Study design and population
A retrospective cohort study included consecutive patients who underwent clinically indicated exercise treadmill stress testing at King Abdulaziz cardiac center between April 2001 and December 2016. Baseline demographics, past medical history, and medications at the stress test were obtained by reviewing the electronic medical record and database search using the ICD-9 coding system. Data about the exercise stress testing were collected from the stress testing unique system (MUSE). Patients who had diabetes at the baseline before exercise stress testing, patients younger than 18 years, records with incomplete stress testing information, and non-Bruce protocol exercise stress testing were excluded from the study analysis. The study was conducted in full accordance with the protocol and the current revision of the declaration of Helsinki, the Good Clinical Practice. The study was a part of the Saudi CArdioRespiratory Fitness (SCARF) project (Study protocol: RC16/103/R - approved by King Abdullah International Medical Research Center (KAIMRC)).
Exercise treadmill stress testing
Patients underwent symptom-limited maximal treadmill stress testing, which followed the standard Bruce protocol. The test day was pointed as the individual study baseline. Individuals’ results of the initial exercise test were included in the database. Resting measures, heart rate, and blood pressures were measured in the seated position and were recorded immediately before each test initiation. Supervised clinicians were following American Heart Association/American College of Cardiology (AHA/ACC) guidelines for terminating the test if the patient had Exercise limiting symptoms: chest pain, shortness of breath, significant arrhythmias, abnormal hemodynamic responses, diagnostic ST-segment changes, other limiting symptoms independent of the achieved heart rate or if the participant was unwilling or unable to continue. Otherwise, patients could reach their peak attainable workload independent of the heart rate achieved. Target heart rate was calculated as 85% of the age-predicted maximal heart rate, which is the patient age subtracted from a constant value of 220. Metabolic equivalents (METs) were adopted to represent cardiorespiratory fitness status based on the workload derived from the maximal speed and grade achieved during the total treadmill time. METs results were categorized into four groups: <6, 6-9, 10-11, and ≥12 METs.
Study definitions for risk factors
The history of hypertension was defined as a prior diagnosis of hypertension or the use of antihypertensive medications at the time of stress testing. Dyslipidemia was defined as the prior diagnosis of any significant lipid abnormality in the medical records or lipid-lowering medication use. On a prior angiogram, patients with obstructive coronary artery disease (CAD), prior myocardial infarction, coronary angioplasty, or coronary artery bypass surgery are considered known coronary artery disease. Prior congestive heart failure was defined as a prior clinical diagnosis of systolic or diastolic heart failure.
Study outcome: Incident Diabetes
Incident diabetes was determined among patients without diabetes at baseline and defined as clinical diagnosis of diabetes in the medical records or clinical problem list, use of anti-hyperglycemic medications including insulin, or had lab results suggestive of diabetes post-exercise testing date. Time-to-incident diabetes was based on the time between treadmill testing and the date of the first encounter with a new diabetes diagnosis.
Statistical analysis plan
Study participants were divided into four groups based on their METs (<6, 6-9, 10-11, and ≥12). Categorical variables were presented in frequencies and percentages. Continuous variables were presented base on the normality of distribution as mean ± standard deviation or median and interquartile ranges, overall and in categories of METs. The four groups were compared using Chi-square or Fisher exact test for categorical variables. Analysis of Variance (ANOVA) and Kruskal Wallis test were used for continuous variables compersion, as appropriate. Cumulative incidence was presented at 5-, 10-, and 15-yr intervals via a bar graph.
Kaplan-Meier cumulative incident diabetes was computed for different exercise capacity groups, and they were compared using the log-rank test. Cox regression was used to compute hazard ratios (HR) and 95% confidence intervals (CI). A forward selection technique was used to demonstrate independent predictors of incident diabetes. In each forward step, we added a related set of variables to improve our model. Therefore, gender, age, and heart rate were included in the baseline model (model 1). Subsequently, hypertension, hyperlipidemia, smoking, lung disease, previous coronary artery disease and previous congestive heart failure (model 2), antihypertensive medications, lipid-lowering medications, and proton pump inhibitors (model3), METs achieved, chronotropic incompetence, and Duke score (model 4). The primary analysis assessed the association between exercise capacity and time-free event of incident diabetes after adjusting for baseline clinical and cardiovascular risk factors. In the baseline model, we included age, gender. Subsequently, cardiovascular risk factors and related cardiac medications (Model 2), METs groups (Model 3) were added to the baseline model. We computed Harrell's concordance index (C-index) area under the Curve (AUC) and Akaike information criterion (AIC) to compare the models using this methodology developed for incidence diabetes. The selection of variables for entry consideration was based on clinical judgment, results of previous publications, and the expertise of the investigators. We also plotted a restricted cubic spline model to show the shape of the continuous relationship between METs and incident diabetes after adjustment for covariates. Finally, we examined the association between METs and incident diabetes in the subset of participants with a BMI measurement (N=6,539) (Supplementary Table 1). All Statistical analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA). Statistical significance was defined as P ≤ 0.05.