Effects of Fluoride Intake on Cortical and Trabecular Bone Microstructure at Early Adulthood using Multi-Row Detector Computed Tomography (MDCT)

Abstract

Purpose:

The aim of this study was to examine the effects of period-specific and cumulative fluoride (F) intake on bone at the levels of cortical and trabecular bone microstructural outcomes at early adulthood using emerging multi-row detector computed tomography (MDCT)-based novel techniques.

Methods:

Ultra-high resolution MDCT distal tibia scans were collected at age 19 visits under the Iowa Bone Development Study (IBDS), and cortical and trabecular bone microstructural outcomes were computed at the distal tibia using previously validated methods. CT scans of a tissue characterization phantom were used to calibrate CT numbers (Hounsfield units) into bone mineral density (mg/cc). Period-specific and cumulative F intakes from birth up to the age of 19 years were assessed for IBDS participants through questionnaire, and their relationships with MDCT-derived bone microstructural outcomes were examined using bivariable and multivariable analyses, adjusting for height, weight, maturity offset (years since age of peak height velocity (PHV)), physical activity (questionnaire for adolescents (PAQ-A)), healthy eating index version 2010 (HEI-2010) scores, and calcium and protein intakes.

Results:

MDCT distal tibia scans were acquired for 324 participants from among the total of 329 participants at age 19 visits. No motion artifacts were observed in any MDCT scans, and all images were successfully processed to measure cortical and trabecular bone microstructural outcomes. At early adulthood, males were observed to have stronger trabecular bone microstructural features, as well as thicker cortical bone (p < 0.01), as compared to age-similar females; however, females were found to have less cortical bone porosity as compared to males. Among participants with available F intake estimates (75 to 91% of the 324 with MDCT scans, depending on the period-specific F intake measure), no statistically significant associations were detected between any period-specific or cumulative F intake and bone microstructural outcomes of the tibia at the p < 0.01 level. Only for females, statistically suggestive associations (p < 0.05) were found between recent F intake (from 14-19 years) and trabecular mean plate width and trabecular thickness at the tibia. Those associations became somewhat weaker, but still statistically suggestive, for trabecular thickness in fully adjusted analysis with height, weight, PHV, calcium and protein intake, and HEI-2010 and PAQ-A scores as covariates.

Conclusion:

The findings show that the effects of lifelong or period-specific F intake from combined sources for adolescents typical to the United States Midwest region are not strongly associated with bone microstructural outcomes at age 19 years. These findings are generally consistent with previously reported results of IBDS analyses, which further confirms that effects of lifelong or period-specific F intake on skeletons in early adulthood are absent or weak, even at the levels of cortical and trabecular bone microstructural details.

1. INTRODUCTION

Osteoporosis is a common age-related disease characterized by reduced bone mineral density (BMD) and microstructural deterioration, which reduces bone strength and elevates risk of fractures [1]. Two hundred million people in the world and 54 million people in the United States have osteoporosis or low BMD [2]. Fracture incidence increases progressively with age [3]. The total number of hip fractures in 1990 was estimated as 1.26 million and the continued increase in life expectancy predicts approximately double the number of hip fractures to 2.6 million by the year 2025, and 4.5 million by the year 2050 [4]. Nearly, 40% of women and 13% of men suffer at least one osteoporotic fragility fracture in their lifetime [5]. Osteoporotic fractures are one of the most common causes of disability and a major contributor to medical care costs [1]. Osteoporotic hip fractures are especially devastating, reducing life expectancy by 10 to 20% [6], and more than three-quarters of all hip fractures occur in women [1].

Studies have shown that the antecedents of osteoporosis are established during childhood and adolescence [7]. Late adolescence/early adulthood is a period of major lifestyle behavior changes; however, major gaps exist in our understanding of the effects of modifiable and non-modifiable factors on bone development during the adolescent-adult transition period. This period is important in determining future risk for osteoporosis and bone fractures [8, 9]. Consequently, assessment of bone health in adolescence and early adulthood is important to better understand the relationships among nutrition, physical activity, and life-style factors and bone measures, and to identify factors associated with poor bone mineral accretion and microstructural development, producing individuals at elevated risk of developing osteoporosis at later ages and, thus, the need for preventive intervention [9]. Few studies have assessed bone outcomes and their determinants longitudinally from middle childhood to early adulthood. The Iowa Bone Development Study (IBDS) investigated the effects of fluoride (F), dietary factors, physical attributes, physical activity, lifestyle behaviors, and other factors on bone development and microstructure [10].

Few agents are capable of inducing marked anabolic effects on bone formation. F is one of the agents that is incorporated into bone mineral and has an anabolic effect [11, 12]. Sodium F (NaF) is a clinically-available F-based compound that stimulates bone formation [13]. A few studies have shown that long-term NaF treatment is capable of restoring bone mass at sites of trabecular bone [14]. A subgroup meta-analysis of F treatment studies [12] showed that a “low” therapeutic F dose (≤ 20 mg/day of F equivalents) was associated with a significant reduction in fracture risk. In a longitudinal study of normal and osteoporotic females (n = 219), Resch et al. [15] concluded that the osteogenic action of F therapy is not limited to the axial skeleton and that an increase in trabecular bone density also occurs at peripheral weight-bearing sites. Despite the documented anabolic effects of NaF on trabecular bone mass, controlled trials have failed to demonstrate any therapeutic advantage of NaF over placebo with respect to vertebral fracture rate [16–18]. In a long-term NaF therapy study on post-menopausal osteoporosis, it was observed that NaF had a positive effect on trabecular bone density; however, the number of new crush fractures of the spine during the first year increased in the treatment group [16]. In a histologic study involving biomechanical testing of iliac crest bone biopsies in osteoporotic patients, it was concluded that long-term NaF treatment was detrimental to bone quality [19]. In a meta-analysis of studies from 1991 to 1998 that focused on the effects of non-therapeutic F on bone strength, mass, and fracture rate, a substantial body of evidence was found suggesting that F up to 1 ppm does not have an adverse effect on bone strength, BMD or the incidence of fractures [20]. The United States National Academy of Sciences’ review of F in drinking water found insufficient evidence to establish that water F levels of 2 mg/L are associated with higher fracture rates; however, there was evidence that levels of 4 mg/L are associated with an elevated rate [21]. These conflicting observations on NaF therapy and lower-level exposures justify studying the effects of NaF treatment at the level of trabecular bone microstructure. In a two-year longitudinal histologic study of patients with primary osteoporosis (n = 15), it was observed that NaF treatment was associated with marked increase in bone volume and thickness, but no significant changes were observed in terms of trabecular bone surface area or connectivity number [22].

The human skeleton consists of cortical and trabecular compartments. The outer hard layer of the skeleton is composed of cortical bone (Cb), and the inner section is comprised of a spongy porous network referred to as trabecular bone (Tb). Histologic studies have convincingly demonstrated that Cb and Tb microstructural measures are critical determinants of bone-strength and fracture-risk [23–33]. This current study was designed to examine the effects of normal (non-therapeutic) period-specific and longitudinal F intakes on measures of Cb and Tb microstructure in early adulthood using emerging multi-row detector computed tomography (MDCT)-based techniques.

2. MATERIALS AND METHODS

2.1. Study Design

Longitudinal data on F intakes during different age periods and cumulatively from birth to age 19, and emerging MDCT-derived trabecular and cortical bone microstructural metrics of the distal tibia at age 19 were used. The goal of our study was to determine if there was any association between non-therapeutic F intake from typical diets from the United States Midwest region, including many with fluoridated water, and measures of cortical and trabecular bone microstructure, and we hypothesized that there is no negative effect on bone microstructure of these typical F intakes.

2.2. Study Participants and Procedures

The study sample comprised participants in the Iowa Bone Development Study (IBDS) at age 19 who formed a subset from the longitudinal Iowa Fluoride Study (IFS) cohort [34]. Families were invited into the IFS during 1992 to 1995 from eight postpartum hospitals in Iowa, accounting for approximately 20% of the Iowa births [34]. From 1998 until 2000, the IFS participants were invited to participate in the IBDS. Participating cohort members underwent several clinical and radiographic examinations, questionnaire data collection, and accelerometry measurements beginning at age 5 years [10]. MDCT scans of the distal tibia were obtained between 2011 and 2014 at approximately age 19 years. All components of the IFS and the IBDS were approved by the Institutional Review Board of the University of Iowa. Informed consent and assent for participation in the study were obtained from the study participants; consent was obtained from parents when the participants were younger.

Weight and height were measured by research staff during the visit at which the MDCT images were obtained. Weight was measured in tenths of kilograms with a Healthometer physician scale (Continental, Bridgeview, IL). Height was measured in tenths of centimeters with a Harpenden stadiometer (Holtrain, United Kingdom). Measurements were acquired for the participants while they were wearing indoor clothing without shoes.

Cross-sectional dietary intakes of calcium and protein for the participants were assessed from the Diet History Questionnaire II (DHQ-II) [35], which was completed at the age 19 visit. DHQ-II is a freely available food questionnaire developed by the Risk Factor Monitoring and Methods Branch (RFMMB) staff at the National Cancer Institute. The Healthy Eating Index version 2010 (HEI-2010), calculated from DHQ-II dietary outcomes, was used to measure diet quality in terms of conformance to federal dietary guidance [36]. HEI-2010 can be utilized to assess compliance with the 2010 Dietary Guidelines for Americans, monitor changes in dietary patterns, and to understand relationships between diet and health-related outcomes. HEI-2010 uses a caloric density approach to determine the overall quality of the diet thus eliminating the need to assess various subgroups based on age, sex, and activity level. Specifically, HEI-2010 has 12 components – nine components focus on adequacy (food to eat enough of to get the nutrients needed for overall good health) and three components focus on moderation or those dietary components that should be limited or consumed in small amounts.

Accelerometry measurements were collected close to the times of bone scans to objectively assess the physical activity (PA) of study participants [37, 38]. During the autumn, at or close to age 19, the Model GT3X ActiGraph accelerometer (Pensacola, Florida, USA) was mailed to participants with directions to wear the monitor for 5 consecutive days, including both weekend days. Moderate-to-vigorous-physical activity (MVPA) was defined as total time (minutes) with ≥2,296 counts per minute. Accelerometer data were used after being re-integrated to 1 min epochs. Because at age 19, objectively assessed PA was available for a substantially-reduced study sample, the PAQ-A (Physical Activity Questionnaire for Adolescents) score was used as the main measure of PA (score range 1-5, with a higher score corresponding to a higher level of activity). As we previously reported [39] for the IBDS study cohort, this questionnaire showed good internal consistency and acceptable concurrent validity compared with ActiGraph accelerometry measures. The PAQ-A score was available for almost all participants.

Age of Peak Height Velocity (PHV) is an indicator of somatic maturity that represents the time of maximum growth in stature during adolescence. The age of PHV was estimated from the age 11,13 and 15 assessment visit data using Mirwald’s predictive equations for maturity offset [40]. These equations included age, sex, weight, height, sitting height, and leg length as predictors. The clinical examination (between age 11 and age 15) which provided an estimate of age at PHV that was closest to the actual clinical examination age was used as the best estimate. Maturity offset at age 19 (the focus of this analysis) was defined as years since age of peak height velocity ((age at MDCT scan)-(best estimate of age at PHV)).

2.3. Fluoride Intake

Frequencies and amounts of F intake among the study participants from different sources (water, other beverages, selected foods, dietary F supplements, and ingested F toothpaste) were assessed through detailed questionnaires which were mailed to parents at ages 1.5, 3, 6, 9 and 12 months, then every four months up to age 4 years, and then every 6 months up to age 19 years [34, 41–47].

Concerning water sources, respondents reported on each questionnaire about use of and daily quantities ingested from tap water from home, childcare, and school, bottled water, and other sources of water, as well as whether they were from public water sources or wells, and filtered or not [42]. Water F levels of non-filtered public water sources were obtained from state records, while individual well sources and filtered public sources were assayed for F content in our laboratories with an F ion-specific electrode by the direct-read method [42]. F intake from water (in mg/day) at each time point was determined for each participant by multiplying the daily intake from each water source by its water F level and summing these source-specific intakes.

Concerning F intake at each time point from ready-to-drink (not reconstituted or diluted) beverages other than water, ingestion of and frequencies and quantities of different beverages were queried. Then they were linked to brand-, flavor-, and product size-specific F assays of many hundreds of beverages from F assays determined in our labs to determine the F intake from each, and they were summed to obtain the total intake from other beverages (in mg/day) [43, 44].

Information about intakes, frequencies, and quantities of ready-to-feed infant foods was queried at each time point and products were assayed for F content using a modified-Taves method of micro-diffusion and F-specific electrode [45]. Infant foods reconstituted with water also were assayed. Quantities of F from infant foods were determined by multiplying each product’s F content by daily intake. The results were summed across all infant products.

F intakes at each time point from selected foods that have substantial water added (e.g., oatmeal, gelatin, pasta, rice), beverages needing reconstitution (e.g., concentrated lemonade, orange juice, Koolaid), and diluted ready-to-drink beverages were determined by analyses of F content of each product when prepared with distilled water, and summing of the F from the water component (specific to each person’s water sources) and the non-water component (based on water F level and quantity of water added). These results then were multiplied by the daily quantity ingested and summed across all such foods and beverages (in mg/day).

At each time point, F intakes from toothpaste were determined through a series of specific questions about frequency of toothbrushing, use of toothpaste, specific toothpaste brand/type to determine F concentration (including regular concentration or prescription, high concentration products), quantity of toothpaste used, and estimated quantity ingested. From these, average daily F intake from F toothpaste was determined (in mg/day) [46].

F intakes from dietary F supplements (drops and tablets) at each time point were determined through a series of specific questions about use of such supplements, weeks of use and frequency of use per week, quantity of product ingested, and product-specific information about F dosages [47]. From these, an average daily F intake from dietary F supplements was determined (in mg/day).

From these sets of questions about F intake from these sources, category-specific F intakes were calculated for each specific age, and then they were summed to obtain age-specific, combined F intakes (in mg/day). Period-specific daily F intakes in mg F/day then were determined [10, 34, 41, 48] using area-under-the-curve (AUC). Each AUC required data at the upper and lower endpoints, with endpoints allowed to be interpolated from estimates within 7 months of the stated endpoints. Three specific multiyear periods of F intake were chosen as the main independent variables to represent early childhood (birth to 8.5 years), late childhood to early adolescence (8.5 to 14 years), and later adolescence to early adulthood (14 to 19 years), as well as lifelong intake (birth to 19 years). The cumulative ‘average’ daily F intake in mg from birth to age 19 years was calculated using AUC, with the requirements that each participant have at least one daily F intake estimate recorded, obtained or interpolated for each of the period-specific F intakes. If a time point was missing, linear interpolation using the nearest two points to the required time point was employed.

2.4. Distal Tibia Ultra-High Resolution CT Imaging

Distal tibia ultra-high resolution (UHR) MDCT scans were acquired at the University of Iowa Comprehensive Lung Imaging Center (ICLIC) on a 128-slice SOMATOM Definition Flash scanner (Siemens, Munich, Germany). The left leg was used for scanning unless a participant reported a prior fracture in the lower left leg, in which case the right leg was scanned. During ankle positioning for distal tibia scans, references of laser rays were used to align the tibial axis with the scanner center. Alignment of the tibial axis with the scanner center is important to achieve the highest spatial resolution for distal tibial UHR CT scans. An anterior-posterior projection CT scout scan was used to ensure the inclusion of the distal tibial end-plateau in the field-of-view (FOV). The reference of the distal tibial end-plateau was used to define different percentages of tibial sites for computation of bone measures, as will be described in the next section (Figure 1). The following CT scan parameters were used — single X-ray source spiral acquisition at 120 kV, 200 effective mAs, 1 second rotation speed, pitch factor: 1.0, number of detector rows: 16, scan time: 23.2 seconds, collimation: 16 × 0.6 mm, total effective dose equivalent: 170 μSv ≈ 20 days of environmental radiation in the Unites States [49]. The Siemens z-UHR scan mode was applied enabling Siemens double z sampling technology to be used. After scanning in a helical mode, images were reconstructed at 400 μm slice-thickness with 200 μm slicespacing and 150 μm pixel-size using a normal cone beam method with a special U70u kernel achieving high structural resolution. A Gammex RMI 467 Tissue Characterization Phantom (Gammex RMI, Middleton, WI) was separately scanned to calibrate CT Hounsfield units into BMD (mg/cc).

Figure 1.

ROIs for cortical and trabecular bone measures. References for distal tibial end-plateau and tibial length and percent peel approaches were used for ROI selection, adjusting for size- and shape-related variations among participants.

2.5. CT Image Processing and Bone Measures

Each CT distal tibia scan was visually inspected for any visible marks of motion and beam hardening artifacts. After clearance through the image quality control step, each CT scan was converted to a BMD (mg/cc) image using the matching Gammex phantom CT scan and an automated algorithm developed in our laboratory generating the calibration function from CT Hounsfield numbers to BMD values. Then the BMD image was interpolated at 150 μm isotropic voxel size using a windowed sync interpolation method [50]. All subsequent image processing operations were applied on BMD images at a 150 μm isotropic voxel size.

2.5.1. Region of Interest (ROI) Selection

A method was developed to generate axial cylindrical ROIs at physiologically consistent tibial sites after adjusting for participant-specific size-and shape-related variation. The method works on a filled-in tibia bone segmentation volume computed using multi-scale morphology and connectivity analyses [51]. This filled-in bone was used to determine – (1) the distal tibial end-plateau, (2) the tibial axis, (3) 2% cylindrical axial ROIs at specific sites of tibial length (Figure 1), and (4) the inner (60% peel) and outer (annular region between 30% and 60% peels) regions. The distal tibial end-plateau was located just above the first image slice containing a 2-D hole within the filled-in bone while tracing slices in the proximal to distal direction. The 60% peel region of filled-in bone proximal to 8% of the tibial length measured from the end-plateau, was used to determine the tibial axis orientation using a mean-square-error line-fitting algorithm. Cylindrical axial ROIs at different tibial locations (Figure 1) were determined after aligning the tibia axis with the image z-axis. A 2-D distance transform analysis of the realigned filled-in bone on individual image slices was applied to generate different percent peels.

2.5.2. Computation of Bone Microstructural Outcomes

The list of Cb and Tb measures examined in this study is presented in Table 1. Skeletonization converts a volumetric representation to digital medial surfaces and curves [52], which is an essential preprocessing step for quantitative analysis of Tb microstructure. In this study, fuzzy skeletonization [53] was used to serve this purpose as it eliminates thresholding-induced data loss associated with binarization needed for conventional skeletonization algorithms [52]. The trabecular volumetric bone mineral density (Tb.vBMD) (Table 1) measure was computed by averaging BMD values over a target ROI. Tensor scale analysis [54] was applied to compute plate-width at individual trabecular locations and distinguish between longitudinal and transverse trabeculae. Tensor scale analysis [54] was applied to identify longitudinal and transverse trabeculae. Tb.tBMD (Table 1) was computed as the volumetric BMD in an ROI contributed by transverse trabeculae. Volumetric topology and tensor scale analysis [55, 54] was applied to characterize trabecular plates and rods. Tb.PW (Table 1) was computed by averaging individual trabecular plate-widths over an ROI. The Tb.Th and Tb.Sp measures (Table 1) were computed using our star-line based algorithm, which was previously demonstrated to yield accurate measures across resolution regimes covering ex vivo and in vivo imaging methods [56]. A validated CT-based cortical bone segmentation algorithm [51] was used to segment Cb regions and compute the measures Cb.Th and Cb.Poro. All Tb measures were computed at the 4 to 6% distal tibial site over inner (60% peel region) and outer (the annular region between 30 and 60% peels) ROIs, while the Cb measures were obtained at the 14 to 16% tibial sites.

Table 1.

List of MDCT-derived trabecular bone (Tb) and cortical bone (Cb) measures examined in this study. The nomenclature of Tb measures used by Bouxsein et al. [76] and Chen et al. [49] is followed here wherever possible.

Parameter (unit)	Description
Trabecular bone
Tb.vBMD (mg/cc)	Volumetric trabecular BMD
Tb.tBMD (mg/cc)	Volumetric trabecular BMD contributed by transverse trabeculae [54]
Tb.pBMD (mg/cc)	Volumetric trabecular BMD contributed by trabecular plates [54]
Tb.NA (cm2/cm3)	Density of trabecular network area, i.e., the area of medial surface of segmented Tb microstructure
Tb.PW (μm)	Mean Tb plate-width [54]
Tb.Th (μm)	Mean trabecular thickness [56]
Tb.Sp (μm)	Mean trabecular separation, i.e., the space between individual trabeculae [56]
Cortical bone
Cb.Poro (no unit)	Mean cortical porosity computed by analyzing BMD over the Cb region [51]
Cb.Th (μm )	Mean cortical thickness computed using a Cb segmentation algorithm [51]

Cb: cortical bone; Tb: trabecular bone; BMD: bone mineral density

2.6. Statistical Analysis

All statistical analyses were stratified by sex, and normality was assessed using the Shapiro-Wilk test and by evaluating histograms. Since many variables of interest like F intakes, dietary calcium intake and protein intake had skewed distributions with high-value outliers, non-parametric (Spearman correlation) methods were used for analysis. Sex-specific Spearman rank–order correlation coefficients were estimated to investigate unadjusted bivariate associations between bone characteristics and F intakes and between bone characteristics and important covariates – weight, height, maturity offset, dietary protein intake, calcium intake from all sources, PAQ-A scores, and HEI-2010. Since calcium and vitamin D dietary measures were highly correlated, analyses using vitamin D intake were not conducted. Partial Spearman rank-order correlation coefficients were used to assess the associations between MDCT-derived outcomes and the estimated F intakes during different time periods and cumulatively, adjusted for covariates.

The significance level was set at p < 0.01 due to the many statistical tests conducted, while suggestive associations were defined as 0.01 < p < 0.05. Analyses were conducted using the Statistical Analysis System (SAS), version 9.4.

3. RESULTS

MDCT scans were obtained for IBDS cohort members at age 19; from among 329 participants with a study visit at age 19; 324 were successfully scanned using MDCT. Cohort members included in the analyses were mostly white (94.8%) and from middle to high socioeconomic status families, with approximately 50% of each of the mothers and fathers having at least a 4-year college degree, and 51% having annual family income of $60,000 or more in 2007. Table 2 shows sex-stratified descriptive statistics for the 324 study participants. Observed mean heights of male and female participants were consistent with average heights of men and women in the United States [57]. The mean BMI of both male and female participants indicated they tended to be moderately overweight as a group. There were significant differences between males and females in their height, weight, years since PHV, HEI-2010, PAQ-A, and calcium and protein intake. Females had better HEI-2010 scores than males. Mean F intakes were slightly greater than medians, and F intakes increased from early to late age-periods for both females and males. F intakes for males were generally higher than for females, with all intakes beyond the age of 8.5 years being statistically significantly greater.

Table 2.

Descriptive statistics of the study population stratified by sex.

	Females			Males
Variable	N	Mean (SD)	Median (Range)	N	Mean (SD)	Median (Range)
Age (years)	178	19.8 (0.7)	19.7 (18.9, 22.1)	146	19.8 (0.7)	19.7 (18.6, 22.1)
Height (cm)	178	166.4 (6.9)	166.3 (44.3, 181.0)	146	180.2 (7.7)**	179.8 (162.0, 200.9)
Weight (kg)	178	70.6 (19.3)	65.3 (44.5, 159.8)	146	84.7 (19.9)**	80.0 (51.3, 166.0)**
Maturity offset (yrs.)	177	7.9(1.0)	7.9(4.6,10.8)	144	6.1(1.1)**	5.9 (3.6,9.4)
HEI-2010 score (0-100)	177	59.5 (12.2)	60.5 (30.3, 90.8)	146	53.9 (11.8)**	53.7 (27.5, 87.8)
PAQ score	174	2.08 (0.72)	2.04 (1.00, 4.16)	139	2.44 (0.83)**	2.48 (1.00, 4.43)
Calcium (mg)	177	1004.4 (580.5)	865.9 (3.9, 3116.9)	146	1924.4 (1534.9)**	1582.0 (176.7, 9613.7)**
Protein (g)	177	58.2 (30.1)	51.1 (4.8, 227.1)	146	122.3 (78.0)**	104.6 (14.3, 506.0)**
Period-Specific Fluoride Intake (AUC mg F)
0 to 8.5 years	160	0.63 (0.27)	0.57 (0.17, 2.36)	135	0.69 (0.27)	0.66 (0.12, 1.55)*
8.5 to 14 years	149	0.68 (0.34)	0.62 (0.17, 2.57)	123	0.79 (0.35)**	0.73 (0.24, 2.21)**
14 to 19 years	137	0.80 (0.42)	0.73 (0.16, 2.48)	107	1.00 (0.51)**	0.89 (0.22, 3.03)**
0 to 19 years	143	0.69 (0.27)	0.64 (0.28, 2.33)	113	0.82 (0.33)**	0.74 (0.21, 1.92)**

^*p-value < 0.05,

^**p-value < 0.01 for sex comparisons; t-tests were used for normally distributed variables; both t-tests and non-parametric Wilcoxon rank sum tests were used for variables with skewed distribution.

PHV = peak height velocity; HEI-2010 = healthy eating index version 2010; PAQ = physical activity questionnaire; AUC

Maturity offset = (age-(best estimate of age of peak height velocity))

F = fluoride

Results of distal tibia CT imaging, cortical bone segmentation, and 3-D microstructural analysis for three female and three male participants with different vBMD are illustrated in Figure 2 and Figure 3, respectively. No visible marks of motion blur or beam hardening artifacts were observed in any of the 324 MDCT distal tibia scans acquired for this study. All scans were processed using fully automated and previously validated image processing software [49], and expected Cb and Tb microstructural measures were computed. A fixed scaling for all 2-D displays, and another fixed scaling for all 3-D displays, was used in both Figure 2 and Figure 3. The same display intensity setting and color-coding map were used for these figures. Surface rendition of the 3-D Tb microstructures and color-coded displays of individual trabecular plate-rod classifications [49, 55, 54] are shown in the last two columns. In general, for both sexes, participants with higher vBMD displayed more trabecular plates (green). Between males and females, visual inspection for these specific participants suggests that males have more plates as compared to their vBMD group-matched female counterparts. Table 3 presents sex-specific descriptive statistics for the MDCT-derived trabecular and cortical bone outcomes.

Figure 2.

MDCT-based bone microstructure in females with different vBMD. From top to bottom: Results for randomly selected female participants with vBMD close to the female-specific mean−SD; mean; and mean+SD, respectively. Columns 1 and 2 from left: Axial and sagittal slices from MDCT images after conversion of CT values into BMD; segmented cortical bone over 10 to 16% of distal tibia is overlaid on sagittal slices. Column 3: Surface rendition of MDCT-based 3-D trabecular microstructure reconstructed over 4 to 6% distal tibia with 60% peel from the outer cortex. Column 4: Color-coded display of individual trabecular plate-rod characterization.

Figure 3.

Same as Figure 2, but for three randomly selected male participants with vBMD close to male-specific mean−SD; mean; and mean+SD, respectively.

Table 3.

Distributions of sex-specific MDCT-derived trabecular and cortical bone measures.

Bone measures (units)	Mean (SD)	Min	5th pctl	10th pctl	25th pctl	Median	75th pctl	90th pctl	95th pctl	Max
Females (N=178)
Trabecular (Tb)
Tb.vBMD (mg/cc)	1164.4 (32.1)	1086.9	1113.2	1125.3	1144.6	1160.9	1185.8	1203.0	1213.1	1333.1
Tb.tBMD (mg/cc)	312.1 (86.3)	102.5	192.2	212.9	255.9	297.2	362.3	418.2	450.8	759.6
Tb.pBMD (mg/cc)	928.2 (120.9)	497.9	727.6	774.3	862.9	920.0	1016.5	1075.0	1107.8	1284.4
Tb.NA (cm2/cm3)	0.05(0.01)	0.0	0.0	0.0	0.0	0.0	0.1	0.1	0.1	0.1
Tb.PW (μm)	1211.0 (315.9)	658.2	836.9	888.3	997.3	1146.9	1375.4	1534.7	1762.1	3108.7
Tb.Th (μm)	161.6 (23.8)	116.0	134.8	139.9	149.1	158.8	169.9	181.5	188.2	337.8
Tb.Sp (μm)	439.6 (85.1)	252.2	314.6	342.8	375.4	440.0	484.0	546.4	594.5	835.5
Cortical (Cb)
Cb.Poro (no unit)	0.21 (0.02)	0.18	0.20	0.20	0.20	0.21	0.22	0.23	0.24	0.43
Cb.Th (μm)	2.01 (0.23)	1.45	1.67	1.73	1.84	1.97	2.17	2.33	2.42	2.63
Males (N=146)
Trabecular (Tb)
Tb.vBMD (mg/cc)	1182.9 (27.7)**	1125.5	1137.5	1148.2	1165.0	1180.7	1203.8	1217.6	1225.4	1256.7
Tb.tBMD (mg/cc)	360.4 (74.2)**	187.4	252.0	268.3	309.5	350.2	410.8	466.6	479.9	600.1
Tb.pBMD (mg/cc)	1005.0 (103.1)**	616.4	819.9	867.3	941.7	1013.8	1079.0	1125.0	1148.8	1211.9
Tb.NA (cm2/cm3)	0.06 (0.01)**	0.03	0.04	0.05	0.05	0.06	0.07	0.08	0.08	0.11
Tb.PW (μm)	1375.7 (300.2)**	775.6	944.5	1028.4	1175.7	1337.7	1554.4	1782.6	1868.6	2595.8
Tb.Th (μm)	173.2 (24.9)**	128.6	143.2	149.8	159.3	171.1	183.0	195.9	201.0	387.4
Tb.Sp (μm)	398.0 (63.9)**	281.9	304.3	318.0	348.5	397.1	432.0	474.1	529.7	569.9
Cortical (Cb)
Cb.Poro (no unit)	0.23 (0.02)**	0.19	0.21	0.21	0.22	0.23	0.23	0.24	0.24	0.47
Cb.Th (μm)	2.32 (0.27)**	1.71	1.89	1.99	2.13	2.30	2.54	2.64	2.77	3.09

^**p-value < 0.01 for comparisons between sexes using t-tests.

Tb.vBMD: volumetric BMD, Tb.tBMD: transpose BMD, Tb.pBMD: plate trabecula BMD, Tb.NA: trabecular network, Tb.PW: plate width, Tb.Th: trabecular thickness, Tb.Sp: trabecular spacing, Cb.Poro: cortical porosity, Cb.Th: cortical thickness pctl = percentile

Spearman correlations between F intakes for specific periods and cumulatively from birth to age 19 and MDCT-derived trabecular and cortical bone outcomes are shown in Table 4. Table 5 shows sex-specific associations between period-specific and cumulative F intakes and MDCT-derived bone outcomes fully adjusted for height, weight, time since PHV, calcium and protein intake, HEI-2010, and PAQ-A score (measure of physical activity). We obtained similar results using accelerometry-based MVPA instead of PAQ-A score, but the MVPA adjustment analysis was based on a substantially reduced sample size (data not shown). There were no statistically significant associations between any of the period-specific or cumulative F intakes and MDCT-derived bone outcomes for females or males in either the unadjusted or fully adjusted analyses.

Table 4.

Associations (Spearman correlation coefficient (p-value)) between period-specific fluoride intake (AUC) and MDCT-derived trabecular and cortical bone measures.

Bone Measures (units)	Daily Fluoride Intake
	Females				Males
	mg F 0-8.5 years (n=160)	mg F 8.5-14 years (n=149)	mg F 14-19 years (n=137)	mg F 0-19 years (n=143)	mg F 0-8.5 years (n=135)	mg F 8.5-14 years (n=123)	mg F 14-19 years (n=107)	mg F 0-19 years (n=113)
Trabecular (Tb)
Tb.vBMD (mg/cc)	0.02 (0.82)	0.00 (1.00)	0.08 (0.36)	0.06 (0.48)	0.10 (0.25)	0.04 (0.64)	0.05 (0.60)	0.11 (0.23)
Tb.tBMD (mg/cc)	0.03 (0.73)	0.00 (0.97)	0.10 (0.25)	0.07 (0.38)	0.09 (0.29)	0.05 (0.58)	0.04 (0.68)	0.12 (0.21)
Tb.pBMD (mg/cc)	−0.01 (0.88)	0.07 (0.40)	0.16 (0.07)	0.13 (0.13)	0.05 (0.54)	0.00 (0.99)	0.07 (0.46)	0.09 (0.33)
Tb.NA (cm2/cm3)	−0.04 (0.63)	0.02 (0.77)	0.17 (0.05)	0.10 (0.26)	0.11 (0.20)	0.03 (0.72)	0.08 (0.42)	0.17 (0.07)
Tb.PW (μm)	−0.00 (0.97)	0.08 (0.31)	0.18 (0.03)*	0.16 (0.05)	0.05 (0.58)	0.01 (0.95)	0.07 (0.45)	0.10 (0.29)
Tb.Th (μm)	−0.03 (0.71)	0.09 (0.29)	0.20 (0.02)*	0.16 (0.05)	0.04 (0.68)	−0.01 (0.94)	0.05 (0.60)	0.06 (0.50)
Tb.Sp (μm)	−0.03 (0.72)	0.03 (0.75)	−0.03 (0.75)	−0.02 (0.81)	−0.13 (0.12)	−0.10 (0.27)	−0.05 (0.60)	−0.15 (0.11)
Cortical (Cb)
Cb.Poro (no unit)	0.03 (0.68)	0.04 (0.60)	0.01 (0.91)	0.02 (0.82)	0.09 (0.32)	0.08 (0.36)	0.07 (0.49)	0.07 (0.47)
Cb.Th (μm)	−0.13 (0.09)	−0.01 (0.88)	0.11 (0.18)	0.01 (0.87)	−0.08 (0.36)	−0.05 (0.60)	0.02 (0.88)	−0.02 (0.85)

^*Suggestive (0.01<p-value <0.05).

Tb.vBMD: volumetric BMD, Tb.tBMD: transpose BMD, Tb.pBMD: plate trabecula BMD, Tb.NA: trabecular network, Tb.PW: plate width, Tb.Th: trabecular thickness, Tb.Sp: trabecular spacing, Cb.Poro: cortical porosity, Cb.Th: cortical thickness.

F = fluoride

Table 5.

Adjusted associations between period-specific fluoride intakes and MDCT-derived bone measures (Partial Spearman correlation coefficient (p-value) with height, weight, years since age of peak height velocity (PHV), calcium and protein intakes, Healthy Eating Index version 2010 (HEI-2010), and PAQ-score as covariates).

Bone Measures (units)	Daily Fluoride Intake
	Females				Males
	mg F 0-8.5 years (n=155)	mg F 8.5-14 years (n=144)	mg F 14-19 years (n=134)	mg F 0-19 years (n=141)	mg F 0-8.5 years (n=126)	mg F 8.5-14 years (n=114)	mg F 14-19 years (n=99)	mg F 0-19 years (n=105)
Trabecular (Tb)
Tb.vBMD (mg/cc)	0.02 (0.85)	−0.02 (0.82)	0.01 (0.89)	0.02 (0.85)	0.01 (0.89)	−0.02 (0.87)	−0.05 (0.61)	0.04 (0.72)
Tb.tBMD (mg/cc)	0.02 (0.77)	−0.01 (0.87)	0.04 (0.65)	0.03 (0.69)	0.01 (0.93)	−0.00 (0.96)	−0.06 (0.55)	0.05 (0.63)
Tb.pBMD (mg/cc)	−0.01 (0.91)	0.05 (0.54)	0.12 (0.18)	0.10 (0.24)	0.00 (0.99)	−0.02 (0.82)	0.01 (0.90)	0.08 (0.44)
Tb.NA (cm2/cm3)	−0.05 (0.57)	−0.01 (0.95)	0.12 (0.19)	0.05 (0.54)	0.02 (0.83)	−0.02 (0.82)	−0.03 (0.81)	0.12 (0.25)
Tb.PW (μm)	−0.00 (0.96)	0.07 (0.39)	0.14 (0.12)	0.14 (0.11)	−0.01 (0.88)	−0.04 (0.69)	0.01 (0.90)	0.07 (0.48)
Tb.Th (μm)	−0.02 (0.77)	0.09 (0.31)	0.18 (0.04)*	0.16 (0.07)	−0.02 (0.86)	−0.0 2(0.80)	0.01 (0.92)	0.0 5(0.61)
Tb.Sp (μm)	−0.02 (0.84)	0.06 (0.47)	0.05 (0.60)	0.04 (0.64)	−0.04 (0.64)	−0.04 (0.67)	0.06 (0.56)	−0.06 (0.56)
Cortical (Cb)
Cb.Poro (no unit)	−0.02 (0.83)	−0.01 (0.95)	−0.08 (0.35)	−0.04 (0.64)	0.01 (0.92)	0.03 (0.73)	0.03 (0.81)	−0.01(0.91)
Cb.Th (μm)	−0.12 (0.13)	0.01 (0.90)	0.10 (0.25)	−0.01 (0.91)	−0.12 (0.20)	0.01 (0.94)	−0.02 (0.83)	−0.05 (0.61)

^*Suggestive (0.01 < p-value < 0.05).

Tb.vBMD: volumetric BMD, Tb.tBMD: transpose BMD, Tb.pBMD: plate trabecula BMD, Tb.NA: trabecular network, Tb.PW: plate width, Tb.Th: trabecular thickness, Tb.Sp: trabecular spacing, Cb.Poro: cortical porosity, Cb.Th: cortical thickness.

F = fluoride

4. DISCUSSION

The aim of this study was to investigate associations of F intakes with bone microstructural measures during the bone development phase throughout childhood, adolescence, and early adulthood. Previous analyses from the IBDS longitudinal study examined associations between F intakes and dual-energy X-ray absorptiometry (DXA)-based bone density and content outcomes at ages 11 [10] and 15 [48], and between F intakes and Peripheral Quantitative Computed Tomography (pQCT)-based cortical and trabecular measures at ages 11 [58] and 17 [59]. Although reported results at ages 11 and 17 involved analysis of pQCT-based cortical and trabecular bone characteristics, there was no assessment at the level of cortical or trabecular bone microstructure.

Results of a meta-analysis have shown that only about 60% of bone’s mechanical competence is explained by variations in BMD [60]. Histologic studies and mechanical modelling studies have demonstrated convincingly that Cb and Tb microstructural quality are additional critical determinants of bone-strength and fracture-risk [61, 23–33, 62].

Volumetric bone imaging modalities, including magnetic resonance imaging (MRI) [63, 64, 60, 65] and high-resolution peripheral quantitative computed tomography (HR-pQCT) [66–68], have been investigated for in vivo assessment of bone microstructure. Use of clinical whole-body MDCT scanners for in vivo bone microstructural analysis at a peripheral site is relatively recent [49]. The high spatial resolution achievable by emerging MDCT scanners at peripheral sites (e.g., distal tibia, wrist, or knee) allows segmentation and quantitative characterization of Cb and Tb microstructure at a relatively low radiation dose, and previous studies have demonstrated strong associations between several CT-based Tb microstructural measures and their micro-CT derived values and experimental bone-strength [49]. Modern MDCT scanners are capable of imaging 10 cm of scan-length at an UHR mode in just 6 seconds, as compared to ~2.9 minutes for 0.9 cm of scan-length using HR-pQCT [69], thus reducing substantially motion blur artifacts. Additionally, emerging MDCT technology offers major radiation dose reduction, while simultaneously increasing spatial resolution and signal-to-noise ratio (SNR) [70]. Wide availability of whole MDCT imaging facilities in clinical environments and the low associated radiation dose will add strengths to MDCT-based multi-center musculoskeletal studies.

All Tb measures examined in this study, except the trabecular separation measure Tb.Sp, are positively associated with bone strength; Tb.Sp is negatively associated with bone strength [49]. The Cb thickness measure Cb.Th has a positive association, while the Cb porosity measure Cb.Poro has a negative association with fracture risk. Results from Table 3 suggest that young adult males have stronger Tb as compared to age-similar females. Observed relationships of Cb measures with sex are more interesting. As observed in Table 3, young adult males have thicker, but more porous, Cb as compared to age-similar females. It will be interesting to conduct future mechanical modelling and finite element analysis to examine implications of these mixed relationships of cortical bone features by sex on bone strength, skeletal geometry, and weight.

Effects of cumulative and period-specific F intakes on bone characteristics were analyzed. For females, no statistically significant associations were detected between any period-specific or cumulative F intake and Tb or Cb measures of the tibia at the p < 0.01 level. Statistically-suggestive associations (p < 0.05) were found only between period-specific F intake from 14-19 years and trabecular mean plate width (Tb.PW) and trabecular thickness (Tb.Th) of the tibia. Those associations became somewhat weaker, but still statistically suggestive, for Tb.Th in fully adjusted analyses with height, weight, PHV, calcium and protein intake, and HEI-2010 and PAQ-A scores as covariates.

For males, there were no statistically significant associations at the 0.01 level or statistically suggestive associations at the 0.05 level between F intakes and MDCT-derived bone outcomes. Previous analyses from the IBDS longitudinal study of boys and girls at ages 11 [10], 15 [48], and 17 [59] reported weak or no associations between F intakes and bone density and mineral content outcomes. At ages 11 and 17, F intakes showed weak associations with pQCT-based cortical bone characteristics for both boys and girls. The lack of any statistically significant unadjusted or adjusted associations between period-specific or cumulative F intakes from birth to age 19 and the MDCT microstructural measures is consistent with the previously reported outcomes at the level of bone mineral density and content. A few studies have reported significant associations of F intake with bone development. For example, Grobler el al. [71] investigated BMD of children from two areas with similar nutritional, dietary habits and similar ethnic and socioeconomic status, but different drinking water F levels and found that the BMD in the left radius increased with age in the high F (3.00 mg/L F) area, but decreased significantly (p < 0.05) with age in the low F (0.19 mg/L F) study area. They observed differences in a comparison between the two different F areas for the BMD of boys and girls at a specific age group. These differences were not significant in the 10-11 year age group, while in the age group 12-13, girls from the high F area had significantly higher BMD than boys from the low F area. In the age group 14-15, the BMD of boys and girls in the high F area was significantly higher than for the boys and girls in the low F group. In a study involving healthy young women from two regions with different F levels (0.1 mg/L F versus 1.0 mg/L F), Arnold et al. [72] women from the high F region had significantly higher mean BMD at total anterior-posterior lumbar spine and estimated volumetric BMD at L3, with no difference in mean BMD of total body or proximal femur (PF). In a Swedish adolescent bone study, Lötborn et al. [73] observed that the mean total bone mineral content and bone mineral area of the boys and girls from Uppsala (1.1 mg/L F) were significantly higher than those of the boys and girls from Trollháttan (< 50.1 mg/L F). In a population-based study of peri-menopausal women, Kröger et al. [74] found that women exposed to a high F concentration (1.0-1.2 mg/L F) in their drinking water had a slightly higher BMD of the lumbar spine and femoral neck compared with women who lived in a low-F (0.3 mg/L F) region. The results observed in the current study are novel because they show that community water F levels typical to the United States Midwest region have no effects on the developing skeleton through early adulthood at age 19 at the levels of Cb and Tb bone microstructure.

Most of the previously-reported studies investigating the effects of F intake on bone outcomes were ecological, i.e., F intakes were not estimated at the individual level, but rather bone outcomes were compared among participant groups living in communities with different levels of water F. In this study, F intake levels for individual participants were assessed using questionnaire data. Almost 70% of participants in this study had access to optimally fluoridated water. High standard deviations (from 0.29 to 0.49 mg F for male and 0.25 to 0.44 for female intakes) indicate considerable variation in daily F intakes which were not assessed in previous ecological studies.

F plays an important role in the mineralization of body tissues. Community water fluoridation is judged as a cost-effective method to deliver F to people of all ages, education levels, and income levels in a community (CDC, 2016), and many health organizations and government agencies, including the Centers for Disease Control and Prevention (CDC), recommend fluoridation of community water supplies for caries prevention (CDC, 2014). Previously-analyzed results from the IBDS [10, 48, 59] have supported that the levels of F that are currently present in community waters are not associated with any harm to bone density and content of the developing skeleton of children and adolescents. This study further confirms that current community water F levels pose no harm to the developing skeleton through early adulthood at age 19 at the level of cortical and trabecular bone microstructural details, and therefore, efforts should continue to be focused on preserving implementation of community water fluoridation and expanding its use.

There are several limitations of this study. All IFS/IBDS participants were recruited from the state of Iowa, and almost all were non-Hispanic whites from relatively high socioeconomic families. Therefore, the results are limited in regard to the external validity to other races and ethnicities [75] and a wide socioeconomic status range.

Because of some missing data for F intake at different time points, and requirements for adequate AUC F calculation, each period-specific F intake estimate was available for a somewhat different sub-sample of our total study sample (we used all available data for each analysis of associations between MDCT outcomes and period-specific F intake). F intakes for the study participants were based on parent and adolescent reports of intakes of foods and beverages and ingested F-containing toothpaste, which is an indirect method of quantifying intake and possesses several limitations in terms of its reliability and validity. Also, F intakes generally were modest to moderate and not extreme, with mean intakes generally consistent with the recommended levels. Thus, generalizing study findings to other geographic areas—especially areas with higher water F levels—should be done with caution.

On the other hand, several strengths of this study should be recognized. Study participants were followed longitudinally from birth. Detailed, age-specific individual data were collected for each participant. Several other relevant factors influencing bone development were included in the study, including height, weight, years since PHV, calcium and protein intake, and HEI-2010 and PAQ-A scores. High resolution MDCT images were acquired for participants at age 19 years, and associations of F intake during several time periods and cumulatively with detailed bone outcomes at the cortical and trabecular bone microstructural level were examined.

In summary, the findings of this study show that the effects of lifelong or period-specific F intake from combined sources for adolescents typical to the United States Midwest region are not strongly associated with MDCT-based cortical and trabecular bone microstructure at age 19 years. These findings are generally consistent with previously-reported results of IBDS analyses, and these findings further confirm that effects of lifelong or period-specific F intake on developing skeletons of children, adolescents, and young adults to age 19 at the cortical and trabecular bone microstructural levels is absent or weak. Thus, the study findings provide support to the assertion that F intakes, within these ranges, are not associated with adverse consequences on bone outcome measures at age 19. Consequently, efforts should be focused toward preserving and expanding community water fluoridation, as F intakes within ranges presented in this paper pose no adverse effects on bone health.

Highlights.

No evidence of a significant effect of lifetime fluoride intake on bone microstructure at early adulthood

Iowa Bone Development Study cohort at age 19 years

High resolution multi-row detector CT imaging at the distal tibia

Quantitative measures of cortical and trabecular bone microstructure

Individual trabecular plate-rod and longitudinal-transverse characterization of the distal tibia