Association of bone mineral density with perioperative hidden blood loss in older adults undergoing intramedullary nailing for unstable intertrochanteric fractures Authors: Yan Qu, Kairui Pan, Shuo Wang, Changwen Gu, Ruru Wang, Hua Wang Published: 2026 ABSTRACT BackgroundThe influence of bone mineral density (BMD) on hidden blood loss (HBL) in older patients with unstable intertrochanteric fractures treated with intramedullary nailing (IMN) is underexplored. The current retrospective cohort study aimed to evaluate the relationship between BMD and perioperative HBL in these cases.Materials and methodsA total of 115 patients aged > 60 years undergoing IMN between January 2022 and December 2024 were included. Data included demographics, BMD T score, preoperative hemoglobin (Hb), medical history, and HBL(calculated using the Nadler formula). Spearman correlation, Mann-Whitney U tests, and multiple linear regression were applied to identify factors associated with HBL.ResultsPatients (mean age 82.90 ± 9.48 years, 29.57% male) had a mean BMD T-score of -3.29 ± 2.17. Lower BMD T-score (β = -14.35, p < 0.001) and lower preoperative Hb (β = -1.282, p = 0.033) were independently associated with greater HBL on postoperative day 3; a history of hypertension was independently associated with lower HBL after adjustment (β = -24.56, p = 0.016). Similar associations were observed on postoperative day 7 (BMD T-score: β = -17.14, p < 0.001; hypertension: β = -23.22, p = 0.045). Other factors showed no significant association.ConclusionIn this retrospective cohort study of older adults undergoing intramedullary nailing for unstable intertrochanteric fractures, lower BMD T-score and lower preoperative Hb were independently associated with greater perioperative HBL, while a history of hypertension appeared to be independently associated with lower HBL after adjustment. These associations were moderate in magnitude and should be interpreted with caution given the retrospective design, imbalanced BMD subgroups, and missing postoperative day-7 data. Preoperative BMD assessment and Hb optimization may inform perioperative risk stratification, though prospective validation is required. Introduction Unstable intertrochanteric fractures in older adults are associated with substantial morbidity, mortality, and healthcare costs, and they frequently occur in patients with osteoporosis [1, 2]. Intramedullary nailing (IMN) is widely used because it provides reliable biomechanical stability with relatively minimal invasiveness [3]. Nevertheless, perioperative hidden blood loss (HBL) remains common after IMN and may contribute to postoperative anemia, increased transfusion requirements, and prolonged hospitalization [4, 5]. HBL is defined as the discrepancy between visible blood loss and the actual perioperative reduction in hemoglobin/hematocrit, and previous studies in hip fracture surgery have shown that HBL can account for a substantial proportion of total blood loss [6, 7]. Multiple patient-related, laboratory, and surgical variables have been reported as potential predictors of HBL [8–10]. However, evidence specifically clarifying the role of bone mineral density (BMD) in older patients with unstable intertrochanteric fractures treated with IMN remains limited, and the association between BMD and HBL has not been fully elucidated in this specific population [11]. Osteoporosis has been associated with increased hidden blood loss in other orthopedic procedures as well, supporting biological plausibility for a link between low BMD and greater perioperative blood loss [12, 13]. Therefore, we hypothesized that patients with lower BMD (i.e., a lower T-score) would have greater perioperative HBL following IMN for unstable intertrochanteric fractures. Because HBL is derived from perioperative hematocrit/hemoglobin changes, it was quantified postoperatively (on postoperative day 3 and postoperative day 7) rather than preoperatively. The aim of this retrospective cohort study was to evaluate the association between preoperative BMD and postoperatively assessed perioperative HBL (POD3 and POD7) and to identify factors independently associated with HBL in this population. Materials and methods Study design This study is a retrospective cohort study aimed at evaluating the impact of BMD on perioperative HBL in older patients with unstable intertrochanteric fracture of the femur, with HBL assessed postoperatively on postoperative day 3 and postoperative day 7, providing a basis for optimizing clinical management. Inclusion and exclusion criteria Inclusion criteria: Patients (1) age > 60 years old; (2) diagnosed with unstable intertrochanteric fractures through imaging examination (classified as AO/ASIF 31-A2 and AO 31-A3) [14]; (3) treated with IMN fixation; (4) with complete medical records; (5) the total liquid infusion volume is ≤ 2000 mL on the surgery day. Exclusion criteria: Patients (1) with multiple fractures, open fractures, or pathological fractures; (2) with coagulation dysfunction; (3) using medications affecting blood coagulation within two weeks prior to surgery. Patients receiving anticoagulant or antiplatelet therapy within two weeks prior to surgery were excluded to minimize confounding effects on perioperative hemoglobin changes and hidden blood loss. Patient inclusion/exclusion and postoperative laboratory-data completeness are summarized in Fig. 1. Fig. 1Study flowchart. Patients using anticoagulants or antiplatelet agents within two weeks preoperatively were excluded (n = 34). Postoperative day-7 Hb/HCT was unavailable in 40 patients (34.8%); thus, postoperative day-7 HBL was calculated in the remaining 75 patients. To assess potential selection bias from missing postoperative day-7 data, we compared baseline characteristics between patients with and without day-7 measurements. No significant differences were observed in age (p = 0.263), BMI (p = 0.980), BMD T-score (p = 0.426), preoperative hemoglobin (p = 0.102), sex distribution (p = 1.000), or hypertension prevalence (p = 0.695) between the two groups, supporting the assumption that the missing data were likely missing at random (MAR) and that the available day-7 sample was broadly representative of the full cohort Study flowchart. Patients using anticoagulants or antiplatelet agents within two weeks preoperatively were excluded (n = 34). Postoperative day-7 Hb/HCT was unavailable in 40 patients (34.8%); thus, postoperative day-7 HBL was calculated in the remaining 75 patients. To assess potential selection bias from missing postoperative day-7 data, we compared baseline characteristics between patients with and without day-7 measurements. No significant differences were observed in age (p = 0.263), BMI (p = 0.980), BMD T-score (p = 0.426), preoperative hemoglobin (p = 0.102), sex distribution (p = 1.000), or hypertension prevalence (p = 0.695) between the two groups, supporting the assumption that the missing data were likely missing at random (MAR) and that the available day-7 sample was broadly representative of the full cohort The current study included a total of 115 patients treated in the institution between January 2022 and December 2024. The sample size is based on all eligible patients during the time period to maximize statistical power. As this was a retrospective study, no a priori sample-size calculation was performed; all eligible patients during the study period were included. A post hoc power analysis was not performed; instead, we report effect estimates with corresponding confidence intervals where applicable. Data collection The following data were collected: (1) Demographic and clinical characteristics: age, gender, height, weight, BMI; (2) Laboratory examinations: hemoglobin (Hb) and hematocrit (HCT) on the day before surgery and on the 1st, 3rd, and 7th postoperative days; (2) coagulation parameters (D-dimer, prothrombin time [PT], activated partial thromboplastin time [APTT], thrombin time [TT], fibrinogen [Fbg]) on the day before surgery; (3) Medical history: hypertension, hyperlipidemia, diabetes, hypercortisolism, fibrous osteitis, myeloma, bone tumor, osteomyelitis; (4) Medication history: (i) preoperative prophylactic antibiotic administration (yes/no), defined as receipt of prophylactic antibiotics given routinely before incision according to our institutional hip-fracture protocol; and (ii) perioperative analgesic use (yes/no), defined as receipt of routine perioperative pain-control medications (e.g., NSAIDs and/or opioids as clinically indicated). Both items were extracted independently from medication orders/administration records. (5) Bone mineral density (BMD): BMD was evaluated using dual-energy X-ray absorptiometry (DXA) before surgery. According to the WHO criteria, patients were classified as osteoporosis if the T-score ≤ -2.5, osteopenia if -2.5 < T-score < -1.0, and normal BMD if T-score ≥ -1.0 [15]. In our institution, assessment of bone mineral density is incorporated into the routine management pathway for geriatric fragility hip fractures (including unstable intertrochanteric fractures) to support perioperative decision-making and to facilitate early initiation of anti-osteoporosis therapy for secondary fracture prevention. BMD testing was performed during the index hospitalization according to this departmental protocol (preoperatively when feasible after analgesia and stabilization; otherwise early postoperatively prior to discharge), and BMD data were available for all patients included in the final analysis. This practice was implemented throughout the entire study period of the present retrospective cohort. BMD was measured using a dual-energy X-ray absorptiometry (DXA) densitometer (MEDIX 90; MEDILINK, France). Formal global health-status scores (e.g., ASA classification) and comprehensive geriatric/frailty indices (e.g., Nottingham Hip Fracture Score or Charlson Comorbidity Index) were not routinely recorded in our electronic medical records during the study period and therefore were not available for analysis; instead, baseline clinical status was characterized using the documented comorbidities and laboratory parameters collected above. Intraoperative variables—including hemodynamic parameters (e.g., blood pressure), operative time, incision length, and degree of soft-tissue compromise—were not consistently available in a structured format in this retrospective dataset and therefore were not included in the current analyses. All procedures were performed by the same attending chief orthopedic surgeon (experienced hip surgeon/consultant), therefore surgeon-related variability was minimized and surgeon factors were not modeled as independent predictors in the present analyses. Clinical outcomes The primary outcome was HBL on postoperative day 7, and the secondary outcome was HBL on postoperative day 3. HBL was calculated as: HBL = total blood loss (TBL) - visible blood loss (VBL) + allogeneic blood transfusion + autologous blood reinfusion [16].TBL was estimated using the patient blood volume (PBV) and perioperative hematocrit change [17]. PBV was calculated using Nadler’s formula: for male, PBV = 0.3669 × height (m)^3 + 0.03219 × weight (kg) + 0.604; for female, PBV = 0.3561 × height (m)^3 + 0.03308 × weight (kg) + 0.1833 [18]. VBL was defined as intraoperative blood loss plus other measurable losses (drainage volume, dressing seepage, and blood drawn for tests). The choice of postoperative day 7 as the primary outcome reflects a stabilized phase of perioperative hidden blood loss, as HBL typically peaks within 3–5 days after injury and surgery and then declines with hematopoietic recovery. Statistical analysis Continuous variables are expressed as mean ± standard deviation (Mean ± SD) if normally distributed or as median (interquartile range, IQR) if non-normally distributed. The normality of continuous variables was tested by the Shapiro-Wilk test. For normally distributed data, group differences were analyzed using independent t-tests for two groups or one-way ANOVA for three or more groups. For non-normally distributed data, the Mann-Whitney U test was used for two groups, and the Kruskal-Wallis test for three or more groups. When an overall between-group difference was detected in analyses involving three BMD categories (osteoporosis, osteopenia, and normal BMD), post hoc pairwise comparisons were performed (Dunn’s multiple-comparisons test) with Bonferroni-adjusted p values reported. To quantify the magnitude of overall between-group differences for the Kruskal–Wallis test, epsilon-squared (ε²) was computed as (H - k + 1) / (N - k), where H is the Kruskal–Wallis statistic, k is the number of groups, and N is the total sample size; ε² values were interpreted as negligible (< 0.01), small (0.01–0.06), moderate (0.06–0.14), or large (> 0.14). For pairwise post hoc comparisons, Cliff’s delta (δ) was computed as the effect size, interpreted as negligible (|δ| < 0.15), small (0.15–0.33), medium (0.33–0.47), or large (> 0.47). The median difference between groups was estimated using the Hodges–Lehmann estimator with its 95% confidence interval. Given the markedly unequal group sizes in the three BMD categories (osteoporosis, n = 70; osteopenia, n = 24; normal BMD, n = 21), no formal a priori power calculation was performed for between-group comparisons, and the statistical power for pairwise comparisons involving the smaller subgroups—particularly the normal-BMD group—was inherently limited; these comparisons should therefore be interpreted with appropriate caution. Categorical variables are expressed in frequency and percentage. Associations between categorical variables and outcomes were evaluated using the chi-square test or Fisher’s exact test (for expected frequencies < 5). Spearman correlation analysis was used to evaluate the correlation between continuous variables and HBL. Mann Whitney U test was used to evaluate the association between categorical variables and HBL. For Spearman correlation coefficients, the magnitude of association was interpreted according to conventional thresholds as negligible (|ρ| < 0.10), weak (0.10 ≤ |ρ| < 0.30), moderate (0.30 ≤ |ρ| < 0.50), or strong (|ρ| ≥ 0.50). Because Spearman’s ρ is a rank-based statistic for which the Fisher z transformation provides only an approximation, 95% confidence intervals for Spearman’s ρ were calculated using non-parametric bootstrap resampling (10,000 resamples; bias-corrected percentile method). Variables with p < 0.2 in the univariable correlation analysis were included into the multiple linear regression analysis to explore independent risk factors for HBL. This threshold was chosen to avoid premature exclusion of clinically plausible predictors while keeping the model parsimonious. Candidate variables were further guided by a priori clinical reasoning: BMD T-score and preoperative hemoglobin were included based on established biological plausibility (bone fragility increasing bleeding risk; preoperative anemia reducing tolerance to blood loss), and hypertension was included given its known association with vascular changes and perioperative blood pressure fluctuation. Variables not reaching the p < 0.2 threshold in univariable analysis (e.g., diabetes, hyperlipidemia, coagulation parameters) were excluded, consistent with the absence of strong a priori mechanistic rationale for their direct influence on perioperative HBL in this population. The analysis was completed using GraphPad Prism (version 9.5.1; GraphPad Software, San Diego, CA, USA), with a significance level set at p < 0.05. Results Patient characteristics A total of 115 patients were included, with an average age of 82.90 ± 9.48 years. Males accounted for 29.57% (n = 34) and females accounted for 70.43% (n = 81). The average BMI is 22.53 ± 2.69 kg/m2, and the average BMD T score is -3.29 ± 2.17. Hypertension (60.00%), diabetes (28.70%) and hyperlipidemia (20.87%) were the most common disease histories. Preoperative prophylactic antibiotics and perioperative analgesics were used in 78.26% of patients, respectively. The identical proportions reflect our standardized perioperative order set in which prophylactic antibiotics and baseline perioperative analgesia are commonly prescribed together; these two variables were nevertheless extracted as separate medication items from the electronic medication records. In addition, 66.09% used antihypertensive drugs. The detailed characteristics are shown in Table 1. Table 1Patient characteristicTotal (N = 115)Age(year)Mean ± SD82.90 ± 9.48SexMale (n,%)34 (29.57)Female (n,%)81 (70.43)Height(cm)Mean ± SD161.17 ± 7.54Weight(Kg)Mean ± SD58.75 ± 7.43BMI(kg/m2)Mean ± SD22.53 ± 2.69BMD T scoreMean ± SD-3.29 ± 2.17History of hypertensionYes (n,%)69 (60.00)No (n,%)46 (40.00)History of hyperlipidemiaYes (n,%)24 (20.87)No (n,%)91 (79.13)History of diabetesYes (n,%)33 (28.70)No (n,%)82 (71.30)History of hypercortisolismYes (n,%)2 (1.74)No (n,%)113 (98.26)History of fibrous osteitisYes (n,%)3 (2.61)No (n,%)112 (97.39)History of MyelomaYes (n,%)1 (0.87)No (n,%)114 (99.13)History of bone tumorsYes (n,%)2 (1.74)No (n,%)113 (98.26)History of osteomyelitisYes (n,%)2 (1.74)No (n,%)113 (98.26)Preoperative prophylactic antibiotic useYes (n,%)90 (78.26)No (n,%)25 (21.74)Perioperative analgesic useYes (n,%)90 (78.26)No (n,%)25 (21.74)History of using antihypertensive drugsYes (n,%)76 (66.09)No (n,%)39 (33.91)History of antidepressant useYes (n,%)3 (2.61)No (n,%)112 (97.39)History of use of inhaled corticosteroidsYes (n,%)4 (3.48)No (n,%)111 (96.52)History of using bronchodilatorsYes (n,%)7 (6.09)No (n,%)108 (93.91)Single-surgeon procedureYes (n,%)115 (100.00)No (n,%)0 (0.00)Lead surgeon: attending chief hip surgeon/consultantYes (n,%)115 (100.00)No (n,%)0 (0.00)History of diuretic useYes (n,%)8 (6.96)No (n,%)107 (93.04Postoperative 3 day HBLMean ± SD110.65 ± 61.82Postoperative 7 day HBLMean ± SD132.56 ± 71.75Height, weight, BMI and BMD T-score were assessed for normality using the Shapiro-Wilk test and are presented as mean ± SD when normally distributed; non-normally distributed variables are presented as median (IQR). Perioperative antibiotic use refers to preoperative prophylactic antibiotics routinely administered for hip-fracture surgery according to the institutional protocol. Perioperative analgesic use refers to routine perioperative pain-control medications (e.g., NSAIDs/opioids as clinically indicated). All procedures were performed by a single attending chief orthopedic surgeon (experienced hip surgeon/consultant)HBL Hidden blood loss, BMI Body mass index, BMD Bone mineral density Table 1Patient characteristicTotal (N = 115)Age(year)Mean ± SD82.90 ± 9.48SexMale (n,%)34 (29.57)Female (n,%)81 (70.43)Height(cm)Mean ± SD161.17 ± 7.54Weight(Kg)Mean ± SD58.75 ± 7.43BMI(kg/m2)Mean ± SD22.53 ± 2.69BMD T scoreMean ± SD-3.29 ± 2.17History of hypertensionYes (n,%)69 (60.00)No (n,%)46 (40.00)History of hyperlipidemiaYes (n,%)24 (20.87)No (n,%)91 (79.13)History of diabetesYes (n,%)33 (28.70)No (n,%)82 (71.30)History of hypercortisolismYes (n,%)2 (1.74)No (n,%)113 (98.26)History of fibrous osteitisYes (n,%)3 (2.61)No (n,%)112 (97.39)History of MyelomaYes (n,%)1 (0.87)No (n,%)114 (99.13)History of bone tumorsYes (n,%)2 (1.74)No (n,%)113 (98.26)History of osteomyelitisYes (n,%)2 (1.74)No (n,%)113 (98.26)Preoperative prophylactic antibiotic useYes (n,%)90 (78.26)No (n,%)25 (21.74)Perioperative analgesic useYes (n,%)90 (78.26)No (n,%)25 (21.74)History of using antihypertensive drugsYes (n,%)76 (66.09)No (n,%)39 (33.91)History of antidepressant useYes (n,%)3 (2.61)No (n,%)112 (97.39)History of use of inhaled corticosteroidsYes (n,%)4 (3.48)No (n,%)111 (96.52)History of using bronchodilatorsYes (n,%)7 (6.09)No (n,%)108 (93.91)Single-surgeon procedureYes (n,%)115 (100.00)No (n,%)0 (0.00)Lead surgeon: attending chief hip surgeon/consultantYes (n,%)115 (100.00)No (n,%)0 (0.00)History of diuretic useYes (n,%)8 (6.96)No (n,%)107 (93.04Postoperative 3 day HBLMean ± SD110.65 ± 61.82Postoperative 7 day HBLMean ± SD132.56 ± 71.75Height, weight, BMI and BMD T-score were assessed for normality using the Shapiro-Wilk test and are presented as mean ± SD when normally distributed; non-normally distributed variables are presented as median (IQR). Perioperative antibiotic use refers to preoperative prophylactic antibiotics routinely administered for hip-fracture surgery according to the institutional protocol. Perioperative analgesic use refers to routine perioperative pain-control medications (e.g., NSAIDs/opioids as clinically indicated). All procedures were performed by a single attending chief orthopedic surgeon (experienced hip surgeon/consultant)HBL Hidden blood loss, BMI Body mass index, BMD Bone mineral density Patient characteristic Height, weight, BMI and BMD T-score were assessed for normality using the Shapiro-Wilk test and are presented as mean ± SD when normally distributed; non-normally distributed variables are presented as median (IQR). Perioperative antibiotic use refers to preoperative prophylactic antibiotics routinely administered for hip-fracture surgery according to the institutional protocol. Perioperative analgesic use refers to routine perioperative pain-control medications (e.g., NSAIDs/opioids as clinically indicated). All procedures were performed by a single attending chief orthopedic surgeon (experienced hip surgeon/consultant) HBL Hidden blood loss, BMI Body mass index, BMD Bone mineral density The correlation between BMD and HBL Based on the T-value, 70 patients were assigned to the osteoporosis group, 24 to the osteopenia group, and 21 to the normal BMD group. We found that postoperative day 3 HBL and postoperative day 7 HBL were highest in the osteoporosis group, whereas both time points were lowest in the normal BMD group (Fig. 2A and 2B). These observed group differences should, however, be interpreted with caution given the markedly unequal subgroup sizes (osteoporosis, n = 70; osteopenia, n = 24; normal BMD, n = 21), which limit the statistical power of pairwise comparisons, particularly for those involving the small normal-BMD subgroup. Post hoc pairwise comparisons using Dunn’s test with Bonferroni correction were performed to identify which specific group comparisons drove the overall significant difference. The osteoporosis group showed significantly higher HBL than both the osteopenia and normal BMD groups at both time points. At postoperative day 3, the Kruskal–Wallis test revealed a significant overall between-group difference (H = 30.16, p < 0.0001, ε² = 0.251, large effect). The median [IQR] HBL was 132.0 [106.2–156.8] mL in the osteoporosis group, 64.5 [48.5–122.5] mL in the osteopenia group, and 55.0 [18.0–100.0] mL in the normal BMD group. Bonferroni-corrected Dunn’s post hoc analysis demonstrated that the osteoporosis group had significantly higher HBL than the osteopenia group (adjusted p = 0.008; Cliff’s δ = 0.41 [medium effect]; Hodges–Lehmann difference = 42.0 mL, 95% CI [16.0, 68.0]) and the normal BMD group (adjusted p = 0.001; Cliff’s δ = 0.51 [large effect]; Hodges–Lehmann difference = 54.0 mL, 95% CI [25.0, 85.0]). The difference between the osteopenia and normal BMD groups was not statistically significant after Bonferroni correction (adjusted p = 1.000; Cliff’s δ = 0.12 [negligible effect]; Hodges–Lehmann difference = 13.5 mL, 95% CI [-22.0, 47.0]), likely reflecting the limited statistical power available for this comparison given the small normal-BMD subgroup (n = 21). At postoperative day 7, the overall between-group difference was also significant (H = 20.27, p < 0.0001, ε² = 0.254, large effect). The median [IQR] HBL was 161.0 [119.0–203.0] mL in the osteoporosis group, 96.0 [81.2–171.0] mL in the osteopenia group, and 50.5 [33.0–91.8] mL in the normal BMD group. Post hoc comparisons showed that the osteoporosis group had significantly higher HBL than the osteopenia group (adjusted p = 0.027; Cliff’s δ = 0.43 [medium]; Hodges–Lehmann = 50.0 mL, 95% CI [14.0, 86.0]) and the normal BMD group (adjusted p = 0.0002; Cliff’s δ = 0.77 [large]; Hodges–Lehmann = 97.0 mL, 95% CI [57.0, 137.0]). The osteopenia versus normal BMD comparison did not reach significance after correction (adjusted p = 0.088; Cliff’s δ = 0.48 [large effect]; Hodges–Lehmann = 48.5 mL, 95% CI [10.0, 93.0]). Full pairwise comparison results are also illustrated in Fig. 2A and 2B. We then performed Spearman correlation analysis to evaluate the association between the BMD T-score and HBL. The results showed that the BMD T-score was significantly correlated with postoperative day 3 HBL (Spearman’s ρ = -0.492, 95% bootstrap confidence interval [CI]: -0.631 to -0.322, p < 0.001) and postoperative day 7 HBL (ρ = -0.513, 95% bootstrap CI: -0.678 to -0.309, p < 0.001) (Fig. 2C). These correlation coefficients indicate a moderate negative association between BMD T-score and perioperative HBL (|ρ| ≈ 0.50, explaining approximately 25% of HBL variance), and should be interpreted as a meaningful but partial statistical relationship rather than a strong or deterministic association. Fig. 2The correlation between BMD and HBL. A Comparison of postoperative day 3 (POD3) HBL across BMD groups (osteoporosis, n = 70; osteopenia, n = 24; normal BMD, n = 21). B Comparison of postoperative day 7 (POD7) HBL across the same BMD groups (n = 50, 16, and 9, respectively, after exclusion of patients without POD7 measurements). Overall between-group differences were assessed by the Kruskal–Wallis test, and post hoc pairwise comparisons were performed using Dunn’s test with Bonferroni correction; full pairwise adjusted p values, Hodges–Lehmann median differences with 95% CIs, and Cliff’s δ effect sizes are reported in the Results section. Asterisks shown above the boxplots indicate the conventional graphical thresholds (* p < 0.05, ** p < 0.01, *** p < 0.001) corresponding to the Bonferroni-adjusted p values from Dunn’s test and are provided for visual reference only; precise adjusted p values and effect sizes should be interpreted as reported in the text. C Scatter plot of BMD T-score versus HBL at POD3 (blue circles, n = 115) and POD7 (red triangles, n = 75). Solid lines represent ordinary least-squares (OLS) linear regression fits; shaded bands (or dotted lines) indicate the 95% confidence bands of the best-fit line. Spearman’s rank correlation coefficients (ρ) with 95% bootstrap confidence intervals (10,000 resamples) were: POD3, ρ = -0.492 (95% CI: -0.631 to -0.322), p < 0.001; POD7, ρ = -0.513 (95% CI: -0.678 to -0.309), p < 0.001. Note that Spearman’s ρ quantifies the monotonic association, whereas the regression line and confidence bands depict the OLS linear fit; these are complementary descriptors. HBL, hidden blood loss; BMD, bone mineral density; POD, postoperative day; CI, confidence interval; OLS, ordinary least squares The correlation between BMD and HBL. A Comparison of postoperative day 3 (POD3) HBL across BMD groups (osteoporosis, n = 70; osteopenia, n = 24; normal BMD, n = 21). B Comparison of postoperative day 7 (POD7) HBL across the same BMD groups (n = 50, 16, and 9, respectively, after exclusion of patients without POD7 measurements). Overall between-group differences were assessed by the Kruskal–Wallis test, and post hoc pairwise comparisons were performed using Dunn’s test with Bonferroni correction; full pairwise adjusted p values, Hodges–Lehmann median differences with 95% CIs, and Cliff’s δ effect sizes are reported in the Results section. Asterisks shown above the boxplots indicate the conventional graphical thresholds (* p < 0.05, ** p < 0.01, *** p < 0.001) corresponding to the Bonferroni-adjusted p values from Dunn’s test and are provided for visual reference only; precise adjusted p values and effect sizes should be interpreted as reported in the text. C Scatter plot of BMD T-score versus HBL at POD3 (blue circles, n = 115) and POD7 (red triangles, n = 75). Solid lines represent ordinary least-squares (OLS) linear regression fits; shaded bands (or dotted lines) indicate the 95% confidence bands of the best-fit line. Spearman’s rank correlation coefficients (ρ) with 95% bootstrap confidence intervals (10,000 resamples) were: POD3, ρ = -0.492 (95% CI: -0.631 to -0.322), p < 0.001; POD7, ρ = -0.513 (95% CI: -0.678 to -0.309), p < 0.001. Note that Spearman’s ρ quantifies the monotonic association, whereas the regression line and confidence bands depict the OLS linear fit; these are complementary descriptors. HBL, hidden blood loss; BMD, bone mineral density; POD, postoperative day; CI, confidence interval; OLS, ordinary least squares The correlation between general characteristics with HBL Age was positively correlated with postoperative day 3 HBL (ρ = 0.229, 95% CI: 0.0422 to 0.400, p = 0.014) and day 7 HBL (ρ = 0.215, 95% CI: 0.028 to 0.388, p = 0.021). There was no significant correlation between height, weight, BMI and HBL (p > 0.05), as shown in Table 2. Table 2The correlation between general characteristics with HBLPostoperative 3 day HBLPostoperative 7 day HBLρ95% CIp valueρ95% CIp valueAge0.2290.0422 to 0.4000.0140.2150.028 to 0.3880.021Height-0.011-0.199 to 0.1780.906-0.005-0.194 to 0.1830.955Weight-0.035-0.222 to 0.1540.710-0.008-0.196 to 0.1810.934BMI-0.141-0.321 to 0.0480.132-0.113-0.295 to 0.0770.229Hb Hemoglobin, HCT Hematocrit, HBL Hidden blood loss Table 2The correlation between general characteristics with HBLPostoperative 3 day HBLPostoperative 7 day HBLρ95% CIp valueρ95% CIp valueAge0.2290.0422 to 0.4000.0140.2150.028 to 0.3880.021Height-0.011-0.199 to 0.1780.906-0.005-0.194 to 0.1830.955Weight-0.035-0.222 to 0.1540.710-0.008-0.196 to 0.1810.934BMI-0.141-0.321 to 0.0480.132-0.113-0.295 to 0.0770.229Hb Hemoglobin, HCT Hematocrit, HBL Hidden blood loss The correlation between general characteristics with HBL Hb Hemoglobin, HCT Hematocrit, HBL Hidden blood loss Association of medical and medication history with HBL Chi-square test or Fisher’s exact test (for expected frequencies < 5) was used to analyze the association of medical and medication history with HBL. Results showed that the disease history (such as hypertension, diabetes, hyperlipidemia, fibro osteitis, bone tumor, osteomyelitis) and medication history (such as perioperative antibiotics, analgesics, antihypertensive drugs, antidepressants, inhaled glucocorticoids, bronchodilators, diuretics) were not significantly associated with HBL on the third and seventh days after surgery (p > 0.05, Table 3). Notably, although the unadjusted comparison did not reach statistical significance, patients with hypertension showed a higher unadjusted HBL than normotensive patients (POD3: 117.97 ± 64.83 vs. 100.06 ± 56.20 mL, p = 0.155; POD7: 140.82 ± 74.43 vs. 120.60 ± 66.67 mL, p = 0.168). However, after adjustment in the multivariable model, a history of hypertension was independently associated with lower HBL (Tables 6 and 7). Table 3Hidden blood loss by medical history and medication useFactorGroupPOD3 HBL (mL) p POD7 HBL (mL)p valueHistory of hypertensionNo100.06 ± 56.20120.60 ± 66.67History of hyperlipidemiaYes118.05 ± 62.750.505136.45 ± 70.690.824History of hyperlipidemiaNo108.70 ± 61.78131.52 ± 72.38History of diabetesYes104.74 ± 63.190.308133.02 ± 75.560.775History of diabetesNo113.02 ± 61.50132.37 ± 70.65History of hypercortisolismYes100.70 ± 29.780.914115.76 ± 20.450.974History of hypercortisolismNo110.97 ± 62.27132.85 ± 72.33History of fibrous osteitisYes80.56 ± 8.200.623104.15 ± 11.390.824History of fibrous osteitisNo111.45 ± 62.44133.32 ± 72.55History of bone tumorsYes87.06 ± 8.830.923101.41 ± 3.110.809History of bone tumorsNo111.07 ± 62.29133.11 ± 72.27History of osteomyelitisYes115.37 ± 100.470.809130.81 ± 113.860.763History of osteomyelitisNo110.56 ± 61.64132.59 ± 71.59History of perioperative antibiotic useYes111.41 ± 62.430.734135.59 ± 74.130.507History of perioperative antibiotic useNo107.89 ± 60.77121.65 ± 62.59History of perioperative analgesic useYes107.23 ± 61.280.252132.21 ± 74.740.684History of perioperative analgesic useNo122.78 ± 63.50133.79 ± 61.16History of using antihypertensive drugsYes113.76 ± 63.270.376135.88 ± 72.310.408History of using antihypertensive drugsNo104.59 ± 59.23126.08 ± 71.14History of antidepressant useYes155.62 ± 108.800.444176.65 ± 114.260.475History of antidepressant useNo109.44 ± 60.47131.38 ± 70.70History of use of inhaled corticosteroidsYes120.03 ± 64.840.843132.55 ± 70.820.947History of use of inhaled corticosteroidsNo110.31 ± 61.99132.56 ± 72.11History of using bronchodilatorsYes101.52 ± 55.770.594113.84 ± 60.510.445History of using bronchodilatorsNo111.24 ± 62.39133.77 ± 72.50History of diuretic useYes116.12 ± 52.430.616145.92 ± 67.460.506History of diuretic useNo110.24 ± 62.66131.56 ± 72.27Values are presented as mean ± SDHBL Hidden blood lossp values compare groups (Yes vs. No) for each factor Table 3Hidden blood loss by medical history and medication useFactorGroupPOD3 HBL (mL) p POD7 HBL (mL)p valueHistory of hypertensionNo100.06 ± 56.20120.60 ± 66.67History of hyperlipidemiaYes118.05 ± 62.750.505136.45 ± 70.690.824History of hyperlipidemiaNo108.70 ± 61.78131.52 ± 72.38History of diabetesYes104.74 ± 63.190.308133.02 ± 75.560.775History of diabetesNo113.02 ± 61.50132.37 ± 70.65History of hypercortisolismYes100.70 ± 29.780.914115.76 ± 20.450.974History of hypercortisolismNo110.97 ± 62.27132.85 ± 72.33History of fibrous osteitisYes80.56 ± 8.200.623104.15 ± 11.390.824History of fibrous osteitisNo111.45 ± 62.44133.32 ± 72.55History of bone tumorsYes87.06 ± 8.830.923101.41 ± 3.110.809History of bone tumorsNo111.07 ± 62.29133.11 ± 72.27History of osteomyelitisYes115.37 ± 100.470.809130.81 ± 113.860.763History of osteomyelitisNo110.56 ± 61.64132.59 ± 71.59History of perioperative antibiotic useYes111.41 ± 62.430.734135.59 ± 74.130.507History of perioperative antibiotic useNo107.89 ± 60.77121.65 ± 62.59History of perioperative analgesic useYes107.23 ± 61.280.252132.21 ± 74.740.684History of perioperative analgesic useNo122.78 ± 63.50133.79 ± 61.16History of using antihypertensive drugsYes113.76 ± 63.270.376135.88 ± 72.310.408History of using antihypertensive drugsNo104.59 ± 59.23126.08 ± 71.14History of antidepressant useYes155.62 ± 108.800.444176.65 ± 114.260.475History of antidepressant useNo109.44 ± 60.47131.38 ± 70.70History of use of inhaled corticosteroidsYes120.03 ± 64.840.843132.55 ± 70.820.947History of use of inhaled corticosteroidsNo110.31 ± 61.99132.56 ± 72.11History of using bronchodilatorsYes101.52 ± 55.770.594113.84 ± 60.510.445History of using bronchodilatorsNo111.24 ± 62.39133.77 ± 72.50History of diuretic useYes116.12 ± 52.430.616145.92 ± 67.460.506History of diuretic useNo110.24 ± 62.66131.56 ± 72.27Values are presented as mean ± SDHBL Hidden blood lossp values compare groups (Yes vs. No) for each factor Hidden blood loss by medical history and medication use Values are presented as mean ± SD HBL Hidden blood loss p values compare groups (Yes vs. No) for each factor The correlation between perioperative Hb/HCT and HBL Spearman correlation analysis showed a significant negative correlation between Hb on the day before surgery and HBL on day 3 (ρ=-0.281, p = 0.002) and day 7 (ρ=-0.243, p = 0.009) after surgery. There was also a significant negative correlation between preoperative HCT and postoperative HBL on day 3 (ρ=-0.270, p = 0.004) and day 7 (ρ=-0.217, p = 0.020). There was no significant correlation between postoperative Hb, HCT, and HBL (p > 0.05, Table 4, Fig. 2). Table 4The correlation between perioperative Hb/HCT and HBLPostoperative 3 day HBLPostoperative 7 day HBLρ95% CIp valueρ95% CIp valuePreoperative 1 day Hb-0.281-0.446 to -0.0980.002-0.243-0.412 to -0.0570.009Postoperative 1 day Hb-0.014-0.202 to 0.1750.885-0.019-0.170 to 0.2060.844Postoperative 3 days Hb-0.051-0.237 to 0.1390.590-0.054-0.240 to 0.1360.565Postoperative 7 days Hb-0.043-0.147 to 0.2290.6520.051-0.138 to 0.2380.585Preoperative 1 day HCT-0.270-0.437 to -0.0860.004-0.217-0.389 to -0.0300.020Postoperative 1 day HCT-0.051-0.237 to 0.1390.5910.009-0.179 to 0.1970.921Postoperative 3 days HCT-0.114-0.296 to 0.0760.225-0.042-0.229 to 0.1480.656Postoperative 7 days HCT0.205-0.3788 to -0.0170.028-0.152-0.331 to 0.0370.104Hb Hemoglobin, HCT Hematocrit, HBL Hidden blood loss Table 4The correlation between perioperative Hb/HCT and HBLPostoperative 3 day HBLPostoperative 7 day HBLρ95% CIp valueρ95% CIp valuePreoperative 1 day Hb-0.281-0.446 to -0.0980.002-0.243-0.412 to -0.0570.009Postoperative 1 day Hb-0.014-0.202 to 0.1750.885-0.019-0.170 to 0.2060.844Postoperative 3 days Hb-0.051-0.237 to 0.1390.590-0.054-0.240 to 0.1360.565Postoperative 7 days Hb-0.043-0.147 to 0.2290.6520.051-0.138 to 0.2380.585Preoperative 1 day HCT-0.270-0.437 to -0.0860.004-0.217-0.389 to -0.0300.020Postoperative 1 day HCT-0.051-0.237 to 0.1390.5910.009-0.179 to 0.1970.921Postoperative 3 days HCT-0.114-0.296 to 0.0760.225-0.042-0.229 to 0.1480.656Postoperative 7 days HCT0.205-0.3788 to -0.0170.028-0.152-0.331 to 0.0370.104Hb Hemoglobin, HCT Hematocrit, HBL Hidden blood loss The correlation between perioperative Hb/HCT and HBL Hb Hemoglobin, HCT Hematocrit, HBL Hidden blood loss The correlation between preoperative coagulation function and HBL Spearman correlation analysis was used to assess the correlation between preoperative coagulation function and HBL. There was no significant correlation between preoperative coagulation function parameters (D-dimer, PT, APTT, TT, Fbg) and postoperative HBL on the 3rd and 7th day (p > 0.05, Table 5). Table 5The correlation between preoperative coagulation function and HBLPostoperative 3 day HBLPostoperative 7 day HBLρ95% CIp valueρ95% CIp valuePreoperative 1 day D-dimer0.021-0.169 to 0.2080.828-0.027-0.214 to 0.1620.775Preoperative 1 day PT0.134-0.056 to 0.3140.1550.085-0.105 to 0.2700.364Preoperative 1 day APTT-0.013-0.201to 0.1760.8940.044-0.145 to 0.2310.637Preoperative 1 day TT0.103-0.087 to 0.2860.2750.100-0.090 to 0.2830.287Preoperative 1 day Fbg0.124-0.067 to 0.3060.1890.078-0.113 to 0.2630.410PT Prothrombin time, APTT Activated partial thromboplastin time, TT Thrombin time, Fbg Preoperative 1 day fibrinogen, HBL Hidden blood loss Table 5The correlation between preoperative coagulation function and HBLPostoperative 3 day HBLPostoperative 7 day HBLρ95% CIp valueρ95% CIp valuePreoperative 1 day D-dimer0.021-0.169 to 0.2080.828-0.027-0.214 to 0.1620.775Preoperative 1 day PT0.134-0.056 to 0.3140.1550.085-0.105 to 0.2700.364Preoperative 1 day APTT-0.013-0.201to 0.1760.8940.044-0.145 to 0.2310.637Preoperative 1 day TT0.103-0.087 to 0.2860.2750.100-0.090 to 0.2830.287Preoperative 1 day Fbg0.124-0.067 to 0.3060.1890.078-0.113 to 0.2630.410PT Prothrombin time, APTT Activated partial thromboplastin time, TT Thrombin time, Fbg Preoperative 1 day fibrinogen, HBL Hidden blood loss The correlation between preoperative coagulation function and HBL PT Prothrombin time, APTT Activated partial thromboplastin time, TT Thrombin time, Fbg Preoperative 1 day fibrinogen, HBL Hidden blood loss Analysis of factors independently associated with HBL Based on the variables with P < 0.2 in the correlation analysis, factors associated with HBL were examined using a multiple linear regression model. The results showed that BMD T score (β = -14.35, p < 0.001) and preoperative 1-day Hb (β = -1.282, p = 0.033) were independently associated with greater postoperative day 3 HBL, while a history of hypertension was independently associated with lower postoperative day 3 HBL after adjustment (β = -24.56, p = 0.016; R^2 = 0.405). For postoperative day 7 HBL, BMD T score (β = -17.14, p < 0.001) remained independently associated with greater HBL, and a history of hypertension was independently associated with lower HBL after adjustment (β = -23.22, p = 0.045; R^2 = 0.354). The association between preoperative Hb and postoperative day 7 HBL approached statistical significance (p = 0.056), suggesting its potential relevance (Tables 6 and 7). Table 6Multiple linear regression predicting hidden blood loss on postoperative day 3Predictorβ95% CIp valueAge-0.64-1.818, 0.5310.280BMD T score-14.35-19.35, -9.338< 0.001Preoperative 1 day Hb-1.28-2.459, -0.1040.033Preoperative 1 day HCT127.7-250.9, 506.20.505Postoperative 7 day HCT-132.2-501.7, 237.30.480History of hypertension-24.56-44.45, -4.6750.016BMI-3.43-7.083, 0.2150.065Preoperative 1 day PT-5.33-17.18, 6.5200.375Preoperative 1 day Fbg13.29-0.904, 27.490.066Categorical variables were coded as: history of hypertension (Yes = 1, No = 0). Negative β coefficients for hypertension therefore indicate lower HBL in patients with hypertension relative to those without, after adjustment for other predictorsHBL Hidden blood loss, BMD Bone mineral density, Hb Hemoglobin, HCT Hematocrit, BMI Body mass index, PT Prothrombin time, Fbg Fibrinogenp values are two-tailed Table 6Multiple linear regression predicting hidden blood loss on postoperative day 3Predictorβ95% CIp valueAge-0.64-1.818, 0.5310.280BMD T score-14.35-19.35, -9.338< 0.001Preoperative 1 day Hb-1.28-2.459, -0.1040.033Preoperative 1 day HCT127.7-250.9, 506.20.505Postoperative 7 day HCT-132.2-501.7, 237.30.480History of hypertension-24.56-44.45, -4.6750.016BMI-3.43-7.083, 0.2150.065Preoperative 1 day PT-5.33-17.18, 6.5200.375Preoperative 1 day Fbg13.29-0.904, 27.490.066Categorical variables were coded as: history of hypertension (Yes = 1, No = 0). Negative β coefficients for hypertension therefore indicate lower HBL in patients with hypertension relative to those without, after adjustment for other predictorsHBL Hidden blood loss, BMD Bone mineral density, Hb Hemoglobin, HCT Hematocrit, BMI Body mass index, PT Prothrombin time, Fbg Fibrinogenp values are two-tailed Multiple linear regression predicting hidden blood loss on postoperative day 3 Categorical variables were coded as: history of hypertension (Yes = 1, No = 0). Negative β coefficients for hypertension therefore indicate lower HBL in patients with hypertension relative to those without, after adjustment for other predictors HBL Hidden blood loss, BMD Bone mineral density, Hb Hemoglobin, HCT Hematocrit, BMI Body mass index, PT Prothrombin time, Fbg Fibrinogen p values are two-tailed Table 7Multiple linear regression predicting hidden blood loss on postoperative day 7Predictorβ95% CIp valueAge-0.66-1.987, 0.6620.324BMD T score-17.14-23.05, -11.24< 0.001Preoperative 1 day Hb-1.34-2.719, 0.0360.056Preoperative 1 day HCT106.1-336.5, 548.70.636Postoperative 7 day HCT42.87-389.3, 475.00.845History of hypertension-23.22-45.91, -0.5360.045Categorical variables were coded as: history of hypertension (Yes = 1, No = 0). Negative β coefficients for hypertension therefore indicate lower HBL in patients with hypertension relative to those without, after adjustment for other predictorsHBL Hidden blood loss, BMD Bone mineral density, Hb Hemoglobin, HCT Hematocrit, BMI Body mass index, PT Prothrombin time, Fbg Fibrinogenp values are two-tailed Table 7Multiple linear regression predicting hidden blood loss on postoperative day 7Predictorβ95% CIp valueAge-0.66-1.987, 0.6620.324BMD T score-17.14-23.05, -11.24< 0.001Preoperative 1 day Hb-1.34-2.719, 0.0360.056Preoperative 1 day HCT106.1-336.5, 548.70.636Postoperative 7 day HCT42.87-389.3, 475.00.845History of hypertension-23.22-45.91, -0.5360.045Categorical variables were coded as: history of hypertension (Yes = 1, No = 0). Negative β coefficients for hypertension therefore indicate lower HBL in patients with hypertension relative to those without, after adjustment for other predictorsHBL Hidden blood loss, BMD Bone mineral density, Hb Hemoglobin, HCT Hematocrit, BMI Body mass index, PT Prothrombin time, Fbg Fibrinogenp values are two-tailed Multiple linear regression predicting hidden blood loss on postoperative day 7 Categorical variables were coded as: history of hypertension (Yes = 1, No = 0). Negative β coefficients for hypertension therefore indicate lower HBL in patients with hypertension relative to those without, after adjustment for other predictors HBL Hidden blood loss, BMD Bone mineral density, Hb Hemoglobin, HCT Hematocrit, BMI Body mass index, PT Prothrombin time, Fbg Fibrinogen p values are two-tailed Discussion The aim of this study is to evaluate the association between BMD and postoperative HBL in older patients with unstable intertrochanteric fractures and to inform perioperative management. The results indicate that BMD T-score, preoperative hemoglobin level, and a history of hypertension showed statistically significant associations with HBL in multivariable analysis. Specifically, lower BMD and lower preoperative Hb were moderately correlated with higher HBL (Spearman’s ρ ≈ -0.50, explaining approximately 25% of HBL variance), while a history of hypertension was independently associated with lower HBL after adjustment (β = -24.56 at POD3; β = -23.22 at POD7). No significant associations were observed between other medical histories or medication use and HBL, though statistical power for detecting such associations was limited by small subgroup sizes. These findings suggest that BMD and preoperative Hb may be relevant considerations for perioperative risk assessment; however, the moderate magnitude of the observed associations and the limitations inherent to the retrospective design call for cautious clinical interpretation. The results of the current study shows that low BMD is moderately associated with higher HBL (Spearman’s ρ ≈ -0.50, reflecting a moderate association that explains approximately 25% of HBL variance), which is consistent with recent literature [12]. Increased intraoperative blood loss in patients with osteoporosis may be related to increased surgical complexity and operation time caused by bone fragility [19, 20]. These studies support the findings of this study, suggesting that osteoporosis may affect HBL by increasing surgical difficulty or bleeding risk. In addition, osteoporosis may delay fracture healing, prolong postoperative recovery time, and even increase the risk of readmission and death [21]. Preoperative BMD assessment may help identify patients who require closer perioperative monitoring, and calcium supplementation, vitamin D, or anti-osteoporosis therapy may be considered as part of comprehensive perioperative and secondary-fracture prevention strategies. However, whether these interventions can reduce HBL or improve postoperative outcomes requires prospective validation. It should be noted that this moderate level of correlation should not be overinterpreted as evidence of a strong or causal relationship; other perioperative factors likely contribute to the remaining variance in HBL, and the group-level differences across BMD categories—while statistically significant overall—must be viewed in the context of the markedly unequal subgroup sizes and the limited power of pairwise comparisons involving the small normal-BMD subgroup (n = 21). Although the absolute difference in HBL across BMD categories was modest (maximum approximately 50–80 mL, as illustrated in Fig. 2A and B), we consider this finding clinically relevant in the context of older adults with hip fractures who have limited physiologic reserve. Even a relatively small additional occult blood loss may contribute to postoperative anemia when combined with other perioperative factors, which may prompt closer hemoglobin monitoring, earlier anemia correction, and more cautious perioperative management in patients with low BMD. Nevertheless, we acknowledge that this magnitude of difference alone may not invariably change transfusion decisions, and its clinical impact should be interpreted alongside overall blood loss, baseline hemoglobin, symptoms, and transfusion thresholds. To better contextualize this magnitude of difference against established clinical decision-making thresholds, contemporary geriatric hip-fracture management typically employs restrictive transfusion strategies at a hemoglobin threshold of 70–80 g/L, with symptom-driven transfusion often considered at Hb < 90–100 g/L in patients with cardiovascular comorbidities. In our cohort, the mean preoperative Hb was approximately 110 g/L (Table 1), so that an additional ~ 80 mL of occult blood loss—corresponding to roughly 5–10 g/L of Hb reduction in a typical 4–5 L blood volume—could plausibly shift selected patients with already-low preoperative Hb (e.g., Hb < 100 g/L) into the transfusion-decision zone or trigger earlier iron supplementation or erythropoietin therapy. Accordingly, the observed 50–80 mL differential between osteoporotic and non-osteoporotic patients is unlikely, in isolation, to change transfusion thresholds, but becomes clinically meaningful when combined with low preoperative Hb, symptomatic anemia, or comorbidities that limit oxygen delivery. This supports preoperative BMD assessment and Hb optimization in patients with low T-scores as part of routine perioperative planning. We also found a negative correlation between preoperative Hb levels and HBL, particularly significant on postoperative day 3 (p = 0.033) and approaching significant levels on postoperative day 7 (p = 0.056). This suggests that preoperative Hb levels may be a relevant factor for anticipating HBL. Low Hb levels may reflect preoperative anemia status, weaken patients’ tolerance to intraoperative and postoperative blood loss, thereby increasing HBL. Previous studies have shown in orthopedic surgery that low preoperative Hb levels are associated with increased postoperative blood loss and transfusion needs, especially in older patients who have lower tolerance to blood loss [22, 23]. Another meta-analysis further indicates that preoperative anemia is an important risk factor for blood loss [24]. The results of this study are consistent with this association, underscoring the potential value of preoperative Hb optimization. In clinical practice, optimizing Hb levels before surgery (such as through iron supplementation or erythropoietin therapy) may represent one component of a comprehensive perioperative management strategy, though prospective studies are needed to confirm its impact on HBL and postoperative outcomes. In our geriatric cohort (mean age > 80 years) with multiple comorbidities, multidisciplinary perioperative management commonly required approximately one week of hospitalization; however, as acknowledged in the Limitations section, postoperative day-7 hemoglobin measurements were unavailable for 34.8% of patients, which represents a limitation of the current study. We found that a history of hypertension was independently associated with lower HBL after multivariable adjustment (β = -24.56, p = 0.016 at POD3; β = -23.22, p = 0.045 at POD7), although the unadjusted comparison did not reach statistical significance (Table 3). The relationship between hypertension and perioperative blood loss is likely complex and may be influenced by antihypertensive medications, intraoperative blood pressure management, and vascular changes associated with chronic hypertension [25, 26]. Previous studies have reported conflicting findings: some have suggested that hypertension is associated with increased postoperative complications [9], while others have reported reduced operative blood loss in hypertensive patients undergoing arthroplasty, possibly related to better vascular control or the effects of antihypertensive therapy [25]. Given the retrospective design and the lack of standardized data on antihypertensive medication classes, perioperative blood pressure control, and hemodynamic management in the current study, the observed association between hypertension and HBL should be interpreted cautiously and considered an exploratory finding. Further research incorporating detailed antihypertensive medication and perioperative hemodynamic data is needed to clarify the mechanisms underlying this association. This study did not find a significant correlation between other disease history (such as diabetes, hyperlipidemia), medication history (such as anticoagulants) and preoperative coagulation function indicators and HBL. This result is consistent with some literature, indicating that routine coagulation function indicators may not be sufficient to predict HBL in older patients with fractures. Previous studies have found no significant correlation between preoperative APTT and transfusion requirements following major surgery [27]. However, considering the retrospective design and sample size limitations of this study, the potential role of these factors cannot be completely ruled out, and further validation is needed in future research. Notably, some comorbidity and medication subgroups were small, and patients receiving anticoagulant or antiplatelet therapy within two weeks prior to surgery were excluded; therefore, the statistical power to detect associations between diabetes, hyperlipidemia, anticoagulant-related factors, preoperative coagulation indicators, and HBL may have been limited. Similarly, the observed association between hypertension and lower HBL after adjustment should be interpreted cautiously, as it may partly reflect residual confounding given the lack of standardized data on antihypertensive medication classes and perioperative hemodynamic management, and the discrepancy between unadjusted and adjusted results warrants further investigation in larger cohorts. Larger multicenter studies incorporating standardized comorbidity/frailty measures are warranted to confirm these relationships. The high prevalence of perioperative antibiotic and analgesic use in this cohort reflects our routine prophylactic antibiotic policy and standardized perioperative pain-management practice for geriatric hip-fracture surgery, rather than selective prescribing. This study has several limitations. First, the retrospective design may lead to missing data or selection bias, affecting the robustness of the results. Second, the sample size is modest; the 115 patients included may not be sufficient to detect weak associations between certain variables and HBL. Third, some error in HBL estimation may exist because HBL was calculated using Nadler’s formula and changes in hematocrit (HCT), which may be affected by measurement variability and perioperative fluid shifts. Fourth, several potentially important perioperative factors were not captured or were unavailable in a structured form and thus were not included in the analyses, including operative time, incision length, degree of soft-tissue compromise, and intraoperative hemodynamic management (e.g., blood pressure control); accordingly, residual confounding cannot be excluded. All procedures were performed by a single experienced attending chief orthopedic surgeon, which reduces inter-surgeon variability but may limit generalizability to settings with different surgeon skill levels or operative techniques. To further validate and expand the results of this study, prospective studies with standardized data collection are needed. Randomized controlled trials may better control confounding and clarify the relationships among BMD, preoperative Hb, hypertension, and HBL. Furthermore, mechanistic studies are warranted to explore the physiological pathways by which osteoporosis and hypertension may influence perioperative blood loss. In particular, hypertension-related factors (e.g., baseline blood pressure level, perioperative variability, and antihypertensive medication classes) were not standardized, which may partly explain the discrepancy between unadjusted comparisons and adjusted associations. A further limitation is that we did not include a validated general health or frailty score (such as ASA class, NHFS, or CCI), which may limit characterization of the cohort’s overall physiologic reserve and comparability with other geriatric hip-fracture studies. Fifth, a substantial proportion of patients (40/115, 34.8%) lacked complete postoperative day-7 hemoglobin/hematocrit measurements and were therefore excluded from POD7 HBL calculations, raising the possibility of selection or attrition bias in the POD7 analyses. To evaluate this, we compared baseline characteristics between patients with and without POD7 measurements (Fig. 1); no significant differences were observed in age (p = 0.263), BMI (p = 0.980), BMD T-score (p = 0.426), preoperative hemoglobin (p = 0.102), sex distribution (p = 1.000), or hypertension prevalence (p = 0.695), supporting a missing-at-random (MAR) assumption with respect to the key clinical covariates examined. Nevertheless, residual bias arising from unmeasured factors (e.g., early discharge prior to POD7, postoperative complications, or transfer to another facility) cannot be entirely excluded, and the smaller POD7 sample (n = 75) carries inherently lower precision than the full POD3 cohort (n = 115); this is reflected in the wider confidence intervals reported for POD7 estimates. Multiple imputation was not performed because the primary Spearman correlation analyses are non-parametric and were considered robust to complete-case restriction under MAR; however, the consistency of the BMD–HBL association across both time points (ρ = -0.492 at POD3 vs. ρ = -0.513 at POD7) argues against substantial bias from the POD7 missingness. Prospective studies with standardized POD7 follow-up would further mitigate this limitation. Importantly, subgroup sizes were markedly unequal (osteoporosis, n = 70; osteopenia, n = 24; normal BMD, n = 21), and the normal-BMD group was relatively small; this imbalance substantially reduces the statistical power for detecting differences in pairwise comparisons involving the smaller subgroups and increases the risk of type II error. Accordingly, the observed group differences in HBL across BMD categories (Fig. 2A and B) should be interpreted with caution, and the moderate Spearman correlations (ρ ≈ -0.50) should not be overinterpreted as evidence of a strong or deterministic relationship. No formal a priori power calculation was performed for this retrospective study. As a post hoc contextualization, the achieved statistical power for the primary Spearman correlation analysis (BMD T-score vs. HBL) was greater than 0.999 for both postoperative day 3 (n = 115, ρ = -0.492) and day 7 (n = 75, ρ = -0.513), indicating that the primary analyses were well-powered to detect the observed associations. At these sample sizes, the minimum detectable correlation at 80% power was |ρ| = 0.26 (day 3) and |ρ| = 0.32 (day 7). However, the normal-BMD subgroup (n = 21) was notably small; the minimum detectable correlation for this subgroup at 80% power is |ρ| = 0.58, indicating that subgroup-level analyses involving this stratum were substantially underpowered and should be interpreted with caution. Future prospective studies should include pre-specified sample size calculations based on the effect sizes reported here to ensure adequate statistical power for both overall and subgroup comparisons. Larger multicenter studies with more balanced BMD-group representation are warranted to confirm these findings. Conclusion In this retrospective cohort study of older patients undergoing intramedullary nailing for unstable intertrochanteric fractures, lower BMD T-score and lower preoperative Hb were independently associated with greater perioperative HBL, while a history of hypertension appeared to be independently associated with lower HBL after multivariable adjustment. These findings suggest that preoperative BMD assessment and Hb optimization may be considered in perioperative risk stratification for geriatric hip-fracture patients. However, given the study’s retrospective design, modest overall sample size, markedly imbalanced BMD subgroups, and missing postoperative day-7 data in 34.8% of patients, these results should be interpreted with caution and do not establish causality. The modest magnitude of the observed differences in HBL across BMD categories (approximately 50–80 mL) suggests that clinical relevance is greatest when interpreted alongside low baseline Hb, symptomatic anemia, or cardiovascular comorbidities. Prospective, multicenter studies with pre-specified sample-size calculations and balanced BMD-group enrollment are warranted to confirm these associations and to evaluate whether incorporating BMD into perioperative risk assessment can improve clinical decision-making and patient outcomes.