August 21, 2026 · Science · Screening
What the study found
Researchers at Karolinska Institutet and Umeå University analyzed the Stockholm-Gotland colorectal cancer screening program and report that participating in it was associated with a 43% lower rate of death from colorectal cancer. The paper, "Fecal Occult Blood Screening Outcomes Adjusted for Contamination Bias and Nonadherence" by Johannes Blom, Lennarth Nyström and Håkan Jonsson, was published in JAMA Network Open on August 20, 2026 (DOI 10.1001/jamanetworkopen.2026.29856). It is a prospective cohort study covering the Stockholm-Gotland region of Sweden from 2008 to 2021, including 376,511 people — 203,692 in the study group and 172,819 in the control group — among whom 1,668 colorectal cancer deaths were recorded, with follow-up of up to 14 years.
The program offers people aged 60 to 74 a stool sample kit to complete at home every two years, with a positive result triggering a colonoscopy. Two estimates come out of the analysis. Being invited to screen was associated with a rate ratio of 0.74 (95% CI, 0.58–0.94), a 26% reduction. Actually participating was associated with a rate ratio of 0.57 (95% CI, 0.40–0.80), the 43% figure. The gap between the two is largely the roughly one third of people who, per the researchers, do not send a sample back. An earlier evaluation of the same program had put the invitation figure at 14%; the increase reflects the authors' statistical adjustment for contamination and nonadherence rather than any change in the program itself.
The short version
What it means for you
If you are weighing a multi-cancer blood test or a whole-body scan, the honest read of this paper is not "the cheap test wins." The two categories are not answering the same question, and a study of one cancer in one age band cannot adjudicate a product designed to look across many. The useful read is about sequencing. A screening pathway with a mortality endpoint behind it — an endpoint measured in deaths, not in cancers detected or signals flagged — is a different class of evidence from a detection statistic, and the paper is a clean illustration of how long and how large a study has to be to produce one. Thirteen years of program data and 1,668 deaths yielded a confidence interval that still runs from 0.40 to 0.80.
That framing is already how our Galleri review treats the question, and this study is the evidence behind the position rather than a change to it: established single-cancer screening carries mortality-benefit evidence that multi-cancer early detection has not yet established for itself, which is why the review treats broader testing as something layered on top of established screening rather than a substitute for it. The same logic applies to whole-body MRI, where the open question has always been whether finding more things earlier translates into fewer deaths. Nothing here answers that question for imaging. It does show what answering it looks like.
There is a second, less comfortable point in the numbers. The single largest determinant of benefit in this program was not the technology — it was whether people returned the kit. Roughly a third did not, and the difference between the 26% invitation estimate and the 43% participation estimate is mostly that gap. For a reader who is drawn to longevity testing because it feels thorough, that is worth sitting with: the unglamorous test that is easy to defer has a larger documented effect on this particular cause of death than anything currently available to add on top of it. Whether and when you should screen, by which method, is set by national guidelines that differ between countries and is a question for your clinician or your national program, not one to settle from a Swedish cohort.
How this fits what we already know
Our platform reviews have consistently separated two claims that screening products tend to present together: that a test can detect something, and that using it makes people live longer. The first is a performance statistic and is often well documented. The second requires a study shaped like this one. This paper does not change what Prenuvo, Ezra or Galleri measure, what they cost, or how to read their results, and we are not revising those reviews on the back of a single cohort in a different modality. What it does is give a concrete, citable shape to the evidentiary gap those reviews already describe. We will fold the comparison into the screening guides on their next review cycle.
Sources
- Blom J., Nyström L., Jonsson H., "Fecal Occult Blood Screening Outcomes Adjusted for Contamination Bias and Nonadherence," JAMA Network Open, August 20, 2026, DOI 10.1001/jamanetworkopen.2026.29856 (abstract accessed August 21, 2026).
- MedicalXpress, "Colorectal cancer screening could reduce mortality by over 40%," medicalxpress.com, August 20, 2026.
Frequently Asked Questions
What exactly did the 43% figure measure?
It is the reduction in death from colorectal cancer associated with actually participating in the screening program, expressed as a rate ratio of 0.57 with a 95% confidence interval of 0.40 to 0.80. The separate figure for simply being invited — which includes everyone who was sent a kit and never returned it — was a rate ratio of 0.74, or a 26% reduction, with a confidence interval of 0.58 to 0.94. Both come from the same analysis. Which number is the relevant one depends on whether you are asking about a public health program or about an individual who takes the test.
Does this mean a stool test is better than a multi-cancer blood test or a full-body MRI?
It does not, because the study does not compare them and they are not aimed at the same thing. A fecal occult blood program targets one cancer in one age band with a defined follow-up pathway to colonoscopy. Multi-cancer early detection tests and whole-body imaging cast a much wider net. What the study does establish is the standard of evidence a single-cancer screening pathway has now reached: a mortality endpoint, hundreds of thousands of people, confidence intervals. That is the bar against which broader tests are still working.
Is this a randomized trial?
No. It is a prospective cohort study of the Stockholm-Gotland program between 2008 and 2021, comparing 203,692 people in the study group against 172,819 in the control group. The authors used statistical adjustment for contamination bias — people in the comparison group who got screened anyway — and for nonadherence. They state plainly that because the results rest on those adjustments, some uncertainty remains. A cohort study shows association, not cause.
Why did the estimate go up from earlier evaluations?
An earlier evaluation of the same program found a 14% lower risk of colorectal cancer death among people invited to screen. The new analysis reports 26% for invitation after adjusting for contamination and nonadherence. The change comes from the statistical correction, not from new participants: when people in the unscreened comparison group get screened through other routes, an unadjusted comparison understates the difference. This is a methodological argument, and it is the paper's actual contribution.
What should I take from this before paying for a screening product?
A useful question rather than an answer: for the specific cancer you are worried about, is there an established screening pathway with a mortality endpoint behind it, and are you current on it? Screening intervals, starting ages and preferred methods are set by national guidelines and differ between countries, so that is a conversation with your clinician or your national program rather than something to settle from a study of one Swedish region. Our reviews of the self-pay screening products cover what each one does and does not claim to have shown.