Cannabinoid Antitumor Activity in Preclinical Cancer Models

Key Findings

• Cannabinoid monotherapy produced statistically significant tumor reductions in glioblastoma (−981 mm³), breast cancer (−403 mm³), and lung cancer (−562 mm³).
• CBD plus doxorubicin in breast cancer produced the largest combination effect (−1295 mm³).
• THC monotherapy in breast cancer trended protumoral (+98 mm³); THC:CBD plus chemotherapy in prostate cancer produced statistically significant antagonism (+306 mm³).
• Hepatocellular carcinoma, pancreatic, colon, and skin cancers showed essentially null cannabinoid effects.
• Risk of bias was predominantly high, heterogeneity ranged from I² 34 to 100 percent, and findings are hypothesis-generating rather than definitive.
• Glioblastoma remains the most defensible translational target, supported by Phase 1b clinical data (Twelves 2021) and the ongoing ARISTOCRAT Phase 2 trial.

Introduction

Cannabinoids have attracted growing attention as candidate antitumor agents, with two decades of preclinical work suggesting they can influence proliferation, apoptosis, angiogenesis, and metastasis across diverse cancer cell types. The mechanistic foundations are reasonably well documented, the body of supportive data is substantial, and small clinical trials have begun translating signals from the bench into early human evidence. Yet the field's most careful practitioners have consistently flagged a difficult problem: the preclinical literature is heterogeneous, methodologically variable, and at times contradictory, with some studies showing protumoral effects from the same compounds that produce tumor regression elsewhere. Pooling that body of work into a coherent quantitative picture has been overdue.

A 2026 systematic review and meta-analysis published in Pharmaceuticals by Creangă-Murariu and colleagues at Semmelweis University and Grigore T. Popa University of Medicine and Pharmacy provides exactly that synthesis, integrating 189 preclinical studies into pooled estimates across multiple tumor types [1]. The analysis is prospectively registered (PROSPERO CRD42025543744), PRISMA-compliant, and explicit about its limitations. Its findings are notable for both what they confirm and what they qualify, and the authors describe their results as hypothesis-generating rather than definitive, a framing this article preserves throughout.

What follows is an evidence-grounded survey of the meta-analysis findings, organized by tumor type and compound class, with translational context drawn from the small but growing body of clinical trial data. The picture that emerges is neither uniformly favorable nor uniformly discouraging. Cannabinoids show context-dependent antitumor activity, and that context matters in ways that should shape both future research design and the interpretation of any clinical claims advanced ahead of confirmatory evidence.

The Multi-Tumor Landscape

The Creangă-Murariu meta-analysis searched MEDLINE, Embase, and Cochrane CENTRAL through April 2024, identifying 27,690 records that were narrowed to 189 eligible articles for the systematic review, with 52 of those contributing data to the quantitative meta-analysis [1]. The included studies covered both in vitro tumor cell lines and in vivo animal models, with cannabinoid interventions ranging from purified CBD and THC to THC:CBD-rich preparations and synthetic cannabinoid agonists such as WIN-55,212-2 and JWH-133. Comparators were vehicle-treated controls or standard chemotherapeutic agents, and outcomes were standardized to percent cell viability for in vitro work and tumor volume in cubic millimeters for in vivo studies.

The breadth of tumor types represented is striking. Breast cancer accounted for the largest group of studies, followed by glioblastoma, with substantial representation also for colorectal, hepatocarcinoma, pancreatic, lung, prostate, and skin cancers. Bladder, cervical, cholangiocarcinoma, gastric, leukemia, ovarian, mesothelioma, head and neck, and several other tumor types appeared in smaller numbers. This range matters because the meta-analysis is, in effect, the most comprehensive attempt yet to determine where cannabinoid activity is most reliably observed and where it is not.

When findings are organized by tumor type, the pattern is unmistakably uneven. Statistically significant tumor volume reductions emerged consistently in glioblastoma, breast cancer, lung cancer, and certain prostate cancer subgroups. By contrast, hepatocarcinoma showed essentially no effect, colon cancer responses were modest and nonsignificant, pancreatic cancer findings were variable, and skin cancer effects were near-neutral. Even within tumor types where signals appeared, the magnitude of effect varied substantially across compound classes, dosing regimens, and model systems. The meta-analysis quantifies this variability through I² statistics that range from 34 to 100 percent, indicating moderate to extreme heterogeneity in most subgroup analyses.

Hytiva Research's earlier coverage in Cancer Care Evolution mapped the broader landscape of cannabis in cancer care qualitatively. The 2026 meta-analysis tightens what was previously a directional picture into a quantitative one. Where prior reviews identified glioblastoma as the highest-priority translational candidate based on consistent preclinical responses and early clinical signals, the pooled effect sizes and confidence intervals from the new analysis confirm that judgment under formal statistical scrutiny.

Glioblastoma: The Most Reproducible Signal

Glioblastoma emerged as the tumor type with the most reproducible cannabinoid response in the 2026 meta-analysis. Cannabinoid monotherapy produced a pooled tumor volume reduction of 980.58 mm³ (95% CI: 1270.28 to 690.88 mm³) compared with vehicle controls, a statistically significant effect across the overall analysis [1]. Disaggregated by compound class, CBD-rich preparations, high-THC formulations, and THC:CBD combinations all reached statistical significance individually. Synthetic cannabinoid subgroups (WIN-55,212-2 and JWH-133) trended in the same favorable direction but did not reach significance on their own, with wide confidence intervals reflecting the small number of contributing studies. The largest pooled point estimates in absolute terms were observed for high-THC drugs at 1082.49 mm³ reduction and for WIN-55,212-2 at 1259.11 mm³ reduction.

The mechanistic foundations for this signal have been characterized in detail across multiple research groups. THC and CBD both engage cannabinoid receptors that glioma cells overexpress relative to healthy brain tissue, triggering sustained ceramide accumulation, endoplasmic reticulum stress, and autophagy-mediated apoptosis [2,3,4]. CBD additionally exerts effects through multiple non-receptor pathways, including TRPV2 channel modulation, RAD51 suppression, and inhibition of mitochondrial respiration, providing convergent routes toward tumor cell death [5,6].

The combination findings are more nuanced. When cannabinoids were added to temozolomide, the standard chemotherapy for glioblastoma, the meta-analysis observed a tumor volume reduction of 220.65 mm³ compared with temozolomide alone (95% CI: 579.34 to 138.03 mm³). The direction of effect favored combination therapy, but the confidence interval crossed zero, and the result did not reach statistical significance. The authors note that this analysis drew on limited combination-therapy data, primarily from López-Valero 2018 work targeting glioma initiating cells with cannabinoid and temozolomide combinations [7]. Earlier work by Torres and colleagues had demonstrated that THC plus temozolomide produced strong antitumor activity in glioma xenografts, including in tumors resistant to temozolomide monotherapy [3], but the meta-analytic pooling reflects the heterogeneity that emerges when multiple combination-strategy studies are aggregated under a single estimate.

In vitro pooled analyses showed a moderate reduction in glioblastoma cell viability, with a mean difference of 18.77 percentage points (95% CI: 27.15 to 10.39), and CBD appeared to drive a more pronounced effect than other cannabinoid classes [1]. Heterogeneity was very high, with I² in the high 90s, reflecting the diverse cell lines, exposure durations, and dosing protocols across studies. This pattern of strong directional consistency with substantial magnitude variability is characteristic of cannabinoid oncology research, and it points to the importance of standardized experimental design for future work.

Breast Cancer: Combination Strategies Take the Lead

Breast cancer findings in the 2026 meta-analysis produced one of the most clinically interesting patterns in the dataset, with sharply divergent effects depending on cannabinoid class and combination strategy. Cannabinoid monotherapy in breast cancer xenograft models produced a pooled tumor volume reduction of 402.64 mm³ (95% CI: 671.84 to 133.45 mm³), a statistically significant effect [1]. Disaggregation by compound class, however, revealed substantial heterogeneity within that summary number.

Synthetic cannabinoids achieved the largest absolute monotherapy effect at 1703.90 mm³ reduction (95% CI: 2863.43 to 544.37 mm³), again with wide confidence intervals reflecting limited study counts. CBD and THC:CBD preparations produced moderate, statistically meaningful reductions of 470.33 mm³ and 397.12 mm³, respectively. THC-rich preparations, by contrast, produced a pooled effect in the protumoral direction, with mean tumor volume increasing by 97.79 mm³ relative to vehicle controls.

This last finding warrants careful attention. McKallip and colleagues reported in 2005 that THC enhanced breast cancer growth and metastasis in murine models through suppression of antitumor immune responses, a result subsequently corroborated by additional work suggesting that CB1-mediated signaling can attenuate cytotoxic immunity against certain tumor types [8,9]. The meta-analytic pooling captures this concern at the aggregate level: in breast cancer specifically, THC alone may not behave as a reliable tumor suppressor and may actively work against tumor control depending on the experimental context.

The combination findings reverse this picture. When CBD was added to doxorubicin, a standard chemotherapy for several breast cancer subtypes, the pooled effect was a tumor volume reduction of 1295.19 mm³ compared with doxorubicin alone (95% CI: 1664.33 to 928.05 mm³). This was the largest and tightest combination-therapy effect observed anywhere in the meta-analysis. THC added to chemotherapy produced a similarly favorable shift at 1100.39 mm³ reduction (95% CI: 1503.79 to 698.99 mm³), reversing its problematic monotherapy profile. THC:CBD combinations with chemotherapy produced a more modest but still statistically significant additional benefit of 70.06 mm³ reduction.

These patterns align with mechanistic reports describing cannabinoid sensitization of breast cancer cells to cytotoxic agents through autophagy induction, ceramide accumulation, and endoplasmic reticulum stress [10]. Caffarel and colleagues had earlier demonstrated that cannabinoids can interfere with ErbB2/HER2-driven breast cancer progression through Akt inhibition in animal models [11], and Blasco-Benito and coworkers showed in 2018 that botanical THC:CBD preparations exerted antiproliferative effects across ER+, HER2+, and triple-negative breast cancer models, with additive activity alongside tamoxifen and lapatinib [12].

The clinical implication is that cannabinoid roles in breast cancer may be most defensible as chemotherapy potentiators rather than as standalone interventions, particularly given the evidence that THC alone can act against tumor control in this tumor type. No completed clinical trials have evaluated cannabinoid antitumor activity in breast cancer patients to date.

Lung, Prostate, and the Null or Antagonistic Findings

Lung cancer findings produced a clear positive signal at the pooled level. Cannabinoid monotherapy reduced tumor volume by 562.17 mm³ in xenograft models (95% CI: 693.99 to 430.35 mm³) [1]. Disaggregated by compound class, individual subgroup effect estimates for CBD and synthetic cannabinoids both trended favorable, with CBD producing the largest absolute reduction at 659.92 mm³, but neither subgroup reached statistical significance on its own under the wide confidence intervals reflecting limited per-subgroup study counts. Mechanistic work supporting the overall favorable lung cancer signal has implicated CB1 and CB2 receptors as targets for inhibition of non-small cell lung cancer growth and metastasis, with effects observed on epithelial-to-mesenchymal transition, tumor invasion, and angiogenesis [13,14]. Recent work has also demonstrated synergistic activity between CBD and tyrosine kinase inhibitors such as dasatinib through SRC/PI3K/AKT signaling in lung cancer models [15].

The prostate cancer findings are more complicated. Cannabinoid monotherapy produced a pooled tumor volume reduction of 394.32 mm³ in xenograft models, but the confidence interval crossed zero (95% CI: 793.91 to 5.26 mm³), and the overall effect did not reach statistical significance. CBD and synthetic cannabinoids individually produced nonsignificant reductions. THC:CBD combinations, however, produced a substantial and statistically significant tumor volume reduction of 1136.59 mm³ (95% CI: 1320.97 to 952.21 mm³). This result rests on a single contributing study by Motadi and colleagues in PC3 xenografts [16], a limitation worth flagging given that meta-analytic estimates drawn from one study should be interpreted as preliminary observations rather than robust pooled findings.

The combination-with-chemotherapy data in prostate cancer reveal a cautionary signal that deserves clinical attention. THC:CBD added to chemotherapy significantly increased tumor growth in pooled analyses, with a mean difference of +306.34 mm³ (95% CI: 231.29 to 381.39 mm³), an antagonistic interaction that the meta-analysis authors note explicitly. CBD with chemotherapy showed nonsignificant tumor reduction, and CBD monotherapy compared with chemotherapy actually increased tumor burden modestly. The picture from these data is that cannabinoid-chemotherapy interactions in prostate cancer are not uniformly favorable and may, in some compound combinations and contexts, work against tumor control.

Several tumor types showed essentially null or minimal cannabinoid effects. In hepatocellular carcinoma, cannabinoid monotherapy produced a pooled tumor volume reduction of only 2.01 mm³, with a tight confidence interval (95% CI: -4.55 to 0.53 mm³) that crosses zero and indicates a precisely estimated absence of effect rather than insufficient statistical power. Pancreatic cancer findings were modest and nonsignificant. Skin cancer responses were near-neutral, with cannabinoid effects shifting slightly in the protumoral direction when combined with other agents. Colon cancer monotherapy responses were modest and inconsistent, and direct comparison with chemotherapy showed worse tumor control with cannabinoids in this tumor type. Hytiva Research's prior coverage of cannabinoids in skin cancer provides additional context on the underlying biology and the limited clinical translation in this area.

These null results matter as much as the positive ones. Cannabinoids are not universal antitumor agents, and the meta-analysis confirms what mechanistic work has long suggested: receptor distribution, microenvironmental context, and oncogenic signaling differ enough between tumor types to produce divergent therapeutic outcomes.

Compound-Specific Patterns

The 2026 meta-analysis offers a clearer view of how compound class shapes cannabinoid response than any prior synthesis. CBD demonstrated the broadest favorable profile across tumor types, with consistent antitumor activity, the cleanest combination effects (particularly with doxorubicin in breast cancer and with temozolomide in glioblastoma), and the advantage of a well-documented human safety record from epilepsy use [1]. CBD exerts antitumor effects through multiple parallel pathways, including direct effects on mitochondrial function, autophagy induction, modulation of TRPV channels, and inhibition of pro-tumor signaling through pathways such as PI3K-Akt and Wnt/β-catenin [17,18].

THC produced more variable effects across the meta-analysis. In glioblastoma, where the mechanistic case is strongest and the receptor expression profile of tumor cells supports robust CB1 and CB2 engagement, THC produced significant tumor reductions. In breast cancer, THC monotherapy trended protumoral, and in colon cancer it produced nonsignificant effects with directional movement toward tumor growth in some studies. The picture suggests that THC's antitumor utility is genuinely context-dependent, with CB1-mediated immunosuppressive signaling potentially offsetting direct cytotoxicity in certain tumor types [9].

Synthetic cannabinoid agonists, including WIN-55,212-2 and JWH-133, showed promising effects in several tumor types but with substantially less clinical infrastructure behind them. These compounds offer pharmacological advantages, particularly enhanced blood-brain barrier penetration relevant for glioblastoma applications, but their limited human exposure data means any translational pathway will require extensive early-phase safety evaluation.

THC:CBD combinations, particularly in defined ratios, may mitigate some of THC's tumor-type-specific liabilities. The Blasco-Benito 2018 work demonstrated that whole-plant THC:CBD preparations produced antitumor effects across breast cancer subtypes, sometimes outperforming pure cannabinoids on the same models [12]. In glioblastoma, the THC:CBD nabiximols preparation tested in the Twelves 2021 Phase 1b trial achieved survival improvements that pure THC alone would not be expected to deliver based on the preclinical data, though the small sample size limits the inference that can be drawn from a single early-phase trial.

These compound-specific patterns suggest that future cannabinoid oncology research should not treat the cannabinoid class as a unitary therapeutic entity. The pharmacology of CBD, THC, synthetic agonists, and ratio-defined combinations differs in ways that meaningfully shape therapeutic outcome, and trial design will need to reflect that.

From Preclinical Signal to Clinical Translation

The preclinical evidence base now has its first peer-reviewed clinical companions. In 2021, Twelves and colleagues published a Phase 1b trial of nabiximols, an oromucosal THC:CBD spray, combined with dose-intense temozolomide in patients with first recurrence of glioblastoma [19]. The trial enrolled 27 patients across UK and German centers in two parts: an open-label Part 1 (n=6) and a randomized double-blind, placebo-controlled Part 2 (n=21, comprising 12 nabiximols and 9 placebo). The most common treatment-emergent adverse events were vomiting, dizziness, fatigue, nausea, and headache, most graded as grade 2 or 3. In the randomized Part 2, one-year survival was 83 percent in the nabiximols arm compared with 44 percent in the placebo arm (p=0.042), although the authors noted that two patients in the placebo arm died within the first 40 days of enrollment. There were no apparent effects of nabiximols on temozolomide pharmacokinetics. The trial was not powered to demonstrate efficacy definitively, but it established the safety of combining nabiximols with dose-intense temozolomide in a recurrent glioblastoma population, with sufficient signal to motivate larger randomized work.

A Phase 2 trial led by Schloss and colleagues evaluated two different ratios of medicinal cannabis in patients with high-grade gliomas, with tolerability as the primary outcome [20]. The trial confirmed that cannabis preparations were generally well-tolerated in this population, though longer-term efficacy outcomes remain forthcoming. The ARISTOCRAT trial, a Phase 2 multi-center, double-blind, placebo-controlled randomized evaluation of nabiximols plus temozolomide in patients with recurrent MGMT-methylated glioblastoma suitable for temozolomide treatment, is currently underway with 2:1 randomization, and represents the most substantial randomized evaluation of cannabinoid antitumor activity in any cancer indication to date.

These early clinical signals align with the meta-analytic preclinical picture in one important respect: glioblastoma is consistently the most defensible translational target. The receptor biology of glioma cells, the mechanistic depth of the preclinical evidence, the consistent direction of in vivo effects, and the early clinical safety and survival signals converge on glioblastoma as the area where cannabinoid oncology research has the strongest case for advancement.

For other tumor types, the translational pathway is much less developed. No completed clinical trials have evaluated direct antitumor effects of cannabinoids in breast, lung, or prostate cancer patients, despite favorable preclinical data. The clinical evidence that does exist for cannabinoids in cancer care has focused primarily on palliative outcomes: pain management, chemotherapy-induced nausea and vomiting, appetite stimulation, and quality of life. Recent clinical work on oral CBD in patients with several solid tumor types, including prostate, breast, colorectal, and gynecological cancers, showed no significant effect on survival or tumor progression over a short observation window, a result that does not invalidate the preclinical signal but underscores how difficult translation has been [21].

The translational challenge ahead is to design rigorous trials that test the specific cannabinoid-tumor combinations preclinical data identifies as most promising, with attention to compound class, ratio, dosing, and combination with conventional therapy. Generic cannabis-for-cancer trial designs are unlikely to capture the context-dependent effects the preclinical literature consistently demonstrates.

The Limits of Preclinical Evidence

The Creangă-Murariu meta-analysis is methodologically among the strongest preclinical syntheses in cannabinoid oncology to date, and it nevertheless arrives at conclusions that the authors themselves frame as hypothesis-generating rather than definitive. Three structural limitations of the underlying evidence base shape that posture and warrant explicit attention.

First, risk of bias was predominantly high across included studies, with frequent under-reporting of allocation concealment, blinding, randomization sequence generation, and outcome selection. These were largely reporting deficiencies rather than confirmed methodological flaws, but they limit the inferential weight that can be placed on individual effect estimates. The SYRCLE tool used to evaluate animal studies and the adapted framework used for in vitro studies both flagged this pattern across tumor types [1].

Second, heterogeneity was substantial across nearly all analyses, with I² statistics ranging from 34 to 100 percent. Cannabinoid compound, dose, schedule, route, formulation, cell line, animal model, tumor implantation site, and outcome measurement all varied across studies, and subgroup analyses could only partially explain the resulting variability. This level of heterogeneity means pooled effect estimates are rough approximations of central tendency rather than precise quantifications of any specific therapeutic effect.

Third, the underlying experimental designs themselves carry limitations that shape what the data can support. Most in vivo work relied on xenograft models, which restrict evaluation of immune-mediated mechanisms and limit insight into tumor microenvironment interactions. Cell line authentication procedures were inconsistently reported, raising the possibility of model identity issues in some studies. Combination-therapy studies were relatively underrepresented despite being arguably the most clinically relevant configuration for cannabinoid translation, since cannabinoids are most plausibly useful as chemotherapy adjuncts rather than standalone treatments.

These limitations do not invalidate the preclinical signal. Consistent directional trends across diverse models, supported by mechanistic plausibility and early clinical safety data, justify continued investigation. They do, however, place specific constraints on what can be claimed. Cannabinoids are not universal antitumor agents. Their effects are context-dependent, sometimes contradictory between compound classes, and occasionally protumoral or antagonistic depending on tumor type and combination partner. The clinical case for cannabinoid oncology rests most defensibly on glioblastoma at present, with breast cancer as the next-most-supported target, and with substantial additional clinical evidence required before broader indications can be considered.

What rigorous future research requires is now reasonably well-defined: standardized cannabinoid formulations with documented content and purity, dose-response designs that capture the biphasic effects characteristic of cannabinoid pharmacology, biomarker-driven patient stratification incorporating CB1 and CB2 receptor expression profiles, and clinical trials that test specific cannabinoid-tumor-chemotherapy combinations rather than generic cannabis interventions. The 2026 meta-analysis provides a quantitative foundation for designing such studies. The work of conducting them remains ahead.