their intact parathyroid levels were included in this study. The patient cohort was subdivided based on the intact parathyroid hormone and alkaline phosphatase levels. The mean hemodialysis duration within this cohort was 3.5 years. The mean (standard deviation) age was 62 (14) years. After adjusting for age, sex, diabetes, the hemodialysis duration, serum albumin levels, hematocrit levels, calcium levels, phosphate levels, and the hemodialysis treatment adequacy score, the single-pool Kt/V, the crude and adjusted all-cause mortality rates increased when alkaline phosphatase levels were higher or intact parathyroid hormone levels were lower. In general, at any given level of serum calcium or phosphate, patients with low intact parathyroid hormone levels had higher mortality rates than those with normal or high iPTH levels. At a given alkaline phosphatase level, the hazard ratio for all-cause mortality was 1.33 (p < 0.01, 95% confidence interval 1.27–1.39) in the group with intact parathyroid hormone levels < 150 pg/mL and serum calcium levels > 9.5 mg/dL, but in the group with intact parathyroid hormone levels > 300 pg/mL and serum calcium levels > 9.5 mg/dL, the hazard ratio was 0.92 (95% confidence interval 0.85–1.01). Hence, maintaining albumin-corrected high serum calcium levels at > 9.5 mg/dL may correlate with poor prognoses for patients with low intact parathyroid hormone levels.


Introduction
Chronic kidney disease (CKD)-mineral and bone disorder (MBD) increases morbidity and mortality in end-stage renal disease (ESRD) patients.
[1] Renal osteodystrophy may be initiated at the onset of CKD because of an increase in urinary phosphate (P) excretion. However, low parathyroid hormone (PTH) levels have long been associated with aluminum intoxication, malnutrition, older age, race, diabetes, peritoneal dialysis (PD), calcium (Ca)-containing P binders, and vitamin D overuse. [2,3] Therefore, it is difficult to determine the true association between low PTH levels and mortality.
Vascular calcification may be associated with low PTH levels in ESRD patients, [4,5] and coronary artery disease is the leading cause of death in these patients. However, normal PTH levels do not prevent low turnover disease. [6] In addition, low PTH levels may be linked to hypercalcemia. [7] Hypercalcemia may increase the mortality of ESRD patients with abnormal PTH levels at any given serum P level. [8] Therefore, risk stratification of patients based on the Ca or P levels as well as the PTH levels is crucial. [9] In this study, we analyzed data from a large, nationwide, and established ESRD registration database, the Taiwan Renal Registry Data System (TWRDS), [10] to determine predictors of mortality that included PTH, serum Ca, and P levels.

Materials and Methods
This study was approved by the ethics committee of Taipei Medical University's institutional review board (Number: 201411017), and was conducted in accordance with the principles of the Declaration of Helsinki. The requirement for written informed consent was waived.

The Taiwan Renal Registry Data System
The TWRDS was initially established in 1987 for the accreditation of dialysis therapy. All dialysis units in Taiwan must provide the appropriate information for inclusion in the TWRDS to obtain national health insurance (NHI)-associated reimbursements. Every dialysis unit submits a quarterly report. In 1996, a self-developed software program, HOPE, was used for computerized data collection. In 1997, additional data were gathered, including those relating to comorbidities, the patients' rehabilitation statuses, and the dialysis adequacy indices, which included the biochemical and hematological parameters, the hepatitis serological results, hypertension control, anemia management, and mineral bone indices. [11] The data within the TWRDS provide a robust foundation for the continual quality control of dialysis at the national level. [12][13][14][15] We analyzed the data in the TWRDS database that were generated between 2005 and 2012. During this period, 119,115 dialysis patients were registered within the database.
in December 2008. The dialysis units submit quarterly and annual reports to the TWRDS, and the data we obtained from these reports are described next.
Patients registered with the TWRDS from 2005 to 2012 were included in the analysis (n = 115,565). Patients who had received HD or PD for > 1 month were assigned to either the HD group or the PD group, respectively. After excluding 4,661 patients who had changed their dialysis modality, the final sample that went forward for analysis comprised 110,994 patients (Fig 1). Of these patients, 101,672 patients (91.6%) opted for HD and 9,232 patients (8.3%) opted for PD as their initial renal replacement therapy (RRT) modality between 2005 and 2012 (Fig 1). Among the HD patients, 6,113 (6%) had data missing in relation to their intact PTH (iPTH) levels, and 586 (0.6%) had data missing in relation to their alkaline phosphatase (ALP) levels; therefore, 94,983 patients were included in this analysis. Most patients were lost to follow up because of death, which was determined based on the complete national coverage provided by the NHI policy for all RRT expenditure. Most of the hospitals in Taiwan (97.9%) for intact PTH analysis was sent to the Union Clinical Laboratory, where second generation PTH Chemiluminescence assay ADVIA Centaur (Siemens Healthcare Diagnostics, Tarrytown, NY) was employed since 2007 (normal range 14-72 pg/mL). KDIGO guidelines on CKD-MBD were published in 2009 with second-generation intact PTH Allegro Nichols IRMA, and the commercial assays on average showed 60%-152% variation compared with the standard Nichols Allegro IRMA, [16] Therefore, the Union Clinical Laboratory still reported original data to the "HOPE" database of TWRDS.

Statistical analysis
The descriptive statistics were expressed as the means (standard deviations [SD]), medians (ranges), or frequencies (percentages). The average PTH, Ca, P, and ALP levels were used. Serum Ca levels were corrected using the following formula: corrected Ca = (0.8 × [normal albumin level − the patient's albumin level]) + serum Ca. We performed Cox regression analysis to estimate the hazard ratios (HR) for PTH or ALP. The case-mix adjusted model included the following confounding factors: age, sex, the presence of diabetes mellitus, and the laboratory data, which included hematocrit levels, albumin levels, Ca levels, P levels, and the HD treatment adequacy score, the single-pool Kt/V. Low PTH levels were defined as those that were lower than the KDIGO-recommended PTH level, which is between two-and ninetimes the upper limit of the normal PTH level. Since the upper limit for PTH in our assay was 72 pg/mL, we chose 150 pg/mL as the cutoff value for a low PTH level. Although the KDIGO guidelines suggest regular ALP measurements, a target range has not been specified. Hence, we used an ALP cutoff value of 120 U/L, because a higher ALP level correlated with mortality in a study from the University of California, Los Angeles (UCLA) study group that comprised a large population of HD patients. [17] In addition, another study found that an ALP level of 120 U/L was a robust predictor of coronary artery calcification. [18] All of the descriptive and multivariate analyses were performed using the Statistical Package for the Social Sciences software version 17.0 for Windows XP (SPSS Inc., Chicago, IL, USA) and SAS version 9.1 (SAS Institute, Cary, NC). Table 1 presents the study population's demographic data. A total of 94,983 HD patients with valid data relating to iPTH and ALP levels were included in this study. The mean duration of HD in this cohort was 3.5 years. The mean (SD) age of the patients was 62 (14) years. Type 2 diabetes was present in 51% of the patients. We divided each categories of iPTH by a cut-off value of ALP 120 U/L. There were no obvious difference in characters between different ALP levels in the same iPTH group except the percentage ALP 120 U/L group was higher in the iPTH > 300 pg/mL group. Patients with high iPTH levels ( 300 pg/mL) tended to have longer dialysis durations, and higher serum P and ALP levels (Table 1). Patients with low iPTH levels (< 150 pg/mL) had low albumin and hematocrit levels, were older, and tended to have diabetes.

Serum calcium and phosphate levels and their correlations with death
The risk of death increased when the serum Ca level reached 9.5 mg/dL (HR 1.05, 95% confidence interval [CI] 1.02-1.08) or the serum P level reached 5.5 mg/dL (HR 1.15, 95% CI 1.12-1. 19). The risk of death also increased when the serum Ca levels were < 8.5 mg/dL or the serum P levels were < 3.5 mg/dL ( Table 2).
Hazard ratio of all-cause mortality between different groups of calcium and phosphate and intact parathyroid hormone Table 3 shows that the risk of all-cause mortality was highest in the group with low iPTH, high Ca, and low P levels (HR 2.62, 95% CI 2.43-2.81), while the risk of all-cause mortality was lowest in the group with high iPTH, high Ca, and normal P levels (HR 0.86, 95% CI 0.81-0.91). In general, those patients with low iPTH levels had higher mortality rates than those with normal or high iPTH levels at any given serum level of Ca or P. All-cause mortality and its associations with intact parathyroid hormone and alkaline phosphatase levels One-year average baseline iPTH or ALP level and 3-year all-cause mortality. Fig 2A  illustrates the incremental trends in the three-year crude and adjusted HRs for mortality in ESRD patients in association with higher baseline one-year ALP levels. Fig 2B illustrates that the three-year crude and adjusted HRs for mortality in ESRD patients were higher when the iPTH levels were lower.
Time-averaged iPTH or ALP level and all-cause mortality. Fig 3A and 3B illustrates the incremental trends in the crude and adjusted HRs for mortality in ESRD patients in association with the time-averaged ALP levels within the different iPTH categories, especially in patients with low (< 150 pg/mL) (Fig 3A) or normal (150-300 pg/mL) (Fig 3B) iPTH levels. Fig 4A and  4B illustrates that the crude and adjusted HRs for mortality in ESRD patients within the  Fig 4A) and 120 U/L (Fig 4B), increased when the iPTH levels were lower.

Calcium and phosphate as predictors of high or low serum intact parathyroid hormone levels
For patients with ALP levels of 50-100 U/L and low iPTH levels, the cumulative HR for the mortality associated with high serum Ca levels (> 9.5 mg/dL) was 1.33 (95% CI 1.27-1.39) (p < 0.01) ( Table 4). In contrast, for patients with ALP levels of 50-100 U/L and high iPTH levels, the cumulative HR for the mortality associated with high serum Ca levels (> 9.5 mg/dL) was 0.92 (95% CI 0.85-1.01) ( Table 5), a result that was not statistically significant. The same results were obtained for patients with ALP levels of > 100 U/L (data not shown).

Discussion
In this large, long-term, nationwide study, approximately 45% of the HD patients had low iPTH levels (< 150 pg/mL). The prevalence of hypoparathyroidism was similar to that reported from a cohort study undertaken in south Asia. [19] However, the frequency of hypoparathyroidism in this study was much higher than the frequencies reported from the Dialysis Outcomes and Practice Patterns Study (DOPPS) that was undertaken in the United States in which 29% of the patients had iPTH levels of < 100 pg/mL, [8] and the United Kingdom Renal Registry (UKRR) study in which 32% of the patients had iPTH levels of < 150 pg/mL. [20] Furthermore, the low iPTH levels did not pose a threat to all-cause mortality in the DOPPS or the UKRR study, which is in contrast to our study's findings. However, the investigators participating in the UCLA study [21] and the ARNOS study, which was undertaken in France, suggested that low or very low PTH levels (< 50 pg/mL) may correlate with poor survival in HD patients. [22] In addition, the Cox regression analysis showed that higher ALP or lower iPTH levels were associated with higher levels of mortality in HD patients. Of particular interest is that a U-shaped association was determined between PTH levels and the risk of death for dialysis patients in a recent cohort study. [21] As a potential biomarker of CKD-MBD, ALP may be superior compared with PTH because it shows a lower level of intra-individual biological variation. [23] A publication from a Japanese study reported that high ALP levels in patients with low iPTH levels were strongly associated with hip fractures and all-cause or cardiovascular  mortality. [24] In this study, we confirmed that high ALP levels were harmful and were associated with mortality in HD patients. Since ALP is expressed in the liver, kidneys, intestines, bones, and leukocytes, high ALP levels may be a marker of inflammation and neutrophil activation. [25] Another publication has suggested that low PTH levels may be associated with inflammation or malnutrition, an association that confounded the relationship between serum PTH and ALP levels. [26] A recent investigation that involved a large sample of American patients who were undergoing HD, demonstrated the importance of controlling PTH levels  between 150 pg/mL and 300 pg/mL, [27] but the KDIGO guidelines only recommend regular measurements of ALP and no target range has been specified. From a recent large cohort study [28], albumin-corrected calcium level over 10.2 mg/dL or phosphate level over 6.2 mg/dL were correlated with higher adjusted mortality. However, in our study, given that high calcium 9.5 mg/dL and high phosphate 5.5 mg/dL, mortality was increased only in low iPTH group, but the a phenomenon that was not apparent in patients with normal or high iPTH levels. It is possible that hypercalcemia suppresses PTH secretion, and that its feedback inhibition may provide a beneficial physiological balance in hyperparathyroidism. In addition, the relatively higher values for iPTH levels produced by secondgeneration iPTH assays compared with those produced by third-generation bio-iPTH assays [29] might mean that the high iPTH levels are overestimations in some patients; hence, hypercalcemia may not be a problem in these patients.
The findings from the Dialysis Outcomes and Practice Patterns Study indicated that serum calcium levels of between 9.6 and 10.0 mg/dL may increase mortality, but statistical significance was not achieved [8]. In this study, we demonstrated that serum calcium levels of more than 9.5 mg/dL will gradually increase mortality (HR 1.05), but the clinical significance was not pronounced until the serum calcium levels reached 10.5 mg/dL (HR 1.77).
Vascular calcification had gained attention because of its associated morbidity and mortality [30], and avoiding Ca loads is crucial to prevent cardiovascular events in HD patients [31]. Most of the studies have focused on high PTH levels, and Ca and P levels and their correlations with mortality in dialysis patients [8]. The findings from our study indicate that high serum Ca levels may be harmful to HD patients with low PTH levels. Two recent meta-analyses of populations with CKD and ESRD determined that higher serum phosphate levels were correlated with higher cardiovascular mortality rates [1,32]; therefore, further investigations into the use of non-calcium-based phosphate binders in HD patients with low iPTH levels are warranted. No associations were found between high serum Ca levels and increasing mortality in ESRD patients with high iPTH levels in this study. However, in the current study's population, the patients with high iPTH levels and high P levels but relatively stable Ca levels, were younger and they had the optimal outcomes. Hypercalcemia may also reduce the calcemic effect of PTH, which may benefit secondary hyperparathyroidism. [33] Finally, patients with high PTH levels may have a wider normal range of serum Ca levels because only hyperphosphatemia and the Ca-P product can predict mortality [34].

Limitations
This study has several limitations that are described next. First, no data were available that described the patients' body mass indices. However, epidemiologic studies have shown that obese dialysis patients may have better outcomes [35]. Second, when mortality occurred it was recorded as a study drop out, and this occurred because almost all dialysis patients in Taiwan only receive one payment from the government to enable them to use the health care system, hence all patients who died were considered to be study drop outs, except those patients who received renal transplantations. Furthermore, we did not have accurate medical records regarding the use of orally administered calcium-containing phosphate binders, calcimimetics, or vitamin D, but, as yet, no studies have described reductions in mortality associated with the use of these drugs. Finally, the absence of bone specific alkaline phosphatase is also an important limitation in our study, further evaluation and study for bone specific alkaline phosphatase is warranted.

Conclusions
In this retrospective, large, nationwide, population-based cohort study, we determined that low PTH levels correlated significantly with mortality in HD patients. In addition, low PTH levels and high serum Ca levels were significantly associated with mortality. Additional prospective studies are warranted to investigate the use of non-Ca-containing P binders for the prevention of vascular calcification and complications among uremia patients.