Proper handling of continuous variables is crucial in healthcare research, for example, within regression modelling for descriptive, explanatory, or predictive purposes. However, inadequate methods are commonly used. This article highlights the importance of appropriately handling continuous variables, and illustrates the consequences of categorisation. This article also explains why assuming a linear relationship between the independent and dependent variable might be inappropriate, and describes how to use splines or fractional polynomials to model non-linear relationships.
Background: The aim of this study was to validate the clinical performance of the Beckman Access high-sensitivity cardiac troponin I (hs-cTnI) assay. Methods: We enrolled patients presenting to the emergency department with symptoms suggestive of acute myocardial infarction (AMI). Final diagnoses were centrally adjudicated by 2 independent cardiologists with all clinical information including cardiac imaging twice: first, using serial hs-cTnT (Elecsys, primary analysis), and second, using hs-cTnI (Architect, secondary analysis) measurements in addition to the clinically used hs-cTn. hs-cTnI Access was measured at presentation and at 1 h. The primary objective was a direct comparison of diagnostic accuracy as quantified by the area under the ROC curve (AUC) of hs-cTnI Access vs the hs-cTnT Elecsys and hs-cTnI Architect assays. Secondary objectives included the derivation and validation of an hs-cTnI Access-specific 0/1-h algorithm. Results: AMI was the adjudicated final diagnosis in 243 of 1579 (15.4%) patients. The AUC at presentation for hs-cTnI Access was 0.95 (95% CI, 0.94–0.96), higher than hs-cTnI Architect [0.92 (95% CI, 0.91–0.94; P < 0.001)] and comparable to hs-cTnT Elecsys [0.94 (95% CI, 0.93–0.95; P = 0.12)]. Applying the derived hs-cTnI Access 0/1-h algorithm (derivation cohort n = 686) to the validation cohort (n = 680), 60% of patients were ruled out [sensitivity, 98.9% (95% CI, 94.3–99.8)], and 15% of patients were ruled in [specificity, 95.9% (95% CI, 94.0–97.2)]. Patients ruled out by the 0/1-h algorithm had a survival rate of 100% at 30 days. Findings were confirmed in the secondary analyses by the adjudication including serial measurements of Architect hs-cTnI. Conclusions: Diagnostic accuracy and clinical utility of the Beckman hs-cTnI Access assay are very high and at least comparable to Roche hs-cTnT and Abbott hs-cTnI assays. ClinicalTrials.gov Identifier: NCT00470587.
BACKGROUND: The clinical utility of procalcitonin in the diagnosis and management of pneumonia remains controversial. METHODS: We assessed the clinical utility of procalcitonin in 2 prospective studies: first, a multicenter diagnostic study in patients presenting to the emergency department with acute dyspnea to directly compare the diagnostic accuracy of procalcitonin with that of interleukin 6 and C-reactive protein (CRP) in the diagnosis of pneumonia; second, a randomized management study of procalcitonin guidance in patients with acute heart failure and suspected pneumonia. Diagnostic accuracy for pneumonia as centrally adjudicated by 2 independent experts was quantified with the area under the ROC curve (AUC). RESULTS: Among 690 patients in the diagnostic study, 178 (25.8%) had an adjudicated final diagnosis of pneumonia. Procalcitonin, interleukin 6, and CRP were significantly higher in patients with pneumonia than in those without. When compared to procalcitonin (AUC = 0.75; 95% CI, 0.71–0.78), interleukin 6 (AUC = 0.80; 95% CI, 0.77–0.83) and CRP (AUC = 0.82; 95% CI, 0.79–0.85) had significantly higher diagnostic accuracy (P = 0.010 and P < 0.001, respectively). The management study was stopped early owing to the unexpectedly low AUC of procalcitonin in the diagnostic study. Among 45 randomized patients, the number of days on antibiotic therapy and the length of hospital stay were similar (both P = 0.39) in patients randomized to the procalcitonin-guided group (n = 25) and usual-care group (n = 20). CONCLUSIONS: In patients presenting with dyspnea, diagnostic accuracy of procalcitonin for pneumonia is only moderate and lower than that of interleukin 6 and CRP. The clinical utility of procalcitonin was lower than expected. SUMMARY: Pneumonia has diverse and often unspecific symptoms. As the role of biomarkers in the diagnosis of pneumonia remains controversial, it is often difficult to distinguish pneumonia from other illnesses causing shortness of breath. The current study prospectively enrolled unselected patients presenting with acute dyspnea and directly compared the diagnostic accuracy of procalcitonin, interleukin 6, and CRP for the diagnosis of pneumonia. In this setting, diagnostic accuracy of procalcitonin for pneumonia was lower as compared to interleukin 6 and CRP. The clinical utility of procalcitonin was lower than expected. ClinicalTrials.gov Identifier: NCT01831115.
BACKGROUND: We aimed to validate the clinical performance of the high-sensitivity cardiac troponin I [VITROS® Immunodiagnostic Products hs Troponin I (hs-cTnI-VITROS)] assay. METHODS: We enrolled patients presenting to the emergency department with symptoms suggestive of acute myocardial infarction (AMI). Final diagnoses were centrally adjudicated by 2 independent cardiologists considering all clinical information, including cardiac imaging: first, using serial hs-cTnT-Elecsys (primary analysis) and, second, using hs-cTnI-Architect (secondary analysis) measurements in addition to the clinically used (hs)-cTn. hs-cTnI-VITROS was measured at presentation and at 1 h in a blinded fashion. The primary objective was direct comparison of diagnostic accuracy as quantified by the area under the ROC curve (AUC) of hs-cTnI-VITROS vs hs-cTnT-Elecsys and hs-cTnI-Architect, and in a subgroup also hs-cTnI-Centaur and hs-cTnI-Access. Secondary objectives included the derivation and validation of an hs-cTnI-VITROS-0/1-h algorithm. RESULTS: AMI was the adjudicated final diagnosis in 158 of 1231 (13%) patients. At presentation, the AUC for hs-cTnI-VITROS was 0.95 (95% CI, 0.93–0.96); for hs-cTnT-Elecsys, 0.94 (95% CI, 0.92–0.95); and for hs-cTnI-Architect, 0.92 (95% CI, 0.90–0.94). AUCs for hs-cTnI-Centaur and hs-cTnI-Access were 0.95 (95% CI, 0.94–0.97). Applying the derived hs-cTnI-VITROS-0/1-h algorithm (derivation cohort n = 519) to the validation cohort (n = 520), 53% of patients were ruled out [sensitivity, 100% (95% CI, 94.1–100)] and 14% of patients were ruled in [specificity, 95.6% (95% CI, 93.4–97.2)]. Patients ruled out by the 0/1-h algorithm had a survival rate of 99.8% at 30 days. Findings were confirmed in the secondary analyses using the adjudication including serial measurements of hs-cTnI-Architect. CONCLUSIONS: The hs-cTnI-VITROS assay has at least comparable diagnostic accuracy with the currently best validated hs-cTnT and hs-cTnI assays. ClinicalTrials.gov Identifier: NCT00470587.
1 October 2019 in International Journal of Cardiology