Comment on “Physiological phenotypes of obstructive sleep apnea identified by cluster analysis: Association with type 2 diabetes” Authors: Huijuan Cheng, Zhe Zhang, Feng Lu Published: 2026 Dear Editor, We read with great interest the study by Osawa et al., who used polysomnographic data and unsupervised clustering to identify physiological phenotypes of obstructive sleep apnea (OSA) and examine their associations with type 2 diabetes mellitus (T2DM) 1 . This multidimensional approach offers a valuable perspective beyond conventional apnea–hypopnea index (AHI)‐based severity classification. However, we believe that the further interpretation that nocturnal hypoxemia plays a “predominant role” over sleep fragmentation or overall AHI severity in glucose dysregulation warrants more cautious qualification. In the multivariable model including age, sex, body mass index, CT90, estimated arousal threshold (ArTH), and AHI, CT90 remained significantly associated with T2DM (OR 1.02 per 1% increase, 95% CI 1.00–1.05; P = 0.020), whereas ArTH (OR 1.00, 95% CI 0.95–1.05; P = 0.984) and AHI (OR 1.00, 95% CI 0.98–1.02; P = 0.745) were not statistically significant 1 . The authors subsequently interpreted these findings as indicating that nocturnal hypoxic burden was more strongly related to T2DM and ultimately suggested that nocturnal hypoxemia, rather than sleep fragmentation or overall AHI severity, plays a predominant role in glucose dysregulation. However, a key premise underlying this interpretation deserves further consideration. Estimated respiratory ArTH reflects the physiological threshold or propensity for respiratory events to trigger cortical arousal. Although it is related to sleep fragmentation, it is not equivalent to the actual burden of sleep fragmentation 2 . Notably, the study itself reported a directly observed arousal index, which differed substantially across the identified OSA phenotypes 1 . Therefore, the absence of an independent statistical association between estimated ArTH and T2DM does not establish that sleep fragmentation itself lacks an independent association with T2DM. Even if ArTH is provisionally accepted as a surrogate for sleep fragmentation, the reported statistical results still do not establish the comparative claim that CT90 is “more strongly related” to T2DM. Statistical significance for one variable and non‐significance for another do not themselves demonstrate a statistically significant difference between their associations 3 . In addition, CT90, AHI, and ArTH were entered into the model on different measurement scales, so their odds ratios cannot be directly compared by magnitude. The current findings therefore more securely support the conclusion that CT90 remained independently associated with prevalent T2DM within the specified multivariable model. They do not establish that the association of CT90 with T2DM is stronger than that of ArTH or AHI, nor do they by themselves establish a predominant pathophysiological role for nocturnal hypoxemia in glucose dysregulation. A sensitivity analysis incorporating the directly observed arousal index could further clarify the relationship between sleep fragmentation‐related measures and T2DM. If the terms “more strongly related” or “predominant role” are to be retained, a formal comparison of the associations or model contributions of these physiological measures would be preferable to inference based solely on whether individual P values cross the conventional significance threshold. These considerations do not diminish the value of data‐driven PSG phenotyping. Rather, distinguishing the robust finding of an independent association between CT90 and prevalent T2DM from stronger claims regarding the relative importance of individual OSA mechanisms would make the interpretation of this important study more precise. AUTHOR CONTRIBUTIONS Huijuan Cheng, Zhe Zhang, and Feng Lu: Conceptualization, Methodology, Writing—original draft, Writing—review and editing. FUNDING This work received no specific funding. 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 because no new data were generated or analyzed. References Associated Data Data Availability Statement Data sharing is not applicable to this article because no new data were generated or analyzed.