Summary of Identified Studies
The search strategy identified a total of 977 articles to review across the three databases (see the PRISMA [Preferred Reporting Items for Systematic Reviews and Meta-Analyses] diagram in Fig. 1). No additional studies were found through the grey literature search. Of the 977 articles, 959 were excluded because they failed to meet the inclusion criteria or met the exclusion criteria, leaving 18 studies for full-text review. Six narrative reviews were identified and excluded because they did not present or assess economic evaluations or considerations. One article was excluded because it provided a cost listing for a take-home naloxone program without an economic evaluation. One article was excluded because it was a minor correction to another article already included and the correction did not change the findings of the original article. One article was excluded because it was an abstract from a conference presentation of a separate study for which no further information could be found, even after contacting the corresponding author. A total of nine studies were included in the review. Appendix C in the ESM provides a list of all articles that were assessed in the full-text review, with an explanation for their exclusion.
The included studies consisted of one CEA, eight CUAs, and one CBA (ten analyses from nine studies—one study conducted both a CBA and a CEA). Studies covered a variety of settings, including the USA (n = 5), Canada (n = 1), Russia (n = 1), Scotland (n = 1), and the UK (n = 1), where five were focused at the national level and four on specific cities. Articles were published from 2013 to 2020, and all declared funding except for two, which had no conflicts of interest to disclose. All studies included community distribution of naloxone for lay administration as a component of their intervention, though some articles differed on specific populations targeted. Seven studies looked at users or hypothetical users, one study looked at non-users, and one study looked at counties with at least five opioid overdose deaths each year. Eight studies used the comparator of no distribution, with one using an additional comparator of pre-exposure prophylaxis, and one study comparing a combination of distribution to laypeople, police and fire, and EMS. All studies used a Markov and/or decision-analytic model, apart from one study that was trial based. Five studies used a societal perspective, three studies used a healthcare perspective, and one study used both a societal and a healthcare perspective. Six studies used a lifetime time horizon, one study used a 20-year time horizon, and two studies did not specify a time horizon. After analysis, all studies found positive benefits related to health, where community distribution of naloxone either prevented or reduced the number of overdose-related deaths. Results were rated as worthwhile on the basis of each article’s willingness-to-pay (WTP) threshold. If no WTP threshold was stated, we used a conservative WTP threshold of $US50,000.
Table 1 provides a summary of the results from the data extraction.
Table 1 Summary of included studies [20,21,22,23,24,25,26,27,28] Cost-Utility Analyses
The study conducted in the USA by Townsend et al. [20] observed that community distribution of naloxone was only worthwhile at a WTP threshold of $US50,000 (year 2017 values; $US53,000 in 2020) and if the kits cost less than $US2200 (year 2017 values; $US2332 in 2020) and was the most worthwhile when community distribution was combined with high EMS distribution and low police officer and firefighter distribution; returning an ICUR of $US12,880–15,950 (year 2017 values; $US13,568–16,907 in 2020) per QALY gained. Further, 5% more overdose deaths were prevented with high EMS distribution and low police officer and firefighter distribution compared with overdose death rates with low distribution in all three groups [20]. Both deterministic and probabilistic sensitivity analyses were performed, which considered ranges for the price of naloxone, the percentage of people who intervened in overdose, a hypothetical moral hazard, and the rates of distribution [20]. Although the rate of distribution had the largest impact, community distribution still proved worthwhile throughout the range considered [20]. Therefore, none of the evaluated variables changed the conclusion [20]. This study did not report justification for the form of economic evaluation or the quantities of resources separately reported from their unit cost. This study was classified as high quality.
The study by Langham et al. [21] was set in the UK and found that, although the intervention increased overdoses by 2.7%, it reduced overdose death by 6.6% and increased lifetime QALYs by 0.164, with an ICUR of £899 (year 2016 values; $US1312 in 2020) per QALY gained. They determined naloxone distribution to be worthwhile, assuming a WTP threshold of £20,000 (year 2016 values; $US29,189 in 2020) [21]. In their probabilistic sensitivity analysis, they considered ranges for the price of naloxone, additional societal costs, rates of distribution, and witness to overdose [21]. However, none of the variables had a substantive impact so did not change the conclusion [21]. This study clearly reported all elements, with the exception of the rationale for the comparison intervention, justification for the form of economic evaluation, the quantities of resources or unit cost, or justification for the variables in the sensitivity analysis. Overall, this study was classified as high quality.
In the Canadian study by Cipriano and Zaric [22], community distribution of naloxone was considered worthwhile at a WTP threshold of CAN$50,000 (year 2018 values; $US36,525 in 2020) per QALY gained. This school-based naloxone distribution program was found to be worthwhile in the probabilistic sensitivity analysis, which considered the number of overdoses per year and the effectiveness of the program at reducing mortality [22]. The scenarios considered ranged from 15 to 97% program effectiveness. In the worst-case scenario of 15% effectiveness, the program would be worthwhile with an ICUR >CAN$50,000 (year 2018 values; $US36,525 in 2020) per QALY gained at a minimum overdose rate of 2.7 per year; in the best-case scenario of 97% effectiveness, the program would be worthwhile with an ICUR of > CAN$50,000 (year 2018 values; $US36,525 in 2020) per QALY gained at a minimum overdose rate of 0.4 per year [22]. This study was classified as high quality, with only the details of statistical test and confidence intervals for scholastic data, justification for the form of economic evaluation, and the rationale for the comparator not reported.
Uyei et al. [23] based their study in the USA and ran their model for 20 years, concluding that the intervention reduced overdose deaths by 6% but increased HIV deaths. Considering a WTP threshold of $US100,000 (year 2015 values; $US109,216 in 2020), they concluded that naloxone distribution was worthwhile with an ICUR of $US323 (year 2015 values; $US352 in 2020) per QALY gained and recommended that the program be funded [23]. Their one-way sensitivity analysis considered ranges for the price of naloxone, survival rates, and the percentage of people who intervene in overdose [23]. No variables had a substantial impact or changed the conclusion [23]. This study was determined to be high quality, missing a justification for the form of economic evaluation and quantities of resources reported separately from their unit costs.
In the USA, Coffin and Sullivan [24] found that the intervention prevented 6.5% of overdose deaths, with an ICUR of $US14,000 (year 2012 values; $US15,784 in 2020) per QALY gained. All in all, naloxone distribution was worthwhile at a WTP threshold of $US50,000 (year 2012 values; $US56,500 in 2020), with funding being recommended [24]. Both deterministic and probabilistic sensitivity analyses were performed, which considered ranges for the price of naloxone, rates of bystander response, and justice system rates [24]. In the deterministic sensitivity analysis, bystander response rate made the largest impact on the ICUR but was not important enough to alter the conclusion [24]. This study was found to be of high quality, with elements of study viewpoints and justification, rationale for comparator, justification of form of economic evaluation, quantities of resources reported separately from unit costs, and justification of discount rate not included.
The Russian study by Coffin and Sullivan [25] built off their previous research by using the same model and variables in the sensitivity analysis. They found that the intervention reduced overdose deaths by 13.4% in the first 5 years and by 7.6% over a lifetime, with an ICUR of $US94 (year 2010 values; $US112 in 2020) per QALY gained. Again, they found that naloxone distribution was worthwhile at a WTP threshold of $US1500 (year 2010 values; $US1785 in 2020) and recommended funding the program [25]. In the deterministic sensitivity analysis, bystander response rate again made the largest impact on the ICUR, but the impact was not substantial enough to change the conclusion [25]. This study was found to be of high quality, only missing elements of study viewpoints and justification, rationale for choosing comparator, justification of form of economic evaluation, and quantities of resources reported separately from unit costs.
Results of the study by Acharya et al. [26], set in the USA, showed that intervention modestly reduced overdose deaths, with an ICUR of $US56,699–76,929 (year 2018 values; $US58,400–79,237 in 2020) per QALY gained. Thus, naloxone distribution was worthwhile, assuming a WTP threshold of $US100,000 (year 2018 values; $US103,083 in 2020) [26]. Both deterministic and probabilistic sensitivity analyses were performed, which considered the price of naloxone, naloxone effectiveness, the proportion of overdoses witnessed, the probability of EMS intervention, overdose risk based on the specific opioid, and overdose survival rates [26]. In the deterministic sensitivity analysis, naloxone effectiveness and the proportion of overdoses witnessed had the largest impact on biannual distribution but did not have a large enough impact to change the conclusion [26]. This study quality was assessed as moderate because of the lack of clear reporting of the viewpoints of the analysis, the rationale for choosing the alternative program or intervention, the justification of the form of economic evaluation, the quantities of resources, and price data.
The Scottish study by Bird et al. [27] resulted in a 3.5% decrease in overdose deaths and had an ICUR of £560–16,900 (year 2015 values; $US769–23,209 in 2020) per QALY gained. No WTP threshold was specified, but naloxone distribution was deemed worthwhile upon conclusion, with continued funding recommended [27]. This study performed a one-way sensitivity analysis on averted overdose numbers [27]. This study did not report the economic importance of the research question, justification for the form of economic evaluation, details of currency price adjustment for inflation or currency conversion, time horizon of costs and benefits, justification of the discount rate, incremental analysis, the approach to sensitivity analysis, or justification of variables chosen for sensitivity analysis. These missing elements resulted in a quality rating of moderate.
Cost-Benefit and Cost-Effectiveness Study
In their study set in the USA, Naumann et al. [28] deemed community distribution of naloxone to be a good investment. They found that, over the course of 3 years, 352 overdose deaths were prevented, leading to an ICBR of $US1:$US2742 (year 2019 values; $US1:$US2769 in 2020). This study based the one-way sensitivity analysis solely on the price of naloxone [28]. They also performed a CEA and deemed community distribution of naloxone to be cost effective [28]. The ICER was $US1605 (year 2019 values; $US1621 in 2020) per death avoided from opioid overdose [28]. No WTP threshold was specified; however, the conclusion of cost effectiveness was stated. This study was found to be of low quality as the rationale for choosing the comparison intervention, justification for the form of economic evaluation, details of currency price adjustments for inflation or currency conversion, the choice of model used and its key parameters, the time horizon of costs and benefits, a discount rate, an explanation for not discounting costs, or the approach to sensitivity analysis were not clearly reported.