Abdominal aortic calcification is associated with hip fracture in the elderly

Abstract

Hip fracture, the most serious fragility fracture in elderly osteoporotic patients, is critical to address as it decreases levels of activities of daily living, worsens life expectancy and increases risk of requiring long-term care. Many patients with hip fracture are unaware of their risk at the time of fracture and therefore do not use osteoporosis medications. Thus, a tool is needed to raise patient awareness of hip fracture risk. Here we enrolled 1395 hip fracture cases and 1075 non-hip fracture controls and evaluated their respective abdominal aortic calcification semiquantitatively using lumbar spine lateral X-ray images. That analysis revealed that the degree of abdominal aorta calcification (AAC) seen on lateral lumbar spine X-ray images, as indicated by a score of 5, was significantly and positively correlated with hip fracture occurrence. In particular, we show that calcification of the wall of the abdominal aorta at the level of the fourth lumbar vertebra correlated with hip fracture development. Lower back pain is a common complaint among the elderly, often prompting patients to undergo lumbar spine radiography. Thus, AAC score could serve as a potential surrogate marker to identify patients at potential risk of hip fracture as a means of encouraging them to take osteoporosis medications.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-53150-x.

Introduction

In Japan, the proportion of individuals aged 65 years or older is the highest in the world, and the incidence of hip fractures continues to increase with population aging¹. Although most clinical guidelines recommend osteoporosis treatment, including bisphosphonates, to prevent fragility fractures, previous study have shown that the majority of patients who sustain hip fractures are not receiving osteoporosis treatment at the time of injury². This suggests that many patients are unaware of their fracture risk until the fracture occurs.

To prevent subsequent fragility fractures, it is essential to establish a system that enables physicians managing conditions other than osteoporosis, such as spinal disorder, to recognize patients at risk of hip fracture and to recommend appropriate anti-osteoporosis therapy. Ideally, such a system should be simple impose minimal burden on physicians.

Several fracture risk assessment tools, including FRAX (fracture risk assessment tool), are currently available³. Although FRAX provides a convenient, questionnaire-based method for estimating fracture risk, its predictive accuracy declines with advancing age³. Given that most hip fractures occur in individuals older than 80⁴, there is a need for a more effective risk assessment tool specifically tailored to the elderly population. Recently, a new questionnaire-based tool, the Kumamoto-STudy for Osteoporotic-hip fracture Prevention (K-STOP) scoring system, has been developed to assess hip fracture risk in older adults, with or without measurements of serum 25(OH)D levels and bone mineral density⁴. However, this tool does not incorporate lumbar spine radiographic findings^5–7.

An association between osteoporosis and ectopic calcification has been reported previously^8–11. Abdominal aortic calcification (AAC), a form of vascular calcification, can be evaluated using Kauppila’s semiquantitative visual score based on lateral lumbar spine radiographs¹². Because lower back pain is highly prevalent among older adults^13,14, lumbar spine X-ray images are frequently obtained in clinical practice, providing an opportunity to assess hip fracture risk during routine evaluation. Therefore, a simple radiographic assessment of the lumbar spine may serve as a practical tool for identifying patients at risk. Although several studies gave reported an association between AAC, as assessed by Kauppila’s score, and fracture risk^37–40, data in the Japanese population remain limited.

In the current study, we show that abdominal aortic calcification is significantly associated with the occurrence of hip fracture in elderly patients. We analyzed 1395 patients with hip fracture and 1075 control subjects without hip fracture, and semi-quantitatively assessed AAC using lateral lumbar spine X-ray images. Calcification of the anterior and posterior walls of the abdominal aorta adjacent to each vertebral body was semi-quantitatively evaluated from the 1st to the 4th lumbar vertebrae. Each spinal segment was scored up to 6 points, yielding a total possible score of 24 points. Using this scoring system, we found that an AAC score of 5 or higher was significantly associated with hip fracture. In particular, calcification of the wall of the abdominal aorta anterior to the fourth lumbar vertebra was associated with the occurrence of hip fracture (Fig. 1).

Fig. 1.

A semi-quantitative Abdominal Aortic Calcification (AAC) score. AAC score was assessed by Kauppila’s semiquantitative score from lateral radiography of lumbar vertebrae. Briefly, on the X-ray image, calcification of the anterior and posterior walls of the abdominal aorta running anterior to each vertebral body was evaluated semi-quantitatively, with 1 point awarded for every 1/3 of the length of each vertebral body in either the anterior or posterior wall. Thus, if calcification was found along the entire length of both the anterior and posterior walls of the abdominal aorta anterior to a single vertebral body, the score was 6. Circled numbers indicate scores for each site. In this image, the total score is 13.

Subjects and methods

Study design and participants

The Kumamoto STudy for Osteoporotic-hip fracture Prevention (K-STOP) group was established to assess factors associated with hip fractures and develop a way to prevent their occurrence. Twenty-three hospitals associated with the Department of Orthopaedic Surgery at Kumamoto University participated in K-STOP. Of the 23 hospitals, which treat a total of approximately 1.8 million people, 22 were located in municipalities in Kumamoto Prefecture, and 1 in Omuta City, Fukuoka Prefecture.

To enroll case group participants (group 1), hip fracture patients and their families were informed of the study and recruited by the attending physician on a hospital visit. Although the enrollment start date varied, at all sites participants were enrolled for a one-year period between March 2021 and December 2022. We then excluded patients who: (1) were under 20 years of age, (2) refused to participate, (3) had pathological fractures, or (4) had fractures due to occupational or traffic accidents. Consecutive hip fracture patients were invited to participate. Among patients who met inclusion criteria and did not meet exclusion criteria, none who were invited to participate in the study refused to enroll. Patients who agreed to participate were asked to complete questionnaires and undergo examinations, including X-rays, for data collection. Overall, a total of 1395 hip fracture patients were enrolled in group 1. Then, 40 were excluded from the case group due to incomplete data, leaving 1355 hip fracture patients (Fig. 2). We excluded from the analysis subjects in which AAC assessment was difficult due to implants, as well as individuals in which the imaging field was insufficient for AAC assessment. There were no missing values for age, sex, or BMI, and all cases could be evaluated. No imputation was used in the regression analysis.

Fig. 2.

Flowchart showing selection of cases and controls. We enrolled 1395 hip fracture patients (group 1) plus 2 groups of non-fracture controls: 338 subjects for group 2 and 737 for group 3. Note that 40 individuals for group 1 and 2 for group 2 were eliminated based on exclusion criteria. Also, 682 and 376 in respective hip fracture and control groups were excluded after subjects were matched for age, gender and BMI. Final hip fracture and control groups included 673 participants each.

For the control group, we combined two subgroups to reduce selection bias. One (group 2) consisted of in- or out-patients who were visiting one of the 23 hospitals described above for conditions other than hip fracture where they were recruited by attending physicians. The other (group 3) consisted of older community-dwelling residents in Arao City, Kumamoto Prefecture who had participated in the 2nd survey of the Japan Prospective Studies Collaboration for Aging and Dementia (JPSC-AD) between October 1, 2022, and December 18, 2022. JPSC-AD is a multisite, population-based prospective cohort study of dementia. In both subgroups 2 and 3, subjects agreeing to participate were enrolled and administered questionnaires and examinations comparable to case group subjects and excluded if they (1) refused to participate (2) or undergo X-ray, (3) had a history of hip fracture, or (4) were under 65 years of age. For group 2, we enrolled a total of 338 individuals, and for group 3, we enrolled 737 subjects. Group 3 subjects were informed of the study by an examiner, recruited, and, if they agreed to participate, enrolled. They then responded to questionnaires and underwent examinations, as described above. For group 3, we enrolled a total of 713 individuals using exclusion criteria comparable to those used in group 2.

After combining groups 2 and 3, 1049 control subjects remained. Comparison of age, sex and BMI revealed significant differences in these three parameters between groups 2 and 3 (Supplementary Table 1). Finally, control group subjects were matched 1:1 using propensity score matching to 673 cases based on age (± 2 years), gender and BMI to eliminate these differences (Table 3).

Table 3.

Matched age, sex and BMI characteristics of cases and controls.

Characteristics	Controls (N = 673)	Cases (N = 673)	p-value
Male	27.0 (182)	26.4 (178)	0.853
Female	73.0 (491)	73.6 (495)
Age, years (mean (SD))	80.5 (6.5)	80.2 (9.3)	0.365
BMI (kg/m2) (mean (SD))	22.3 (3.4)	22.1 (3.6)	0.343
Diabetes % (N)
No	86.8 (223)	84.9 (570)	0.533
Yes	13.2 (34)	15.1 (101)
History of CVD % (N)
No	90.0 (606)	91.3 (606)	0.454
Yes	10.0 (67)	8.7 (58)
eGFR (ml/min/1.73㎡) ( (mean (SD))	61.0 (19.2)	67.0 (25.4)	0.001
Hypertension % (N)
No	34.3 (87)	48.4 (324)	< 0.001
Yes	65.7 (167)	51.6 (346)
Smoking habits % (N)
No	97.3 (250)	95.8 (643)	0.342
Yes	2.7 (7)	4.2 (28)
Statin use % (N)
No	66.5 (422)	76.7 (516)	< 0.001
Yes	33.5 (213)	23.3 (157)
Osteoporosis medications % (N)
No	70.4 (474)	82.8 (557)	< 0.001
Yes	29.6 (199)	17.2 (116)

Control: non-hip fracture, Case: hip fracture.

BMI = body mass index.

The study was first approved by the Institutional Review Board of Kumamoto University and then by the respective research ethics committees. All methods were performed in accordance with the relevant guidelines and regulations. Informed consent was obtained from all participants, including JPSC-AD participants, after providing sufficient information about the study and protocol prior to their participation. If a participant was unable to consent for any reason (e.g., severe dementia), consent was obtained from a family member or by proxy.

Assessment of abdominal aortic calcification by AAC score

Calcification of the abdominal aorta was assessed using Kauppila’s visual semiquantitative score by evaluating spinal X-ray lateral images at the 1st to 4th lumbar vertebrae¹². Briefly, in either the anterior or posterior walls of the abdominal aorta, the presence of ectopic calcification over 1/3 of the length of the vertebral body was scored as an AAC score 1, and the presence of ectopic calcification over the entire length of each vertebrae in both the anterior and posterior walls was scored as 6. Thus, the maximum score possible following assessment of four lumbar vertebrae was 24. To assess distribution and trends in ectopic calcification, the score for each segment (anterior or posterior wall, L1, 2, 3 or 4 vertebrae) was assessed separately. Lateral lumbar radiographs of fracture cases were taken approximately 7 days postoperatively in the lateral decubitus position. Cases in which the AAC was difficult to identify due to the presence of implants or other factors, or in which the field of view was insufficient for AAC assessment, were excluded from the evaluation. As noted above, X-ray images were obtained from all the subjects in both the case group and the control group after individuals were accepted into the study regardless of whether they had low back pain or not, and evaluated by radiologists using standardized protocols.

AAC score reproducibility was analyzed as a log-transformed continuous variable using the intraclass correlation coefficient (ICC). Inter- and intra-rater reliabilities were determined as described¹². Briefly, to evaluate inter-rater reliability, two orthopedic specialists certified by the Japanese Orthopaedic Association (one of whom was a spine specialist certified by The Japanese Society for Spine Surgery and Related Research) first evaluated AAC using a set of 30 randomly selected “baseline” films individually and determined the correlation for each AAC score. These same two specialists then performed comparable analysis of another 30 “follow-up” films individually. Thus, a total of 60 films were assessed by both specialists to evaluate the AAC. Inter-reader ICC was 0.94 (95% CI 0.81–0.88) (Table 1). To evaluate intra-rater reliability, AAC was first evaluated in 30 “baseline” films by one specialist (the one described above who was also a spine specialist) and then a second AAC evaluation was performed by the same specialist on the same baseline films and scores were correlated. Intra-reader ICC was 0.97 (95% confidence interval, 95% CI 0.95–0.98) (Table 1). Thus, the remaining X-rays were assessed by the orthopedic specialist described above who was also a spine specialist.

Table 1.

Intraclass correlations of intra- and inter-rater agreement of aortic calcifications.

	Inter-rater baseline films	Follow-up films	Intra-rater baseline films
Affected segments score	0.66	0.87	0.97
Anterior or posterior affected segments score	0.82	0.92	0.98
Antero-posterior score	0.71	0.94	0.97

Inter-rater reliability: Two examiners, both orthopedic specialists certified by the Japanese Orthopaedic Association and one a spine specialist certified by The Japanese Society for Spine Surgery and Related Research, evaluated two different sets (Baseline and Follow-up) of 30 films per set.

Intra-rater reproducibility: A set of 30 randomly selected images was evaluated a second time by the orthopedic specialist described above who was also a certified spine specialist.

Data collection

All participants underwent questioning and were examined. Group 1 participants were interviewed and tested during hospitalization, Group 2 participants were treated during hospitalization or as out-patients, and Group 3 participants were interviewed and examined in the second JPSC-AD survey.

The first part of the questionnaire included self-reported answers to questions administered by medical staff at each site. Proxy respondents, such as family members or other caregivers, responded only if the subject had other health problems, including cognitive or visual impairment. The second part of the questionnaire was administered by medical staff at each site based on subjects’ medical records and examinations, including use of statins. That questionnaire included height and weight and BMI, the latter as weight divided by height squared (kg/m²). Previous hip fractures were defined as bilateral hip fractures detected within the case group on radiographs of bilateral hips in an anteroposterior view. Radiographs of the bilateral hips were also reviewed in the control group (Table 2).

Table 2.

Participant characteristics.

Characteristics	Controls (N = 1049)	Cases (N = 1355)	p-value
Sex % (N)
Male	33.2 (348)	20.4 (276)	< 0.001
Female	66.8 (701)	79.6 (1079)
Age, years (mean (SD))	79.2 (6.3)	85.0 (8.9)	< 0.001
BMI (kg/m2) (mean (SD))	23.1 (3.4)	20.6 (3.4)	< 0.001

Control: non-hip fracture, Case: hip fracture.

BMI = body mass index.

Statistical analysis

For all characteristics, univariate comparisons were undertaken between case and control groups using a chi-squared test for categorical and the Student’s t-test for continuous variables. A p value < 0.05 was considered significant. Data assembly and analysis were performed in R version 4.2.1.

Characteristics at the baseline exam were summarized by AAC value as mean ± SD (%).

An association between AAC values and hip fracture was examined using logistic regression in 673 matched pairs as a case–control study, and findings were presented as odds ratios with their 95% confidence intervals (OR, 95% CI). For all hip fracture outcomes, AAC was analyzed as a continuous variable (1 unit increase). Multivariable models included potential confounder factors (age, BMI and gender) of associations between AAC and hip fracture. All associations were examined with each value of AAC to determine whether the results were robust to higher AAC values. Then, the AUC value was calculated at each step to determine the cut-off value. The most discriminating threshold was chosen using the Youden’s index.

Results

Patient enrollment and basic characteristics of cases and controls

We enrolled 1395 hip fracture patients and 1075 non-hip fracture controls in this study as previously reported⁴. The proportion of women was significantly higher in the hip fracture case group (79.6% in cases, 66.8% in controls, p < 0.001) (Table 2). The mean age of the case group (85.0 ± 8.9 years) was significantly higher (p < 0.001) than that of the control group (79.2 ± 6.2 years) (p < 0.001) (Table 2). BMI was significantly higher (p < 0.001) in the control (23.1) versus case (20.6) group (Table 2).

After matching by sex, age and BMI, 673 subjects remained in both groups. Note that after matching, sex, age, and BMI did not differ significantly between control and case groups, nor did diabetes, a history of CVD, hypertension, or smoking habits. By contrast, when renal function was evaluated by eGFR, the case group exhibited a significantly better eGFR value. The percent statin use by subjects was significantly higher in the case versus control group.

Here, prior to matching, AAC was significantly higher as eGFR decreased (correlation coefficient − 0.145, p value 9.21 × 10^–13); however, after matching, both groups exhibited comparable normal eGFR values of 60 ml/min/1.73m² or higher (case 67.0 vs. control 61.0 ml/min/1.73m²). Relevant to hypertension, AAC was significantly higher in the case group with current or past hypertension than in the control group (AAC; current/past hypertension 4.80 vs control 3.49, p value 1.49 × 10^–13). However, post-matching comparisons showed that the proportion of subjects with current or past hypertension was comparable in case and control groups (case 65.3% vs. control 66.1%). Statin use was lower in the case versus control group (case 23.3% vs. control 33.5%), but AAC was significantly higher in the statin group (treatment 3.58 vs. no treatment 2.77).

After matching, the proportion of subjects with a history of CVD or diabetes did not differ between case and control groups (in the former, case 8.7% vs. control 10.0%; in the latter, 15.1% vs. control 13.2%). There was no significant difference in smoking habits after matching (case 4.2% vs. control 2.7%). By contrast, although there was no difference in the proportion of subjects receiving anti-osteoporosis medications between case and control groups prior to matching, after matching the proportion of patients already treated with anti-osteoporosis medications was higher in the fracture group after matching (case 17.2% vs. control 29.6%, p value < 0.001) (Table 3).

Abdominal aortic calcification (AAC) is significantly associated with hip fracture occurrence

We calculated AAC scores for L1 to L4 semi-quantitatively using lateral X-ray images of the lumbar spine in both hip fracture patients and non-hip fracture controls, and then correlated AAC score with occurrence of hip fracture (Table 4). Note that the lowest score possible was 0 and the highest (a summation of L1 to L4 scores) was 24. We observed a significant correlation between AAC score and occurrence of hip fracture, with an odds ratio (OR) of 1.47 (95% CI 1.13–1.91), p value 0.00333 for an AAC score of 5 or higher and an OR of 1.59 (95% CI 1.19–2.12), p value 0.00178 for an AAC score of 6 or higher after adjustment for age, gender and BMI. When AAC was analyzed as a continuous variable, we still observed a significant correlation between AAC score and occurrence of hip fracture, with an OR of 1.040 (95% CI 1.00–1.07) and p value 0.0398.

Table 4.

Odds ratio (OR) analysis of association of abdominal aortic calcification score (AAC) with hip fracture occurrence.

	OR	(95% CI)	p-value
AAC ≥ 5	1.47	(1.13–1.91)	0.00333
AAC ≥ 6	1.59	(1.19–2.12)	0.00178

Data were adjusted for age, sex and BMI.

When we evaluated the percentage of individuals achieving a particular AAC score in both hip fracture and control groups, we observed a greater percentage of control over hip fracture subjects scoring from 0 to 4 (Fig. 3) but this trend reversed at an AAC score of 5, at which point we consistently observed a greater percentage of hip fracture patients versus non-hip fracture controls (Fig. 3). Based on this, we considered the AAC ≥ 5 is a suitable score for predicting a clinical association with hip fractures. We therefore performed a ROC analysis. ROC analysis indicated an area under the curve of 0.664 (95% CI 0.642–0.685) at a cut-off value of 5 and a sensitivity of 0.482 and specificity 0.783 (Fig. 4).

Fig. 3.

Percentage of AAC scores in control and hip fracture groups. Shown is the percentage of cases showing the AAC score indicated on the x-axis, subdivided based on hip fracture and control groups. Note that as AAC score increases, the percentage awarded that score in the control group decreases, while the hip fracture group showed a multimodal pattern, with the least difference at score 3 and the greatest difference at score 5. Also, a larger number of scores of 5 or higher were seen in the hip fracture group.

Fig. 4.

Receiver operating characteristic (ROC) analysis and optimal cutoff score for hip fracture occurrence. Using AAC score, we performed ROC analysis to obtain the optimal cutoff score. The optimal cutoff score was 5, with a specificity of 0.783, sensitivity of 0.482, and an AUC of 0.664.

The AAC score at the 4th lumbar vertebrae correlates best with hip fracture occurrence

We then evaluated lumbar segments separately by subdividing all AAC scores according to L1-L4 levels of the lumbar spine (Fig. 5). At the L1 and L2 levels, the mean AAC score in the case group was less than 1, thus, there were many cases with no abdominal aortic calcification at all, therefore, these were excluded from the analysis. At the L3 level, the mean AAC score in the case group was 1.44, exceeding 1, but there was no significant difference in AAC scores between the case group and the control group. In contract, the highest mean score, 1.96, was seen in L4 in the hip fracture group. The difference in mean AAC score between hip fracture patients and non-hip fracture controls was also greatest at L4 (Fig. 5). Moreover, evaluation of L4 AAC scores revealed that a score of 4 or more was significantly correlated with occurrence of hip fracture with an OR of 1.50 (95% CI 1.02–2.22), p value 0.0404 (Table 5, Fig. 6). These results suggest that it may be possible to assess the risk of hip fracture based solely on the AAC score for L4, without the need to evaluate all levels from L1 to L4.

Fig. 5.

Mean AAC score in each lumbar segment (L1-4). Total AAC scores were subdivided based on lumbar vertebral levels. The lower the level evaluated, the greater the calcification, with scores above 1 at L3 or 4 levels in the hip fracture group (mean score at L4 was 1.96).

Table 5.

Odds ratio (OR) of hip fracture occurrence with abdominal aortic calcification score at lumbar level 4.

	OR	(95% CI)	p-value
AAC (L4) ≥ 4	1.50	(1.020–2.22)	0.0404

Date was adjusted for age, sex and BMI.

Fig. 6.

AAC score frequencies in L4 in control and hip fracture groups. Shown is the frequency of individual AAC scores in both control and hip fracture groups at the L4 level. Note that the greatest difference in frequency was seen at a score of 3, and hip fracture case group more frequent than control group for scores of 3 or more points.

Discussion

Hip fracture is the most serious fragility fracture seen in the elderly in terms of ADL limitation and worsening life expectancy^1,15. Hip fracture patients are most often treated surgically, which is an important consideration from a health economic perspective¹⁶. Ideally, hip fracture patients should be aware of their risk of fracture prior to a fracture, but that is often not the case, as many patients are not under treatment for osteoporosis at the time of fracture². Low back pain is frequently reported by older adults^13,14, and patients seeking treatment for this condition often undergo lumbar spine X-ray examinations. Based on our observations, we feel that physicians evaluating those X-ray images would have an opportunity to become aware of their patients’ possible risk of hip fracture and inform them about preventative measures. Here, we show that AAC seen anteriorly on a lateral lumbar spine X-ray image could serve as a convenient tool to assess possible hip fracture risk.

Other fracture risk assessment tools for the elderly include the FRAX and the K-STOP scoring system^3,4. Both assess hip fracture probability without bone density and blood or urine testing or X-ray images and can be implemented with a questionnaire. We compared the predictive ability of the AAC score, FRAX, and K-STOP for hip fracture risk in the 84–86-year-old age group, since the mean age in this cohort of hip fracture cases was 85.0 years. The AAC score serving as the cutoff value was ≥ 5, and the cutoff value for K-STOP was demonstrated as ≥ 5⁴. For FRAX, although no international cutoff value is recommended, we used the 15% cutoff value based on Japanese guidelines. As a result, the AAC score, FRAX and K-STOP identified 47.2, 46.1 and 71.4%, respectively, of individuals as being at risk of hip fracture, suggesting that AAC is as effective as FRAX and K-STOP in identifying individuals at risk. However, the time required per case was 10.2, 15.5 and 242.4 s for AAC, FRAX and K-STOP, respectively, indicating that the AAC score takes significantly less time than the other two methods. Since FRAX and K-STOP can be either administered by medical staff or self-administered by patients, they do necessarily not require physician time; however, those patients must assess their own scores, taking into account the possibility that they may be at risk for a hip fracture. However, in our AAC assessment, we found that 93% of patients with hip fractures were not undergoing osteoporosis treatment at the time of fracture and had been unaware of fracture risk. Nonetheless, while patients cannot assess fracture risk by AAC score, it requires only ~ 10 s for a physician and offers that physician has an opportunity to raise awareness of the patient regarding hip fracture prevention. The AAC differs from these risk assessment tools in that the physician is made aware of the possible risk by an unrelated test frequently administered to elderly individuals.

The AAC assessment takes only a few seconds to complete and is not a significant burden for physicians in low back pain practice. The SQ grade for vertebral fracture is known as a semi-quantitative assessment of severity of vertebral fracture but it is not a tool to assess hip fracture risk. To simplify AAC evaluation, we have also found that the presence of AAC at L4 indicates an increased probability for hip fracture.

AAC evaluates arterial calcification and is thus confounded by the presence of conditions that promote such calcification, including renal insufficiency^17–19, as well as hypertension^20–22, statin use^23,24, diabetes^25,26 or smoking habits^27,28. In fact, hypertension and chronic kidney disease are reportedly significantly correlated with development of hip fracture^22,29,30. AAC is known to be associated with cardiovascular disease (CVD)^31–33. Thus, we examined the association between AAC and hip fractures using a model that extended the adjustment beyond age, sex, and BMI to include confounding factors such as eGFR, hypertension, statin use, diabetes, smoking and osteoporosis medication, all potentially confounding the analysis. As noted above, in the analysis after matching with age, sex and BMI, higher AAC score was significantly associated with an increased probability of hip fracture. However, we found that in the fully adjusted model, for an AAC score ≥ 5, the OR was 0.705, the CI was 0.492–1.01, and the p value was 0.0563. Also, in the L4 model without adjustment the AAC score ≥ 4, the OR was 2.91, the CI was 2.21–3.82, and the p value was 2.17e−14, while in the fully adjusted model, the AAC was ≥ 4, the OR was 0.561, the CI was 0.316–0.998, and the p value was 0.049. Thus, the association between AAC and hip fracture was weaker after adjustment for confounding factors, indicating that AAC is unlikely to be an independent determinant of fracture but rather reflects the cumulative burden of underlying conditions such as cardiovascular or metabolic disease. Nonetheless, from a clinical perspective, AAC may still be useful as a potential surrogate marker to identify individuals at higher fracture risk.

Systems used to assess risk of osteoporosis in patients presenting for non-osteoporosis care have been reported, such as chest X-ray or CT images that estimate bone density by artificial intelligence (AI)^34–36. However, tools to predict occurrence of hip fracture are less well known. In addition, AAC is currently difficult to evaluate due to imaging of intestinal gas and fecal stones, which complicates assessment of lumbar spine X-rays. AAC assessment requires a trained expert; however, recently machine-learning assessment of AAC score based on dual-energy X-ray absorptiometry (DXA)-derived lateral spine images indicated an association of high AAC with long-term risk of falls and fractures in community-dwelling older women³⁶.

AAC is also reportedly associated with increased fracture risk, indicating that vascular disease is considered associated with decreased skeletal health. AAC assessment in conjunction with BMD testing has been shown to improve fracture risk prediction³⁷. Szulc et al. reported on the relationship between AAC severity and fracture risk prospectively^38,39; however, those subjects were aged 65 years or older and the mean age was below 75. Our study aimed at enrolling older subjects, with an average age of over 80 years. Lewis et al. suggested that older women (> 75 years) with pronounced AAC are at greater risk of fracture and proposed that vascular calcification and bone pathology may share similar causal mechanisms⁴⁰. Although the AAC score is evaluated on a scale of 0 to 24, we categorized it as AAC ≥ 1, AAC ≥ 2, AAC ≥ 5, and AAC ≥ 9 to assess its association with hip fractures³⁹. A significant correlation between AAC and hip fracture was confirmed for AAC ≥ 5 and AAC ≥ 9, with a higher OR observed for AAC ≥ 9 compared to AAC ≥ 5 (Supplementary Table 2). However, cases with AAC ≥ 5 accounted for 48% of the total, while those with AAC ≥ 9 for 16%. Therefore, given the larger number of cases, AAC ≥ 5 was considered clinically more appropriate for identifying cases with potential hip fracture risk. Again, we note that in older subjects enrolled in our study we observed that an AAC value of 5 was the cutoff for association with hip fracture.

There are several limitations to this study. First, because this is a cross-sectional study, the temporal ordering between AAC and fracture cannot be established. In this study, lumbar spine X-rays were taken at a single time point during hospitalization for hip fracture, suggesting that AAC was unlikely to have been caused by the hip fracture. However, since lumbar spine X-rays were not taken prior to hip fracture, reverse causation cannot be ruled out. Future longitudinal studies will be needed to verify the utility of AAC scores as a probability factor for development of hip fracture. Although we enrolled study cases at multiple centers, the geographical area was limited, and thus it is not known whether our results can be generalized to other areas. However, the hip fracture cases enrolled in this study are similar to those enrolled in the Japanese nationwide hip fracture registry by the Japanese Orthopaedic Association in terms of age and gender⁴¹. In this study, to reduce selection bias, two different control groups (Group 2 and Group 3) were combined and used as the control group, but we performed AAC comparisons between each group and the case group. After matching group 2 (the hospital control group) with the case group (Group 1) based on age, sex, and BMI, we analyzed data for 333 controls in group 2 and 333 subjects in the case group in the overall model for AAC ≥ 5. We found that the OR was 2.080, the CI was 1.500–2.88, and the p value was 0.00000997. Similarly, after matching group 3 (the community control group) with the case group by age, sex, and BMI, and analyzing the overall model with 506 individuals from group 3 and 506 cases from the case group for AAC ≥ 5, the OR was 1.860, the CI was 1.400–2.47, and the p value was 0.0000169. Overall, we conclude that the AAC score defined in this study can be standardized for to assess potential hip fracture risk in the elderly.

Taken together, we propose use of a tool to assess possible risk of hip fracture in the practice of lumbar spine disease. We expect that the present method will serve as a reference for the development of other methods to assess hip fracture risk in the elderly, a population that will likely increase in the future.

Author contributions

Conceptualization: MT, TM (Miyamoto); Date curation: YS, MU, NK; Formal analysis: YS; Funding acquisition: MT, TM (Miyamoto); Investigation: YS; Project administration: MT, TM; Supervision: MT, TM (Miyamoto); Validation: TN, MU, ST (Takata), TK, TM (Masuda), TT, SH, MY, KS, KT, YU, KM (Maeda), NY, HM, ST (Tanimura), MS, MT, KM (Miyamoto), XT, KY; Writing—original draft: YS; Writing—review and editing: TM (Miyamoto)

Data availability

Date will be made available on request.

Declarations

Competing interests

The authors declare no competing interests.