This study plans to recruit 440 older adult participants aged 50 years or older, with at least 50% over the age of 65, and with approximately 50% of the sample being male. We will try to achieve the balanced ratio by monitoring sample demographics throughout the study and adjusting recruitment techniques as needed to target necessary demographics. Among these, 330 participants will be seeking to start MM treatment during the study period, as indicated at baseline when they are enrolled in the study, and 110 participants will have no intention of starting MM treatment for at least 6 months. However, participants are free to initiate or stop MM use at any time during the study and continue participation after they join the study. If a participant decides to withdraw from the study or is lost to follow-up at any point, data will be analyzed up to the point of withdrawal.

Potential participants will be eligible for the Study on Medical Marijuana and Its Long-Term Effects on Older Adults (SMILE) if they (1) are 50 years or older; (2) have had noncancer chronic musculoskeletal or neuropathic pain [52] for 3 or more months [53], with a pain intensity level in the past week greater than zero (“0=no pain” to “10=most pain imaginable”) [54]; (3) are seeking MM treatment (for the MM group) or have no intention to start MM for at least 6 months (for the comparison group); and (4) have access to a smartphone.

Individuals aged 50 years and older were chosen due to the rising use of MM for chronic pain and the need for research on this population [5,6]. Participants must have current noncancer chronic musculoskeletal or neuropathic chronic pain, confirmed by verification of pain for 3 or more months at an intensity level greater than zero. Intention to initiate MM treatment is assessed at recruitment to determine which group enrolled participants will be in. To obtain baseline data before the initiation of MM treatment, participants must not be receiving MM treatment at the time of enrollment. Individuals who are not interested in MM should ideally not initiate MM treatment within the 6 months following enrollment. Access to a smartphone is necessary for completing the EMA and Fitbit aspects of the study. Because all assessments and questionnaires that will be used are in English, participants must be able to read English to complete study measures.

Potential participants will be excluded if they (1) are currently pregnant or nursing (by self-report); (2) self-report marijuana or CBD use in the past 30 days, or if the urine test is positive for THC; (3) report any history of using MM regularly (daily use); (4) used marijuana regularly (daily use) for over a month in the past 10 years; (5) have current or past substantial or severe substance use disorders [55], as assessed by the Drug Abuse Screening Test [56,57]; (6) are undergoing active cancer treatment, excluding certain cancers such as skin cancer; (7) have a terminal illness; (8) have neurological conditions that impact their ability to consent or complete study procedures (eg, Alzheimer disease, Parkinson disease, multiple sclerosis, head injury, stroke, and/or seizures); or (9) have active, uncontrolled psychiatric disorders (eg, schizophrenia, bipolar disorder, excluding anxiety and depression). Many of these exclusion criteria are imposed to ensure participant safety, the ability to provide consent, and the completion of study procedures.

All participants will complete an in-person screening, including a urine sample to confirm their marijuana use status, verification of musculoskeletal or neuropathic chronic pain using medication bottles or medical records, and a screening questionnaire. Participants will complete in-person assessments at baseline and 12 months, including survey questionnaires, cognitive testing, blood pressure measurement, and biomarker collection (blood samples for cannabis metabolite tests, telomere length, and inflammation/inflammasome markers). After the initial baseline visit, participants will complete technology-based real-time assessments using a smartphone-based EMA app to measure subjective elements of pain intensity, emotional and physical functioning, and sleep. Participants will also be provided with a wearable Fitbit sensor to record objective data on heart rate, physical activity, and sleep. The EMA surveys and Fitbit data will capture real-time short-term effects during the first month and for 1 week every 3 months over 12 months. Biospecimen collection (urine tests and blood samples) and a follow-up survey will be repeated every 3 months. Participants are encouraged to provide feedback regarding study procedures throughout their participation to inform future revisions to the protocol. Figure 1 provides an overview of the study procedures. A detailed timeline for each measure and assessment conducted throughout the course of the study can be found in Table 1.

Participants will be assigned to the MM group or the comparison group based on their responses to baseline screening questions regarding current interest or disinterest in starting MM treatment. Group assignment will remain consistent for each participant, regardless of whether they change their mind throughout the course of the study. Participants interested in initiating MM treatment will be given information about local MM physicians who can certify them for a Florida MM card. Additional compensation for delaying MM treatment to enroll in the study will also be provided to participants who begin MM treatment within 4 weeks after their baseline visit to cover most of the cost for MM physician visits and the initial charge to obtain their MM card in Florida.

Surveys at all time points take approximately 30 to 45 minutes for participants to complete. The length of the visit can be a barrier for older adults, especially those with chronic pain. In the interest of saving participants’ time and preventing fatigue during the in-person assessment, participants will have the option to complete the survey remotely before their visit. Participants are offered snacks and beverages during in-person visits and can take as many breaks as needed. The survey includes questions such as demographics, health status measured using the Charlson Comorbidity Index and health history [67], medication and supplement use measured using the Medication Quantification Scale, patient-reported medication symptoms adapted for MM, nonpharmacological treatments [68-70], self-rated health, mental well-being, history of and current nonmedical marijuana use, and substance use. Established scales will be used to assess variables such as pain measured by the Brief Pain Inventory, Graded Chronic Pain Scale (GCPS), Patient Global Impression of Change, and Patient-Reported Outcomes Measurement Information System (PROMIS) Pain Interference Scale [71-74]; marijuana use measured by the comprehensive marijuana motives questionnaire adapted by Lankenau et al [75-78], Medical Cannabis Expectancy Questionnaire, and Cannabis Use Disorders Identification Test-Revised; substance use measured by the World Health Organization-Alcohol, Smoking and Substance Involvement Screening Test [79]; quality of life measured by the PROMIS Global-10 [80]; depression and anxiety measured by the PROMIS Anxiety and Depression Scales [81]; social desirability measured by the Marlowe-Crowne Social Desirability Scale [82]; post-traumatic stress disorder measured by the Primary Care-Post-Traumatic Stress Disorder Scale [83]; and sleep quality measured by the Pittsburgh Sleep Quality Index [84]. A full list of core survey assessments can be found in Table 2, and information on how these measures are related to each study aim is outlined in the Analytical Strategy section. Participants will complete the survey independently or with assistance from research staff.

Follow-up surveys will be conducted at 3, 6, 9, and 12 months after the baseline assessments. Quarterly follow-up surveys at 3, 6, and 9 months will collect information on MM products (if applicable), current medications, and any health updates. At the 12-month visit, participants will complete the same core survey as the baseline, excluding demographic questions. Follow-up surveys may be completed remotely.

At every visit, participants will be asked to provide a urine sample to test for THC. For remote follow-up visits, a drug test cup will be mailed with instructions on how to upload a photo of the results.

We will also collect approximately 4 mL of blood every 3 months from each participant completing an in-person visit. Plasma will be shipped to the University of Colorado laboratory for analysis of various cannabinoids and associated metabolites (6a-OH-CBD, 7-OH-CBD, 11OH-THC, THC, CBD, CBC, CBD-gluc, THC-gluc, 7-CBD-COOH, THC-COOH, THC-COOH-Gluc, CBN, CBG, THCV, CBDV) [61-63]. Backup plasma samples may also be shipped to the University of Miami for inflammasome and/or neurodegeneration biomarker tests for exploratory analysis.

For the blood drawn at baseline and 12-month visits, the buffy coat will be frozen for analysis of telomere length, a biological measure of cellular aging [58-60]. Collected samples will be sent to the Blackburn Lab at the University of California, San Francisco, for analysis at the end of the study.

Participant’s height and weight will be measured at the baseline and 12-month study visits. Every 3 months, study staff will measure the participant’s blood pressure using a vital sign monitoring device during in-person visits.

Clinical pain measures will include the GCPS [72], which assesses chronic pain severity, and the Chronic Pain Stage, which is a stage measure based on pain frequency, intensity, time, and total number of pain sites [37,58,85]. Experimental pain measures will include punctate mechanical stimuli and pressure pain threshold to assess pain sensitivity, which has shown predictive utility for pain severity in chronic pain patients [64].

After completing the baseline and 12-month follow-up surveys, participants will complete cognitive assessments using the NIH Toolbox Cognition Battery, a standardized neurocognitive assessment tool available as an iPad app [65]. It contains 7 cognition battery tests assessing 5 cognitive subdomains: Language, Executive Function, Episodic Memory, Processing Speed, and Working Memory [65]. Table 1 shows the full list of assessments used. Three composite scores are created from these tests (Fluid, Crystallized, Total Composites), with normative standards available for comparison [86,87]. At the 12-month follow-up visit, only the fluid composite score will be used.

Participants starting MM treatment will complete a 1-week EMA using a smartphone app (mEMA, illumivu Inc. or MetricWire) before they start MM and a 3-week EMA after they begin using it. They will also complete a 1-week EMA at 3, 6, 9, and 12 months to track their MM use and longitudinal real-time outcomes. The comparison group will complete an approximately 4-week EMA at baseline and an approximately 1-week EMA at 3, 6, 9, and 12 months to answer questions regarding medication and nonpharmaceutical use.

This study will use 4 random prompts (roughly 1 every 4 hours) to assess real-time pain levels, mood (measured by the Brief Symptom Inventory[66]), and other symptoms (eg, nausea, sleepiness, dizziness) and 1 daily prompt to assess overall pain levels in the past 24 hours, the previous night’s sleep quality, and medications taken in the past 24 hours besides MM. After the baseline survey and cognitive assessment, study staff will train participants on how to use the EMA with practice questions. Participants starting MM will also be asked to upload a picture of the label on the MM products they have been using to validate the product information.

Participants will also wear a Fitbit Charge 5 or 6 (the latest model available) in parallel with EMA for objective measures of physical activity, heart rate, and sleep duration and quality. Data will be accessed via Fitabase, a service that allows centralized management of Fitbit data.

The research team will not be permitted to recruit participants or conduct visits until they receive training and pass evaluation by the study coordinators or principal investigator (PI). EMA and Fitbit data will be assessed multiple times per week to monitor for any missing data or issues with participant compliance. Weekly meetings with the research team and biweekly meetings with the investigators will be held to discuss any protocol deviations or discrepancies. The research coordinators and data manager will conduct biweekly checks of all survey data and visit log notes for discrepancies.

The total sample size was calculated based on Aims 1 and 2, for which statistical analysis will use a “difference-in-differences” approach. Power was calculated to ensure the detection of differences between the baseline and 12-month follow-up survey in Aim 2, which requires a larger sample size than the intensive longitudinal data from EMA assessments. Assuming a 3:1 ratio of MM versus comparison group and within-subject correlation of variables of 0.5, the sample size required to detect an effect size of 0.3 SD between-group difference (small to medium effect size) [88] in the outcome measured by standardized scales (eg, reduction in pain intensity measured by GCPS), with 80% power (type I error=0.05), is 352 (264 vs 88). Assuming 20% attrition, based on our pilot study (79% retention without a preplanned 12-month follow-up), the final sample size required to achieve the power is 440 (330 vs 110) [89]. For Aim 1, we will have more than 80% power to detect an effect size of 0.3 SD in between-group difference because there will be more repeated measurements in Aim 1. For Aim 3, we will have 264 participants for the analysis since only the MM group will be analyzed. After adjusting for multiple testing with the conservative Bonferroni method [90], the adjusted type I error rate becomes 0.05/7=0.007, under which we will have 99% power to detect the Pearson correlation of 0.3.

Aim 1 is to determine whether MM use leads to short-term changes in pain intensity and physical and emotional functioning, as measured in real time by EMA (subjective data) and the Fitbit device (objective data).

Key outcomes of interest include pain intensity, physical functioning, emotional functioning, and sleep. Pain intensity will be assessed from the EMA surveys using self-reported real-time pain intensity on a 0 to 100 visual analog scale. Physical functioning will be assessed from the EMA surveys using self-reported real-time pain interference and from the Fitbit using step counts. Emotional functioning will be assessed from the EMA surveys using self-reported real-time anxiety and depression symptoms (Brief Symptom Inventory).

Separate generalized linear mixed-effects models (GLMM) or generalized estimating equations (GEE) will be used to compare the temporal changes in each of the main outcomes (pain intensity level, physical and emotional functioning) between the MM and comparison groups. A difference-in-difference analysis will be conducted. A propensity score method will account for potential baseline differences between the 2 groups for analysis with inverse probability of treatment weighting. Daily MM use measured via EMA will be a time-varying exposure variable in the GLMM or GEE models to account for crossover between groups. The research team will also include time-varying covariates in these models to account for potential impacts of other non-pharmaceutical pain treatments (eg, category of treatments and number of sessions for each category) and non-MM use (self-reported frequency/quantity).

Aim 2 is to evaluate whether MM use leads to longer-term changes over a 12-month period based on self-reported pain intensity, emotional and physical functioning, health-related quality of life, and cellular aging markers such as telomere length.

Key outcomes of interest include pain intensity and interference, physical functioning, emotional functioning, cognitive function, opioid, benzodiazepine, and other medication use, sleep quality, quality of life, and cellular aging. Pain intensity will be assessed from EMA surveys, similar to Aim 1, as well as Brief Pain Inventory and GCPS responses from the core survey. Physical functioning will be assessed from EMA surveys, similar to Aim 1, and the PROMIS Pain Interference Scale from the core survey. Emotional functioning will be assessed from EMA surveys, similar to Aim 1, and the PROMIS Anxiety and Depression Scales from the core survey. Cognitive function will be assessed using the NIH Toolbox Cognition Battery and self-reported changes in cognition from the core survey. Opioid, benzodiazepine, and other medication use will be assessed using the Medication Quantification Scale from the core survey. Sleep quality will be assessed using self-reported sleep duration and quality from the EMA surveys, the Pittsburgh Sleep Quality Index from the core survey, and Fitbit-recorded sleep duration. Quality of life will be assessed using the PROMIS Global-10 and Patient Global Impression of Change from the core survey. Cellular aging will be assessed by measuring telomere length from buffy coat samples.

The research team will analyze 2 types of data: (1) intensive longitudinal data (EMA, Fitbit, opioid, and benzodiazepine dose) and (2) data collected only at baseline and 12 months (core survey, cognitive tests, biomarkers). For intensive longitudinal data, separate GLMM or GEE models will be built to compare longitudinal trajectories in each outcome (pain intensity, emotional and physical functioning, opioid dose) measured at multiple time points over 12 months between the MM group and the comparison group. As in Aim 1, we will treat MM use as a time-varying exposure variable to account for crossover and include other nonpharmaceutical pain treatments and nonmedical marijuana as time-varying covariates. The analysis will use the exact time (eg, hours) of each measurement (instead of days) for constructing the trajectory and thus does not need to average measurements within each day into a single value. For measures collected only at baseline and 12 months, generalized linear regression models will be used to compare group differences in the survey data (eg, anxiety/depression, disability, quality of life), cognitive test results, and telomere length. Additionally, milligram morphine equivalent opioid dosage from the prescription drug monitoring program will be obtained through medical records to assess differences in the risk of unintentional overdose between the MM and comparison groups.

Aim 3 is to examine among individuals initiating MM which MM product characteristics (ie, THC:CBD ratio, administration route, dose) are associated with greater improvements in the outcomes defined in Aims 1 and 2, or with a greater incidence of side effects, and explore whether individual differences (eg, sex, baseline pain phenotype) moderate these relationships. Key outcomes of interest are defined in Aims 1 and 2.

Separate GLMM or GEE models using each product characteristic as a predictor will be developed. As in Aims 1 and 2, we will treat MM use as a time-varying exposure variable to account for crossover and include other nonpharmaceutical pain treatments and nonmedical marijuana as time-varying covariates. The research team will construct a multivariate model to include all 3 product characteristics (ie, THC:CBD ratio, administration route, dose) as predictors for each main outcome using stepwise model selection or least absolute shrinkage and selection operator to account for potential interactions within these product characteristics. To determine significant moderating effects of interpersonal differences, separate GLMM or GEE models will be conducted by gender (male vs female) or at each level of the moderator (low/moderate/high stage of pain, low/high marijuana expectancy, low/high social desirability bias) to assess whether the main effect of the MM product characteristic on each main outcome differs by group membership or the level of the moderator.

Missing data will be examined to determine whether information is missing at random by comparing participants with and without missing data. If no significant differences are observed in key demographics, missing at random will be assumed. If missingness exceeds 5%, we will use multiple imputation methods to impute missing responses. Sensitivity analyses will be conducted to assess the impact of the imputed data.

Adverse events and discomforts may occur due to participation in the study. In the event of adverse events, the PI will determine whether the event is serious and related to the study procedures. If this is the case, the PI will report the event to the NIH and IRB within 72 hours of the event. If the event is not serious, it will be reported to the NIH and IRB during regular status reports.

Our study will collect data on participants’ medications, including MM. Breaches of confidentiality could occur through communications with participants, loss of research data, or transfer of data.

The assessments used in this study address some sensitive issues, such as mental well-being and drug use. The major disadvantages of these assessments are the time required to complete them and the potential breach of confidentiality. Our past experience with these measures indicates that they are acceptable to participants.

There are minimal risks associated with these tests. Participants may experience fatigue and feelings of frustration while completing the cognitive function tests. Some participants may also experience stress associated with being tested, though this tends to be quite limited.

The EMA involves minimal risk as it only collects self-reported data related to participants’ symptoms and their relief, experienced side effects, and other related feelings (eg, anxiety, depression). The major disadvantage of this assessment is the time required to complete daily reports. The possibility of a confidentiality breach is very low because the data collected from the mobile app will be encrypted and pushed to the cloud-based storage database. No data are ever stored on the server’s file system but always in the database. Access to the database is gated, so entry is only permitted to users entering through the approved route (ie, you cannot hack your way into the database by guessing the URL).

Procedures may be uncomfortable or unpleasant. Participants may experience some temporary discomfort from the pressure and mechanical pain testing. However, if they feel the pain is greater than they wish to tolerate, they can stop any of the procedures at any time.

The risk is minimal for urine sample collection, although some people may experience embarrassment when providing the sample.

The risks of drawing blood from a vein include discomfort at the site of puncture; possible bruising and swelling around the puncture site; rarely, an infection; and uncommonly, faintness from the procedure.

The risk is minimal when wearing a Fitbit during the EMA period. All data from Fitbit do not include personal identifiers. Data will be stored temporarily on cloud-based storage and will be downloaded regularly onto a secure UF server. Data will be encrypted during the transmission process to ensure security.

The risk is minimal for collecting these measures, including height, weight, blood pressure, and heart rate. Some participants may feel slightly uncomfortable or embarrassed about having some of these measures taken.