We’re seeing promising human results from a colorectal cancer study that added mebendazole, discussed in the paper below. However, potential discrepancies have raised questions about how to interpret those findings.
“The appropriate interpretation is a substantial signal requiring clarification and independent confirmation”...wait…what?
It feels like one step forward and two steps back. What do you think?
The paper is Hegazy and colleagues, Mebendazole; from an anti-parasitic drug to a promising candidate for drug repurposing in colorectal cancer, Life Sciences (2022), DOI 10.1016/j.lfs.2022.120536.
The appraisal concludes that the trial provides a promising, statistically noticeable add-on efficacy signal for Mebendazole in combination with FOLFOX4 and bevacizumab.
Summary and contribution
The question is clinically meaningful: does adding mebendazole improve outcomes beyond the same background chemotherapy? The abstract describes a randomized, double-blind, placebo-controlled study of metastatic colorectal cancer, with 20 participants per group. Both received six cycles of FOLFOX4 plus bevacizumab; the experimental group additionally received mebendazole 500 mg twice daily for 12 weeks, while controls received placebo. The relevant treatment effect is therefore the incremental effect of this add-on regimen, not mebendazole monotherapy or an effect across all cancers.
The abstract reports response in 13/20 experimental patients versus 2/20 controls and median progression-free survival of 9.25 versus 3 months. It also reports reduced VEGF, no statistically significant difference in CEA, and describes the treatment as well tolerated. These findings are potentially important, but their credibility depends on trial conduct and analysis details not available in this appraisal.
Strengths
The described add-on design is stronger than a case series because it compares otherwise similar treatment regimens and, if randomization was properly implemented, can estimate incremental benefit. A true matched-placebo design would further reduce performance and assessment bias, but the registry disputes that feature and has not independently verified it. The outcomes include imaging response and progression-free survival rather than relying solely on tumor-marker changes.
Major comments
1. Masking and placebo conflict with the registry
Verified discrepancy; a major concern that requires clarification. The abstract’s Materials and Methods describe the study as double-blind and placebo-controlled. The linked registry, NCT03925662, instead lists no masking and an active comparator of FOLFOX/bevacizumab only. A registry mirror also explicitly describes open-label treatment and names investigators overlapping with the paper.
This matters because unblinded care, decisions to discontinue therapy, and imaging interpretation could influence measured outcomes. It does not establish that the paper’s description is false: an unrecorded protocol amendment or stale registry is a plausible explanation.
Decisive remedy: provide the dated protocol and amendments, ethics approvals, placebo manufacturing/matching details, allocation-concealment procedure, identities of masked personnel, and whether radiologists were independently blinded. Reconcile these with the registry history before assigning low risk of bias.
2. The registered response timing differs from the emphasized result
Verified discrepancy; major concern requiring clarification. The registry specifies tumor response at six months; the abstract emphasizes response at 12 weeks. The paper explicitly labels the registration as retrospective. The registry mirror lists study start as April 1, 2019, and first posting as April 24, 2019; those dates alone do not establish when the first participant was randomized.
This is a risk of outcome/timepoint selection, not proof of selective reporting. The unavailable manuscript may have reported the six-month endpoint elsewhere.
Decisive remedy: identify the prespecified primary endpoint and assessment window, provide dated amendments and the analysis plan, and show response results at both timepoints with complete denominators. Clarify when registration occurred relative to enrollment and access to outcome data.
3. The response counts support a strong nominal association, not an independently verified causal effect.
Verified calculation; major uncertainty about underlying conduct. From 13/20 versus 2/20 responses, the absolute difference is 55 percentage points. A two-sided Fisher exact recalculation gives p ≈ 0.00077, and a Newcombe interval using Wilson score limits gives an approximate 95% confidence interval of 25–73 percentage points for the response-rate difference. These calculations use only the published counts and assume all randomized participants are correctly represented; they do not reproduce the authors’ full analysis.
Thus, “only 40 patients” is not a sufficient statistical rebuttal: the response contrast is too large to dismiss merely as a weak nominal signal. Equally, a small p-value cannot repair flawed randomization, missing outcomes, imbalanced prognosis, or unblinded response assessment.
Decisive remedy: provide sequence generation and concealment, participant flow, intention-to-treat denominators, missing-data handling, response criteria and confirmation, baseline tumor burden, treatment line, performance status, and actual chemotherapy exposure by group. These are not assessable, not established defects.
4. Progression-free survival cannot be adequately audited from the reported medians
Major concern requiring clarification. The abstract reports median progression-free survival of 3 versus 9.25 months and says Kaplan–Meier methods were used, but provides no hazard ratio, confidence intervals, event counts, or numbers at risk. It reports one-year overall survival but does not provide an overall-survival result in the abstract.
The ratio of medians should not be interpreted as a corresponding reduction in progression risk. Scan schedules, censoring rules, deaths, follow-up maturity, subsequent therapies, and informative discontinuation require inspection. Overall survival cannot be inferred from the response and progression-free-survival claims.
Decisive remedy: provide complete survival curves with numbers at risk, events, and censoring; a prespecified between-group test and effect estimate with uncertainty; assessment schedules; and the overall-survival analysis. Explain the control group’s outcomes using actual baseline and treatment-delivery data rather than unsupported historical comparisons.
5. Safety and biological mechanism are overstated if inferred from the abstract alone
Major concern requiring clarification. “Well tolerated” is the only substantive safety conclusion in the abstract; it does not supply graded adverse-event frequencies or treatment discontinuations. VEGF decreased, but CEA did not differ significantly. Neither biomarker establishes that a particular mechanism caused clinical benefit.
Decisive remedy: report adverse events by grade and attribution; laboratory changes; dose interruptions and reductions; chemotherapy dose intensity; adherence; and concomitant drugs. Separate biomarker hypotheses from demonstrated patient benefit. A small add-on trial cannot establish absence of uncommon or delayed toxicity.
Minor comments
Key Findings: report a small positive p-value rather than “p = 0.000.” The recalculation above supports a low nominal p-value, but cannot establish the authors’ exact test.
Statistical Methods: identify which endpoint used each listed test and how primary versus exploratory analyses and multiple comparisons were handled. The abstract lists several tests without this mapping; adequacy is not assessable without the full analysis.
Terminology: the registry’s “adjuvant” label should not be read as evidence in postoperative, potentially cured colorectal cancer; its inclusion criterion is stage IV disease and the abstract concerns metastatic cancer.
Internal-consistency findings
Verified: 13/20 and 2/20 correspond to the reported 65% and 10%; the response comparison is nominally statistically significant under an exact test. Verified discrepancies concern masking/placebo and response timing. The registry still shows estimated enrollment, recruiting status, and a future completion date, whereas the paper reports results; therefore, registry status should not be used as current operational information without confirmation. It does not independently invalidate the published data.
Not checked: baseline table balance, CONSORT flow, imaging examples, response adjudication, survival curves, safety tables, supplements, protocol amendments, or individual participant data. No claim is made that absent abstract details are absent from the full paper.
Overall assessment and recommendation
Evidence-use recommendation: treat the study as a promising but insufficiently verified add-on efficacy signal. It warrants a protocol-level audit and an adequately powered, prospectively registered replication with independently blinded imaging review before influencing routine treatment. No fatal flaw or misconduct is established from the accessible record.
Publication recommendation: withheld. An accept/revise/reject judgment would require the full manuscript and supporting materials. The most consequential unresolved issue is reconciliation of the actual trial conduct with the registered masking, comparator, and endpoint timing.
Assessment: The reported response difference is large and statistically persuasive if the published counts reflect a valid randomized comparison. However, verified registry–publication discrepancies in masking and response timing, along with unavailable trial-level methods, prevent us from treating this as practice-changing evidence. The appropriate interpretation is a substantial signal requiring clarification and independent confirmation—not proof that mebendazole improves colorectal cancer outcomes.
Partial appraisal—abstract and registry assessed; full manuscript, tables, figures, supplements, protocol, and statistical analysis plan unavailable.
Source
Mebendazole; from an anti-parasitic drug to a promising candidate for drug repurposing in colorectal cancer, Life Sciences (2022), DOI 10.1016/j.lfs.2022.120536.