Introduction
If statins have been demonstrated to be efficient in preventing cardiovascular events in patients with normal kidney function, several studies showed that they are less efficacious in reducing cardiovascular disease risk in patients with CKD, particularly in patients on dialysis therapy (1). Recently, the authors of observational analyses on the basis of the United States Renal Data System, including Medicare claims, demonstrated that statins might not meaningfully reduce atherosclerotic cardiovascular events even among patients on incident dialysis with established atherosclerotic heart disease and those likely to receive kidney transplants (2).
Beyond generally normal or even low LDL levels in these patients, the reasons for the absence of statin efficacy in treating or preventing cardiovascular disease remain obscure. Several hypotheses have been advanced to explain this lack of efficacy, such as a more limited number of atheromatous events and/or survival selection in those patients who reach kidney failure (3). An additional explanation is the observation of low or no statin effects on vascular stiffness, cardiovascular calcification, or heart failure, which are predominant pathologies in patients on dialysis (3). Other potential reasons are more severe comorbidities and extremely high pill burden, both favoring poor adherence.
A recent experimental study (4,5) showed that phosphate excess promotes de novo cholesterol synthesis in vascular smooth muscle cells and macrophages through 3-hydroxy-3-methylglutaryl CoA (HMG-CoA) reductase activation. This observation led us to hypothesize that this newly described mechanism might also account for vascular smooth muscle cell resistance to statins in patients with CKD who might therefore require higher doses of statins to overcome the phosphate-induced HMG-CoA reductase activation.
The aim of this post hoc analysis is therefore to explore the efficacy of statin treatment on major adverse cardiovascular events (MACEs) and all-cause death according to serum phosphate levels (at baseline and its change during trial course) in the patients on dialysis who participated in the A Study to Evaluate the Use of Rosuvastatin in Subjects on Regular Hemodialysis: An Assessment of Survival and Cardiovascular Events (AURORA) trial (6) and to use the Deutsche Diabetes Dialyze Studie (4D) trial as a validation cohort (7).
Materials and Methods
Study Cohort and Design of the AURORA and the 4D Trials
The AURORA design, baseline data, and main results have been previously described (6). In short, it was a randomized, double-blind, parallel group study that recruited 2776 patients receiving maintenance hemodialysis (for at least 3 months), aged 50–80 years, from 280 nephrology centers in 25 countries. Differing levels of serum phosphate levels were not an exclusion criterion. Eligible patients were randomly assigned to receive either rosuvastatin 10 mg daily or matching placebo. The composite primary end point was MACE defined as death from cardiovascular causes, nonfatal myocardial infarction, or nonfatal stroke. All myocardial infarctions, strokes, and deaths were reviewed and adjudicated by a clinical end point committee blind to treatment allocation. Secondary end points included death from all causes and death from cardiovascular and noncardiovascular causes.
This post hoc analysis was performed in the 2457 patients with baseline phosphate data and baseline and 3-month LDL data (Supplemental Figure 1). Median follow-up of this subpopulation was 3.83 years (interquartile range, 1.79–4.46 years). Follow-up visits occurred 3 months after randomization, and then every 6 months for 4 years, with a closeout visit. However, serum phosphate was mostly available at baseline, at 1 year, and at closeout.
The design and main results of the 4D trial have also been reported previously (7). Briefly, 4D was a prospective, randomized controlled trial including 1255 patients with type 2 diabetes mellitus, aged 18–80 years, receiving hemodialysis for <2 years. The patients were recruited from 178 dialysis centers in Germany. Differing serum phosphate levels were not considered as exclusion criteria. This post hoc analysis was performed in the 1255 patients of the initial trial in whom information on baseline phosphate and LDL was available. Median follow-up was 2.46 years. Follow-up visits occurred 1 month after randomization, and then every 6 months for 5 years maximum. In 4D, serum phosphate data were available for most patients at each visit.
Both studies were conducted in accordance with the ethical principles of the Declaration of Helsinki, the International Conference of Harmonization Good Clinical Practice guidelines, and local regulatory requirements. All patients provided written informed consent.
Statistical Analyses
Changes of Serum Phosphate and Low-Density Lipoprotein during Follow-Up in Relation to Treatment in the AURORA Trial.
We described phosphate changes during follow-up and checked—by testing an interaction between longitudinal serum phosphate and treatment—that the changes were not affected by treatment by running a participant-specific random intercept and slope linear mixed model with lmerTest (8) and lme4 packages (9). As serum phosphate changes during follow-up were nonlinear, we included a restricted cubic spline in the model using the rms package (10).
We then aimed to classify patients by groups of similar phosphate trajectories over time. To identify those trajectory groups, we used a latent class growth curve model with the lcmm package (11). Because there was no effect of treatment on serum phosphate levels over time (P value for interaction =0.42) (Supplemental Figure 2), we conducted the latent class mixed modeling approach across treatment arms. We used a β-link function with two trajectory groups as this approach proved to yield the best fit for the data (Supplemental Table 1).
Baseline characteristics were described for the overall population and according to the combination of treatment and phosphate trajectory groups. Percentages were given for categorical variables, and mean (SD) or median [quartile 1, quartile 3] was given for continuous variables depending on normality.
Next, we tested whether phosphate as a longitudinal exposure (summarized by the identified trajectory groups) modulated LDL changes over time in relation to treatment. Using a mixed model similar to the one stated above, we investigated the interaction between LDL over time and phosphate trajectories while stratifying by treatment arm to reduce model complexity. As LDL changes over time were nonlinear, we included a restricted cubic spline in the model. For AURORA, the following adjustment variables were included: age, sex, serum C-reactive protein (CRP), diabetes, and Kt/V at baseline, with LDL at baseline added in a second time. We imputed the missing values for CRP (one patient) and Kt/V (103 patients) with their respective median.
Association between Outcomes and Serum Phosphate Variations in Relation to Treatment in the AURORA Trial.
Finally, we tested whether the interaction between phosphate and treatment affected the two outcomes of interest (MACE and all-cause death) using Cox proportional hazards models with phosphate considered as a time-dependent covariate using the survival package (12), as well as with baseline phosphate. The assumptions of the Cox proportional hazard model (linearity and proportional hazards) were checked and, if necessary, issues were addressed correspondingly. All relevant variables chosen a priori on the basis of a pathophysiologic standpoint were included. Thus, for AURORA, the following adjustment variables were used: age, sex, smoking status, CRP, diabetes, cardiovascular history, and mean BP; time-varying LDL was added in a second step to check that it did not influence the results. We imputed the missing values for mean BP (three patients). Results of the interactions were presented in two complementary ways: the effect of time-dependent phosphate variations in both the placebo and rosuvastatin arms and the effect of treatment for specific values or groups of values of time-dependent phosphate variations. We reanalyzed the same models with adjustment for sevelamer, vitamin D, calcium substitution use, and time-dependent albumin levels (dichotomized along a 3.5-g/dl cutoff) and compared them with current results. We also provided visualization of the functional relationship between phosphate and risk by using a restricted cubic spline.
In addition to these main analyses, we also analyzed cardiovascular and noncardiovascular death separately (using the same approach).
Replication of the Analyses in the 4D Trial
The analyses described in the last subsection were also performed in the 4D trial. All of the patients initially included in the 4D trial were included in our post hoc analysis. For 4D, we used the same adjustment variables except that we replaced diabetes by diabetes duration. However, model assumption checks showed that the hazards of treatment and time-dependent phosphate were significantly nonproportional for MACE, and an interaction with the log of time was found. On the basis of a graphical inspection of the Schoenfeld residuals, we decided to split time in two periods (the first 2.5 years and after 2.5 years) and tested the interaction between time-dependent phosphate and treatment within each time period.
All analyses were conducted in R 4.0.4 (R Core Team 2021).
Results
Changes of Serum Phosphate and LDL during Follow-Up in Relation to Treatment in the AURORA Trial
The two trajectories identified corresponded to patients with high serum phosphate levels at baseline whose levels remained high during the course of the study (13% of patients) and to patients with lower baseline serum phosphate levels that remained low (Supplemental Figure 3). The two treatment arms were evenly spread across these groups, with 1070 patients from the placebo arm and 1068 patients from the rosuvastatin arm in the low-phosphate group and 162 patients from the placebo arm and 157 patients from the rosuvastatin arm in the high-phosphate group.
The baseline characteristics of the population according to phosphate trajectory groups are presented in Table 1. Cause of kidney failure was more likely to be diabetes for the low-phosphate trajectory group and more likely to be hereditary or tubulointerstitial kidney diseases for the high-phosphate trajectory group. Diabetes, overall cardiovascular history (more specifically, coronary artery bypass graft, peripheral artery disease, and carotid stenosis ≥50%), and the use of angiotensin-converting enzyme inhibitor or angiotensin receptor blocker were more likely in the low-phosphate trajectory group, whereas sevelamer use was more likely in the high-phosphate trajectory group. There were no differences between trajectory groups in terms of serum LDL changes at 3 months in the rosuvastatin group. In contrast, the patients on placebo of the high-phosphate trajectory group had higher serum LDL levels at 3 months than those in the low-phosphate trajectory group.
Table 1. -
Baseline characteristics in relation to serum
phosphate trajectories and the treatment arm (A Study to Evaluate the Use of Rosuvastatin in Subjects on Regular Hemodialysis: An Assessment of Survival and Cardiovascular Events trial)
Population Characteristic |
Overall |
N
|
Low-Phosphate Trajectory |
High-Phosphate Trajectory |
Placebo |
Rosuvastatin |
Placebo |
Rosuvastatin |
n
|
2457 |
|
1070 |
1068 |
162 |
157 |
Age at randomization, yr |
64 (9) |
2457 |
65 (9) |
64 (9) |
61 (8) |
62 (8) |
Women |
943 (38) |
2457 |
387 (36) |
427 (40) |
74 (46) |
55 (35) |
Origin
|
|
2457 |
|
|
|
|
European |
2087 (85) |
|
910 (85) |
893 (84) |
139 (86) |
145 (92) |
Sub-Saharan |
90 (4) |
|
40 (4) |
43 (4) |
5 (3) |
2 (1) |
Asian |
124 (5) |
|
54 (5) |
60 (6) |
8 (5) |
2 (1) |
Hispanic |
96 (4) |
|
43 (4) |
43 (4) |
7 (4) |
3 (2) |
Other |
60 (2) |
|
23 (2) |
29 (3) |
3 (2) |
5 (3) |
Body mass index, kg/m2
|
25 (5) |
2430 |
25 (5) |
25 (5) |
26 (5) |
26 (5) |
Systolic BP, mm Hg |
137 (24) |
2455 |
136 (24) |
137 (24) |
136 (26) |
132 (24) |
Diastolic BP, mm Hg |
76 (13) |
2454 |
75 (12) |
76 (13) |
76 (12) |
76 (14) |
Current smoker |
377 (15) |
2457 |
173 (16) |
148 (14) |
32 (20) |
24 (15) |
Total cholesterol, mg/dl |
175 (42) |
2457 |
173 (41) |
176 (43) |
181 (45) |
175 (43) |
LDL cholesterol, mg/dl |
99 (34) |
2457 |
98 (33) |
100 (35) |
104 (35) |
99 (36) |
HDL cholesterol, mg/dl |
45 (15) |
2457 |
45 (16) |
45 (15) |
45 (13) |
44 (14) |
Triglycerides, mg/dl |
156 (97) |
2457 |
153 (88) |
157 (96) |
162 (143) |
159 (96) |
High-sensitivity C-reactive protein, mg/L |
0.5 [0.2, 1.4] |
2456 |
0.5 [0.2, 1.4] |
0.5 [0.2, 1.3] |
0.5 [0.2, 1.8] |
0.5 [0.2, 1.5] |
Hemoglobin, g/dl |
11.7 (1.6) |
2355 |
11.7 (1.6) |
11.8 (1.6) |
11.4 (1.7) |
11.5 (1.7) |
Albumin, g/dl |
4.0 (0.3) |
2457 |
4.0 (0.3) |
4.0 (0.3) |
4.0 (0.3) |
4.0 (0.3) |
Calcium, mg/dl |
9.4 (0.9) |
2456 |
9.4 (0.8) |
9.4 (0.9) |
9.2 (1.0) |
9.4 (1.1) |
Kt/V measured |
1.4 (0.6) |
2051 |
1.5 (0.6) |
1.4 (0.6) |
1.3 (0.4) |
1.4 (0.5) |
Kt/V calculated |
1.2 (0.3) |
2354 |
1.2 (0.3) |
1.2 (0.3) |
1.1 (0.3) |
1.1 (0.3) |
Duration of treatment with hemodialysis, yr |
3.5 (3.9) |
2457 |
3.6 (4.0) |
3.5 (3.9) |
3.1 (2.6) |
3.4 (3.7) |
Type of kidney disease
|
|
2457 |
|
|
|
|
Nephrosclerosis |
484 (20) |
|
223 (21) |
206 (19) |
24 (15) |
31 (20) |
GN or vasculitis |
455 (19) |
|
198 (19) |
204 (19) |
33 (20) |
20 (13) |
Diabetes |
474 (19) |
|
205 (19) |
226 (21) |
19 (12) |
24 (15) |
Tubulointerstitial disease |
363 (15) |
|
141 (13) |
156 (15) |
34 (21) |
32 (20) |
Hereditary |
316 (13) |
|
137 (13) |
127 (12) |
28 (17) |
24 (15) |
Other cause |
365 (15) |
|
166 (16) |
149 (14) |
24 (15) |
26 (17) |
Diabetes |
643 (26) |
2457 |
272 (25) |
301 (28) |
33 (20) |
37 (24) |
Cardiovascular history |
953 (39) |
2457 |
436 (41) |
412 (39) |
49 (30) |
56 (36) |
Prior myocardial infarction |
243 (10) |
2457 |
114 (11) |
103 (10) |
9 (6) |
17 (11) |
Prior coronary angioplasty or stent |
73 (3) |
2457 |
40 (4) |
25 (2) |
4 (2) |
4 (3) |
Coronary artery bypass graft |
92 (4) |
2457 |
47 (4) |
39 (4) |
2 (1) |
4 (3) |
Peripheral artery disease |
363 (15) |
2457 |
171 (16) |
166 (16) |
12 (7) |
14 (9) |
Carotid artery disease |
125 (5) |
2456 |
58 (5) |
59 (6) |
2 (1) |
6 (4) |
Carotid stenosis ≥50% |
42 (2) |
2455 |
15 (1) |
27 (3) |
0 (0) |
0 (0) |
Carotid endarterectomy |
17 (1) |
2457 |
6 (1) |
11 (1) |
0 (0) |
0 (0) |
Prior ischemic vascular accident |
206 (8) |
2457 |
86 (8) |
98 (9) |
9 (6) |
13 (8) |
History chronic stable/unstable angina |
458 (19) |
2457 |
200 (19) |
193 (18) |
27 (17) |
38 (24) |
Transient ischemic attack |
107 (4) |
2457 |
42 (4) |
56 (5) |
3 (2) |
6 (4) |
Angiotensin-converting enzyme inhibitor or angiotensin receptor blocker |
898 (37) |
2447 |
412 (39) |
392 (37) |
50 (31) |
44 (28) |
β-blocker |
912 (37) |
2447 |
384 (36) |
407 (38) |
58 (36) |
63 (40) |
Calcium channel blocker |
875 (36) |
2457 |
393 (37) |
365 (34) |
56 (35) |
61 (39) |
Diuretic |
759 (31) |
2457 |
333 (31) |
327 (31) |
46 (28) |
53 (34) |
Platelet inhibitor |
1023 (42) |
2457 |
440 (41) |
441 (41) |
69 (43) |
73 (46) |
Vitamin D |
1153 (47) |
2457 |
511 (48) |
493 (46) |
73 (45) |
76 (48) |
Calcium substitution |
1834 (75) |
2457 |
799 (75) |
800 (75) |
123 (76) |
112 (71) |
Sevelamer |
451 (18) |
2447 |
178 (17) |
195 (18) |
35 (22) |
43 (27) |
Erythropoietin |
2156 (88) |
2457 |
945 (88) |
928 (87) |
146 (90) |
137 (87) |
3-mo LDL cholesterol, mg/dl |
77.1 (36.5) |
2457 |
96.2 (33.7) |
56.9 (25.6) |
103.7 (36.2) |
56.3 (28.8) |
LDL change between 3 mo and baseline, mg/dl |
−22.4 (34.2) |
2457 |
−2.0 (22.9) |
−43.1 (30.6) |
−0.8 (25.9) |
−43.0 (31.1) |
For continuous variables, data are presented as mean (SD) except for C-reactive protein, for which we presented median [quartile 1, quartile 3]. For categorical variables, data are presented as N (percentage).
LDL changes over time in relation to treatment only have been described before (6). LDL changes over time did not differ in relation to phosphate trajectories, regardless of treatment arm (both P values for interaction ≥0.09) (Figure 1). Additionally, adjusting for baseline LDL did not change the results (both P values for interaction ≥0.10).
Figure 1.: No effect of phosphate trajectories on serum LDL changes over time, stratified by treatment arm, in the A Study to Evaluate the Use of Rosuvastatin in Subjects on Regular Hemodialysis: An Assessment of Survival and Cardiovascular Events trial. (A) Placebo. (B) Rosuvastatin. 95% CI, 95% confidence interval.
Association between Outcomes and Time-Dependent Serum Phosphate Variations in Relation to Treatment in the AURORA Trial
Modeling phosphate during the course of the trial using splines, we observed a protective effect of statin treatment on MACE and all-cause death (although only significant for all-cause death) in patients with lower phosphate levels (<5 mg/dl) (Figure 2, Supplemental Figure 4). No treatment effect was observed above 5 mg/dl. A similar pattern was observed when considering tertiles of serum phosphate during the course of the trial; the protective effect of statin treatment on MACE and all-cause death was found only in the lowest tertile, but the interaction was significant for all-cause death only (Table 2).
Figure 2.: Protective effect of rosuvastatin treatment on clinical outcomes for low values of time-dependent serum phosphate levels (<5 mg/dl). (A) Major adverse cardiovascular events (MACEs). (B) All-cause death. (C) Hazard ratios (HRs) for specific phosphate values. Models contained the interaction between statin treatment and nonlinear time-dependent serum phosphate levels and were adjusted for age, sex, smoking status, diabetes, cardiovascular history, mean BP, C-reactive protein, and time-dependent LDL changes. Serum phosphate was modeled using a restricted cubic spline, and median phosphate concentration of 5.29 mg/dl was used as a reference value.
Table 2. -
Interaction between linear time-dependent serum
phosphate and rosuvastatin for both major adverse cardiovascular events and all-cause death outcomes
Model and Effect |
Major Adverse Cardiovascular Events, n=692 Events |
All-Cause Death, n=1115 Events |
N Event per Group |
Hazard Ratio [95% Confidence Interval] |
P Value |
N Event per Group |
Hazard Ratio [95% Confidence Interval] |
P Value |
Without time-dependent LDL adjustment
|
Treatment effect for T1 time-dependent phosphate |
213 |
0.74 [0.57 to 0.97] |
0.03 |
406 |
0.74 [0.61 to 0.91] |
0.005 |
Treatment effect for T2 time-dependent phosphate |
238 |
1.17 [0.90 to 1.52] |
0.24 |
371 |
1.18 [0.96 to 1.45] |
0.11 |
Treatment effect for T3 time-dependent phosphate |
241 |
0.97 [0.76 to 1.24] |
0.81 |
338 |
1.04 [0.85 to 1.27] |
0.72 |
P value for interaction (cat. phosphate) |
|
|
0.06 |
|
|
0.005 |
Effect of time-dependent phosphate in placebo group |
357 |
1.07 [1.00 to 1.14] |
0.04 |
573 |
1.02 [0.97 to 1.07] |
0.53 |
Effect of time-dependent phosphate in rosuvastatin group |
335 |
1.16 [1.10 to 1.23] |
<0.001 |
542 |
1.11 [1.06 to 1.16] |
<0.001 |
P value for interaction (cont. phosphate) |
|
|
0.05 |
|
|
0.01 |
With time-dependent LDL adjustment
|
Treatment effect for T1 time-dependent phosphate |
213 |
0.69 [0.52 to 0.91] |
0.01 |
406 |
0.69 [0.56 to 0.84] |
<0.001 |
Treatment effect for T2 time-dependent phosphate |
238 |
1.07 [0.81 to 1.39] |
0.64 |
371 |
0.98 [0.79 to 1.22] |
0.88 |
Treatment effect for T3 time-dependent phosphate |
241 |
0.88 [0.68 to 1.15] |
0.36 |
338 |
0.88 [0.7 to 1.09] |
0.24 |
P value for interaction (cat. phosphate) |
|
|
0.07 |
|
|
0.04 |
Effect of time-dependent phosphate in placebo group |
357 |
1.07 [1.00 to 1.14] |
0.03 |
573 |
1.02 [0.97 to 1.07] |
0.52 |
Effect of time-dependent phosphate in rosuvastatin group |
335 |
1.16 [1.09 to 1.23] |
<0.001 |
542 |
1.1 [1.05 to 1.15] |
<0.001 |
P value for interaction (cont. phosphate) |
|
|
0.07 |
|
|
0.03 |
Major cardiovascular events included nonfatal myocardial infarction, nonfatal stroke, or death from cardiovascular causes. Models were adjusted for age, sex, smoking status, diabetes, cardiovascular history, mean BP, and C-reactive protein. Serum phosphate was considered as a time-dependent covariate in those models, meaning that a given patient could have multiple values of serum phosphate over the course of the trial. When considering time-dependent phosphate in tertiles, a given patient could then move from one tertile to another during the course of the trial. cat., categorical; cont., continuous.
We reanalyzed the same models with adjustment for sevelamer, vitamin D, calcium substitution use, and time-dependent albumin levels (dichotomized along a 3.5-g/dl cutoff) and compared them with current results. Such adjustments did not strongly affect the results; for instance, the P value of the interaction between phosphate and treatment changed from 0.07 (Table 2) to 0.10 (categorical) and from 0.07 to 0.01 (continuous) for MACE, and from 0.04 to 0.07 (categorical) and from 0.03 to 0.03 (continuous) for all-cause death.
Separating cardiovascular and noncardiovascular death, the effects of treatment on all-cause death could be driven by noncardiovascular death. The interaction between treatment and nonlinear phosphate variations observed during the trial was of marginal significance for noncardiovascular death (P=0.09) and nonsignificant for cardiovascular death (P=0.58) (Figure 3, Supplemental Figure 5).
Figure 3.: Weak protective effect of rosuvastatin treatment on noncardiovascular death for low values of time-dependent serum phosphate levels (<5 mg/dl), but no effect on cardiovascular death. (A) Cardiovascular (CV) death (n=558). (B) Non-CV death (n=452). (C) HRs for specific phosphate values. Models contained the interaction between statin treatment and nonlinear time-dependent serum phosphate levels and were adjusted for age, sex, smoking status, diabetes, CV history, mean BP, C-reactive protein, and time-dependent LDL. Phosphate was modeled using a restricted cubic spline with the median value of 5.29 mg/dl as the reference.
Association between Outcomes and Baseline Phosphate in Relation to Treatment in the AURORA Trial
Using baseline phosphate as tertiles adjusted for several comorbidities factors, including CRP, and with LDL time-dependent adjustment, we observed similar results to the analyses with time-dependent phosphate, as rosuvastatin had a protective effect only in the lowest tertile for both MACE (hazard ratio, 0.71; 95% confidence interval, 0.53 to 0.94) and all-cause death (hazard ratio, 0.69; 95% confidence interval, 0.56 to 0.86), although the interactions between baseline phosphate and treatment did not reach the significance threshold (Supplemental Table 2). Using baseline phosphate as a continuous variable, again the interactions were not significant (Supplemental Table 2), but results were similar to the time-dependent analyses with a stronger effect, if anything, of phosphate in the rosuvastatin group as compared with the placebo group.
Association between Outcomes and Time-Dependent Serum Phosphate Variations in Relation to Treatment in the 4D Trial
When trying to replicate the analyses in the 4D trial, we found a tendency toward a protective effect of statin treatment in patients with lower phosphate levels or in the lowest tertile for MACE (before 2.5 years of follow-up), but it was far from significant (Supplemental Figure 6, Supplemental Table 3). For MACE after 2.5 years of follow-up and all-cause death, there was no indication of a protective effect of statin treatment, regardless of how phosphate was considered (linear or nonlinear) or whether we adjusted for time-dependent LDL or not (Supplemental Figure 7, Supplemental Table 3). This may be due to the more limited number of patients and events in this trial. When separating cardiovascular and noncardiovascular death, we found no interactions between time-dependent phosphate and treatment (Supplemental Figure 8).
Discussion
Our post hoc analysis of the AURORA trial demonstrates that the treatment effect of statin on MACE and all-cause death was significant and protective in patients with low values of serum phosphate and gradually faded for higher phosphate levels (>5 mg/dl). The fact that this finding could not be reproduced in the 4D trial may be due to a more limited number of patients and events in the 4D trial.
The post hoc analysis of the AURORA trial is in agreement with the recently described mechanism of phosphate-induced HMG-CoA reductase activation in uremic animals (4,5). Therefore, this mechanism may also account for the vascular smooth muscle cell resistance to statins observed in patients with CKD. The phosphate-induced HMG-CoA reductase activation leads to a higher intracellular cholesterol production and, subsequently, a lower cell membrane expression of LDL receptors, limiting the action of statins (13). The use of higher doses of statins than those used in the AURORA or 4D trial might overcome the consequences of phosphate-induced HMG-CoA reductase activation, but this would be associated with a higher risk of side effects in patients on dialysis (14). Whether the use of phosphate binders in statin-treated patients will enhance the efficacy of statins in preventing cardiovascular events remains to be explored.
In the experiments by Zhou et al. (4), a correlation between hyperphosphatemia and atheroma burden was also observed in orally phosphate-loaded, nonuremic apoE knockout mice. Because serum phosphate levels within the normal range were also found to be associated with a higher cardiovascular events in the general population (15,16), further studies should be undertaken to examine the hypothesis that even mild elevation of serum phosphate could interfere with statin actions in people with normal kidney function.
Serum phosphate independently correlates with inflammation in CKD, and phosphate enhances vascular smooth muscle cell's proinflammatory cytokines release (17). Because statins are powerful modulators of inflammation, the possibility that part of the observed stimulatory effect of phosphate on HMG-CoA reductase activity in vascular smooth muscle cells is linked to a higher vascular inflammation should also be considered. However, in the study by Zhou et al. (4), hyperphosphatemia was not linked to a higher systemic inflammation in patients with CKD or in nonuremic mice. Moreover, in this post hoc analysis of AURORA, the gradual attenuation of statin treatment efficacy regarding MACE and all-cause death with increasing serum levels of phosphate was independent of serum CRP levels.
Our analyses on cardiovascular and noncardiovascular death could suggest that the effects of treatment on all-cause death could be driven by noncardiovascular death. The interaction between treatment and nonlinear phosphate variations observed during the trial was of marginal significance for noncardiovascular death (P=0.09) and nonsignificant for cardiovascular death (P=0.58) (Figure 3, Supplemental Figure 5). Therefore, the lack of clear significant results prevents us from having a definitive conclusion, and further studies are needed to confirm this result. However, we could not exclude that off-target effects of statins on inflammatory markers could explain this difference, although we adjusted for CRP.
Limitations of the study include the study nature as a post hoc analysis, including predominantly European patients with a limited number of patients with diabetes, and the lack of adjustment for not available nutrition-related factors but serum albumin. The fact that this finding could not be reproduced in the 4D trial may be due to a more limited number of patients and events in the 4D trial.
To conclude, the observation of limited treatment efficacy of statins in patients on dialysis may be due, at least in part, to a higher intracellular cholesterol production by hyperphosphatemia, possibly via a lower membrane LDL receptor expression. The use of higher doses of statins may overcome this hyporesponsiveness, but the risk of side effects will probably hamper this treatment approach. Normalizing serum phosphate levels may help to ameliorate the efficacy of statin treatment in patients on dialysis. We suggest that this hypothesis be tested in future clinical trials.
Disclosures
M. Essig reports consultancy agreements with MedinCell. B.C. Fellstrom reports honoraria for consulting or lecturing or research funding from Astellas, AstraZeneca, Bristol Myer Squibb, Calliditas, CSL Behring, Novartis, and Sandoz over the past 5 years, and also reports consultancy agreements with Alexion, Astellas, AstraZeneca, BMS, Calliditas, CSL Behring, Pharmalink, and Sandoz; ownership interest in Calliditas (<1%); research funding from Astellas, BMS, CSL Behring, Pharmalink, and Sandoz; honoraria from Alexion, AstraZeneca, BMS, Calliditas, Novartis, Roche, and Sandoz; a personal pending patent; and an advisory or leadership role for Alexion, Astellas, BioAnalogica AB, BioConcept AB, Calliditas, CSL Behring, Sandoz, and Transcutan AB. N. Girerd reports honoraria from AstraZeneca, Bayer, Boehringer, Lilly, Novartis, and Vifor outside the submitted work. Z.A. Massy reports public funding, grants to charities, travel, and accommodation support from Amgen; public funding, travel, and accommodation support from Sanofi-Genzyme; and grants from AstraZeneca, Baxter, FMC, the French Government, GSK, Lilly, MSD, and Outsuka outside this work. He reports research funding from Amgen, Baxter, Fresenius Medical Care, Genzyme-Sanofi, GlaxoSmithKline, Lilly, Merck Sharp and Dohme-Chibret, and Otsuka; government support for the Chronic Kidney Disease - Réseau Epidémiologie et Information en Néphrologie (CKD REIN) project and experimental projects; honoraria to the charities or for travel from Baxter and Genzyme-Sanofi; and serving in an advisory or leadership role for Journal of Nephrology, Journal of Renal Nutrition, Kidney International, Nephrology Dialysis Transplantation, and Toxins. P. Rossignol reports (outside this work) consulting for Bayer, G3P, Idorsia, and KBP; honoraria from Ablative Solutions, AstraZeneca, Bayer, Boehringer-Ingelheim, Corvidia, CVRx, Fresenius, Grunenthal, Novartis, Novo Nordisk, Relypsa Inc. (a Vifor Pharma Group Company), Sanofi, Sequana Medical, Servier, Stealth Peptides, and Vifor Fresenius Medical Care Renal Pharma; reports being the cofounder of CardioRenal; consultancy agreements with G3P; personal fees from Bayer, Boehringer Ingelheim, CinCor, Idorsia, KBP, NovoNordisk, Sanofi, Sequana Medical, Servier, and Vifor; research funding from Relypsa Inc. (a Vifor Pharma Group Company) and Vifor Fresenius Medical Care Renal Pharma; and serving as an European Society of Hypertension: “Hypertension and the Kidney” working group board member since 2016, an American Society of Nephrology Kidney Health Initiative workgroup board member: Understanding and Overcoming the Challenges to Involving Patients with Kidney Disease in Cardiovascular Trials, a work group board member Heart Failure Association (cardio renal and translational 2016–2020), a work group board member for the Eurecam European Renal Association–European Dialysis and Transplant Association (ERA-EDTA) 2021–2023, and a work group biomarkers board member HFA 2020–2022. C. Wanner reports consultancy agreements with Akebia, Bayer, Boehringer-Ingelheim, Gilead, GSK, MSD, Sanofi-Genzyme, Triceda, and Vifor; an Idorsia grant to the institution and a Sanofi-Genzyme grant to the institution; honoraria from Astellas, AstraZeneca, Bayer, Boehringer-Ingelheim, Chiesi, Eli-Lilly, FMC, Sanofi-Genzyme, and Shire-Takeda; and other interests or relationships with ERA-EDTA. F. Zannad reports steering committee personal fees from Amgen, Applied Therapeutics, Bayer, Boehringer, CVRx, Merck, and Novartis; advisory board and consultancy personal fees from Cardior, Cellprothera, Cereno Pharmaceutical, NovoNordisk, Owkin, and Vifor Fresenius; stock options at Cardior, Cereno Pharmaceutical, and G3Pharmaceutical; and being the owner and founder of the Global Cardiovascular Clinical Trialist Forum. He reports consultancy agreements with Applied Therapeutics, Amgen, AstraZeneca, Bayer, Boehringer Ingelheim, Cardior, Cereno Pharmaceutical, CVRx, Merck, and Vifor-Fresenius; ownership interest in Cardiorenal and CVCT; honoraria from Actelion, Amgen, Applied Therapeutics, AstraZeneca, Bayer, Boehringer Vifor-Fresenius, Boston-Scientific, Cardior, Cellprothera, Cereneo, Corvidia, CVRx, Janssen, Novartis, NovoNordisk, and Owkin; serving in an advisory or leadership role for Actelion, Amgen, Applied Therapeutics, AstraZeneca, Bayer, Boehringer Vifor-Fresenius, Boston-Scientific, Cardior, Cellprothera, Cereneo, Corvidia, CVRx, Janssen, Novartis, Novo Nordisk, and Owkin; and serving on the speakers bureau for Boehringer Ingelheim. All remaining authors have nothing to disclose.
Funding
None.
Acknowledgments
We thank all of the patients and health professionals participating in the AURORA and 4D studies.
Author Contributions
N. Girerd, Z.A. Massy, T. Merkling, P. Rossignol, and S. Wagner conceptualized the study; N. Girerd, Z.A. Massy, T. Merkling, P. Rossignol, and S. Wagner were responsible for data curation; N. Girerd, Z.A. Massy, T. Merkling, P. Rossignol, and S. Wagner were responsible for investigation; N. Girerd, Z.A. Massy, T. Merkling, P. Rossignol, and S. Wagner were responsible for formal analysis; M. Essig, B.C. Fellstrom, N. Girerd, Z.A. Massy, T. Merkling, P. Rossignol, S. Wagner, C. Wanner, and F. Zannad wrote the original draft; and M. Essig, B.C. Fellstrom, N. Girerd, Z.A. Massy, T. Merkling, P. Rossignol, S. Wagner, C. Wanner, and F. Zannad reviewed and edited the manuscript.
Data Sharing Statement
All data used in this study are available in this article. Because of the sensitive nature of the data collected for this study, requests to access the dataset from qualified researchers trained in human subject confidentiality protocols may be sent to the corresponding author.
Supplemental Material
This article contains the following supplemental material online at http://cjasn.asnjournals.org/lookup/suppl/doi:10.2215/CJN.12620921/-/DCSupplemental.
Supplemental Figure 1. Flow chart of the study in the AURORA trial.
Supplemental Figure 2. Evolution of serum phosphate over time in relation to treatment.
Supplemental Figure 3. Serum phosphate trajectories over time using latent class mixed models.
Supplemental Figure 4. Effect of time-dependent serum phosphate levels changes (nonlinear) on outcomes in relation to treatment in the AURORA trial.
Supplemental Figure 5. Effect of time-dependent serum phosphate levels changes (nonlinear) on outcomes in relation to treatment in the AURORA trial.
Supplemental Figure 6. Treatment effect on MACEs for different values of time-dependent serum phosphate levels (nonlinear) and in relation to time period in the 4D study.
Supplemental Figure 7. Treatment effect on all-cause death for different values of time-dependent serum phosphate levels (nonlinear) in the 4D study.
Supplemental Figure 8. Treatment effect on outcomes for different values of time-dependent serum phosphate levels (nonlinear) in the 4D study.
Supplemental Table 1. Comparison of different grouping models for serum phosphate trajectories.
Supplemental Table 2. Interaction between baseline phosphate and rosuvastatin for both MACE and all-cause death outcomes.
Supplemental Table 3. Interaction between linear time-dependent phosphate and atorvastatin for both MACE (by time period) and all-cause death outcomes in the 4D study.
References
1. Massy ZA, de Zeeuw D: LDL
cholesterol in CKD--To treat or not to treat? Kidney Int 84: 451–456, 2013
2. Shavadia JS, Wilson J, Edmonston D, Platt A, Ephraim P, Hall R, Goldstein BA, Boulware LE, Peterson E, Pendergast J, Scialla JJ:
Statins and atherosclerotic cardiovascular outcomes in patients on incident
dialysis and with atherosclerotic heart disease. Am Heart J 231: 36–44, 2021
3. Drüeke TB, Massy ZA: Atherosclerosis in CKD: Differences from the general population. Nat Rev Nephrol 6: 723–735, 2010
4. Zhou C, He Q, Gan H, Zeng T, Liu Q, Moorhead JF, Varghese Z, Ouyang N, Ruan XZ:
Hyperphosphatemia in chronic kidney disease exacerbates atherosclerosis via a mannosidases-mediated complex-type conversion of SCAP N-glycans. Kidney Int 99: 1342–1353, 2021
5. Hénaut L, Massy ZA:
Phosphate meeting
cholesterol-Consequences for cardiovascular disease in chronic kidney disease? Kidney Int 99: 1264–1267, 2021
6. Fellström BC, Jardine AG, Schmieder RE, Holdaas H, Bannister K, Beutler J, Chae DW, Chevaile A, Cobbe SM, Grönhagen-Riska C, De Lima JJ, Lins R, Mayer G, McMahon AW, Parving HH, Remuzzi G, Samuelsson O, Sonkodi S, Sci D, Süleymanlar G, Tsakiris D, Tesar V, Todorov V, Wiecek A, Wüthrich RP, Gottlow M, Johnsson E, Zannad F; AURORA Study Group: Rosuvastatin and cardiovascular events in patients undergoing hemodialysis. N Engl J Med 360: 1395–1407, 2009
7. Wanner C, Krane V, März W, Olschewski M, Mann JF, Ruf G, Ritz E; German Diabetes and
Dialysis Study Investigators: Atorvastatin in patients with type 2 diabetes mellitus undergoing hemodialysis. N Engl J Med 353: 238–248, 2005
8. Kuznetsova A, Brockhoff PB, Rune HBC: lmerTest: Tests in Linear Mixed Effects Models, 2020. Available at:
https://CRAN.R-project.org/package=lmerTest. Accessed January 21, 2022
9. Bates D, Mächler M, Bolker B, Walker S: Fitting linear mixed-effects models using Lme4. J Stat Softw 67: 1–48, 2015
10. Harrell Jr. FE: Rms: Regression Modeling Strategies, 2019. Available at:
https://hbiostat.org/R/rms/. Accessed January 21, 2022
11. Proust-Lima C, Philipps V, Liquet B: Estimation of extended mixed models using latent classes and latent processes: The R Package Lcmm. J Stat Softw 78: 1–56, 2017
12. Therneau TM, Grambsch PM: Modeling Survival Data: Extending the Cox Model, New York, Springer, 2000
13. Rader DJ: A new feature on the
cholesterol-lowering landscape. Nat Med 7: 1282–1284, 2001
14. Massy ZA, Ma JZ, Louis TA, Kasiske BL: Lipid-lowering therapy in patients with renal disease. Kidney Int 48: 188–198, 1995
15. Adeney KL, Siscovick DS, Ix JH, Seliger SL, Shlipak MG, Jenny NS, Kestenbaum BR: Association of serum
phosphate with vascular and valvular calcification in moderate CKD. J Am Soc Nephrol 20: 381–387, 2009
16. Foley RN, Collins AJ, Herzog CA, Ishani A, Kalra PA: Serum phosphorus levels associate with coronary atherosclerosis in young adults. J Am Soc Nephrol 20: 397–404, 2009
17. Gross P, Six I, Kamel S, Massy ZA: Vascular toxicity of
phosphate in chronic kidney disease: Beyond vascular calcification. Circ J 78: 2339–2346, 2014