Rapid and early muscle wasting contributes to the significant long-term functional impairment observed in survivors of critical illness [1-4]. Although muscle and bone mass correlate in healthy individuals [5-7], there are limited data reporting the impact of critical illness on bone health. During acute critical illness, mechanical unloading as a consequence of bed rest [8], inflammation [9], acidaemia [10], vitamin D deficiency [11], corticosteroid use [12] and hypoxia [13] may all worsen bone health and reduce bone mineral density (BMD). Indeed, markers of bone turnover increase during critical illness [11,14]. Such bone demineralization may explain symptoms of generalised musculoskeletal pain in survivors of critical illness [1,2], and the reported increase in fracture risk in elderly females following critical illness [15]. However, the bone response to critical illness has never before been prospectively studied. We thus performed a pilot study to investigate the early effects of critical illness on BMD and fracture risk, and also sought clinical factors that might be associated with early bone demineralization.

Materials and methods

Subjects comprised a subgroup of the Musculoskeletal Ultrasound in Critical Illness: Longitudinal Evaluation cohort (trial registered with, identifier: NCT01106300) [4]. Ethical approval was obtained from the University College London ethics committee. Patients were recruited within 24 hours of admission to a university hospital (Kings College Hospital NHS Foundation Trust) or a community hospital intensive (Whittington Hospital NHS Trust) care unit (ICU) between August 2009 and April 2011. All were anticipated to be invasively ventilated for more than 48 hours, spend more than 7 days in the ICU and survive their ICU stay. Patients who were pregnant or suffering lower limb amputation, primary neuromuscular disease or cancer were excluded. At enrolment, written assent was obtained from the next-of-kin, with retrospective patient consent obtained when full mental capacity was regained.

BMD was assessed using dual X-ray absorptiometry (DXA) imaging on day 1 and day 10 (DXL Calscan, Demetech AB, Sweden), which has a coefficient of variance of 0.9%. Fracture risk was calculated from a change in T-score whereby the relative risk of a major osteoporotic fracture increases 1.5-fold (95% confidence interval (CI): 1.4 to 1.6) per standard deviation below the mean T-score [16]. Detailed clinical and physiological bedside data were collected, as previously described [4].

Statistical analysis

All data were assessed for normality using D’Agostino and Pearson omnibus normality tests, and analyzed using Student’s t-test, Pearson’s coefficient, Mann-Whitney U test and Wilcoxon’s signed rank tests, as appropriate. For the purposes of hypothesis generation in this pilot study, we sought a parsimonious model of associated physiological factors. Age, sex and chronic disease were forced into bivariable logistical regression (Statistical Package for the Social Sciences version 17; SPSS Inc., Chicago, IL, USA) using a threshold of 2% loss of BMD, which is twice the expected loss from bed rest alone [8]. Statistically significant independent variables from the bivariable analysis were entered into a backward multivariable analysis if the P value was 0.10 or less.


Fifty-seven patients assented to serial DXA scanning. Of these, seven did not survive ten days, one was transferred to another hospital, one withdrew from the study, one was discharged before day ten and one was unable to have serial scans for technical reasons. Forty-six patients were included in the final analysis. The characteristics of these 46 patients, shown in Table 1, did not differ from those withdrawn, except for a higher Simplified Acute Physiology Score (SAPS II) (40.9 (95% CI: 37.5 to 44.3, n = 46) versus 53.3 (95% CI: 44.2 to 62.4, n = 10); P <0.01). Four patients had pre-morbid conditions associated with possible disrupted calcium homeostasis (one with hypothyroidism, one with Crohn’s disease, two with hyperthyroidism). Their baseline DXA measurement was no different from the remaining cohort (0.387 ± 0.07 versus 0.437 ± 0.01, P = 0.350). Twenty eight patients (61%) were defined as having osteopenia on day one of the study; however, no patients received renal replacement therapy using citrate anticoagulation with calcium replacement, and no subjects received regular selective serotonin reuptake inhibitors or serotonin norepinephrine reuptake inhibitors (associated with lower BMD) prior to critical illness. Further data regarding recruitment and survival, as well as baseline laboratory values, are available in Additional file 1.

Table 1 Characteristics of patients who had serial measurements versus those who only had admission measurements

Clinical associations with change in bone mass density

BMD data were non-normally distributed prior to and following log transformation. There was no change in BMD between day 1 and 10 in the cohort overall (0.434 (95% CI: 0.405 to 0.463) versus 0.425 g/cm2 (95% CI: 0.399 to 0.450), P = 0.58). Multivariable logistical regression, adjusted for age, sex and chronic disease, was performed using a threshold of 2% loss of BMD. The model (overall model fit P <0.001, Hosmer-Lemeshow test P = 0.90, Table 2) demonstrated that admission calcium corrected for serum albumin (odds ratio (OR): 1.980 (95% CI: 1.089 to 3.609), P = 0.026) and the ratio of arterial partial pressure of oxygen (PaO2) to fraction of inspired oxygen (FiO2) on admission (OR: 0.916 (95% CI: 0.833 to 0.998), P = 0.044) were associated with a greater than 2% loss of BMD. Patient with acute respiratory distress syndrome (ARDS) (PaO2-to-FiO2 ratio of less than 300 mmHg [17] demonstrated greater BMD loss than those without (−2.81% (95% CI: −5.73 to 0.118%), n = 34 versus 2.40% (95% CI: 0.204 to 4.586%), n = 12, P = 0.029).

Table 2 Bivariable and multivariable logistical analysis of bedside physiology versus 2% loss of bone mineral density by day 10

Change in T-score

Admission T-score in those with ARDS did not differ from those without (−1.259 (95% CI: −1.737 to −0.781), n = 34 versus −1.325 (95% CI: −1.969 to −0.681), n = 12, P = 0.617), but ARDS patients had a change in T-score from day 1 to day 10 (−1.259 (95% CI: −1.737 to −0.781), n = 34 versus −1.518 (95% CI: −1.922 to −1.114), P = 0.047) compared to those without (−1.325 (95% CI: −1.969 to −0.681) versus −1.200 (95% CI: −1.798 to −0.602), P = 0.101). In the 34 patients with ARDS, fracture risk increased by 19.4% (95% CI: 13.9 to 25.0%) in the first 10 days of critical illness in comparison to those without (9.35% (95% CI: −2.1 to 20.9%), P = 0.012, Figure 1).

Figure 1
figure 1

Change in T-score and percentage increase in fracture risk in patients with (n = 34) and without (n = 12) acute respiratory distress syndrome. FiO2 = Fraction of inspired oxygen, PaO2 = Partial pressure of oxygen in blood. Mann-Whitney U test was performed between groups, *P <0.05.


Clinical relevance

The relationship between loss of BMD and rise in serum calcium concentration is to be expected. However, rapid early bone demineralization, of similar magnitude to that observed after much more prolonged weightlessness in space [18], occurs rapidly and early in critically ill patients with ARDS. The association of ARDS with bone demineralization is physiologically plausible: ARDS results in systemic release of inflammatory cytokines [19] such as interleukin-6 [20], TNFα [20,21], interleukin-1 [21] and interleukin-8 [22], which stimulate osteoclastogenesis and bone resorption with calcium mobilized into the circulation from these bone stores [23]. The scale of this loss is associated with an increase in calculated prospective fracture risk; an issue of some importance given that ARDS survivors have a number of independent risk factors for falls [24], including accelerated skeletal muscle wasting [4] and marked loss of executive function [25]. Bone is certainly able to respond rapidly to remodelling forces: changes are observed within 15 to 21 days in rodent models of unloading [26,27], and markers of bone turnover alter with single bouts of exercise in humans [28]. However, we are unaware of studies of the human skeletal response to illness over timeframes a short as that which we have addressed. Whilst in vitro evidence exists that hypoxia is detrimental to skeletal physiology and has both an inhibitory effect on osteoblastogenesis [29] and an activator-enhancing effect on osteoclastogenesis [30], no relationship was seen with admission hypoxia, although intermittent hypoxia as a stimulus cannot be excluded.


Whilst acknowledging that the first day of ICU admission is not the first day of critical illness, the median time to ICU admission was 24 hours, and 23 patients were admitted after a sudden acute event (for example trauma, myocardial infarction or intracranial bleed) with no antecedent decline. Nonetheless, these data should be considered hypothesis-generating. The limited sample size also precludes detailed exploration of other risk factors, such as osteopaenia, and generalisation to specific patient groups. Regrettably, we are unable to determine whether the observed changes translated into clinical skeletal events, given that this pilot study was not funded for the post-discharge follow-up of patients. Larger observational cohort studies, with extended follow-up periods, are required to determine the impact of critical illness on BMD and actual fracture risk. The determination of calcaneal BMD using DXA is valid and comparable to hip and spine DXA in determining fracture risk [16,31], and is more readily performed than the assessment of spine and hip bone densitometry in a remote imaging facility during the early stage of critical illness when the patient is most unstable. The coefficient of variance of DXA measures was 0.9%. A loss of more than 2% in calcaneal BMD (as we sought) was thus likely to represent true loss as opposed to measurement error.


Rapid bone demineralization, associated with an increase in fracture risk, was observed in critically ill patients with ARDS. More extensive and extended hypothesis-driven epidemiological cohort studies are required to confirm this finding, and to determine whether bone demineralisation represents a new therapeutic target in reducing morbidity following critical illness.

Key messages

  • Loss of bone mineral density occurs rapidly in critically ill patients with acute respiratory distress syndrome.

  • This is likely to be associated with an increase in fracture risk.