Intended for healthcare professionals

Letters

Meta-analysis of cigarette smoking, bone mineral density, and risk of hip fracture

BMJ 1998; 316 doi: https://doi.org/10.1136/bmj.316.7136.1017 (Published 28 March 1998) Cite this as: BMJ 1998;316:1017

Three studies were omitted from meta-analysis

  1. David J Torgerson, Senior research fellow,
  2. David M Reid, Reader in rheumatology,
  3. Marion K Campbell, Senior statistician
  1. Centre for Health Economics, University of York, York YO1 5DD
  2. Department of Medicine and Therapeutics, University of Aberdeen, Aberdeen AB25 2ZD
  3. Health Services Research Unit, University of Aberdeen
  4. Wolfson Institute of Preventive Medicine, St Bartholomew's and the Royal London School of Medicine, London EC1M 6BQ

    EDITOR—Law and Hacksaw's meta-analysis of cigarette smoking, bone mineral density, and risk of hip fracture seems to be incomplete.1 At least three relevant studies,24 including our own,4 were not included. One of these studies,3 however, would need reanalysis of the data. In our study a random sample of 1227 premenopausal women aged 45-49 (including 311 smokers) had their bone mineral density measured with dual energy x ray absorptiometry, and several risk factors were then related to bone mineral density. This study was more than three times bigger than the largest study of premenopausal women included in the meta-analysis. Although smoking was significantly associated with low bone mineral density at all sites, this difference was abolished after adjustment for current body weight. There was, however, a suggestion of a link between low bone mineral density and smoking during pregnancy.4

    Using the search terms “tobacco” and “smoking” may be inefficient and lead to bias. Studies that looked at a range of risk factors for osteoporosis may not use the key words of smoking or tobacco because they were not shown to be important risk factors. Hence, Law and Hacksaw may be more likely to have identified papers that showed a negative effect of smoking on bone mass and missed those in which no association was found, 3 4 thus exaggerating the risk of smoking on fracture risk. Better search terms might have included “lifestyle” or “risk factors.”

    Even if the conclusion of the meta-analysis is correct, women may not always follow advice to stop smoking. In a randomised trial of bone density screening among perimenopausal women, those with low bone mineral density increased their use of hormone replacement therapy to reduce their risk of fracture.5 We have re-examined our data and compared those women who thought that they had low hip bone mineral density with those who thought that they had high hip bone mineral density; there was no difference in smoking levels two years after screening (21% v 20%, respectively; 95% confidence interval of difference −10% to 12%). Hence, even among women who believed themselves to be at high risk of osteoporosis (and most women had individual preventive advice from their general practitioners as well as general lifestyle advice from the screening unit) there was no significant change in smoking patterns. Thus the meta-analysis may be prone to selection bias and overestimate the effects of smoking on risk of fracture; even women who believe that they are at high risk of fracture and who smoke may not give up.

    References

    1. 1.
    2. 2.
    3. 3.
    4. 4.
    5. 5.

    Authors' reply

    1. Malcolm Law, Reader,
    2. Allan Hackshaw, Lecturer
    1. Centre for Health Economics, University of York, York YO1 5DD
    2. Department of Medicine and Therapeutics, University of Aberdeen, Aberdeen AB25 2ZD
    3. Health Services Research Unit, University of Aberdeen
    4. Wolfson Institute of Preventive Medicine, St Bartholomew's and the Royal London School of Medicine, London EC1M 6BQ

      EDITOR—Torgerson et al say that we did not include three studies in our meta-analysis, but the only one of the three that we overlooked in our search was theirs. Like the other studies in premenopausal women it showed little effect of smoking and its inclusion did not alter the finding that there was no material change in bone density in premenopausal smokers (estimated change per 10 years 0.02 SD (95% confidence interval −0.09 to 0.12 SD). The two other studies were identified but excluded. One measured bone density only in the spine (we stated that our analysis applied to long bones). The other reported no quantitative data on smoking, only that there was no significant association with bone density. As the 23 quantitative estimates of the relation in premenopausal and early postmenopausal women showed clearly that smoking had little or no effect on bone density, another study would not have altered the conclusion.

      We do not think that failure to identify studies has affected our estimates of the greater postmenopausal bone loss in smokers and their increasing risk of hip fracture with age relative to that in non-smokers. We did conduct a search using the term “risk factors.” More importantly, we derived two independent estimates of the relative risk of hip fracture in smokers according to age, one based directly on studies of smoking and hip fracture and the other based on studies of the difference in bone density between postmenopausal smokers and non-smokers using the association between bone density and risk of hip fracture. The two yielded similar results; it would be a remarkable coincidence if bias altered both estimates to the same extent.

      The main conclusion from our paper is not that people should not smoke (which is already clear) or that certain strategies effectively dissuade people from smoking. It is the clarification that smoking increases postmenopausal bone loss and causes hip fracture even though it has no detectable effect on bone density before the menopause.