Database, confidence

The uncertainty factors used in the development of the RfC represent a specific area of uncertainty inherent in the extrapolation from the available data. The basis for the application of these uncertainty factors are (1) human variability (2) animal to human extrapolation (3) subchronic to chronic extrapolation (4) use of an LOAEL instead of an NOAEL and (5) database confidence. These uncertainty factors can be 1, 3, or 10, depending on the amount of uncertainty. 

Software tools for computational chemistry are often based on empirical information. To use these tools, you need to understand how the technique is implemented and the nature of the database used to parameterize the method. You use this knowledge to determine the most appropriate tools for specific investigations and to define the limits of confidence in results. 

The presence of errors within the underlying database fudher degrades the accuracy and precision of the parameter e.stimate. If the database contains bias, this will translate into bias in the parameter estimates. In the flash example referenced above, including reasonable database uncertainty in the phase equilibria increases me 95 percent confidence interval to 14. As the database uncertainty increases, the uncertainty in the resultant parameter estimate increases as shown by the trend line represented in Fig. 30-24. Failure to account for the database uncertainty results in poor extrapolations to other operating conditions. 

Table 4.1-5 shows in these records, there has been 3 control rod drive failures. Assume 100 plants in the U.S. with an average of 30 control rods/plant and 10.7 years of experience in this database. Estimate, the mode, 90% and 10% confidence limits for the failure rate. 

Two sensibly priced commercial databases for solubility exist [366,507], An article in the journal Analytical Profiles of Drug Substances carries solubility data [496]. Abraham and Le [508] published a list of intrinsic aqueous solubilities of 665 compounds, with many ionizable molecules. It is difficult to tell from published lists what the quality of the data for ionizable molecules is. Sometimes, it is not clear what the listed number stands for. For example, Sw, water solubility, can mean several different things either intrinsic value, or value determined at a particular pH (using buffers), or value measured by saturating distilled water with excess compound. In the most critical applications using ionizable molecules, it may be necessary to scour the original publications in order to be confident of the quality of reported values. 

No reliable, quantitative human data exist. Human data are limited to descriptive effects from accidental exposure and are thus inappropriate for derivation of AEGL values. There is, however, a plethora of acute inhalation data in many experimental species. The database is sufficient to have good confidence in AEGL-2 and AEGL-3 values. 

Figure 6. Statistics for 5 selected substructures of the 500 tested on the EPA IR database. Values of the Reliability, False Positives, and Recall (see text) are compared at the 45% confidence level. The number of compounds in the database containing each substructure is given beneath the substructure name. Note the expanded scale used to plot the False Positive measure.

Average results for 500 IR-active substructures are shown in Figure 7 at four different confidence levels. The average compound in the database contains 8.1 of the 500 substructures. At a confidence level of > 45%, only 1.4 (of 492) incorrect substructures are reported, while 4.6 of 8.1 substructures actually present are reported. In other words, a t3rpical" analysis will report 6.0 substructures at > 45% confidence, of which 4.6 are correct. 3.5 substructures actually present in the compound will fail to be reported. In an actual analysis, infrared data is combined with other t3rpes of data, so that many of the substructures undetected by infrared would be found by other techniques. 

The results of data treatment are documented and evaluated in ES 5 and the interpretation in ES 6 is guided by the analyst s constraints and requirements. For instance, simple visual pattern comparisions may be acceptable for sample identification, or a combined database (GC-FTIR/GC-MS), (PGC/FTIR), (GC/TA), etc., analysis may be required. Judgmental decisions must be trained into the system as to depth of analysis, its acceptability and reliability (e.g., the hit quality index (HQI) of the MS search combined with that from the FTIR search may confirm within a 95% confidence level the GC peak or sample identity). 

In the hazard assessment, it is important to evaluate the toxicological database with regard to its adequacy. The adequacy of a study includes its validity and its relevance. The relevance refers to what has been studied in relation to what is needed for the hazard and risk assessment, and the validity refers to how the study was performed, e.g., conforming with a particular test guideline. The validity and the relevance of a study, or a whole database, has to be considered in relation to the reliability and thus the confidence. The data for hazard assessment are described in detail in Chapter 3. 

It has been suggested to apply an assessment factor for the confidence in the database in case there are limitations in the database, including lack of data for children, which are important in relation to the purpose of the assessment. This section gives an overview of such proposals and evaluations. Then, the key issues are summarized and our recommendations are presented. The question of an extra assessment factor in the hazard and risk assessment for chemicals of concern for children is specifically addressed in Section 5.2.1.13. 

High degree of confidence The database contains high quality human or animal studies, i.e., two or more studies with the same endpoint. The database should be sufficiently extensive to give confidence that the correct critical effect has been selected, and that there are no major uncertainties in this respect. No additional numerical UF required, i.e., the default factor is 1. 

Medium degree of confidence The database falls short of the quality described above in some significant respect, which limits the overall confidence to medium . Assess on a case-by-case basis, perhaps consider to use a low numerical UF, in the range of 1-2. 

The EU TGD (EC 2003) pointed out that the overall confidence in the database needs to be considered in the evaluation of the MOS (can be interpreted as an overall assessment factor). 

---