Introduction

Insulin resistance is described as a reduced biological effect at a given level of serum insulin and consequently results in hyperinsulinemia to maintain glucose homeostasis [1]. Altered glucose metabolism and insulin resistance are recognized at all stages of chronic kidney disease (CKD) and end-stage renal disease (ESRD) [2, 3]. In ESRD, insulin resistance is an independent non-traditional risk factor for cardiovascular mortality and is associated with protein energy wasting and malnutrition [4]. While the exact mechanism remains unclear, a post-receptor defect in the insulin receptor signaling pathway in skeletal muscle is the likely primary abnormality. Other suggested contributors include adipose tissue dysregulation, inflammation, anemia, metabolic acidosis, uremic toxins and vitamin D deficiency [4]. Of these, vitamin D requires particular attention because of the kidney’s intricate role in vitamin D metabolism, the high prevalence of vitamin D deficiency in CKD, the availability of safe vitamin D and its analogs and the potential pleiotropic effects of vitamin D which may include CKD-related insulin resistance [5, 6].

Animal studies have demonstrated improvement in insulin resistance with administration of vitamin D with both increased insulin sensitivity and insulin secretion being affected [7, 8]. The presence of vitamin D receptors on pancreatic beta cells leading to increased intracellular free calcium and thereby insulin secretion, and immune modulation that prevents beta-cell apoptosis are proposed non-classical mechanisms by which vitamin D may improve insulin resistance [9, 10].

Observational studies and small randomized controlled trials (RCTs) have evaluated the link between vitamin D deficiency and insulin resistance in CKD. These studies have shown both an improvement in glucose metabolism with vitamin D supplementation in ESRD patients [11–13] and no improvement [14]. To clarify this uncertainty, we conducted a systematic review and meta-analysis to obtain a summary understanding and effect size of the impact of vitamin D supplementation (both nutritional and active) on glucose metabolism in dialysis patients. To our knowledge, this is the first meta-analysis of its kind.

Subjects, materials and methods

Our search strategy was developed with the help of a medical librarian and included a search of PubMed, EMBASE and The Cochrane Library CENTRAL Register of Controlled Trials through Aug 31, 2013. Limits were preset to studies conducted in adult humans (18 years and older) and manuscripts published in the English language. The search strategy for all databases was built on MeSH terms for “vitamin D” and “renal dialysis” and “insulin resistance” or “blood glucose,” with related keywords in the Title/Abstract added to the search.

Two authors (H. S. and V. P.) conducted the search and reviewed all abstracts independently. Manuscripts of potential relevance were retained for a review of the full text. Additional publications were identified from citations of manuscripts, review articles and personal reference lists. Only original manuscripts available in full text were included. Authors were contacted when full texts were needed. Disagreements regarding final inclusion of a study were resolved by consensus or by a third author (M. L. M.).

Data were abstracted in accordance with eligibility criteria set a priori. Studies were included if (1) any vitamin D analog or derivative was administered as an intervention, by any route and for any duration, (2) study participants receiving the intervention were on hemodialysis (HD) or peritoneal dialysis (PD), (3) primary or secondary outcomes involved the measurement of fasting blood glucose or fasting serum insulin as surrogate measures of insulin resistance at baseline and the end of study. We accepted the following study designs: prospective or retrospective, RCTs or non-randomized intervention studies (NRIS) or observational studies. Additional data extraction using standardized abstraction forms included age, sex, weight/BMI, diabetes mellitus status, type of vitamin D, route, dose and duration, type and average duration (months) of dialysis, serum vitamin D, PTH, calcium and phosphate levels as well as randomization, blinding procedures and loss to follow-up.

Statistical analyses

We analyzed the pooled sample of RCTs separately from that of NRIS [15]. Meta-analyses were performed to quantify the change in mean levels of fasting serum glucose, fasting serum insulin and PTH. Most of the NRIS were reported as single-group studies with pre- and post-intervention results. Pre-effect size and post-effect size estimations require knowledge of pre–post correlations (ρ 12) that were not reported in the primary studies. Therefore, we analyzed these paired comparisons in sensitivity analyses under assumptions of low (ρ 12 = 0.2), moderate (ρ 12 = 0.5) and high (ρ 12 = 0.8) correlations.

A priori random-effects models were employed, and standardized mean differences (SMDs) with 95 % confidence intervals (CI) were generated for continuous outcomes using the Dersimonian–Laird model. The SMD is the difference in means between the two groups divided by study-specific standard deviation [16, 19]. The SMD value should be interpreted as the number of standard deviations between the means being compared and is independent of measurement scale [16, 19]. A negative SMD indicates lower levels, whereas a positive SMD indicates higher levels of outcome. Cohen’s rule of thumb guides interpretation of magnitude of effect size, SMD 0.2: small, SMD 0.5: moderate, SMD > 0.8: large [17].

Heterogeneity across studies was assessed by the Cochran Q statistic and I 2 statistic of measured inconsistency (the percentage of total variance across studies attributable to real differences between studies than by chance). The magnitude of heterogeneity was categorized as I 2 = 25 %: low, I 2 = 50 %: moderate and I 2 = 75 %: high [18]. Heterogeneity was expected given the wide variation in study design. Strategies to address heterogeneity included use of random-effects modeling that assumes both within-study and between-study variance, and sensitivity analyses excluding 1–2 studies with outlying effect sizes [19]. Funnel plots of effect size against study-level standard error were constructed using the Begg–Mazumdar method to evaluate publication bias. Risk of bias in RCTs was assessed by the tool provided by Cochrane Back Review Group [20]. Statistical significance was set at two-sided p value of 0.05 for all analyses. Statistical analyses were performed with Comprehensive Meta-Analysis software version 2.

Results

Figure 1 provides a summary of the search and manuscript retrieval for this review. The initial literature search yielded a total of 223 articles from PubMed and EMBASE; no new studies were identified from Cochrane CENTRAL. Of note, one paper suggested by personal reference was added to this review. This study was not retrieved by any database search [14]. The final systematic review was performed on 17 studies (Fig. 1) [11–14, 21–33].

Fig. 1
figure 1

Flow diagram of studies identified for systematic review and meta-analysis

Full size image

Study methodology

Tables 1 and 2 provide a summary of the reviewed studies. Most of the studies included in this review were small. Of the 17 studies, four were RCTs [14, 23, 28, 31]. While Mak [13] did not report a randomization procedure, HD patients were divided into treatment and placebo groups, and therefore, the study was included as an RCT. The remaining 12 studies were NRIS that also reported a control group of healthy volunteers who served as comparison for demonstrating improvement from baseline values in the HD group after vitamin D treatment [11, 12, 21, 22, 24–27, 29, 30, 32, 33].

Table 1 Descriptive characteristics of randomized controlled trials (RCTs) of vitamin D supplementation with insulin resistance as an outcome
Full size table
Table 2 Descriptive characteristics of non-randomized intervention studies (NRIS) of vitamin D supplementation with insulin resistance as an outcome
Full size table

Intervention

Vitamin D formulations varied widely, with the majority of the older studies employing calcitriol (Tables 1, 2). The duration and dose of vitamin D were also variable; most studies evaluated vitamin D effects after 4–12 weeks, though this ranged widely from the shortest duration being 2 h after intravenous calcitriol [11, 12], to the longest duration of 24 weeks (6 months) [30]. In the NRIS, healthy controls did not receive any intervention. In the RCTs, control groups received dihydrotachysterol [13], cinacalcet [14] or placebo [23, 28, 31].

Demographic characteristics

Baseline patient characteristics were not well reported especially in the older NRIS (Table 1). Age diversity was notable, but the mean age typically ranged from 40 to 50 years across most studies. While the eligibility criterion only included adult participants, we included two 1992 studies by Mak that specified the age range as 16–22 years. Sex was reported in seven studies, and the majority were male (>50 %). Diabetic patients were included in three studies [14, 30, 32]. The mean BMI reported in four studies was between 20.6 and 34.6 kg/m2 [14, 27, 29, 33]. Body weight was reported as within 115–120 % of ideal weight for height [11, 12, 24, 25]. Ibrahim et al. [32] studied the impact of cholecalciferol on outcomes in HCV seropositive and seronegative HD patients; for this review, we only used data from HCV seronegative patients for generalizability. Hung et al. [14] studied a cohort of African–American HD patients; race was not reported in the other studies.

ESRD and renal dialyses

Ulutas et al. [33] was the only study done in PD patients. The mean duration of HD, if reported, ranged from 23 to 47 months. The etiology of ESRD was variable; dietary restrictions were variably reported and typically were for fluids, sodium, potassium and phosphate but not protein. Most studies excluded patients if they had a history of acute or chronic inflammatory conditions, malignancy, were scheduled for or received a renal transplant, had abnormally elevated serum phosphate and calcium levels or took medications affecting liver function or glucose metabolism (except studies that included diabetics). Participants in most studies were on phosphate binders (calcium carbonate or aluminum hydroxide).

Baseline secondary hyperparathyroidism and vitamin D status

The severity of secondary hyperparathyroidism varied across the studies, with baseline mean intact PTH (iPTH) levels ranging from mildly (<400 pg/ml) to severely elevated (>1,000 pg/ml) (Tables 3, 4). Exposure to medications to treat secondary hyperparathyroidism was variable and reported only in the Blair et al. [30] (cinacalcet) and Hung et al. [14] (paricalcitol, cinacalcet) studies. Participants in Ulutas et al. [33] study continued to take 1-α-calciferol for secondary hyperparathyroidism along with the intervention [33]. Other studies had protocols, where participants stopped taking vitamin D supplements 2–4 weeks prior to the start of the study [11–13, 26, 32], or were never on vitamin D supplementation [22, 31]. Of note, in the two Mak studies [11, 12], participants discontinued oral calcitriol only 3 days prior to study intervention. Baseline vitamin D levels were variably reported (Tables 1, 2).

Table 3 Baseline and post-treatment glycemic parameters for randomized controlled trials (RCTs)
Full size table
Table 4 Baseline and post-treatment glycemic parameters for the non-randomized intervention studies
Full size table

Outcomes and meta-analyses

Short-term vitamin D supplementation was associated with decreased fasting serum glucose levels in three of five RCTs, whereas NRIS demonstrated non-significant decrements; fasting serum insulin outcomes were variable (Tables 3, 4). Fasting plasma glucose values were fairly normal in most studies, especially in the NRIS (Table 3, 4). Improved hemoglobin A1C (HbA1C %) values were also observed (Tables 3, 4) [21, 23, 31, 32]. There was greater evidence for the improvement of glucose-stimulated insulin secretion based on hyperglycemic clamp testing or glucose tolerance testing (Table 4) [11, 12, 23–26]. Similar results were obtained with testing for HOMA-IR, except in the Hung et al. [14] study (Tables 3, 4) [14, 31–33]. Changes in serum PTH and calcium levels were associated with changes in the primary outcome and also noted independently.

For the meta-analyses, we excluded the Blair et al. study to avoid bias from the large attrition between the pre- and post-treatment groups (Table 4) [30]. Meta-analyses of RCTs showed significantly decreased SMD of post-treatment fasting serum glucose levels [SMD −1.13, 95 % CI (−2.11 to −0.11); p value 0.03; heterogeneity: I 2 = 83.03], non-significant increase in fasting serum insulin values [SMD 1.32, 95 % CI (−0.15 to 2.79); p value 0.08; heterogeneity: I 2 = 87.47] and significant decrease in PTH levels [SMD −1.50, 95 % CI (−2.95 to −0.04); p value 0.04; heterogeneity: I 2 = 90.03] when compared to the placebo group (Table 3; Fig. 2a, b, c). Meta-analyses of pooled NRIS showed significant improvement only in serum PTH levels after vitamin D treatment: fasting serum glucose levels [SMD −0.22, 95 % CI (−0.51 to 0.07) p value 0.14; heterogeneity: I 2 = 57.80], fasting serum insulin levels [SMD 0.45, 95 % CI (−0.17 to 1.08) p value 0.16; heterogeneity: I 2 = 88.23] and serum PTH levels [SMD −1.68, 95 % CI (−2.55 to −0.82) p value <0.001; heterogeneity: I 2 = 88.52] (Table 4; Fig. 3a, b, c). Sensitivity analyses at different levels of correlation coefficient (0.2, 0.8) showed similar results (data not shown). In RCTs and NRIS that demonstrated improvement in glucose levels, baseline fasting glucose levels were elevated compared to studies that did not show significant change (Tables 3, 4).

Fig. 2
figure 2

Forest plot of pooled randomized controlled trials to evaluate the summary effect (standardized mean difference) of the vitamin D intervention group compared to the placebo group on fasting serum glucose (a), fasting serum insulin (b), serum PTH levels (c)

Full size image
Fig. 3
figure 3

Forest plot of pooled non-randomized intervention studies (NRIS) to evaluate the summary effect (standardized mean difference) comparing pre- and post-vitamin D fasting serum glucose levels (a), fasting serum insulin levels (b), serum PTH levels (c)

Full size image

Heterogeneity and sensitivity analyses

The Q values and I 2 values revealed significant heterogeneity as expected (Table 5). There were too few RCTs for sensitivity analyses. For NRIS, after removing three studies with outlying effect sizes [12, 26], with a very different study methodology [11, 12] or with additional exclusion of the trial of PD patients [33], heterogeneity was reduced significantly but the summary effect size for serum glucose levels did not change overall (SMD −0.18, 95 % CI (−0.40 to 0.04) p value 0.11, heterogeneity I 2 = 13.4); similar sensitivity analyses did not affect summary effect sizes for serum insulin levels (SMD 0.58, 95 % CI (−0.21 to 1.38) p value 0.15; heterogeneity I 2 = 89.00) or serum PTH levels (SMD −1.53, 95 % CI (−1.96 to −1.07) p value <0.001; heterogeneity I 2 = 90.08).

Table 5 Summary effect size [standardized mean difference (SMD)] from meta-analysis of randomized controlled trials (comparing intervention group to placebo group) and non-randomized intervention studies (comparing pre- and post-values)
Full size table

Risk of bias assessment

Most of the RCTs were single-blinded studies with insufficient reporting on randomization, allocation concealment and blinding. There was no selective reporting or loss to follow-up in most studies; however, methods to deal with attrition were not reported. The Blair et al. [30] study used as-treated analysis to deal with attrition. Only Hung et al. [14] had a low risk of bias based on the Cochrane Back Review Group tool [20]. Funnel plots demonstrated a deficiency of studies with null or negative results suggesting a high grade of publication bias.

Discussion

To our knowledge, this is the first systematic review and meta-analysis that evaluates the impact of short-term supplementation with vitamin D or its analogs on glucose metabolism in dialysis patients. A large meta-analysis of observational studies had demonstrated a significant reduction in mortality in CKD and ESRD patients who received calcitriol or synthetic vitamin D analogs over 3–5 years [34]. Prior to this, a meta-analysis of RCTs failed to demonstrate a protective effect of vitamin D for mortality; however, mortality was not the primary outcome of those primary trials [35]. We found reasonably strong evidence, largely driven by five RCTs, of short-term (up to 12 weeks) vitamin D supplementation associated improvement in glucose metabolism in dialysis patients. Based on Cohen’s rule of thumb, the meta-analysis of RCTs demonstrated a significantly large improvement in fasting glucose levels; this effect was not significant among NRIS. A significant decrement in serum PTH was seen in meta-analyses of both RCTs and NRIS. Long-term glycemic outcomes (HbA1C %) also showed an improvement with associated decreases in PTH, though these results could not be meta-analyzed.

Fasting insulin levels are an accepted marker for insulin resistance in normoglycemic subjects; the meta-analysis population was mostly non-diabetic [1]. Review of studies that tested glucose-responsive insulin sensitivity and insulin secretion revealed an increase in these measures and thereby improvement in insulin resistance after vitamin D supplementation (Table 4) [11, 12, 22, 24–26]. Marked variation in testing of HOMA-IR precluded an interpretable quantification of summary effect size.

There was a large degree of heterogeneity in the meta-analysis. However, meta-regression to explain these differences was not done due to risk of serious bias given few RCTs [16, 36]. Unexplained heterogeneity could be secondary to residual confounding and/or bias, especially in NRIS [15].

Paricalcitol, a vitamin D analog, was found to have a null impact in the Hung et al. [14] study. It also had a null effect on short-term glucose metabolism among non-diabetic, non-dialysis stage 3–4 CKD patients (deBoer et al., 8 weeks) and on left ventricular mass (PRIMO trial, 48 weeks) [37, 38]. Paricalcitol has been found to suppress 25(OH)D levels, raising the possibility that vitamin D analogs may function differently from compounds that raise 25(OH)D levels [37]. As non-classical effects of vitamin D are likely dependent on circulating levels of 25(OH)D and local levels of extra-renal 1-α-hydroxylase enzyme, further clinical trials are needed that study compounds which raise 25(OH)D levels [5].

Autier et al. [39] examined the association of vitamin D with various health outcomes in extensive meta-analyses of prospective cohort studies and RCTs but did not find compelling evidence of a beneficial effect of vitamin D supplementation on glucose metabolism outcomes. However, these trials involved non-ESRD patients who had higher baseline levels of vitamin D and physiological renal activity enabling vitamin D activation, a significant difference from patients in our studies. Moreover, an interesting observation in our meta-analysis was that vitamin D-associated improvement in fasting glucose levels was seen in participants with impaired glucose tolerance at baseline, a common feature of the ESRD population. Also, fasting insulin levels were notably lower or borderline in most of these HD study populations than would be expected for insulin resistance in the general population (variably defined to be >12–>17 mU/ml) [40]. This could be explained by the fact that insulin is removed by HD, again attesting to the special status of this population [41].

It remains uncertain whether correction of hyperparathyroidism is a mediator for vitamin D-associated changes in glucose metabolism or whether these changes in PTH levels are merely expected side effects of supplementation. In the study demonstrating protective effect of vitamin D supplementation on mortality in ESRD patients, meta-regression demonstrated greater risk reduction among those with higher baseline PTH levels [34]. There is evidence to suggest an independent association of PTH and insulin resistance [42]. It is beyond the scope of this meta-analysis to establish the mechanisms of the observed differences.

Our study has its limitations. Firstly, outcomes varied by definition and measurement, only a few studies measured outcomes of interest such as HOMA-IR [14, 31, 32]. Hyperglycemia and glucose intolerance only manifest at later stages of insulin resistance when it cannot be overcome by insulin secretion [4]. Hyperinsulinemic euglycemic clamp studies and intravenous glucose tolerance tests are cumbersome and were infrequently conducted, though they are the gold standard test for insulin resistance [11–14]. Secondly, the design and quality of studies varied extensively. Despite this, RCTs demonstrated a significant summary effect of vitamin D supplementation on fasting serum glucose levels. Thirdly, most of the studies had small sample sizes and were focused on short-term outcomes (4–12 weeks) that are not long enough to adequately study relevant clinical outcomes.

In conclusion, improved fasting glucose levels are observed after short-term vitamin D supplementation in dialysis patients with associated decreases in serum PTH levels. However, well-designed larger clinical trials are needed to focus on long-term clinical outcomes of insulin resistance and related cardiovascular outcomes in ESRD.