Vitamin D Deficiency and Treatment Versus Risk of Infection in End-Stage Renal Disease Patients Under Dialysis

A Systematic Review and Meta-analysis

Guobin Su; Zhuangzhu Liu; Xindong Qin; Xu Hong; Xusheng Liu; Zehuai Wen; Bengt Lindholm; Juan-Jesus Carrero; David W Johnson; Nele Brusselaers; Cecilia Stålsby Lundborg

Disclosures

Nephrol Dial Transplant. 2019;34(1):146-156.

Results

Study Selection

We identified 2158 records in English databases, 250 records in Chinese databases and 32 records from additional sources. Of these, 2140 records remained after removing duplicates. We excluded 2063 records based on titles and abstracts screening. Of the remaining 77 articles, we excluded 60 articles that did not meet inclusion criteria after full-text screening (21 articles not including dialysis patients; 35 articles without infection-related data; 1 article restricted to children; 3 articles^[27–29] solely reported 25(OH)D as a continuous variable). Seventeen studies {11 for vitamin D use^{[8–10,30–37]} and 6 for different levels of 25(OH)D^{[11–13,15,38,39]}} met the eligibility criteria and were considered for meta-analysis (Figure 1).

(Enlarge Image)

Figure 1.

Flowchart.

Study Characteristics

The six studies^{[11–13,15,38,39]} selected for evaluation of different serum levels of 25(OH)D enrolled a total of 5714 participants (Table 1) and were cohort studies. Three studies^[12,13,15] provided data on infection-related death including 5220 HD patients. The remaining three studies^[11,38,39] provided data for PD-related peritonitis enrolling 494 PD patients. The HD studies were from the Hemodialysis (HEMO) Study,^[13] an international multicentre study that examined whether augmented dialytic urea removal or the use of high-flux dialysis could improve outcomes in HD patients (n = 1) and patients from Germany^[12,15] (n = 2) and China^[11,38,39] (n = 3). The mean age ranged from 48 to 71 years and the proportion of men ranged from 27.5% to 63.9%. The duration of follow-up varied between 6 months and 4 years. Categorization of low serum 25(OH)D levels varied among the six studies from <5.4 to <15 ng/mL. The most common testing method was enzyme-linked immunosorbent assay (ELISA).

The 11 studies^{[8–10,30–37]} assessing the use of vitamin D included a total of 92 309 patients (Table 2). Among these studies, two^[32,33] evaluated nutritional vitamin D supplementation (both RCTs) and nine evaluated VDRA (seven observational cohort studies^{[8,9,30,31,35–37]} and two case–control studies^[10,34]). Four studies^{[30,31,36,37]} provided data for infection-related mortality and included 78 981 patients. Three studies^[9,10,32] provided data for IRH and included 12 315 patients. The remaining studies provided data for infections, all types of infection,^[33] PD-related peritonitis^[8,35] and herpes zoster.^[34] These studies were from the USA^[32,33,37] (n = 3), Japan^[9,30,31] (n = 3), Latin America^[36] (n = 1), Austria^[8,35] (n = 2), Canada^[10] (n = 1) and Taiwan^[34] (n = 1). The mean age ranged from 48 to 70 years and the proportion of men ranged from 52% to 84%. The duration of follow-up varied between 16 weeks and 5 years. Among those using VDRA, six^{[8–10,34–36]} used oral formulations, one^[37] used an intravenous formulation and two^[30,31] used mixed formulations.

Risk of Bias Assessment

In general, the quality of the included observational studies was moderate according to the NOS criteria (4 studies with moderate quality and 11 studies with high quality). The common item underscored in this scale was an unclear definition of the outcomes of infection, which might have been related to the fact that infection was usually not the primary outcome. One RCT^[32] provided only its primary outcome (the change in epoetin dose of >6 months) in the protocol. Since they did not mention infection as an outcome in the study protocol, this raised the possibility of selective outcome reporting. Another RCT^[33] had an imbalance in missing data between groups, thereby implicating a high risk of incomplete outcomes. The full quality assessment is detailed in Supplementary data, Table S1, Table S2 and Table S3.

Effects of 25(OH)D on Infection-related Outcomes

When compared with individuals with low serum levels of 25(OH)D, the pooled adjusted risk for composite infection was 39% lower [RR 0.61 (95% CI 0.41–0.89), all were cohort studies] in those with high/normal levels, with moderate heterogeneity (I ² = 60.5%, P = 0.03; Figure 2).

(Enlarge Image)

Figure 2.

Forest plot depicting the meta-association between high/normal versus low level of 25(OH)D and risk for infection-related outcomes using the DerSimonian and Laird random effects model. All the included studies were cohort studies.

In patients undergoing PD, the risk of PD-related infections was 66% lower [RR 0.34 (95% CI 0.20–0.58)] in those with high/normal levels of 25(OH)D than in those with low levels of 25(OH)D, with minimal heterogeneity (I ² = 0%, P = 0.71; Supplementary data, Figure S1). In patients undergoing HD, the risk of infection-related mortality was 12% lower [RR 0.88 (95% CI 0.70–1.08)] in those with high/normal levels of 25(OH)D than in those with a low level, with minimal heterogeneity (I ² = 0%, P = 0.43; Supplementary data Figure 1S).

(Enlarge Image)

Supplementary Figure 1.

Forest plot depicting the meta-association between different level of 25(OH)D and risk for infection-related outcomes by different types of dialysis, using the DerSimonian and Laird random-effects model

Effects of Vitamin D use on Infection-related Outcomes

The pooled adjusted risk for infection-related outcomes was 41% lower in those who used vitamin D [RR 0.59 (95% CI 0.43–0.81)] when compared with those who did not use vitamin D (nutritional supplement or VDRA), with high heterogeneity (I ² = 91.1%, P < 0.001; Figure 3).

(Enlarge Image)

Figure 3.

Forest plot depicting the meta-association between use and non-use of vitamin D and risk for infection-related outcomes by study design using the DerSimonian and Laird random effects model. Observational studies groups included cohort study and case–control study exploring the association between use of VDRA and risk for infection-related outcomes. RCT groups explored the association between supplemental nutritional vitamin D and risk for infection-related outcomes.

There was no difference in infection-related outcomes between supplemental nutritional vitamin D users and non-users [RR 1.22 (95% CI 0.74–2.01), all were RCTs; Figure 3]. The pooled adjusted risk for infection-related outcomes was 48% lower in those who used VDRA [RR 0.52 (95% CI 0.36–0.73), all were observational studies; Figure 3], with high heterogeneity (I ² = 92.6%, P < 0.001; Figure 3).

When compared with those who did not use VDRA, the pooled adjusted risks for composite infection-related outcomes were 46% and 61% lower in those taking oral [RR 0.53 (95% CI 0.35–0.82), all were observational studies] and intravenous [RR 0.39 (95% CI 0.31–0.48)] VDRA, respectively (Supplementary data, Figure S2). A lower risk of infection-related mortality and a lower risk of infection were observed in those who used VDRA (Supplementary data, Figure S3).

(Enlarge Image)

Supplementary Figure 2.

Forest plot depicting the meta-association between use and non-use of Vitamin D receptor activator (VDRA) and risk for infection-related outcomes by different types of administration, using the DerSimonian and Laird random-effects model

(Enlarge Image)

Supplementary Figure 3.

Forest plot depicting the meta-association between use and non-use of Vitamin D receptor activator (VDRA) and risk for different infection-related outcomes by different types of administration, using the DerSimonian and Laird random-effects model

Meta-regression and Sensitivity Analyses

A sensitivity analysis examining VDRA use included seven cohort studies and demonstrated a similar association between the use of VDRA and a lower risk of infection-related outcomes (Supplementary data, Figure S4). Meta-regression analyses suggested that PD patients in studies with a larger sample size and using intravenous VDRA showed lower RRs when compared with their counterparts (Table 3).

(Enlarge Image)

Supplementary Figure 4.

Sensitivity analysis: forest plot depicting the meta-association between use and non-use of Vitamin D, high/normal vs. low level of 25(OH)D and risk for infection-related outcomes

Publication Bias

Publication bias was indicated in included studies investigating the level of 25(OH)D (P = 0.007 in Egger's test; Supplementary data, Figure S5) but not the use of vitamin D (P = 0.12 in Egger's test; Supplementary data, Figure S6).

(Enlarge Image)

Supplementary Figure 5.

Funnel plot depicting publication bias in included studies investigating the association between different level of 25(OH)D and risk for infection-related outcomes

(Enlarge Image)

Supplementary Figure 6.

Funnel plot depicting publication bias in included studies investigating association between use and non-use of Vitamin D and risk for infection-related outcomes

1 2 3 4 5

Next Section

Table 1. Characteristics of studies reporting on the association between 25(OH)D and risk of infection-related outcomes
Study	Region	Study design	Population	n	Age [mean ± SD; median (quartile)]	Male (%)	High 25(OH)D	Low 25(OH)D	Testing methods (25(OH)D)	When measured 25(OH)D	Types of infection-related outcomes	Definition of outcomes	Number of events	Follow-up duration
Chonchol et al. [13]	Multi-centres in HEMO study	Cohort in a RCT	HD	670	58 ±13 in group 25(OH)D <14 ng/mL; 55 ± 16 in group 25(OH)D > 26 ng/mL	27.5% in group 25(OH)D <14 ng/mL; 63.9% in group 25(OH)D >26 ng/mL	>26 ng/mL	<14 ng/mL	Immunoaffinity purification and liquid chromatography tandem mass spectrometry	Baseline	Composite of infection-related hospitalization and infection-related death	Composite of infectious hospitalization and infectious death	245	2.80 ± 1.7 years
Krause et al. [12]	Germany	Cohort	HD	4259	71 (20–94) in <12.5 ng/mL group; 71 (20–96) in ≥30 ng/mL group	35.5% in <12.5 ng/mL group; 26.7 in ≥30 ng/mL group	≥30 ng/mL	<12.5 ng/mL	A competitive protein-binding assay that was developed in-house or Nichols Advantage 27-hydroxyvitamin D assay	The last observation	Infection-related death	Causes of death according to schemes defined by the EDTA (EDTA code)	366	January 1997–December 2006
Drechsler et al. [15]	Germany	Cohort in a RCT	HD	291	66 ± 8	54%	≥30 ng/mL (≥75 nmol/L)	≤10.0 ng/mL (≤25 nmol/L)	Chemiluminescence assay (IDS, iSYS 25(OH)D; Immunodiagnostic Systems, Boldon, UK)	Baseline	Infection-related death	Death owing to infection, centrally adjudicated by three members of the endpoint committee blinded to study treatment and according to pre-defined criteria	Unclear	4 years
Pi et al.[39]	China	Cohort	PD	230	59.99 ± 12.13 in the low-value group; 56.97 ± 13.68 in the high-value group	42.61% in the low-value group; 46.96% in the high-value group	>7.7 ng/mL (19.13 nmol/L)	<5.4 ng/mL (13.51 nmol/L)	ELISA (Immunodiagnostic Systems, Bolden, UK)	Baseline	PD-associated peritonitis	The presence of at least two of the following conditions: (i) abdominal pain or tenderness, (ii) presence of white blood cells (100 cells/mL) in peritoneal effluent, with at least 50% polymorphs and (iii) positive dialysate culture results	Unclear	March 2011–November 2011
Fengwei et al. [11]	China	Cohort	PD	124	48 ± 13.2	58.90%	≥15 ng/mL	<15 ng/mL	N/A	Before PD catheterization	PD-associated peritonitis	The presence of at least two of the following conditions: (i) abdominal pain or tenderness, (ii) presence of white blood cells (100 cells/mL) in peritoneal effluent, with at least 50% polymorphs and (iii) positive dialysate culture results	56	>6 months
Zhang [38]	China	Cohort	PD	140	51 ± 15 in the normal 25(OH)D group; 45 ± 16 in the low 25(OH)D group	32.1% in the normal 25(OH)D group; 27.9% in the low 25(OH)D group	N/A	N/A	ELISA	Unclear	PD-associated peritonitis	The presence of at least two of the following conditions: (i) abdominal pain or tenderness, (ii) presence of white blood cells (100 cells/mL) in peritoneal effluent, with at least 50% polymorphs and (iii) positive dialysate culture results	16	January 2012–December 2015

Table 2. Characteristics of studies reporting on the association between use of vitamin D and risk of infection-related outcomes
Study	Region	Study design	Population	Total number	Age [mean ± SD; median (quartile)]	Male (%)	Types of VitD	Administration	Intervention	Comparison	Types of infection-related outcomes	Definition of outcomes	Number of events	Follow-up duration
Miskulin et al. [32]	USA	RCT	HD	276	61.16 ± 13.6	Unclear	Supplementation	Oral	Baseline serum 25(OH)D ≤15 ng/mL received 50 000 IU weekly for 6 months; 25(OH)D 16–30 ng/mL received 50 000 IU every week for the first 3 months followed by 50 000 IU monthly for 3 months	Placebo	Infection-related hospitalization	Unclear	Unclear	6 months
Bhan et al.[33]	USA	RCT	HD	105	58 ± 16 for monthly group; 53±17 for weekly group; 59 ± 17 for placebo group	69.4–84.9	Supplementation	Oral	Four pills of 50 000 IU ergocalciferol (weekly arm), one ergocalciferol and three placebo pills (monthly arm)	Placebo	All types of infection	Unclear	30	16 weeks
Obi et al.[30]	Japan	Cohort	PD/HD	8568	65 ± 13	65	VDRA	Oral or intravenous	Oral calcitriol, oral alfacalcidol, oral falecalcitriol, intravenous calcitriol and intravenous maxacalcitol; unclear dose and frequency	Non-users based on the 2009 survey	Infection-related mortality	Death owing to sepsis, pneumonia, gastrointestinal infection, cholecystitis/cholangitis, peritonitis, tuberculosis, HIV/AIDS, hepatitis virus infection, influenza virus infection, or other infection	229	January 2007–10
Tanaka et al.[31]	Japan	Cohort	HD	3372	64.5 ± 12.5 in no use group; 64.2 ± 12.8 in oral group; 60.3 ± 12.8 in intravenous group	54.6–61.9	VDRA	Oral or intravenous	VDRA (oral or intravenous) at baseline; unclear dose and frequency	No administration of VDRA	Infection-related mortality	Unclear	118	4 years
Normand et al. [10]	Canada	Nested case–control study	HD	11 532	68.1 ± 12.3 in patients with IRH; 69.6 ± 10.0 in control group	59.3–60.8	VDRA	Oral	VDRA (alfacalcidol, calcitriol or doxercalciferol) at the date of the admission for an IRH, calculated cumulative dose received by each patient during the 6 months preceding the index date	Patients were not using VDRA at the date of the admission for an IRH	IRH	ICD-9 code (or ICD-10 after 2006) indicating an infection as main diagnosis on the hospital's discharge sheet	1136	January 2001–December 2007
Tsujimoto et al. [9]	Japan	Cohort	HD	508	59.6 ± 11.3	60.80	VDRA	Oral	Oral alfacalcidol (0.25–1.0 g/d) or oral calcitriol (0.25–0.5 g/d or 1.0–3.0 g three times a week) at the start of the study	Not on active vitamin D treatment	IRH	Hospitalization because of acute respiratory infections (ARIs; e.g. pneumonia and bronchitis)	57	5 years
Naves-Diaz et al. [36]	Latin America	Cohort	HD	16 004	55.6 ± 16.0 for non-use and 53.9 ± 15.9 for use group	58.3–58.4	VDRA	Oral	Oral active vitamin D (calcitriol or alphacalcidol) for more than 1 month	Patients receiving oral active vitamin D for a period of less than 1 month	Infection-related mortality	Unclear	Unclear	January 2000 and June 2004
Teng et al.[37]	USA	Cohort	HD	51 037	61 for non-use and 63 for use group	52–54	VDRA	Intravenous	Once initiated activated injectable vitamin D (calcitriol, paricalcitol or doxercalciferol); unclear dose, frequency and duration	The time after starting dialysis and before therapy was credited to the untreated group	Infection-related mortality	Unclear	Unclear	January 1996–December 2002
Chao et al.[34]	Taiwan	Matched case–control study	HD	126	62.0 (27.0–85.0) for the case group; 63.0 (29.0–88.0) for the control group	53.70%	VDRA	Oral	1α-hydroxylated vitamin D; unclear dose, frequency and duration	No 1α-hydroxylated vitamin D	Herpes zoster	Diagnosed by dermatologists with the characteristic grouped vesicles within dermatomal distribution, with or without Tzanck smear	63	January 2000–December 2009
Kerschbaum et al. [8]	Austria	Cohort	PD	726	55.4 (16.1–87.0)	58.80%	VDRA	Oral	Oral active vitamin D; unclear dose, frequency and duration	No oral active vitamin D	PD-associated peritonitis	The diagnosis of peritonitis was established in all centres based on two or more of the following findings: (i) abdominal pain, (ii) cloudy effluent, (iii) effluent white blood cell count exceeding 100/mL with at least 50% neutrophils and (iv) positive Gram staining	400	January 2000 –December 2009
Rudnicki et al. [35]	Austria	Cohort	PD	55	52.3 ± 13.4 for the non-user and 48.0±13.5 for the user group	Unclear	VDRA	Oral	Oral active vitamin D; unclear dose, frequency, duration	No oral active vitamin D after insertion of the CAPD catheter	PD-associated peritonitis	The diagnosis of an episode of peritonitis was established when two of the following three findings were present in the electronic chart: abdominal pain, cloudy effluent and effluent white blood cell count exceeding 100 μL, with at least 50% neutrophils	Unclear	January 2000–December 2007

Table 3. Meta-regression analyses on association between use of VDRA and risk of infection-related outcomes
Use versus non-use of VDRA	No. of studies	Empirical Bayes meta-regression pooled HR (95% CI)	P-value	I ²(%)
All infection-related outcomes
HD versus PD	8	0.61 (0.20–1.83)	0.32	93.40
Oral versus intravenous	9	0.66 (0.24–1.80)	0.36	86.50
Sample size <500 patients versus ≥500	9	0.31 (0.09–1.09)	0.06	93.00
Non-fatal infection
HD versus PD	5	0.57 (0.06–5.53)	0.49	76.97
Sample size <500 patients versus ≥500	5	0.27 (0.04–1.82)	0.12	86.42

Supplementary Table 1. Quality assessment of randomized controlled trials using Cochrane Collaboration's tool
Study	Random sequence generation	Allocation concealment	Blinding of participants and personnel	Blinding of outcome assessment	Incomplete outcome data	Selective reporting	Other bias
Miskulin 2016	Low risk: Randomization schedules were generated by ordering blocks of eight according to a random number generator	Low risk: Only the DCI Pharmacy was aware of which medication (A versus B) was ergocalciferol versus placebo	Unclear	Unclear	Unclear	High risk; It only provided primary outcome in the protocol, other outcomes in the final study were not mentioned.	Unclear
Bhan 2015	Low risk: assigned by a pharmacy maintained computer-generated random-number list	Unclear	Low risk: Patients, research staff, laboratory technicians, and clinical providers were blinded to study arm enrollment	Unclear	High risk; Imbalance in numbers for missing data	Unclear	Unclear

Supplementary Table 2. Quality assessment of cohort study using Newcastle-Ottawa Quality Assessment Scale
	Selection				Comparability	Outcome
Study	1) Representativeness of the exposed cohort	2) Selection of the non exposed cohort	3) Ascertainment of exposure	4) Demonstration that outcome of interest was not present at start of study	1) Comparability of cohorts on the basis of the design or analysis	1) Assessment of outcome	2) Was follow-up long enough for outcomes to occur	3) Adequacy of follow up of cohorts	Total Star ★
Obi 2017	★	★	★	★	★	★	★	★	8
Tanaka 2016	★	★	★	★	★	Unclear	★	★	7
Tsujimoto 2011	★	★	Unclear	★	★	★	★	★	7
Naves-Diaz 2008	★	★	★	★	★	Unclear	★	★	7
Teng 2005	★	★	★	★	Unclear	Unclear	★	★	6
Kerschbaum 2013	★	★	★	★	★	★	★	★	8
Rudnicki 2010	★	★	Unclear	★	★	★	★	★	7
Chonchol 2016	★	★	★	★	★	★	★	★	8
Krause 2012	★	★	★	★	★	★	★	★	8
Drechsler 2010	★	★	★	★	★	★	★	★	8
Pi 2015	★	★	★	★	★	★	☆	☆	6
Nong 2017	★	★	★	★	★	★	★	★	8
Zhang 2016	★	★	★	☆	Unclear	★	☆	☆	4

★Indicate low risk of bias, ☆ Indicate high risk of bias

Supplementary Table 3. Quality assessment of Case-controlled study using Newcastle-Ottawa Quality Assessment Scale
	Selection				Comparability	Exposure
Study	1) Is the case definition adequate?	2) Representativeness of the cases	3) Selection of Controls	4) Definition of Controls	1) Comparability of cases and controls on the basis of the design or analysis	1) Ascertainment of exposure	2) Same method of ascertainment for cases and controls	3) Non-Response rate	Total star ★
Normand 2014	★	★	★	★	★	★	★	☆		7
Chao 2012	★	★	☆	★	★	★	★	☆		6

★Indicate low risk of bias, ☆ Indicate high risk of bias

Authors and Disclosures

Guobin Su^1,2, Zhuangzhu Liu³, Xindong Qin², Xu Hong⁴, Xusheng Liu², Zehuai Wen⁵, Bengt Lindholm⁶, Juan-Jesus Carrero^3,4, David W Johnson^7,8,9, Nele Brusselaers¹⁰ and Cecilia Stålsby Lundborg¹

¹Global Health–Health Systems and Policy, Department of Public Health Sciences, Karolinska Institutet, Stockholm, Sweden, ²Department of Nephrology, Guangdong Provincial Hospital of Chinese Medicine, Second Affiliated Hospital, Guangzhou University of ChineseMedicine, Guangzhou, Guangdong Province, China, ³Department of Emergency, Guangdong Provincial Hospital of Chinese Medicine, Second Affiliated Hospital, Guangzhou University of Chinese Medicine, Guangzhou, Guangdong Province, China, ⁴Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden, ⁵Key Unit of Methodology in Clinical Research, Guangdong Provincial Hospital of ChineseMedicine, Second Affiliated Hospital, Guangzhou University of Chinese Medicine, Guangzhou, Guangdong Province, China, ⁶Division of Renal Medicine and Baxter Novum, Department of Clinical Science, Intervention and Technology, Karolinska Institutet, Stockholm, Sweden, ⁷Department of Nephrology, Princess Alexandra Hospital, Brisbane, Queensland, Australia, ⁸Centre for Kidney Disease Research, University of Queensland, Brisbane, Queensland, Australia, ⁹Translational Research Institute, Brisbane, Queensland, Australia and ¹⁰Centre for Translational Microbiome Research, Department of Microbiology, Tumor and Cell Biology, Karolinska Institutet, Stockholm, Sweden

Correspondence and offprint requests to
Xusheng Liu; E-mail: xushengliu801@126.com; liuxusheng@gzucm.edu.cn

Authors' Contributions
All authors contributed to the study design, data acquisition or data analysis and interpretation. Specifically, G.S., X.Q. and Z.W. developed and ran the search strategy. G.S. and Z.L. independently extracted data and obtained copies of the studies. X.S.L. checked the data extraction. X.D.Q. and Z.Z.L. assessed studies for inclusion and assessed the risk of bias. N.B. was the third review author who selected which studies to include when disagreements existed. C.S.L., J.-J.C., D.W.J. and B.L. contributed to the development of data analysis strategies. G.B.S. and H.X. performed the data analysis. G.S., X.L., N.B. and C.S.L. drafted the manuscript. All authors provided comments on this review and approved the final version.

Conflict of Interest Statement
B.L. is employed by Baxter Healthcare. The other authors declare no conflicts of interest.