We read with great interest the study by Chen et al1 entitled “Osteosclerotic changes on computed tomography predict disease progression and poor survival in prostate cancer with osteoblastic metastases.” The authors conducted an analysis of changes in bone attenuation on computed tomography (CT) among patients with prostate cancer and osteoblastic bone metastases, and found that increases in sclerosis during systemic therapy were statistically associated with inferior radiographic progression-free survival (rPFS) and overall survival. Osteosclerosis can be used as a biologically potential marker of poor outcome, providing a new approach to disease assessment. We would like to offer several comments on how to translate these findings into a clinically applicable risk stratification framework for patients with metastatic prostate cancer. Osteosclerosis is a time-dependent outcome, but the covariates did not incorporate time. These changes do not exist at baseline; they emerge only after therapy has begun. Patients must survive long enough – and remain under imaging surveillance – to be classified accordingly. Conventional proportional hazards models, however, freeze this evolving feature into a static covariate. The result is subtle but consequential: exposure status becomes entangled with future survival time. In longitudinal observational settings, such design choices can quietly bend the data, allowing differences in follow-up duration or disease course to masquerade as prognostic signal rather than reflecting the independent effect of OCs2,3. From an inferential standpoint, only by explicitly accounting for the timing of OC emergence through time-updated covariate modeling or clinically defined landmark analyses can we clarify whether sclerosis precedes subsequent disease progression or simply coincides with disease that has already entered a more aggressive phase. In this context, the effect estimates derived from standard survival models may partly reflect subsequent clinical decision-making behavior rather than the prognostic role of osteosclerotic changes per se. Sensitivity analyses that consider treatment regimen adjustments as endpoint events, or alternative analytical frameworks that take into account competing risks, can provide a more adequate basis for interpreting the strength and direction of the observed associative effects3,4. The attenuation threshold appears to be informed, at least in part, by evidence derived from osteolytic metastases in other malignancies5, and the increase in CT attenuation values as a criterion for classification of osteosclerosis, but there are more causes of osteosclerosis in prostate cancer metastases, such as bone remodeling and pathological osteogenesis. Osteosclerosis CT attenuation values need to be corrected in conjunction with specific tumor data to determine baseline levels, which were not adjusted by the authors. The single lesion results with the most pronounced attenuation values are included in the text, which may increase the sensitivity and bias the results to one side of the spectrum, ultimately leading to a lack of robustness6,7. On further strengthening research: (1) Establish large-scale, prospective cohort studies incorporating temporal and sensitivity models to validate the value of osteosclerosis in predicting prostate cancer prognosis8; (2) Conduct tumor-specific calibration and robust analyses at the lesion level to validate the efficacy of imaging biomarkers; (3) Incorporate survival analysis frameworks to clarify definitions of rPFS and select more appropriate censoring rules9; and (4) Establish clinically actionable thresholds for CT-based bone sclerosis assessment, providing quantitative evidence for treatment response evaluation and therapeutic decision-making. The team of Chen et al provided a comprehensively designed and statistically analyzed study of prostate cancer osteosclerosis to assess prognosis, quantifying prognosis by means of CT, and the study provides new perspectives and a valuable research base. Establishing a large-scale prospective sample repository, adopting rigorous statistical design protocols, and precisely defining diagnostic criteria for osteosclerosis will play a significant role in the future10.
Zhou et al. (Thu,) studied this question.