Can blood chemistry years earlier predict prostate cancer death?
Among Finnish men who later got prostate cancer, dozens of blood metabolites measured years before diagnosis - led by choline and glutamate - were linked to a higher chance of dying from the cancer.
Source
Metabolomic profile of prostate cancer-specific survival among 1812 Finnish men
Study at a glance
- Design
- Cohort — Case-only prospective survival analysis: 961 serum metabolites measured in fasting blood drawn at ATBC trial baseline (1985-1988) were related by Cox regression to prostate cancer death among men later diagnosed with prostate cancer, with Bonferroni correction and a 70/30 discovery-replication split.
- N
- N=1812 · 1812 incident prostate cancer cases from the 29,133-man ATBC cohort, of whom 472 died of prostate cancer during follow-up to the end of 2016.
- Population
- Finnish male smokers aged 50-69 at enrolment in the Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study who later developed prostate cancer.
- Outcome
- Prostate cancer-specific mortality after diagnosis.
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
The researchers used frozen fasting blood samples collected in 1985-1988 from Finnish male smokers in a large vitamin-supplement trial. For the 1812 men later diagnosed with prostate cancer, they measured 961 small molecules (metabolites) and used Cox survival models, adjusted for age, stage and Gleason score, to see which predicted death from prostate cancer. They applied a strict Bonferroni threshold for multiple testing, analysed metabolic pathways, and built a metabolite risk score in a 70% discovery set that they tested in the remaining 30%.
What they found
Of the 961 metabolites, 49 were linked to prostate cancer death after Bonferroni correction. The strongest were choline (hazard ratio 2.07, top versus bottom quarter) and glutamate (hazard ratio 2.31), plus gamma-glutamyl amino acids, dipeptides and endocannabinoids; some glutathione-related metabolites were protective. Adding five selected metabolites to clinical factors improved discrimination from an AUC of 0.82 to 0.86. In the replication set, each standard-deviation rise in the risk score raised the hazard of prostate cancer death by 48%.
The limits
What it doesn't show
The men were all Finnish smokers aged 50-69, so results may not apply to non-smokers, younger men or other ethnic groups. Replication was only internal (a random split of the same cohort), with no external validation, and tumour molecular subtypes were unavailable. As an observational study, residual confounding cannot be excluded, and associations do not show that changing these metabolites would alter survival. Metabolites were measured once, often more than a decade before diagnosis.
Key terms
- Metabolomics
- The large-scale measurement of small molecules (metabolites) in blood or tissue, giving a snapshot of body chemistry shaped by genes, diet and environment.
- Bonferroni correction
- A way to limit false positives when testing many hypotheses by dividing the significance threshold by the number of tests (here 0.05/961).
- Hazard ratio
- From a survival (Cox) model, how much faster an event such as death occurs in one group than another; 2 means roughly double the rate.
- Discovery and replication sets
- Splitting data so that findings chosen in one part are tested in another, reducing the chance that results reflect random quirks of one sample.
- Area under the curve (AUC)
- A measure of how well a model separates people who have an outcome from those who do not, from 0.5 (chance) to 1.0 (perfect).
- Reverse causality
- When the outcome, or an early form of it, causes the apparent exposure; measuring blood long before diagnosis reduces this risk.
Flashcards
0 of 11 answers reviewed
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
What design best describes this study?
Common questions
Why does it matter that blood was taken so long before diagnosis?
Samples were drawn a median of about 14 years before diagnosis, so the metabolite levels are unlikely to be a consequence of an already-growing tumour, reducing reverse causality.
With 961 metabolites tested, couldn't some findings be flukes?
Yes, which is why the authors required a Bonferroni-corrected threshold of 0.05 divided by 961 and checked a risk score in a separate replication subset; external validation in another cohort is still needed.
Does this mean men should cut choline from their diet?
No. The study shows an association in blood levels, not that changing diet or metabolite levels would change survival; that would need intervention studies.
More on Oncology outcomes