INTRODUCTION

The osteoporosis is an illness linked to a high morbimortality that increases as the population grows older. It has been defined as a systemic skeletal disease characterized by a deterioration of bone micro-architecture and a decrease of bone tissue, with a consequent increase in bone fragility and a higher susceptibility to fracture¹. It is a clinically silent disease that is not manifested by other signs but for its complications, fractures.

The main consequences of osteoporosis are fragility fractures that can appear in different locations, though they typically happen on the vertebrae, distal radius and proximal extremity of the femur^2,3. They are fractures with a high economic cost and are associated with a higher morbimortality, specifically those on the vertebrae and the proximal femur. Hip fracture mortality, the most serious manifestation of osteoporosis, is 8% during the first month after the fracture (acute mortality). It rises to 30% after a year⁴. Furthermore, the recovery of patients who do not pass away is poor. Only 30% of patients suffering a hip fracture return to the baseline situation⁵. The vertebral fracture shows a higher incidence than the hip fracture. While the hip fracture shows a yearly incidence of 1.3-1.9 cases/1,000 inhabitants/year, the incidence of vertebral fractures is 13.6/1,000 inhabitants/year in males and 29.3/1,000 inhabitants/year in females². Although its mortality is lower than that of hip fracture, it is not despicable, especially in patients also presenting a respiratory disease^6,7. Therefore treatments are designed to prevent its appearance through adequate therapeutic measures. In order to establish the most appropriate treatment it is necessary to dispose of stand-alone diagnostic factors that help identify the every patient’s individual risk through additional tests or risk scales.

Bone mineral density (BMD) has demonstrated an ability to predict fracture risk and also to prove the efficacy against fractures of different treatments⁸. This has been proven by a meta-analysis, although its usefulness in individual patients is less measurable⁹. Moreover, there are a high number of fractures whose BMD levels are in the range of osteopenia¹⁰.

Due to the current difficulty in some countries to carry out densitometries, different clinical procedures have been developed to establish fractures risks and indication of densitometry¹¹. These procedures have not been clearly implemented owing to the development of the FRAX^® tool, which is a risk scale sponsored by the World Health Organization consisting of a very simpleto-use computer tool whose risk has been adjusted by country. Its purpose is to identify the risk of fracture in men and women between 40 to 90 years¹². However, it also poses problems by not including important fracture risk factors such as falls, the lack of definition of osteoporosis secondary causes or the corticosteroid dosing, the underestimation of osteoporotic major fracture risk in some populations and its invalidity on treated osteoporosis patients. On the other hand, simple clinical data reportedly have prediction ability comparable to that of the FRAX¹⁰.

The objective of this study, then, was to assess the fracture prediction ability of BMD, of the FRAX with or without BMD and of clinical data in an osteoporosis-treated female population and in a non-osteoporotic female population showing risk factors of the disease over an eight-year follow-up, by using the classification and regression tree method (CART). Its ultimate goal is to identify those patients with the highest risk of osteoporotic fracture.

PATIENTS AND METHODS

RESULTS

Our work included the study and follow-up of 276 postmenopausal patients whose mean age at the start of the study was 61.08±8.43 years. The BMI was 25.67±4.04. Almost all the patients were Caucasian (n=274, 99.6%), only one patient was of South American ethnicity (0.4%). Regarding gynecological data, the age at menarche was 13.03±1.46 years and the age at menopause was 47.99±5.75 years. The patients had a mean of 2.07±1.3 children. 56.5% of the patients (n=156) were diagnosed with osteoporosis before the start of our study, and all of them were treated. After eight years of follow-up, 72 patients (24.6%) had suffered a fracture and 17 (6.2%) had also suffered a second fracture; 61 patients (22.1%) had suffered a fracture being over 55 years of age. The data for the global population are shown in table 1.

Tabla 1.  General characteristics of the population included in this study 

BMI: body mass index; BMD: bone mineral density. Data expressed as means ± standard deviation and absolute number (percentage).

There were 16 (8.6%) patients treated with thiazides, 32 (17.1%) with serotonin receptor inhibitors, 1 (0.5%) with androgen inhibitors, 8 (4.3%) with beta-blockers, 20 (10.7%) with thyroid hormones, 74 (27%) with antiresorptive agents, 4 (1.5%) with hormone replacement therapy, 11 (4%) with anabolic therapy, 25 (9%) with corticosteroids and 10 (3.6%) with strontium ranelate. Table 2 shows the characteristics of the included patients divided into two groups: patients with fractures occurred during the eight years of follow-up and patients without fractures.

Tabla 2.  Clinical characteristics of the patients included in this study, comparing those who have suffered fragility fractures to those who have not 

BMI: body mass index; BMD: bone mineral density. Data expressed as means ± standard deviation and absolute number (percentage).

The results of the CART analysis showed that the main risk factor for suffering an osteoporotic fracture after 8 years of follow-up was having suffered previous fractures. Among patients who had suffered a previous fracture, having a femoral neck BMD less than 0.67 was the main risk factor. Among the patients who had not suffered previous fractures, the main risk factors were BMI and having a FRAX without BMD (major osteoporotic fracture) greater than 9.30 (Figure 1). Figure 2 shows the results of the CART analysis in patients diagnosed with osteoporosis. The main risk factors for suffering a fracture after 8 years of follow-up were having suffered previous fractures and presenting a BMD of the femoral neck less than 0.663. Among the patients who had not suffered previous fractures, the main risk factor was being over 67 years of age (Figure 2). Finally, the results of the CART analysis in patients not diagnosed with osteoporosis showed that the main risk factors for suffering an osteoporotic fracture were being over 55 years of age and having previously suffered the associated non-bonerelated disease, comorbidity (Figure 3).

Figure 1.  CART analysis (Classification and regression tree) in order to study the main risk factors associated with the risk of suffering an osteoporotic fracture in all the patients included in our study 

Figure 2.  CART analysis (Classification and regression tree) in order to study the main risk factors associated with the risk of suffering an osteoporotic fracture in those patients diagnosed with osteoporosis 

Figure 3.  CART analysis (Classification and regression tree) in order to study the main risk factors associated with the risk of suffering an osteoporotic fracture in osteoporosis-free patients 

DISCUSSION

In our study, the data that allow us to pin-point patients who have suffered an osteoporotic fracture during follow-up are the existence of a previous fracture, age, and FRAX with or without BMD. This last data loses part of its value in the CART analysis, which is why it is probably a marker of disease severity rather than a predictor in our series. In the CART analysis, which takes into consideration the factors that may influence the subsequent appearance of fractures, we witnessed that the key element for the total population is the existence of a previous fragility fracture supplemented with a BMD lower than 0.67 in the femoral neck, while in the absence of fractures, a BMI of 28 or lower and a FRAX for a major osteoporotic fracture above 9.3 would be the determining elements. In patients diagnosed with osteoporosis, under active treatment, the data are similar to those described for the general population except that, in those cases without a previous fracture, FRAX did not play any role in determining that 67 years of age is the distinguishing data. This is logical given that FRAX is of no use in treated osteoporotic patients. In patients without osteoporosis, the factor that determines the appearance of fractures is an age of over 55 years and the associated non-bone-related pathology, comorbidity.

The importance of secondary prevention is based on the fact that prior fractures constitutes the first determinant in the appearance of posterior fractures¹⁶,¹⁷. A high percentage of men and women are considered at high risk, which has been determined by a previous fracture and not diagnosed or treated¹⁸. This occurs for any fragility fracture location, being especially significant for vertebral and hip fracture. In the first case, in addition to the fracture itself, its severity according to Genant's classification and the number of fractures increase the risk¹⁹. Hip fracture is the most serious complication of osteoporosis, not only because of its ability to predict subsequent risk, but also because of its high morbimortality. Furthermore, in this case, BMD is a strong and consistent predictor of posterior fracture²⁰. New studies carried out in real clinical practice to assess the efficacy of different treatments have shown that the increase in BMD is a determining factor in its efficacy reducing fractures²¹. In our study, the existence of a BMD in the femoral neck below 0.660 was a strong predictor consistent with these data. Another important factor in the appearance of new fractures is the time factor. This is where the concept of imminent risk of fracture has arisen, which is greater in the first two years after the fracture²²,²³. In a study carried out in a group that included 377,561 older women, the greatest risk of fracture was found to occur in the following 5 years, and especially frequent in the first two. These authors found that the determining factors were age, the location of the fracture and the associated bone disease²⁴. In our study, this imminent risk and the location of the posterior fractures have not been determined. However, in patients not diagnosed with osteoporosis, age and associated pathology were present, coinciding with the data from this large series.

Age was another important risk factor in our CART analysis, especially in patients without prior osteoporosis and in patients with osteoporosis but without prior fracture. Numerous published studies have confirmed this fact. Ensrud et al. ²⁵, in a group of 6,652 women and with a 10-year follow-up, concluded that models based on age and BMD, or on age and fractures, predicted risk in a similar way to FRAX. Similar data were obtained by Bolland et al.²⁶ when comparing two scales, FRAX and Garvan, with age and BMD. In the Glow study, with 19,586 female participants over 60 years of age, it was observed that a model based on age and previous fractures was superior to FRAX and Q-Fracture without BMD²⁷.

Another relevant fact of our study is the small role of FRAX in our CART analysis. We can see that the FRAX, both with BMD and without BMD, makes differences between the patients who do and do not fracture, but when conducting a statistical analysis based on algorithms, its value disappears. Only when analysing the total population, a FRAX for major osteoporotic fracture without BMD, higher than 9.3, has a role in patients without fracture. This FRAX value coincides with the FRAX thresholds established by Azagra et al.²⁸ in their analysis of the Fridex cohort. These authors divided the patients into three groups, considering those with a FRAX greater than 10 as high risk. This may be due to several facts. In the Spanish population, FRAX underestimates the risk of major osteoporotic fracture. Several cohorts have tried to validate it by analyzing the differences between predicted and observed fractures²⁹-³¹. The results can be considered acceptable in predicting the risk of hip fracture, but not for that of a major osteoporotic fracture, probably due to the lack of robust epidemiological data for this type of fracture. Comparing the FRAX with other simpler tests, including only age, has not shown a greater predictive ability³². Another fact of our study is that osteoporotic patients were receiving active treatment, therefore, it was not possible to validate FRAX in their cases.

Another noteworthy fact in non-osteoporotic patients is the importance of non-bone-related pathological records, comorbidity in the appearance of fractures. Unfortunately, there is no single approved index to assess fragility in clinical practice³³. Some used indices have been associated with an increased risk of fractures³⁴ and falls³⁵. The Glow cohort and the CaMos cohort used different indices to assess the relationship between fragility and fractures, but both include many of the parameters assessed in our study. The variables of the Glow fragility fracture index included 15 items on comorbidity, 12 on basic activities of daily living (similar to the Barthel index), 6 items on signs and symptoms (fullness of life, energy, exhaustion, fatigue, non-intentional self-assessed pain/disconfort, weight loss). The fragility index used in the CaMos cohort included 30 items, 13 referring to pathology, 5 to functional aspects (vision, hearing, gait, manual dexterity/use of tools and cognition) and 12 to general health and daily activities. However, the Charlson index was not associated with the risk of fracture, although some studies, such as the SIDIAP registry, which included 186,171 men, found that a Charlson index ≥3 was linked to an increased risk of hip fracture³⁶. The newest finding in this study is the use of the CART statistical methodology to establish the risk of posterior fracture in a heterogeneous population that included osteoporotic patients under active treatment and non-osteoporotic patients.

In conclusion, the use of a binary statistical procedure (CART) in a cohort of patients allows us to identify patients with a higher risk of fractures based on clinical parameters and complementary tests that are simple to carry out and establish more effective therapeutic measures.