This study was approved by the West Virginia School of Osteopathic Medicine Institutional Review Board. Participants were recruited between June 2015 and June 2016 from the Robert C. Byrd Clinic, a rural, primary care facility in Lewisburg, West Virginia. Patients aged 18 years or older scheduled for DXA were eligible for inclusion in the study. Patients wearing stockings that could not be removed were excluded from the study because stockings interfere with US. Research personnel obtained a weekly schedule of appointments for DXA scans and recruited individuals into the study at the beginning of their appointment. Data were not collected regarding patients who did not agree to participate. The clinic staff was aware of the study, but referring physicians were likely unaware of the project. Hence, the participants scheduled for DXA most likely reflected individuals who either met screening guidelines or whose BMD was being analyzed for some other reason (eg, recent fracture, medications).
Informed written consent was obtained from all participants. Participants could agree to calcaneus US only, or calcaneus US and fingerstick blood testing in addition to the DXA scan. At the time of screening, participants also filled out a questionnaire. Information that was collected included some parameters associated with the Fracture Risk Assessment Tool (FRAX) calculation (age, sex, weight, height, previous fracture, parental hip fracture, current smoker, glucorticoid use, rheumatoid arthritis, 3 or more alcoholic drinks per day). The collected information also included the first day of last menstrual period for premenopausal women, calcium or vitamin D supplement use, and use of drugs that could affect BMD (eg, steroids, diuretics, anticoagulants, thyroid hormone, proton pump inhibitors).
An Achilles bone ultrasonometer (GE Healthcare) was used to assess the density of the calcaneus (
Figure 1). The shoe and sock were removed from the right foot and the heel was sprayed with rubbing alcohol. The foot was then placed in the cradle of the US machine, and after recording the BMD T score and Z score readings, the foot was removed and dried, and the sock and shoe were donned. This procedure was then repeated with the left foot.
Blood was collected for vitamin D analysis using the standard sterile protocol for daily assessment of circulating concentrations of glucose. The blood was placed on a ZRT blood spot card and allowed to dry (
Figure 2). Blood spot cards were stored at −80°C until analyzed for 25-hydroxyvitamin (25[OH]D) D
2 and D
3 by ZRT Laboratories using liquid chromatography–tandem mass spectrometry.
11 The intra-assay coefficient of variation for 25(OH)D D
2 and D
3 was 8.1% and 9.2%, respectively; the interassay coefficient of variation was 13% and 12%, respectively. The limit of sensitivity for 25(OH)D D
2 and D
3 was 0.5 and 1.9 ng/mL, respectively.
After US data were recorded and blood collected by research personnel, DXA was performed by trained technologists Lunar Prodigy Advance (GE Healthcare). The computed T and Z scores for the spine and left and right femurs were recorded by research personnel.
Data were analyzed using SAS software (SAS Institute). Criteria were established to define a criterion standard of poor vs good BMD: a DXA BMD T score of the spine less than −1.0 was defined as poor, and a DXA BMD T score greater than or equal to −1.0 was defined as good. This cut point reflects the current diagnostic guideline from the World Health Organization in which DXA BMD T scores of −1.0 and above are categorized as normal, and DXA BMD T scores between −1.0 and −2.5, as osteopenia.
12 Bone mineral density T scores higher (more negative) than −2.5 correspond to osteoporosis.
12
Descriptive statistics were calculated. Pearson correlation coefficients were calculated within and between the DXA and US measurements, between the DXA and vitamin D levels, and between the US measurements and vitamin D levels. The Wilcoxon signed rank test was used to compare the right vs left sides within the DXA and US measurements. A logistic regression model was fit using the new criterion standard as the outcome and the US measurements as explanatory variables. Area under the curve analysis using receiver operator characteristic curves was used to assess the predictive performance of the US readings on bone health as determined by DXA scan. Optimal cut points in the US readings, maximizing sensitivity and specificity, were determined for assigning poor vs good bone quality using our criterion standard. All tests were 2-sided, with significance level α=.05 and
P<.05 considered statistically significant. The receiver operator characteristic curves were compared using the nonparametric approach of DeLong et al.
13