Background
Low back pain (LBP) contributed the highest years lived with disability among a total of 291 conditions investigated in the Global Burden of Disease 2010 study, resulting in 83 million years lived with disability [
1]. One in ten people suffer from LBP worldwide at any point in time [
1,
2] and 70–85% of people have a LBP episode at some time in their life [
2]. There has been an increase in disability, chronicity and work absenteeism attributable to this condition [
3], which has had an enormous negative economic impact on individuals, families, communities, industries and governments [
2,
4]. As such, understanding the aetiology and risk factors for LBP is important in reducing the significant burden accountable to this condition.
The prevalence of overweight and obesity that has contributed to 3.4 million deaths in 2010 is escalating in many countries and has been labelled as a global pandemic [
5]. Overweight and obesity are associated with several musculoskeletal diseases including LBP. A systematic review [
6] and afterwards a meta-analysis [
7] of data from small cross-sectional and prospective cohort studies showed that both overweight and obesity increased the risk of LBP. These findings were supported by the large population-based Nord-Trøndelag Health (HUNT) Study that included over 25,000 participants [
8]. Most of the studies included in the meta-analysis used body mass index (BMI) cutoff points to define overweight and obesity, while some studies used body weight and a few studies used waist circumference in combination with waist to hip ratio [
7]. However, no studies in the systematic review or the meta-analysis examined fat mass or distribution.
Although obesity is a risk factor for LBP in adults, the mechanism is unclear. Most research has used weight and BMI as a measure of obesity. However, these measures do not take into account body composition i.e. fat and muscle mass. There is increasing evidence for a differential effect of fat and muscle on the risk of musculoskeletal diseases [
9,
10]. For instance, there are negative effects of excess adiposity on movement patterns and on body structure [
11,
12] that contribute to different disabilities including LBP. Two previous cross-sectional studies examining the association between body composition and LBP have suggested an effect of fat mass on LBP. One study of 135 participants (83.1% women) found that greater fat mass, but not lean tissue mass, was associated with high levels of LBP intensity and disability [
13]. A Spanish study of 1128 twin women, with 38% unavailable data on fat mass, reported a weak association of fat percent and fat mass with LBP which was confounded by genetic and early shared environmental factors [
14]. Neither of these studies had longitudinal data. Similarly, both the studies included women only [
14] or women mostly [
13]. However, men and women differ substantially in regard to body composition, especially adipose tissue distribution [
15,
16]. Determining the gender-specific effect of body composition on LBP has the potential to improve our understanding of the mechanisms by which obesity affects LBP in men and women, and thus has the potential to provide novel approaches to managing this condition. Therefore the aim of this study was to examine the relationship between gender-specific body composition and LBP intensity and disability longitudinally in a national, population-based cohort of men and women.
Results
The characteristics of the participants are presented in Table
1. The prevalence of low-intensity LBP was 62% (n = 3085) and high-intensity LBP was 20% (n = 1001). Most of the participants had no LBP disability 73% (n = 3061), while 16% (n = 651) reported low disability and 11% (n = 482) high disability. Participants with high LBP intensity or disability were likely to be older, less educated, current smokers, from lower SEIFA and scored lower in the SF-36 MCS component compared to those who had no LBP intensity and disability. All obesity and body composition measurements were greater in those with LBP intensity or disability compared with those without LBP symptoms. All the obesity measures (BMI, waist circumference) and body composition measures (percent fat, fat mass and muscle mass) are highly correlated at baseline and at first follow-up (correlation coefficient >0.89,
p = <0.001) (Additional file
1: Table S2).
Table 1
General characteristics of participants
Low back pain intensitya
|
| No intensity n = 900 | Low intensity n = 3085 |
P
No vs low | High intensity n = 1001 |
P
No vs high |
Age at baseline, years | 48.8 (11.1) | 47.3 (10.9) | 0.001 | 51.1 (11.8) | <0.001 |
Female, n (%) | 497 (55.2) | 1682 (54.6) | 0.75 | 598 (60.1) | 0.03 |
Current smoker, n (%) | 89 (10.1) | 357 (11.2) | 0.26 | 182 (18.6) | <0.001 |
University degree, n (%) | 354 (39.6) | 1175 (38.4) | 0.54 | 226 (22.9) | <0.001 |
SEIFA (in lowest tertile), n (%) | 258 (29.2) | 918 (30.3) | 0.86 | 397 (40.4) | <0.001 |
MCS (SF-36) | 51.0 (8.5) | 48.9 (9.6) | <0.001 | 46.3 (11.2) | <0.001 |
BMI, kg/m2
| 26.0 (4.7) | 26.6 (4.8) | <0.001 | 27.8 (5.3) | <0.001 |
Waist circumference, cm | 87.9 (13.7) | 89.5 (13.7) | 0.002 | 92.7 (14.3) | <0.001 |
Percent of fat (%) | 31.7 (11.8) | 32.6 (11.6) | 0.03 | 36.2 (13.3) | <0.001 |
Fat mass, kg | 24.3 (12.4) | 25.7 (12.4) | 0.004 | 29.1 (14.6) | <0.001 |
Fat-free mass, kg | 50.6 (11.8) | 51.4 (12.2) | 0.10 | 49.3 (12.6) | 0.03 |
Low back pain disabilityb
|
| No disability n = 3,061 | Low disability n = 651 |
P
No vs low | High disability n = 482 |
P
No vs high |
Age at baseline, years | 46.5 (10.7) | 47.7 (11.5) | 0.01 | 51.4 (10.9) | <0.001 |
Female, n (%) | 1650 (54.0) | 352 (54.2) | 0.94 | 61.6 (61.2) | 0.002 |
Current smoker, n (%) | 335 (11.1) | 97 (15.6) | 0.01 | 108 (23.1) | <0.001 |
University degree, n (%) | 1244 (41.0) | 233 (36.1) | 0.02 | 98 (20.6) | <0.001 |
SEIFA (in lowest tertile), (%) | 888 (29.5) | 212 (33.0) | 0.20 | 194 (41.0) | <0.001 |
MCS (SF-36) | 49.3 (9.2) | 47.2 (10.6) | <0.001 | 46.2 (11.7) | <0.001 |
BMI, kg/m2
| 26.3 (4.7) | 27.2 (5.0) | <0.001 | 28.6 (5.8) | <0.001 |
Waist circumference, cm | 88.7 (13.5) | 91.2 (13.4) | <0.001 | 94.1 (14.9) | <0.001 |
Percent of fat (%) | 32.1 (11.5) | 33.7 (12.6) | 0.001 | 37.4 (13.3) | <0.001 |
Fat mass, kg | 25.0 (12.5) | 27.2 (14.0) | <0.001 | 30.6 (14.9) | <0.001 |
Fat-free mass, kg | 51.4 (12.3) | 51.4 (12.3) | 0.95 | 49.1 (12.5) | <0.001 |
Table
2 shows the association of LBP intensity with all anthropometric measures in men and women. All the obesity measures including BMI, waist circumference and body composition measures (percent fat, fat mass) were positively associated with LBP intensity and disability in both men and women after adjustment for age, education, smoking status, SEIFA and MCS of SF-36 (fat mass and fat-free mass were adjusted for each other). The associations between adiposity measures and high-intensity LBP were stronger than with lower-intensity LBP. Fat-free mass was negatively associated with LBP intensity only in women (low-intensity LBP: OR 0.72, 95% CI 0.54–0.96; high-intensity LBP: OR 0.75, 95% CI 0.56–1.01;
p for between group variability <0.001 (low vs high intensity). The association did not change after introducing physical activity to the model (data not shown).
Table 2
Relationship between obesity and body composition and intensity of low back pain
Model 1 |
Obesity measures |
BMI (per 5 kg/m2) | 1 | 1.24 (1.07–1.44) | 1.42 (1.18–1.71) | 0.08 | 1 | 1.17 (1.05–1.30) | 1.43 (1.26–1.62) | <0.001 |
Waist circumference (per 10 cm)a
| 1 | 1.11 (0.98–1.26) | 1.26 (1.08–1.47) | 0.02 | 1 | 1.14 (1.04–1.25) | 1.38 (1.24–1.54) | <0.001 |
Body composition measures |
Percent fat (per 10%) | 1 | 1.27 (1.08–1.49) | 1.45 (1.20–1.76) | 0.06 | 1 | 1.11 (1.00–1.23) | 1.41 (1.25–1.59) | <0.001 |
Fat mass (per 10 kg)a
| 1 | 1.10 (0.96–1.26) | 1.22 (1.04–1.42) | 0.01 | 1 | 1.10 (1.01–1.19) | 1.28 (1.16–1.41) | <0.001 |
Fat-free mass (per 10 kg)a
| 1 | 1.20 (0.98–1.48) | 1.20 (0.93–1.56) | 0.69 | 1 | 1.18 (0.94–1.48) | 0.72 (0.54–0.96) | <0.001 |
Model 2 |
Obesity measures |
BMI (per 5 kg/m2) | 1 | 1.25 (1.07–1.45) | 1.41 (1.17–1.70) | 0.01 | 1 | 1.17 (1.05–1.30) | 1.39 (1.22–1.59) | <0.001 |
Waist circumference (per 10 cm)a
| 1 | 1.11 (0.98–1.27) | 1.25 (1.07–1.46) | 0.06 | 1 | 1.13 (1.03–1.24) | 1.36 (1.22–1.52) | <0.001 |
Body composition measures |
Percent fat (per 10%) | 1 | 1.28 (1.09–1.51) | 1.45 (1.19–1.77) | 0.01 | 1 | 1.41 (1.25–1.59) | 1.39 (1.22–1.57) | <0.001 |
Fat mass (per 10 kg)b
| 1 | 1.11 (0.97–1.27) | 1.23 (1.05–1.44) | 0.18 | 1 | 1.28 (1.16–1.41) | 1.27 (1.15–1.40) | <0.001 |
Fat-free mass (per 10 kg)b
| 1 | 1.19 (0.97–1.47) | 1.15 (0.88–1.50) | 0.77 | 1 | 0.72 (0.54–0.96) | 0.75 (0.56–1.01) | <0.001 |
The association of LBP disability with all obesity and body composition measures is shown in Table
3. Similar negative associations for all the obesity measures and body composition measures except fat-free mass were observed for LBP disability in both men and women after adjustment for age, education, smoking status, SEIFA and MCS of SF-36 (fat mass and fat-free mass were adjusted for each other). The association did not change after introducing physical activity or change in obesity and body composition measures to the model (data not shown).
Table 3
Relationship between obesity and body composition and low back disability
Model 1 |
Obesity measures |
BMI (per 5 kg/m2) | 1 | 1.04 (0.88–1.22) | 1.44 (1.20–1.74) | 0.002 | 1 | 1.25 (1.13–1.39) | 1.49 (1.33–1.66) | 0.01 |
Waist circumference (per 10 cm)a
| 1 | 1.15 (1.00–1.32) | 1.36 (1.17–1.61) | 0.05 | 1 | 1.23 (1.12–1.35) | 1.39 (1.27–1.54) | 0.03 |
Body composition measures |
Percent fat (per 10%) | 1 | 1.12 (0.94–1.34) | 1.42 (1.14–1.77) | 0.05 | 1 | 1.21 (1.08–1.35) | 1.52 (1.35–1.72) | 0.002 |
Fat mass (per 10 kg)b
| 1 | 1.08 (0.94–1.24) | 1.30 (1.11–1.52) | 0.03 | 1 | 1.18 (1.09–1.28) | 1.34 (1.23–1.47) | 0.02 |
Fat-free mass (per 10 kg)b
| 1 | 0.92 (0.73–1.17) | 0.86 (0.65–1.14) | 0.66 | 1 | 1.05 (0.82–1.33) | 0.89 (0.69–1.16) | 0.28 |
Model 2 |
Obesity measures |
BMI (per 5 kg/m2) | 1 | 1.02 (0.86–1.21) | 1.39 (1.15–1.68) | 0.004 | 1 | 1.25 (1.12–1.39) | 1.45 (1.29–1.62) | 0.02 |
Waist circumference (per 10 cm)a
| 1 | 1.13 (0.98–1.30) | 1.32 (1.12–1.56) | 0.08 | 1 | 1.23 (1.11–1.35) | 1.37 (1.23–1.52) | 0.04 |
Body composition measures |
Percent fat (per 10%) | 1 | 1.11 (0.92–1.32) | 1.37 (1.10–1.72) | 0.07 | 1 | 1.20 (1.07–1.35) | 1.48 (1.31–1.68) | 0.004 |
Fat mass (per 10 kg)b
| 1 | 1.07 (0.93–1.23) | 1.28 (1.09–1.51) | 0.04 | 1 | 1.18 (1.09–1.28) | 1.32 (1.21–1.45) | 0.04 |
Fat-free mass (per 10 kg)b
| 1 | 0.93 (0.73–1.18) | 0.85 (0.63–1.13) | 0.61 | 1 | 1.09 (0.85–1.39) | 0.93 (0.72–1.22) | 0.29 |
BMI, waist circumference, percent fat and fat mass were significantly and positively related to LBP intensity and disability for both men and women for sex-specific quartile cutoff points of these measures after adjustment for confounders (Table
4,
p <0.04). For example, the adjusted OR for women for fat mass in relation to LBP intensity increased from quartile 1 to quartile 4 (Q1: reference category, Q2: OR 1.05, 95% CI 0.84–1.32; Q3: OR 1.25, 95% CI 1.00–1.57; and Q4: OR 1.78, 95% CI 1.42–2.24;
p <0.001). A similar trend was observed for LBP disability (Q1: reference category, Q2: OR 1.07, 95% CI 0.79–1.45; Q3: OR 1.37, 95% CI 1.03–1.83; and Q4: OR 2.30, 95% CI 1.74–3.04;
p <0.001). In contrast, fat-free mass was negatively associated with LBP intensity in women. No such association was observed for either LBP disability in women, or for LBP intensity or LBP disability in men.
Table 4
Low back pain intensity and disability in relation to measures of sex-specific quartiles
Low back pain intensity, OR (95% CI) |
Men |
Obesity measures |
Body mass index | 1 | 1.47 (1.14–1.88) | 1.75 (1.36–2.25) | 1.50 (1.16–1.93) | 0.001 |
Waist circumferencea
| 1 | 1.43 (1.10–1.84) | 1.78 (1.36–2.34) | 1.50 (1.12–2.00) | 0.003 |
Body composition measures |
Percent fat | 1 | 1.58 (1.23–2.03) | 1.64 (1.28–2.12) | 1.61 (1.25–2.08) | <0.001 |
Fat massb
| 1 | 1.78 (1.38–2.30) | 1.91 (1.46–2.51) | 1.51 (1.14–1.99) | 0.02 |
Fat-free massb
| 1 | 0.88 (0.68–1.14) | 0.89 (0.68–1.16) | 1.00 (0.76–1.33) | 0.43 |
Women |
Obesity measures |
Body mass index | 1 | 1.25 (1.01–1.56) | 1.25 (1.00–1.56) | 1.92 (1.53–2.41) | <0.001 |
Waist circumferencea
| 1 | 1.20 (0.96–1.50) | 1.42 (1.13–1.78) | 2.09 (1.65–2.65) | <0.001 |
Body composition measures |
Percent fat | 1 | 1.01 (0.81–1.26) | 1.21 (0.96–1.51) | 1.79 (1.42–2.24) | <0.001 |
Fat massb
| 1 | 1.05 (0.84–1.32) | 1.25 (1.00–1.57) | 1.78 (1.42–2.24) | <0.001 |
Fat-free massb
| 1 | 0.85 (0.68–1.06) | 0.94 (0.75–1.17) | 0.79 (0.63–0.99) | 0.01 |
Low back pain disability, OR (95% CI) |
Men |
Obesity measures |
Body mass index | 1 | 1.12 (0.82–1.53) | 1.13 (0.83–1.54) | 1.31 (0.96–1.79) | 0.001 |
Waist circumferencea
| 1 | 1.59 (1.15–2.21) | 1.29 (0.91–1.83) | 1.85 (1.28–2.66) | 0.003 |
Body composition measures |
Percentage fat | 1 | 1.12 (0.82–1.54) | 1.20 (0.87–1.64) | 1.37 (1.00–1.88) | 0.04 |
Fat massb
| 1 | 1.20 (0.87–1.66) | 1.26 (0.90–1.76) | 1.45 (1.02–2.05) | 0.04 |
Fat-free massb
| 1 | 1.00 (0.73–1.37) | 0.75 (0.54–1.05) | 0.78 (0.55–1.12) | 0.07 |
Women |
Obesity measures |
Body mass index | 1 | 1.36 (1.01–1.83) | 1.58 (1.18–2.12) | 2.56 (1.93–3.40) | <0.001 |
Waist circumferencea
| 1 | 1.16 (0.86–1.58) | 1.53 (1.15–2.06) | 2.50 (1.86–3.34) | <0.001 |
Body composition measures |
Percent fat | 1 | 1.08 (0.80–1.45) | 1.49 (1.12–1.99) | 2.15 (1.63–2.85) | <0.001 |
Fat massb
| 1 | 1.07 (0.79–1.45) | 1.37 (1.03–1.83) | 2.30 (1.74–3.04) | <0.001 |
Fat-free massb
| 1 | 1.20 (0.91–1.58) | 1.16 (0.88–1.54) | .92 (0.69–1.22) | 0.86 |
Repeating the analysis in Tables
2,
3 and
4 including only those with no bodily pain at baseline (n = 3961) yielded similar results (data not shown).
Discussion
The results from this large, prospective, population-based cohort study showed that obesity measures (BMI and waist circumference), percent fat, and fat mass were positively associated with LBP intensity and disability, independent of fat-free mass. With increasing sex-specific quartiles of obesity measures, fat percent and fat mass, the risk of LBP intensity and disability increased in a linear manner. Fat-free mass was negatively associated with LBP intensity in women only, with LBP intensity reduced in a linear manner in relation to increasing sex-specific quartiles of fat-free mass, independent of the potential confounders and fat mass.
The results of the current study, that LBP is associated with overweight and obesity measures, are consistent with a recently published systematic review [
6] and a meta-analysis [
7] of mostly cross-sectional studies and the results from the large HUNT cross-sectional study [
8]. The present study extends these findings by providing evidence from a longitudinal study, thus strengthening the evidence for this relationship. In addition we also demonstrated that obesity measures are associated with both LBP intensity and disability and that the strength of association increases with the severity of LBP, so a dose–response relationship.
The current study demonstrated a longitudinal relationship between fat mass and percent fat and LBP. The only previous data were from two cross-sectional studies [
13,
14], including one that was a twin study [
14]. Furthermore, this study showed a dose–response relationship of increased LBP intensity and disability in relation to increasing quartile of fat percent and fat mass measures. In contrast, the effect of fat-free mass was less consistent, with some evidence for reduced fat-free mass as a risk factor of LBP intensity in women. While most previous studies examining the association between muscle mass and LBP did not report any significant findings [
25‐
27], only one study reported that reduced trunk and lower extremity muscle mass were associated with chronic LBP measured by a negative straight leg raise test, in a subgroup of 71 women [
28]. The straight leg raise test has been reported to have limited diagnostic accuracy [
29]. Moreover, the study did not adjust for potential confounders, particularly the independent effect of muscle and fat mass, even though both are strongly associated with BMI and individually may simply reflect obesity. Though there is a paucity of evidence, it is possible that fat-free mass may have a beneficial effect on reducing LBP through increasing back muscle strength and back muscle endurance, which ultimately improve spinal health [
30]. However, this needs further exploration.
By imposing continuous high biomechanical loading on the intervertebral disc of the lower back, obesity may result in structural modification of the disc [
31,
32] that leads to greater low back pain and/or disability. Increased muscle mass having an inconsistent effect on LBP intensity and disability, and increased fat mass associated with LBP intensity and disability both in men and women suggest that the effect of obesity in the pathogenesis of LBP is mainly mediated through fat mass, despite differences in degrees of obesity. The mechanism of involvement of fat mass in the pathogenesis of LBP is likely to be multifactorial: it occurs via excess fat placing increased load on the spine; alternatively, the effect of fat mass may occur through systemic processes. There is evidence that fat mass is metabolically active and may adversely affect structures through systemic inflammatory processes induced by pro-inflammatory molecules i.e. tumour necrosis factor, adiponectin and interleukins released by adipose tissue [
33] as well as decreased nutrition to the intervertebral disc via atherosclerosis [
34]. Indeed, fat mass or adipose tissue secret pro-inflammatory cytokines, such as tumor necrosis factor [
35] that have been shown to be more prevalent in people with modic change [
36] and muscle catabolism [
37]. Likewise, in patients with chronic LBP, interleukins were present in abundance in those with mild disc degeneration diagnosed by MRI [
38]. It is possible that changes in disc height, modic changes and muscle destruction play a distinctive role in the degeneration of tissue surrounding the spine and thus promote development of chronic pain conditions [
13]. This is supported by the finding that C-reactive protein, a marker of chronic systemic inflammation, was associated with the odds of reporting LBP [
39]. Moreover, atherosclerosis, largely driven by chronic inflammation [
40], may decrease blood supply to the lumbar region, in the vertebrae and surrounding muscles [
34]. These might result in reduced nutrition of the lumber intervertebral disc and result in tissue degeneration [
34,
41]; which is in the pathway of LBP. For example, aortic atherosclerosis was shown to increase the risk for development of disc degeneration and was associated with the occurrence of LBP in participants of the Framingham cohort after 25-year follow-up [
41]. Furthermore, central and peripheral pain sensitization also has a role in LBP [
42,
43]. There is increasing evidence that systemic inflammation affects central and peripheral pain sensitization [
44], which may be another mechanism for the adiposity-related meta-inflammatory processes [
45]. Furthermore, it has been suggested that LBP could also cause less physical activity and thereby result in obesity, however in our analysis, to eliminate the effect of less physical activity, we have adjusted all our analyses.
The results of our study should be considered within the context of its limitations. We have used bio-impedance to measure body composition. Bio-impedance is fast, inexpensive, and does not require extensive operator training or cross-validation [
46], however, bio-impedance depends on body hydration, which is difficult to assess and has a strong effect on the estimation of fat mass based on bioelectric impedance analysis [
47]. As a result, any between-subject variability in hydration level in this study would have resulted in attenuation of the observed association between body composition, and LBP intensity and disability. Depression has been shown to predict the development of LBP and obesity [
48], thus it is a confounding factor. Though we do not have data on depression as part of the AusDiab study, we have controlled our analysis for the mental component score of the SF-36, which can correctly identify 87% of cases of depression [
49]. Furthermore, participants who responded to the Chronic Pain Grade Questionnaire had better health and were of lower socio-economic status compared with those who did not respond to the questionnaire, which is likely to have underestimated the association between measures of adiposity and LBP we observed. It is also possible that residual confounding may have accounted for the associations. However in this study, we were able to adjust for most of the well-known risk factors for LBP. Strengths of our study include the large sample size and wide age range of the cohort, and use of a validated measure of LBP intensity and disability. Furthermore, in this longitudinal study we have performed subgroup analysis leaving those experiencing moderate to very severe bodily pain at baseline and have observed the similar effect. This rules out the possibility of reverse causality that suffering from LBP at study induction may have been responsible for elevated adiposity measures.
Acknowledgements
The AusDiab study, co-coordinated by the Baker IDI Heart and Diabetes Institute, gratefully acknowledges the support and assistance given by: B Atkins, E Barr, A Cameron, S Chadban, M de Courten, D Dunstan, S Murray, N Owen, S Tanamas, T Welborn, P Zimmet and all the study participants. We would also like to thank Dr Stephane Heritier for his suggestions on statistical analysis.
Funding
For the Australian Diabetes, Obesity and Lifestyle (AusDiab) Study, funding support has been received from the National Health and Medical Research Council (NHMRC grant 233200), Australian Government Department of Health and Ageing, Abbott Australasia Pty Ltd, Alphapharm Pty Ltd, AstraZeneca, Bristol-Myers Squibb, City Health Centre-Diabetes Service-Canberra, Department of Health and Community Services – Northern Territory, Department of Health and Human Services – Tasmania, Department of Health – New South Wales, Department of Health – Western Australia, Department of Health – South Australia, Department of Human Services – Victoria, Diabetes Australia, Diabetes Australia Northern Territory, Eli Lilly Australia, Estate of the Late Edward Wilson, GlaxoSmithKline, Jack Brockhoff Foundation, Janssen-Cilag, Kidney Health Australia, Marian and FH Flack Trust, Menzies Research Institute, Merck Sharp & Dohme, Novartis Pharmaceuticals, Novo Nordisk Pharmaceuticals, Pfizer Pty Ltd, Pratt Foundation, Queensland Health, Roche Diagnostics Australia, Royal Prince Alfred Hospital, Sydney, Sanofi Aventis, sanofi-synthelabo, and the Victorian Government’s Operational Infrastructure Support (OIS) Program.
SMH is the recipient of AFA-ARA Heald Fellowship funded by the Australian Rheumatology Association. DU, YW and AEW are the recipients of an NHMRC Career Development Fellowship (Clinical level 1, #1011975 Clinical level 1, #1065464, and Clinical level 2, #1063574 respectively). JES is supported by an NHMRC Research Fellowship. DWD is supported by an NHMRC Senior Research Fellowship.