Helle Holst

Do clinical databases render population-based cancer registers obsolete? The example of breast cancer in Denmark.

AbstractObjective: Clinical databases have been invented to monitor treatment outcomes, therapies or diseases, often in great detail. The traditional population-based cancer registry has been invented to collect a minimum of information about all incident cancers. Do clinical databases render population-based cancer registers obsolete as sources of cancer cases for epidemiological study? Methods: We compared the study base of first incident breast cancer cases in Denmark in 1978–1994 known from the national cancer register and from the national clinical database on breast cancer patients. The clinical database is used for monitoring protocoled treatment. Results: Combining the two data sources we found 48,522 first primary breast cancers in Denmark 1978–1994. Of these, 37,640 were included in both data sources, 2151 were included only in the clinical database, and 8731 were included only in the cancer register. A major part of the difference between the two data sources was due to treatment-focused data collection in the clinical database, and a minor part due to differences in the registration of second primaries, date of diagnosis and invasiveness. Conclusions: Cancer incidence data are sensitive to registration procedures and definitions. Clinical cancer databases cannot generally replace the traditional cancer register as a reliable data source for incident cancer cases in a national population.

Age–period–cohort modelling of breast cancer incidence in the Nordic countries

The Nordic countries have experienced a steady increase in breast cancer incidence throughout the past 35 years. We analysed the incidence in Denmark, Finland, Norway and Sweden during the period 1958 to 1992 using age-period-cohort models and taking the systematic mammography screening into account. Assuming the age dependency of the incidence pattern in old age to be common for the Nordic countries, an internal comparison could be made among the four countries of the cohort effects and the period effects. The study indicated that the period effects have been of importance for the increase in breast cancer incidence seen in the Nordic countries. The widespread practice of neglecting the period effects in age-period-cohort analysis of time trends in breast cancer incidence therefore probably needs reconsideration. A key finding was that Danish women born in the 20th century seem to have been exposed to an increasing load of cohort borne breast cancer risk factors not experienced to the same extent by Norwegian women, whereas they were seemingly subjected to the same period effects.

Continuous Ecotoxicological Data Evaluated Relative to a Control Response

This article describes a generalized nonlinear regression method that models continuous ecotoxicological response data evaluated relative to a control response. The major improvement and difference from the methods described in the literature is consideration of the covariance between data, which is induced when modeling responses relative to a control response. The capability to estimate ECps (the concentration that corresponds to p % effect), such as ECIo and EC50, and their confidence limits is evaluated through a simulation study. The method is illustrated with real data from growth inhibition tests with freshwater microalgae. Recommendations are given on the choice of regression function, on how to handle initial stimulation and on test designs.

Comparison of Different Calibration Methods Suited for Calibration Problems with Many Variables

This paper describes and compares different kinds of statistical methods proposed in the literature as suited for solving calibration problems with many variables. These are: principal component regression, partial least-squares, and ridge regression. The statistical techniques themselves do not provide robust results in the spirit of calibration equations which can last for long periods. A way of obtaining this property is by smoothing and differentiating the data. These techniques are considered, and it is shown how they fit into the treated description.

Modelling the growth of methane-oxidizing bacteria in a fixed biofilm

Abstract Methane-oxidizing bacteria were grown in a fixed biofilm reactor in order to study their ability to degrade chlorinated aliphatic hydrocarbons. Focus is on the growth behaviour of the mixed culture. The growth is described by a model that includes methanotrophic bacteria in the active biomass fraction. The inactive biomass fraction consists of exocellular polymers and biodegradable and inert particulate biomass. The model describes the oxygen respiration in detail. Yield coefficients, decay constants and hydrolysis constants are estimated based on the oxygen respiration. An analysis of the observability of the system reveals that several of the coefficients cannot be determined explicitly due to the complexity of the model and the limited amount of variables measured. Estimation procedures based on least squares methods are employed and parameter estimates and confidence intervals are computed. The study forms the basis for setting up a future measuring programme. This work indicates a yield coefficient for methanotrophs of 0.36 mg biomass COD per mg CH 4 . In weight units this corresponds to approx. 1.44 mg biomass per mg CH 4 . This is close to the theoretical maximum growth yield for methanotrophs.

Examination of reproducibility in microbiological degradation experiments.

Experimental data indicate that certain microbiological degradation experiments have a limited reproducibility. Nine identical batch experiments were carried out on 3 different days to examine reproducibility. A pure culture, isolated from soil, grew with toluene as the only carbon and energy source. Toluene was degraded under aerobic conditions at a constant temperature of 28 °C. The experiments were modelled by a Monod model – extended to meet the air/liquid system, and the parameter values were estimated using a statistical nonlinear estimation procedure. Model reduction analysis resulted in a simpler model without the biomass decay term. In order to test for model reduction and reproducibility of parameter estimates, a likelihood ratio test was employed. The limited reproducibility for these experiments implied that all 9 batch experiments could not be described by the same set of parameter values. However, experiments carried out the same day (within the same run) were more uniform than experiments carried out on different days (between runs), and a common set of parameter estimates could be accepted for experiments within runs, but not for experiments from different runs. The limited reproducibility may be caused by variability in the preculture, or more precisely, variations in the physiological state of the bacteria in the precultures just before used as inoculum.

Nonlinear parameter estimation in microbiological degradation systems and statistic test for common estimation

Abstract Three identical microbiological experiments were carried out and analysed in order to examine the variability of the parameter estimates. The microbiological system consisted of a substrate (toluene) and a biomass (pure culture) mixed together in an aquifer medium. The degradation of the substrate and the growth of the biomass are described by the Monod model consisting of two nonlinear coupled first-order differential equations. The objective of this study was to estimate the kinetic parameters in the Monod model and to test whether the parameters from the three identical experiments have the same values. Estimation of the parameters was obtained using an iterative maximum likelihood method and the test used was an approximative likelihood ratio test. The test showed that the three sets of parameters were identical only on a 4% α level.

The Influence of Design Characteristics on Statistical Inference in Nonlinear Estimation: A Simulation Study Based on Survival Data and Hazard Modeling

This paper describes the influence of design characteristics on the statistical inference for an ecotoxicological hazard-based model using simulated survival data. The design characteristics of interest are the number and spacing of observations (counts) in time, the number and spacing of exposure concentrations (within cmin and cmax), and the initial number of individuals at time 0 in each concentration. A comparison of the coverage probabilities for confidence limits arising from the profile-likelihood approach and the Wald-based approach is carried out. The Wald-based approach is very sensitive to the choice of design characteristics, whereas the profile-likelihood approach is more robust and unbiased. Special attention is paid to estimating a parametric no-effect concentration in realistic small-sample situations since this is the most interesting parameter from an environmental protection point of view.