Vitamin D supplementation in patients with cystic fibrosis: A systematic review and meta-analysis

Despite routine supplementation, vitamin D insufficiency is often seen in cystic fibrosis (CF) patients on account of pancreatic insufficiency. Vitamin D is a crucial component of bone health and affects nearly all cells of the immune system. However, clinical benefits or harms associated with supplementation are poorly documented. In this systematic review, we included randomized controlled trials (RCTs) that compared vitamin D supplementation with placebo (i.e. ‘non-increased dose’) in CF patients. Analysing the 8 included RCTs, the intervention group had significantly higher serum 25-hydroxyvitamin D (se25OHD) levels, but there were no significant differences found in the quantitative synthesis of clinical outcomes, including bone disease-, respiratory statusand immunological status-related outcomes. Based on our current results, while a higher vitamin D dose elevates se25OHD, it does not seem to influence clinical outcomes. Future RCTs should include outcomes of past studies and apply longer follow-up periods to document long-term patient-important outcomes. © 2020 The Authors. Published by Elsevier B.V. on behalf of European Cystic Fibrosis Society. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ )


Introduction
Cystic fibrosis (CF) (OMIM: #219700), is a hereditary disease caused by mutations in both alleles of the cystic fibrosis transmembrane conductance regulator (CFTR) gene located on the long arms of chromosome 7. It is one of the most common autosomal recessive disorders, affecting approximately 1 out of every 3,0 0 0 live-born worldwide [1] . Mutations -more than 2,0 0 0 documented to date, divided into six classes -generally result in the absence or the reduced/annulled function of the CFTR-protein, a transmembrane Cl − -channel present in the apical surface of epithelial cells Abbreviations: Ca, calcium; CI, confidence interval; CF, cystic fibrosis; CFTR, cystic fibrosis transmembrane conductance regulator; D2, ergocalciferol; D3, cholecalciferol; FEF25%, forced expiratory flow at 25% of FVC; FEV1, forced expiratory volume in the first second; FVC, forced vital capacity; GI, gastrointestinal; LL-37, cathelicidin; PTH, parathyroid hormone; RCT, randomized controlled trial; se25OHD, serum 25-hydroxyvitamin D; SMD, standardised mean difference; WMD, weighted mean difference. throughout the body. The defect of this channel causes the damage of multiple organs, mainly: the airways, pancreas, gastrointestinal (GI) tract, liver and reproductive system [2] .
In the pancreatic and biliary ducts, the defective CFTR-channels will result in thickened secretion production, chronic obstruction of these ducts, leading to chronic pancreatitis and exocrine insufficiency. Loss of exocrine function will lead to fat malabsorption and the deficiency of fat-soluble vitamins, including vitamin D [2][3][4] . Nutritional status and the level a micro-and macronutrients have a strong association with lung function and greatly determine the morbidity and mortality of CF patients [5] .
Vitamin D is crucial for bone health: mainly stimulating the calcium (Ca) absorption from the gut, its deficiency can lead to secondary hyperparathyroidism and bone loss [6] . Vitamin D also affects nearly all cells of both the innate (monocytes, macrophages, dendritic cells, etc.) and the adaptive (B-and T-cells) immune system, with sufficient vitamin D levels decreasing the risk of respiratory infections, thus possibly delaying CF progression and mortality [ 5 , 7 ]. As of 2019, the Cystic Fibrosis Foundation recommends different, consensus-based doses of cholecalciferol (D3) rather than ergocalciferol (D2) in different age groups of CF patients to be el-evated when necessary, with the goal of reaching a serum 25hydroxyvitamin D (se25OHD) concentration of at least 30 ng/ml (75 nmol/l) [8] . However, clinical benefits and harms of vitamin D supplementation in this population are poorly documented. The last systematic review of controlled trials was conducted in 2014 [9] with only three of the included randomized controlled trials (RCTs) supplying useful information and with inconclusive results on account of the heterogeneity and small sample size of the RCTs.
As more RCTs have been published since, our aim was to reevaluate the clinical benefits and possible adverse events accompanying a higher vitamin D dose in CF patients by conducting a systematic review and meta-analysis of RCTs in the topic.

Protocol and registration, reporting
This meta-analysis was registered with PROSPERO under registration number CRD42020155847. There have been no deviations from the registered protocol. We adhered to the recommendations of the 'Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)'-group in reporting our findings and writing the review [10] .

Eligibility criteria
Only randomized controlled trials were considered eligible for inclusion in this systematic review. We determined eligibility based on the PICO of trials in the following manner: P -Population: We included studies that examined CF patients, either paediatric or adult or both. There were no other restrictions based on the examined participants of these trials -exacerbation and stable disease were both allowed, also comorbidities and subpopulations of CF patients with the intent of conducting post-hoc analyses if feasible and necessary. Trials where only an examined subpopulation of participants have CF were also allowed; however, at the end, no such trials were eligible for inclusion. I -Intervention: We included studies that, as an intervention, supplemented vitamin D to the participants in any dose, any form (vitamin D3 or D2) and for any duration. Compounds consisting of multiple active substances including vitamin D were also allowed. Studies utilizing sunshine as a means of vitamin D supplementation were not eligible for inclusion. C -Comparator: Placebo or if the intervention in given study consisted of a compound with multiple active substances including vitamin D, any otherwise identical therapy lacking vitamin D.
Continued or basal vitamin D supplementation was allowed if it did not categorically differ between the intervention and comparator groups.
O -Outcome: There were no restrictions applied based on examined outcomes in the individual studies. Our primary interests were as entailed by our PICO: Bone mineral density and other quantifiable bone disease-related outcomes; mortality; height and weight Z-score; se25OHD, Ca, and parathyroid hormone (PTH) concentrations; respiratory status (forced expiratory volume in the first second (FEV1), forced vital capacity (FVC), forced expiratory flow at 25% of FVC (FEF25%), other assessed respiratory parameters); measures of immunological status (immunological markers, exacerbation due to infection, antibiotic use); quality of life; adverse events. Apart from these, all other examined outcomes were considered valuable and if a study failed to report on any of the listed outcomes it was still not to be excluded from our systematic review.
Studies were not restricted in eligibility based on length of follow-up.

Systematic search and selection
We conducted the systematic search -using the same search key as detailed in supplementary material Suppl1. -in 4 databases: Embase, MEDLINE (via PubMed), Cochrane Central Register of Controlled Trials (CENTRAL) and Web of Science. The date of last systematic search was 9 th October 2019. There were no restrictions imposed on the search. Citations were exported as a shared pool to a citation manager software, EndNote X9 (Clarivate Analytics). Two independent reviewers (MFJ and OV) conducted the selection by title, abstract and full-text based on the previously disclosed predetermined set of rules. After selecting by title and abstract -in an inclusive manner -, the rate of agreement was determined and documented by calculating the Cohen's kappa coefficient ( κ) and the two citation pools were merged before screening by full text. Any disagreements were settled by an independent third party (ZS). Exclusions made in the full text phase of selection were documented, studies of the same population were linked together. In case of population overlaps for an outcome, the article with the higher number of participants was preferred.
Citations of the studies inspected for eligibility in the full-text phase were reviewed in order to identify any additional eligible trials.

Data extraction
Information on data extraction is available in the supplementary material.

Risk of bias assessment
The Revised Risk of Bias Assessment Tool (RoB 2) [11] was used to assess the risk of bias in the individual studies (all RCTs). MFJ and OV independently conducted the assessment, final results are based on consensus. Assessments were made on a study level but the nature of the examined outcomes were consistent -mostly results of laboratory examinations at predetermined points in time.
Results of the risk of bias assessment are available in the supplementary material.

Statistical analysis
All statistical analyses in this study were carried out using Stata 15 SE (StataCorp). Pooled weighted mean difference (WMD) or standardized mean difference (SMD) were calculated with 95% confidence intervals (CI) for continuous outcomes. A random effects model was applied for all analyses, with the DerSimonien-Laird estimation. Statistical heterogeneity was analysed using the I 2 statistic and the chi-square to gain probability-values; I 2 represents the magnitude of the heterogeneity (moderate: 30-60%, substantial: 50-90%, considerable: 75-100%) [12] .

Determination of quality of evidence
Recommendations of the 'Grading of Recommendations Assessment, Development and Evaluation' (GRADE) working group [13] were followed upon assessing the quality of the evidence (MFJ and VI, independently). The GRADEPro Guideline Development Tool [14] was used for preparing the Summary of Findings table and accessory tables.

Systematic search and selection
The systematic search yielded 2,738 hits, 1,011 after duplicate removal. These 1,011 records were screened. 931 studies were ex-  cluded based on title and abstract (Cohen's kappa: 0.86), leaving 80 studies to be assessed based on reviewing the full text. Out of the 80 articles, 72 were excluded, mostly because the full-text revealed that they did not contain original data or that they were alternative reports / conference abstracts of identical / overlapping study populations. At the end, 8 articles remained of which 2 were conference abstracts not providing useful data for the quantitative syntheses but that could be included in the systematic review ( Fig. 1 ).

Study characteristics
The main characteristics of included studies are gathered in Table 1 .   2 presents the quantitative synthesis of intervention effects on serum bone alkaline phosphatase (BALP), serum osteocalcin and lumbar spine Z-score -these were the only three parameters that were consistently measured in both studies -with no significant difference between groups neither in the individual studies, nor overall. It is to be noted that Hillman 2008 examined children while Haworth 2004 examined adults. Hillman 2008 also assessed BMC, BMD and Z-score in: whole body, lumbar spine, 1/3 radius and hip -with no significant differences between groups. Haworth 2004 also assessed total hip and distal forearm Z-scores with no significant differences between groups.  Table 1 Characteristics of included studies. RCT = randomized controlled trial, se25OHD = serum 25-hydroxyvitamin D, IU = international units, SD = standard deviation, med = median, IQR = interquartile range, FEV1 = forced expiratory volume in the first second, FVC = forced vital capacity, FEF25% = forced expiratory flow at 25% of FVC, Ca = serum calcium, PTH = serum parathyroid hormone, NA = not available. * = only patients with se25OHD < 30 ng/ml were accepted. 3

Respiratory status-related outcome measures.
The two studies examining pulmonary exacerbation (Grossman 2012, Tangpricha 2019) reported on return to baseline FEV1% (patients whose FEV1% of predicted returned to within 95% of baseline i.e. the best lung function in the 6 months / 1 year before the study) but in different subgroups of patients, thus results were not pooled. Grossman 2012 analysed patients whose FEV1% of predicted decreased greater than 10% from baseline to admission -90 vs 50% returned to baseline in the vitamin D vs placebo groups, respec-tively (p = 0.12). Tangpricha 2019 reported 35.3% of vitamin D and 25.3% of placebo patients returning to baseline at month 3, among all patients. Tangpricha 2019 also reported on mean FEV1% (no significant differences between groups), Pincikova 2017 on FVC (significantly increased in D3 group compared to baseline) and FEF25% (no significant differences). Table 2 .

Additional outcomes
All outcomes pre-planned in our protocol and outcomes reported on by multiple studies were assessed. Pooled analysis was possible in the case of serum Ca, PTH, cathelicidin (LL-37) and albumin, without significant differences between groups. These outcomes are available in the supplementary material.

Quality of evidence
We included seven outcomes in our main Summary of Findings table: long-term survival (not reported), 12-month mortality after exacerbation (not pooled, opposing results, very low certainty), return to baseline lung function after exacerbation (not pooled, favoured vitamin D, very low certainty), adverse event rate (not pooled, no difference, very low certainty), quality of life (not reported), lumbar spine Z-score (no significant difference, very low certainty) and se25OHD (significantly higher in intervention group, moderate certainty). The main Summary of Findings table, and the one of additional outcomes are available in the supplementary material.    Grossman 2012 "There were no reported symptoms of vitamin D toxicity as assessed by patient questionnaire at any study visit, no clinical signs of hypercalcemia, no significant changes in mean serum calcium or PTH concentrations."

ARTICLE IN PRESS
Haworth 2004 "The patients did not report any significant adverse events with drug or placebo, and there were no documented episodes of hypercalcaemia." Manshadi 2012 "No adverse events were identified during the study period." Brown 2006 nephrolithiasis (n = 1), asymptomatic hypercalcaemia (n = 1) nephrolithiasis (n = 1), persistent hypercalciuria (n = 2)

Pincikova 2017
No reports on presence / absence of adverse events.

Hillman 2008
No reports on presence / absence of adverse events.

Discussion
The meta-analysis of the included RCTs demonstrated significantly higher se25OHD levels in the intervention group and no significant differences between intervention and comparator -neither on a study-level nor in total -regarding: lumbar spine Z-score, serum levels of BALP, osteocalcin, Ca, PTH, LL-37 and albumin. All of the included studies applied a basal vitamin D dose, meaning that only the comparison of this basal to a higher vitamin D dose could serve as a means to evaluate the intervention.
Even though the number of RCTs in the field of interest has doubled since the last systematic review, the lack of significant dif-  ferences between groups should still not be considered as definite ineffectiveness, the quality of evidence being very low. The main reason behind this is the low number of participants in each study and the heterogeneous choice of reported outcomes resulting, in most cases, less than 100 participants per comparison.

ARTICLE IN PRESS
However, the tendency of our current results suggests that an increased vitamin D dose (usually an additional 1,60 0-5,0 0 0 IU/day in the included studies) when compared with placebo plus the patients' continued, basal vitamin D dose (usually 40 0-1,80 0 IU/day, which is roughly equivalent to the initial regimen recommended by the CF Foundation ( Table 3 )) while significantly raises se25OHD, does not influence clinical or other laboratory outcome measures.
Based on our results, it seems that a higher dose is unnecessary, as it poses no additional benefits. But it is also of note that, -while only three of the RCTs excluded patients with se25OHD ≥30 ng/mL, -six studies experienced baseline mean se25OHD < 30 ng/ml, which is the target minimum se25OHD value recommended by the CF Foundation. This value, in contrast with the > 20 ng/ml recommendation for the general population [ 27 , 28 ], was modelled after endocrinology and osteoporosis guidelines given the frequent vitamin D deficiency, lower bone density and bone health markers, and higher rate of fractures among CF patients [ 29 , 30 ]. Accordingly, similar target se25OHD levels are recommended by European ( ≥30 ng/mL) and Australasian ( ≥20 ng/mL end of winter, ≥24-28 ng/mL rest of the year) CF Societies [ 25 , 26 ]. It seems that the recommended initial dose is often inadequate for achieving the target se25OHD, thus, even though we observed no clinical benefits, we think it would be sensible to consider a higher initial vitamin D dose for CF patients -more in line with the European recommendations, or the second dose step of the CF Foundation. A higher vitamin D dose, as described by the individual studies, was also not accompanied by a higher rate of adverse events ( Table 2 ).
We would also like to call on future RCTs of vitamin D therapy in CF to include outcomes of preceding studies, enabling the rise of high-quality scientific evidence.

Strengths and limitations
This is the largest systematic review to date to examine vitamin D supplementation in CF. To our knowledge, this is the first review to include only RCTs, in order to analyse only the highest quality clinical trials. We were able to identify 8 RCTs, of which 6 supplied useful information for meta-analyses -3 more than in the last systematic review.
Our study has several limitations. First of all, while more RCTs emerged since the last meta-analysis, the choice of outcomes, in most cases, did not overlap. While we understand the importance of presenting novel results, we highly recommend including outcomes of past studies in order to achieve high quality scientific evidence. Partly because of this, and partly because of the low participant number in the studies themselves, high risk of imprecision was noted, all comparisons were well below the optimal information size. Indirectness was also present as all studies applied a basal vitamin D dose. The research question at hand should have been vitamin D versus no vitamin D, which of course, would be unethical in such a high-risk population. We would also like to point out that due to the short follow-up periods, long-term patient-important outcomes, such as mortality and exacerbations, are yet to be documented. While the mean age in most studies was between 20 and 35 years, Hillman 2008 examined a younger (mean age: 9.1 years) population. Also, this was a crossover study, while others applied a parallel design.

Implications…
… for practice: Additional vitamin D does not seem to influence clinical and laboratory outcomes. However, CF patients receiving vitamin D, especially those receiving their initial dose, should be closely monitored as insufficiency is frequent. … for research: Outcomes of past studies should be included. RCTs with longer follow-up periods (5 years or more) should also be conducted to observe the long-term effects of supplementation.

Authors' contributions
MFJ wrote the manuscript and took part in the selection and data extraction process. OV, ÁM, VI took part in the selection and extraction process, and contributed to the manuscript. DN conducted the statistical analysis of the data. ZS and SK provided methodological guidance. AP and PH contributed to and supervised the writing of the manuscript and gave medical insight. All authors have (1) contributed to the concept of the study; (2) re- The funders had no role in designing the study; collecting, analysing and interpreting the data; writing the report; deciding to submit the manuscript for publication.

Declaration of Competing Interest
Authors disclose no conflicts of interest.

Supplementary materials
Supplementary material associated with this article can be found, in the online version, at doi: 10.1016/j.jcf.2020.12.008 .