ORIGINAL ARTICLE
Vitamin D Deficiency in Children With Fractures Jamie Jaqua Contreras, MD,* Brian Hiestand, MD, MPH,* James C. O'Neill, MD,* Robert Schwartz, MD,† and Milan Nadkarni, MD* Objective: This study aimed to determine whether healthy children with fractures resulting from minor accidental trauma have a higher prevalence of vitamin D deficiency than that of healthy children without fractures. Methods: This was a prospective case-control study of ambulatory children younger than 18 years with and without fractures in a pediatric emergency department. Evaluation included serum 25-hydroxyvitamin D (25 (OH)D) level, complete metabolic panel, and phosphorus level. Vitamin D deficiency was defined as a 25(OH)D level of less than 20 ng/mL and insufficiency less than 30 ng/mL but 20 ng/mL or greater. A level of 30 ng/mL or greater was considered sufficient. Fisher exact test was used to test for association between 25(OH)D level and fracture status. Logistic regression was used to examine the relationship between 25(OH)D levels and the odds of fracture, conditioned on season, age, race, body weight percentile, history of fracture, multivitamin use, and estimated daily milk intake. Results: The sample included 100 case and 100 control patients. There was no statistical difference in median 25(OH)D levels between fracture and control groups (26.7 vs 25.45 ng/mL, P = 0.84). There was no difference in the proportion of patients with sufficient 25(OH)D levels or in the distribution of sufficient, insufficient, and deficient. After adjusting for male sex and season of enrollment, vitamin D sufficiency was not a significant predictor of fracture status in a multiple variable logistic model (odds ratio, 0.94; 95% confidence interval, 0.51–1.77; Wald P = 0.859). Conclusions: We found no relationship between vitamin D deficiency and fracture risk in our study population.
age 8 to 24 months with vitamin D deficiency. James et al10 found that 24% of children with upper extremity fractures were vitamin D deficient and 41% were insufficient (a 25-hydroxyvitamin D (25(OH)D) level of 20–32 ng/mL). One study by Ryan et al11 suggest that vitamin D deficiency in an African American population is associated with higher odds of forearm fracture. It is the only case-control study, to our knowledge, thus far, which compared fracture risk to vitamin D levels. However, the study was limited to African American children age 5 to 9 years, and fracture type was limited to forearm fractures in case patients. A study performed by Schilling et al12 reported that it was unlikely that a strong association exists between low vitamin D levels and multiple fractures or fracture types concerning for nonaccidental trauma. However, this study was limited to children younger than 2 years and did not include a control group of children without fractures. Schilling et al categorized 25(OH)D levels of 30 ng/mL or greater as sufficient and levels of 20 ng/mL or greater but less than 30 ng/mL as insufficient. All of these studies defined vitamin D deficiency as a 25(OH)D level of less than 20 ng/mL. The objective of this study was to determine the 25(OH)D levels in healthy children presenting to an emergency department with fractures resulting from accidental minor injuries and compare them with a nonfracture control group. Our hypothesis was that children with fractures would have a higher prevalence of vitamin D deficiency than children without fractures.
Key Words: vitamin D deficiency, fractures, injury (Pediatr Emer Care 2014;30: 777–781)
V
itamin D is an essential nutrient for calcium absorption, the formation of bones, and maintaining bone health during childhood.1 Extreme vitamin D deficiency, leading to rickets, is a well-defined disease process in pediatrics. Rickets became much less common once nutrition standards led to food items (primarily dairy products) fortified with vitamin D.2 In recent years, a high prevalence of vitamin D deficiency has been reported in children of all ages throughout the United States and various countries around the world.3 Obesity and other childhood disease processes may limit vitamin D absorption and lead to deficient bone growth and risk for fracture (sickle cell, chronic inflammatory conditions, and childhood cancers).4–7 Dark skin pigmentation, limited exposure to sunlight, season, age, as well as inadequate dietary intake and milk consumption have also been shown to be associated with vitamin D deficiency.1,2,8 It is unclear, however, if vitamin D deficiency increases the risk of fracture in the pediatric population. There have been very few studies thus far that have looked into the correlation between vitamin D deficiency and risk of fracture in the pediatric population. Perez-Rossello et al9 found a prevalence of zero fracture in 35 otherwise healthy infants and toddlers From the Departments of *Emergency Medicine, and †Pediatrics, Wake Forest University Health Sciences, Winston-Salem, NC. Disclosure: The authors declare no conflict of interest. Reprints: Jamie Jaqua Contreras, MD, Department of Emergency Medicine, Wake Forest University Health Sciences, Medical Center Blvd, Winston-Salem, NC 27157 (e‐mail: [email protected]). Copyright © 2014 by Lippincott Williams & Wilkins ISSN: 0749-5161
Pediatric Emergency Care • Volume 30, Number 11, November 2014
METHODS Study Design and Setting We performed a prospective case-control study approved by the institutional review board at Wake Forest Baptist Health in Winston-Salem, NC. Subjects were enrolled in the pediatric emergency department, which serves as a tertiary referral center and level I trauma center for a multicounty catchment area.
Patient Population All participants were recruited through the pediatric emergency department. All patients were younger than 18 years. Control patients were those being evaluated for various problems, including but not limited to minor injuries that did not result in fractures, abdominal pain, headaches, and so on, often requiring laboratory studies or intravenous medications. Case patients were required to have an examination for a history of an injury resulting in a fracture confirmed by x-ray as interpreted by a radiologist. Exclusion criteria for both groups included children who were nonambulatory, malnourished, on steroidal or anticonvulsant medication; children with osteogenesis imperfecta, rickets, or other bone disorders; those with diabetes mellitus, kidney disease, cystic fibrosis, malabsorptive, or chronic gastrointestinal illnesses; those on any medications determined to affect vitamin D levels; or those who have any other chronic illness that may affect nutrition and/or vitamin D levels. Nonambulatory children were excluded because of the higher likelihood of nonaccidental or pathological fractures in infants or children with musculoskeletal disorders. Exclusion criteria for case patients also included those with pathological fractures or www.pec-online.com
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
777
Pediatric Emergency Care • Volume 30, Number 11, November 2014
Contreras et al
those involved in level I or II trauma activations. Written informed consent was obtained from every participant's guardian, and written assent was obtained for all children 7 years and older.
Data Collection We interviewed the patients and/or their guardians to obtain information regarding race/ethnicity, medical history, medications, mechanism of injury (for case patients), daily intake of milk, soda, daily use of vitamin supplements, and history of fracture. Age, sex, and weight were obtained from the patient's chart at the time of enrollment. Height was recorded based either on patient report or measured height.
Laboratory Evaluations Peripheral venous blood samples were obtained by a nurse in the pediatric emergency department and sent to 1 of 2 Laboratory Corporation of America locations, either in Winston-Salem, NC, or Burlington, NC, for analysis. Both locations are accredited by the College of American Pathologists and the Centers for Medicare and Medicaid Services Clinical Laboratory Improvement Amendments. A 25(OH)D level, phosphorus level, and a complete metabolic panel were obtained on all participants enrolled in the study. 25-hydroxyvitamin D level was measured using immunochemiluminometric assay. Phosphorus and calcium levels were measured using colorimetric methodology. Alkaline phosphatase was measured using kinetic methodology. Laboratory results were not available to the treating clinicians.
Study Variables and Definitions Vitamin D deficiency was defined as a serum 25(OH)D level of less than 20 ng/mL and vitamin D insufficiency as a level less than 30 ng/mL but greater than or equal to 20 ng/mL. A level of 30 ng/mL or greater was considered sufficient. Given the distribution of race and ethnicity in the study sample, race was dichotomized for analytic considerations as white versus non-white, although discrete categorizations were also reported. Body weight percentile was obtained from About.com Pediatrics Growth Chart Percentiles Calculator (http://pediatrics.about.com/cs/growthcharts2/ a/percentiles.htm). Given that the study period ran over several months with seasonal variation in sunlight intensity and ultraviolet light exposure, a categorical variable defining seasons as fall (September, October, and November), winter (December, January, and February), spring (March, April, and May), and summer (June, July, and August) was established based on the month of enrollment to explore the potential interplay of season, 25(OH)D levels, and fracture incidence.
examination and by the Shapiro-Wilk test.14 We did not adjust the α for multiple comparisons because these represent exploratory analyses and not the primary hypothesis. Finally, a multiple variable logistic regression model was constructed to examine the relationship between 25(OH)D levels and the odds of fracture, conditioned on season, age, race, body weight percentile, history of fracture, multivitamin use, and estimated daily milk intake. The fractional polynomial method15 and graphic examination were used to examine continuous variables in the model, and no reason was found to treat these variables as anything other than linear and continuous. There were no missing data in the study sample, so manual backward variable elimination was used to produce a parsimonious model, requiring a Wald P < 0.05 for variable retention in the model for linear or dichotomous categorical variables and likelihood ratio testing at the same P value for categorical variables with more than 2 levels. Vitamin D was forcibly retained in the model regardless of significance because of the nature of the primary hypothesis. The model was examined for multicollinearity, overly influential or outlier covariate patterns, and specification errors. We used Pearson goodness-of-fit testing to evaluate the final model because of the predominance of categorical variables in the model. Given the estimated presence of 100 cases (fractures), we felt that our model would not be overfit, based on the standard “rule of 10” for predictor covariate degree-of-freedom ratios to cases.16 All statistics were calculated using Stata/IC 11.2 (College Station, TX).
RESULTS Convenience samples of both case and control patients were collected from October 2012 through January 2013. During the study period, 222 potential subjects were approached for consent/ assent. Sixteen subjects declined to participate, and we were unable to obtain blood samples on 6 subjects. This resulted in 100 cases and 100 controls enrolled with usable blood specimens (Fig. 1). Table 1 provides the demographic and historical features of the 2 samples. The children in the fracture arm were slightly younger and consisted of more males than the children in the control arm. Table 2 presents the relationship between 25(OH)D levels and fracture status. There was no statistical difference in median 25(OH)D levels between fracture and control groups (26.7 vs 25.45 ng/mL, P = 0.84). Likewise, when approached categorically, there was no difference in the proportion of patients with sufficient 25(OH)D levels (≥30 ng/mL) or in the distribution of sufficient, insufficient, and deficient. We constructed a multiple variable logistic model examining the relationship of sufficient 25(OH)D levels with fracture status.
Statistical Considerations To examine the primary hypothesis, we compared the proportion of patients with sufficient 25(OH)D levels between the fracture group and the control group. Using Fisher exact test, with an α of 0.05 and a β of 0.20, we estimated that 91 subjects per group would be required to reliably detect a statistically significant difference in proportions of subjects with vitamin D sufficiency of 20% (80% vs 60%), based on previous reports in the literature.12,13 To account for enrollment vagaries and screen failures, we enrolled 100 subjects in each group. To explore secondary associations, univariate relationships between 25(OH)D levels and patient-level covariates were compared using Fisher exact test for categorical variables and Wilcoxon rank sum analysis for continuous variables. We used nonparametric statistics for continuous variables after examination of 25(OH)D levels revealed a nonparametric distribution, both by direct graphic
778
www.pec-online.com
FIGURE 1. Enrollment flow diagram. © 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Pediatric Emergency Care • Volume 30, Number 11, November 2014
TABLE 1. Demographic and Historical Characteristics of the Cohort Fractures (n = 100) Controls (n = 100) Age Male Race White Black Hispanic Other Season Summer Fall Winter Spring Previous fracture Milk intake (glasses/day) Multivitamin use Weight percentile
10.1 (9.2–11.0) 70 (61–79)
11.0 (10.0–11.9) 50 (40–60)
54 (44–64) 22 (14–30) 19 (11–27) 5 (2–11)
57 (47–67) 31 (22–40) 11 (5–17) 1 (0–5)
15 (8–22) 32 (23–41) 32 (23–41) 21 (13–29) 34 (25–43) 1.8 (1.5–2.0) 24 (15–33) 64 (58–69)
7 (2–12) 34 (25–43) 20 (12–28) 39 (29–49) 33 (24–42) 1.6 (1.3–1.8) 21 (13–29) 69 (64–75)
Categorical variables are presented as percentage and 95% confidence intervals, and continuous variables are presented as means with 95% confidence intervals.
Table 3 contains the parsimonious model retained after the elimination of nonexplanatory variables. After adjusting for male sex and season of enrollment, vitamin D sufficiency was not a significant predictor of fracture status (odds ratio [OR], 0.94; 95% confidence interval [CI], 0.51–1.77, Wald P = 0.859). The area under the curve of the model was 0.68 (95% CI, 0.60–0.75). The Pearson goodness-of-fit test provided no reason to reject the model as poorly fit (P = 0.70). When the model was constructed using severe vitamin D deficiency as the primary predictor, there was no substantive change in the model, nor was severe vitamin D deficiency any more predictive in the multiple variable model (OR, 1.01; 95%CI, 0.48–2.11; Wald P = 0.979). In terms of secondary hypotheses exploring the relationship between 25(OH)D levels and patient-level covariates, we noted the following results (Table 4). Median daily reported milk intake was higher in children with sufficient 25(OH)D levels than those with any degree of insufficiency (2 glasses daily vs 1 glass daily, P = 0.024 by Wilcoxon rank sum test). Multivitamin use was higher in children with sufficient 25(OH)D levels than in those with insufficiency (28 [38.4%; 95% CI, 27.2–50.5%] vs 17 [13.4%; 95% CI, 8.0–20.6%], P < 0.001 by Fisher exact test). We did not find a statistically significant relationship between body weight percentile TABLE 2. Association Between 25(OH)D Levels and Fracture Status Fracture (n = 100) Control (n = 100) 25(OH)D level, ng/mL 25(OH)D categories Sufficient (≥30 ng/mL) Insufficient (20–29.99 ng/mL) Deficient (
Vitamin D Deficiency in Children With Fractures Jamie Jaqua Contreras, MD,* Brian Hiestand, MD, MPH,* James C. O'Neill, MD,* Robert Schwartz, MD,† and Milan Nadkarni, MD* Objective: This study aimed to determine whether healthy children with fractures resulting from minor accidental trauma have a higher prevalence of vitamin D deficiency than that of healthy children without fractures. Methods: This was a prospective case-control study of ambulatory children younger than 18 years with and without fractures in a pediatric emergency department. Evaluation included serum 25-hydroxyvitamin D (25 (OH)D) level, complete metabolic panel, and phosphorus level. Vitamin D deficiency was defined as a 25(OH)D level of less than 20 ng/mL and insufficiency less than 30 ng/mL but 20 ng/mL or greater. A level of 30 ng/mL or greater was considered sufficient. Fisher exact test was used to test for association between 25(OH)D level and fracture status. Logistic regression was used to examine the relationship between 25(OH)D levels and the odds of fracture, conditioned on season, age, race, body weight percentile, history of fracture, multivitamin use, and estimated daily milk intake. Results: The sample included 100 case and 100 control patients. There was no statistical difference in median 25(OH)D levels between fracture and control groups (26.7 vs 25.45 ng/mL, P = 0.84). There was no difference in the proportion of patients with sufficient 25(OH)D levels or in the distribution of sufficient, insufficient, and deficient. After adjusting for male sex and season of enrollment, vitamin D sufficiency was not a significant predictor of fracture status in a multiple variable logistic model (odds ratio, 0.94; 95% confidence interval, 0.51–1.77; Wald P = 0.859). Conclusions: We found no relationship between vitamin D deficiency and fracture risk in our study population.
age 8 to 24 months with vitamin D deficiency. James et al10 found that 24% of children with upper extremity fractures were vitamin D deficient and 41% were insufficient (a 25-hydroxyvitamin D (25(OH)D) level of 20–32 ng/mL). One study by Ryan et al11 suggest that vitamin D deficiency in an African American population is associated with higher odds of forearm fracture. It is the only case-control study, to our knowledge, thus far, which compared fracture risk to vitamin D levels. However, the study was limited to African American children age 5 to 9 years, and fracture type was limited to forearm fractures in case patients. A study performed by Schilling et al12 reported that it was unlikely that a strong association exists between low vitamin D levels and multiple fractures or fracture types concerning for nonaccidental trauma. However, this study was limited to children younger than 2 years and did not include a control group of children without fractures. Schilling et al categorized 25(OH)D levels of 30 ng/mL or greater as sufficient and levels of 20 ng/mL or greater but less than 30 ng/mL as insufficient. All of these studies defined vitamin D deficiency as a 25(OH)D level of less than 20 ng/mL. The objective of this study was to determine the 25(OH)D levels in healthy children presenting to an emergency department with fractures resulting from accidental minor injuries and compare them with a nonfracture control group. Our hypothesis was that children with fractures would have a higher prevalence of vitamin D deficiency than children without fractures.
Key Words: vitamin D deficiency, fractures, injury (Pediatr Emer Care 2014;30: 777–781)
V
itamin D is an essential nutrient for calcium absorption, the formation of bones, and maintaining bone health during childhood.1 Extreme vitamin D deficiency, leading to rickets, is a well-defined disease process in pediatrics. Rickets became much less common once nutrition standards led to food items (primarily dairy products) fortified with vitamin D.2 In recent years, a high prevalence of vitamin D deficiency has been reported in children of all ages throughout the United States and various countries around the world.3 Obesity and other childhood disease processes may limit vitamin D absorption and lead to deficient bone growth and risk for fracture (sickle cell, chronic inflammatory conditions, and childhood cancers).4–7 Dark skin pigmentation, limited exposure to sunlight, season, age, as well as inadequate dietary intake and milk consumption have also been shown to be associated with vitamin D deficiency.1,2,8 It is unclear, however, if vitamin D deficiency increases the risk of fracture in the pediatric population. There have been very few studies thus far that have looked into the correlation between vitamin D deficiency and risk of fracture in the pediatric population. Perez-Rossello et al9 found a prevalence of zero fracture in 35 otherwise healthy infants and toddlers From the Departments of *Emergency Medicine, and †Pediatrics, Wake Forest University Health Sciences, Winston-Salem, NC. Disclosure: The authors declare no conflict of interest. Reprints: Jamie Jaqua Contreras, MD, Department of Emergency Medicine, Wake Forest University Health Sciences, Medical Center Blvd, Winston-Salem, NC 27157 (e‐mail: [email protected]). Copyright © 2014 by Lippincott Williams & Wilkins ISSN: 0749-5161
Pediatric Emergency Care • Volume 30, Number 11, November 2014
METHODS Study Design and Setting We performed a prospective case-control study approved by the institutional review board at Wake Forest Baptist Health in Winston-Salem, NC. Subjects were enrolled in the pediatric emergency department, which serves as a tertiary referral center and level I trauma center for a multicounty catchment area.
Patient Population All participants were recruited through the pediatric emergency department. All patients were younger than 18 years. Control patients were those being evaluated for various problems, including but not limited to minor injuries that did not result in fractures, abdominal pain, headaches, and so on, often requiring laboratory studies or intravenous medications. Case patients were required to have an examination for a history of an injury resulting in a fracture confirmed by x-ray as interpreted by a radiologist. Exclusion criteria for both groups included children who were nonambulatory, malnourished, on steroidal or anticonvulsant medication; children with osteogenesis imperfecta, rickets, or other bone disorders; those with diabetes mellitus, kidney disease, cystic fibrosis, malabsorptive, or chronic gastrointestinal illnesses; those on any medications determined to affect vitamin D levels; or those who have any other chronic illness that may affect nutrition and/or vitamin D levels. Nonambulatory children were excluded because of the higher likelihood of nonaccidental or pathological fractures in infants or children with musculoskeletal disorders. Exclusion criteria for case patients also included those with pathological fractures or www.pec-online.com
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
777
Pediatric Emergency Care • Volume 30, Number 11, November 2014
Contreras et al
those involved in level I or II trauma activations. Written informed consent was obtained from every participant's guardian, and written assent was obtained for all children 7 years and older.
Data Collection We interviewed the patients and/or their guardians to obtain information regarding race/ethnicity, medical history, medications, mechanism of injury (for case patients), daily intake of milk, soda, daily use of vitamin supplements, and history of fracture. Age, sex, and weight were obtained from the patient's chart at the time of enrollment. Height was recorded based either on patient report or measured height.
Laboratory Evaluations Peripheral venous blood samples were obtained by a nurse in the pediatric emergency department and sent to 1 of 2 Laboratory Corporation of America locations, either in Winston-Salem, NC, or Burlington, NC, for analysis. Both locations are accredited by the College of American Pathologists and the Centers for Medicare and Medicaid Services Clinical Laboratory Improvement Amendments. A 25(OH)D level, phosphorus level, and a complete metabolic panel were obtained on all participants enrolled in the study. 25-hydroxyvitamin D level was measured using immunochemiluminometric assay. Phosphorus and calcium levels were measured using colorimetric methodology. Alkaline phosphatase was measured using kinetic methodology. Laboratory results were not available to the treating clinicians.
Study Variables and Definitions Vitamin D deficiency was defined as a serum 25(OH)D level of less than 20 ng/mL and vitamin D insufficiency as a level less than 30 ng/mL but greater than or equal to 20 ng/mL. A level of 30 ng/mL or greater was considered sufficient. Given the distribution of race and ethnicity in the study sample, race was dichotomized for analytic considerations as white versus non-white, although discrete categorizations were also reported. Body weight percentile was obtained from About.com Pediatrics Growth Chart Percentiles Calculator (http://pediatrics.about.com/cs/growthcharts2/ a/percentiles.htm). Given that the study period ran over several months with seasonal variation in sunlight intensity and ultraviolet light exposure, a categorical variable defining seasons as fall (September, October, and November), winter (December, January, and February), spring (March, April, and May), and summer (June, July, and August) was established based on the month of enrollment to explore the potential interplay of season, 25(OH)D levels, and fracture incidence.
examination and by the Shapiro-Wilk test.14 We did not adjust the α for multiple comparisons because these represent exploratory analyses and not the primary hypothesis. Finally, a multiple variable logistic regression model was constructed to examine the relationship between 25(OH)D levels and the odds of fracture, conditioned on season, age, race, body weight percentile, history of fracture, multivitamin use, and estimated daily milk intake. The fractional polynomial method15 and graphic examination were used to examine continuous variables in the model, and no reason was found to treat these variables as anything other than linear and continuous. There were no missing data in the study sample, so manual backward variable elimination was used to produce a parsimonious model, requiring a Wald P < 0.05 for variable retention in the model for linear or dichotomous categorical variables and likelihood ratio testing at the same P value for categorical variables with more than 2 levels. Vitamin D was forcibly retained in the model regardless of significance because of the nature of the primary hypothesis. The model was examined for multicollinearity, overly influential or outlier covariate patterns, and specification errors. We used Pearson goodness-of-fit testing to evaluate the final model because of the predominance of categorical variables in the model. Given the estimated presence of 100 cases (fractures), we felt that our model would not be overfit, based on the standard “rule of 10” for predictor covariate degree-of-freedom ratios to cases.16 All statistics were calculated using Stata/IC 11.2 (College Station, TX).
RESULTS Convenience samples of both case and control patients were collected from October 2012 through January 2013. During the study period, 222 potential subjects were approached for consent/ assent. Sixteen subjects declined to participate, and we were unable to obtain blood samples on 6 subjects. This resulted in 100 cases and 100 controls enrolled with usable blood specimens (Fig. 1). Table 1 provides the demographic and historical features of the 2 samples. The children in the fracture arm were slightly younger and consisted of more males than the children in the control arm. Table 2 presents the relationship between 25(OH)D levels and fracture status. There was no statistical difference in median 25(OH)D levels between fracture and control groups (26.7 vs 25.45 ng/mL, P = 0.84). Likewise, when approached categorically, there was no difference in the proportion of patients with sufficient 25(OH)D levels (≥30 ng/mL) or in the distribution of sufficient, insufficient, and deficient. We constructed a multiple variable logistic model examining the relationship of sufficient 25(OH)D levels with fracture status.
Statistical Considerations To examine the primary hypothesis, we compared the proportion of patients with sufficient 25(OH)D levels between the fracture group and the control group. Using Fisher exact test, with an α of 0.05 and a β of 0.20, we estimated that 91 subjects per group would be required to reliably detect a statistically significant difference in proportions of subjects with vitamin D sufficiency of 20% (80% vs 60%), based on previous reports in the literature.12,13 To account for enrollment vagaries and screen failures, we enrolled 100 subjects in each group. To explore secondary associations, univariate relationships between 25(OH)D levels and patient-level covariates were compared using Fisher exact test for categorical variables and Wilcoxon rank sum analysis for continuous variables. We used nonparametric statistics for continuous variables after examination of 25(OH)D levels revealed a nonparametric distribution, both by direct graphic
778
www.pec-online.com
FIGURE 1. Enrollment flow diagram. © 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Pediatric Emergency Care • Volume 30, Number 11, November 2014
TABLE 1. Demographic and Historical Characteristics of the Cohort Fractures (n = 100) Controls (n = 100) Age Male Race White Black Hispanic Other Season Summer Fall Winter Spring Previous fracture Milk intake (glasses/day) Multivitamin use Weight percentile
10.1 (9.2–11.0) 70 (61–79)
11.0 (10.0–11.9) 50 (40–60)
54 (44–64) 22 (14–30) 19 (11–27) 5 (2–11)
57 (47–67) 31 (22–40) 11 (5–17) 1 (0–5)
15 (8–22) 32 (23–41) 32 (23–41) 21 (13–29) 34 (25–43) 1.8 (1.5–2.0) 24 (15–33) 64 (58–69)
7 (2–12) 34 (25–43) 20 (12–28) 39 (29–49) 33 (24–42) 1.6 (1.3–1.8) 21 (13–29) 69 (64–75)
Categorical variables are presented as percentage and 95% confidence intervals, and continuous variables are presented as means with 95% confidence intervals.
Table 3 contains the parsimonious model retained after the elimination of nonexplanatory variables. After adjusting for male sex and season of enrollment, vitamin D sufficiency was not a significant predictor of fracture status (odds ratio [OR], 0.94; 95% confidence interval [CI], 0.51–1.77, Wald P = 0.859). The area under the curve of the model was 0.68 (95% CI, 0.60–0.75). The Pearson goodness-of-fit test provided no reason to reject the model as poorly fit (P = 0.70). When the model was constructed using severe vitamin D deficiency as the primary predictor, there was no substantive change in the model, nor was severe vitamin D deficiency any more predictive in the multiple variable model (OR, 1.01; 95%CI, 0.48–2.11; Wald P = 0.979). In terms of secondary hypotheses exploring the relationship between 25(OH)D levels and patient-level covariates, we noted the following results (Table 4). Median daily reported milk intake was higher in children with sufficient 25(OH)D levels than those with any degree of insufficiency (2 glasses daily vs 1 glass daily, P = 0.024 by Wilcoxon rank sum test). Multivitamin use was higher in children with sufficient 25(OH)D levels than in those with insufficiency (28 [38.4%; 95% CI, 27.2–50.5%] vs 17 [13.4%; 95% CI, 8.0–20.6%], P < 0.001 by Fisher exact test). We did not find a statistically significant relationship between body weight percentile TABLE 2. Association Between 25(OH)D Levels and Fracture Status Fracture (n = 100) Control (n = 100) 25(OH)D level, ng/mL 25(OH)D categories Sufficient (≥30 ng/mL) Insufficient (20–29.99 ng/mL) Deficient (
