Authors’ reply

We thank Denegri et al¹ for their thoughtful critique of our recent paper² and welcome the opportunity to respond.

Regarding the comment on coronary vs peripheral blood sampling, lesion‑adjacent coronary sampling was a deliberate and mechanistically principled design choice. Because optical coherence tomography—the study’s reference standard for high‑risk plaque (HRP) identification—requires invasive catheterization, sampling blood proximally to the target lesion during the same procedure captures the local microenvironment most directly relevant to plaque biology, avoiding the confounding dilution inherent in peripheral venous sampling for a lesion‑specific pathological process. Demonstrating a copper–HRP association at the lesion level is therefore a necessary first step before peripheral validation can be meaningfully designed. We fully agree that peripheral venous testing is the indispensable next phase for clinical translation, and a follow‑up study with paired coronary and peripheral venous samples in an expanded cohort is already underway.

With respect to specificity of total serum copper, the most direct evidence against a nonspecific acute‑phase interpretation is the pattern of cytokine associations: copper correlated positively with matrix metalloproteinase‑9, a protease directly implicated in fibrous cap degradation, and inversely with interleukin‑10, a plaque‑stabilizing cytokine.³ This biological profile is inconsistent with copper functioning as a mere systemic inflammatory marker. Supporting this, high‑sensitivity C‑reactive protein (hs‑CRP) levels did not differ significantly between the HRP and stable plaque groups (P = 0.59) in our study, arguing against copper elevation driven by acute‑phase response. To minimize confounding, the patients with hs‑CRP levels equal to or greater than 10 mg/l, active infections, autoimmune diseases, or liver disease were excluded at enrollment. Copper levels were also invariant across subgroups stratified by sex, age, hypertension, or diabetes, further arguing against a general comorbidity or nutritional confounder. We agree that ceruloplasmin quantification would sharpen interpretation, and this approach will be incorporated in our ongoing and future investigations.

We appreciate the point on model development and validation and wish to clarify an important conceptual distinction—the TRIPOD (Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis) statement itself distinguishes between clinical prediction models and exploratory risk‑factor discovery models, and different validation standards apply to each.⁴ Mandatory external validation is required for clinical prediction models intended to guide individual patient decisions in new populations. Our model was constructed as an exploratory risk‑factor discovery model: its purpose was to identify independent associates of HRP and to quantify the incremental discriminatory gain from adding serum copper to established markers. For such hypothesis‑generating analyses, external validation is not a methodological prerequisite.⁵ Nonetheless, several design features provide internal reassurance: the model retained only 3 predictors via stepwise regression with stringent entry and retention criteria, the Hosmer–Lemeshow test confirmed satisfactory calibration (P = 0.8), and a sensitivity analysis applying a stricter HRP definition yielded essentially unchanged results. We openly acknowledge single‑cohort testing as a limitation and have clearly stated so in the manuscript. External validation across geographically and ethnically diverse populations is planned within our ongoing multicenter study at 3 centers in eastern China.

We appreciate this constructive exchange and look forward to reporting the results of our peripheral validation studies.