High Mortality Among Non–HIV-Infected People Who Inject Drugs in Bangkok, Thailand, 2005–2012

Abstract

Objectives. We examined the causes of hospitalization and death of people who inject drugs participating in the Bangkok Tenofovir Study, an HIV preexposure prophylaxis trial.

Methods. The Bangkok Tenofovir Study was a randomized, double-blind, placebo-controlled trial conducted during 2005 to 2012 among 2413 people who inject drugs. We reviewed medical records to define the causes of hospitalization and death, examined participant characteristics and risk behaviors to determine predictors of death, and compared the participant mortality rate with the rate of the general population of Bangkok, Thailand.

Results. Participants were followed an average of 4 years; 107 died: 22 (20.6%) from overdose, 13 (12.2%) from traffic accidents, and 12 (11.2%) from sepsis. In multivariable analysis, older age (40–59 years; P = .001), injecting drugs (P = .03), and injecting midazolam (P < .001) were associated with death. The standardized mortality ratio was 2.9.

Conclusions. People who injected drugs were nearly 3 times as likely to die as were those in the general population of Bangkok and injecting midazolam was independently associated with death. Drug overdose and traffic accidents were the most common causes of death, and their prevention should be public health priorities.

People who inject drugs (PWID) are at higher risk for death than are people of similar age and gender who do not inject drugs.^1,2 Studies in Europe and North America have demonstrated high mortality rates and standardized mortality ratios among cohorts of PWID.^3–8 Trauma, bacterial and HIV infection, suicide, and drug overdose account for much of the increased mortality risk in these studies.^9–13 There have been, however, few reports describing the mortality and morbidity of PWID in Southeast Asia and Thailand.^14,15

The Bangkok Tenofovir Study was a phase-3, randomized, double-blind, placebo-controlled, endpoint-driven, HIV preexposure prophylaxis trial with variable participant follow-up time conducted from June 2005 to July 2012 that demonstrated that a daily oral dose of tenofovir disoproxil fumarate (tenofovir) can reduce HIV transmission among PWID by 49%.¹⁶ We have summarized the mortality and morbidity data of 2413 non–HIV-infected PWID who participated in the study. We examined participant demographic characteristics and risk behavior to determine predictors of all-cause mortality and calculated the standardized mortality ratio compared with that of the general population of Bangkok, Thailand.

METHODS

Descriptions of community engagement and enrollment and the safety and efficacy results of the Bangkok Tenofovir Study have been published.^16–18 The study was conducted at 17 Bangkok Metropolitan Administration drug treatment clinics in densely populated urban communities of Bangkok. The clinics offer a range of services, including HIV counseling and testing, risk-reduction counseling, social and welfare services, health education, medical care and referrals, methadone treatment, condoms, and bleach to clean injection equipment, with demonstrations of appropriate use. These services are provided free of charge. Thailand’s narcotics law prohibits the distribution of needles to inject illicit drugs, and needles are not provided in the clinics¹⁹; however, sterile needles and syringes are available to the public over the counter at low cost (5–10 baht, which equals US $0.12–US $0.25) in pharmacies in Bangkok.

Non–HIV-infected individuals aged 20 to 60 years who reported injecting drugs during the previous year were candidates for the study. We randomly assigned participants in a 1:1 ratio to receive 300 milligrams of oral tenofovir daily or placebo.

Participants completed a questionnaire assessing injecting drug use, needle sharing, sexual activity, and incarceration during the previous 3 months at enrollment and every 3 months thereafter using an audio computer-assisted self-interview. At enrollment and monthly (every 28 days) visits, we assessed participants for adverse events, provided individualized risk-reduction counseling, and tested oral fluid for HIV antibodies (OraSure Technologies, Inc., Bethlehem, PA). We referred newly infected individuals for care according to national guidelines.²⁰ At enrollment, months 1, 2, 3, and every 3 months thereafter, we collected blood for hematologic, hepatic, and renal safety assessment.

Research staff telephoned participants before monthly visits to encourage them to attend the visit and contacted participants who missed visits by telephone or with a home visit if they could not be reached by telephone. Staff reviewed medical records of hospitalized participants to abstract data on the cause of hospitalization. We used autopsy reports (n = 42) or death certificates (n = 65) if an autopsy was not performed to define the cause of death, and we checked the Thailand Public Health Statistics database²¹ every 6 months during the trial and for 1 year after the trial ended to see if participants who were lost to follow-up had died.

We used the Kaplan–Meier product-limit method to estimate participant survival. We censored data when participants died or at their final study visit. We compared the mortality rate of participants in the study with the rate in the general population of Bangkok during 2008 to 2009 using indirect standardization.^21–23 We calculated death rates per 1000 person-years of observation and exact 95% Poisson confidence intervals (CIs) for the death rates and the standardized mortality ratios.²⁴ We used a time-varying covariate proportional hazards model to evaluate baseline demographic characteristics and risk behaviors reported each 3 months during follow-up as predictors of death.²⁵

Time-varying variables included injecting drugs, sharing needles, incarceration, sexual activity, and participating in a methadone program. We included data from participants who died within 3 months of their final study visit; we censored data from participants who died more than 3 months after their final visit at the time of their last visit. We evaluated variables that were associated with death in bivariable analysis (P ≤ .1) using a backward stepwise multivariable model. We included treatment group in all models to control for the randomization of participants to tenofovir or placebo. We used SAS version 9.3 (SAS Institute, Cary, NC) for statistical analyses.

RESULTS

We described study results in previous publications.^16–18 Briefly, from June 2005 through July 2010, we evaluated 4094 people for enrollment; 2413 (59.0%) were eligible and enrolled. We followed enrolled participants for an average of 4.0 years (maximum 6.9 years), and participants contributed 9786 person-years of follow-up time. The median age of participants was 31 years (range = 20–59), and 1924 (79.7%) were male. A total of 107 participants died during the trial, 5 (4.7%) of whom we identified in the Public Health Statistics database. The 7-year survival rate using Kaplan–Meier analysis was 87.9% (95% CI = 80.3%, 96.1%; Figure 1).

FIGURE 1—

Kaplan–Meier product-limit survival estimate for injection drug users (n = 2413): Bangkok Tenofovir Study; Bangkok, Thailand; 2005–2012.

At enrollment, 610 (25.3%) participants reported they had been incarcerated in the previous 3 months, 552 (22.9%) in jail, 389 (16.1%) in prison, and 331 (13.7%) in both. Among participants who reported injecting at baseline, 801 (53.2%) injected methamphetamine, 559 (37.1%) midazolam, and 527 (35.0%) heroin. Fifty participants became infected with HIV during follow-up: 17 in the tenofovir group and 33 in the placebo group, indicating a 48.9% reduction in the HIV incidence (95% CI = 9.6, 72.2; P = .01) among participants randomized to tenofovir.¹⁶

The frequency of deaths, serious adverse events, and grades 3 and 4 laboratory results were similar among the tenofovir and placebo groups.¹⁶ During the trial, 107 participants died: 22 (20.6%) from drug overdose, 13 (12.2%) from traffic accidents, and 12 (11.2%) from sepsis (Table 1). Among the 13 participants who died in traffic accidents, 8 died of head injuries sustained in motorcycle accidents, 3 of other injuries from motorcycle accidents, 1 from a car accident, and 1 from being hit by a train. Three of the 13 had blood or urine examined for drugs and alcohol: 2 had alcohol in their blood, and 1 had methamphetamine in his urine. Nine participants died of cancer: 3 of cancer of the liver, 3 of lung cancer, 1 of rectal cancer, 1 of bone cancer, and 1 of brain cancer.

TABLE 1—

Causes of Death and Hospitalization of Participants (n = 2413): Bangkok Tenofovir Study; Bangkok, Thailand; 2005–2012.

Variable	No. (%)	Rate per 1000 person-years (95% CI)
Cause of death
All causes	107 (100.0)	10.9 (9.0, 13.2)
Drug overdose	22 (20.6)	2.3 (1.4, 3.4)
Suspected drug overdose	3 (2.8)	0.3 (0.1, 0.9)
Traffic accident	13 (12.2)	1.3 (0.7, 2.3)
Sepsis	12 (11.2)	1.2 (0.6, 2.1)
Cancer	9 (8.4)	0.9 (0.4, 1.8)
Respiratory or circulatory failure	8 (7.5)	0.8 (0.4, 1.6)
Intracranial hemorrhage	8 (7.5)	0.8 (0.4, 1.6)
Pneumonia	6 (5.6)	0.6 (0.2, 1.3)
Myocardial infarction	3 (2.8)	0.3 (0.1, 0.9)
Assault	3 (2.8)	0.3 (0.1, 0.9)
Drowning	3 (2.8)	0.3 (0.1, 0.9)
Suicide	3 (2.8)	0.3 (0.1, 0.9)
Arrhythmia	2 (1.9)	0.2 (0.1, 0.7)
Peritonitis	2 (1.9)	0.2 (0.1, 0.7)
Upper gastrointestinal bleed	2 (1.9)	0.2 (0.1, 0.7)
Other	8 (7.5)	0.8 (0.4, 1.6)
Cause of hospitalization
All hospitalizations	525 (100.0)	53.7 (49.2, 58.4)
Traffic accident	114 (21.7)	11.7 (9.6, 14.0)
Infection	104 (19.8)	10.6 (8.7, 12.9)
Drug use	78 (14.9)	8.0 (6.3, 10.0)
Assault	54 (10.3)	5.5 (4.2, 7.2)
Gastrointestinal	24 (4.6)	2.5 (1.6, 3.7)
Psychiatric	16 (3.1)	1.6 (0.9, 2.7)
Cancer	14 (2.7)	1.4 (0.8, 2.4)
Heart disease	14 (2.7)	1.4 (0.8, 2.4)
Other	107 (20.4)	10.9 (9.0, 13.2)

Note. CI = confidence interval.

During the study, 407 participants were hospitalized 525 times (range = 1–5 hospitalizations). The most common causes of hospitalizations were traffic accidents (n = 114; 21.7%), infections (n = 104; 19.8%), drug-related causes (n = 78; 14.9%), and assaults (n = 54; 10.3%; Table 1). The most common grade 3 or 4 laboratory test results were aspartate aminotransferase (182 [7.5%] participants), amylase (153 [6.3%] participants), and alanine aminotransferase (144 [6.0%] participants; Table 2). Grade 4 laboratory events were reported as serious adverse events, and alcohol was recorded as a contributing cause in 75 of 112 (67.0%) grade 4 aspartate aminotransferase, alanine aminotransferase, and amylase results; participants could have more than 1 serious adverse event.

TABLE 2—

Grades 3 and 4 Laboratory Results Reported in 12 (0.5%) or More Participants (n = 2413): Bangkok Tenofovir Study; Bangkok, Thailand; 2005–2012

	No. (%)
Lab Result	Grade 3	Grade 4	Grade 3 or 4
AST	159 (6.6)	64 (2.7)	182 (7.5)
Amylase	150 (6.2)	6 (0.3)	153 (6.3)
ALT	127 (5.3)	38 (1.6)	144 (6.0)
Bilirubin	25 (1.0)	6 (0.3)	31 (1.3)
Phosphorus	21 (0.9)	0 (0.0)	21 (0.9)
Platelet	14 (0.6)	1 (0.1)	14 (0.6)
Hemoglobin	12 (0.5)	3 (0.1)	13 (0.5)

Note. ALT = alanine aminotransferase; AST = aspartate aminotransferase.

The all-cause mortality rate of participants in the Bangkok Tenofovir Study was 10.9 per 1000 person-years (95% CI = 9.0, 13.2). Multivariable analysis showed that participants aged 40 to 59 years (hazard ratio [HR] = 2.5; 95% CI = 1.4, 4.3) were more likely to die than were participants aged 20 to 29 years and that participants who reported injecting drugs (HR = 2.4; 95% CI = 1.1, 5.4) and, controlling for injecting, those who injected midazolam (HR = 3.6; 95% CI = 1.8, 7.1) were more likely to die than were those who did not (Table 3). Participants who reported sex with a live-in partner were less likely to die (HR = 0.6; 95% CI = 0.4, 1.0) than were those who did not. Among participants who died of overdose or suspected overdose, 20 (80.0%) reported injecting midazolam during the study, 13 (52.0%) injected heroin, and 7 (28.0%) injected methamphetamine.

TABLE 3—

Results of Bivariable and Multivariable Analysis of Baseline Demographic Characteristics and Risk Factors as Predictors of Death Among Injection Drug Users (n = 2413): Bangkok Tenofovir Study; Bangkok, Thailand; 2005–2012

				Bivariable Analysis		Multivariable Analysis
Variable	Deaths	Person-Years	Deaths per 1000 Person-Years (95% CI)	HR (95% CI)	P	HR (95% CI)	P
Placebo	52	4207	12.4 (9.2, 16.2)	1.0 (Ref)		1.0 (Ref)
Tenofovir	42	4183	10.0 (7.2, 13.6)	0.8 (0.5, 1.2)	.308	0.9 (0.6, 1.3)	.506
Gender
Female	11	1745	6.3 (3.2, 11.3)	1.0 (Ref)		1.0 (Ref)
Male	83	6646	12.5 (10.0, 15.5)	2.0 (1.1, 3.7)	.035	1.2 (0.6, 2.5)	.617
Age at enrollment, y
20–29	22	3250	6.8 (4.2, 10.3)	1.0 (Ref)		1.0 (Ref)
30–39	37	3162	11.7 (8.2, 16.1)	1.8 (1.0, 3.0)	.038	1.4 (0.8, 2.4)	.188
40–59	35	1979	17.7 (12.3, 24.6)	2.6 (1.6, 4.5)	< .001	2.5 (1.4, 4.3)	.001
Education
≤ grade 6	36	4112	8.8 (6.1, 12.1)	1.0 (Ref)		1.0 (Ref)
> grade 6	58	4279	13.6 (10.3, 17.5)	1.6 (1.0, 2.4)	.035	1.2 (0.8, 1.9)	.353
In jail in past 3 mo
No	84	7444	11.3 (9.0, 14.0)	1.0 (Ref)
Yes	10	948	10.6 (5.1, 19.4)	0.9 (0.5, 1.7)	.755	. . .
In prison in past 3 mo
No	86	7379	11.7 (9.3, 14.4)	1.0
Yes	8	1013	7.9 (3.4, 15.6)	0.7 (0.3, 1.4)	.299	. . .
Injected drugs in past 3 mo
No	33	6700	4.9 (3.4, 6.9)	1.0 (Ref)		1.0 (Ref)
Yes	61	1692	36.1 (27.6, 46.3)	7.2 (4.7, 11.0)	< .001	2.4 (1.1, 5.4)	.028
Shared needles in past 3 mo
No	88	8209	10.7 (8.6, 13.2)	1.0 (Ref)		1.0 (Ref)
Yes	6	183	32.9 (12.1, 71.6)	2.4 (1.0, 5.7)	.045	1.0 (0.4, 2.2)	.906
In methadone program
No	67	6407	10.5 (8.1, 13.3)	1.0 (Ref)
Yes	27	1985	13.6 (9.0, 19.8)	1.3 (0.8, 2.1)	.229	. . .
Injected heroin in past 3 mo
No	69	7808	8.8 (6.9, 11.2)	1.0 (Ref)		1.0 (Ref)
Yes	25	584	42.8 (27.7, 63.3)	4.5 (2.8, 7.2)	< .001	1.4 (0.8, 2.3)	.211
Injected methamphetamine in past 3 mo
No	79	7768	10.2 (8.1, 12.7)	1.0 (Ref)		1.0 (Ref)
Yes	15	624	24.0 (13.5, 39.7)	2.1 (1.2, 3.6)	.014	0.7 (0.4, 1.2)	.167
Injected midazolam in past 3 mo
No	43	7383	5.8 (4.2, 7.9)	1.0 (Ref)		1.0 (Ref)
Yes	51	1009	50.6 (37.6, 66.5)	8.5 (5.7, 12.8)	< .001	3.6 (1.8, 7.1)	< .001
Had sex in past 3 mo
No	57	3790	15.0 (11.4, 19.5)	1.0 (Ref)		1.0 (Ref)
Yes	37	4602	8.0 (5.7, 11.1)	0.5 (0.3, 0.8)	.002	0.8 (0.5, 1.3)	.324
Had sex with a live-in partner in past 3 mo
No	66	4233	15.6 (12.1, 19.8)	1.0 (Ref)		1.0 (Ref)
Yes	28	4158	6.7 (4.5, 9.7)	0.4 (0.3, 0.7)	< .001	0.6 (0.4, 1.0)	.041
Had sex with a casual partner in past 3 mo
No	75	6713	11.2 (8.8, 14.0)	1.0 (Ref)
Yes	19	1679	11.3 (6.8, 17.7)	1.0 (0.6, 1.6)	.897	. . .

Note. CI = confidence interval; HR = hazard ratio. The treatment group was included in bivariable and multivariable analyses to control for the randomization of participants to tenofovir or placebo. We analyzed data from participants who died within 3 months of their final study visit (n = 94).

The standardized mortality ratio of participants was 2.9 (95% CI = 2.4, 3.6) compared with the mortality rate of the general population of Bangkok, controlling for gender and age group with indirect standardization.^21–23 Consistent with the findings of the multivariable analysis, the standardized mortality ratio comparing women in the study with women in the general population of Bangkok (3.4; 95% CI = 1.7, 6.1) was similar to the ratio of men in the study compared with the general population of men in Bangkok (2.9; 95% CI = 2.3, 3.5).

DISCUSSION

The 7-year Kaplan–Meier survival estimate of participants in this cohort of non–HIV-infected PWID was 87.9%; approximately 1 in 8 participants died during the study. Participants were nearly 3 times as likely to die as were those in the general population of Bangkok. The death rate was higher in men than in women but did not differ significantly in multivariable analysis.

Similar to other studies among PWID, drug overdose, traffic accidents, and sepsis were the most common causes of death.^9–12 Eleven of the 13 participants who died in traffic accidents died from motorcycle accidents, and most of these were because of head injuries. Traffic accidents, infections, and drug-related causes were the most common causes of hospitalization. Traffic accidents cause a substantial burden of injury and death in Thailand.^26,27 The World Health Organization reports that an estimated 38.1 per 100 000 Thais died in traffic accidents in 2010, the third highest death rate among the 182 countries analyzed.²⁸ Consistent with our findings among PWID, most (74%) traffic accident deaths in Thailand were because of motorcycle accidents.

Controlling for demographic characteristics and injection practices, injecting midazolam—a benzodiazepine with sedative, anxiolytic, and amnestic properties that can cause respiratory depression and arrest²⁹—was independently associated with death. Midazolam use was reported by 80.0% of participants who died of overdose, suggesting that the midazolam that was injected may have been particularly potent, mixed with other drugs, or used in combination with alcohol or other sedative hypnotics, or that the amnestic actions may have led to impaired judgment and overdosing.

Participants reporting sex with a live-in partner were less likely to die than were those who did not report sex with a live-in partner. A previous study among PWID in the same clinics³⁰ found that participants who reported having sex were less likely to become HIV-infected than were those who did not report having sex, suggesting that PWID who have a sexual partner, particularly a live-in partner, are less likely to engage in high-risk behaviors that lead to death or HIV infection.

The mortality rate in this cohort of PWID in Bangkok (10.9 per 1000 person-years; 95% CI = 9.0, 13.2) is lower than was that found in a cohort of non–HIV-infected PWID admitted for treatment of opiate or amphetamine dependence at a drug treatment center in Chiang Mai, Thailand, in 1999 to 2000 (38.5 per 1000 person-years; 95% CI = 24.2, 58.3)¹⁴ and a cohort of 690 non–HIV-infected male PWID in Northern Vietnam during 2005 to 2007 (41.1 per 1000 person-years; 95% CI = 26.1, 61.7).¹⁵ This suggests that health and harm reduction services may have been more broadly available to PWID in Bangkok in 2005 to 2012 than to PWID in Chiang Mai and Northern Vietnam when those studies were done. The mortality rate in the Bangkok cohort was similar to rates reported in studies among PWID in Australia (8.9 per 1000 person-years; 95% CI = 8.6, 9.2),⁶ Italy (7.4 per 1000 person-years; 95% CI = 5.9, 8.8),⁹ and the United Kingdom (10.0 per 1000 person-years; 95% CI = 8.6, 11.5).³¹

Our analysis has several limitations. A substantial proportion of morbidity and mortality among PWID stems from HIV and HCV infection.^11,32,33 We conducted the study in a cohort of non–HIV-infected PWID. Thus, our results are limited to non–HIV-related causes of death and illness. We did not test for HCV infection, but HCV antibody testing of 1209 PWID participating in a 1995 to 1999 preparatory cohort study in the same clinics found an HCV prevalence of 95.6%.³⁰ We suspect that a substantial proportion of participants were infected with HCV and that many of the grade 3 and 4 aminotransferase elevations and possibly the 3 liver cancer deaths were caused by HCV infection.

Participants in our study were willing to come to clinics monthly, were willing to report injection practices, and were able to meet study entry criteria. Their death and hospitalization rates may differ from PWID not in the study, limiting the generalizability of the results. However, using respondent-driven sampling, investigators estimated there were 3595 people injecting drugs in Bangkok in 2004,³⁴ and 4200 in 2009,³⁵ suggesting a large proportion of PWID enrolled in the study allowing generalization of results to PWID in Bangkok.

Among 2413 PWID followed for an average of 4 years, there were 107 deaths, 3 times the expected number of deaths on the basis of data from the general population of Bangkok. The most common causes of death were drug overdose, traffic accidents (i.e., motorcycle accidents), and sepsis; traffic accidents, infections, and drug-related causes were the most common causes of hospitalization. Public health programs that aim to prevent drug overdose, promote safe injection practices,³⁶ and promote road safety, particularly the use of motorcycle helmets, may help reduce mortality and morbidity among PWID in Bangkok.