Bulletin of the World Health Organization

Mortality among people who inject drugs: a systematic review and meta-analysis

Bradley M Mathers a, Louisa Degenhardt b, Chiara Bucello b, James Lemon b, Lucas Wiessing c & Mathew Hickman d

a. The Kirby Institute, University of New South Wales, Sydney, NSW 2052, Australia.
b. National Drug and Alcohol Research Centre, University of New South Wales, Sydney, Australia.
c. Scientific Division, European Monitoring Centre for Drugs and Drug Addiction, Lisbon, Portugal.
d. School of Social and Community Medicine, University of Bristol, Bristol, England.

Correspondence to Bradley M Mathers (e-mail: bmathers@kirby.unsw.edu.au).

(Submitted: 31 May 2012 – Revised version received: 26 November 2012 – Accepted: 28 November 2012.)

Bulletin of the World Health Organization 2013;91:102-123. doi: 10.2471/BLT.12.108282

Introduction

People who use drugs, especially by injection, are at higher risk of dying from both acute and chronic diseases, many of which are related to their drug use, than people who do not use these drugs. Fatal overdose and infection with human immunodeficiency virus (HIV) and other blood-borne viruses transmitted through shared needles and syringes are the most common causes of death in this group.1 Understanding causes of death is important when setting priorities for programmes designed to reduce deaths from the use of drugs. Longitudinal studies of people who inject drugs are critical for assessing the magnitude, nature and correlates of the risk of death in this population.

A systematic review conducted in 2004 identified 30 prospective studies published between 1967 and 2004 that dealt with “problematic drug users” or people who inject drugs.2 These reviews have consistently shown that the practice of injecting drugs is associated with an elevated risk of death, particularly from the complications of HIV infection, drug overdose and suicide. Since these reviews were conducted, the number of studies examining mortality among cohorts of people who inject drugs has risen substantially. This has made it possible to perform fine-grained analyses that were not feasible in earlier reviews. Furthermore, those earlier reviews did not examine the potential impact of study-level variables, variation across countries, or of participant-level variables that could affect both mortality rates and differences in causes of death, yet study-level evidence suggests that males who inject drugs may be at higher risk of dying than females and that different types of drugs are associated with different risks of death.35 Findings from other reviews have also suggested that rates of death among people who are dependent on opioids are different from the rates of death observed in people who are dependent on stimulants such as cocaine and amphetamine type stimulants.35

In recent years the number of studies reporting on mortality among people who inject drugs has increased. Hence, the objective of this review was to determine the following:

  • overall crude mortality rates (CMRs) and excess deaths across cohorts of people who inject drugs, by sex;
  • causes of death across studies, particularly from drug overdose and acquired immunodeficiency syndrome (AIDS); differences in mortality rates and causes of death among HIV-positive (HIV+) and HIV-negative (HIV−) people who inject drugs;
  • differences in mortality rates across cohorts by geographical location and country income level;
  • mortality rates by type of drug injected (e.g. opioids versus stimulants);
  • mortality rates during in-treatment and off-treatment periods.

Methods

Identifying studies

A recent series of reviews identified cohort studies among opioid, amphetamine and cocaine users to examine mortality.35 In these reviews tailored search strings were used to search three electronic databases for studies published between 1980 and 2012: Medline, EMBASE and PsycINFO. The search strings contained keywords and database-specific terms (MeSH headings, EMTREE terms and explode terms; Box 1). All results were limited to human subjects. We identified grey literature sources reporting on mortality by searching online grey literature databases, library databases and the web sites listed in a published technical report.6 To make sure that no relevant papers had been missed, we sent the draft lists of the papers identified through these searches to experts for their review.

Box 1. Strategy for search of the peer-reviewed literature

Database specific search terms were developed and combined using Boolean operators as follows:

( < opioids > OR < cocaine > OR < amphetamine type stimulants > ) AND < drug use > AND < mortality > AND < longitudinal studies > 

All results were limited to human subjects and publication years between 1980 and 2012. The full search strings used for each database were as follows:

Medline: ((heroin or opiate$ OR opium OR opioid$ OR Exp Opium/ OR exp Narcotics/ OR exp Heroin Dependence/ OR exp Heroin/ OR exp Morphine/ OR exp Opioid-Related Disorders/ OR exp Opiate Alkaloids/ OR exp Methadone/ OR exp Analgesics, Opioid/) OR (Cocaine exp Cocaine-Related Disorders/ or exp Cocaine/ or exp Crack Cocaine/) OR (ATS OR amphetamine type stimulant$ OR amphetamine$ OR methamphetamine OR deoxyephedrine OR desoxyephedrine OR Desoxyn OR madrine OR metamfetamine OR methamphetamine hydrochloride OR methylamphetamine OR n-methylamphetamine OR d-amphetamine OR dextroamphetamine sulfate OR dexamphetamine OR dexedrine OR dextro-amphetamine sulfate OR dextroamphetamine sulfate OR d-amphetamine sulfate OR stimulant$ exp amphetamines/ or exp amphetamine/ or exp dextroamphetamine/ or exp p-chloroamphetamine/ or exp 2,5-dimethoxy-4-methylamphetamine/ or exp p-hydroxyamphetamine/ or exp iofetamine/ or exp methamphetamine/ or exp benzphetamine/ or exp phentermine/ or exp chlorphentermine/ or exp mephentermine/ or exp amphetamine-related disorders/)) AND (drug abuse$ OR drug use$ OR drug misuse$ OR drug dependenc$ OR substance abuse$ OR substance use$ OR substance misuse$ OR substance dependenc$ OR addict$ OR Exp Substance-related disorders/) AND (Mortal$ OR fatal$ OR death$ OR exp “death and dying”/ OR exp mortality/ OR exp hospitalization) AND (“cohort” OR “longitudinal” OR “incidence” OR “prospective” OR “follow-up” OR exp cohort studies/ OR exp longitudinal studies/ OR exp follow-up studies/ OR exp prospective studies/)

EMBASE: ((heroin OR opioid$ OR opiate$ OR opium OR exp Diamorphine/ OR exp Opiate/ OR exp Methadone treatment/ OR exp Methadone/) OR (Cocaine exp Cocaine-Related Disorders/ or exp Cocaine/ or exp Crack Cocaine/) OR (ATS OR amphetamine type stimulant$ OR amphetamine$ OR methamphetamine OR deoxyephedrine OR desoxyephedrine OR Desoxyn OR madrine OR metamfetamine OR methamphetamine hydrochloride OR methylamphetamine OR n-methylamphetamine OR d-amphetamine OR dextroamphetamine sulfate OR dexamphetamine OR dexedrine OR dextro-amphetamine sulfate OR dextroamphetamine sulfate OR d-amphetamine sulfate OR stimulant$ exp amphetamines/ or exp amphetamine/ or exp dextroamphetamine/ or exp p-chloroamphetamine/ or exp 2,5-dimethoxy-4-methylamphetamine/ or exp p-hydroxyamphetamine/ or exp iofetamine/ or exp methamphetamine/ or exp benzphetamine/ or exp phentermine/ or exp chlorphentermine/ or exp mephentermine/ or exp amphetamine-related disorders/)) AND (Drug abuse OR drug use$ OR drug misuse OR drug dependenc$ OR substance abuse OR substance use$ OR substance misuse OR substance dependenc$ OR addict$ OR exp substance abuse/ OR exp drug abuse/ OR exp analgesic agent abuse/ OR exp drug abuse pattern/ OR exp drug misuse/ OR exp drug traffic/ OR exp multiple drug abuse/ OR exp addiction/ OR exp drug dependence/ OR exp cocaine dependence/ OR narcotic dependence/ OR exp heroin dependence/ OR exp morphine addiction/ OR exp opiate addiction/) AND (Mortal$ OR fatal$ OR death$ OR exp death/ OR exp “cause of death”/ OR exp accidental death/ OR exp sudden death/ OR exp fatality/ OR exp mortality/ OR exp hospitalization/) AND (“cohort” OR “longitudinal” OR “incidence” OR “prospective” OR “follow-up” OR exp cohort analysis/ OR exp longitudinal study/ OR exp prospective study/ OR exp follow up/)

PsychINFO: ((“heroin” OR “opium” OR “opiate$” OR “methadone” OR exp Opiates/ OR exp METHADONE/ OR exp HEROIN ADDICTION/ OR exp HEROIN) OR (Cocaine exp Cocaine-Related Disorders/ or exp Cocaine/ or exp Crack Cocaine/) OR (ATS OR amphetamine type stimulant$ OR amphetamine$ OR methamphetamine OR deoxyephedrine OR desoxyephedrine OR Desoxyn OR madrine OR metamfetamine OR methamphetamine hydrochloride OR methylamphetamine OR n-methylamphetamine OR d-amphetamine OR dextroamphetamine sulfate OR dexamphetamine OR dexedrine OR dextro-amphetamine sulfate OR dextroamphetamine sulfate OR d-amphetamine sulfate OR stimulant$ exp amphetamines/ or exp amphetamine/ or exp dextroamphetamine/ or exp p-chloroamphetamine/ or exp 2,5-dimethoxy-4-methylamphetamine/ or exp p-hydroxyamphetamine/ or exp iofetamine/ or exp methamphetamine/ or exp benzphetamine/ or exp phentermine/ or exp chlorphentermine/ or exp mephentermine/ or exp amphetamine-related disorders/)) AND (Drug abuse OR drug use$ OR drug misuse OR drug dependenc$ OR substance abuse OR substance use$ OR substance misuse OR substance dependenc$ OR addict$ OR Exp drug abuse/ OR exp drug addiction/ OR exp addiction/ OR exp drug usage) AND (Mortal$ OR fatal$ OR death$ OR exp “death and dying”/ OR exp mortality/ OR exp hospitalization) AND (“cohort” OR “longitudinal” OR “incidence” OR “prospective” OR “follow-up” OR Exp age differences/ OR exp cohort analysis/ OR exp human sex differences)

For the current study we examined all papers found in the reviews of drug-related mortality but selected only cohorts composed of people who injected opioids and other drugs. We used the strategy outlined in the preceding paragraph to further search for these cohorts. We included in the analysis only studies of drug users that included mortality data disaggregated by participants’ injecting drug use; studies were included only if more than 70% of the cohort was composed of people who injected drugs.

The searches yielded a total of 5981 studies of mortality related to the use of opioids, amphetamines and cocaine. We identified another 79 articles by searching the reference lists of reviews on mortality related to drug use. Experts provided additional studies for 16 cohorts. From these 5981 articles we excluded a total of 5762: 4999 did not focus on drug dependence or mortality, 118 did not include raw data, 292 were case series, and 600 had insufficient mortality data on people who inject drugs. In total, we selected 67 cohort studies for inclusion in the analyses (Fig. 1). These studies were further assessed using STROBE reporting guidelines.7

Fig. 1. Flowchart showing study selection process for systematic review of studies on mortality in people who inject drugs
Fig. 1. Flowchart showing study selection process for systematic review of studies on mortality in people who inject drugs

Data extraction

Once we had identified all studies, one of the authors (JL) extracted the data into an Excel database (Microsoft, Redmond, United States of America) and two others rechecked them (BM, CB). This yielded the basic data set for the statistical analyses. We extracted information on the location of each study, the period of recruitment and duration of follow-up, the number of people in the cohort, the percentage of people in the cohort who injected drugs, the number of person–years (PY) of follow-up and the number of deaths.

We extracted CMRs and standardized mortality ratios (SMRs). We expressed CMRs as the number of deaths per 100 PY of follow-up. We reported SMRs as calculated in the source papers. In several cases standard errors, confidence intervals (CIs) and CMRs were not reported, so we estimated them using standard calculations. We also put into the database CMRs and SMRs that were reported according to sex, HIV status, treatment status and type of drug injected, as well as data on deaths from drug overdose or AIDs-related causes.

We included in the analyses studies that specified treatment status if they classified the data by mutually exclusive treatment groups or periods. We only included studies in which the exact dates of entry into and exit from the study had been recorded and used to calculate the number of PYs, the number of deaths and mortality rates.

Statistical analysis

We performed meta-analyses to estimate pooled all-cause CMRs and all-cause SMRs, and pooled estimates of deaths from specific causes, as in previous reviews.8 To perform the meta-analyses we used the “metan” command in STATA version 10.1 (StataCorp LP, College Station, USA). The “metan” command uses inverse-variance weighting to calculate random effects pooled summary estimates and their confidence limits, true effect differences between studies and study heterogeneity.9,10 Random effects models allow for heterogeneity between and within studies. We expected high levels of heterogeneity between studies because of the marked differences between the samples of people injecting drugs; accordingly, we applied a random effects model to all analyses. The appropriateness of this a priori decision was confirmed by the resultant χ2 and the I-squared statistic. To further investigate this heterogeneity, when the data permitted we divided the cohorts into subgroups and used CMR ratios to compare differences in mortality.11 We made comparisons between subgroups as follows: sex (male versus female); primary drug injected at baseline (opioids versus stimulants); HIV status (HIV+ versus HIV−); and treatment for drug dependence (in-treatment period versus off-treatment period).

We examined the following as potential sources of heterogeneity in CMRs or SMRs using random effects univariate meta-regressions in STATA: geographic region, country income group (based on World Bank categories), percentage of sample that injected drugs, were male or were HIV+ at baseline; presence of opioid users in the cohort; and the year in which the follow-up period ended.12,13

Results

We included 67 cohorts in the analysis; 14 were from low- and middle-income countries (Table 1). Studies from Europe, North America and Australasia were the most common; nine studies were from Asia and one was from South America. The pooled CMR across the 65 cohorts for which a CMR was provided was 2.35 deaths per 100 PY (Fig. 2). Cohorts from Asia had the highest pooled CMRs (5.25), followed by the cohorts from North America (2.64) and western Europe (2.31); cohorts from Australasia had the lowest pooled CMR (0.71).

Fig. 2. Crude mortality rates for people who inject drugs, by region
Fig. 2. Crude mortality rates for people who inject drugs, by region
Image produced using Stata (StataCorp. LP, College Station, TX, United States of America).CI, confidence interval; CMR, crude mortality rate.

SMRs were reported for 31 cohorts; their pooled SMR was 14.68 (Fig. 3). Since the heterogeneities (I2) of the pooled CMR and SMR were both very high (98.6% and 98.3%, respectively), we stratified estimates by subgroups.

Fig. 3. Standardized mortality ratios for people who inject drugs, by region
Fig. 3. Standardized mortality ratios for people who inject drugs, by region
Image produced using Stata (StataCorp. LP, College Station, TX, United States of America).CI, confidence interval; SMR, standardized mortality ratio.

Sex differences in mortality

Thirty-seven studies presented CMRs by sex.14,1719,2124,26,28,3032,34,3638,43,4550,52,54,57,58,62,6672,74 The pooled CMR ratio for males versus females was 1.32 (Fig. 4), which suggests that crude mortality was higher among males. Nineteen studies reported SMRs by sex;14,17,21,24,26,28,30,32,34,38,43,46,52,54,58,66,67,70,74 the pooled CMR ratio suggests that females had significantly greater excess mortality than males in similar age groups in the general population (Fig. 5). Only two of the nineteen studies presented SMRs for males that were greater than those for females.30,74

Fig. 4. Ratios of crude mortality rates in males versus females who inject drugs
Fig. 4. Ratios of crude mortality rates in males versus females who inject drugs
Image produced using Stata (StataCorp. LP, College Station, TX, United States of America).CI, confidence interval; RR, relative risk.
Fig. 5. Ratios of standardized mortality ratios for males versus females who inject drugs
Fig. 5. Ratios of standardized mortality ratios for males versus females who inject drugs
Image produced using Stata (StataCorp. LP, College Station, TX, United States of America).CI, confidence interval; RR, relative risk.

Causes of death

Several studies reported specific causes of death. The pooled CMR for death from drug overdose was 0.62 per 100 PY across 43 studies (Fig. 6). Eleven studies reported CMRs for death from drug overdose by sex: overall the CMR was 1.38 times higher (Fig. 7) among males than among females.14,17,19,21,24,32,38,49,52,57,58

Fig. 6. Crude mortality rates for death from drug overdose in people who inject drugs
Fig. 6. Crude mortality rates for death from drug overdose in people who inject drugs
Image produced using Stata (StataCorp. LP, College Station, TX, United States of America).CI, confidence interval; CMR, crude mortality rate.
Fig. 7. Ratios of crude mortality rates for death from drug overdose in males versus females who inject drugs
Fig. 7. Ratios of crude mortality rates for death from drug overdose in males versus females who inject drugs
Image produced using Stata (StataCorp. LP, College Station, TX, United States of America).CI, confidence interval, RR, relative risk.

In 20 studies CMRs were provided separately for people who inject drugs according to their HIV status.15,16,18,28,34,3640,45,48,49,55,57,60,65,70,72,74 All-cause mortality was three times higher among HIV+ than among HIV− subjects (CMR ratio: 3.15) (Fig. 8). Much of this elevated mortality appeared to result from AIDS deaths among HIV+ users of injecting drugs. The pooled estimate of AIDS-related mortality for the 16 studies for which data were available was 2.55 per 100 PY (Fig. 9).28,3336,3841,48,49,60,65,70,72,74 When we examined mortality from causes other than AIDS, we found it to be 1.63 times higher among HIV+ than among HIV− people who inject drugs (Fig. 10).28,34,36,3840,48,49,60,65,70,72,74

Fig. 8. Ratios of crude mortality rates in HIV-positive versus HIV-negative people who inject drugs
Fig. 8. Ratios of crude mortality rates in HIV-positive versus HIV-negative people who inject drugs
Image produced using Stata (StataCorp. LP, College Station, TX, United States of America).CI, confidence interval; HIV, human immunodeficiency virus; RR, relative risk.
Fig. 9. Crude mortality rates for AIDS-related deaths in people injecting drugs who were HIV-positive at baseline
Fig. 9. Crude mortality rates for AIDS-related deaths in people injecting drugs who were HIV-positive at baseline
Image produced using Stata (StataCorp. LP, College Station, TX, United States of America).AIDS, acquired immunodeficiency syndrome; CI, confidence interval; CMR, crude mortality rate; HIV, human immunodeficiency virus.
Fig. 10. Ratios of crude mortality rates for non-AIDS-related deaths in HIV-positive versus HIV-negative people who inject drugs
Fig. 10. Ratios of crude mortality rates for non-AIDS-related deaths in HIV-positive versus HIV-negative people who inject drugs
Image produced using Stata (StataCorp. LP, College Station, TX, United States of America).AIDS, acquired immunodeficiency syndrome; CI, confidence interval; HIV, human immunodeficiency virus; RR, relative risk.

Mortality from drug overdose was presented by HIV status in 9 studies.28,34,36,38,39,49,65,70,74 Pooled estimates showed mortality to be twice as high among HIV+ than among HIV− people who inject drugs (CMR ratio: 1.99) (Fig. 11). Further analyses across 13 studies conducted on HIV+ people who inject drugs showed no significant difference in deaths from drug overdose and from AIDS in this group (CMR ratio: 1.35; P = 0.554) (Fig. 12).28,3336,38,39,41,49,65,70,71,74

Fig. 11. Ratios of crude mortality rates for death from drug overdose in HIV-positive versus HIV-negative people who inject drugs
Fig. 11. Ratios of crude mortality rates for death from drug overdose in HIV-positive versus HIV-negative people who inject drugs
Image produced using Stata (StataCorp. LP, College Station, TX, United States of America).CI, confidence interval; HIV, human immunodeficiency virus; RR, relative risk.
Fig. 12. Ratios of crude mortality rates for AIDS-related death versus death from drug overdose in HIV-positive people who inject drugs
Fig. 12. Ratios of crude mortality rates for AIDS-related death versus death from drug overdose in HIV-positive people who inject drugs
Image produced using Stata (StataCorp. LP, College Station, TX, United States of America).AIDS, acquired immunodeficiency syndrome; CI, confidence interval; HIV, human immunodeficiency virus; RR, relative risk.

Only four studies presented data by sex and HIV status.37,47,49,74 They showed no significant difference in CMRs between HIV+ males and HIV+ females who inject drugs (CMR ratio: 1.13; Fig. 13), but HIV– males had a pooled CMR 1.81 times greater than that of HIV– females who inject drugs (Fig. 14).

Fig. 13. Ratios of crude mortality rates in HIV-positive males versus HIV-positive females who inject drugs
Fig. 13. Ratios of crude mortality rates in HIV-positive males versus HIV-positive females who inject drugs
Image produced using Stata (StataCorp. LP, College Station, TX, United States of America).CI, confidence interval; HIV, human immunodeficiency virus; RR, relative risk.
Fig. 14. Ratios of crude mortality rates in HIV-negative males versus HIV-negative females who inject drugs
Fig. 14. Ratios of crude mortality rates in HIV-negative males versus HIV-negative females who inject drugs
Image produced using Stata (StataCorp. LP, College Station, TX, United States of America).CI, confidence interval; HIV, human immunodeficiency virus; RR, relative risk.

Mortality by primary drug injected at baseline

Five studies estimated mortality by primary drug injected (opioids versus stimulants) (Table 2). Pooled estimates of all-cause mortality by primary type of drug injected showed no overall difference across studies (CMR ratio: 1.25; 95% CI: 0.60–2.61; P = 0.553).36,46,57,70,71 The same was true of studies of mortality resulting from drug overdose (CMR ratio: 1.85; 95% CI: 0.75–4.56; P = 0.18). In three of the four studies mortality associated with drug overdose was higher among people injecting opioids than among those injecting stimulants.36,46,71 In the fourth study, people who injected primarily stimulants had higher rates of drug overdose; however, the deaths from overdose in this group were later shown to have been caused by opioid use.57

Mortality according to treatment

Six studies provided information on mortality during in-treatment and off-treatment periods at follow-up: the meta-analysis suggested that mortality was 2.52 times higher during off-treatment periods than during in-treatment periods (Fig. 15).22,23,25,26,35,37

Fig. 15. Ratios of crude mortality rates in people who inject drugs during in-treatment period versus off-treatment period
Fig. 15. Ratios of crude mortality rates in people who inject drugs during in-treatment period versus off-treatment period
Image produced using Stata (StataCorp. LP, College Station, TX, United States of America).CI, confidence interval; RR, relative risk.

Heterogeneity in mortality

We performed univariate analyses to determine if the heterogeneity in overall CMRs and SMRs could be explained by participant characteristics and methodological variables. The results showed that high-income countries had lower CMRs than low- and middle-income countries (Fig. 16). Cohorts with greater proportions of males and HIV+ participants at baseline also had higher CMRs. Cohorts whose follow-up periods ended in more recent years had lower SMRs (Table 3). Study data were not sufficient to allow for multivariate analyses.

Fig. 16. Crude mortality rates for people who inject drugs, by country income group
Fig. 16. Crude mortality rates for people who inject drugs, by country income group
Image produced using Stata (StataCorp. LP, College Station, TX, United States of America).CI, confidence interval; CMR, crude mortality rate.

Discussion

Although previous reviews have examined mortality among people who inject drugs, to our knowledge this is the most comprehensive systematic review of the topic and the first to employ novel approaches to search for the available evidence. These approaches ranged from standard searches of the peer-reviewed literature to comprehensive searches of the non-peer reviewed literature and multiple expert consultations, as well as examination of both participant- and study-level factors potentially associated with the risk of death.

The pooled CMR of 2.35 deaths per 100 PY provides evidence of the high mortality associated with injecting drug use. The pooled SMR of 14.68 also shows that mortality is much higher in those who inject drugs than in the general population. Differences by sex were evident: across all studies that reported mortality by sex, males had higher CMRs, yet females who inject drugs had a much higher elevation in mortality relative to their age-matched peers in the general population than did males who inject drugs.

Most of the cohorts identified were from 14 high-income countries; together these 14 counties represent 78% of the total estimated population of people who inject drugs in such countries.76 Studies from only 11 low or middle income countries were identified; these countries account for only 40% of the estimated number of people injecting drugs in low- or middle-income countries.76

Although pooled CMRs were higher among people injecting drugs in low- and middle-income countries rather than high-income countries, we observed no significant difference in pooled SMRs. This suggests that the higher CMRs may reflect higher overall mortality in the general population in low- and middle-income countries than in high-income countries. The lowest and highest mortality rates were documented in cohorts in Australasia and Asia, respectively. Differences across high-income countries probably reflect differences in HIV infection prevalence, coverage of HIV prevention and coverage of opioid agonist maintenance treatment.76,77

Drug overdose and AIDS-related mortality were by far the most common causes of death. The pooled CMR for death from drug overdose was 0.62 per 100 PY, higher among males than females who inject drugs, and higher among HIV+ people who inject drugs than among those who were HIV−. In three of the four studies comparing drug overdose among people injecting opioids compared to those injecting stimulants, CMRs were higher among the former group, as expected.36,46,71 In a fourth study, however, drug overdose was higher among people who injected stimulants,57 although further investigation revealed that the deaths from overdose in this group were more often linked to the injection of opioids than to the injection of stimulants. This finding highlights the fact that people who inject drugs often use more than one drug type, even if they have a particular drug of choice.

The prevalence of HIV infection varied widely. As expected, overall mortality was much higher among HIV+ than among HIV− people who inject drugs (pooled CMR ratio: 3.15), but mortality from causes other than AIDS was also higher among those who were HIV+. Overdose-related mortality was also higher among HIV+ people who inject drugs in many cohorts. These differences in mortality may reflect differences in risky behaviour, physical health and social disadvantage.

The observational evidence examined in this review is consistent with the evidence from randomized controlled trials that opioid agonist maintenance treatment is associated with a reduced risk of death.78 Among cohorts for which in-treatment and off-treatment periods were carefully tracked, mortality rates were around 2.5 times higher in off-treatment periods than in in-treatment periods. Variation in exposure to treatment could also explain differences between cohorts in mortality from drug overdose, although this variation was not explicitly measured across cohorts.

The prevention of HIV transmission among people who inject drugs is clearly a public health priority.79,80 There is growing evidence that opioid agonist maintenance treatment, antiretroviral treatment and needle and syringe programmes reduce HIV transmission.8183 These interventions have been implemented in many countries, but often on a limited scale only.77 Clearly, however, AIDS is only one of several common causes of death in this group: a comprehensive approach to improving health outcomes among people who inject drugs must also include efforts to reduce other causes of death frequently found among them, particularly drug overdose.84

Limitations of the evidence

Evidence on mortality rates among users of injecting drugs is still predominantly from high-income countries, particularly in western Europe. Interestingly, however, this review has shown that despite marked differences in CMRs across countries, the extent to which this mortality exceeds that of the general population may show less pronounced differences. It would be inappropriate to assume that mortality is equally high among all people who inject drugs. New research in this area is needed, especially in countries where drug injecting is taking place but little research has been conducted about it.

In this review we found no significant differences in the risk of death by type of primary drug injected. This contrasts with the findings of other reviews of people dependent on different drug types, which, despite their own limitations, have suggested differences in mortality among opioid-, amphetamine- and cocaine-dependent persons.35 An explanation for this discrepancy might lie in the extent of drug injection among the groups examined, whether people used multiple drugs (polydrug use being the norm), or the possibility, seldom examined, that some people in the cohorts switched from one primary drug to another during the follow-up period. All of these factors would have reduced our capacity to detect any differences in mortality among people injecting different types of drugs.

The ability to detect differences in mortality in cohort studies according to HIV status is subject to limitations. HIV status was typically measured at baseline only, and some subjects who contracted HIV infection during follow-up would remain assigned to the HIV− group for the entire follow-up period. Nonetheless, this would only serve to underestimate the relative differences in mortality between HIV+ and HIV− people who inject drugs. The markedly higher all-cause mortality that we observed among HIV+ people who inject drugs is therefore probably a conservative estimate of the elevation in mortality in that group. Misattribution of cause of death as either AIDS- or non-AIDS-related could have occurred as well.

Treatment for HIV infection has improved greatly and has become more widely available. In some cohorts, mortality was examined for the periods before and after highly active antiretroviral therapy (HAART) was introduced. The findings suggest that mortality among HIV+ people who inject drugs decreased after the widespread introduction of HAART.55 Unfortunately, we were unable to examine the impact of treatment for HIV infection across studies because mortality was rarely reported separately for the periods before and after the introduction of HAART.77 However, the observed association between cohorts with more recent follow-up periods and lower SMRs might have to do with the greater availability in recent years of effective interventions for the prevention and treatment of HIV infection.

Reporting quality was poor. Few studies met criteria in consensus statements for the reporting of observational studies.7 Mortality estimates were reported in various forms, including odds ratios, relative risks, hazard ratios and CMRs. Most studies did not report SMRs and many failed to report standard parameters such as PY, or were seldom easy to calculate, particularly for disaggregated mortality estimates. As a result, only a subset of studies could be included in many of the analyses.

Causes of death were not uniformly or consistently coded. Deaths from drug overdose might have been missed in countries with limited capacity to conduct toxicological tests or where recording a death as being from a drug overdose is surrounded by stigma. As a result, we may have underestimated CMRs and SMRs for death from drug overdose. Misattribution of deaths by HIV status may have occurred, since most cohorts were assessed for HIV status at the beginning of the study only and people infected during follow-up could have been missed. Again, this may have resulted in conservative estimates of mortality among HIV+ people who inject drugs and in lower effect sizes. In future research, assessing individuals’ HIV status at several time points during the follow-up period would allow a more accurate measurement of mortality in relation to HIV status.

Limitations of the review and meta-analysis

Our review has limitations. The lag time between the date when the studies were conducted and when they were published in peer-reviewed journals was generally long. In light of this we used several methods to search for published and unpublished studies. We reviewed primarily English-language papers, although we also reviewed the abstracts of non-English-language peer-reviewed articles when they were available in English. When studies seemed relevant, we had them translated; we engaged experts from a range of different countries and language groups to review these reference lists. Meta-analytical methods were originally developed to aggregate the findings of randomized controlled trials,85 which have the advantage of allowing for control or adjustment of pre-conditions and sample-related factors that could influence the outcomes of interest. Controlling for such factors is not possible in observational studies, like the ones included in our review.

Conclusion

People who inject drugs have a much higher risk of death than those who do not. Major causes of death in this group are often poorly specified, but death from drug overdose is common, as is AIDS-related mortality in settings with a high prevalence of HIV infection. HIV+ people who inject drugs have higher mortality not just from HIV-related causes but also from drug overdose. Mortality varies by participant- and study-level characteristics, which suggests that multiple factors contribute to the higher risk of death observed in people who inject drugs. Many of these factors are probably modifiable, since certain predominant causes of death account for most of the mortality observed in this group.


Competing interests:

None declared.

References

Share