Long game
What is the cancer signal, good or bad?
Kept deliberately as one outcome covering both directions, because a compound being trialled as a cancer treatment and a compound carrying a cancer warning both belong in front of you.
251,345
people in trials
20
human studies
5
compounds, animal or cell only
2
trials found nothing
Measured in people
Ranked by how many people, which is a fact about the evidence and not a judgement about the compound. The one at the top is the most studied, which is not the same as the best. Counts are never added across compounds, because the same person can appear in more than one trial.
- 199,698 people1 human study
The epidemiology that any page about giving people more IGF-1 has to state. 199,698 men in UK Biobank followed a mean of 6.9 years, 5,402 diagnosed with and 295 dying from prostate cancer. Higher circulating IGF-1 was associated with prostate cancer diagnosis (hazard ratio 1.09 per 5 nmol/L increment, 95 percent CI 1.05 to 1.12) and with prostate cancer mortality (1.15, 1.02 to 1.29), with hazard ratios corrected for regression dilution using repeat measurements. A two-sample Mendelian randomisation analysis using genetic instruments and outcome data from 79,148 cases and 61,106 controls supported a causal role (cis-MR odds ratio 1.34 per 5 nmol/L, 1.07 to 1.68). This is an association study in men with naturally varying IGF-1, not a study of mecasermin, and it does not establish that treating a deficient child raises cancer risk.
International Journal of Cancer, 2021 · 199,698 people
- 25,871 people1 human study
VITAL. 25,871 participants randomised to 2,000 IU daily or placebo, median 5.3 years. Neither primary endpoint was met: no reduction in invasive cancer, no reduction in major cardiovascular events. Mean baseline 25-hydroxyvitamin D was 30.8 ng/mL and only 12.7% of participants were below 20.
New England Journal of Medicine, 2019 · 25,871 people
- 12,537 people1 human study
ORIGIN: 12,537 people with cardiovascular risk factors plus impaired fasting glucose, impaired glucose tolerance or type 2 diabetes randomised to insulin glargine or standard care, median 6.2 years. Cardiovascular outcomes were neutral (hazard ratio 1.02, 95 percent confidence interval 0.94 to 1.11). Severe hypoglycaemia was 1.00 versus 0.31 per 100 person-years, and median weight rose 1.6 kg on insulin while falling 0.5 kg on standard care. No excess cancer (hazard ratio 1.00). Funded by Sanofi.
New England Journal of Medicine, 2012 · 12,537 people
- 9,463 people1 human study
9,463 adults aged 40 and over with type 2 diabetes and cardiovascular disease, 610 sites, 28 countries, randomised 1:1 to albiglutide 30 to 50 mg weekly or placebo, median follow-up 1.6 years. Primary composite (cardiovascular death, myocardial infarction, stroke): 338 of 4,731 (7 percent) against 428 of 4,732 (9 percent), hazard ratio 0.78 (95 percent CI 0.68 to 0.90), superiority P equals 0.0006. Pancreatitis 10 against 7, pancreatic cancer 6 against 5, medullary thyroid carcinoma 0 against 0. Funded by GlaxoSmithKline.
Lancet, 2018 · 9,463 people
- 1,075 people1 human study
1,075 postmenopausal women randomised to extract with 843 mg EGCG a day or placebo for one year. Adverse events 75.6 versus 72.8 percent; serious 2.2 versus 1.5 percent; more nausea and skin events on extract. ALT elevation in 36 (6.7 percent) versus 4 (0.7 percent), p < 0.001; 1.3 percent had ALT-related serious adverse events. US National Cancer Institute funded.
Food and Chemical Toxicology, 2015 · 1,075 people
- 1,075 people1 human study
1,075 postmenopausal women randomised to an extract delivering 843 mg EGCG a day or placebo for one year. Alanine aminotransferase elevation in 36 women on extract (6.7 percent) against 4 on placebo (0.7 percent), p < 0.001, with 1.3 percent having an ALT-related serious adverse event. Overall adverse event rates were similar, 75.6 against 72.8 percent, with more nausea and skin events on extract. US National Cancer Institute funded.
Food and Chemical Toxicology, 2015 · 1,075 people
- 513 people2 human studies
The only randomised human data on glutaminase inhibition. 444 patients with metastatic renal cell carcinoma. Median progression-free survival 9.2 months with telaglenastat plus cabozantinib versus 9.3 months with placebo plus cabozantinib, hazard ratio 0.94, P = 0.65. Manufacturer funded. Oncology, not senescence.
JAMA Oncology, 2022 · 444 people
69 patients randomised 2:1. Median progression-free survival 3.8 months versus 1.9 months, hazard ratio 0.64, one-sided P = 0.079 against a pre-specified one-sided alpha below 0.2. Grade 3 to 4 events in 74 percent versus 61 percent. Manufacturer funded.
Clinical Cancer Research, 2022 · 69 people
- 513 people2 human studies1 found nothing
found nothing444 patients randomised, double blind, placebo controlled. Median progression-free survival 9.2 versus 9.3 months, hazard ratio 0.94, 95 percent confidence interval 0.74 to 1.21, P = 0.65. Overall response 31 versus 28 percent. Grade 3 to 4 adverse events in 71 versus 79 percent. Manufacturer funded. The primary endpoint was missed.
JAMA Oncology, 2022 · 444 people
69 patients randomised 2:1 after a median of three prior lines. Median progression-free survival 3.8 versus 1.9 months, hazard ratio 0.64, 95 percent confidence interval 0.34 to 1.20, one-sided P = 0.079 against a pre-specified one-sided alpha below 0.2. One partial response on telaglenastat. Grade 3 to 4 events 74 versus 61 percent. Manufacturer funded.
Clinical Cancer Research, 2022 · 69 people
- 187 people1 human study
187 patients with cancer cachexia and GDF15 of at least 1,500 pg/mL (40 percent non-small-cell lung, 32 percent pancreatic, 29 percent colorectal) randomised 1:1:1:1 to ponsegromab 100, 200 or 400 mg or placebo every 4 weeks for three doses. Median weight gain versus placebo at 12 weeks: 1.22 kg (95 percent credible interval 0.37 to 2.25), 1.92 kg (0.92 to 2.97) and 2.81 kg (1.55 to 4.08). Appetite, cachexia symptoms and physical activity improved at 400 mg. Adverse events in 70 percent on ponsegromab and 80 percent on placebo. Funded by Pfizer.
New England Journal of Medicine, 2024 · 187 people
Also measured, in animals or cells
An antibody against GFRAL prevented cachexia in tumour-bearing mice, reversing excess lipid oxidation, and identified a sympathetic lipolytic axis by which GDF15 wastes fat and muscle independently of anorexia. The preclinical basis for blocking the pathway in people.
Nature Medicine, 2020 · animal
- 170 people1 human study
170 adults randomised to placebo (n = 55) or one of three concentrations of broccoli sprout beverage for 10 consecutive days. Urinary benzene mercapturic acids rose significantly only in the high dose group, by 63.2 percent. The one half dose (+11.3 percent) and one fifth dose (-6.4 percent) were not significantly different from placebo. Median 24 hour urinary sulforaphane metabolite output was 24.6, 10.3 and 4.3 micromol across the three doses. The endpoint is a urinary detoxification biomarker, not a cancer rate. NIH funded, registered as NCT02656420.
American Journal of Clinical Nutrition, 2019 · 170 people
- 159 people1 human study1 found nothing
found nothing159 patients with cancer and at least 2% weight loss, randomised to enobosarm 1 mg, 3 mg or placebo for up to 113 days. Total lean body mass rose significantly on both doses (median 1.5 kg on 1 mg, 1.0 kg on 3 mg) while placebo did not change. This is the Phase 2 trial that preceded the POWER Phase 3 program.
The Lancet Oncology, 2013 · 159 people
- 46 people1 human study
46 women with platinum-resistant high-grade serous ovarian cancer, median four prior lines, single-arm. Three-month progression-free survival 22.7 percent, median progression-free survival 1.64 months, one partial response and 15 stable diseases. Grade 3 or 4 thrombocytopenia in 12 of 46, causing dose reduction in 8 and discontinuation in 3. The authors concluded activity was poor.
Gynecologic Oncology, 2022 · 46 people
- 38 people2 human studies
20 people with metastatic sarcoma received 20 mg subcutaneously daily. Median progression-free survival 2.7 months (95 percent confidence interval 1.4 to 4.1) and median overall survival 10.2 months (5.3 to 18.3). The placental growth factor biomarker effect seen in phase 1 was not confirmed. Two people with vascular sarcomas had prolonged disease stabilisation at 10 and 19 months.
Sarcoma, 2016 · 20 people
18 adults with advanced solid tumours received escalating subcutaneous doses daily for 5 days on a 3 week cycle. Dose-limiting toxicities at 700 micrograms per kilogram were a grade 4 stroke and a grade 3 reversible cranial neuropathy. One patient had a 19 percent tumour reduction and 3 more had disease stabilisation beyond 3 months. Recommended phase 2 dose 400 micrograms per kilogram.
Clinical Cancer Research, 2009 · 18 people
- no headcount stated1 human study
Academic phase 2 in head and neck cancer, 17 enrolled, terminated. Reason recorded by the sponsor: unmet primary endpoint.
ClinicalTrials.gov, Cliniques universitaires Saint-Luc, 2020 · no headcount in the line
- no headcount stated1 human study
The CLARINET randomised phase 3 in metastatic enteropancreatic neuroendocrine tumours, showing prolonged progression-free survival on lanreotide. A hard clinical endpoint in a randomised trial, which very few peptides documented on this site can claim.
New England Journal of Medicine, 2014 · no headcount in the line
- no headcount stated1 human study
A randomised phase 3 trial in breast cancer patients receiving adjuvant radiotherapy, reporting prevention of dysphagia. A real randomised result with a symptom endpoint that matters to patients, in a specific population receiving a specific mucosal insult.
The Breast Journal, 2020 · no headcount in the line
- no headcount stated1 human study
The five-year follow-up of the same myelofibrosis trial, reporting durability of spleen response and long-term safety in that population. Still a cancer population, still no ageing endpoint.
Journal of Hematology and Oncology, 2017 · no headcount in the line
Measured in animals or cells, not yet in people
These are separated from the trials above rather than ranked against them, because they answer a different question. A result in a rat is a real result and a reason somebody ran the study; it is not a measurement of what happens in you. Where the human work is missing, the reason is usually money rather than failure, and that is set out at the bottom of this page.
- 2 preclinical studies
Measured in animals or cells
Human cell lines and mouse xenografts. DT2216 was less toxic to platelets than ABT-263 in vitro, more potent against BCL-xL-dependent cancer cells, and inhibited xenograft tumour growth in mice without appreciable thrombocytopenia. The design rationale is that VHL is poorly expressed in platelets.
Nature Medicine, 2019 · animal
Cells and mice. Activity against BCL-xL-dependent T cell lymphoma models, extending the preclinical case for selective degradation over inhibition.
- 2 preclinical studies
Measured in animals or cells
Azoxymethane-induced colon cancer in wild-type and PPAR-beta/delta-null mice. GW0742 had inhibitory effects on colon carcinogenesis, acting through mechanisms independent of COX2 inhibition. Mouse work, and it points opposite to the class cancer fear.
Carcinogenesis, 2008 · animal
found nothingGW0742 and GW501516 tested in five human cancer cell lines with and without serum. Neither increased cell growth, Akt phosphorylation, VEGF or COX2, and liver, colon and colon polyps from treated mice showed no change in these markers either. A published argument against the tumour-promotion hypothesis for this class.
Carcinogenesis, 2007 · in vitro
- 1 preclinical study
Measured in animals or cells
Mouse, from the Yamada and Yamanaka laboratories. Transient expression of reprogramming factors followed by doxycycline withdrawal produced tumours in various tissues, consisting of undifferentiated dysplastic cells with global DNA methylation changes. Kidney tumours shared characteristics with Wilms tumour. Stem cells derived from those tumour cells gave rise to non-neoplastic kidney cells, proving the tumours were epigenetically rather than genetically driven.
Cell, 2014 · animal
- 1 preclinical study
Measured in animals or cells
Comparative analysis of naked mole rat and human plasma. Naked mole rat plasma carried A2M at 8.3 plus or minus 0.44 mg/mL versus 4.4 plus or minus 0.20 mg/mL in human plasma, with lower total plasma protein (38.7 versus 61.7 mg/mL), higher antitryptic activity, lower proteolytic activity, and a different predicted glycosylation giving a higher molecular weight. The authors suggest a possible contribution to that species' cancer resistance and longevity. This is comparative biochemistry, not an intervention.
PLoS One, 2015 · in vitro
- 1 preclinical study
Measured in animals or cells
Mouse. Genetic or senolytic ablation of senescent cells, or macrophage depletion, reduced tumour burden and increased survival in KRAS-driven lung cancer models. Macrophages with senescent features were also found in human pre-malignant lung lesions, an observation rather than an intervention.
Cancer Cell, 2023 · animal
Why the trial record looks like this
A missing trial is usually a missing sponsor. Nobody runs a nine-figure programme on a molecule they cannot own, so the compounds with the thinnest human records are often the oldest and most freely available ones rather than the weakest. Taking DT2216 as the example, because it states the problem most clearly:
DT2216 is the clearest case in this field of a rational fix to a mechanism problem being carried into people. The problem, stated in 2010, was that BCL-xL inhibition kills platelets. The fix, published in 2019, was to degrade BCL-xL through a ligase platelets barely express. The first human data, published in 2025, are consistent with the fix: a single dose-limiting thrombocytopenia that resolved in 48 hours, against 29 of 55 patients with grade 3 or 4 thrombocytopenia on navitoclax.
That is a pharmacology success, not a geroscience result. Every DT2216 trial is in cancer, no ageing indication is registered, and the senolytic case for it remains entirely in mice. A Children's Oncology Group trial is suspended for unacceptable toxicity, which belongs on the same page as the phase 1 tolerability finding.
A compound appears here only when a citation on its own page states this outcome and reports a result for it; naming it is not enough. Human, animal and cell work are counted separately and never merged. Participant counts are read from the study lines and deliberately undercount. Nothing here is advice, and nothing here is a recommendation to take or to avoid anything.