Apolipoprotein B Attenuates Albuminuria-Associated Cardiovascular Disease in Prevention of Renal and Vascular Endstage Disease (PREVEND) Participants

Abstract

Whether urinary albumin excretion relates to higher levels of atherogenic apolipoprotein B fractions in the nondiabetic population is uncertain. Such a relationship could explain, in part, the association of elevated urinary albumin excretion with cardiovascular disease risk. We assessed the relationship of urinary albumin excretion with apolipoprotein B fractions and determined whether the association of elevated urinary albumin excretion with incident cardiovascular events is modified by high apolipoprotein B fraction levels. We performed a prospective study on 8286 nondiabetic participants (580 participants with cardiovascular disease; 4.9 years median follow-up time) with fasting lipids, apolipoprotein B, and urinary albumin excretion determined at baseline. With adjustment for sex and age, micro- and macroalbuminuria were associated with increased apolipoprotein B fractions (non-HDL cholesterol, LDL cholesterol, triglycerides, and apolipoprotein B). All four apolipoprotein B fractions modified associations of urinary albumin excretion with incident cardiovascular disease (hazard ratios for interaction terms ranged from 0.89 to 0.94 with 95% confidence intervals ranging from 0.84 to 0.99 and P values ranging from 0.001 to 0.02 by Cox proportional hazards modeling). These interactions remained present after additional adjustment for conventional risk factors, eGFR, cardiovascular history, and lipid-lowering and antihypertensive drug treatments. Such modification was also observed when urinary albumin excretion was stratified into normo-, micro-, and macroalbuminuria. We conclude that there is an association between elevated urinary albumin excretion and apolipoprotein B fraction levels and a negative interaction between these variables in their associations with incident cardiovascular events. Elevated urinary albumin excretion may share common causal pathways with high apolipoprotein B fractions in the pathogenesis of cardiovascular disease.

Robust associations of both microalbuminuria and macroalbuminuria with cardiovascular disease (CVD) are well established.^1–5 Elevated plasma levels of atherogenic apolipoprotein B lipoproteins (apoB fractions; i.e., LDL cholesterol [LDL-C], very LDL-C [collectively designated non-HDL], triglycerides, and apoB) are inherent features of overt proteinuria.^6–10 Interestingly, increased apoB fractions have been well established in microalbuminuric participants with types 1 and 2 diabetes mellitus^11–13 as well as participants with impaired fasting glucose¹⁴ and obesity.¹⁵ Less is known on the association between moderately elevated urinary albumin excretion (UAE) and abnormalities in apoB fractions in unselected populations. Equivocal associations of elevated UAE with higher total cholesterol and LDL-C, higher triglycerides, and higher numbers of small-sized LDL particles have been observed in population-based cohorts.^16–18 Thus, the extent to which elevated UAE is associated with consistent changes in various apoB fractions is still unclear.

If higher plasma apoB fraction levels were present in nondiabetic participants with elevated UAE, it would be clinically relevant to test whether the strength of the relationship of elevated UAE with cardiovascular (CV) risk is modified by concurrently high plasma levels of apoB fractions. Previous studies have already shown that the association of elevated UAE with incident CVD remains present, even when taking into account plasma total cholesterol.^3,19–21 Nonetheless, it has not yet been explored whether high plasma apoB fractions modify the association of elevated UAE with incident CVD. If such associations were to be attenuated in the context of high apoB fraction levels, then it would be consistent with the hypothesis that elevated UAE and increased apoB fraction levels share pathogenic mechanisms that may contribute to the development of clinically manifest atherosclerotic CVD.²²

In view of the ongoing debate as to whether a common causal pathway can be identified that underlies the relationship of microalbuminuria with atherosclerotic CVD,¹ we decided to first test the associations between UAE and various apoB fractions in nondiabetic participants. Then, we examined the extent to which the relationship of incident CVD with elevated UAE is modified by high plasma apoB fraction levels. To this end, we conducted an investigation on participants of the Prevention of Renal and Vascular Endstage Disease (PREVEND) study.

Results

Table 1 gives a summary of clinical and laboratory parameters for study participants stratified according to UAE as follows: normoalbuminuria (<30 mg/24 h), microalbuminuria (30–300 mg/24 h), and macroalbuminuria (>300 mg/24 h). Unadjusted ANOVA and chi-squared testing results showed significant differences among the UAE groups for all parameters in Table 1. Most parameters traditionally associated with CVD risk tended to be increased in the higher UAE groups, namely men, age, body mass index, systolic and diastolic BPs, creatinine, glucose, C-reactive protein (CRP), total cholesterol, non-HDL cholesterol (non–HDL-C), LDL-C, triglycerides, apoB, total cholesterol:HDL-C ratio, and apoB:apoA-I ratio, whereas eGFR, HDL-C, and apoA-I were found to be lower. Results of statistical comparisons between groups revealed generally little difference in parameters for the microalbuminuric and macroalbuminuric groups, whereas these parameters in both groups differed compared with the normoalbuminuric group. With adjustment for sex and age, all lipid and lipoprotein markers remained significantly associated with the UAE groups, except for apoA-I (Table 1); with additional adjustment for eGFR, all lipid and lipoprotein markers remained significantly associated with UAE, except for apoA-I and LDL-C, although the association for LDL-C approached statistical significance (Table 1).

Table 1.

Clinical covariates and biomarker levels as mean±SD or median (interquartile range) in the total study population stratified according to 24-hour UAE as normoalbuminuria (<30 mg/24 h), microalbuminuria (between 30 and 300 mg/24 h), and macroalbuminuria (>300 mg/24 h)

Parameter	Total Population	Normoalbuminuria (86.1% of Subjects)	Microalbuminuria	Macroalbuminuria	P Valuea	P Valueb	P Valuec
N	8286	7137	1036	113
Women, %	50.4	52.6	36.5	34.5	<0.001 (chi square)
Cardiac history, %d	4.2	3.2	10.5	9.7	<0.001 (chi square)
Statins, %	4.6	3.7	9.5	12.5	<0.001 (chi square)
Antihypertensives, %	15.7	13.5	27.3	40.9	<0.01 (chi square)
Smoking status, %
Never	29.5	31.0	21.1	15.3	<0.001 (chi square)
Former	36.2	35.1	41.4	55.9
Current	34.3	33.9	37.5	28.8
Age, yr	48.9±12.6	47.8±12.3	55.5±12.6	56.9±12.9	<0.001
Body mass index, kg/m2	26.00±4.17	25.72±4.01	27.61±4.69	28.62±4.47	<0.001	<0.001	<0.001
Systolic BP, mmHg	128±20	126±18	142±23	150±25	<0.001	<0.001	<0.001
Diastolic BP, mmHg	74±10	73±9	80±11	82±11	<0.001	<0.001	<0.001
UAE, mg/24 h	9.25 (6.27–16.9)	8.21 (5.96–12.4)	54.6 (38.6–90.4)	551 (379–1077)	<0.001	<0.001	<0.001
Creatinine, μmol/L	83.6±14.5	82.6±13.5	89.8±26.0	113.4±96.2	<0.001	<0.001	<0.001
eGFR, ml/min per 1.73 m2	84.0±15.3	85.1±14.8	78.3±16.9	67.9±21.8	<0.001	<0.001	<0.001
Glucose, mmol/L	4.74±0.66	4.70±0.63	4.99±0.75	4.99±0.78	<0.001	<0.001	<0.001
CRP, mg/L	1.25 (0.55–2.87)	1.13 (0.50–2.64)	2.17 (1.00–4.43)	2.48 (1.20–5.22)	<0.001	<0.001	<0.001
Cholesterol, mmol/L	5.63±1.12	5.60±1.12	5.85±1.12	6.13±1.35	<0.001	0.003	<0.01
Non–HDL-C, mmol/L	4.30±1.21	4.24±1.19	4.63±1.20	4.95±1.39	<0.001	<0.001	<0.001
LDL-C, mmol/L	3.67±1.05	3.64±1.04	3.89±1.05	4.10±1.13	<0.001	0.04	0.08
HDL-C, mmol/L	1.33±0.40	1.35±0.40	1.22±0.38	1.16±0.31	<0.001	<0.001	<0.001
Triglycerides, mmol/L	1.14 (0.84–1.66)	1.11 (0.82–1.60)	1.38 (0.98–2.07)	1.46 (1.08–2.17)	<0.001	<0.001	<0.001
ApoB, g/L	1.03±0.30	1.02±0.30	1.11±0.32	1.18±0.32	<0.001	<0.001	<0.001
ApoA-I, g/L	1.39±0.29	1.40±0.29	1.36±0.29	1.35±0.28	<0.001	0.32	0.28
Cholesterol/HDL-C	4.65±1.82	4.54±1.77	5.27±1.99	5.66±2.04	<0.001	<0.001	<0.001
ApoB/ApoA-I	0.77±0.27	0.76±0.26	0.85±0.29	0.89±0.28	<0.001	<0.001	<0.001

^aChi-squared test or unadjusted ANOVA for comparisons of continuous variables in the microalbuminuric versus the normoalbuminuric group and the macroalbuminuric versus the normoalbuminuric group revealed statistically significant differences (P<0.001) for all parameters except for apoA-I in the macroalbuminuric versus normoalbuminuric comparison (P=0.09). For the macroalbuminuric versus microalbuminuric comparison, significant differences resulted for smoking score (P=0.01), systolic BP (P<0.001), diastolic BP (P<0.01), UAE (P<0.001), creatinine (P<0.001), eGFR (P<0.001), cholesterol (P=0.01), and non–HDL-C (P<0.01).

^bANOVA adjusted for sex and age.

^cANOVA adjusted for sex, age, and eGFR.

^dCardiac history included hospitalization for myocardial infarction, revascularization procedures, or obstructive coronary artery disease.

Risk Associations and Interactions of UAE with Lipids and Lipoproteins

There were 580 participants with CV events in the study population (n=8286) identified with a median follow-up time of 4.9 (interquartile range=2.2–7.5) years. Of these events, 417, 143, and 20 events occurred in the normo-, micro-, and macroalbuminuric groups (corresponding to event rates of 5.84%, 13.80%, and 17.70%, respectively). To examine potential associations of UAE and lipids with CVD risk, Cox multivariable proportional hazards models adjusted for sex and age were formulated with risk as a function of UAE and the individual lipid and lipoprotein parameters shown in Table 1. Results (Table 2) showed significant risk associations for all lipid and lipoprotein parameters (direct associations for cholesterol, non–HDL-C, LDL-C, triglycerides, apoB, cholesterol:HDL-C ratio, and apoB:apoA-I ratio and inverse association for HDL-C and apoA-I).

Table 2.

Interaction of UAE with lipid parameters as assessed by Cox proportional hazards models adjusted for sex and age

Parameters	Model without Interaction			Model with interaction
	HR	95% CI	P Value	HR	95% CI	P Value
UAE	1.19	1.12 to 1.28	<0.001	1.25	1.17 to 1.35	<0.001
Cholesterol	1.27	1.17 to 1.37	<0.001	1.33	1.23 to 1.44	<0.001
Interaction				0.92	0.87 to 0.97	0.002
UAE	1.17	1.10 to 1.25	<0.001	1.25	1.16 to 1.35	<0.001
Non–HDL-C	1.37	1.27 to 1.48	<0.001	1.43	1.32 to 1.54	<0.001
Interaction				0.92	0.87 to 0.97	0.001
UAE	1.20	1.12 to 1.28	<0.001	1.25	1.16 to 1.35	<0.001
LDL-C	1.30	1.20 to 1.41	<0.001	1.36	1.25 to 1.49	<0.001
Interaction				0.94	0.87 to 0.99	0.02
UAE	1.17	1.10 to 1.25	<0.001	1.25	1.16 to 1.34	<0.001
Triglycerides	1.28	1.17 to 1.39	<0.001	1.37	1.25 to 1.49	<0.001
Interaction				0.89	0.84 to 0.95	0.001
UAE	1.20	1.12 to 1.28	<0.001	1.25	1.16 to 1.35	<0.001
ApoB	1.27	1.18 to 1.37	<0.001	1.30	1.21 to 1.40	<0.001
Interaction				0.93	0.88 to 0.98	0.01
UAE	1.18	1.10 to 1.26	<0.001	1.22	1.13 to 1.32	<0.001
HDL-C	0.71	0.63 to 0.79	<0.001	0.69	0.61 to 0.77	<0.001
Interaction				1.07	0.98 to 1.16	0.12
UAE	1.21	1.13 to 1.29	<0.001	1.21	1.12 to 1.30	<0.001
ApoA-I	0.76	0.70 to 0.84	<0.001	0.76	0.69 to 0.84	<0.001
Interaction				1.00	0.94 to 1.07	0.92
UAE	1.19	1.11 to 1.26	<0.001	1.17	1.08 to 1.26	<0.001
Cholesterol/HDL-C	1.27	1.22 to 1.32	<0.001	1.28	1.22 to 1.34	<0.001
Interaction				1.01	0.98 to 1.05	0.46
UAE	1.19	1.11 to 1.28	<0.001	1.21	1.12 to 1.31	<0.001
ApoB/ApoA-I	1.40	1.31 to 1.50	<0.001	1.40	1.31 to 1.50	<0.001
Interaction				0.98	0.93 to 1.03	0.41

Interactions between UAE and the lipid and lipoprotein parameters were also examined by inclusion of corresponding interactions terms into the sex- and age-adjusted models. Results showed significant negative interactions (hazard ratio [HR], <1.00) for the apoB-associated fractions (cholesterol, non–HDL-C, LDL-C, triglycerides, and apoB), which is also shown in Table 2. Interactions were nonsignificant for HDL-C and apoA-I as well as the cholesterol:HDL-C ratio and the apoB:apoA-I ratio.

Interaction of UAE with ApoB Fractions

To further investigate resultant negative interactions of UAE and apoB-associated fractions on risk, individual Cox models as functions of cholesterol, non–HDL-C, LDL-C, triglyceride, and apoB levels as continuous variables were developed with the addition of adjustments beyond sex and age, including CV history, statin use, anti-hypertensive use, smoking, systolic BP, eGFR, and glucose. As a representation of apoA-I–containing particles, corresponding apoA-I modeling was also generated for comparison with the apoB parameters. For each of the apoB fractions (Table 3), significant negative interaction with UAE was maintained. Again, for apoA-I, the interaction term was nonsignificant. Negative interactions between UAE and representative apoB fractions (non–HDL-C, LDL-C, triglycerides, and apoB) are illustrated graphically in Figure 1. Figure 1 presents HR surfaces generated analytically from the corresponding Cox regression equations including interaction terms.

Table 3.

Interaction of UAE with apoB-associated parameters and apoA-I as assessed by Cox proportional hazards models adjusted for sex, age, CV history, statin use, anti-HTN use, smoking, systolic BP, eGFR, and glucose

Parameters	Model with Interaction
	HR	95% CI	P Value
UAE	1.10	1.00 to 1.19	0.05
Cholesterol	1.40	1.28 to 1.54	<0.001
Interaction	0.92	0.86 to 0.98	<0.01
UAE	1.10	1.00 to 1.20	0.05
Non–HDL-C	1.49	1.35 to 1.63	<0.001
Interaction	0.91	0.86 to 0.97	<0.01
UAE	1.07	0.98 to 1.17	0.13
LDL-C	1.40	1.27 to 1.54	<0.001
Interaction	0.96	0.89 to 1.01	0.10
UAE	1.11	1.01 to 1.21	0.03
Triglycerides	1.25	1.13 to 1.39	<0.001
Interaction	0.91	0.84 to 0.98	0.01
UAE	1.07	0.98 to 1.17	0.14
ApoB	1.28	1.18 to 1.39	<0.001
Interaction	0.94	0.88 to 1.00	0.06
UAE	1.02	0.94 to 1.12	0.61
ApoA-I	0.80	0.72 to 0.90	<0.001
Interaction	0.96	0.88 to 1.04	0.30

HTN, hypertensive.

Figure 1.

Plots illustrating negative interaction effects in terms of continuous variables (Table 2) of UAE with apoB fractions (non–HDL-C, LDL-C, triglycerides, and apoB). Plots give HRs for CV events adjusted for sex and age as a function of UAE and apoB fractions as follows: (A) non–HDL-C, (B) LDL-C, (C) triglycerides, and (D) apoB. HRs were derived analytically using Cox modeling results including UAE, apoB fraction, and the respective interaction term.

We next examined whether negative interactions of UAE with apoB fractions on risk would also be manifested in Cox models using stratified variables: UAE (<30, 30–300, and >300 mg/d) and apoB fractions (tertiles). In each case, participants with UAE<30 mg/d and lowest tertile of the respective representative apoB fraction served as reference groups. Table 4 gives corresponding HRs, 95% confidence intervals (95% CIs), P values, and numbers of participants in each UAE category. Results again showed relatively lower sex- and age-adjusted risk for participants with the highest levels of UAE and apoB fraction. Negative interactions of UAE with the apoB fractions for the stratified variable models are illustrated in the bar plots in Figure 2. For comparison, corresponding results for apoA-I are given in Supplemental Figure 1 and Supplemental Table 1.

Table 4.

HRs, 95% CIs, P values, and N values from Cox proportional hazards modeling of time to event as a function of stratified UAE (<30, 30–300, and >300 mg/d) and tertiles of apoB fractions (non–HDL-C, LDL-C, triglycerides, and apoB) adjusted for sex and age

Model	UAE (<30 mg/d)				UAE (30–300 mg/d)				UAE (>300 mg/d)
	HR	95% CI	P Value	N	HR	95% CI	P Value	N	HR	95% CI	P Value	N
Non–HDL-C
Tertile 1	1.00	—	—	2429	1.48	0.86 to 2.52	0.15	235	5.09	1.82 to 14.25	0.002	18
Tertile 2	1.34	0.96 to 1.86	0.08	2300	1.79	1.15 to 2.79	0.01	344	2.19	0.97 to 4.95	0.06	39
Tertile 3	2.47	1.82 to 3.37	<0.001	2191	2.65	1.81 to 3.89	<0.001	437	1.90	0.91 to 3.97	0.09	54
LDL-C
Tertile 1	1.00	—	—	2376	1.32	0.81 to 2.17	0.27	253	3.43	1.24 to 9.50	0.02	20
Tertile 2	1.08	0.79 to 1.49	0.62	2277	1.36	0.88 to 2.09	0.16	336	1.76	0.78 to 3.94	0.17	36
Tertile 3	1.93	1.44 to 2.59	<0.001	2199	2.22	1.53 to 3.22	<0.001	402	1.77	0.86 to 3.66	0.12	50
Triglycerides
Tertile 1	1.00	—	—	2447	1.37	0.81 to 2.31	0.23	228	3.28	1.19 to 9.06	0.02	16
Tertile 2	1.47	1.10 to 1.96	0.01	2323	1.84	1.24 to 2.75	0.003	316	3.56	1.63 to 7.78	0.001	39
Tertile 3	2.12	1.61 to 2.78	<0.001	2156	2.31	1.64 to 3.25	<0.001	472	1.58	0.77 to 3.22	0.21	56
ApoB
Tertile 1	1.00	—	—	2349	1.54	0.94 to 2.52	0.09	226	2.92	1.04 to 8.15	0.04	18
Tertile 2	0.99	0.73 to 1.36	0.97	2217	1.29	0.85 to 1.98	0.23	337	1.93	0.83 to 4.51	0.13	34
Tertile 3	1.83	1.37 to 2.43	<0.001	2125	1.95	1.34 to 2.83	<0.001	411	1.57	0.78 to 3.15	0.20	57

Reference group taken as UAE<30 mg/d and lowest tertile of respective apoB fraction.

Figure 2.

Plots showing negative interaction effects in terms of stratified variables (Table 3) of UAE with apoB fractions (non–HDL-C, LDL-C, triglycerides, and apoB). Plots give HRs for CV events adjusted for sex and age as a function of UAE (<30, 30–300, and >300 mg/d) and tertiles of apoB fractions as follows: (A) non–HDL-C, (B) LDL-C, (C) triglyceride, and (D) apoB. The reference group in each case was taken as UAE<30 mg/d and the lowest tertile of respective apoB fraction.

Mediation Analyses

Mediation analyses were performed to elucidate potential mechanistic effects of apoB fractions on the association of UAE with CVD outcomes. Thus, in Cox models of incident CVD as a function of UAE, changes in the coefficient of UAE (continuous) were estimated for models adjusted for apoB fractions separately. In a sex-, age-, and eGFR-adjusted model, adjustments for non–HDL-C, LDL-C, triglycerides, and apoB resulted (Table 5) in significant reductions in the coefficient of the UAE-incident CVD association of 18% (P=0.002), 9% (P=0.02), 17% (P<0.01), and 8% (P=0.001), respectively.

Table 5.

Mediation analysis showing changes in the coefficient relating the association of UAE with incident CVD risk in a Cox proportional hazards model adjusted for sex, age, and eGFR after separate entry of apoB fractions into the models

Mediation Modeling	Coefficient	P Value	Percent Coefficient Reduction	P Value
Model 1a	0.18 (0.11–0.25)	<0.001	NA
Model 1+non–HDL-C	0.15 (0.08–0.21)	<0.001	18 (7–30)	0.002
Model 1+LDL-C	0.16 (0.10–0.23)	<0.001	9 (2–16)	0.02
Model 1+triglycerides	0.15 (0.08–0.22)	<0.001	17 (5–29)	<0.01
Model 1+apoB	0.16 (0.10–0.23)	<0.001	8 (2–15)	0.01

NA, not applicable.

^aModel 1 is a base model consisting of Cox proportional hazards regression with time to incident CVD event as a function of UAE (continuous) with adjustment for sex, age, and eGFR.

Sensitivity Analyses

Sensitivity analyses were performed to assess the robustness of our findings. We addressed whether the enrichment of the PREVEND study with participants with higher UAE affected the findings of negative interactions of UAE with apoB fractions. Similar Cox analyses were, therefore, performed on a subsample (n=3338) of the PREVEND cohort representative of the general population²³ that was generated by removal of the microalbuminuric participants who were initially added for enrichment to the PREVEND cohort of this study. Essentially similar results were obtained with models including UAE and apoB fractions adjusted for sex and age, showing continued significance of HRs: UAE (P=0.002) with non–HDL-C (P<0.001), UAE (P=0.002) with LDL-C (P<0.001), UAE (P=0.002) with triglycerides (P<0.001), and UAE (P<0.01) with apoB (P<0.001). Inclusion of respective interaction terms between UAE and apoB fractions to base models showed comparable negative interaction terms with the respective apoB fractions, giving results for the interaction term with UAE as follows: non–HDL-C (HR, 0.89; 95% CI, 0.81 to 0.98; P=0.02), LDL-C (HR, 0.83; 95% CI, 0.73 to 0.95; P<0.01), triglycerides (HR, 0.90; 95% CI, 0.79 to 1.02; P=0.11), and apoB (HR, 0.79; 95% CI, 0.70 to 0.90; P<0.001).

Discussion

This study shows that elevated UAE is associated with four measures of plasma apoB fractions (non–HDL-C, LDL-C, triglycerides, and apoB) in analyses in which sex and age were taken into account. Although both UAE and each of these apoB fractions predicted CV outcome in sex- and age-adjusted analysis, we were also able to show that these apoB fractions modified the extent to which elevated UAE predicts incident CVD. In particular, there were negative interactions of each of these apoB fractions with elevated UAE on CV outcome, irrespective of elevated UAE being expressed as a continuous variable or stratified into normo-, micro-, and macroalbuminuria. Such effect modification by these apoB fractions on incident CVD remained present after additional adjustment for individual conventional CVD risk factors (smoking, systolic BP, and glucose), kidney function (eGFR and serum creatinine), a positive CVD history, and lipid-lowering and antihypertensive drug treatment. If elevated UAE and apoB fractions were totally unrelated factors that affect incident CVD, then it would be anticipated that there were no interactions on CV outcome.²² Additionally, mediation analysis also indicated that part of the association of elevated UAE with incident CVD was attributable to apoB fractions as mediating factors. We interpret these results to be consistent with our hypothesis that elevated UAE and increased apoB fraction levels share, to some degree, causal pathways that are involved in the pathogenesis of atherosclerotic CVD.

This population-based study revealed that elevated UAE confers consistent proatherogenic changes in plasma lipids and (apo)lipoproteins, including higher total cholesterol, LDL-C, non–HDL-C, triglycerides, and apoB. Remarkably, higher apoB fraction levels were already evident in conjunction with microalbuminuria, with only minor additional changes being observed between micro- and macroalbuminuria. Previous population-based studies variably showed slightly higher total cholesterol, LDL-C, and triglycerides.^16–18 We believe that the currently observed increases in the various apoB fractions with elevated UAE in nondiabetic participants emphasize the pathophysiologic relevance of elevated UAE per se for alterations in the metabolism of apoB lipoproteins. Our observations, therefore, raise the possibility that abnormalities in the metabolism of apoB lipoproteins, such as increased LDL production, impaired LDL clearance, or decreased very LDL catabolism, which are all well documented in overt proteinuria,^{7–9,24–27} may already be operative in conjunction with moderately elevated UAE.

Of potential interest, evidence has accumulated in support of the possibility that elevated UAE does not only represent an early sign of renal disease but also that it should be regarded as a marker of generalized endothelial damage.^1,28,29 Thus, it is plausible to propose that apoB-containing lipoproteins are more readily retained in the arterial wall in the context of a defective endothelial barrier,^11,29 which was more recently evidenced by a diminished glycocalyx barrier.^30,31 It is, furthermore, likely that atherosclerotic vascular diseases are also featured by enhanced tissue accumulation of advanced glycation end products (AGEs),^32–34 a process that includes glycation of apoB lipoproteins.³⁵ AGE accumulation in tissue may further aggravate vascular leakiness and precipitate tissue damage.³⁶ Of additional note, enhanced tissue accumulation of AGE, at least in diabetic subjects, is dependent on the degree of albuminuria.³⁷ Elevated UAE and increased apoB fractions could, therefore, act in concert in promoting the development of atherosclerosis.

Our choice for selecting non–HDL-C, LDL-C, triglycerides, and apoB as representative apoB fractions related to elevated UAE was driven by the statistically significant differences in these apoB fractions according to UAE groups after adjustment for sex and age. The relationship of elevated UAE with non–HDL-C, triglycerides, and apoB remained significant after additional adjustment for eGFR, the latter being an important covariate, because elevated UAE coincides with changes in kidney function^3,21 and mildly impaired kidney function is associated with atherogenic changes in apoB lipoproteins.³⁸ Although the association of LDL-C with elevated UAE did not reach formal statistical significance after additional adjustment for eGFR, all four fractions as well as total cholesterol were found to interact negatively with elevated UAE on incident CVD. From a clinical perspective, it is relevant that non–HDL-C and apoB levels are considered to be at least as good as LDL-C in CV risk prediction.^19,39–43

Several other methodological aspects and limitations of our study need to be considered. First, the PREVEND cohort is enriched with participants with microalbuminuria.⁴⁴ However, relevant bias caused by overrepresentation of participants with higher albuminuria is very unlikely, because a sensitivity analysis on a subsample of participants representative of the general population yielded similar point estimates of the negative interactions between UAE and the four apoB fractions. Second, reliable information regarding the use of lipid-lowering and antihypertensive medication was available at baseline but not during follow-up. Because changes in the relationship of UAE with apoB fractions in response to antihypertensive treatment cannot be excluded, they may have resulted in some underestimation in the effect modifying of apoB fractions on UAE-related CV risk.⁴⁵ Additionally, it can be envisaged that statin treatment initiated during follow-up would weaken the relationship of incident CVD with LDL-C, because statin administration results in a shift in the relationship of LDL-C with apoB to lower LDL-C for a given apoB concentration.^46,47 Third, although our study was focused on the association of apoB lipoproteins with UAE, additional analyses showed that neither HDL-C nor apoA-I were found to modify the relationship of incident CVD with elevated UAE.

In conclusion, this study shows that, in nondiabetic participants, elevated UAE is consistently associated with higher levels of non–HDL-C, LDL-C, triglycerides, and apoB taking into account sex and age. Furthermore, we showed that the increased CVD risk attributable to elevated UAE is attenuated in the context of concurrently high apoB fraction levels. It is, therefore, conceivable that elevated UAE and increased apoB fractions share, at some level, common pathophysiologic mechanisms that contribute to the pathogenesis of atherosclerotic CVD.

Concise Methods

Study Population

PREVEND is a large, general population-based prospective cohort study that was started in 1997 with the goal of investigating relationships of increased albuminuria with CV and renal disease. A complete description of the PREVEND study protocol has been presented previously.⁴⁴ In summary, a one-page questionnaire was sent by mail to all inhabitants of the city of Groningen, The Netherlands, ages 28–75 years old, of whom 48% responded (n=40,856). Exclusions included pregnant women and type 1 diabetic individuals. Enrichment of the cohort with microalbuminuric participants was achieved by inviting participation of all individuals having urinary albumin concentration≥10 mg/L (6000 of 7768 participated). In addition, a random sample of participants with urinary albumin concentration<10 mg/L was also invited (2592 of 3395 participated), resulting in a total of 8592 participants from the PREVEND cohort. The study population of this work (n=8286) was derived by exclusion of type 2 diabetic participants (defined by use of glucose-lowering drugs according to questionnaires or pharmacy data or an increased glucose concentration). The PREVEND study was approved by the Medical Ethics Committee of the University of Groningen. Informed written consent was obtained from all participants. The study adhered to the ethical principles set by the Declaration of Helsinki.

Measurements and Definitions

UAE was determined at baseline on entry into the study as the mean value of two 24-hour urine collections taken over 2 consecutive days. Urinary albumin concentration was measured by nephelometry (BNII; Dade Behring, Marburg, Germany), with lower limit of sensitivity of 2.3 mg/L and intra- and interassay coefficients of variation of 2.2% and 2.6%, respectively. UAE blood biomarker analyses were performed on plasma and serum venous samples from participants fasted overnight, with collection after 15 minutes of rest. Levels of creatinine, glucose, CRP, total cholesterol, HDL-C, triglycerides, apoB, and apoA-I were determined by standard techniques as described previously.^48,49 Interassay coefficients of variation for apoB and apoA-I analyses were <5% in each case. eGFR was determined using the serum creatinine-based Chronic Kidney Disease Epidemiology Collaboration equation.⁵⁰ Non–HDL-C was calculated as the difference between total cholesterol and HDL-C. LDL-C levels were estimated using the Friedewald equation.

Smoking status was assessed using a trilevel variable (never been a smoker, past smoker, and current smoker). Body mass index was calculated as weight divided by height squared (kilograms per meter²). BP was measured for 10 minutes at 1-minute intervals in the supine position using automatic instrumentation (Dinamap XL Model 9300; Johnson-Johnson Medical, Tampa, FL); reported values were the means of the last two recordings.⁵¹ Initial exclusion of type 2 diabetic participants was on the basis of subject medical history, use of glucose-lowering drugs, and/or fasting plasma glucose level>7.0 mmol/L (according to the 1997 American Diabetes Association criteria). Information on medication use was checked by the use of pharmacy-dispensing data from all community pharmacies in the city of Groningen, which covers complete information on drug use in 95% of PREVEND participants.¹⁹

Outcome

History of previous CVD (hospitalization for myocardial infarction, revascularization procedures, or obstructive coronary artery disease) was acquired from questionnaires. Incident outcome events included CV-related mortality and hospitalization. Outcome events (International Classification of Diseases [ICD], ninth revision and classification of interventions) were followed over time and included acute myocardial infarction (ICD-410), acute and subacute ischemic heart disease (ICD-411), coronary artery bypass grafting, and percutaneous transluminal coronary angioplasty. Hospitalizations for heart failure and stroke were not counted as outcome events, because it is known that, in general, these events show no or only weak associations with lipid parameters.^41,52–56 Survival time was defined as date of initial urine collection (1997–1998) to date of first CV event or December 31, 2008. If a person was lost to follow-up (396 of the overall cohort of 8592 participants), censoring date was date of removal from the municipal registry. Mortality data and causes of death were acquired by record linkage with the Dutch Central Bureau of Statistics. CV morbidity data were obtained from PRISMANT, the Dutch national registry of hospital discharge diagnoses.

Data Analyses

Statistica 10.0 (StatSoft, Inc., Tulsa, OK) was used for statistical and graphical analyses. Results are reported as means±SDs for normally distributed variables and medians (interquartile ranges) for non-normal variables (UAE, triglycerides, and CRP). ANOVA was used to assess differences among groups with Bonferroni correction for multiple comparisons. Non-normal variables were conditioned for analysis by logarithmic transformation. Chi-squared testing was used to assess differences among groups for categorical variables. Multivariable Cox–proportional hazards modeling was used to follow CV outcomes over time, with continuous independent variables standardized by transformation to distributions with means=0 and SDs=1. HRs are per 1 SD unit. Interactions between risk variables were assessed by inclusion of multiplicative interaction terms into base models. The proportional hazard assumption was verified by evaluation of correlation between survival time and scaled Schoenfeld residuals.

Illustration of interactions of UAE with apoB fractions (non–HDL-C, LDL-C, triglycerides, and apoB) for continuous variable models was performed by analytically graphing Cox regression equations (including interaction terms) to give plots of HR as a function of UAE and apoB fraction. Using apoB as an example, it would involve evaluation of the expression exp(β_UAE×UAE+β_apoB×apoB+β_interact×UAE×apoB). Results are presented as three-dimensional plots, showing HR as a surface extending over clinically relevant domains of UAE and apoB fraction. Interactions were also illustrated for stratified variable models. Risk variables were stratified into three levels of UAE as 30, 30–300, and >300 mg/d and apoB fractions as tertiles. Interactions were shown by calculating HRs of the resulting nine cells using as reference the cell with the lowest UAE and the lowest tertile of apoB fraction (HR, 1.00). HRs as a function of UAE and apoB fractions were plotted as a three-dimensional bar graph.

Mediation analysis was performed to estimate the change in the coefficient of the UAE versus incident CVD association when adjusted for apoB fractions (i.e., what percentage of the association between UAE and CV events could be explained by apoB fractions as mediators).^22,57 Sensitivity analyses were performed on a subsample of participants representative of the general population to assess the robustness of our findings regarding whether the enrichment of the PREVEND study with participants with higher UAE values affected findings of interactions between UAE and apoB fractions.

Two-sided P values <0.05 were considered statistically significant except for interaction terms, where a more conservative P value <0.10 was taken to indicate statistical significance as proposed by Selvin⁵⁸ and advocated by the Food and Drug Administration.⁵⁹

Disclosures

None.