Evaluating treatment outcomes across four substance use treatment settings: a 12-month longitudinal study in Iran Authors: Saeid Mirzaei, Vahid Yazdi-Feyzabadi, Mohammad Hossein Mehrolhassani, Nouzar Nakhaee, Nadia Oroomiei Published: 2025 ABSTRACT AbstractBackgroundSubstance use disorder is a persistent public-health challenge in Iran, where multiple treatment settings operate in parallel. This study assessed within-modality changes over 12 months among clients in Methadone Maintenance Treatment (MMT), Drop-in Centers (DICs), Therapeutic Communities (TCs), compulsory (Article 16) treatment.MethodsA prospective, naturalistic cohort was conducted across 15 routine-care centers in Kerman Province (May 2023–May 2024). Of 493 enrolled, 470 completed follow-up (MMT/DIC at 12 months; TC/compulsory at end-of-treatment [EOT]). MMT/DIC were assessed at baseline, 2 months, and 12 months; TC/Article 16 at baseline and EOT. Outcomes were assessed using the Maudsley Addiction Profile (MAP). Within-modality changes from baseline were estimated with mixed-effects models in Stata 15, using two-sided tests (α = 0.05).ResultsMMT was associated with reductions in alcohol, opium, and heroin use, lower odds of interpersonal conflicts, fewer physical/psychological problems, and fewer days of illness-related unemployment. DIC showed decreases in physical/psychological problems, drug-related harms (e.g., dealing), and fighting; changes in spouse/relative/friend-conflict indicators were not statistically significant. TC showed short-term reductions in several substances; employment decreased during the residential phase (a programmatic constraint rather than a change in employability). Article 16 showed declines in selected substances and health problems during observation; durability post-discharge remains uncertain.ConclusionsFindings indicate domain-specific, within-modality changes rather than comparative effectiveness between settings. In TCs, short-term within-modality declines were observed in alcohol and selected illicit opioids at EOT alongside limited employment opportunities during residence due to programmatic restrictions. Within MMT, improvements were observed in selected psychosocial indicators at follow-up, consistent with stabilization during ongoing care. Within DICs, reductions were observed in health problems and certain risk behaviors, aligning with their harm-reduction and linkage role. Within Article 16, short-term declines were observed in selected substances and health problems during observation; durability after discharge is uncertain without structured aftercare. Overall, no single modality appears universally superior; matching clients to treatment according to target outcomes, and implementing stepped-care or blended pathways with robust post-treatment supports—particularly following compulsory programs—may optimize long-term impact.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12954-025-01346-1.Keywords: Substance use disorder, Methadone maintenance treatment (MMT), Drop-in centers (DICs), Therapeutic communities (TCs), Article 16, Compulsory treatment Background Substance use disorder (SUD) is a major public health concern with significant social and economic repercussions worldwide [1]. According to the United Nations Office on Drugs and Crime (UNODC), global substance use has risen, with the number of drug users reaching 292 million by 2022—an increase of 20% over the past decade. Cannabis remains the most commonly used substance globally, followed by opioids, amphetamines, cocaine, and ecstasy. Despite an estimated 64 million people living with SUDs, only one in eleven individuals receives appropriate treatment [2]. Alcohol use disorder (AUD) is a major component of the global substance-use burden and a key driver of avoidable morbidity and mortality. In line with WHO/UNODC frameworks, alcohol is treated within the same continuum of care as other substances, ranging from risk reduction and managed consumption to abstinence-oriented recovery [3]. Historically, public-health and clinical responses to substance use have emphasized two broad approaches: abstinence-oriented recovery/rehabilitation and harm reduction. The relative emphasis between these approaches has varied across jurisdictions and periods; in many settings, recovery-oriented strategies were prominent, and over time they were complemented by harm-reduction components as evidence accumulated regarding benefits and limitations [4]. Harm reduction refers to pragmatic, evidence-informed policies and services that aim to reduce adverse health, social, and legal consequences associated with substance use. It encompasses a spectrum of goals and practices—from safer use and managed consumption to abstinence—and focuses on risk reduction rather than on a particular moral stance toward use [5]. Recovery, by contrast, is a clinical and psychosocial objective that shapes treatment planning and service delivery. The concepts are related but distinct: harm reduction targets risk and immediate harms, whereas recovery articulates longer-term change and well-being. In this study, “recovery” denotes a person-centered, voluntary process involving sustained remission (or non-problematic use, depending on the framework) together with improvements in health, social functioning, and quality of life. Harm-reduction services can stabilize risk, facilitate engagement, and support progress toward recovery, while also having value in their own right [6]. To avoid conflating distinct domains, we distinguish law-enforcement–led drug-control measures from clinical/public-health treatment and harm-reduction services. Where relevant, we describe the former as court-mandated or law-enforcement–led pathways, and the latter as treatment and harm-reduction services. In this framework, coercion refers to law-enforcement–led pathways (e.g., Article 16) that can channel individuals into abstinence-oriented care; it is conceptually distinct from both recovery (a clinical/psychosocial objective) and harm reduction (risk mitigation), and we interpret outcomes within each domain accordingly. Within this paradigm, alcohol is not peripheral: harm-reduction services commonly address risky drinking alongside drug use through brief interventions, counseling, and linkage to care. Recovery-oriented programs (e.g., TCs) typically adopt abstinence for alcohol and drugs, whereas low-threshold services (e.g., DICs) prioritize risk mitigation and engagement over immediate abstinence [7]. Over the past decade, community-based recovery supports—peer-led groups, counseling, and family-centered services—have expanded and now complement both harm-reduction and abstinence-oriented pathways. At the system level, many countries, including Iran, have integrated a mix of services—MMT, DICs, and community-based treatments—into national drug policies [8, 9]. The relative emphasis on abstinence-oriented versus harm-reduction approaches has differed by jurisdiction and over time, including during different phases of policy development in Iran. Iran, situated in the Middle East and North Africa (MENA) region and bordering Afghanistan—one of the world’s largest producers of illicit drugs—lies along a major drug trafficking route from east to west [10]. The country has grappled with substance use issues for decades. The prevalence of SUDs in Iran increased from 0.5% in 1990 to 0.8% in 2021 [11]. From 1979 to 2004, Iran’s national drug policy primarily emphasized abstinence- and recovery-oriented approaches within a law-enforcement framework, with limited integration of harm-reduction services. Over time, policy and service provision expanded to include harm-reduction components [10]. This transition led to the establishment of a range of services, including MMT, DICs, and therapeutic communities (TCs), collectively referred to as “Article 15” centers. Although the Iranian treatment system has historically focused on opioids, polysubstance use that includes alcohol is common in clinical practice. Consequently, alcohol outcomes are relevant across modalities—even when programs are primarily designed around opioid use [12]. Despite this policy evolution, law-enforcement–led drug-control measures continue alongside harm-reduction initiatives. Under Article 16, individuals may be court-mandated to short-term, abstinence-oriented residential care (typically ≥ 30 days), including assessment, withdrawal management/detoxification, abstinence-oriented counseling and psychoeducation, and basic medical care, with variable linkage to ongoing outpatient care after discharge [13–15]. In Iran, voluntary Article 15 services (MMT, DICs, TCs) operate alongside Article 16. Mandatory admissions do not occur in TCs; TC participation is voluntary under Article 15, whereas court-mandated treatment is implemented exclusively under Article 16 [16–18]. In this manuscript, we therefore treat Article 15 services as treatment and harm-reduction modalities, and Article 16 as a law-enforcement–led pathway for court-mandated, abstinence-oriented residential care [16–18]. Although production and sale are legally restricted, unrecorded alcohol use persists and is likely underreported due to stigma and regulatory constraints. This context complicates measurement, service uptake, and risk communication; nonetheless, alcohol remains a critical outcome for evaluating treatment impact across settings [19]. Despite a growing body of work on substance use in Iran, few studies have prospectively followed clients across multiple, structurally distinct service modalities with repeated assessments, while treating alcohol as a primary outcome alongside drug use and social/health indicators. Prior evaluations often focus on single modalities (typically opioid-centered) or cross-sectional snapshots, limiting inferences about within-modality change over time in LMIC settings [20–23]. A longitudinal, multimodality perspective is needed to inform person-centered policy and service planning. Consistent with this rationale, we designated alcohol use as a primary outcome, alongside selected drug, risk, health, and social-functioning indicators. Our aim is to describe within-modality change from baseline at the observed assessment points—MMT/DIC: 2 and 12 months (T2/T3); TC/Article 16: end-of-treatment (EOT)—across substance-use, health, risk, and social indicators. We do not estimate causal between-group effects; any cross-modality contrasts are descriptive given nonrandomized allocation, baseline heterogeneity, unequal sample sizes, and unbalanced follow-up schedules. RQ1: (Substance use): Within each modality, what is the change from baseline in substance-specific indicators (alcohol, opium, heroin, hashish/cannabis, methamphetamine, prescribed methadone, illegal methadone)? RQ2: (Health status): Within each modality, what is the change from baseline in physical- and psychological-problem scores? RQ3: (Risk behaviors): Within each modality, what is the change from baseline in unsafe sexual intercourse and interpersonal conflicts (with spouse, relatives, friends)? RQ4: (Social/crime & employment): Within each modality, what is the change from baseline in days unemployed, unemployment due to illness, days selling drugs, days in physical conflicts/fights, and days employed? RQ5 (exploratory, descriptive): Do within-modality temporal patterns differ across domains (substance use, health, risk, social/employment) or by service type (residential: TC/Article 16 at EOT; outpatient/low-threshold: MMT/DIC at T2/T3), acknowledging that these contrasts are descriptive and non-causal? This study contributes to the global discourse on substance use treatment in low- and middle-income countries (LMICs) by documenting treatment outcomes across multiple service modalities currently operating in Iran. Rather than testing head-to-head comparative effectiveness, within-modality outcome patterns—ranging from harm-reduction (MMT, DIC) to recovery-oriented residential (TC) and compulsory abstinence-based (Article 16) approaches—are described here. Direct statistical cross-modality comparisons were not undertaken because allocation was nonrandomized with baseline heterogeneity; assessment schedules were unbalanced (MMT/DIC: three waves; TC/Article 16: two waves); sample sizes were unequal (smaller TC/Article 16 cohorts); and structural differences between residential and outpatient settings (e.g., restricted employment during residence) would undermine causal interpretability. This framing reflects the practical and ethical complexities of implementing structurally diverse interventions in sociopolitical environments where law-enforcement–led and harm-reduction strategies coexist. These findings offer important insights for policymakers aiming to align national strategies with evidence-based, person-centered approaches recommended by the WHO and UNODC. Methods Design and setting A prospective, observational, naturalistic cohort study was conducted in routine-care settings across Kerman Province, Iran, from May 2023 to May 2024. Treatment allocation and content were determined by usual care, with no investigator manipulation. Outcomes were operationalized as within-person change in substance use and individual functioning. Participants were recruited from MMT centers, DICs, TCs, and Article 16 compulsory treatment centers. Because TCs operated only in Kerman and Bam during the study period, centers in both cities were included. Relation to the parent project and analytic scope This manuscript is part of a larger multi-phase evaluation. The baseline, cross-sectional characterization of clients at intake across MMT, DIC, TC, and Article 16 centers has been published elsewhere and is not repeated here [18]. The present report focuses on longitudinal effectiveness within each modality—i.e., change over time in MAP domains (substance use, health-risk behaviors, physical/psychological well-being, and social functioning) at 2 and 12 months for MMT/DIC and at end-of-treatment for TC/Article 16—using mixed-effects models. Analyses that are outside the scope of this paper include: (i) detailed baseline risk-profile comparisons already reported in the cross-sectional publication; (ii) biological validation analyses (patterns of urine toxicology and concordance with self-report), which will be presented in a separate methods-focused paper; and (iii) health-related quality of life trajectories, which are being developed as an independent manuscript with a distinct conceptual framework and endpoints. This scoping reflects the present paper’s rationale—estimating treatment outcome trajectories in routine care while minimizing redundancy, multiple-testing burden, and word-count constraints. Participant flow and missing data A total of 493 participants were enrolled at baseline (MMT 230; DIC 114; Article 16 101; TC 48). For descriptive transparency, interim retention counts were: MMT 227 at 2 months (attrition from baseline: 3/230, 1.30%) and 220 at 12 months (overall attrition from baseline: 10/230, 4.35%; additional attrition between 2 and 12 months: 7/227, 3.08%). DIC 112 at 2 months (attrition from baseline: 2/114, 1.75%) and 101 at 12 months (overall attrition from baseline: 13/114, 11.40%; additional attrition between 2 and 12 months: 11/112, 9.82%). TC and Article 16 are short-term residential programs with end-of-treatment (EOT) assessments only; EOT completion was 100% (TC 48/48; Article 16 101/101). Overall attrition percentages use the baseline denominator; additional attrition uses the 2-month denominator. For all analyses reported in this paper (and in the prior baseline paper), the analytic cohort was restricted a priori to completers—i.e., participants who finished their modality’s last scheduled assessment (MMT/DIC: 12 months; TC/Article 16: EOT). Consequently, 23 baseline enrollees (MMT 10; DIC 13) who did not reach their last scheduled wave were excluded from analyses. The resulting analytic cohort comprised 470 completers: MMT 220; DIC 101; Article 16 101; TC 48. All descriptive statistics, tables, and model estimates pertain exclusively to these 470 completers; interim retention counts are provided only for transparency and were not used in the analyses. TC and Article 16 operate as short-term residential programs. Post-discharge long-term follow-up infrastructure was not available for TC and Article 16; therefore, these modalities were assessed at baseline and EOT only. Attrition observed in MMT/DIC among all 493 enrollees largely reflected treatment discontinuation or loss to follow-up under routine care. Accordingly, all estimates reflect within-modality change among completers. Study population and sampling The study population comprised adults (≥ 18 years) initiating index treatment at participating routine-care centers across Kerman Province (MMT, DIC, TC, and Article 16). Treatment allocation and content were determined by usual care; investigators did not assign modality. Detailed eligibility and consent procedures are provided under Screening and data-collection procedures. Sampling followed World Health Organization (WHO) guidelines [24] for evaluating substance-use treatment programs, which recommend enrolling ≥ 100 participants per modality. Using Cochran’s formula for an unknown population (95% confidence level, 5% margin of error, assumed prevalence 0.5), the minimum total sample was 384; to allow subgroup analyses and potential non-response, the target was increased to 470 (Table 1). Table 1Participating centers and final analytic sample by treatment modalityCenter namesArticle 16TCsDICsMMT centersNumber of centers1239Number of participants10148101220Counts reflect the final analytic sample (last available wave). Baseline enrollments were reported in Methods (Participant flow and missing data). TC and Article 16 are residential programs with no mid-course exits; end-of-treatment completion was 100% Participating centers and final analytic sample by treatment modality Counts reflect the final analytic sample (last available wave). Baseline enrollments were reported in Methods (Participant flow and missing data). TC and Article 16 are residential programs with no mid-course exits; end-of-treatment completion was 100% A multistage sampling strategy was implemented: stratified random selection of MMT centers (two per urban district), subject to operational feasibility (e.g., supervisor willingness); all eligible DIC, TC, and Article 16 centers operating during the study period were approached, and participating sites were included. Within each participating site, consecutive eligible new admissions were enrolled. The final analytic cohort included 220 participants from MMT centers, 101 from DICs, 48 from TCs, and 101 from Article 16 centers. Although TC enrollment fell below the WHO-recommended threshold, all 48 available clients were included—exceeding the minimum acceptable subgroup size of 30. To minimize recall bias, only newly admitted clients were enrolled at baseline. Sex distribution within each modality reflected routine service populations in 2022–2023 and informed center selection. In Kerman’s five urban districts, two MMT centers were randomly selected per district; in the district with only one eligible center, that site was included (total nine MMT centers). All eligible DIC, TC, and Article 16 centers operating in Kerman at the time were included. Due to low enrollment at the Kerman TC, the Bam TC was also added to the sampling frame. During data collection, TC and Article 16 centers were authorized to serve only men; no female-only facilities under these modalities were operational. In total, 15 centers participated: nine MMT centers, three DICs, one Article 16 center, and two TC centers (Kerman and Bam). Eligibility screening and MAP interviews were conducted by center psychologists under a uniform SOP; see Screening and data-collection procedures for details. Screening and data collection procedures At each participating center, consecutive new admissions were identified during routine intake and were screened for eligibility by center psychologists. Eligibility required age ≥ 18 years (verified with official ID where available), a clinician-documented substance use disorder (SUD) diagnosis in the medical record per routine practice, initiation of the index treatment at the time of screening, and capacity to provide written informed consent. Written informed consent was obtained prior to the baseline assessment. Individuals who had already initiated the index treatment prior to screening or declined participation were excluded. Eligibility verification and consent were documented on standardized forms. Each site maintained a screening log (approached/eligible/consented; reasons for ineligibility/refusal). No study data were collected from individuals who did not consent. Pre-launch standardization Prior to study initiation, two members of the research team visited collaborating centers, delivered a uniform briefing and a written standard operating procedure (SOP) covering eligibility checks, informed-consent procedures, administration of the Maudsley Addiction Profile (MAP; 30-day recall), urine toxicology procedures (collection, recording), and kit storage. A study contact phone number was provided for real-time procedural questions throughout fieldwork. All centers implemented the same procedures under a uniform SOP, ensuring comparable screening and data collection across modalities. Interview setting and administration Baseline and follow-up interviews (MMT/DIC: T1, T2, T3; TC/Article 16: T1, EOT) were conducted in person, in a private room on-site, by the center psychologists using the interviewer-administered MAP and a short sociodemographic form, following the written SOP. To mitigate potential interviewer or social-desirability bias, interviews were conducted without other clinical staff present. Urine toxicology was obtained at each time point during the same visit; results were recorded on the case-report forms, and kit lot numbers and storage logs were maintained on-site by the psychologists. Modality-specific follow-ups For MMT/DIC, assessments were scheduled at baseline (T1), ~ 2 months (T2), and ~ 12 months (T3) during routine visits; missed visits triggered phone reminders and rescheduling windows (e.g., T2 ± 14 days; T3 ± 30 days). For TC/Article 16, assessments were conducted at baseline (admission) and end-of-treatment (EOT) within 24–72 h prior to discharge; post-discharge long-term follow-up was not available in these modalities. At baseline, consent included permission for recontact, and contact details (e.g., phone numbers) were recorded. For MMT/DIC follow-ups, center psychologists issued reminder calls/SMS within the predefined rescheduling windows; all assessments were administered in person, and no telephone interviews were conducted. Data handling and quality assurance Center psychologists completed paper case report forms (CRFs), which were de-identified and transferred to the coordinating team for data entry/cleaning. The coordinating team conducted periodic checks of incoming forms against site logs (screening, kit inventory, urine results) and resolved discrepancies via phone with the center psychologists. These procedures ensured consistent implementation across centers and supported data completeness and fidelity. Paper forms were digitized into a secure database with predefined range and logic checks and an audit trail; any discrepancies were resolved via data queries with the center psychologists. Ethics The study protocol was approved by the Ethics Committee of Bam University of Medical Sciences, Iran (IR.MUBAM.REC.1404.005). All participants provided written informed consent prior to enrollment. Procedures adhered to the principles of the Declaration of Helsinki. Data were de-identified for analysis, and participants were informed that participation was voluntary and that they could withdraw at any time without consequences. Data collection and instruments Data were collected using the Maudsley Addiction Profile (MAP) questionnaire [25], a structured, interviewer-administered instrument that evaluates functioning among individuals with substance use disorders. The tool covers four domains: (1) substance use behavior, (2) health-related risk activities, (3) physical and psychological well-being, and (4) social functioning. Items are categorical or frequency-based and refer to experiences/behaviors over the preceding 30 days (MAP 30-day recall). Unless otherwise specified, higher problem scores indicate worse status. The validated Persian version of the MAP has demonstrated strong psychometric properties in Iran (e.g., Sadeghi et al., 2013; Afshari et al., 2011) [26, 27]. In the present study, internal consistency was excellent (Cronbach’s α = 0.92). MAP was selected based on operational feasibility in routine-care settings, its documented psychometrics in Iran, and alignment with our primary outcomes. The Persian ASI was considered; however, its higher administrative burden was less feasible for a 15-center longitudinal protocol, with greater risk of attrition and interviewer-level variability. MAP captures past-30-day physical and psychological symptoms but does not assess clinical diagnoses, concurrent psychiatric/medical treatment, or treatment motivation; these constructs were not measured systematically in this study. Participants in MMT and DIC were assessed at three time points (baseline, 2 months, 12 months). In TC and Article 16, only two waves were available because both modalities are delivered as short-term residential programs. Once clients are discharged, centers do not maintain systematic follow-up infrastructure, precluding a 12-month assessment; therefore, data collection was limited to baseline and end-of-treatment in these groups. Urine drug testing was conducted at each time point to assess the accuracy of self-reported substance use; however, these biological data were collected as ancillary checks rather than primary endpoints and are not analyzed in this paper. Measures Alcohol and substance use were measured with the MAP consumption items for the past 30 days. Participants self-reported frequency (days used; range 0–30) and typical quantity per day in their customary units (e.g., glasses, grams, milliliters, joints). For analysis, all substance-use measures were expressed in standardized units as follows: alcohol in standard drinks/day (1 standard drink = 10 g pure ethanol); opium/opium paste in grams/day; heroin in grams/day; methadone (prescribed and illegal) in milligrams/day (mg/day; entries originally recorded in mL/day were converted using the site-specific formulation concentration, mg/mL); cannabis in grams/day; and methamphetamine in grams/day. Where both joints and grams were available for cannabis, grams/day took precedence; otherwise, the reported unit was retained. Higher values indicate greater consumption. Continuous outcomes in Table 2 are reported as ΔMean (95% CI) in these standardized units (negative values indicate reductions from baseline). Table 2Substance use outcomes—ΔMean vs baseline (MMT/DIC: T2/T3; TC/ Article 16: EOT)OutcomeMMT (T2)MMT (T3)DIC (T2)DIC (T3)Article 16 (EOT)TC (EOT)Alcohol − 4.79 (− 7.83 to − 1.75), p < 0.001 − 5.41 (− 8.75 to − 2.08), p < 0.001 + 2.24 (− 7.30 to 11.78), p = 0.65 + 6.99 (− 3.33 to 17.30), p = 0.18 − 0.83 (− 1.76 to 0.10), p = 0.08 − 1.63 (− 2.64 to − 0.61), p < 0.001Opium − 0.92 (− 1.35 to − 0.50), p < 0.001 − 0.57 (− 1.04 to − 0.10), p = 0.02 − 0.44 (− 1.20 to 0.32), p = 0.26 − 0.54 (− 1.35 to 0.26), p = 0.18 − 0.66 (− 0.94 to − 0.38), p < 0.001 − 0.75 (− 1.21 to − 0.30), p < 0.001Heroin − 0.49 (− 0.84 to − 0.13), p = 0.01 − 0.55 (− 0.94 to − 0.16), p = 0.01 − 0.75 (− 1.65 to 0.16), p = 0.11 − 1.21 (− 2.18 to − 0.24), p = 0.01 − 0.59 (− 0.85 to − 0.34), p < 0.001 − 1.19 (− 2.29 to − 0.09), p = 0.04Prescribed methadone + 3.47 (0.61 to 6.33), p = 0.017 + 4.97 (2.01 to 7.93), p < 0.001 + 2.66 (0.90 to 4.42), p < 0.001 + 3.67 (1.81 to 5.54), p < 0.001 + 9.45 (2.26 to 16.64), p < 0.001 − 3.59 (− 6.80 to − 0.38), p = 0.03Illegal methadone − 1.21 (− 2.02 to − 0.41), p < 0.001 − 1.18 (− 2.06 to − 0.29), p = 0.01 − 0.79 (− 1.88 to 0.29), p = 0.15 − 0.70 (− 1.85 to 0.45), p = 0.23 − 0.30 (− 0.57 to − 0.03), p = 0.03 − 2.83 (− 5.70 to 0.03), p = 0.05Cannabis (hashish) − 0.62 (− 1.11 to − 0.17), p = 0.01 − 0.36 (− 0.84 to 0.13), p = 0.15 − 0.91 (− 1.59 to − 0.23), p = 0.01 − 0.91 (− 1.63 to − 0.19), p = 0.01 + 0.26 (0.06 to 0.46), p < 0.001 − 0.65 (− 1.62 to 0.33), p = 0.40Methamphetamine − 0.02 (− 0.34 to 0.29), p = 0.89 + 0.22 (− 0.12 to 0.57), p = 0.21 − 0.56 (− 1.25 to 0.13), p = 0.11 − 0.56 (− 1.30 to 0.18), p = 0.14 − 0.73 (− 1.88 to 0.44), p = 0.22 − 0.20 (− 0.34 to − 0.06), p = 0.01Time points: T2 = 2 months; T3 = 12 months; EOT = end of treatment.Estimates are from linear mixed-effects models with time nested within participants; values are ΔMean (95% CI) vs. baseline (negative = reduction). Unadjusted estimates are presented.Units: alcohol = standard drinks/day (1 drink = 10 g ethanol); opium/opium paste, heroin, methamphetamine = grams/day; methadone (prescribed/illegal) = mg/day; cannabis = grams/day(joints converted to grams where available) Substance use outcomes—ΔMean vs baseline (MMT/DIC: T2/T3; TC/ Article 16: EOT) Time points: T2 = 2 months; T3 = 12 months; EOT = end of treatment.Estimates are from linear mixed-effects models with time nested within participants; values are ΔMean (95% CI) vs. baseline (negative = reduction). Unadjusted estimates are presented.Units: alcohol = standard drinks/day (1 drink = 10 g ethanol); opium/opium paste, heroin, methamphetamine = grams/day; methadone (prescribed/illegal) = mg/day; cannabis = grams/day(joints converted to grams where available) Risk behaviors and count outcomes. Binary risk behaviors were defined as any occurrence in the past 30 days (MAP). Count outcomes were operationalized as the number of days in the past 30 days. Logistic generalized linear mixed models report odds ratios (OR), and Poisson/negative binomial generalized linear mixed models report incidence rate ratios (IRR). Data analysis Rationale and specification Primary analyses were conducted separately by modality to describe within-modality trajectories and were restricted to the 470-participant completer cohort. Within this cohort, mixed-effects models used all available wave-level observations without imputation, assuming missing at random (MAR) for intermittent missingness (e.g., a completer missing the 2-month visit). Time was modeled categorically (baseline, 2 months, 12 months for MMT/DIC; baseline and EOT for TC/Article 16), with baseline as the reference where applicable. All models included a participant-level random intercept; random slopes for time were examined and not retained based on AIC/BIC and variance components. Continuous endpoints were analyzed using linear mixed-effects models (REML); binary endpoints were analyzed using logistic GLMMs (logit link; maximum likelihood with Laplace approximation; standard convergence diagnostics applied); and count endpoints were analyzed using Poisson or negative binomial GLMMs (log link), selecting the latter when overdispersion was detected. Two-sided tests (α = 0.05) were applied. GLMMs were selected to handle intra-subject correlation, non-Gaussian endpoints, and unequal visit spacing; for two-wave modalities (TC/Article 16), the random-intercept specification reduces to the usual within-subject paired contrast. Mixed-effects models were used not merely because groups had different numbers of waves, but because within each modality the data are hierarchical—repeated measurements are nested within persons—creating intra-subject correlation, intermittent missingness, and unequally spaced visits; accordingly, mixed models are warranted irrespective of cross-modality pooling. We sought subject-specific (conditional) estimates of within-person change, for which LMM/GLMM are appropriate; in contrast, GEE targets population-averaged effects and can be less stable with small cluster counts and two-wave designs. Relative to paired/change-score approaches, mixed models (i) use all available waves, (ii) accommodate unequal visit spacing, and (iii) include cases with intermittent missingness under the MAR assumption without ad-hoc imputation, while mapping naturally to continuous (β), binary (OR = expβ), and count (IRR = expβ) endpoints. Reporting Continuous endpoints (e.g., physical/psychological problem scores; alcohol, opium, heroin, cannabis (hashish), and methamphetamine quantities; prescribed and illegal methadone doses (mg/day) are reported as mean differences (β). Binary endpoints (e.g., unsafe sexual intercourse; conflicts) are reported as odds ratios (OR = expβ). Count endpoints (e.g., days employed/unemployed, fighting) are reported as incidence rate ratios (IRR = expβ). For the two-wave modalities, the random-intercept specification reduces to the usual paired within-subject contrast for continuous outcomes and to the corresponding conditional analyses for binary and count outcomes; interpretations are therefore aligned with those approaches. Why primary analyses were stratified by modality A pooled mixed-effects model with modality × time interactions was considered. Primary analyses were stratified by modality to align with the asymmetric design (MMT/DIC: baseline, 2 months, 12 months; TC/Article 16: baseline and EOT with variable duration), male-only enrollment in two modalities, heterogeneous outcome families/dispersion (e.g., Poisson versus negative binomial), and modality-specific missingness patterns. A pooled specification would require collapsing/extrapolating non-overlapping time points, estimating high-order interactions with limited cells (particularly for Article 16), and imposing common variance assumptions that are unlikely to hold. Stratified models therefore yield design-congruent, within-modality estimands with clearer clinical interpretation under fewer cross-modality assumptions. To align the estimand with the observed data structures, a pooled modality × time model was not fitted or reported, and all primary inferences were drawn from modality-specific models. Covariate handling Primary mixed-effects models estimated within-person change (time nested within participants with a participant-level random intercept). This specification accounts for stable between-person heterogeneity via the random intercept; because time-invariant characteristics do not vary within individuals, they cannot confound the within-person time contrasts in these models. Given the short follow-up windows (2 and 12 months), age is effectively constant within participants and thus not a source of within-person confounding. Moreover, TC and Article 16 enrolled men only, making a pooled sex adjustment across modalities ill-defined and potentially misleading. For parsimony and transparency, unadjusted primary estimates are reported, and baseline differences are explicitly described to contextualize findings. Our primary estimands are within-modality, within-person changes over time; between-modality comparisons are descriptive only given nonrandomized allocation. Center effects Given the small number of centers within some modalities (e.g., a single Article 16 site), center-level random effects were not retained; estimates are reported from the main specification. Model assumptions The Poisson specification was used only when the mean and variance of the outcome were approximately equal; when variance exceeded the mean, the negative binomial model was preferred. Overdispersion was assessed via variance-to-mean comparisons and the likelihood-ratio test of the overdispersion parameter (α = 0). Treating time categorically avoided imposing a linear trend. All analyses were conducted in Stata 15 (StataCorp, College Station, TX). Models were checked for convergence and sparse-data issues; estimates for outcomes with very low event counts (and correspondingly large standard errors) were reported but interpreted cautiously. Where nested, random-slope terms were compared with random-intercept models using AIC/BIC and likelihood-based criteria; additional random slopes were not retained when improvements were unsupported. For LMMs, conditional residuals (QQ plots and residuals-versus-fitted) were visually inspected to assess approximate normality and homoscedasticity; influential observations were screened with standard diagnostics. For GLMMs, conditional Pearson residuals and dispersion checks were reviewed; model selection relied on AIC/BIC and convergence diagnostics. These diagnostics supported the reported specifications. Multiplicity Given the descriptive scope across multiple endpoints, no formal multiplicity adjustment was applied. Primary interpretation emphasizes patterns and effect sizes within outcome families and 95% confidence intervals (CIs) rather than isolated p-values. Results Overview and baseline context At baseline, the modalities served distinct populations. Approximately 90% of participants were male, and during data collection TC and Article 16 enrolled men only; most were aged 26–45, and about 70% reported onset of use after age 18. Primary substance profiles differed by setting: opium/opium paste was most common in MMT (about 68%), heroin was more prevalent in DIC (about 31%), and methamphetamine was relatively more common as a primary substance in Article 16 (about 23%). For past-30-day use, crystal methamphetamine prevalence was ~ 42% in DIC and TC, and lowest in MMT (~ 12%). Prior treatment experience also varied by modality—TC 56.3%, MMT 40.9%, Article 16 36.6%, and DIC 19.8% reported at least one previous episode—providing additional context for the modality-specific trajectories. Markers of social vulnerability clustered in DIC (~ 53%) and were also elevated in Article 16 (~ 47%), alongside higher rates of high-risk behaviors. These baseline differences contextualize the longitudinal changes summarized below; full cross-sectional profiles are available in the published baseline report [18]. Baseline characteristics by modality are summarized in Supplementary Table S1. At study entry, mean MAP physical-problem scores were highest in Article 16 (13.50 [SD 7.05]) and lowest in TC (9.12 [7.39]); MMT and DIC were similar (12.35 [6.75] and 12.56 [6.97], respectively). Mean psychological-problem scores were highest in DIC (16.89 [11.13]) and lowest in Article 16 (4.93 [7.05]); MMT and TC averaged 11.94 (7.93) and 9.39 (7.28), respectively. These values refer to past-30-day symptoms; higher scores indicate worse status. Together, they indicate greater baseline physical burden in Article 16 and greater baseline psychological burden in DIC, providing clinical context for the modality-specific trajectories reported in Tables 2, 3, 4 and 5. Table 3Physical & psychological problems—ΔMean vs baselineOutcomeMMT (T2)MMT (T3)DIC (T2)DIC (T3)Article 16 (EOT)TC (EOT)Physical problems − 4.58 (− 5.62 to − 3.54), p < 0.001 − 2.51 (− 3.67 to − 1.35), p < 0.001 − 5.63 (− 7.25 to − 4.01), p < 0.001 − 5.36 (− 7.09 to − 3.64), p < 0.001 − 10.71 (− 13.69 to − 7.73), p < 0.001 − 0.85 (− 3.45 to 1.74), p = 0.52Psychological problems − 4.78 (− 5.82 to − 3.73), p < 0.001 − 3.57 (− 4.74 to − 2.40), p < 0.001 − 9.40 (− 11.40 to − 7.39), p < 0.001 − 10.75 (− 12.90 to − 8.60), p < 0.001 − 7.99 (− 9.79 to − 6.19), p < 0.001 − 2.14 (− 5.18 to 0.90), p = 0.17Time points: T2 = 2 months; T3 = 12 months; EOT = end of treatment.Outcomes are MAP problem scores (past 30 days); higher scores indicate worse status.Estimates are from linear mixed-effects models with time nested within participants; values are ΔMean (95% CI) vs. baseline (ΔMean < 0 = improvement). Unadjusted estimates are presented Physical & psychological problems—ΔMean vs baseline Time points: T2 = 2 months; T3 = 12 months; EOT = end of treatment.Outcomes are MAP problem scores (past 30 days); higher scores indicate worse status.Estimates are from linear mixed-effects models with time nested within participants; values are ΔMean (95% CI) vs. baseline (ΔMean < 0 = improvement). Unadjusted estimates are presented Table 4Risk behaviors—odds ratios vs baseline (OR, 95% CI, p)OutcomeModalityTimeOR (95% CI)pUnsafe sexual intercourseMMTT23.06 (1.17 to 8.08)0.02DICT22.64 (1.00 to 6.96)0.05Conflict with spouseMMTT20.32 (0.18 to 0.55) < 0.01Conflict with relativesMMTT20.39 (0.23 to 0.66) < 0.01Conflict with friendsMMTT20.23 (0.11 to 0.50) < 0.01Time points: T2 = 2 months; T3 = 12 months; EOT = end of treatment.Unsafe sexual intercourse: any unprotected vaginal/anal sex in the past 30 days. Conflict measures: any past-30-day episodes (spouse, relatives, friends). OR < 1 indicates lower odds vs. baseline.Estimates are from logistic generalized linear mixed models (logit link) with time nested within participants; OR = exp(β). Unadjusted estimates are presented Risk behaviors—odds ratios vs baseline (OR, 95% CI, p) Time points: T2 = 2 months; T3 = 12 months; EOT = end of treatment.Unsafe sexual intercourse: any unprotected vaginal/anal sex in the past 30 days. Conflict measures: any past-30-day episodes (spouse, relatives, friends). OR < 1 indicates lower odds vs. baseline.Estimates are from logistic generalized linear mixed models (logit link) with time nested within participants; OR = exp(β). Unadjusted estimates are presented Table 5Social & crime/employment—incidence rate ratios vs baseline (IRR, 95% CI, p)OutcomeModalityTimeIRR (95% CI)pDays unemployedMMTT31.49 (1.04 to 2.12)0.03TCEOT2.36 (1.30 to 4.31)0.01Unemployed due to illnessMMTT20.37 (0.16 to 0.86)0.02DICT20.33 (0.12 to 0.90)0.03Days selling drugsDICT20.14 (0.03 to 0.63)0.01DICT30.02 (0.00 to 0.10) < 0.01Days in conflicts/fightsDICT20.28 (0.11 to 0.71)0.01DICT30.10 (0.03 to 0.35) < 0.01Time points: T2 = 2 months; T3 = 12 months; EOT = end of treatment.Counts refer to the number of days in the past 30 days (MAP items). IRR < 1 indicates fewer days vs. baseline (IRR > 1 = more days).Estimates are from Poisson/negative binomial generalized linear mixed models (log link) with time nested within participants; IRR = exp(β). Unadjusted estimates are presented.NE = not estimable due to model non-convergence or unstable CIs Social & crime/employment—incidence rate ratios vs baseline (IRR, 95% CI, p) Time points: T2 = 2 months; T3 = 12 months; EOT = end of treatment.Counts refer to the number of days in the past 30 days (MAP items). IRR < 1 indicates fewer days vs. baseline (IRR > 1 = more days).Estimates are from Poisson/negative binomial generalized linear mixed models (log link) with time nested within participants; IRR = exp(β). Unadjusted estimates are presented.NE = not estimable due to model non-convergence or unstable CIs Follow-ups were conducted at 2 months (T2) and 12 months (T3) for MMT and DIC; for TC and Article 16, outcomes are reported at end of treatment (EOT). Primary analyses used mixed-effects models with time treated categorically and nested within participants (random intercept). Continuous outcomes are reported as mean change from baseline (ΔMean) with 95% CIs (confidence intervals); binary outcomes as odds ratios (OR); and count outcomes as incidence rate ratios (IRR). All substance-use quantities are expressed in the standardized units specified in Methods; negative ΔMean values indicate reductions. Unadjusted estimates are presented. The within-person specification accounts for time-invariant heterogeneity, and two modalities enrolled men only, making a global sex adjustment not meaningful across modalities; age is effectively stable over the 2–12-month windows. Baseline differences are summarized above to contextualize the findings. Cross-modality differences reported in Tables 2, 3, 4 and 5 are descriptive only; primary inferences pertain to within-modality change rather than pooled modality × time contrasts. Across modalities, MMT shows consistent reductions in alcohol, opium, and heroin at both Follow-ups. Within TC, short-term declines were observed at EOT (including alcohol, heroin, and methamphetamine); these signals are descriptive and within-modality rather than comparative. Prescribed methadone increases in MMT and DIC but decreases in TC, whereas illegal methadone declines in Article 16 (and is borderline in TC). Cannabis decreases in DIC but increases in Article 16. These patterns suggest modality-specific trajectories rather than a uniform effect across substances. Physical and psychological problems decline significantly for MMT and DIC at both Follow-ups, with Article 16 also showing within-modality reductions at EOT. In contrast, TC does not exhibit statistically significant change in these domains. Overall, improvements in DIC and MMT appeared more consistently across waves, indicating benefits that extended beyond short-term endpoints; these signals are descriptive and within-modality rather than comparative. Risky sexual behaviors display a transient increase at T2 in MMT and DIC. Conversely, MMT is associated with significant reductions in interpersonal conflicts (spouse, relatives, friends). No other modality shows consistent or statistically significant shifts across these indicators, underscoring that behavioral risks may not improve uniformly alongside substance-use outcomes. Days unemployed increase in MMT (T3) and TC (EOT), whereas unemployment due to illness decreases in both MMT and DIC. In DIC, days selling drugs and days in physical conflicts drop sharply at both Follow-ups. Notably, days of employment do not change significantly across modalities, suggesting that improvements in risk and health domains may not immediately translate into gains in formal employment. Discussion This observational, nonrandomized cohort was not designed for between-modality comparisons. Cross-modality differences are described for context only; no claims of comparative effectiveness are made. All statements below are within-modality and descriptive; apparent differences across settings may reflect selection, assessment schedules, and residential containment rather than treatment effects. In the MMT group, within-modality declines in alcohol, opium, and heroin use were observed over time. In our nonrandomized, within-modality data, at EOT the TC group showed within-modality declines in alcohol, heroin, and methamphetamine, with a borderline decline in illegal methadone; these patterns are descriptive, not causal, and may partly reflect residential containment and selection. In Iran, this pattern may reflect programmatic differences in design and accessibility—MMT as widely scaled, medication-anchored outpatient care versus TC as structured, residential care. Within-modality reductions observed in MMT align with prior literature reporting decreases in alcohol and illicit substance use among MMT clients [28, 29]. This is consistent with strong HIC evidence on OAT benefits (e.g., reduced all-cause and overdose mortality) [30, 31], and with Iran’s large opioid agonist therapy (OAT) footprint, where stabilization under routine MMT can translate into gradual reductions in selected substances. In practice, coverage and dosing policies for OAT in Iran may facilitate stabilization while leaving broader social determinants (e.g., employment) less directly affected in the short term [15]. Within this harm-reduction framing and the multi-modality context in Iran, we prioritized outcomes that are both policy-relevant and measurable across modalities: (i) substance use by drug class (proximal clinical risk), (ii) risk and safety indicators (unprotected sex, days involved in fights/selling drugs), and (iii) functional/health domains (interpersonal conflict, physical/psychological symptoms, employment, and treatment engagement/retention). These endpoints align with program goals in Iran (stabilization and risk reduction alongside treatment engagement) and are feasible to collect repeatedly in routine care. Comparable multi-domain outcome tracking has been used in UK cohorts (e.g., NTORS) and in Iranian OAT evaluations, supporting repeated-measure assessment in routine settings [32–35]. Interpersonal conflicts with spouses, relatives, and friends declined significantly at the second measurement in the MMT group. Comparable changes were not observed for TC or Article 16. DIC did not show significant effects on these specific conflict indicators, although other conflict-related outcomes (e.g., days involved in physical altercations) declined significantly in DIC. This aligns with the findings of Ganesh et al., who reported improved family functioning and quality of life among individuals receiving MMT [36]. Furthermore, average physical and psychological health problems decreased, and the number of days unemployed due to illness declined at the second measurement. Furthermore, unemployment increased in the MMT group at the third measurement. Across modalities, days employed did not change significantly; unemployment increased in TC at the second measurement and rose slightly in MMT by the third, which should be interpreted cautiously given TC’s residential restrictions. By program policy, TC is a residential setting where paid employment is typically not permitted during residence; therefore, EOT employment indicators are not directly comparable to outpatient settings. These null or adverse employment trends may reflect macroeconomic headwinds and stigma; future work should test integrated vocational packages (job-readiness training, employer engagement, and legal support) within each modality. In the Iranian labor market, employer screening practices and stigma toward people with a history of substance use plausibly dampen near-term gains, underscoring the need for structured placement partnerships alongside clinical care. Coordination with Iran’s welfare and labor agencies (e.g., job centers and small- and medium-sized enterprises [SMEs]) could be tested as an implementation strategy in routine care, rather than inferred as an explanation for (in)effectiveness in our data [37–40]. These changes suggest within-modality improvements in psychosocial measures in MMT, with TC showing within-modality declines in several substances at EOT; all signals are descriptive rather than comparative efficacy. Taken together, the patterns were domain-specific (e.g., within-modality declines at EOT were observed in TC; several psychosocial indicators improved within MMT), effect sizes were generally modest, and several 95% CIs crossed conventional significance thresholds; accordingly, no single modality can be identified as superior, particularly within Iran’s mixed service configuration. Beyond substance-use metrics, a strengths-based lens indicates gains in recovery capital—e.g., reduced interpersonal conflict, declines in physical/psychological problems, fewer days selling drugs or involved in fights, and sustained engagement with services—which may represent meaningful functional improvement even when abstinence is incomplete. Notably, an increase at T2 was observed in MMT (and borderline in DIC); persistence beyond T2 was not consistent across time points in our analyses and should be interpreted cautiously. This transient rise suggests unmet sexual-health needs early in treatment; integrating condoms, STI testing, and brief risk-reduction counseling into MMT intakes may mitigate this [41]. This may be particularly salient in Iran, where integrated sexual-health services are not uniformly embedded within routine OAT visits. Similar service-package gaps are noted in Iranian OAT guidance, motivating pilots that integrate sexual-health supports within intake workflows [42]. Pilot integration of STI services into MMT intakes at high-volume clinics in Iran would be a feasible, low-cost test of this proposition. Framed positively, early identification and rapid integration of sexual-health supports within OAT can enhance self-efficacy and patient-reported quality of care—key elements of recovery capital in the Iranian context. It should also be acknowledged that the selected outcome measures carry important limitations. Several measures are brief and self-reported, and therefore susceptible to recall and social-desirability bias; some indicators are dichotomous (e.g., unprotected sex) and may obscure frequency or context; and the employment measure used here did not capture informal work or underemployment. In addition, MAP items do not capture treatment motivation, program fidelity, or concurrent psychiatric care; these constructs were outside the scope of this study and may moderate trajectories. “Days selling drugs” and “days in fights” are partly contingent on exposure opportunity (for example, constraints associated with residential TC), and assessments at EOT may partly reflect residence/containment effects rather than durable change. Even with these constraints, this outcome portfolio maps onto widely used recovery domains—health and safety (symptoms, substance use, violence), community/relationships (conflict), purpose (employment), and treatment engagement—offering a pragmatic, harm-reduction–consistent view of progress beyond abstinence, in line with HIC guidance on realistic treatment outcomes [43]. In this light, apparent discrepancies with prior reports on employment should be interpreted cautiously, as they may reflect differences in measurement scope as well as structural labor-market constraints in Iran—points elaborated in the following paragraph. While Ganesh et al. reported improved employment and social connections following MMT initiation [36], the current study’s results align regarding interpersonal relationships but differ in employment outcomes. This disparity may reflect broader socioeconomic constraints in Iran that limit job opportunities, particularly for people with a history of substance use [44]. This interpretation is consistent with Iranian analyses on employment stigma and screening practices in hiring, which document sustained barriers despite clinical stabilization [37]. In the DIC treatment group, prescribed methadone consumption increased at both the second and third measurements, while heroin and cannabis use declined. However, reductions were not consistent across other illicit substances: methamphetamine and illegal methadone did not change significantly, whereas cannabis declined significantly at both follow-ups. This pattern is compatible with the harm-reduction orientation of DICs, which prioritizes lowering risks from injecting and unsafe sex over complete abstinence. Within Iran, DICs are low-threshold gateways; variable counseling uptake and limited linkage to structured care may yield partial risk reduction without immediate abstinence. Iranian DIC program reports similarly highlight partial risk reductions alongside variable linkage to structured care, underscoring the value of low-threshold engagement in this context [15, 45, 46]. These proximal, harm-reduction gains function as recovery capital (e.g., service contact, safer-use knowledge, and warm hand-offs), which can scaffold later transitions to structured treatment and broader social stability. Together, these outcomes offer a pragmatic bridge between harm reduction and recovery by tracking proximal risk reduction alongside components of recovery capital. Many DIC clients are not yet committed to abstinence and may have limited uptake of therapeutic services (e.g., family counseling or psychotherapy). This modality also functions as a gateway to other treatment approaches. Embedding standardized readiness checks and warm hand-offs from DIC to MMT/TC—as operational protocols—could be tested as implementation strategies in Iran’s routine care, rather than inferred as explanations for effectiveness in our data [47–49]. Slesnick et al. emphasized that to enhance the effectiveness of DICs in reducing substance and alcohol use and improving mental health, comprehensive support is required, including financial resources, psychological services, residential facilities, and access to employment and educational opportunities [4]. The current study found that physical and psychological health problems, along with the number of days spent selling drugs and involved in fights or conflicts, decreased significantly in the DIC group. These findings are consistent with those of Sahebi et al., who also observed reductions in crime rates and improvements in health outcomes among DIC clients. The decline in criminal behavior was partly attributed to reduced heroin use and its neurological effects associated with aggression and psychological disturbances [50]. In the TC group, the second measurement (i.e., end of treatment, EOT) showed reductions in alcohol, opium, heroin, prescribed methadone, and methamphetamine, with a borderline decline in illegal methadone, alongside an increase in unemployment. At EOT, TCs showed short-term within-modality declines in several substances; however, given the residential containment nature of TC and non-random allocation, these patterns should not be interpreted as causal superiority over other modalities. In Iran, implementation heterogeneity and fidelity gaps across TC sites may shape outcomes, and standardized core components and post-discharge aftercare could be tested as implementation strategies in routine care, consistent with the continuing-care evidence base [51], rather than inferred as reasons for (in)effectiveness in our data. Given these reported gaps, routine monitoring of adherence to core TC components could clarify site-level variation in outcomes. This aligns with site-level assessments from Iran that emphasize fidelity monitoring and standardized components as prerequisites for durable gains [15]. Contextual implementation factors in Iran—such as resource constraints or variation in program components—have been reported and may help explain this pattern; however, our study was not designed to test these mechanisms [27, 52–54]. Potential cross-cultural differences warrant explicit consideration. Implementation features in Iranian TCs—shorter average length of stay, staffing/training mix, family-involvement norms, gender-segregated delivery, compulsory referral streams, and limited community aftercare—may moderate the short-term EOT signals observed here and limit generalizability beyond Iranian TC settings. These features can dampen psychosocial change during residence while accentuating short-term substance-use reductions at EOT. Variability in peer-leadership structures and vocational opportunities could also influence functional outcomes and relapse risk post-discharge. These contextual elements suggest that TC fidelity and outcome expectations should be calibrated to local conditions, with culturally adapted core components and standardized aftercare. Accordingly, TC findings should be interpreted as short-term end-of-treatment (EOT), within-modality signals conditioned on residential containment and local implementation, rather than comparative effectiveness [55–60]. Alternative explanations merit consideration: non-random selection into TC may reflect higher baseline motivation or family leverage; time-varying exposure (e.g., varying treatment intensity/length) may confound short-term reductions. Measurement limitations (self-report, differential attrition) and regression-to-the-mean could also inflate apparent early gains. Despite supportive global evidence for TC efficacy, transferability may be constrained by context—differences in length of stay, staffing, and post-discharge supports—particularly in Iran [15, 17, 42, 61]. A study by Sadeir et al. conducted in Kerman showed that those who completed TC treatment experienced better physical and psychological outcomes than those who dropped out. This was attributed to greater adherence to TC principles in previous years [27]. Additionally, research by Shaver et al. reported significant reductions in substance misuse, improvements in mental health, and positive behavioral changes [56] Similarly, Perfas et al. highlighted the importance of full program implementation and adherence to TC principles in determining success, noting reductions in alcohol and illicit drug use during treatment [62]. Maliwart et al., in a systematic review, also concluded that while TC participation consistently led to reductions in substance use, relapse often occurred after discharge [63]. In this observational, nonrandomized cohort, the Article 16 group showed within-modality declines at the second measurement (EOT) in opium, heroin, and illegal methadone use, as well as reductions in physical and psychological problems; however, these were short-term signals and long-term effectiveness remains uncertain. Estimates reflect short-term change during residence; durability after discharge cannot be inferred from these data. This concern may be magnified in Iran, where structured aftercare following compulsory episodes has historically been limited. Local evaluations have reported limited post-discharge continuity and elevated relapse risk in compulsory pathways [60, 64, 65]. Linking discharge from compulsory episodes to immediate OAT enrollment and scheduled follow-ups may mitigate rapid relapse observed locally. This should be viewed as a testable implementation strategy rather than an inference of effectiveness from our analyses. Nakhaee et al. found that compulsory detoxification centers were ineffective in maintaining abstinence, underscoring the need to shift toward evidence-based, voluntary alternatives. They identified the absence of post-treatment support as the main driver of relapse [16]. Similarly, studies by Jiang et al. [66] and Yang et al. [67] in China found high relapse rates among individuals in compulsory treatment—up to 36% within six months. A systematic review reported that 46% of individuals released from compulsory treatment relapsed within 1–6 months, with a further 10% relapsing within one year [68]. Long-term effectiveness cannot be adjudicated on the basis of three time points of follow-up data; nevertheless, alignment with prior evidence questioning post-discharge durability indicates that compulsory pathways may require robust aftercare to prevent rapid attenuation. While demand for rapid, visible results may shape their prioritization, our data and prior evidence point to the importance of post-treatment supports and voluntary, evidence-based pathways. Overlooking the short-lived nature of such interventions risks long-term harm, resource inefficiency, and growing mistrust among people who use drugs toward treatment programs. Within Iran’s mixed system—large OAT footprint, short-stay residential programs with work restrictions, variable TC fidelity, and historically limited aftercare in compulsory pathways—the observed patterns are likely shaped by both program design and implementation context rather than by modality effects per se. Given domain-specific gains, a stepped-care approach may be appropriate: (i) sustaining MMT while systematically bundling family counseling, vocational linkage, and sexual-health services; (ii) using DICs as intentional gateways with warm hand-offs to MMT/TC when readiness rises; (iii) ensuring TC programs include standardized components and planned aftercare; and (iv) prioritizing voluntary, evidence-based pathways with continuity of care after any compulsory episode. In practice, embedding vocational services and sexual-health packages into routine MMT, formalizing DIC-to-MMT/TC referral protocols, and standardizing TC components with planned aftercare are actionable levers for Iran. Taken together, these patterns suggest that functional domains (family functioning, health symptoms, and conflict reduction) are responsive even when employment lags, indicating scope for layered, strengths-focused packages. Structuring follow-up around recovery capital milestones (service retention, family engagement, vocational readiness) may yield durable improvements beyond abstinence alone. Future analyses should explicitly model potential effect modification (e.g., by gender, age, baseline severity) and implementation dose (fidelity, session counts, medication adherence) to disentangle program effects from residence and selection effects. Future evaluations should incorporate longer follow-ups and implementation metrics to distinguish residence effects from program effects. Including basic implementation indicators (coverage, fidelity, and aftercare contact rates) in Iranian evaluations would aid interpretation of effectiveness signals. Interpretations are constrained by non-random modality selection, unequal group sizes, reliance on self-report, differential attrition by time-point, and possible regression-to-the-mean; accordingly, effect estimates should be viewed as within-modality signals rather than comparative efficacy, and should be interpreted in light of Iran-specific service configurations and implementation realities. Conclusion Across modalities, effects were domain-specific rather than indicative of a single superior approach. Within our nonrandomized, within-modality data, at EOT the TC group showed short-term within-modality declines in several substances; MMT showed within-modality improvements in selected psychosocial indicators; DICs primarily delivered harm-reduction and linkage benefits consistent with their low-threshold mandate; and Article 16 showed short-term gains predominantly at EOT with uncertain durability beyond discharge. These are descriptive patterns, not comparative-effectiveness conclusions. These patterns are hypothesis-generating and suggest that, in future evaluations, matching clients to modalities by target outcomes—substance-use reduction versus stabilization and social functioning—may be associated with greater benefit; this requires testing. Strengthening post-treatment supports is a priority hypothesis to test, particularly for compulsory programs. Comparisons of employment outcomes should be interpreted cautiously because TC residents typically cannot work during residential treatment. Taken together, our findings support a stepped-care or blended-pathway approach in which DICs facilitate engagement, MMT addresses ongoing stabilization and family functioning, and TC is considered when intensive, abstinence-oriented change is the primary goal, with robust aftercare to sustain benefits. Future multi-province, prospective studies that track multi-domain outcomes over longer follow-up—and include basic implementation indicators (coverage, fidelity, aftercare contact rates)—can clarify durability and inform finer-grained matching of clients to modalities. These findings are observational and descriptive, not comparative effectiveness; they motivate testing of tailored post-discharge supports and stepped-care pathways in future controlled evaluations. Limitations, biases, and transferability Although these findings illuminate within-modality changes over time in routine-care settings in southeastern Iran, several limitations and potential biases warrant consideration. The study was conducted in a single province, and center selection partly depended on supervisor willingness; despite a multistage design (stratified random selection of MMT centers and consecutive enrollment at participating sites), operational constraints limit external validity, and selection bias cannot be excluded. Treatment allocation was nonrandomized; thus, confounding by indication and baseline imbalances (e.g., motivation, severity, socioeconomic factors) may persist despite consistent reporting of effect sizes and statistical significance. In the primary mixed-effects models, age and sex were not included as covariates. Although some residual confounding by age cannot be ruled out, this risk is mitigated by the within-person specification and short follow-up. Moreover, TC and Article 16 enrolled men only, which both limits the interpretability of any global sex adjustment across modalities and constrains the generalizability of results. Sample sizes were unequal across modalities—TC had 48 participants, below the a priori target of ≥ 100—reducing precision and power for subgroup contrasts. Follow-up was also unbalanced: MMT and DIC had three assessments (baseline, 2-month, 12-month), whereas TC and Article 16 had two time points, which may attenuate detection of later changes and complicate interpretation across modalities. Analyses were conducted within the 470-participant completer cohort using mixed-effects models that incorporated all available wave-level observations without imputation, implicitly assuming missing at random (MAR) for intermittent missingness; this remains a limitation and indicates that estimated trajectories reflect participants who remained engaged through the final assessment rather than all treatment initiators. Outcomes relied largely on retrospective self-report. Urine toxicology screens were obtained at each wave to support the validity of self-report, but toxicology data were collected as ancillary checks rather than primary endpoints and were not analyzed here; therefore, recall and social desirability bias cannot be excluded. The sample was predominantly male—TC and Article 16 served men only—and the small number of women precluded gender-stratified analyses, limiting generalizability to women and gender-diverse populations. Structural differences between residential (TC/Article 16) and outpatient (MMT/DIC) settings limit comparability for certain indicators (e.g., employment), as residents typically cannot work during residential treatment. Implementation fidelity (e.g., presence/quality of core TC components) was not assessed, leaving mechanisms—particularly for TC—uncertain. Multiple outcomes were tested without formal multiplicity correction; combined with unequal group sizes, this increases the risk of type I error and warrants cautious interpretation. Finally, despite use of a standardized, domain-specific reporting frame to avoid emphasizing any single modality, residual interpretive bias cannot be entirely excluded. Taken together, these constraints support cautious interpretation and underscore the need for prospective, multi-site studies with longer follow-up, gender-stratified analyses, fidelity assessments, and covariate-adjusted models to strengthen causal inference and enhance transferability. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements The authors would like to thank the Bam University of Medical Sciences for the financial support of this research. We also sincerely appreciate all those who participated in this study. Author contributions Overall design of the study: NO, SM, MHM, VYF. Data collection activities: NO, SM. Analysis, conceptualization of the manuscript: NO, SM, MHM. Writing the original manuscript draft: NO, NN. Review and editing of subsequent manuscript drafts: MHM, VYF, NN. All authors read the manuscript and approved the final version of the manuscript. Funding This paper was funded by Bam University of Medical Sciences. The Bam University of Medical Sciences did not have any role in the design of the study, collection, analysis, interpretation of data, or writing of the manuscript. Data availability The data that support the findings of this study are available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate This paper was approved by the Ethics Committee of Bam University of Medical Sciences, Iran (IR.MUBAM.REC.1404.005). All participants provided written informed consent. Authors are responsible for correctness of the statements provided in the manuscript. Authors have consented to the submission of the manuscript to the journal. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request.