![\"Hybrid](\"https:\/\/rtmedical.com.br\/wp-content\/uploads\/2026\/08\/pet-ct-her2-cancer-de-mama.jpg\")
What the study measured<\/h2>\n
The work was carried out by a group from the Fourth Hospital of Hebei Medical University and Hebei Provincial Tumor Hospital in Shijiazhuang, China. Forty-five patients with newly diagnosed HER2-positive breast cancer underwent serial PET\/CT with the 68<\/sup>Ga-HER2 tracer across neoadjuvant therapy. In the pathologic outcome analysis, 44 patients were evaluable: 31 (70%) achieved pathologic complete response (pCR) and 13 (30%) did not.<\/p>\n The serial design is the crux. This is not a single scan at diagnosis but a sequence: a baseline study before treatment, an early second study during the regimen and a third in the preoperative phase. The question the study asks is not “how much does this tumor take up” but “how much has uptake fallen since last time.”<\/p>\n That distinction matters because today’s clinical alternative is to wait. In the standard pathway, the patient completes the full neoadjuvant course, goes to surgery, and only then does pathology reveal whether there was a complete response. Bringing that information forward by a few weeks opens the door to intensifying, switching or de-escalating for those who are not responding.<\/p>\n The most instructive finding is a negative one: at the first PET, maximum standardized uptake value (SUVmax) did not differ significantly between patients who would go on to achieve pCR and those who would not. In other words, baseline uptake intensity alone does not predict response. That makes sense \u2014 every patient was HER2-positive by definition, so all expressed the target before treatment began.<\/p>\n The signal lived in the dynamics. Optimal cutoffs for predicting pCR were SUVmax \u2264 1.15 at the second scan, with an area under the ROC curve (AUC) of 0.738, and SUVmax \u2264 0.50 at the third scan, with an AUC of 0.735. Applying the second-scan cutoff, the group identified a therapy-sensitive subgroup with a positive predictive value of 85% for pCR \u2014 roughly 15 percentage points of absolute probability above the non-sensitive subgroup \u2014 with a negative predictive value around 60%.<\/p>\n It is worth recalling how SUV is constructed, because that explains its fragility as an absolute number:<\/p>\n $$\\mathrm{SUV} = \\frac{C_{tissue}}{D_{inj}\/m}$$<\/p>\n where $C_{tissue}$ is the activity concentration measured in tissue, $D_{inj}$ the injected dose and $m$ the patient’s body mass. Because the denominator depends on weight and on decay between injection and acquisition, absolute SUV varies with protocol, scanner and reconstruction. That is why the ratio between successive scans of the same patient on the same scanner is intrinsically more robust than any isolated value \u2014 and that is precisely this study’s logic.<\/p>\n Eighty-five percent positive predictive value sounds strong, and the number deserves careful reading. The pCR prevalence in this cohort was 70%. That means that with no imaging at all, betting “complete response” on any patient in the series would already be right 70% of the time. The scan raises that probability to 85% in the subgroup flagged as sensitive \u2014 a real gain, but incremental over an already high baseline.<\/p>\n The negative predictive value of about 60% is the weaker side and the more consequential one for clinical decisions. If the scan suggests a patient is not sensitive, it is wrong in roughly 40% of cases. No oncologist should abandon an ongoing regimen with that degree of uncertainty. In practice, this is a test that reinforces confidence in patients doing well, but does not yet authorize changing course in those who appear not to be.<\/p>\n It is fair to note that an AUC of about 0.74 places the method in the moderate discrimination range. It is not a reliable dichotomous test; it is a probability enrichment. For clinical trial design \u2014 selecting de-escalation candidates, for instance \u2014 that level of discrimination already has immediate use.<\/p>\n Routine oncologic molecular imaging uses FDG, a glucose analog that labels metabolically active tissue. It is sensitive and nonspecific: inflammation takes it up, infection takes it up, repair tissue takes it up. In a treatment-response setting, that nonspecificity is a genuine problem, because the therapeutic effect itself generates inflammation and can mimic persistent disease.<\/p>\n The anti-HER2 tracer works on different logic. It binds the HER2 receptor expressed on the tumor cell membrane, so uptake reflects target availability rather than metabolic activity. When uptake falls, the biological reading is direct: either the receptor-expressing cells were eliminated, or the receptor was blocked by the therapeutic antibody. Both indicate that targeted therapy is finding its mark.<\/p>\n Gallium-68 also carries a logistical advantage: with a half-life of roughly 68 minutes and production from a germanium-68 generator, it requires no on-site cyclotron \u2014 which makes the technique feasible in mid-sized services, unlike fluorine-18 tracers dependent on regional production. Global radiopharmaceutical dependence, incidentally, explains market moves such as Curium’s talks to acquire Lantheus for $7 billion<\/a>.<\/p>\n For the oncologist, the most immediate application is not deciding treatment but stratifying. A patient flagged as sensitive at the second scan has a high probability of pCR, which supports conversations about surgical extent and protocol-based de-escalation strategies. For the nuclear medicine physician, the message is about standardization: because interpretation depends on comparison between scans, identical acquisition protocol, the same injection-to-acquisition interval and the same reconstruction stop being good practice and become a validity requirement.<\/p>\n Access is the binding constraint before technique is. FDG PET\/CT already faces coverage and regional distribution limits in many health systems; HER2-directed tracers are not commercially available and would require local production and regulatory approval. The realistic short-term path is research centers and institutional protocols, not broad adoption.<\/p>\n That does not diminish the finding’s interest. The search for imaging markers that anticipate response is a consistent line in recent literature, including MRI approaches \u2014 such as the radiomic model predicting PD-L1 expression in breast cancer from MRI<\/a> \u2014 and AI-assisted detection tools like DeepHealth’s FDA-cleared breast ultrasound software<\/a>. The common denominator is clear: extract from imaging the information only pathology delivers today, and deliver it earlier.<\/p>\n The limitations are those expected of a study of this nature: small cohort, single center, no external validation, and cutoffs derived from the sample itself \u2014 which typically overestimates performance when applied to new populations. The authors themselves condition broader adoption on validation in larger multicenter cohorts.<\/p>\n Practically relevant questions also remain open. The ideal timing of the second scan is not established, and it determines the entire clinical utility: too early and the signal has not formed; too late and there is no time left to change course. Direct comparison with cheaper alternatives, such as contrast-enhanced MRI or FDG-PET in the same serial design, is also missing. Without it, we cannot know how much the targeted tracer adds to what is already available.<\/p>\n The authors’ conclusion is appropriately restrained: dynamic changes in 68<\/sup>Ga-HER2 PET\/CT uptake during neoadjuvant therapy constitute an effective early indicator of therapeutic sensitivity in HER2-positive breast cancer. This is not yet a routine exam \u2014 it is a proof of concept with plausible numbers and a well-defined validation path.<\/p>\nSUVmax, cutoffs and AUC<\/h2>\n
An 85% PPV and what it does not solve<\/h2>\n
Why HER2-PET differs from FDG<\/h2>\n
Clinical implications and access<\/h2>\n
Limits and next steps<\/h2>\n