The reviewed paper on using PyCaret for predicting Type 2 diabetes mellitus (T2DM) demonstrates methodological and interpretative shortcomings that undermine its credibility. Several inconsistencies were found in the representation of cited studies, particularly overstating the novelty and capabilities of PyCaret and downplaying the role of laboratory data critical to diabetes prediction. The dataset’s limited diversity (NHS and HPFS cohorts) raises concerns about generalizability, while methodological flaws—such as reporting zero AUC values for top-performing models—highlight evaluation inconsistencies. Additionally, gender-specific analyses fail to provide distinct insights, as feature importance overlaps significantly across groups, contradicting claims of strong gender-specific predictors. Transparency issues, including a lack of reproducible code and ambiguous feature selection methods, further weaken the study’s findings. The omission of key limitations and selective emphasis on favorable metrics (e.g., AUC) suggest a possible bias in reporting.
