Rapid Responses to:

EDITORIALS: Kay-Tee Khaw Hormone replacement therapy again BMJ 1998; 316: 1842-1844 [Full text]

Rapid Responses: Submit a response to this article

Rapid Responses published:

Hormone replacement therapy in anorexia nervosa

Jim Bolton   (29 June 1998)

Confounding is mistakenly called selection bias

John Newton   (18 August 1998)

Hormone replacement therapy in anorexia nervosa 29 June 1998
Jim Bolton, Lecturer Department of General Psychiatry, St George's Hospital Medical School, London. Send response to journal: Re: Hormone replacement therapy in anorexia nervosa	EDITOR - Khaw-Tee1 advocates for a risk-benefit balance in the use of hormone replacement therapy (HRT) in postmenopausal women. Anorexia nervosa is another hypo-oestrogenic state which carries an increased risk of osteoporosis. Compared with the paucity of appropriate research into HRT in asymptomatic postmenopausal women, there is even less data to guide an evidence based approach to the prevention and treatment of osteoporosis in anorexia. As a consequence there is a temptation to extrapolate from the postmenopausal state and to prescribe HRT. However, bone turnover in anorexia appears to differ from that following the menopause. Anorexia is associated with a decrease in the rate of bone turnover, compared to the increase in turnover associated with the menopause.2 This suggests that in anorexia other factors are of increased importance in the aetiology of osteoporosis, particularly calorific and nutritional deficiency. This, in turn, has implications for management. The optimum management is the restoration of an adequate diet with a return to normal body weight and a resumption of menstruation. What is unclear is the role of HRT in this group. Oestrogen supplementation is commonly used, often in the form of the oral contraceptive pill. However, this alone is not effective in preventing osteopenia in young women with anorexia.3 Indeed, trabecular bone loss may progress despite oestrogen therapy. A probable explanation for this finding is that HRT alone cannot correct the multiple factors which contribute to bone loss in anorexia. An additional factor to consider in prescribing HRT is the danger that hormone supplementation will be seen by the patient as an alternative to definitive treatment of the anorexia. Just as HRT is unlikely to be a solution to healthy ageing in women1, it is unlikely to be the universal solution to osteoporosis in anorexia nervosa. There is a need for research into the aetiology of bone loss and the management of osteoporosis associated with anorexia nervosa. References 1Khaw-Tee K. Hormone replacement therapy again. Risk-benefit relation differs between populations and individuals. BMJ 1998;316:1842-1843. 2Serpall L, Treasure J. Osteoporosis - a serious health risk in chronic anorexia nervosa. European Eating Disorders Review 1997;5:149-157. 3Klibanski A, Biller B M K, Schoenfeld D A, Herzog D B, Saxe V C. The effects of estrogen administration on trabecular bone loss in young women with anorexia nervosa. Journal of Clinical Endocrinology and Metabolism. 1995;80:898-904. Dr Jim Bolton Lecturer Department of General Psychiatry, St George's Hospital Medical School, London SW17 ORE. jgbolton@sghms.ac.uk
Confounding is mistakenly called selection bias 18 August 1998
John Newton, Consultant Epidemiologist University of Oxford Send response to journal: Re: Confounding is mistakenly called selection bias	In her editorial[1] Professor Kay-Tee Khaw expertly reviewed the case that HRT can prevent coronary heart disease. Unfortunately, she has reproduced the common error of referring to the fact that HRT users are likely to be healthier than non-users as “selection bias”[2-4] whereas it is in fact an example of confounding. Selection bias arises when unrepresentative individuals are selected for study leading to spurious results.[5] Confounding is not a spurious phenomenon: the factor being studied is in truth associated in the target population with another factor that influences the outcome of interest.[5] Errors due to confounding arise because of misinterpretation of essentially valid results. Thus, women in general who take HRT probably do experience less heart disease; the question is whether this is due to their HRT.[6] Does terminology matter that much? In this case I believe it does. Bias is the more damning criticism because design errors make study results unpredictably unreliable. Confounding is a reproducible effect that can be quantified and allowed for in the design or the analysis of studies. Of course there are limits: it is logically impossible to distinguish the effects of two very closely correlated factors and investigators can only allow for confounders that are known and measured accurately. In the case of HRT, adjustment for measurable confounders hardly changed the size of the observed effect on coronary heart disease.[7] A difficult judgement then has to be made as to whether all potentially important confounders have been adequately measured.[8] Randomised controlled trials using clinical endpoints (now underway) are the only studies that can deal with known and unknown confounders. Ironically, bias could still be a problem, for example, due to different levels of compliance in placebo and treatment arms.9 Proper debate of these issues requires careful use of appropriate terms. 1. Khaw KT. Hormone replacement therapy again. risk-benefit relation differs between populations and individuals. BMJ 1998;316:1842-4. 2. Hemminki E, Sihvo S. A review of postmenopausal hormone therapy recommendations: potential for selection bias. Obstet Gynecol 1993;82:1021-8. 3. Posthuma W, Westendorp R, Vandenbroucke J. Cardioprotective effect of hormone replacement therapy in postmenopausal women: is the evidence biased? Br Med J 1994;308:1268-9. 4. Grodstein F. Invited commentary: can selection bias explain the cardiovascular benefits of estrogen replacement therapy? Am J Epidemiol 1996;143:979-82. 5. Last JM. A dictionary of epidemiology. 3rd ed. Oxford: Oxford University Press, 1995. 6. Vandenbroucke JP. How much of the cardioprotective effect of postmenopausal estrogens is real? Epidemiology 1995;6:207-8. 7. Stampfer M, Grodstein F. Cardioprotective effect of hormone replacement therapy. Is not due to selection bias. BMJ 1994;309:808-9. 8. Willett WC. Re: "Prior to use of estrogen replacement therapy, are users healthier than nonusers?" Am J Epidemiol 1997;146:283-4. 9. Petitti DB. Coronary heart disease and estrogen replacement therapy. Can compliance bias explain the results of observational studies? Ann Epidemiol 1994;4:115-8.