An index drawn from a baseline PET/CT scan was associated with poor pathological response after pre-surgery chemoradiotherapy for locally advanced rectal cancer, a retrospective study found. The study included 48 patients. Among the scan features examined, the heterogeneity index had the highest numerical AUC, but its result was comparable with metabolic tumor volume and total lesion glycolysis.
The researchers wanted to know whether information from a baseline 18F-FDG PET/CT scan could predict a Ryan 3 poor pathological response after neoadjuvant, or pre-surgery, chemoradiotherapy. Patients were classified as Ryan 3 poor or non-response, or Ryan 0–2 response, according to surgical pathology.
All 48 patients had histopathologically confirmed locally advanced rectal cancer. Each underwent baseline PET/CT before chemoradiotherapy and then curative-intent low anterior resection. The imaging measures were compared with the response category recorded from the surgical pathology.
Sixteen patients had Ryan 3 poor pathological response, while 32 had Ryan 0–2 response. The Ryan 3 group was therefore the smaller of the two groups used in the analysis.
What the scan captured
From the primary tumor, the team calculated SUVmax, SUVmean, metabolic tumor volume, or MTV, total lesion glycolysis, or TLG, and a heterogeneity index, or HI. HI was derived with linear regression from MTV values measured at thresholds of 40%, 60% and 80% of SUVmax.
The statistical work included ROC analysis, logistic regression and bootstrap internal validation. ROC analysis was used to assess how well the PET measures distinguished the two pathology-defined response groups, while the regression models examined whether the measures remained associated with poor response after adjustment for T stage.
Heterogeneity led, but not by much
MTV, TLG, HI and clinical or radiological T stage were significantly higher in patients with Ryan 3 poor response. Age, sex, CEA, clinical or radiological N stage, SUVmax and SUVmean did not differ significantly between the two groups, and SUVmax and SUVmean showed no significant discriminatory value.
The study used AUC as its measure of discrimination, meaning how well a scan feature separated the poor-response group from the response group. HI had an AUC of 0.760, with a bootstrapped 95% confidence interval of 0.586 to 0.899. MTV and TLG had similar AUCs of 0.749 and 0.748. HI was numerically highest, but the reported results do not show that it was statistically superior to either metric.
After adjustment for T stage in restricted models, HI, MTV and TLG each remained independently associated with Ryan 3 poor response. Bootstrap optimism-corrected AUCs were 0.847 for the HI plus T stage model, 0.842 for MTV plus T stage and 0.813 for TLG plus T stage.
The HI plus T stage model had the highest reported corrected AUC, but the difference from the MTV plus T stage model was modest. Confidence intervals for the optimism-corrected AUCs were not reported in the supplied analysis.
What the result does not show
The authors say baseline HI, MTV and TLG may help identify patients at risk of Ryan 3 poor pathological response, with workstation-calculable HI potentially adding complementary information for pretreatment risk stratification. Because the study was retrospective, the result is an association, not evidence that the imaging features caused poor response or that using them would improve care.
That design narrows how far the result can be taken. It was a retrospective, single-center cohort of 48 patients, including 16 with Ryan 3 poor response and 32 with Ryan 0–2 response. Bootstrap internal validation was reported, but no external validation cohort was described, so the findings do not establish how the models would perform in other centers or populations.
The supplied analysis does not give odds ratios, confidence intervals or p-values for the adjusted associations. It also gives no confidence intervals for the optimism-corrected AUCs. That makes the precision of those estimates harder to judge.
The remaining questions are whether the pattern holds in larger, multicenter cohorts, whether HI adds meaningful information beyond T stage and the other PET metrics, and whether such classifications could affect treatment decisions or patient outcomes. Those questions were not answered by this study's pathological-response analysis.
The paper reports publication and version-of-record dates of 21 August 2026. It says the study received no funding and that the authors had no conflicts of interest.
Paper data and sources
Original title: BaselineF-FDG PET/CT-derived heterogeneity index predicts poor pathological response after neoadjuvant chemoradiotherapy in locally advanced rectal cancer.
Authors: Yiğithan Okar, Ömer Faruk Şahin, Seval Erhamamcı et al.
Journal/Repository: Annals of nuclear medicine
Status: Peer-reviewed
First online: 2026-08-21
DOI: 10.1007/s12149-026-02270-5
Original paper · Full text