Introduction

Falling and fractures are an important public health challenge for elderly population whose proportion is growing worldwide. In older population, falling and fractures are important determinants of functional decline and disability [1, 2], increase the mortality rate, and lead to dramatic medical costs to society [3]. It is estimated that about 30 % of elderly people living at home fall at least once each year [4, 5]. The risk and severity of injuries after a fall are important, and falling appears to be a key determinant of fracture risk [6]. In elderly people, some body sites of fractures, such as the hip, the wrist, or the spine, are a clinical manifestation of osteoporosis [7]. Other bones common to fracture in older people are the arms or legs. The prevention of falling could decrease the occurrence of fractures. Among modifiable risk factors for falls and fractures, malnutrition could play a role through its impact on muscle and bone health [8]. Adequate protein intake is recommended for optimal muscle function with aging and to maintain strength and physical endurance [9], while the role of protein intake on bone health is more controversial [10]. A low protein intake has been associated with a decrease of bone mineral density [11] and with higher risk of bone fractures in women [12]. Among other nutrients, adequate calcium and vitamin D intake have been consistently associated with preserved bone health [13, 14]. We have previously shown that a nutrient pattern with higher intakes of calcium, phosphorus, vitamins B12, proteins, unsaturated fats, and moderate alcohol was associated with a lower risk of fracture [15]. Furthermore, malnutrition has been associated with lower bone mineral density [16] that could have an effect on fracture. Taken altogether, these findings suggest that global malnutrition could be a risk factor for both falling and fracture.

Malnutrition, as evaluated by the Mini Nutritional Assessment (MNA), is present in 4 to 10 % of seniors living at home [17]. The MNA, sign of energy and multiple nutrient deficiencies and particularly of a low intake of protein, has been validated specifically in older persons to identify people with a poor nutritional status [18]. Malnutrition has been associated with poor muscle strength [19] and impaired balance [20], two conditions that could lead to falling. Saka et al. have shown that 43 % of people with a poor nutritional status experience one or more falls in the next year, which is significantly higher when compared to 31 % of people with a satisfying MNA score (p = 0.02) [21]. Other studies have shown an association between malnutrition and the risk of falling [22–26], but most of them examined subjects living in hospitals or in nursing homes and during a short follow-up. At our knowledge, no study has explored the association between the nutritional status assessed by the MNA and the occurrence of fracture. The aim of this study was to determine whether nutritional status (evaluated by the MNA) is associated with the overall and fracture site-specific (i.e., the hip, wrist, spine, upper and lower limbs) risk of falling and fracture in community-dwelling older people.

Methods

Study population

The Three-City (3C) study is an observational cohort study of vascular risk factors for dementia; the methodology has been described elsewhere [27]. In 1999–2001, 9294 community-dwellers aged 65 years and over were randomly recruited from electoral rolls of three French cities: Bordeaux (n = 2104), Dijon (n = 4931), and Montpellier (n = 2259). The study was approved by the Ethics Committee of the Kremlin-Bicêtre Hospital. Written consent was provided by all participants. At baseline, psychologists collected sociodemographic and medical history of participants by a questionnaire. To date, five follow-up examinations were performed at about 2, 4, 7, 10, and 12 years after inclusion.

Two subsamples were analyzed. For the falling subsample, from the 7464 participants with no recent falls at baseline, we included 6040 individuals with available nutritional status information at baseline, who were followed at least once during the 12 years of follow-up, and had no missing data in the covariates. We employed the same selection criteria for the fracture subsample: from the 8560 participants with no recent fractures at baseline, 6839 subjects were included.

Outcomes

Self-reported history of falling was provided by participants at baseline and at each follow-up visit (except at the 7 years follow-up). Self-reported history of fractures and information about the body sites of the fractures (the hip, wrist, spine, upper limb (arm, shoulder, collarbone, excluding wrist), lower limb (excluding hip), and other site) were collected at baseline and at each follow-up visit. The two main outcomes of interest in this analysis were the occurrences of a first fall and of a first fracture. In secondary analyses, we studied fractures for each site separately.

Nutritional status assessment

MNA is a questionnaire identifying elderly people who are at risk of malnutrition or malnourished [18]. It is composed of 18 items grouped into four parts (anthropometric measurements, a global assessment, a dietary questionnaire, and a self-assessment) and provides a total score on 30 points. A score superior to 23.5 indicates a good nutritional status, inferior or equal to 23.5 a risk of malnutrition, and inferior to 17 indicates malnutrition. In the present study, the MNA was not strictly administered but similar questions from the baseline survey allowed us to build an equivalent from proxy items. This form has shown a good agreement with the standard form in a previous study [28]. In 3C, the proportion of malnourished individuals was only small; therefore, these malnourished were grouped with those at risk of malnutrition representing people in a “poor nutritional status” opposed to a “good nutritional status.”

Other variables

Whatever the outcome studied, i.e., risk of falling or risk of fracture, a set of common adjustment variables, including gender; study center (Bordeaux, Dijon, or Montpellier); marital status (married, widowed, or other); education level (defined as high for subjects with at least a high school diploma vs the others); smoking status (never smoker, former smoker, or current smoker); cognitive function evaluated by the Mini Mental State Examination (MMSE) [29]; using more than five drugs, hypertension (defined as having systolic blood pressure >160 mmHg or diastolic blood pressure >95 mmHg or use of antihypertensive drugs); and the presence of dyspnea, arteritis, and heart failure were considered. Regarding the risk of fracture, treatments for osteoporosis (bisphosphonates, raloxifene, strontium ranelate, teriparatide, and calcitonin) and supplementation with calcium and/or vitamin D were considered in addition to the abovementioned common covariates.

Statistical analyses

A description of sociodemographic and medical information was performed in the two subsamples according to their nutritional status. The χ 2 test and Student’s t test were used to determine bivariate differences. Cox proportional hazard regression models with delayed entry and taking age as the time scale were performed to estimate the associations between poor nutritional status and the risks of incident fall and the overall and site-specific fractures. We secondarily tested effect modification by gender, by including a malnutrition-by-gender interaction term in our model. Statistical analyses were conducted using SAS statistical package release 9.3 (SAS Institute Inc., Cary, NC).

Results

Subjects characteristics

Table 1 presents the characteristics of the sample included in the “falling subsample.” These subjects were 73.5 years old on average and 58.2 % were female. At baseline, 12.0 % were in poor nutritional status according to the MNA, with a twofold higher prevalence in women compared to men (15.5 vs 7.1 %, respectively). During a maximum follow-up of 12 years (mean 7.8 years), 3369 subjects (55.8 %) reported at least one incident fall. The incidence rates was 14.3/100 person years (PY) (95 % confidence interval (95 % CI) 13.8–14.9) among women and 8.8/100 PY (95 % CI 8.3–9.3) among men. People who reported a fall during the follow-up were significantly older, more often female, widowed, never smoker, user more than five drugs, and had more often history of heart failure compared with those without experience of falls.

Table 1 Baseline characteristics according to the incidence of a first fall, the 3C Study, n = 6040
Full size table

Table 2 presents the characteristics of the sample included in the “fracture subsample.” These subjects were 73.7 years old on average and 59.8 % were female. At baseline, 12.8 % had a poor nutritional status according to the proxy MNA. Over 12 years, 1264 subjects (18.5 %) had at least one incident fracture at any site. Among this sample, 2.3 % had a hip fracture, 1.8 % a spinal fracture, 4.6 % a wrist fracture, 4.6 % a lower limb fracture, 3.6 % an upper limb fracture, and 5.5 % a fracture at other body sites as first fracture. The incidence rate for fracture at any site was 2.6/100 PY (95 % CI 2.5–2.7), without significant differences between men and women. People with an incident fracture were significantly older, more often female, widowed, and never smokers. They also reported more often supplementation with vitamin D and/or calcium and medication for osteoporosis than those without experience of fracture.

Table 2 Baseline characteristics according to the incidence of a first fracture at any site, the 3C Study, n = 6839
Full size table

Poor nutritional status and risk of falling

The results of the Cox proportional hazards model analyses of the association between nutritional status and the incidence of falling are presented in Table 3. As the interaction between nutritional status and gender was statistically significant (p < 0.01), we presented results stratified by gender. A poor nutritional status was significantly associated with an increased risk of falling in both genders. Indeed, the association appeared stronger in males (hazard ratio (HR) = 1.67; 95 % CI = 1.36–2.05) than in females (HR = 1.21, 95 % CI = 1.08–1.36).

Table 3 Hazard ratios (and 95 % confidence intervals) of falling associated with baseline poor nutritional status according to the MNA, stratified by gender, the 3C Study, n = 6040
Full size table

Poor nutritional status and risk of fracture

A poor nutritional status was significantly associated with a higher risk of fracture at any site (Table 4) after controlling for potential confounders (HR = 1.28, 95 % CI 1.09–1.49). There was no significant interaction between the nutritional status and gender. The analysis by site-specific fractures did not show significant association. Also, lower and upper limbs were close to significance.

Table 4 Hazard ratios (and 95 % confidence intervals) of having a fracture and by sites associated with baseline poor nutritional status according to the MNA, the 3C Study, n = 6839
Full size table

Discussion

In this large population-based cohort of elderly individuals followed for up to 12 years, a baseline poor nutritional status, assessed by the proxy MNA, was significantly associated with an increased risk of falling, especially in males. Poor nutritional status was also significantly associated in both genders with a higher risk of fracture. These associations were independent of major potential confounders.

Our results are in accordance with different population-based studies using MNA to assess risk of malnutrition. First, Chien et al. examined the association between the baseline nutritional status and the incidence of falls between the third and the fourth year after baseline [22] among 4440 community-dwelling Taiwanese aged 53 years and older. A poor nutritional status was associated with a higher risk of falling (odds ratio (OR) = 1.73; 95 % CI 1.23–2.42) in particular among men. Second, as reported by Visvanathan et al., elderly people receiving domiciliary care services in Australia and identified in poor nutritional status were more likely to fall (HR = 1.65; 95 % CI 1.13–2.41) over a 12-month period [30]. However, in an Australian study on a sample of 254 subjects with only 10 defined at risk of malnutrition, there were no association with the risk of falls over a 1-year period [31].

Other studies have investigated the association between nutritional status and the risk of falling in different populations, i.e., in frail elderly people [23], in elderly receiving home care [24], in residential long-term care [25], and during a hospital admission [26], and generally found consistent associations between poor nutritional status and higher prevalence of falls. Yet, these studies were mainly cross-sectional and did not observed incident falling. Thus, no causal link between nutritional status and falls could be derived from these studies, where malnutrition could be a consequence of falling in these elderly people due to for example inability to go out for shopping and prepare adequate meals. Our study adds to this literature showing a prospective relation between nutritional status and the occurrence of fall, an association which support different hypotheses.

With aging, muscle mass and muscle strength decrease, a phenomenon also known as sarcopenia [32] with multiple related factors such as inadequate nutritional intake, imbalance protein metabolism, and malnutrition [33, 34]. The risk of falling associated with malnutrition could be related to muscle weakness and deficiencies in the coordination of movement as well as protective mechanisms such as reaction time and muscle strength [35]. In our study, unexpectedly, the relation between malnutrition and a greater risk of falling was stronger in men than in women; we could hypothesize that the hormone secretion plays a role on the muscle metabolism involving a difference according to the gender.

To our knowledge, no published study has investigated the association between nutritional status and the risk of total fractures (i.e., fractures at any site) since most studies were restricted to hip fractures [36–38], which is the most frequent fracture in the elderly. Wilsgaard et al. showed in older non-smokers Norwegian that loss of BMI over 10 years was significantly associated with a risk of non-vertebral fracture [39]. Weight loss has been associated with a higher risk of hip fracture [36–38] and frailty fractures [40].

As the risk factors differed according to the body sites of fracture [3], we have explored the association between poor nutritional status and the risk of fracture by subtypes. However, no association was observed between poor nutritional status and the risk of fracture by sites. These analyses were performed on a limited number of events with a maximum of 300 fractures, and a lack of power is expected. Nevertheless, the risks of fracture of the hip, lower limbs, and upper limbs ranged from 1.3 to 1.4, and the last two were close to significance. Surprisingly, no study so far has been conducted to assess a potential detrimental effect of poor nutritional status evaluated by the MNA on the risk of fracture over time, which precludes comparisons with existing literature. However, our results were consistent with the other analyses from a subsample of the 3C study, reporting that a higher adherence to the Mediterranean diet [41] and a “nutrient-dense” pattern rich, among others, in calcium and protein [15] were significantly associated with a lower risk of fracture. MNA reflects multiple deficiencies, including proteins. This is a more general and easy tool to assess nutritional deficiencies in older adults compared to the dietary surveys that are more difficult to administer.

Regarding the underlying mechanisms linking malnutrition to falling and fracture, we could hypothesize that the effect of poor nutritional status on fracture relies on the loss of muscle strength [42] and that falls might mediate the relationship between malnutrition and the risk of fractures. However, it could be also due to a bone metabolism disturbance since Bischoff-Ferrari et al. have shown that a vitamin D supplementation decreased the risk of hip fracture and any non-vertebral fracture [43]. A study has found that a higher protein intake, rather than calcium or vitamin D, was a preventive factor of hip fracture in postmenopausal women [44]. Moreover, multiple intertwined biological mechanisms between the metabolisms of bone and muscle may underlie relationships between sarcopenia and frailty fractures [45]. There are many common pathogenic pathways leading to osteoporosis and sarcopenia including the decrease of anabolic hormone secretion, increase of inflammatory cytokine activity, and decreased physical activity. The decrease in muscle strength in sarcopenia leads to lower bone strength. Yu et al. has studied the association between sarcopenia and incident fracture among older men aged 65 years and over living at home [46] and found that sarcopenia was associated with an increased fracture risk.

While the strengths of our study are the large sample size of community-dwellers in relatively good health, not yet malnourished, and that were followed for more than a decade, there are also some limitations. By selecting subjects with no missing data in the covariates, we excluded individuals who were older, in poorer health, and with a higher risk of fracture. This may have resulted in possible underestimation of observed associations. In addition, the assessments of fall and fracture were self-reported and some events may have been missed. However, the frequency of falls, even if underestimated, appears high (45 % in men and 63 % in women) in this community-based sample. It is likely that only the most significant events were reported and reinforces the fact that poor nutritional status measured with MNA is associated with major clinical events.

In conclusion, this longitudinal population-based study suggests that a poor nutritional status is associated with a higher risk of falling and fractures in elderly community-dwellers. MNA could be a good indicator for the clinicians to target people at risk before malnutrition is established. Early and routinely screening of elderly people living at home and management of their nutritional status may be useful to reduce the frequency of these events. Further studies to identify the deficiencies of specific nutrients, the mechanisms involved, and the impact of nutritional intervention are needed.