From the Muirhouse Medical Group (Ms Copeland and Drs Budd, Robertson, and Elton), Department of Community Health Sciences, University of Edinburgh (Dr Robertson), and Statistical Services (Dr Elton), Edinburgh, Scotland. The authors have no relevant financial interest in this article.
High mortality among drug users has been widely recognized. This study investigates, in a large family practice of 10 000 patients in Edinburgh, Scotland, whether there has been a change in causes of mortality over time. Patients known to have ever injected drugs were recruited into a cohort study from 1980 until 2001.
Death certificates and clinical notes were scrutinized and data relating to demographic features, drug use, and causes of death were recorded.
Of 667 patients, there were 153 deaths at follow-up (110 men and 43 women). Average annual mortality rate was 2.3%. Death rate peaked in the early to mid-1990s, reflecting the development of advanced human immunodeficiency virus (HIV) from the early epidemic in 1982-1984 and the onset of the effect of antiviral chemotherapy. Drug deaths and suicide were the same in both sexes but tended to occur in younger subjects. Principal cause of death was overdose in the early years and HIV/AIDS in later years. Toward the close of the study period, hepatitis C emerged as a cause of death.
Injecting drug users have a very high risk of mortality. Infectious diseases from nonsterile injecting are the most obvious preventable cause of death. Use of death certificate information alone is inaccurate in analyzing drug-related deaths and greatly underestimates the full impact of the HIV epidemic. This study provides some of the most convincing evidence so far that harm minimization, in its broadest sense, is effective in reducing drug-related mortality.
Death rates and causes of death are important as outcome measures of drug use for various reasons. In practical terms, they are hard outcome measures compared with the notoriously difficult-to-assess values of abstinence, remission rates, and quality of life. Deaths can also be counted from national databases derived for other purposes and can, in theory at least, be compared among regions and countries. However, there are various problems using mortality data. Coding systems (International Classification of Diseases [ICD]) have few specific codes relating to drug use, reporting of causes is inconsistent and variable, and most studies probably underestimate effects of drugs on suicide, violence, and early death due to accelerated common degenerative diseases.1,2
Important data from contemporary research give information on increased mortality due to drugs. Reports show that the number of deaths in drug users compared with non–drug-using populations is approximately 13 times higher,3 with causes of death varying among different centers. In Western Europe and North America, overdose remains the predominant cause of death. Tuberculosis is more common in certain countries,4- 6 and minor epidemics, usually due to bacterial contamination, although infrequent are important.7,8 Human immunodeficiency virus (HIV)/AIDS is rapidly becoming the most important cause of mortality in many southern European and Asian centers.9,10 Several studies have shown the impact of HIV and AIDS, with many presentations at the 14th International AIDS Conference in Barcelona, Spain, providing as yet unpublished information on epidemics of drug injecting and HIV in many Asian countries. These developments have been identified at an early stage10 compared with the later rising impact of HIV in Western countries identified by other studies.11 However, certain studies12 have identified a comparatively small impact of HIV in cohorts of injecting drug users.
Studying deaths can inform policy in the area of prevention and treatment. Mortality rates can be used to identify the success or failure of drug policy and treatments. Changing patterns of drug misuse give rise to changing patterns of morbidity and mortality, such as epidemics of bloodborne virus13 or deaths from overdose, which vary over time depending on availability of illegal and legal drugs.14- 16
It is in this context that cohort studies are of particular use in that they show patterns of drug-related risk behavior and the associated morbidity and mortality in a population. Cohort studies like this allow comparisons with survivors and highlight any differences there might be between this group and those who have died, thus providing valuable information for health and social policy makers. One of the values of this study is that it is rare to have a prospective study that can study deaths as they occur and explore changing causes over time. The primary medical care setting is also significant in that it allows access to extensive clinical information, making it easier to identify the real pattern of risk-taking behavior for individuals, compared with other studies that may have no data on ongoing risk behavior. Clinical material enables greater accuracy in determining the main cause of death.
The objectives of this study were to identify the mortality risk and explore the causes of death in a group of known injecting drug users during long-term follow-up for a period of 21 years (1980-2001). In previous communications, the unfolding HIV epidemic in the cohort and trends in death have been presented.17,18 However, new patterns emerge with time as risk-taking behavior changes, and this study was designed to investigate these patterns.
Since 1980 all known drug injectors who attended a family medical practice in northwest Edinburgh, Scotland, have been recorded in a database that includes a range of demographic and clinical information. Given this community-based setting, rather than a specialist drug misuse treatment center, the study is potentially less open to bias through selection than other cohort studies. Although the study participants were resident in the practice catchment area at the time of enrollment, many had resided in other parts of the city before this, thus giving a more representative sample of Edinburgh's drug injectors.
Ethical approval for this long-running cohort study was granted by the Lothian General Practice and Community Care Ethical Committee and consent obtained from patients when entering the study. At enrollment into the study, patients were flagged at the General Register Office for Scotland (the department of the devolved Scottish Administration responsible for the registration of births, marriages, deaths, divorces, and adoptions in Scotland). This office also maintains the National Health Service Central Register (the central register that maintains records of all National Health Service patients and birth records of those registered with a general medical practitioner), and research study populations are flagged with a cipher representing that particular study so that when deaths notified to this register are found to relate to such entries, researchers can be informed of the death and supplied with copies of certified causes of death. (All information provided by the General Register Office is treated in a highly confidential manner.) Thus, for this particular study, flagging ensured that medical records and death certificates were returned to the researcher, even if the patients have moved from the practice population.
At the end point for the current study (December 2001), the cohort size was 667, of whom 7 were lost to follow-up. Full and valid information was available for 445 men and 215 women. Of this group, 153 (23%) were dead (110 men and 43 women). Information was available for all of the 153 persons. Descriptive data were gathered and analyzed to give the mean age of first injecting drugs and the mean age of entry into the cohort, identifying any differences between the dead and those still alive. Mean age at death, mean number of years of follow-up for the dead, and mean years of death from year of first injecting were also calculated.
The annual mortality rate and the mortality rate per 100 years of cohort membership were calculated, along with the overall standardized mortality ratio for the cohort. The mortality rates for different age groups within the cohort were then compared with age- and sex-matched groups in the general population and the relative risk calculated.
A close scrutiny of causes of death was undertaken by examining the principal or underlying cause as noted on the actual death certificate and supplementing this information with that from patients' clinical records when required. This information was available for all of the dead patients. This gave a clearer and more accurate picture of causes of death when ICD codes were found to be misleading for both drug-related and HIV-related deaths. The following 6 categories of principal causes were identified by the researchers: HIV related/AIDS, drug related (mainly overdose and short-term drug misuse problems), hepatitis C, suicide, mixed (a variety of causes such as chronic alcoholism, cerebral hemorrhage, epilepsy, and 1 murder), and liver disease other than hepatitis C.
Statistical associations between different variables and mortality were determined using Poisson regression test to test for association with age and logistic regression to test for associations among age at death, sex, and the 6 principal identified causes of death.
Descriptive characteristics are shown in Table 1. For age of entry into the cohort, which might be taken as a marker for entry into treatment services, there is no significant difference between those who are still alive and those who have died for both sexes. This is also the case for the age of first injection. Mean age at death was 32.9 years, mean time of follow-up for those who died was 7.6 years, and mean time from first injection until death was 13.9 years. All these factors were similar for men and women (Table 1).
There were no fatalities in the cohort between 1980 and 1982. Annual mortality rates (Table 2) ranged from 1.2% in 1984 to 4.4% in 1994. The average annual mortality rate from 1983 to 2001 was 2.3%. The mortality rate per 100 years of cohort membership was 2.45% (153/6244). The male patients had 110 deaths in 4214 person-years of follow-up, and the female patients had 43 deaths in 2030 person-years of follow-up.
Table 2 gives a breakdown of the number of deaths in the study per annum and shows that the annual numbers of deaths were highest in the early to mid-1990s. However, in the table high numbers reflect the increasing annual cohort population size and show that the annual mortality rate is in fact fairly constant throughout the study in the 1980s, increasing in the mid-1990s (reflecting the high numbers of HIV deaths in the cohort), and then declining again the late 1990s.
Comparisons were made with death rates for an age- and sex-matched population drawn from the General Register for Scotland figures for 1990, the middle of the cohort's follow-up period. An overall standardized mortality ratio for the cohort was calculated at 1745 (95% confidence interval [CI], 1459-2030). For both sexes, the number of deaths was highest in the 25- to 34-year age group, with 60% of male deaths and 42% of female deaths occurring in this age group. The relative mortality risk was calculated, and it was apparent that young female patients in the cohort (age 15-24 years) were at a very high relative risk of dying (increased by a factor of 76.3; 95% CI, 34.9-144.9), whereas young male patients were not (increased by a factor of 5.4; 95% CI, 1.5-13.8). Poisson regression to test for association with age was performed for both sexes, and it was found that for female patients there was no significant trend in the crude mortality rate but a highly significant downward trend in relative risk (P<.001). This confirmed the impression that female death rates in the cohort are mainly independent of age and that, as expected, the relative risk decreased with age, due to the natural increase in mortality rate in the general population with age. For male patients, it was found that there was a significantly nonlinear trend in both the crude rate and relative risk (P<.001 in both cases), and although the crude rate rises with age, the relative risk fails to show such a clear trend (Table 3).
Interestingly, in this cohort there was a significantly downward trend in the standardized mortality ratio over time once adjusted for age and sex (P<.001), perhaps reflecting the success of harm minimization in curbing the HIV epidemic. Thus, for patients recruited before 1985, the ratio of observed deaths to expected is 19 for the male patients and 29 for the female patients compared with 13 and 18 for those recruited later than this. For the whole study period, from 1980 to 2001, HIV related/AIDS was the principal cause of death (82, 53%), followed by drug related (38, 25%), hepatitis C (11, 7%), suicide (9, 6%), mixed (9, 6%), and liver disease other than hepatitis C (4, 3%).
However, when examining the 2 periods of the study, 1980-1990 and 1991-2001, it is clear that drug-related deaths was the principal cause in the cohort in the 1980s and that HIV-related deaths became dominant in the mid-1990s (Table 4 and Figure 1). Taken together, deaths from all causes in the first period of the study (n = 28) and deaths from bloodborne viruses (n = 85) account for 74% of all deaths in the cohort (acquisition bloodborne viruses occurred in the first period). Early indications from the 15 deaths in the period 2000-2001 are that HIV is still dominant, followed by drug-related causes, but the researchers expect that hepatitis C will become increasingly significant (Figure 2).19
Changing patterns in causes of death for the 2 periods of the study. HIV indicates human immunodeficiency virus.
Changes in causes of death over time. HIV indicates human immunodeficiency virus.
To determine whether age at death or sex is associated with the proportion of deaths from each of the 6 causes identified, logistic regression was performed. There were no significant differences according to sex, but drug deaths and suicides tended to occur in younger subjects (P = .004 and P = .02, respectively). Seventy-six percent of the total of drug-related deaths occurred in the 15- to 34-year age group compared with 61% of HIV-related/AIDS deaths (Table 5).
One of the interesting aspects of the Edinburgh study is that the principal cause of death differs from that of many other European studies of drug users. Although many other studies12,14- 16,20 have found that most deaths were associated with drug overdoses, the main cause of death in this cohort was HIV related/AIDS.
Comparing this cohort with the Glasgow drug injector cohort,21 mean age at entry into the Edinburgh cohort is slightly older at 26.3 years compared with 24.4 years. Mean age at first injection is very similar at 19.9 years for Glasgow compared with 19.2 years for those still alive and 19.4 years for the dead in the Edinburgh cohort. Where the 2 cohorts differ is in the mean age at death, which is 32.9 years for the Edinburgh cohort and 26.3 years for the Glasgow cohort. This might be explained by the difference in the principal cause of death in the 2 cohorts (HIV related in Edinburgh and drug related in Glasgow), reflecting the long incubation period of AIDS, which became so dominant in the Edinburgh cohort in the second decade of the study. The mortality rate per 100 years of cohort membership was 2.45% (153/6244) compared with 2.08% (53/2547) in the Glasgow cohort. The annual mortality rate for this study is 2.3% compared with 1.8% for the study by Oppenheimer et al,14 3.3% for the study by Segest et al,22 2.2% for the study by Fugelstad,23 1.2% for the study by Gossop et al,16 and 2.7% for the study by Sanchez-Carbonell and Vilaregut.24
The nature of drug misuse exposes misusers to health risks they would not normally incur. This risk behavior in drug users contributes to their causes of death being dissimilar to the general population, where cancer and circulatory disease predominate, and these causes of death in drug-using populations are influenced by changes in drug-using behavior over time. This study population has been followed up for a considerable period, and full and valid information was available for most of the group, giving the advantage of a high degree of reliability in identifying and analyzing the changes over time. The researchers acknowledge that since the period of enrollment covers more than 20 years, people entering the study at different time points could be different with respect to sociodemographic characteristics and drug use habits, for example, primary drug injected. The possibility of a selection bias giving rise to an apparently greater mortality was also considered. Most studies reported from specialist centers or clinics have similar populations, although some have access only to older individuals. Our general medical practice setting provides no additional services or opportunities other than proximity to the individual's home address. It seems unlikely that this study attracted individuals with an increased risk of medical complications.
Most deaths in the cohort occurred in the 25- to 34-year age group, indicating that younger drug users are at greater risk in the early stages of their addiction. Most of the cohort was enrolled into the study approximately 7 years after first injecting drugs and died approximately 7 years after entering the cohort, with a mean age at death of 32.9 years. Many of these deaths were caused by drug overdose, clearly highlighting the high risk involved in misusing drugs at the beginning of the young user's career. In particular, young female drug users in the 15- to 24-year age group have a very high relative risk of dying (increased by a factor of 76.3) compared with young females in the general population, and in nearly 90% of these cases drug overdose was the principal cause.
Taking into account the deprived inner-city lifestyle with high crime rates and unemployment, it is perhaps surprising that there was only 1 violent death and comparatively few suicides. There was an increase in the number of suicides in the second decade of the study, and although several of these cohort members had HIV-positive status, the researchers could only speculate whether this was a significant factor in their deaths.
Although drug-related deaths were the most prominent in the early years of the study, consistent with other cohort studies from the United Kingdom14,21 deaths from HIV-related causes predominated in the mid-1990s. A well-described epidemic of HIV, hepatitis B, and hepatitis C occurred in Edinburgh between 1982 and 198425 after which time transmission, of HIV at least, has been minimal in the drug-using population. The long incubation period of HIV means that the mortality due to this virus increased throughout the early 1990s and was interrupted only by the introduction of highly active antiretroviral therapy (HAART) in 1996, showing a marked decline after this time. Despite the increase in deaths in the latter period of the study, most of the risk exposure occurred in the early years, some even before recruitment. Taken together, the deaths from all causes in the first period and the deaths from bloodborne viruses in the second period account for 74% of all deaths. This further supports the hypothesis that patterns of behavior, particularly damaging injecting, have changed and that this change occurred some time in the mid to late 1980s. Clinical experience reinforces this message, since there have been no recent HIV seroconverters among the cohort. This study provides some of the most convincing evidence so far that harm minimization, in its broadest sense, is effective.
It is important that this study was able to access additional information about causes of death to that on the formal death certificate. It was clear in many cases that the given diagnosis deliberately or accidentally concealed the diagnosis of HIV/AIDS as the principal cause of death. Diagnoses such as pneumonia, bronchopneumonia, hematological disorder, cerebral neoplasm, lymphoma, and others were clearly, when investigated, symptomatic of HIV in advanced stages. This obscuring of diagnosis of the cause of death leads to a significant underestimate of the impact of the HIV epidemic and has been noted in at least one other study.26
Although this is essentially a study of deaths in drug users, it became apparent that there are many long-term HIV survivors, some undergoing treatment and some who have deliberately stayed out of treatment. (The researchers are currently planning a new follow-up study that will focus on this aspect.) This is an important observation, because conventional wisdom reports that those infected early in the HIV epidemic are all likely to be dead. This is clearly not true among this cohort. There are many long-term survivors and a small number of nonprogressors.
A significant change in the pattern of causes begins to emerge in the later years of the study with the emergence of hepatitis C. The identification of hepatitis C as a cause on death certificates became apparent in the second decade of the study (from death certificates dated 1996 onward), having not been identified earlier. In a previous study of this cohort, the prevalence of hepatitis C virus infection was 65% compared with 43% for HIV. Approximately 80% of those who tested HIV positive were also positive for hepatitis C virus. Most deaths in the cohort occurred in the early to mid-1990s, and the hepatitis C prevalence rate for the living contemporaries of those who died during this period is more than 80%. Although the researchers, in a previous study, have been unable to identify an entry point for the hepatitis C virus into the main cohort (due to it predating any stored early blood samples), it is likely to have entered the cohort in the early 1980s.19 It is, therefore, reasonable to assume that the prevalence among the 153 dead would be similar to their contemporaries. The comparatively recent identification of hepatitis C as an infectious disease means that, although many of the dead in the cohort may have had hepatitis C, this would not have been recorded as a cause of death. The evident decline in HIV and AIDS deaths in the cohort in recent years due to effectiveness of HAART means that we are now seeing liver damage caused by hepatitis C identified more frequently as the cause of death. Other studies27,28 have shown the very high transmission rate of hepatitis C infection in the first year or two after staring to inject. Therefore, to address the trend of increasing hepatitis C–related mortality, as identified by this study, it is again vital for services to engage with young drug users at an early stage.
Many harm reduction measures were brought into place (needle and syringe exchanges, increased access to methadone maintenance programs, peer group, and other sources of behavior education) to tackle the predicted epidemic of HIV. Declining prevalence and incidence of HIV among injecting drug users are further evidence that these measures have had considerable success. It is, however, difficult to determine the true effect of harm reduction on the transmission of hepatitis C virus because investigations of its epidemiology did not generally start until after these harm reduction measures were in place. What seems apparent at present is that strategies that have been effective in controlling the spread of HIV in injecting drug users, particularly needle and syringe exchanges, have not had similar success with the spread of hepatitis C virus in this group. Emerging data from Asian centers demonstrates how HIV/AIDS will be a major problem among drug users and their heterosexual partners in the next few years. As the HIV/AIDS epidemic unfolds in these developing countries, driven often by injecting drug use,29 strategies for prevention should become increasingly important given the evidence for the impact of the harm minimization approach.
This study demonstrates that data from long-term follow-up can provide information on a number of important issues, such as whether people addicted to drugs are at greater risk of dying in the early or late phases of their addiction and treatment. It also highlights the factors involved in their deaths, such as bloodborne virus infection, drug overdose, mental health problems, and alcohol abuse, that marginalize them from the general population and ultimately result in loss of life. Data on abstinence or continued drug use are an important part of this study.13 No relationships have been identified among abstinence, continued use, and mortality. It is apparent that more detailed studies are required that gather data on the cumulative effect of high-risk behavior; this study, however, confirms that drug users are exposed to much greater risk of death than the general population.
Clear indications are that injecting drug users need to be targeted at a younger age by government and health care services so they can be offered harm minimization advice with substitute prescribing if appropriate, physical health care, and counseling support when they are most at risk. By adopting a longitudinal perspective and studying specific cohorts, information can be obtained about the natural history and career of addiction that will be vital in the areas of prevention, treatment, and social policy. Cohort studies such as this one can be invaluable in identifying the key determinants involved in injecting drug use and assessing the impact of this behavior in terms of the individual and society as a whole.
Corresponding author and reprints: J. Roy Robertson, FRCP(Edin), FRCGP, Edinburgh Drug Addiction Study, 1 Muirhouse Ave, Edinburgh EH4 4PL, Scotland (e-mail: firstname.lastname@example.org).
Accepted for publication June 27, 2003.
We thank the staff and physicians of the Muirhouse Medical Group, Edinburgh, for their support in performing this research study. Ethical approval for the study was granted by the Lothian General Practice and Community Care Ethical Committee and consent obtained from patients when entering the study.
Thank you for submitting a comment on this article. It will be reviewed by JAMA Internal Medicine editors. You will be notified when your comment has been published. Comments should not exceed 500 words of text and 10 references.
Do not submit personal medical questions or information that could identify a specific patient, questions about a particular case, or general inquiries to an author. Only content that has not been published, posted, or submitted elsewhere should be submitted. By submitting this Comment, you and any coauthors transfer copyright to the journal if your Comment is posted.
* = Required Field
Disclosure of Any Conflicts of Interest*
Indicate all relevant conflicts of interest of each author below, including all relevant financial interests, activities, and relationships within the past 3 years including, but not limited to, employment, affiliation, grants or funding, consultancies, honoraria or payment, speakers’ bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued. If all authors have none, check "No potential conflicts or relevant financial interests" in the box below. Please also indicate any funding received in support of this work. The information will be posted with your response.
Register and get free email Table of Contents alerts, saved searches, PowerPoint downloads, CME quizzes, and more
Subscribe for full-text access to content from 1998 forward and a host of useful features
Activate your current subscription (AMA members and current subscribers)
Purchase Online Access to this article for 24 hours
Some tools below are only available to our subscribers or users with an online account.
Download citation file:
Web of Science® Times Cited: 45
Customize your page view by dragging & repositioning the boxes below.
and access these and other features:
Enter your username and email address. We'll send you a link to reset your password.
Enter your username and email address. We'll send instructions on how to reset your password to the email address we have on record.
Athens and Shibboleth are access management services that provide single sign-on to protected resources. They replace the multiple user names and passwords necessary to access subscription-based content with a single user name and password that can be entered once per session. It operates independently of a user's location or IP address. If your institution uses Athens or Shibboleth authentication, please contact your site administrator to receive your user name and password.