Methodological considerations in interpreting serum galectin‐3 as a predictor of incident type 2 diabetes

To the Editor,

We read with great interest the article by Takubo‐Yoshimura et al. ¹ , which reported that higher baseline serum galectin‐3 was associated with incident type 2 diabetes mellitus over 10 years, particularly among participants with lower BMI or waist circumference. This study provides valuable longitudinal evidence regarding galectin‐3 as a potential metabolic risk marker in a Japanese population. However, we would like to highlight two methodological considerations that may be relevant to the interpretation of the findings.

First, loss to follow‐up raises the possibility of selection bias in the longitudinal analysis. Among 433 participants at baseline, only 307 underwent repeat examination; 87 had no follow‐up examination, 15 relocated, and 24 died, leaving 275 participants without baseline type 2 diabetes mellitus for analysis. Moreover, baseline serum galectin‐3 levels were associated with age, hypertension, cardiovascular disease, creatinine, and eGFR—characteristics that may also influence survival and participation. Restricting analysis to participants with available follow‐up data could therefore introduce informative selection, which may affect the estimated association between galectin‐3 and incident type 2 diabetes mellitus ² . Additional analyses, such as comparing baseline characteristics between retained and non‐retained participants, applying inverse‐probability‐of‐participation weighting, or conducting sensitivity analyses, could further clarify the robustness of the observed association.

Second, the association between galectin‐3 and incident type 2 diabetes mellitus may require further consideration regarding its functional form. The comparison between the highest and lowest galectin‐3 quartiles yielded an adjusted OR of 5.75 (95% CI 1.08–30.65), whereas the adjusted estimate per 1‐SD increase was not statistically significant (OR 1.49, 95% CI 0.89–2.49; P = 0.13). Event proportions were also not fully monotonic across quartiles (2.9%, 11.6%, 7.1%, and 17.7%). These differences between categorical and continuous analyses do not necessarily indicate inconsistency, as they may reflect nonlinear or threshold associations ³ . Categorizing a continuous predictor may obscure its underlying functional form, and modeling galectin‐3 continuously using restricted cubic splines, accompanied by overall and nonlinearity tests, could help determine whether a threshold, nonlinear, or null association is supported.

In summary, the study offers an important perspective on galectin‐3 and future type 2 diabetes mellitus risk. Further evaluation of potential selection mechanisms and the functional form of galectin‐3 may improve the interpretation of the observed association and help guide future biomarker research and preventive strategies.

DISCLOSURE

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.

DATA AVAILABILITY STATEMENT

Data sharing is not applicable to this article as no new data were created or analyzed in this study.