Letter to the Editor in Response to ‘A prediction model for diabetes complications using the Kokuho Database and its application to public health services in Japan’

Abstract

Dear Editor,

We read with great interest the article ‘A prediction model for diabetes complications using the Kokuho Database and its application to public health services in Japan’¹. This study represents a valuable endeavor in developing a diabetes complication prediction model leveraging large-scale data, with substantial implications for public health practice in Japan. However, two methodological considerations warrant further clarification to enhance the scientific rigor and clinical translatability of the findings.

First, the selection of a 6-year historical stratification window for ischemic heart disease and cerebrovascular disease requires mechanistic and empirical justification. Epidemiological evidence consistently indicates that diabetic macrovascular complications exhibit stronger associations with short-term metabolic control parameters (1–3 years) such as glycemic variability and blood pressure trajectories, rather than distant historical data^{2, 3}. A 6-year window may introduce collinearity by over-including less relevant historical records, potentially attenuating the weight of dynamic, time-sensitive risk factors. In addition, long-term indicators may not directly reflect the current risk status of the patient. The overall risk factor status of the patient during this period may change due to various factors (such as improved lifestyle, changes in medication, and the occurrence of other comorbidities). We urge the authors to clarify whether systematic comparisons of alternative time frames (1, 3, and 6 years) were performed to validate the optimal stratification criteria for predictive accuracy.

Second, the comparative prognostic value of early lifestyle determinants vs late biochemical markers in the model merits discussion. Lifestyle factors (sedentary behavior, dietary patterns, and smoking) typically precede biochemical abnormalities by 5–10 years, offering a critical window for primordial prevention. From a translational perspective, these modifiable behaviors represent more actionable targets for public health interventions compared to established metabolic derangements. Did the authors perform subgroup analyses to quantify the relative predictive power of these variables, particularly regarding their incremental value in early risk stratification?

Addressing these points would significantly strengthen the study's methodological robustness and public health relevance. We commend the authors for their contribution to this important field and anticipate their clarifications.

The authors declare no conflict of interest.

Approval of the research protocol: N/A.

Informed consent: N/A.

Registry and the registration no. of the study/trial: N/A.

Animal studies: N/A.

None.

Yanna Le and Qingyun Xu: methodology, writing—review and editing. Feiqi Xu: methodology, writing.