A Comparative Study of Serum Phosphate and Related Parameters in Chronic Kidney Disease between the USA and Japan

Chronic kidney disease (CKD) is an independent risk of developing cardiovascular disease (CVD) [1-3]. Recent studies have suggested that there is a wide variation in the incidence of CVD across CKD cohorts worldwide and that individuals with non-dialysis-dependent CKD (NDD-CKD) in the USA are more likely to develop CVD than the Japanese counterparts [4-6]. Similar findings have been reported in end-stage renal disease (ESRD) patients, and an ecological study suggested underlying associations with genetic and environmental factors [7], while apparent differences in clinical practice patterns in CKD-mineral and bone disorder remained [8]. It is therefore clinically important to elaborate international differences in risk factors for CVD between NDD-CKD cohorts, as a first step, to investigate what makes the difference in outcomes.

One such risk is serum phosphate. Many reports have shown that hyperphosphatemia is strongly associated with all-cause mortality [9, 10], cardiovascular events [11], vascular calcification [12], and the development of ESRD [13] among individuals with CKD. However, serum phosphate levels are associated with many factors, such as parathyroid hormone (PTH) [14], fibroblast growth factor-23 (FGF23) [15], and vitamin D status, in addition to renal function. Since these parameters are associated with one another and change nonlinearly as renal function declines [16], direct comparison of summarized figures may produce misleading interpretation. Simultaneous measurements and patient-level analyses enable extensive comparison of these phosphate parameters and thus can provide more detailed information than ecological studies.

Such an international comparison study in NDD-CKD has not been performed due to difficulties in standardizing different measurements and laboratory assays between cohorts. With the international collaboration between the Chronic Renal Insufficiency Cohort (CRIC) Study [17-19] in the USA and the Chronic Kidney Disease-Japan Cohort (CKD-JAC) Study [20, 21] in Japan, we could investigate phosphate parameters in NDD-CKD patients with different backgrounds. For example, Americans consume nearly 1.5 times as much phosphorus (1687 mg/day on average for men aged 40–49 years) as their Japanese counterparts [22], and vitamin D deficiency is much rarer in North America where vitamin D-fortified products are available than in Japan and other countries in Southeast Asia [23]. These differences may mediate international variation in phosphate parameters in CKD. This study aimed to investigate whether there are substantial differences in clinical parameters that affect phosphate metabolism between American and Japanese NDD-CKD patients, and, if so, to identify factors that might explain them.

CRIC is an American CKD cohort study of 3,939 participants. Eligibility criteria and baseline characteristics have been published elsewhere [17-19]. Briefly, it includes racially and ethnically diverse patients who were aged 21–74 and had mild to moderate CKD (estimated glomerular filtration rate [eGFR] ≤70 mL/min/1.73 m² for 21–44 years, eGFR ≤60 for 45–64, and eGFR ≤50 for 65–74). Approximately, half of study participants had diabetes. Exclusion criteria included prior kidney transplantation, polycystic kidney disease, primary kidney disease requiring active immunosuppressive therapy, and significant coexisting illness, such as liver cirrhosis, HIV infection, or cancer requiring chemotherapy.

CKD-JAC is a Japanese CKD cohort study of 2,977 participants. Enrollment began at 17 sites across Japan in September 2007, with inclusion criteria being (1) Japanese or Asian patients living in Japan, (2) age 20–75 years, and (3) CKD stage 3–5 with eGFR of 10–59 mL/min/1.73 m². Major exclusion criteria included (1) polycystic kidney disease, HIV infection, liver cirrhosis, or cancer and (2) prior transplantation or prior receipt of dialysis therapy. More detailed information about CKD-JAC has been published [20, 21].

Since serum phosphate was the main outcome of interest in this study, we only included participants with serum phosphate levels at enrollment (N = 6,502) in the analyses. All study protocols were approved by local institutional review boards. All study participants provided written informed consent.

In both CRIC and CKD-JAC, clinical data were collected at enrollment, and laboratory parameters of blood and urine samples were measured centrally using standardized assays, as reported elsewhere [17, 20]. Serum inorganic phosphate levels were measured with phosphomolybdate assay and enzymatic assay in CRIC and CKD-JAC, respectively. Both studies measured PTH with second-generation assays, but with different methods: an immunoradiometric assay (Total PTH, Scantibodies Laboratories, Santee, CA, USA) in CRIC and an electrochemiluminescence immunoassay (iPTH; Intact PTH, Roche® Elecsys 2010 System, Roche Diagnostics) in CKD-JAC. Plasma FGF23 levels were measured in CRIC using the second-generation C-terminal assay (cFGF23; Immunotopics, San Clemente, CA, USA), while serum FGF23 levels were measured in CKD-JAC using a sandwich-antibody method that only detects the biologically active form of FGF23 (iFGF23; Intact FGF23, Kyowa Medex Inc., Tokyo, Japan). Serum 25-hydroxy vitamin D (25OHD) levels were also determined differently in the two studies: a more sensitive liquid chromatography and mass spectroscopy method in CRIC and a conventional radioimmunoassay (DiaSorin® LIAISON 25OH Vitamin D Total Assay, Stillwater, MN, USA) in CKD-JAC.

For harmonization of the laboratory assays for phosphate biomarkers, we used three separate approaches. We originally created the conversion equations from CKD-JAC to CRIC (serum phosphate and PTH), directly measured a random subset of CRIC baseline samples in the same laboratory as CKD-JAC (FGF23), or used external conversion equations reported in previous literatures (25OHD) [24, 25] (for detailed information and for all online suppl. material, see www.karger.com/doi/10.1159/000521386).

Serum phosphate levels are associated positively with nutritional status and negatively with obesity [26]. We regarded serum albumin as a nutritional marker and body mass index (BMI) as a marker of obesity. Due to different body composition at a given BMI between Asians and non-Asians in general, we applied the World Health Organization Western Pacific Region classification [27] of obesity to Asian participants, while utilizing the World Health Organization criteria [28] for other races. We assessed dietary phosphate intake by either dietary frequency questionnaires (DHQ) optimized for the corresponding populations or urinary phosphate-to-creatinine ratio (uPhos/uCr), which was a surrogate of daily urinary excretion of phosphate [29]. Using the DHQ, we also assessed the frequency of carbonated beverage consumption as a proxy for the habit of consuming phosphate-enriched food products.

This is a cross-sectional study comparing serum phosphate levels in predialysis CKD patients between the USA and Japan using baseline data of CRIC and CKD-JAC, respectively. Data are expressed as a mean ± standard deviation and a median (interquartile range) for normally and non-normally distributed continuous variables, respectively. All continuous variables with right-skewed distribution (e.g., FGF23 and PTH) were logarithmically transformed prior to analysis. Differences in categorical, normal, and non-normally distributed continuous variables between the two cohorts were assessed using the χ² test, Student’s t test, and Wilcoxon rank-sum test, respectively.

Serum phosphate and PTH were missing in 6.0% and 4.3% of total participants, respectively. 25OHD was measured only in random samples (N = 1,690) in CRIC, while it was measured in 80.3% of participants in CKD-JAC. The original cFGF23 in CRIC and iFGF23 in CKD-JAC had missing values in 2.9% and 19.7%, respectively. There were 0.5–28.6% (9.5% on average) of missing values in other variables, including laboratory measurements and questionnaire data. We generated 20 multiply-imputed datasets in advance of the analyses, using chained equations [30] with predictive mean matching [31] and logistic regression for continuous and categorical variables, respectively. Before imputation, continuous variables with a normal distribution were converted to Z-scores to avoid multicollinearity, while right-skewed data were logarithmically transformed in advance.

To facilitate nonlinear relationships between eGFR and the parameters of interest, we modeled eGFR using restricted cubic spline (RCS) analysis with five interior knots (at 5, 25, 50, 75, and 95 percentiles of the combined population of CRIC and CKD-JAC) and performed multivariable linear regression analyses, including an interaction term of cohort-by-eGFR, adjusting for age, gender, albuminuria (urinary albumin-to-creatinine ratio [uACR]), and diabetes. Then, we stratified the subjects into five eGFR categories (≤20, 21–30, 31–40, 41–50, and >50; each category included one knot of RCS) for further analyses. To examine whether the associations between renal function and serum phosphate are consistent between the two cohorts (CRIC vs. CKD-JAC), we fit linear mixed-effects models with the harmonized serum phosphate level as the dependent variable, the cohort and an interaction term of eGFR category-by-cohort as the independent variables, and facility as a random intercept (base model). Then, we added potential confounders to the model in a stepwise manner adjusting for demographics (age, gender, albuminuria, and diabetes), nutritional status including obesity (albumin and BMI), medications (any phosphate binders and active vitamin D), and 25OHD as vitamin D status (Model 1). To evaluate the dietary effect, we included indices of calcium (Ca) and phosphorus load (dietary Ca and phosphorus intake, carbonated beverage consumption, and urinary phosphate excretion). Because of the similarity among the phosphorus load indices, we restricted the model to have either a combination of DHQ phosphorus and carbonated beverage intake or uPhos/uCr (Model 2) in addition to DHQ Ca. We, then, added the two phosphaturic hormones, PTH and FGF23 (Model 3). Given the limited capacity of urinary phosphate excretion in advanced CKD stages, we hypothesized that the same amount of phosphate load might differently raise serum phosphate levels across eGFR categories, and we additionally employed an interaction term, uPhos/uCr-by-eGFR category (Model 4).

For sensitivity analyses, we performed a subgroup analysis with only Asian participants in CRIC and CKD-JAC in the same manner, to minimize the influence of racial differences in phosphate parameters. We also repeated the analyses by using original phosphate values as the dependent variable, by using all the data including participants without phosphate measurements, and by using subgroups of the participants according to vitamin D status.

Statistical tests were two-sided and p values <0.05 were considered significant. We used Stata/SE 13.1 software for Windows (Stata Corp., College Station, TX, USA) for statistical analyses.

The characteristics of the study subjects and their original laboratory data before standardization (Table 1). CRIC participants were significantly younger, more were females and diabetic than their Japanese counterparts. Due to the different inclusion criteria, eGFR levels in CRIC were significantly higher than those in CKD-JAC (42.8 ± 13.5 vs. 28.7 ± 12.2 mL/min/1.73 m² in CRIC and CKD-JAC, respectively). In addition, significantly less albuminuria in CRIC (median uACR 52 vs. 497 mg/gCr) suggested less severe CKD in CRIC in general. Corrected Ca levels were slightly, but statistically significantly, higher in CRIC than in CKD-JAC (9.4 ± 0.5 vs. 9.2 ± 0.5 mg/dL). Before harmonization, serum phosphate levels were 3.7 ± 0.7 and 3.5 ± 0.7 mg/dL in CRIC and CKD-JAC, respectively. The scatter plot demonstrated different distributions of serum phosphate across eGFR levels between the two cohorts (online suppl. Fig. S2).

Notably, the participants in CRIC were much more frequently obese than those in CKD-JAC (BMI 32.1 ± 7.8 vs. 23.5 ± 3.8 kg/m² in CRIC and CKD-JAC, respectively). Even with stratification accounting for race (the World Health Organization for non-Asian and the World Health Organization Western Pacific Region for Asian), a considerable number of participants (29.8%) in CRIC were classified as “Obese II and or more,” whereas in CKD-JAC, only 5.6% were so classified. Twice as many participants in CRIC received phosphate binders as in CKD-JAC, but active vitamin D was prescribed more often in CKD-JAC participants than in CRIC (8.8 vs. 3.2%, respectively). All three indices for phosphate load (DHQ phosphorus intake, carbonated beverage consumption, and uPhos/uCr) were significantly higher in CRIC.

After harmonization of the parameters associated with phosphate metabolism (serum phosphate, PTH, and 25OHD) (Table 2), we performed multivariable regression analyses that incorporated RCS with an interaction term, eGFR-by-cohort, to broadly characterize the patterns of each parameter across eGFR levels (Fig. 1). All of the parameters showed nonlinear associations with eGFR; most of them (except for 25OHD) showed a significant interaction with the cohort, indicating that the associations between the phosphate parameters and eGFR were significantly different between CRIC and CKD-JAC. Even after adjustment for demographics, serum phosphate levels were significantly higher in CRIC across all eGFR levels, with a similar inflexion point at eGFR of 30 mL/min/1.73 m² (Fig. 1a). These patterns did not change substantially in the sensitivity analysis with original values of serum phosphate (online suppl. Fig. S2). Interestingly, the two phosphaturic hormones, PTH and FGF23, also demonstrated the different patterns between CRIC and CKD-JAC. The CRIC participants had lower PTH levels than their counterparts at a given level of eGFR (Fig. 1b); whereas, in FGF23 levels, they surpassed CKD-JAC participants at eGFR <30 mL/min/1.73 m² (Fig. 1c). Serum 25OHD levels were significantly higher in CRIC participants except for those with advanced CKD, while those in CKD-JAC were consistently low at about 15 ng/mL, half of the targeted level recommended by K/DOQI guidelines [32] (Fig. 1d).

We stratified the study participants by eGFR with cut-points at 20, 30, 40, and 50 mL/min/1.73 m² to elucidate the factors associated with the different patterns in phosphate-related biomarkers (Table 3). Due to the different eGFR distributions by design, we observed disproportionate numbers of participants in each study at the lowest and highest eGFR categories (86 and 837 participants with eGFR ≤20 in CRIC and CKD-JAC, respectively; and 837 and 131, with eGFR >50). As for medication (Fig. 2), CRIC participants were more likely to be treated with phosphate binders; whereas, they had higher phosphate levels across all eGFR categories. In contrast, CKD-JAC participants were more likely to be treated with active vitamin D analogues from earlier stages of CKD. Approximately, 10% of CRIC participants, except for those with eGFR ≤20, were receiving native vitamin D, while none in CKD-JAC. Interestingly, the prescription patterns of native vitamin D correlated well with serum 25OHD levels; we, thus, excluded them from the analytical model. DHQ phosphorus intake, carbonated beverage consumption, and uPhos/uCr were considerably higher in CRIC enrollees than CKD-JAC enrollees, consistent with higher daily phosphate load in CRIC study participants.

After multiple imputations, we fit multivariable linear mixed-effects models in a stepwise manner to evaluate the difference in serum phosphate levels across eGFR categories between CRIC and CKD-JAC and to explore the factors that could mediate or explain these differences (Table 4 or online suppl. Table S2 in detail). The exposure of interest was a combination of cohort and eGFR category, including the interaction of the two. The interaction was significant (p < 0.001) throughout all models. Serum phosphate levels in CRIC were consistently higher than those in CKD-JAC, and the difference between the cohorts increased significantly as eGFR declined, while the average phosphate levels rose with renal dysfunction, as had been formerly reported [16, 33]. This association remained significant even after adjustment for participants’ demographics, nutritional status, medications, and vitamin D status (Model 1; the intermediate models demonstrated identical results to Model 1, thus being omitted). The addition of Ca intake and indices of phosphate load significantly attenuated the mean difference in serum phosphate between CRIC and CKD-JAC at eGFR >50 mL/min/1.73 m² (0.48 mg/dL in Model 1 to 0.38 mg/dL in Model 2); however, the increasing difference at lower eGFR categories remained statistically significant. Although CRIC and CKD-JAC demonstrated different trends in PTH and FGF23 across eGFR levels, the addition of these variables did not alter the results substantially (Model 3). Furthermore, we added an interaction term, uPhos/uCr-by-eGFR category to account for the hypothesis that the same amount of phosphate load might differently raise serum phosphate levels across eGFR categories. The interaction was significant (p < 0.001), and it attenuated the incremental difference in serum phosphate toward lower eGFR levels between CRIC and CKD-JAC (Model 4, Fig. 3).

In the present study, we explored the differences in the parameters associated with serum phosphate levels between CRIC and CKD-JAC, both of which include a broad representation of CKD patients in their respective countries. The harmonization of measurements and the patient-level data integration enabled such analyses and, thus, helped explore the differences in serum phosphate and related parameters. Our results showed that serum phosphate levels in CRIC were significantly higher than those in CKD-JAC across all eGFR ranges, independent of demographic, nutritional, dietary, and treatment differences between the two cohorts. Intriguingly, they demonstrated different patterns in the two phosphaturic hormones, PTH and FGF23. CKD-JAC participants had significantly higher PTH levels than their counterparts, while CRIC participants showed relatively high FGF23 levels, especially among those with advanced CKD stages. Although FGF23 levels in CRIC were reevaluated with the same assay as in CKD-JAC by using deep-frozen aliquots of the same serums, based on the report by El-Mouche et al. [34], the measurement of iFGF23 of deep-frozen sample after 36–40 months tended to show lower values. Therefore, the difference might have been larger if we had measured iFGF23 simultaneously. According to the recent reports in humans and rodents, FGF23 is not only a biomarker for kidney disease progression and CVD or death, but it also directly induces hypertrophy in cardiomyocytes and promotes left ventricular hypertrophy [35-37]. The findings in the present study, that the elevation of FGF23 and PTH showed different patterns between American and Japanese CKD cohorts, are of potentially great importance in understanding the differences in CVD risk and burden across these two national settings.

The difference between CRIC and CKD-JAC was not limited to the patterns of the two phosphaturic hormones. The 25OHD levels in CKD-JAC were significantly lower than in CRIC, where 71% of the CKD-JAC participants had insufficient vitamin D levels of <20 ng/mL, while only 44% of CRIC participants had similarly low levels. This difference may have resulted from different policies regarding vitamin D supplementation in the two countries. In the US, many vitamin D-fortified food products and prescription/OTC native vitamin D supplements were available, while in Japan, none of these are generally accessible. Furthermore, higher uACR levels in CKD-JAC as compared to CRIC might also have contributed to its lower 25OHD levels. Proteinuria-promoted tubular injury has been demonstrated in patients with nephrotic syndrome, leading to a loss of endocytic apparatus components [38], which are responsible for a receptor-mediated uptake of the complex of 25OHD and vitamin D-binding protein in the proximal tubules. It has been reported that vitamin D status critically determines whether FGF23 or PTH becomes elevated first in the context of lower GFR [39]; however, low vitamin D status in CKD-JAC participants partially but not fully explained their tendency to have higher PTH and lower FGF23 levels than CRIC counterparts (online suppl. Fig. S5).

In addition, there was a remarkable difference in the indices of phosphate load. Dietary phosphorus intake, carbonated beverage consumption, and uPhos/uCr were all significantly greater in CRIC than in CKD-JAC. Although dietary recall data could not fully capture the amount of phosphorus intake including food additives [40], these data, along with the urinary phosphate excretion data, support that American CKD patients consume more phosphorus than their Japanese counterparts. This finding was consistent with the fact that FGF23 in CRIC was higher than that in CKD-JAC. Intriguingly, the addition of the interaction term of uPhos/uCr-by-eGFR category significantly improved the model fitness with the attenuation of incremental difference in serum phosphate between the two cohorts. This indicated that the same amount of phosphate load would cause greater burden at lower eGFR levels and that the incremental difference in serum phosphate was partially explained by the difference in the phosphate load between CRIC and CKD-JAC.

All these adjustments for demographic factors, drugs, phosphate load, uACR, 25OHD levels, and the two phosphaturic hormones, however, could not fully explain the different serum phosphate levels between the two cohorts, suggesting that there were other unmeasured explanatory factors. One such factor is the racial differences. Due to the lack of racial variation in CKD-JAC, we could not evaluate the association of race with serum phosphate levels sufficiently in this study. Instead, we performed a subgroup sensitivity analysis with Asian CRIC participants (N = 105) and CKD-JAC (N = 2,140) in the same manner to minimize racial differences (online suppl. Fig. S3); however, the differences between the cohorts were not eliminated. This finding implies that race did not play a critical role in the variation of serum phosphate levels. This is in line with the prior report from the CRIC study that revealed that the differences in serum phosphate levels between blacks and whites were largely modified by socioeconomic status and that they were negligibly small within the lowest income stratum [41]. We also performed a sensitivity analysis, adjusting for annual income level, although the income categories were not standardized between the two cohorts, and the result remained unchanged (data not shown). Interestingly, the association between serum phosphate and income level was not remarkable in CKD-JAC, suggesting the difference in habits and access to healthcare or social support system may be associated.

To the extent that the difference in serum phosphate levels between the two cohorts were not the result of different racial or genetic backgrounds, but rather from habits or behaviors that could be modified, the higher phosphate levels in the US could potentially be improved by a variety of interventions, such as diet, the main source of phosphate load, and access to healthcare. Our findings are in line with a recent report in healthy subjects that demonstrated significantly different serum FGF23 levels across differently industrialized countries [42]; however, Japan and the US are both well industrialized. Therefore, further investigations are warranted to identify factors responsible for these patterns. These differences in phosphate parameters may impact the international variations in the incidence of CVD, ESRD, and even mortality [43] and should be investigated in a longitudinal study. Given that CRIC and CKD-JAC are prospective cohort studies, the impact of other factors, such as the change in biomarkers, could also be explored.

The present study also highlights the importance of simultaneous measurements and interpretation of parameters associated with serum phosphate. As shown in Table 4, two phosphaturic hormones, FGF23 and PTH, were both significantly associated with serum phosphate levels, but in opposite directions. Regarding FGF23, the positive association could be explained by the loss of function of FGF23 with reduced klotho expression in the kidney [44] and the subsequent, ineffective elevation of FGF23 in response to phosphate load. PTH, on the other hand, demonstrated an effective, negative feedback against it. These bidirectional associations strongly suggest that there could be a masked phosphate load in patients with high FGF23 and high PTH levels simultaneously for a given normal serum phosphate level and imply the importance of controlling phosphate load, since continuous stimulation of the parathyroid glands may cause diffuse hyperplasia and adenomas in the long run [45], and the persistent elevation of FGF23 could be detrimental to renal function [46], cardiovascular health, and for survival [47]. As for serum phosphate, we were unable to confirm that the thresholds for cardiovascular risk were consistent across the cohorts; however, it is plausible to think that elevated serum phosphate levels should be disadvantageous in the same manner or more so in the CRIC participants because it is widely accepted that a higher phosphate concentration physiochemically promotes the formation of secondary calciprotein particles that lead to vascular calcification [48], and that Caucasians are more likely to develop atherosclerotic cardiovascular events than Asians [49]. Furthermore, it is noteworthy that a recent clinical trial in Japanese patients with ESRD revealed that the strict phosphate control of 3.5–4.5 mg/dL resulted in the less progression of coronary artery calcification scores than the modest control group (5.0–6.0 mg/dL) [50], suggesting even more benefit in American counterparts by lowering serum phosphate levels.

Our study had limitations that should be considered. Although we tried to minimize the differences in measurements, the harmonization methods we took in the present study were imperfect. We lack information about iron status, which, recent data suggest, impacts FGF23 levels. However, we measured iFGF23, which was less susceptible to iron status than cFGF23 [51], resulting in attenuation of the effect of iron deficiency. Additionally, because of uneven distribution of races between the cohorts and the limited variation in races in CKD-JAC, the association of race with serum phosphate levels could not be fully evaluated. Therefore, our finding of no association between race and serum phosphate was not conclusive. In addition, the participants in CRIC were not randomly chosen from all outpatients with CKD in the US. The participants of CKD-JAC were recruited only in secondary or tertiary hospitals in Japan and no one was enrolled in a clinic. Therefore, the generalizability of our findings needs to be interpreted with caution.

The present study showed that serum phosphate levels in CRIC were significantly higher than those of CKD-JAC and that there were significantly different patterns in phosphate parameters. Our patient-level analyses suggested that different backgrounds, dietary habits, medication patterns, and other clinical parameters associated with serum phosphate could partially explain the differences in phosphate levels. Other factors, such as genetics, might have affected phosphate levels; however, our analysis, which was limited to Asian participants, did not support that assumption. Further analyses, including direct genetic information, should be performed to explain the gap between the two cohorts. These findings may shed light on the international variations in serum phosphate and related parameters and thus in cardiovascular risk among CKD patients.

We thank all the investigators and participants of the CRIC, CKD-JAC, and Hyogo Cohort for their contributions. We hereby express our deepest condolences for the loss of Dr. Ohashi, who had been working enthusiastically for CKD-JAC as one of the biostatistical advisors. This international comparison project had not been successful without his great input about the harmonization of different measurements between the two cohort studies.

Both CRIC and CKD-JAC have been conducted ethically in accordance with the World Medical Association Declaration of Helsinki and the appropriate guidelines for human studies. The Institutional Review Board assessed the procedures and the conduct of the study, and the written informed consent has been obtained from each participant in both studies.

Dr. Naohiko Fujii, Dr. Takayuki Hamano, Dr. Takeshi Watanabe, Dr. Kosaku Nitta, Dr. Seiichi Matsuo, and Dr. Hirofumi Makino received a research support grant from Kyowa Kirin (KK). Dr. Satoshi Iimuro has consulted for KK and has been a member of the Cardiovascular Function Evaluation Committee. Dr. Enyu Imai has consulted for and received a research support grant from KK. Dr. Tadao Akizawa has consulted for, received a research support grant from, and has been a member of the speakers’ bureau of KK. Dr. Harold I. Feldman consulted for KK. Other authors have declared no conflict of interest regarding this manuscript.

The CKD-JAC study was financially supported by KK Co., Ltd., Tokyo, Japan and by the Japanese Society of Nephrology, Tokyo, Japan. Funding for the CRIC study was obtained under a cooperative agreement from the National Institute of Diabetes and Digestive and Kidney Diseases (U01DK060990, U01DK060984, U01DK061022, U01DK061021, U01DK061028, U01DK060980, U01DK060963, and U01DK060902). In addition, this work was supported in part by the Perelman School of Medicine at the University of Pennsylvania Clinical and Translational Science Award NIH/NCATS UL1TR000003, Johns Hopkins University UL1 TR-000424, University of Maryland GCRC M01 RR-16500, Clinical and Translational Science Collaborative of Cleveland, UL1TR000439 from the National Center for Advancing Translational Sciences component of the National Institutes of Health and NIH roadmap for Medical Research, Michigan Institute for Clinical and Health Research UL1TR000433, University of Illinois at Chicago CTSA UL1RR029879, Tulane COBRE for Clinical and Translational Research in Cardiometabolic Diseases P20. GM109036, Kaiser Permanente NIH/NCRR UCSF-CTSI UL1 RR-024131.

Dr. Naohiko Fujii and Dr. Takayuki Hamano designed the study; Dr. Harold I. Feldman, Dr. Yasuo Ohashi, Dr. Enyu Imai, and Dr. Tadao Akizawa supervised the elaboration of the study design and the data harmonization; Dr. Naohiko Fujii analyzed the data; Dr. Jesse Hsu, Dr. Wei Yang, Dr. Takayuki Hamano, and Dr. Harold I. Feldman supervised the data analysis; Dr. Naohiko Fujii made the figures, drafted, and revised the manuscript; All authors advised, revised, and approved the final version of the manuscript.

Since participants of CKD-JAC did not agree for their data to be shared publicly, supporting data are not available without the conclusion of a specific data use agreement with the steering committee of the CKD-JAC. The data on participants of CRIC may be partially available in NIDDK Central Repository under the name of CRIC or via the conclusion of a specific data use agreement with the steering committee of the CRIC Study.

The authors Naohiko Fujii and Takayuki Hamano contributed equally to this work.Yasuo Ohashi deceased during the submission process of this manuscript.