Introduction

Vertebral fractures are the most common osteoporotic fractures, and their consequences in morbidity and mortality are now well recognized [1–3]. They have an important impact on patient’s health related quality of life [4], and cause great health care utilization [5]. The number of elderly people at risk for such fractures is expected to increase in the next decades; accurate identification of osteoporotic fractures and appropriate treatment are needed to reduce the impact of the disease on patients and on the health care system.

The diagnosis of vertebral fractures is based on spine X-rays, but underdiagnosis has been recognized [6–8]. This is explained in part by the absence of gold standard for definition of fractures, particularly the mild vertebral fractures, and the disagreement about the clinical relevance of such deformities. Some studies [9], but not all [10], found a correlation between mild decreases in vertebral heights and back pain, height loss, and low lumbar spine bone density. But, besides clinical symptoms, the clinical significance of vertebral fractures is related to the increased risk of subsequent fractures that they induce [11–13], and this subsequent risk is influenced by the baseline deformity definition [14]. In post-menopausal osteoporotic women, both the severity and the number of prevalent fractures are determinants of the risk of incident ones [11, 15].

The assessment of prevalent mild vertebral fractures is problematic, as although they represent a large proportion of the vertebral fractures [14], such deformities are sometimes considered as an expected effect of aging [8]. Facing a mild vertebral deformity, the best mean to distinguish between a mild fracture, and a degenerative change, is to have a previous X-ray to compare as aging changes are unlikely to develop within a short-time period. Prospective follow-up studies are an opportunity to study such mild incident deformity, and their role in subsequent risk of fracture. The aim of our prospective study was to assess the effect of mild vertebral fractures, both prevalent and incident, on the risk of subsequent vertebral and non-vertebral fractures in a large population of post-menopausal patients with osteoporosis, followed for 4 years.

Material and methods

Study subjects

The study population consisted of women who were randomly assigned to placebo in two randomized double blind clinical trials evaluating the efficacy of strontium ranelate for treatment of post-menopausal osteoporosis [16, 17]. In the Spinal Osteoporosis Therapeutic Intervention (SOTI) [16], patients were included if they were 50 years old or more and post-menopausal for at least 5 years, had at least one vertebral fracture, and had a lumbar spine BMD of 0.840 g/cm2 (Hologic device) or less. In the TReatment Of Peripheral Osteoporosis (TROPOS) [17], patients were included if they were ≥ 74 years of age (or between 70 and 74 years old with one additional fracture risk factor, i.e. history of osteoporotic fracture after menopause, residence in a retirement home, frequent falls, or a maternal history of osteoporotic fractures at the hip, spine, or wrist), had been post-menopausal for at least 5 years, and had a femoral neck BMD ≤ 0.600 g/cm2 (Hologic devices). Before inclusion either in the SOTI or the TROPOS trial, patients were subjected to a run-in study to initiate normalization of their calcium and vitamin D status [16, 17]. Patients in the placebo groups received daily calcium supplements at lunchtime (up 1,000 mg of elemental calcium, depending on their dietary calcium intake) and vitamin D (400–800 IU depending on the baseline serum concentration of 25 hydroxyvitamin D). Because the study design, centers, BMD central reading center, and X-ray central reading center were common to both SOTI and TROPOS studies, data were pooled [18]. The current study has been conducted in patients randomly assigned to placebo arms, followed for 4 years.

Assessment outcomes

The assessment of vertebral fracture was similar in SOTI and TROPOS studies. Three lateral radiographs of the spine (thoracic and lumbar radiographs and an image of the thoracolumbar junction) were obtained at baseline and annually, according to standardized procedures, and if there were indications of symptomatic vertebral fracture (acute back pain, a decrease in body height of at least 1 cm, or both). At baseline, anteroposterior radiographs of the spine were also obtained. All radiographs were assessed at a central facility, where they were assessed by a single investigator (JF) told of the time sequence of each radiograph but unaware of the treatment assignment and clinical symptoms. A semiquantitative visual assessment of each vertebra, from T4 to L4, was performed by the same reader throughout the study. The semiquantitative grading scale was as follows: normal, grade 0; mild fracture: a decrease in the height of any vertebra of 20 to 25 percent (grade 1); moderate fracture: a decrease of 25 to 40 percent (grade 2); and a severe fracture: a decrease of 40 percent or more (grade 3) [19–21]. Vertebrae with deformities of non-osteoporotic origin (degenerative changes) or exhibiting potential misleading appearances because of an obliquity of their projection, were not given a grade > 0. A new fracture was defined by a change in the score of a vertebra from grade 0 at baseline to a subsequent grade of 1 or more.

The assessment of non-vertebral fractures was similar in both studies. They were reported by study investigators based on written documentation provided and documented in the source document (radiograph, radiological report, copy of the hospitalization/emergency department report). Only documented non-vertebral fractures were taken into account in the statistical analysis. Fractures of the coccyx, skull, jaw, face, phalanx (fingers and toes), and ankle were not regarded as being related to osteoporosis and were not considered.

Bone mineral density at the lumbar spine and proximal femur was measured by dual energy X-ray absorptiometry at baseline and at 6-month intervals (Hologic devices). All the scans were analysed centrally [16, 17]. Baseline data were used in this study.

Statistical analysis

Data corresponding to the first 4 years of SOTI and TROPOS have been studied. The analyses were restricted to patients randomised in placebo group (N = 3358). Analyses on vertebral fractures were based on patients having a baseline and at least one post-baseline assessable X-ray (N = 2551). Similarly, analyses on non-vertebral fractures were performed on patients having a baseline assessable X-ray and a follow-up of non-vertebral fracture (N = 2954). Besides, non-vertebral fractures were also analysed in patients older than 70 years.

The incidence of patients experiencing a first new vertebral or non-vertebral fracture was estimated according to Kaplan--Meier method. A Cox regression model was used to estimate the relative risk (and its 95% confidence interval) of new fracture, in subjects with only mild prevalent fractures as compared to those without prevalent fracture, and in subjects with at least one grade ≥ 2 prevalent fracture as compared to those without grade ≥ 2 prevalent fractures. The same analysis over the third and fourth years of the study was used to compare patients with only mild incident fractures to patients without incident fracture during the first two years of the study, and with at least one grade ≥ 2 incident fracture to patients without grade ≥ 2 incident fractures during the first two years of the study. The Cox model was adjusted on potential baseline covariates (age, body mass index, lumbar spine bone mineral density and then adding prevalent vertebral fractures), in the analysis of vertebral fractures; the model was adjusted on age, body mass index and femoral bone mineral density in the analysis of non-vertebral fractures.

Results

A population of 3,358 post-menopausal women with osteoporosis is the basis of this study.

Vertebral fracture risk

The risk of incident vertebral fracture was analyzed in 2,551 patients having baseline and post-baseline X-rays. Their baseline characteristics are in Table 1. Baseline characteristics such as age, years since menopause, weight and height were similar in patients having no prevalent vertebral fractures, in patients having only mild fractures and in patients having at least one vertebral fracture of higher severity. The lumbar spine T score, but not the femoral neck T score, was lower in patients having grade 2 fractures as compared to the two other groups. Among the 215 patients with grade 1 vertebral fractures at baseline, 169 (78.6%) had only one fracture; 38 (17.7%), 7 (3.3%) and one patient had 2, 3 and 4 grade 1 vertebral fractures, respectively.

Table 1 Characteristics of the study population for vertebral fracture risk according to severity of vertebral fractures at baseline
Full size table

Effect of prevalent fractures

The Kaplan--Meier estimate of the vertebral fracture incidence over 4 years was 26.7% (24.9%–28.6%) in this population. In the 1,292 patients without prevalent vertebral fracture, this incidence was 16.1% (14.0–18.3). In the 215 patients with only mild vertebral fractures at baseline, this incidence was 27.4% (21.0–33.8). In the 1,044 patients with at least one grade ≥ 2 vertebral fracture at baseline, the incidence was 39.8% (36.6–43.0) (Fig. 1).

Fig. 1
figure 1

Incidence of new vertebral fracture by severity of prevalent and incident vertebral fractures. a-no fracture. b-Only mild fractures. c-At least one grade ≥ 2 fracture

Full size image

The relative risk (adjusted on age, body mass index and lumbar spine BMD) of sustaining a vertebral fracture in 4 years was 1.8 (1.3–2.4) (p < 0.001), for patients having only mild fractures at baseline as compared to patients without vertebral fracture (Table 2). In patients with at least one grade ≥ 2 fracture, this adjusted relative risk was 2.7 (2.3–3.3) (p < 0.001); as compared to patients with only mild vertebral fractures, the adjusted relative risk of incident fracture in these patients was 1.5 (1.1–2.1) (p = 0.002). As sensitivity analysis, new vertebral fractures have been studied in patients with only one mild fracture at baseline (N = 169) as compared to patients without fractures, which yielded similar results (RR = 1.85 (1.32–2.60) p < 0.001).

Table 2 Risk of new vertebral fractures
Full size table

Considering the whole SOTI and TROPOS database, we assessed, in a post-hoc analysis, the effect of strontium ranelate in the sub group of patients having only mild vertebral fractures at baseline; over 4 years strontium ranelate decreased the risk of incident vertebral fractures by 42% (RR = 0.58 (0.37–0.90), p = 0.014).

Effect of incident fractures

During the first 2 years of this four-year study, 156 patients had only mild incident fractures, without fracture of higher severity, and 173 other patients had at least one incident grade ≥ 2 fracture (Fig. 1). The relative risk (adjusted on age, BMI and lumbar spine BMD) of an incident vertebral fracture in the 2 subsequent years was 1.7 (1.1–2.6) (p = 0.015) and 1.9 (1.3–2.6) (p < 0.001) in these two groups respectively (versus without fractures during the first two years) (Table 2). There was no difference in the relative risk of having a vertebral fracture in th 3rd and 4th years between patients having only mild incident fractures during the first 2 years, and those having at least one grade ≥ 2 incident fracture during the same period (RR = 1.1 (0.7–1.9) (p = 0.62).

In a second step, we adjusted the Cox model on a 4th parameter, i.e., the number of prevalent (baseline) fractures. With such a model, those relative risks to have a new vertebral fracture in the 3rd and 4th years were respectively 1.4 (0.9–2.2) (p = 0.11) for a mild incident fracture, and 1.5 (1.1–2.2) (p = 0.02) for an incident grade ≥ 2 fracture during the 2 first years.

During the 4 years of follow-up, symptomatic vertebral fractures occured in 228 patients. Among them, the proportion of patients with a first radiological mild fracture was 29.9%.

Non-vertebral fracture risk

The risk of incident non-vertebral fracture was analysed in 2,954 patients having a baseline X-ray and a follow-up on non-vertebral fractures. Their baseline characteristics were not different as those of patients analysed for the vertebral fracture risk.

Effect of prevalent fractures

The Kaplan--Meier estimate of the non-vertebral fracture incidence over 4 years was 16.5% (14.9%–18.0%) in this population. In the 1,511 patients without prevalent vertebral fracture, the incidence was 12.9% (10.9–14.8). In the 240 patients having only mild vertebral fractures at baseline, this incidence was 17.6% (12.1–23.1). In the 1,203 patients with at least one grade ≥ 2 vertebral fracture at baseline, the incidence was 20.8% (18.1–23.5). These incidences in patients over the age of 70 were 16.7% (15.1%–18.4%) (N = 2,550), 12.6% (N = 1429), 19.0% (N = 224) and 23.0% (N = 897) in these four groups respectively (Fig. 2). The effect of prevalent vertebral fractures on the subsequent risk of non-vertebral fractures was estimated by relative risks adjusted on age, BMI and femoral neck BMD; these RR were 1.32 (0.91–1.9) p = 0.15 and 1.68 (1.36–2.09) p < 0.001 in patients having only mild fractures and at least one grade ≥ 2 fracture at baseline, as compared to patients without prevalent vertebral fractures. This analysis was also restricted to the 2550 patients (among 2,954) aged of at least 70 years; the results are in Table 3. In this population the risk of non-vertebral fracture induced by mild vertebral fractures, calculated before adjustements, was statistically significant (RR = 1.57 (1.08–2.29), p = 0.024) and approached the significance (p = 0.06) after adjustements (Table 3).

Fig. 2
figure 2

Incidence of new non-vertebral fracture by severity of prevalent and incident vertebral fractures. a-no fracture. b-Only mild fractures. c-At least one grade ≥ 2 fracture

Full size image
Table 3 Risk of new non-vertebral fractures in osteoporotic women over 70 years old
Full size table

Effect of incident fractures

The relative risks of non-vertebral fractures during the 3rd and 4th years of follow-up were 1.32 (0.91–1.92), p = 0.15 and 1.68 (1.36–2.09) p < 0.001 for patients having only mild incident fractures, and patients having at least one grade 2 incident fractures, (versus without fractures), respectively. These results in patients aged of at least 70 years are reported on Table 3.

Discussion

This post hoc analysis of the placebo groups of two trials shows that mild vertebral fractures in post-menopausal osteoporotic women are a risk factor for subsequent vertebral fractures. Incident mild fractures induce an immediate risk: among patients having at least one mild fracture in 2 years, 28% will fracture again in the 2 next years. Our results indicate also that an incident vertebral fracture increases the short-term, i.e., 2-year, risk of non-vertebral fracture.

Our data are consistent with previous studies demonstrating that the risk of vertebral fractures is increased in patients with prevalent vertebral fractures [12, 22]. Few studies considered the severity of prevalent vertebral fracture as a risk factor. In the placebo group of the MORE trial [11], the proportion of women with at least one incident vertebral fracture in 3 years was 10.5% in patients having mild fractures at baseline, as compared to 4.3% in patients without prevalent vertebral fractures. In the EPOS study, a vertebral deformity with a severity between 3 and 4 standard deviations according to morphometric definition induces a significant increased risk for an incident fracture by 2.2 (1.2–4.2) [23]. Our data confirm that mild fractures, and even only one, increase significantly the risk of incident fracture in osteoporotic women, including after adjustements on other risk factors as age and BMD. Moreover, an incident mild fracture increases the risk of a subsequent vertebral fracture and thus such a diagnosis has clinical application in the management of these patients. Thus, mild vertebral fractures participate to the “fracture cascade” phenomenon which has been described either in clinical studies [13] or in mathematical models [22].

Our data confirm that prevalent vertebral fractures are a risk factor for non-vertebral fractures ; above all, we have shown prospectively that an incident fracture of grade ≥ 2 in 2 years increases the risk of a non-vertebral fracture in the 2 subsequent years. Such relationship was not so strong for mild incident fractures. This may be explained by the low sample size and the low incidence of non-vertebral fracture in this sub-population. It is possible that mild fractures represent a local risk on the spine through mechanical factors [24], or that the alterations of bone they represent are more prominent in bones of trabecular histology. On the other hand patients with such fractures may have less decrease in physical activity than patients with more severe fractures [25], and, thus, a different risk of falls, but we have no data to test this hypothesis.

Our data emphasize the relevance of the diagnosis of mild vertebral fractures. Clinical symptoms cannot help in this matter, and the discrepancy between radiological and clinical vertebral fractures is well-known, including for incident ones [26]. On the other hand, the missing of diagnosis of mild fractures even on X-rays is frequent (7. 8). Some of them may be missed because physicians consider that there are only an expected effect of age, and because their clinical significance has been a subject of controversy [7, 8]. Actually this point is raised when such deformities are used to discriminate patients with or without osteoporosis, but our results have been obtained in a population of patients having osteoporosis according to BMD data. Vertebral deformities can be due to developmental abnormalities, Scheuermann’s disease sequelae, and degenerative changes [27]. Attention must be paid on osteoporotic depressions of the central end plates of the vertebrae [27, 28], as osteoarthritic changes occur only on the anterior part of the vertebrae [29]. The main difficulties are related to isolated short anterior vertebral heights at the mid-thoracic spine. The interpretation of such deformities must take into account degenerative changes of adjacent discs, and the presence of deformities of similar appearances on contiguous vertebrae, both signs being more in favor of a non-osteoporotic origin of the deformity. Vertebral fractures are unlikely to have identical aspect and occur more frequently to non-contiguous vertebrae, although these signs are not specific of osteoporosis. Thus, the use of the semi quantitative method for fractures assessment needs a first step of identification of non-fracture deformities or normal variants, otherwise the method may be less specific than the quantitative approach [28]. Finally, an incident vertebral deformity occurring in 2 years in an osteoporotic post-menopausal woman cannot be attributed to degenerative change, and our results on prevalent deformitites are confirmed and reinforced by the results on incident ones.

Our study has strengths and limitations. The assessment of fracture was carefully conducted using standard procedures of acquisition, and standard centralized reading of all X-rays. All the semiquantitative assessments were made by a single experienced investigator, throughout the study, unaware of treatments and clinical data, but including comparison with the patient’s previous X-rays. Parallax potential effects and patients positioning errors were limited in the context of a clinical trial, as written procedures of acquisition were followed by all centers. Before diagnosis of fracture, a non-osteoporotic origin was considered for each deformity. Our results were obtained in a population of post-menopausal women with osteoporosis, and thus cannot be generalizable to other patients including women with osteopenia or normal BMDs, and men. We did not analyse the risk according to the type of the deformity (biconcave, wedge), which may influence the outcome of the patients [23, 30] as this is not included in the semiquantitative method we used [19]. It has been shown previously that shape of the prevalent fractures is important in predicting incident vertebral fractures [30]. Moreover we did not analyse falls in this population, which may influence the relationship between mild vertebral fractures and non-vertebral fractures.

In conclusion, mild vertebral fractures (and even only one) must be considered as a risk factor for subsequent fractures in post-menopausal women with osteoporosis. Efforts must be made for accurate diagnosis of these fractures and there is a need for training and awareness in this field. Moreover, the finding that one out of four patients with an incident mild vertebral fracture in 2 years will fracture again in the next 2 years justifies a fast and effective anti-osteoporotic treatment in these patients.