Search results and study characteristics
A total of 2085 titles and abstracts were retrieved from electronic database searches, and after removing 254 duplicates, 1831 were screened based on their titles and abstracts. This yielded 61 full-text studies that were subsequently evaluated for eligibility. Six supplementary articles were also found to be eligible from the reference lists of the included studies. After reviewing the full texts, 45 studies were ultimately included in the meta-analysis. A summary of the selection process for the studies is presented in Fig. 1
Descriptions of included studies
Out of the 45 studies, 19 had cross-sectional designs [22-40], 23 were case–control studies [41-63], 2 had baseline cross-sectional data from a longitudinal study [64,65] and one was based on the baseline data of a cohort study [66]. The reported data also included 6,995 participants, mostly aged ≤ 18 years old, of which 2,436 were children/adolescents with T1D as well as vitamin D deficiency (sample size n=13~1,426). Overall, T1D cases were mainly ascertained on the basis of criteria laid out by the World Health Organization (WHO), the American Diabetes Association and the European Diabetes (EURODIAB) collaboration, while levels of 25-hydroxyvitamin D (25(OH)D) were measured using a radioimmunoassay kit or by high-performance liquid chromatography (HPLC). Similarly, vitamin D status was mainly ascertained based on the Endocrine Society Clinical Practice Guideline, the Institute of Medicine guidelines, the Australian consensus statement criteria and the Central European Guidelines. Regarding the countries covered in the studies, seven were conducted in America, four were conducted in Turkey, three were carried out each in Korea, Iran and India, two studies each were conducted in Australia, the UK, Egypt, Spain, Italy and the Kingdom of Saudi Arabia, and one study was performed in China, Indonesia, Poland, Kuwait, Canada, Bangladesh, Slovakia, Switzerland, Boston, Ukraine, Tunisia, Iraq and Germany. The main characteristics of the 45 included studies are shown in Table 1. In accordance with the recommended NOS and AHRQ criteria, only studies of acceptable quality were included in the present meta-analysis; eight studies received 9 stars [51,53,56-58,62,63,66], ten studies received 8 stars [43,48-50,52,54,55,59-61], five studies received 7 stars [42,44-47] and one study received 6 stars [41]. When using the quality assessment criteria from the AHRQ, three studies received a score of 11 [24,29,65], ten received a score of 10 [22,25,28,29,31,33,34,36,39,40], three received a score of 9 [23,32,35], one received a score of 8 [27], one received a score of 7 [37] and two received a score of 5 [26,38], the quality assessment are shown in Appendix S2. Therefore, no article from the meta-analysis was excluded for quality reasons.
Meta-analyses and data synthesis
For the whole sample of 6,995 individuals, the proportion of vitamin D deficiency in children and/or adolescents with T1D was 45% (95% CI; 37-54%; P < 0.01; Fig. 2). The analyses further indicated the heterogeneity between studies (I-square [I2] = 97.94%, P < 0.001), and publication bias could be observed on the funnel plot. Publication bias in studies assessing the total proportion of vitamin D deficiency in T1D was analyzed using Begg’s test (z= 1.88; P= 0.060), Egger’s test (P= 0.000) and the funnel plot (Fig. 3).
Subgroup analyses were carried out depending on the publication year, study design, classification of vitamin D, season and geographical region of the studies, with Table 2 presenting the estimated proportion of vitamin D deficiency after the analysis.
All included studies were from 2009 to 2022. Twenty-one studies were published between 2009 and 2015, and 24 were published between 2016 and 2022. In contrast with the literature data of the previous six years (48%, 95% CI; 36-59%), more recent publications tended to yield a low proportion of vitamin D deficiency (43%, 95% CI; 31-56%). By comparing study designs, the subgroup analysis showed that a higher proportion of vitamin D deficiency could be noted in case‒control studies (58%, 95% CI; 45-72%), followed by one cohort study (51%, 95% CI; 45-58%) and 19 cross-sectional studies (31%, 95% CI; 22-40%), with the lowest proportion identified for 2 longitudinal studies (22%, 95% CI; 20-25%), but with significant heterogeneity. The proportion of vitamin D deficiency in children and/or adolescents with T1D was highest in Africa (65%, 95% CI; 42-85%), followed by Asia (54%, 95% CI; 40-68%), Europe (50%, 95% CI; 32-69%), North America (24%, 95% CI; 15-34%) and Oceania (15%, 95% CI; 12-18%), with a significant difference among the five subgroups (P < 0.01). The proportion of vitamin D deficiency in children and/or adolescents with T1D in low-mid latitudes was 56% (95% CI; 38-72%), followed by low latitudes (50%, 95% CI; 12-88%), mid-high latitudes (42%, 95% CI; 37-47%) and middle latitudes (39%, 95% CI; 29-50%). By vitamin D status, the results showed that a higher proportion of vitamin D deficiency could be noted at 30 ng/ml (87%, 95% CI; 82-92%), followed by one at 25 ng/ml (80%, 95% CI; 71-87%), 10 ng/ml (67%, 95% CI; 26-97%), 20 ng/ml (49%, 95% CI; 39-60%), and 15 ng/ml (24%, 95% CI; 11-41%), with the lowest proportion identified at12 ng/ml (14%, 95% CI; 9-20%). The subgroup analyses for different seasons showed that the proportion of vitamin D deficiency in winter tended to be significantly higher than that in summer (50%, 95% CI; 37-64% vs. 17%, 95% CI; 8-27%). In addition, studies conducted in spring reported a higher proportion of vitamin D deficiency (28%, 95% CI; 23-33%) than those conducted in autumn (20%, 95% CI; 12-29%), but this was not significant (P > 0.01).
Sensitivity analysis was carried out to examine the influence of any particular study. To determine whether potential publication bias existed in the reviewed literature, Egger’s test was also carried out. The results of Egger’s test (P < 0.05) did suggest the existence of publication bias. Thus the publication bias of this study was corrected using the trim-and-fill method. The results showed that publication bias had little effect on the combined amount of results, indicating that the robustness of the results of this study was high.
Thirty-five studies involving 5,862 participants were included in the meta-analysis of the rate of vitamin D insufficiency among children and/or adolescents with T1D. In this case, the random effects model indicated that the cumulative proportion was 33.0% (95% CI; 27-38%). Considerable heterogeneity was also observed across studies (I2 = 94.27%, P < 0.01). Analyses of publication bias for studies estimating the total proportion of vitamin D insufficiency were also conducted, with biases determined based on Begg’s test (z = 0.67; P = 0.504), Egger’s test (P = 0.614) and the funnel plot.
Thirty-nine studies, grouping 6,490 individuals from Europe (n = 11), Asia (n = 17), Africa (n = 1), North America (n = 9), and Oceania (n = 1), assessed the proportion of vitamin D sufficiency in children and/or adolescents with T1D. In this case, the proportion was estimated to be 27% (95% CI; 19-35%; I2 = 97.87%). Analyses of publication bias for studies estimating the total proportion of vitamin D sufficiency were also performed, with biases determined as before (i.e., with Begg’s test (z = 0.11; P = 0.913) and Egger’s test (P= 0.007) and the funnel plot). Sensitivity analyses further revealed 2 studies were found to be off-center, and after omitting it [38,65], the biases were again determined by both Begg’s test (z= 0.29; P= 0.773) and Egger’s test (P = 0.509).