|Home | About | Journals | Submit | Contact Us | Français|
The monitoring time interval (MTI) is the expected time in years necessary to identify a change between two measures that exceeds the measurement error. Our purpose was to determine MTI values for dual-energy X-ray absorptiometry (DXA) scans in normal healthy children according to age, sex, and skeletal site.
2014 children were enrolled in the Bone Mineral Density in Childhood Study and had DXA scans of the lumbar spine, total hip, non-dominant forearm, and whole body. Measurements were obtained annually for seven visits from 2002 to 2010. Annualized rates of change were calculated by age and sex for all bone regions. A subgroup of 155 children ages 6 to 16 years (85 boys) had duplicate scans for calculation of scan precision. The bone mineral density (BMD) regions of interest included spine, total body less head (TBLH), total hip, femoral neck, and 1/3 radius. Bone mineral content (BMC) was also evaluated for spine and TBLH. Percent coefficient of variation (%CV) and MTI were calculated for each measure as a function of age and sex.
MTI values were substantially less than 1 year for TBLH and spine BMD and BMC for boys ≤ 17 years and girls ≤ 15 years. Hip and 1/3 radius MTIs were generally 1 year in the same group. MTI values as low as 3 months were found during the peak growth years. However, MTI values in late adolescence for all regions were substantially longer and became nonsensical as each region neared the age for peak bone density.
All 4 DXA measurement sites had reasonable (< 1 year) MTI values for boys ≤ 17 years and girls ≤ 15 years. MTI was neither useful nor stable in late adolescence and young adulthood. Alternative criteria to determine scan intervals must be used in this age range.
Dual-energy X-ray absorptiometry (DXA) is the most widely used technique for measuring bone mineral content (BMC) and areal bone mineral density (BMD) in children and adolescents (1-2). Although there are limitations to DXA-derived measures including inaccuracies due to bone size and fat inhomogeneity overlaying bone, DXA is low in cost, accessible, easy to use, provides precise quantification, and is the most accurate measure of BMC and BMD (2-3) clinically available. There are multi-ethnic reference datasets available for both boys and girls from ages as young as 3 years old (4-6). The International Society for Clinical Densitometry (ISCD) recommends that DXA measurements be obtained in children with increased risk of fracture, including patients with primary bone diseases or potential secondary bone diseases (e.g., due to chronic inflammatory diseases; endocrine disturbances; history of childhood cancer; and immobilization) (7). In these children, the ISCD recommends measuring the lumbar spine and total body less head (TBLH) BMD and BMC. The ISCD also recommends a minimum monitoring time interval (MTI) of six months for repeating a BMD measurement, not specific to age, sex, or skeletal site. Owing to lack of data, development of the ISCD recommendation was largely based on expert opinion (7). However, factors like measurement precision and rate of change in BMD must be considered for accurate interpretation at different measurement sites and with different populations.
The precision error of DXA measurements is important to differentiate between measurement error and genuine bone changes (8). BMD precision errors differ in children and adults due to children’s smaller bone size and lower BMD (9). Minimizing pediatric precision errors is critical for detecting low-density bone changes.
There has been ongoing development of pediatric scan and analysis algorithms for DXA images that may impact precision and accuracy (10). For example, the APEX software (Hologic, Inc., Bedford, MA) has special decision-making features called “Automatic Low Density analysis” to better identify and correct bone maps that previously required operator intervention, thus yielding more accurate and consistent results in low BMD and pediatric subjects. Fan et al. also found that the APEX software improved the precision of adults scans (10).
Precision error estimates have been used to clinically determine how much change needs to occur to have confidence that the change is genuine and not just due to measurement error. The least significant change (LSC) is defined as the change that can be recognized with 95% confidence with a given scan precision. The LSC is mathematically calculated as 2.77 times the precision error (11). Comparing the change in serial DXA bone measures to the LSC is recommended by the ISCD for reporting clinically significant change (8). The LSC concept was developed for monitoring bone loss in adults after reaching peak BMD where a clinically significant bone loss is based upon statistically detectable loss. In healthy children, the BMD and BMC normally increase every year until peak bone mass is attained in late adolescence or young adulthood. The positive changes in BMC and BMD during growth are physiologic and not considered of clinical concern. The median BMD differences between two adjacent ages can vary from 1% (total hip BMD of girls from 15 to 16 years old) to 12% (spine BMD of girls from 11 to 12 years old)(4). The LSC has been further used in adults to derive the MTI: the time needed to pass before the difference between two measures is expected to surpass the LSC. The MTI is simply the ratio of the LSC to the median annual change in BMD for a specific age and sex group and specific measurement site. Half of the population will experience a change in BMD equal to the LSC for measures taken in the time interval defined by the MTI (12). The MTI may be particularly relevant in pediatric clinical practices to help in planning longitudinal follow-up intervals and in selecting the BMD measurement sites most relevant for serial monitoring for children. The objective of this study was to determine the MTI for five skeletal sites in normal healthy children according to age and sex.
The Bone Mineral Density in Childhood Study (BMDCS) is a multi-center, longitudinal study of bone accrual in 2014 healthy children and adolescents. Detailed information about the study participants, inclusion/exclusion criteria, and study procedures have been previously published (4). In brief, a total of 1554 healthy participants were enrolled in the BMDCS in 2002 and 2003, including an equal number of boys and girls from 6 to 16 years of age and a mix of ethnic groups. In 2006, 460 additional participants were recruited aged 5 and 19 years old to increase the sample sizes at the lowest and highest ages. Serial measures were acquired over 7 annual visits. Most participants had more than 3 measures with many having 7 annual measures. For this study, we excluded participant visits that did not occur within an 11 to 13 month follow-up window. Individual longitudinal changes were used to derive annual changes in BMC and BMD by age and sex.
A subset of 155 children (85 boys and 70 girls) was enrolled in a precision sub-study, comprising about 30 children from each of the five participating clinical centers. Subjects were recruited from three age groups: 6-9 years (preteens), 10-13 years (early teens), and 14-16 years (late-teens).
Written informed consent was obtained from the study participants ≥18 years. For participants <18 years of age, consent was obtained from the parent or guardian and assent was obtained from participants. The protocol was approved by the Institutional Review Boards of each Clinical Center.
DXA scans of the posteroanterior lumbar spine (L1-L4), the left total hip, the non-dominant forearm, and total body were acquired using either a Hologic 4500A or 4500W system (Hologic Inc., Bedford, MA). In addition to BMD and BMC for the total spine and TBLH mentioned in the ISCD pediatric guidelines (7), BMC and BMD for the total hip, femoral neck, and the 1/3 radius BMD were examined for their potential monitoring utility. The software acquisition versions were either version 8.6 or 12.3.The fast-scan mode was used for spine and total hip scans.
A duplicate set of scans using the same scanner was acquired for the precision sub-study subjects at either their baseline or one-year follow-up visit. Subjects got off the scanning table and were repositioned before obtaining the second set of scans.
All BMDCS DXA scans were analyzed centrally. Duplicate scans were analyzed by one trained research assistant using Hologic software APEX (version 3.0).
For each subject and each measurement site, the annual percent changes in BMD and BMC were calculated using two consecutive visits. Median annual rates were calculated for each skeletal site in 1-year age intervals for each sex. From duplicate scans, precision error was quantified as the root mean square error (RSME) and percent coefficient of variation (%CV) (12) expressed as a percentage of the sample mean. Bootstrapping techniques were used to estimate the 95% confidence interval for each average precision error by age group (13). The LSC for the 95% confidence level was defined as 2.77 times the precision error (14). Precision error and LSC were calculated separately for three age groups (6-9 y, 10-13 y, 14-16 y) and for all ages combined.
The MTI in years was defined as the LSC (derived from the %CV) divided by the absolute value of the age- and sex-specific median annual percent rate of change in BMD: MTI = LSC / rate of annual BMD change (12). The percent error versions of the LSC and annual change were used to define MTI instead of the absolute error and annual change values since percent change is generally considered the clinically-significant indicator of bone disorders than absolute change. However, both definitions result in MTI values in years. All statistical calculations were performed using SAS 9.2 (SAS Institute, Cary, NC).
Table 1 summarizes the precision error findings by age group and regions of interest. There were only unique differences in precision error expressed as a RSME by age for spine and TBLH BMC. When precision error was expressed as a %CV, all bone measures, except for femoral neck and 1/3 radius BMD, had unique precision error values. Pooled across all ages, BMD precision values were less than 1.5% for all sites except the forearm. The oldest group (14-16 years) had the best precision (smallest errors) compared to the youngest (6-9 years). The forearm and femoral neck BMD had the poorest precisions (largest errors) for all ages.
Median annual percent changes in BMD varied dramatically by age and skeletal site. (See Table 2.) The annual percent changes ranged from 0% to 10.3% for boys and from 0% to 10.4% for girls, depending on the skeletal site and age. The BMC changes for spine and TBLH were even more dramatic than BMD and ranged from 0.2% to 18.6% for boys and from 0% to 18% for girls. BMC accrual was over 10% annually for up to 5 consecutive years in both the spine and whole body.
Using our precision estimates and median annual percent changes in BMD, we calculated MTIs for the five skeletal sites for boys and girls (Table 3 and Figure 1). For ages above 16 we used the precision measured on the 14-16 year olds. MTI values less than one year were the norm for ages between 6 and 15 years for both boys and girls. In the older ages, however, where annual rates of change were either 0% or close to it, the MTI values became much larger and in some cases nonsensical. In general, the MTI became unstable and nonsensical when the annual change approached zero. For example, the MTI for TBLH BMC for 20 year old girls was 926 years. This large value is because there was no net change in TBLH BMC over a year. The values considered too large to be meaningful are highlighted in grey in Table 3. The spine and TBLH BMC measures consistently produced the shortest MTI values. MTI values for girls and boy are, for the most part, indistinguishable until pubertal ages. MTI for girls lengthens before boys as their annual rate of change begins to decelerate sooner.
In this study, we have shown that the MTI varies greatly by skeletal region, sex, and age. MTI was a concept primarily used in adults where peak bone mass has already been achieved and where the annual rates of loss were considered to be similar over many years. In general, the variation in these pediatric MTI values was mostly driven by the large, age-specific differences in the annualized changes in BMD and BMC. The range of annualized rates of change varied from 0 to 18% per year while precision error never varied more than about a factor of 3 between age groups and regions. When BMD and BMC accretion was at its highest in the early teen years, the MTI values were as short as 3 months. As bone accretion rate decreased in the late teen years, the MTI values increased to values substantially higher than 1 year, sometimes over 10 years, due to the rate of change approaching zero. Hence, in healthy children and younger adolescents (less than15 to 16 years of age) statistically and clinically meaningful changes in BMC and BMD are expected at all skeletal sites. The information presented here may provide valuable information in clinical situations to determine how long to wait before a follow-up DXA scan is acquired. MTI is not useful for older adolescents and young adults around the age of peak bone mass when MTI values exceed 1 year and are thus unreliable due to low annual rate of change.
This is one of the largest precision studies to date in children. We found precision errors varied in a different way than the annualized rates of change. For example, the worst precision error in terms of %CV, albeit still reasonable, was found in the youngest ages while the oldest children had the best precision. The absolute precision error in RSME was similar in all age groups except for the spine and TBLH BMC values. Spine and TBLH had worse precision for the youngest age group for BMC and BMD. We attribute the spine and TBLH age-based differences in precision to known reproducibility challenges to DXA systems. For example, DXA precision is generally worse in low density, small bones due to contrast and edge detection issues. Both of these situations are found in the young children. For the same reasons, the older post-pubertal adolescents had precision error values slightly better than previously reported for adults. The older study participants, with bones close to adult size and at lifetime peak bone density levels, had precision error values of 0.6, 0.6, and 1.2 %CV at the spine, total hip, and femoral neck. For comparison, the precision error in 90 postmenopausal women (full adult size bones but bone density lower than peak) using similar Hologic scanners and analysis software was 1.0, 1.1, , and 2.3 %CV for the same sites (10).
For the other sites besides spine and TBLH BMC, the precision was not unique by age group for the absolute RMSE error, only for %CV. In these BMD measures (spine, TBLH, femoral neck, total hip, and 1/3rd radius forearm), the differences in the age-specific mean values caused the %CVs to be unique even though the RMSE values were quite similar. Thus, for these measures, the absolute BMD errors appear not to be driven by issues related to bone size or density, and the overall absolute precision values may be used for LSC calculations (not shown.)
This study confirms the conclusion of Leonard et al. (9) that DXA scanning is more challenging in younger children than older children. In that study, 32 children had DXA scans of the spine BMD, total hip BMD, and whole body BMD and BMC with a Hologic Discovery A (version 12.3.3). Leonard found worse %CV precision errors for younger group (less than 10 years) versus older group (10-18 years). The %CVs for spine, total hip and whole body BMD were 1.2, 1.6 and 1.0 respectively in the younger children compared to 1.1, 1.2, and 1.2 in our study for the youngest group. For the older children, Leonard found 0.7, 1.0, and 0.9 as compared to our study’s 0.6, 0.7, and 0.6. Thus the precision error values for the Leonard study were similar to our study. However, because Leonard used fewer subjects, significant differences were only seen by age for the spine. In our study, significant differences by age were seen for spine, total body, and total hip. Although the Leonard study speculated about the ability to monitor changes in BMD after 3, 6, and 12 months, our study was able to calculate the MTI based on longitudinal changes observed in the whole BMDCS population.
In a study by Margulies et al. (15) whole body DXA precision error was measured in 49 children aged 5 to 17 years using a GE Lunar Prodigy (GE Lunar, Madison, WI) with software version 6.6 and 6.7. The whole body BMD %CV was found to be 0.73. This compares to our whole population TBLH BMD %CV of 0.95 on the Hologic systems. Even though the technology for X-ray generation, bone identification, are different between GE and Hologic (16), the precision was similar for the two studies.
Given these findings, what are the implications for clinical DXA scans in children? The ISCD Official Pediatric Positions on DXA Interpretation and Reporting (17) stated that “The posterior-anterior (PA) spine and total body less head (TBLH) are the most accurate and reproducible skeletal site.” Here we demonstrate that the spine and TBLH were the most reproducible skeletal sites with the largest annual changes for both BMC and areal BMD measurements. Thus, the MTIs of spine and TBLH were also consistently the shortest. The spine and TBLH had MTIs of 6 months or less for BMD and BMC for both boys and girls less than 14 years old. Femoral neck, total hip, and 1/3rd distal radius have not been previously recommended in young children due to either significant morphological changes taking place during growth (femur) and thus the lack of consistent regions of interest over time (17), or the lack of pediatric reference data (radius). However, pediatric reference data for 1/3 radius BMD and total femur have recently been published (4). Furthermore, in this study, although the MTIs of these sites were consistently longer than that of the spine and total body measures, the MTIs were still reasonable with values around 1 year. Thus, our results support the ISCD position that the spine and TBLH are preferred sites of measure. Nevertheless, reasonable monitoring times and precision errors can still be found with the femoral neck, total femur, and 1/3rd radius.
Which skeletal sites should be used for monitoring bone change in children? Bonnick (8) outlined in the following four general rules to follow regarding monitoring that were derived primarily with adults in mind: 1. Determine the skeletal sites or bone type considered most affected by the disease or its treatment. 2. Determine the site with the greatest expected change in BMD. 3. Determine the site with the best precision, and 4. Avoid peripheral sites. In adults, the sites that fit these rules in most situations have been the spine and total hip. We studied only healthy children and therefore cannot directly address rule 1. However, the spine and TBLH had both the largest annual changes and best precisions and fit the criterion of rules 2 and 3. Rule 4 may not be applicable to children. When studying pediatric metabolic bone disorders, such as rickets, wrist and knee radiographs are typically acquired since these are the fastest growing areas of the skeleton and, therefore, where a deficiency in mineralization is first likely to be depicted. Although this study does not address specific bone disorders of children, there may be further utility to peripheral measures beyond what this study can demonstrate. It would be reasonable to expect that peripheral skeletal sites would provide unique and valuable information for monitoring bone disease in children. In adults, whole body BMD and BMC are typically not good choices for monitoring because of their low annual change rate. But because of skeletal growth, TBLH in children has very large annual changes.
Our study had the following limitations. First, all of our subjects were scanned on systems made by one DXA manufacturer. It is not possible to generalize these MTI values to scans acquired on systems from other manufacturers without further validation. The primary driver of the MTI was the large differences in the age-related changes. It is reasonable to expect that our results would generalize across different DXA manufacturers. Second, these data were acquired on healthy children. The precision would most likely be worse on children with chronic diseases with lower bone mass and density. More importantly, the annual rates of change were also for healthy children, and these are likely to be affected by disease and/or medications. Third, our study utilized short-term precision estimates from scans acquired on the same day. The long-term precision, where time intervals between scans are several days to months, is affected by both patient positioning, soft-tissue changes around the bone, and machine stability (18). In a study of 40 postmenopausal women over a 7-year period, Patel (19) found the long-term precision error values to be similar to previously reported short-term precision error values for Hologic systems, 1.12%, 2.21%, 1.32% for spine, femoral neck, and total hip respectively. These values are similar to our findings and support the use of short-term precision error values when estimating MTI. Lastly, DXA is used on children of all ages including newborns and infants (20). Our study did not include any participants less than 6 years old. It is likely that the precision errors and MTI values for younger children are less favorable.
We conclude that DXA precision errors and annual rates of change for BMD and BMC in children vary with region of interest, age, and sex. The resulting MTI measures are similar for boys and girls at or below pubertal ages, and provide useful guidelines for scanning intervals. The definition of MTI is not applicable for young adults at ages around the occurrence of peak bone mass. For older adolescents and young adults, some other clinical criteria must be used to determine DXA scanning interval.
Authors’ roles: Study design: JS, YL, and BF. Study conduct: VG, HK, TH, JL, SO, MF, and KW. Data collection: MF, JS, and BF. Data analysis: JS, BF, LY, and BZ. Data interpretation: JS, LW, BF, YL, HK, and BZ. Drafting manuscript: JS, LW, YL, and BF. Revising manuscript content: BZ, HK, VG, JL, TH, YL, and KW. Approving final version of manuscript: JS, LW, BF, VG, HK, JL, TH, MF, BZ, SO, and KW. JS takes responsibility for the integrity of the data analysis. This study was funded by the National Institute of Child Health and Human Development Bone Mineral Density in Childhood Study, NICHD N01-HD-13328, USPHS Grant #UL1 RR026314 from the National Center for Research Resources (NIH), and CTSA grant number UL1-RR-024134. The authors wish to thank the investigators and staff at Cincinnati Children’s Hospital Medical Center, Creighton University, Children’s Hospital of Philadelphia, St. Luke’s-Roosevelt Hospital, and Children’s Hospital of Los Angeles. We acknowledge the editorial support of Erin Reese.
Funding: This study was funded by the National Institute of Child Health and Human Development Bone Mineral Density in Childhood Study, NICHD N01-HD-13328, USPHS Grant #UL1 RR026314 from the National Center for Research Resources (NIH), and CTSA grant number UL1-RR-024134.
DISCLOSURES/CONFLICT OF INTEREST All authors state that they have no conflicts of interest.
AVAILABILITY/DISCLOSURE OF MATERIALS AND METHODS AND AUTHOR ACCESS TO DATA