Osteoporotic fractures are a major clinical challenge due to their substantial impact on health and quality of life.1 They are a leading cause of pain, functional decline, and excess mortality—particularly in the case of hip and vertebral fractures. From a healthcare system perspective, their economic burden is considerable, driven by hospitalization, long-term care needs, and loss of productivity. In aging populations, the rising incidence of these fractures represents a growing strain on both clinical services and public health budgets.2

From the age of 50,3 the lifetime risk of sustaining a major osteoporotic fracture (MOF)—including hip, forearm, humerus, and clinical vertebral fractures—is estimated at 46% in women and 22% in men. Accurately predicting who will sustain a fragility fracture is essential to guide preventive strategies and prioritize healthcare interventions. In this context, the development and refinement of reliable fracture risk assessment tools4 are critical to identify high-risk individuals and ensure efficient use of diagnostic and therapeutic resources.

Bone mineral density (BMD), assessed by dual-energy X-ray absorptiometry (DXA), is the most widely used tool for estimating fracture risk, despite its limited sensitivity.5 While BMD shows high specificity, many fragility fractures occur in individuals without densitometric osteoporosis. Incorporating clinical risk factors alongside BMD enhances predictive performance, prompting the development of integrated tools that offer a more accurate assessment of individual fracture risk.

FRAX is an algorithm available on the Internet that calculates the absolute probability of sustaining a MOF or a hip fracture (HF) over the next 10 years.6 Developed in 2008, it combines clinical risk factors with or without BMD to estimate fracture risk. Its main advantage lies in allowing risk stratification even in the absence of BMD measurements, thereby facilitating decision-making in diverse clinical settings.

In Spain, however, its application in daily practice remains limited. While estimates for HF risk (HFR) are considered valid and clinically useful, the algorithm tends to underestimate the probability of MOF in the population.7–9 This limitation has prompted efforts to adapt FRAX-based strategies to local epidemiology, aiming to improve their accuracy and clinical applicability.

In this context, our group previously proposed10 a classification algorithm based on locally adapted MOFR thresholds to improve the clinical applicability of FRAX in Spain. In the present study, we aimed to assess its performance in a real-world clinical setting.

The final study sample consisted of 837 women. Follow-up was incomplete in 10 cases due to relocation outside Catalonia or loss of contact with primary care. Fifty-eight women died during the follow-up period.

At baseline (2008), the mean age was 62±9 years. Eighty percent of participants had at least one fracture risk factor. BMD was normal in 20%, osteopenic in 55%, and osteoporotic in 25%. Detailed baseline characteristics are shown in Table 1.

The mean MOFR was 6.21%±5.39, with a median of 4.23% (IQR 2.56–8.00), and the mean HFR was 2.08%±3.20, with a median of 0.90% (IQR 0.40–2.40). Both estimates were lower when BMD was included in the FRAX calculation.

More than 80% of women received calcium and/or vitamin D supplementation, and 25% (n: 209) received BA. Among the latter, 91% were treated with bisphosphonates, 14% with selective estrogen receptor modulators, 14% with denosumab, 11% with strontium ranelate, and 5% with teriparatide.

A total of 251 fractures were identified in 207 women: 169 due to fragility, 80 traumatic, and 2 pathological (Table 2). Thirty-five women experienced two fractures, eight had three, and one patient had four, including a hip fracture as the final event. Among the pathological cases, one woman developed jaw osteonecrosis following radiotherapy for a tongue-base neoplasm (rheumatoid arthritis, low-dose glucocorticoids, methotrexate, leflunomide, no BA in prior 5 years); another sustained a pelvic fracture during the course of multiple myeloma.

A hundred and forty-four women experienced a MOF (66 vertebral, 39 distal forearm, 10 hip, 9 humerus). The median time to first MOF was 64±36 months (range 0–119). Notably, 64% of these women did not meet densitometric criteria for osteoporosis. Based on baseline MOFR values from 2008, the expected number of MOF was 52 (or 44 when BMD was included in the calculation), underscoring the underestimation by FRAX.

Women who sustained a MOF were significantly older (mean 65 vs. 61 years; p<0.001), had more fracture risk factors (age>65, previous fracture), lower BMD at all sites, and higher MOFR at baseline. They also received BA more frequently during follow-up (101/118 vs. 407/704; p<0.001).

Among the 58 women who died, eight had experienced a MOF (including two hip fractures), with no significant differences compared to the rest of the cohort.

Twelve women sustained a hip fracture, with a median time to event of 82 months (IQR 46–112). The number of expected hip fractures based on 2008 HFR was 17 (or 12 with BMD included).

The area under the ROC curve (AUC) for MOFR in predicting MOF was 0.643 (95% CI 0.592–0.694), indicating low discriminative ability. For hip fracture prediction, the AUC for HFR was 0.740 (95% CI 0.632–0.849), indicating moderate accuracy. BMD alone had no predictive value at any site (AUC<0.5). The negative predictive value for hip fracture in women with HFR<3% was 99%.

Applying the proposed MOFR thresholds, 331 women (40%) were classified as low risk (31 MOFR; incidence 9.36%, 95% CI 6.22–12.5), 357 (42%) as intermediate (58 MOF; incidence 16.25%, 95% CI 12.42–20.07), and 149 (18%) as high risk (35 MOF; incidence 23.49%, 95% CI 16.68–30.30).

A FRAX value of 7% or the presence of osteoporosis was identified as the most discriminative threshold for differentiating between low- and high-risk groups. After reclassification of the intermediate group, 574 women (69%) were categorized as low risk (63 MOF; incidence 10.98%, 95% CI 8.42–13.53), and 263 (31%) as high risk (61 MOF; incidence 23.19%, 95% CI 18.09–28.29).

The proposed algorithm is shown in Fig. 2.

Fig. 2.

Classification algorithm.

The negative predictive value for fracture in women classified as low risk by the algorithm was 89%.

Kaplan–Meier curves for time to first MOF by risk category are shown in Fig. 3.

Fig. 3.

Time until first MOF for the initial classification of MOFR (A) and after reclassification with BMD results (B) for low, medium and high MOFR.

This study assessed the clinical performance of a FRAX-based classification algorithm adapted to the Spanish population. Using prospectively collected data and a 10-year follow-up, we evaluated its ability to stratify fracture risk and guide decision-making in a real-world setting.

Our results confirm that the Spanish version of FRAX predicts HFR with acceptable discriminative power and high negative predictive value, particularly when the HFR is below 3%. However, its usefulness is limited in younger women, where hip fracture risk is generally low despite the possibility of other major fractures.

Consistent with previous observations, FRAX underestimates MOFR in this population. Nevertheless, applying the proposed thresholds—3.5% for low risk and 10% for high risk—allowed effective stratification into clinically meaningful categories. Rather than quantifying precise risk, the algorithm helps classify patients as low, intermediate, or high risk, and supports the indication for densitometry or treatment. Reclassification of intermediate-risk women using BMD further refines this process, particularly in identifying additional candidates for intervention. Although it may also be applied after a fracture, the algorithm is intended for primary prevention. Therefore, in the absence of fracture risk factors, it would be advisable to apply it periodically in postmenopausal women and in men over the age of 50.

To date, only one study has proposed FRAX-based thresholds for low (5%) and high (7.5%) MOFR in Spain. This was published by Azagra et al.,19 based on a retrospective 10-year follow-up of 816 women from the FRIDEX cohort, in whom 76 MOF were identified. Only 8.9% of the women fell into the intermediate-risk category, and thus had an indication for BD. The high-risk group comprised 70 women (8.6% of the cohort), among whom 15 MOF occurred. The authors proposed that BD should also be performed in these high-risk women, bringing the total percentage of candidates for densitometry to 17.5%. This strategy was considered cost-effective when compared with a model assuming BD in all women and treatment of all osteoporotic patients—an approach that no longer aligns with current clinical guidelines.

Validation of these thresholds in the FROCAT cohort,20 confirmed the distribution into low (67.8%), medium (11.9%) and high (20.3%) MOFR categories, although BMD was only available in 21.5% of subjects, limiting further analysis.

Compared with the FRIDEX cohort, our study population showed a higher incidence of MOF. Although the mean age of our participants was higher (62 vs. 57 years, with 38% vs. 18% aged ≥65), the main reason likely lies in our inclusion of women who received BA either before or after study entry. Theoretically, these patients would have been treated due to high risk or high fracture incidence. In contrast, the FRIDEX cohort had a greater proportion of women with previous fractures (49% vs. 26%), many of whom had not received treatment despite being at higher risk.

The recommendations from the Spanish Society of Rheumatology,21 advise performing BD in patients with a fragility fracture, two or more major risk factors, or an MOFR≥5%. However, data from EPISER201622—the first study to evaluate FRAX in a representative Spanish population—showed that only 10–13% of women fell within the 5–10% MOFR range, depending on the threshold used. This suggests that using 5% as a lower limit may be overly restrictive, limiting the identification of women who could benefit from further assessment.

One of the key contributions of Azagra et al.19,20 was to show that the high MOFR threshold in Spain is far below the 20% proposed by the NOF. Nonetheless, applying their algorithm strictly would substantially reduce the number of BDs performed and may contradict the widely accepted principle of treating high-risk patients directly, as is common in cardiovascular risk models. Recommending BD even in clearly high-risk individuals could delay necessary treatment.

The 10% MOFR threshold, as adopted by the Spanish Society of Rheumatology,21 is more conservative than the 7.5% proposed by Azagra et al.,19 but more compatible with clinical practice. It increases the proportion of women classified as intermediate risk—where BMD is most informative—while allowing direct treatment in those exceeding 10%, even without densitometry.

Data from EPISER2016 further support the use of a 10% MOFR threshold in Spain. Among women aged ≥65, all those with MOFR ≥10% also had HFR ≥3%—a value considered valid in the Spanish FRAX model. In contrast, 8% of those with MOFR ≥7.5% had HFR values below 3%, suggesting potential overestimation of risk at that lower threshold. Moreover, 13% of women in the 7.5–10% range would reasonably undergo BD and could then be reclassified based on BMD.

Establishing a high-risk threshold inevitably raises the question of how many older women would qualify for treatment. In EPISER2016, applying a 10% MOFR threshold would categorize most women over the age of 75 as high risk. Nonetheless, pharmacological intervention in this age group is considered cost-effective.23 Moreover, clinical evidence in women aged 65 years or older with osteopenia has shown that treatment with zoledronic acid significantly reduces both vertebral and non-vertebral fractures.24 Therefore, in older adults with osteopenia and high fracture risk, initiating treatment—even in the absence of densitometric osteoporosis—appears reasonable and appropriate when guided by individualized clinical judgment.

The main limitations of this study relate to the recruitment of patients from a BD unit and the inclusion of women treated with BA, factors that may limit generalizability. However, by including only referrals from primary care, we likely excluded patients with secondary osteoporosis due to severe comorbidities, and the population studied closely reflects the clinical setting where the algorithm would be applied. Excluding treated women would have led to a lower-risk sample, less representative of routine practice. Regarding the influence of BA during follow-up on fracture outcomes, previous research involving over 35,000 women aged ≥50 years −65% of whom received BA- showed that FRAX predictions remained well-calibrated in both treated and untreated individuals, with similar observed and predicted fracture rates.25 Since the only FRAX variable affected by treatment is BMD, and the model incorporates only femoral neck BMD (which tends to change modestly compared to spinal BMD), the overall impact of treatment on FRAX-based fracture risk estimation appears limited. Although baseline spine radiographs were not available, all recorded vertebral fractures were clinically relevant, suggesting they represent true incident or progressive events. Lastly, the results cannot be extrapolated to men.

Importantly, the study has several notable strengths. It was conducted in a real-world clinical setting, without interference in therapeutic decisions. Fracture identification was exhaustive, combining diagnostic codes and free-text searches, and confirmed through radiology or emergency reports when necessary. The age distribution of the cohort—mean age 62 years, with over one-third aged ≥65—enhances external validity, better reflecting the profile of women at increased fracture risk. As highlighted in EPISER2016, age is a key determinant of fracture incidence and high-risk classification. Finally, the exceptionally low loss to follow-up over a 10-year period (<2%) provides a robust foundation for the longitudinal analyses.

In conclusion, our results support the use of the Spanish version of FRAX to estimate hip fracture risk, applying the internationally accepted 3% threshold to identify high-risk individuals. To assess overall osteoporotic fracture risk, we propose a validated classification algorithm based on adapted MOFR thresholds (3.5% and 10%) that allows effective stratification of Spanish women into low, intermediate, and high-risk categories. This model offers a pragmatic approach to guide the use of bone densitometry and treatment decisions in routine clinical practice. By improving the alignment between estimated risk and observed fracture incidence, it contributes to more efficient resource allocation and optimized patient care in the management of osteoporosis.

The proposed algorithm may be particularly valuable in primary care settings and in contexts with limited access to bone densitometry. It facilitates clinical decision-making, especially for physicians who are not experts in osteoporosis, by providing a structured and evidence-based framework for risk assessment. A frequently underestimated strength of FRAX is that it prompts clinicians to systematically evaluate fracture risk factors, thereby improving the quality of risk stratification. Furthermore, in areas where DXA availability is restricted or subject to long waiting times, our findings support the appropriateness of initiating treatment in high-risk individuals based solely on FRAX estimates.

No funding was received for this study.

None.