We appreciate the interest of He et al. (1) and Zhao et al. (2) in our article and are happy to clarify and discuss further the aspects they raise. Our analysis was designed primarily to assess possible roles for osteoprotegerin, osteopontin, sclerostin, and osteocalcin values in the stratification of cardiovascular risk for people with type 2 diabetes by leveraging biomarker data assayed in a subgroup of EXSCEL (Exenatide Study of Cardiovascular Event Lowering) participants. We provided a detailed comparison of the available baseline features of participants who were included, or not included, in the analyses in Supplementary Table 1 of our article (3), which showed no clinically meaningful differences. EXSCEL biomarkers were assayed with the SomaScan platform, which quantifies the relative concentration of ∼5,000 proteins using DNA-based binding reagents (modified aptamers called SOMAmers). This assay has demonstrated optimal analytic performance, has high reproducibility with a median 5% coefficient of variation (4), and has been validated in several cardiovascular disease studies (4,5). A full description of the SomaScan technology, including operating procedures, reliability, comparison across different proteomic platforms, and differences from conventional clinical detection methods, has been published (5,6). Information concerning the quality controls used in the EXSCEL biomarker substudy was included in our original article (3). As the SomaScan assay measures relative, rather than absolute, concentrations, we cannot suggest optimal cutoff values describing the nonlinear relationship between osteocalcin and all-cause mortality. A previous study showing a similar J-shaped relationship between osteocalcin values and retinopathy reported 15–25 ng/mL as the osteocalcin range for minimal risk of retinopathy in postmenopausal women with type 2 diabetes (7). Whether such values could also be used for mortality risk stratification, and across different populations, remains to be evaluated.
Our results suggest a role of circulating osteokines in the pathophysiology of cardiovascular disease in diabetes, and we agree that mechanistic studies specifically investigating the biology of the associations we found are warranted. Although our results could prompt further research looking at whether osteokine change over time results in cardiovascular risk modifications, as suggested by He et al. (1), it is important to note that the addition of the bone biomarkers we investigated to existing cardiovascular disease prediction models for type 2 diabetes provided only minimal incremental value. Accordingly, we do not anticipate a potential role for longitudinal measurements of such molecules for the clinical monitoring of cardiovascular risk in type 2 diabetes.
We value the suggestions pertaining to possible unmeasured confounders, which should be considered when interpreting results of all observational studies (8). In this regard, both osteoporosis and vitamin D deficiency have been associated with increased cardiovascular risk. A role for circulating osteokines in such relationships could be hypothesized and should be evaluated in studies with the requisite data. In EXSCEL, baseline bone mineral density and history of fractures were not collected, but notably no study participants reported the use of antiosteoporosis drugs. Lifestyle, including physical activity and medical nutrition therapy, might theoretically affect the bone-vascular axis, as suggested by Zhao et al. (2), but evidence in this regard is lacking, as discussed in previous reports (9).
Whether the role of circulating osteokines differs among different subtypes of type 2 diabetes could be evaluated in future studies. Data in people with type 1 diabetes (10), or even people without diabetes (11), support the existence of a bone-vascular axis that is independent of insulin-deficient and insulin-resistant phenotypes. Zhao et al. (2) suggest a possible influence of incretin therapies on bone metabolism, which is evident in our analysis that showed an impact of once-weekly exenatide on osteoprotegerin and osteopontin concentrations. However, including a randomization arm as a covariate in our multivariate analysis did not appear to affect the relationships we described. Of note, this multivariate analysis also included estimated glomerular filtration rate as a covariate, which should limit a possible impact of nephropathy as a confounder. We evaluated the association between bone biomarkers and prevalent retinopathy in an earlier study conducted in a different cohort of people with type 2 diabetes (12).
Article Information
Funding and Duality of Interest. The EXSCEL trial was conducted jointly by the Duke Clinical Research Institute and the University of Oxford Diabetes Trials Unit, in collaboration with the sponsor, Amylin Pharmaceuticals, a wholly owned subsidiary of AstraZeneca. E.M. reports consulting fees and payment or honoraria for lectures, presentations, or advisory boards from Abbott, AstraZeneca, Eli Lilly, Merck-Serono KgA, MSD, MTD, NovoNordisk, and PikDare. M.N. reports funding for this study from AstraZeneca and stock or stock options with AstraZeneca, Moderna, and Pfizer. S.H.S. reports funding for this study from AstraZeneca. R.R.H. reports consulting fees from AstraZeneca, Novartis, and Lilly. No other potential conflicts of interest relevant to this article were reported.
Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Steven E. Kahn and Vanita R. Aroda.
