DOI: 10.1177/25785125261475334 ISSN: 2578-5125

Development and Internal Validation of a Clinical Prediction Tool for Cannabis Dosage Requirements in Chronic Low Back Pain: The Cannabis Dosage Risk Score

Muhammad Khatib, Hamza Murad, Dror Robinson, Eitan Lavon, Feras Qawasmi, Mustafa Yassin
Show PDF Cite

Introduction:

Medical cannabis dosage for chronic low back pain (LBP) varies substantially between patients, yet clinicians lack validated tools to predict individual dosage requirements. We developed and internally validated the Cannabis Dosage Risk Score (CDRS), a practical bedside prognostic tool to identify patients likely to require high-dose (≥60 g/month) cannabis therapy.

Materials and Methods:

We conducted a retrospective Target Trial Emulation analysis of a prospectively followed cohort of 225 patients with chronic LBP who consented to research participation and completed baseline assessments between January 1 and May 31, 2020, at an Israeli pain clinic. Patients were categorized by stabilized 5-year cannabis dosage as low dose (<40 g/month, n = 121), medium dose (40–59 g/month, n = 41), or high dose (≥60 g/month, n = 63). Primary analysis compared low-dose versus high-dose groups ( n = 184); medium-dose patients were analyzed separately. Multivariable logistic regression identified independent predictors, which were converted to a weighted point-based scoring system. Model performance was assessed using bootstrap-validated area under the receiver operating characteristic curve (AUC-ROC), Hosmer–Lemeshow calibration, and decision curve analysis.

Results:

The final CDRS incorporates 4 baseline clinical variables: disability status (odds ratio [OR] 5.10, 95% confidence interval [CI]: 2.42–10.76), absence of baseline opioid use (OR: 2.38, 95% CI: 1.06–5.26; equivalently, opioid-use OR: 0.42, p = 0.035), lower Patient-Reported Outcomes Measurement Information System (PROMIS) Sleep T-score (OR: 0.95 per point, 95% CI: 0.92–0.98), and lower Hospital Anxiety and Depression Scale (HADS) combined score (OR: 0.92 per point, 95% CI: 0.86–0.98). With an events-per-variable ratio of 15.8 and variance inflation factors <1.3, the model achieved AUC-ROC of 0.80 (optimism-corrected 0.78) with adequate calibration (Hosmer–Lemeshow p = 0.24). Ordinal logistic regression confirmed predictor validity across all three dose categories. Decision curve analysis demonstrated net clinical benefit across threshold probabilities of 10–70%. Risk stratification yielded: low risk (0–3 points, 17.2% high-dose rate), moderate risk (4–7 points, 24.7% high-dose rate), and high risk (≥8 points, 76.7% high-dose rate). At 5-year follow-up, all dose groups showed significant pain reduction (Brief Pain Inventory severity Δ: low −4.9, medium −4.7, high −6.2; analysis of variance p = 0.0005); high-dose patients had the greatest improvement, refuting the hypothesis of dose escalation by nonresponders. Among patients using baseline opioids, 93.1% (135/145) achieved opioid cessation at 5 years.

Conclusions:

The CDRS provides a practical, internally validated four-item bedside tool for predicting cannabis dosage requirements among patients with chronic LBP who initiate and continue medical cannabis therapy. Patients with disabilities who do not have baseline opioid use and who experience less sleep disturbance or psychological distress are most likely to require high-dose therapy. The 100% 5-year follow-up reflects the cohort’s intake-based definition, its early-2020 (pandemic-onset) enrollment period, and structural features of the Israeli regulatory framework rather than post hoc selection of responders; retention in the same clinic’s other cohorts was lower (82.5–92.9% at 5 years), and generalization to jurisdictions without comparable regulatory follow-up structures should be made cautiously. External validation in diverse populations is warranted.

More from our Archive