The debate over whether every woman should get an annual mammogram, versus a screening schedule personalized to individual risk, has run for years largely on modeling and opinion. The WISDOM study (Women Informed to Screen Depending On Measures of risk) is the national randomized trial actually built to answer it, and a 2025 paper describes how it integrated polygenic risk scores at scale to make personalized screening operational rather than theoretical.
What is WISDOM actually comparing?
WISDOM randomizes participants to one of two screening strategies: annual mammography starting at age 40, the conventional approach in much of US practice, or a personalized screening strategy that adjusts starting age, frequency, and imaging modality based on an individual risk assessment combining traditional risk factors, breast density, and a polygenic risk score. Source: Integrating breast cancer polygenic risk scores at scale in the WISDOM Study, Genome Medicine 2025.
This is a pragmatic, large-scale randomized trial, not a retrospective modeling exercise, which is what separates it from most of the prior literature arguing for or against risk-stratified screening. The trial's primary hypothesis is that a personalized approach can achieve non-inferior or better cancer detection, with fewer total screening exams and less overdiagnosis of low-risk, indolent disease, compared to a uniform annual schedule applied to everyone regardless of individual risk.
What did the 2025 paper specifically report?
The 2025 Genome Medicine paper is primarily a methods and implementation paper, describing how the trial integrated polygenic risk scoring into its risk-assessment pipeline at the scale required for a national study enrolling tens of thousands of participants. This included validating the polygenic score's performance across the diverse population WISDOM enrolls, addressing the ancestry-portability problem that affects most polygenic scores derived primarily from European-ancestry reference populations, and operationalizing the score into a composite risk calculation clinicians and participants could act on in real time.
This kind of implementation science, how do you actually run a validated genetic risk score inside a live clinical trial pipeline at scale, is a necessary and underappreciated precondition for personalized screening to ever become standard practice. A risk model that performs well in a research paper but cannot be deployed reliably and equitably inside a live screening workflow is not yet a clinical tool.
Why has this question been so hard to settle without a trial like this?
Retrospective and modeling studies on risk-stratified screening intervals have produced conflicting recommendations for over a decade, because the answer depends heavily on assumptions that differ across models: the assumed sensitivity of mammography at different densities, the assumed rate of interval cancers under extended screening intervals, and the assumed harm from overdiagnosis and false positives under annual screening. Different modeling groups have made different reasonable assumptions and arrived at different guideline recommendations, which is part of why US and European screening guidelines have historically diverged on starting age and interval.
A pragmatic randomized trial removes the dependency on modeling assumptions by measuring actual outcomes under each strategy in the same population, concurrently. That is a slower and more expensive way to get an answer, but it is the only way to get one that does not depend on which model's assumptions you trust.
What are the plausible outcomes, and what would each mean?
If personalized screening achieves non-inferior detection with meaningfully fewer total screens and lower false-positive recall rates, it would support a shift away from uniform annual screening toward risk-based intervals, a change several other countries' screening programs have already begun making administratively, ahead of definitive US trial data. If personalized screening underperforms on interval cancer rate for a specific risk subgroup, most plausibly women classified as lower risk who are then screened less frequently, it would argue for caution in extending longer intervals to that group specifically, rather than abandoning risk stratification altogether.
Either outcome will be more informative than the current state, where screening intervals differ across guideline bodies without a shared trial base to arbitrate the disagreement.
What this does not prove, yet
The 2025 publication describes trial implementation and score integration, it is not yet the outcome readout. WISDOM has not yet published its primary comparative endpoint data on cancer detection, interval cancer rate, or overdiagnosis between the two screening strategies at the scale needed to change guidelines. The polygenic score validation work addresses technical implementation and some ancestry-related performance, but full outcome equivalence across all enrolled subgroups, including historically underrepresented populations in genomic research, remains to be reported. And like MASAI, any mortality benefit or harm signal will require years of additional follow-up beyond what has been published so far.
The takeaway
WISDOM is the trial actually designed to resolve the personalized-versus-uniform breast screening debate with outcome data rather than competing models, and the 2025 publication demonstrates the polygenic risk integration required to make that comparison operational at national scale. The comparative outcome data is still to come. Health systems and guideline committees should treat this as one of the most consequential ongoing trials in cancer screening, and should wait for its primary endpoint results before treating either annual or personalized screening as the settled answer.
