Rapid Responses to:

PRIMARY CARE: Anders Beich, Thorkil Thorsen, and Stephen Rollnick Screening in brief intervention trials targeting excessive drinkers in general practice: systematic review and meta-analysis BMJ 2003; 327: 536-542 [Abstract] [Full text]

Rapid Responses: Submit a response to this article

Rapid Responses published:

ERRATUM table 5

Anders Beich   (5 September 2003)

Role of Public Health community in identifying the problem drinkers

Ediriweera B.R., Desapriya   (10 September 2003)

Number needed to screen and the prevention paradox

Daniel C. Vinson   (11 September 2003)

Transcription errors and erroneous assumptions

Peter Anderson   (13 September 2003)

Why do Beich et al. ignore selective screening?

Nick Heather   (13 September 2003)

Neither errors nor erroneous assumptions – only disagreement and emotional misunderstanding

Anders Beich, Thorkil Thorsen   (15 September 2003)

‘The Teachable Moment’ – opportunistic intervention for alcohol misuse.

Robin Touquet, Robert Patton, Michael Crawford, and James S Huntley.   (15 September 2003)

Bad news for GP’s about alcohol counseling

Jean-Bernard Daeppen, Nicolas Bertholet, and Bernard Burnand   (18 September 2003)

Screening and brief intervention in primary health care - worth doing in real life situations

Kaija Seppä, Mauri Aalto   (20 September 2003)

Good doctors still enquire about their patient's drinking habits

John R Kemm   (20 September 2003)

Misleading ‘Editorial’

Tim Rapley   (23 September 2003)

INEFFECTIVE ALCOHOL SCREENING – BAD NEWS OR GOOD NEWS?

Anders Beich, Thorkil Thorsen   (26 September 2003)

Alcohol screening as good as other recommended screens

Richard Saitz   (28 September 2003)

Problem drinkers in Primary Care have complex needs

Francis Labinjo   (30 September 2003)

Screening effect ten times greater than calculated by Beich et al.

Nick Heather, Nick Heather, Robyn Richmond   (1 October 2003)

Doctors should ask every patient about alcohol

Katherine M Conigrave, Elizabeth M Proude and John B Saunders   (3 October 2003)

Re: Screening effect ten times greater ... (by Heather and Richmond)

Anders Beich, Thorkil Thorsen   (5 October 2003)

PROBLEMS IN USING NNT FOR STUDIES OF EXCESS DRINKING

Ivar S. Kristiansen, Torbjorn Wisloff   (8 October 2003)

Re: PROBLEMS IN USING NNT FOR STUDIES OF EXCESS DRINKING

Anders Beich, Thorkil Thorsen   (9 October 2003)

Re: Saitz R. Alcohol screening as good as other recommended screens

Anders Beich, Thorkil Thorsen   (20 October 2003)

Beich et al. have got it wrong again

Nick Heather   (24 October 2003)

Further errors in paper by Beich et al

Paul G Wallace, Andy Haines   (24 October 2003)

Re: Beich et al. have got it right again

Anders Beich   (24 October 2003)

Systematic screening for alcohol hazardous consumption :

Philippe Michaud, Jean-Pierre Aubert   (30 October 2003)

Re: Further errors in paper by Beich et al (Response to Wallace and Haines)

Anders Beich, Thorkil Thorsen   (31 October 2003)

Misrepresentation of published research

Michael F. Fleming   (5 November 2003)

Re: Misrepresentation of published research (reply to Fleming)

Anders Beich, Thorkil Thorsen   (10 November 2003)

Still serious reservations regarding this meta-analysis' conclusions

Richard Saitz   (13 November 2003)

Calculations correct - conclusions depend on cultural background (reply to Saitz)

Anders Beich, Thorkil Thorsen   (30 November 2003)

Intention-to-treat analysis was over-zealous - but this does not affect findings

Ian R. White   (9 December 2003)

Beich et. al. Need to Make Data Analyses Available for Review

Katharine A Bradley   (14 December 2003)

Does Beich et al add to the existing evidence?

Olivia Wu, Robin Knill-Jones   (17 December 2003)

On accuracy of findings, the need for further analyses, and generalization beyond the trial context

Anders Beich, Thorkil Thorsen   (30 January 2004)

On our use of the word "opponents" - we meant critics

Anders Beich   (2 February 2004)

Playground football and the decline of respectful debate: alcohol screening and general practice

Stephen Rollnick   (7 March 2004)

ERRATUM table 5 5 September 2003
Anders Beich, research fellow Central Res.Unit Gen.Pract., University of Cph., Panum Institute, 2200 Copenh.N, Denmark Send response to journal: Re: ERRATUM table 5	As a strange result of the editing process a few minor errors have been introduced into the Table 5 (screening effect for non-binge drinking) of our paper. In the study by Fleming (ref42) the screening effect (last column) figures should be: 3.7 (-0.3 to 7.5). In the study by Ockene (ref30) the max.no.of drinks (2nd column) should be: 5 drinks for men and 4 for women. In the study by Fleming (ref41) the max.no.of drinks (2nd column) should be: 5 drinks for men and 4 for women. In the study by Anderson (ref39) the max.no.of drinks (2nd column) should be: 11 drinks (men). On behalf of the authors I apologize for these disturbing errors. Anders Beich Competing interests:   None declared
Role of Public Health community in identifying the problem drinkers 10 September 2003
Ediriweera B.R., Desapriya, Research Associate BC injury Research and Prevention Unit, Centre for community Child health Research, BC, V6H 3V4 Send response to journal: Re: Role of Public Health community in identifying the problem drinkers	Role of public health community in identifying problem drinkers Beich, A., Thorkil Thorsen, T., Rollnick, S., (1) recent systematic review shows that screening in general practice does not seem to be an effective precursor to brief interventions targeting excessive alcohol use. Approximately 78% of Canadians consult a physician each year. Of these 6% are heavily dependent on alcohol, and up to 25% have or are at risk for alcohol-related health problems. (2)About 10% of premature death in Canada is caused by hazardous drinking, and more than 50% of fatal traffic crashes involve alcohol.(3) By assessing and treating alcohol abuse and alcoholism in their patients physicians can play an important role in preventing some traffic deaths and injuries. Individuals arrested for DWI have been targeted for preventive intervention of alcohol related problems. However in literature it was found that the high rates of diagnosed alcoholism among DWI arrestees suggest a need to identify individuals at risk earlier in the developmental process. Highway safety advocates and public health professionals have therefore suggested that an arrest for driving while impaired (DWI) offers a valuable opportunity to identify persons with substance-abuse disorders and refer them for treatment. But identifying offenders with these problems is challenging. Studies suggest that offenders in court-mandated screening programs often provide inaccurate information about their alcohol use and consequences and criminal histories. Offenders may fear recriminations from admitting to illegal drug use. Dissimulation may be motivated by fear of being labeled or mandated to undergo treatment for alcohol and drug problems. Denial, hostility, and suspicion also may contribute to under-reporting. The health, social and economic costs of alcohol abuse may be as high as $8.6 billion, of which $1.3 billion is spent on direct health care costs.(4)Physicians often fail to identify or are reluctant to intervene in cases of problem drinking, perhaps because of a lack of awareness, uncertainty about how best to intervene or doubts about the effectiveness of intervention.(5,6)There is, however, evidence that efforts to identify and treat problem drinkers can be cost effective (7) and that even brief low-cost behavioral interventions suitable for use in general health care settings are often helpful. (8-10) Although relapses are common following abstinence-oriented treatments many studies have shown that treatment can substantially reduce alcohol use and improve functioning in other areas of life as well.(5,6,7,8910) REFERENCES: 1. Beich, A., Thorkil Thorsen, T., Rollnick, S., Screening in brief intervention trials targeting excessive drinkers in general practice: systematic review and meta-analysis BMJ 2003;327:536-542 2.Alcohol Risk Assessment and Intervention (ARAI): A resource manual for family physicians. Mississauga (ON): College of Family Physicians of Canada; 1994 3. Smart RG, Ogborne AC. Northern spirits: A social history of drinking in Canada. Toronto: Addiction Research Foundation; 1998 4.Single E, Robson L, Xie X, Rehm J. The costs of substance abuse in Canada. Ottawa: Canadian Centre on Substance Abuse; 1996. Available: www.ccsa.ca/costhigh.htm (accessed 2003 Sep.7) 5.Negrete JC. The role of medical schools in the prevention of alcohol- related problems. CMAJ 1990;143(10):1048-53 6.Rankin JG, Ashley MJ, Brewster JM, Chow YC, Single E, Skinner HA. Preventing alcohol problems: preparing physicians for their roles and responsibilities [editorial]. CMAJ 1990;143(10):1005-6 7.Holder H, Longabaugh R, Miller WR, Rubonis AV. The cost effectiveness of treatment for alcoholism: a first approximation. J Stud Alcohol 1991;52:517-40 8.Hester RK, Miller WR. Handbook of alcoholism treatment approaches: effective alternatives. 2nd ed. New York: Plenum Press; 1994 9.Bien TH, Miller WR, Tonigan JS. Brief interventions for alcohol problems: a review. Addiction 1993;315-35 10.Roberts G, Ogborne AC. Best practices in substance abuse treatment and rehabilitation. Ottawa: Ministry of Public Works and Government Services Canada; 1999. Competing interests:   None declared
Number needed to screen and the prevention paradox 11 September 2003
Daniel C. Vinson, Professor University of Missouri-Columbia Send response to journal: Re: Number needed to screen and the prevention paradox	Beich and colleagues(1) have raised an important question. What is a reasonable number needed to screen to improve one person’s health? For alcohol problems, that number is about 400. Is that too high? Firstly, consider the complexity of screening. A single question is reasonably sensitive and specific.(2) Its ease of use would reduce the work of screening, and that might make the large number needed to screen more acceptable. Secondly, consider the number needed to screen for other diseases, for example, hypercholesterolemia. If the prevalence of that condition is 25%, if half of those who screen positive accept treatment, and if the number needed to treat is 40 to 50 (it was 44 in the West of Scotland study(3)), one would improve the health outcomes of 2 or 3 patients if 1000 were screened, a number needed to screen of about 400. Beich and colleagues’ study is an example of the prevention paradox,(4,5) that few patients personally benefit from preventive interventions. Although there are fewer studies that brief alcohol interventions change patients’ health outcomes(6) than in pharmacotherapy for cholesterol, screening for alcohol problems is neither more nor less caught in that paradox. Dan Vinson, M.D. Department of Family and Community Medicine University of Missouri-Columbia Columbia, MO 65212 U.S.A. REFERENCES 1. Beich A, Thorsen T, Rollnick S. Screening in brief intervention trials targeting excessive drinkers in general practice: systematic review and meta-analysis. BMJ. 2003;327(7414):536-542. 2. Williams RH, Vinson DC. Validation of a single question screen for problem drinking. J Fam Pract. 2001;50(4):307-312. 3. Shepherd J, Cobbe SM, Ford I, et al. Prevention of coronary heart disease with pravastatin in men with hypercholesterolemia. N Engl J Med. 1995;333(20):1301-1308. 4. Rose G. Strategy of prevention: lessons from cardiovascular disease. Br Med J (Clin Res Ed). 1981;282(6279):1847-1851. 5. Rose G. Sick individuals and sick populations. Int J Epidemiol. 1985;14(1):32-38. 6. Fleming MF, Mundt MP, French MT, Manwell LB, Stauffacher EA, Barry KL. Brief physician advice for problem drinkers: long-term efficacy and benefit-cost analysis. Alcohol Clin Exp Res. 2002;26(1):36-43. Competing interests:   None declared
Transcription errors and erroneous assumptions 13 September 2003
Peter Anderson, Independent Consultant in Public Health Nijmegen University 6525HC Netherlands Send response to journal: Re: Transcription errors and erroneous assumptions	Incorrect data extraction Two of the eight studies that form part of the meta-analysis were mine (Scott & Anderson 1990; Anderson & Scott 1992). There are substantive errors in the numbers extracted from these studies and reported by the authors in tables 1, 2, and 3. Consequent to this there is an inaccuracy in the proportion of screened patients given the intervention reported in table 4. For example, according to the methodology of the authors, the figure for the men (Anderson & Scott 1992) should be 1.8% and not 3.05%, which actually makes the screening effect lower than that calculated in table 4 (but see below). The number of standard drinks has been calculated at 12g per drink (see footnote to table 2). This is certainly incorrect for the UK studies. Erroneous assumptions when calculating the screening effect The authors confuse research studies with normal practice, which affects their estimates of the screening effect, thus leading to flawed conclusions. For exampl, in our study of men (Anderson & Scott 1992), we used a screening questionnaire which included quantity frequency questions to measure alcohol consumption to identify the sample to include in the trial. 8483 men from 8 general practices completed the questionnaire. In normal general practice (and certainly how my own general practice operated at the time), interventions would be based on the results of this screening questionnaire. That is those people who had high alcohol consumption would either be offered advice to reduce their consumption or have their notes tagged, so that the next time they visited the practice they would be offered advice. 6% of the men scored positive on high alcohol consumption, so the proportion of screened patients given the intervention would normally be 6% or very close to it, and not the 3.05% reported in table 4 by the authors (which is inaccurate anyway - see above). What we did in the study was first to remove 20% (105) of the 524 identified heavy drinkers who then formed a part of another study. The remaining 419 were invited to an assessment interview for the design purposes of the study (such an interview would not be part of normal practice). 49% (205) attended the asssessment interview. Of these, 194 were heavy drinkers during the past week (a different measure to the quantity frequency questionnaire). We removed 40 of the 194 heavy drinkers during the past week who consumed more than 1050 grammes per week, as the protocol stipulated that we were studying the impact of the intervention on those consuming 350-1050 grammes of alcohol during the previous week. Again these heavier drinkers would normally be part of an intervention in normal general practice. This left us with 154 heavy drinkers randomly allocated to a control group and an intervention group. Thus the correct denominator to calculate the proportion of the screened patients given the intervention is not 8343, but just less than one third, 2640, adjusting for the above study designs. Thus the proportion of screened patients given the intervention is 154/2640=5.8%, similar to the 6% above. But, even without these errors, the conclusions by the authors are based on other erroneous assumptions. Health screening in general practice is not just about alcohol. It is about smoking, overweight, raised blood pressure etc. So the utility of screening has to be based on broader assumptions. A general practice can target its methods of screening to particular population groups or at particular times, such as new patient registration, which might decrease its workload and increase its efficiency. It is not correct to state that screening is not an effective precursor to brief interventions, without either undertaking cost effectiveness analysis or comparing it to other health interventions. In terms of health gain, screening and brief interventions for hazardous/harmful alcohol use might be a very cost effective intervention. No information is given about this. References Scott, E. & Anderson, P. (1990) Randomized controlled trial of general practitioner intervention in women with excessive alcohol consumption. Drug and Alcohol Review 10, 313-321; Anderson, P. & Scott, E. (1992) The effect of general practitioners advice to heavy drinking men. British Journal of Addiction 87,891-900. Competing interests:   None declared
Why do Beich et al. ignore selective screening? 13 September 2003
Nick Heather, Emeritus Professor School of Psychology & Sport Sciences, Northumbria University, Newcastle upon Tyne NE1 8ST Send response to journal: Re: Why do Beich et al. ignore selective screening?	It seems that the article by Beich and colleagues [1] is riddled with errors [2] and is seriously flawed in other ways [3]. Accepting for the moment the validity of their findings, however, the authors base their conclusions on a misleading dichotomy of “universal” screening versus no screening at all. As was pointed out [4] in relation to a previous publication from this group [5], a middle way between these extremes is to target screening at specific types of consultation – special clinics in which the proportion of positive cases can be expected to be higher than average and general health checks, new patient registrations etc. at which enquiries about drinking are more acceptable to both patients and health professionals. A recent Delphi study of relevant expert opinion in the UK [6] concluded that selective screening of this kind was to best way to promote the widespread implementation of screening and brief intervention for excessive alcohol consumption in primary health care. In the Accident and Emergency setting, Huntly and colleagues [7] described a list of the “top ten” indications for the identification of excessive drinkers and a similar list could usefully be provided for the primary health care setting. By ignoring the possibility of selective screening, Beich et al. imply that the only alternative to screening all patients attending the practice is to confine attention to those with obvious alcohol-related problems. This is a surrender to a practice in which advice and help is offered only when it is already too late to prevent harm. The consequences of this for alcohol policy in the UK would be disastrous and it can only be hoped that those responsible for finalising the National Alcohol Harm Reduction Strategy are made aware of the flaws in Beich et al.’s article. In their keenness to reduce the GPs’ workload, Dr. Beich and his colleagues do a grave disservice to patients whose excessive drinking put them at risk of medical and other kinds of damage. Nick Heather PhD, Emeritus Professor of Alcohol & Other Drug Studies, School of Psychology & Sport Sciences, Northumbria University, Newcastle upon Tyne NE1 8ST REFERENCES 1. Beich A, Thorsen T, Rollnick S. Screening in brief intervention trials targeting excessive drinkers in general practice: systematic review and meta-analysis. BMJ 2003;327:536-542. 2. Beich, A. Erratum table 5. Rapid response bmj.com 2003; 5 September. 3. Anderson, P. Transcription errors and erroneous assumptions. Rapid response bmj.com 2003; 12 September. 4. Heather N, Anderson P, Gual A & Seppa K. Some screening is necessary to identify excessive drinkers early in primary care. BMJ 2003;326:550. 5. Beich A, Gannik D and Malterud K. Screening and brief intervention for excessive alcohol use: qualitative interview and study of the experiences of general practitioners. BMJ 2002;325:870-872. 6. Heather N, Dallolio E, Hutchings D, Kaner E & White M. Implementing routine screening and brief alcohol intervention in primary health care: a Delphi survey of expert opinion. J Substance Use; in press. 7. Huntly J S, Blain C, Hood S & Touquet R. Improving detection of alcohol misuse in patients presenting to accident and emergency departments. Emergency Medicine Journal,2001;18:99-104. Competing interests:   None declared
Neither errors nor erroneous assumptions – only disagreement and emotional misunderstanding 15 September 2003
Anders Beich, Research fellow Central Res.Unit Gen.Pract., University of Cph., Panum Institute, 2200 Copenh.N, Denmark, Thorkil Thorsen Send response to journal: Re: Neither errors nor erroneous assumptions – only disagreement and emotional misunderstanding	Unfortunately, it seems hard to have an open and non-emotional discussion about the evidence base for screening and brief intervention seen from different perspectives. Pressure and desperation characterizes Peter Anderson’s response to our article (1). He claims that we have not used the right data from his study for our analysis (which we have) and claims that we have made miscalculations (which we have not). Had Dr. Anderson taken his time to read our article thoroughly and open-minded, he would have known the premises for our calculations. And he might better have understood our conclusions. For instance, he might then have been able to see that in order to make comparisons easier, all units of alcohol were correctly converted into 12 grams standard drinks, depending on the number of grams of alcohol per unit stated in each paper, his own included. We have used exactly the same numbers as he puts forward himself in his response: the data that are available in his paper (2). The only real disagreement here is the classical one about the denominator. The facts are once again: 8483 men were screened, 524 (6.2%) screened positive, 419 were invited for an assessment interview, whereas 105 were put in another arm of the study. Of the 419 less than half (205) showed up, another 40 were excluded because they drank too much (!!). The remaining 154 were randomized into two groups; the third arm of the study had 105. To render unto Caesar the things that are Caesar's we added 154 and 105 to get 259 (3.05%) who might have been receptive to the brief intervention offered at baseline. Peter Anderson says that the 105 should not have been included in this figure, so that the percentage is only 1.8%. On the other hand he also claims that under normal circumstances all 524 that screened positive would have been offered brief intervention. But the truth is that at least 245 of them were non-receptive as defined by Feinstein (3) and 40 were excluded because they drank much too much. Anderson had drinking data on 100 patients at follow-up. To generalize the findings from these 100 patients to everyday practice to everyone who screen positive would make no sense. We said in our review: according to the data available 259 patients (3.05%) might have been receptive to the brief intervention offered in this study. Peter Anderson correctly states that health screening in general practice is not only about alcohol. Other public health experts have trotted out their favourite ideas for things to screen for. We propose that the utility of screening should be studied for each of these other areas mentioned, as well. Dr. Anderson finds it incorrect of us to state that screening is not an effective precursor to brief interventions, without either undertaking cost effectiveness analysis or comparing it to other health interventions. Again, we can refer to our article where the issue of cost-effectiveness is touched upon. Screening 1000 to find (at best) two or three that report drinking less after one year is not effective seen from the consultation room. It is not possible to conduct reliable cost-effectiveness analyses on screening and brief intervention in general practice before we know the effectiveness of usual case finding and usual care within a framework that respects the dynamics of a clinical encounter. To date no one has compared the two. It has been assumed that the effectiveness of usual (excessive drinking) case finding and care was zero. We regard this assumption as wrong. And we urge others to embark upon scrutinizing the whole issue further. 1. Beich A, Thorsen T, Rollnick S. Screening in brief intervention trials targeting excessive drinkers in general practice: systematic review and meta-analysis. BMJ 2003; 327: 536-42. 2. Anderson P, Scott E. The effect of general practitioners' advice to heavy drinking men. Br J Addict 1992; 87: 891-900. 3. Feinstein AR. Clinical epidemiology. The architecture of clinical research. Philadelphia: Saunders, 1985. Anders Beich Thorkil Thorsen Competing interests:   None declared
‘The Teachable Moment’ – opportunistic intervention for alcohol misuse. 15 September 2003
Robin Touquet, Consultant Accident & Emergency, St Mary's Hospital, Paddington, London, W2 1NY, Robert Patton, Michael Crawford, and James S Huntley. Send response to journal: Re: ‘The Teachable Moment’ – opportunistic intervention for alcohol misuse.	EDITOR - Getz et al are right to say it is good medical practice to support health-promoting activities (1). One example quoted is the ‘teachable moment’ for alcohol misuse. This uses the presenting complaint, e.g. fall, collapse, head injury, assault, accident, for the Accident and Emergency Department (A&E) as the ‘learning opportunity’ (2). Thereby the patient may start to develop insight into the consequences of their drinking behaviour. Brief intervention (BI) is reported to be effective in the Emergency Room (ER) following injury, especially when carried out by Alcohol Health Workers secondary to initial detection from the medical or nurse practitioner dealing with the patient’s presenting complaint (3). Although the consequences of alcohol misuse are explained to the patient, to encourage the take up of the offered appointment with the AHW (4), the actual BI is not carried out in the initial consultation. Therefore this is not an additional role for the practitioner - that will create extra work and therefore possible stress – be it the General Practitioner or the A&E practitioner (nurse or doctor). Rather it is stress relieving to refer the patient on for BI – which is time consuming and requires special skills – in anticipation that the likelihood of reattendance and thereby further work is lessened. We question whether the above has been taken into account by Beich et al (5), who do not specify 1.) If screening was selective for presenting conditions known to be associated with alcohol misuse (e.g. indigestion or lack of sleep) and 2.) If the BI - secondary to initial positive screening - was carried out by AHWs. Screening needs to have a logical link to the patient’s reason for initial consultation(1) if it is to be a true ‘teachable moment’; it also then ethically correct, perhaps medico-legally correct as well! Many ‘Teachable moments’ for underlying alcohol misuse are available in hospital, e.g. A&E, Facio-maxillary clinics, fracture clinics, Sexually Transmitted Disease clinics, to name but a few, as well as in General Practice. However these must not place additional workload on the practitioner, thereby risking clinical inertia. Rather extra resources are needed to fund AHWs, not only to provide expert BI, but also to provide training, encouragement and feedback for the referring doctors and nurses. Robin Touquet Professor Accident & Emergency Medicine, St Mary’s Hospital, Praed Street, London W2 1NY Robert Patton Research Associate Michael Crawford Senior Lecturer Psychological Medicine, Imperial College London, Charing Cross site, W6 8RP James S Huntley Lecturer Orthopaedic Surgery New Royal Infirmary of Edinburgh, Old Dalkeith Road, Edinburgh EH16 4SU 1 Getz L, Sigurdsson JA, Hetlevik I. Is opportunistic disease prevention in the consultation ethically justifiable? BMJ 2003;327:498- 500. (30th August.) 2 Huntley JS, Blain C, Hood S, Touquet R. Improving detection of alcohol misuse in patients presenting to an accident and emergency department. EMJ 2001;18:99-104. 3 Longabaugh R, Woolard RF, Niremberg TD, Minugh AP, Becker B Clifford PP et al. Evaluating the effects of a brief intervention for injured drinkers in the emergency department. Journal of studies on alcohol 2001;62:806- 816. 4 Patton R, Crawford MJ, Touquet R. Impact of health consequences feedback on patients acceptance of advice about alcohol consumption. EMJ 2003:20:451-452. 5 Beich A, Thorsen T, Rollnick S. Screening in brief intervention trials targeting excessive drinkers in general practice: systematic review and meta-analysis. BMJ 2003;327:536-540. (6th September) Competing interests:  The first 3 authors are in receipt of a grant from the Alcohol Education & Research Council (AERC) for a RTC trial of written advice for alcohol misuse compared to written advice and BI (work in preparation).
Bad news for GP’s about alcohol counseling 18 September 2003
Jean-Bernard Daeppen, Alcohol Treatment Center 1011 Lausanne, Nicolas Bertholet, and Bernard Burnand Send response to journal: Re: Bad news for GP’s about alcohol counseling	Beich and colleagues give bad news to primary care doctors concluding their meta-analysis of the brief alcohol intervention literature saying that only two to three patients per thousand screened will benefit from brief alcohol interventions (1). We question the rationale of using data extracted from studies aimed at assessing the efficacy of brief intervention to conclude on the effectiveness of the screening procedure. In studies designed to evaluate the efficacy of a therapeutic intervention, it is important to ensure internal validity, and thus inclusion and exclusion criteria have to be restrictive. For instance, in order to guarantee a high proportion of follow-up data, subjects are required to have a stable home address, provide significant others’ addresses, and speak the official trial language. Some of the brief intervention trials excluded patients who received advice to cut down on drinking in the last 6 months or last year. Nevertheless, their exclusion could not be considered as lack of efficacy of the screening process. We agree that the implementation of screening for hazardous alcohol use in primary care is a major challenge. A recent study conducted by Beich and colleagues described some of the difficulties GPs might experience when conducting systematic screening in their daily medical practice, including that GPs were convinced that patients did not respond honestly to screening, heavy drinkers refused screening, screening was associated to adverse effects on the doctor-patient relationship (2). Additional studies are needed to optimize systems of screening, improve acceptability both from patients and health care providers’ perspectives. Because of a 25 % prevalence of hazardous drinkers in primary care (3) and screening performance that seems similar to other screening procedures – typical of community based preventive interventions-, the further development and implementation of screening and brief alcohol intervention is a promising area of development for primary care. In other word we suggest that Beich and colleagues’ study does not adequately address the question of the effectiveness of screening and that one major challenge here is the optimal teaching strategies aimed at improving students’ and GPs skills to conduct this kind of prevention measures (4). Jean-Bernard Daeppen, MD Alcohol Treatment Center Mont-Paisible 16 University Hospital 1011 Lausanne Switzerland Jean-Bernard.Daeppen@inst.hospvd.ch Nicolas Bertholet, MD Alcohol Treatment Center and Clinical Epidemiology Center, University Hospital, Lausanne, Switzerland Bernard Burnand, MD, MPH Clinical Epidemiology Center and Health Care Evaluation Unit, University of Lausanne, Switzerland References 1. Beich A, Thorsen T, Rollnick S. Screening in brief intervention trials targeting excessive drinkers in general practice: systematic review and meta-analysis. Bmj 2003;327(7414):536-42. 2. Beich A, Gannik D, Malterud K. Screening and brief intervention for excessive alcohol use: qualitative interview study of the experiences of general practitioners. Bmj 2002;325(7369):870. 3. Reid MC, Fiellin DA, O'Connor PG. Hazardous and harmful alcohol consumption in primary care. Arch Intern Med 1999;159(15):1681-9. 4. Yedidia MJ, Gillespie CC, Kachur E, Schwartz MD, Ockene J, Chepaitis AE, et al. Effect of communications training on medical student performance. Jama 2003;290(9):1157-65. Competing interests: None declared
Screening and brief intervention in primary health care - worth doing in real life situations 20 September 2003
Kaija Seppä, Professor Medical School, Department of General Practice, FIN-33014 University of Tampere, FINLAND, Mauri Aalto Send response to journal: Re: Screening and brief intervention in primary health care - worth doing in real life situations	Beich et al. report figures which are very discouraging in relation to screening and brief alcohol intervention. (1) However, their technically excellent results do not take into account that in real life situations 1) patients do not have to sign informed consents and participate in studies, which presumably makes them more receptive to intervention, 2) they may (after intervention) reduce their drinking even though the amount is not big enough to classify them as ‘successful’ according to the scientifically strict criteria used in Beich et al.’s analysis, but in which a reduction may be meaningful in the process of changing behaviour or in preventing alcohol-related harm, 3) a smaller reduction may additionally have an impact on other alcohol-related symptoms e.g. depression. We calculated the total cost in Finland, if all adult patients registered with one GP (the number of patients used in this example was 1600) were screened either opportunistically or systematically. (unpublished) The cost of the more expensive choice, systematic screening, is of the same magnitude as the treatment cost of one acute pancreatitis at a university hospital (about 30 000€). (2) The corresponding cost of opportunistic screening preferred, for example, by GPs in Finland, is less than half of that. This comparison suggests that even the published result, (2.6/1000 patients benefit from brief intervention after systematic screening, which makes 4.2 patients in a population of 1600), although apparently an underestimation, would have a remarkable effect on public health. REFERENCES 1. Beich A, Thorsen T, Rollnick S. Screening in brief intervention trials targeting excessive drinkers in general practice: systematic review and meta-analysis. BMJ 2003;327:536-542. 2. Sillanaukee PA, Kääriäinen J, Sillanaukee P, Poutanen P, Seppä K. Substance use-related outpatient consultations in specialized health care; an underestimated entity. Alcohol Clin Exp Res 2002;26:1359-1364. Competing interests:   None declared
Good doctors still enquire about their patient's drinking habits 20 September 2003
John R Kemm, Public Health Physician Birmingham Send response to journal: Re: Good doctors still enquire about their patient's drinking habits	The useful review by Beich, Thorsen and Rollnick (BMJ September 2003) provides no grounds to justify the sweeping conclusion that “Only two or three patients per thousand will benefit from the laborious activities entailed in screening” or the comment on the front cover of the issue that “Screening for alcohol problems is largely ineffective in primary care”. These fallacious conclusions appear to rest on the assumption that unless the process leads to an intervention, which causes the patient to reduce their drinking, no benefit has resulted. The conclusions are wrong for the following reasons. First knowledge of drinking habits is a valuable aid to clinical assessment. Faced with a clinical problem the most probable diagnosis is often different for a heavy drinker and a light or non drinker. It is therefore as important for the doctor to collect and record information on drinking habits as for them to do so for smoking habits. Second it is important that patients should realise that their doctors are interested in drinking habits that can affect health. “Just asking can make a difference”. Third the paper confuses screening to select patients for entry into a trial of a particular intervention with screening to identify patients who might benefit from a whole range of different interventions. Fourth in areas as complex as human behaviour it is unreasonable to classify all outcomes other than reduced drinking as failure. At the very least brief interventions are likely to produce shift from pre- contemplation to contemplation and it may well move the decisional balance in a direction that produces change at a later date. Finally the use of the term laborious to describe screening is unfortunate. Given a reasonably organised surgery it is very simple to ask patients to complete lifestyle questionnaires and to file these with the notes. Of course there is scope for discussion as to how the systematic collection and recording of information on drinking habits can best be integrated into the daily work of a busy general practice. There is also need for discussion as to how select patients for intervention and how to make those interventions more effective. However it is not reasonable to suggest that primary care should be taking less interest in patient’s drinking behaviour or be less concerned to help those, who drink in a risky or health damaging way, to change. Competing interests:   I am a board member of Medical Council on Alcohol and of Alcohol Concern
Misleading ‘Editorial’ 23 September 2003
Tim Rapley, Research Associate Centre for Health Services Research, University of Newcastle upon Tyne, NE2 4AA Send response to journal: Re: Misleading ‘Editorial’	The leader (1) that introduces the article by Beich and colleagues (2) is misleading. The opening sentence of the leader, in the section ‘This week in the BMJ’, notes that “Screening for excessive alcohol use and then providing brief interventions is not effective in general practice”. This statement suggests that both screening *and* brief interventions for excessive alcohol use are ineffective. Beich et al’s article provides no grounds to justify the description that brief interventions are ineffective in general practice or that brief interventions conducted after screening are ineffective in general practice. Their article was only concerned to engage with and comment on the research on the *screening component* of efficacy brief intervention trials. References 1. Anonymous. Alcohol screening in general practice is not effective. BMJ 2003; 327 2. Beich A, Thorsen T, Rollnick S. Screening in brief intervention trials targeting excessive drinkers in general practice: systematic review and meta-analysis. BMJ 2003;327:536-542. Competing interests:   None declared
INEFFECTIVE ALCOHOL SCREENING – BAD NEWS OR GOOD NEWS? 26 September 2003
Anders Beich, Research fellow Central Res.Unit Gen.Pract., University of Cph., Panum Institute, 2200 Copenh.N, Denmark, Thorkil Thorsen Send response to journal: Re: INEFFECTIVE ALCOHOL SCREENING – BAD NEWS OR GOOD NEWS?	Ediriweera and Desapriya (1) remind us that excessive drinking and alcohol-related problems are common and that even brief low-cost behavioural interventions suitable for use in general health care settings can be helpful. We agree on both issues. We found the average NNT for consumption below weekly limits in the trials meta-analysed to be 10 (7- 14). How the general practitioners can identify subjects among his/her patients for whom such an outcome of brief intervention in decency could be expected is another issue. Judging from the available evidence we conclude that screening does not seem to be an attractive alternative. Vinson (2) points out that our analysis exemplifies a more general prevention paradox: that only few patients personally benefit from preventive interventions. We believe that this paradox is actually generated by the systematic approaches to prevention suggested. Most of the preventive work that goes on already in general practice is not covered by this paradox. An experienced GP will typically use several types of knowledge and different sources of information (for example about the background, personality and present state of mind of the individual and functional capacity of his or her family at the moment) to decide when and where to address for example alcohol issues. Such information and knowledge cannot be provided by screening for risk factors extracted from large population studies. Daeppen (3) questions the rationale of using data extracted from studies aimed at assessing the efficacy of brief intervention to conclude on the effectiveness of the screening procedure. As examples he points out that some subjects might have been excluded from trials because they had no permanent address, did not speak the language or they had already received advice and not responded to it. In fact such cases were often lost or excluded even before the screening took place in the trials reviewed (as they might be in real life too). Anyway, we question that homeless people with alcohol problems would response to a GPs advice about drinking less (experiment or not). We also question that people who don’t speak the language would fit in the condensed SBI concept in everyday circumstances. And do we have reason to believe that subjects who have not yet responded to customized interventions will benefit from a subsequent brief standard intervention? We have pointed out, that in the trials we reviewed the considerable loss of 3 out of 4 subjects who had screened positive was first and foremost due to non-receptiveness, patients who screened false positive or “too positive” (or in some cases it was not specified why the subjects were lost). Only a few patients were lost because they were not accessible (i.e. died or moved away). Seppä and Aalto (4) as well as Kemm (5) put forward expectations for side benefit that might emerge in real life screening and brief intervention programmes. Intervention might cause some patients to reduce their drinking just a little bit, whereas others might move from precontemplation to contemplation in a process of change, and maybe screening has a beneficial effect in itself, they say. We have no wish to repudiate such speculative positive side effects and it was not within the aims of our meta-analysis to deal with them. We can however, easily think of negative side effects that might have to be considered as well; alienation, stigmatization, distrust and activation of defence reflexes being some of them. Touquet et al. (6) suggest that screening needs to have a logical link to the patient’s reason for initial consultation if it is to be a true ‘teachable moment’; it is also then ethically correct, and perhaps medico-legally correct as well, they say. We agree about the need for logic, timing and legitimacy, but we are not sure that screening is a very precise term to use then. Screening is by definition directed at (large) groups while the strategy rolled out by Touquet is targeted at individual cases. In the individual case we believe that the communication process (including establishing rapport, agenda setting, information exchange, and lowering resistance) is the hard part. This part requires skills and should be developed further in our opinion. We won’t comment on Nick Heather’s (7) unfriendly sidestepping the issue by attempting to raise doubts about the rigour of our review by referring to trivial errors that emerged from the editing process in one electronically published table. Regarding the opinion of Heather that our study “is seriously flawed in other ways”, we refer to our reply to Peter Anderson (8). We have reviewed universal screening as a precursor to brief intervention because it has been the widespread recommendation by WHO and others (Heather included) for several years. Now he suggests selective screening. Little is known about selective screening, but we urge Heather to review the evidence for selective screening in general practice and to include in his review the relation between case finding approach, the level of acceptance among patients and the overall effectiveness of such programmes. By referring to a study on expert opinions (by himself, not yet published) he takes us into a level in the hierarchy of evidence often flawed by attitude and personal interests and therefore usually given little attention in contemporary decision-making. We have taken note of the fact that the real disagreement is about the interpretation of the result of our meta-analysis rather than about the results as such. Among the responses to our meta-analysis we have not yet seen any substantial arguments against our conclusion that, seen from the consultation room, screening is at best a low-effective and not very attractive alternative for excessive drinking case-finding. Before any reasonable cost effectiveness calculations can be made we suggest that screening based programmes should be compared to patient-centred clinical approach like the one taking place already. Comparing screening only with screening and brief intervention is comparing two aliens. We consider our conclusion to be good news for anyone who truly believes that preventive work in health care is something else and much more than an increasing number of single risk factor or single disease focused standardized screenings and interventions. We believe that GPs should be much more directly involved in the process of quality improvement regarding the big issue of alcohol. No doubt that more could be done in general practice, especially if GPs were given influence on and ownership of future strategies, in stead of just being exposed to advanced implementation and marketing strategies for the dissemination of prevention programmes that might turn out to severely lack compatibility when put into practice (9). 1. Ediriweera BR, Desapriya. Role of Public Health community in identifying the problem drinkers. Rapid response bmj.com 2003; 10 September. 2. Vinson DC. Number needed to screen and the prevention paradox. Rapid response bmj.com 2003; 11 September. 3. Daeppen J-B. Bad news for GP’s about alcohol counselling. Rapid response bmj.com 2003; 18 September. 4. Seppä K et al. Screening and brief intervention in primary health care - worth doing in real life situations. Rapid response bmj.com 2003; 20 September. 5. Kemm JR. Good doctors still enquire about their patient's drinking habits. Rapid response bmj.com 2003; 20 September 6. Touquet R et al. ‘The Teachable Moment’ – opportunistic intervention for alcohol misuse. Rapid response bmj.com 2003; 15 September. 7. Heather N. Why do Beich et al. ignore selective screening? Rapid response bmj.com 2003; 13 September. 8. Beich A, Thorsen T. Neither errors nor erroneous assumptions – only disagreement and emotional misunderstanding. Rapid response bmj.com 2003; 15 September. 9. Beich A, Gannik D, Malterud K. Screening and brief intervention for excessive alcohol use: qualitative interview and study of the experiences of general practitioners. BMJ 2002; 325: 870-872. Anders Beich Thorjil Thorsen Competing interests:   None declared
Alcohol screening as good as other recommended screens 28 September 2003
Richard Saitz, Associate Professor of Medicine and Epidemiology Boston Medical Center + Boston University Schools of Medicine & Public Health, Boston, MA, 02118,USA Send response to journal: Re: Alcohol screening as good as other recommended screens	Beich et al. (1) set out to test the effectiveness of screening for decreasing excessive drinking, using efficacy trials. They conclude that screening is not an effective precursor to brief intervention and raise questions about the efficiency of universal screening in general practice. While I do not question the results of their meta-analysis, I believe they have drawn some incorrect conclusions. The proper study design to answer the question posed would be a randomized study of screening and brief intervention versus no screening (usual care case finding). No such studies were identified. Without citing such a study (e.g. Welte et al is an example though not randomized (2)) it is difficult to draw conclusions about the effectiveness of screening on clinical outcomes. In lieu of citing appropriate studies to answer the question, the authors reviewed randomized trials of brief intervention. Clinical trials are well known to sacrifice generalizeability for internal validity. It should be no surprise that many potential subjects are excluded. Beich et al. might say this is the point. But I believe clinicians are accustomed to generalizing the results of clinical trials beyond the highly selected populations studied. In the studies reviewed, a research screen preceded the brief intervention. The authors justify this choice of study by making the assumption that reasons for exclusion and dropout after a positive screening test, but before intervention, would be similar to what would occur in practice. This assumption is almost certainly incorrect. In clinical practice, there is no informed consent for research, no necessity to return for research visits for intervention, and no need to exclude patients with comorbidity or more severe alcohol problems. A screening test by a clinician could immediately be followed by feedback of results and further assessment and discussion (e.g. brief intervention). For example, I ask the screening questions and begin talking with the patient about their answers immediately. I do this with homeless individuals, people who do not speak my language (via an interpreter), and others who would not be in clinical trials. I am not unlike other primary care physicians in that respect. Admittedly, however, there are no randomized trials, nor will there likely ever be trials that prove effectivness in these patients. The authors state that a physician finding that 9% of those screened will screen positive could be disappointed, and imply that the 0.2-0.3% of those screened who will benefit from intervention is a small number (by using the word “only” and by their conclusions and recommendations). But these findings differ little from those for other conditions. In a randomized trial of hypertension treatment, 1% of those screened were enrolled, and the absolute risk reduction (ARR) for stroke was 3% in 5 years (number needed to treat [NNT] 33, compared to ARR 3% and NNT 10 for alcohol)(3). Similarly, a randomized trial of colorectal cancer screening found that fewer than half completed recommended annual screens, and the ARR for colorectal cancer mortality was 0.3% over 13 years (NNT 333)(4). As with Beich et al.’s finding of a “low” number who benefit, these data should not and have not dissuaded physicians from considering screening for hypertension and colon cancer to be standard practice. A seemingly low number who benefit in a clinical trial does not mean that screening is ineffective. Classic randomized trials are clearly not the right sources of data for drawing conclusions regarding the efficiency or effectiveness of screening. And doctors do not seem disappointed when they do not find colon cancer or benefit from hypertension screening. They are familiar with delivering preventive services routinely to all patients and know they are doing the right thing in clinical practice screening for colon cancer, breast cancer, hypertension, hypercholesterolemia, and providing vaccinations--all interventions guided not by specific patient characteristics beyond gender and age, and all proven effective by large population-based clinical trials. Universal screening with validated instruments is much better for identifying patients with alcohol problems than any currently known alternative. And the data summarized by Beich et al. support the notion that brief intervention after screening is efficacious. Pragmatic trials (5) and cost-effectiveness analyses in the future will likely find that alcohol screening followed by brief intervention is effective and cost- effective, much like other medical conditions routinely identified by screening. I do agree with Beich et al. that these data are needed, and that even when they are available, we will find that many people need to be screened to benefit few (as is true for most preventive interventions). Selective screening is a possible option but unlikely to be the best choice since risky drinking is best identified before consequences that would lead selective screening to occur. Scientists can respectfully disagree about the implications of valid results. The discussion about this paper seems to be just that kind of disagreement. Until further data are available for alcohol screening and intervention (cost-effectiveness analyses and effectiveness trials), I respectfully disagree with the conclusions drawn by Beich et al. and see no reason for recommendations for universal screening to change. 1. Beich A, Thorkil T, Rollnick S. Screening in brief intervention trials targeting excessive drinkers in general practice: systematic review and meta-analysis. 2003;327:536-42. 2. Welte JW, Perry P, Longabaugh R, Clifford PR. An outcome evaluation of a hospital-based early intervention program. Addiction. 1998 Apr;93(4):573 -81. 3. SHEP Cooperative Research Group. Prevention of stroke by antihypertensive drug treatment in older persons with isolated systolic hypertension. Final results of the Systolic Hypertension in the Elderly Program. JAMA 1991;265:3255-64. 4. Mandel JS, Bond JH, Church TR, et al. Reducing mortality from colorectal cancer by screening for fecal occult blood. N Engl J Med 1993;328:1365-1371. 5. Tunis SR, Stryer DB, Clancy CM. Practical clinical trials: increasing the value of clinical research for decision making in clinical and health policy. JAMA 2003;290:1624-1632. Competing interests:   None declared
Problem drinkers in Primary Care have complex needs 30 September 2003
Francis Labinjo, Consultant in Substance Misuse West Kent Health & Social Care NHS Trust 4 Manor Road Chatham ME4 6AG Send response to journal: Re: Problem drinkers in Primary Care have complex needs	Beich et al (1) raise the important question about the screening for excessive alcohol use in general practice. Wu et al(2)convey an even more serious message about problem drinkers who have other problems such as mental illness, where they found that the rate of help seeking is so low as to be a public health concern. The report (3) of an inquiry by the Advisory Council on the Misuse of Drugs, after assessing the impact of problem alcohol use on children, makes a key recommendation that problem alcohol use by pregnant women should be routinely recorded at the antenatal clinic and these data linked to stillbirths, congenital abnormalities in the new-born, and subsequent developmental abnormalities in the child. Johnson et al (4) in a study in South London of the co- occurrence of severe mental illness and substance use disorder found that 34% of patients misused alcohol only and a further 22% used alcohol and cannabis. The reality in primary care settings is therefore complex and multifaceted. Kendrick et al (5)looked at patient and practice factors associated with contact with specialist services. The only predictors they found, were whether or not the patient's practice offered a special service on-site and greater patient needs of care on the Camberwell Assessment of Need(CAN). In my opinion the way forward is to approach this problem from a multi-agency perspective within primary care, including general practice. References: 1. Beich A, Thorsen T, and Rollnick S. Screening in brief intervention trials targeting excessive drinkers in general practice: systematic review and meta-analysis BMJ 2003; 327: 536-542 2. Wu LT, Ringwalt CL, Williams CE. Use of substance abuse treatment services by persons with mental health and substance abuse problems. Psychiatric Service 2003; 54(3): 363-9 3. Hidden Harm. Responding to the needs of children of problem drug users. The report of an Inquiry by the Advisory Council on the Misuse of Drugs. Home Office. 2003 4. Miles H, Johnson S, Amponsah-Afuwape S. Finch E, Leese M, Thornicroft G. Characteristics of subgroups of individuals with psychotic illness and a comorbid substance use disorder. Psychiatric Services. 2003; 54(4):554-61 5.Kendrick T, Burns T, Garland C, Greenwood N, Smith P. Are specialist mental health services being targeted on the most needy patients? The effects of setting up special services in general practice. British Journal of General Practice 2000; 150: 121-126. Competing interests:   None declared
Screening effect ten times greater than calculated by Beich et al. 1 October 2003
Nick Heather, Emeritus Professor School of Psychology & Sport Sciences, Norhtumbria University, Newcastle upon Tyne, NE1 8ST, Nick Heather, Robyn Richmond Send response to journal: Re: Screening effect ten times greater than calculated by Beich et al.	After reading the recent article by Beich and colleagues(1), we went back to our 1995 article reporting an evaluation of the “Alcoholscreen” brief intervention programme in Sydney, Australia(2). We found that the findings of our study were seriously misrepresented in the Beich et al. article. In Table 4 they report the "proportions of sensible drinkers at follow-up" as 16/96 (16.67%) for the intervention group and 13/93 (13.98%) for the control group. The first of these is clearly the percentage of those in the intervention group reporting drinking under recommended levels at initial assessment, not at follow-up. The provenance of the second figure for the non-intervention control group is completely mysterious and does not correspond with any of the outcome percentages we reported in our paper. In fact, the proportions of good outcomes (i.e., those who had changed from being hazardous drinkers at initial assessment to “sensible” drinkers at 12-month follow-up) in our study were 7.3% in the intervention group and –5.4% in the control group, giving an Absolute Risk Reduction (ARR) of 12.7 and a Number Needed to Treat of roughly 8, rather than 2.7 and 37 respectively as calculated by Beich et al.. Equally misleading is the calculation of the prevalence of excessive drinking in the practice which the authors define as the number randomized to study groups (and therefore available for brief intervention) divided by the number screened. As has been pointed out(3), this ignores patients who were not offered brief intervention in the trials in question purely because of the requirements of a research study (i.e., those withholding informed consent to participate in research, those missed by the consent procedure or research assessment and those who would be difficult to trace for follow-up), but who would be offered brief intervention in real world conditions of general practice. Admittedly, we do not know what the outcome of intervention among such patients is likely to be, but it is clearly illegitimate to exclude them from the calculation of a screening effect in routine practice, as do Beich et al., because “the reasons for exclusion and dropout after a positive result on screening in the studies were similar to the reasons for the practitioner or the patient choosing to undergo no further assessment or intervention..." (p.541). One might also argue that patients excluded by the research protocol because of too high levels of dependence or problems and those who had previously received treatment for alcohol problems could also benefit from routinely- offered brief intervention by being encouraged to seek (further) treatment. In any event, subtracting from the number screened only those patients in our trial who met exclusion criteria (high dependence or problems, illness indicating abstinence, psychiatric disturbance, current or previous alcohol treatment, pregnant or planning pregnancy) gives a prevalence of 6.4%. Multiplying this by the ARR gives a screening effect of 8.1, 10 times the figures given by Beich et al.! The assumption on which our calculation is based, that patients who did not receive intervention in our trial for research-related reasons would have as good an outcome as those who did, is somewhat questionable, but far less questionable than the opposite assumption made by Beich et al. which is completely arbitrary, misleading and obviously tendentious. We do not know whether other trials used by the authors in their meta -analysis were as poorly represented as ours, but think this quite possible. Even if they were not, however, we find it depressing that an article founded on the kind of inaccurate calculations and highly dubious assumptions we have demonstrated in relation to our own study, should have been published in the BMJ, especially in view of the incalculable damage that this publication, if taken seriously, might do to the effort to reduce alcohol-related harm in Britain, Australia and elsewhere. Nick Heather, Emeritus Professor, School of Psychology & Sport Sciences, Northumbria University, Newcastle upon Tyne NE1 8ST, United Kingdom Robyn Richmond Professor School of Public Health and Community Medicine, University of New South Wales, Sydney NSW 2052, Australia REFERENCES 1. Beich A, Thorsen T, Rollnick S. Screening in brief intervention trials targeting excessive drinkers in general practice: systematic review and meta-analysis. BMJ 2003;327:536-542. 2. Richmond R, Heather N, Wodak A, Kehoe L, Webster I. Controlled evaluation of a general practice-based brief intervention for excessive drinking. Addiction 1995;90: 119-132. 3. Anderson, P. Transcription errors and erroneous assumptions. Rapid response bmj.com 2003; 12 September. Competing interests:   None declared
Doctors should ask every patient about alcohol 3 October 2003
Katherine M Conigrave, Staff specialist Drug Health Services, Royal Prince Alfred Hospital, Missenden Rd, Camperdown, Sydney NSW 2050, Elizabeth M Proude and John B Saunders Send response to journal: Re: Doctors should ask every patient about alcohol	Re: Screening in brief intervention trials targeting excessive drinkers in general practice: systematic review and meta-analysis Beich et al question the feasibility of screening in general practice for excessive alcohol use as a precursor to brief interventions. Most of the screening in the Beich meta-analysis was conducted during the consultation. In our experience doctors don't like using questionnaires, and will often selectively apply them rather than screening all patients. The questionnaires were identified in the report only as ‘general health or lifestyle questionnaires’: their validity is not clear or how long they took to complete. The rate of drinking problems detected was less than 1%, while a WHO study in primary health care settings identified 22% of subjects ‘at risk’ using the using the well validated 10 item AUDIT questionnaire. These figures are in keeping with a prevalence of alcohol dependence of 5% in the adult population and at least double that number drinking above recommended limits. [1] Rather than disrupt the medical consultation, we prefer the 3-minute WHO AUDIT questionnaire, which is completed by every new patient in the waiting room. The experience of family doctors who have used this model is that most patients are happy to complete the questionnaire and to receive advice. [2] The University of Sydney has developed a brief and simple intervention kit (Drink-less) which provides a practical and non- confronting means for the doctor to provide advice to the patient on drinking, modelled on the WHO's successful five minute intervention. Studies have consistently shown that brief intervention reduces alcohol consumption. The meta-analysis points out that we shouldn’t expect consumption to necessarily return to within recommended limits. With the known association between level of consumption and social and physical problems, any reduction is desirable. Several studies have documented reduced problems after intervention. [3, 4, 5] We know that as little as five minutes of advice on drinking has proven benefit in reducing consumption. This is acceptable to the patient and can be integrated into a busy practice. If doctors were to stop providing brief advice on drinking as a result of reading this meta-analysis it would clearly be a huge mistake. If we used the parallel of hypertension, it would be like GPs no longer checking and treating blood pressures because the simplest available treatment only reduced BP by 30%, but didn't reduce necessarily reduce the diastolic back to 80 mm Hg. There is a role for maximising the efficiency of screening and constantly improving treatment modalities, as there is with any condition, but in the meantime doctors should ask every patient about alcohol consumption. Yours sincerely A/Prof. Katherine M. Conigrave, Staff Specialist, Drug Health Services, Central Sydney Area Health Service Building 82, Royal Prince Alfred Hospital, Missenden Rd, Camperdown NSW 2050 Phone 61 (02) 9515 8650 Fax (02) 9515 8970 email: katec@med.usyd.edu.au Prof. John B. Saunders, Professor of Alcohol & Drug Studies, University of Queensland Mental Health Centre K Floor, Royal Brisbane Hospital Email jbsaunders@compuserve.com Dr Elizabeth M. Proude, Research Officer, University of Sydney Drug Health Services, Building 82, Royal Prince Alfred Hospital, Missenden Rd, Camperdown NSW 2050 Email: elizabeth.proude@email.cs.nsw.gov.au References: 1. WHO Brief Intervention Study Group. A randomised cross-national clinical trial of brief interventions with heavy drinkers. American Journal of Public Health 1996; 86(7): 948-955 2. Gomel MK, Saunders JB, Wutzke SE, Hardcastle DM, Carnegie MA. Implementation of early intervention for hazardous and harmful alcohol consumption in general practice. Final report for the Research into Drug Abuse program. Department of Human Services & Health, July 1996 3. Berglund G, Nilsson P, Eriksson K-F, Nilsson J-A, Hedblad B, Kristenson H, Lindgarde F. Long-term outcome of the Malmo Preventive Project: mortality and cardiovascular morbidity. Journal of Internal Medicine 2000; 247(1): 19-29 4. Senft RA, Polen MR, Freeborn DK, Hollis JF. Brief intervention in a primary care setting for hazardous drinkers. American Journal of Preventive Medicine 1997; 13(6): 464-70 5. Moyer A, Finney JW, Swearingen CE, Vergun P. Brief interventions for alcohol problems: a meta-analytic review. Addiction 2002; 97(3): 293-4 Competing interests:   None declared
Re: Screening effect ten times greater ... (by Heather and Richmond) 5 October 2003
Anders Beich, Research fellow Central Res.Unit Gen.Pract., University of Cph., Panum Institute, 2200 Copenh.N, Denmark, Thorkil Thorsen Send response to journal: Re: Re: Screening effect ten times greater ... (by Heather and Richmond)	Heather and Richmond have read their own paper (1) from 1995 once again and they now ask us how we have calculated proportions of sensible drinkers and risk reductions in our meta-analysis (2) in which we included their paper. We have already stated in this our paper: we have determined the risk reduction (benefit increase) as the difference between proportions of sensible drinkers at follow-up in the respective groups, i.e. proportions of the ones randomized. In their report Heather and Richmond have unfortunately omitted to state the number of subjects drinking below limits at follow-up, i.e. only percentages in the text were available for our purpose, but facts taken from their paper are; At follow-up 24.0% were within limits in the intervention group. 66 intervention patients participated in the follow- up. 24% of 66 equal 16 patients. In the control group 61 patients participated in the follow-up of which 21.5% =13 were within limits. The intervention group (Alcoholscreen) counted 96 subjects, the control group 93. Unfortunately the randomization failed to distribute excessive drinkers evenly among groups in this particular study: at baseline there were more excessive drinkers in the intervention group than in the control group. Heather and Richmond propose that absolute risk reduction can simply be calculated as the difference between the proportion changes (baseline minus follow-up) in the respective groups. Unfortunately this is not a legitimate way to calculate risk reduction because we are no longer dealing with a binomial outcome: subjects can either become excessive drinkers, stay sensible drinkers, give up excessive drinking, stay excessive drinkers, or they can get lost between baseline and follow-up. The problem can be easily illustrated by trying to comprehend what it really means that “those who had changed from being hazardous drinkers at initial assessment to “sensible” drinkers at 12-month follow-up” ends up being -5.4% in the control group, as suggested by Heather and Richmond. It does not make any sense, does it? We admit that this misapprehension of risk reduction (ore benefit increase) is not uncommon. Nevertheless, we are somewhat embarrassed to have to put this in the open and we only do so in order to avoid that the discussion of our paper is disturbed by too much noise. Anders Beich Thorkil Thorsen References 1. Richmond R, Heather N, Wodak A, Kehoe L, Webster I. Controlled evaluation of a general practice-based brief intervention for excessive drinking. Addiction 1995;90:119-132. 2. Beich A, Thorsen T, Rollnick S. Screening in brief intervention trials targeting excessive drinkers in general practice: systematic review and meta-analysis. BMJ 2003;327:536-542. Competing interests:   None declared
PROBLEMS IN USING NNT FOR STUDIES OF EXCESS DRINKING 8 October 2003
Ivar S. Kristiansen, Senior reseracher Norwegian Centre for Health Technology Assessment, Torbjorn Wisloff Send response to journal: Re: PROBLEMS IN USING NNT FOR STUDIES OF EXCESS DRINKING	In a recent paper in the BMJ, Beich and co-workers use the number-needed-to-treat (NNT) to express the effectiveness of interventions aimed at reducing excessive drinking (1). Unfortunately, their use of NNT may be misleading. In table 4 of the paper, confidence intervals (CI) for NNT are incorrectly computed (2). For example, the 95% CI of the absolute risk reduction in the Manwell study was (-0.009, 0.225) and Beich claims that the corresponding CI for NNT is (4, -113). It's not! NNT has two CIs: one at (-infinity, -111) and another at (4, infinity). Four other CIs are subject to the same type of miscalculation. More important, the authors compute the NNT at an arbitrary length of follow-up (12 months). It is unlikely that the effects of "brief interventions" are constant in time. Rather it seems plausible that the proportions that abandon excessive drinking are changing over time. While some give it up forever, others give it up for shorter or longer periods, or not at all. In this case, the effect would be better expressed in terms of average number of years without excessive drinking. It would require repeated measurements and time-to-event (survival) analysis for proper expression of the benefit. Finally, NNT cannot be used to infer the probability that the individual patient will benefit (3). It is therefore misleading to claim that "only two to three patients per thousand screened will benefit". Small reductions in problem-free-drinking time in many patients can create the same risk reduction and NNT as large reductions in a few patients. NNT is a problematic effect measure for interventions that postpone rather than totally prevent adverse outcomes. Ivar Sonbo Kristiansen, MD PhD MPH Torbjorn Wisloff, MSc (Stat) The Norwegian Centre for Health Technology Assessment References 1. Beich A, Thorsen T, Rollnick S. Screening in brief intervention trials targeting excessive drinkers in general practice: systematic review and meta-analysis. BMJ 2003; 327: 536-42. 2. Altman DG. Confidence intervals for the number needed to treat. BMJ 1998; 317: 1309-12. 3. Kristiansen IS, Gyrd-Hansen D, Nexoe J, Nielsen JB. Number needed to treat: easily understood and intuitively meaningful? Theoretical considerations and a randomized trial. J Clin Epidemiol 2002; 55: 888-92. Competing interests:   None declared
Re: PROBLEMS IN USING NNT FOR STUDIES OF EXCESS DRINKING 9 October 2003
Anders Beich, research fellow Central Res.Unit of General Pract., Panum Inst., Unversity of CPH, DK-2200 Copenhagen, Denmark, Thorkil Thorsen Send response to journal: Re: Re: PROBLEMS IN USING NNT FOR STUDIES OF EXCESS DRINKING	Sometimes details mean everything, sometimes they are trivial. Kristiansen and Wisloff(1) reproach us in a didactic tone (talking about miscalculations?) for choosing a way of reporting confidence intervals for NNT that can at worst be characterized as imprecise. We all know that a confidence interval for NNT cannot be a continuous interval when it covers both benefit and harm (when the interval for ARR includes 0). We would have preferred to report the interval example (4, -113) as (NNTB 4 to infinity to NNTH 113) like proposed by Altman et al.(2), and we actually did our first draft to BMJ. We were asked by the editorial board not to use infinity in our confidence intervals, thus we had no choice but to revert to a notation used before in BMJ (3) in our final manuscript. We would prefer to discuss this somewhat subtle matter further some other time. Kristiansen and Wisloff then reproach us that we have not used for example a time-to-event outcome in stead of just sensible drinking at 12 month follow-up (“an arbitrary length of follow-up”) , thereby demonstrating that they have not read the papers included in our meta- analysis and have no specific knowledge on the subject of our analysis. Others might prefer a continuous registration of blood alcohol concentration for each subject in the trial sample to monitor the intervention effect, but unfortunately there is not a buffet of outcome measures out there. Finally, Kristiansen and Wisloff put forward the public health maxim that if enough patients make small reductions even in problem free drinking this might create a large “amount of risk reduction” all in all. Again, had they read the papers in question they would have known that quite a few of these small reductions in question would be taking place in the flat neighbourhood of nadir on a consumption-mortality (or consumption -morbidity) curve. In one study mentioned by Kristiansen and Wisloff (4) the mean consumption in the trial sample was 14 drinks per week (SD=9). The possibility of confusion of statistical significance and clinical significance should be kept in mind every time the public health maxim mentioned above is put forward. We deliberately chose a medically and clinically relevant outcome measure like (at least) drinking within sensible weekly limits (at a time after intervention comparable between studies), because this makes some sense when you are facing the individual patient in a clinical context. Anders Beich Thorkil Thorsen REFERENCES: 1. Kristiansen IS et al. Problems in using NNT for studies of excess drinking. bmj.com, 8 Oct 2003 2. Altman DG, Machin D, Bryant TN, Gardner MJ. Statistics with confidence. London: BMJ Books, 2000. 3. Rembold CM. Number needed to screen: development of a statistic for disease screening. BMJ 1998; 317: 307-12. 4. Manwell LB, Fleming MF, Mundt MP, Stauffacher EA, Barry KL. Treatment of problem alcohol use in women of childbearing age: results of a brief intervention trial. Alcohol Clin Exp Res 2000; 24: 1517-24 Competing interests:   None declared
Re: Saitz R. Alcohol screening as good as other recommended screens 20 October 2003
Anders Beich, research fellow Central Res.Unit of General Pract., Panum Inst., Unversity of CPH, DK-2200 Copenhagen, Denmark, Thorkil Thorsen Send response to journal: Re: Re: Saitz R. Alcohol screening as good as other recommended screens	Richard Saitz scrutinizes our conclusions and raises a few quite important issues in his response. In fact he has in a very nice way summarized the most important criticisms raised in different rapid responses to our paper (1) and we appreciate very much the thoroughly constructive tone of his response. Our answers here are meant in the same spirit. WHICH STUDIES TO INCLUDE ANSWERING OUR QUESTION: We would have preferred to review studies that compared screening and brief intervention with usual case finding and usual care, but we haven’t been able to locate any taking place in general practice. It seems that screening for hazardous drinking became an imperative a long time ago and therefore part of the control conditions. The study cited by Saitz (2) is an evaluation example of a hospital-based early intervention program (screening and intervention versus usual case finding and care) that seems to have had some impact but only on subjects with signs of alcohol dependence, that is: a different context and a different target group. Until studies on program effectiveness, compatibility and suitability are available we can disagree on how to use results from efficacy studies. Unlike Saitz, we believe that clinicians should demand that effectiveness studies of prevention programs are carried out before implementation is recommended. Besides, we should realize that the justification of such programs is depending on compatibility, suitability and other matters as well; and somehow it seems a naturalistic fallacy to recommend general practitioners to take on new tasks just because evidence of cost-efficacy is available. An implicit value judgment has been put in between efficacy and implementation somewhere. We believe such judgments should be put in the open and not be ignored or concealed. GENERALIZING FROM STUDIES ON HIGHLY SELECTED POPULATIONS: Saitz believes that clinicians are accustomed to generalizing the results of clinical trials beyond the highly selected populations studied. When practicing evidence based medicine it should include that the clinician considers the following factors (3): “- Is the relative risk reduction that is attributed to the intervention likely to be different in this case because of the patient's physiological or clinical characteristics? - What is the patient's absolute risk of an adverse event without the intervention? - Is there significant comorbidity or a contraindication that might reduce the benefit? - Are there social or cultural factors that might affect the suitability of treatment or its acceptability? - What do the patient and the patient's family want?” Why is it that lifestyle preventive initiatives should not trigger the very same questions within the clinician? Why should good clinical practice be overruled by preventive automatics? Saitz believes that the drop-out between screening positive and receiving brief intervention in real life practice will not be similar to the drop-out in a trial (no informed consent needed, no need to return for further visits, and no need to exclude patients with comorbidity, severe alcohol problems or homelessness) If the intervention is to be motivational and of use to the patient we believe there needs to be a clinical parallel to informed consent (asking permission to touch upon lifestyle issues, and explaining why), especially when there is no immediate logical link to the patient's reason for coming to see the doctor. In clinical situations the need for further visits would likely emerge if the first visit was successful in establishing rapport and an agenda that included drinking. But routinely giving brief intervention to alcohol dependent patients regardless is in our view likely to be a mismatched intervention. Giving brief intervention to people who are seriously ill or living in the street we consider to be ill-timed or even unethical. COMPARABILITY OF NNTs AND DOCTORS DISSAPOINTMENTS: We do not believe that NNTs can be compared the way Saitz does it. NNT for reducing alcohol consumption to within sensible limits should not be compared to NNT for avoiding one stroke or one case of advanced colorectal cancer. It certainly matters which benefit we are talking about. Proxy measures can be relevant but they should not be confused with consequences like these. Saitz says that “doctors do not seem disappointed when they do not find colon cancer or benefit from hypertension screening”. But doctors would probably be disappointed if they were screening for colorectal cancer and found themselves unable to offer effective treatment to more than 3 out of 100 early cancers diagnosed. Such a screening would not fit well with the rationale for screening as defined by WHO (4). According to our analysis, the factual documentation in trials on screening and brief intervention suggests that 3 out of 100 screened positive (2-3 out of 1,000 screened) benefit from the “therapy”. SAITZ CONCLUDES: “Universal screening with validated instruments is much better for identifying patients with alcohol problems than any currently known alternative”. We find this conclusion to be somewhat tautological. Better implies a reference point, but the reference point (usual case finding as it takes place in most countries) has not been established yet. Saitz also concludes that our review supports the notion that brief intervention after screening is efficacious. We kindly remind him that we discovered abundant sources of bias in the trials included for meta- analysis, all tending towards overestimation of effect. We omitted to give guesstimates of the size of these biases in order not to blur the screening message. Nevertheless, we are somewhat surprised that nobody, not even the authors of the most biased trials, have taken a position in regard to our scrutiny of the internal validity of these trials. Anders Beich Thorkil Thorsen REFERENCES 1. Beich A, Thorsen T, Rollnick S. Screening in brief intervention trials targeting excessive drinkers in general practice: systematic review and meta-analysis. BMJ 2003;327:536-42. 2. Welte JW, Perry P, Longabaugh R, Clifford PR. An outcome evaluation of a hospital-based early intervention program. Addiction. 1998;93:573-81. 3. Sheldon TA, Guyatt GH, Haines A. Getting research findings into practice - When to act on the evidence. BMJ 1998;317:139-42. 4. Wilson JMG, Jungner G. Principles and practice of screening for disease. Geneva: World Health Organization, 1968. Competing interests: None declared