Long-Term Inhaled Cannabis Therapy for Chronic Low Back Pain: A Five-Year Retrospective Analysis of Prospectively...

Background/Objectives: Chronic low back pain (CLBP) affects approximately 20% of the global population and is a leading cause of years lived with disability. Long-term, real-world evidence for inhaled cannabis in patients refractory to conventional multimodal therapy remains scarce. We assessed the five-year efficacy and safety of inhaled cannabis in CLBP patients who had documented failure of ≥1 year of opioid analgesics, anticonvulsants, antidepressants, NSAIDs, and physiotherapy, with each patient serving as their own historical control. Methods: We analyzed prospectively collected clinical data from 241 consecutive adults with treatment-refractory CLBP (mean age 49.3 ± 14.9 years; 37.8% female; mean pain duration 15.1 years) initiated on inhaled medical cannabis (predominantly smoking, THC 4-22%, CBD 2-22%) in a single-center tertiary orthopedic clinic between 2020 and 2025 (Hasharon Hospital, Rabin Medical Center, Israel; IRB protocols 0807-21-RMC and 0634-25-RMC). Year-0 outcomes during conventional therapy were compared with outcomes at Years 1-5 on cannabis. Primary outcomes were the Numeric Rating Scale (NRS), Oswestry Disability Index (ODI), and Brief Pain Inventory severity/interference (BPI-S/BPI-I). Concomitant-medication trajectories were a secondary outcome. The primary analysis was a mixed model for repeated measures (MMRM) with random intercept and slope, REML estimation, and time as a categorical fixed effect. Multiple imputation (MAR, m = 20, Rubin's rules) was the primary missing-data approach; complete-case and tipping-point pattern-mixture sensitivity analyses were used. A multivariate Hotelling T² provided a joint test across the four correlated PROMs. Concomitant-medication discontinuation was modeled with GEE logistic regression and exact McNemar tests. Time to discontinuation was estimated by Kaplan-Meier and Cox regression. The Bonferroni-adjusted significance threshold for the four primary outcomes was α = 0.0125. BioWell gas-discharge-visualization (GDV) parameters were exploratory only. Results: Of 241 patients, 238 (98.8%) provided Year-5 data and 224 (92.9%) remained on cannabis at Year 5; only five patients (2.1%) discontinued for adverse events or inefficacy. All four primary PROMs improved markedly and durably. MMRM-estimated Year-5 minus Year-0 changes were: NRS -5.36 (95% CI -5.65, -5.07), ODI -17.68 (95% CI -19.73, -15.63), BPI-S -6.73 (95% CI -6.99, -6.47), and BPI-I -3.41 (95% CI -3.65, -3.16); all four contrasts had |z| ≥ 16.9 and p < 10^-20. MI-pooled estimates were within 0.05 of MMRM (FMI < 0.03 for all outcomes). Hotelling T² was F(4, 232) = 872.8, p < 10^-20. At Year 5, 89.2% achieved ≥30% NRS reduction, 77.2% ≥ 50%, and 93.4% met the NRS minimum clinically important difference (MCID); ODI MCID 65.6%, BPI-S MCID (≥1 pt) 98.3%, BPI-I MCID (≥1 pt) 91.3%. Concomitant opioid use fell from 100% at baseline to 4.6% at Year 5 (within-patient absolute risk reduction 95.4%, McNemar exact p = 1.16 × 10^-69), NSAID from 100% to 7.1%, SSRI/SNRI from 80.5% to 5.4%, and gabapentinoid from 38.6% to 2.5%. The ARR-derived NNT for opioid discontinuation was 1.05; this NNT is referenced to each patient's own documented maximal-conventional-therapy state and is not equivalent to a between-arm randomized-trial NNT. Cannabis dose × time interaction was consistent with no pharmacological tolerance (β = -0.0044 per gram-month per year, p = 0.074). Across 1205 patient-years of cannabis exposure (calculated as 241 patients × 5 follow-up years from Year 1 through Year 5; baseline Year 0 represents pre-cannabis state and is not included in person-time on cannabis), 1338 organ-system AE events were recorded at 1.110/patient-year (Poisson 95% CI 1.05-1.17); 99.8% of graded events were mild (grade 1), with ocular (476 events, 0.40/PY), cognitive (460, 0.38/PY), and gastrointestinal (368, 0.31/PY) reactions predominating. The Year-3 retention dip reflected a documented telemedicine-clinic phenomenon during 2022-2024, with patients returning to in-person follow-up by Year 4-5. BioWell GDV discriminated NRS ≥ 4 only at chance level (BWS AUC 0.574, 95% CI 0.54-0.60; BWV AUC 0.51). Conclusions: In a treatment-refractory CLBP cohort with five-year longitudinal follow-up, inhaled cannabis was associated with large, sustained, and statistically robust improvements in pain, disability, and pain interference, accompanied by near-total displacement of opioids, NSAIDs, antidepressants, and gabapentinoids. These observational associations, although mechanically less susceptible to bias for the binary medication-discontinuation outcomes than for self-reported PROMs, cannot be interpreted causally in the absence of a concurrent randomized control arm and may reflect a combination of pharmacological effect, regression to the mean from a high pre-treatment baseline, expectancy and self-selection effects intrinsic to an actively chosen open-label therapy, and secular trends in pain reporting. The within-patient benefit-risk profile-ARR-derived NNT ≈ 1 for opioid sparing against a predominantly mild adverse-event burden-supports consideration of cannabis as a potentially clinically meaningful, opioid-sparing option in patients who have failed multimodal conventional therapy, pending confirmation in randomized comparative trials.