Important risk factors and attributable risk of vertebral fractures in the population-based Tromsø study

Fragility fracture is highly prevalent in the general population, and is increasingly becoming a public health concern [1, 2]. From the age of 60, the lifetime risk of any fracture in Norway is 63% in women and 34% in men [3]. Although hip fracture is the most serious consequence of osteoporosis [4], other osteoporotic fractures are also associated with important health issues as pain [5], loss of function [6], and increased risk of new fractures [7]. Some studies have suggested that clinical vertebral fractures are associated with increased risk of subsequent vertebral fractures [8‐10], non-vertebral fractures [11‐14], and mortality [15]. Vertebral fracture is often underestimated due largely to the problem of under-diagnosis. Indeed, only one in three vertebral fractures is clinically diagnosed [5, 16], and the majority is either undetected or incidentally detected through radiographs. Recent development in dual x-ray absorptiometry (DXA) technology has allowed a population based assessment of vertebral fracture to be carried out by common DXA densitometers [17]. The vertebral fracture assessment (VFA) method has been used in many population settings, and its sensitivity and specificity are comparable to spinal radiographs in the ability to diagnose grade 2 and 3 (moderate and severe) vertebral fracture [13, 17]. These developments provide opportunities for studies of vertebral fracture risk.

Norway is among the countries in the world with the highest incidence of fragility fractures, including hip [18] and forearm [19]. According to the FRAX risk calculator, age, sex, body mass index, previous fracture, parental hip fracture history, alcohol, tobacco, arthritis, and use of glucocorticoids constitute important risk factors for osteoporotic fractures [20]. Studies on vertebral fractures indicate that among these, age, bone mineral density (BMD), and previous fractures are the most significant risk factors [21, 22]. Data recently reported from the Tromsø Study in Norway indicate that the prevalence of vertebral fractures do not differ between women and men [23]. The aim of the present study was therefore to address the important risk factors for vertebral fractures in a Norwegian population, and to estimate the percentage of fracture cases that can be explained by these risk factors.

The study cohort included 2887 individuals (1206 men and 1681 women) aged between 38 and 87 years, with the majority being post-menopausal (women) or aged 50+ years.

Overall, approximately 64% of men and 61% of women reported good health status, with a majority (more than 78%) in the sedentary or moderately physical activity group.

Vertebral fracture was found in 166 (13.8%) men and 199 (11.8%) women (p = 0.07). Baseline characteristics of participants stratified by fracture status and sex are shown in Table 1. Men and women with a vertebral fracture were on average older than those without and had lower total hip BMD. Among women, those with a fracture also had lower levels of education and poor overall self-reported health status compared to those without a fracture in addition to lower weight and stature. The results did not change when we included only participants above 50 years of age in analyses.

In either sex, age was significantly associated with the prevalence of vertebral fracture (p < 0.001) (Figure 1). When data were combined and analyzed in a bivariate logistic regression model (Table 2), the risk of vertebral fracture increased with advancing age, OR 1.43; 95% CI: 1.31 to 1.56 in women and OR 1.28; 95% CI: 1.17 to 1.40 in men, per 5-year increase. Decreasing total hip BMD, increased body weight and higher stature were each associated with reduced vertebral fracture risk, but after adjusting for age, only BMD was associated with vertebral fracture risk in men (OR 1.40; 95% CI: 1.18 to 1.67 per −0.14 g/cm² decrease), and in women only BMD (OR 1.60; 95% CI: 1.34 to 1.90 per −0.13 g/cm² decrease) and weight (OR 0.93; 95% CI: 0.88 to 0.99 per 5 kg increase) (Table 2).

Because sex was not associated with the prevalence of vertebral fractures in any age group (p > 0.12) (Figure 1), we wanted to explore the association between sex and vertebral fracture risk. BMD was inversely related to the prevalence of vertebral fracture in both sexes (p < 0.001) (Figure 2). After adjusting for age and BMD, the odds of vertebral fracture in men was 1.87-fold (95% CI 1.4 to 2.5) higher than in women. Testing for interaction between the variables showed that there was no interaction between sex and age (p = 0.08), but a significant interaction between sex and BMD (p = 0.005). According to the WHO definition [32], 1.5% of all men and 6% of all women had osteoporosis (T-score ≤ −2.5 at the total hip). Categorizing BMD into T-scores, men had greater risk of vertebral fracture than women in the normal and normal + groups (T scores > −1.5) (p < 0.012) and in the osteoporosis group (p = 0.011), but the numbers were low. We therefore categorized the BMD levels into quartiles, kept the lowest quartile of less than 0.85 g/cm² as one group, and the rest as another. With this stratification, at BMD levels higher than 0.85 g/cm², men had a greater risk of fracture than women (OR 1.52; 95% CI 1.14-2.04), after adjusting for age.

A multivariable logistic regression model was conducted to search for independent risk factors for each sex separately, and the results are shown in Table 3. For both women and men the “final” and most optimal model for predicting fracture risk included age and total hip BMD. For a given age group, the risk of fracture increased as BMD decreased. Conversely, for a given BMD category, the prevalence of fracture increased with advancing age. Attributable risk analysis (Table 4) indicated that in women 45.9% and in men 33.4% of all vertebral fractures cases were attributable to advancing age (i.e. more than 70 years) and lower total hip BMD (i.e. less than 0.85 g/cm²).

In line with other studies [21, 22, 33], we found that advancing age and lower BMD at the total hip were independent determinants of vertebral fracture risk, whereas all other background- and lifestyle variables did not contribute statistically in the final model. The estimated population attributable risk fraction for these two factors combined was 45.9% in women and 33.4% in men.

At the highest BMD levels, the risk of vertebral fracture at given age was greater in men than in women. Few population-based studies have included vertebral fracture assessment in both women and men. It is therefore difficult to compare the present sex-related vertebral fracture risk. There has been some controversy in the literature about whether there is a higher risk of vertebral fractures in women than in men. Our finding of non-significant difference in vertebral fracture risk between men and women (before adjustment of age and BMD) is not consistent with a previous review which suggested that the incidence of vertebral fractures in men is about one third to one half of that in women [34], but of course, our estimates are based on prevalent vertebral fractures. However, our results are in line with findings from the Dubbo Osteoporosis Epidemiological Study [35] where the prevalence of vertebral deformities was higher in men than in women. In that study the higher prevalence in men was observed regardless of diagnostic criteria, suggesting, as in our study, that vertebral fractures may be overlooked in men.

Several other studies report, similar to the present study, that low BMD measured at the femoral sites are associated with prevalent radiographic vertebral fractures [21, 22, 28]. We have not seen other studies comparing sex difference at normal and normal + levels (not osteoporosis nor osteopenia). The higher prevalence of vertebral deformities at these BMD levels in men indicate, as also suggested by others [16], that these deformities could be of other origin than osteoporosis, possibly mechanic or due to childhood diseases [34] and should be explored in follow up studies. Studies defining the proportion of fractures attributable to trauma in childhood and young adulthood are lacking [34]. As Seeman et al. claims, if these youth fractures do make a contribution to the numbers of elderly men with fractures, this will exaggerate the prevalence of vertebral fractures and by that mask a sex difference in the prevalence of osteoporotic vertebral fractures [34], something that should be examined in follow-up studies. The finding from our study indicating that at a given age and given BMD level the OR for having a vertebral fracture is higher in men than in women, somehow corresponds to how the risk of hip fractures seems to be similar in men and women for any given BMD [16, 32, 36, 37]. Although an unknown proportion of the vertebral fractures in men may be of mechanic origin, our study highlights the importance of osteoporosis and vertebral fracture risk in men, and that vertebral fractures should definitely not be considered a women health problem only.

The present findings should be viewed within the context of strengths and potential limitations. The Tromsø Study is a population-based, longitudinal study with a high participation rate. The present study is a cross-sectional survey within the framework of the Tromsø Study, where vertebral morphometry was done for the first time. Because of its cross-sectional design, causal inference cannot be drawn from the findings and the results will need confirmation within a longitudinal design. For logistic reasons, we did not perform quality control using x-ray technology. It is, however, reported that DXA scans are more precise in measuring moderate and severe than mild deformities [13], and 99% of the identified deformities in our data set were either moderate or severe. The intra-class correlation coefficient for determination of average height of the vertebra was good. Ideally, we should have compared determination of fracture severity in the sample. This was, however, difficult to attain, since determination is done electronically by the software, based on identified vertebral heights. Despite of random selection in the Tromsø Study, the morphometry group was younger with a slightly lower proportion of women compared to the group not selected to VFA. However, when we compare the morphometry group with the phase 2 participants of the Tromsø VI survey, whom to our best knowledge should be a representative sample [24], the morphometry sample of women and men was slightly older (3 years) with lower height (2 cm), but did not differ significantly in any other way. Despite high rates of hip- and forearm fractures in Norway, the prevalence of vertebral fractures in this population was comparable to reports from others [38‐40]. Although we should be careful drawing firm conclusions from our prevalence estimates, we still feel comfortable to compare between women and men, especially in age stratified analyses. It is a major limitation to our study that we lack information of important risk factors included in FRAX [20], especially the history of vertebral fractures [21, 33]. Although BMI [41] and smoking [42] are considered to be independent predictors of fracture risk, they did not contribute to the final model. The predictive value of physical activity, self-perceived health, and education to fracture risk is uncertain [43] and did not affect the results although the self-reported nature and our dichotomization of the variables may have precluded possible associations. However, this study confirms that age and BMD are important predictors of vertebral fractures in women and men [21, 22].

Advancing age and declining BMD are independent determinants of vertebral fracture risk in women and men, but accounting for less than half of the total risk. The predictive value of prevalent vertebral fractures on subsequent vertebral fractures and other types of fractures should be explored in longitudinal studies.