Kathrin Hohl

Boehringer Ingelheim Veterinary Research Center Gmbh & Co. Kg, Hanover, Lower Saxony, Germany

Are you Kathrin Hohl?

Claim your profile

Publications (18)33.26 Total impact

  • [Show abstract] [Hide abstract]
    ABSTRACT: Objective: The objectives of the studies reported here were to determine the relative bioavailability of linagliptin and metformin when administered in a fixeddose combination (FDC) tablet with and without food, and to investigate the relative bioavailability of linagliptin and metformin FDC tablets from two treatment batches with different dissolution behavior. Methods These studies were open-label, singledose, randomized, two-way crossover trials. After an overnight fast, healthy volunteers received an FDC tablet once (with/without food in the food-effect study; or from one of two batches with differing dissolution behavior in the tablet-dissolution study). On a separate visit, following a washout period of 35 days, participants received the alternative treatment. In the food-effect study the primary endpoints were maximum measured concentration in plasma (Cmax) for linagliptin and metformin, area under the plasma concentration-time curve from 0 to 72 hours (AUC0-72) for linagliptin and from 0 to infinity (AUC0-∞) for metformin. In the tablet-dissolution study the primary endpoints were Cmax for both analytes, AUC0-72 for linagliptin, and from 0 to the time of the last quantifiable data point (AUC0-t) for metformin. Results: The administration of the FDC tablet with food had no influence on the relative bioavailability of linagliptin and metformin with regard to the extent of exposure as determined by AUC0-72 (linagliptin) and AUC0-∞ (metformin) compared with FDC tablet administration while fasting. After food intake, peak plasma concentrations of linagliptin were slightly lowered (from 4.99 to 4.56 nmol L-1), but the 90% confidence interval (CI) of the geometric mean test/reference ratio was still located within the generally applied bioequivalence acceptance limits of 80 - 125%. The median time from dosing to the maximum concentration of linagliptin in plasma (tmax) was similar under both conditions. Administration with food reduced the rate of absorption of metformin indicated by a prolongation in median tmax (from 2 to 4 hours) and a decrease in Cmax by ~ 18%. There were no notable differences between the two treatment groups with respect to safety and tolerability. In the tablet-dissolution study, bioequivalence was demonstrated between linagliptin/metformin FDC tablets with normal and slower dissolution characteristics. For both linagliptin and metformin, the 90% CI of all pharmacokinetic (PK) parameters were well within the bioequivalence acceptance limits of 80 - 125%. Tablets from both batches were well tolerated with no unexpected adverse events. Conclusions: Food did not have a relevant impact on the bioavailability of linagliptin from the FDC tablet. The effect of food on the metformin component was comparable to that previously demonstrated. Furthermore, differences in tablet-dissolution characteristics did not have an impact on the bioavailability of linagliptin or metformin from the FDC tablet.
    International journal of clinical pharmacology and therapeutics 04/2014; 52(7). DOI:10.5414/CP201961 · 1.04 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Objectives: Empagliflozin is an orally available, potent and highly selective inhibitor of the sodium glucose cotransporter 2 (SGLT2). This study was undertaken to investigate the effect of food on the pharmacokinetics of 25 mg empagliflozin and to assess dose proportionality between 10 mg and 25 mg empagliflozin under fasted conditions. Materials and methods: In this open-label, 3-way, cross-over study, 18 healthy volunteers received 3 single doses of empagliflozin in a randomized sequence (25 mg empagliflozin under fasted conditions, 25 mg empagliflozin after a high-fat, high-calorie breakfast and 10 mg empagliflozin under fasted conditions), each separated by a washout period of at least 7 days. Serial plasma samples were collected at selected time points over a period of 72 hours. Results: Administration with food had no clinically relevant effect on the area under the plasma concentration-time curve (AUC0-∞) of empagliflozin (geometric mean ratio (GMR): 84.04, 90% confidence interval (CI): 80.86 - 87.34). The decrease observed in the maximum plasma concentrations (Cmax) of empagliflozin (GMR: 63.22, 90% CI: 56.74 - 70.44) when administered with food was not considered clinically meaningful. The increases in AUC0-∞ and Cmax for 10 mg vs. 25 mg empagliflozin administered under fasting conditions were roughly dose-proportional, as demonstrated by the slope β of the regression lines being slightly less than 1 (slope β for AUC0-∞: 0.94, 95% CI: 0.90 - 0.97; slope β for Cmax: 0.91, 95% CI: 0.80 - 1.01). Empagliflozin was well tolerated under fed and fasting conditions. Conclusions: The results support administration of empagliflozin tablets independently of food. Increases in empagliflozin exposure under fasting conditions were roughly dose-proportional between 10 mg and 25 mg empagliflozin.
    International journal of clinical pharmacology and therapeutics 11/2013; 51(11):873-9. DOI:10.5414/CP201948 · 1.04 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: To investigate training effects of two different resistance and proprioceptive exercising concepts of neck muscles. Twenty-six healthy women participated in a randomized pilot trial. The test persons were randomized to two different neck-training programs (resistance training (RT) and proprioceptive resistance training (PRT)). They performed a standardized training program for the duration of ten weeks two times weekly. The neck strength, the cross-sectional area of three neck muscle groups (1. sternocleidomastoid muscles; 2. multifidus and semispinalis cervicis muscles; 3. semispinalis capitis and splenius muscles) and the proprioceptive capability evaluated by the dynamic joint repositioning error (DJRE) of the head were assessed pre- and post-intervention. Strength gain did not differ significantly between the two resistance training groups (PRT group: 8.2% to 29.3%; RT group: 1.4% to 19.8%). Change of hypertrophy of all neck muscle groups was significantly (p< 0.001 to p=0.013) greater in the PRT group (18.9% to 32.3%) than in the RT group (1.5% to 12.9%). The DJRE deteriorated with 35% in the RT group and did not change in PRT group (-2.0%). In combination with resistance training, proprioceptive training led to a significantly higher muscle hypertrophy and didn't effect a significant deterioration of the proprioceptive capability compared to isolated resistance training.
    Journal of Back and Musculoskeletal Rehabilitation 01/2013; 26(2):189-97. DOI:10.3233/BMR-130368 · 1.04 Impact Factor
  • Source
    Benjamin Mayer, Rainer Muche, Kathrin Hohl
    [Show abstract] [Hide abstract]
    ABSTRACT: In empirical, data driven research missing values often arise in the course of a data analysis. This fact constitutes a problem for different reasons, so e.g. standard methods for analyzing data lead to biased estimates and a loss of statistical power due to missing values, since those methods require complete data sets and therefore omit incomplete cases for the analyses. Furthermore missing values imply a certain loss of information, for that reason the validity of results of a study with missing values has to be rated less than in a case where all data had been available. For years there are methods present for replacement of missing values (Rubin, Schafer) to tackle these problems and solve them in parts. Hence in this article we want to present an overview of existing software to handle and replace missing values on the one hand and give an outline about strategies to get information about software solutions on the other hand. The methodological aspects of the imputation strategies therefore are delineated just briefly in this article. Citation: Mayer B, Muche R, Hohl K (2012) Software for the Handling and Imputation of Missing Data – An Overview. J Clinic Trials 1:103. doi:10.4172/ jctr.1000103 Copyright: © 2012 Mayer B, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The aim of the study was to investigate the need for pressure change in patients with sleep-disordered breathing (SDB) several weeks after therapy initiation. We prospectively studied 905 consecutive patients (740 men and 165 women) with SDB and therapeutic intervention with continuous positive airway pressure (CPAP)/bilevel PAP. Several weeks after therapy initiation, patients were restudied for control, and pressure was optimized if it was necessary. The differences in CPAP pressure from initial treatment and control night were assessed. Anthropometric data, polysomnography data, Epworth sleepiness scale, and Berlin questionnaire scores were correlated to pressure differences from the first and control titration nights. Pressure change was needed in 511 patients (58.2%). Pressure increase was more frequent than pressure reduction (41.7% vs. 11.7%). Mean pressure increase in CPAP was 1.3 mbar, and mean decrease, 1.6 mbar. In the bilevel PAP group, the mean increase in inspiratory pressure was 1.2 mbar, and in expiratory pressure, 0.8 mbar; the mean decrease in inspiratory pressure was 1.9 mbar, and in expiratory pressure, 1.4 mbar. No correlation was found between anthropometric data, sleep efficacy, the amount of rapid eye movement sleep per night, or questionnaire scores and pressure change. Our results show that pressure changes are necessary in the majority of patients several weeks after therapy initiation. Therefore, re-evaluation of therapy pressure is useful.
    Sleep And Breathing 03/2010; 15(1):107-12. DOI:10.1007/s11325-010-0332-9 · 2.87 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: To evaluate [(11)C]choline positron emission tomography/computed tomography ([(11)C]choline PET/CT) for the detection of a biochemical recurrence of prostate cancer after radical prostatectomy. Retrospective analysis of [(11)C]choline PET/CT performed in 41 consecutive prostate cancer patients with a rising PSA. The mean time to biochemical relapse was 24 months. PSA levels were determined at time of examination, and patients received either a targeted biopsy or surgery. Histopathology reports served as reference for the evaluation of the [(11)C]choline PET/CT findings. Mean PSA in [(11)C]choline PET/CT positive patients was 3.1 ng/ml (median 2.2 ng/ml, range 0.5-11.6 ng/ml) and 0.86 ng/ml in [(11)C]choline PET/CT negative patients (median 0.83 ng/ml, range 0.41-1.40 ng/ml). Six of 12 patients with PSA < 1.5 ng/ml [(11)C]choline PET/CT revealed a pathological uptake. Histopathology was positive in 6/12 patients in this group. At PSA levels ranging from 1.5 to 2.5 ng/ml all [(11)C]choline PET/CT were positive (n = 16), a positive histology was found in 12/16 patients (75%) and at PSA 2.5-5 ng/ml [(11)C]choline PET/CT was positive in 8/8 patients, confirmed by histology in 7/8 patients. Finally, at PSA higher than 5 ng/ml [(11)C]choline PET/CT identified 5/5 patients positive all confirmed by histology. The sensitivity of [(11)C]choline PET/CT for the detection of recurrence at PSA < 2.5 ng/ml was 89% with a positive predictive value of 72%. [(11)C]choline PET/CT is useful for re-staging of prostate cancer in patients with rising PSA even at levels below 1.5 ng/ml. Our study confirms results from other published studies on [(11)C]choline PET/CT in prostate cancer relapse.
    World Journal of Urology 03/2009; 27(5):619-25. DOI:10.1007/s00345-009-0371-7 · 3.42 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: In this study, we investigated the influence of sublingual nitroglycerine (NTG) on the peripheral diameter, intraluminal contrast agent density, and image quality of coronary arteries during computed tomography coronary angiography (CTCA). Thirty patients with sublingual NTG application were matched to 30 patients without NTG. The diameters of the left anterior descending coronary artery (LAD), the left circumflex coronary artery and the right coronary artery were measured at 1-, 4-, and 8-cm length of each vessel as well as the intraluminal contrast agent density along the LAD. Vessel diameters and contrast attenuation at 4 and 8 cm were referenced against the values at 1 cm and processed as percentage reduction. Image quality of the posterior descending artery was assessed subjectively by 2 independent observers. The percentage of peripheral vessel diameter reduction and the peripheral attenuation of contrast agent density for all measured coronary arteries was significantly smaller in the group with NTG administration. The image quality of the posterior descending artery was significantly higher in the group with NTG. Sublingual administration of NTG before CTCA results in improved diagnostic image quality because of a significant dilatation and improved intraluminal contrast agent density of the peripheral vessels.
    Journal of computer assisted tomography 01/2009; 33(2):199-203. DOI:10.1097/RCT.0b013e31817c6b33 · 1.60 Impact Factor
  • Source
    Benjamin Mayer, Rainer Muche, Kathrin Hohl
    [Show abstract] [Hide abstract]
    ABSTRACT: In medical research missing values often arise in the course of a data analysis. This fact constitutes a problem for different reasons, so e.g. standard methods for analyzing data lead to biased estimates and a loss of statistical power due to missing values, since those methods require complete data sets and therefore omit incomplete cases for the analyses. Furthermore missing values imply a certain loss of information for what reason the validity of results of a study with missing values has to be rated less than in a case where all data had been available. For years there are methods for replacement of missing values (Rubin, Schafer) to tackle these problems and solve them in parts. Hence in this article we want to present the existing software to handle and replace missing values on the one hand and give an outline about the available options to get information on the other hand. The methodological aspects of the replacement strategies are delineated just briefly in this article.
  • [Show abstract] [Hide abstract]
    ABSTRACT: The aim of this study was to investigate the cervicocephalic kinaesthesia of healthy subjects for gender and age effects and its reliability in a new virtual reality test procedure. 57 healthy subjects (30 male, 27 females; 18-64 years) were immersed into a virtual 3D scene via a headmounted display, which generated specific head movements. The joint repositioning error was determined in a static and dynamic test at the times T0, T1 (T0+10 minutes) and T2 (T0+24 hours). The intrasession reliability (T0-T1) and the intersession reliability (T0-T2) were analysed. In both tests no gender- or age-specific effects were found. In the overall group the means of the static test were 6.2 degrees -6.9 degrees and of the dynamic test were 4.5 degrees -4.9 degrees . The intratest difference in the static test was -0.16 degrees and the intertest difference was 0.47 degrees . The intratest difference in the dynamic test was 0.42 degrees and the intertest difference was 0.37 degrees . The static and dynamic test was reproducible in healthy subjects, with minor deviations, irrespective of gender and age. The smaller interindividual differences in the dynamic test could be beneficial in the comparison of healthy individuals and individuals with cervical spine disorders.
    Journal of electromyography and kinesiology: official journal of the International Society of Electrophysiological Kinesiology 09/2008; 19(5):e353-61. DOI:10.1016/j.jelekin.2008.05.005 · 1.73 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The incidence of acute pancreatitis varies from 5 to 80 per 100,000 inhabitants throughout the world. The most common cause of death in these patients is infection of pancreatic necrosis by enteric bacteria, spurring the discussion of whether or not prophylactic antibiotic administration could be a beneficial approach. We therefore analyzed randomized clinical trials, which form the basis of guidelines and recommendations on this topic. One trial demonstrated that antibiotic prophylaxis reduces mortality, but the statistical design of this trial was questionable. Another important trial, showing an effect of antibiotic prophylaxis on the incidence of pancreatic sepsis, used the wrong statistical test to analyze their data. An analysis with the correct test could not confirm this effect. Three randomized clinical trials demonstrated that antibiotic prophylaxis in severe acute pancreatitis could reduce the incidence of extrapancreatic infections. Two trials showed a significant reduction of the overall infection rate; while in one of them peripancreatic and extrapancreatic infections alone were not significantly different. Two double blinded studies could not demonstrate a significant effect of antibiotic prophylaxis on pancreatic/peripancreatic infection, extrapancreatic infection or mortality. Our analysis shows that some of the reported significant effects of prophylactic antibiotic treatment are either questionable or less clinically relevant. With regards to reduction in mortality and the incidence of infected pancreatic necrosis, no convincing evidence exists which supports the routine administration of prophylactic antibiotics in severe acute pancreatitis.
    Hepato-gastroenterology 01/2008; 55(88):2233-7. · 0.91 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: In many circumstances of data fitting one has to choose the optimal fitting function or model among several alternatives. Criteria or tests on which this decision is based are necessary and have to be well selected. In this preliminary analysis the application of the corrected Akaike information criterion is demonstrated considering the example of determining pharmacokinetic parameters for the blood serum time activity curves of 111In-labeled anti-CD66 antibody. Another model selection criterion, the F-test, is used for comparison. For the investigated data the corrected Akaike information criterion has proved to be an effective and efficient approach, applicable to nested and non-nested models.
    Medical Physics 12/2007; 34(11):4285-92. DOI:10.1118/1.2794176 · 3.01 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: To evaluate [(11)C]-choline positron-emission tomography (PET)/computed tomography (CT) for detecting clinical recurrence after primary treatment for prostate cancer. In all, 50 patients with prostate cancer who had had initial therapy (radical prostatectomy in 40, external beam radiation in three and interstitial brachytherapy in seven) had PET/CT using [(11)C]-choline in the presence of an increased or increasing prostate-specific antigen (PSA) level. The mean (range) time to biochemical progression was 22 (2-136) months. Current PSA levels were determined in all patients at the time of examination. The results were correlated with the histopathology reports after targeted biopsy or surgery, and with the clinical follow-up. The mean (median, range) PSA level in patients with positive PET/CT was 3.62 (2.42, 0.5-13.1) ng/mL, and that in patients with a negative scan was 0.90 (0.95, 0.41-1.40) ng/mL. PET/CT was positive in seven of 13 patients with a PSA level of <1.5 ng/mL, and histology was positive in this group in nine. In 17 patients with PSA levels of 1.5-2.5 ng/mL PET/CT was positive in all and the histology was positive in 13; in 11 men with a PSA level of 2.5-5 ng/mL PET/CT was positive in all 11 and the histology was positive in 10; in nine men with PSA levels of >5 ng/mL PET/CT identified all as positive and the histology was positive in eight. The sensitivity at a PSA level of <2.5 ng/mL of PET/CT for detecting recurrence was 91% (95% confidence interval, 71-99%) with a specificity of 50% (16-84)%. [(11)C]-choline PET/CT seems to be useful for re-staging prostate cancer after curative therapy and with increasing PSA levels; this was verified by histological examination. We recommend this method at PSA levels of <2.5 ng/mL.
    BJU International 10/2007; 100(4):786-93. DOI:10.1111/j.1464-410X.2007.07083.x · 3.13 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: To evaluate and compare the role of (11)C-choline positron emission tomography (PET) and transrectal ultrasonography (TRUS) in the preoperative staging of clinically localized prostate cancer. Fifty-five consecutive patients with biopsy-confirmed prostate cancer had TRUS and (11)C-choline PET as a part of their clinical staging programme before radical retropubic prostatectomy (RP). The PET images were prospectively interpreted by a consensus decision of two nuclear medicine physicians and one radiologist with special expertise in the field. The TRUS was done by one experienced urologist. The criteria evaluated prospectively in each patient were extracapsular extension (ECE), seminal vesicle invasion (SVI) and bladder neck invasion (BNI). The results were compared with the histopathological findings after RP. At pathology, 32 patients were classified pT2, 16 as pT3a and three had pT3b lesions. In four patients the histopathological examination showed pT4 with BNI. The overall accuracy of PET in defining local tumour stage (pT2 and pT3a-4) was 70%; the overall accuracy by TRUS was 26%. PET was more sensitive than TRUS for detecting ECE (pT3a) and SVI (pT3b) in advanced stages, and in pT4 stages. The sensitivity and positive predictive value (PPV) (95% confidence interval) in stages pT3a-pT4 for PET were 36 (17-59)% and 73 (39-89)%. The sensitivity and PPV in stages pT3a-pT4 for TRUS were 14 (3-35)% and 100 (29-100)%. (11)C-choline PET and TRUS tended to understage prostate cancer. This series shows the current limited value of TRUS and PET for making treatment decisions in patients with clinically localized prostate cancer, especially if a nerve-sparing RP is considered. Treatment decisions should not be based on TRUS and (11)C-choline PET findings alone. In future studies, the combination of metabolic and anatomical information of PET and endorectal magnetic resonance imaging should be evaluated, as this might optimize the preoperative staging in prostate cancer.
    BJU International 07/2007; 99(6):1421-6. DOI:10.1111/j.1464-410X.2007.06776.x · 3.13 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In this prospective study, reliability of integrated (18)F-FDG PET/CT for staging of NSCLC was evaluated and compared to MDCT or PET alone. 240 patients (pts) with suspected NSCLC were examined using PET/CT. Of those patients 112 underwent surgery comprising 80 patients with NSCLC (T1 n = 26, T2 n = 37, T3 n = 11, T4 n = 6). Imaging modalities were evaluated independently. MDCT, PET and PET/CT diagnosed the correct T-stage in 40/80 pts (50%; CI: 0.39-0.61), 40/80 pts (50%; CI: 0.39-0.61) and 51/80 pts (64%; CI: 0.52-0.74), respectively, whereas equivocal T-stage was found in 15/80 pts (19%; CI: 0.11-0.19), 12/80 pts (15%; CI: 0.08-0.25) and 4/80 pts (5%; CI: 0.01-0.12), respectively. With PET/CT, T-stage was more frequently correct compared to MDCT (p = 0.003) or PET (p = 0.019). Pooling stages T1/T2, T-stage was correctly diagnosed with MDCT, PET and PET/CT in 54/80 pts (68%; CI: 0.56-0.78), 56/80 pts (70%; CI: 0.59-0.80) and 65/80 pts (81%; CI: 0.71-0.89). T3 stage was most difficult to diagnose. T3 tumors were correctly diagnosed with MDCT in 2/11 pts (18%; CI: 0.02-0.52) versus 0/11 pts (0%; CI: 0.00-0.28) with PET and 5/11 pts (45%; CI: 0.17-0.77) with PET/CT. In all imaging modalities, there were no equivocal findings for T4 tumors. Of these, MDCT found the correct tumor stage in 4/6 pts (67%; CI: 0.22-0.95), PET in 3/6 pts (50%; CI: 0.12-0.88) and PET/CT in 5/6 pts (83%; CI: 0.36-0.99). Integrated PET/CT was significantly more accurate for T-staging of NSCLC compared to MDCT or PET alone. The advantages of PET/CT are especially pronounced combining T1- and T2-stage as well as in advanced tumors.
    Nuklearmedizin 02/2007; 46(1):9-14; quiz N1-2. · 1.67 Impact Factor
  • Source
    Sleep Medicine 02/2007; 8(1). DOI:10.1016/S1389-9457(07)70237-X · 3.10 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The outcome of hematopoietic cell transplantation for hematologic malignancies may be improved by delivering targeted radiation to hematopoietic organs while relatively sparing nontarget organs. We evaluated the biodistribution of 111In-labeled anti-CD45 antibody in humans using the rat IgG2a monoclonal antibody YAML568 that recognizes a common CD45 epitope present on all human leukocytes. Eight patients undergoing bone marrow transplantation received YAML568 labeled with 122 +/- 16 MBq of 111In intravenously followed by serial blood sampling, urine collection, and conjugated view planar gamma-camera imaging up to 144 h after injection. Time-activity curves were obtained using region-of-interest analysis in the accumulating organs and residence times were calculated. An estimate for the radiation-absorbed doses for each organ per unit of administered activity of 90Y was calculated using software for internal dose assessment. The first patient received no unlabeled antibody preloading. The second 2 patients received a preloading dose of 10 mg (0.15 mg/kg). The last 5 patients received a preloading dose of 30-47 mg (0.5 mg/kg). No significant administration-related side effects were seen. The 3 patients receiving no antibody or low antibody preloading had an unfavorable biodistribution with a high initial accumulation of activity in the liver (37%) and the spleen (34%). For the patients receiving 0.5-mg/kg antibody preloading, the estimated radiation-absorbed doses for red bone marrow, spleen, liver, kidney, and total body were 6.4 +/- 1.2, 19 +/- 5, 3.9 +/- 1.4, 1.1 +/- 0.4, and 0.6 +/- 0.1 mGy/MBq, respectively, demonstrating preferential red marrow targeting. A linear regression model showed that the amount of unlabeled antibody preloading per body weight has a strong influence on the estimated red marrow absorbed dose (P = 0.003, R2 = 0.80). This study shows that the anti-CD45 monoclonal antibody YAML568 is suitable for delivering selectively radiation to hematopoietic tissues when labeled with 90Y provided that a preloading dose of about 0.5 mg/kg unlabeled antibody is given.
    Journal of Nuclear Medicine 09/2006; 47(8):1335-41. · 5.56 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Zusammenfassung Ein generelles Problem bei der Auswertung von Datensätzen ist das Vorkommen von fehlenden Werten. Besonders bei multivariaten Regressionsmodellen kann es bei einer hohen Anzahl an fehlenden Werten zu einer drastischen Fallzahlreduktion kommen, da solche Modelle auf der Analyse einer vollständigen Datenmatrix beruhen. Etliche Erset-zungsmethoden sind in den letzten Jahren in der Literatur vorgeschlagen worden. Um eine geeignete Methode zu bestimmen, ist es notwendig, den Datensatz zuerst rein de-skriptiv auf fehlende Werte zu untersuchen und Kenntnisse über die Struktur fehlender Werte zu gewinnen. Mögliche Ersetzungsstrategien sind die Single oder Multiple Im-putation. Für die komplexe Auswertung von Datensätzen mit fehlenden Werten wurden von uns zwei SAS-Makros für eine detaillierte Deskription bzw. für die Ersetzung feh-lender Werte entwickelt. Die Makros vereinfachen den Umgang mit fehlenden Werten erheblich und werden im Folgenden kurz vorgestellt.
  • Source
    Kathrin Hohl
    [Show abstract] [Hide abstract]
    ABSTRACT: Zusammenfassung Fehlende Werte in Datensätzen können zu Fehlinterpretationen der Daten führen. Deshalb muss der Umgang mit fehlenden Werten wohlüberlegt gewählt werden. Eine Möglichkeit ist die Ersetzung dieser Werte durch plausible geschätzte Werte. Es stehen hierfür dem Anwender in SAS in der Prozedur PROC MI mehrere Methoden zur Verfügung. Die Eig-nung der Methoden hängt von mehreren Faktoren ab, unter anderem vom Merkmalstyp der Variablen mit fehlenden Werten. Seit SAS Version 9.1 sind die beiden Ersetzungsmetho-den logistische Regression und Discriminant Function Method speziell für kategoriale Va-riablen implementiert. In einer Simulationsstudie wurde untersucht, welche Konsequenzen es hat, wenn fehlende Werte kategorialer Variablen durch die beiden neuen Ersetzungs-methoden oder durch Ersetzungsmethoden primär für stetige Variablen ersetzt werden. Die Simulationsergebnisse zeigen, dass sich die Ersetzungsmethoden nur geringfügig im Hin-blick auf die Übereinstimmung der ersetzten mit den Originalwerten und der Validität der Ergebnisse einer statistischen Auswertungsmethode unterscheiden.

Publication Stats

278 Citations
33.26 Total Impact Points

Institutions

  • 2014
    • Boehringer Ingelheim Veterinary Research Center Gmbh & Co. Kg
      Hanover, Lower Saxony, Germany
  • 2007–2013
    • Universität Ulm
      • • Clinic of Trauma, Hand, Plastic and Reconstructive Surgery
      • • Clinic of Nuclear Medicine
      Ulm, Baden-Württemberg, Germany