Rate of tarsal and metatarsal bone mineral density change in adults with diabetes mellitus and peripheral neuropathy: a longitudinal study

This is a prospective, longitudinal clinical trial of secondary outcomes included in a larger clinical trial (ClinicalTrials.gov: NCT02616263). Participants with type 2 DM and PN were randomized in one of two intervention cohorts, which received targeted strengthening and stretching interventions on either the foot or shoulder. For the purposes of this analysis, all participants were assessed as one cohort. A sensitivity analysis was completed to ensure there were no significant differences in BMD between the intervention groups prior to collapsing the groups. The detailed findings of the sensitivity analysis are reported in the results section. Data were collected over the course of three to four years at four-time points: baseline (T1), 0.5 years (T2), 1.5 years (T3), and at the conclusion of the study (3–4 years after baseline, T4) [28, 29]. Time was treated as a continuous variable to account for variation in the actual date of reassessment for each participant.

Participants were recruited from the Recruitment Enhancement Core of the Institute of Clinical and Translational Sciences at the Washington University School of Medicine in St. Louis, email blasts to a senior center, along with several databases (e.g., Research Participant Registry through Volunteers for Health, and patient databases of the Applied Biomechanics and Human Biodynamics Laboratories). Inclusion criteria were type 2 DM diagnosed by the participants’ healthcare providers and PN assessed by the study team. In order to capture the multi-system effects of PN, presence of PN was defined as any of the following: inability to feel a 5.07 Semmes Weinstein Monofilament in at least one location on the plantar aspect of the foot, a vibration perception threshold greater than 25 V on biothesiometry, or a score of greater than or equal to 2 on the Michigan Neuropathy Screening Instrument (1 point is given for each foot abnormality including deformities, dry skin, calluses, infections, fissures, and ulcers) [30]. For the purposes of analysis, PN severity was represented by each participant’s Michigan Neuropathy Screening Instrument score. Exclusion criteria were PN of non-DM etiology, ankle brachial index of < 0.9 or > 1.3 indicating severe arterial disease [11, 31], amputation of more than one toe, acute shoulder injury, pregnancy, > 180 kg—the weight limit capacity of the magnetic resonance imaging (part of the parent study) scanner, active neuropathic ulceration, inability to walk without a device or complete required testing, presence of non-magnetic resonance imaging compatible metal implants, age greater than 75 years, or completion of less than two computed tomography (CT) scans over the course of the study. Race was self-reported on intake. Each participant completed the Foot and Ankle Ability Measure [32] for descriptive purposes. The Foot and Ankle Ability Measure is a self-reported measure of foot and ankle specific symptoms and function, with higher scores indicating less disability/symptoms (maximum score of 100). No participants experienced non-healing plantar ulceration or amputation during the course of the study. While exclusion criteria included history of non-healing ulceration, participants were not withdrawn from the study if they acquired wounds during the study. Nine participants self-reported occurrence of foot wounds (from non-traumatic causes as well as known trauma – cuts, scratches, etc.) during the course of the study, which were able to heal. Wounds were defined as any break in skin continuity of any depth. All wounds reported by participants were non-tunneling. Wound management for all instances of foot wounds required neither immobilization nor offloading. The participant demographics and characteristics are in Table 1.

A spiral CT scanner (Siemens Biograph 40 TruePoint Tomograph) was used to scan each participant’s lower extremity capturing the entire ankle mortise of the selected side as well as all tarsals and metatarsals. Scans were performed at all four time points. Participants underwent CT scans of the foot in 30 degrees of plantarflexion and metatarsophalangeal joints in resting position (Fig. 1: Panel A). A CT calibration phantom (Mindways, Austin, TX, USA; Calibration value: L11G 11F1 1171 611L) containing multiple discs of calcium hydroxyapatite of different densities was placed in front of the foot for all scans. The CT scanning parameters were 0.5 s rotation time, 64 × 0.6 mm collimation, 220 mAs, 120 kVp, a pitch of 1, a 512 × 512 matrix, and a voxel size of 0.6 × 0.6 × 0.6 mm³.

BMD of each of the 12 tarsals and metatarsals was obtained using a previously published protocol [17, 27]. Each tarsal and metatarsal was individually segmented, and Hounsfield Unit (HU) was converted to BMD (mg/cm³) using the calibration phantom (Fig. 1: Panel B).

The physical activity level of each participant was obtained at baseline (T1) using an ActiGraph wGT3X-BT activity monitor (ActiGraph, Pensacola, FL). Participants were asked to wear the ActiGraph full-time for 5 days, of which the middle 3 days were used for analysis (to ensure only days with a full 24-h wear time were included in analysis). The ActiGraph was placed on the participant’s wrist of the non-dominant hand using a band that prevented removal during the 5-day collection period. Daily step count and sedentary time (minutes) were calculated using ActiLife software and were included in statistical analysis.

Participants’ medication use and diabetes management strategies were obtained via self-report at baseline (T1). Medications were classified as statins or non-statins for use in statistical analysis because statin use is known to increase BMD [33]. No participants reported taking bisphosphonate medications, a common class of medications used to treat osteoporosis [34]. The diabetes management strategies were collected with four choices (i.e., diet, exercise, oral medications, and insulin) and allowed multiple selections.

All BMD measures were checked for normality assumptions via Shapiro–Wilk test and visual inspection of QQ plots and histograms. All data met normality assumptions. Alpha level of significance was set at 0.05 a priori for all analyses.

A random coefficients model (linear mixed model) was used to estimate the annual rate of change (slope) of BMD (mg/cm³ per year) in the calcaneus and mean of all tarsal and metatarsal bones. This model allowed both intercept and slope to vary randomly between subjects and fitted a separate regression line for each individual subject. A separate random coefficients model analysis was performed to predict annual rate of BMD change for individual tarsals and metatarsals.

To determine if there are any significant differences in rate of BMD change between the bones of the foot, an additional random coefficients model was built to test for significance of bone name by time interaction with significant interaction meaning a different rate of BMD change between bones. In the current body of literature, the calcaneus is the most frequently used bone to assess foot BMD [19, 35], so the calcaneus was selected as the referent value for this analysis. The estimated rate of BMD change of each individual tarsal and metatarsal are reported.

Personal factor variables that were hypothesized to affect mean BMD (mg/cm³) and calcaneal BMD (mg/cm³) included age, sex, step count, sedentary time, and Michigan Neuropathy Screening Instrument score. A covariance pattern model (linear mixed model) was fit to assess the association between the personal factors and BMD variables and accounted for the correlation of repeated measures of BMD from the same subject over time. Relationships between personal factor variables and rate of BMD change (mg/cm³ per year) were tested using Pearson correlation coefficient.

The results of a linear mixed model analysis using each tarsal and metatarsal bone as a single input variable and calcaneal BMD as a reference variable indicate that the rate of BMD change (mg/cm³ per year) was not significantly different between the calcaneus and any other bone included in the analysis (p = 0.9966). Descriptive statistics for rate of BMD change (mg/cm³ per year) for individual bones are reported in Table 3.

The rate of mean BMD change (mg/cm³ per year) and rate of calcaneal BMD change (mg/cm³ per year) were not impacted by sex, PN severity as determined by score on the Michigan Neuropathy Screening Instrument, daily step count, sedentary time, or statin use (Table 4). Data for daily step count and sedentary time was obtained from ActiGraph data, which was initiated after the beginning of the study, resulting in an incomplete baseline data set for these variables. There was a significant main effect of age on mean (p = 0.0086) and calcaneal (p = 0.0256) BMD (mg/cm³) with lower BMD associated with older age. The linear mixed effects model analysis suggests that the average mean BMD decreases by 2.1793 mg/cm³ (p = 0.0086) and calcaneal BMD decreases by 2.19 mg/cm³ (p = 0.0256) with every one-year increase in age. Rate of change of BMD (mg/cm³ per year) was also negatively related to age.