Institute of Medical Statistics

Köln, Germany

Institute of Medical Statistics

Köln, Germany
Time filter
Source Type

Puhr R.,University of New South Wales | Heinze G.,Medical University of Vienna | Nold M.,Institute of Medical Statistics | Lusa L.,University of Ljubljana | Geroldinger A.,Medical University of Vienna
Statistics in Medicine | Year: 2017

Firth's logistic regression has become a standard approach for the analysis of binary outcomes with small samples. Whereas it reduces the bias in maximum likelihood estimates of coefficients, bias towards one-half is introduced in the predicted probabilities. The stronger the imbalance of the outcome, the more severe is the bias in the predicted probabilities. We propose two simple modifications of Firth's logistic regression resulting in unbiased predicted probabilities. The first corrects the predicted probabilities by a post hoc adjustment of the intercept. The other is based on an alternative formulation of Firth's penalization as an iterative data augmentation procedure. Our suggested modification consists in introducing an indicator variable that distinguishes between original and pseudo-observations in the augmented data. In a comprehensive simulation study, these approaches are compared with other attempts to improve predictions based on Firth's penalization and to other published penalization strategies intended for routine use. For instance, we consider a recently suggested compromise between maximum likelihood and Firth's logistic regression. Simulation results are scrutinized with regard to prediction and effect estimation. We find that both our suggested methods do not only give unbiased predicted probabilities but also improve the accuracy conditional on explanatory variables compared with Firth's penalization. While one method results in effect estimates identical to those of Firth's penalization, the other introduces some bias, but this is compensated by a decrease in the mean squared error. Finally, all methods considered are illustrated and compared for a study on arterial closure devices in minimally invasive cardiac surgery. © 2017 The Authors.

Thadhani R.,Harvard University | Karumanchi S.A.,Institute of Medical Statistics | Karumanchi S.A.,University of Cologne | Wenger J.,Harvard University | And 6 more authors.
Journal of the American Society of Nephrology | Year: 2016

Preeclampsia is a devastating complication of pregnancy. Soluble Fms-like tyrosine kinase-1 (sFlt-1) is an antiangiogenic protein believed to mediate the signs and symptoms of preeclampsia. We conducted an open pilot study to evaluate the safety and potential efficacy of therapeutic apheresis with a plasmaspecific dextran sulfate column to remove circulating sFlt-1 in 11 pregnant women (20-38 years of age) with very preterm preeclampsia (23-32 weeks of gestation, systolic BP 140 mmHg or diastolic BP 90 mmHg, new onset protein/creatinine ratio .0.30 g/g, and sFlt-1/placental growth factor ratio .85). We evaluated the extent of sFlt-1 removal, proteinuria reduction, pregnancy continuation, and neonatal and fetal safety of apheresis after one (n=6), two (n=4), or three (n=1) apheresis treatments.Mean sFlt-1 levels were reduced by 18% (range 7%-28%) with concomitant reductions of 44% in protein/creatinine ratios. Pregnancy continued for 8 days (range 2-11) and 15 days (range 11-21) in women treated once and multiple times, respectively, compared with 3 days (range 0-14) in untreated contemporaneous preeclampsia controls (n=22). Transient maternal BP reduction during apheresis wasmanaged by withholDing pre-apheresis antihypertensive therapy, saline prehydration, andreducingbloodflowthrough the apheresis column.Compared with infants born prematurely to untreated women with and without preeclampsia (n=22 per group), no adverse effects of apheresis were observed. In conclusion, therapeutic apheresis reduced circulating sFlt-1 and proteinuria in women with very preterm preeclampsia and appeared to prolong pregnancy without major adverse maternal or fetal consequences. A controlled trial is warranted to confirm these findings.

Kiffmeyer T.K.,Institute of Energy and Environmental Technology | Tuerk J.,Institute of Energy and Environmental Technology | Hahn M.,Institute of Medical Statistics | Stuetzer H.,Institute of Medical Statistics | And 3 more authors.
Annals of Occupational Hygiene | Year: 2013

A large-scale study was carried out in order to determine the contamination level of antineoplastic drugs in pharmacies and to investigate the suitability and effects of wipe sample monitoring at regular intervals. A specific study design was developed. The 130 participating pharmacies were divided into a study and a control group, carrying out five and two wipe sampling cycles, respectively. The work practice was analyzed using questionnaires to identify factors that influence the contamination level. From 1269 wipe samples, 774 (61%) were contaminated with at least one of the analyzed cytotoxic drugs: cyclophosphamide, docetaxel, etoposide, 5-fluorouracil, gemcitabine, ifosfamide, methotrexate, and paclitaxel. A significant decrease of the contamination with cyclophosphamide and 5-fluorouracil was observed in the study group. The Monitoring-Effect Study of Wipe Sampling in Pharmacies method has proven to be a reliable and affordable tool for contamination control. Based on the 90th percentile of the contamination values, a substance-independent performance-based guidance value of 0.1ng cm-2 has been derived. © 2012 The Author.

Hornig-Do H.-T.,University of Cologne | Tatsuta T.,University of Cologne | Buckermann A.,University of Cologne | Bust M.,University of Cologne | And 6 more authors.
EMBO Journal | Year: 2012

Respiratory chain (RC) complexes are organized into supercomplexes forming 'respirasomes'. The mechanism underlying the interdependence of individual complexes is still unclear. Here, we show in human patient cells that the presence of a truncated COX1 subunit leads to destabilization of complex IV (CIV) and other RC complexes. Surprisingly, the truncated COX1 protein is integrated into subcomplexes, the holocomplex and even into supercomplexes, which however are all unstable. Depletion of the m-AAA protease AFG3L2 increases stability of the truncated COX1 and other mitochondrially encoded proteins, whereas overexpression of wild-type AFG3L2 decreases their stability. Both full-length and truncated COX1 proteins physically interact with AFG3L2. Expression of a dominant negative AFG3L2 variant also promotes stabilization of CIV proteins as well as the assembled complex and rescues the severe phenotype in heteroplasmic cells. Our data indicate that the mechanism underlying pathogenesis in these patients is the rapid clearance of unstable respiratory complexes by quality control pathways, rather than their impaired assembly. © 2012 European Molecular Biology Organization | All Rights Reserved.

Uhlig U.,John Radcliffe Hospital | Uhlig U.,Martin Luther University of Halle Wittenberg | Kostev K.,Institute of Medical Statistics | Schuster V.,University of Leipzig | And 3 more authors.
Pediatric Infectious Disease Journal | Year: 2014

Background: Although rotavirus (RV) vaccination was licensed in 2006, it was not included into the officially recommended German childhood vaccination schedule until 2013. Local differences in health policies in the past led to large differences in vaccination coverage rate among the federal states of Germany. This enables an ecologic study of RV vaccine effectiveness. Methods: We performed a population-based retrospective analysis of RV vaccination use, RV notification and hospitalization among 0 to 5-year-old children in Germany during 2006 to 2011/2012. We compared effectiveness of the 2 RV vaccines, Rotateq and Rotarix, in an ambulatory setting and analyzed potential side effects. Results: We observed a significant reduction in RV notifications since introduction of RV vaccination. In areas attaining vaccine coverage of 64%, RV-related hospital admissions of 0 and 1-year-old children decreased by 60% compared with 19% reduction in the low vaccination coverage area. Decrease in RV-related hospitalizations of 0 and 1-year-old children was specific and significantly associated with vaccination coverage of the individual federal state (P < 0.0001, r = -0.68). There was no overall increase in intussusception rate or Kawasaki disease-related hospital admissions since introduction of RV vaccination. The 2 licensed RV vaccines had similar effectiveness in the ambulatory setting. Conclusions: Postmarketing data suggest that RV vaccination is efficient in reducing RV-related hospitalizations. There is no apparent difference in effectiveness for Rotarix and Rotateq. Copyright © 2014 by Lippincott Williams & Wilkins.

Milde T.,Institute of Medical Statistics | Schwab K.,Institute of Medical Statistics | Walther M.,Institute of Medical Statistics | Eiselt M.,Institute of Medical Statistics | And 3 more authors.
Physiological Measurement | Year: 2011

Time-variant partial directed coherence (tvPDC) is used for the first time in a multivariate analysis of heart rate variability (HRV), respiratory movements (RMs) and (systolic) arterial blood pressure. It is shown that respiration-related HRV components which also occur at other frequencies besides the RM frequency (= respiratory sinus arrhythmia, RSA) can be identified. These additional components are known to be an effect of the 'half-the-mean-heart- rate-dilemma' ('cardiac aliasing' CA). These CA components may contaminate the entire frequency range of HRV and can lead to misinterpretation of the RSA analysis. TvPDC analysis of simulated and clinical data (full-term neonates and sedated patients) reveals these contamination effects and, in addition, the respiration-related CA components can be separated from the RSA component and the Traube-Hering-Mayer wave. It can be concluded that tvPDC can be beneficially applied to avoid misinterpretations in HRV analyses as well as to quantify partial correlative interaction properties between RM and RSA. © 2011 Institute of Physics and Engineering in Medicine.

Fitzek S.,DRK Kliniken Berlin Kopenick | Leistritz L.,Institute of Medical Statistics | Witte O.W.,Friedrich - Schiller University of Jena | Heuschmann P.U.,Charité - Medical University of Berlin | Fitzek C.,Asklepios Kliniken Brandenburg
Cerebrovascular Diseases | Year: 2011

Background and Purpose: We aimed to validate the usefulness of the Essen Stroke Risk Score (ESRS) to predict stroke recurrence in a hospital-based follow-up study. Methods: We followed up 730 consecutive patients admitted to a neurological stroke unit in Berlin, Germany, with ischemic stroke (IS) or transient ischemic attack (TIA). The mean follow-up time was 13.4 months (SD 5.9). We calculated the time of event-free survival by Kaplan-Meier estimates. Results: The cumulative rate of a recurrent stroke was 10.4% (76 of 730). The risk of a recurrent stroke was significantly higher in patients with an ESRS >2 (p = 0.01; log rank). The area under the curve (AUC) was 0.59. The risk of a vascular event (VE), defined as a fatal or nonfatal IS or myocardial infarction, was significantly higher in patients with an ESRS >2 (p = 0.003; log rank). The AUC was 0.58. Stratifying the data at exact 1-year follow-up, 9 of 185 (10.3%) patients with an ESRS ≤2 suffered a recurrent stroke, while 56 of 311 (18%) patients with an ESRS >2 did so (p < 0.02; χ2). Thirty-six of 196 (18.4%) patients with an ESRS ≤2 suffered a VE, while 156 of 353 (35.7%) patients with an ESRS >2 did so (p < 0.001; χ2). Conclusions: The ESRS identifies groups of patients with TIA or IS who are at significantly increased risk for a recurrent stroke and cardiovascular event. Patients with a high ESRS require short-term follow-up and reevaluation of secondary prevention strategies. Copyright © 2011 S. Karger AG.

Krug K.B.,University of Cologne | Ulhaas A.,University of Cologne | Hellmich M.,Institute of Medical Statistics | Schwabe H.,University of Cologne | And 5 more authors.
Investigative Radiology | Year: 2013

PURPOSE: Preoperative magnetic resonance (MR)-guided wire localizations are warranted in patients with suspicious focal breast lesions on MR mammographic findings without equivalent in x-ray mammography and ultrasonography. The study was performed to assess the impact of clinical parameters, tumor size, and target localization on the procedural characteristics in magnetic resonance imaging (MRI)-guided wire localizations of breast lesions using an open 1.0-T open MR system. MATERIAL AND METHODS: The clinical, radiological, and histological characteristics of all 347 patients and all 394 interventional procedures performed in a 6-year interval were extracted from the clinical files. Two board-certified senior radiologists evaluated the impact of target localization and the size on the interventional results in the available 302 image data sets. Patient characteristics, lesion characteristics, and interventional results were statistically correlated in subgroup analyses. RESULTS: A total of 387 of the 394 MR-guided wire localizations (98.2%) were technically successful. In 7 cases (2.3%), the intervention was aborted because the suspicious finding of the diagnostic MR mammography could not be visualized during the intervention. Minor complications occurred in 13 interventions (3.3%). The histological workup of the operative specimen showed benign results in 226 of the 394 interventions (57.4%) and malignant findings in 154 wire localizations (39.1%). The mean (SD) length of the interventional procedure time defined as the time interval between the start of the first and of the last MRI sequence as documented in the electronic MRI data sets was 24.6 (8.4) minutes. Patient age, medical history, and the anticipated risk for developing breast cancer and a simultaneous known carcinoma did not affect the technical success and complication rates and the interventional procedure time. A total of 60 targets (19.5%) were located in the retromamillary zone, 89 targets (28.9%) in the peripheral zone, and 1 target (0.3%) near the chest wall. The maximum diameter was 1 to 5 mm in 64 lesions (21.2%), 6 to 10 mm in 136 lesions (45.0%), 11 to 15 mm in 56 lesions (18.6%), and 16 mm or greater in 46 lesions (15.2%). A total of 23 of the 100 histologically proven invasive carcinomas had a maximum MRI diameter of 1 to 5 mm (23.0%) and 38 (38.0%) of 6 to 10 mm. CONCLUSIONS: Magnetic resonance-guided wire localizations of suspicious breast lesions using an open high-field MR system are a clinically safe and feasible method even in small target lesions and anatomical regions that are usually considered difficult to access. Copyright © 2013 Lippincott Williams & Wilkins.

Wolff S.,Evangelisches Krankenhaus Kalk | Terheggen G.,Evangelisches Krankenhaus Duesseldorf | Mueller R.,Dr. Falk Pharma GmbH | Greinwald R.,Dr. Falk Pharma GmbH | And 2 more authors.
Inflammatory Bowel Diseases | Year: 2013

Background: The consistency of endoscopic and histologic findings in patients with ulcerative colitis (UC) has not been elucidated. Choice of assessment may affect study outcomes. Methods: Post hoc analyses were performed using data from 2 randomized, controlled multicenter trials: (1) SAG-26, mesalazine granules for induction of remission in UC (n = 380), and (2) SAG-27, mesalazine granules for maintenance of UC remission (n = 647). Assessments included Clinical Activity Index, Endoscopic Index, modified Disease Activity Index, and Histology Index. Results: In SAG-26, 52 of 380 patients (13.7%) with clinically (Clinical Activity Index >4) and endoscopically (Endoscopic Index ≥4) active UC showed no histological signs of active inflammation (Histology Index ≤1) at baseline. Among endoscopically and histologically active patients, 246 of 327 (75.2%) reached clinical remission versus 48 of 52 patients (92.3%) with active endoscopy but no inflammation on histology (difference, 17.1%; P = 0.006). The difference in the proportion of patients achieving clinical remission according to endoscopy and histology in clinically inactive (Clinical Activity Index ≤4) patients was 30.8% in SAG-26 (at the study end) and 28.1% in SAG-27 (at baseline). In SAG-27, clinical relapse occurred in 21.2% of patients with endoscopic and histologic remission at baseline and 27.1% of patients with some histological inflammation at baseline (P = 0.111). Using the modified Disease Activity Index ≤1 (mucosal healing) instead of the Endoscopic Index score, the difference was similar (21.2% versus 28.0%, P = 0.073). Conclusions: Endoscopic and histologic assessments differ in both active and inactive UC. Overdiagnosis of inflammation using endoscopy versus histology can significantly affect outcomes, at least in studies using induction of clinical remission as an endpoint. The assessment criteria for trials in UC should be reconsidered. Copyright © 2013 Crohn's & Colitis Foundation of America, Inc.

Er F.,University of Cologne | Nia A.M.,University of Cologne | Dopp H.,University of Cologne | Hellmich M.,Institute of Medical Statistics | And 10 more authors.
Circulation | Year: 2012

Background-Contrast medium-induced acute kidney injury is associated with substantial morbidity and mortality. The underlying mechanism has been attributed in part to ischemic kidney injury. The aim of this randomized, double-blind, sham-controlled trial was to assess the impact of remote ischemic preconditioning on contrast medium-induced acute kidney injury. Methods and Results-Patients with impaired renal function (serum creatinine >1.4 mg/dL or estimated glomerular filtration rate <60 mL • min • 1.73 m) undergoing elective coronary angiography were randomized in a 1:1 ratio to standard care with (n=50) or without ischemic preconditioning (n=50; intermittent arm ischemia through 4 cycles of 5-minute inflation and 5-minute deflation of a blood pressure cuff). Overall, both study groups were at high risk of developing contrast medium-induced acute kidney injury according to the Mehran risk score. The primary end point was the incidence of contrast medium-induced kidney injury, defined as an increase in serum creatinine ≥25% or ≥0.5 mg/dL above baseline at 48 hours after contrast medium exposure. Contrast medium-induced acute kidney injury occurred in 26 patients (26%), 20 (40%) in the control group and 6 (12%) in the remote ischemic preconditioning group (odds ratio, 0.21; 95% confidence interval, 0.07-0.57; P=0.002). No major adverse events were related to remote ischemic preconditioning. Conclusions-Remote ischemic preconditioning before contrast medium use prevents contrast medium-induced acute kidney injury in high-risk patients. Our findings merit a larger trial to establish the effect of remote ischemic preconditioning on clinical outcomes. Clinical Trial Registration-URL: Unique identifier: U1111-1118-8098. © 2012 American Heart Association, Inc.

Loading Institute of Medical Statistics collaborators
Loading Institute of Medical Statistics collaborators