DRUG ABUSE

DRUG ABUSE

746 USING CONFIDENCE INTERVALS SIR,-We were pleased to see that the paper by Dr Bulpitt (Feb 28, p 494) and the accompanying editorial echoed much of...

200KB Sizes 0 Downloads 163 Views

746 USING CONFIDENCE INTERVALS

SIR,-We were pleased to see that the paper by Dr Bulpitt (Feb 28, p 494) and the accompanying editorial echoed much of what we had written last year in the British Medical JournaU It is most encouraging, that medical journals are moving to a more consistent presentation of information in relation to statistical estimation. We would, however, like to mention some small difficulties and

discrepancies.

spectrometry. The patient was prescribed methadone, which appeared in the urine 2 days later and persisted. Chloroquine and its metabolite were identified in all subsequent specimens. The half-life of chloroquine is 2 */2 to 5 days and detection is thus likely up to 25 days. He had not been prescribed chloroquine. The patient stated that the drug had an unusual pungent bitter mass

,

We have reservations about, your editorial’s "firm recommendation" to standardise on a 95% confidence interval. This may be unwise, though we recognise the potential problems if authors choose different confidence levels for different purposes. Whatever level of confidence is used, however, the mean difference and its standard error, for example, should be given to enable readers to make their own choice. Bulpitt’s graphical presentation differs from our proposa1,1 Bars at the extremities of confidence intervals give a sense of total inclusion; arrows’ indicate that values outside the confidence interval are also possible. It would be preferable for journals to use the same representation. The methods of calculating confidence intervals that we detailed1 for common situations will be supplemented by a forthcoming paper in the BMJ that will include methods for all the further statistics mentioned by Bulpitt as well as others. These are being brought together since they are often difficult to find or understand in books and other sources. We were sorry to read of the examples of misuse of confidence intervals in the BMJ since our paper appeared, especially since the errors described were ones we had warned against. Bulpitt incorrectly defines the p value. The p value is the probability that, if two groups are random samples from populations equal in terms of the factor being studied, the group (mean) difference will differ by the amount observed or more (regardless of the direction of the difference for a two-sided test) by chance due to sampling variation. Bulpitt states that in well-designed studies "false positive results are very rare". Evidence of publication bias in favour of statistically significant ("positive") fmdings2,3 suggests that this may not be true. The notion that studies, experimental or observational, can be classified as positive or negative on the basis of statistical significance is unhelpful.4 The use of confidence intervals in journals should avoid this dichotomy. MRC Environmental Epidemiology Unit, Southampton General Hospital, Southampton SO9 4XY

MARTIN J. GARDNER

Division of Medical Statistics, MRC Clinical Research Centre, Harrow, Middlesex

DOUGLAS G. ALTMAN

MJ, Altman DG. Confidence intervals rather than P values: Estimation rather than hypothesis testing. Br Med J 1986; 292: 746-50. 2. Chan SS, Facks HS, Chalmers TC. The epidemiology of unpublished randomized control trials. Clin Res 1982; 30: 234A. 3. Pocock SJ. Clinical trials: A practical approach. Chichester: Wiley, 1983: 240. 4. Chalmers I. Proposal to outlaw the term "negative trial". Br Med J 1985; 290: 1002. 1. Gardner

taste, for which reason it could not be smoked, and was irritant when taken by the nasal route which he used. He complained of malaise and excessive sweating, which has continued up to now, 3 weeks after the drug was withdrawn. He said that the drug was thought to come from Karachi and was being extensively used. Assuming a minimum adulteration of 10% chloroquine, the patient believed he would have taken between 800 and 1000 mg chloroquine in 5 days. The usual malarial prophylactic oral dose is 500 mg weekly. Toxic effects of chloroquine include headache, confusion, visual disturbance, and an increase in serum creatinine at blood levels of chloroquine above 0’6 mg/1. Our patient’s serum creatinine was 99 umol/1. Acute overdose may produce respiratory depression and

shock. Identification of unusual spots on TLC as chloroquine and its metabolite may avoid further exhaustive investigations. We thank Dr Judith Morgan for permission to study her patient, and Dr Brian Widdop, National Poisons Centre, New Cross Hospital, for GC/MS,

Department of Pathology, Brook General Hospital, London SE18 4LW

DRUG ABUSE

SiR,—Your Feb 14 editorial implies that widespread availability screening would reduce the need for treatment centres by enabling general practitioners once more to manage maintenance as in former years. This change is widely to be desired in view of the restricted access patients have to treatment centers, the stigma that

of drug

attaches to attendance at such centres, and the increased numbers of addicts. However, a "complete drug screen... on at least two occasions" is neither clinically necessary, financially prudent, nor desirable. Screening is useful, but to insist that a heavily addicted patient, with several track marks and perhaps indurated groins, goes through the cumbersome "London clinics" procedure of Connell and Mitcheson is clinically redundant. At c5 per drug tested such a policy would be a profligate waste of National Health Service resources. It would also deter many addicts from making contact with services: our experience in Liverpool shows how important this can be.1 As Willis has said,2 the chief obstacle to clear thinking about drugs of addiction appears to be doctors themselves. Liverpool Drug Dependency Clinic, 30 Hope Street, Liverpool Ll 9BX

SIR,-Your editorial (Feb 14, p 365) referred to the adulteration of heroin by dilution ("cutting") with barbiturates, which may lead to addiction to barbiturate as well as to heroin. Although we have regularly observed phenobarbitone in association with morphine in the urine of heroin abusers this association has been seen less often during the past 2 months. However, a new adulterant has appeared, and we report the cutting of heroin with chloroquine. A narcotic abuser was admitted to the addiction unit at Bexley Hospital. He had changed his supplier of heroin 5 days before admission. Urine samples, taken under supervision,’ were submitted to the laboratory on ten occasions up to 18 days after

admission. Morphine was identified by thin-layer chromotography (TLC) in the first sample together with nicotine and two unidentified spots. These were identified as chloroquine and its metabolite. The identity was confirmed by gas chromatography/

JOHN MARKS

1. Marks J, Parry A. Syringe exchange programme for drug addicts. Lancet 1987; i: 69l. 2. Willis JH. Prescription of controlled drugs to addicts. Br Med J 1983, 287: 500.

ADULTERATION OF "STREET" HEROIN WITH

CHLOROQUINE

P. O’GORMAN SHANTA PATEL SUSAN NOTCUTT JENNIFER WICKING

SYRINGES FOR DIABETICS

SiR,—The Government’s decision (March 14, p 637) to provide disposable syringes for patients with diabetes is welcome. However, the stated cost of this decision is around six times greater free

than the likely true cost. Most patients using disposable syringes reuse them until the needle is blunt, a practice which is safe and economical.’ This practice is recommended to patients by all physicians caring for diabetics in North East Thames Region, and the same is probably true elsewhere. A patient on twice-daily insulin will use only 100-150 syringes in a year, and often many fewer. Someone on once-daily insulin uses only half as many. Very few diabetics use glass syringes nowadays. The remainder either buy their own disposable syringes or receive them from the hospital pharmacy. In this region, 62% of clinics already provide free disposable syringes. We have calculated that, aside from the fact that glass syringes need replacing for breakages,’ the loss of insulin