Porta · Diabetes care 2014 · retrospective cross-sectional modeling study · n=12074

Estimating the delay between onset and diagnosis of type 2 diabetes from the time course of retinopathy prevalence.

Cited 146 times in the scientific literature.

Level 4 - case-series / case-control

Cross-sectional modeling and retrospective analysis of clinic records.

PubMed 24705614 · doi:10.2337/dc13-2101 · record verified 2026-08-27

What was done

Researchers evaluated 12,074 diabetic outpatients (35,545 fundus examinations) screened for retinopathy at a hospital-based diabetes clinic to recalculate the duration of undiagnosed type 2 diabetes. Patients were stratified by age of onset (younger onset <30 years [YO] vs. older onset ≥30 years [OO]), treatment (insulin-treated [IT] vs. not insulin-treated [NIT]), and retinopathy severity (any diabetic retinopathy [AnyDR] vs. moderate or more severe diabetic retinopathy [ModDR]). Mathematical models were applied to extrapolate the time from retinopathy emergence to diagnosis in the OO-NIT group and the time required for retinopathy to develop after diabetes onset in the YO-IT group.

What was found

The OO-NIT cohort included 1,719 patients with AnyDR and 685 with ModDR, while the YO-IT cohort included 756 with AnyDR and 385 with ModDR. In the OO-NIT group, a linear model estimated ModDR appeared 2.66 years prior to clinical diagnosis. In the YO-IT group, a quadratic model estimated ModDR appeared 3.29 years after diabetes onset. Combining these yielded an estimated delay of 6.05 years (2.66 + 3.29) between type 2 diabetes onset and diagnosis, compared with 13.36 years derived from standard traditional criteria.

Why it matters

This updated estimate reduces the presumed duration of undiagnosed type 2 diabetes by more than half, providing a revised timeline for disease progression and screening window calculations.

Limits

The findings rely on mathematical extrapolation rather than prospective longitudinal tracking from biological onset. The model assumes progression dynamics in younger-onset insulin-treated patients accurately reflect onset-to-retinopathy intervals for older-onset diabetes. Sourcing from a hospital-based clinic may introduce referral bias compared to general population cohorts.

Cited by