The advanced-cancer drop that got smaller under a magnifying glass
Adjunctive ultrasound cut advanced breast cancer by 1.08% vs 1.14% at 180 months in Japan’s J-START trial.
By The Weekend · · 5 min read
Add ultrasound to a mammogram and, over 15 years, 1.08% of women developed advanced breast cancer instead of 1.14%. That is the entire headline finding of Japan's J-START trial, and three separate letters to The Lancet spent the past month arguing about what a 0.06 percentage-point gap is actually worth.
The number the whole argument turns on: 0.83
The trial's authors, led by Narumi Harada-Shoji, reported a hazard ratio of 0.83 for advanced breast cancer (stage 2 or higher) in women aged 40 to 49 who received adjunctive ultrasonography alongside mammography, compared with mammography alone. The 95.6% confidence interval ran from 0.70 to 0.98 — a result that clears statistical significance, but only just. A hazard ratio of 0.83 means the ultrasound group had roughly 17% lower relative risk of advanced cancer over the follow-up period. That is the number that made headlines when the extended follow-up data published.
J-START began enrolling in 2007 and randomized roughly 72,000 women in that age band, one of the few trials large enough to test adjunctive screening with any statistical power. Extending follow-up to 180 months — 15 years — let the researchers accumulate enough advanced-cancer cases to report a hazard ratio at all. Short screening trials rarely can; advanced breast cancer in women under 50 is uncommon enough that early cutoffs produce noise, not signal.
A relative risk reduction and an absolute one tell different stories
Here is where the correspondence gets interesting, and where three separate groups of clinicians — Anna Macios and colleagues, Saisai Jing and Jiazhao Song, and breast surgeon Atsushi Fushimi — converged on the same objection from different angles. A 17% relative reduction sounds like a meaningful clinical win. The absolute reduction was 1.14% down to 1.08%, a difference of 0.06 percentage points across the full trial population.
Translate that into what it means for a person: out of roughly 1,700 women screened with adjunctive ultrasound for 15 years, about one additional case of advanced breast cancer was prevented compared with mammography alone. That is the everyday-life outcome hiding inside the hazard ratio — not a doubling of protection, not a dramatic shift in outcomes, but a small number of cases avoided across a very large group of women, each of whom also underwent more scans, more callbacks, and more biopsies to get there.
One of the letters goes further, disputing a specific claim in the original paper: that the proportion of advanced cancers out of all diagnosed cancers was lower in the ultrasound group, and that this proportion itself demonstrates a reduction in advanced disease. The correspondents call this framing incorrect, arguing that a shift in the mix of cancer stages diagnosed can reflect earlier detection of cancers that would have surfaced anyway — not necessarily fewer advanced cancers overall. It is the classic screening-trial trap: finding more cancer, and finding it earlier, are not automatically the same as finding it in time to change the outcome.
The strongest case for the ultrasound program, taken seriously
The rebuttal deserves its full weight before the caveats pile up. Breast tissue in women under 50 is denser, and mammography's sensitivity drops in dense tissue — a mechanical limitation, not a screening-philosophy dispute. Ultrasound uses sound waves rather than X-rays and can pick up masses that dense tissue hides from a mammogram. Japan's population skews toward denser breast tissue at younger ages than some Western cohorts, which is precisely why J-START enrolled the 40-to-49 bracket rather than an older group.
Fushimi, writing as a breast surgeon inside Japan's population-based screening system, frames the hazard ratio as clinically real even if the absolute numbers are modest — advanced-stage diagnosis carries worse five-year survival and harsher treatment regardless of how rare it is at the population level. Preventing even a small number of stage 2-or-later diagnoses matters enormously to the specific women who avoid that diagnosis, even if the population-level percentage barely moves. A trial doesn't have to produce a large absolute number to justify a policy, if the harm avoided per case is severe enough and the added cost per screen is tolerable.
The authors' reply, published after the three critical letters, takes this position directly: they clarify that the reduction in advanced breast cancer should inform screening policy at the population level, not be read as a guarantee for any individual woman, and that hazard ratios and absolute risk differences are answering different questions — one about relative protection, one about the scale of benefit across a screened population. Both figures are correct simultaneously; the disagreement is about which one should drive a national screening recommendation.
What a 0.06 percentage-point gap does to a screening budget
This is where the arithmetic stops being academic. Adjunctive ultrasound means more machines, more trained sonographers, more follow-up imaging for false positives, and more biopsies for masses that turn out to be benign — a known cost of any screening method with imperfect specificity. Japan's health system would need to weigh that infrastructure and false-positive burden against preventing roughly one additional advanced-stage case per 1,700 women screened over 15 years.
No funding disclosure in the correspondence suggests industry money shaped the trial's design or interpretation — J-START is a publicly registered, government-supported Japanese trial, which is part of why the debate is being conducted in a major journal's letters pages rather than dismissed as marketing. That is not nothing: a screening trial with no manufacturer stake behind it is exactly the kind of evidence that should carry weight in a policy decision, even when the effect size it produces is this narrow.
What happens next is a policy question, not a scientific one. Japan's screening guidelines committees will have to decide whether a hazard ratio of 0.83, sitting on an absolute difference of 0.06 percentage points, clears the bar for a national recommendation extending ultrasound to all women aged 40 to 49 — knowing that every screening program funded is a screening program not funded elsewhere, and that the number everyone quotes in headlines is rarely the number that should decide the budget.