Choosing Chemicals for Precautionary Regulation: A Filter Series Approach (2005)

Choosing Chemicals forPrecautionary Regulation: A FilterSeries ApproachU L R I C H M U L L E R - H E R O L D , *M A R C O M O R O S I N I , A N DO L I V I E R S C H U C H T

Swiss Federal Institute of Technology Zurich (ETH),8092 Zurich, Switzerland

The present case study develops and applies a systematicapproach to the precautionary pre-screening of xenobioticorganic chemicals with respect to large-scale environmentalthreats. It starts from scenarios for uncontrollable harmand identifies conditions for their occurrence that then arerelated to a set of amplifying factors, such as characteristicisotropic spatial range F. The amplifying factors relatedto a particular scenario are combined in a pre-screeningfilter. It is the amplifying factors that can transform a potentiallocal damage into a large-scale threat. Controlling theamplifying factors means controlling the scope and rangeof the potential for damage. The threshold levels for theamplifying factors of each filter are fixed through recourseto historical and present-day reference chemicals so asto filter out as many as possible of the currently regulatedenvironmental chemicals and to allow the economicallyimportant compounds that pose no large-scale environmentalconcern. The totality of filters, with each filter correspondingto a particular threat scenario, provides the filter seriesto be used in precautionary regulation. As a demonstration,the filter series is then applied to a group of nonreferentialchemicals. The case study suggests that the filterseries approach may serve as a starting point forprecautionary assessment as a scientific method of itsown.

IntroductionOn February 2, 2000, the European Commission informedthe interested parties of the manner in which the Commissionapplied or intended to apply the precautionary principle (1).In its attempts to compose a European policy on theapplication of the precautionary principle, the Commissionhas funded PrecauPri, a thematic network project conductedunder the auspices of the EC’s STRATA Programme. Theparticular aim of the project was to develop a scientificallysound, politically feasible, legally unambiguous, and demo-cratically legitimated concept of precaution. The conceptshould provide a policy framework for the implementationof the precautionary principle in different risk areas andensure specificity and predictability for the various actorsinvolved. Within PrecauPri, the regulation of chemicals wasselected as a test case for the design of appropriate proceduresin the application of precautionary reasoning.

Although there is no single authoritative definition of theprecautionary principle, in many of its various formulations

four dimensions can be identified (2): (i) a dimension ofthreat; (ii) a dimension of ignorance, concerning the limitsof scientific knowledge; (iii) a dimension of action, concerningthe response to the threat; and (iv) a dimension of command,concerning the way in which the action is prescribed. By“threat”, in this context, is meant one or another undesiredstate of the world. “Ignorance” has a wide range of differentmeanings, from milder forms of uncertainty about prob-abilities or incomplete proof of supposed cause-and-effectrelations to the most extreme form of “ignorance of igno-rance” where the kind of possible unwanted effects itself isunknown (3). This last possibility typically applies toenvironmental chemicals, where the complete spectrum ofpossible negative effects is but insufficiently known.

However, even if negative effectsssuch as plant toxicity,thinning of eggshells of sea eagles, endocrine disruption,weakening of the immune systemsare largely unknown,there may be sufficient knowledge of so-called amplifyingfactors that can serve as a basis for precautionary action. Itis the amplifying factors that transform, for example, toxicityinto a large-scale environmental threat. Toxicity, by its verynature, is first of all a local phenomenonsone organism isexposed to a possibly toxic dose of a substance at a giveninstance and a given place. However, amplifying factors suchas mobility, bioaccumulation, and persistence of a substancecan transform its toxicity into a nonlocal, possibly globallarge-scale problem. The amplifying factors then generatethe large-scale nature of the respective management prob-lems. If, eventually, perhaps even a long time after release,an apparently innocuous persistent and mobile chemicalhas negative biological effects, it is impossible to eliminateit from the environment. The resulting situation would beuncontrollable because even immediate phasing out maynot ameliorate the situation quickly enough for some species.The PCBs and the extinction of the European otter (Lutralutra) can be regarded as an example of this behavior (4).

This leads to the central idea of precautionary regulation.It considers large-scale threats arising from uncontrollablesituations in case of eventually discovered adverse effects.It is a regulation based on reliable scientific knowledge ofamplifying factors but prior to knowledge of adverse effects.It is tailored to situations where no immediate action wouldsolve the problem if any novel adverse effect is discovered.

Although persistence is an amplifying factor of this kind,one has to recognize that the long-term presence of achemical or product alone does not lead to possiblyuncontrollable environmental situations of said type. (Thiscan be seen from the examples of concrete, bitumen, plastics,etc.) Only in combination with other amplifying factors, suchas mobility, does persistence play a significant role asindicator for large-scale chemical threats. These observationsare used to propose a general approach to precautionaryregulation by controlling amplifying factors of adverse effectsinstead of controlling adverse end points directly (5).

Precaution and Chemical Risk AssessmentThe current practice of chemical risk assessment centersaround the identification of risks for human health and theenvironment (6). The detailed outcome of the assessmentprocedure then leads to specific regulations depending onexposure, tonnage, and use pattern. The present case studyaims at complementing this procedure by a precautionarypre-screening stage (Figure 1).

Regulations on the basis of precautionary assessment arenecessarily controversial, mainly due to the dimension ofignorance and uncertainty. Since stakeholders in public

* Corresponding author phone: +41-44-6324403; fax: +41-44-6331136; e-mail: [email protected].

Environ. Sci. Technol. 2005, 39, 683-691

10.1021/es049241n CCC: $30.25 2005 American Chemical Society VOL. 39, NO. 3, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 683Published on Web 12/24/2004

debates or in courts of law are not experts in this field,arguments and methods have to be transparent, simple, andintuitivesin addition to being scientifically correct. Thisexcludes black-box-type computer calculations, highbrowmathematics, and esoteric chemical details. Furthermore,measures should concentrate on serious threats where “wait-and-see” strategies cannot be justified. The history of theMontreal Protocol on ozone-depleting substances shows thatrestrictions on the basis of precautionary arguments can beaccepted even by stakeholders with opposite interests if thethreats are important and action is urgent.

Along these lines, we present a filter series procedure forprecautionary pre-screeening of organic chemicals withrespect to large-scale environmental threats. Each filter isdesigned to screen for one particular threat scenario (whichhelps intuition). Accordingly, there is a one-to-one cor-respondence between threat scenarios and filters. Theamplifying factors entering the filters are calculated byinserting measurable physicochemical constants into simpleand theoretically sound formulas (transparence), and thecalibration of the filters makes optimal use of historicalexperience with environmental chemicals (relevance). Theoverall outcome is independent of the filter ordering, andnew threat scenarios can be taken into account withoutquestioning earlier results that led to the elimination ofsuspect compounds (upward compatibility). Substances notfiltered out by any of the filters continue on to standardchemical assessment.

Large-Scale Threat Scenarios and FiltersOriginally, the PrecauPri case study provided a four-membered filter series for pre-screening: Pandora, ColdCondensation, Transformation Pandora, and Bioaccumu-lation (Figure 2). The Pandora scenario relates to enduringubiquity of environmental chemicals. The Cold Condensationor Cold Trap scenario considers the selective accumulationof environmental chemicals in low temperature areas, firstof all in the polar regions. In addition to a domain of directimpact, environmental chemicals have a second, naturallymore extended domain of influence due to their transforma-tion products in the environment. The TransformationPandora scenario, accordingly, deals with the enduring

ubiquity of these secondary compounds, using the results ofQuartier and Muller-Herold (7) and of Fenner et al. (8). Thethreat scenario related to Bioaccumulation is given throughthe possibility of substances to have adverse effects on livingorganisms even if their concentration in the oceans, lakes,rivers, or in the atmosphere is extremely low. If a bioaccu-mulating and persistent chemical has negative biologicaleffects, it is impossible to eliminate it from the biosphere,and the resulting situation is as uncontrollable as in thePandora scenario.

Due to the data situation, the present case study restrictsitself to a combination of only Pandora and Bioaccumulation.(An introduction to Transformation Pandora and Cold Trapis provided in Section 5 of the Supporting Information, andan example of a three-filter sequence including Transforma-tion Pandora is provided in Section 6 of the SupportingInformation.)

Pandora. The Pandora scenario is named after the Greekmyth of Pandora’s box, which held all evils and complaints.When the box was opened, its contents were unleashed uponthe world, causing irreversible harm. The enduring ubiquityof persistent organic pollutants (POPs) is regarded as theepitome of the Pandora scenario (9). For the construction ofa related filter, one observes that the Pandora scenario isessentially due to the interplay of mobility and longevity.The potential for mobility and longevity is expressed by twoproxy measures: characteristic isotropic spatial (CIS) range(F) and characteristic isotropic global (CIG) half-life (τ).

Characteristic isotropic spatial (CIS) range F is the typicaldistance a molecule would travel before degradationsunderearth-like but spatially isotropic conditions where concen-trations quickly equilibrate between the atmosphere, thesurface layer of the oceans, and the upper layer of the soil(see Appendix).

Characteristic isotropic global (CIG) half-life τ is the typicaloverall lifetime of a molecule under conditions as for F (seeAppendix). (The joint use of spatial range and persistence inchemical assessment goes back to Scheringer and Berg (10).(For details of the subject and its history, see Scheringer (11)and references therein.)

Bioaccumulation. Bioaccumulation (12) is a phenomenoncombining bioconcentration and biomagnification. Biocon-centration relates to the partition of a chemical between anorganism and a surrounding inorganic medium (e.g., leaves/air, fish/water). Biomagnification denotes the heterotrophicenhancement of concentration in subsequent elements ofthe food chain (grass/cow, cow/man).

As fat tissue is the relevant storage medium in livingorganisms and as octanol is the chemical proxy usuallyrepresenting organismic fat, bioconcentration is related eitherto a chemical’s octanol-water partition coefficient (Kow) orto its octanol-air partition coefficient (Koa ) Kow/K′), with K′) KH/RT being the chemical’s dimensionless Henry’s lawconstant. Kow is a direct measure for bioaccumulation fromwater into aquatic species, whereas Koa is a direct measurefor bioaccumulation into plants from air. In order not toclassify the Montreal gases as bioaccumulatingswhich theydefinitely are notsKoa is preferred to Kow. (For details of thischoice, see Section 8.1 of the Supporting Information.)Analogous to the Pandora scenario, the Bioaccumulationfilter is based on two amplifying factors: a combination ofhigh Koa values and increased global characteristic persistence(τ). (To bioaccumulate, a chemical has to survive a minimalperiod of time before degradation.)

Filters and Filter SeriesIn the case study, the individual filters were realized as two-parameter classification schemes with three outcomes:“green” (“unconditional clearance”), “yellow” (“conditionalclearance”), and “red” (“no clearance”). For filters based on

FIGURE 1. Extended chemical assessment including pre-screening.A chemical not screened out by one of the filters proceeds to standardchemical assessment.

FIGURE 2. Scheme of a series of four filters for pre-screening withrespect to large-scale environmental impact.

684 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 3, 2005

two parameters x and yswith each parameter x or y havingthe grades high/medium/lowsthe outcomes are definedusing these grades of the two parameters (Figure 3): green(medium/low, low/low, low/medium); yellow (high/low,medium/medium, low/high); red (high/medium, high/high,medium/high).

The calibration of filters now consists of defining theparameter grades leading to the filter outcomes green, yellow,and red. For two-parameter filters with three grades for eachparameter, one has to find limiting values separating low/medium and medium/high for the respective filter param-eters. If x and y denote the two parameters and xlm, ylm, xmh,and ymh denote the corresponding limiting valuesswherexlm signifies a limiting value separating low values of x frommedium ones and xmh is the corresponding mark at the borderbetween medium and high, etc.sthen the two points, (xlm,ylm) and (xmh, ymh), define a partition of the x-y plane intothe required nine rectangular filter domains (see Figure 3),which then are grouped into the three filter scores: green(low/low, low/medium, medium/low), yellow (low/high,high/low, medium/medium), and red (medium/high, high/medium, high/high).

In the case of a series of several filters, the above procedureapplies to each filter separately. The outcome of the seriesas a whole then consists of a list of results for the individualfilters that subsequently have to be combined to an overallresult. This is performed along the following rules:

1. A substance classified as red by at least one filterdefinitively constitutes a serious threat, which triggersprevention. Such a chemical should be eliminated (with thepossible exception of “life-saving” pharmaceuticals or someintermediates in industrial synthesis if contained under strictsafety standards).

2. Green results in all filters open the way to standardchemical risk assessment. Such a result implies that thesubstance is inconspicuous with respect to the threatscenarios under consideration.

3. The intermediate cases (i.e., yellow results with orwithout green scores) trigger a variety of procedures,depending on the intended modes of use.

Rule 1 is the epitome of the precautionary approach,whereas rule 2 opens the door to current practice. Rule 3may result in requirements relating to chemical modification(pesticides), restriction of production volume (consumerproducts), or containment charges (intermediates in chemicalsynthesis), etc.

Case Study with Two Filters: Pandora andBioaccumulationThe essential difference between one single filter and a seriesof filters is most easily illustrated by a combination of onlytwo filters: A precarious chemical should get a red score byat least one of the filters. For precautionary regulation, thereis no need to receive a red score from more than one filtersince one red is regarded as a sufficient condition forpreventive measures. Since this distinction becomes trivialin the case of one single filter, one needs at least two filtersfor its nontrivial demonstration. Additionally, for obviousreasons it has to be required that chemicals known asinconspicuous should be stopped by none of the filters.

As an example, the amplifying factors for both the Pandorafilter and the Bioaccumulation filter were calculated. For thecalculation of characteristic isotropic spatial (CIS) range,characteristic isotropic global (CIG) half-life, and octanol-air partition coefficient (Koa) of a chemical, four substance-related input data are needed (see Appendix): KH, Henry’slaw constant (air-water partition coefficient); Kow, octanol-water partition coefficient (descriptor of lipophilicity); ka,degradation rate constant in air; and kw, degradation rateconstant in water.

On the basis of the data of the top 35 U.S. High ProductionVolume (organic) Compounds (HPVCs) (13) as paradigmaticexamples for chemicals not posing large-scale threats in theenvironment and a relevant selection of 43 Montreal/Kyoto/Stockholm compounds as paradigmatic examples for pre-carious chemicals, the output parameters τ, F, and Koa werecalculated. The results are shown in Figures 4 and 5. It turnsout that in both scenarios the regulated compounds are wellseparated from the HPVCs.

Filter Calibration and Filtering ResultsFollowing the Filters and Filter Series section, one now hasto find the limiting values defining the filter grades forPandora and Bioaccumulation (which at the same timecorresponds to the specification of a level of protection).Generally, limiting values are directly discussed in purelyscientific terms. Along these lines, one could try to fix thefilter calibration directly. However, the history of medicaland environmental threshold values shows that the way tofirm, lasting agreements is long and troublesome. To cometo a first meaningful estimate, we look at limiting valuesoptimally separating the two sets of reference substancessthe 35 U.S. HPVCs and the 43 chemicals of the Montreal/

FIGURE 3. Two-parameter filter with three grades.

VOL. 39, NO. 3, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 685

Kyoto/Stockholm group. Though rather indirectly, economi-cal and political facts thus enter calibration and complement(pure) science. Algorithms solving the separation problemare the first Jarimo Procedure (Section 3 of the SupportingInformation), its refinements (14), and the geometricalmethod by Schucht (15). The first Jarimo algorithm gives thefollowing separating values: Pandora: F: low/medium, 340km; medium/high, 8600 km. τ: low/medium, 9 d; medium/high: 50 d. Bioaccumulation: log Koa: low/medium, 3.27;medium/high, 6.89. τ: low/medium, 6.31 d; medium/high,641 d. As the filters are calibrated independently, it is hardlysurprising that the threshold value for high persistence isdifferent for Pandora (50 d) and Bioaccumulation (641 d).

With respect to the large-scale threats in question, thereare four basic outcomes. A substance can be

classified as (a) inconspicuous (two green scores) when beinginconspicuous (HPVCs); (b) inconspicuous (two green scores)though being precarious (Montreal/Kyoto, etc.); (c) precari-ous (at least one red score) though being inconspicuous(HPVCs); or (d) precarious (at least one red score) when beingprecarious (Montreal/Kyoto, etc.).

With the above calibration, the pre-screening filteringcompletely reproduces the present situation (see Table 1;the details are contained in Tables 1 and 2 of the SupportingInformation): no HPVC received a red score (which wouldstop it), and most of them (80%) even were given two greenscores (unconditional clearance). Only seven substances(20%) received a yellow score (conditional clearance),indicating that closer examination should follow. Concur-rently, each of the universally itemized Montreal/Kyoto/

FIGURE 4. Outcome of the Pandora amplifying factors τ (characteristic isotropic global half-life) and G (characteristic isotropic spatialrange) for the HPVCs and a relevant selection of the Montreal/Kyoto/Stockholm chemicals. The regulated compounds are well separatedfrom the HPVCs. The dotted line gives the theoretical maximum spatial range at given half-life, obtained by combining CIG half-life withmaximal mobility (i.e., eddy diffusion in air). Realistic (i.e., lower) mobility in water and soil leads to points exclusively at the left of thedotted line.

FIGURE 5. Outcome of the Bioaccumulation amplifying factors Koa (octanol-air partition coefficient) and CIG half-life τ for the HPVCsand a relevant selection of the Montreal/Kyoto/Stockholm chemicals. The Montreal/Kyoto/Stockholm chemicals are well separated fromthe HPVCs. (References and details of the data selection are given in Sections 2 and 4 of the Supporting Information.)


Stockholm chemicals was given one or two red scores (noclearance), completely in line with the outcome of the aboveconferences.

The calibration for the two filters was validated separatelyby the repeated 2-fold cross-validation method (16). For thispurpose the set of reference chemicals was randomly dividedinto two approximately equal-size halves, referred to as thetraining set and the test set. The filter at hand was thencalibrated using the chemicals in the training set as thereference chemicals. The desired statistics were calculatedby filtering the chemicals of the test set using the filtercalibration obtained with the training set. For both filters,the 2-fold cross-validation was repeated N ) 5 times, eachtime with a different randomly generated training set andtest set. Within each run, the following statistics for the testset were calculated: (a) number of HPVCs in green; (b)number of HPVCs in red; (c) number of regulated compounds(Montreal, etc.) in green; and (d) number of regulatedcompounds (Montreal, etc.) in red. Table 2 shows the resultsof the five test runs.

The repeated 2-fold cross-validation method was chosendue to the small size of the set of reference data. By dividing

the set of reference chemicals in two equal size halves, weobtain the largest simultaneous training set and test set.Repeating the cross-validation several times compensatesfor the small size of the data set. (As for the question whetherequivalent results can be achieved using a so-called BooleanOR operation, see Section 8.2 of the Supporting Information.)

Special Chemicals of Environmental InterestAside from the two sets of referential chemicals, a selectionof chemicals was put together showing some a priori evidenceof persistence, bioaccumulation or long-range transport.Some of these special chemicals might be regulated onnational levels. As an application of the filter series technique,they were submitted to precautionary pre-screening. Theresults are shown in Table 3 and Figures 6 and 7. The inputparameters of these chemicals and the calculated values ofthe amplifying factors are shown in Table 4.

The three stereoisomers R-HCH, â-HCH, and γ-HCH(lindane) of the insecticide hexachlorocyclohexane are themajor components of the once widely used so-called“technical HCH” (benzene hydrochloride, BHC). They arealso the most frequently detected HCH isomers in environ-mental samples and in human fat and milk. Technical HCHis now banned in most industrialized countries, where incontrast lindane, the only insecticidal isomer, is used as analmost pure substance. In the United States, the productionof lindane ceased in 1976. R-HCH and γ-HCH are almostubiquitous in environmental samples from every continent,including polar and pristine regions (17). The three chemicalsreceived a red score both in the Pandora and the Bioaccu-mulation filters. Although they are widely considered as POPsin scientific literature, the HCHs are not included in theStockholm Convention.

Endosulfan is a polychlorinated cyclodiene insecticidewhose use is permitted in most countries because of its rel-atively rapid degradation in air and water and because of itslower tendency to bioaccumulate if compared to DDT or theHCHs. It passes both filters, receiving a green score fromboth the Pandora and the Bioaccumulation filters. For anextended appraisal of endosulfan, its transformation pro-ductssendosulfan diol, endosulfan sulfate, and endosulfanendolactonesshould also be considered (i.e., endosulfanitself should be sent through the Transformation Pandorafilter). At present, however, physicochemical input param-

FIGURE 6. Outcome of the Pandora parameters τ (characteristic isotropic global half-life) and G (characteristic isotropic spatial range)for 11 chemicals of special interest (see Table 3). The dotted straight lines denote the limiting values of 9 and 50 days, respectively, forCIG half-life and 340 and 8600 km, respectively, for CIS range.

TABLE 1. Result of the Chemical Classification Problema

reference chemicals

classification HPVCs Montreal, Kyoto, Stockholm

inconspicuous (green) 80% 0%precarious (red) 0% 100%

a As green + yellow + red add up to 100%, green + red can add toless than 100%, i.e., to 80%.

TABLE 2. Statistics of 2-Fold Cross-Validationa

filter

HPVCsin

green

HPVCsin

red

regulatedcompdsin green

regulatedcompdsin red

Pandora 12.4 ( 1.82 1.0 ( 1.41 0.6 ( 0.55 19.4 ( 1.14Bioaccumulation 13.4 ( 2.51 1.6 ( 1.14 0.0 ( 0.00 21.2 ( 0.84

a Average ( SD of five different runs in absolute numbers. Theaverage number of HPVCs was 17.5, while the average number ofregulated (Montreal/Kyoto/Stockholm) compounds was 21.5.


eters for the respective transformation products are notavailable.

Carbaryl and carbofuran are the most widely usedcarbamate insecticides. Because of their rapid degradationin air and water (due to photooxidation, photolysis, hy-drolysis, and biodegradation) and their low tendency to

bioaccumulate, their potential for persistence and long-rangetransport is supposed to be low. Both chemicals pass thefilters, receiving a green score in the Pandora filter and ayellow one in the Bioaccumulation filter. For an extendedappraisal of carbaryl and carbofuran, their transformationproducts carbofuran phenol, 3-hydroxycarbofuran, and

FIGURE 7. Outcome of the Bioaccumulation parameters Koa (octanol-air partition coefficient) and CIG half-life τ for 11 chemicals of specialinterest (see Table 3). The dotted straight lines denote the limiting values of 3.27 and 6.89, respectively, for log Koa and 6.31 and 641 days,respectively, for CIG half-life.

TABLE 3. Filter Series Performance of 11 Chemicals of Special Environmental Interesta

chemicals of specialenvironmental interest

CIS range(km)

Pandora CIGhalf-life (d)

Pandorafilter log Koa

BioaccumulationCIG half-life (d)

Bioaccumulationfilter

medium:340-8600 km

medium:9-50 d

medium:3.27-6.89

medium:6.3-641 d

R-HCH medium high red high medium redâ-HCH medium high red high medium redγ-HCH medium high red high medium redendosulfan medium low green medium low greencarbaryl low low green high low yellowcarbofuran low low green high low yellowHMDS medium low green medium low greenOMCTS (D4) medium low green medium low greenDMCPS (D5) medium low green medium low greenHBB medium high red high medium redDBDE medium high red high high red

a The lower and upper limiting values of the amplifying factors are listed in the second row.

TABLE 4. Physicochemical Input Parameters and Calculated Values of the Amplifying Factors for 11 Chemicals of EnvironmentalInterest

chemicals of specialenvironmental interest

KHenry(Pa‚m3/mol) log Kow log Koa kair (1/s) kwater (1/s)

CIS range(km)

CIG persistence(d)

R-HCH 1.24E+00 3.80 7.10 1.36E-07 1.08E-07 6209 79.6â-HCH 7.53E-02 3.78 8.30 1.32E-06 6.32E-08 2169 101.7γ-HCH 5.21E-01 3.72 7.40 1.84E-07 6.32E-08 5332 113.6endosulfan 6.59E+00 3.83 6.41 8.00E-5 1.73E-06 428 0.2carbaryl 3.31E-04 2.36 9.23 5.15E-05 1.89E-06 200 4.3carbofuran 3.13E-04 2.32 9.22 2.80E-05 2.14E-06 188 3.8HMDS 4.59E+03 4.20 3.93 1.34E-06 0.00E+00 3321 6.0OMCTS (D4) 1.19E+04 5.10 4.42 9.80E-07 0.00E+00 3883 8.2DMCPS (D5) 3.10E+04 5.20 4.10 1.50E-06 0.00E+00 3139 5.4HBB 2.21E+00 6.07 9.12 1.12E-08 5.35E-07 4577 636.3DBDE 1.21E-03 5.24 11.55 1.69E-07 2.94E-08 1648 1937.2


3-ketocarbofuran should be considered. Again, the necessaryinput parameters for these compounds are not available.

On the basis of environmental monitoring and generalecotoxicological considerations, a possible role of siliconcompounds as a general new class of “environmentalchemicals” has been postulated (18). It is then interesting totest the precautionary filter procedure on some of thesecompounds, such as hexamethyldisiloxane (HMDS), octa-methylcyclotetrasiloxane (OMCTS or D4), and decameth-ylcyclopentasiloxane (DMCPS or D5), which are man-madespecial representatives of the silicones, commonly referredto as polymethylsiloxanes.

HMDS is a constituent of cosmetic and personal careproducts, hydraulic fluids, and serves as a starting materialin the production of other silicone compounds such as D4.D4 is found in soft drinks, cosmetics, detergents, and polishes,whereas D5 is an ingredient of hair care products, antiper-spirants, cosmetics, and toiletries.

Their environmental behavior and fate is characterizedby moderate volatility, low reactivity in soil and water, andan estimated high potential for bioaccumulation. Environ-mental degradation seems to occur only in the air (throughphotooxidation by hydroxyl radicals). In water and soil, theyare considered nonreactive with respect to hydrolysis andbiodegradation. They all pass the filters with a green score.

Brominated compounds such as polybrominated diphenylethers (PBDE) are widely used as flame retardants inconsumer products. They have been detected in environ-mental and human milk samples in industrialized countries,with increasing concentrations in the past decades. Two ofthem are submitted to the two-filter procedure. Hexabro-mobenzene (HBB) is used as flame retardant in polymers.It is not expected to be degraded by direct photolysis,hydrolysis, chemical oxidation, or biological activity. Somedegradation in seawater inocula was reported. Its slowdegradation in air through photooxidation by hydroxyradicals could be retarded, hexabromobenzene being ex-pected to exist solely on the particles of the troposphere(insofar preventing the reaction with hydroxy radicals). Dataon bioaccumulation are controversial, showing potentialbioaccumulation only in long-time studies. It was suggestedthat nonaccumulation was due to the size of hexabro-mobenzene, resulting in lack of membrane permeation.Hexabromobenzene is retained by both the Pandora and theBioaccumulation filters (two red scores).

Decabromodiphenyl ether (DBDE) is used as flameretardant in textiles, rubbers, and virtually every class ofpolymers (ABS, PVC, polyamides, polyesters, polyolefins, etc.)It degrades in air, water, and soil only in the presence ofsunlight. Hydrolysis and biodegradation have not beenreported. Statements concerning the potential for bioaccu-mulation are inconsistent (19). It is retained in both thePandora filter (red score) and the Bioaccumulation filter (redscore). Debromination of decabromodiphenyl ether leads tothe lower brominated congeners, tetra- to hexabrominateddiphenyls, which readily bioaccumulate. It is unclear whatproportion of the lower brominated congeners in theenvironment are breakdown products of DBDE and whatproportion comes from the commercial penta-BDE mixture.

What Has Been Achieved?First, a kind of scenario technique was used as a basis forprecautionary regulation: Scenarios for uncontrollable harmwere identified as situations to be avoided. The quantitativerepresentation of scenarios is achieved through filters. Eachfilter is defined via a small set of relevant assessmentparameters.

Then, a filter series approach was presented, which is analternative to the familiar risk-benefit valuations in situationswhere risks (i.e., probability times magnitude of adverse

effects) cannot be specified because the spectrum of theadverse effects is largely unknown. As a formal scheme thefilter series procedure is independent of particular hazards.

Next, in a case study dealing with special features of large-scale hazards of organic chemicals, two types of two-parameter filters have been constructed and suitably cali-brated. The sequence of two filters was shown to reproducein a shortcut essential results of a long and cumbersomehistorical development. (A short preview on precautionaryfilters and Pandora filtering was provided by Muller-Herold;20.) In the given context of large-scale threats, the respectiveassessment parameters play the role of amplifying factors.

The interplay of amplifying factors in the diverse threatscenarios is then taken into account using two-parameterfilters. Two-parameter filters compensate for the one-sidedness of limiting values for single assessment param-eters: In the Pandora scenario, the interplay of the twoparameters prevents concrete, bitumen, and plastics frombeing eliminated on the basis of persistence (as their mobilityis too low), and in the Bioaccumulation scenario they keepthe silicones from being eliminated on the basis of high Kow

values (as their lifetime is too short).The usual practice of defining limiting values for individual

parameters through a body of experts was then comple-mented by a kind of self-calibration of filters on the basis ofreference chemicals with broadly accepted, unequivocalinternational regulatory status. These sets of chemicals arecomparably small and cannot be easily extended withoutloss of regulatory status. Calibration and validation have toproperly deal with this situation. However, if industry findsthat thresholds thus obtained are too low or NGOs thinkthey are too high, calibration can be altered by politicaldecision makers (without questioning the precautionary pre-screening procedure as a whole.) Such new calibrations,though, would not be based on the Montreal/Kyoto/Stockholm Protocols or the U.S. HPVCs, and a new consensuswould have to be found at an international level (due to theWTO).

In cases of several scientifically equivalent methods, weconsistently chose the one that was likely to be more suitablefor public debate, as citizen participation is one of thedeclared objectives of the EU. Accordingly, closed analyticalformulas were preferred to numerical computer calculationswhenever possible. For this purpose we developed conceptssuch as CIG range, CIS lifetime, CCP cold condensationpotentials, secondary ranges, etc. The references cited andthe Supporting Information allow the interested reader toget an idea of these concepts. The mathematics for theirderivation can be found in more technical papers inEnvironmental Science and Technology and Ecological Mod-elling, respectively. Finally, a first look on a group ofnonreferential chemicals of special environmental interestlinks up to the discussion of nonreferential compounds.

To conclude, a procedure is presented that fits into thegeneral architecture of the PrecauPri model, building on thethree pillars of screening, appraisal, and management (21).The model was developed in a cooperation of social scientistsspecialized in risk and uncertainty issues, natural scientists,and a legal scholar with special expertise in risk regulation.It honors and carries forward the EU’s philosophy ofprecautionary policies and good governance and may beused as a template for precautionary risk regulation withinand beyond the EU context.

OutlookAlthough the approach to precautionary pre-screeningpresented here was developed as an answer to the needs ofregulative authorities, a far more extended application isconceivable: Ideally, a chemist designing a new compoundon paper could directly “send it through the filters”. At this


early stage, of course, the measurable input parameters haveto be replaced by theoretical or estimated values. Incombination with a suitable software solution, a firstpreliminary precautionary pre-screening could be under-taken directly after the molecule has first appeared on achemist’s drawing table. In this way, precaution could comeinto playsprior to the synthesis of one single molecule of aprecarious substance. This would be prevention at the source.

Note Added in ProofThe authors want to draw the readers’ attention to a recentlypublished paper by P. Sandin et al. (27), which opens adifferent perspective to precautionary regulation of chemi-cals.

AcknowledgmentsThis work was funded under Grant BBW-NR. 00.0487 of theSwiss Federal Office for Education and Science. The authorsare indebted to Susanna Bucher (Zurich) for technicalsupport, Martin Scheringer (Zurich) for valuable discussions,and Toni Jarimo (Helsinki) for his contributions to calibrationand validation.

AppendixSpatial Range and Persistence. A closed analytical formulafor characteristic isotropic spatial (CIS) range F has beenderived by Muller-Herold and Nickel (22):

with

For characteristic isotropic global (CIG) half-lives (τ), theformula

was used. The symbols denote the relevant unit worldparameters and the substance-related quantities.

Unit World Parameters. r ) 6381 km is the radius of theearth, which entails that πr ) 20 037 km is the maximallypossible spatial range. The calibration of the unit world’srelative compartment volumes (Vi) and eddy diffusionconstants (Di) are taken as

Substance-Related Input Quantities. ka, kw, and ks arethe degradation rate constants for air, water, and soil,respectively. If Kij ) cieq/cjeq denotes the equilibrium partitionbetween compartments i and j, then Kwa and Ksa are thewater-air and soil-air partition coefficients. Kwa and Ksa areobtained from a chemical’s Henry’s law constant KH (in Pam3 mol-1) and octanol-water partition coefficient Kow by

Following Karickhoff (23), the fraction foc of organic carbonin soil is set to 0.02. The factor 0.41 converts the octanol-water partition coefficient into the organic carbon-waterpartition coefficient Koc; Ksw is the soil-water partitioncoefficient. The Henry’s law constants are taken for distilledwater. Seawater corrections, usually giving an increase of20-40%, are neglected. As this applies to all substances, itenters the filter calibration and does not lead to arbitrarydistortions.

For legal considerations, degradation constants (ks) insoil are set to zero. As soil is a highly inhomogeneous medium,degradation constants in soil are not justiciable (i.e., liableto be tried in a court of justice). Their inclusion wouldundermine legal certainty. This choice of ks leads to slightlyincreased CIS ranges and CIG half-lives. In the context ofprecautionary pre-screening, it always leads to results onthe safe side, accordingly. As the assumption applies to allsubstances, it enters the filter calibration. Testing its influenceon the output, results have shown that in most cases it hasno visible effect. (The inhomogeneity argument is not appliedto the soil-water partition coefficient as the Karickhoffprocedure seems to be generally accepted. Soil, accordingly,enters the scenario as a lipophilic storage medium.)

Comments: CIS Ranges. The CIS ranges are based on athree compartment isotropic global unit-world scenarioinvolving the main global compartments: the troposphere,the surface water of the oceans, and the upper layer of thesoil. The concept of CIS ranges was first introduced by oneof the present authors (U.M.H.) together with M. Scheringerand M. Berg 10 years ago (24) and is preferred to simplermethods based on single media lifetimes, which can givewrong results. (For details, see Section 8.3 of the SupportingInformation.)

Comments: CIG Half-Lives. The τ formula with k∞ hasbeen used for a long time in environmental and other multi-compartment models. It is a direct consequence of the so-called instant equilibrium assumption presuming rapidequilibration of the chemical potentials of a substance in therespective compartments. A widely known application of theinstant equilibrium assumption is gas chromatography. Ithas been demonstrated by Muller-Herold (25) and Muller-Herold et al. (26) that half-lives based on the instantequilibrium assumption (i) are highly precise in the case ofrapid exchange between the compartments; and (ii) in allcases they give an upper value to real half-lives calculatedwithout the instant equilibrium assumption in more extendedmodels with corresponding input parameters. If used inprecautionary pre-screening, the formula always gives resultson the safe side, accordingly.

The CIG half-lives as used in the present setup are basedon a three-compartment isotropic global unit-world scenarioinvolving the main global compartments: the troposphere,the surface water of the oceans, and the upper layer of thesoil.

Supporting Information AvailablePhysicochemical input parameters, calculated values of theamplifying factors, and filtering results of the referencechemicals; details of the first Jarimo procedure for filtercalibration and a digression on uncertainty aspects of thepresent approach; a sketch on complementing filters (Trans-formation Pandora, Cold Condensation) and on a three-filtersequence; an outlook on REACH, the three-level testing andregulatory system presently discussed in the EU; an accountof several discussions with reviewers of this paper. Thismaterial is available free of charge via the Internet at http://pubs.acs.org.

F ) exD/k tanh(πrxk/D) exp{π/2 - 2 arctan[eπrxk/D]

sinh (πrxk/D) }D/k )

DaVa + DwKwaVw + DsKsaVs

kaVa + kwKwaVw + ksKsaVs

τ ) ln 2k∞

, k∞ )def kaVa + kwKwaVw + ksKsaVs

Va + KwaVw + KsaVs

compartment Di (km2 s-1) Vi (m3)

water (w) 0.01 233air (a) 2 200 000soil (s) 0 1

Kwa ) RT/KH

Ksw ) focKoc ) 0.02 × 0.41Kow

Ksa ) KswKwa


Literature Cited(1) Commission of the European Communities. Communication

from the Commission on the Precautionary Principle; COM 2000No. 1; Brussels, February 2000.

(2) Sandin, P. Dimensions of the precautionary principle. Hum.Ecol. Risk Assess. 1999, 5, 889-907.

(3) Ravetz, J. R. The sin of sciencesignorance of ignorance.Knowledge 1993, 15, 157-165.

(4) Weber, D. Das Ende des Fischotters in der Schweiz; SchriftenreiheUmwelt Nr. 128, Bundesamt fur Umwelt, Wald und Landschaft(BUWAL), 1990. (The End of the Otter in Switzerland; SwissFederal Environmental Protection Agency Series No. 128, 1990).From the Abstract: “The PCB- (polychlorinated biphenyl-) levelsin fish from Switzerland are generally several times higher thanthose which would allow reproduction of the otter. Switzerlandhas to be considered as a country where viable otter populationscannot exist now and in the foreseeable future because PCBlevels in fish are much too high.”

(5) Muller-Herold, U. Measures of endangerment. Geneva Pap. RiskIns. 1996, 21, 383-392.

(6) In the EU the current practice is codified in the EuropeanCommunities 1997 Technical Guidance Document in supportof Commission Directive 93/67/EEC on risk assessment for newnotified substances and in the Commission Regulation (EC)No. 1488/94 on risk assessment for existing substances,Luxembourg Office for Official Publications of the EuropeanCommunities.

(7) Quartier, R.; Muller-Herold, U. On secondary spatial ranges oftransformation products in the environment. Ecol. Modell. 2000,135, 187-198.

(8) Fenner, K.; Scheringer, M.; Hungerbuhler, K. Persistence ofparent compounds and transformation products in a level IVmultimedia model. Environ. Sci. Technol. 2000, 34, 3809-3817.

(9) Klinke, A.; Renn, O. A new approach to risk evaluation andmangement: risk-based, precaution-based and discourse-basedmanagement. Risk Anal. 2002, 22, 1071-1094.

(10) Scheringer, M.; Berg, M. Spatial and temporal range as measuresof environmental threat. Fresenius Environ. Bull. 1994, 3, 493-498.

(11) Scheringer, M. Persistence and Spatial Range of EnvironmentalChemicals: New Ethical and Scientific Concepts for Risk Assess-ment; Wiley-VCH: Weinheim, 2002.

(12) Connell, D. W. Bioaccumulation of Xenobiotic Compounds; CRCPress: Boca Raton, FL, 1990.

(13) American Chemical Society. Chem. Eng. News 1996, June 24;http://pubs.acs.org/hotartcl/cenear/960624/prod.html.

(14) Jarimo, T. Two Algorithms for the Calibration of PrecautionaryFilters; unpublished (available as pdf at http://e-collection.ethbib.ethz.ch/cgi-bin/show.pl?type ) bericht&nr ) 247).

(15) Schucht, O. Partial order theory in the assessment of environ-mental chemicals: formal aspects of a precautionary pre-

selection procedure. Ph.D. Thesis, ETH, Zurich, 2004.(16) Efron, B.; Tibshirani, R. An Introduction to the Bootstrap;

Chapman and Hall: New York, 1993.(17) Calamari, D.; Bacci, E.; Focardi, S.; Gaggi, C.; Morosini, M.; Vighi,

M. Role of plant biomass in the global environmental partition-ing of chlorinated hydrocarbons. Environ. Sci. Technol. 1991,25, 1489-1495.

(18) Reller, A.; Braungart, M.; Soth, J.; von Uexkull, O. Siliconeseinevollsynthetische Materialklasse macht Geschichte(n). GAIA 2000,9, 13-24.

(19) National Library of Medecine, National Institutes of Health.Hazardous Substances Data Bank; http://toxnet.nlm.nih.gov/cgi-bin/sis/htmlgen?HSDB; enter search term decabromobi-phenyl ether.

(20) Muller-Herold, U. Quantitative Vorsorgebewertung neuer Chemi-kalien im Hinblick auf globale Gefahrdungsszenarios. GAIA 2002,11, 46-47.

(21) Renn, O.; Dreyer, M.; Klinke, A.; Losert, C.; Stirling, A.; vanZwanenberg, P.; Muller-Herold, U.; Morosini, M.; Fisher, E. TheApplication of the Precautionary Principle in the EuropeanUnion; Final document of the EU Project Regulatory Strategiesand Research Needs to Compose and Specify a European Policyon the Application of the Precautionary Principle; http://www.sussex.ac.uk/spru/environment/precaupripdfs.html.

(22) Muller-Herold, U.; Nickel, G. A closed analytical formula for thecharacteristic spatial range of persistent organic pollutants. Ecol.Modell. 2000, 126, 191-200.

(23) Karickhoff, S. W. Semi-empirical estimation of sorption ofhydrophobic pollutants on natural sediments and soils. Chemo-sphere 1981, 10, 833-846.

(24) Scheringer, M.; Berg, M.; Muller-Herold, U. Jenseits derSchadenfrage: Umweltschutz durch Gefahrdungsbegrenzung.In Was ist ein Schaden? Zur normativen Dimension desRisikobegriffes; Berg, M., et al., Eds.; VdF: Zurich, 1994; pp 115-146.

(25) Muller-Herold, U. A simple general limiting law for the overalldecay of organic compounds with global pollution potential.Environ. Sci. Technol. 1996, 30, 586-591.

(26) Muller-Herold, U.; Caderas, D.; Funck, P. Validity of globallifetime estimates by a simple general limiting law for the decayof organic compounds with long-range pollution potential.

(27) Sandin, P.; et al. Precautionary defaults-a new strategy forchemical risk management. Hum. Ecol. Risk Assess. 2004, 10,1-18. Environ. Sci. Technol. 1997, 31, 3511-3515.

Received for review May 21, 2004. Revised manuscript re-ceived September 24, 2004. Accepted October 13, 2004.

ES049241N
