Blackwell Science, LtdOxford, UKADDAddiction0965-2140 2003 Society for the Study of Addiction to Alcohol and Other Drugs98419425Original ArticleTrends

in drug overdose deathsMatthew Hickman et al.

RESEARCH REPORT

Trends in drug overdose deaths in England and Wales 199398: methadone does not kill more people than heroin
Matthew Hickman1, Peter Madden1, John Henry2, Allan Baker3, Chris Wallace4, Jon Wakeeld5, Gerry Stimson1 & Paul Elliott6
Centre for Research on Drugs and Health Behaviour, Social Science and Medicine, Imperial College, London, UK1, Division of Surgery, Imperial College, London, UK2, Ofce for National Statistics (ONS), London, UK3, London School of Hygiene and Tropical Medicine, London, UK4, Department of Biostatistics, University of Washington, Seattle, WA, USA and 5 Centre for Epidemiology and Biostatistics, Primary Care and Population Health Sciences, Imperial College, London, UK6

Correspondence to: Matthew Hickman Centre for Research on Drugs and Health Behaviour Social Science and Medicine Imperial College, Reynolds Building Charing Cross Campus St Dunstans Road London W6 8RP UK Tel: + 44 20 7 594 0784 / 0811 Fax: + 44 20 7 594 0866 E-mail: m.hickman@ic.ac.uk Submitted 26 March 2002; initial review completed 19 June 2002; nal version accepted 19 August 2002

ABSTRACT Aims To test the hypothesis that methadone is responsible for a greater increase in overdose deaths than heroin, and causes proportionally more overdose deaths than heroin at weekends. Design and setting Multivariate analysis of 3961 death certicates mentioning heroin, morphine and/or methadone held on the Ofce for National Statistics drug-related poisoning mortality database from 1993 to 1998 in England and Wales. Measurements Percentage increase in deaths by year by drug, odds ratio (OR) of dying at the weekend from methadone-related overdose compared to dying from heroin/morphine overdose. Findings From 1993 to 1998, annual opiate overdose deaths increased from 378 to 909. There was a 24.7% (95% condence interval (CI) 2228%) yearly increase in heroin deaths compared to 9.4% (95% CI 613%) for methadone only. This difference was signicant (P < 0.001 by test of interaction) after adjustment for sex, age group, polydrug use, area of residence and underlying cause of death. The largest number of deaths occurred on Saturday (673). The OR of death from methadone overdose on Saturday and Sunday was 1.48 (95% CI 1.291.71) for methadone-only deaths compared to dying from heroin/ morphine at the weekend after adjustment for other covariates, but the OR was not signicant (1.09, 95% CI 0.951.25) if the weekend was dened as Friday and Saturday. Conclusions There was no evidence that the threefold increase in deaths over time was due to methadone. There was equivocal support only for the hypothesis that there was an excess of deaths from methadone at weekends. Increased interventions to prevent overdose among injectors in England and Wales are long overdue. KEYWORDS Heroin, methadone, mortality, overdose.

RESEARCH REPORT

INTRODUCTION Overdose-related mortality is a major public health harm associated with injecting drug use (IDU) and a signicant cause of mortality in young adults in the general popula 2003 Society for the Study of Addiction to Alcohol and Other Drugs

tion [1,2]. In England and Wales, as in some other countries, overdose mortality is the most signicant cause of death among injectors, far outnumbering deaths from HIV infection. For instance, in England and Wales it was estimated that from the early 1980s to the end of 1996
Addiction, 98, 419425

420

Matthew Hickman et al.

less than 500 injectors had died of HIV-related causes [3]less than the number of opiate overdose-related deaths in 1996 alone [4]. Injecting drug users risk of death from overdose has been estimated to be 0.8% per year injecting [2]. In the United Kingdom long-term mortality cohort studies suggest that there has been little or no progress in reducing overdose mortality among intravenous drug injectors since the late 1970s, when the availability of barbiturates on prescription was greatly reduced [5]. Compared to the general population, injecting IDU have a risk of premature death of between six and 30 times higher [1,2]. In the United Kingdom opiate-related overdose is the largest cause of death through poisoning [4]. In 1996, of nearly 3000 deaths mentioning a drug, heroin and/or methadone were cited in nearly 30%, with antidepressants and paracetamol containing medications at 20% and 18%, respectively. Opiate overdoses are often underascertained [6]. Among young adults, for example, in 1996 at least 6% of the 9000 deaths among males and females aged 1534 in England and Wales were certied as opiate overdose, although the true level may be nearly twice as high, as reported recently in Italy [7]. The risk of death from opiate-related overdose is affected by a number of factors [8]. Mortality is increased in combination with other drugs, in particular alcohol and benzodiazepines [9,10], or when tolerance has fallen, for example after release from prison [11] or interruptions in, or cessation of treatment [12], and tolerance can potentially fail under unusual environmental conditions [13]. Several studies suggest that treatment, especially with methadone, is protective from death from overdose, with a meta-analysis suggesting that heroin users out of treatment are four times more likely to die compared to those on a treatment programme [2,14]. In the United Kingdom most methadone administration is not supervised, but is prescribed for patients to take at home. There is also some variation in the frequency with which prescriptions are given: some patients are expected to pick up prescriptions on a daily basis, others less frequently, and often with increased supplies prescribed to cover the weekend and bank holidays [15]. An ongoing concern in the United Kingdom is that the peculiarity of the British system of prescribing has been a risk factor, rather than a protective factor, for opiate-related deaths [1624]. That is that methadoneeither through diversion of pharmaceutical supplies and/or overprescribing to an individualis responsible for an increase in opiate-related overdoses. The UK Advisory Council on the Misuse of Drugs (ACMD) recently gave further support for this hypothesis, arguing that: agency approaches to prescribing and dispensing of methadone have often been too lax [25].
2003 Society for the Study of Addiction to Alcohol and Other Drugs

Two previous analyses have addressed this problem. The rst, by Neeleman & Farrell, analysed routine mortality statistics from 1974 to 1992 comparing the proportional mortality ratio (i.e the number of methadone or heroin poisonings/total number of poisonings) over time [26]. During this period poisonings attributed to methadone had increased from 26 to 240 and deaths from heroin alone had increased from 7 to 90. They concluded that although there was a considerable increase over time there was no evidence to suggest that the rate of methadone poisoning was any higher than that for heroin. Hall et al. compared United Kingdom and Australia population rates of opiate overdose mortality between 1985 and 1995 [27]. They suggested that the overall rate of opiaterelated overdose was substantially lower in the United Kingdom than in Australia (by a quarter in 1995: 0.12% and 0.47%, respectively), albeit the proportion of methadone deaths was higher in the United Kingdom. They suggested that the scale of the difference between the two countries was unlikely to be explained by differences in the size of the injecting population, or the proportion of heroin users that are injectors. They considered that a contributory factor (both to the higher proportion of methadone deaths and lower overall mortality rate) might be the increased availability of methadone treatment in the United Kingdom compared to Australia. Despite these analyses, concern over the role of methadone in drug-related deaths remain controversial and has not diminished [25]. Therefore, we examined again the claim that methadone kills more people than heroin [17]. In particular, two hypotheses were assessed, whether: (a) methadone overdose deaths were increasing at a higher rate than heroin because of increased availability of treatment over time; and (b) people were more likely to die from methadone than heroin at the weekend because of prescribing practice that gives methadone to take at home with increased supplies for the weekend. If supported these hypotheses would suggest that uncontrolled prescribing and the potential diversion of methadone could be responsible for an increase in overdose deaths [28].

METHODS Counting overdose deaths is not straightforward [29,30]. First, they may be classied under several underlying causes of death: drug dependence or drug abuse, or as poisonings as an accident, suicide or open verdict. Secondly, the International Classication of Disease (ICD) diagnostic codes cover several drugs, e.g. in ICD-9, code 965.0 for opiates and related narcotics covers codeine, heroin, methadone, morphine, pethidine and opium. For this study we used a new database compiled by the Ofce
Addiction, 98, 419425

Trends in drug overdose deaths

421

for National Statistics (ONS) that collates deaths by the individual drugs specied on the death certicate, rather than causes of death classied by ICD [4]. Deaths mentioning heroin, morphine and/or methadone were extracted from the ONS database of drugrelated deaths for the years 199398 (n = 4070). Deaths among those under 15 and over 54 were excluded (12 and 96, respectively) because they were less likely to be drug users (and their deaths were less likely to be due to dependent opiate use). One death mentioning chronic ingestion of kaolin and morphine and alcoholism also was excluded. Deaths mentioning morphine were assumed to be due to heroin, as it is metabolized rapidly into morphine and most tests for presence of drugs will detect only morphine [9]. Univariate analyses assessed differences between type of drug (methadone versus heroin) by year, and day of week of death. The choice of analysis (e.g. Poisson regression, ordinary least squares regression, or logistic regression) was complicated by the lack of an offset or denominator (i.e. number at risk in the population) to generate rates over time. The fact that the data were simple counts also affected the interpretation of statistical measures of comparison. Consequently, Poisson regression was undertaken, with deaths grouped by year into Poisson frequency records, to test for differences and adjust the analysis for other covariates including: sex, age-group, polydrug use (dened as use of benzodiazepines and/or alcohol), area of residence (dened as London or outside London) and underlying cause of death (coded as drug misuse or poisoning as dened on the death certicate). The number of methadone and heroin deaths by calendar year were examined by tting dummy intercepts and slope variables and testing whether the lines were signicantly different (i.e. whether the rates of increase were signicantly different). Separate intercepts and slopes also were allowed for each covariate. Analysis of drug and day of the week were compared using logistic regression, which tested whether a person who died from a death mentioning methadone or heroin was more likely to die from methadone at the weekend. The analysis also adjusted for other covariates. Given the variation in prescribing practice it was not possible to specify the weekend prior to analysis (i.e. it was not known whether patients were given an increased prescription on Friday or Saturday, or equally if any excess methadone was diverted or taken on specic days of the weekend). Therefore, we dened the weekend in three ways: FridaySunday, FridaySaturday and Saturday Sunday. Further, each set of analyses was carried out twice. This was because some death certicates mentioned both heroin/morphine and methadone, and it was not possible
2003 Society for the Study of Addiction to Alcohol and Other Drugs

to determine which was primarily responsible. Thus, one set of analyses grouped deaths mentioning both drugs with heroin deaths (heroin +), and in the other they were grouped with methadone deaths.

RESULTS Opiate overdose deathscharacteristics From 1993 to 1998 there was a total of 3961 opiaterelated overdose deaths among people aged 1554. Of these, 2278 (58%) mentioned heroin and/or morphine with 1252 (32%) heroin; 1944 (49%) mentioned methadone; and 261 (7%) mentioned both methadone and heroin/morphine. These accounted for 89% of all opiate overdose deaths. Over the same period there were an additional 493 deaths that cited opiate overdose without mentioning a specic drug. There were no differences between deaths mentioning opiate only or specifying heroin and/or methadone over time (c2 independence test, 5.6, df 5, P = 0.34). The following analyses were carried out on only those deaths where a specic drug (methadone/heroin or morphine) was mentioned. Overall, the mean age was 29.7 (28.9 for deaths mentioning methadone and 30.0 for those mentioning heroin/morphine); 3390 (86%) of the deaths were among males, and 738 (19%) resident in London. In addition, 1785 (45%) of the death certicates mentioned other drugs, 34% mentioned alcohol or benzodiazepines (coded as polydrug use). In 1660 (42%) deaths the underlying cause of death was coded as misuse of drugs and in 2301 (58%) as poisoning.

Opiate overdose deathsanalysis by calendar year Over the study period sex, age and polydrug use did not vary signicantly by year. Apart from drug type there were signicant variations over time by area of residence from 26% of deaths resident in London in 1993 to 15% in 1996 and 18% in 1998 (c2 (5) 27.9, P < 0.0001). Underlying cause of death given as drug misuse or dependence increased from 31% in 1993 to 50% in 1998 (c2 (5) 63.8, P < 0.0001). Figure 1 shows the number of deaths mentioning methadone only and those mentioning heroin+ (including heroin and/or morphine and methadone) over time. Overdose deaths increased nearly threefold from 378 in 1993 to 909 in 1998. Figure 1 also shows a tted line using Poisson regression. Poisson regression analysis indicated that for heroin+ deaths from 1993 to 1998 there was a 24.7% (95% condence interval (CI), 22 28%) yearly increase in deaths compared to 9.4% (95% CI, 613%) for methadone only. This difference was sigAddiction, 98, 419425

422

Matthew Hickman et al.

Table 1 The Odds Ratio of someone dying from a methadone overdose at the weekend compared to dying from an opiate overdose during the week. FridaySunday (95% CI) Unadjusted 1.29 (1.141.47) 1.24 (1.091.40) Adjusted* 1.26 (1.111.44) 1.21 (1.061.37) FridaySaturday (95% CI) Unadjusted 1.12 (0.981.28) 1.09 (0.951.24) Adjusted* 1.09 (0.951.25) 1.07 (0.931.22 SaturdaySunday (95% CI) Unadjusted 1.51 (1.321.73) 1.48 (1.291.69) Adjusted* 1.48 (1.291.71) 1.44 (1.261.66)

Weekend: Comparison Heroin versus. methadone Heroin versus methadone

*Adjusted for covariates: age, sex, area of residence, polydrug use, underlying cause of death. Plus deaths citing heroin/morphine and methadone.

700 600 500 400 300 200 100 0 1993 Heroin+ 1994 1995 1996 Methadone only 1997 1998

pared to 11.1% (95% CI, 814%) for those mentioning methadone. The difference was signicant after adjustment for other covariates (P < 0.001 by test of interaction).

Opiate overdose deathsanalysis by day of the week By day of the week there were no signicant differences by sex, polydrug use, area of residence, or underlying cause of death. Drug overdose deaths at the weekend were slightly younger than those dying during the week (28.7 versus 30.1); although statistically signicant, such a difference was not considered to be of public health signicance. Figure 2 shows that overdose deaths increased progressively by day of the week from 431 on Monday to peak at 673 on Saturday. One-third (1305/3961) of the deaths occurred on Friday and Saturday, with 29% (1149/3961) on Saturday and Sunday, and 45% (1781/ 3961) on Friday to Sunday. Table 1 shows the odds ratio (OR) of dying at the weekend from an overdose involving methadone compared to heroin. Adjusting for covariates and the grouping of deaths that cite both heroin and methadone with heroin or with methadone deaths has little effect on the size or signicance of the estimates. The signicance and size of the difference does depend, however, on the denition of the weekend. Compared to heroin/morphine deaths the adjusted OR of someone dying from a methadone-only overdose on Saturday and Sunday was 1.48 (95% CI, 1.291.71), whereas the adjusted OR of someone dying from a methadone-only overdose on Friday and Saturday was not signicant at 1.09 (0.951.25).

Poisson (heroin+)

Poisson (methadone)

Figure 1 Trends in methadone and heroin/morphine deaths over time

Figure 2 Methadone and heroin/morphine deaths by day of the week: 199398

nicant (P < 0.001 by test of interaction) and remained signicant after adjustment for sex, age-group, polydrug use, area of residence, and underlying cause of death (P < 0.001). Similarly, where deaths were coded as heroin and morphine (excluding methadone) there was a 24.9% (95% CI, 2228%) yearly increase in deaths com 2003 Society for the Study of Addiction to Alcohol and Other Drugs

DISCUSSION More people die from heroin-related overdose than methadone-related overdose, contrary to previous comAddiction, 98, 419425

Trends in drug overdose deaths

423

munications [16,17,26]. There is only equivocal support for the hypothesis that there is an elevated risk of overdose involving methadone at the weekend. Dening the weekend as Saturday and Sunday suggests that the OR of death involving methadone is nearly 1.5 compared to deaths involving heroin, which would equate to an excess of approximately 200 deaths (5% of the total) over the 6year period. However, dening the weekend as Friday and Saturday suggests that there is no difference between the likelihood of dying at the weekend from a heroin or methadone overdose. Between 1993 and 1998 there has been a marked and dramatic threefold rise in opiate overdose deaths in England and Wales. The analysis does not provide evidence to support the notion that methadone is mainly responsible, supporting the previous studies that concluded that the rate of methadone overdose death was not higher than that of heroin [26,27]. Rather, in our analysis the number of deaths attributed to heroin increased at a signicantly higher rate than those attributed to methadone. The AMCD report and others have blamed the system of methadone prescribing in the United Kingdom for contributing to overdose mortality and recommended further controls and supervision of methadone prescribing [16,17,28]. Although all methadone in circulation is from NHS prescriptions, to concentrate solely on controlling methadone supply and administration to reduce drug-related deaths may be misguided [31]. Clearly, the public health focus should be on reducing opiate-related mortality overall, and the benets of imposing greater control on methadone prescriptions should be considered in the light of this overall target [27]. Furthermore, it is vital that new and/or enhanced interventions to prevent overdose are introduced, especially given the background of low transmission of HIV infection due to public health action and investment in needle exchange and other harm-reduction initiatives [32,33]. In the United Kingdom preventing overdose mortality must now be seen as the key public health priority. Other factors have driven the increase in overdose deaths that may need to be tested by further analytical studies. There are a number of potential explanations, including changes in recording, overdose risk and the underlying population of injectors. An increase in the proportion of overdose deaths recorded cannot be ruled out, although we believe it is unlikely. We believe the increase in the proportion of deaths allocated to misuse of drugs instead of poisoning reects a change in practice among coronersand that the choice of verdict should not affect the overall count of deaths. Moreover, the number of opiate overdose deaths that did not specify a substance also increased threefold from 38 in 1993 to 115 in 1998, with the overall propor 2003 Society for the Study of Addiction to Alcohol and Other Drugs

tion of opiate overdose deaths mentioning either methadone or heroin/morphine remaining the same over time (at 89%). Also, an increase in overdose risk cannot be ruled out, as the potential contribution of changes in injectors behaviour and environment towards the overall risk of overdose remain important questions that have not been investigated fully and will need to be explored through mortality cohort studies [34]. Most studies reject the notion that overdose is related to drug purity or adulterants [1,8]. Polydrug use, as measured on the death certicates, did not change over time, although it is unlikely that the data are complete. An increase in imprisonment increasing the disruption of injectors tolerance and increasing overdose risk has been raised as a cause for concern in Australia , although there is no evidence currently available to support such an hypothesis in the United Kingdom; neither is there any support for a change in policing policy or attitudes to emergency services increasing the risk of fatal overdose among opiate users. A more likely explanation for the increase lies in changes in the underlying population of injectors. The issue of whether overdose risk increases with injecting career or remains constant is an important epidemiological consideration, because if risk was constant then changes in overdose statistics could be a rapid indicator of changes in the incidence of injecting. In theory, as the incidence of injecting increased and with it the number of overdose deaths, the average age of overdose death would be expected to fall. However, this did not happen during the threefold increase in overdose deaths in our study, as the average age at death did not change over time. More compelling is the view that the risk of death may increase with length of injecting career [8,9]. White & Irvine have argued that tolerance to the euphoric or intoxicating effects rises more quickly and may be more complete in individuals than tolerance to respiratory depression [9]. Thus, as the amount of drug consumed increases the likelihood of overdose also increases; and as heavy users moderate their consumption their tolerance to respiratory depression may fall more rapidly than tolerance to the toxicity of the drug. Warner-Smith and colleagues also have argued that systemic disease, which increases with length of injecting career, may reduce tolerance to respiratory depression thereby increasing the likelihood of overdose [8]. Under this hypothesis the increase in deaths (if it is related to the underlying population) is due to a cohort effect, pointing to an ageing population of injectors from an earlier epidemic of heroin use 5 or more years previously in the early 1990s and late 1980s. Statistical models that use trends in overdose deaths to predict the incidence of injecting heroin use also
Addiction, 98, 419425

424

Matthew Hickman et al.

suggest that a rise in the number of injecting drug users occur a number of years before any observed rise in overdose deaths [35].

ACKNOWLEDGEMENTS The Centre for Research on Drugs and Health Behaviour acknowledges the nancial support of both the London Regional Ofce of the NHS Executive and the Department of Health. Dr Matthew Hickman is funded through a National Public Health Career Scientist Award from the Department of Health. P. Elliott, J. Henry and J. Wakeeld were involved with the conception, design and nal manuscript. P. Madden and C. Wallace undertook the statistical analysis under the supervision of J. Wakeeld, wrote the methods and contributed to the nal paper. A. Baker extracted and provided the data, and contributed to the analysis and nal manuscript. G. Stimson contributed to the analysis and nal manuscript. M. Hickman co-ordinated the work and drafted the manuscript.

REFERENCES
Advisory Council on Misuse of Drugs (1998) Drug Misuse and the Environment. London: HMSO. Advisory Council on Misuse of Drugs (2000) Reducing Drug Related Deaths. London: Home Ofce. Baker, A. (2000) ONS drug-related deaths database: rst results for England and Wales, 199397. Health Statistics Quarterly, 5, 5760. Bentley, A. & Brusuttil, A. (1996) Deaths among drug abusers in South-east Scotland 198994. Medical Science and Law, 36, 231236. Cairns, A., Roberts, I. S. D. & Benbow, E. W. (1996) Characteristics of fatal methadone overdose in Manchester, 198594. British Medical Journal, 313, 246265. Capelhorn, J. R. M., Dalton, M. S. Y. N., Haldar, F., Paterenas, A. M. & Nisbet, J. G. (1996) Methadone maintenance and addicts risk of fatal heroin overdose. Substance and Misuse, 31, 177196. Cassidy, M., Curtis, M., Muir, G. & Oliver, J. (1995) Drug abuse deaths in Glasgow in 1992a retrospective study. Medicine Science Law, 35, 207212. Christopherson, O., Rooney, C. & Kelly, S. (1998) Drug related mortality: methods and trends. Population Trends, 93, 29 37. Clark, J. C., Milroy, C. M. & Forrest, A. R. W. (1995) Deaths from methadone use. Annals of Clinical Forensic Medicine, 2, 143 144. Darke, S. & Zador, D. (1996) Fatal heroin overdose: a review. Addiction, 91, 17651772. Darke, S., Sunjic, S., Zador, D. & Prolov, T. (1997) A comparison of blood toxicity of heroin related deaths and current heroin users in Sydney. Australian Journal of Drug and Alcohol Dependence, 47, 4553. Davoli, M., Perucci, C. A., Forastiere, F., Doyle, P., Rapiti, E., Zaccarelli, M. & Abeni, D. D. (1993) Risk factors for overdose mortality. a casecontrol study within a cohort of intravenous
2003 Society for the Study of Addiction to Alcohol and Other Drugs

drug users. International Journal of Epidemiology, 22, 273 277. Day, E., Kahn, A. & Georgiou, G. (1999) Misusers may receive methadone from drug centres and GPs simultaneously. British Medical Journal, 318, 395396. De La Fuente, L., Barrio, G., Vicente, J., Bravo, J. K. & Santacreu, J. (1995) The impact of drug-related deaths on mortality among young adults in Madrid. American Journal of Public Health, 85, 102105. Gabbay, M. B., Carnwath, T., Ford, C. & Zador, D. A. (2001) Reducing deaths among drug misuserstighter controls on drug prescribing are not the answer. British Medical Journal, 332, 749750. Ghodse, H., Oyefeso, A. & Kilpatrick, B. (1998) Mortality of drug addicts in the United Kingdom, 19671993. International Journal of Epidemiology, 27, 473478. Hall, W., Lynskey, M. & Degenhardt, L. (2000) Trends in opiate related deaths in the United Kingdom and Australia, 1985 95. Drug and Alcohol Dependence, 57, 247254. Hammersley, R., Cassidy, M. & Oliver, J. (1992) Drugs associated with drug-related deaths in Edinburgh and Glasgow, November 1990October 1992. Addiction, 90, 959965. Henry, J. (1999) Methadone: where are we now? Hospital Medicine, 60, 161164. Hickman, M., Bardsley, M., De Angelis, D. & Ward, H. (1999) Impact of HIV on adult mortality rates in London: 1979 1996. Sexually Transmitted Infections, 75, 385388. Hulse, G. K., English, D. R., Milne, E. & Holman, D. J. (1999) The quantication of mortality resulting from the regular use of illicit opiates. Addiction, 94, 221229. Ingold, F. R. (1986) Study of deaths related to drug abuse in France and Europe. Bulletin on Narcotics, 38, 8189. Law, M., Lynskey, M., Ross, J. & Hall, W. (2001) Backprojection estimates of the number of dependent heroin users in Australia. Addiction, 96, 433443. Marks, J. (1994) Deaths from methadone and heroin. Lancet, 343, 976. Neeleman, J. & Farrell, M. (1997) Fatal methadone and heroin overdoses: time trends in England and Wales. Journal of Epidemiology and Community Health, 51, 435437. Newcombe, R. (1996) Live and let die: is methadone more likely to kill you than heroin? Druglink, January/February, 912. Obafunwa, J. & Brusuttil, A. (1994) Deaths from substance overdose in the Lothian and Borders Region of Scotland 198391. Human and Experimental Toxicology, 13, 401406. Quaglio, G., Talamini, G., Lechi, A., Venturini, L., Lugoboni, F., Gruppo Intersert di Collaborazione Scientica & Mezzelani, P. (2001) Study of 2708 heroin-related deaths in North Eastern Italy 198598 to establish the main causes of death. Addiction, 96, 11271137. Seaman, S. R. & Gore, S. (1998) Mortality from overdose among injecting drug users recently released from prison: database linkage study. British Medical Journal, 316, 426428. Siegel, S. (1984) Pavlovian conditioning and heroin overdose: reports by overdose victims. Bulletin of the Psychonomic Society, 22, 428430. Stimson, G. V. (1995) AIDS and injecting drug use in the United Kingdom, 198793: the policy response and the prevention of the epidemic. Social Science and Medicine, 41, 699716. Strang, J., Sheridan, J. & Barber, N. (1996) Prescribing injectable and oral methadone to opiate addicts: results from the 1995 national postal survey of community pharmacies in England and Wales. British Medical Journal, 313, 270272. Unlinked Anonymous Surveys Steering Group (1999) Prevalence
Addiction, 98, 419425

Trends in drug overdose deaths

425

of HIV in England and Wales: 1998. Annual report of the Unlinked Anonymous Prevalence Monitoring Programme. London: Department of Health. Warner-Smith, M., Darke, S., Lynskey, M. & Hall, W. (2001) Her-

oin overdose:causes and consequences. Addiction, 96, 1113 1125. White, J. M. & Irvine, R. J. (1999) Mechanisms of fatal opioid overdose. Addiction, 94, 961972.

2003 Society for the Study of Addiction to Alcohol and Other Drugs

Addiction, 98, 419425