## Abstract

Objective: To estimate the mortality rates from drug-related deaths and other causes among problem drug users and population attributable risk of death due to opiate use in eight study sites in Europe. Methods: Opiate users were recruited from drug treatment centres during the period 1990–1998 and deaths followed up through national or local mortality registries. Gender-specific overall mortality rate, proportion of deaths by cause (drug-related, HIV, other), standardized mortality ratios (SMRs), and the attributable risk fraction (ARF) were estimated. Results: Crude mortality rates varied from 1 per 100 person-years in the Dublin and London cohorts to 3.8 per 100 person-years in Barcelona. The highest drug-related mortality rate was 10 per 1000 person-years in Barcelona; the rates were ∼7 per 1000 person-years in Denmark, London, Rome, and Vienna, and <3.5 per 1000 person-years for the others cohorts. The mortality rate for AIDS was <2 per 1000 person-years in all the cohorts except Lisbon, Rome, and Barcelona, for which it was ∼6 per 1000 person-years. The highest SMR among males was 21.1 in Barcelona, and among females the highest SMRs were 53.7 and 37.7 in Barcelona and Rome, respectively. In Denmark the ARF was 5%, whereas it was >10% in all other study sites and 24% in Barcelona. Conclusion: Cohort mortality studies, especially in combination with estimates of prevalence, provide useful insights into the impact of opiate use on mortality across European countries and emphasize how preventing overall and drug-related deaths among opiate users can significantly improve the health of the population.

The advantage of mortality statistics over most surveillance systems is that they are invariably complete;1 information on cause of death can also be obtained, although only on the condition that one can identify the population one wants to be informed about. Although drug poisoning/overdose and drug dependence are key causes of death directly attributable to problem drug use, problem drug users can die from many other causes which are not readily visible in routine population mortality statistics because we cannot identify problem drug users in the register. Furthermore, any interpretation of drug-related deaths2 over time has to distinguish between at least two potential explanations: changes in the prevalence of opiate use and in mortality risk. Mortality cohort studies which follow up samples of problem drug users identify all causes of death and measure both the overdose and all cause mortality rate. If successful, they offer the opportunity of a full picture of drug-related mortality and support the interpretation of mortality statistics.3

The occurrence of heroin use and the mortality associated with it vary between geographical areas.410 In general, the opiate overdose mortality rate will be higher in populations with a higher prevalence of injecting drug use. However, other factors may govern the impact of opiate and injecting drug use on opiate poisoning and all cause mortality, notably HIV prevalence among injecting drug users (IDUs) and also the average age of the cohort and availability of substitution treatment. Many longitudinal studies show IDUs to have higher overall and cause-specific mortality rates than the general, gender- and age-matched, population,7,1113 with meta-analysis estimating that the risk of death among injectors was over 13 times higher and that the drug-related mortality rate was ∼0.8% per annum.14 In this study we estimate the excess of mortality among opiate users enrolled in eight European cohorts compared with the general population and the proportion of deaths in the general population that may be attributable to opiate use.

## Methods

The study population included eight cohorts of opiate users collected by the European Monitoring Centre for Drugs and Drug Addiction (EMCDDA) working group on cohort mortality studies among opiate addicts. All cohorts consisted of opiate users aged 15–69 years entering treatment during the period 1990–1999 and followed up using a comparable methodology. Deaths among the subjects were traced by linking the cohorts with national or municipal registers. A more detailed description of the different cohorts and the methodology has been published elsewhere.15 Participant study sites were selected on the basis of the availability of basic information on patients entering treatment and access to population and/or mortality registries. Subjects contributed person-years at risk up to the age of 69 years, the end of the study period, or date of death, whichever came first. Causes of death where available were retrieved according to the following selection of ICD-IX and ICD-X codes: The remaining causes were classified as ‘other’.

• Drug-related deaths: 292; 304.0–.9; 305.2–.9; 965.0–.9; 969.0–.9; E850–E858; E980.0-5; E980.9; E950.0-.5; E950.9; E962; F11–F16; F18–F19; X40–X44, X49; Y10–Y14, Y19; X60–X64; X85.

• HIV-related deaths: 042.0–044.9; 279.1; 279.5; 279.8; B20–B24.

Standardized mortality ratios (SMRs) were calculated in order to estimate the overall mortality excess among opiate users compared with the general population of the same age and gender. SMRs were calculated separately for males and females, with the expected number of deaths in the cohort generated from the mortality rate of the individual country by 5-year age group (for the year in the middle of the follow-up period).16 Two-sided 95% confidence intervals for SMRs were based on the Poisson distribution. The attributable risk fraction (ARF) was calculated as follows (for adults aged 15–49 years because prevalence estimates were available this age group):

$\mathrm{ARF}\ =\ P_{\mathrm{opiate}}\left(\mathrm{SMR}\ {-}\ 1\right)/\left(P_{\mathrm{opiate}}\left(\mathrm{SMR}\ {-}\ 1\right)\ +\ 1\right).$
17 Estimates of the prevalence of opiate use were obtained from the literature and EMCDDA sponsored studies.1822 Credible estimates for Lisbon were not available, so the ARF was not calculated.

## Results

The proportion of males in the cohorts ranged from 69% in Vienna to 82% in Rome, and the average age at recruitment was 25–29 years in six of the cohorts and >30 years in the Danish and Dutch cohorts. Table 1 shows the number of subjects in the cohort, the study period, and crude death rates overall and among males and females aged 15–69 years. The mortality rate varied from just over 1 per 100 person-years in Dublin and London to 2 per 100 person-years in Rome and 3.8 per 100 person-years in the Barcelona cohort (i.e. nearly 1 in 25 cohort subjects dying per year). Death rates for males and females were similar in the cohorts from Amsterdam, Denmark, Barcelona, Lisbon, and Rome, whereas in Dublin, Vienna, and London the death rate was 2–3-fold lower among females in the cohorts than among male opiate users (i.e. 0.5–0.8 per 100 person-years compared with 1.3–1.6 per 100 person-years).

Table 1

Number of subjects enrolled in the eight cohorts and crude death rates (per 1000 person-years)

Study site Study period Number of subjects Person-years Number of deaths Crude death rate

Overall

Males

Females

Amsterdam 1996–2002 2575 10 576.31 174 16.45 16.72 15.39
Barcelona 1992–2001 5037 30 237.06 1137 37.60 38.94 33.38
Denmark 1996–2002 8808 40 317.80 701 17.39 18.33 14.78
Dublin 1994–1997 5285 10 345.27 114 11.02 13.17 5.30
Lisbon 1992–2003 3275 28 619.70 440 15.37 16.13 12.35
London 1997–2001 881 2850.39 35 12.28 13.82 7.97
Rome 1992–1997 5924 21 248.39 425 20.00 19.92 20.37
Vienna 1990–1998 4150 14 834.62 195 13.14 16.41 6.01
Study site Study period Number of subjects Person-years Number of deaths Crude death rate

Overall

Males

Females

Amsterdam 1996–2002 2575 10 576.31 174 16.45 16.72 15.39
Barcelona 1992–2001 5037 30 237.06 1137 37.60 38.94 33.38
Denmark 1996–2002 8808 40 317.80 701 17.39 18.33 14.78
Dublin 1994–1997 5285 10 345.27 114 11.02 13.17 5.30
Lisbon 1992–2003 3275 28 619.70 440 15.37 16.13 12.35
London 1997–2001 881 2850.39 35 12.28 13.82 7.97
Rome 1992–1997 5924 21 248.39 425 20.00 19.92 20.37
Vienna 1990–1998 4150 14 834.62 195 13.14 16.41 6.01

Age range 15–69 years for both calculation of person-years at risk and number of deaths

Table 2 shows the mortality rates and proportion of deaths by cause. The proportion of drug-related deaths varies from ∼7% in Lisbon to 60% in the London cohort. The highest mortality rate from fatal overdose was 13 per 1000 person-years in Barcelona, compared with ∼7 per 1000 person-years in Denmark, London, Rome, and Vienna. No deaths from AIDS were found in the cohort recruited in London, while the proportion of AIDS deaths varied considerably across the other cohorts, from ∼2% in Denmark to 41% in Lisbon, with similar death rates in Rome and Lisbon of ∼6 per 1000. Several of the cohorts had substantial numbers of deaths with missing or unascertained causes, e.g. in Lisbon, London, and Denmark >10% were missing information on cause of death.

Table 2

Number and proportion of deaths, and mortality rate by cause, all ages (15–69 years), by cohort

Study site Drug-related deaths

AIDS deaths

Other causes

Missing causes

n

%

Crude rate/1000

n

%

Crude rate/1000

n

%

Crude rate/1000

n

%

Crude rate/1000

Amsterdama —   —   —
Barcelona 392 34.5 12.96 421 37.0 13.9 280 24.6 9.3 44 3.9 1.46
Denmark 285 40.7 7.07 17 2.4 0.42 169 24.1 4.19 230 32.8 5.70
Dublin 32 28.1 3.09 24 21.1 2.32 54 47.4 5.22 3.5 0.39
Lisbon 32 7.3 1.12 179 40.7 6.25 181 41.1 6.32 48 10.9 1.68
London 21 60.0 7.37 0.0 0.00 10 28.6 3.51 11.4 1.40
Rome 141 33.2 6.64 135 31.8 6.35 142 33.4 6.68 1.7 0.33
Vienna 98 50.3 6.61 37 19.0 2.49 60 30.8 4.04
Study site Drug-related deaths

AIDS deaths

Other causes

Missing causes

n

%

Crude rate/1000

n

%

Crude rate/1000

n

%

Crude rate/1000

n

%

Crude rate/1000

Amsterdama —   —   —
Barcelona 392 34.5 12.96 421 37.0 13.9 280 24.6 9.3 44 3.9 1.46
Denmark 285 40.7 7.07 17 2.4 0.42 169 24.1 4.19 230 32.8 5.70
Dublin 32 28.1 3.09 24 21.1 2.32 54 47.4 5.22 3.5 0.39
Lisbon 32 7.3 1.12 179 40.7 6.25 181 41.1 6.32 48 10.9 1.68
London 21 60.0 7.37 0.0 0.00 10 28.6 3.51 11.4 1.40
Rome 141 33.2 6.64 135 31.8 6.35 142 33.4 6.68 1.7 0.33
Vienna 98 50.3 6.61 37 19.0 2.49 60 30.8 4.04

a: No information on causes of death available

Table 3 shows the SMR comparing the mortality rate in the cohort with the general population. In four of the eight sites the mortality rate was 10 times higher than in the general population. In Barcelona the SMR was 21.1 in males and 53.7 in females; in Rome female opiate users were nearly 40 times more likely to die than females in the general population. In Lisbon the SMR among males was just over 6 times higher than in the general population, and in Vienna the SMR among females was just over 10 times higher. Given that death rates in the adult population are higher in males than in females, the largest differences in SMR between males and females were observed for those cohorts where the mortality rate of opiate users by gender was similar (see table 1), whereas cohorts with higher death rates among male opiate users reported similar SMRs for males and females.

Table 3

All causes standardized mortality ratios (SMRs) and 95% confidence intervals (95% CI), all ages (15–69 years), by cohort, 1990–2000

Study site Observed

Expected

SMR 15–69 (95% CI)

Males

Females

Males

Females

Males

Females

Amsterdam 141 33 19.7 2.7 7.2 (6.1–8.4) 12.2 (8.7–17.2)
Barcelona 894 243 42.4 4.5 21.1 (19.8–22.5) 53.7 (47.4–60.9)
Denmark 543 158 68.4 15.2 7.9 (7.3–8.6) 10.4 (8.9–12.1)
Dublin 99 15 9.2 1.3 10.7 (8.8–13.1) 11.4 (6.9–18.9)
Lisbon 369 71 58.3 4.3 6.3 (5.7–7.0) 16.7 (13.2–21.0)
London 29 2.4 0.4 12.2 (8.5–17.6) 15.8 (7.1–35.1)
Rome 347 78 25.5 2.1 13.6 (12.2–15.1) 37.7 (30.2–47.1)
Vienna 167 28 16.8 2.7 9.9 (8.5–11.6) 10.2 (7.1–14.8)
Study site Observed

Expected

SMR 15–69 (95% CI)

Males

Females

Males

Females

Males

Females

Amsterdam 141 33 19.7 2.7 7.2 (6.1–8.4) 12.2 (8.7–17.2)
Barcelona 894 243 42.4 4.5 21.1 (19.8–22.5) 53.7 (47.4–60.9)
Denmark 543 158 68.4 15.2 7.9 (7.3–8.6) 10.4 (8.9–12.1)
Dublin 99 15 9.2 1.3 10.7 (8.8–13.1) 11.4 (6.9–18.9)
Lisbon 369 71 58.3 4.3 6.3 (5.7–7.0) 16.7 (13.2–21.0)
London 29 2.4 0.4 12.2 (8.5–17.6) 15.8 (7.1–35.1)
Rome 347 78 25.5 2.1 13.6 (12.2–15.1) 37.7 (30.2–47.1)
Vienna 167 28 16.8 2.7 9.9 (8.5–11.6) 10.2 (7.1–14.8)

Table 4 shows estimates of the ARF (i.e. the proportion of deaths in the general population that may be attributable to opiate use if the total surplus of deaths in the cohort were causally related to opiate use). Denmark is the only country-level cohort; it has a low prevalence of opiate use and therefore <5% of adult mortality at ages 15–49 attributable to opiate use. In contrast almost one in four adult deaths in Barcelona could be a consequence of opiate use. In the other study sites the ARF ranged from 10 to 17%.

Table 4

Attributable risk fraction (ARF) of all cause mortality by opiate use, ages 15–49 years, by cohort

Study site

Population

Estimated number of opiate users

Prevalence of opiate use (%)

Standardized mortality ratio (15–49 years)

ARF (%)

Study period

Amsterdam 369 000 5000 1.4 10.1 10.9 1996–2002
Barcelona 746 318 9176 1.2 25.6 22.8 1992–2001
Denmark 2 553 000 14 000 0.5 9.2 4.3 1996–2002
Dublin 588 000 12 250 2.1 10.9 17.0 1994–1997
London 3 919 000 62 000 1.6 13.6 16.6 1997–2001
Rome 1 317 000 14 500 1.1 15.4 13.7 1992–1997
Vienna 786 000 10 200 1.3 9.8 10.3 1990–1998
Study site

Population

Estimated number of opiate users

Prevalence of opiate use (%)

Standardized mortality ratio (15–49 years)

ARF (%)

Study period

Amsterdam 369 000 5000 1.4 10.1 10.9 1996–2002
Barcelona 746 318 9176 1.2 25.6 22.8 1992–2001
Denmark 2 553 000 14 000 0.5 9.2 4.3 1996–2002
Dublin 588 000 12 250 2.1 10.9 17.0 1994–1997
London 3 919 000 62 000 1.6 13.6 16.6 1997–2001
Rome 1 317 000 14 500 1.1 15.4 13.7 1992–1997
Vienna 786 000 10 200 1.3 9.8 10.3 1990–1998

## Discussion

This descriptive study emphasizes the public health importance of drug-related mortality, the scale of risk experienced by opiate users (mostly heroin injectors), and the potential impact of opiate use on the health of the adult population, especially in cities in Europe. The evidence is clear: if we manage to prevent opiate use, transmission of HIV among IDUs, and fatal overdose we will improve significantly the overall health of the population. The findings also underline the value of mortality cohort studies. Further, the difference between the sites suggests that there is unlikely to be any standard or inherent drug mortality rate or SMR among problem drug users. Nevertheless, most of the cohort estimates suggested a higher mortality rate and SMR than an earlier meta-analysis14—i.e. that the annual mortality rate of opiate users is >1 in 100 and could be as high as 3.7 per 100, with an SMR ranging from 6 to 54 and 10–20% of adult mortality at ages 15–49 years attributable to opiate use. In Barcelona nearly one in four deaths among young adults are potentially due to opiate use. The drug-related mortality rate across the sites was more consistent, ranging from 0.7 to 1% in five of the sites. However, the lower rate, represented in Lisbon, may reflect problems with the coding/certification of deaths. Part of the variation in overall mortality between the cohorts reflects the uneven distribution of HIV infection in Europe, with high death rates and SMRs found in Southern European countries where the prevalence of HIV also is higher.3

Although they illustrate the scale of the problem, the study estimates need to be treated cautiously, and our analyses were limited by a lack of information on the generalizability of the cohorts to all opiate users and IDUs in the respective sites. There are several potential biases that could lead to an under- or overestimate of the different mortality measures. First, the mortality rates observed among the cohorts may not be representative of the rates of the total population of opiate users. The study populations were recruited through treatment centres, which under-represent recent opiate users. Moreover, different criteria to enter treatment may result in differential selection across countries. Cohort mortality may be higher if treatment attracts those drug users with the most serious morbidity or it may be lower (immediately after recruitment) because of the protective effect of treatment.

Second, prevalence estimates and thus ARF are based on the official population. Larger cities, however, may attract drug users who do not belong to the registered population of that city.

Third, not all excess mortality is likely to be caused by opiate use: strong risk factors for mortality such as deprivation, smoking, drinking, and other risky behaviour are strongly associated with opiate use; controlled for these might reduce the SMR.

Fourth, the SMR underestimates the risk ratio based on the general population rather than a non-opiate using cohort. However, it must be considered that because we use national reference rates rather than local ones this effect will be small.

In addition to potential methodological issues, the information about the cohorts is too limited to draw conclusions concerning drug policy. Data on injecting status and drug profile were not collected from enough of the subjects in the cohorts to test whether these could explain differences in the mortality rate between the cohorts. Moreover, although all subjects were selected from treatment sites they also differ across sites in terms of age distribution, years of follow-up, and length of time in contact with treatment services, all of which can have an impact on the risk of death.2326 Thus, any differences in the distribution of these factors between subjects in the cohort and the underlying population would add bias to the findings.

Ideally the ARF would be estimated separately for males and females; however, the prevalence estimates were not sufficiently robust and we did not want to compound the uncertainty. In general, the ARF would be larger for males than for females, given that it is likely that about three-quarters of opiate users would be male. Finally, the current cohorts include too many unknowns to permit a comparative assessment of the risk of death among countries. For instance, do the sites with higher mortality rates indicate inequalities in provision of specialist drug treatment or other harm reduction/prevention initiatives and/or greater environmental risk?

Joint evaluation of the mortality rates, SMRs, and prevalence of opiate use, however, could provide useful insights into the impact of opiate use on mortality across European countries and should be a goal of future surveillance of drug-related deaths. Our data corroborate reports that in many European countries drug users have a high mortality risk of death from infectious diseases, AIDS, drug-related deaths, and violence.2730 In conclusion, our results suggest that public health interventions targeted towards opiate users/IDUs, though a comparatively small proportion of the population, may produce a major effect on the health of the general population overall, especially in cities. Moreover, we urge investment in standardized mortality cohorts of opiate users across Europe and reliable estimates of the prevalence of opiate use. These instruments will enable a clearer comparison among countries of the scale of different causes of death among problem drug users in order both to monitor this important public health problem and to identify sites which have reduced the risk of death and improved the health of the population.

Key points

• The study estimates mortality rates of opiate users in Amsterdam, Barcelona, Denmark, Dublin, Lisbon, London, Rome, and Vienna.

• Mortality rates range from 1 per 100 person-years in Dublin and London to 3.8 per 100 in Barcelona and are 6–54 times higher than expected in the general population.

• 10–20% of adult mortality at ages 15–49 years is attributable to opiate use.

• Public health interventions targeted towards opiate users/IDUs may produce a major effect on the health of the general population overall.

• Investment in mortality cohort studies among drug users is required to monitor temporal and geographical trends in overall and cause-specific mortality.

This study has been carried out within the EMCDDA project aimed at promoting cohort studies on the mortality of problem drug users in Europe. The authors thank the EMCDDA for supporting and facilitating the work of the study group.

## References

1
Doll R. Progress against cancer: an epidemiological assessment.
Am J Epidemiol

1991
;
7
:
675
–8.
2
European Monitoring Centre for Drugs and Drug Addiction. Standard protocol for the EU Member States to collect data and report figures for the Key Indicator Drug-Related Deaths by the Standard Reitox tables. EMCDDA project CT.02.P1.05. EMCDDA Scientific Report. Luxembourg: EMCDDA,
2002
.
3
European Monitoring Centre for Drugs and Drug Addiction. Annual Report 2003: the state of the drugs problem in the European Union and Norway; available at http://www.emcdda.eu.int. Luxemburg: EMCDDA,
2003
.
4
Cherubin CE, Sapira JD. The medical complications of drug addiction and the medical assessment of the intravenous drug user: 25 years later.
Ann Intern Med

1993
;
119
:
1017
–28.
5
de la Fuente L, Barrio G, Vicente J, et al. The impact of drug-related deaths on mortality among young adults in Madrid.
Am J Public Health

1995
;
85
:
102
–5.
6
Selwyn PA. Injection drug use, mortality, and the AIDS epidemic.
Am J Public Health

1991
;
81
:
1247
–9.
7
Frischer M, Goldberg, D, Rahman M, Berney L. Mortality and survival among a cohort of drug injectors in Glasgow, 1982–1994.

1997
;
92
:
419
–27.
8
Ortì RM, Domingo-Salvany A, Munoz A, et al. Mortality trends in a cohort of opiate addicts, Catalonia, Spain.
Int J Epidemiol

1996
;
25
:
545
–53.
9
Bargagli AM, Sperati A, Davoli M, et al. Mortality among problem drug users in Rome: an 18 years follow-up study 1980–1997.

2001
;
96
:
1455
–63.
10
Warhen AC, Brandt L, Allebeck P. Has mortality in drug addicts increased? A comparison between two hospitalised cohorts in Stockholm.
Int J Epidemiol

1997
;
26
:
1219
–26.
11
Engström A, Adamsson C, Allebeck P, Rydberg U. Mortality in patients with substance abuse: a follow-up in Stockolm County, 1973–1984.

1991
;
26
:
91
–106.
12
Quaglio G, Talamini G, Lechi A, et al. Study of 2708 heroin-related deaths in north-eastern Italy 1985–98 to establish the main causes of death.

2001
;
96
:
1127
–37.
13
Hickman M, Carnwath Z, Madden P, et al. Drug-related mortality and fatal overdose risk: pilot cohort study of heroin users recruited from specialist drug treatment sites in London.
J Urban Health

2003
;
80
:
274
–87.
14
Hulse GK, English DR, Milne CD, Holman CD. The quantification of mortality resulting from the regular use of illicit opiates.

1999
;
94
:
221
–9.
15
European Monitoring Centre for Drugs and Drug Addiction. Mortality of drug users in the EU: co-ordination of implementation of new cohort studies, follow-up and analysis of existing cohorts and development of new methods and outputs. EMCDDA Scientific Report. Luxembourg: EMCDDA,
2002
.
16
United Nations. Demographic yearbook 1999. New York: United Nations, Statistics Division,
2001
.
17
English DR, Holman CD, Milne E, et al. The quantification of drug-caused morbidity and mortality in Australia 1995. Canberra: Commonwealth Department of Human Services and Health,
1995
.
18
Hickman M, Higgins V, Hope V, et al. Injecting drug use in Brighton, Liverpool, and London: best estimates of prevalence and coverage of public health indicators. J Epidemiol Community Health 2004;58:766–71.
19
Brugal MT, Domingo Salvany A, et al. A small area analysis estimating the prevalence of addiction to opiates in Barcelona, 1993.
J Epidemiol Community Health

1999
;
53
:
488
–94.
20
Comiskey CM, Barry JM. A capture-recapture study of the prevalence and implications of opiate use in Dublin.
Eur J Public Health

2001
;
11
:
198
–200.
21
Buster MC, van Brussel GH, van den Brink W. Estimating the number of opiate users in Amsterdam by capture-recapture: the importance of case definition.
Eur J Epidemiol

2001
;
17
:
935
–42.
22
European Monitoring Centre for Drugs and Drug Addiction. National prevalence estimates of problem drug use in the European Union, 1995–2000. EMCDDA project CT.00.RTX.23, Final report, vol. I: Results. Luxembourg: EMCDDA,
2002
.
23
van Haastrecht HJA, van Ameijden EJC, van den Hoek JAR, et al. Predictors of mortality in the Amsterdam cohort of human immunodeficiency virus (HIV)-positive and HIV-negative drug users.
Am J Epidemiol

1996
;
143
:
380
–91.
24
McLellan AT, Luborsky L, O'Brien CP. Alcohol and drug abuse treatment in three different populations: is there improvement and is it predictable?
Am J Drug Alcohol Abuse

1986
;
12
:
101
–20.
25
Strain EC, Stizer ML, Bigelow GE. Early treatment time course of depressive symptoms in opiate addicts.
J Nerv Ment Dis

1991
;
179
:
215
–21.
26
Galai N, Safaeian M, Vlahov D, et al. Longitudinal patterns of drug injection behaviour in the ALIVE Study cohort, 1988–2000: description and determinants.
Am J Epidemiol

2003
;
158
:
695
–704.
27
Robertson JR, Ronald PJ, Raab JM, et al. Deaths, HIV infection, abstinence, and other outcomes in a cohort of injecting drug users followed up for 10 years.
Br Med J

1994
;
309
:
369
–72.
28
Oppenheimer E, Tobutt C, Taylor C, Andrew T. Death and survival in a cohort of heroin addicts from London clinics: a 22-year follow-up study.

1994
;
89
:
1299
–1308.
29
Marzuk PM, Tardiff K, Leon AC, et al. Fatal injuries after cocaine use as a leading cause of death among young adults in New York City.
N Engl J Med

1995
;
332
:
1753
–7.
30
Sanchez-Carbonell X, Seus L. Ten-year survival analysis of a cohort of heroin addicts in Catalonia: the EMETYST project.