Anemia is associated with increased risk of non-vertebral osteoporotic fractures in elderly men: the MrOS Sweden cohort

Osteoporosis is an important public health issue because of potential devastating results of osteoporotic fractures. Major osteoporotic fractures are typically defined as fractures of the hip, vertebral body, distal forearm, or humerus [1, 2]. They are associated with increased morbidity and mortality for the individual and are a major economic burden for the society [3]. The risk of osteoporosis is highly dependent on age [4]. Anemia is another public health issue that gets increasingly common in older age [5]. Anemia is defined by the World Health Organization (WHO) as hemoglobin (Hb) < 130 g/L in men and < 120 g/L in women [6]. Blood and bone cells live in close proximity in the bone marrow and all three bone cells, osteoblasts [7], osteoclasts [8, 9], and osteocytes [10, 11], have been implicated to play a role in the hematopoietic stem cell niche and in the differentiation of blood cells. A recent meta-analysis reported that hematopoietic disorders with increased marrow cell proliferation are associated with significant deterioration of bone health, independent of the lineage of the affected blood cells [12]. The association between blood and bone health is also seen in non-proliferative hematopoietic diseases such as Diamond-Blackfan anemia and Fanconi anemia [13]. In recent years, accumulating evidence suggests that anemia is a risk factor for osteoporotic fractures, with most evidence available for non-vertebral fractures in men [14‐18]. The primary aim of the present study was to investigate if anemia could predict fracture in a well-defined population-based cohort of ambulatory elderly men, with a long follow-up time.

The current cohort consisted of 1005 men with a mean age of 75.3 years (SD 3.2) and a mean BMI of 26.2 kg/m² (SD 3.5). Hb values were normally distributed. The mean value of Hb was 147 g/L (SD 12). The cohort included 66 men with anemia, of whom 14 had a Hb < 120 g/L. Baseline demographic and laboratory values for the whole study group and the subjects with or without anemia are presented in Table 1. Men with anemia had experienced more falls in the previous year before baseline and had lower hand strength and slower walking speed. Men with anemia had a higher prevalence of hypertension, stroke, diabetes, cancer, and prostate cancer. The 10-year survival rate was 31% in the anemic group and 49% in the non-anemic group (p = 0.009). We were neither able to establish a statistically significant difference in total hip BMD between the anemic and the non-anemic group (0.94 vs. 0.96, p = 0.159), nor in the difference between bone remodeling markers in anemic and non-anemic subjects. Subjects with anemia had higher EPO, iFGF23, CRP, and phosphate and lower eGFR, while we were not able to show statistical difference regarding PTH, D vitamin, and calcium levels. The association between anemia and iFGF23 was independent of age, EPO, and eGFR (β = 6.56, standard error = 2.84, p = 0.021).

In this prospective observational study of 1005 ambulatory elderly men, we found that anemia increased the risk of any fracture independently of age, falls, total hip BMD, and various comorbidities and conditions, even with a long follow-up time of up to 16 years. The increased fracture risk was driven by non-vertebral fractures, where the 10-year risk was increased by 115%, independently of age and total hip BMD. The age-adjusted hip fracture risk was increased by 132%, 10 years from baseline, but, probably due to few fracture events and few anemic men, the risk increase was not statistically significant after adjusted for total hip BMD.

Our results are in agreement with previous studies. A population-based study from Tromsö, Norway, found that anemia was associated with an increased risk of non-vertebral fractures but not for vertebral fractures, in men but not in women, after adjustment for BMD [15]. Similar information was reported from MrOS USA, which only studied men [17]. The National Health Screening Program in Korea reported an association between anemia and both vertebral and non-vertebral fractures, although the risk increase was higher for non-vertebral fractures [16]. The Korean cohort included 72,131 subjects, while our cohort, as well as the MrOS USA and the Tromsö cohorts, was smaller, thus making it harder to show a statistically significant risk difference. Furthermore, no adjustments were made for BMD in the Korean cohort [15‐17]. Conflicting results are published on the association between Hb values or anemia and fracture in women [14‐16].

In agreement with other studies, we showed that anemia is associated with increased frailty, morbidity, and mortality [5, 25]; however, the risk for fractures was independent of baseline falls and various diseases and conditions. Men with anemia had a higher prevalence of diabetes, as previously described [26]. Diabetes is known to be associated with increased risk of osteoporotic fractures [27], as well as being associated with hypogonadism [28]. The prevalence of prostate cancer was markedly increased in the anemic population compared to the non-anemic population, as well as the prevalence of hypogonadism. Hypogonadism is associated with prostate cancer, partly due to its treatment with androgen deprivation therapy (ADT), as well as the prostate cancer in itself being associated with lower testosterone values [29]. Both low testosterone and low estradiol in men are known to be associated with decreased BMD and increased risk for fractures [19, 30], as well as being associated with anemia, in our, as well as previous studies [23, 31]. Chronic kidney disease (CKD) and inflammation are known causes of secondary anemia and have both been associated with increased risk of fracture [32, 33], which is in agreement with our results. The mechanism by which anemia increases fracture risk is probably multifactorial and it is likely that anemia, to some extent, serves as a proxy for the underlying disease burden. Another mechanisms through which anemia might be associated with fracture risk is via higher phosphate and iFGF23, both related to anemia in our study. Both are associated with decreased renal function [34]. Hyperphosphatemia has been related to anemia [35] and increased fracture risk [36]. FGF23 is produced by osteocytes/blasts and is a known effector of bone mineralization, but our group has previously shown that Hb as a continuous variable correlates with iFGF23 [37]. We further demonstrated that anemia was associated with iFGF23, independently of renal function, EPO, and age. Contradicting results are published on the association between FGF23 and fracture risk [38, 39]. Another possible mechanism through which anemia affects bone is via elevated levels of EPO in the blood. Our group has previously shown that increased levels of plasma EPO are associated with increased risk of fracture in elderly men, although most noticeable for vertebral fractures [21]. EPO production is stimulated via hypoxia and bone homeostasis is effected by oxygen tension. Preclinical studies have shown that hypoxia has a negative effect on bone both by inhibiting osteoblast formation and stimulating osteoclast formation [40]. Clinical studies have shown that short-term exposure of high altitude, which is associated with hypoxia, can lead to decreased BMD [41, 42], and hypoxia during sleep in elderly men leads to increased risk for falls and fractures [43]. Thus, it is possible that anemia increased fracture risk partly because of relative hypoxia. In the bone marrow, EPO controls the proliferation and differentiation of erythrocytes [44]. Several preclinical studies have shown that EPO has a detrimental effect on bone mass in adult rodents, but other studies in growing mice and using traumatic models have reported a stimulation of bone formation by EPO [45]. Our group has recently shown that EPO is associated with increased BMD in elderly men [21]. It has been suggested that EPO is a cross-link between the hematopoietic stem cell niche and bone cells. Osteoblasts can produce EPO, which results in the expansion of the hematopoietic stem cell niche and is associated with selective expansion of erythroid cells [46].

In our present study on anemia, we were not able to establish a statistically significant association with lower BMD. The reason for this can be either a real lack of association or a lack of power in our study owing to the small number of men with anemia. The latter would be supported by the loss of statistical significance after adding BMD in multivariate Poisson regression for hip fractures. Our group has recently shown that Hb analyzed as a continuous variable is associated with total hip BMD after adjustment for age and BMI. When adjusted for estradiol or osteocalcin, the association was no longer significant [47]. Hb has been associated with lower bone mass in two Italian cohort studies [48, 49]. In these studies, bone mass was measured with peripheral quantitative computed tomography (pQCT) or ultrasound-derived T- and Z-score, respectively. In another study where BMD has been measured with DXA, no association between anemia and cross-sectional BMD was seen. In that study, an association between fast annual change in BMD and anemia was seen [50]. It is therefore possible that anemia is associated with bone mass in a way that was not captured by a single DXA or anemia is associated with bone fragility not captured by DXA. In a recent meta-analysis by Steer et al., an association between bone marrow cellularity evaluated with a trephine biopsy, and bone density change was seen, whereas no association was seen with Hb measurement and bone density change [12]. Peripheral blood cell count may not be the best way to evaluate hematopoiesis, rather a trephine biopsy would be preferable; however, this was not collected in the MrOS cohort.

An important strength of our study is the long follow-up time for fractures and a relatively large cohort. The fracture data are reliable owing to the unique personal registration number of all Swedish citizens, making it possible to access information in X-ray archives about the time and the site of fracture. Fasting and standardized sampling of data were used for all blood samples. We acknowledge several limitations in our study. First, it is a cross-sectional study, with only single measurements of blood variables and BMD. A significant limitation is that the anemic group only contains 66 subjects. Thus, the total amount of events was few in the anemic group which makes statistical analysis more vulnerable and true associations can be missed. Since this is an observational study, causality cannot be determined. Although multiple covariates were tested, we cannot rule out the possibility of residual confounding.

In conclusion, our findings indicate that anemia, after controlling for multiple factors, is predictive for any fracture and non-vertebral osteoporotic fractures in elderly men. The underlying mechanism is probably multifactorial but measuring Hb might be useful for long-term risk prediction for fractures.