German DVO risk score identified more patients requiring treatment compared to FRAX

Introduction

Worldwide, 6% of men and 21% of women between 50 and 84 years of age suffer from osteoporosis. Along with cardiovascular disease, stroke and cancer, osteoporosis is one of the most common health problems (1, 2). The disease is characterised by a pathological increase in fracture risk resulting from a reduction in bone mineral density (BMD) and bone quality.

The determination of individual fracture risks is becoming more relevant to the assessment and timely treatment of osteoporosis. Lately, the availability of anabolic therapeutic agents has increased the necessity to improve the prediction of fragility fractures and thus target therapeutic interventions in postmenopausal women more effectively (3). Risk factors other than BMD influence the fracture risk. Different tools have thus been developed to assess osteoporosis-associated fracture risk. (4, 5). The FRAX score is one such tool that was established in 2008 and updated several times to adapt to new developments (6). The FRAX score calculates an individual 10-year fracture risk by considering defined clinical risk factors and optionally the BMD at the femoral neck. The probability calculated is a country-specific risk of suffering a major osteoporotic fracture (MOF), including spinal, hip, shoulder and wrist fracture, or a hip fracture (HF), using an algorithm. There are no official therapy thresholds above which scores patients are recommended treatment. However, the internationally most commonly implemented therapy thresholds for FRAX are ≥20% for MOF and ≥3% for HF (7, 8, 9). Nowadays, age-dependent therapy thresholds are much more the norm, but were not available at the time of our data collection (10, 11, 12).

The score commonly used in German-speaking countries is based on the German guidelines for diagnostics and treatment of osteoporosis developed by the Dachverband Osteologie (German Confederation of Osteologists, DVO). The DVO score is a step-by-step approach (Table 1) that classifies patients in groups of 10-year fracture-risk incidences of suffering a hip or clinical vertebral fracture. Based on forty risk factors, patients are assessed whether they need further diagnostics. Starting from a 10-year fracture incidence >20%, further diagnostics including BMD measurement via dual-energy absorptiometry (DXA) are advised. Patients are stratified into groups with either a risk between 20 and 30% or >30%, depending on the combination of risk factors and BMD. Specific treatment for osteoporosis is recommended for patients with a 10-year fracture risk >30% (13, 14, 15). The first edition of these guidelines was published in 2003 and was revised in 2006, 2009, 2014 and 2017, in accordance with the relevant literature. In 2023, a new DVO score was published, representing a major change in risk assessment by calculating the 3-year fracture risk for 10-year fracture-risk probabilities of suffering a hip or clinical vertebral fracture and adapting different risk categories (16, 17).

Aims

In respect of developing the new DVO score in 2023, it was discussed whether German-speaking countries would also apply the FRAX score in the clinical care of patients with osteoporosis. We were therefore interested in whether the FRAX score would have characterised women treated according to the DVO score before 2023. In this retrospective study, we compared risk scores and treatment recommendations of female patients in Germany based on DVO score, with the FRAX score effective at the time of clinical presentation. We adapted therapy thresholds frequently employed internationally (≥20% for MOF and ≥3% for HF) (15) for FRAX to our study group and contrasted them with therapy recommendations based on the DVO score. Owing to the fact that our patients were treated according to DVO score during the subsequent years, it is not possible to conclude which score is more accurate regarding fracture risk prediction.

Material and methods

Results

In total, 555 female patients with a complete set of data were analysed. Table 2 summarises the baseline characteristics of these patients. The two scores approach risk factors differently. Smoking, glucocorticoid intake, rheumatoid arthritis and alcohol consumption are treated as causes of secondary osteoporosis in DVO, whereas these risk factors have to be entered separately into the FRAX interface. Viewing the secondary osteoporosis causes according to FRAX, we see that the onset of menopause before the age of 45 was the most common (95.23%). The most common secondary cause in DVO was vitamin-D deficiency in 28 of 116 patients (24.13%).

The average T-score at the femoral neck was −1.86 ± 0.98, and thus higher than the average T-score at the lumbar spine of −2.21 ± 1.23. The lowest T-score (−2.51 ± 0.99) of all measurement sites (femoral neck, total femur and lumbar spine) was even lower than at the lumbar spine. When only patients with any prior fracture were analysed, the minimum T-score at the femoral neck was significantly lower (P < 0.01, Wilcoxon test) than in those without any fracture (femoral T-score −1.99 ± 0.93 versus −1.72 ± 1.01).

Figure 1 illustrates fracture risk probabilities (vertical axis) according to the DVO score in our study population. The DVO score identified 52.8% (293 out of 555) as having a 10-year fracture risk of suffering a hip or vertebral fracture >30%. We identified a secondary cause of osteoporosis in 22.9% (67 out of 293) of these patients (12.1% of the total 555 patients).

View Full Size

10-year fracture probabilities (vertical axis) for hip or vertebral fractures according to the DVO score, split into primary (black) and secondary (grey) causes of osteoporosis.

Citation: Endocrine Connections 14, 5; 10.1530/EC-25-0048

• Download Figure
• Download figure as PowerPoint slide

The mean 10-year fracture probabilities according to FRAX were significantly higher when including BMD, both for MOF (P < 0.001) and HF (P < 0.001). We additionally calculated the mean individual FRAX 10-year fracture probability within the subgroups of <20%, 20–30%, and >30% fracture risk according to the DVO score. The mean FRAX score fracture probabilities with and without BMD proved to be lower than those determined for the DVO scores in all groups. Hence, the FRAX score identified fewer patients at risk. The mean FRAX score fracture probabilities for patients with a DVO score >30% without BMD proved to be 18.2 ± 11.4 for FRAX MOF and 8.9 ± 9.5 for FRAX HF. Including BMD, they were 17.4 ± 10.8 for FRAX MOF and 8.1 ± 9.1 for FRAX HF.

We calculated FRAX score thresholds that would identify the same percentage of patients (52.5%) as the DVO score. We found the adapted therapy thresholds for MOF to be markedly lower (without BMD 11.0%, with BMD 10.0%) in our female study population than internationally employed thresholds of ≥20%. Looking at the internationally common therapy threshold of ≥3% for HF, our adapted therapy threshold without BMD was higher (3.4%), whereas the value with BMD (2.6%) was below the common internationally implemented threshold.

We applied the therapy threshold of ≥3% for HF and ≥20 for MOF to compare the therapy indications for both the DVO and FRAX scores (Fig. 2). The most patients were identified by the FRAX score as requiring treatment adopting a threshold of 3% for HF without BMD followed by the DVO score, with the small difference being statistically insignificant (P = 0.705). The fewest patients were recommended treatment by the FRAX score based on a therapy threshold of≥ 20% for MOF with or without BMD. All FRAX scores, except FRAX for HF without BMD, showed a significant (P < 0.05) difference regarding their therapy recommendations in respect to the DVO score.

View Full Size

Therapy recommendations ‘yes’ (black) and ‘no’ (grey) based on DVO guidelines, FRAX score (grouped by therapy thresholds for MOF and HF). The figure displays a difference in the total number of therapy recommendations based on the score.

Citation: Endocrine Connections 14, 5; 10.1530/EC-25-0048

• Download Figure
• Download figure as PowerPoint slide

Aiming to determine the overlap of patients identified by each score, the greatest percentage concordance of 39.3% for women with therapy recommended was found between FRAX HF ≥ 3% without BMD and DVO. However, both scores identified different individual women in need of therapy. The FRAX score classified 16.8% of patients as in need of therapy who were not identified by DVO score. In comparison, 13.5% were identified by DVO but not by FRAX. Furthermore, FRAX HF ≥ 3% with BMD identified 72.4% of the patients recommended treatment according to the DVO score. The FRAX score for MOF ≥20% with BMD identified the fewest individual patients also identified by DVO score as requiring treatment.

Table 3 depicts specific differences for patients identified as in need of therapy by the DVO score and the FRAX score HF with BMD. These scores reveal the greatest overlap (38.2% of the total population) of patient-specific therapy decisions. We compared patient-specific parameters of females identified by both DVO and FRAX HF with BMD with those of each score to describe the difference in the risk profile of each score.

We conclude that certain risk factors have differing influence on the calculation of the DVO and FRAX scores, resulting in different treatment recommendations to patients. The number of prior spinal fractures was more than twice as high in the females identified only by the DVO score (38.3 vs 18.6%). Furthermore, those patients only identified by DVO score were revealed to have a higher percentage of parental HFs in their history (13.6 vs 8.5%). Patients only identified by the FRAX score were also revealed as having a higher prevalence of peripheral fractures (40.7 vs 29.6%).

Looking at risk factors other than fracture history, those patients only identified by FRAX HF with BMD were found to have a higher percentage prevalence of the risk factor ‘current smoking’ (22 vs 16.6%). In addition, the risk factor ‘rheumatoid arthritis’ was 5.1 vs 2.5% in DVO score. Nevertheless, patients identified by the DVO score presented secondary osteoporosis more often (28.4 vs 15.3%).

We used a confusion matrix to analyse the respective sensitivities and specificities of the two scores in regard of the predictive value of prevalent spinal fractures. In comparison, the DVO score proved to have a greater and significant sensitivity (0.8, P < 0.001), precision (0.41, P = 0.002), negative predictive value (0.86, P < 0.001) and accuracy (0.63, P < 0.001) than the FRAX score for HF with BMD (sensitivity (0.67, P < 0.001), precision (0.37, P < 0.001), negative predictive value (0.82, P < 0.001) and accuracy (0.60, P < 0.001)). Specificity was nearly identical between FRAX (0.58, P = 0.002) and DVO score 0.57 (P = 0.004).

Discussion

This single-centre retrospective study compared the respective fracture risks calculated by the DVO and FRAX scores with and without BMD measurements in 555 female patients evaluated for osteoporosis in a specialised endocrinological centre between 2007 and 2014. The study population was not a representative sample of the female population, presenting a higher risk of suffering from osteoporosis compared to the general population. The FRAX and DVO scores applied for these women differ in the number of risk factors assessed and their algorithms of 10-year fracture-risk calculation. The FRAX score implements an algorithm to calculate 10-year fracture probabilities of a MOF or HF. The score may be used without any DXA measurements, or with BMD or T-score values (which are based on a female reference population that is applied to both men and women alike). In contrast, the DVO score used during the investigated period presents a step-by-step, replicable 10-year fracture assessment. The risk of suffering an osteoporotic hip or vertebral fracture is calculated based on the T-score (using a female reference population). The DVO guideline recommends specific anti-osteoporotic therapy at a fixed threshold of a 10-year fracture risk >30%. On the other hand, the FRAX score does not supply any official therapy threshold. A systematic review of numerous countries revealed a therapy threshold for MOF is commonly ≥20% and ≥3% for HF (15, 18).

There are hardly any data published with the German DVO score in comparison with the FRAX score (19, 20). To better analyse the similarities and differences of the scores, we compared both scores with respect to their indication for anti-osteoporotic treatment. As already stated above, we were not able to determine which score predicted the 10-year fracture risk more accurately with this approach.

In accordance with the version of the DVO guidelines applied, patients were recommended further treatment when the estimated 10-year fracture risk is ≥30%. Comparing the 10-year fracture risks that FRAX calculated to those determined by the DVO guidelines, only the FRAX score for HF without BMD was similar. The FRAX score for MOF (with and without BMD) indicated a lower risk than the DVO score. We assume the following reasons for this observation: the DVO score and the FRAX score do not estimate the same fracture outcomes. The DVO score estimates fracture risks based on prevalent and clinical hip and vertebral fractures, whereas FRAX only estimates the risk dependent on clinical fractures including those of the spine, shoulder and forearm in the MOF category. Furthermore, our study population represented females suffering from severe forms of osteoporosis, with a large prevalence (52.6%) of prior fractures including those of the spine.

Interestingly, below the age of 65 years, the fracture risk calculated by FRAX was higher when including BMD compared to FRAX without BMD (data not shown). This relationship reversed for women above the age of 65 years, with higher percentages for FRAX without BMD. This may reflect the changing influence of BMD in FRAX score calculation, possibly caused by the increasing importance of the different risk factors during ageing. There was no further increase in FRAX scores after the age of 80 years, perhaps owing to the inclusion of mortality risk in the FRAX score calculation.

Taking the internationally common therapy thresholds of 10-year fracture risk ≥20% for MOF and ≥3% for HF into account, therapy recommendations did not differ significantly with BMD scores. However, the individual patients identified were not the same and there were patients only identified either by FRAX or by DVO as requiring treatment.

If exactly the 52.8% of patients identified by the DVO score as in need of therapy were to be treated according to the FRAX scores, the therapy thresholds would need to be ≥2.6% with BMD and ≥3.4% without BMD for HF. The adapted therapy thresholds nearly match or even exceed the internationally recommended therapy threshold of ≥3% for HF. The adapted therapy thresholds for the 52.8% of patients identified by the DVO score as requiring therapy strongly differ (≥10% without BMD; ≥11% with BMD) from the internationally recommended therapy threshold of ≥20% for FRAX MOF. This implies that female patients in Germany are treated at an earlier time point and at lower fracture risk compared to other European countries implementing FRAX. We concluded that the FRAX score appears to discriminate better for HF than for MOF. This observation is in accordance with the findings of other studies (20, 21, 22).

There might be several reasons for the difference between the FRAX and DVO scores and the differing therapy recommendations for the patients in our study population. One explanation might be that the DVO implements the lowest T-score of all measurement sites, and specifically including the spine. DXA measurement of the total femur or femoral neck has the best predictive value for HFs, whereas the measurement of the lumbar spine has a better predictive value for vertebral fractures (23). There is no evidence of any improvement in stratification of fracture risk by using the lowest T-score of several measurement sites including the spine versus taking T-score of the hip alone (24, 25). An epidemiological study from Canada even revealed an overestimation of the fracture risk by taking the lumbar spine instead of femoral neck as a BMD parameter for FRAX. One might discuss whether the DVO score overestimates 10-year fracture risks by using the lowest T-score of all measurement sites.

To analyse the scores further, we contrasted the patient-specific sets of risk factors of patients identified by each score. Spinal fractures were identified as being twice as common in women identified only by the DVO score as requiring therapy. The possible assumption is that the DVO score is more sensitive when detecting vertebral fractures in female patients by considering the lowest T-scores. Women only identified by the DVO score as in need of treatment were positive for more secondary osteoporosis risk factors. In contrast, those only identified by the FRAX HF score as in need of treatment had more peripheral fractures at the age of 50 were more likely to be smokers, took more glucocorticoids and the T-score at the femoral neck was nearly one point lower compared to those only identified by the DVO score (−2.3 versus −1.4, respectively). We applied a confusion matrix to analyse the respective sensitivities and specificities of each score in terms of the predictive value of recording prevalent spinal fractures. Whereas the DVO score was more sensitive (0.8 versus 0.67) for patients with prevalent spinal fractures, both scores demonstrated similar specificities (0.57 versus 0.58).

In summary, the DVO score and FRAX score do not identify the same individual female patients at risk of fracture in our study population. In this retrospective study of females presenting a higher risk of suffering from osteoporosis compared to the general population, only the FRAX score for HF without BMD (internationally common therapy threshold ≥3%) identified a similar percentage of patients compared to the DVO score. The patients in our study who were identified by the DVO score as requiring treatment corresponded to a FRAX with BMD hip-fracture risk threshold of ≥2.6%. However, using the FRAX score for MOF with and without BMD for our study population, the results demonstrate that the German therapy threshold corresponds to ≥11% (MOF with BMD) or ≥10% (MOF without BMD). Hence, therapy was clearly recommended at a lower risk compared to the internationally common therapy threshold of ≥20%.

Our study results have a number of practical implications for the treatment of patients with osteoporosis in German-speaking countries. The DVO score was better at identifying particularly women with a high-risk profile of vertebral fractures. Given the fact that the German version of the FRAX score is used in some areas of Austria and Switzerland, our data increase the awareness of differences in treatment decision. Our results also had major impact on the development of the new DVO guidelines. Initially, there was discussion on using the German version of the FRAX score in Germany also; however, our comparison of these two scores in men (20) and women supported the revision of the DVO score instead. More generally, our data may also contribute to the discussion on country-specific solutions to identify high-risk patients with osteoporosis as early as possible (26). Furthermore, the German government and institutions responsible for treatment decisions and reimbursement policies accept the DVO score in the version valid at any specific time.

There are a number of limitations to our study. The study group does not reflect epidemiological data, but women evaluated for osteoporosis in a specialised osteoporosis centre. In addition, there are no valid data on fracture incidence in these women, because the patients included in this study were treated based on their individual risk after being evaluated. Therefore, we cannot conclude which score predicts the actual fracture incidence more accurately.

Since the analysis of the women in our study, FRAX and the DVO score have been continuously undergoing revision. It will be interesting to compare these tools in future trials, especially with respect to the identification of high-risk patients. The ultimate priority of any of these risk scores should be to identify the high-risk patients as early as possible to prevent fractures.

In conclusion, the comparison of the international FRAX score and the German DVO score valid before 2023 in a preselected female study population with a higher risk of suffering from osteoporosis compared to the general population revealed clear differences. The DVO score proved to be better at identifying particularly women with a high-risk profile of vertebral fractures. Given the goal is to identify women at high risk of fracture as early as possible to prevent fractures, it would be highly advisable to use the score most appropriate in the respective country.

Declaration of interest

Anna Frank, Judith Charlotte Witzel, Christina Heppner, Annette Lamersdorf and Andreas Leha declare that they have no conflict of interest. Heide Siggelkow is a member of the committee of the German guidelines for diagnostics and treatment of osteoporosis developed by the Dachverband Osteologie (German Confederation of Osteologists, DVO).

Funding

We acknowledge support of open access publication charges by Göttingen University.