New OHTS analyses revisit protective effect of diabetes

June 1, 2008

In contrast to the original finding in the Ocular Hypertension Treatment Study, re-analyses using new data collected during follow-up shows a history of diabetes mellitus does not protect patients with ocular hypertension from progression to glaucoma.

Washington, DC-An updated analysis of data from the Ocular Hypertension Treatment Study (OHTS) has failed to confirm the original finding that a history of diabetes mellitus protects against progression from ocular hypertension to primary open-angle glaucoma (POAG), said Michael A. Kass, MD, at the annual meeting of the American Glaucoma Society.

The analyses used new self-reported information that was collected after publication of the original OHTS predictive model, said Dr. Kass, OHTS chairman and professor and chairman, Department of Ophthalmology and Visual Sciences, School of Medicine, Washington University, St. Louis. Patients were categorized as having a history of diabetes based on their responses to three different questions, including the question posed originally in the OHTS and two others considered to offer greater specificity but less sensitivity. Then, the original predictive model was rerun, modifying only the information on diabetes history.

In univariate and multivariate analysis using the refreshed data, a history of diabetes was not associated with the development of POAG for any of the three definitions tested. The multivariate hazard ratios for the three analyses ranged from 0.73 to 0.84, and all had 95% confidence intervals including "1."

"The original finding of a protective effect of diabetes mellitus was unexpected and puzzling," Dr. Kass said. "Considering the potential limitation of our ascertainment method, we decided to get additional data on self-reported diagnosis to evaluate this finding further. The OHTS also excluded participants with diabetes with any degree of retinopathy, so it is likely that we enrolled an atypical group of diabetics.

"Using new definitions to categorize patients as having diabetes showed no evidence that diabetes had a protective effect," he said. "These updated findings are consistent with previous studies that reported that diabetes mellitus either increased the risk of developing POAG or had no effect."

According to the OHTS predictive model published in 2002, diabetes mellitus appeared protective against progression from ocular hypertension to open-angle glaucoma, with a multivariate hazard ratio of 0.37 and 95% confidence interval of 0.15 - 0.90. The ascertainment method for identifying a history of diabetes involved asking patients whether they ever were told by their doctor that they had diabetes or sugar in their blood.

Collection of information for the new analyses began in February 2003. In addition to the original medical history question on diabetes, patients were asked whether a doctor or health professional had recommended a special diet to lower their blood sugar and whether they currently were taking insulin or diabetic pills to lower their blood sugar.

"Again, the information came from the patients, because we did not have the resources to review the medical records and did not perform laboratory tests," Dr. Kass said.

Three definitions

The three definitions used to identify diabetes status varied in sensitivity and specificity. The original question about patients being told they had diabetes or sugar in their blood was considered to have high sensitivity and low specificity. Dr. Kass described the question about diet as having moderate sensitivity and specificity and the question about interventions as having low sensitivity but high specificity.

Originally, in OHTS 191 patients were categorized as having a history of diabetes. All of those patients plus an additional 218 responded positively when the original question used to classify patients as having diabetes was asked during follow-up. The number of patients who responded affirmatively to the questions about diet and insulin/medication use was 277 and 256, respectively.

Dr. Kass noted that the reason for the discrepancy between the refreshed analysis and the original model is not known, but he suggested that it may reflect more complete ascertainment of diabetes mellitus. He also mentioned an editorial by Alfred Sommer, MD (Arch Ophthalmol. 2008;126:265-266), however, that admonished the OHTS investigators for "taking the giant leap of assuming that the subjects' responses were sufficiently precise and valid indicators of diabetes status to draw meaningful conclusions about the relationship between diabetes and the subsequent development of glaucoma."

"Dr. Sommer's concerns are valid but difficult to circumvent," Dr. Kass said. "Most medical history information is gathered from patient self-report. The only way to get around this is to conduct all research in a closed health system where there is access to electronic medical records. Even then, there are potential shortcomings relating to the completeness of testing in all patients, and there also are limitations when performing large studies concerning excessive cost and patient burden."

Dr. Kass is one of the authors of a paper reporting on the new analyses (Gordon MO, Beiser JA, Kass MA: Ocular Hypertension Treatment Study Group. Arch Ophthalmol. 2008;280-281). The glaucoma 5-year risk estimator based on the results of OHTS and the European Glaucoma Prevention Study is available online (http://ohts.wustl.edu/risk/calculator.html) and does not include a history of diabetes mellitus.OT