By Dr. Benjamin Caplan, MD | Board-Certified Family Physician, CMO at CED Clinic | Evidence Watch
A Phase II randomized, double-blind, active-controlled trial published in the Journal of Cannabis Research found that a traditional Thai cannabis-based multi-herbal formulation matched lorazepam – a commonly prescribed benzodiazepine – for improving sleep quality in chronic insomnia over four weeks, with a comparable safety profile and no clinically significant adverse effects on liver, kidney, or cardiac function.
Cannabis Insomnia Treatment Matches Lorazepam in Head-to-Head RCT
In the first randomized controlled trial directly comparing a cannabis-based formulation to an active benzodiazepine comparator for chronic insomnia, researchers from Suan Sunandha Rajabhat University and Mahidol University in Thailand found that the herbal formula produced PSQI improvements of 9 points over baseline, larger than those reported in most prior cannabinoid sleep trials, while matching lorazepam on the predefined non-inferiority criterion and on every safety measure tested.
#89 Exceptional Clinical Interest
Sleep is among the top three reasons patients at CED Clinic seek cannabis care; this is the first RCT design to pit a cannabis formulation head-to-head against an active benzodiazepine comparator, making it directly relevant to the clinical conversations that happen in this practice every week.
Sleep is the condition I hear about most often in cannabis medicine. Patients arrive after years of Ambien, or Ativan, or nothing that works, and they want to know if cannabis is a real option or just wishful thinking. The honest answer has always been that the evidence is promising but incomplete, most cannabis sleep trials compare against placebo, not against the medications patients are actually trying to stop taking. This trial changes the comparison. It puts a cannabis formulation head-to-head against lorazepam under rigorous blinded conditions, and it finds non-inferiority on the primary endpoint. That is a genuinely useful piece of information, and it deserves a careful reading from anyone counseling patients on sleep.
This Phase II randomized, double-blind, active-controlled non-inferiority trial was conducted by Kamoltham, Chokchaisiri, Yongram, Sripan, Im-iam, and colleagues at Prapokklao Hospital in Chanthaburi, Thailand, and published in the Journal of Cannabis Research on February 24, 2026. One hundred adults with chronic insomnia (PSQI score greater than 5 at baseline) were randomized to receive either the traditional Thai cannabis-based Anti-Pom-Leung Fever medicine or lorazepam 0.5 mg for four weeks under double-dummy blinding. The herbal formulation is an eight-herb combination with Cannabis sativa leaves constituting 50% of total mass by weight, codified in the National Thai Traditional Medicine Pharmacopeia. The study was funded by the Thai Traditional Medical Knowledge Fund and conducted under full IRB oversight. (Kamoltham T et al., J Cannabis Res. 2026;8:46. DOI: 10.1186/s42238-026-00415-x)
Of the 100 enrolled participants, 82 completed the study (41 per group). Mean PSQI scores at baseline were 12.44 in the herbal group and 13.05 in the lorazepam group — comparable and well above the clinical threshold for insomnia. At week 4, mean PSQI scores had fallen to 3.44 and 4.78, respectively. The mean difference of negative 1.34 points (95% CI: negative 2.99 to 0.31) fell below the predefined non-inferiority margin of 2.1, demonstrating non-inferiority. There was no significant time-by-treatment interaction on the PSQI, meaning the trajectory of sleep improvement was statistically similar between the two groups across the four-week period. Quality of life and stress scores improved significantly in both arms with no between-group differences. Safety monitoring — covering liver enzymes, renal function, hematology, lipid profiles, blood glucose, cardiac EKG, and blood pressure — found no clinically significant differences between groups at any time point. Minor adverse events (dizziness, drowsiness, gastrointestinal upset) occurred in comparable proportions in both groups and resolved spontaneously.
The Benzodiazepine Problem in a Cannabis Practice, and Why This Trial Answers a Different Question
Most of the patients I see who want to use cannabis for sleep are not coming from a baseline of nothing. They are coming from benzodiazepines. Or from Z-drugs like zolpidem. Sometimes from alcohol. More often from a combination of two of those things, alternated across years of trying to manage sleep that stopped working on its own. When they ask me whether cannabis is a reasonable alternative to their current prescription, I want to be able to tell them something specific. Not “there are studies showing cannabis helps with sleep” that is true but unhelpful in a clinical context where the comparison is an actual drug they are currently taking. This Thai trial matters precisely because it makes that comparison directly.
What did they find? Both treatments produced large PSQI improvements, 9 points in the cannabis group, 8.3 in the lorazepam group, from baselines solidly in the clinical insomnia range. Both groups crossed below the PSQI threshold of 5 (indicating good sleep quality) by the end of the study. The cannabis formulation met its non-inferiority criterion. Safety was equivalent across a rigorous set of laboratory and cardiac endpoints. That is a real result. It is not trivial.
Now the harder interpretation. This formulation is not a dispensary product. It is a standardized multi-herb capsule with cannabis constituting 50% of the mass, the predominant cannabinoids being THC at 6.6 mg/g, followed by CBN and THCV, manufactured under pharmacopoeial standards by researchers using an identified Thai cannabis strain. That is a different product from what a Massachusetts patient buys at a licensed dispensary, in terms of cannabinoid profile, dosing precision, and the presence of the other seven herbal co-constituents (sandalwood, neem, five-leaved chaste leaf, ginger, pepper, and long pepper). How much of the effect comes from the cannabis? How much from the herbal matrix? The study is not designed to answer that. The discussion notes that the other herbs carry their own anxiolytic and sedative histories in traditional pharmacopoeias. That is worth acknowledging honestly: this is not a pure-cannabinoid trial. It is a trial of a complex plant-based formulation that happens to contain a significant amount of cannabis.
That complexity does not diminish the result. If anything, it raises an interesting question about whether the entourage effect, the idea that cannabis works differently in the context of a broader phytochemical matrix than as an isolated cannabinoid, has clinical legs in a sleep context. The authors invoke it, reasonably. The biological plausibility is there. CB1 receptor activation modulates orexin-mediated arousal circuits. THC reduces sleep latency. CBN, less studied, has traditional use and emerging preclinical signals for sedation. THCV’s role in this context is speculative, but its presence within a complex herb is not disqualifying. Whether the entourage framework explains the full 9-point PSQI drop or whether the other herbs contribute meaningfully is a genuinely open question, and it would take a different trial design, factorial, with arm-by-arm dismantling of the formula, to answer it.
The most significant limitation is the absence of a placebo arm. A non-inferiority design does what it says: it tests whether treatment A is not worse than treatment B. It does not tell you whether B is doing something beyond placebo. The authors address this thoughtfully, noting that published placebo responses on the PSQI typically fall in the range of 1 to 2 points, making the 8 to 9 point drops observed here unlikely to be attributable to placebo alone. That argument is reasonable but not definitive. The most honest reading is: the placebo component almost certainly explains some of the improvement; the active treatment likely explains most of it; and only a placebo-controlled arm would let us quantify the split. The researchers call for exactly that in future work, and they are right.
The safety data deserve particular attention for clinicians. Both groups showed minor EKG changes at week 4, five participants in the herbal group (mostly sinus bradycardia) and eight in the lorazepam group (primarily conduction abnormalities). Neither set of findings was statistically significant or led to discontinuation. Liver enzymes, kidney function, lipid panels, and blood glucose remained within normal ranges in both groups. For patients asking whether cannabis is safer than their current benzodiazepine, this trial does not produce data showing cannabis is safer, it produces data showing it is not detectably worse across the endpoints measured. That is an important distinction. It is also, frankly, more than we usually have.
What this means for a Massachusetts patient who is using lorazepam or another benzodiazepine for sleep is nuanced. The study tells us a cannabis formulation can match a short-acting benzodiazepine on sleep quality metrics over 28 days in a clinical trial setting. It does not tell us whether cannabinoid-based sleep therapies produce fewer long-term adverse effects, dependence, cognitive effects, rebound insomnia on discontinuation, because it did not follow patients past four weeks. Those long-term concerns are exactly the clinical reasons many patients want to get off benzodiazepines. This trial does not resolve them. What it does is provide the first randomized evidence that a cannabis formulation is at least as effective as the medication they are trying to stop, over the time window studied. That is a meaningful starting point for a clinical conversation, not a sufficient basis for a unilateral switch.
The broader context here matters too. The Thai regulatory environment permitted this research in a way that remains genuinely difficult in the United States. Thailand legalized medical cannabis in 2018, and traditional cannabis-based formulas codified in the national pharmacopeia can be studied with active comparators at government-funded hospitals. The fact that this trial happened at all is partly a function of that regulatory flexibility. It is worth noting as we observe the slow movement toward Schedule III reclassification in the United States and the downstream research opportunities that reform would open. This is the kind of trial that the American sleep medicine literature still lacks, not because researchers have not wanted to run it, but because the regulatory infrastructure has not allowed it.
The existing cannabinoid insomnia literature has shown consistent positive effects against placebo, though effect sizes have varied widely. Walsh et al. (2021, Sleep) found that a THC/CBD oromucosal spray significantly reduced the Insomnia Severity Index versus placebo in a randomized crossover trial. Ried et al. (2023, J Sleep Res) reported improvements of approximately 3 to 4 points on the ISI with medicinal cannabis in a randomized placebo-controlled crossover trial. Prior THC-focused trials, including work with nabilone by Ware et al. in fibromyalgia patients, showed PSQI improvements of roughly 3 points — substantially less than the 9-point drop seen in the Thai trial. The magnitude of improvement in the Kamoltham et al. study is larger than virtually all prior cannabis sleep studies, possibly reflecting the complex herbal matrix, the specific THC/CBN/THCV combination, or simply higher baseline severity in this population. The PSQI reduction also exceeds what cognitive behavioral therapy for insomnia (CBT-I) typically produces in the short term, though CBT-I has superior long-term outcomes and remains the first-line recommendation in most clinical guidelines.
From a Massachusetts clinical practice standpoint, the relevant patient population includes adults who have been on benzodiazepines long-term, those whose physicians are trying to taper them, and those who have heard that cannabis helps with sleep and want informed guidance on whether and how to try it. The drug-interaction profile is worth explicit attention: the herbal formulation used in this trial is not a product available in Massachusetts dispensaries, and the cannabinoid interaction profile with CNS depressants (additive sedation) differs meaningfully from the interaction profile of lorazepam with the same class of drugs. Patients combining cannabis with any benzodiazepine or Z-drug should be counseled carefully about CNS depression risk, and any transition should be supervised rather than self-directed.
| Study Type | Phase II randomized, double-blind, active-controlled non-inferiority trial. Double-dummy blinding with identical capsule appearance. CONSORT-compliant reporting. |
| Population | 100 enrolled; 82 completed (41 per group). Thai adults aged 25–70 with chronic insomnia (PSQI >5). Excluded: hepatic/renal disease, cardiovascular disease, anticoagulant or antihypertensive medications, alcohol use within 1 month, illicit drug history, pregnancy. |
| Intervention | Anti-Pom-Leung Fever medicine (500 mg capsules, 8 capsules/day; 4 before breakfast + 4 at bedtime) vs. lorazepam 0.5 mg at bedtime for 28 days. Cannabis sativa leaves = 50% of formulation by weight. Predominant cannabinoids: Δ9-THC (6.645 mg/g), CBN (5.759 mg/g), THCV (4.103 mg/g). |
| Primary Outcome | Pittsburgh Sleep Quality Index (PSQI) at week 4. Non-inferiority margin: 2.1 points (upper bound of two-sided 95% CI for mean difference). |
| Key Finding | Week 4 PSQI: 3.44 (cannabis group) vs. 4.78 (lorazepam group). Mean difference −1.34 (95% CI: −2.99 to 0.31). Non-inferiority demonstrated. Both groups crossed below PSQI threshold of 5. No significant between-group differences in quality of life, stress, or any safety parameter. |
| Authors | Thavatchai Kamoltham, Suwadee Chokchaisiri, Chawalit Yongram, Panupan Sripan, Surasak Im-iam, and colleagues (Suan Sunandha Rajabhat University; Prapokklao Hospital; Mahidol University; Thailand) |
| Journal / DOI | Journal of Cannabis Research, 2026;8:46. DOI: 10.1186/s42238-026-00415-x |
| Limitations | No placebo arm; active comparator design cannot rule out placebo contribution. 28-day follow-up only; no data on dependence, tolerance, or rebound after discontinuation. Subjective sleep outcomes only; no polysomnography or actigraphy. Multi-herb formulation; cannot isolate cannabis contribution. Thai regulatory context limits direct extrapolation to other markets. 18% dropout rate. Per-protocol analysis only. |
Phase II randomized controlled trials occupy a strong position in the evidence hierarchy. This one used double-dummy blinding (both groups received identical capsule regimens) to maintain allocation concealment, block randomization for balanced group assignment, and a predefined non-inferiority margin with appropriate statistical power. Those are markers of methodological rigor. The limitation is the non-inferiority design itself: it establishes equivalence with a known treatment but cannot isolate active treatment effects from placebo. The 28-day window is adequate for short-term sleep outcomes but insufficient for assessing dependence, withdrawal, tolerance, or long-term safety. The study is better described as Phase II evidence of short-term non-inferiority — which is genuinely useful — than as definitive proof of long-term efficacy or safety superiority over benzodiazepines.
The Kamoltham et al. trial extends a growing body of randomized evidence for cannabinoid-based therapies in insomnia. Where it breaks new ground is the active comparator design — prior trials have compared cannabis to placebo, but none has run a blinded head-to-head against a benzodiazepine. The 9-point PSQI drop is substantially larger than what most prior cannabis sleep trials have achieved (typically 3 to 4 points), which may reflect the higher baseline severity in this cohort, the complex herbal formulation, or both. The PSQI reductions also exceed what most short-term behavioral interventions produce, though without a placebo arm the active treatment contribution cannot be precisely quantified. The trial aligns with Thailand’s broader movement toward evidence-based integration of traditional cannabis medicine, paralleling similar research trajectories in Israel, Australia, and Canada where regulatory environments have enabled more rigorous cannabis research than has been possible in the United States. The 2022 Medical Marijuana and Cannabidiol Research Expansion Act created some structural openings for similar research in the US, but the infrastructure is still scaling.
Yes. An intention-to-treat analysis (including the 18 dropouts, conservatively attributed as treatment failures) would likely have modestly reduced the within-group PSQI improvements and may have narrowed the confidence interval around non-inferiority. A placebo arm would have allowed decomposition of active treatment vs. spontaneous remission vs. expectation effects. A factorial design with isolated cannabis extract versus the full multi-herb formulation would have clarified the cannabis-specific contribution. A longer follow-up period — 12 weeks or more — would have revealed whether the improvements were durable and whether either treatment produced tolerance or withdrawal effects. None of these analytical choices invalidates the core finding, but each would sharpen the interpretation in ways the current design cannot provide.
Misreading 1: “Cannabis is now proven to be as good as benzodiazepines for sleep.” This trial demonstrates short-term non-inferiority for a specific multi-herb cannabis formulation against lorazepam 0.5 mg in a Thai population over 28 days. That is not the same as proving equivalence for all cannabis products, all populations, and all time horizons. The absence of a placebo arm also limits causal attribution of the full treatment effect.
Misreading 2: “Cannabis is safer than my sleeping pill, so I should switch.” The trial showed comparable safety on laboratory and cardiac markers over four weeks. It did not compare long-term dependence risk, cognitive effects, or discontinuation syndrome — the very reasons most patients want off benzodiazepines. Any transition should be discussed with and supervised by a physician familiar with both agents.
Misreading 3: “I can replicate this with a dispensary product.” The formulation tested is a standardized multi-herb capsule manufactured under pharmacopoeial conditions in Thailand. Its cannabinoid profile (THC-dominant with CBN and THCV, within a herbal matrix of seven other plants) is not equivalent to a typical retail cannabis product. Product selection and dosing for sleep remain highly individualized and require clinical guidance.
A Phase II RCT published in February 2026 found that a traditional Thai cannabis-based formulation matched lorazepam 0.5 mg on the primary sleep quality endpoint over four weeks, with a comparable safety profile across all laboratory and cardiac measures tested. The 9-point PSQI improvement in the cannabis group is larger than most prior cannabinoid sleep trials have reported. The critical limitation is the absence of a placebo arm, which leaves the active treatment contribution partially unquantified. What this trial adds to the clinical conversation is direct — a cannabis-based formulation head-to-head against a benzodiazepine, under blinded conditions, in a randomized controlled design. That is a data point clinicians and patients have needed. It does not answer the long-term questions. It answers the short-term comparison question, and it answers it clearly.
