Title: EPA-HQ-OPP-2013-0821-0008_content.pdf

URL Source: https://downloads.regulations.gov/EPA-HQ-OPP-2013-0821-0008/content.pdf

Published Time: Sat, 12 Apr 2025 00:19:18 GMT

Number of Pages: 9

Markdown Content:
# 2 . De Minimis and the Threshold of Regulation  

> Alan M . Rulis Food and Drug Administration

INTRODUCTION 

This chapter addresses the concept of a "threshold of regulation" for 

food additives . The Food and Drug Administration (FDA) has appreci-

ated the need for such a concept since the early years after the passage of the 1958 Food Additives Amendment to the Federal Food, Drug and 

Cosmetic Act (the FD&C Act, or the Act) . The approach put forward in this chapter is based upon the premise that, through an examination of a sufficiently large sample of toxicological data from both classical 

toxicological feeding studies and from carcinogenicity bioassays, some global delimiters of risk and exposure can be determined to define levels of human exposure and/or levels of migration of substances to food that 

can be said to fall below some "threshold of regulation ." When this is the case, the substance in question would not necessarily need to undergo the rigors of the premarket safety evaluation requirements of the Act . Inste2d, the particular use of the substance could be acceded to by the FDA after an abbreviated review of pertinent information, thus 

avoiding the need for the submission and agency approval of a food additive petition covering the use of the substance . (Implicit in such a 

process is the absence of any indication that the substance in question is in fact a carcinogen or other potent toxin . Known carcinogens would 

need to be handled in a considerably more formal manner, and would be subjected to more formal risk assessment and risk management deci-

sionmaking.) 

THE PROBLEM 

Section 201(s) of the FD&C Act defines a food additive as : 

> . . . any substance the intended use of which results or may reasonably
> - be expected to result, directly or indirectly, in its becoming a compo-nent or otherwise affecting the characteristics of any food (including 29

# http://legacy.library.ucsf.edu/tid/rdc05c00/pdf 30 TOXICOLOGY any substance intended for use in producing, manufacturing, packing, processing, preparing, treating, packaging, transporting, or holding food . . . . I Because this broad definition includes food packaging and other food-contact materials and their components that might migrate unin-tentionally into food, it is easy to see why the need has arisen for a policy concerning a "threshold of regulation ." With the development of ever more sensitive and discriminating analytical capabilities, the ana-lyst today can confirm the presence in food, or in food-simulating sol-vents, of extremely small amounts of substances . In some cases, these amounts may be as small as a few parts per billion (ppb) . The FDA is often presented with situations in which a chemical substance used or present in a food-contact material may indeed theo-retically migrate to food, but is predicted to do so at levels that are so low as to arguably not satisfy the above statutory definition . Often such substances are not known to be carcinogens . Perhaps little else is known about their toxic potential either . The question that often arises for these very low level migrants is, "Is it necessary for the agency to request and approve a full-blown petition allowing the presence of the sub-stance in food?" Over the years, the agency has made many threshold-of-regulation decisions on a case-by-case basis, although some people have cited inconsistencies in these decisions and some have called for a more formal policy statement by the agency, indicating the principles and procedures it would follow in granting threshold-of-regulation approvals . If the agency wishes to continue to use its discretion not to require a petition in certain cases, is there some consistent basis on which this can be done, while still protecting the public health in the (unlikely) event that the substance in question turns out to be a carcino-gen or other potent toxin? (At exceedingly low exposure levels -say, less than 10 ppb--it is likely that potential carcinogenesis would be almost the only toxic phenomenon capable of producing any concern .) Stating the problem another way, we may ask, "How can the agency determine under a general policy that a given use of a food-contact material or a component thereof, not presently known to be a carcino-gen, does indeed not require full petition review and regulation, and will, even in the worst case, present no unwarranted risk?" Further-more, more, we might ask, "How can the agency build consistency and scien-tific credibility into the threshold-of-regulation decisionmaking pro-cess?" It is with one form or another of this problem, and with the above questions, that the agency, the regulated industry, and the courts have grappled for nearly three decades .~ Arguably, requiring the regulation of low ppb levels of substances places significant and costly burdens on both the agency and the affected industries, while the human exposure 

# http://legacy.library.ucsf.edu/tid/rdc05c00/pdf A .M . RULIS 31 ng, ing : other ; unin-ifora 'ient of ie ana-ng sol-', these emical i theo-are so i~ such eknown ses for '~equest e sub-,shold-'people i for a '-iciples ilation not to isis on in the iicino-iy, less ;3lmost rgency ' ontact ircino-n, and trther-seien-3 pro-=-above -s have tlation ens on posure to such low levels of substances may have negligible public health signif-icance . Conversely, however, even small exposures to highly toxic sub-stances or to potent carcinogens might sometimes be found to produce significant human risk . The agency must navigate carefully between these two potentially unpleasant consequences . Obviously, no simple solutions to this problem will materialize over-night . However, recent progress in understanding the global trends and distributions of risk and exposure may help the agency to sufficiently clarify the scope of available options, so that a consistent policy can finally be achieved . This chapter addresses some areas of recent think-ing about the problem at the FDA and offers some recommendations for making further progress . 

## A PROBABILISTIC APPROACH 

The complexities of the biological interactions of any chemical with a living system are enormous . To construct a threshold-of-regulation pol-icy that anticipates all potential adverse responses that a compound may induce, including potential carcinogenesis, is clearly impossible . Fortunately, however, the present toxicological literature contains an enormous amount of data on the toxicity of chemical substances . The premise of this chapter is that a proper summarization and analysis of these data may provide important needed insights into the threshold-of-regulation question . For example, Figure 1 displays potency data on a subset of 343 oral carcinogens from animal studies recently compiled by Gold et al .y For the purpose of the present discussion, "potency" is defined as the slope of a straight line connecting the point representing TD,, dose (toxic dose to 50% of the test animals) of Gold et al . with the point representing zero risk and zero dose . (For a more complete description of the defini-tion, refer to Flamm et al . 10 Figure 1 shows that when carcinogen potencies are analyzed by grouping them into ranges and plotting them as a probability distribution, they form, on a semilogarithmic scale, a curve that is remarkably Gaussian (normal) in shape . The existence of such a curve for carcinogen potencies demonstrates not only the well known fact that chemical carcinogens present greatly different poten-cies (probabilities of inducing cancer per unit level of intake), but also that a single chemical carcinogen selected at random can be predicted a priori to have a potency that falls within fairly well defined limits . It seems that such knowledge ought to be applicable to the threshold-of-regulation question . Therefore, let us now try to apply this knowledge to the problem at hand, namely the delineation of options for determin-

# http://legacy.library.ucsf.edu/tid/rdc05c00/pdf 32 TOXICOLOGY 

> Figure 1 . Histogram and nonlinear least squares best fit Gaussian to the potencies of Gold et al .9
> 10

~w~ .,r. e.. . ...:~  

> ~ .. . .rno ..w., 10

# g 10 4 10 2 1- 10 /0 4

> Pa.ney IntpykWdi''

ing exposures to chemicals that might be said to fall below a threshold of regulation . First, in order to be useful, the Gaussian distribution of Figure 1 must 

be transformed into an exposure distribution at a constant assumed risk 

of I x 10$ per lifetime . (This process has been previously described .10) 

(A risk level of 1 x 10 e per lifetime has been chosen because it is the upper bound level of risk identified by the agency recently as de 

minimis for the purpose of regulating the carcinogen methylene chlo-ride as a coffee decaffeinater.)" The transformation applicd to the potency distribution of Figure 1 results in curve number 5 of Figure 2 . 

We call such a curve a"risk equivalent exposure distribution ." It describes the relative probability that a carcinogen selected at random 

from the universe of known carcinogens will be one that presents a risk 

of 1 x 10 e per lifetime at the exposure level indicated on the horizontal axis across the top of Figure 2 . As can be seen, about half the area of 

curve 5 falls on either side of the 1 ppb line . Therefore, at a human 

dietary intake of 1 ppb, about half of all carcinogens are predicted, under the very simplified assumptions of this analysis, to present a risk 

of greater than 1 x 10 e per lifetime ; the other half, lower than 1 x 10-a . 

> rv
> 10

# http://legacy.library.ucsf.edu/tid/rdc05c00/pdf an to 

gshold of 

`e 1 must med risk iibed .10) ' it is the 

y as de 

ne chlo-1 to the },gure 2 . 3on :' It iandom its a risk irizontal 

i"area of human edicted, pt a risk 'an I x 

A .M . RULIS Figure 2 . Relative probability distributions for various groupings of toxicity data. Curves are arbitrarily vertically scaled and are described in the key. Upper abscissas refer to various measures of intake . "PP" refers to the "packaging factor" defined in the text . Lower abscissa values are negative base-10 logarithms of effect levels in mg/kglday for nonacute data, mg/kg for acute data, and TD50s for the carcinogen data of Gold et al? Below lower abscissa are minimum lethal doses for selected toxins for comparison purposes . 

> obxin

# Zsvichnme      

> `olphlherle -Iny,r EMM LwN ONMEI fw Mkl

KEY: 1 18000 RTECS LD50s Rat/Mouse ; 2 130 LD5°s for Gold et al . carcinogens ; 3 PAFA LD50s (295 cpds) ; 4 PAFA LELS (159 cpds) ; 5 Risk-Equivalent Exposure Distribution 10~ Risk for 343 Gold et al . carcinogens . Based on assumption of linear extrapolation from TD5°s. 33 A threshold-of-regulation level that precludes almost any measurable lifetime risk from carcinogenesis (except for risk from substances like aflatoxin B-1 or 2,3,7,8-tetrachlorodibenzo-p-dioxin [TCDD]), say-one part per trillion (ppt) in the human diet -would in all likelihood be so low that it could not be reliably measured analytically, and certainly could not be easily and consistently enforced . To insist on such a strin-gent level of risk protection would, in effect, preclude any reasonable and practical threshold-of-regulation policy . The use of a 5-ppb dietary level as a threshold of regulation presents an interesting example, because this level has recently been suggested by 

# http://legacy.library.ucsf.edu/tid/rdc05c00/pdf some as an appropriate level to use for deciding on a threshold for regulation .12 About 60% of the area under curve 5 of Figure 2 falls to the right of the 5-ppb line . Therefore, it can be predicted on a probabi-listic basis that should a substance permitted under a threshold-of-regulation decision unknowingly be a carcinogen, it would, under the present analysis, have roughly a 60% chance of presenting greater than a 1 x 10~ per lifetime level of risk at the 5-ppb level of exposure . Some have suggested that an appropriate threshold-of-regulation level of exposure would be the 5-ppb human dietary exposure as above, but "modulated" by known classical (noncarcinogenic) toxicity infor-mation . For example, one could add to the 5-ppb requirement the additional constraint that the LDso of the migrant be no lower than 105 times the threshold-of-regulation level (5-ppb) . In other words, any candidate for a threshold-of-regulation decision whose LDso falls to the right of about -1 on the abscissa of Figure 2 would be disqualified for consideration at the 5-ppb exposure level on the grounds that its (non-carcinogenic) toxic potency is too high . Certain aspects of such an approach have recently been put forward .12 In order to analyze this type of proposal, data on toxicity of numerous chemical substances were compiled . The curves numbered 1, 2, 3, and 4 on the left of Figure 2 describe these data . Curve 1(arbitrarily scaled) depicts a nonlinear least squares best fit of a Gaussian curve to data compiled from 18,000 oral (rat or mouse) LDsos contained in the Regis-try of Toxic Effects of Chemical Substances (RTECS) . Curve 2 (arbi-trarily scaled) represents a similar envelope of LDsos for 130 compounds identical to the carcinogens of Gold et al . comprising curve 5 . Curves 3 and 4 (also arbitrarily scaled) derive from data analyzed in the FDA's Priority-Based Assessment of Food Additives (PAFA) project ."-" Curve 3 represents the probability distribution of LDsqs for 295 regulated food additives analyzed in the PAFA project . Curve 4 represents an envelope of lowest effect levels from subehronic or chronic feeding studies on food additives studied in PAFA . As can be seen, the "realm" of classical toxicity (except for pesticides, economic poisons, or some exquisitely toxic substances arrayed along the abscissa for comparison) is fairly well delineated by these curves, with no sizeable probability of such effects occurring in rodents at doses lower than about I mg/kg/day (to the right of about 0 on the abscissa) . (Such knowledge is actually not new at all, having been observed and duly noted by Frawley in 1967, using a different data base .2) The noteworthy point for our purposes, however, is that the require-ment that the acute toxic dose of a substance be no less than five orders of magnitude higher than the benchmark 5-ppb dietary intake level is no requirement at all, because virtually all chemicals can meet that criterion . To see this, just step five decades to the left of the 5-ppb arrow ~ . . Wv7-° ~

# http://legacy.library.ucsf.edu/tid/rdc05c00/pdf A .M . RULIS 35 

'thold for 

"2 falls to 

,probabi-

shold-of-

mder the ater than Isure . 

: :gulation 

is above, tv infor-nent the than 10' 

~ .rds, any 

Ils to the 

tlified for 

its (non-

such an umerous 

:, 3, and ' y scaled) to data ie Regis-22 (arbi-apounds 's~urves 3 e FDA's 

e1s Curve I ted food 

":nvelope ;itlies on 

sticides, 

d along 

curves, ' at doses bScissa) . 

ved and require-2 orders 

level is -ae t that 

b arrow 

in Figure 2 . Analysis of relationships between acute toxicity and carcin-ogen potency in recent work in progress at the FDA reveals that use of such comparisons is of dubious practical value when applied to threshold-of-regulation questions where the acceptable upper bound risk value is set at 1 x 10 ° or lower . Recently, Wilson et al .16 have dealt with this relationship and have demonstrated its potential utility in the decisionmaking process . They too, however, note that such an approach is of limited value when the acceptable risk level is set at 1 x 10-e per lifetime or Iower .1F So far, we have shown how compilations of potency and toxicity data can be used to rule out certain threshoid-of-regulation scenarios . They can also be used to support other choices . For example, suppose we define a threshold-oFregulation level to be 50 ppt in the human diet . Such a level, although quite low, would correspond to a migration level to food of 1 .0 ppb, assuming a 5% "packaging factor ." (A "packaging factor" may be defined as the weight fraction of the human diet likely to be packaged in a given material .) One ppb, while on the borderline of reliable analytical detection, is not outside the realm of measurability . This level divides curve 5 of Figure 2 in such a way as to exclude about 85% of carcinogens that produce greater than I x 10-' per lifetime risk, if held at that level of exposure. If we conservatively assume that one out of five chemicals assented to by the agency under a threshold-of-regulation policy at this exposure level is unknowingly a carcinogen, then on a purely probabilistic basis we can argue that 95 out of every 100 such threshold-of-regulation decisions will result in no more risk than 1 x 10 6 per lifetime . Furthermore, that level of risk is a maxi-mum, with the vast majority yielding far lower levels of potential risk . 

SUMMARY AND RECOMMENDATIONS 

A probabilistic approach has been described for analyzing large amounts of toxicological data on numerous chemical substances . The application of such an analysis to the problem of the threshold of regu-lation for food-contact materials was described . While such an approach certainly does not address the immense bio-chemical complexity inherent in the interactions of individual chemical substances with living systems, it may prove to be a useful tool for FDA in evaluating the viability of options for, and in providing a scientific basis of support for, choices of threshold-of-regulation migration levels . Once the agency is able to support a given migration level or range of levels as acceptable for a threshold of regulation, it will be in a better position than ever before to make threshold-of-regulation judgments on a consistent basis . 

# http://legacy.library.ucsf.edu/tid/rdc05c00/pdf Once the principles of such a policy have been established, it would remain to devise appropriate procedures for implementing the policy . It has been suggested . that the agency could make such threshold-of-regulation judgments based on a sufficient summary of both chemical and toxicological information provided by an applicant . Certainly some sort of abbreviated agency review of summary information may be an important aspect of such procedures . In order to be able to efficiently review its accumulated decisions over the years, the agency could keep records of all threshold-of-regulation actions . Such accumulated deci-sions could form the basis for a data base of information on low level migration of chemical substances to food . It appears that there presently exists an adequate scientific basis of data and information on which to construct a threshold-of-regulation policy relating to food-contact substances . By applying this information to the problem and implementing appropriate policies and procedures, the agency will be in a position to better delineate the boundaries of the food additive definition of the FD&C Act as it relates to food-contact substances . 

## REFERENCES 

1 . Federal Food, Drug and Cosmetic Act, as amended, (Title 21 U .S . Code) . 1958 . U.S . Government Printing Office, Washington, DC . 2 . Frawley, J .P . 1967 . Scientific Evidence and Common Sense as a Basis for Food-Packaging Regulations . Food . Cosmet . Toxicol . 5 : 293-308.               

> 3 . Checchi, A .A . 1959 . Developments Under the National Pure-Food Law Affecting the Packaging Industry . Food Drug Cosmetic Law Journal 14 : 527-533 . 4 . Checchi, A .A . 1959 . Food Additives Procedures and Policies . Food Drug Cosmetic Law Journal, 14 : 591-596 .

5 . Rankin, W .B . 1959 . Incidental Food Additives . Food Drug Cosmetic Law Journal 14 : 768-773, 777 . 6 . Kirk, J .K . 1960 . Food Additive Developments . Food Drug Cosmetic Law Journal 15 :755-760 . 7 . Harvey, J .L . 1962. Food Additives and Regulations . Food Drug Cosmetic Law Journal 17 : 272-281 . 8 . DC Cir . 1979 . Monsanto v . Kennedy, 613 F. 2d 947 .   

> 9 . Gold, L .S., et al . 1984. A Carcinogenic Potency Data Base of the Stand-ardized Results of Animal Bioassays . Environ . Health Pers. 58 : 9-314 . 10. Flamm, W.G ., Lake, L .R ., Lorentzen, R .J ., Bulis, A .M ., Schwartz, P .S .,

and Troxell, T.C . Carcinogenic Potencies and Establishment of a Thresh-

old of Regulation for Food Contact Substances . Submitted for Publication in Risk Assessment . 

# http://legacy.library.ucsf.edu/tid/rdc05c00/pdf A .M . RULIS 37 it would poliey. It NsTrold-of-

emical 

y some ;ry be an 

iently 1d keep deci-level basis of ation ation 

ures, of the 

contact 

11 . Cosmetics : Proposed Ban on the Use of Methylene Chloride as an Ingredi-ent of Aerosol Cosmetic Products . Dec. 18, 1985 . Federal Register 50 : 51551-51559 . 

12 . Threshold of Regulation for Packaging Materials Proposed . Dec . 2, 1985 . Food Chem . News 29-30 . 13 . Rulis, A .M ., Hattan, D .G ., and Morgenroth, V.M . 1984 . FDA's Priority-Based Assessment of Food Additives: Preliminary Results . Regul . Toxicol . Pharmacol . 4 : 37-56. 

14 . Bulis, A .M ., and Hattan, D .C . 1985 . FDA's Priority-Based Assessment of Food Additives : General Toxicity Parameters . Regul . Toxicol . Pharmacol . 5 : 152-174 . 

15 . Hattan, D .C . and Rulis, A .M . 1986 . FDA's Priority-Based Assessment of Food Additives : Specific Toxicity Parameters . Regul . Toxicol, Pharmacol . 6 : 181-191 . 

16 . Zeise, L ., Wilson, R ., and Crouch, E . 1984 . Use of Acute Toxicity to Estimate Carcinogenic Risk . Risk Analysis 4 : 187-199 . 

# http://legacy.library.ucsf.edu/tid/rdc05c00/pdf