Higher dialysate calcium concentration is associated with incident myocardial infarction among diabetic patients with low bone turnover: a longitudinal study

This is a longitudinal study on 53,560 hemodialysis patients from the Japan Renal Data Registry. Predictor was D[Ca] ≥3.0 vs 2.5 mEq/L. Outcomes were the first CV events during 1-year observation period. Association of D[Ca] with CV events and effect modifications were tested using multivariate logistic regression analyses. Diabetes mellitus (DM) was a significant effect modifier for association of higher D[Ca] and myocardial infarction (MI) (OR: 1.26 (1.03–1.55) among DM and 0.86 (0.72–1.03) among non-DM, p for interaction <0.01). The effect size was not affected by further adjustment for serum albumin-corrected Ca or intact parathyroid hormone (iPTH) levels, but was attenuated by adjustment for intradialytic change in serum Ca concentration (ΔCa) (1.16 [0.89–1.51]). Among DM, D[Ca] ≥3.0 mEq/L was significantly associated with MI in the first tertile of corrected Ca or iPTH ≤60 pg/ml (p for interaction 0.03 and 0.03, respectively). In conclusion, higher D[Ca] was associated with incident MI in DM, especially with low serum Ca or iPTH levels. Attenuation of the effect size by adjustment for ΔCa and stratified analyses suggest that larger Ca influx during dialysis with higher D[Ca] in patients suggestive of low bone turnover leads to vascular calcification and subsequent MI in DM.

Subjects. The inclusion criteria were prevalent hemodialysis patients who have been on hemodialysis for more than 1 year at the end of 2008, age 20-100 years old. The exclusion criteria were as follows: subjects who withdrew from dialysis, changed their treatment modality to peritoneal dialysis, or were transplanted, with missing data for D [Ca], history of CV events, or covariates used for adjustment, or apparent errors (ex. history of a CV event was present in 2008 but not in 2009).

Exposure of interest and outcomes. Exposure of interest was D[Ca]
≥3.0 mEq/L compared with 2.5 mEq/L in 2008, as D[Ca] of majority of commercially available dialysate at the time of the study in Japan was either 2.5 or 3.0 mEq/L. Outcome variables were the first episodes of CV events during the 1-year observation period in 2009. Subjects were considered to have incident events if they had no histories of these diseases in the 2008 database and did have histories of these diseases in the 2009 database. As the JRDR contains data on the histories of CV events at the end of each year, whether subjects had events during 2009 could not be determined for those with histories of CV events at the end of 2008. Thus, subjects with histories of MI or ischemic stroke at baseline were excluded from the analyses when the outcome variables were incident MI or ischemic stroke, respectively (Fig. 1).
Definition. Serum albumin-corrected Ca was calculated as follows: corrected Ca (mg/dL) = serum calcium (mg/dL) + 0.8* [4-serum albumin (g/dL)] if serum albumin is less than 4 g/dL. Intradialytic change in serum  M: mean, SD: standard deviation, n: number Associations of D[Ca] and CV events were evaluated using multivariate logistic regression analyses adjusted for the following variables: age, sex, causes of end-stage renal disease, dialysis vintage, albumin, phosphate, quintiles of Kt/V, total cholesterol, C-reactive protein, ultrafiltration rate, body mass index, hemoglobin, and history of CV events other than the outcome variable. Corrected Ca and intact parathyroid hormone (iPTH) levels were not included in the initial model as they could be in the causal pathway between higher D[Ca] and CV events. Intact PTH levels were imputed from the 2007 database (last observation carried forward) as the data were not available in the 2008 database. Intact PTH were divided into four categories (≤60, 61-150, 151-240, >240 pg/ml) according to guidelines by Japanese Society for Dialysis therapy 18 . Logistic regression analyses were performed instead of survival analyses because the data indicated that the subjects had histories of each disease at the end of the year, but not when the events occurred during the year. Mediation analysis was performed using incident MI, D[Ca], and delta Ca as a dependent variable, an independent variable, and a mediator variable, respectively. Descriptive statistics and logistic regression analyses were performed using SPSS version 19.0 (SPSS Inc, Chicago, IL). Mediation analysis was performed using Stata MP version 15.0 (Stata Corp,College Station, Texas).

Results
At the end of 2008, data were collected on 301,649 patients treated with dialysis in Japan. Among them, 292,301 patients met the inclusion criteria. After the exclusion of patients by the exclusion criteria, data from 53,560 patients were analyzed (Fig. 1). Excluded patients were older by 1.5 years and vintage was shorter by 1 year. The causes of end-stage renal diseases were similar and the proportion of patients using higher dialysate Ca concentration was also similar. The characteristics of the subjects included in the analyses were shown in Table 1. The proportion of patients using D[Ca] 2.5, 3.0, 3.3, and 3.5 mEq/L was 38.8%, 60.8%, 0.2%, and 0.2%, respectively. The patients using D[Ca] ≥3.0 mEq/L had significantly higher corrected Ca and lower iPTH levels. There were statistically significant differences in age, causes of end-stage renal disease, dialysis vintage, body mass index, total cholesterol, C-reactive protein and KT/V among users of D[Ca] ≥3.0 mEq/L and 2.5 mEq/L, but the standardized difference was less than 10%. Serum phosphate was significantly lower among users of D[Ca] ≥3.0 mEq/L and  , 304 (0.6) and 5,095 (9.9), respectively. There were 186 death due to coronary artery disease. As the incidence of hemorrhagic stroke and limb amputation was low, subsequent subgroup analyses were performed for MI, ischemic stroke, and CV composite outcomes. In the total cohort, there was no significant association between D[Ca] and CV events in both univariate and multivariate models ( Table 2). Association of D[Ca] and CV events in different subgroups were examined. D[Ca] ≥3.0 mEq/L compared with 2.5 mEq/L was significantly associated with MI among patients with diabetes mellitus (DM) (Fig. 2). The ORs (D[Ca] ≥3.0 mEq/L compared with 2.5 mEq/L) for MI among diabetics and non-diabetics were 1.26 (1.03-1.55) and 0.86 (0.72-1.03), respectively (p for interaction = 0.007). There was no significant interaction in other subgroups (Fig. 2). Also, higher D[Ca] was not associated with ischemic stroke or CV composite outcomes in any subgroups (data not shown). Therefore, we focused on association of D   As the use of D[Ca] ≥3.0 mEq/L was associated with significantly higher corrected Ca and lower iPTH levels (Table 1), there was a possibility that association of D[Ca] and MI were through its effect on serum Ca or iPTH levels. Thus the model was further adjusted for corrected Ca and iPTH levels ( Table 3). There was no change in the effect size. When the model was adjusted for ΔCa (post-dialysis serum Ca -pre-dialysis serum Ca level), the association of D[Ca] and MI was attenuated (Table 3). Mediation analysis was performed using incident MI, D[Ca], and delta Ca as a dependent variable, an independent variable, and a mediator variable, respectively. Data were adjusted for the same covariates as logistic regression analyses. Proportion of total effect mediated by delta Ca was 52%. These results suggest that D[Ca] ≥3.0 mEq/L was associated with MI through raising serum Ca rapidly during dialysis. Among non-diabetics using D Patient with low serum corrected Ca should have high ΔCa due to large concentration gradient of Ca between serum and dialysate. Also, patients with low iPTH might not be able to buffer Ca influx during hemodialysis due to low bone turnover. Association of D[Ca] and MI among diabetics and non-diabetics, stratified by corrected Ca and iPTH was examined and the results were shown in Fig. 3

Discussion
In this study, the association of D[Ca] and CV events and its effect modifiers were examined. The major findings were as follows: higher D[Ca] was significantly associated with MI only among diabetics, especially with low corrected Ca or low iPTH levels, and this association was not through higher serum corrected Ca or lower iPTH levels but through rapid increase in serum Ca during hemodialysis with the use of higher D [Ca].
Previous studies showed variable association of D[Ca] and CV morbidity or mortality 7,8,15,16 . In our study, D[Ca] was not significantly associated with CV events in total cohort. These different associations of D[Ca] and CV events could be due to the different patient population. In fact, previous studies have showed that impact of D[Ca] may be different according to patient background. For example, when D[Ca] was changed from 3.0 to 2.75 mEq/L, iPTH levels increased among patients with low baseline iPTH, but iPTH levels did not change or slightly decreased among patients with high baseline iPTH 12 . Another study showed that progression of coronary calcification score was significantly larger among high D[Ca] users if baseline serum phosphate was ≥4.7 mg/dL but not if baseline serum phosphate was <4.7 mg/dL 17 . These studies unfortunately did not examine the association of higher D[Ca] and CV events as the number of patients included was small and also the follow-up was short. In our study, association of D[Ca] and different CV events in different subgroups were examined. D[Ca] was significantly associated with MI, but not other CV events, only among diabetics (Fig. 2). The difference in prevalence of DM and the incidence of MI might be an explanation for different association of D[Ca] and CV mortality in previous studies. Also, association of D[Ca] only with MI was similar to our previous study showing that Ca, phosphate, and iPTH levels were associated most with MI compared to other CV events 19 . Unlike a previous study 17 , phosphate level was not an effect modifier for the association of D[Ca] and CV events in our study (Fig. 2).
Despite the expectation that physicians would try to control corrected Ca and iPTH levels in target range recommended by the guideline 18 irrespective of D[Ca], the users of higher D[Ca] had significantly higher corrected Ca and significantly lower iPTH levels (Table 1). Similar results were reported in previous studies 12   or iPTH, as the adjustment for these variables did not change the effect size (Table 3). Rather, adjustment for ΔCa attenuated the association of D[Ca] and MI among diabetics ( Patients with low serum Ca should have larger ΔCa due to large concentration gradient of Ca between serum and dialysate. Also, patients with low iPTH levels are likely to have low bone turnover and may not buffer large Ca influx during dialysis. Thus, we performed stratified analyses by corrected Ca and iPTH. Higher D[Ca] was associated with MI especially in patients with low corrected Ca and iPTH among diabetics (Figs 3 and 4), suggesting that diabetic patients likely with low bone turnover could not buffer larger Ca influx with the use of higher D [Ca] and that might lead to vascular calcification and subsequent MI. Of note, the use of higher D[Ca] suppresses iPTH levels. Thus, the use of higher D[Ca] increases the number of patients with low iPTH, for whom, the use of higher D[Ca] was associated with increased incidence of MI. As this study is an observational study, low iPTH levels caused by higher D[Ca] and low iPTH with the use of lower D[Ca] were treated equally, which would underestimate the effect size. Actually, previous studies have shown that low iPTH induced by higher D[Ca] was associated with increased incidence of CV mortality 23 .
The significant effect modification by corrected Ca and iPTH and association of ΔCa and MI among higher D[Ca] users were only seen among diabetics (Fig. 3, Table 4). It is known that iPTH secretion is impaired in diabetics 24 and it leads to low bone turnover. However, in our study, even among same strata of iPTH ≤60 pg/ mL, D[Ca] was associated with MI only among diabetics. It has been reported that function and differentiation of osteoblasts and ostelclasts are impaired and these lead to low bone turnover in DM [25][26][27] . For example, advanced glycation end products (AGEs) act through receptor for AGEs on osteoblasts and inhibit osterix and runt-related transcription factor 2, which are necessary for osteoblast differentiation 25 and induce transforming growth factor-β, which inhibits osteoblast differentiation 26 . AGEs also inhibit receptor activator of nuclear factor-κB ligand, which are necessary for osteoclast differentiation 27 . Our results further reinforce the notion that DM itself is the cause of low bone turnover, independent of iPTH in chronic kidney disease.
Strength of our study is as follows: this is a large cohort of hemodialysis patients and associations of D[Ca] with hard outcomes were examined. In Japan, more than 90% dialysis facility use central dialysate supply system. This means that the choice of dialysate Ca concentration is center-based and not based on patient characteristics such as serum Ca or iPTH levels and there is little indication bias for D[Ca] for each individual patient. Thus, JRDR is a unique population for investigating the association between D[Ca] and CV events. Also, to our knowledge, this is the first study showing that rapid increase in serum Ca during hemodialysis probably from large concentration gradient of Ca between serum and dialysate and also from low buffering capacity of bone is a possible mechanism to explain association of higher D[Ca] and CV events.
There are several limitations in our study. As this was an observational study, there was a possibility of residual confounders. Data for the use of vitamin D, cinacalcet and phosphate binders were not available in the database. It is known that more vitamin D is used among patients dialyzed by lower D[Ca] 20 . However, the association between dialysate Ca concentration and MI was mediated through intradialytic change in serum Ca concentration and it is unlikely that vitamin D, cinacalcet, and phosphate binders have direct impact on intradialytic change in serum Ca concentration. Also, although we do not have exact data, cinacalcet was introduced to Japanese market in 2008, and the number of patients on cinacalcet was presumed to be small at the time of the study. Also, as we could not determine whether patients with history of CV events at baseline had recurrent events during the observation period, the incidence of CV events should have been underestimated. Large number of subjects was excluded from analyses due to missing data and demographics of excluded and included subjects were different. However, it is unlikely that data were systematically missed in relation to exposure (i.e. D[Ca]) and it is unlikely to affect the results. The data for D[Ca] and post-dialysis serum calcium concentration were only available in 2008 in JRDR database. As the use of lower D[Ca] increased over time since then in Japan, we limited the observation period to one year to examine the effect of D[Ca] and delta Ca on outcomes.
In conclusion, we showed that higher D[Ca] was associated with incident MI among diabetics, especially with low serum Ca and iPTH levels. The attenuation of effect size by adjustment for ΔCa and stratified analyses suggest that larger Ca influx with the use of higher D[Ca] in patients likely with low bone turnover lead to vascular calcification and subsequent MI. The results suggest that D[Ca] ≥3.0 mEq/L should be avoided in diabetic patients with low serum Ca or iPTH levels.