Background

Mineral and bone disorders (MBD) are early and common complications of CKD, and progress as glomerular filtration rate (GFR) declines. Multiple factors contribute to the development and maintenance of CKD-MBD, but principally involve phosphate retention and vitamin D metabolism abnormalities. Kidney Disease: Improving Global Outcomes defines chronic kidney disease-mineral and bone disorder (CKD-MBD) as a systemic syndrome characterized by abnormalities in serum calcium, phosphorus and parathyroid hormone (PTH) concentration, vitamin D metabolism, and bone turnover [1]. This syndrome is common among CKD patients and has been associated with an increased risk of cardiovascular calcification [2, 3] and mortality [4]. The Third National Health and Nutrition Examination Survey (NHANES III) reported 15068 adults patients with vitamin D deficiency and demonstrated a higher prevalence of cardiovascular disease and mortality in untreated patients [5]. An association between vitamin D deficiency and other traditional cardiovascular risk factors, such as hypertension, insulin resistance, diabetes, and dyslipidemia, has also been reported [6, 7]. The recognition of biochemical components of CKD-MBD associated with increased mortality in dialysis patients [8] and in patients with CKD not treated with dialysis [9] has provided an impetus to explore the effect of these factors on survival and associated treatment modalities. Numerous reports have characterized the nonskeletal benefits of vitamin D [10].

Wang et al. and Pittas et al. reported the benefits of vitamin D supplementation on cardiovascular disease (CVD) in the general population [11, 12]. Nutritional vitamin D supplementation has also been reported to be beneficial to CKD patients [13]. Most reviews, however, had few participants, short follow-up, and lacked survival analyses. We conducted a systematic review of the literature to assess whether vitamin D supplements reduced mortality in patients with ESRD on dialysis and patients with CKD not requiring dialysis.

Methods

Data sources and Search strategy

MEDLINE (1966 to March 2013), EMBASE (1980 to March 2013) and the Cochrance Controlled Trials Register (CCTR-Specialized Renal Registry) were searched. Relevant studies were identified [14, 15]. References from identified studies were reviewed to find additional relevant studies. This systematic review was planned, conducted, and reported following the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines [16].

Eligibility criteria

Studies were included in the meta-analysis if they met the following criteria: (1) cohort study design and follow-up duration was at least 1 year; (2) patients had chronic kidney disease or renal replacement treatment; (3) patients were treated with active vitamin D sterols (alfacalcidol, doxercalciferol, calcitriol, maxacalcitol, falecalcitriol and paricalcitol), but not native vitamin D (ergocalciferol and cholecalciferol); (4) the outcome of interest was all-cause mortality or cardiovascular mortality; (5) there was quantitative data (i.e., events rates, risk ratio [RR] or hazard ratio [HR]). If data were duplicated in more than 1 study, we included the study with the largest number of patients.

Data extraction

All data were independently abstracted by 2 investigators (Z.F.Z. and H.L.S) using a standardized data collection form. Discrepancies were resolved through discussion with other investigators (D.L. and J.Y.J.) and through reference to the original articles.

Quality assessment

Two authors (Z.F.Z. and H.L.S.) independently evaluated the quality of each study using the 9-star Newcastle-Ottawa Scale (NOS) [17]. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist for cohort studies was used to limit heterogeneity resulting from differences in study design [18]. Disagreements were resolved by consensus.

Statistical analysis

Studies that provided relative risk (RR) or hazard ratios (HR) were used directly in the pooled meta-analysis calculations. Overall crude (unadjusted) HR and adjusted HR were calculated. Adjusted variables included demographic and clinical values, biochemical indices and erythropoietin and phosphate binder use. The overall pooled-effect estimates were calculated using DerSimonian & Laird random-effect models. The Q test was used to assess the presence of heterogeneity and the I2 index was used to quantify the extent of heterogeneity [19, 20]. I2 values of 25% or less indicated low heterogeneity, values near 50% indicated moderate heterogeneity, and values 75% or greater indicated high heterogeneity [21]. Publication bias was assessed using funnel plots for each outcome and ln(HR) was plotted against its standard error. The Begg rank correlation test was used to examine asymmetry of the funnel plot [22]. The Egger weighted linear regression test was used to examine the association between mean effect estimate and its variance [23]. If an asymmetric funnel plot was found, a contour-enhanced funnel plot was used to further explore the source of bias [24]. P<0.05 was considered statistically significant. All tests were 2-sided. All analyses were conducted using STATA version 12.0 (StataCorp, College Station, Texas).

Results

Literature search

Our initial literature search yielded 2483 citations. 2319 articles were excluded. The majority of these citations were excluded at the level of title or abstract review. There were 164 citations which were considered to be potentially eligible. 144 articles were excluded after reviewing the article. Excluded articles included 37 narrative reviews, 31 duplication studies, 23 without vitamin D treatment, 20 without survival outcome, 15 without survival outcome data, 9 systematic reviews or meta-analyses, 5 author replies, 2 comments, 2 editorials and 1 letter. Twenty studies were considered eligible to be included in the meta-analysis [25–44]. The overall search flow is presented in Figure 1.

Figure 1
figure 1

Selection process for studies included in the meta-analysis.

Full size image

Study characteristics

The characteristics of eligible studies are summarized in Table 1. Of the 20 included observational studies, eleven were prospective cohort studies. [26, 28, 31, 34–36, 39, 41–44]. The remaining 9 consisted of 6 historical cohort studies [25, 27, 29, 30, 32, 33] and 3 retrospective cohort studies [37, 38, 40]. Seventeen studies reported ESRD patients on dialysis [25–30, 32, 33, 35–39, 41–44] and three reported CKD patients not on dialysis [31, 34, 40]. Five studies compared calcitriol to no active vitamin D treatment [28, 31, 34, 38, 42], two studies compared paricalcitol to no active vitamin D treatment [33, 42] and four studies compared alfacalcidol to no active vitamin D treatment [26, 40, 41, 44]. Nine studies did not report the specific analogues used and compared active vitamin D compounds with no treatment [27, 29, 30, 32, 35–37, 39, 43]. Two studies compared the survival benefits of paricalcitol and calcitriol [25, 30]. Several sophisticated statistical models were used in these observational studies. Fifteen studies used a fixed covariate baseline Cox model [25, 26, 30–35, 37–41, 43, 44], two studies used a time-dependent Cox model [27, 42], and three studies used both Cox models [28, 29, 36]. Only 4 studies were confirmed by intention to treat (ITT) analysis [27, 30, 32, 34].

Table 1 Observational studies examining active vitamin D administration in patients with CKD or on dialysis
Full size table

Vitamin D and all-cause mortality

14 studies examined the association between active vitamin D treatment and crude all-cause mortality. Patients that received alfacalcidol had a 46% (HR, 0.54; 95% CI, 0.37-0.80) lower overall mortality risk compared to untreated patients. Calcitriol, paricalcitol and not otherwise specified active vitamin D treated patients had a 43% (HR, 0.57; 95% CI, 0.46-0.70), 27% (HR, 0.73; 95% CI, 0.62-0.87) and 36% (HR, 0.64; 95% CI, 0.57-0.72) lower overall mortality risk. Similar results were observed with the crude time-dependent Cox model. All-cause mortality risk with calcitriol, paricalcitol and not otherwise specified active vitamin D was 26% (HR, 0.74; 95% CI, 0.55-0.99), 39% (HR, 0.61; 95% CI, 0.58-0.64) and 30% (HR, 70; 95% CI, 0.63-0.79) lower, respectively, than that found patients without active vitamin D treatment (Figure 2).

Figure 2
figure 2

Pooled crude hazard ratio of all-cause mortality for vitamin D treatment vs. no treatment in CKD patients. (A) baseline Cox model; (B) time-dependent Cox model.

Full size image

Ten studies reported vitamin D intake and risk for all-cause mortality using an adjusted case mixed baseline model. The risk of all-cause mortality was reduced 39% (HR, 0.61; 95% CI, 0.50-0.73) with calcitriol and 14% (HR, 0.86; 95% CI, 0.83-0.90) with paricalcitol. Using the adjusted case mixed time-dependent Cox model, patients who received active vitamin D treatment had a survival benefit (HR, 0.71; 95% CI, 0.57-0.89) (Figure 3).

Figure 3
figure 3

Pooled case mixed adjusted hazard ratio of all-cause mortality for vitamin D treatment vs. no treatment in CKD patients. (A) baseline Cox model; (B) time-dependent Cox model.

Full size image

We pooled data for ESRD on dialysis patients and CKD not on dialysis patients. Three studies evaluated patients with CKD that were not on dialysis. The survival advantage was similar in both the crude model (HR, 0.61; 95% CI, 0.43-0.77) and the adjusted model (HR, 0.59; 95% CI, 0.35-0.99). Patients with ESRD on dialysis had less survival benefit in the adjusted model (HR, 0.80; 95% CI, 0.63-0.94) than in the crude model (HR, 0.65; 95% CI, 0.58-0.73) (Table 2).

Table 2 Pooled hazard ratio for ESRD on dialysis and CKD not on dialysis
Full size table

Vitamin D and cardiovascular mortality

Four studies reported the HR between active vitamin D treatment and cardiovascular mortality using a crude Cox model and five using an adjusted baseline Cox model. A significant survival advantage was found in patients receiving active vitamin D using an unadjusted analysis (HR, 0.41; 95% CI, 0.28-0.59) and an adjusted analysis (HR, 0.59; 95% CI, 0.41-0.86). Similar results were found with calcitriol and paricalcitol. The adjusted baseline Cox model analysis found the reduction of cardiovascular mortality with calcitriol and paricalcitol to be 37% (HR, 0.63; 95% CI, 0.50-0.79) and 57% (HR, 0.43; 95% CI, 0.29-0.63), respectively. There was no survival difference associated with alfacalcidol treatment (HR, 0.45; 95% CI, 0.14-1.47) (Table 3).

Table 3 Pooled hazard ratio for cardiovascular mortality in patients receiving vitamin D or no treatment
Full size table

Calcitriol vs paricalcitol and all-cause mortality

Three studies reported hazard ratios that compared calcitriol and paricalcitol treatment. The crude baseline Cox model found a survival advantage with paricalcitol treatment (HR, 0.80; 95% CI, 0.75-0.86). In contrast, the adjusted baseline Cox case mixed and malnutrition-inflammation-cachexia syndrome (MICS) model demonstrated a survival advantage with calcitriol treatment (HR, 0.95; 95% CI, 0.91-0.99) in (Table 4).

Table 4 Comparison of all-cause mortality with paricaltitol and calcitrol
Full size table

Vitamin D dosage and all-cause mortality

Three studies reported the relationship between active vitamin D dose and all-cause mortality. Calcitriol treatment was associated with a dose dependent decrease in all-cause mortality. There was no survival advantage when calcitriol dose exceeded 7 ug per week. A dose dependent response was not found with paricalcitol (Table 5).

Table 5 Vitamin D dosage and all-cause mortality risk
Full size table

Assessment bias and meta-regression analysis

A publication bias was identified using an Egger regression asymmetry test (β=−3.81, P=0.01) and a funnel plot (Figure 4). A contour-enhanced funnel plot was used to explore the source of the bias. The contour-enhanced funnel plot demonstrated that the majority of studies had a high statistical significance. Therefore, publication bias was a less likely cause of the funnel plot asymmetry (Figure 4).

Figure 4
figure 4

Funnel plot and contour-enhanced funnel plot used to explore the source of publication bias. (A) funnel plot; (B) contour-enhanced funnel plot.

Full size image

Within study heterogeneity was evaluated using covariate meta regression analysis. Of the seven covariates, publication year (t=−2.19, P=0.049) and study participants (t=2.52, P=0.027) had the greatest between study variance. The proportion of within-study variance explained by publication year and study participants was 24.14% and 36.20%, respectively (Figure 5).

Figure 5
figure 5

Meta-regression graph of hazard ratio for all-cause mortality in vitamin D treated vs. no treatment patients. (A) meta-regression by publication year; (B) meta-regression by number of study patients.

Full size image

Discussion

Active vitamin D compounds were associated with a reduced risk of mortality in patients with ESRD on dialysis and patients with CKD not requiring dialysis. Several mechanisms may explain how vitamin D can modify risk for mortality. Vitamin D down regulates the renin-angiotensin system [45], improves insulin secretion and sensitivity [46], inhibits vascular smooth-muscle cell proliferation [47], protects normal endothelial cell function [48], modulates inflammatory processes [49], inhibits anticoagulant activity [50], and inhibits myocardial cell hypertrophy and proliferation [51]. These findings suggest that vitamin D may decrease mortality through multiple pathways. Although the actual mechanism of mortality is unclear, patient death has been associated with vascular calcifications, left-ventricular hypertrophy and left-ventricular dysfunction. The multi-organ protective effects of vitamin D may explain the lower mortality rate found in these patients.

A fixed covariate baseline Cox model was used in the majority of included studies. Only 5 studies used a time-dependent Cox model to analyze the relationship between active vitamin D use and survival. Although a standard baseline Cox proportional regression model is usually used to analyze cohort studies, it may be inadequate to evaluate active vitamin D treatments due to the presence of time-dependent variation in outcome. Higher serum calcium and phosphorus levels were consistently associated with increased risk of death [4, 52]. Elevated serum PTH levels have also been associated with increased mortality [4, 33]. The serum levels of calcium, phosphorus and PTH are affected by vitamin D therapy. Serum levels of calcium, phosphorus and PTH vary during the course of vitamin D therapy and affect patient outcome. These mineral metabolism indexes are recognized as time-dependent confounders. Time-dependent confounders cannot be controlled by conventional survival analysis methods [53]. Marginal structural modeling (MSM) can control for time-dependent confounders affected by prior treatment [54]. Under some conditions, the treatment estimate from a MSM can have the same causal interpretation as an estimate from a randomized clinical trial [55]. Only the Tentori et al. study reported detailed data regarding the survival advantage of patients treated with active vitamin D. The unadjusted baseline Cox model and time-varying MSM models demonstrated a 16% and 22%, respectively, reduction of all-cause mortality associated with active vitamin D treatment. Most studies included in this meta-analysis had some selection bias. For example, the study of Teng et al. [27] had statistical differences in the baseline characteristics of patient age, primary cause of renal failure, body mass index, blood pressure, and intact parathyroid hormone and hemoglobin levels. Several studies included in the meta-analysis used sophisticated statistical techniques, such as adjustment for time-dependent confounders, propensity score-matching or marginal structural models, to mimic the design of randomized controlled trials. Only the characteristics of patients that were treated with vitamin D analogues were known to the researchers. Any confounding factors would be controlled by these statistical methods and the results would be comparable to randomized controlled trials. The problem with the observational studies was that such knowledge was not available. The potential presence of unmeasured confounders prevented any conclusions of causation, even when sophisticated statistical methods were used. The survival advantage associated with active vitamin D treatment occurred in a dose-dependent manner. This phenomenon has been supported by two studies [29, 32]. There has been no well-designed dose gradient study to test this hypothesis. Although we do not have higher quality evidence to prove this association, we believe that vitamin D will improve survival.

The meta-analysis detected slight differences in survival associated with different analogues of active vitamin D. The baseline case mixed and MICS Cox models detected a 5% lower mortality with paricalcitol treatment than with calcitriol treatment. This slight survival difference may be explained by differential effects of calcitriol and its analogue, paricalcitol on vascular calcification. In vitro studies have demonstrated that calcitriol is a growth factor for vascular smooth muscle cells, while the analogue, paricalcitol, is not [56]. In vivo studies have shown that vitamin D sterols have a differential effect on vascular calcification. 1-α-hydroxy vitamin D (calcitriol) was associated with greater vascular calcification than paricalcitol, even though there was equivalent suppression of PTH in these animal models [57]. Only two well-designed cohort studies or randomized controlled trials, Teng et al. [25] and Tentori et al. [30], have evaluated the mortality risk associated with different active vitamin D analogues. Further studies are needed to clarify the survival difference before one vitamin D analogue is recommended over another in clinical practice.

Three studies included in the meta-analysis reported mortality risk associated with different mean daily or weekly doses of vitamin D. In the Naves-Diaz et al. study, the maximum reduction of mortality occurred when the mean daily dose of oral calcitriol was less than 0.25 ug. This survival benefit was lost as the mean daily calcitriol dose was increased to more than 1.0 ug. This dose-dependent benefit effect was also reported with paricalcitol. Kalantar-Zadeh et al. reported patients treated with mean weekly doses of 1.0 ug to 5.0 ug of paricalcitol. Mean weekly doses of paricalcitol above 15.0 ug were associated with an 18% reduction of mortality risk. A possible explanation is that low-dose vitamin D exerts weaker anti-vascular calcification effects than higher doses in CKD patients. High doses of vitamin D could be associated with adverse effects, such as hypercalcemia, that would overwhelm its protective effects.

There were several limitations in our meta-analysis. First, only a few of the included studies used a time-dependent or marginal structural model to analyze the follow-up data. The majority of studies had limited power to draw a definitive conclusion on the effects of vitamin D supplements on all-cause or cardiovascular mortality. Second, there was high heterogeneity in the meta-analysis. Sample size and publication year were the sources of heterogeneity. Third, the possible sources of heterogeneity could not be carefully examined. This included observational studies of the use of recombinant erythropoietin to correct anemia and studies of phosphorus binders to ameliorate hyperphosphatemia in patients with CKD that showed beneficial effects on mortality, CVD outcome, and progression of renal disease. Fourth, we did not seek to identify unpublished studies and several studies were excluded because the published data were not suitable for meta-analysis.

Conclusions

Active vitamin D compounds used to treat abnormal calcium, phosphorus and PTH levels in patients with either ESRD on dialysis or CKD not requiring dialysis. Active vitamin D compound treatment was associated with decreased all cause and cardiovascular mortality. Low dose active vitamin D compounds were associated with improved survival. Large, well designed randomized trials of active vitamin D supplements with different doses are needed to elucidate the role of vitamin D supplementation in reducing mortality.