Results of our study revealed that over two years of follow-up, changes in trabecular bone microarchitecture are not different in women with and without type 2 diabetes. This is the first study to explore longitudinal changes in trabecular bone microarchitecture in women with type 2 diabetes. Our study provides important feasibility data which should be considered when planning and conducting subsequent longitudinal studies on trabecular bone microarchitecture change in older women with type 2 diabetes. The women with type 2 diabetes who did not attend the follow-up study visit had a different baseline trabecular bone microarchitecture phenotype than those who did attend the follow-up study visit. We demonstrated that for participants with type 2 diabetes who dropped out of the study, trabecular bone was less intact (i.e., larger trabecular bone holes, lower BVTV, greater branch density). Thus, our preliminary internal validity data suggests that it may be important to increase the frequency of study visits (i.e., annually or semi-annually) for older participants with type 2 diabetes and multiple comorbidities in order to retain participants and obtain data that is more generalizable to patients with type 2 diabetes of varying disease severity. Subsequent studies with adequate follow-up are needed to verify our observations. In particular, higher resolution imaging systems should be used to examine changes in trabecular bone microarchitecture relative to changes in cortical bone structure (i.e., cortical thickness, porosity), as cross-sectional studies suggest that cortical bone is compromised in women with type 2 diabetes .
Skeletal change in adults with type 2 diabetes has been limited to the description of BMD change; however, whether individuals with diabetes lose bone at a faster rate than non-diabetics is unclear [32–34]. Younger women with diabetes  and women with newly diagnosed diabetes  experienced greater losses in hip BMD, whereas in women with diabetes for more than twelve years  and in postmenopausal women , no differences in the rate of BMD loss have been reported. It is possible that the greatest losses in bone occur during the years surrounding the diagnosis of type 2 diabetes when the likelihood of hyperglycemia, hypercalcuria, and generation of reactive oxygen species (ROS) is high [35–37]. The negative impact of hyperglycemia and ROS on osteoblasts has been demonstrated in vitro [38, 39], and is a potential mechanism causing diabetic bone fragility . Variability exists in the concentration of these factors in adults with type 2 diabetes, depending on duration and control of diabetes . This may explain the discrepancy in BMD change in women with diabetes, and supports our finding of no difference in the change in trabecular bone microarchitecture variables in older postmenopausal women with long-standing type 2 diabetes given that the average length of time since diabetes diagnosis in our study was over 18 years. Weight loss is another factor involved in proximal femur bone loss in adults with type 2 diabetes [41, 42]. In a large prospective study, Caucasian women with type 2 diabetes lost more femoral neck BMD over 4 years than Caucasian women without type 2 diabetes, and weight loss was an important mediator of this relationship . The mechanism involved is likely related to the reduction in skeletal loading with weight loss . In our study, weight was approximately stable over the two years in both groups, and weight change was not related to the change in trabecular bone hole size (data not shown). Further, we assessed trabecular bone at a non-weightbearing site, which is unlikely to be impacted by weight change.
In studies with similar sample sizes to our study, non-diabetic women taking alendronate  and estrogen supplementation  experienced no change in some microarchitectural variables assessed with MRI at the radius. Similarly, nasal calcitonin does not change trabecular bone microarchitecture at the more distal sites of the radius, but does preserve microarchitecural quality at proximal radius sites . We speculate that losses in trabecular bone microarchitecture at more proximal sites might be apparent in women with newly diagnosed type 2 diabetes, which should be investigated in the future. While measures of trabecular bone microarchitecture can be assessed using MRI, peripheral quantitative computed tomography (pQCT) and high-resolution pQCT (HR-pQCT), the superior signal to noise ratio with MRI allows for improved differentiation between bone and marrow [45–47]. Studies have however shown moderate to strong correlations between measures derived by pQCT or HR-pQCT and MRI [10, 48].
There were several study limitations. First, our sample size was small as approximately 50% of participants with type 2 diabetes either dropped out of the study, were lost to follow-up or died after the baseline assessment. To assess the internal validity of the study, we compared the baseline descriptive characteristics and trabecular bone microarchitecture variables for the participants who dropped out to those who remained in the study. In women with diabetes who dropped out, the trabecular bone microarchitecture appeared less intact, and in women without diabetes who dropped out, the trabecular bone network appeared to have more holes compared to those who returned. It is possible that the individuals who dropped out had more comorbidities that were not assessed in this study. For example, subclinical peripheral arterial disease, which we did not assess, has been linked to reduced bone mineral content  and to osteoporotic fractures in adults with type 2 diabetes . Due to the large number of participants who did not return for follow-up assessment, our results may have been biased towards not detecting a difference in trabecular bone microarchitectural changes, given the baseline differences between those who dropped out and returned to complete the study. Our analysis should be considered exploratory and will need to be confirmed in larger studies. We were also unable to accurately capture diabetes-related complications in this study, however future studies should consider the role of comorbidities, particularly neuropathy and nephropathy  when examining bone loss in diabetics. While multivariable linear regression models were used to account for the differences in ethnicity between women with diabetes and controls, the study would have been strengthened if participants were matched based on ethnicity. In addition, the resolution of the images acquired with our 1 Tesla MRI system restricts our analyses to trabecular bone, and is not appropriate for the assessment of distal radius cortical bone. A 1 Telsa MRI system is limited by signal strength, although we attempted to optimize signal strength by using a small (100 mm diameter) radio-frequency coil to enhance the signal-to-noise ratio. The lack of overall microarchitectural change may have been due to the lower signal strength of the 1 Tesla MRI system and it’s limited ability to detect small changes in microarchitecture, such as the thickness of trabeculae. Future studies should explore longitudinal changes in trabecular bone microarchitecture using 1.5, 3 or 7 Tesla systems which have superior signal-to-noise ratio and higher image resolution. Finally, no prospective data were available at study inception on the change in the size or number of trabecular bone holes, therefore we were unable to estimate an ideal sample size required at follow-up to capture differences in these key variables. Given our study limitations, larger studies with more complete follow-up, the ability to look at multiple study outcomes and those that assess potential confounders (i.e., diabetes related complications) are needed prior to making definitive conclusions about the lack of change in trabecular bone microarchitecture in women with type 2 diabetes.