Focus group and pilot survey participation
Five participants were included in the focus group, with at least one participant from each didactic year. Analysis of the transcript indicated that participants were initially slow to respond, and the length of responses were short, but as the session proceeded, participants showed more openness, shared longer responses (including personal stories), and built off one another’s responses to the questions in the guide sheet (See ; Appendix C, Additional File 1). By minute 11, all participants had spoken at least once and both verbal and non-verbal agreements were witnessed among the group. Seven emerging themes were extracted using the word similarity query in NVivo (See Supplemental Figure 1, Additional File 1;) and identified strong participation amongst all study group members (See Supplemental Figure 2, Additional File 1) on most topics (See Supplemental Figure 3, Additional File 1), with direct discussions of opioid overdose having more limited participation (See Supplemental Figure 4, Additional File 1). Three of five participants in the focus group fully answered the pilot survey, and no comprehension issues were noted in their overall response. The primary feedback on the content of the pilot survey was that there were too many items, and so the final student and practitioner surveys were shortened to accommodate possible mental fatigue, tailoring the included content to address the differential level of expertise and background between the student pharmacist and practitioner surveys.
Survey sample characteristics
Out of 525 possible respondents, a total of 37 first year students, 53 second year students, and 26 third year students took the student survey (22.1% response). A sample of 364 practicing pharmacists with a Wisconsin state license were invited to complete the practitioner survey, and 38 were taken (10.4% response). The vast majority of student respondents had some experience working in a pharmacy, with 1-2 years of experience being the most common condition, and outpatient sites being the most common work setting (Table 1). Practitioner experience was skewed toward more recent graduates, with less than 5 years in practice being the most frequent condition, and outpatient being the most common practice setting. When considering self-reported practice areas, a small minority of respondents did not associate their practice with any of pain management, addiction therapy, or palliative care, with 44.7% indicating they associated at least one of these with their area of practice (Table 2).
Opinions on addressing opioid use disorder
To first measure how respondents perceived OUD and its management overall, we assessed their perceptions of interventions at the individual or societal level. When asked to rank order the importance of possible outcomes for individual-level interventions, both students and pharmacists indicated that preventing overdose is significantly more important than other patient-level outcomes (p<0.05 for all comparisons, Figure 2A-D). Although the overall ranking was quite similar overall, pharmacists prioritized craving management over social function, whereas students ranked these two of nearly equal importance. Both groups ranked craving management as significantly more important than decreasing drug reward, which was considered to be the least important outcome. When ranking approaches taken at the societal level, both students and pharmacists ranked the development of new treatments for OUD as the least important, but pharmacists ranked promoting treatment access as most important, while students ranked educating providers as most important, perhaps reflecting their own present roles.
Students were additionally asked to indicate how severe they perceived the opioid epidemic to be relative to other current healthcare issues. On average, students indicated the opioid epidemic was in the top 30% of national healthcare issues, although variability in this measure covered the full range from 0 to 100% (See Supplemental Figure 5, Additional File 1). Students were also asked to rank the categories of OUD treatment interventions from most to least important. As a group, students ranked safe medication storage and disposal as significantly less important than any other intervention, with behavioral therapy ranked as the most important overall (p<0.0001, p=0.0048, and p=0.0004 for behavioral therapy, drug use monitoring, and detox vs. safe medication disposal respectively).
Familiarity with and Use of OUD therapeutics
Given that one of the initial criteria for choosing pharmacists as a sample was their high baseline level of knowledge regarding medication interventions, respondents were also asked to assess their knowledge regarding OUD therapeutics and anti-opioid vaccine approaches (Figure 3A-D). Both students and pharmacists indicated a greater degree of familiarity with current treatment options when compared to anti-opioid vaccines. This difference was apparent when using a bipolar scale for the student survey, but adjustment to a unipolar scale for the practitioner survey further highlighted the large number of individuals with no prior knowledge of anti-opioid vaccine approaches whatsoever. Stratification of student data by year in school and pharmacist data by years in practice revealed students’ confidence in knowledge of current therapeutics increased with years of training (p = 0.03). This significance was not reflected in the pharmacist population. Additionally, there was no equivalent trend regarding familiarity with VBTs in either students or pharmacists.
In order to more closely measure functional familiarity with current resources and OUD medications, and additionally to provide a baseline analysis for which types of resources were most likely to penetrate into a pharmacy practice environment, respondents were further asked about their use frequency for other potential interventions, including non-pharmacologic interventions (Figure 3E-H). Compared to all other available resources, the prescription drug monitoring program (PDMP; i.e. controlled substance tracking database) was both the most available resource and was reported as significantly more frequently utilized by pharmacists than all other resources (p<0.05 for all comparisons) and significantly more utilized at students’ places of work (p<0.0001 for all comparisons). For the examined resources, availability was closely matched with frequency of use overall, with the notable exception that dropboxes appear more frequently used than take-back programs, despite being less popular.
With this information regarding non-medication resource use in hand, we further assessed respondents’ perceptions regarding availability and utilization of specific medications for management of OUD and its associated risks (Figure 4A-F). Overall, naloxone and buprenorphine-naloxone were available at roughly 80% of pharmacies where students worked and were used weekly or more. Naltrexone (oral tablets) and Vivitrol® were available at roughly 60% of these pharmacies and were used about monthly or less. For the practitioner survey, respondents were specifically asked to assess the availability and utilization of different dosage forms of these medications, as a means to identify whether an injectable dosage form seemed to be a structural barrier to use overall. Injectable forms of naltrexone and naloxone were available at just over 50% of pharmacies while injectable buprenorphine, and methadone were available at about 25% or less of pharmacies. Conversely, oral naltrexone and intranasal naloxone were available at 89% and 86% of pharmacies respectively while sublingual/buccal buprenorphine and oral methadone were available at 81% and 62% of pharmacies. Utilization of injectable dosage forms was lower overall, aside from IM naltrexone, but was significantly lower for only buprenorphine (p=0.0003). In contrast to the results with the resources, availability of medications was not as closely associated with frequency of use, particularly for the oral medication formulations.
Efficacy of Current Interventions and Perceptions of Anti-Opioid Vaccine Utility
The overall perceived utility of current OUD therapeutics and anti-opioid vaccines was then assessed (Figure 5A-D). When students were asked to consider how helpful or unhelpful currently available interventions were for meeting patient-specific outcomes (abstinence, social functioning, craving management, overdose prevention, and decreased drug reward) there was general agreement that current treatments are helpful for meeting these outcomes overall, using a bipolar scale. After being asked how helpful or unhelpful they believed that anti-opioid vaccines will be as an addition to current OUD therapies, the majority of students also indicated vaccines would likely be either helpful or very helpful. Given the limited difference observed between efficacy in treating specific outcomes, and the overall bias toward helpfulness that was observed when using the bipolar scale with students, the question addressing utility of current OUD treatments was amended when asked to pharmacists. Specifically, it was adjusted to get additional information regarding the relative helpfulness of OUD treatments as compared to available treatments for other conditions. When asked the question using this approach, about one-third of pharmacists believed current treatments are slightly worse or much worse than treatments available for other therapeutic areas. Considering the difficulties of predicting relative efficacy for available versus hypothetical treatments, pharmacists were still presented with a bipolar question regarding their perception of how helpful or harmful anti-opioid vaccines would be for treatment of OUD. The majority of respondents indicated that vaccines would be slightly, somewhat, or very helpful for OUD treatment outcomes.
Perceived logistical barriers and ethical concerns with current OUD treatments and anti-opioid vaccines
When asked about logistical barriers to both current OUD treatments and anti-opioid vaccines, student responses did not differentiate greatly between these two classes. In terms of the logistical and ethical barriers considered, all of them were considered significant barriers or concerns, although variability was greater within those responses that concerned anti-opioid vaccines (See Supplemental Figure 6, Additional File 1). This observed bias toward one half of the bipolar scale again prompted the use of a modified, unipolar scale when administering the analogous questions on the pharmacist survey. In order to improve quantification of differences, exact matching of items for current OUD therapeutics and anti-opioid vaccines was introduced as well. Using this modified scale and approach, pharmacists indicated time was the least worrisome logistical barrier, as compared to medication access, provider availability, affordability, and patient refusal for the vaccine-based therapies (p<0.05 for all comparisons) and medication access or provider availability for current therapies (p<0.05 for both comparisons) (Figure 6A, B). Affordability and patient refusal were noted as more serious barriers for anti-opioid vaccines than for current therapeutics, while the magnitude of concern for the other barriers was similar on average. A larger degree of variability was again seen for the anti-opioid vaccine items. When assessing ethical considerations, pharmacists also ranked access inequality as a very important ethical concern for current and vaccine-based treatments (Figure 6C, D). For current therapies, access inequality was ranked as a significantly more important treatment barrier than any of inefficient use of scarce resources, promotion of risky behavior, reduced patient autonomy, and potential confidentiality breaches (p<0.05 for all comparisons). For vaccine-based therapies, access inequality was also ranked as a significantly more important barrier than both reduced autonomy (p=0.03) and confidentiality breaches (p=0.0004). Reduced autonomy was rated as more concerning for anti-opioid vaccines than current therapeutics, while the other items were rated similarly. Amongst logistical and ethical considerations tested pairwise, only affordability was found to be significantly different with respect to current therapeutics versus anti-opioid vaccines (p=0.0085).
Indeed, using the directly matched items presented to pharmacists, it was then considered whether pharmacists were considering anti-opioid vaccines as having equivalent logistical and ethical dimensions to current OUD therapeutics, a hypothetical cognitive approach referred to here as the ‘reference equivalency’ model. If application of this ‘reference equivalency’ cognitive model was the case, as hypothesized, the values provided by pharmacists should be most closely matched between identical items presented across both the current therapeutic and anti-opioid vaccine matrices, as compared to any other possible pairwise combination of these items, and not allowing each item to be matched with itself. That is, if the values for the five items in the logistical barrier matrix for current therapeutics are [●, ■, ▲, ▼, ♦], and the values given for the identical items in the logistical barrier matrix for anti-opioid vaccines are [○, □, △, ▽, ◇], then there are nine possible differences for each item in the matrix (e.g, ●-■, ●-▲, ●-▼, ●-♦, ●-○, ●-□, ●-△, ●-▽, ●-◇), resulting in 95 possible sums of differences across all five items in the matrix. Using this notation, the ‘reference equivalency’ model is represented as [(●-○) + (■-□) + (▲-△) + (▼-▽) + (♦-◇)] (Figure 7A, B).
To test whether this ‘reference equivalency’ model yields the smallest sum of differences for both logistical barriers and ethical concerns, the differences in values for each of the five items collected in the logistical barrier and ethical concern matrices, respectively, were summed to generate distributions of all 95 possible resulting values (Figure 7C, D). In this analysis, a smaller sum of differences represents a higher correlation between the paired combinations for current therapeutics and anti-opioid vaccines, demonstrating which items are most closely matched overall. With this analysis, the resulting means of the distributions for all sums of differences in each matrix (Logistics: 5.63 ± 0.60; Ethics: 5.78 ± 0.64) were significantly smaller than the sum of differences expected (10.00 ± 3.14; p<0.0001 for both) if subjects were applying a random approach to consideration of these items overall.
Furthermore, when looking within the 95 possible pairwise comparisons that were possible within the observed dataset, the sum of differences for the ‘reference equivalency’ cognitive model was among the smallest 0.0001% of all possible differences that could be generated for these two datasets, a result that is significantly different than expected by chance alone (Logistics: 3.65, Z = -3.30, p < 0.001; Ethics: 3.65, Z = -3.16, p = 0.0016). However, it should be noted that the ‘reference equivalency’ sum of differences for each domain was not the absolutely smallest of all possible sums of differences resulting from the observed data set. This finding indicates that while respondents were treating current OUD therapeutics and anti-opioid vaccines as having similar logistical and ethical dimensions overall, there were still specific areas that merited differential consideration between the two approaches. The two examined areas which differed most between current OUD therapeutics and anti-opioid vaccines were cost and patient refusal.
Product-specific Characteristics for Anti-Opioid Vaccine Development
When developing a new therapeutic approach, there are pragmatic considerations that arise regarding product development in addition to the broader logistical and ethical considerations considered above. To this end, students and pharmacists were presented with a matrix of questions related to vaccine product preferences across several dimensions. These dimensions included the generic considerations of cost and storage requirements, the vaccine-specific considerations of effective population coverage, time to onset of full protection, and duration of protection, and the product-specific consideration of breadth of opioid blockade (See Supplemental Figure 7, Additional File 1). Students were asked fewer questions in this domain, given their relative lack of expected experience with product handling and recommendation. Furthermore, as a means to understand prioritization of broad-scale efforts, students and pharmacists were also asked to rank which of three domains were most important to consider in the development of VBTs. Across both populations, establishing efficacy was ranked significantly higher than management of either ethical or logistical considerations (p<0.05 for all comparisons).
Application of Anti-Opioid Vaccines in Ethically-Variable Clinical Practice Settings and Populations
Determination of which clinical population to target with a given intervention is a critically important decision during the development of a new therapeutic intervention, as it informs the design of clinical trials, directly constrains the indications considered for approval by regulatory bodies, and often determines coverage access to the treatment itself. Selection of an appropriate population for anti-opioid vaccine use is potentially even more fraught than the average case, as considerations of mandated use and application to vulnerable populations have been a challenge in the context of both vaccination and substance use disorder treatment. One relevant example of mandatory therapy in a vulnerable population is court-ordered medical treatment subsequent to a drug possession and distribution offense; such policies lead to potentially coercive use of MAT for SUDs in incarcerated populations. These types of programs have been studied since the late 1990s and have grown substantially in the populations they serve.38, 39 Therefore, we surveyed pharmacist-rated support or opposition for the mandatory and voluntary use of vaccine-based therapies across various clinical scenarios and in different vulnerable populations (Figure 8). A global preference for voluntary use as compared to mandatory use was reported (p<0.05 for all comparisons).