Treating endothelial dysfunction with vitamin D in chronic kidney disease: a meta-analysis

Chronic kidney disease (CKD) is a worldwide health issue, affecting 10–15% of the population with high costs both for patients and society [1, 2]. The main reasons for death is not, however, end stage renal disease, but more often cardiovascular events [2, 3]. The acknowledgement of kidney dysfunction as a strong risk factor for cardiovascular (CV) disease is highlighted in the European Society of Cardiology (ESC) [4] and the Kidney Disease Improving Global Outcome (KDIGO) guidelines [5]. Even so, treatment options to affect outcome are few, and evidence of these on cardiovascular hard end points are sparse.

Most CKD patients suffer a pronounced vascular disease, with endothelial dysfunction from early stages [6], and later on a marked vascular stiffening and arterial calcification [1, 3]. The reasons are multifactorial though with emphasis on chronic inflammation and disturbances in the hormonal mineral bone disorder (MBD) axis, with vitamin D deficiency, secondary hyperparathyreoidism (hPTH), high phosphate, and FGF-23 levels and downregulation of Klotho [1‐3] [7].

Vitamin D has been shown to have anti-inflammatory and anti-oxidative properties ([8]). Vitamin D also down-regulates the expression of renin and has therefore gained interest as a possible treatment option in CKD [9, 10]. Meta-analyses from the last decade show that vitamin D affects residual albuminuria/proteinuria, on top of RAAS blockade [11‐14] probably due to anti-inflammatory effects, such as downregulation of the TGF-beta pathway [13], downregulation of renin expressing genes [9, 10], and based on synergy with the AT1-receptor [15]. Glucose metabolism is another interesting area, where one meta-analysis shows positive effect on glucose control by treatment [16].

There has been some concern that active vitamin D compounds might cause a deterioration of renal function. Zhang et al. [17] performed a meta-analysis that showed higher creatinine levels with treatment, but no effect on eGFR when measured without creatinine, interpreted as no real effect on eGFR but probably higher creatinine due to an altered creatinine metabolism with active compounds [12]. These results are in line with other meta-analyses investigating the same area [12, 14].

Meta-analyses on cardiovascular risk and mortality show effect of treatment with vitamin D in CKD patients in observational studies [18, 19]. One meta-analysis [20] investigated the effect on cardiovascular endpoints in controlled trials, but could not show any benefit of treatment. These results may be questioned however, since none of the included studies had these endpoints as à priori primary or secondary endpoints, and the study durations varied from 3 weeks to 2 years.

There is a lack of interventional studies of vitamin D in CKD with sufficient power to answer questions on hard endpoints. In the absence of hard endpoints, surrogate markers of cardiovascular risk have been used in interventional trials, such as flow mediated vasodilation (FMD) since endothelial dysfunction precedes manifest vascular disease [21, 22]. In vitro data support the notion of a direct effect of vitamin D on endothelial function, with decreased oxidative stress and augmented levels of eNOS [23‐25]. This makes endothelial function measured by FMD an interesting topic for a meta-analysis in the small, short duration studies performed in the area.

This meta-analysis of existing publications on intervention with vitamin D on measures of FMD shows significant effect on endothelial function, an important factor in vascular disease. Two studies using FMD as outcome were not included due to the lack of a control group. Chitalia et al. [23] showed positive effects using Cholecalciferol 300,000 IU, given as two doses at the beginning and at 8 weeks, in a 16 week duration trial. Kendrick et al. [32] compared cholecalciferol 2000 IU with calcitriol 0.5 μg daily for 6 months and did not detect any change in FMD.

There is a lack of interventional vitamin D studies with enough power to investigate hard endpoints. Instead, surrogate markers of cardiovascular risk have been used, such as pulse wave velocity and pulse wave-form analysis (PWV/PWA) and flow mediated vasodilatation (FMD) [21, 22, 33, 34]. Whereas PWV/PWA are complex measures of both beta-2 induced vasodilation [34], arterial stiffening and calcification [3], FMD is primarily a measure of the capacity of the endothelial cells to produce Nitic Oxide (NO) [35]. Since FMD is a measure of function and not structure, it is an earlier sign of vascular disease, and likely easier to affect by shorter duration of treatment. Even so, PWV and FMD are interrelated [36] and both are predictors of cardiovascular risk [6, 22, 33, 34, 37].

There seem to be a discrepancy between the results for PWV/PWA and FMD after treatment with vitamin D compounds to CKD patients. The reasons are probably many, but one important factor might be, that PWV/PWA is also a measure of arterial remodelling. Although probably due to inflammation in the first place [7], when established the structural changes in the vasculature are likely harder to affect and reverse. CKD patients have an accelerated remodelling with fibrosis and calcification [1, 3], and concerns have been raised [7] that we may intervene too late in the process in our attempts to ameliorate CKD associated vascular disease.

For these reasons we chose to use FMD as the only outcome. It is a more direct measure of NO availability, with clear antioxidant, anti-inflammatory and eNOS upregulating pathways for vitamin D actions [23‐25].

We found 14 studies measuring the effect of vitamin D on different aspects of vascular function in CKD. Of these, 10 studied the effect on PWV and or PWAix with 26–120 participants, duration of 8 to 44 weeks, with CKD stage 3–5 and various vitamin D compounds and doses. Three studies reported positive effects on PWV [38‐40], the rest did not detect any change after treatment. Two articles [41, 42] assessed iontophoresis by acetylcholine showing ameliorated microvascular function with treatment, though interpreted with caution due to the small number of patients and short duration. One study investigated reactive hyperaemia index, with no significant change in treated patients [43].

We chose a fixed model statistical analysis as our primary model. The reason was the few and small studies performed, which makes the results hard to generalize and thus indicated the use of a fixed model [27]. Even so, the random model performed as a secondary analysis also showed significant results of vitamin D intervention on FMD, which may allow generalization of our results.

There was, not surprisingly, a substantial heterogeneity in the fixed model. There were too few studies to perform a meta-regression of significant cofactors, but when the studies were inspected for clinical heterogeneity there were some important differences. The largest study [38] had the strongest effect size and the youngest population, originating from India, while the other populations were from western countries. In this study cholecalciferol was used as treatment in comparison with paricalcitol 1 to 2 μg in the other studies. This study also had the longest duration (16 weeks), and the highest Jadad score (5p). There was also a difference in baseline 25OH-vitamin D with significantly lower levels in the two studies with the strongest effect sizes [38, 44]. Theti et al. [45], who failed to show significant results, included only patients with diabetic nephropathy, in contrast to Kumar et al. [38] and Lundwall et al. [42] who excluded diabetics.

Even though the number of studies were too few for subgroup analyses, it is interesting to discuss the fact that the two studies that did not show positive effect of treatment with paricalcitol both used 1 μg [42, 45], in comparison with the others using 2 μg. One of them, Theti et al., [45] was also the only one investigating the effects on diabetic nephropathy. Highly interesting is also the fact that the study with the strongest effect size [38], used inactive treatment with cholecalciferol, to patients substantially younger than in the other studies, but in the same CKD stage (3–4).

Of note, there seem to be a tendency to less negative effects on calcium and phosphate metabolism with treatments using precursors of vitamin D [11‐14, 46]. This might be favourable in treating vascular disease, and is also supported by a study of Zoccali et al. [47]. They saw, in a sub-study of the PENNY trial [44], that the effect of paricalcitol on FMD was most pronounced in patients with no change in phosphate during the study, and abolished in patients with the highest rise in phosphate levels, indicating the importance of different aspects of CKD mineral and bone disorders. These results might imply the use of phosphate binders in combination with active vitamin D.

Even though our results are hard to generalize due to the small number of studies and patients included, we show favourable effects in both the fixed and the random model, suggesting benefits of vitamin D intervention on endothelial function. Our results also indicate that the highest impact is seen in younger patients, probably due to an earlier stage of disease, where vascular remodelling has not yet been established. It might also be more favourable with the precursor than with active treatment, especially since there seemed to be a need of high doses of active compounds for effects. This is possibly due to less increased calcium and phosphate levels with inactive treatment. There is still a great need for larger and longer studies on this topic, to a proper selection of CKD patients at earlier stages of their vascular disease, and with sufficient power to assess hard endpoints.