Beyond PAID ≥40: Preserving the clinical meaning of diabetes distress in type 1 diabetes

To the Editor,

We read with interest the study by Sakane et al. reporting a machine‐learning approach to predicting severe diabetes distress in adults with type 1 diabetes mellitus (T1D) using continuous glucose monitoring (CGM) metrics. ¹ We commend the authors for addressing the clinically important relationship between glycemic patterns and diabetes distress. However, we believe that the way diabetes distress was operationalized warrants particular consideration.

The study used the 20‐item Problem Areas in Diabetes (PAID‐20), with a score ≥40 defining the outcome. PAID‐20 was originally developed as a measure of diabetes‐related emotional distress, with 20 items representing distinct problem areas and a total score ranging from 0 to 100. ² Importantly, although ≥40 is an established threshold for identifying high diabetes distress, subsequent latent class analysis supports three clinically meaningful levels: low (0–16), moderate (17–39), and high (≥40) distress. ³ Thus, dichotomizing PAID at 40 is useful for identifying high distress but necessarily discards information about the continuum of distress.

This distinction is particularly relevant to the interpretation of the present machine‐learning model. In Sakane et al., none of the major CGM metrics—including time in range (TIR), time below range (TBR), coefficient of variation (CV), Glycemia Risk Index (GRI), Average Daily Risk Range (ADRR), or Low Blood Glucose Index (LBGI)—differed significantly between participants with and without PAID ≥40. Nevertheless, LBGI emerged as the most important feature in the neural‐network model. ^{1 , 4} This finding is intriguing, but it should not be interpreted as evidence that hypoglycemia causes diabetes distress. Diabetes distress may affect self‐management and glycemic patterns, while concerns about hypoglycemia may themselves contribute to distress; the cross‐sectional design cannot establish directionality.

Moreover, the neural‐network model achieved a recall of only 0.50. ¹ CGM‐derived variables may therefore be useful as signals that prompt psychosocial assessment, but they should not replace direct assessment with PAID.

Future studies should consider preserving PAID as a continuous outcome, while also examining individual PAID items or empirically validated dimensions. The clinically relevant question may not simply be whether CGM metrics can predict high diabetes distress, but which aspects of diabetes‐related distress are associated with specific glycemic patterns. Such an approach could move the field from predicting a binary PAID category toward understanding the particular burdens experienced by individuals living with T1D.

CGM should therefore be viewed as complementary to, rather than a substitute for, patient‐reported assessment of diabetes distress.

DISCLOSURE

S.M.K. has received lecture fees from Novo Nordisk, Mitsubishi Tanabe Pharma, and Kowa, and participated in a training program supported by Eli Lilly with travel support. The author declares no other competing interests relevant to this Letter.

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

Research data are not shared.