The ACCORD Study is a double 2x2 factorial trial designed to test the effect of an intensive glucose control strategy defined as a goal hemoglobin A1C ≤ 6.0%, versus a standard control strategy (goal hemoglobin A1C between 7.0 and 7.9%), intensive blood pressure control strategy (SBP < 120 mmHg) versus standard control strategy (SBP < 140 mmHg), and a lipid treatment strategy that utilizes fenofibrate plus a statin, compared to one that uses a statin alone, on the composite outcome of myocardial infarction, stroke or cardiovascular death . The intensive glycemia intervention was stopped early due to increased mortality in this group, and all participants were transitioned to the standard glycemia intervention for the remainder of the trial. This analysis is limited to the period of time prior to this transition. Main results have been previously reported [6–8], as has a full description of the methodology and the rationale for the trial [5, 9]. All clinical sites obtain Institutional Review Board approval for the ACCORD Study.
Study participants and design
Briefly, participants were eligible to enroll in ACCORD if they had type 2 diabetes, a hemoglobin A1C of 7.5-11%, and were either between the ages of 40 and 79 years with cardiovascular disease, or between the ages of 55 and 79 years with evidence of significant atherosclerosis, albuminuria, left ventricular hypertrophy, or two or more additional risk factors for cardiovascular disease (dyslipidemia, hypertension, current smoker, or obesity). Participants also needed to meet inclusion criteria for blood pressure and/or lipid trials. Participants were excluded if they had a history of frequent or recent serious hypoglycemic events, history of coma/seizure within the last 12 months, unwillingness to do home glucose monitoring or inject insulin, a body-mass index (the weight in kilograms divided by the square of the height in meters) of more than 45 kg/m2, a serum creatinine level of more than 1.5 mg per deciliter (133 μmol per liter), or other serious illness.
Participants were randomly assigned to intensive or standard glycemia intervention strategy. All participants received education and counseling about diabetes care. Supplies for glucose monitoring and glucose-lowering medications (from a study-supervised formulary of FDA-approved medications) were provided. Glucose lowering medications not on the ACCORD formulary could be prescribed to participants but were not provided free of cost to participants. Medications were utilized at the discretion of the study physician, with a suggested algorithm of order of use provided in the protocol. Use of medication class was not limited by study assignment. Intensive group participants were expected to be seen and medications adjusted every 2 months with at least one phone contact between visits while standard group participants were expected to be seen every 4 months (or more frequently as needed). Self-monitoring of blood glucose (SMBG) was advised for participants in both groups with more frequent monitoring recommended for intensive group participants and less frequent for standard group participants. Participating clinical sites included diabetes clinics, primary care clinics and dedicated research clinics. For this report, we used data for all follow-up contacts where ACCORD physicians actively managed the participant’s glycemia medications through February 5, 2008, the date that participants were notified that intensive glycemia participants were to be transitioned to standard glycemia therapy.
Definition and reporting of hypoglycemia
Participants were asked at every visit if they had experienced episodes of low blood sugar. A full description of the review of such events, including the adjustment of therapeutic goals in response to severe hypoglycemia, has been previously reported . The present report focuses on symptomatic, severe hypoglycemic events requiring medical assistance (hospitalization, visit to the emergency room, treatment by medical personnel including emergency medical technician either in a clinical setting or at home) which was defined in ACCORD as either a blood glucose less than 50 mg/dl (2.8 mmol/L) or symptoms that promptly resolved with oral carbohydrate, intravenous glucose, or glucagon. NOTE: Prior to March 2003, clinic sites were not required to document the blood glucose level of the participant during a hypoglycemic episode requiring medical assistance. Fifteen percent of the episodes used in the analyses in this paper do not have this documentation. Exclusion of events without this documentation does not qualitatively change the conclusions. (Data not shown.)
Following a report of a severe hypoglycemic episode by the participant, research staff inquired about the circumstances surrounding the episode. Standardized forms that utilized lists, check boxes and free text boxes were used to collect information. The time of the event (4 hour intervals check boxes) and whether the event had occurred while the participant was asleep was collected. Symptoms that the participant experienced prior to treatment of the hypoglycemia were ascertained and recorded. Symptom choices included both physical symptoms (e.g., shakiness, fast heartbeat) as well as psychological (e.g., anxiety) and neuroglycopenic (e.g., confusion). More than one symptom could be reported.
Activities and circumstances immediately antecedent to and consequences of the event were also obtained. Antecedents included changes in diet and exercise, changes in both glucose lowering medications and other medications, and intercurrent illnesses. Potential consequences included physiological responses (loss of consciousness, seizure), emergency treatment, and accident or injury. For both antecedents and consequences, more than one response was possible.
All statistical analyses were conducted at the coordinating center with the use of SAS software, version 9.2 (SAS Institute). All analyses were conducted on the first reported hypoglycemic event requiring medical assistance (HMA). Symptoms reported at the time of the event and immediate antecedents to events are summarized as the number and percent of all initial events.
Use of glucose-lowering medications was summarized according to study group as the total person-years that participants were prescribed medications. Person-years were determined by using clinic reports of what was prescribed for the interval of time between each clinic visit. Rates of HMA per 100 person-years of medication use was obtained by taking the number of initial events that occurred while participants were prescribed a medication divided by the total number of person-years of use.
Hazard ratios (95% CI) relating the risk of HMA to prescription of medications were obtained within glycemia treatment groups using Cox proportional hazards (PH) regression models allowing for medication use as time-dependent covariates and controlling for the baseline covariates identified in Miller et al.  as predictive of future episodes of severe hypoglycemia. These baseline covariates included: age, gender, race, education, BMI, history of neuropathy/nerve problems, time since diagnosis of diabetes, A1C, albumin to creatinine ratio, serum creatinine, LDL-C, as well as factors used to stratify randomization (treatment groups within the BP and Lipid trials and the presence of clinical cardiovascular disease). Series of models that were fitted included baseline covariates plus each of the following as a single additional factor entered into the model: 1) each medication; 2) the oral/insulin medication category variables; 3) the number of medications. In addition to these models, we fitted models with all medications in the model and a priori selected the following interactions to investigate: oral agents and insulin, insulin and insulin sensitizers; metformin, sulfonylurea and TZDs; insulin, sulfonylurea and TZD.