Please activate JavaScript!
Please install Adobe Flash Player, click here for download
ePaper created 2014-10-06, 13:17:01 | version 1.33.00
[8, 23]), and to be able to assess any differences between devices during the more stable phase associated with better INR control after the first 3 months [23]. Main endpoints in the statistical analysis are TTR values. TTR values were computed by the Rosendaal method [24], using the predefined therapeutic range of INR values. In addition, the percentage of values inside the therapeutic range, and the frequency of critical values (INR \1.5 or INR [5.0) were included as endpoints. Further, critical values were analyzed as number of events observed per month. The occurrence of minor clinical events was self-reported by patients. Serious clinical events and deaths were captured and kept on record by the NTS, and mortality data are included in the analysis. TTR values and the number of critical values per month were treated as interval scaled variables in subsequent analyses. Location differences between subgroups were tested using Wilcoxon’s ranksum test with adjustment for tied values. In case of more than two subgroups, this test was replaced by the Kruskal–Wallis ranksum test, also adjusted for ties. Least squares multiple regression was used to explain the influence of the treatment (i.e. the device in use), demographic data and additional medical variables (indication, therapeutic range, medication) on TTR, with dichotomous and polytomous variables being transformed into groups of dummy indicator variables. For each group of indicator variables, the largest subgroup was used as reference group. Cox proportional hazard regression was used to assess the multivariate influence on the time to fatal events. The regression took account of repeated events per patient applying the Conditional Risk Set modelling technique proposed by Prentice et al. [25]. As standard, the computation of standard errors used the robust clustering technique proposed by Lin and Wei [26]. The simple effect of the device on the time between fatal events was determined graphically using the Kaplan–Meier curve and an associated logrank test. Statistical analysis as conducted in STATA/SE v13.0 by StataCorp LP, College Station, USA (2013). This article does not contain any new studies with human or animal subjects performed by any of the authors. RESULTS Of the 5,108 patients enrolled in the NTS program from June 2011–February 2014, 2,210 used CoaguChek XS and 2,898 used INRatio2 as an INR home monitoring device. Out of these, 4,326 patients (85%) had INR measurements recorded in the database after the 90-day familiarization period and were included in the further analysis (1,961 CoaguChek XS and 2,365 INRatio2). For these patients a total of 217,369 valid INR measurements were available for analysis (104,295 CoaguChek XS and 113,074 INRatio2), of which 156,507 INR values (72%) were obtained after the 90-day familiarization period (79,418 CoaguChek XS and 77,089 INRatio2, reflecting the longer average observation period for patients using CoaguChek) and these were included in the further analysis. Patient Characteristics Table 1 shows a summary of patient characteristics. Males made up 67.6% of the patients, the majority of the patients were between 45 and 74 years of age, and the leading indication for anticoagulation was atrial fibrillation (54.3%) (Table 1). The predominant target range for the INR was Adv Ther