Epidemiology of falls and osteoporotic fractures: a systematic review

Abstract

Background and methods: Fractures in elderly populations result from the combination of falls and osteoporosis. We report a systematic review of studies indexed in PubMed reporting annual rates of low-trauma falls and associated osteoporotic fractures among older community-dwelling people (age ≥ 50 years). An osteoporotic fracture was defined as either a fracture resulting from a low-impact fall in subjects with clinical osteoporosis, a fall resulting in an investigator-defined osteoporotic fracture, or a fall resulting in a low-trauma fracture. Rates are presented using descriptive statistics. Meta-analysis was conducted for statistically homogeneous data sets.

Results: The median (range) annual fall prevalence rates (median proportion of people who experienced one or more falls during the past year) for cohorts of women and men (10 determinations), women alone (seven determinations), and men alone (four determinations) were, respectively, 0.334 (0.217-0.625), 0.460 (0.372-0.517), and 0.349 (0.284-0.526). In studies that reported fall prevalence rates for Western men and women separately (four determinations), the pooled risk ratio (95% confidence interval [CI]) for men versus women was 0.805 (95% CI 0.721-0.900). The ranges of fall prevalence rates in East Asian women (two studies) and East Asian men (two studies) were, respectively, 0.163-0.258 and 0.087-0.184. The risk ratio (95% CI) for fall prevalence in East Asian men versus women was 0.634 (0.479-0.838) in studies (two determinations) reporting results for East Asian men and women separately. In cohorts of Western women and men (five determinations), the pooled rate (95% CI) of low-impact falls resulting in fractures was 0.041 (0.031-0.054). The proportion of low-trauma fractures attributable to falls among the Western community-dwelling elderly was within the range of 0.860-0.950 for fractures at all sites or the hip (five determinations). A range of 0.716-0.924 of all fractures were osteoporotic (eight determinations).

Conclusion: Fall rates are higher in women than in men in Western community-dwelling populations and lower in East Asian populations. Extrapolated to the US population, the statistics imply that low-impact falls cause approximately 0.53 million osteoporotic fractures annually among the US community-dwelling elderly.

Keywords: accidental falls; aged; bone density; fractures; osteoporosis.

Figure 1

Prevalence of low-impact falls in prospective cohorts of older community-based Western populations. Notes: The event rate shown is the proportion of subjects with a low-impact fall per year. Abbreviations refer to women and men combined (F, M), women (F), and men (M). The median (range) fall prevalence rates for women/men (10 studies), women (seven studies), and men (four studies) are, respectively, 0.334 (0.217–0.625), 0.460 (0.372–0.517), and 0.349 (0.284–0.526). The pooled mean fall prevalence rates (95% confidence interval) for women/men (10 studies), women (seven studies), and men (four studies) are, respectively, 0.316 (0.303–0.329), 0.426 (0.407–0.445), and 0.316 (0.289–0.344) in the fixed-effects model and 0.343 (0.293–0.398), 0.437 (0.395–0.479), and 0.351 (0.288–0.431) in the random-effects model. P values of the Q statistic for the pooled studies of women/men, women, and men were, respectively, 0.000, 0.000, and 0.001; corresponding I² values were 92.5%, 76.6%, and 81.2%. Abbreviation: CI, confidence interval.

Figure 2

Incidence of low-impact falls in prospective cohorts of older community-based Western populations. Notes: The median (range) fall incidence rates (95% confidence interval) for women/men (eight studies), women (five studies), and men (four studies) are, respectively, 0.530 (0.229–0.948), 0.687 (0.536–1.028), and 0.534 (0.368–1.630). Among studies of men, the fall incidence rate of 1.630 per patient per year reported by Studenski et al (a study of male veterans) is a statistical outlier. The median (range) for studies of men excluding Studenski et al (three determinations) is 0.428 (0.368–0.640). The pooled fall incidence rates (95% confidence interval) for women/men (eight studies), women (five studies), and men (four studies) are, respectively, 0.498 (0.476–0.519), 0.688 (0.669–0.706), and 0.557 (0.51–0.601) in the fixed-effects model and 0.574 (0.453–0.694), 0.710 (0.601–0.820), and 0.763 (0.327–1.199) in the random-effects model. P values of the Q statistic for the pooled studies of women/men, women, and men were, respectively, 0.000, 0.000, and 0.000; corresponding I² values were 93.9%, 96.4%, and 98.9%. The pooled fall incidence rate (95% confidence interval) in studies of men with Studenski et al excluded (three determinations) was 0.448 (0.402–0.494) in the fixed-effects model and 0.477 (0.309–0.645) in the random-effects model; P of Q 0.000, I² 91.6%. Abbreviation: CI, confidence interval.

Figure 3

Prevalence of low-impact falls in prospective cohorts of older community-based East Asian populations. Notes: The event rate is the proportion of subjects with a low-impact fall per year. Abbreviations refer to women and men combined (F, M), women (F), and men (M). The prevalence rate for the single study of women/men is 0.195. The ranges of prevalence rate values for women (two studies) and men (two studies) are, respectively, 0.163–0.258 and 0.087–0.184. The pooled mean fall prevalence rates (95% confidence interval) for women (two studies) and men (two studies) are, respectively, 0.206 (0.183–0.232) and 0.143 (0.121–0.169) in the fixed-effects model and 0.206 (0.128–0.315) and 0.129 (0.006–0.256) in the random-effects model. P values of Q statistic for the pooled studies of women, and men were, respectively, 0.000 and 0.000; corresponding I² values were 93.1% and 94.0%. Abbreviation: CI, confidence interval.

Figure 4

Incidence of low-impact falls in prospective cohorts of older community-based East Asian populations. Notes: The fall incidence rates in the two studies of women were 0.276 and 0.324; the median (range) of values in the three studies of men are 0.436 (0.434–0.771). The pooled fall incidence rates (95% confidence interval) for women (two studies) and men (three studies) are, respectively, 0.300 (0.271–0.347) and 0.215 (0.192–0.237) in the fixed-effects model, and 0.300 (0.253–0.347) and 0.261 (0.126–0.397) in the random-effects model. P values of Q statistic for the pooled studies of women and men were, respectively, 0.104 and 0.000; corresponding I² values were 62.3% and 97.0%. Abbreviation: CI, confidence interval.

Figure 5

Proportion of low-impact falls resulting in fractures in prospective cohorts of older community-based Western populations. Notes: Shown is the proportion of falls resulting in fracture(s) in prospective community-dwelling cohorts of women and men, where both the numbers of falls and resulting fractures were reported. The attribution of fractures as osteoporotic is based on a fall defined as low-trauma in an elderly population. The pooled event rate (95% confidence interval) for the six cohorts of women/men is 0.085 (0.074–0.098) in the fixed-effects analysis and 0.048 (0.026–0.087) in the random-effects analysis; P of Q 0.000, I² 88.8%. For the five cohorts of women/men after excluding Tinetti et al, which is a statistical outlier, the pooled event rate (95% confidence interval) is 0.041 (0.031–0.054) in the fixed-effects analysis and 0.040 (0.027–0.058) in the random-effects analysis; P of Q 0.154, I² 40.1%. Note that all the fractures in the Berg et al cohort occurred in women, and none occurred in men. Abbreviation: CI, confidence interval.