A Systematic Review of the Efficacy of Contingency Management for Substance Use Disorders in Low and Middle Income Countries

Abstract

Background

The impact of illicit substance use is especially devastating in low-resourced countries where factors such as poverty, unemployment, and inadequate services impede successful treatment. Contingency management (CM) is a treatment for substance use disorders that has shown to be effective in eliciting behaviour change. The efficacy of CM interventions in low and middle income countries (LMICs) has been under explored.

Methods

The aim of this systematic review of randomized controlled trials was to assess measures of CM efficacy in addressing substance use disorders, while also considering contextual moderators of CM in LMICs. A search of PubMed, Scopus, and Cochrane library databases yielded 18 studies for inclusion, from which relevant data were extracted using modified versions of the Cochrane Characteristics of Studies tool.

Results

Two studies were located in a low-income country, two in lower-middle income countries, and fourteen in upper middle-income countries. Overall, estimated efficacy estimates were similar to those from higher income countries. However, context-specific challenges that warrant further investigation included limited access to trained staff and structural and financial constraints.

Conclusions

While CM in LMICs is in its early stages of development, efficacy estimates were not substantially different compared to high income countries. Challenges such as costs, willingness to implement, and the stigma associated with addiction sets the stage for further research in these contexts.

Introduction

Effective interventions are essential to addressing substance use disorders due to the social, economic, and personal consequences associated with them [1]. Extensive research has been conducted to assess the efficacy of Contingency Management (CM), a behavioural intervention aimed at facilitating abstinence from and reducing high-risk behaviour associated with substance use disorders [2–4]. While there is an abundance of CM studies in mostly high-income countries, there is little research available on the efficacy of CM in lower-resourced settings. Furthermore, implementation of CM interventions in low and middle income countries (LMICs) to reduce substance use may require adaptation based on the contextual factors specific to this setting.

CM is founded on the principles of operant conditioning [5]. These principles suggest that behaviour has a greater chance of being repeated if followed by a positive consequence, while behaviour with negative consequences will likely diminish. CM interventions provide incentives for a specific behaviour (e.g., biologically confirmed drug abstinence), with this incentive withheld if the behaviour was not elicited (e.g., confirmed substance use). CM has been applied in a variety of settings; examples of specific goals include: improving adherence to medication [6], reducing substance use [7], and eliciting behavioural changes in interventions with children [8].

The objectives of this systematic review were to, (1) explore the efficacy of CM interventions that have been used to address substance use disorders in LMICs and, (2) identify possible contextual moderators which may influence the efficacy of these interventions. This review evaluated randomized controlled trials and other non-randomized experimental designs. To the best of our knowledge, this paper is the first review of CM for the treatment of substance use disorders in LMICs. By describing the efficacy of CM interventions for substance use disorders that have been implemented in these contexts, we hope to aid intervention design in lower-resourced countries.

Methods

Prior to completing this review, a protocol describing the methodology to be employed was completed, and the study was registered with the International Prospective Register of Systematic Reviews for publication (PROSPERO; https://www.crd.york.ac.uk/prospero/; Registration #: CRD42018106102). The database searches, selection of studies, assessment for eligibility, and extraction of relevant data were conducted independently by two researchers with extensive experience in the CM field to ensure reliability and consistency of the search results.

Given the small number of included studies, the heterogeneity of the outcomes assessed, and the variability of methods and clinical populations a meta-analysis was deemed not feasible. CM interventions for the treatment of substance use disorders in LMICs have taken many forms and used diverse methodologies; therefore, we utilized a systematic review to gather, combine, and appraise the research. The following questions were posed to gain an understanding of how CM, or similar behaviour interventions that provide a reinforcement for desired behavioural change, was used to treat substance use disorder in LMICs:

1. What were the efficacy outcomes of these interventions?

2. What were the characteristics of CM or similar interventions in low resourced countries?

3. What strategies were employed to counter any challenges that arose when implementing CM interventions in an LMIC context?

Eligibility Criteria

Selection of studies was based on specific inclusion and exclusion criteria and was limited to studies published through December 2022. This review primarily considered studies utilizing a randomised controlled trial (RCT) design. Due to their rigorous design, studies incorporating RCTs are considered to contribute high quality and reliable evidence [9, 10], however including only RCTs could result in skewed recommendations [11]. Therefore, selected studies had to employ either an RCT or non-randomized experimental study design with the primary outcome related to reduction of substance/alcohol/tobacco use in order to be considered for inclusion in this review. Selected studies must have been peer-reviewed and published for which it was possible to retrieve full text articles. Only studies with human adult participants, aged 18 and above, were considered for inclusion. Other inclusion criteria were that study participants must have been seeking or receiving treatment for substance/alcohol/tobacco use, and that the CM studies were located in LMICs. Studies were excluded if they did not include an experimental research design (e.g., observational cohort, case-control, cross-sectional survey, acceptability and feasibility study, review, commentary, or protocol description), were animal studies, did not include a CM intervention, did not include a primary outcome related to reduction of substance use, were not conducted in a LMIC, were not published in a peer-reviewed journal (e.g., conference papers, theses or dissertations, and reports), were published after December 2022, or had insufficient data for consideration.

Search Strategy

For this review, we used the following three databases to search for relevant studies published through December 2022: (1) PubMed; (2) SCOPUS; (3) The Cochrane library databases. An initial broad search was carried out to gauge the breadth of the literature available. The search strategy was subsequently adjusted, on the basis of common terms within the initial papers, in order to garner more relevant papers while minimizing the number of irrelevant papers [12]. We used search terms related to CM, substance use disorders, and names of LMICs. The final search terms, using Boolean operators, can be found in Appendix A.

To describe CM or related interventions we used terms related to interventions that provided a reward for desired behaviour. This review used Substance Use Disorder (SUD) as defined by the DSM-5, in addition to using substance classes and commonly used substance names [13]. Specific country names were included in the search queries in order to identify all relevant LMICs, as classified by the World Bank [14].

Selection of Studies

The study selection followed a three-step approach. First, databases were searched. The searches of the databases retrieved 4,125 papers (see Fig. 1). Second, the screening process consisted of a) screening titles and abstracts and selecting potential studies for full text screening, then b) screening full text articles for potentially eligible papers by considering the eligibility criteria. After screening titles and abstracts, 322 papers were retained (Pubmed n = 72; Scopus n = 217; Cochrane = 33). After full text screening, 27 papers were identified as potentially eligible. Studies included from reviews and reference lists were also perused which yielded an additional seven papers. Thirdly, a more detailed and in-depth paper assessment was completed to assess whether the 34 papers met the eligibility criteria. Sixteen papers were excluded as they did not meet all the criteria, leaving 18 papers retained for inclusion in the review (Fig. 1). All papers were published in English.

Fig. 1.

Flow Diagram of Study Selection Process

Data Extraction

A critical part of systematic reviews is assessing what the study identifies as an efficacious outcome and extracting essential data. Data extraction was performed using a modified version of the Cochrane Data Collection Form. Outcome data used to determine efficacy included primary and secondary outcomes as specified in the selected studies. These outcomes included abstinence or substance use reduction, treatment attendance, treatment retention, compliance with treatment, and reduction of cravings.

Result

Efficacy Outcome of Interventions

Inclusion of CM in a substance use disorder intervention was reported as having beneficial effects in at least one outcome for 17 of the 18 studies included in this review.

Hser et al. found there were a significantly larger number of opioid negative urine samples among persons in the intervention group in Shanghai when compared to persons receiving treatment as usual (TAU) group (74% vs 68% respectively, p < 0.05) [15]. A reduction in positive opioid urine samples in the intervention group was reported in Kunming as well (27% IG vs 18% TAU), although the significance level was not reported [15]. Otiashvili et al. found that of the population of men injecting opioids, persons in the comprehensive intervention group had a significantly greater number of negative urine samples during the intervention period compared to persons in the usual care group (mean of 7.0 versus 1.4 respectively; p < 0.001) [16]. In Chen et al., the authors reported that there were higher rates of abstinence for persons in the intervention group as evident in a larger percentage of opioid-negative urine samples when compared to persons in the standard treatment group (68% vs 58% respectively; p < 0.001); there was also a greater difference between groups in rural compared to urban settings (68% vs 53%; p < 0.001) [17]. In Jiang et al. compliance to methadone maintenance treatment was assessed as the number of negative urine samples (mean of 7.9 for the intervention group versus 7.6 for the standard treatment group) [28]. The authors therefore concluded that adding CM to standard treatment had no significant effect on retention and compliance.

In Chawarski et al., the study was designed for female sex workers who use stimulants [18]. Persons in the enhanced services group, which included counselling and CM components, had a greater proportion of negative urine samples (87% vs. 69%, p = 0.04) and longer duration of abstinence (10.3 vs. 7.8 weeks) compared to persons in the standard services group. Carrico et al. assessed treatment retention associated with an intervention to decrease stimulant use and risky behaviour in female sex workers [19]. The study found that eligible participants attended a median of 25 of the 36 urine screenings and that 68% of eligible participants attended three or more of the cognitive behavioural aftercare sessions [19]. Feasibility of the conditional cash transfer (CCT) intervention was set at participant enrolment of 70% or more. The study found that 75% of participants enrolled in the intervention [19]. The authors therefore concluded that the specific intervention implemented was feasible. In a follow-up study among women in this same population, participants receiving CM had a 60% lower odds of having amphetamine-positive urine screenings at six months, but these results did not persist at 12 and 18 month follow-up time points [20].

In a population of individuals with crack-cocaine use disorder, Miguel et al., 2016 found that when comparing a standard treatment plus CM group to a standard treatment only group, there was a significant difference between the groups in mean attendance (7.7 weeks vs. 3.0 weeks, respectively; p < 0.01) [21]. Following this trial, a crossover trial in which participants randomized to usual care in the initial trial were subsequently offered CM showed that CM resulted in a higher percentage of cocaine-negative urine samples (45% vs. 12%; t = 3.21; p < 0.01) [22]. Similarly, a follow-up study in Brazil confirmed these findings with those receiving CM demonstrating a 20% higher mean percentage of negative cocaine urines submitted, and were retained in treatment on average for three weeks longer compared to individuals receiving usual care [23]. A small pilot study conducted among persons experiencing homelessness in Brazil additionally showed positive effects of CM on treatment retention and cocaine abstinence; however, effects were not statistically significant due to small sample size [24].

In Pirnia et al., 2016 the control group had a mean of 19.7 cocaine-negative tests over the 12-week period compared to 15.4 for the CM group [25]. However, there is an apparent discrepancy in the study report, with the authors concluding that CM had a “productive effect on increasing the rate of negative urine tests”. At the time of writing, further requests to the authors for clarification have not been successful. In another study addressing cocaine use disorder, Pirnia et al., 2017 found that when comparing the means of pre and post cocaine craving test results between groups, Topiramate (14.84 vs 12.04), CM (13.90 vs 13.89), and Topiramate plus CM (15.81 vs. 10.92) resulted in significant decrease in cocaine craving (p = 0.01), with the largest effect in the Topiramate plus CM group (p < 0.01) [26]. In a third study addressing cocaine and opioid use during Methadone Maintenance Treatment, Pirnia et al., 2018 found that neither the cash-based reinforcement therapy group nor the voucher-based reinforcement therapy group reported a significant effect on abstinence, with a mean number of negative cocaine tests of 13.6 vs. 14.3 in the reinforcement groups, respectively, and a mean of 15.1 in the control group [27].

Giné et al. employed a smoking cessation intervention in the Philippines that randomized individuals to three groups: the contingency management group, a group that received cue cards as motivation, and a control group that received no intervention (n = 616) [29]. The results of the study indicated that at 6 months, those receiving contingency management were 3%–6% more likely to be abstinent (p = 0.709), and at 12 months 4%–6% (p = 0.515) when compared to the other groups; however, these results were not statistically significant [29]. In another smoking cessation study, White et al. found a significant effect on abstinence in the experimental group with 46% maintaining abstinence when compared to the 15% of the control group (p < 0.003) [30]. White et al. found that the intervention performed better when calculating “cost per quitter” of $281, which with minor adjustments, can be effectively employed at a cost of $195 per person [30]. In Hofmeyr et al., CM was found to be effective in reducing smoking among university students by the end of the treatment period (45% of treatment subjects were abstinent compared to only 6% of control subjects (p < 0.001); but these effects were diminished at the end of the 6-month follow-up period [31].

In the only study assessing the effect of CM on alcohol abstinence, Jirapramukpitak et al. found that although CM did not have a significant effect on continuous alcohol abstinence, those in the high CM group did have a lower number of alcohol-positive breath samples and had higher abstinence rates during the follow-up period [32].

Characteristics of Studies

A reasonable degree of consistency was observed in study and sample characteristics for the included studies (see Table 1). The majority (14/18) of the studies were located in upper-middle income countries: Malaysia [18], China [15, 17, 28], Brazil [21–24], Islamic Republic of Iran [25–27], Thailand [30, 32], and South Africa [31]. Two of the included studies were located in Cambodia, a country classified as a low-income country [19, 20], while two were located in lower-middle income countries: the Philippines [29], and The Republic of Georgia [16].

Table 1.

Characteristics of Included Studies

Study / Country (Classification)	Setting	Substance	N / Groups	Contingency	Outcomes	Duration / Follow Up (FU)	Findings	Cost	Limitations
Chawarski et al. [18] / Malaysia (UMIC)	Community-based outpatient center	Opioids	24/ST (12) CM (12)	Take home doses of Bup	Reduction of illicit drug use & high risk sexual behaviour	12 weeks / no FU	Significant reduction in opioid use in CM; insignificant reduction in high risk sexual behaviour; excellent retention in both groups	Not specified	Small sample size; intervention composed of multiple components including increased social contact; self-report for buprenorphine abuse not reliable
Giné et al. [29] / Philippines (LMIC)	Field sites	Nicotine	2000/ST (616) Cue cards (603) CM (781)	Personal cash deposits	Smoking cessation; demand for CARES	12 months / no FU	Those with CARES more likely to abstain at 6 months (3.3–5.8%) and 12 months (3.5–5.7%); majority of CARES clients failed to quit	Cost / quit $700	Few clients were willing to sign up for CARES; different randomization procedures between the 3 waves
Hser et al. [15] / China (UMIC)	MMT outpatient clinic	Opioids	320 / ST (160) CM (160)	Prize draw	Retention in MMT programme; abstinence	12 weeks / FU 1, 3, 6 months post study	Site differences with unusually high retention rates in Shanghai created ceiling effect (88% retention); % of negative urine samples in Shanghai 74% CM vs 68% ST and in Kunming 27% CM vs 18% ST	Average cost of incentive $55 / person	Site differences: Shanghai setting included social worker support which could have promoted overall outcomes; Shanghai’s higher economic status may have reduced attractiveness of incentives
Jiang et al. [28] / China (UMIC)	MMT outpatient clinic	Heroin / Opioids	160 / ST (80) CM (80)	Prize draw	Retention; abstinence	12 weeks / FU at 16, 20, 24 weeks post baseline	CM ineffective at improving retention and abstinence due to unusually high rates of retention (86%) and weak incentive	Cost of MMT max. $48 / month / person; average value of prizes $83.60 / person	Small sample size; financial incentives not sufficient to motivate behaviour change given higher resourced Shanghai
Otiashvili et al. [16] / Republic of Georgia (LMIC)	Field site	Opioids	40 / ST (20) CM (20)	Cash	Treatment impact; reduction in HIV drug risk behaviours	12 weeks / FU at 1, 3, and 6 months	CM more likely to enter detoxification and Naltrexone treatment, higher rate of negative urine samples; significant reduction in benzodiazepine & buprenorphine in CM	Earning potential $194 in CM group	Small sample size; sample not based on a power analysis; single counsellor providing service new to MI technique; sizeable attrition occurred during the study
Chen et al. [17] / China (UMIC)	MMT outpatient clinic	Heroin / Opioids	246/ST (120) CM (126)	Prize draw	Abstinence; adherence to treatment	12 weeks / no FU	CM improved treatment attendance & drug abstinence	Not specified	Site differences due to improved services and better access in the urban areas may have contributed to difference between groups
White et al. [30] / Thailand (UMIC)	Field sites	Nicotine	201 / ST (69) CM (132)	Personal cash deposits	Smoking cessation	12 weeks / FU 14 months post baseline	CM increased likelihood of quitting smoking at 6 months by 91% – 136%	Cost of $195 / quitter	Convenience sample therefore likely not an accurate reflection of the population; unclear which of the multiple elements included in intervention contributed most to effect; self-report on cessation at the 14 month FU not an objective measure of abstinence
Carrico et al. [19] / Cambodia (LIC)	Field sites	Amphet-amine type stimulants (ATS)	158 / All persons received intervention	Conditional Cash Transfer	Feasibility of intervention implementation	16 weeks in total: 12 weeks CCT + 4 weeks after care / no FU	Median of 25 (IQR: 9–32) of the 36 urine screening visits completed; 68% completed 3 or more aftercare sessions	Max value of CCT: $120	Shortened cognitive behavioural model may not be sufficient to affect behaviour change; reinforcers may not be sufficient to offset the loss of earnings for the female entertainment sex workers
Miguel et al., 2016 [21] / Brazil (UMIC)	Outpatient addiction clinic	Crack-cocaine	65/ST (32) CM (33)	Vouchers	Abstinence	12 weeks / FU at 3 and 6 months	Greater retention in treatment and abstinence in crack cocaine, THC, and alcohol for CM	Max value of $235; Average cost of $29 / person	Many missing urine test results, especially in ST; few patients completed FU
Pirnia et al., 2016 [25] / Iran (UMIC)	Outpatient addiction clinic	Cocaine	50/ST (25) CM (25)	Vouchers	Abstinence; reduction of cocaine craving	12 weeks / no FU	CM improved abstinence; CM helped reduce cocaine cravings	Max value of $550; average cost of $22 / person	Study sample not representative of general population; no attendance records; use of self-reports impacts objectivity of results
Pirnia et al., 2017 [26] / Iran (UMIC)	Outpatient addiction clinic	Cocaine	100 / ST (25) TP (25) CM (25) TP + CM (25)	Vouchers	Abstinence; reduction of cocaine craving	12 weeks / no FU	Significant reduction in craving in TP + CM	Max value of $550	Study sample not representative of general population; use of self- reports impacts the objectivity of results
Pirnia et al., 2018 [27] / Iran (UMIC)	Outpatient addiction clinic	Cocaine; Opioids	75/ST (25) VBRT(25) CBRT (25)	Vouchers; Cash	Abstinence; reduction of cocaine craving	12 weeks / no FU	VBRT and CBRT effective at reducing cocaine cravings; neither VBRT and CBRT increased abstinence	Not specified	Study sample not representative of general population; cash not provided immediately following the target behaviour, which likely minimized the power of the reinforcer
Miguel et al., 2019 [22] Brazil (UMIC)	Outpatient addiction clinic	Crack-cocaine	16 / all received Usual care followed by CM after 6 months	Vouchers	Treatment retention; abstinence	12 weeks of CM 6 months after initial usual care	Greater retention in treatment and abstinence of crack cocaine use during CM treatment vs usual care	Max value of CM: $225	Small sample size; high attrition from initial UC treatment group; study sample not representative of general population
Page et al. [20] Cambodia (LIC)	Field sites	Amphet-amine type stimulants (ATS)	1198 / CCT (281)	Conditional Cash Transfer	Sexual risk ATS Tox +urine results Severity of ATS use Binge ATS use	12 weeks / FU assessed at 6, 12, and 18 months	60% lower odds (statistically significant) of being ATS Tox + at 6-months, and continued but non-significant reductions at 12- and 18-months	Max value of CCT: $120	Multi-level intervention design makes disentangling which components are responsible for effect challenging
Hofmeyr et al. [31] South Africa (UMIC)	University	Nicotine	87/CM (40)	Cash	Smoking cessation at end of intervention period and a 6 months; repeated measure of smoking intensity of non-abstinent subjects	6 weeks / 6 months FU	CM effective at reducing smoking at end of intervention period, but not at 6 months; no effect of CM on smoking intensity	Max value of CM: $95	Small sample size; use of CO to measure tobacco exposure limited to recent time period
Jirapramukpitak et al. [32] Thailand (UMIC)	Catchment communities surrounding university hospital	Alcohol	161/CM-L (42) CM-H (37)	Cash	Continuous abstinence over 12 weeks and during FU; number of positive breath samples over 12 weeks	12 weeks / FU at 16 weeks	No effect of CM on continuous abstinence over 12 week intervention period, but CM-H group had higher abstinence during FU; reduction in number of positive breath samples submitted in CM-H group	Average of $68 in CM-H	Use of BAC to measure alcohol exposure limited to recent time period
Miguel et al., 2022a [23], Int J Drug Policy Brazil (UMIC)	Outpatient addiction clinic	Crack cocaine	98/CM (50)	Vouchers	Abstinence	12 weeks / FU at 3 months	Compared to usual care, the CM group was significantly more likely to submit a negative cocaine urinalysis during treatment	Max value of CM: $185	High attrition during follow-up period; study sample not representative of general population
Miguel et al., 2022b [24], Exp Clin Psy-chopharmacol Brazil (UMIC)	Abstinent-Contingent Housing treatment program	Crack cocaine	21 / CM (12)	Vouchers	Abstinence and treatment retention	12 weeks	Compared to the usual care, CM group had consistently better treatment retention and cocaine use outcome measures, but were nonsignificant due to small sample size		Small sample size; study sample not representative of general population

World Bank Classification: LIC low income country, LMIC lower middle income country, UMIC upper middle income country, ST standard treatment/ control group, CM contingency management/intervention group, FU follow up, $ cost labelled in USD

Most studies were reportedly located in urban areas, with the exception of Carrico et al. and White et al., whose participants were located in rural areas [15, 19, 30]. Most interventions took place in an outpatient setting, either at clinics or at field sites (research site associated with a particular organisation, typically outside of its central offices), and at times due to difficulty with participants accessing study locations, accommodations were made to assist them. Hofmeyr et al. was conducted in a university setting [31]. Carrico et al. provided off site screening if required [19]. In White et al., the study took place at participants’ homes if they were unable to reach the study locations [30]. Jirapramukpitak et al. employed a home-based approach to deliver the CM intervention to individuals with alcohol use disorder [32]. A small pilot study among persons experiencing homelessness in Brazil used an abstinent-contingent housing treatment program as the study setting [24].

Across all the included studies, recruited participants mainly sought treatment for their substance use disorders, with the exception of participants in Giné et al., who were invited to participate regardless of their intention to stop smoking [29]. Interventions targeted treatment of opioid use disorder [15–18, 28], stimulant use disorder [19–27] tobacco use disorder [29–31], and alcohol use disorder [32]. Abstinence or substance use reduction was outlined as an outcome measure in the majority of the studies [15, 17, 21–32]. Other outcome measures included programme retention to outpatient methadone maintenance treatment and adherence to methadone [15, 17, 28]; adherence to buprenorphine [18]; reduction of cravings [25–27]; reduction of high-risk behaviour associated with substance use [16, 18, 20]; and feasibility of adapting and implementing an intervention in a challenging population [19, 24].

Reinforcements were provided for the following target behaviours: substance-negative urine samples, a decrease in high-risk substance use and sexual behaviours, treatment attendance, and adherence to medication. Prize draws were utilized in studies located in China. Treatment cost can be very high in China; therefore, participants used their prizes to offset treatment cost [17, 28]. While prizes were also provided for drug negative urine samples, Hser et al. did not specify how the prizes were utilized [15]. Vouchers were often provided as an alternative to cash. In all of the studies conducted in Brazil the vouchers earned could be used to exchange for goods or to pay bills [21–24]. Likewise, vouchers earned in the Pirnia et al., 2017 study could be exchanged for goods or services [26]. In Pirnia et al., 2018 the efficacy of cash reinforcement was compared to that of vouchers [27].

Six studies used cash as reinforcement [16, 19, 20, 26, 31, 32]. Given that the population of female sex workers targeted in Carrico et al. were reimbursed by their clients with cash, cash was deemed an appropriate reinforcement to award desired behaviour as part of the CM intervention reported in that study [19]. In Chawarski et al., take-home doses of buprenorphine was deemed as an attractive reinforcement for abstinence [18]. Most studies used an escalating form of reinforcement, with rewards increasing with consecutive negative substance tests. In Otiashvili et al. [16], the weekly possible reward was consistent and did not increase with consecutive negative testing. Giné et al. and White et al. utilized fixed-schedule cash deposits [29, 30]; participants contracted to deposit consistent amounts of their own money into an account, to which the study added bonuses when specific treatment goals were reached. The deposits were refunded at the end of the study if participants were able to meet these goals.

In none of the included studies was CM used as a stand-alone intervention, but rather used as adjunct to standard treatment, or combined with complimentary treatment approaches. In Carrico et al., a brief cognitive-behavioural model was added to conditional cash transfer to boost intervention efficacy [19]. Chawarski et al. provided “behavioural drug and risk reduction counselling” for the enhanced services group [18]. In Jiang et al. and Otiashvili et al., motivational interviewing was provided in addition to CM, with Otiashvili et al. also providing the option for a 14-day detox if the participants of the experimental group requested it [16, 28]. Topiramite, a pharmaceutical treatment, and brief behavioural compliance enhanced treatment were adjuncts to standard treatment in Pirnia et al., 2017 [26].

Eleven studies employed follow-up assessment post intervention. Jirapramukpitak et al., the sole study targeting alcohol use abstinence, had a 1-month follow-up period [32]. Giné et al. reported a 12-month surprise follow-up [29], while Hser et al., Jiang et al., and Otiashvili et al. conducted follow-up testing at 1, 3 and 6 months post intervention [15, 16, 28]. Both White et al. and Hofmeyr et al. conducted follow-up testing at 6 [30, 31]. Three out of the four studies conducted in Brazil reported attempting follow-ups at 3 and 6 months post-intervention, but were not able to interpret the results meaningfully given the sizeable attrition that occurred [21–23]. Page et al. [20] had the longest follow-up at 12 and 18 months.

Adaptions of CM Intervention to LMIC Context

While many of the contingency management interventions employed in these LMICs resembled those of high income countries (HICs), some studies chose to adapt the intervention based on context and population of interest. Fifteen of the 18 studies employed a 12-week intervention period, which was based on CM interventions originating predominantly in North America. However, Otiashvili et al. utilized a 22-week intervention period, Giné et al. a 24-month intervention period, and Hofmeyr et al. a 6-week intervention period [16, 29, 31]. Giné et al. [29] and White et al. [30] utilised self-deposits as reinforcement for abstinence but, given the potential challenges with accessibility from more rural areas, some participants were offered a deposit collection service as a means of addressing these possible barriers.

Given the cost of MMT in China that is carried by the individuals accessing treatment, participants were able to offset some of the cost of treatment with the earnings from their prizes [17, 28], which may be favourable in terms of uptake in CM interventions among populations in LMICs. In Chen et al. [17] and Pirnia et al., 2016 [25] attempts were made to lower intervention costs by reducing the frequency of urine tests thus placing less pressure on the staff in implementing the intervention. By reducing the deposit period from 6 to 3 months, White et al. [30] were able to reduce the reinforcer-related cost. An important element of interventions is the staff required to implement it. White et al. [30] attempted to address this by utilising the staff resources available in the intervention setting, and training local community health workers to implement the intervention, which provided an added advantage of familiarity to the participants. Jirapramukpitak et al. employed a home-based intervention provided by community health workers, an approach that may be more feasibly delivered and scaled while still providing beneficial reductions in alcohol or substance use [32].

Discussion

In these 18 studies reviewed, CM in LMIC settings proved efficacious in addressing substance use disorders. There were improved substance use outcomes across the various substance classes: opioids [15–18, 28], stimulants [19–27], tobacco [29–31], and alcohol [32]. In terms of abstinence or reduction in substance use, significant treatment effects were found in the intervention groups receiving a CM component [15–18, 20–23, 25, 29–32], and these reflected the findings of CM studies in HICs [33–34]. Hser et al. found significant site differences in terms of abstinence and reported that better quality services may be associated with improved outcomes [15]. This was demonstrated by patients in the CM group situated in the higher resourced setting of Shanghai (as compared to the lower resourced rural areas) achieving significantly improved abstinence and retention [15]. In Miguel et al., 2016 while abstinence was the primary outcome, the authors found a knock-on beneficial effect also on programme attendance and retention, which demonstrates that improvement in one area can lead to improvement in other areas of functioning [21]. CM was also shown by Carrico et al. and Page et al. to be efficacious in reducing high-risk sexual behaviour, which is often associated with substance use disorders [19, 20]

The CM interventions in LMICs were similar to those in HICs such as 12-week intervention durations, addressing similar substance categories (opioid, stimulant, tobacco, and alcohol), and utilizing similar reinforcements (vouchers, prize draws, cash, take home doses and self-deposits). The co-interventions, structure of the groups, and challenges unique to LMICs (e.g. resources), highlighted differences between CM studies in HICs and LMICs. Outpatient clinic settings were considered ideal locations for the LMIC interventions given that they provided an existing infrastructure, which included staff and potential participants. Sufficient heterogeneity was observed for the methodological and sample characteristics of the included studies to rule out use of a meta-analysis to rate reinforcement approaches in this review in terms of efficacy.

It is difficult to make recommendations regarding the optimal magnitude of reinforcers in CM interventions. This depends largely on the specific settings of the interventions; feasibility studies would be useful in determining whether different settings are more suited to particular value reinforcers. Indeed, Jiang et al. may not have been able to reproduce positive outcomes of CM typically reported in studies conducted in HICs (apart from unusually high levels of retention and compliance) because the reinforcers were not of sufficient value to elicit significant change in this population [28]. This echoes the findings of Petry et al. where lower value reinforcers resulted in treatment effects similar to that of the control group, compared to higher value reinforcers that yielded significant changes in a cocaine-dependent population [35].

Given that the vast majority of CM trials reflect HIC circumstances, aspects of the intervention can be creatively adapted to the LMIC context. In an attempt to address some of the issues around intervention costs, patients were allowed to use self-deposits in White et al. and Giné et al. [29, 30]. In terms of this method of reinforcement, the outcomes in these two studies differed. In White et al., the authors indicated that having a shorter deposit period of three months, versus the six months used in Giné et al., may have contributed to its success [29, 30]. Depositing one’s own money may increase motivation to meet the target behaviour and contribute to reducing the cost of running the intervention [36]. Alternatively, the use of self-deposits may provide a barrier to accessing treatment in poorly-resourced settings and may be the key reason why, despite having the largest sample of the included studies, Giné et al. reported a low uptake level of the intervention by participants [29].

Low-resourced countries often exhibit large treatment gaps between those requiring treatment and those who are able to receive it. Attempts were made in the studies included in this review to bridge treatment gaps by extending interventions to rural areas [19, 30]. In White et al., community health workers that were part of an existing treatment structure attended to participants at their homes if necessary; thereby lowering potential challenges of attendance which could impact the treatment efficacy [30]. Another way of adapting CM interventions to LMIC contexts utilizes available resources, such as staff. White et al. and Jirapramukpitak et al. utilised local community health workers already present in the study setting instead of having to employ and train field staff [30, 32]. Chen et al. found that by implementing behavioural drug and risk reduction counselling, there was no need to train or employ specialised staff as the intervention could easily be administered by existing staff [17].

Limitations in the studies included in this review provide important information about areas that need improvement for future interventions. CM interventions are frequently combined with other treatment elements, as was the case for all studies included in this review. This has the knock-on effect of making it difficult to understand which aspect of treatment contributed most to the impact of the intervention.

As delineated by Skinner, an essential element in reinforcement is the immediate provision of a consequence following the target behaviour [5]. When comparing cash and voucher reinforcements, Pirnia et al., 2016 were not able to provide the cash incentive immediately after the target behaviour was met [25], thereby most likely losing the strong reinforcing value of cash that previous studies have reported [37]. In addition, low power to detect effects of the interventions due to small sample size was identified as a limitation in several studies [15, 16, 22, 24, 28, 31].

A number of the included studies did not employ follow-up testing after the intervention had ended, focusing solely on the duration of the intervention. Follow-up assessments should be an integral part of these interventions to have a more realistic understanding of the efficacy of these interventions beyond the intervention period. Hser et al. reported a decrease in treatment efficacy at six months post intervention in the group receiving the active intervention [15]. This is similar to what has been reported in HIC studies [38–40]. Another shortcoming was that not all of the included studies reported effect sizes, which can provide important insight into the likely clinical effectiveness of CM, particularly for low-powered studies employing small sample sizes. As such, effect sizes should be included in future trials of CM and other interventions for SUDs.

This study employed three comprehensive research databases to identify controlled clinical trials of CM in LMICs. However, the eligibility criteria were narrow such that some studies with relevant findings were excluded from inclusion because they did not assess efficacy of the CM intervention. For instance, two studies evaluated outcomes other than efficacy of CM on substance use reduction or abstinence among participants of the original trial conducted in Sao Paulo, Brazil [21]. In a survey study of this same population, CM was found to be acceptable with the majority of participants endorsing it as a helpful intervention to reduce their cocaine use [41]. A secondary analysis of the originally included study found that the CM intervention was associated with lower levels of depressive and anxiety symptoms compared to usual care [42]. These studies offer evidence of important benefits of CM beyond that of traditional efficacy.

Future work should evaluate subjective value of reinforcers used in CM trials and investigate what participants spend their reinforcers on to optimize intervention design. CM in LMICs is in its early stages of development. Challenges such as intervention costs, willingness to implement this effective intervention in LMICs, and the stigma associated with addiction sets the stage for the next round of research in these contexts. In sum, this review found that CM utilised to address substance use disorders in LMICs had parallel outcomes to those from HICs when they had access to sufficient resources, providing support to the use of this intervention in a range of settings