Response to the comment of Wasti on “Association between severity of diabetic complications and risk of cancer in middle-aged patients with type 2 diabetes”

Abstract

Dear Dr. Hotta,

We appreciate Wasti's thoughtful comment¹ on our research recently published in the Journal of Diabetes Investigation. In this study, we demonstrated a positive association between the severity of diabetic complications and cancer risk in middle-aged patients with type 2 diabetes². We carefully evaluated concerns related to cohort selection, measurement of complication severity, and potential confounding factors, and we appreciate the opportunity to clarify these issues.

Our study focused on participants aged 40–65 with newly diagnosed type 2 diabetes who received antidiabetic treatment. By excluding younger individuals—who may have type 1 or atypical early-onset diabetes—and elderly patients with multiple comorbidities and prolonged disease durations, we established a more homogeneous cohort. It was particularly crucial to exclude patients who did not use medication, as untreated individuals are more likely to be misclassified or to represent only very mild diabetes managed solely by diet. We minimized misclassification bias and ensured that our cohort consisted exclusively of patients with clinically active, pharmacologically treated diabetes. This careful selection strategy offers a more comprehensive understanding of the incidence of cancer following the initial diabetes diagnosis^{3, 4}.

We measured the severity of diabetes complications using the adapted Diabetes Complications Severity Index (aDCSI), a validated tool for claims data that assigns scores across seven major complication categories. By updating the aDCSI annually as a time-dependent measure, we were able to track the cumulative and evolving effects of diabetes over time. This approach more accurately reflects the progression of disease burden and monitors changes in patient conditions even without direct clinical measures such as HbA1c. Although our dataset lacked comprehensive lifestyle information (e.g., body mass index, smoking status, or alcohol consumption), we adjusted for available proxies, including age, sex, income, urbanization, hypertension, hyperlipidemia, and medication usage. While these adjustments may not eliminate residual confounding, they provide a robust basis for the validity of our findings.

The overall hazard ratio of roughly 1.2 may appear modest; however, it signifies a substantial risk elevation given the high prevalence of both diabetes and cancer. Notably, our subgroup analysis revealed that younger patients—particularly those aged 40–44—experienced an almost 1.8-fold increased risk, underscoring our work's significant public health implications. To mitigate potential surveillance bias and reverse causality, cancer outcomes were obtained from a national registry that requires pathological confirmation, and we imposed a one-year lag period to exclude cancers present at the time of diabetes diagnosis. Furthermore, we utilized a competing risks model to account for the higher mortality rate observed in patients with severe complications. This method ensures unbiased estimates by effectively censoring individuals who die from causes other than cancer.

These methodological refinements underscore our confidence in the observed association, even after adjusting for multiple confounding factors and biases inherent in retrospective analyses. We aim to address the remaining uncertainties by conducting a prospective trial to more thoroughly assess the temporal relationship between cancer development and diabetic complications.

The authors declare no conflict of interest.

Approval of the research protocol: This study used de-identified, retrospective claims data; hence, formal ethical approval was not required.

Informed consent: N/A.

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

Animal Studies: N/A.