Abstract
Context:
Cannabis use for symptom management among individuals with cancer is increasingly common.
Objectives:
We sought to describe the (1) prevalence and characteristics of cannabis use, (2) perceived therapeutic benefits of cannabis use, and (3) examine how use of cannabis was associated with self-reported pain, mood, and general health outcomes in a representative sample of patients treated at a National Cancer Institute (NCI)-designated Oregon Health and Science University Knight Cancer Institute.
Methods:
We conducted a population-based, cross-sectional survey developed in conjunction with 11 other NCI-designated cancer centers and distributed to eligible individuals. The survey inquired about characteristics of cannabis use, perception of therapeutic benefits, pain, mood, and general health. Responses were population weighted. We examined the association of cannabis use with self-reported pain, mood, and general health using logistic regression controlling for relevant sociodemographic and clinical characteristics.
Results:
A total of 523 individuals were included in our analytic sample. A total of 54% endorsed using cannabis at any time since their cancer diagnosis and 42% endorsed using cannabis during active treatment. The most commonly endorsed reasons for use included the following: sleep disturbance (54.7%), pain (47.1%), and mood (42.6%). We found that moderate pain was associated with more than a twofold (odds ratio = 2.4 [95% confidence interval = 1.3–4.6], p = 0.002) greater likelihood of self-reported cannabis use. Depressed mood and general health were not associated with cannabis use.
Conclusions:
In a state that had early adoption of medical and recreational cannabis legislation, a high number of cancer survivors report cannabis use. Moderate or more severe pain was associated with an increased likelihood to use cannabis, while mood and general health were not associated. Oncologists should be aware of these trends and assess use of cannabis when managing long-term symptoms of cancer and its treatments.
Introduction
Cancer and its treatments are qualifying conditions in the majority of the 37 states that have legalized medical cannabis. 1 Cannabis use in patients with cancer is common. Depending on the legal status where it is measured, 8% 2 to almost a quarter of individuals with cancer report current cannabis use3,4 and more than 90% of cancer survivors view cannabis as potentially beneficial for symptom management and support its legalization. 5
More than one-third of patients will report moderate-to-severe pain due to cancer, its treatments, or both.6,7 Among patients with cancer who use cannabis, data suggest that approximately 75% use it for symptom management, most commonly pain, nausea, and sleep disruption.8,9 Unfortunately, there is a paucity of data about the characteristics of cannabis use and limited evidence about its safety and efficacy for noncancer pain management, 10 with a recent systematic review on cancer pain concluding limited effectiveness of cannabis to improve pain. 11
Several cross-sectional surveys have linked cannabis use among cancer survivors to the presence of pain.2,12 In addition, prior work among cancer survivors has associated characteristics such as male sex, lack of insurance, younger age, and lower education to cannabis use.2,4 One study found that those with cancer who use cannabis are more likely to have chronic pain that impacts their daily activities (high-impact chronic pain). 12 In addition, those with high-impact chronic pain are more likely to perceive a therapeutic benefit of cannabis and less likely to report adverse events related to cannabis compared with those without high-impact chronic pain. 12 Data are still lacking about whether cannabis use is associated with other important outcomes such as mood and perception of general health across different time points of cancer survivorship. Importantly, among cancer survivors, depression and cannabis use have been associated,13,14 particularly among those who use cannabis more regularly. 14 Perception of general health has also been associated with poor outcomes in cancer survivors, generally,15–17 but limited data are available about the relationship between general health and cannabis use among cancer survivors.
Within the current climate of increasing cannabis availability and acceptance, we sought to obtain a more complete and nuanced understanding of cannabis use, including frequency of use, modes of use, and reasons for use among a large representative sample of cancer patients. As oncologists increasingly encounter patients using cannabis, knowledge about patient’s use and perceived therapeutic benefit is clinically informative. Thus, our goal was to describe the (1) prevalence and characteristics of cannabis use, (2) perceived therapeutic benefits of cannabis use, and (3) examine how use of cannabis was associated with self-reported pain, mood, and general health outcomes in a representative sample of cancer patients treated in Oregon, a state that has long-standing recreational and medical cannabis. We hypothesized that cannabis would be associated with higher pain intensity and depressed mood. The relationship with perception of general health was exploratory.
Methods
We conducted a cross-sectional population-based survey (see Supplementary Data S1). This study was approved by our Institutional Review Board (OHSU 21714/VA #4679). This project was one of 12 that were funded as part of a National Cancer Institute (NCI) initiative across NCI-designated cancer centers. 18
Study setting and sample population
The Oregon Health and Science University (OHSU) Knight Cancer Institute (Knight) is the only NCI-designated comprehensive cancer center in Oregon. Along with the main campus location in Portland, the Knight also has 5 community cancer clinics, and its catchment area is defined as the entire state of Oregon. We obtained a sample of patients treated at OHSU Knight from the OHSU Cancer Registry, which collects and maintains information on the demographics, diagnosis, treatment, and outcomes of all patients diagnosed with cancer from all of its affiliate clinics. 19 For this study, we included adults (age 21+) who had been diagnosed with any type of cancer, received some treatment at the OHSU or its affiliate clinics, and completed treatment within the last 18 months. Data collection occurred between August 2021 and April 2022. There were no study exclusionary criteria.
Recruitment processes
We used a validated survey method (Dillman Tailored Design Method) to inform the survey procedures. 20 The survey invitations and reminders were sent through electronic health record (EHR)-based electronic messaging. We also added a mailed invitation option for those who were not enrolled in the EHR-based electronic messaging. Patients identified as eligible received electronic or mailed invitations containing (1) a study description, (2) study team contact information, and (3) an opt-out response form. To those who did not opt-out, a letter with a link to the REDCap questionnaire and a monetary incentive of $5 gift card was mailed 2 weeks after the first mailing. Upon request, participants could also receive a paper copy of the questionnaire with a prepaid return envelope. All electronic surveys were completed anonymously. Paper surveys were removed from their envelope and assigned a random code to deidentify.
Sampling framework
We used random sampling stratified by rural versus urban locality defined by patient’s living location using Rural Urban Community Codes. 21 Patients were sampled in a ratio of 48.0% urban to 52.0% rural to ensure rural representation; actual percentages of eligible patients were 65.0% urban and 35.0% rural. Out of 7,356 potentially eligible patients, a total of 3,725 were sampled.
Survey development and measures
Through discussion and consensus, we collaboratively developed a set of core data elements along with a national consortium of investigators from 12 funded NCI Comprehensive Cancer Centers in collaboration with ICF Next™ (https://www.icf.com), a global marketing and research consultation group. A full copy of the survey instrument can be found in Supplementary Data S1 and at https://epi.grants.cancer.gov/clinical/#initiatives.18
Current and past use of cannabis
Cannabis use was defined as “any marijuana, cannabis concentrates, edibles, lotions, ointments, tinctures, or other products made with cannabis, as well as cannabidiol (CBD) only products.” We inquired about cannabis use in participants’ lifetime, use during cancer treatment, and use in the last 30 days. Among participants who endorsed any use since their cancer diagnosis, follow-up questions assessed cannabis use characteristics, including frequency of cannabis use (e.g., multiple times per day to monthly) and mode of use (e.g., route of cannabis administration such as smoked, tincture, edibles, vaping products). We defined cannabis use as endorsement of cannabis any time since cancer diagnosis.
Therapeutic reasons for use
We inquired about which symptoms cannabis was used to ameliorate (e.g., pain, chemotherapy-induced neuropathy, nausea, lack of appetite, disrupted sleep, and depressed or anxious mood).
Pain and mood
In addition, while not part of the multisite data collection, our site included the Pain, Enjoyment, General Activity (PEG) scale, which is a 3-item questionnaire assessing pain intensity, interference with enjoyment of life, and interference with activity. 22 The Patient Health Questionnaire-2 (PHQ-2) is a validated 2-item measure of depression severity. 23 Higher scores on the PEG and PHQ-2 indicate more severe symptoms.
General health
We included a single-question inquiring about perception of general health in which the participants were asked to rate their health from Excellent to Poor on a 5-point Likert scale. 24
Opioid questions
We also assessed whether individuals were using prescription opioids (current, in the last 3 months, more than 3 months in the past, or never). Response to these questions was included as a dichotomous covariate (current vs. all other time points) to represent current opioid use in the regression models.
Sociodemographic characteristics
Information about age, sex, gender identity, race/ethnicity, income, education level, and insurance status was obtained via self-report. We did not obtain data on cancer type and treatment.
Data analysis
Population survey weights were used to ensure that the sample of survey respondents was representative of the OHSU cancer patient population and to reduce potential survey bias. Rural/urban status was not collected from survey respondents, and so, equal probability of selection was assigned to each respondent. From nonresponse analysis, significant differences in the response rates by age category and race were detected. A raking algorithm was used to iteratively adjust the base weights to population totals using age, sex, race, and ethnicity to account for potential nonresponse bias. Appropriate statistical adjustment was used to incorporate the estimated weights into all analyses.
Descriptive statistics were used to summarize the survey responses (means and standard deviations for continuous variables; counts and percentages for categorical variables). Participants were divided into two groups based on whether they endorsed cannabis use after their cancer diagnosis. T-tests and chi-square tests were used to compare the demographic characteristics based on self-reported cannabis use status. Descriptive statistics were also used to summarize barriers to using cannabis, therapeutic reasons for using, characteristics of use (e.g., frequency, amount, routes of administration), and adverse events experienced.
We used multivariable logistic regression modeling to examine the relationships between cannabis use (as the outcome) and pain, general health, and mood. Each of these main effects of interest was modeled separately to independently assess their association with cannabis use at any time since cancer diagnosis. These models also included additional covariates to adjust for demographic and personal characteristics: age, race/ethnicity, sex, education level completed, household income, employment status, prior cannabis use, opioid use, and health coverage. Models were assessed for multicollinearity, and no issues were found. Statistical analyses used R: A Language and Environment for Statistical Computing. 25 We used the srvyr 26 and survey packages 27 for weighted analysis of the population survey data. We did not use imputation in the weighted analysis. A missing category is shown in the tables. For the regression models, those missing outcomes (cannabis uses) were excluded, and a missing category was used for the predictors.
Results
We reached out to 3,725 unique individuals of 7,356 potentially eligible patients (2,625 via EHR messaging and 1,100 via mailed invitation; Fig. 1). A total of 541 patients met the study inclusion criteria and answered the survey (14% response rate); due to missing data or skip pattern failures, 523 were retained for the final analytic sample. We examined sociodemographic variables (age, sex at birth, race, and ethnicity) that were associated with nonresponse (data not shown). Individuals who were more likely to complete the survey were non-White and older than 80 years.
Participant flowchart. This figure demonstrates the number of participants that we identified as initially eligible and ends with the number of participants who had completed data and were included in our analytic sample.
The population weighted and unweighted characteristics of our enrolled sample are summarized in Table 1. Fifty-four percent of our participants endorsed using cannabis at some point after their cancer diagnosis (hereafter referred to as cannabis users). Forty-two percent endorsed using during treatment and an additional 26.6% reported considering cannabis use at some point since their cancer diagnosis. In addition, 71.8% endorsed using cannabis at any time over the course of their life.
Participant Characteristics by Endorsement of Cannabis Use Any Time Since Their Cancer Diagnosis
In unadjusted analyses, the following sociodemographic characteristics were associated with cannabis use at any time following diagnosis: younger age (61.4 vs. 65.8, p = 0.001), driven by a higher percentage of 18- to 44-year olds who reported cannabis use (16.3% vs. 8.1%). A significantly higher proportion of those who used cannabis endorsed “finding it difficult to get by on my present income” (12.4% vs. 4.4%, p = 0.016). Finally, there was an association with the primary source of health insurance and cannabis use, with a notable difference in the percentage of those who used cannabis and had Medicaid compared with those who did not use cannabis and had Medicaid (15.1% vs. 5.7%, p = 0.005). There was also a higher percentage with private health insurance plans among those who did not use cannabis compared with those who did use cannabis (38.8% vs. 31.8%, p = 0.005). There were no other significant differences based on demographic characteristics between the two groups (Table 1).
Differences between the two groups on measures of pain, perception of general health, and mood are displayed in Table 2. Of these variables, higher pain severity, pain interference, and depressive symptomatology were all associated with cannabis use in univariate analyses (all p ≤ 0.001). General health was not associated with endorsement of cannabis use.
Pain, General Health, and Depression by Cannabis Use Since Diagnosis
Note: PHQ-2 scores of 3 or greater are indicative of the likely presence of a major depressive disorder (and would warrant further clinical assessment).
The most commonly endorsed reasons for current cannabis use included the following: sleep disturbance (54.7%, n = 1663), pain (47.1%, n = 1432), mood changes, stress, anxiety, or depression (42.6%, n = 1296), and recreation (35.6%, n = 1083; Table 3). Among those who reported using cannabis during or after cancer treatment, the majority did so multiple times per week or more (Table 4). The four most common routes of administration endorsed were as follows: eating (i.e., candy, brownies) (44.5%, n = 1355), smoking (i.e., in a joint, bong, pipe, or blunt) (34.7%, n = 1056), taking by mouth (i.e., tincture, pills) (32.6%, n = 991), and applying topically (i.e., lotion or cream) (23.7%, n = 720; Fig. 2).
Self-reported routes of cannabis administration. This figure demonstrates the most commonly reported routes of administration (used and tried). Respondents could select more than one response.
Self-Reported Reasons for Cannabis Use
Note: participants were informed to “select all that apply” so numbers exceed 100%.
Frequency of Cannabis Use During and After Treatment
Logistic regression analyses were conducted to examine variables significantly associated with cannabis use at any time since cancer diagnosis. Covariates included in these analyses were age, sex, race/ethnicity, education level completed, household income, occupational status, prior cannabis use, opioid use, and health coverage status. Results indicate that participants with moderate past-week average pain severity were more likely to use cannabis (odds ratio [OR] = 2.4; 95% confidence interval [CI] = 1.3–4.6], p = 0.002). In this model, non-Hispanic ethnicity and prior cannabis use were also associated with an increased likelihood of cannabis use (p < 0.001). A second logistic regression was conducted to examine the extent to which self-rated health status was associated with current cannabis use; this model controlled for the same variables. We did not find an association between self-reported poor or fair health (vs. excellent health) and cannabis use (OR = 0.6 [95% CI = 0.2–1.6], p = 0.316). A final model examined the extent to which depressive symptomatology was associated with cannabis use. After controlling for covariates, depressed mood (as indicated by a score of 3 or greater on the PHQ-2) was not associated with an increased risk of cannabis use (OR = 0.8 [95% CI = 0.4–1.7], p = 0.566, Fig. 3).
Forest plot of regression analyses. This figure shows the adjusted ORs for the associations with cannabis use, pain, depression, and general health. ORs, odds ratios.
Discussion
Cancer survivorship is often fraught with management of ongoing symptoms, including pain, depression, and impairments in general health. With changes in legalization, cancer survivors are increasingly using cannabis as a strategy for managing symptoms, but the extent to which cannabis is effective is unclear. A more complete picture of therapeutic reasons for use, routes of administration, and patterns of use are clinically important and can help oncologic clinicians evaluate potential risks and benefit. In a representative sample of individuals receiving care at an NCI-designated cancer center in Oregon, a state with long-standing medical and recreational cannabis legislation, we found that 54% of patients used cannabis at some point after their diagnosis and 42% used during treatment. This prevalence is higher than the prevalence of cannabis use among cancer survivors, which was estimated to be about 8% 2 and the 12.4% reported by a survey conducted in 2020 by an NCI cancer center spanning multiple states (Minnesota, Florida, and Arizona) 12 and even higher yet than a study estimating 25%, which included participants from an NCI-designated cancer center in a state where recreational and medical cannabis were legal. 4 A likely driver of the notably higher prevalence in our sample is the duration of legal status in Oregon, which has allowed medical cannabis since 1998 and recreational cannabis (use for anyone 21+) since 2015. 28
Nuanced information about how patients are using cannabis is important to inform clinical conversations around patient safety and to screen for potential risks. Our study adds to the current literature by describing more details about cannabis use patterns and reasons for use among cancer survivors. In our sample, among those who use cannabis, about a quarter of individuals reported using cannabis daily and an additional 14% used multiple times per day. Daily or more than daily use has been linked to the development of cannabis use disorder, although in a younger population 29 ; the relationship between cannabis use frequency and potential development of cannabis use disorder has not been sufficiently detailed among cancer survivors. In addition, we found that consuming cannabis via edibles was the most regularly used route of administration (endorsed by about 44% of participants) followed by smoking (35%). These frequencies and preferred routes of cannabis use are similar to what has been documented in other research. 4 Inhalation of cannabis has been linked to bronchial infections, which may have particular relevance to those with lung or upper aerodigestive tract cancers.29,30 These findings point to particular areas of additional clinical inquiry needed to effectively counsel patients using cannabis on potential harm.
Our study adds to current literature by exploring association of cannabis use during treatment with pain, mood, and perception of general health. We found that the presence of moderate pain severity (compared with those without pain) was associated with 2.4 times the likelihood of using cannabis at any point after their cancer diagnosis, when controlling for covariates. This finding is somewhat consistent with other recently published data from another NCI-designated cancer center that found those who use cannabis were more likely to have high-impact chronic pain 12 (defined as chronic pain of moderate-to-severe intensity that limits life or work activities on most day or every day in the last 3 months). However, interestingly in our data, those with the most severe pain were not more likely than those with no pain to use cannabis. The nuance of this finding can likely be explained because some groups of cancer survivors perceive opioid medications as essential in managing severe pain 31 and thus may be using opioids instead of cannabis in cases of severe pain.
We also found that mood improvement, including depression, was the third most common therapeutic use of cannabis among our cohort; however, contrary to our hypotheses, depressive symptoms were not associated with cannabis use in multivariate analyses. Although a link between depression and cannabis use has been previously described13,14 among cancer survivors, differences in sampling and measurement of constructs may explain disparate findings. Finally, in exploratory analyses, general health was not found associated with cannabis use. In the context of long-term cancer symptom management, further inquiry could be focused on understanding the prospective effect of cannabis on self-reported symptoms such as pain and mood. Such information is essential for oncology care providers to counsel patients effectively about cannabis use. 32
There are several limitations that should be considered when interpreting the results from this study. First, these data are cross-sectional, and so, temporal relationships between variables cannot be inferred. Second, we did not collect data on cancer type and treatment, which limits our ability to contextualize these findings based on cancer variables. Third, our response rate was relatively low, yet similar to other broad, “cold call” EHR-based recruitment survey response rates, potentially limiting generalizability due to issues related to response bias.
Conclusion
We found that about half of cancer survivors used cannabis at some point following their diagnosis, with many choosing edibles and using multiple times per week. In our sample, the presence of moderate pain was associated with cannabis use, whereas the presence of depressed mood or poor health was not. Clinicians should assess for cannabis use status among all patients during and after cancer treatment. Prospective research is needed to understand the clinical benefits and harms from use.
Footnotes
Acknowledgments
Kara Winchell, MA; Richard Torres, MA.
Authors’ Contributions
S.N.: Funding acquisition, conceptualization, project administration, supervision, investigation, interpretation of data, and writing—original draft. J.S.: Conceptualization, investigation, interpretation of data, writing—original draft, and writing—review and editing. E.L.: Conceptualization, investigation, data curation, formal analysis, interpretation of data, and writing—original draft. J.L.: Conceptualization, investigation, formal analysis, interpretation of data, and writing—original draft. B.M.: Conceptualization, investigation, interpretation of data, writing—original draft, and writing—review and editing.
Disclosure
This article was released as a preprint on Research Square. www.researchsquare.com, 
Author Disclosure Statement
The authors do not have any relevant conflicts of interest to disclose.
Funding Information
This work was supported by a National Cancer Institute P30 supplement (
Abbreviations Used
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
