Herwig Friedl

Negative binomial loglinear mixed models

The Poisson loglinear model is a common choice for explaining variability in counts. However, in many practical circumstances the restriction that the mean and variance are equal is not realistic. Overdispersion with respect to the Poisson distribution can be modeled explicitly by integrating with respect to a mixture distribution, and use of the conjugate gamma mixing distribution leads to a negative binomial loglinear model. This paper extends the negative binomial loglinear model to the case of dependent counts, where dependence among the counts is handled by including linear combinations of random effects in the linear predictor. If we assume that the vector of random effects is multivariate normal, then complex forms of dependence can be modelled by appropriate specification of the covariance structure. Although the likelihood function for the resulting model is not tractable, maximum likelihood estimates (and standard errors) can be found using the NLMIXED procedure in SAS or, in more complicated examples, using a Monte Carlo EM algorithm. An alternate approach is to leave the random effects completely unspecified and attempt to estimate them using nonparametric maximum likelihood. The methodologies are illustrated with several examples.

The assessment of airborne microorganisms in large-scale composting facilities and their immediate surroundings

The number of airborne microorganisms in the area of large-scale composting facilities with different composting techniques (A: open facility using the intensive decomposition process [4000 t/year]. B: closed facility with compost containers [7000 t/year], C: closed facility with table-pile compositing and automatic turning equipment [22 000 t/year]) was investigated using impactor sampling systems (Andersen samplers). All counts carried out inside the closed facilities, especially during the turning process, showed values of >5.0 × 105 CFU/m3 for viable bacteria and moulds with a proportion ofAspergillus fumigatus of up to 64%. Depending on the type of facility, different median values were determined inside the plant area. Counts were highest in the immediate area around the biofilter outside of Facility C (1.7 × 104 CFU/m3 for bacteria and 9.5 × 103 CFU/m3 for moulds). In view of the high load of ambient airborne microorganisms inside the composting facilities, adequate occupational health measures are urgently required. Counts determining the hazard to neighbourhood residents at distances of between 150 and 2000 m showed, depending on the facility, annual median values of 170–330 CFU/m3 for bacteria, 75–340 CFU/m3 for moulds, and 15–52 CFU/m3 forA. fumigatus. Higher individual counts — up to 3 × 103 CFU/m3 for moulds and up to 350 CFU/m3 forA. fumigatus — were found as a result of specific climatic influences, (e.g. winds) and activities as well poor operation. Given the high proportion ofA. fumigatus in the exhaust air, this mould can serve as an indicator for the evaluation of the health risk. However, the maximum values found in the present study, may also be caused by other events in rural areas, (e.g. agricultural activities). With regard to neighbourhood residents, odour complaints are more important than pollution by microorganisms.

Onset and duration of effect of incobotulinumtoxinA, onabotulinumtoxinA, and abobotulinumtoxinA in the treatment of glabellar frown lines: a randomized, double-blind study

Background Three botulinum neurotoxin type A preparations (incobotulinumtoxinA, onabotulinumtoxinA, and abobotulinumtoxinA) are widely approved in Europe and in the US for the treatment of glabellar frown lines. The purpose of this study was to determine and compare the time to onset and duration of treatment effect of incobotulinumtoxinA, onabotulinumtoxinA, and abobotulinumtoxinA for the treatment of glabellar frown lines. Subjects and methods Subjects aged 20–60 years with moderate to severe glabellar frown lines received one treatment of either 21 units (U) incobotulinumtoxinA, 21 U onabotulinumtoxinA, or 63 U abobotulinumtoxinA. Assessments were made over a period of 180 days. Onset of treatment effect was defined as the day that the observer noted a decrease in glabellar muscle activity compared with baseline photographs and videos. Duration of treatment effect was defined as the time until glabellar muscle action returned to the baseline level. Analyses were performed using a Weibull log(T) regression model. Results The study enrolled 180 subjects; 60 per group. For all three products, onset of treatment effect occurred earlier in female subjects compared to male subjects. For both sexes, a significantly earlier time to onset of treatment effect was seen for incobotulinumtoxinA compared to onabotulinumtoxinA and abobotulinumtoxinA; in female subjects these times were 3.02 days, 5.29 days, and 5.32 days, respectively. The duration of treatment effect was longer for incobotulinumtoxinA compared to onabotulinumtoxinA and abobotulinumtoxinA; for all products, treatment effect duration was longer in females than in males. Time to onset was not a predictor of treatment duration. Conclusion IncobotulinumtoxinA demonstrated a more rapid onset and a longer duration of treatment effect than onabotulinumtoxinA (1:1 dose ratio) and abobotulinumtoxinA (1:3 dose ratio). Onset of effect was faster and duration of effect was longer in female subjects compared to male subjects.

A comparison between Andersen (ACFM) and Reuter Centrifugal Sampler (RCS-plus) for indoor sampling of airborne molds

This paper shows first results regarding thecollection efficiency and suitability of theAndersen-6-stage cascade impactor (ACFM,Graseby Andersen) and the Reuter CentrifugalSampler (RCS-plus, Biotest Diagnostics Corp.,Denville, NJ) in measuring concentrations ofindoor airborne viable mold spores. The ACFMshowed significantly higher numbers of colonyforming units than the RCS-plus. For exposureassessment the lower sampling efficiency of theRCS-plus in comparison to the ACFM is not soimportant, as no sampler is able to record theactual concentration of airbornemicroorganisms. For fungal exposure assessmentit is more important to compare indoor andoutdoor mold spore concentration and theindoors and outdoors occuring species. The ACFMdoes not overload as quickly as the RCS-plus,so there is not so much overgrowing andinhibition of colonies and it is easier tomaintain pure cultures. The advantages of theRCS plus are the simpler handling and, the optionof changing flow-rate. Even though each sampler hasits set of advantages and disadvantages, bothcan serve a useful function in airborne fungalspore sampling.

Variance estimates in logistic regression using the bootstrap

Large-sample confidence intervals for the parameter β under the binomial and extra-binomial variance model are presented. Alternative estimates of Var( ) are discussed, which all have a nice bootstrap interpretation in the context of resampling from residuals or score components. The latter approach yields both a model-based and a robust estimate. Some properties of these estimates and their corresponding confidence intervals are also discussed. In an extensive simulation study we compare the coverage probabilities of the intervals supposing binomial variation as well as overdispersion.

Standardizing the complication rate after breast reduction using the Clavien-Dindo classification

Background. Published complication rates for breast reduction surgery, also known as reduction mammaplasty, vary between 4% and 54%. This wide range of complication rates could be attributable to the lack of a standardized classification of complications in plastic surgery. The aim of this study was to analyze our single‐center complication rates after reduction mammaplasty using the Clavien‐Dindo classification. Methods. We performed a retrospective chart review studying 804 patients between the ages of 18 and 81 years old who underwent breast reduction between 2005 and 2015 at our institution. Patients with a history of breast cancer, a previous breast operation, who did not undergo bilateral reduction mammaplasty, or who required systemic immunodeficiency/immunosuppressive drugs were excluded from our analysis. Complications were classified according to the Clavien‐Dindo classification from Grades I to V. Results. A total of 486 patients met the inclusion criteria for the analysis. Patients had an age (mean ± standard deviation) of 39 ± 13 years and a body mass index of 26 ± 4 kg/m2. Median follow‐up was 274 days (interquartile range: 90.5–378). The overall rate of complications of reduction mammaplasty was 63%, with the majority of those being Grades I (48%) and II (9%), comprising 92% of all the complications. Operative revisions were required in 6% (1% Grade IIIA and 5% Grade IIIB). There were no complications graded in categories IV and V. Conclusion. Although complications occurred in more than half of the cases, the majority did not require operative reintervention. The Clavien‐Dindo classification can classify the severity of complications and serve as a benchmark to compare complication rates between different practices. We believe that grading of complications should distinguish between those that do and do not require operative reinterventions.

Modelling over- and underdispersed frequencies of successful ink transmissions onto paper

This work focuses on the statistical modelling of successful or failed ink transmission during the printing of packaging paper. The main aim is to explain the probability of successful ink transmission with the help of a regression model. But many of the applied logistic regression models show that the variabilities of various samples from the printability data set range from much smaller to much larger values than the variability assumed in a binomial model. Hence, the main part of this paper concentrates on the discussion of distribution families that are capable of handling such a wide spectrum of different variations of frequency data.

Complications in tissue expansion: A logistic regression analysis for risk factors

BACKGROUND Tissue expansion is frequently used in reconstructive surgery. Although the surgical procedure is typically considered simple, reported complication rates of tissue expansions exceed 40%. There is little evidence concerning risk factors for complications in tissue expansion in body regions other than breast. The aim was to determine risk factors for complications in non-breast tissue expansion. METHODS 34 patients treated with subcutaneous tissue expanders between 2005 and 2014 were analyzed. Demographic data, body-mass index (BMI), mean arterial blood pressure (MAP), treatment indications, expansion site, previous expansion therapies in the same body region, smoking history, as well as expander characteristics (shape, volume, and filling mechanism) were ascertained. Complications were assessed and ranked according to severity based on the Clavien-Dindo classification. Binary logistic regression analysis adjusted for clinical characteristics was used. A p<0.05 was considered as statistically significant. RESULTS Complications were observed in 26 out of 71 expanders analyzed (36.6%), of whom 10 led to therapy failure. Expanders used in the limbs, female gender, and high expander volume turned out as significant risk factors. Patients with both a high MAP and low BMI developed tissue necrosis significantly more often (p=0.002). The use of tissue expansion after a burn was not associated with an increased risk for complications. CONCLUSIONS This is the first study revealing female gender and low BMI as risk factors in tissue expander surgery. Thus, careful patient selection is mandatory to avoid complications in tissue expansion. Burn patients do not develop complications more often.

Insulin receptor binding to erythrocytes in the first half of pregnancy is increased in healthy pregnant women as compared with non-pregnant or gestational diabetic women

Insulin binding to erythrocytes was measured longitudinally by a competitive radioreceptor assay in 21 healthy pregnant (HP) and 20 well-controlled gestational diabetic women (GD) in 4-week intervals throughout pregnancy and at day 4 post-partum. Maximum insulin binding (maxbdg) at weeks 8-14 was increased (P < 0.001) in HP (median: 6.0%) but not in GD (median: 2.7%) as compared with non-pregnant control subjects (C) (median: 3.6%; previously reported: Clin. Chim. Acta 1992;207:57-71) due to an increased number of high-affinity insulin receptors. Throughout gestation the binding decreased continuously, to reach at term the levels found in C. In GD maxbdg remained close to the level of C throughout pregnancy. Binding differences between HP and GD were independent of the body mass index. Maxbdg did not differ between diet- and insulin-treated patients. It was higher in women whose offspring had low umbilical cord insulin levels (< 10 mu units/ml). The findings suggest that (a) higher insulin binding in HP could contribute to the improved glucose tolerance in early pregnancy and (b) the lack of increase in insulin binding during early pregnancy in gestational diabetes might be one factor leading to the manifestation of the disease in late pregnancy. However, it must be kept in mind that insulin receptors on erythrocytes do not necessarily resemble those on the major target tissues of insulin.

Resampling Methods in Generalized Linear Models Useful in Environmetrics

Generalized linear models are important tools for analysing relationships between binary, count or continuous response variables and predictors with fixed effects. In this paper we present a survey on bootstrap methods based on (extended) quasi-likelihood assumptions. We discuss two approaches: one-step residual resampling and score resampling to estimate the variability of functions in the linear parameters of the model, and an iterative procedure which allows us to define replicates of the dependent variate. With the latter we are able to estimate non-linear parameters in the variance function and to compare non-nested models. The power of these resampling schemes is illustrated by air sampler data concentrating on the number of bacteria colonies observed at outdoor sites in the area of Graz.