This prospective follow-up study includes subjects from a database of criminal justice clients with substance use problems. The subjects were interviewed with a standardized and structured interview instrument, the Addiction Severity Index (ASI) (29). The ASI assessment is used on people with a known or suspected misuse of drugs or alcohol, with the aim to evaluate alcohol- and drug-related problems and transferring the clients to appropriate treatment units within the prison and probation system (28, 30). The interview examines different areas such as substance use, including alcohol and tobacco use, medical and psychiatric illness history, as well as social and legal problems (29, 31).

The ASI is a well-documented instrument, developed by researchers in The United States in the seventies and now used globally (29). It was first introduced in the Swedish Prison and Probation Service in 2001 and has been in increasing use since then (31). The present ASI data used in this study has been the subject of several other longitudinal studies (28, 30, 32–34). The ASI version used in this setting, ASI-X (35), is a developed version of the European standard version of the instrument, EuropASI (36).

The interviews were performed in the Swedish Prison and Probation Service between January 2001 and August 2006. The ASI is voluntary, and approximately 6% of the clients addressed for an ASI interview in this data set refused to participate (33). Only Swedish-speaking clients were examined with the ASI. Fifty individuals were interviewed before intake to the criminal justice facility with a median of 10 days before intake. The remaining 2,372 clients had a mean number of days between intake to the time when the interview was performed of 55 days (median 27 days; 98% were interviewed within the first year) (32). A total of 2,608 participants (82%) were interviewed in custody or prison [median time of incarceration 4.8 months (range 0–166.5 months)].

The inclusion criteria for this study were self-reported regular injection use with amphetamine, heroin, or polysubstance use as the main problem. Regular injection use was defined as any injection drug use for at least six months before admission to the criminal justice facility. The main drug was defined as the substance that the interviewed subject reported as “the major problem” (37). Polysubstance use was defined as the reporting that the major problem was two or more different drugs, again during at least six months prior to admission, of which at least one drug was an illicit substance (and again including the fulfilment of the injection drug use criterion of the study).

A total of 7,085 individuals were interviewed during the study period. Fifty subjects were excluded due to misinterpretation of the questions, inability to understand, refusing the interview, inability to conduct the interview, or interrupted the interview. A total of 4,588 individuals were excluded due to not fulfilling the inclusion criteria of at least 6 months of injection use with predominant (main drug) use of heroin, amphetamine, or polysubstance use. The cohort then consisted of 2,447 clients. For three subjects the interview instrument used was the Adolescent Drug Abuse Diagnosis (ADAD) (38) and not a full ASI interview, and these clients were excluded. Eight individuals lacked information regarding age, these were also excluded. Lastly, 14 individuals had an event occurring before the ASI-interview was performed, and these were excluded. The study was finally based on 2,422 clients (Figure 1).

The independent variables were retrieved from the ASI database. The main variables were self-reported injection drug (predominant use of amphetamine, heroin, or polysubstance use). Amphetamine was set as the reference variable. The control variables were sex (female, male), age, tobacco use (current tobacco use or not), whether the interview took place in incarceration (custody/prison vs. not), and time in prison or custody.

Follow-up data on cardiovascular and cerebrovascular outcomes were collected from two registers: The National Patient Register (NPR) and The Causes of Death Register (CDR), held by The Swedish National Board of Health and Welfare. The NPR provides the primary and secondary diagnoses for hospital-based patients as well as patients in the specialist out care. The diagnoses are set by physicians according to the International Classification of Diseases 10 (ICD-10). The register covers 99% of all somatic and psychiatric hospital discharges and 80% of all specialized outpatient care, primary health care is not included (39). The lower coverage of the outpatient care is probably due to missing data from private caregivers and some psychiatric outpatient treatment units (40).

Underlying causes of death, and death dates, were collected from the CDR. The causes of death in the register are stated according to ICD-10 and are based on death certificates set by physicians. Deaths amongst people with known or suspected drug or alcohol use disorders, shall according to Swedish legislation, routinely be investigated through forensic autopsy, and the cause of death is set by the physician performing the examination. The CDR covers more than 99% of deaths occurring in Sweden and includes all deceased individuals that at the time of death were registered in The Swedish Population Register, between the years 1961 and 2011. From 2012 the register also includes those not registered in the Swedish Population Register (41).

The outcome variables were: (1) time to the first CVD event, and (2) time to the first CBD event. An event was defined as the first occurrence of the diagnoses ICD-codes I20–I25 (CVD events) or the diagnoses ICD-codes I60–I69 (CBD events) in the NPR or the CDR (Table 1). Both primary and secondary diagnoses were considered in the NPR as well as the underlying causes of death in the CDR.

The Prison and Probation Service and the Swedish National Board of Health and Welfare linked the data in the ASI database with the diagnosis and causes of death data from 2001 to 2014 through national identification numbers. The endpoint of the observational time was either cardiovascular or cerebrovascular diagnosis or death, or 31 December 2014, whichever occurred first.

Initially, bivariate analyses were performed. Dichotomous variables are presented as absolute and relative frequency and comparisons between groups were calculated with Fisher's Exact test. Differences in median age were assessed using the Mann Whitney U test. We used extended Cox regression models to study time to the first CVD and CBD event during the observational time. As noted earlier, an event is defined as either morbidity or mortality resulting from CVD or CBD, respectively. The presented hazard ratios (HR) from the Cox regression therefore reflect both morbidity and mortality associated with CVD and CBD. Age of the study participants was used as the time scale, as it's considered preferable to use time-on-study with age as a covariate in observational studies when the influence of age is not of primary interest (42). Incarceration (including time in custody) was set as a time-varying variable due to the fact that the follow-up data includes both time in prison and the time after release from prison, until the first occurring event. In the unadjusted model we tested every variable separately and in the adjusted model all variables were controlled for one another as they were entered simultaneously into the model.

We used the Huber Sandwich Estimator to obtain robust estimates of the standard errors in order to account for the multiple observations of some study participants. Variance inflation factors were calculated and found to be below 1.1 for every variable studied.

In order to obtain estimates for all pairwise comparisons of main drug in the Cox regression models, we performed a contrasts analysis using the multcomp package in R, with p values being adjusted using the single-step method (43). Statistical analyses were performed in R.3.3.2 (44) and in IBM Statistical Package for Social Sciences (SPSS) versions 25 and 29.

Due to the study design and following the approval by the Lund Regional Ethics Board (file number 2014/478), active informed consent was not obtained. The ASI interviews were conducted routinely as part of the efforts of the criminal justice service. The ASI material did not contain any personal identifiers such as names or social security numbers when accessed by the researchers. Both our internal assessment and the evaluation conducted by the Ethics Review Board indicated that the research entailed a minimal likelihood of harm to the research subjects. Consequently, it was determined that written approval was deemed not necessary in retrospect. However, the planned study was announced in a free-of-charge newspaper in the three major cities in Sweden (Stockholm, Gothenburg, and Malmö), to give the participants an opportunity to decline contributing to the study. No individuals chose to opt-out.

The studied cohort consisted of 2,422 subjects, 339 were women {14%, median age 36 [interquartile range (IQR) 29–43 years]}, 2,083 were men [86%, median age 36 (IQR 28–43 years)] (Table 2). Self-reported predominant (main drug) use in the selected cohort was amphetamine in 51.5% (n = 1,247), polysubstance use in 33% (n = 799) and heroin in 15.5% (n = 376).

Total observational time for the entire cohort was 23,911 person-years [median 10.3 years (IQR 9.3–11.2 years)]. Total observational time for the outcome variable CVD event was 23,679 person-years [median 10.3 years (IQR 9.2–11.2 years)] and 23,767 person-years [median 10.3 years (IQR 9.2–11.2 years)] for CBD event.

Bivariate analyses showed a significantly higher percentage of predominant amphetamine users experiencing a CVD event compared to other substance users, 3% compared to 1.7% (p < 0.044) (Table 2). There was a significantly lower proportion women experiencing a CVD event, 0.6%, compared to the proportion men experiencing a CVD event, 2.6% (p = 0.019). The median age was significantly higher amongst those who experienced an event (CVD or CBD) than those who did not experience an event, 44 (IQR 39–50) and 45 (IQR 38–50) years respectively, compared to 36 years (IQR 28–43) (p < 0.001).

There were no significant differences in proportions of CVD or CBD events between tobacco users and non-users and individuals serving time in prison or custody and those who did not (Table 2).

The total number of CVD events was 57, of those were 9 events CVD-related deaths. The total number of CBD events was 41 and 9 events of those were CBD-related deaths. Person-time incidence rate of first episode of CVD event was 2.41 per 1,000 person-years for the entire study population. Person-time incidence rate of first episode of CBD events was 1.73 per 1,000 person-years for the total sample. The highest incidence rates of both CVD and CBD events were found among subjects reporting amphetamine as predominant (main drug) use (2.99 and 2.09 respectively).

In the unadjusted model, female sex was found to be associated with lower risk of CVD event (HR 0.21 [95% CI 0.05–0.87], p = 0.032 (Table 3). No other variable was significantly associated with CVD or CBD event. In the adjusted analysis, none of the outcome variables CVD or CBD events were significantly associated with any of the predominant (main) drugs, neither in the comparison of amphetamine with heroin or polysubstance drug use, nor in the comparison between heroin and polysubstance use. Female sex was significantly associated with lower risk of a CVD event [aHR 0.21 (95% CI 0.05–0.87), p = 0.032].

The ASI database used in this study is based on self-reports which implies a risk of recall bias. Self-reports in this setting have, however, been shown to have high reliability (54, 55). Furthermore, our knowledge is limited to the self-reported main drug at baseline, and the study participants may have changed their main drug during the course of the study. Moreover, this study includes injecting substance users in the criminal justice system and the result should not be generalized to substance users with other routes of administration. Also, this study was performed in a criminal justice setting and the result should therefore not be generalized to other settings of substance users. Importantly, the ASI database used here has been shown to include a certain overrepresentation of women and clients sentenced with the types of crimes which are more typically associated with the drug use itself (56). Also, the present sample includes a high proportion of clients with predominant use of amphetamine. However, although data from the full general population are lacking and cannot describe the full picture of injection drug use in the whole population, our predominance of clients using amphetamine is consistent with other more clinical samples from the present setting, such as in data from needle exchange programs (2). In the latter study, past-12-month drug use was amphetamine in 43% of cases (compared to 37% for heroin), and in a needle exchange sample from a different region, amphetamine was the primary drug reported by 56% of clients and heroin by 42% (18, 57). In the database used for the present study, 52% of clients in the whole criminal justice dataset had a history of amphetamine use, compared to 18% for heroin, and 30% of clients reporting a primary drug reported that to be amphetamine, compared to 9% for heroin (56, 58). Thus, the high proportion of predominant amphetamine use in the present study can be considered to reflect the clinical picture of illicit drug use in the present geographical setting.

Moreover, 20% of all specialized outpatient care as well as the primary health care is not covered by NPR (39) and is thereby not represented in the study. The IPR have an overall validity of 85%–95% when compared to inpatient records and the proportion of valid diagnoses is probably higher in patients with severe diseases in contrast to mild or more chronic diseases like hypertension and lipid disorders (39). However, in this study we mainly included acute and severe diseases (Table 1), and the validity of the diagnosis from the IPR could therefore be assumed to be high.

This study on substance-using criminal justice clients did not provide enough evidence to conclude any differences in the risk of CVD or CBD events between the substance-using groups. However, the highest incidence rates of both CVD and CBD events were found among predominantly amphetamine-using individuals. Implications of the present paper include the need to disseminate knowledge about somatic risk factors in both amphetamine users and users of other illicit drugs, including the larger risk seen in absolute numbers among amphetamine users, but also the need for clinicians and researchers to highlight co-factors which contribute to this risk increase beyond the substance itself. More studies are needed to assess whether amphetamine use is a risk factor for CVD or CBD in comparison to other drugs and distinguish substance-specific pathology from the impact of other unhealthy lifestyle factors among substance-using individuals.