Deficiency and Insufficiency of Vitamin D in Women of Childbearing Age: A Systematic Review and Meta-analysis

Objective  To estimate the prevalence of inadequate vitamin D level and its associated factors for women of childbearing age in Brazil. Methods  A systematic review was conducted (last updated May 2020). Meta-analyses were performed using the inverse-variance for fixed models with summary proportion calculation by Freeman-Tukey double arcsine. Reporting and methodological quality were assessed using the Joanna Briggs Institute tool for prevalence studies. Results  Our review identified 31 studies, comprising 4,006 participants. All the studies had at least one weakness, mainly due to the use of convenience sampling and small sample size. The overall prevalence of vitamin D deficiency, insufficiency, and both deficiency and insufficiency were 35% (confidence interval, 95%CI: 34–37%), 42% (95%CI: 41–44%), and 72% (95%CI: 71–74%), respectively. Conclusion  Although the magnitude of the prevalence of inadequate levels of vitamin D is uncertain, the evidence suggests that presence of vitamin D deficiency or insufficiency in women of reproductive age can cause moderate to severe problems.


Introduction
The deficiency and insufficiency of 25-hydroxyvitamin D, also known as 25(OH)D or vitamin D, is a worldwide issue: less than 50% of the world population has an adequate level of vitamin D, but in older people, pregnant women, and non-Western immigrants the proportion is smaller. 1 In pregnant women, for instance, the prevalence of insufficiency (25(OH) D < 50 nmol/L) and deficiency (25(OH)D < 25 nmol/L) ranged from 46% to 87% and 9% to 79%, respectively. 2 Even in warmer countries, such as Brazil, there is an alarming prevalence of vitamin D deficiency (28%) and insufficiency (45%), reaching 85% in pregnant women. 3,4 Recent studies suggested that vitamin D homeostasis may be important for several nonskeletal outcomes, including cardiovascular and respiratory diseases, neuromuscular function, psoriasis, falls, obesity, type 2 diabetes mellitus, multiple sclerosis, colorectal cancer, and coronavirus disease . [5][6][7][8][9][10][11][12] Vitamin D deficiency also causes a series of poor gestational outcomes, 13 increasing the risk of preeclampsia and gestational diabetes mellitus, as well as the production of maternal inflammatory cytokines, 13,14 insulin resistance, 13,15 and postpartum depression. 13,16 In Brazil, there is a great variability in studies assessing insufficiency and deficiency of vitamin D in women of childbearing age (12-68%), [17][18][19] but there is also a lack of evidence that systematically summarizes their prevalence. A systematic review (2019) evaluated the deficiency and insufficiency of vitamin D in Brazil, with no specific analysis for women of childbearing age. 4 The present systematic review aimed to identify the prevalence and factors associated with inadequate levels of vitamin D in women of childbearing age in Brazil.

Study Design, Protocol, and Registration
A systematic review was performed in accordance with the Meta-analysis of Observational Studies in Epidemiology (MOOSE) group, 20 and Joanna Briggs Institute recommendations, 21 and reported following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). 22 The protocol of this review is available at Center for Open Science 23 and PROSPERO (CRD42020221605). This study is part of a larger project that evaluated vitamins A, B, C, D, and E, calcium, iodine, iron, and zinc deficiencies in women of childbearing age in Brazil.

Information Sources, Search Strategy, and Eligibility Criteria
Electronic searches were conducted in the following databases: PubMed, Scopus, LILACS, World Health Organization (WHO), and CAPES' dissertations and theses (gray literature). The selection of these sources ensured including EMBASE, Medline, open access sources, scientific websites, and gray literature, 24 through a predefined search strategy (available in the protocol) 23 from their inception to May 2020. An additional manual search was performed using reference lists of reviews and included studies.
Studies that fulfilled the following criteria according to the CoCoPop acronym were included 25 : i) Condition: vitamin D deficiency or insufficiency; ii) Context: Brazil, without restriction of setting; iii) Population: women of childbearing age (15-49 years old) without any restriction of diseases or physiological status (e.g., nonpregnant, pregnant, postpartum). Data from studies that reported the deficiencies of interest, using a different population classification (e.g., premenopausal women), or different laboratory parameters were separated for appropriate subgroup analyses. All types of articles were included, except for reviews, letters, comments, case reports, and case series. No language restriction was applied.

Study Selection and Data Extraction
Two researchers screened the titles and abstracts and evaluated the full-text articles independently. Discrepancies were

Synthesis of Results
Although predefined cutoffs for the assessment of deficiencies and insufficiencies of vitamin D were not considered inclusion criteria in the present review, only studies that considered identical cutoffs were grouped. The data synthesis was primarily done by meta-analysis. Transitivity assessment was performed by comparing the CoCoPop acronym for each study. 25 Once important discrepancies were identified, sensitivity analyses with the exclusion of the study in question were performed (i.e., leave-oneout method). Proportion meta-analyses were conducted in the RStudio IDE (RStudio, PBC. Boston, MA, USA) software, version 3.6.3, 1.2.5033, 26 using the READR (RStudio, PBC.) 27 and META packages (RStudio, PBC.). 28 In the base-case, direct proportion meta-analyses were conducted using the inverse variance method. 28 To calculate the weighted summary proportion, the Freeman-Tukey double arcsine (PFT) was considered in the fixed effects model. 22,28 Although high heterogeneity is expected and, therefore, a random effects model could be considered appropriate, a fixed effects model is preferred for the assessment of prevalence, because otherwise the weighting will not properly consider the weight of the studies. 29 The result of the meta-analysis was given by the proportion combined with 95% confidence interval (95% CI), as well as the list of proportions (presented as a percentages), with their respective 95%CIs found in the individual studies. A Higgins inconsistency test (I 2 ) with an estimator for tau 2 was considered using the DerSimonian-Laird method.
Cumulative meta-analyses were also performed to assess changes and trends over time, and to highlight emerging or decreasing deficiency or insufficiency. Potential publication bias was assessed using rank tests with at least ten studies by meta-analysis. 28 Sensitivity analyses were performed by the leave-one-out method. Subgroup and meta-regression analyses, considering the publication year, state and region of Brazil, comorbidities, age, or status (i.e., not pregnant, pregnant, postpartum) were planned for meta-analyses with at least ten studies. Alternative statistical methods were also con-ducted to validate the conclusions (i.e., GLMM, Logit transformation, random effects, and Hartung and Knapp for random models).

Methodological Quality in Individual Studies
An assessment of methodological and reporting quality based on the JBI Critical Appraisal Checklist for studies reporting prevalence data was conducted. 30,31 Two reviewers performed the assessment, independently. In the absence of consensus, points of disagreement were resolved by a third investigator.

Data Sharing and Data Accessibility
The data that support the findings of this study are openly available in OSF at http:doi.org/10.17605/OSF.IO/J9QMH. 23

Results
Our systematic review identified 1,977 records in the electronic databases after duplicate removal (PubMed, LILACS, and Scopus) and 91 additional records identified through other sources (manual search, WHO, and CAPES' dissertations and theses databases). After selection process, 31 studies were included, published between 2008 and 2020, reporting deficiency or insufficiency of vitamin D. The list of included and excluded studies, as well as a PRISMA flowchart, are available in the OSF. 23 Of the 31 studies selected, 23 were cross-sectional, 4 prospective, 2 retrospective cohorts, and 2 were case-control studies. The studies were conducted between 1995 and 2017 (six studies did not report inclusion period), in cities in the Southeast (n ¼ 18), South (n ¼ 7), Northeast (n ¼ 5), and Center-west (n ¼ 2) Brazilian regions, with women selected mainly from outpatient care (n ¼ 20). Araújo et al., 32 Queiroz, 33 Queiroz et al., 34 de Oliveira et al., 35 and dos Santos et al. 36 used a random probabilistic sampling, while Martins et al. 37 used convenience sampling (►Chart 1). 3,[17][18][19] Most studies assessed women of childbearing age (n ¼ 13), followed by pregnant women (n ¼ 10), adolescents (n ¼ 6), and postpartum women (n ¼ 4). Two studies assessed pregnant and nonpregnant women, concomitantly. Therefore, 4,006 participants were included, mainly women of childbearing age (n ¼ 1,239), with a mean age ranged from 13 to 46 years old, and mean body mass index ranged from 22 to 46 kg/m 2 . The majority of studies included women with a medical condition (e.g., HIV þ , gestational diabetes mellitus, hypertension) or post Roux-en-Y gastric bypass surgery (RYGB, n ¼ 18). Although drug therapy use was not reported is most studies, nutrient supplementation (n ¼ 11) or no supplementation (n ¼ 11) use were reported. The main characteristics of the participants are described in ►Chart 2. 3,[17][18][19][32][33][34][35] In the quality assessment, all studies had at least one 'No' answer, which suggests an overall poor reporting or methodological quality. The main questions with 'No' answers were regarding sample size (n ¼ 30) and sampling method (n ¼ 29). Questions with 'Yes' answers were about sample frame and valid methods used for the identification of the  Most studies (n ¼ 26) used common cutoff values (vitamin D deficiency: < 50 nmol/L or < 20 ng/mL; vitamin D insufficiency: 50-75 nmol/L or 20-30 ng/mL; and vitamin D deficiency or insufficiency: < 75 nmol/L or 30 ng/mL) (Chart 1) and were included in meta-analyses.
Five studies assessed vitamin D in women post-RYGB, and two of them analyzed pregnant women after RYGB. No metaanalysis was possible due to the different cutoff values and categories used. The deficiency, insufficiency, and deficiency or insufficiency ranged from 12 to 39%, 41 to 54%, and 60 to 91%, respectively. Cumulative meta-analyses were performed considering the year of publication, showing a trend toward a lower prevalence of vitamin D deficiency, and higher prevalence of vitamin D insufficiency and vitamin D deficiency or insufficiency, with a slight join point in 2017 (►Appendix B, supplementary material). 23 Meta-regression analyses were conducted for publication year, and a moderator effect was not identified (p > 0.05) (►Appendix C, supplementary material). 23 Meta-regression or subgroup analyses for other variables were not possible, and neither were cumulative meta-analyses regarding gestational trimesters, because the minimum number of studies required was not met.
Sensitivity analyses by the leave-one-out method were not able to reduce heterogeneity (93-96%) and the overall preva-lence ranged from 32 to 37% for vitamin D deficiency, 41 to 44% for vitamin D insufficiency, and 71 to 73% for vitamin D deficiency or insufficiency (►Appendix D, supplementary material). 23 The study with more influence in the variations was Prado et al., 3 conducted in Minas Gerais, in 2012, with postpartum women taking supplements (97%). Sensitivity analyses with alternative statistical methods identified values of prevalence ranging from 35 to 37% for vitamin D deficiency, 41 to 43% for vitamin D insufficiency, and 69 to 72% for vitamin D deficiency or insufficiency (►Appendix A, supplementary material). 23 It was not possible to conduct sensitivity analyses regarding gestational trimesters.
Potential publication biases were not identified in vitamin D deficiency (p ¼ 0.84), insufficiency (p ¼ 0.60), or deficiency or insufficiency (p ¼ 0.54) in statistical or visual analyses (►Appendix E, supplementary material). 23 It was also not G -Mothers' full-term births, and mothers' preterm births. H -Women with smoking; eating disorders; major depression; any metabolic disease, such as diabetes mellitus or hypothyroidism; any chronic diseases severely affecting the CV, gastrointestinal, and renal systems; and pregnancy or lactation were excluded; I -850 mg of calcium carbonate and 600 IU of vitamin D3. J -Women with malignant tumors or infectious diseases; were postmenopausal; were taking drugs that affect bone metabolism (bisphosphonates, estrogens, anticonvulsants, glucocorticoids); were pregnant; had malabsorption syndrome, primary hyperparathyroidism, renal, or liver failure; or weighed > 120 kg were excluded. K -Women with multiple pregnancies, preeclampsia, kidney problems, HIV and women who had not fasted for the blood collection were excluded. L -Chronic illness, pregnancy, and obesity were excluded. M -Alcohol use, hyperglycemia, hypertension, preterm/post-term deliveries and adolescent pregnancy, were excluded. Were the study's subjects and setting described in detail? 5. Was data analysis conducted with sufficient coverage of the identified sample? 6. Were valid methods used for the identification of the condition? 7. Was the condition measured in a standard, reliable way for all participants? 8. Was there appropriate statistical analysis? 9. Was the response rate adequate, and if not, was the low response rate managed appropriately? a -Not reported, convenience sampling was considered. b -Reported convenience sampling. c -The target sample size reported was low. d -A target sample size was not reported. e -Did not report at least two of the following information: ethnicity, comorbidities, medicines/supplementation, body mass index, age, educational level, or income per capita. f -Vitamin D cutoff different than usual. g -Not reported numerator (n) or denominator (N) of prevalence). h -Not reported numerator (n) or denominator (N) of prevalence. i -The studies presented a response rate below 176 participants to vitamin D assessment. j -The studies presented a response rate between 176 and 345 (vitamin D), therefore is unclear if the sample size is appropriate, since a reliable estimate was not possible. k -The studies presented a response rate higher than 345, therefore, high confidence about good response was achieved.
possible to conduct statistical and visual analyses of publication bias for meta-analysis along gestational trimesters. Four studies reported different cutoff values and were not included in any meta-analysis. They identified prevalence values ranging from 11% to 75%: Duran de Campos et al. 39 identified serum 25(OH)D levels between 12.5 and 25 nmol/L (5-10 ng/mL) in 50% of the participants, and between 25 and 50 nmol/L (10-20 ng/mL) in 40% of the participants; Mendes et al. 47 identified 11% of the participants with values between 25 and 50 nmol/L; Peters et al. 51 identified 61% of the participants with values between 25 and 75 nmol/L (11-30 ng/mL); and Rosa et al. 68 identified 55% and 75% of the participants with values between 15 and 90 ng/mL in preand postoperative RYGB, respectively.
Redundant evidence of vitamin D levels was identified, especially for women of childbearing age in Brazil, to the detriment of population subgroups such as pregnant women, women who have recently given birth, and adolescents. In 2019, Pereira-Santos et al. 4 identified 72 studies that reported prevalence of vitamin D deficiency (28%) and insufficiency (45%) in the general population, and five studies that reported prevalence of 33% and 49%, respectively, in pregnant women. Although our systematic review identified the double of studies in pregnant women and 22 studies with women of childbearing age, our prevalence is similar to the Pereira-Santos' et al. 4 study, confirming the findings of our cumulative meta-analysis that new studies (published after 2017) had little impact on the prevalence estimates. At the same time, all the included studies showed weaknesses and high heterogeneity, which reduced the confidence on the prevalence rates reported.
Although little variation on the estimates has been added in the last years for women of childbearing age, when considering population subgroups (e.g., adolescents, pregnant women, postpartum women) the uncertainty still exists. For instance, when considered vitamin D deficiency in postpartum women (48%, 95% CI 44-52%, I 2 99%), Martins et al. 37 identified prevalence of 19%, whereas Prado et al. 3 described it as 85%. While Martins et al. 37 included 79% of dark-skin women (variable associated with deficiency), 64% using supplement (variable associated with sufficiency), and during spring and summer (variable associated with sufficiency); Prado et al. 3 included 52% of white women (variable associated with sufficiency), 97% using supplement (variable associated with sufficiency), and throughout the year. Moreover, it was not possible to conduct a robust subgroup analysis to explore the heterogeneity, as well as to identify possible associated factors to deficiency or insufficiency of vitamin D, since most studies did not report the characteristics of the participants, nor population subgroup analysis using common categories. Primary studies should appropriately report the findings according to common subgroups, and minimally, season, skin pigmentation, WHO standardized age group, 69 and supplement use.
In comparison with international data for inadequate vitamin D levels, our prevalence estimates are lower than estimates for women in Iran (44% deficiency), 70 and for women of childbearing age in in Saudi Arabia (77% deficiency or insufficiency), 71 but higher than estimates for adolescent girls in India (26% deficiency). 72 Several factors can explain the differences between the estimates, such as age, latitude, skin pigmentation, dietary habits, fortification of foods with vitamin D, use of vitamin D supplements, sunlight exposure, and cultural factors. 1,73-75 To exemplify, Gomes et al. 76 identified a seriously inadequate intake of vitamin D among Brazilian pregnant women in the primary healthcare network.
It is important to highlight that our systematic review identified several studies evaluating nonpregnant and nonlactating women, which were grouped as women of childbearing age. Notwithstanding, it was noted that many of these women had conditions associated with inadequate levels of vitamin D, such as overweight or obesity, [77][78][79] gestational diabetes mellitus, 80 preeclampsia, 81,82 cardiovascular disease, 83 breast cancer, 84 polycystic ovarian  syndrome, 85 and infertility, 86 among others, which may overestimate the identified prevalence.
Another important consideration is that despite the variation in cutoff values used by studies to define vitamin D deficiency, most studies included in this meta-analysis considered the threshold recommended by the US Institute of Medicine (< 50 nmol/L of 25 (OH)D) as opposed to the threshold recommended by the Endocrine Society Practice Guidelines (< 75 nmol/L of 25(OH)D). The generally accepted cutoff levels consider the values necessary to ensure optimal effects in the calcium economy and skeletal health, 87 and studies designed to assess the correlation of clinical responses with clinically relevant vitamin D deficiency suggest that depending on the physiological parameters considered (e.g., pregnancy outcomes, cardiometabolic risk) the results may differ and be even greater than those mentioned above, [88][89][90] resulting in the identification of larger populations with vitamin D deficiency. Although it is not possible to be sure about the magnitude of deficiency/insufficiency of vitamin D in some subgroups among Brazilian women, current evidence suggests that this is a public health problem, given the Institute of Medicine's (US) recommended cutoff values. 91 In this sense, some preventive strategies for adequate vitamin D levels include fish consumption, food fortification, 92 and advice on moderate sunlight exposure. 1, 93 Among the few countries with specific policies, the United Kingdom and Finland stand out with the recommendation of 10 μg of vitamin D daily intake for general population, and the mandatory food fortification programs, respectively. 94 In pregnant women, conflicting evidence suggests the benefit of supplementation, despite the documented negative clinical, humanistic, and economic impact of the deficiency or insufficiency of vitamin D, mainly, during the first trimester of pregnancy. 95 The hesitation about the recommendation of supplement intake may be justified by the limited evidence on the safety of vitamin D supplements, which could explain the reason why WHO does not recommend the supplementation during pregnancy as part of routine antenatal care. 96 Conversely to WHO, the Brazilian consensus recommends supplementation in pregnant women at risk of deficiency. 11 However, the Brazilian consensus does not recommend generalized vitamin D supplementation for the entire population, while it recommends the assessment of serum levels in obese patients. 11 Despite several options of vitamin supplements containing vitamin D being available in Brazil, with some of them included in Brazilian National List of Essential Medicines (Rename), 97 no national policy to prevent vitamin D insufficiency or deficiency in any women subgroup exists. In addition to funding studies to estimate the prevalence of micronutrient deficiencies in women of childbearing age, 98 a government policy is needed to avoid vitamin D inadequate levels, as well as excessive intake by self-medication or inappropriate prescription. 99 As any systematic review, one limitation of this study is that missing studies could exist. To overcome this limitation, extensive gray literature and manual searches to find unpublished and published studies were conducted, having found a few studies not retrieved by electronic searches. Although a high number of studies were identified through manual search, which could be seen as a limitation of the search strategy, one hypothesis is that many studies may not have properly written titles and abstracts, or are not correctly indexed, hindering the automatic search algorithm's ability to retrieve them. Finally, another limitation was the absence of a robust analysis about potential associated factors of inadequate levels of vitamin D, due to the poor reporting of the compiled studies.

Conclusion
Although the magnitude of the prevalence of inadequate levels of vitamin D is uncertain, the evidence found in the literature suggests a moderate to severe problem with a prevalence of vitamin D deficiency (35%), insufficiency (42%), and deficiency or insufficiency (72%) in women of reproductive age. Future studies about vitamin D levels should consider random probabilistic sampling, appropriate sample sizes and reporting of findings. Furthermore, vitamin D studies should consider estimates according to the season, skin pigmentation, age range standardized by WHO, and use of supplements, to better inform potential health policies.