Lipoprotein(a) concentrations, apolipoprotein(a) isoforms and clinical endpoints in haemodialysis patients with type 2 diabetes mellitus: results from the 4D Study

Abstract

Background

High lipoprotein(a) [Lp(a)] concentrations and low molecular weight (LMW) apolipoprotein(a) [apo(a)] isoforms are associated with cardiovascular disease and mortality in the general population. We examined the association of both with all-cause mortality and cardiovascular endpoints in haemodialysis patients with diabetes mellitus.

Methods

This is a post hoc analysis of the prospective 4D Study (German Diabetes Dialysis Study) that evaluated atorvastatin compared with placebo in 1255 haemodialysis patients with type 2 diabetes mellitus (median follow-up 4 years). The association of natural logarithm-transformed Lp(a) concentrations (increment one unit) and apo(a) isoforms with outcomes was analysed by Cox proportional hazards regression. The influence of age (median 66 years) was evaluated by stratified survival analyses.

Results

The median baseline Lp(a) concentration was 11.5 mg/dL (IQR 5.0–41.8). A quarter of patients had at least one LMW apo(a) isoform. Increased Lp(a) concentrations were associated with all-cause mortality in the total group [hazard ratio (HR) 1.09 (95% CI 1.03–1.16), P = 0.004]. LMW apo(a) isoforms were only associated with all-cause mortality in patients ≤ 66 years [HR 1.38 (95% CI 1.05–1.80), P = 0.02]. The strongest association for Lp(a) concentrations and LMW apo(a) isoforms was found for death due to infection in patients ≤ 66 years [HR 1.39 (95% CI 1.14–1.71), P = 0.001; HR 2.17 (95% CI 1.26–3.75), P = 0.005]. Lp(a) concentrations were also associated with fatal stroke in patients ≤66 years of age [HR 1.54 (95% CI 1.05–2.24), P = 0.03]. Neither Lp(a) nor LMW apo(a) isoforms were associated with other atherosclerosis-related events.

Conclusions

High Lp(a) concentrations and LMW apo(a) isoforms are risk predictors for all-cause mortality and death due to infection in haemodialysis patients with diabetes mellitus. These associations are modified by age.

all-cause mortality, apolipoprotein(a) isoforms, cardiovascular disease, diabetes mellitus, infection, lipoprotein(a)

INTRODUCTION

Lipoprotein(a) [Lp(a)] is a highly genetically determined lipoprotein. It consists of a low-density lipoprotein (LDL) molecule and an additional glycoprotein named apolipoprotein(a) [apo(a)]. Apo(a) has a high homology with plasminogen and is highly polymorphic in size. Lp(a) plasma concentrations have a wide range between individuals ranging from 0.1 to >200 mg/dL. About 30–70% of the variance of Lp(a) concentrations is explained by the size of the apo(a) isoforms. These isoforms are determined by variation in the number of Kringle IV (K-IV) repeats (11–>50 copies) [1]. An inverse relationship exists between the size of apo(a) isoforms and Lp(a) concentrations. The pathophysiological role of Lp(a) is not fully elucidated. Increased Lp(a) concentrations and low molecular weight (LMW) apo(a) isoforms (with ≤22 K-IV repeats) are strongly associated with the development of atherosclerosis and may be one contributing cause of atherosclerosis [1–3].

Chronic kidney disease (CKD), glomerular filtration rate (GFR) and proteinuria considerably influence Lp(a) concentrations [4]. The concentrations start to rise in the earliest phase of renal impairment [5]. This increase was only detected in carriers of large apo(a) isoforms compared with isoform-matched controls in non-nephrotic kidney disease and in haemodialysis patients in whom a block in the catabolism of Lp(a) is assumed. This is different than patients who lose a large amount of proteins either in urine (nephrotic syndrome) or into the peritoneal cavity, as is the case in patients treated by peritoneal dialysis: an increase of Lp(a) was observed for all apo(a) isoform groups, which is probably caused by an overproduction of Lp(a) due to the enormous protein loss in these conditions [6].

Only a few large studies have examined the association of Lp(a) concentrations and apo(a) isoforms with clinical outcomes in dialysis patients. It became obvious in the 1990s that especially the LMW apo(a) isoforms seem to be an interesting predictor of cardiovascular events. Cross-sectional as well as longitudinal studies from our group demonstrated that LMW apo(a) isoforms are associated with a high risk for cardiovascular disease in haemodialysis patients [7, 8]. The CHOICE Study in incident dialysis patients reported that LMW apo(a) isoforms and high Lp(a) concentrations are associated with cardiovascular events, with a stronger association for LMW apo(a) isoforms. Only LMW apo(a) isoforms were related to total mortality in these patients [9, 10]. Smaller and older studies in dialysis patients showed that increased Lp(a) is a risk factor for fatal and non-fatal cardiovascular events including silent stroke and/or overall death [11–15].

Diabetes mellitus is the leading cause of end-stage renal disease (ESRD). It was one of the major surprises of recent years that very low Lp(a) concentrations in the general population are associated with an increased risk for diabetes mellitus [16–18]. Independent from that, it has been described that high Lp(a) concentrations and/or LMW apo(a) isoforms predict cardiovascular events and mortality in the subgroup of haemodialysis patients with diabetes mellitus [7–10].

In earlier studies in haemodialysis patients, we found that LMW apo(a) isoforms showed a strong association with atherosclerosis-related outcomes, particularly at a young age [7, 19]. This is in line with findings in the general population where Lp(a) was primarily a risk factor for early coronary artery disease (CAD) [20–22].

This is the first prospective study in haemodialysis patients with type 2 diabetes mellitus (T2DM) with the aim to (i) evaluate the association between Lp(a) concentrations and apo(a) isoforms with all-cause mortality and fatal and non-fatal endpoints and (ii) analyse whether these associations are modified by age.

MATERIALS AND METHODS

Study design and participants

The study design and characteristics of the 4D Study have been described in more detail previously [23]. Briefly, the 4D Study is a prospective, randomized controlled trial with the aim of evaluating the efficacy and safety of atorvastatin in 1255 patients with T2DM ages 18–80 years and <2 years on maintenance haemodialysis at baseline. Patients were recruited between March 1998 and October 2002 in 178 dialysis centres in Germany and randomly assigned, based on a double-blind design, to receive either atorvastatin (n = 619) or placebo (n = 636). Patients were regularly followed until death or censoring or until the end of the study in March 2004, with a median observation period of 4 years.

Endpoints

Death from all causes, death from cardiac causes, all cardiac events combined, all cerebrovascular events combined and combined cardiovascular events as well as death due to infection, sudden cardiac death, myocardial infarction and fatal stroke were considered as endpoints. All these were predefined endpoints of the original 4D Study that were centrally adjudicated by three members of the endpoint committee blinded to study treatment according to the criteria previously defined [24]. The primary endpoint was ‘combined cardiovascular events’; the others were secondary outcome measures of the original 4D Study. The following detailed definitions were used during the 4D Study: death from cardiac causes included fatal myocardial infarction, sudden cardiac death due to congestive heart failure, death due to coronary heart disease (CHD) during or within 28 days after an intervention and all other deaths ascribed to CHD. In addition to death from cardiac causes (see above), all cardiac events combined comprised non-fatal myocardial infarction and non-fatal percutaneous transluminal coronary angioplasty (PTCA) or coronary artery bypass graft (CABG) interventions. Combined cerebrovascular events included fatal and non-fatal ischaemic and haemorrhagic strokes as well as strokes of unclear type and transient ischaemic attack (TIA)/prolonged ischaemic neurological deficit (PRIND). Death from cardiac causes (see above), non-fatal myocardial infarction and fatal and non-fatal stroke were part of the combined cardiovascular events endpoint.

Data collection

A more comprehensive description of the baseline data collection was reported earlier [25]. Baseline comorbidities such as CAD and congestive heart failure and information on the duration of diabetes and dialysis treatment were reported by the patient's nephrologists.

Lp(a) concentrations and apo(a) isoforms

Lp(a) was quantified with a double-antibody ELISA, and apo(a) phenotyping was done by SDS agarose gel electrophoresis followed by immunoblotting using a specific monoclonal antibody against apo(a) [26]. Patients were categorized into low (LMW) and high molecular weight (HMW) isoform groups. The LMW group is defined by patients with at least one apo(a) isoform with 11–22 K-IV repeats; the HMW group included patients having only isoforms with >22 K-IV repeats. In the case of two detectable apo(a) isoforms, the smaller apo(a) isoform was used for categorization [5].

For all major analyses, we used Lp(a) measured in baseline samples. In addition, Lp(a) measurements were available in samples from the 6-month follow-up visit. We compared baseline and follow-up measurements to test whether atorvastatin treatment influenced Lp(a) concentrations. Measurements of Lp(a) concentrations and apo(a) isoforms were done at the Medical University of Innsbruck and all other laboratory measurements centrally at the Department of Clinical Chemistry, University of Freiburg, Germany. Measurements of Lp(a) and apo(a) isoforms and all additional blood parameters used in these analyses are based on blood samples collected before randomization and before the start of a dialysis session (except those from the 6-month follow-up visit).

Statistical analysis

Baseline characteristics of the study population are described according to quartiles of Lp(a). One-way analysis of variance or non-parametric Kruskal–Wallis tests were applied in the case of non-normally distributed variables for univariate comparisons of continuous variables between quartile groups of Lp(a). Categorical variables were compared using the χ² test. Non-parametric independent Wilcoxon tests were performed to compare differences in Lp(a) concentrations between baseline and the first follow-up ∼6 months later [=ΔLp(a)] among the two treatment groups.

Cox proportional hazards regression analyses were performed to calculate hazard ratios (HRs) for Lp(a) (one unit increment) and in separate models for LMW apo(a) isoforms with the HMW isoform group as reference. Lp(a) was transformed based on the natural logarithm (ln) for further analyses due to its right-skewed distribution. All Cox regression analyses were adjusted for age, gender, medication allocation (placebo or atorvastatin) and CHD at baseline, i.e. history of myocardial infarction, CABG, percutaneous coronary intervention and angiographically documented coronary artery disease. As hypoalbuminaemia is a known risk predictor for death in haemodialysis patients [27], we also performed analyses in which we considered albumin as an additional covariable. The proportional hazards assumption was tested by χ² test based on Schoenfeld residuals. Non-linear P-spline analysis was used to check for linearity of Lp(a) on the association with outcomes. We also tested the hypothesis of whether age was a possible effect modifier on the association of Lp(a) or LMW apo(a) isoforms with study endpoints. Therefore, we considered interaction terms of Lp(a) or LMW apo(a) isoforms with two age groups (defined by the median) in the Cox proportional hazards models. In a next step, we stratified the population by the median of age.

All statistical tests were considered to be statistically significant if P <0.05 and were performed with IBM SPSS Statistics for Windows, version 21.0 (IBM, Armonk, NY, USA) and R statistical software, version 3.0.1 (R Project for Statistical Computing, Vienna, Austria).

RESULTS

Baseline analysis

At baseline, Lp(a) concentrations and apo(a) isoforms were available for 1223 patients. The median baseline Lp(a) concentration was 11.5 mg/dL [interquartile range (IQR) 5.0–41.8], and 25% of patients had at least one LMW apo(a) isoform. The percentage of LMW apo(a) isoforms increased in parallel with the quartiles of Lp(a) concentrations (Quartile 1: 7.8%, Quartile 2: 15.4%, Quartile 3: 16.7%, Quartile 4: 59.5%; P > 0.001). Further baseline characteristics of patients according to quartiles of Lp(a) are shown in Table 1, and only triglycerides, LDL cholesterol, VLDL cholesterol and haemoglobin values were significantly different between the four quartiles.

Table 1.

Baseline characteristics stratified by quartiles of lipoprotein(a)

	Lipoprotein(a) quartiles				P-value
	Quartile 1	Quartile 2	Quartile 3	Quartile 4
Number of patients, n	306	306	305	306	–
Lipoprotein(a) range, mg/dL	≤4.99	5.00–11.53	11.54–41.78	>41.78	–
Low molecular apo(a) isoforms, n (%)	24 (7.8)	47 (15.4)	51 (16.7)	182 (59.5)	<0.001
Age, years	65.9 (8.8)	65.7 (8.1)	65.6 (8.2)	65.6 (7.9)	0.97
Female gender, n (%)	127 (42)	146 (48)	140 (46)	152 (50)	0.21
Body mass index, kg/m2	27.8 (4.7)	27.4 (4.6)	27.4 (5.3)	27.5 (4.8)	0.76
Atorvastatin treatment, n (%)	146 (48)	148 (48)	149 (49)	160 (52)	0.68
Smoker and ex-smoker, n (%)	123 (40)	119 (39)	131 (43)	123 (40)	0.77
Systolic blood pressure, mmHg	143 (22)	147 (23)	147 (22)	145 (20)	0.08
Diastolic blood pressure, mmHg	75 (11)	76 (12)	76 (11)	76 (10)	0.60
Time receiving dialysis, months (25th; 50th; 75th percentiles)	8.6 (6.6) (3.6; 6.5; 11.8)	8.8 (7.4) (3.0; 6.4; 12.1)	7.5 (6.1) (2.9; 5.2; 10.7)	7.8 (7.1) (2.7; 5.5; 11.0)	0.05
Duration of diabetes, years	17.3 (8.7)	18.4 (9.0)	18.4 (8.7)	18.3 (8.8)	0.38
Ultrafiltration volumea, kg	2.3 (1.2)	2.2 (1.2)	2.3 (1.1)	2.3 (1.3)	0.85
Baseline comorbidities
Congestive heart failure	103 (34)	117 (38)	94 (31)	119 (39)	0.12
Coronary heart diseaseb	86 (28)	95 (31)	91 (30)	90 (29)	0.89
Stroke/TIA	55 (18)	55 (18)	54 (18)	52 (17)	0.99
Laboratory parameters
Glycated haemoglobin (HbA1c), %	6.78 (1.41)	6.70 (1.20)	6.68 (1.18)	6.72 (1.23)	0.76
Albumin, g/dL	3.85 (0.28)	3.82 (0.30)	3.80 (0.31)	3.79 (0.32)	0.06
Sensitive C-reactive protein, mg/L (25th; 50th; 75th percentiles)	11.3 (24.6) (2.0; 4.7; 11.9)	10.1 (13.2) (2.5; 5.2; 11.9)	11.7 (20.7) (2.3; 5.2; 12.9)	10.9 (16.1) (2.5; 5.3; 13.3)	0.58
Phosphate, mmol/L	5.96 (1.44)	5.97 (1.68)	6.13 (1.70)	6.09 (1.63)	0.49
Total cholesterol, mg/dL	216 (43)	220 (44)	218 (39)	224 (44)	0.18
LDL cholesterol, mg/dL	123 (30)	127 (30)	129 (30)	124 (29)	0.03
HDL cholesterol, mg/dL	35.0 (11.9)	36.1 (13.1)	37.7 (15.0)	36.3 (12.5)	0.08
Triglycerides, mg/dL (25th; 50th; 75th percentiles)	292 (186) (160; 242; 373)	275 (183) (155; 229; 338)	230 (130) (140; 202; 296)	256 (153) (148; 218; 313)	<0.001
VLDL cholesterol, mg/dL (25th; 50th; 75th percentiles)	59 (36) (31; 52; 76)	57 (36) (32; 48; 73)	51 (28) (29; 46; 67)	64 (34) (38; 55; 82)	<0.001
Haemoglobin, g/dL	11.1 (1.4)	11.0 (1.4)	10.8 (1.3)	10.7 (1.4)	0.02

Values are provided as mean (standard deviation) or as 25th, 50th, 75th percentiles in case of non-normal distribution or as number of patients (%).

To convert cholesterol concentrations to mmol/L, multiply by 0.03. To convert triglyceride concentrations to mmol/L, multiply by 0.01. To convert haemoglobin to mmol/L, multiply by 0.62. To convert serum albumin concentrations to g/L, multiply by 10.

^aThe ultrafiltration volume was calculated based on body weight before and after dialysis at randomization.

^bCoronary heart disease defined by a history of myocardial infarction, coronary artery bypass grafting, percutaneous coronary intervention and angiographically documented coronary artery disease.

Impact of treatment groups on Lp(a) concentrations

We observed no significant difference in the median Lp(a) concentrations at baseline between the atorvastatin and placebo group tested by an independent Wilcoxon test (11.9 versus 11.1 mg/dL, P = 0.17). The change in Lp(a) concentrations between the investigations at baseline and ∼6 months later also did not differ between the atorvastatin and placebo group (median change of Lp(a) −0.43 versus −0.22 mg/dL, P = 0.44).

Prospective follow-up in the entire patient group

In the atorvastatin group, the mean follow-up of patients was 3.96 years and in the placebo group 3.91 years, respectively. Of the 1223 patients included in the analyses, 599 died, of whom 265 died from cardiac causes and 157 from sudden cardiac death. There were 123 deaths due to infection. In total, 439 cardiac and 143 cerebrovascular events were observed. The outcome of combined cardiovascular events (irrespective if a cardiac or cerebrovascular event occurred first) was seen in 454 patients.

The association of Lp(a) concentrations and LMW apo(a) isoforms with the main endpoints in the entire population was analysed by Cox proportional hazard regression (Table 2). The proportional hazards assumption was not violated. Non-linear P-spline analysis revealed a linear association of ln-Lp(a) with the major endpoint of interest, i.e. all-cause mortality. Increased ln-Lp(a) concentrations were associated with a higher risk to die during the observation period (HR 1.09, P = 0.004). After including albumin in the statistical model, the association was not markedly attenuated (HR 1.08, P = 0.015). In addition, the hazard to die from death due to infection was 1.23 per increase in one unit ln-Lp(a) (P = 0.002). Further adjustment for albumin did not change this association (HR 1.20, P = 0.006). A borderline significant association of ln-Lp(a) concentrations with fatal stroke was observed in the total group (HR 1.28, P = 0.05). There was no association for LMW apo(a) isoforms with death from all causes (HR 1.13, P = 0.21). This was also the case for death from infections as well as fatal stroke. We did not observe a sex-specific effect for the main endpoints either for Lp(a) concentrations or for LMW apo(a) isoforms.

Table 2.

Association of Lp(a) concentrations (logarithmically transformed) and LMW apo(a) isoforms with outcomes during the prospective follow-up

Endpoints	Number of events	Entire group (n = 1223)
		Association of ln-Lp(a)		Association of LMW apo(a) isoforms
		HR (95% CI)a	P-value	HR (95% CI)a	P-value
Death from all causes	599	1.09 (1.03–1.16)	0.004	1.13 (0.94–1.35)	0.21
Fatal infection/sepsis	123	1.23 (1.08–1.41)	0.002	1.14 (0.76–1.70)	0.52
Death from cardiac causes	265	1.04 (0.95–1.14)	0.39	1.16 (0.88–1.52)	0.29
Sudden cardiac death	157	1.06 (0.95–1.20)	0.30	1.18 (0.83–1.68)	0.36
All cardiac events combined	439	1.01 (0.94–1.08)	0.88	1.01 (0.81–1.26)	0.91
Myocardial infarction	192	0.96 (0.87–1.07)	0.51	0.78 (0.55–1.11)	0.17
All cerebrovascular events combined	143	0.99 (0.88–1.13)	0.93	1.05 (0.72–1.53)	0.79
Fatal stroke	38	1.28 (1.00–1.62)	0.05	1.19 (0.59–2.41)	0.62
Combined cardiovascular events	454	1.04 (0.97–1.11)	0.30	1.02 (0.83–1.27)	0.83

^aCox model adjusted for age, gender and coronary heart disease at baseline and medication (placebo or atorvastatin). LMW apo(a) isoform = 1, HMW apo(a) isoform = 0: reference group; increment for ln-Lp(a): one unit. Death from cardiac causes comprised fatal myocardial infarction, sudden cardiac death, death due to congestive heart failure, death due to coronary heart disease during or within 28 days after an intervention and all other deaths ascribed to coronary heart disease. All cardiac events combined were defined as death from cardiac causes (see above) and additionally non-fatal myocardial infarction and non-fatal PTCA or CABG interventions. Myocardial infarction comprised fatal and non-fatal events. Combined cerebrovascular events included fatal and non-fatal ischaemic, haemorrhagic and strokes of unclear type, and TIA/PRIND. Combined cardiovascular events included death from cardiac causes (see above), non-fatal myocardial infarction, fatal and non-fatal stroke. Small subcategories of events (n < 15) are not listed in this table.

Prospective follow-up stratified by age

In the next step, we stratified the population into two age groups by the median of 66 years. The association of Lp(a) concentrations and LMW apo(a) isoforms with all-cause mortality was only significant in the younger age group [HR 1.12, P = 0.02 for ln-Lp(a) and HR 1.38, P = 0.02 for LMW apo(a) isoforms] but did not reach significance in the older age group [HR 1.07, P = 0.08 for ln-Lp(a) and HR 0.93, P = 0.56, for LMW apo(a) isoforms, respectively] (Table 3). For the younger age group, these associations remained stable when further considering serum albumin (HR for ln-Lp(a) 1.11, P = 0.03 and HR for LMW apo(a) isoforms 1.40, P = 0.01). Associations of LMW apo(a) isoforms with all-cause mortality differed considerably between age groups. A formally tested interaction between the two age groups and LMW apo(a) isoforms was significant (P = 0.03). In addition, an association with fatal stroke was found for ln-Lp(a) in younger (HR 1.54, P = 0.03) but not in older patients (HR 1.05, P = 0.79).

Table 3.

Association of Lp(a) concentrations (logarithmically transformed) and LMW apo(a) isoforms with outcomes during the prospective follow-up stratified by the median of age (66 years) into younger and older age groups

Endpoints	Number of events		Association of ln-Lp(a)				Association of LMW apo(a) isoforms
	Age ≤66	Age >66	Age ≤66		Age >66		Age ≤66		Age >66
	n	n	HR (95% CI)a	P-value	HR (95% CI)a	P-value	HR (95% CI)a	P-value	HR (95% CI)a	P-value
Death from all causes	254	345	1.12 (1.02–1.23)	0.02	1.07 (0.99–1.16)	0.08	1.38 (1.05–1.80)	0.02	0.93 (0.72–1.19)	0.56
Fatal infection/sepsis	55	68	1.39 (1.14–1.71)	0.001	1.14 (0.95–1.36)	0.15	2.17 (1.26–3.75)	0.005	0.60 (0.32–1.15)	0.12
Death from cardiac causes	116	149	0.99 (0.86–1.14)	0.92	1.07 (0.95–1.21)	0.25	1.42 (0.96–2.11)	0.08	0.94 (0.64–1.37)	0.73
Sudden cardiac death	72	85	1.04 (0.87–1.24)	0.67	1.08 (0.92–1.27)	0.35	1.55 (0.95–2.53)	0.08	0.86 (0.51–1.45)	0.57
All cardiac events combined	226	213	0.99 (0.90–1.10)	0.89	1.01 (0.91–1.12)	0.83	1.23 (0.92–1.64)	0.17	0.78 (0.57–1.09)	0.14
Myocardial infarction	105	87	0.93 (0.81–1.09)	0.37	0.97 (0.83–1.14)	0.74	0.88 (0.55–1.39)	0.57	0.65 (0.37–1.13)	0.13
All cerebrovascular events combined	64	79	1.11 (0.92–1.34)	0.30	0.92 (0.78–1.09)	0.32	1.28 (0.75–2.18)	0.37	0.86 (0.50–1.48)	0.59
Fatal stroke	19	19	1.54 (1.05–2.24)	0.03	1.05 (0.75–1.47)	0.79	1.37 (0.53–3.52)	0.52	0.83 (0.28–2.50)	0.74
Combined cardiovascular events	219	235	1.04 (0.94–1.15)	0.50	1.03 (0.94–1.13)	0.56	1.19 (0.88–1.60)	0.25	0.84 (0.61–1.15)	0.27

^aCox model adjusted for age, gender and coronary heart disease at baseline and medication (placebo or atorvastatin). For a description of events, see footnote of Table 2.

The most pronounced association with endpoints was observed for LMW apo(a) isoforms and Lp(a) concentrations in the younger age group for death due to infection. Patients with an LMW apo(a) isoform had a 2.17 times higher risk for death due to infection compared with patients with HMW apo(a) isoforms (P = 0.005). The same was observed for increased ln-Lp(a) concentrations (HR 1.39, P = 0.001). No association was found for patients in the older age group [HR 1.14, P = 0.15 for ln-Lp(a) and HR 0.60, P = 0.12 for LMW apo(a) isoforms] (Table 3). The risk estimates remained unchanged for the younger age group when additionally adjusting for serum albumin [HR for ln-Lp(a) 1.38, P = 0.002 and HR for LMW apo(a) isoforms 2.21, P = 0.005]. Again, we found a significant interaction between LMW apo(a) isoforms and the two age groups in relation to death due to infection (P = 0.004).

DISCUSSION

The present study aimed to examine the impact of Lp(a) concentrations and apo(a) isoforms on fatal and non-fatal endpoints in patients on maintenance haemodialysis with T2DM. Key findings were as follows: (i) increased Lp(a) concentrations were associated with all-cause mortality and death due to infection in the total group, (ii) both increased Lp(a) concentrations and LMW apo(a) isoforms were associated with all-cause mortality and death due to infection, particularly in the younger age group, (iii) increased Lp(a) concentrations, but not LMW apo(a) isoforms, were related to fatal stroke in the younger age group.

In contrast to studies that included incident dialysis patients of various aetiologies [9], we found in the present study of haemodialysis patients with T2DM that only Lp(a) concentrations and not LMW apo(a) isoforms were a risk predictor for all-cause mortality. In younger patients, however, both Lp(a) concentrations and LMW apo(a) isoforms were associated with all-cause mortality, and particularly death due to infection. The finding that LMW apo(a) isoforms were most strongly associated with outcomes in the younger age group in haemodialysis patients supports previous findings from our group [7, 19]. One explanation for this observation in the 4D Study cohort could be a survival bias, as previously suggested [7]. This would mean that patients with LMW apo(a) isoforms are underrepresented in the older age group because of their unfavourable atherogenic risk profile that resulted in premature death. Furthermore, patients who survive with LMW apo(a) isoforms are more likely to have a more favourable risk profile considering other risk factors.

Since not only Lp(a) concentrations but also LMW apo(a) isoforms were associated with mortality and death due to infection, a causal relationship of Lp(a) with these endpoints in the younger age group might be supported: the transmitted apo(a) alleles are of lifelong persistence and also determine to a great extent whether a person is exposed to increased Lp(a) concentrations throughout their entire life. That is the case in persons with LMW apo(a) isoforms. Thus, this association is less likely to be influenced by reverse causation or confounding. This method, known as Mendelian randomization, can be viewed as one methodological approach that can help to further support a causal relationship between risk factors such as Lp(a) concentrations and outcomes [28] and was first applied for Lp(a) concentrations and apo(a) isoforms >20 years ago [29].

There is strong evidence that elevated Lp(a) concentrations are associated with an increased risk for mortality, especially due to cardiovascular causes, in the general population [30]. Interestingly, death due to infections has not yet been related to Lp(a) concentrations and/or apo(a) isoforms. Patients on haemodialysis and patients diagnosed with diabetes mellitus present with a high frequency of infections and states of chronic inflammation. One of the related mechanisms might be an impaired immune system [31–34]. Studies have shown that inflammation together with comorbidities such as diabetes mellitus leads to an increased risk for adverse clinical outcomes in haemodialysis patients [35–37]. In combination with a predisposing genetic background such as LMW apo(a) isoforms, the disadvantageous inflammatory status might lead to a higher incidence of fatal infection-related events compared with the general population. However, these findings have to be confirmed in future studies including patients with less severe multiple general diseases. Previous studies investigating an influence of inflammation, sepsis, infections and acute phase reactions and related phenotypes such as C-reactive protein on Lp(a) concentrations were quite controversial [38–42]. A recent study in almost 35 000 individuals from the general population observed that concomitant inflammation indicated by an increased C-reactive protein does not have a major effect on Lp(a) concentrations, if any [41]. On the other hand, elevated IL-6 concentrations were observed to be associated with increased Lp(a) concentrations, and IL-6 blockade by monoclonal antibodies inhibited apo(a) expression and Lp(a) synthesis [43].

Lp(a) has been associated with stroke in the general population [44]. A large meta-analysis of population-based studies related small apo(a) isoforms to ischaemic stroke [3]. Data on these associations in haemodialysis patients are sparse [15], although cerebrovascular mortality is known to be highly increased in these patients [45]. The marginal positive association of elevated Lp(a) concentrations with stroke and the non-significant finding for LMW apo(a) isoforms in the 4D Study might be due to either heterogeneity in stroke aetiologies [30] or, most likely, the small number of stroke events.

In the 4D Study cohort, we found no indication that Lp(a) concentrations or LMW apo(a) isoforms were associated with atherosclerosis-related endpoints. In the general population, myocardial infarction accounts for most of the deaths in patients diagnosed with diabetes mellitus [46], whereas sudden cardiac death is the major cause of death in haemodialysis patients with diabetes mellitus. Atherosclerotic processes do not seem to account for the high mortality rate in these patients [24, 47]. It has been reported that in CKD patients, plaque growth is slowed down over time when renal function becomes worse [48]. Thus, it might be argued that haemodialysis patients with diabetes mellitus die more frequently from events caused by other cardiovascular causes than from solely atherosclerosis-driven events (e.g. from congestive heart failure or sudden cardiac death rather than myocardial infarction) [47, 49]. Finally, our finding that increased Lp(a) concentrations and/or LMW apo(a) isoforms had a more pronounced effect on outcomes other than myocardial infarction might also be attributable to the rather different pathogenesis of cardiovascular disease in haemodialysis patients with diabetes mellitus. The reason why we did not find an association with atherosclerosis-related endpoints, in contrast to previous findings in dialysis patients [7–10, 19], might be additionally due to different CKD aetiologies of studied patients, prevalence rates of co-morbidities, ethnic differences, different endpoint definitions, the assay used for Lp(a) measurement and/or sample size. However, as discussed previously, the effect modification by age on the association of LMW apo(a) isoforms with outcomes is in agreement with findings from earlier studies in haemodialysis patients [7, 19].

An additional finding of our study was that treatment groups had no influence on Lp(a) concentrations, neither at baseline nor when considering differences in Lp(a) concentrations between the baseline investigation and ∼6 months later. Recent observations that PCSK9 inhibitors have a strong Lp(a)-lowering effect [50] and experimental studies [51] have reinforced the idea that Lp(a) is taken up by the LDL receptor. If that is indeed the case, one would expect also a significant effect of statins on Lp(a) concentrations due to the overexpression of LDL receptors. However, this was not the case in the 4D Study as well as in earlier studies [52, 53]. Even the opposite was observed, as already suggested a long time ago by Kostner et al. [52] and very recently by O'Donoghue et al. [54]: both studies observed an increase in Lp(a) concentrations on statins. This is in line with the results of the JUPITER study showing, for patients treated with rosuvastatin, a statistically significant positive shift in the overall Lp(a) distribution that was not observed in the placebo group [53, 55]. It might well be that PCSK9 targets other receptors besides the LDL receptor that are involved in Lp(a) metabolism.

Strengths and limitations of the study

The main strengths of the 4D Study include the long-term follow-up, the large number of patients and the high incidence of prespecified and centrally adjudicated endpoints [24]. It is a further strength of this study that apo(a) isoforms were measured in addition to Lp(a) concentrations, which helps to provide evidence for causality of the relationships based on the Mendelian randomization approach. The present study is limited by its design, because analyses are performed post hoc and are based on a selected cohort of haemodialysis patients from Germany suffering from T2DM. Whether the results can be replicated in other populations and diseases has to be evaluated.

CONCLUSION

Findings from the 4D Study suggest that elevated Lp(a) concentrations and LMW apo(a) isoforms are associated with an increased risk for all-cause mortality and death due to infection in haemodialysis patients with T2DM. These findings were modified by age. In these patients, Lp(a) and LMW apo(a) isoforms were not associated with atherosclerosis-related events such as myocardial infarction.

CONFLICT OF INTEREST STATEMENT

The authors declare that they have no conflicts of interest and no relevant financial disclosures. The results presented in this paper have not been published previously in whole or part, except in abstract format.

ACKNOWLEDGEMENTS

We thank all investigators and study nurses who participated in the 4D-Study (www.uni-wuerzburg.de/nephrologie), and the laboratory staff. The measurements and analyses of Lp(a) and apo(a) isoforms were supported by a grant from the ‘Standortagentur Tirol’ to F.K. The study was further supported by grants from the interdisciplinary center of clinical research at the University of Würzburg (Habilitationsstipendium Drechsler) and from the German Federal Ministry for Education and Research (BMBF01EO1004).