Benchmark Dose analysis
The Benchmark dose (BMD) approach is a scientifically more advanced method compared to the No Observed Adverse Effect Level (NOAEL) approach for deriving a Reference Point (Point of Departure) for risk assessment (EFSA, 2017). The BMD method pre-defines a specific effect size, referred to as the Benchmark response (BMR) and estimates the dose (BMD) associated with the specified effect. The BMD is estimated from the complete dose‑response dataset by fitting dose-response models. Statistical uncertainties in the data are taken into account in the confidence interval around the BMD, the lower limit of which (denoted as BMDL) is the Point of Departure (or Reference Point) for deriving exposure limits.
Calculation of withdrawal period for residuals in milk
The RIVMNational Institute for Public Health and the Environment web application of PROAST (see below) allows for the calculation of the withdrawal period of cow’s milk needed for residue levels to have reached a sufficiently low level to make it save for consumption. For more information, see the manual under the help button in the RIVM web application.
Key elements of PROAST
PROAST can be applied to:
- dose-response data from all sorts of in vivo studies (animal, human or eco-toxicological)
- concentration-response data from in vitro studies
- high throughput data (e.g. gene expression as a function of dose)
- dose-response data from mixture studies
- combined dose-response datasets for similar endpoints in a single analysis
- non-linear regression in any other field of science, including the option to compare the relationships among subgroups
- the complete data from a toxicological study, by performing a quick (automated) analysis of a whole series of endpoints in a single run.
- decreasing residual concentrations in milk (or other animal products/tissues) for which repeated measurements in individuals are available.
An important feature of PROAST is that it allows for comparing dose-responses among various subgroups, e.g. among sexes, among study durations, or among replicate studies. Based on statistical analysis, PROAST indicates if the various dose-response relationships differ among the subgroups, and if so, in what sense (e.g. in background response, in sensitivity to the chemical, or in dose-response shape).
Main changes with previous version
The current EFSA guidance on applying the BMD approach (EFSA, 2017) differs in various ways from the previous EFSA guidance (2009). These changes have been implemented in PROAST versions from 62.0 onwards. Two major changes resulting from the new EFSA guidance are:
- in continuous data, only models 3 and 5 from the exponential and Hill families of models are considered, and from these two models the one with lowest AIC is used for calculating the BMD confidence interval.
- The BMD confidence interval can be calculated by “Model Averaging’, which is the preferred approach (EFSA 2017) as it takes model uncertainty into account.
PROAST versus BMDS
US-EPA developed the Benchmark Dose Software (BMDS). This package is also suitable for dose-response analysis and deriving a BMDL from dose-response data. RIVM and EPA aim to achieve consistency between the BMDS and PROAST software, but there are still some differences, including a number of default settings for statistical assumptions. Further, the two software packages differ in functionalities (see EFSA, 2017, Appendix A, for a summary of differences between BMDS and PROAST). Examples of useful functionalities in PROAST are the possibility of statistically comparing dose-response relationships among subgroups (covariate analysis), and the larger flexibility in plotting. PROAST closely follows the EFSA guidance on the BMD approach, while BMDS does not allow for performing BMD analyses according to EFSA guidance in all respects.
PROAST as a web application
There are two web applications of PROAST available, either of which may be particularly useful if you want to quickly apply PROAST, avoiding the installation of software on your computer. These web applications are easy to apply, but do not include all functionalities of the R package of PROAST (see below). However, the usual dose-response analyses of toxicity data can be done in these web tools. The URLs are:
- https://efsa.openanalytics.eu/ for the EFSA web application
- https://proastweb.rivm.nl/ for the RIVM web application
For the first URL you need to create an account, the second URL can be used without an account. A manual for the first application can be found on the website itself (see “about” in the upper right corner). For the second web application, see PROAST MANUAL WebApp.pdf. Further, it may be helpful to have a look at the PROAST manual PROAST MANUAL GUI version.pdf. Even though this manual relates to PROAST as an R package, it provides some example analyses which can be mimicked in the PROAST web application. Further, it gives some more background information on the methods used in PROAST.
Either web application will probably satisfy infrequent users of PROAST. However, if you use PROAST on a more regular basis, the stand alone R version of PROAST may be more convenient, making it worthwhile to put some effort in getting acquainted with working in R. Furthermore, not all options in PROAST have been implemented in the current web applications.
The PROAST web application (www.proastweb.rivm.nl) includes a special functionality: the calculation of withdrawal periods based on monitored concentrations in individual animals over time (e.g. in cow’s milk). See Manual web application MILKINGS.
PROAST as an R package
PROAST can be run on your own machine as a package within R, which is a general statistical software environment, freely downloadable (see the PROAST manuals). You need some minimal knowledge of how R works, which is discussed in the manuals as well. Note that R studio does not support all the R functions used by PROAST, so you need to run PROAST directly in R.
To install the PROAST package (copyright RIVM) on your computer, download the R package proast70.0.zip (using save as; do not unzip!) to your computer. MAC users should download and save proast70.0.tgz.zip which needs to be unzipped. Save the package at a convenient location in you folder system. Consult the PROAST manuals to see how PROAST can be installed in R.
First users may go to PROAST quick start.pdf. If you want to use the R-GUI version of PROAST, go to PROAST MANUAL GUI version.pdf. The most comprehensive version of PROAST is the R-menu version, and the associated manual is PROAST MANUAL menu version.pdf.
As PROAST is an R package, it is relatively easy to make a new version available after fixing any bugs in the software code. After any change in the code, the PROAST package receives another version number, whether it relates to a small change (e.g., fixed bug) or a relatively large change (e.g., change in methodological approach due to updated guidance). In this way, outcomes from analyses in the past that differ from those in the present may be checked to be due to a another PROAST version being used, and if so, if this relates to a bug, a different default setting, or a change in methods used. In general, however, the outcomes from different PROAST versions will be similar (although more significant differences may occur before and after version 62.0, due to changes in EFSA guidance).
It is recommended to save all PROAST versions that you ever used (i.e. the zip or tgz file), preferably in a special folder on your computer, so that you can always reproduce calculations that were done in the past. Also, you will need the zip (or tgz) file when you install a new R version. If you find that outcomes differ when using a newer version for the same analysis, and are unable find the cause, you may contact email@example.com.
Updates or questions
If you are interested to be informed about updates of PROAST or other relevant information, send an e-mail to firstname.lastname@example.org. You can also use this email address to ask questions, or report bugs or weaknesses in the software. We will try to answer questions on the use of PROAST to the extent possible.
The folder PROAST exercises contains various exercises, both for the web application and for the GUI version in R, if you want to further train you skills in dose-response modelling. Answers to the exercises can be found in answers.doc.
A concise discussion of the BMD approach can be found in EFSA (2017). See Slob and Setzer (2014; Crit Rev Toxicol 44, 270-297) as an example of the use of PROAST in a more advanced setting, including dose-response analysis with covariates.