Eighteen days separates the two survival figures, a ratio of about 1.43, and both ends of that comparison need careful reading [1][2]. Every treated mouse was alive when observation stopped at 60 days [8], so the treated arm has no failure time in it at all: 60 is the length of the study, not the length of the lives. The other side is reported as a survival of 42 days without saying whether that is a median or the last death [9]. The release also gives no group sizes [11], which matters more than the percentage does. One hundred percent of six animals and 100% of forty animals are the same number and very different evidence.
Then there is the apportionment problem, which is the interesting part. Shi describes a single material doing two jobs in sequence, a guide during the operation and a clean-up afterwards [3]. What was compared is that combined package against surgery alone [10]. If glioblastoma recurrence were mostly a seeing problem, imaging-guided resection on its own should close most of the gap; if it is mostly a residual-biology problem, the phototherapy is doing the work and the 44-micrometer figure is a convenience [5]. The arms that would separate those two readings are not described [10].
The mechanism hints at which way to lean. The postoperative step uses platinum atoms to convert the tumor's own hydrogen peroxide into oxygen, pushing back on the low-oxygen conditions that shield cells from treatment, while the same light generates heat and reactive molecules [7]. That is a therapy whose potency depends on what the residual tissue supplies, not on what the surgeon managed to see. It runs at the same wavelength as the imaging [2], which saves hardware and couples the risks: an optical budget that fails at human dimensions fails for both functions, and Shi names human-brain scale as the outstanding confirmation [13].
For scale, 44 micrometers is 0.044 mm [3], below what the team says current clinical imaging tools resolve [5]. And glioblastoma's roughly 7 percent five-year survival, driven by cells that infiltrate neighbouring tissue and by a blood-brain barrier that limits what drugs and radiotherapy reach, is why a residual-cell treatment is worth this much engineering [14][15]. Safety so far reads clean: follow-up testing found no detectable neurological or motor impairment [12], which is the first thing to look for in anything activated inside a fresh resection cavity, with the caveat that "no detectable" covers the measures used over the window observed.
What the work does establish is that one material can be built to do both jobs and that, bundled, it outperformed surgery alone in mice while reducing recurrence [16]. My view, with its conditions attached: the phototherapy is the half I would want measured on its own next, because that is the half whose failure mode is biology rather than optics.
