Genetically Modified Foods and the Pusztai Affair

Downloaded from bmj.com on 30 June 2009

Genetically modified foods and the Pusztai affair

Jonathan M Rhodes

BMJ 1999;318;1284

Updated information and services can be found at: http://bmj.com/cgi/content/full/318/7193/1284

These include: Rapid responses You can respond to this article at: http://bmj.com/cgi/eletter-submit/318/7193/1284 Email alerting Receive free email alerts when new articles cite this article - sign up in the box at service the top left of the article

Topic collections Articles on similar topics can be found in the following collections

Diet (2132 articles)

Notes

To Request Permissions go to: http://group.bmj.com/group/rights-licensing/permissions To order reprints go to: http://journals.bmj.com/cgi/reprintform To subscribe to BMJ go to: http://resources.bmj.com/bmj/subscribers Downloaded from bmj.com on 30 June 2009 Letters

Website: www.bmj.com Email: [email protected]

Genetically modified foods and the Pusztai affair Vitamin D deficiency

Time for a tablet containing high doses of Editor—In his clinical review on geneti- non-toxic food components, contain vitamin D alone 2 cally modified foods Jones implies that lectins. Some of these, in red kidney beans Editor—I was heartened to read Comp- Pusztai had tested only the effects of potato for example, are toxic and need to be ston’s editorial calling for action on vitamin 3 spiked with concanavalin A (a lectin) at the destroyed by heat before consumption, but D deficiency.1 Another point should be 1 Rowett Institute. The initial dissemination others such as tomato lectin are apparently made to encourage a more active role by the of this incorrect information followed by harmless when eaten raw. Many plant health service. the inappropriate suspension of Pusztai and lectins have an insecticidal or antifungal The British National Formulary does not the suspicion of fraud implied by instituting role in the plant. Some of these food lectins include a tablet containing vitamin D alone an audit according to Medical Research have interesting biological effects. We have in reasonable dose. Prescribers can give a Council guidelines led some scientists, recently shown that the common edible calcium and vitamin D mixture, which may including me, to defend Pusztai, whom we mushroom lectin that is often eaten raw be unpalatable and therefore impair compli- know to be an honourable and careful 4 selectively inhibits nuclear protein import. ance. Alternatively, a vitamin capsule can be scientist. The report of the audit conducted The snowdrop lectin (GNA) binds to prescribed containing many other vitamins. at the Rowett Institute (www.rri.sari.ac.uk/ mannose, which is minimally expressed in If a dose greater than minimal daily require- press) clearly shows that experiments on ments is needed higher doses of unneces- transgenic potatoes containing the gene for mammalian intestine but extensively expressed in the intestine of sap sucking sary vitamins must also be given. Surely it the snowdrop lectin (GNA) had already should be easy to prescribe 500-1000 units insects. Thus expression of this lectin in been performed when Pusztai was inter- of vitamin D daily without any other food plants might render them unattractive viewed for World in Action in August 1998. addition? If the Rowett Institute had released a to insects but safe for human consumption, particularly if the food (potato) is always C H Cheetham Consultant paediatrician statement that the work was preliminary Wycombe General Hospital, High Wycombe, and allowed the work to continue such a cooked before ingestion. Buckinghamshire HP11 2TT potentially damaging media storm would Pusztai’s experiments, whatever their [email protected] probably not have happened. results, would not show that all genetically Not all lectins are toxic. They are modified foods were unsafe. Pusztai’s mes- 1 Compston JE. Vitamin D deficiency: time for action. BMJ 1998;317:1466-7. (28 November.) ubiquitous carbohydrate binding proteins. sage was simply that such foods require All mammalian cells and blood and all careful testing. As with the testing of new plant nuts, seeds, and bulbs, including many pharmaceutical agents, some transgenic Groups at risk need supplementation, and milk could be fortified foods will prove toxic or otherwise unsatis- factory and be discarded at an early stage of Editor—We agree with Compston that the time has come for reappraisal of measures Advice to authors development. aimed at preventing vitamin D deficiency We prefer to receive all responses electronically, The fact that Pusztai has been barred from continuing his experiments since the with consequent bone loss predisposing to sent either directly to our website or to the osteoporotic fracture.1 The high prevalence time of his initial suspension has meant that editorial office as email or on a disk. Processing of vitamin D deficiency and secondary his data remain preliminary, and further your letter will be delayed unless it arrives in an hyperparathyroidism with high bone turn- experimentation will probably be needed electronic form. over emphasises this urgent need. before any final conclusions can be drawn We are now posting all direct submissions to The adequacy of current vitamin D about the effects of the transgenic GNA our website within 72 hours of receipt and our intake recommendations is being ques- intention is to post all other electronic potato or its promoter. tioned. The most sensitive yardstick for submissions there as well. All responses will be There has been little clarity surrounding ascertaining the degree of hypovitaminosis eligible for publication in the paper journal. this debate, and one lesson that needs to be D that predisposes to bone disease is Responses should be under 400 words and learnt is that all scientists need to be careful secondary hyperparathyroidism. Vitamin D relate to articles published in the preceding to ensure that their comments inform rather deficiency is known to occur even at a < month. They should include 5 references, in the than confuse when handling issues that are serum 25-hydroxyvitamin D concentration Vancouver style, including one to the BMJ article of such extreme public interest. previously considered adequate—namely, to which they relate. We welcome illustrations. Professor of medicine 37.5 nmol/l. Over the past two years the Please supply each author’s current Jonathan M Rhodes University of Liverpool, Liverpool L69 3GA threshold has been found to exceed appointment and full address, and a phone or [email protected] 75 nmol/l in both middle aged2 and elderly3 fax number or email address for the people irrespective of calcium intake. corresponding author. We ask authors to declare Calcium intake can modulate parathyroid any competing interest. Please send a stamped 1 Jones L. Genetically modified foods. BMJ 1999;318:581-4. (27 February.) function and bone resorption even in addressed envelope if you would like to know 2 Rhodes JM. Beans means lectins. Gut 1999;44:593-4. subjects who are replete with vitamin D.3 3 Freed DW. Do dietary lectins cause disease? BMJ whether your letter has been accepted or rejected. While the outcome of studies comparing Letters will be edited and may be shortened. 1999;318:1023-4. (17 April.) 4 Yu L-G, Fernig DG, White MRH, Spiller DG, Appleton P, the effectiveness of supplementation with www.bmj.com Evans RC, et al. Edible mushroom (Agaricus bisporus) lec- low dose vitamin D with or without calcium tin, which reversibly inhibits epithelial cell proliferation, [email protected] blocks nuclear localization sequence-dependent nuclear in reducing the incidence of bone fracture is protein import. J Biol Chem 1999;274:4890-9. awaited, two courses of action are required.

1284 BMJ VOLUME 318 8 MAY 1999 www.bmj.com Letters Downloaded from bmj.com on 30 June 2009

The first is supplementing groups at risk taking antiepileptic drugs without supple- made for routine supplementation in such with vitamin D (400-800 IU/day) and mentation. Blanket prescribing of vitamin D subjects, with serum 25-hydroxyvitamin D calcium (500-1000 mg/day), which is safe has economic consequences. Supplementa- concentrations being measured to monitor and effective. The second is fortifying milk tion may unmask undiagnosed primary treatment. There is no reason why children as a public health measure to benefit both hyperparathyroidism. Furthermore, a dis- should not be supplemented with vitamin young4 and old5 people. tinction needs to be made between recom- D, although the dose may require adjust- Rosemarie Freaney Principal biochemist mendations on prophylaxis in patients ment. For other subjects at lower risk Malachi J McKenna Consultant endocrinologist starting treatment with antiepileptic drugs serum 25-hydroxyvitamin D concentrations Metabolism Laboratory, St Vincents Hospital, and the identification and treatment of should be determined to assess the need for Dublin 4, Republic of Ireland osteopenia and osteomalacia in patients supplementation; this should probably already receiving long term treatment, espe- include patients taking the newer anticon- 1 Compston JE. Vitamin D deficiency: time for action. BMJ 1998;317:1466-7. (28 November.) cially in the presence of other risk factors. vulsants since little is known about their 2 Chapuy MC, Preziosi P, Maamer M, Arnaud S, Galen P, Lina Nashef Consultant neurologist effects on vitamin D status. Even the higher Hercberg S, et al. Prevalence of vitamin D deficiency in an adult normal population. Osteoporos Int 1997;7:439-43. [email protected] doses of vitamin D required for patients 3 McKenna M, Freaney R. Secondary hyperparathyroidism Edmund Lamb Consultant clinical biochemist receiving anticonvulsant treatment are safe, in the elderly: means to defining hypovitaminosis D. Kent and Canterbury Hospital, Canterbury and the benefits of preventing hypovitami- Osteoporos Int 1997;8(suppl 2):S3-6. CT1 3NG 4 McKenna M, Freaney R, Byrne P, McBrinn Y, Murray B, nosis D are likely to far outweigh rare events Kelly M, et al. Safety and efficacy of increasing wintertime such as the unmasking of primary hyper- vitamin D and calcium intake by milk fortification. QJMed 1 Compston JE. Vitamin D deficiency: time for action. BMJ 1995;88:895-8. 1998;317:1466-7. (28 November.) parathyroidism. 5 Keane EM, Healy M, O’Moore R, Coakley D, Walsh JB. 2 Valimaki MJ, Tiihonen M, Latinen K, Tahtela R, Karkkainen J E Compston University lecturer Vitamin D—fortified liquid milk: benefits for the elderly M, Lamberg-Allardt C, et al. Bone mineral density measured Department of Medicine, Addenbrooke’s Hospital, community-based population. Calcif Tiss Int 1998;62: by dual-energy x-ray absorptiometry and novel markers of 300-2. bone formation and resorption in patients on antiepileptic Cambridge CB2 2QQ drugs. J Bone Min Res 1994;9:631-7. 3 Collins N, Maher J, Cole M, Baker M, Callaghan N. A pro- Guidelines are needed for treating spective study to evaluate the dose of vitamin D required to 1 Chapuy MC, Arlot ME, Delmas PD, Meunier PJ. Effect of correct low 25-hydroxyvitamin D levels, calcium, and alka- cholecalciferol treatment for three years on hip fractures diseases of bone metabolism in epilepsy line phosphatase in patients at risk of developing anti- in elderly women. BMJ 1994;306:1081-2. Editor—Compston’s editorial is a timely epileptic drug-induced osteomalacia. QJMed1991;78: 2 Dawson-Hughes B, Harris SS, Krall EA, Dallal GE. Effect of 113-22. calcium and vitamin D supplementation on bone density reminder that hypovitaminosis D is a in men and women 65 years of age or older. N Engl J Med common neglected problem in susceptible 1997;337:670-6. Author’s reply 3 Collins N, Maher J, Cole M, Baker M, Callaghan N. A pro- populations, with adverse effects on bone spective study to evaluate the dose of vitamin D required to 1 mass and increased fracture risk. Editor—Cheetham correctly points out the correct low 25-hydroxyvitamin D levels, calcium, and alka- The editorial, however, raises as many inadequacy of the dose of vitamin D line phosphatase in patients at risk of developing antiepileptic drug induced osteomalacia. QJMed1991;78:113-22. questions as it answers. Although acknowl- contained in current formulations. It is also edging uncertainty over the best dose of true that preparations containing calcium and route of administration for vitamin D and vitamin D may be less well tolerated. and whether it should be given with However, the studies showing reduced non- Medical rules are needed in calcium, Compston suggests that routine vertebral fracture risk in elderly people used marathons in the United supplementation (800 IU or 20 ìg daily) both calcium and vitamin D,1 and whether Kingdom should now be given to elderly people and vitamin D alone is effective remains to be other high risk populations. In epilepsy the established. Although higher dose vitamin D Editor—Dyer reports that two junior risk of bony injury is increased because of tablets would be useful in some clinical situ- doctors, Andrew Murphy and Lesley the seizures, and the use of antiepileptic ations, concurrent calcium supplementation McKee, went to the aid of a marathon drugs is also associated with disorders of may be required for protection against frac- runner who collapsed and died as she bone metabolism resulting in osteopenia tures in elderly people. crossed the finishing line of a half mara- and, occasionally, osteomalacia. This is Freaney and McKenna raise the central thon.1 Dyer reports Dr Murphy as saying attributed to vitamin D deficiency as a con- issue of defining hypovitaminosis D. As they that if he was faced with a similar emergency sequence of enzyme induction and state, the current threshold may be too low in future he would “just walk on by.” We increased vitamin D catabolism, although and potential beneficiaries of vitamin D sup- understand his comments. antiepileptic drugs have additional direct plementation may include those who were It should not have been necessary for a effects on bone metabolism.2 Changes seem previously believed to be replete with doctor who had just finished a gruelling half to be influenced by sex, duration of vitamin D. In addition to supplementation of marathon to take part in a resuscitation treatment, whether the drug is given alone groups at risk, fortification of milk would be attempt. Emergency medical assistance or with others, and whether the drug is a step forward, but experience in the United should remain the responsibility of race enzyme inducing. Although Compston States and Canada has shown wide varia- organisers and those hired to provide emer- includes patients with epilepsy among those tions in the vitamin D content of fortified gency medical care. Our daughter collapsed at risk, it is not clear whether all patients milk so the benefits of this approach would at a public event where competitors had taking antiepileptic drugs or only specific be limited to those who drink sufficient paid for medical assistance in their entrance groups should receive supplementation. Is quantities of milk.2 fee. The organisers should have employed a the dose recommended intended as Nashef and Lamb raise several doctor qualified to deal with such an prophylaxis in patients starting treatment questions. Anticonvulsant treatment is a risk emergency, especially as another young per- or to correct established bone disease, in factor for osteomalacia and osteoporosis son had died at the finish line in the previous which hypovitaminosis requires far higher and is associated with an increased preva- year’s event. doses of vitamin D?3 When should bone lence of hypovitaminosis D. The amount of If everything had been done correctly to metabolism or density, or both, be assessed vitamin D required to achieve normal save Anna that would have been the end of before supplementation, and how should serum 25-hydroxyvitamin D concentrations the matter. The absolute tragedy of Anna’s treatment be monitored? Is supplementa- in patients receiving anticonvulsants varies, death is that she collapsed at the finish line tion required in patients receiving new anti- in one study ranging between 400 IU and close to St John Ambulance staff with a epileptic drugs? And, given that epilepsy is 4000 IU daily.3 The use of enzyme inducing modern automatic external defibrillator, common in children when maximum bone drugs, particularly in high doses and with which could have quickly provided the defi- accretion occurs, how early in life should other drugs, and lack of exposure to brillating shock she needed to survive. supplementation start? sunlight as a result of being housebound or Evidence given at the inquest showed Some 400 000 people in the United living in an institution are major risk factors that, although the staff reached her within Kingdom have epilepsy, most of whom are for hypovitaminosis D. A strong case can be less than a minute, they did not diagnose

BMJ VOLUME 318 8 MAY 1999 www.bmj.com 1285 Letters Downloaded from bmj.com on 30 June 2009 cardiac arrest or follow the correct defibrilla- x “The last paper I read on the subject” = concurrent viral hepatitis and differences tor procedures. The first seven minutes of a dimly recalled reference from a dozen have not been significant.4 the resuscitation attempt is well documented years ago The incidence of hepatitis due to isoni- by the internal defibrillator electrocardio- x “Their results may be statistically signifi- azid increases with age over 35. On these graph and audiotape recording, which cant, but they lack clinical relevance” = I grounds it seems reasonable to complete showed continuous ventricular fibrillation, disagree with their findings preventive treatment that has been started but no shock was given despite the x “Existing evidence is as yet too insubstan- but to delay treatment with isoniazid until machine’s prompting the operator to do so. tial to permit definitive conclusions” = I after delivery in those in whom a tuberculin UK Athletics 98 rules require only that haven’t a clue what you’re on about. skin test gives a positive result. Treatment there be a qualified medical officer and an Sam Shortt Director, Queen’s health policy should not, however, be withheld from emergency medical centre to cope with two Queen’s University, Kingston, Canada patients with active disease, nor termination emergencies per 1000 people at the finish. [email protected] advised. There is no information on the degree or Graham Bothamley Consultant respiratory physician quality of care required. Neither do they 1 Molesworth N. Down with EBM. BMJ 1998;317:1720-1. Will Elston Specialist registrar specify the need for advanced life support or (19-26 December.) East London Tuberculosis Service, Homerton advanced cardiac life support, the role of Hospital, London E9 6SR medical staff, or what is meant by emer- 1 Rubin P. Drug treatment during pregnancy. BMJ gency. Pregnancy does not mean that 1998;317:1503-5. (28 November.) In France, medical rules are comprehen- 2 Joint Tuberculosis Committee of the British Thoracic patients with tuberculosis must Society. Chemotherapy and management of tuberculosis sive and legally binding. They call for a high in the United Kingdom: recommendations 1998. Thorax degree of mobility of a doctor skilled in stop treatment 1998;53:536-48. resuscitation, paramedical ambulances, and 3 Brost BC, Newman RB. The maternal and fetal effects of Editor—We were disappointed that tuberculosis therapy. Obstet Gynecol Clin N Am 1997;24: an effective means of radio communication. Rubin’s review of drugs that could safely be 659-73. Since their introduction in 1992 only eight 4 Hamadeh MA, Glassroth J. Tuberculosis and pregnancy. taken in pregnancy did not mention drugs Chest 1992;101:1114-20. people have died in out of stadium running 1 for tuberculosis. Adherence to successful 5 Moulding TS, Bedeker AG, Kanal GC. Twenty isoniazid- events throughout France. associated deaths in one state. Am Rev Respir Dis 1989;140: treatment of tuberculosis is essential. In A campaign was launched on 8 March, 700-5. 1998 three patients with tuberculosis (of the anniversary of Anna’s death, requiring 114 such patients treated at Homerton the sports minister to compel organisers to Hospital) became pregnant; all three were provide advanced life support at these advised by their general practitioner or Cluster randomised trials strenuous events. It is backed by Don Foster, midwife to stop their treatment, and two MP for Bath, and we welcome support from accepted a termination. In addition, two of Standardised approach to analysing and the medical profession. The current haphaz- 71 female patients receiving preventive reporting these trials is misguided ard system which puts bystander doctors in treatment with isoniazid were advised to Editor—In their editorial on cluster ran- such a difficult position should not be stop treatment after a pregnancy test gave a domised trials Campbell and Grimshaw allowed to continue. positive result. correctly identify the importance of the Some 70 athletes have died in the past The British Thoracic Society’s guidelines appropriate choice of the unit of analysis.1 10 years in the United Kingdom at these state: “None of the first line drugs has been Unfortunately, they display a disappoint- events. Information on them is being shown to be teratogenic....Streptomycin ingly poor grasp of the basis of estimation, compiled by Rob Perkins (email: rob@ 2 should be avoided as [it] may be ototoxic.” and the specific recommendations that they newsquick.co.uk; tel 0121 684 3222; fax: In the United States caution is expressed make are inappropriate. 0121 684 3224). Please contact him if you regarding the use of pyrazinamide, but The authors’ plea for a standardised have any information or would like more 34 data do not indicate any adverse effects. approach to the analysis and reporting of details of the campaign. Hamedeh and Glassroth reviewed the litera- cluster randomised trials is misguided. The Phillip Loyley Father of Anna Loyley ture concerning over 2000 pregnancies in framework that they describe has the patient Pauline Loyley Mother of Anna Loyley Bath BA2 2EE mothers with tuberculosis and found no as the principal unit of the experiment, with [email protected] significant difference in side effects in these the correlation between patients in the same women compared with non-pregnant sub- cluster having to be taken into account in the 1 Dyer C. “Good Samaritans” face grilling. BMJ 1998;317: jects and no higher incidence of fetal analysis. Following the recommendations in 1100. (24 October.) abnormalities; they commented that “there is the editorial may, however, make investigators no indication for the interruption of preg- prey to the same errors that the authors are nancy” for those with tuberculosis and that counselling them to guard against. Pseudo-EBM-ers have their “untreated tuberculosis represents a far Although some trials fall naturally into own lexicon greater hazard to a pregnant woman and her the framework that the authors describe, fetus than does treatment of the disease.”4 others do not. It is not helpful in their own Editor—Molesworth is to be commended Theoretical considerations suggest that example of an educational intervention to for drawing our attention to the clear differ- the immune response in pregnancy, which implement a clinical guideline. There, the ences between those who are EBM-ers and shifts T cell responses away from cell clinician (whose behaviour is the target of those who are not.1 But, alas, life is never that mediated immunity (Th1) and towards anti- the intervention) is clearly the principal unit simple. He has failed to acquaint his readers body help (Th2), could allow tuberculosis to of the experiment. We are interested in the with a pernicious class of practitioners develop in those who have been infected change in his or her behaviour as a result of known to students of the field as pseudo- with the tubercle bacillus. The greater the intervention, the average size of the EBM-ers. These wily folk are not easily iden- proportion of pregnant patients with smear change, and how the size varied between cli- tified by appearance or position. However, if positive pulmonary tuberculosis is in keep- nicians. Focusing on the estimation of these listened to carefully—often a taxing experi- ing with this hypothesis (usually this form of quantities of direct interest should lead to a ence—their speech is virtually pathogno- the disease is more common in men)4; active sensible and interpretable analysis. Recog- monic. I provide here a few key phrases with disease should arise in infected pregnant nising that such estimates should be fit for their translations to assist the uninitiated in women more commonly than in others with purpose should lead to a sensible design.2 distinguishing real from pseudo-EBM: a positive tuberculin response. Concerns Adjusting the analysis on the basis of the x “The literature suggests” = I haven’t have been raised about the incidence of intracluster correlation coefficient assumes actually read anything, but, if I had, I bet it hepatitis induced by isoniazid during preg- the observed covariance structure to be the would say . . . nancy,5 although the women may have had true structure, without taking into account

1286 BMJ VOLUME 318 8 MAY 1999 www.bmj.com Letters Downloaded from bmj.com on 30 June 2009 measurement error. It is robust only where a gold standard approach is to measure the ischaemic heart disease. Another possible reasonable estimate of variability at the clini- correlation coefficient directly within the mechanism involves the copper-dependent cian level is available. Thus to suggest that study setting. However, the estimate of this enzyme lysyl oxidase, which has a role in the such methods are more efficient than aggre- coefficient, like other assumptions, is often cross linking of collagen and atherogenesis gated analyses at the level at which the inter- informed by previous research or published in the major vessels. Areca contains a high vention is targeted is incorrect. The authors’ reports. Because of the paucity of published level of copper (302 (SD 92) nmol/g (range quest for information on intracluster corre- information on the potential size of correla- 205-535)) compared with other nut based lation coefficients is a search for an answer to tion coefficients in specific settings and on snacks (range 22-173 nmol/g) and upregu- the wrong problem. Thoughtless focus on factors that might affect their magnitude, we lates lysyl oxidase activity.5 correlation coefficients could lead to the use (with other authors34) are promoting their We have recently described an areca of values estimated in one study for calcula- routine publication. However, to suggest that dependency syndrome in a group of Indian tions of sample size in quite a different area, using an external estimate, even if it is the (Gujarati) immigrants living in west London which is probably worse than doing no such best available, is worse than not calculating who were addicted to areca products.2 As the calculation at all. sample size at all is absurd. The adoption of American study has shown, eating nuts may The authors also suggest incorrectly that such an approach would invalidate most cal- be a safe and healthy habit. However, covariates at the patient level may not be culations of sample size ever undertaken. recommendations encouraging frequent examined in aggregated analyses. General- Marion K Campbell Senior statistician consumption should specify safe nuts and ised linear modelling provides the oppor- Jeremy M Grimshaw Programme director consider deleterious effects of some nuts tunity of specifying covariates at the patient Health Services Research Unit, University of such as areca. Aberdeen, Aberdeen AB25 2ZD level while analysing at the level at which the Chetan Trivedy Research fellow 3 [email protected] intervention is targeted. Saman Warnakulasuriya Senior lecturer in oral medicine Nick Freemantle Senior research fellow 1 Murray DM. The design and analysis of group randomised [email protected] Medicines Evaluation Group, Centre for Health trials. Oxford: Oxford University Press, 1998. Economics, University of York, York YO10 5DD 2 Freemantle N, Mason JM, Haines A, Eccles MP. Timothy J Peters Professor of clinical biochemistry CONSORT: an important step toward evidence-based King’s College Hospital, London SE5 9RW John Wood Principal statistician health care. Ann Intern Med 1997;127:81-3. Statistical Sciences, SmithKline Beecham 3 Donner A, Brown KS, Brasher P. A methodological review Pharmaceuticals, Harlow, Essex CM19 5AW of non-therapeutic intervention trials employing cluster 1 Hu FB, Stampfer MJ, Manson JE, Rimm EB, Colditz GA, randomization, 1979-1989. Int J Epidemiol 1990;19: Rosner BA, et al. Frequent nut consumption and risk of 795-800. coronary heart disease in women: prospective cohort 1 Campbell MK, Grimshaw J. Cluster randomised trials: time 4 Ukoumunne OC, Gulliford MC, Chinn S, Sterne J, Burney study. BMJ 1998;317:1341-5. (14 November.) for improvement. BMJ 1998;317:1171-2. (31 October.) P, Donner A. Evaluations of health care interventions at 2 Winstock AR, Trivedy CR, Warnakulasuriya KAAS, Peters 2 Wood J, Freemantle N. Choosing an appropriate unit of area and organisation level. In: Black N, Brazier J, TJ. A dependency syndrome related to areca nut use: some analysis in trials of interventions that attempt to influence Fitzpatrick R, Reeve B, eds. Health services research methods: medical and psychological aspects among areca nut users practice. J Health Serv Res Policy (in press). a guide to best practice. London: BMJ Publishing, 1998. in the Gujarat community in the UK. Addict Biol (in press). 3 SAS/STAT Softward. Changes and enhancements through 3 Warnakulasuriya KAAS, Trivedy C, Maher R, Johnson NW. release 6.12. Cary, NC: SAS/STAT Softward, 1997. Aetiology of oral submucous fibrosis. Oral Dis 1997;3:286-7. 4 Obeid OA, Mannan N, Perry G, Iles RA, Boucher BJ. Areca nuts can have Homocysteine and folate in healthy east London Bangla- Authors’ reply deleterious effects deshis. Lancet 1998;352:1829-30. Editor—We did not suggest that a standard- 5 Trivedy C, Baldwin D, Warnakulasuriya KAAS, Johnson 1 NW, Peters TJ. Copper content in areca catechu (betel nut) ised approach be adopted for the analysis of Editor—The United States nurses’ study products and oral submucous fibrosis. Lancet cluster trials. Our primary plea is that and the editorial that accompanies it report 1997;340:1447. researchers account for the clustering in some health benefits of eating several their data. Several methods are available to helpings of nuts a week. These benefits can- do this, and considerable debate surrounds not be generalised to all nuts commonly Fractures of the thoracolumbar the choice of method. We emphasised the consumed, and it is important to specify spine in major trauma patients use of hierarchical models; they allow all the which nuts confer them. information and variation at each level of The areca nut (erroneously referred to the data to be explored while retaining the as the betel nut) is chewed by over 200 ATLS manual does recognise importance validity of the analysis. An aggregated million people worldwide—a tenth of the of distracting injuries approach may, however, be more appropri- world’s population. It is chewed either alone Editor—Meek is right to highlight the ate for some analyses. Considerable debate or as part of a quid known as “pan.” This potential danger of missing fractures of the also surrounds the choice of the unit of habit is practised commonly in central, thoracolumbar spine in patients with analysis. As Murray argues, however, this southern, and southeast Asia and is now also trauma if the need for x ray investigations is may be misplaced, and attention should emerging in Western countries. Recently based purely on the clinical examination. rather be focused on the appropriate speci- new information on medical, biochemical, But he is wrong to say that the Advanced fication of the model for the analysis; the and psychological correlates of use of areca Trauma Life Support Manual for Physicians model selected should be well matched to nuts have been reported.2 Several deleteri- “does not recognise the importance of a dis- the underlying structure of the data.1 ous effects on oral and general health are tracting injury in masking the symptoms We were surprised that Freemantle and now linked to use of the nuts. The consump- and signs of spinal fracture.”1 Wood considered a standardised approach tion of areca has been strongly linked to the I believe that the quotation he gave was to the reporting of cluster randomised development of oral submucous fibrosis (a from the introductory section of the chapter trials to be misguided. As Freemantle potentially malignant disorder of the oral on spinal trauma. The following quotes are himself has indicated, implementation of cavity3) and a high incidence of oral cancer from the same chapter.2 “Any patient with the CONSORT guidelines for randomised in the Indian subcontinent and among an altered level of consciousness or cogni- trials should minimise the serious deficien- Indian immigrants. tive dysfunction (GCS < 15), multisystem cies in the design, analysis, and reporting of Consumption of areca has also been injuries, . . . requires spinal protection until such trials.2 Extension of this statement to linked with cardiovascular disease, diabetes, AP [anteroposterior] and lateral [thoraco- include the issues specific to cluster trials and asthma.2 The mechanisms by which it lumbar] spine x-rays are obtained to exclude should yield similar benefits. may induce cardiovascular disease are not any injury.” “In the secondary survey . . . We also cannot agree with the sugges- clear. A recent study in a Bangladeshi popu- thoracolumbar spine films may be obtained tion that we should refrain from seeking fur- lation in east London found raised homo- . . . if the mechanism of injury suggests the ther information on the intracluster correla- cysteine and reduced folate concentrations possibility of spinal injury.” “Other injuries tion coefficient. Reliable estimates of this in 170 healthy regular chewers.4 High serum may mask the physical findings of spinal coefficient are required if robust calculations homocysteine concentrations have been injuries, which may go unsuspected unless of sample size are to be made. Naturally, the associated with an increased risk of acute the doctor obtains the appropriate x-rays.”

BMJ VOLUME 318 8 MAY 1999 www.bmj.com 1287 Letters Downloaded from bmj.com on 30 June 2009

The work of Cooper et al quoted in the Evidence does not exist that (â). These proportions do not vary with the article is also referenced at the end of the number of tests. 3 dyspepsia heralds gastric relevant chapter in the manual. Dredging data is a smart and cost cancer in its earliest stage Ray McGlone Instructor in advanced trauma life effective way of doing research. “Dredge your data and [do] not tell” are Aickin’s support Editor—The editorial claiming that proton Royal Lancaster Infirmary, Lancaster LA1 4RP words; my advice was to “describe what was pump inhibitors may mask early gastric [email protected] done.”3 Besides, Bonferroni adjustments cancer recommends that “dyspeptic patients make no distinction between data dredging over 45 should undergo endoscopy before 1 and multiple planned tests. 1 Meek S. Lesson of the Week. Fractures of the thoracolum- these drugs are started.” The authors argue bar spine in major trauma patients. BMJ 1998;317:1442-3. What tests to include in the adjustment (21 November.) that this policy is justified by the improved is a serious issue since it will determine 2 American College of Surgeons Committee on Trauma. curability of gastric cancer diagnosed at an Advanced trauma life support manual for physicians. Chicago: whether a result will be “significant” or not. early stage. They quote the studies by Hallis- ACS, 1997. Aickin does not seem to have an answer; nor 3 Cooper C, Dunham CM, Rodriguez A. Falls and major sey et al and Fielding et al, which show an does anyone else. Pushing a reasoning to the injuries are risk factors for thoracolumbar fractures: cogni- increased proportion of gastric cancers tive impairment and multiple injuries impede the point of absurdity is a rhetorical device to detection of back pain and tenderness. J Trauma 1995;38: diagnosed at an early stage when all dyspep- show that the reasoning does not hold. 692-6. tic patients aged over 40 are invited to 23 The only advantage of Holm adjust- undergo gastroscopy. No control group of ments over Bonferroni adjustments is that otherwise matched non-dyspeptic patients It happened to me! they inflate â errors less, but the procedure is was studied. These results therefore do little complex.1 I doubt that Holm-adjusted P Editor—Meek’s paper on fractures of the more than show that as a population under- values are understandable to anyone but a thoracolumbar spine in major trauma goes increased gastroscopy so gastric cancer few statisticians. As for the American Journal patients reminded me of my own experiences presents earlier, rather than that dyspepsia is 1 of Public Health—its instructions for authors as a victim of major trauma. In April 1998 I a useful symptom heralding gastric cancer in the December 1998 issue include no was riding my motorcycle to Devon on the in its earliest stage. statement about multiple test adjustments, M5 when I (and six cars) was struck from There may be good reasons for carrying and the three first articles contain respec- behind by a 24.4 tonne lorry. Whiplash back out endoscopy for all patients over 45 with tively 70, 110, and 114 tests, in tables only, and forward occurred several times as I was onset of dyspepsia before treating them with without any multiple test adjustment. knocked into the car in front. I was not struck an acid suppressant; evidence does not yet I am baffled that “whether a given study on the head, and I remained fully conscious exist, however, to show that the ability to should be statistically analysed at all” should and alert throughout. However, I suffered a diagnose gastric cancer early is one of them. fracture of my left femur, a compound concern anyone. We should worry whether to Thomas S Low-Beer Consultant physician do the study in the first place and, when it is fracture of the left tibia and fibula close to the 38 Weoley Park Road, Birmingham B29 6RB knee joint, and an open stab wound to the left [email protected] done, how best to interpret the data. Multiple knee joint. I had pain around the right lower test adjustments help with neither of these. ribs, but at the scene and in casualty I did not 1 Griffin SM, Raimes SA. Proton pump inhibitors may mask Thomas V Perneger Medical epidemiologist early gastric cancer. BMJ 1998;317:1607-8. (12 December.) Institute of Social and Preventive Medicine, have any pain in the back. 2 Hallissey MT, Allum WH, Jewkes AJ, Ellis DJ, Fielding JWL. University of Geneva, CMU, CH-1211 Geneva 4, I was taken to theatre: the femoral Early detection of gastric cancer. BMJ 1990;301:513-5. 3 Fielding JWL, Ellis DJ, Jones BG, Paterson J, Powell DJ, Switzerland fracture was fixed with an intramedullary Waterhouse JAH, et al. Natural history of “early” gastric [email protected] nail, and a temporary external fixator was cancer: results of a 10-year regional survey. BMJ 1980;281:965-7. applied to the fracture of the tibia and fibula. 1 Aickin M, Gensler H. Adjusting for multiple testing when An epidural was inserted for postoperative reporting research results: the Bonferroni vs Holm meth- pain relief. Over the next three days I devel- ods. Am J Public Health 1996;86:726-8. 2 Aickin M. Other method for adjustment of multiple testing oped increasing pain in the area of the lum- Adjusting for multiple testing exists. BMJ 1999;318:127. (9 January.) bar spine that eventually became agonising 3 Perneger TV. What’s wrong with Bonferroni adjustments. in studies is less important BMJ 1998;316:1236-8. and was not relieved by increasing the than other concerns epidural analgesia. After further surgery, mobilisation was ordered. Finally, after I was Editor—In a paper in the American Journal crying out in pain, a routine (non-urgent) of Public Health Aickin stated that “there is Cited studies did not show x ray examination of my lumbar spine was substantial debate . . . concerning when (if relation between maternal ordered, which was not performed until the ever) adjustment for multiple testing is war- anxiety and birth weight next day (five days after the accident); this ranted.”1 I am glad that he has joined the showed a wedge fracture of L1. Only after debate in the BMJ over Bonferroni adjust- Editor—Teixeira et al1 state that “some the x ray film was obtained was my spine ments but find his arguments unconvincing.2 studies have shown that babies of stressed or examined. There were no neurological Yes, “researchers who adjust P values almost anxious mothers have a significantly lower problems after this fracture, which has always present them for their individual than average birth weight for gestational age healed well, as have the other fractures. hypotheses,” as he says. This is why they and tend to be born early” and then I was certainly at risk by Meek’s should not worry about unrelated tests and reference four papers, one of which is mine.2 definitions, because I was riding a motor- renounce a statistical technique that focuses The St George’s birthweight study was a cycle and had other, distracting painful inju- on the largely irrelevant universal null large prospective study of pregnant women ries. Perhaps, though, a medical background hypothesis. which measured anxiety and depression should be added to his list of risk Statistical tests were developed for throughout pregnancy. We found no associ- factors—although I recognised the symp- repeated testing, such as industrial quality ation between anxiety and depression and toms of spinal fracture, I was reluctant to control. The á and â error rates are valid in either prematurity2 or low birth weight.3 Nor make a fuss or be an awkward patient so did the long run, as asymptotic averages. Hence was any association found with life events as not complain nearly loudly or long enough. multiple testing is no violation of test theory, measured by Paykel’s interview for recent Sarah Walters Senior clinical lecturer in public health it is a prerequisite. Industrialists know that life events. and epidemiology over time they will reject a proportion of Interestingly, having demonstrated that University of Birmingham, Birmingham B15 2TT good lots (á) and market a proportion of high anxiety levels are linked to an increased [email protected] bad lots (â) in error; researchers know that uterine artery resistance index, Teixeira et al in their career they will reject a proportion note that the index is predictive of intrauter- 1 Meek S. Lesson of the Week. Fractures of the thoracolum- á bar spine in major trauma patients. BMJ 1998;317:1442-3. of true null hypotheses ( ) and miss a ine growth retardation. Thus it is paradoxi- (21 November.) proportion of true alternative hypotheses cal that one of the largest cohort studies of

1288 BMJ VOLUME 318 8 MAY 1999 www.bmj.com Letters Downloaded from bmj.com on 30 June 2009 anxiety, depression, and stress in pregnancy, concern is to determine the patient’s marital 1 Feely M. Drug treatment of epilepsy. BMJ 1999;318:106-9. (9 January.) which was also referenced, found no associ- status or to rule out marital dysfunction, as 2 Clayton-Smith J, Donnai D. Fetal valproate syndrome. ation between any of the above and suggested in the article, then a gay, lesbian, J Med Genet 1995;32:724-7. 4 3 Ornoy A, Cohen E. Outcome of children born to epileptic intrauterine growth retardation. Although or bisexual patient is immediately alienated. mothers treated with carbamazepine during pregnancy. the results were based on a single psycho- The use of more inclusive questions Arch Dis Child 1996;75:517-20. social assessment at 26 weeks, like the St such as “Who are the most important 4 Morrell MJ. The new antiepileptic drugs and women: efficacy, reproductive health, pregnancy and fetal outcome. George’s study, it found no relation between people in your life?” and, more specifically, Epilepsia 1996;37(suppl 6):S34-44. anxiety and depression and prematurity or “Are you sexually active with men, women, low birth weight. There was, however, an or both?” will allow the patient to answer association between increasing levels of honestly and validate his or her relation- Alcohol intake and cancer of stress and prematurity and low birth weight. ships. The use of such questions will send a Finally, the Danish study of 5873 women clear, normalising message and tell the the upper digestive tract that was cited found that women who had patient that this is a safe place in a largely one or more life events that they perceived hostile health system. Pattern of risk in Italy is different from as being highly stressful had a risk of Peter Meacher Family physician that in Denmark preterm delivery 1.76 times greater than Department of Family Medicine, Montefiore 5 Editor—A Danish study of 156 subjects those without stressful life events. This Medical Center, 3544 Jerome Avenue, Bronx, NY with upper digestive tract cancers showed sounds impressive but equates to a decrease 10467, USA [email protected] that wine drinkers may be at a lower risk in length of gestation of 1.06 days, which is than drinkers with a similar intake of beer or of no clinical importance. Competing interests: Peter Meacher is codirector of spirits.1 Wine is the most common alcoholic The findings of Teixeira et al are the Bronx Lesbian and Gay Healthcare Consortium, beverage in Italy, accounting for over 80% of interesting and warrant further study. How- an activist group in the Bronx. alcohol intake.2 We investigated the separate ever, caution has to be exercised in and combined effect of wine drinking in a interpreting large epidemiological studies 1 Tomlinson J. ABC of sexual health: Taking a sexual history. BMJ 1998;317:1573-6. (5 December.) large dataset.3 that confine psychological assessments of Data were collected between 1984 and pregnant women to the beginning of the last 1993 in Greater Milan and the province of trimester. Women at this stage may already Pordenone, Northern Italy, on 547 histo- be aware, through ultrasound assessments, Drugs for epilepsy have logically confirmed incident cases of cancer fundal measurements, or blood pressure teratogenic risks of the oral cavity and pharynx (466 men, 81 problems, that there is a problem with their women; age 22-74 (median 57)) and 412 pregnancy, which may induce stress and Editor—We would like to draw attention to cases of cancer of the oesophagus (345 anxiety. the fact that no anticonvulsant has been shown to be without risk to the fetus in men, 67 women; age 26-74 (median 60)). Paediatric specialist registrar Michael Perkin 1 Controls were 2375 subjects (1797 men, Mayday University Hospital, Croydon CR7 7YE pregnancy. Associations with neural tube defects and facial clefts are well recognised, 578 women; age 25-74 (median 56)) admit- 1 Teixeira JMA, Fisk NM, Glover V. Association between but congenital abnormalities alone are a ted to the same network of hospitals with maternal anxiety in pregnancy and increased uterine crude indicator of teratogenicity. Of acute non-neoplastic conditions unrelated artery resistance index: cohort based study. BMJ to alcohol or tobacco consumption. Of 1999;318:153-7. (16 January.) increasing concern is the risk of intellectual 2 Perkin MP, Bland JN, Peacock JL, Anderson HR. The effect impairment, which is frequently only these, 570 were admitted for trauma, 593 of anxiety and depression during pregnancy on obstetric for non-traumatic orthopaedic conditions, complications. Br J Obstet Gynaecol 1993;100:629-34. apparent in later childhood. Several studies 3 Brooke OG, Anderson HR, Bland JM, Peacock JL, Stewart that include long term follow up are 475 for acute surgical diseases, 428 for eye CM. Effects on birthweight of smoking, alcohol, caffeine, currently under way. diseases, and 309 for miscellaneous other socioeconomic factors, and psychosocial stress. BMJ illnesses. Trained interviewers questioned 1989;298:795-801. We recognise that anticonvulsant treat- 4 Copper RL, Goldenberg RJ, Das A, Elder N, Swain M, ment during pregnancy is unavoidable for cases and controls, using a structured ques- Norman G, et al. The preterm prediction study: maternal tionnaire; questions on the days per week stress is associated with spontaneous preterm birth at less many women, for whom attempts should be than 35 weeks gestation. Am J Obstet Gynecol made to achieve control with the fewest that each type of alcoholic beverage (wine, 1996;175:1286-92. beer, and spirits) was consumed and the 5 Hedegaard M, Henriksen TB, Secher NJ, Hatch MC, agents and lowest dose possible. Sabroe S. Do stressful life events affect duration of Nevertheless, we also recognise that many average number of drinks per day were gestation and risk of preterm delivery? Epidemiology included. women of childbearing age, despite being 1996;7:339-45. The table gives the distribution of cases seizure free for several years, continue to of oral and pharyngeal and oesophageal take anticonvulsants rather than risk losing cancer and of the comparison group their driving licence should a seizure occur according to consumption of various Inclusive questions are needed when treatment is withdrawn. These women alcoholic beverages. When non-drinkers when taking a sexual history are unable to make informed decisions and moderate drinkers ( < 3 drinks a day) unless they are fully aware of the teratogenic Editor—I welcome the ABC of Sexual were used as the comparison group, the risks of their treatment. Health but found the article on taking a odds ratios of oral and pharyngeal cancer More subtle effects of exposure to sexual history to be limited.1 were 4.95 for very heavy (>10 drinks per anticonvulsants in utero are increasingly Creating the best environment for day) wine drinkers and 4.13 for very heavy recognised, frequently by developmental patients to feel comfortable when discussing 23 wine, beer, and spirit drinkers. Correspond- paediatricians and clinical geneticists. The issues of sex and sexuality starts before the ing findings for oesophageal cancer were experience with newer anticonvulsants is patient enters the doctor’s office. Intake 4 7.53 and 5.52. The odds ratios for beer or limited. Failure to appreciate the risk of forms that ask if the patient is married or spirit drinkers only were 0.9 (95% confi- neurological impairment to the fetus by cli- single and health information pamphlets dence interval 0.3 to 2.4) for cancers of the nicians prescribing anticonvulsants and pro- that do not include information for gay and oral cavity (7 cases, 47 controls) and 1.5 (0.6 viding preconceptional advice can mislead lesbian couples reinforce the real and to 3.9) for oesophageal cancer (7 cases, 47 parents. Indeed, we are aware of several perceived homophobia of the medical controls). These results were consistent instances where this omission has resulted system, making the hurdle of disclosure to across separate diagnostic conditions in in litigation. the provider formidable. The article clearly controls. assumed that patients are heterosexual, as Elisabeth M Rosser Consultant This pattern of risk according to type of most doctors do: a psychiatrist colleague Louise C Wilson Consultant alcoholic beverage is different from that Unit for Clinical Genetics, Institute of Child Health, 1 recently told me that she thought she had no London WC1N 1EH reported in the Danish study. Taken gay or lesbian patients. If the doctor’s first [email protected] together, these data lead us to conclude that

BMJ VOLUME 318 8 MAY 1999 www.bmj.com 1289 Letters Downloaded from bmj.com on 30 June 2009

Distribution of 547 cases of oral and pharyngeal cancer, 412 cases of oesophageal cancer, and 2375 controls, and corresponding multivariate odds ratios* and 95% confidence intervals, according to total number of drinks per week of wine only and wine, beer, or spirits, Italy, 1984-93 Oral cavity and pharynx Oesophagus Wine only Wine and other Wine only Wine and other Cases of Cases of Cases of Cases of cancer: cancer: cancer: cancer: No of drinks/week controls Odds ratio (95% CI) controls Odds ratio (95% CI) controls Odds ratio (95% CI) controls Odds ratio (95% CI) Non-drinkers 37:365† 37:365* 41:365* 41:365* 1† 1† 1† 1† <21 47:541 9:175 46:541 16:175 21-41 59:377 1.49 (1.00 to 2.23) 30:293 0.93 (0.54 to 1.62) 52:377 1.60 (1.06 to 2.40) 39:293 1.45 (0.88 to 2.39) 42-69 102:155 4.94 (3.26 to 7.47) 85:258 2.33 (1.40 to 3.88) 82:155 5.81 (3.81 to 8.86) 62:258 2.76 (1.68 to 4.53) >70 23:18 4.95 (2.32 to 10.56) 155:190 4.13 (2.45 to 6.97) 7:18 7.53 (2.63 to 21.53) 65:190 5.52 (3.16 to 9.65) *Estimates from multiple logistic regression equation including centre, age, sex, education, social class, and smoking habit. The sum of strata do not add up to the total because a few values are missing. †Reference category is non-drinkers and those drinking <21 drinks/week combined. ethanol is the main component of alcoholic Another source of potential residual accurately corrected for, particularly when beverages that determines the risk of cancer confounding is the amount of tobacco smoking habit is only broadly classified. and that the most frequently consumed bev- smoked. Again, to adjust properly for any The term “correction for smoking” is erage in each area tends to be the one with confounding effects arising from smoking ambiguous. In some studies authors have the highest risk.45 Wine consumption is one would like to see finer divisions of intake clearly analysed out the effects of smoking clearly strongly related to cancers of the than those reported in the paper (1-19 g/ and extrapolated back, in order to predict upper digestive tract in individuals and day and 20 g/day). values for non-smokers. The present authors populations with heavy consumption. Finally, the authors mentioned the seem to have corrected to a standard value Carlo La Vecchia Associate professor of epidemiology absence of data on diet. It has been shown for smoking, probably the median of their Istituto di Statistica Medica e Biometria, Università consistently that eating fresh fruit and categories. If this interpretation is correct di Milano, 20133 Milan, Italy vegetables is strongly protective against then their graphs and tables show the Silvia Franceschi Consultant epidemiologist oesophageal cancer.3 In the West Midlands a morbidity standardised for a population who Adriano Favero Staff scientist recent survey on lifestyles showed that wine had formerly smoked for at least five years. Renato Talamini Registrar Centro Regionale Oncologico, 33081 Aviano PN, drinkers tended to have a better diet than The authors suggest that smoking was Italy those who drank beer or spirits (unpub- not the main influence on morbidity since the Eva Negri Head, unit of epidemiological methods lished data). If this was also the case in Den- level of cancer in the heavy smokers (risk 7.1) Istituto di Ricerche Farmacologiche, Mario Negri, mark then it could explain at least part of was lower than that in the heavy drinkers (risk 20157 Milan, Italy the difference observed. The authors’ case 11.7). However, their figure 1 indicates that 1 Grønbæk M, Becker U, Johansen D, Tønnesen H, Jensen G, would be stronger if they can show that such this risk in heavy smokers was in fact higher Sørensen TI. A population based cohort study of the associations do not exist in Denmark. than that for all but one of the alcohol association between alcohol intake and cancer of the upper digestive tract. BMJ 1998;317:844-8. (26 September.) On the whole, we would agree with the consumption categories, which implies that a 2 La Vecchia C. Alcohol in the Mediterranean diet. editorial accompanying the paper that there large proportion of the morbidity may still Assessing risks and benefits. Eur J Cancer Prev 1995;4:3-5. 3 Barra S, Franceschi S, Negri E, Talamini R, La Vecchia C. is “still no clear evidence to link specific have occurred in the smokers. They also Type of alcoholic beverage and cancer of the oral cavity, beverages to specific cancers.”4 point out that heavy smoking and heavy pharynx and oesophagus in an Italian area with high wine drinking tended to go together, which makes consumption. Int J Cancer 1990;46:1017-20. K K Cheng Professor of epidemiology 4 Doll R, Forman D, La Vecchia C, Woutersen R. Alcoholic C Cummins Lecturer accurate correction even harder. beverages and cancers of the digestive tract and larynx. In: R Maric Research fellow Many non-smokers are treated with Verschuren PM, ed. Health issues related to alcohol consump- Department of Public Health and Epidemiology, tion. Washington: International Life Sciences Institute, drugs to reduce their cholesterol concentra- University of Birmingham, Birmingham B15 2TT 1993:126-66. tion or blood pressure, largely on the 5 International Agency for Research on Cancer. Alcohol [email protected] drinking. Lyons: IARC, 1988. (IARC monographs on the evidence of studies that have included evaluation of carcinogenic risks in humans volume 44.) 1 Grønbæk M, Becker U, Johansen D, Tønnesen H, Jensen G, smokers and in which smokers have Sørensen TIA. Population based cohort study of the contributed most of the morbidity. The justi- association between alcohol intake and cancer of the upper fication for treating non-smokers is gener- Confounding in such studies is important digestive tract. BMJ 1998;317:844-7. (26 September.) ally based on an extrapolation that finds a Editor—Grønbæk et al concluded that mod- 2 Cheng KK, Duffy SW, Day NE, Lam TH, Chung SF, Badrinath P. Stopping drinking and risk of oesophageal small residue of morbidity even after smok- erate consumption of wine does not increase cancer. BMJ 1995;310:1094-7. ing has been corrected for. If those the risk of cancer of the upper digestive tract 3 Cheng KK, Day NE. Nutrition and esophageal cancer. Cancer Causes Control 1996;7:33-40. corrections overestimate the true risk to whereas a similar intake of beer or spirits 4 Sabroe S. Alcohol and cancer. BMJ 1998;317:827. 1 non-smokers then these patients might be does. This is a potentially important finding, (26 September.) taking long term medication that may not at least for those of us who mostly drink wine. benefit them since they were not at risk in But several issues concerning confounding the first place. We know that smoking is a warrant some discussion. Such studies should be done in risk factor for most respiratory and cardio- Firstly, there seemed to be fewer heavy non-smokers vascular disease as well as for most cancers drinkers ( > 21 drinks/week) in the group in Editor—Grønbæk et al’s study is a further and some gastrointestinal disorders. Should which wine constituted > 30% of total example of a large epidemiological study in we not perform epidemiological studies alcohol intake than in the group in whom it which smokers have been included even specifically on non-smokers or, at the very constituted 0% of the intake (10.2% v 31.2%, though smoking was likely to be a risk factor 1 least, request authors to provide summary from data in table 2). It has been documented for the disease under study. Statistical tables of their main findings specifically for that there is probably little increase in risk of manipulation was then used to correct for the non-smoking subset of their population? oesophageal cancer below an intake of the effects of smoking. But how reliable are 200 g/week (16.7 drinks from Grønbæk et these corrections, and how much faith Vaughan Reed Freelance statistician 2 should be put in the answers obtained with 4 Inverlea Court, Mickleton, Gloucestershire al’s definition). It would have been useful if GL55 6TZ the authors had subdivided the > 21 drinks/ regard to the specific question of alcohol [email protected] week category to examine the risks associated consumption causing cancer? Given that with higher intake so that residual confound- corrections for relatively straightforward 1 Grønbæk M, Becker U, Johansen D, Tønnesen H, Jensen G, Sørensen TIA. Population based cohort study of the ing could be excluded, although such an data are often only approximate, the association between alcohol intake and cancer of the upper attempt may be hampered by small numbers. development of cancer is unlikely to be digestive tract. BMJ 1998;317:844-7. (26 September.)

1290 BMJ VOLUME 318 8 MAY 1999 www.bmj.com Letters Downloaded from bmj.com on 30 June 2009

Treating upper digestive tract cancers as was hampered by a smaller number of cases medical services would be used to provide a a single entity may be misleading in the heavy drinking categories. Unfortu- service that takes money from the popula- Editor—Grønbæk et al reported that the nately, La Vecchia et al’s study does not tion as a whole. The response from local type of alcohol consumed may be important provide an estimate for non-drinkers and MPs was overwhelmingly supportive, but in determining the risk of cancer of the does not distinguish the effect of drinking Frank Dobson ignored the plea for a upper digestive tract in their population wine from the effect of drinking beer and properly funded service, stating that the based study in Copenhagen.1 In this large spirits. policy of dispersion and the proposed Both Cheng et al and Reed are policy of fast tracking refugees would solve epidemiological study details about the 1 alcohol intake and the particular type of concerned about the possible confounding the problems. alcoholic beverage were obtained by ques- effect of smoking, which we controlled for We remain sceptical. Currently, East tionnaire. Patients with cancer of the upper quite thoroughly. In one analysis (reported Kent has a large number of refugees and digestive tract were then identified from the in the paper) we controlled for smoking by asylum seekers and is popular with London Danish cancer registry. This registry sub- taking into account differences in the boroughs for its cheap and available accom- divides oral cavity, pharyngeal, and oesopha- prevalence of smoking among individuals modation and its proximity to London. In geal cancers, but the authors make no at different levels of wine percentages and addition, landlords with vacant rooms in the comment about whether their results of total alcohol intake in the Cox regression area are encouraging London boroughs to applied to all cancers of the upper digestive models. In another model (not reported) we lodge refugees in their rooms. tract uniformly or not. This information may took into account the number of pack-years In consultation with the local medical be important considering the different and inhalation; this yielded the same results committee and with the support of local pathogenesis of these cancers. as those in the paper. The number of cases general practitioners the health authority Oesophageal cancers, in particular, are did not allow stratified analyses of smokers has instigated a scheme to provide a service made up of two main groups with very dif- and non-smokers. However, the aim of the directed at the specific needs of refugees ferent pathogenesis. Squamous carcinoma study was not to assess smoking or alcohol and asylum seekers. It consists of one and a is declining in incidence in the Western intake as risk factors for cancer of the upper half whole time equivalent general practi- world and is strongly associated with digestive tract (this has been extensively tioners, three nurses, and reception staff in alcohol and tobacco intake, whereas adeno- studied) but to search for potential differ- three surgeries (Dover, Folkestone, and carcinoma (which is associated with the ences in the effects of the three types of Thanet). All staff have appropriate skills and premalignant condition Barrett’s oesopha- alcohol. Unless the effect of smoking and access to translating and other services. The gus) is rapidly increasing in incidence and type of alcoholic beverages interact, the service should lead to the registration of does not appear to be so strongly associated need to estimate risk among smokers and many more refugees and asylum seekers with smoking and alcohol, although further non-smokers is less prominent. with a general practitioner and to an under- information is needed.2 We appreciate that As Cheng et al point out, a few studies standing of the use of services by these in the Danish study it would have been diffi- have shown a protective effect of fruit and patients so that they can in time be properly cult for the authors to obtain information vegetable intake on oesophageal cancer. To integrated into the NHS. on the types of oesophageal cancer, since be a confounder in the present analysis, The scheme is funded by the NHS the cancer registry does not separate the however, the effects of intake of fruit and modernisation fund, and the cost of nursing histopathological subtypes. In a Medline vegetables and its association with wine is met by an allocation from the health search (1995-8) on the relation between intake would have to be unexpectedly strong authority’s service and financial framework, oesophageal cancer and alcohol only five of to explain our findings. with agreement that any overspend will 12 papers specified which histopathological If we had had a larger number of cases come directly from cash limited funds for subtype they were referring to. we would have been able to distinguish general medical services. In the future the It is time that cancer registries classified between these types of cancer and even the local medical committee will need to oeosphageal adenocarcinoma separately histopathological subtypes, as suggested by consult with primary care groups, five in the from squamous carcinoma so that we Fitzgerald and Caygill. If we misclassified case of East Kent. We wonder how each will can gain meaningful information about some of the cases we would expect this to respond as refugees and asylum seekers are the epidemiology of these important have reduced the differences in the effects of currently concentrated in comparatively conditions. beer, wine, and spirits. small and deprived areas. Although general practitioners in Kent Rebecca Fitzgerald Lecturer in gastroenterology Morten Grønbæk Senior research fellow St Bartholomew’s and the Royal London School of Thorkild I A Sørensen Professor might support the government’s aim of dis- Medicine and Dentistry, London E1 2AD Copenhagen Centre of Prospective Population persal, the reality is likely to be much more [email protected] > Studies, Danish Epidemiology Science Centre at difficult to achieve than a properly funded the Institute of Preventive Medicine, Copenhagen Christine Caygill Epidemiologist initiative that addresses the problems faced University Hospital, Kommunehospitalet, DK-1399 National Barrett’s Oesophagus Registry, Wexham Copenhagen K, Denmark by refugees and asylum seekers and the Park Hospital, Slough, Berkshire SL2 4HL [email protected] health and social services that are trying to support them. 1 Grønbæk M, Becker U, Johansen D, Tønnesen H, Jensen G, Sørensen TIA. Population based cohort study of the David Barr Clerk association between alcohol intake and cancer of the upper [email protected] digestive tract. BMJ 1998;317:844-7. (26 September.) 2 Vaughan TL, Davis S, Kristal A, Thomas DB. Obesity, Meeting the healthcare needs J B Ashton Medical secretary alcohol and tobacco risk for cancers of the esophagus and Kent Local Medical Committee, Suite 1, gastric cardia: adenocarcinoma vs squamous cell of refugees and asylum First Floor, Tenacre Court, Herrietsham, Kent carcinoma. Cancer Epidemiol, Biomarkers and Prev 1995;4: ME10 1AH 85-92. seekers Editor—Kent has around 4000 asylum 1 Home Office. Fairer, faster, and firmer—a modern approach to immigration and asylum. London: Stationery Office, 1998. seekers, many of whom are not registered (Cm 4018.) Authors’ reply with a general practitioner. The local medi- Editor—The large number of cases in La cal committee recently wrote to all the Vecchia et al’s case-control study enables members of parliament (MPs) in Kent seek- them to present a detailed analysis of the ing separate central funding to provide effect of wine only and wine together with adequate services for refugees and asylum other types of drinks. Their study shows seekers. The committee was concerned that what ours cannot—namely, that even with an effective service was unlikely to be wine there is an increased risk of the provided unless funds were separately iden- cancers. As Cheng et al point out, our study tified and also that funds for general

BMJ VOLUME 318 8 MAY 1999 www.bmj.com 1291