This secondary analysis of the cross-sectional Athlete PEACE Survey study is the first to identify the mediating effect of cannabis attitudes on knowledge and cannabis use. Cannabis knowledge impacted cannabis use directly and indirectly through cannabis attitudes. Athletes were clustered into three attitudes groups: Conservative, Unsure, and Liberal. Liberal and Unsure athletes tended to score better on the knowledge questions and were more likely to be current cannabis users rather than never users or past users compared to the Conservative cluster.

Previous studies regarding the impact of attitudes and knowledge on cannabis use have examined the direct or mediating effects of parental knowledge on their child’s drug use (Macaulay et al. 2005; Sellers et al. 2018; Wu et al. 2015). Parental knowledge indirectly influenced adolescent cannabis use through the number of offers of cannabis (Siegel et al. 2015), when adolescent-reported parental knowledge increased, adolescent alcohol use decreased (Sellers et al. 2018), and maternal knowledge had direct and indirect effects on adolescent substance use (Wang et al. 2009). In adults, attitudes mediated the relationship between information seeking and intention to use cannabis, but actual cannabis use was not measured (Martinez and Lewis 2016).

While it is not possible to directly measure causality with cross-sectional studies, research examining Knowledge-Attitude-Behavior relationships have been cross-sectional for practical reasons and to test hypotheses that can guide interventional studies (Valente et al. 1998; Baranowski et al. 2003; Fairchild and McDaniel 2017). Cross-sectional studies are considered appropriate for mediation analysis if confounding is addressed and the temporal relationship of the variables can be established given the measured constructs (Fairchild and McDaniel 2017). We addressed the potential confounders of age, gender, and duration of cannabis use by adding them to the mediation model; none of these variables were significant and were not included in the final model. A review of knowledge, attitudes, and substance use disorder included nine epidemiologic studies which measured all three dimensions contemporaneously and concluded “Lack of knowledge on the risks of substance use has contributed to the increasing cases of substance use disorders. Substance use has been attributed to lack of proper knowledge on the associated risks.” (Njoroge and Kenyatta 2017) With respect to this study, knowledge about cannabis which originates from multiple sources throughout an athlete’s lifetime precedes their attitudes toward cannabis which then impacts their decision making on whether to use cannabis. Certainly, this study is a snapshot, taken at one point in time, whereby all three domains could change by shifting knowledge and/or attitudes (e.g. providing accurate knowledge about cannabis may change an athlete’s attitude and thusly change their behavior).

Medical cannabis has received increased attention due to its potential positive impact on chronic pain, fibromyalgia, inflammatory bowel disease, and other difficult to treat conditions (Bachhuber et al. 2019; Sagy et al. 2019; Ahmed and Katz 2016; Park and Wu 2017). Traditionally, the study of cannabis has been through the lens of reducing use and abuse, with only limited research into benefits from cannabis (Satterlund et al. 2015). Even among the Conservative attitudes cluster in these analyses, 79.4% reported that medical cannabis can be beneficial for people with certain medical conditions and 71% supported medical legalization. Thus, there is still a disconnect between knowledge and attitudes regarding cannabis which could impact behavior, particularly among medical professionals who may be asked by patients to give advice on cannabis use (Berlekamp et al. 2019; Moeller and Woods 2015; Chan et al. 2017).

Given the significant impact of knowledge on cannabis use behavior, the source of knowledge is of utmost importance, particularly given the expanding legalization of cannabis at the state level. Often, the primary source of medical cannabis advice comes from dispensary staff (i.e. budtenders) (Peiper et al. 2017; Wang et al. 2009; Martinez and Lewis 2016). Medical training for budtenders is lacking, with only 20% reporting specific medical or scientific training, yet 94% provided cannabis advice to patients (Haug et al. 2016). This lack of training has seen 69% of surveyed budtenders recommend cannabis products as a treatment for morning sickness in pregnant Colorado women (Dickson et al. 2018), despite the known harmful effects of cannabis use in pregnancy in increasing preterm birth (Corsi et al. 2019). Pharmacists in Minnesota believed there was insufficient training in cannabis pharmacotherapy and were not aware of state-wide regulations, however there was interest in educational programs (Hwang et al. 2016). Pharmacy students poorly identified qualifying medical conditions and adverse effects and they did not feel confident about answering questions about medical marijuana; only 13% indicated formal education on medical marijuana during their training (Moeller and Woods 2015). Colorado medical students had favorable attitudes toward cannabis, however, they did not feel comfortable recommending it, due to lack of evidence and training (Chan et al. 2017).

Underscoring how knowledge and attitudes can potentially affect cannabis use, 12.5% of the Conservative attitudes cluster in this sample indicated they do not use cannabis because they are scared but only 2.7% of the Liberal attitudes cluster chose this as a reason not to use cannabis while the Unsure cluster fell in the middle at 8.3% (this difference was significant, p < 0.01). Proper education from informed budtenders and medical professionals could potentially alleviate fear and lack of understanding about the benefits and adverse effects of cannabis. The importance of this undertaking and the need for evidence-based education is clear; unabated or improper cannabis use has increased emergency room visits for cannabis hyperemesis syndrome and psychotic episodes (Monte et al. 2019).

Since this is a convenience sample recruited via multiple sources, it is not possible to know the total number of athletes who saw the call to action and therefore cannot determine the refusal rate. This is a study limitation. The representativeness of this sample to the wider sporting community is unknown. Comparisons to the latest statistics from the governing body of triathlon (USA Triathlon) and cycling (USA Cycling) show that the participant demographics in this sample roughly match the overall populations. The runners in this sample skew male which is different than the sex composition found by Running USA (USA Triathlon 2016; Running 2017; https://legacy.usacycling.org/corp/demographics.php n.d.). Even though the sample demographics roughly reflect those of the greater population of triathletes, runners, and cyclists, the participants are self-selected, therefore the cannabis attitudes clusters and knowledge about cannabis may not be representative of athletes in general. Our data show the following male percentages for the three main sports: 53% running, 81.3% cycling, and 63.45 triathlon. USA Triathlon most recently reported their membership is 65% male (USA Triathlon 2016), while 85.3% of USA Cycling’s member are male (https://legacy.usacycling.org/corp/demographics.php n.d.) and 43% of runners are male (Running 2017). In terms of age of triathletes, 12.8% are 25–29, 35.6% are 30–39, 29.4% are 40–49, and 13.5% are over 50 (USA Triathlon 2016); this compares to our numbers of 14% 21–29, 22.6% 30–39, 28.6% 40–49 and 34.9% over 50. USA Cycling shows that 28.0% of their members are 19–34 and 63.7% are 35 and older (https://legacy.usacycling.org/corp/demographics.php n.d.); 13.5% of our participants who cycled were 21–39. Half of all runners were 25–44 years old (Running 2017) while 62.2% of our runners were over 40.

We recognize that the knowledge and attitudes expressed by this sample may not be representative of other demographic groups. The attitudes observed in this sample were similar to those found in a sample of New York college students (Q Market Research 2015). It is reassuring that the relationship between knowledge, attitudes, and behavior has been observed in other domains (Kuhlemeier et al. 1999; Register-Mihalik et al. 2013; Ozisik et al. 2017). The large sample size allowed for adequate clustering of the data and the mediation regression analysis. In addition, the sample represented a large age-range and varying experience with cannabis use, both of which can impact knowledge and attitudes (Gates et al. 2017; Berlekamp et al. 2019; Moeller and Woods 2015).

It has been noted that cross-sectional data may not be appropriate for mediation analysis, particularly if a process is changing over time (O’Laughlin et al. 2018). The Knowledge-Attitude-Behavior model postulates the temporal ordering, as was succinctly stated by Baranowski et al.: “As knowledge accumulates in a health behavior domain, changes in attitude are initiated. Over some period of time, changes in attitude accumulate, resulting in behavioral change.” (Baranowski et al. 2003) Furthermore, the present analysis provides important insights: (1) cannabis epidemiology is a nascent field, thus obtaining information to direct future epidemiologic studies is an essential first step; (2) longitudinal studies are not necessarily the only way to examine the Knowledge-Attitude-Behavior relationship and often are not feasible to conduct (Buckner et al. 2019; Ball et al. 2018; Stubbs et al. 2018; Buckner et al. 2018); and (3) future interventional study can test this model.

In conclusion, these mediation analyses indicated that cannabis attitudes appeared to mediate the effect of knowledge on use. As more states legalize medical and recreational cannabis, proper consumer, budtender, and medical professional education will become paramount to the safety and efficacy of cannabis use.