Study Design
This study was a retrospectively designed analysis of an ongoing prospective cohort study that aims to identify risk factors associated with diagnosis and outcome of vascular disease. Patients with AAAs were recruited from four outpatient vascular services in Australia, including The Townsville University Hospital, the Mater Hospital Townsville, Gosford Vascular Services and The Royal Brisbane and Women's Hospital. Imaging data collected from patients during AAA surveillance between 2003 and 2018 were used. Written informed consent was obtained from all participants and the study was approved by the relevant human ethics committee (approval numbers HREC/14/QTHS/203; HREC/05/QTHS/29; HREC/09/QTHS/117; HREC/10/QRBW/11; HREC/10/QRBW/208; MHS20100201-01; SSA/10/QTHS/49; MEC/08/08/095; RA/4/1/5765; H5213; H5269 and H6028). The study was performed in accordance with the Declaration of Helsinki. All data regarding the study and participants were stored in a centralised digital database that was accessible by approved personnel only.
Study Population
For inclusion in the study, patients were required to have information available to ascertain diagnosis of gout at recruitment and also had at least two ultrasound scans during AAA surveillance. Gout was defined as either diagnosis by a physician (International Classification of Diseases (ICD)-9 274, ICD-10 M10) or current or prior prescription of allopurinol, colchicine, febuxostat, probenecid or a combination of these medications [14–17]. Follow-up data of included patients were considered for a maximum of six years.
Risk Factors And Medications
Patients were assessed by clinical interview and physical examination in order to collect risk factors and medication history. Risk factors data including age, sex, history of hypertension, diabetes, ischemic heart disease (IHD), smoking and serum concentrations of cholesterol, triglyceride (TG), low density lipoprotein (LDL) and high density lipoprotein (HDL) were collected. Prescriptions of medications, including aspirin, metformin, statins, angiotensin-II inhibitors, beta-blockers, calcium channel blockers, allopurinol and colchicine were also collected at recruitment. Smoking was defined as either ever (current or ex-smoker) or never smoked [18–20]. Hypertension, diabetes and stroke were defined by prior diagnosis or treatment of these conditions at study entry [18–20]. IHD was defined by a history of myocardial infarction, angina or treatment of IHD at study entry [18–20]. Body mass index (BMI) was measured as described previously [21].
Aaa Imaging
Maximum anterior to posterior and transverse infra-renal aortic diameters were measured by experienced sonographers who were unware of the participants diagnosis of gout. Aortic diameter was measured from outer wall to outer wall of the artery. AAA diameter were measured using ultrasound machines employed in the vascular laboratories at each centre including Toshiba Capasee (Toshiba Medical Systems, North Ryde, New South Wales, Australia), Philips HDI 5000 (Philips Medical Systems, Bothell, Washington, USA), GE LOGIQ 9 (GE Healthcare, Chicago, Illinois, USA), Siemens Acuson Antares™ (Siemens Healthcare, Bayswater, Victoria, Australia) and Philips IU22 (Philips Medical Systems) using a standard protocol, as described previously [22–24]. The reproducibility of aortic diameter measurements were assessed in each vascular laboratory, with inter-observer reproducibility coefficients being less than 4 mm as previously reported [22–25].
Data analysis
The primary aim of this study was to analyse the association between diagnosis of gout and AAA growth using linear mixed effects modeling. All continuous variables of the participant data were reported as mean ± standard deviation (S.D.) and categorical variables were reported as percentage (%). Normal distribution of all continuous variables were tested using Shapiro-Wilk normality test. Skewed data were log transformed prior to testing normality distribution. Statistical differences between paired groups of non-parametric data were determined using Wilcoxon test. Differences in nominal variables between groups were compared using chi square test. Continuous variables showed skewed distribution, therefore, Mann-Whitney U test was performed to test the differences between groups and data were reported as median and inter-quartile range (IQR).
Model Development
An unadjusted random intercept with slope model was used to examine the association between AAA growth and diagnosis of gout using unadjusted and multivariate adjusted models. Multivariate model included risk factors selected based on the bivariate comparisons between participants with or without gout that had a p value of < 0.10. Individual patients were treated as random effects in all models. Follow-up period was treated as random effects in both adjusted and unadjusted random intercept with slope model. The interaction of follow-up period and patient group was used as the test statistic for all linear mixed effects (LME) analyses. Model fit was assessed by visual inspection of the standardised residual distribution and q-q norm plots, suggesting the presence of potentially influential outliers. Sensitivity analyses excluding outliers (defined as data points lying > 4 standard deviations from the mean of model residuals) were performed.
Predictions Using Lme Model
Following the model development, a new dataset with all possible combination of variables included in the model were extracted from the original dataset. The predict function was used to predict the AAA diameter using the unadjusted linear mixed effects model. The predicted AAA diameter was designed as a matrix using the Model.matrix function and diagonals were extracted to derive the 95% confidence intervals (CI). Predictions of mean (95% CI) annual increase in AAA diameter were provided for participants with and without gout.
Propensity Score Matched Cohort
A sub-analysis involving propensity-score matching was performed in an effort to balance participants with and without gout for potential confounding variables [26–30]. Propensity scores were estimated using a multivariate logistic regression model at a ratio of 1 and caliper diameter of 0.25 [31, 32]. The model included initial AAA diameter, body mass index (BMI), prior stroke, and angiotensin converting enzyme inhibitor (ACEI), angiotensin receptor blockers (ARB) and calcium channel blockers (CCB) as covariates. A greedy matching algorithm was used to sequentially match each individual with gout to an individual without diagnosis of gout based on their propensity scores as previously described [31–33]. The balance of covariates was assessed by estimating standardised mean differences after matching the participants [33, 34]. An absolute standardised mean difference of < 0.1 was suggestive of limited covariate imbalance [31–33].
R program version 3.4.4 was used to perform the linear mixed effects modeling analysis and propensity-score matching using the ‘nlme’, ‘ggplot2’, ‘MatchIt’ and ‘survey’ packages [35–37]. Sample size estimate was performed using ‘longpower’ package. Statistical significance was assumed if p was ≤ 0.05.