Frontiers in Medicine · Nov 18, 2026
Association of BMP4 rs17563 With 18-Month DXA-Defined Incident Osteoporosis in Older Adults With Poorly Controlled Type 2 Diabetes: A Single-Center Retrospective Cohort Study
By Yanpeng Xu, Lixin Cao, Wei Gao et al.
Background: Older adults with type 2 diabetes mellitus may develop osteoporosis despite non-osteoporotic bone mineral density at baseline. Whether targeted bone morphogenetic protein-related variants are associated with subsequent DXA-defined osteoporosis and provide additional discrimination beyond routine clinical and baseline bone-density measures remains uncertain. Methods: We conducted a single-center retrospective cohort study of hospitalized patients aged ≥60 years with type 2 diabetes, HbA1c ≥7.0%, and non-osteoporotic baseline DXA findings. The primary outcome was DXA-defined incident osteoporosis at approximately 18 months. BMP2 rs1005464, BMP2 rs235768, and BMP4 rs17563 were evaluated as candidate loci. The primary multivariable model included age, sex, body mass index, eGFR, baseline lowest T-score, and the selected BMP4 rs17563 CC-versus-TT/TC term. A five-parameter clinical core model was compared with the corresponding clinical-genetic model as a secondary analysis. Fixed-model and selection-aware bootstrap procedures used 1,000 resamples; the latter repeated comparison of all three candidate loci under additive, dominant, and recessive coding. Sensitivity analyses evaluated additive and dominant BMP4 coding, additional HbA1c adjustment, and exclusion of participants with baseline lowest T-score ≤−2.0. Results: Among 318 screened patients, 255 met baseline eligibility criteria and 240 had an ascertainable follow-up DXA outcome; 61/240 (25.4%) developed incident osteoporosis. BMP4 rs17563 CC genotype was present in 16/61 (26.2%) participants with incident osteoporosis and 10/179 (5.6%) event-free participants. In the six-parameter model, BMP4 rs17563 CC genotype was associated with incident osteoporosis (adjusted OR, 3.68; 95% CI, 1.42–9.53; P = 0.007). Apparent AUCs were 0.84 (95% CI, 0.78–0.90) for the clinical core model and 0.87 (95% CI, 0.81–0.93) for the clinical-genetic model, with an apparent ΔAUC of +0.029 (95% CI, 0.003–0.056; P = 0.028). In the selection-aware bootstrap analysis, corrected AUCs were 0.82 and 0.83, and the corrected ΔAUC was +0.013 (95% empirical bootstrap interval, −0.024 to 0.041). Conclusion: BMP4 rs17563 CC genotype was associated with approximately 18-month DXA-defined incident osteoporosis in this cohort. The incremental discrimination associated with the selected marker was small and remained statistically uncertain after accounting for candidate-locus and coding selection. Independent replication is required, and these findings do not establish causal effects or clinical predictive utility.