Department of Public Health and Primary Care, Institute of Public Health, School of Clinical Medicine, University of Cambridge, Cambridge, United Kingdom
Address correspondence to: Address correspondence to: Kay-Tee Khaw, FRCP, University of Cambridge School of Clinical Medicine, Clinical Gerontology Unit, Box 251, Addenbrooke's Hospital, Cambridge CB2 2QQ, UK
The authors state that they have no conflicts of interest.
Identification of those at high risk of osteoporosis and fractures using clinically available tests beyond BMD measures is a major clinical challenge. We examined forced expiratory volume in 1 s (FEV1), an easily obtainable measure of respiratory function, as a clinical measure for fracture prediction. In the context of the European Prospective Investigation into Cancer-Norfolk Study, 8304 women and 6496 men 42–81 yr of age underwent a health check including spirometry and heel quantitative ultrasonography between 1997 and 2000 and were followed up for incident hip fractures until 2007. The main outcome measures were broadband ultrasound attenuation (BUA) of the heel (cross-sectional analysis) and hip fracture risk (prospective analysis). In multivariate regression models, a 1-liter increase in FEV1 was associated with a statistically significant 2.2-dB/MHz increase in BUA, independent of age, smoking, height, body mass index, history of fracture, and use of corticosteroids. Mean FEV1 was significantly lower among 84 women and 36 men with hip fracture compared with other participants. In multivariate proportional-hazard regression models, the relative risk (RR) of hip fracture associated with a 1-liter increase in FEV1 was 0.5 (95% CI, 0.3–0.9; p < 0.001) for both men and women. RR of hip fracture for a 1 SD increase in FEV1 was approximately equivalent to a 0.5 SD increase in BUA among women (1 SD among men) and an ∼5-yr decrease in age among both men and women. Middle-aged and older people with low respiratory function are at increased risk of osteoporosis and hip fracture. FEV1, an easy, low-cost, and feasible clinical measure, may help improve the identification of high-risk groups.
The aging of the population is a worldwide issue. In the United States, for instance, the number of people ≥65 yr of age is projected to increase from 31 million in 1990 to 77 million in 2040, and the relative growth rate is even faster for people >85 yr of age.(1) The number of elderly individuals is increasing even more rapidly in the developing countries of Asia, the Middle East, Africa, and South America. Bone fractures, resulting from osteoporosis and increased risk of falls, are a leading cause of disability among the elderly.(2) A recent World Health Organization report indicated that osteoporotic fractures are responsible for substantial costs related to hospitalizations, surgery, outpatient care, long-term care, disability, and premature death.(3) Projections suggest that, in the next few decades, numbers of fractures worldwide are likely to increase substantially.(4) Therefore, early identification of groups at high risk of fracture who may benefit most from preventive interventions is a major challenge.(5)
Although low BMD is an established predictor of increased fracture risk, the majority of fractures occur in patients with BMD above the thresholds commonly used to diagnosis osteoporosis. Identification of other factors that independently predict fracture risk may not only help improve identification of high-risk groups but also help understanding of the pathophysiology of the disease. A number of previous studies have suggested a link between respiratory function and BMD.(6–8) Some pathophysiologic mechanisms also plausibly support an association between pulmonary function and bone health. Apart from demographic and anthropometric factors such as age, sex, and height, this association can be mediated through modifiable behavioral risk factors, namely physical activity and smoking.(9–12) In this study, we investigated whether pulmonary function testing is associated with bone characteristics (as assessed by quantitative ultrasound measurement) and prospective risk of hip fracture.
MATERIALS AND METHODS
Study participants and measurements
Participants in this population-based cross-sectional and cohort study were recruited as part of the East Anglian component of the European Prospective Investigation into Cancer-Norfolk (EPIC-Norfolk) study. EPIC is a multicenter collaborative study designed to investigate the relationships among diet, cancer, and chronic diseases.(13) As described elsewhere,(14) the scope of EPIC-Norfolk was extended to exposures other than diet and outcomes other than cancer. The original cohort was comprised of 25,623 men and women 40–79 yr of age, randomly recruited between 1993 and 1997 from general practice age-sex registers in Norfolk region. All participants are being followed up for different health endpoints, including fractures, to the present. The study was approved by the Norwich District Health Authority ethics committee, and all participants signed an informed consent at the beginning of the study.
All participants in the original cohort were invited for a second health examination between 1997 and 2000. A self-administered detailed health and lifestyle questionnaire was completed by 15,028 returning participants, and all of them were examined by trained nurses. Respiratory function was assessed by forced expiratory volume in 1 s (FEV1) using an electronic turbine spirometer (Micro Medical, Rochester, UK). After a practice blow, two measurements were made with the subjects standing and looking forward. The nurses made a subjective judgment of the participants' spirometry technique. The higher of the two values for FEV1 was used for analysis. Forced vital capacity (FVC) and peak expiratory flow (PEF) were also recorded for all participants, but only FEV1 is reported here because the other measures did not add information beyond FEV1. The machine was chosen for portability and simplicity in operation. The reproducibility was ∼2.2% for FEV1, and the device is assessed as having a comparable accuracy to the Vitalograph spirometer.(15) Calibration was performed regularly in a weekly basis to ensure the accuracy and precision of both equipment and personnel.(14)
Height and weight were measured in light clothing without shoes. Height was measured to the nearest millimeter using a free-standing stadiometer (CMS Weighing Equipment, London, UK). Weight was measured to the nearest 100 g using Salter digital scales (Salter Industrial Measurement, West Bromwich, UK). Body mass index (BMI) was calculated as the weight in kilograms divided by the square of the height in meters. Smoking status was derived from the questionnaires, and individuals were categorized as current, former, or never smokers. Total alcohol consumption was estimated from the questionnaires as the total units of drinks (∼8 g of alcohol) consumed in a week.(14) Current or ever use of corticosteroid drugs, bronchodilators, and hormone therapy (HRT), as well as history of respiratory diseases, were also derived from the questionnaires.
Quantitative ultrasound scanning was used to measure broadband ultrasound attenuation (BUA; db/MHz) and speed of sound (SOS; m/s) of the calcaneus with the use of the CUBA sonometer (McCue Ultrasonics, Winchester, UK) at least twice on each foot, as described elsewhere.(16,17) The mean of the measures (left and right foot) was used for analysis. Five machines were used, and each was calibrated daily with its physical phantom and monthly with a roving phantom and on one operator's calcaneus. The CV was 3.5%. Because adjustment for the effect of ambient temperature, machine, or machine drift did not materially influence results, the data are presented unadjusted for these variables.(17)
Participants who were admitted to the hospital were identified using their unique National Health Service (NHS) number by data linkage with ENCORE (East Norfolk health authority database), which identifies all hospital contacts throughout England and Wales for Norfolk residents. International Classification of Diseases (ICD)9 and 10 diagnostic codes were used to ascertain fractures by site occurring in the cohort up to the end of March 2007 for these analyses; mean follow-up time was 7.7 ± 0.8 (SD) yr.
Because bone characteristics differ considerably between men and women, sex-specific analyses were used throughout the paper. For assessment of the association between FEV1 and QUS measures, characteristics of participants in four sex-specific quartiles of FEV1 were compared using one-way ANOVA for continuous variables and χ2 test for categorical variables. Assumption of normality was checked beforehand for all continuous variables. Pearson's correlation coefficients were estimated for the correlations between FEV1 and ultrasound characteristics of the participants. Univariate general linear models were used to assess the linear trend of crude and adjusted BUA in different quartiles of FEV1. To predict the sex-specific difference in BUA, multivariate linear regression models were run with FEV1 with different levels of adjustment. The Wald test was used to test the significance of β coefficients. Prespecified interactions between FEV1 and other factors were checked. Regression models were rerun for different subgroups of participants.
To assess the predictive power of FEV1 for incident osteoporotic fractures, characteristics of those who had developed hip fracture during the follow-up were compared with other participants. FEV1 and known fracture risk factors were entered into a Cox proportional hazards model to determine their independent contribution to the risk of fracture. HRs of hip fracture for sex-specific quartiles of FEV1 were calculated in comparison with the lowest quartiles for men and women. FEV1 was also entered into models as a continuous variable. To enable comparisons between FEV1 and other continuous variables for prediction of fractures, we used intervals of ∼1 SD (0.5 liter). Goodness-of-fit for different models were verified graphically by comparison of Kaplan-Meier curves for observed and predicted values. A set of prespecified interactions between FEV1 and other factors was also checked but not included in the final models because of nonsignificance. Values for continuous variables are expressed as mean ± SD throughout the paper unless otherwise stated. All the analyses were performed using Stata software, version 10.0 (StataCorp, College Station, TX, USA).
Characteristics of the study participants
After exclusion of 228 participants with unsatisfactory spirometry (because of mechanical problems, poor cooperation, coughing, or recent abdominal or chest surgery), 8304 women and 6496 men 42–81 yr of age comprised the study population. Characteristics of the study population are summarized in Table 1. Mean (SD) of FEV1 was 2.1 (0.5) liters among women and 2.6 (0.7) liters among men. Men had significantly higher bone measures (both BUA and SOS) and experienced a lower number of hip fractures during follow-up. Table 1 shows the significant differences between women and men regarding key variables, supporting the need for sex-specific analyses.
Table Table 1.. Baseline Characteristics of Participants in the European Prospective Investigation Into Cancer-Norfolk Study, 1997–2000
Respiratory function and quantitative ultrasound measures
Significant and positive correlations were observed between FEV1 and BUA among both women (Pearson correlation coefficient r = 0.32; p < 0.001) and men (r = 0.11; p < 0.01). The corresponding coefficients for FEV1 and SOS were 0.26 for women and 0.08 for men (p < 0.01 for both). Age, height, and weight also significantly correlated with both FEV1 and ultrasound measures. Given the high correlation of BUA and SOS in both women (r = 0.72) and men (r = 0.69), further analyses used BUA only as the measure of bone health.
Fugure 1 shows the crude and adjusted means of BUA using the generalized linear modeling approach among different quartiles of FEV1 in both sexes. Multivariate-adjusted mean BUA was higher by 3.7 dB/MHz among women and 2.9 dB/MHz among men from the first to fourth quartiles of FEV1. Although the magnitude of the difference was reduced after adjustment, there was still a significant linear trend for increment of BUA across quartiles of FEV1. The trend of increasing BUA with increasing FEV1 quartiles was more noticeable among women.
Results of the multivariate linear regression models to predict heel BUA are summarized in Table 2. A significant and positive relationship between FEV1 and BUA, independent of age, smoking, height, BMI, history of fracture, and use of corticosteroids was observed among both men and women. In the multivariate models, a unit change in FEV1 (1 liter) was associated with a statistically significant 2.21-dB/MHz difference in BUA among women and 1.47-dB/MHz difference in BUA among men. Excluding participants with self-reported respiratory disease or use of corticosteroids or bronchodilators did not materially alter the regression slopes (Table 2).
Table Table 2.. Crude and Adjusted Regression Coefficients of FEV1 for Prediction of Calcaneal BUA in the European Prospective Investigation Into Cancer-Norfolk Study
Respiratory function and hip fractures
A total of 120 participants (84 women) developed a hip fracture during 114,346 person-years of follow-up (64,049 person-years in women). Characteristics of participants who did or did not develop a hip fracture in the study period are summarized in Table 3. Women with subsequent fractures were significantly older, shorter, and lighter and had significantly lower ultrasound measures and FEV1 (1.7 liters on average compared with 2.1 liters for others). Women with hip fractures were less likely to have used hormone therapy, more likely to have a history of fracture earlier in life, and had a lower intake of alcoholic drinks. Smoking was not different among these groups in women (Table 3). Age, past history of fracture, FEV1, and ultrasound measures were significantly different between men with and without subsequent hip fracture (Table 3). Mean age of men with hip fracture was ∼8 yr higher than other participants, and they had a lower FEV1 of ∼0.7 liters on average compared with the others.
Table Table 3.. Comparison of Characteristics of Participants With and Without Subsequent Hip Fracture at the European Prospective Investigation Into Cancer-Norfolk Study
Table 4 shows the results of Cox regression models to predict hip fracture among participants. There was a trend of decreasing risk of hip fracture in subjects with higher FEV1. Multivariate models showed a significant reduction of ∼47% in hip fracture risk per 1 liter increase in FEV1 in both sexes (Table 4, right column).
Table Table 4.. Cox Regression Analyses by FEV1 (95% CIs) for Hip Fractures, the European Prospective Investigation Into Cancer-Norfolk Study
Table 5 shows the results of sex-specific multivariate Cox regression analyses to predict hip fracture. FEV1 was a significant predictor of hip fractures among both men and women, with an HR of ∼0.6/1 SD (0.5 liters). Age, height, alcohol intake, and BUA were the other significant predictors of hip fractures among women. The other significant predictors were age and BUA among men. A history of fracture was associated with a marginally significant 70% increased risk of hip fracture among women and a nonsignificant 130% increased risk among men (Table 5).
Table Table 5.. Cox Proportional Hazard Models to Predict Hip Fracture Among 8304 Women and 6496 Men, the European Prospective Investigation Into Cancer-Norfolk Study
To our knowledge, this is the first population-based study evaluating the association between respiratory function and bone health as assessed by quantitative ultrasound measurement and fracture incidence over time. In our study, there was a significant positive and continuous relationship between FEV1 and BUA of the heel in middle-aged and older women and men. The magnitude of this relationship, however, was not large after adjustment for covariates; the mean BUA measures of women and men in the highest FEV1 quartile were only 5% and 3%, respectively, higher than the mean BUA of women and men in the lowest quartile (Fig. 1). Furthermore, in multiple regression analysis, a 1-liter increase in FEV1 was accompanied by an ∼2.2 dB/MHz increase in BUA for women and a 1.5 dB/MHz increase in BUA for men (Table 2) in comparison with an SD of BUA of ∼17 dB/MHz.
However, there was a significant and strong association between FEV1 and incidence of hip fracture, which was greater than might be predicted from the association with BUA. Indeed, this association remained significant even after adjustment for BUA and other known risk factors including age and history of fracture (Tables 4 and 5). The HR for a 1 SD (0.5 liter) increase in FEV1 was ∼0.6 (95% CI, 0.4–0.9) for both men and women. β coefficients provided in Table 5 can be used to compare the relative effect of different variables for prediction of hip fractures.(18) This shows that a 1 SD increase in FEV1 was equivalent to an ∼0.5 SD increase in BUA among women (1 SD among men) and an ∼5-yr decrease in age among both men and women (Table 5). This suggests that the relationship between respiratory function and bone health is independent of bone characteristics measured by quantitative ultrasound, and FEV1 may be a useful marker of fracture risk independent of bone characteristics in older men and women. Evaluation of the mechanisms by which FEV1 can affect the bone health is beyond the scope of this study, but we can suggest that inclusion of this measure (FEV1) in fracture prediction charts, especially for hip fracture, might be helpful and needs further consideration.
Currently we are facing a universal shift toward use of long-term fracture risk estimation in the field of osteoporosis research and clinical practice guidelines. The FRAX tool, a newly launched online program for estimation of 10-yr absolute risk of fracture for individuals, is likely to be a source for future routine clinical practice in this field.(3,19) This tool currently considers several clinical risk factors and BMD measurements in the femoral neck. The results of this tool can be replicated for different populations using prospective studies with long follow-ups. Moreover, other potential risk factors (including clinical, radiological, and biochemical factors) can be added to or replaced with the current set of risk factors. Whereas use of subjective measures such as history of smoking or physical activity might be prone to several biases, more objective measures such as spirometry results may increase the accuracy of our risk estimates. Future studies need to consider this point and use it to improve the predictive power of forthcoming risk assessment tools.
The first studies examining the association between respiratory function and bone health were in patients with pulmonary diseases. Some clinical studies in patients with cystic fibrosis and bronchial asthma found significant associations between measures of respiratory function and BMD.(20–23) It should be noted, however, that patients with these conditions are exposed to a variety of other factors that might impair their bone health (i.e., cystic fibrosis is associated with pancreatic malabsorption and bronchial asthma is often treated with long-term corticosteroids). Cross-sectional studies among community-dwelling adults have shown a correlation between respiratory function and BMD measured with DXA.(6–8) Two cross-sectional studies from Cambridge, UK, found a positive and continuous relationship between FEV1 and BMD at the hip across the whole normal range of respiratory function in women and men.(7,8) This association was evident in young, middle, and older age groups almost to the same extent. After adjusting for potential confounding factors, mean hip BMD of women in the highest FEV1 quartile was ∼3–5% higher than the mean BMD in women in the lowest quartile.(7) This difference was slightly lower, but still significant, for men (2–3.5%).(8) This magnitude is comparable to that observed for BUA in this study. As far as we know, no prospective study, however, has investigated the predictive power of pulmonary function testing for osteoporotic fractures or assessed the association of respiratory function and QUS measures among healthy members of the community.
Impaired respiratory function is associated with morbidity(24) and mortality.(24,25) Poor respiratory function predicts overall mortality, as well as death caused by cancer,(26) pulmonary disease,(27) cardiovascular disease,(24,26) and stroke.(26) This relation could simply reflect the effect of cigarette smoking, respiratory illness, or other preexisting diseases.(28,29) Researchers have advised that the use of FEV1 as part of any health assessment of middle-aged patients should be considered.(26) This study showed that FEV1 can be used as a marker of bone characteristics as assessed by QUS. Moreover, even after adjustment for BUA in multivariate Cox regression analysis, FEV1 was a significant predictor of hip fracture. This suggests a potential association between respiratory function and some unmeasured bone characteristics or other fracture risk factors such as tendency to falls. One plausible explanation is that respiratory function and bone health both reflect common but as yet unknown determinants.(7)
This study has some limitations. Respiratory function was evaluated using the better of two blow maneuvers in this study. This may induce some imprecision in the estimated respiratory function because most recent guidelines recommend use of at least three blow attempts for determination of FEV1.(30,31) This is mainly because of the fact that the original design and start of the EPIC-Norfolk study goes back to 1992 before development of these guidelines and the investigators chose to continue with a consistent procedure of spirometry throughout the study follow-up.(14) Moreover, random measurement error in FEV1 values is more likely to underestimate the magnitude of the relationship between FEV1 and BUA.(32) Other measures of respiratory function such as FVC, PEF, and FEV1/FVC ratio did not add additional information to our results, and we chose to only report FEV1 as the most widely used and straightforward measure.
Participants in the baseline visit for this study (which was the second health check in EPIC-Norfolk) are likely to be healthier than the general population. About 60% of participants in the original cohort returned for this health check, and this may induce a healthy selection bias. We have previously compared characteristics of those who attended the second health examination with those who did not and, as expected nonattenders were older.(14) However, selection of participants in the first instance and the method of follow-up were not related to or influenced by the exposure level in this study. Moreover, it is unlikely that the association observed in this study between respiratory function and bone health would be different or in the opposite direction in the nonattending population. In fact, pathophysiology would suggest that the link between respiratory function and bone health would be stronger in people with poorer health status because of common risk factors such as smoking and physical activity levels.(9–12) This, however, needs evaluation in further studies. DXA, as the current gold standard for BMD measurement, was not used in this study. Although the method used for ascertainment of fractures (data linkage of all participants with National Health Service hospital records and death certification) has the advantage of ascertainment of all hospitalized fractures and does not rely on follow-up self-reports that can be incomplete, there would be a potential for under-ascertainment of nonhospitalized fractures. Nevertheless, this method identifies the fractures with the most clinical impact. In particular, almost all hip fractures are hospitalized in the United Kingdom.
There is a well-established epidemiological relationship between smoking and respiratory function,(12,33) and several studies have suggested a significant association between smoking and fracture risk.(11,34) In our study, the association between respiratory function and fracture risk seemed to be independent of cigarette smoking habit. Reassessment of the association using pack-years of smoking did not change the results (data not shown). Although the association between respiratory function and fracture risk is independent of major known determinants such as age, smoking, and bone ultrasound measures, we cannot exclude residual confounding from these or other unknown factors. However, the magnitude of the association indicates that residual or unknown confounding factors would have to be substantial to account for this association between respiratory function and fracture risk.
This is the first population-based prospective study examining the association between respiratory function and bone health using both bone measurements (QUS method) and fracture endpoints. There is particularly a paucity of data on fracture risk determinants among men.(35) This study showed that the pattern of association between respiratory function and bone health is similar among men and women. These findings need replication in future prospective studies in different settings and different populations before being generalized and used in fracture risk prediction tools. If the association between FEV1 and hip fracture risk is confirmed, spirometry is a simple, feasible, and low-cost measurement that could be used in general practice to help in fracture risk prediction in older men and women.
EPIC-Norfolk is supported by program grants from the Medical Research Council UK and Cancer Research UK and with additional support from the Research into Aging, Stroke Association, British Heart Foundation, and the Academy of Medical Sciences UK. None of the study sponsors have had any role in study design, collection, analysis, and interpretation of data, writing of the report, or decision to submit the paper for publication.