Skip to: Content | Footer | Accessibility

Air Toxicology and Epidemiology

OEHHA Responses to Public Comments...Cont.

Start | Back | Next

National Particleboard and UF Resin Manufacturers Associations
Up | Down
The main points of the commentators are shown below in detail.

Comment: The formaldehyde Level I REL should be revised. OEHHA has used a Benchmark Dose approach to compute a 0.14 ppm Level I REL for formaldehyde with eye irritation as the toxic endpoint. The Kulle study from which the BC was derived exposed 19 subjects to various formaldehyde concentrations up to 3.0 ppm in a controlled environmental chamber for up to three hours with a number of endpoints examined.² There was no eye irritation at 0 and 0.5 ppm concentrations, but symptoms were noted to increase with dose at 1.0, 2.0 and 3.0 ppm.³ The benchmark dose was derived using a log-probit analysis and a definition of BC as "the 95% lower confidence limit of the concentration expected to produce a response rate of 1%. The resulting BC of 0.25 ppm⁴ was modified to 0.08 ppm through use of an uncertainty factor of 10 for intraspecies variability and a modifying factor of 0.3. That value was further modified to a one-hour exposure, resulting in a final Level I REL of 0.14 ppm.

The Kulle study data should be reinterpreted. The Associations support the use of the Kulle study; it was carefully conceived and carried out and has been widely cited in the literature. However, we believe that the Kulle data has been inappropriately characterized in the TSD due perhaps to inconsistent definitions of "mild irritation’ in the OEHHA document and in the study. Responses of the subjects in Kulle to the effect that formaldehyde was perceived to be present but not annoying have been interpreted as an adverse effect by OEHHA and lumped together with the responses of annoying irritation effects in the derivation of the benchmark dose.

In the Technical Support Document, Level I is defined as follows:

This is the discomfort or mild effect level, and refers to the concentration of an airborne substance (a gas, vapor, aerosol or aerosolized particle) at or below which exposure for one hour may result in some odors, tastes and visual cues but which will cause no adverse health effects in nearly all of the population. Exposures to concentrations above this level may result in minor health effects, such as mild sensory irritation of short duration. It is the exposures above the level I REL which give rise to regulatory responsibilities under the Hot Spots program.⁵ Thus, the mere detection of a substance -- the perception of formaldehyde without an unpleasant sensation -- should not trigger a Level I designation under this standard. The "mild effect level" requires more than minor, non-adverse effects.

Response: OEHHA considers mild irritation to be an adverse effect. Therefore, the mild irritation reported in the Kulle et al. (1987) study was included with moderate irritation for the purpose of the benchmark dose calculation.

Comment: The Associations believe that there may be a typographical error in Figure 2 on page 11 of the TSD where the following description of Level I appears:

The discomfort or mild effect level. The level at or below which no adverse effects are expected. Exposure at or below this level may be perceived by mild irritation of the eyes, nose or throat, or by unpleasant odors, tastes or sight... We believe that the underscored word should be "above".

Response: The comment is correct. This is a typographical error, and has been corrected.

Comment: The definition of "mild" response in Kulle is much different:

The symptoms were scored by each subject as: none, mild (present, but not annoying), moderate (annoying), or severe (debilitating). For statistical analysis a score of 1 was assigned to none, 2 to mild, 3 to moderate, and 4 to severe.⁶

The "present but not annoying" description is directly comparable to OEHHA’s description of "odors, tastes, and visual cues" which do not amount to an adverse health effect covered by Level I.

Response: The question is whether mild irritation that is "present, but not annoying" is adverse. While this is a matter of interpretation and semantics, the authors of the study clearly treat mild irritation as a distinct effect. OEHHA considers mild irritation to be an adverse health effect distinct from odors, tasks, and visual cues. Furthermore, it is clear from page 922 of the study that the mild irritation experienced by subjects included eye irritation in addition to odor detection: "...in our study at 0.5 ppm HCHO, none of our nine subjects had eye irritation, while at 1.0 ppm HCHO three of 19 reported mild eye irritation and one experienced moderate irritation." The mere detection of formaldehyde odor occurred in 4 of 9 subjects at 0.5 ppm, but this was not included in the benchmark analysis since detection of odor is not considered to be an adverse effect and no irritation was observed at that concentration.

Comment: The proper classification of the Kulle responses is critical to the development of an appropriate benchmark dose for formaldehyde. The following is a comparison of the eye irritation data points used in the model fitting and the actual base data from Kulle:

		Kulle Data
Dose	BD Data	Present, but not annoying	Moderate
0.0	0/19	0/19	0/19
0.5	0/9	0/9	0/9
1.0	4/19 (21%)	3/19	1/19 (5%)
2.0	10/19 (53%)	6/19	4/19 (21%)
3.0	9/9 (100%)	5/9	4/9 (44%)

We ask that OEHHA rerun the model with new data points reflecting only those responses from the Kulle study coded number three to indicate annoyance.

Response: As stated above, OEHHA believes it is inappropriate to disregard mild irritation from the Kulle et al. (1987) study in the benchmark analysis. A benchmark analysis using OEHHA methodology (shown below), using the BD₀₅ benchmark reveals that the results of the two analyses are not critically dependent on the interpretation of the mild effects in the study. The Chi-Square analysis indicates an acceptable fit for either data set, and the distance between the MLE and LCL is considerably smaller using the combined data set. The comment’s recommendation would introduce greater uncertainty into the dose-response analysis.

Data set MLE (ppm) 95% LCL (ppm) REL (ppm) Chi-Sq.

Mild + Moderate 0.71 0.43 0.25 2.81

Moderate only 1.05 0.45 0.26 0.11

Comment: The characterization of the Kulle results and the definition of Level I effects underscores another important consideration in the BD approach. The severity of toxic endpoints should be relevant in deciding the degree of safety factors that are appropriate.⁷ It is surely more appropriate to use conservative safety margins in the various component factors of a reference exposure limit when effects such as cancer, reproductive toxicity, or other pathological changes are involved, than when endpoints such as transitory, reversible irritation are involved. This is particularly true if the "irritation" is of a minor nature. As noted in the report of the recent EPA Risk Assessment Forum on The Use of Benchmark Dose Approach in Health Risk Assessment,

The term noncancer effect is nonspecific and encompasses a wide variety of responses, including adverse effects on specific organs or organ systems, reproductive capacity, viability and structure of developing offspring in utero, and survival...Modeling this diversity of response represents a major challenge.⁸

Response: Based on an analysis of LOAEL to NOAEL data (Fowles et al., 1997), we found that for certain endpoints uncertainty factors less than 10 were justified. These endpoints are mild sensory irritation and lethality. In keeping with our analysis, we have used an uncertainty factor of 3 when extrapolating from a LOAEL to a NOAEL for mild irritation or for lethality. An uncertainty factor of 3 is also used in the Benchmark Concentration approach for interspecies uncertainty if the key study is in animals and for intraspecies variability if the key study is in humans.

Comment: The issue of how to account for differences in the severity of toxicological effects is beyond the scope of these comments. It has been explored in the literature⁹ and at a recent conference sponsored by the International Life Sciences Institute’s Risk Science Institute.¹⁰ We note, however, that toxicological severity is not taken into account in OEHHA’s BD approach as an uncertainty factor, modifying factor, or otherwise. The relatively benign effects of low level formaldehyde exposure should be considered by OEHHA in its evaluation of the other inputs in the BD calculation which may be excessively conservative.

Response: The uncertainty factor is designed to account for interspecies uncertainty and intraspecies variability in response to a toxicant. Severity of endpoint is considered by OEHHA in the analysis by designating an effect as mild, severe, or life-threatening. Additionally, as described above, OEHHA has defined specific circumstances in which uncertainty factors less that 10 may be used (see previous response). Severity of effect is taken into consideration for some of these circumstances but is not necessarily related to the degree of uncertainty in extrapolation. While a reduced uncertainty factor of 3 is utilized when deriving a level using the benchmark concentration approach, there are no data to justify additional reductions in the total uncertainty factor based on effect severity.

Comment: The Benchmark Dose is defined as the 95% lower confidence limit of the concentration expected to produce one response for every 100 subjects exposed at this dose. The BD method assumes a log-probit concentration versus response relationship to identify the concentration expected to produce 1% increase in toxic response (TC01) via a maximum likelihood estimate. The incremental increase in risk over background is often termed the Benchmark Response ("BMR"). OEHHA itself has noted, "[f]or a toxic response with a specific threshold, 1 percent approaches the margin of useful extrapolation for acute noncarcinogenic data due to the limited number of animals used in most experiments." (TSD p. 24). Study size is even more problematic under the 1% BMR in human studies because of typically limited participation. The Associations urge OEHHA to use a 5% or 10% BMR in its benchmark dose calculation. Although the EPA Report suggests the use of BMRs between 1% and 10%, it raised a number of cautions about this selection:

...the BMR should be selected near the low end of the range of increased risks that can be detected in a bioassay of typical size. Comparison of the BMD with the NOAEL for a large number of developmental toxicity data sets indicated a BMR in the range of 5 to 10 percent resulted in a BMD that was on average similar to the NOAEL (Allen et al., 1994a; Faustman et al., 1994).

One cannot detect increased risks in the Kulle study doses at the 1% level or anywhere near it. The LOAEL is at 1.0 ppm; the NOAEL is at 0.5 ppm and the indicated BD is 0.25 ppm. The EPA Report also noted the importance of model independence and the fact that various dose response models can fit experimental data well, but "produce widely divergent estimates of risk at doses far below the range that produce measurable increases in response (Crump 1985).’’¹¹ EPA cautions that for the BD approach to be relatively model independent, the BMR cannot be much smaller than increased responses that can be measured reliably in experimental groups of typical size.". Clearly, the 19 subjects in the Kulle study are far fewer than the 100 or more subjects that are mentioned in the EPA Report as a typical minimum for a 1% BMR. A Workshop convened by the International Life Sciences Institute (the "ILSI Workshop’’)¹² took a more decisive stand on this issue. The Workshop participants "...generally agreed that selecting an effective dose (ED) for BMD calculation at the ED01 (01=1% above the experimental background rate or control) was too low...It was acknowledged that use of either an ED₀₅ or an ED₁₀ for BMD calculation was appropriate, recognizing that future research might show the advisability of selecting one value over the other (Faustman et al 1994).¹³

Response: OEHHA agrees that for the sake of consistency with the consensus reached at the above workshops, the BC₀₁ should be changed to BC₀₅. This change will be made in the technical support document for the benchmark calculations, including formaldehyde. However, OEHHA notes that the justification for the BC₀₅ in the USEPA report is based on comparison with NOAELs and uses only developmental toxicity data for such comparisons. Future empirical analyses of acute data sets other than developmental toxicity may necessitate reevaluating the use of the BC₀₅ benchmark.

We wish to emphasize that the analysis presented by Crump and cited in the benchmark dose workshop is limited only to developmental endpoints. An analysis of the effects of changing the benchmark dose calculations in the technical support document from the 1% to the 5% response rate indicates that endpoints other than developmental toxicity (CNS, irritation, and lethality) all yield very similar values for either the 1% or 5% response rates. The magnitude of difference is on average only a factor of about 1.3.

Comment: OEHHA justifies its use of the 1% measure by the fact that it is consistent with its intent to protect "essentially all" individuals from the acute toxic endpoints. There are several concerns with this argument. First, as noted above, mathematically stretching down to a lower BMR impairs the scientific integrity of the result. The issue is not whether a data set can be manipulated to reach the level where the incremental 1% over background is covered, but rather whether there is scientific support for that approach. Moreover, the interest in providing protection to the broad array of the population is addressed in at least three other ways in the REL calculations. A study conducted last year by Allen et al.^l4 reviewed the relation between benchmark doses and NOAELs for over 400 developmental toxicity experiments. The Benchmark doses --which had been determined using the Weibull model, a 1% BMR and 95% lower confidence limits -- were, on average, 30-fold lower than the comparable NOAELs.¹⁵ We presume that similar results would be obtained if the data were analyzed with the log-probit model used by OEHHA. This suggests that the change in methodology, in and of itself, is introducing a 30 fold reduction in the indicated levels. While changes of this magnitude for individual substances may be justified based on improved information, we know of no scientific justification for an average reduction of this size. The combination of conservative assumptions including the 1% BMR are driving the changes. The Associations urge OEHHA to recalculate the formaldehyde Level I REL (and RELs for other chemicals, where appropriate) based either on a 5% or 10% BMR.

Response: As discussed in the response to the Chemical Manufacturers’ Association, OEHHA is changing the benchmark dose methodology from a 1% response to 5%, in accordance with the opinions of the Benchmark Dose Workshop. However, there are several issues raised in the comment in support of this change that should be addressed. First, the "stretching down" of the benchmark response to a 1% response rate is much less influential on the REL than the comment implies, and is only a factor of about 1.3 below an REL based on the 5% response, as discussed above. The data sets analyzed by Allen (1994) contained only developmental defects and may not adequately represent the dose-response slopes or experimental designs of many other acute toxicological responses. In addition, it is inappropriate to judge the benchmark dose solely on its proximity to the NOAEL, since it is clear that the benchmark dose is a methodological improvement over the NOAEL approach. Many data sets do not even establish a NOAEL, but are still useful for benchmark calculations. Finally, the uncertainty factors used in conjunction with the benchmark dose calculations have been reduced by 3.3-fold compared to the NOAEL approach since we believe that some degree of variation in establishing a dose-response threshold has been accounted for by use of the 95% lower bound on the experimental dose-response. OEHHA is therefore reducing uncertainty by use of benchmark dose calculations and is using correspondingly less conservatism in deriving RELs.

Comment: The Associations acknowledge that statistical Lower Confidence Limits ("LCL") have typically been used in expressing benchmark doses rather than maximum likelihood estimates. OEHHA has selected the 95% Lower Confidence Limit expression in its BD calculations including that for formaldehyde. Although the 95% level is not an uncommon confidence interval, selections often range from 90% to 99%. The choice of 95% is a policy decision. Some traditionally noted benefits of confidence limits include their sensitivity to the sample size, their stability in the face of minor changes in the data, and their ability to be determined in some instances when the MLE cannot be. However, like all of the other factors that go into the BC calculation, the appropriateness of the 95% lower confidence limits must be evaluated against the backdrop of the other assumptions that have been made and the scientific knowledge about the substance. If the combination of conservative assumptions leads to results that are out of the range of practical experience, then one must evaluate the justification for the assumptions.

It is undeniable that the calculation of the BC depends heavily on the BMR as well as the size of the statistical confidence bound that is used. The selection of a 1% BMR in the case of formaldehyde as opposed to 5% or 10%, not only drives down the MLE curve but also causes even larger disparities between the MLE expression and the 95% LCL value than occurs at the higher level. The poorer the quality of the data used, the lower the statistical confidence particularly at the lower portions of the curve. The relatively small group sizes employed in most noncancer toxicology studies will also necessarily impart a substantial degree of conservatism to lower bound benchmark dose estimates. Standard carcinogenicity bioassay designs call for 50 animals per sex per treatment group. Human studies for non-cancer endpoints, and particularly endpoints of relatively minor toxic severity are typically much smaller. The Kulle study with 19 participants is not atypical. Although the use of a human study eliminates the use of the default interspecies uncertainty factor of 10, it almost guarantees a penalty through the inflation towards zero of the lower confidence limit. The Associations urge OEHHA to include in the revised TSD a graphic representation of the benchmark dose derivation similar to what has been provided for most other chemicals subject to the benchmark dose approach. This will allow both regulators and the public to better understand the differences between the MLE and the LCL expressions.^l6

Response: OEHHA will provide the graphic display of the formaldehyde benchmark dose to be consistent with the graphic displays already contained in the document for other chemicals, as requested. The use of the lower 95% confidence limit has wide precedent in statistical analyses as well as benchmark dose calculations (Auton, 1994; Malsch et al., 1994; Crump, 1984). Furthermore, as discussed in a previous comment, the magnitude of change between the BC₀₁ and the BC₀₅, including the 95% lower bounds, is only about 1.3 on average. Therefore, the 95% lower bounds do not diverge at an excessive rate in this region. This can be seen in the graphic displays of the benchmark calculations for ammonia, carbon tetrachloride, several of the glycol ethers, and many other chemicals in the technical support document. In general, the minimal distance between the MLE and the confidence bound occurs at the 50% response rate, which is obviously far in excess of an acceptable response rate. As the response rate deviates from that point, the distance between the MLE and confidence bound increases. There is no a priori reason for selecting 5% over 1% response based on distance between confidence bound and MLE. Empirically, the data sets in this document do not show a large deviation from 95% lower bounds of 1% vs. 5% response rates in either Probit or Weibull models.

Comment: Additionally, it would be helpful to have more detail on the computer program that has been employed in the formaldehyde analysis. The TSD references a 1983 unpublished work by Crump -- "Probit-A Computer Program to Extrapolate Quantile Animal Toxicological Data to Low Doses" -- as the program that was used to fit the human data from Kulle. The assumptions that are used in any program are critical to understanding the appropriateness of its use. The process should be totally transparent.

Response: OEHHA will provide the mathematical basis for the calculation of the benchmark dose in the Technical Support Document.

Comment: The Level I REL for formaldehyde uses an uncertainty factor ("UF") of 10 to account for "individual variation." An additional modifying factor of 0.3 was used because "...the BC accounts for some degree of individual variation. The Associations submit that either the uncertainty factor should be lowered or a reduction should be made in the modifying factor -- perhaps to 0.15 or 0.2 rather than the 0.3 that has been used.

Response: The suggested change is not justified by the commentator. The Kulle et al. (1987) study only examined up to 19 healthy subjects. These subjects do not represent the most sensitive individuals in the entire population. The actual range in sensitivity in the general population is likely to be greater than 1.5 or 2-fold.

OEHHA has changed its uncertainty factor methodology since the release of the public comment draft. The interspecies uncertainty factor for formaldehyde is now 3. A modifying factor is no longer applied. This reflects OEHHA’s use of UFs with the BC approach.

Comment: As noted by the National Academy of Sciences in its report, "Science and Judgment in Risk Assessment:"

When reporting estimates of risk to decision-makers and the public, EPA should report not only point estimates of risk but also the sources and magnitudes of uncertainty associated with these estimates. p. 12-19.

Response: The estimates and uncertainty factors for each chemical are specifically given in the Technical Support Document. Statistical uncertainties in exposure estimates are addressed in a separate guideline document, Technical Support Document for Exposure Assessment and Stochastic Analysis.

TOC | Back | Next