Does the Choice of NOEC or EC10 Affect the Hazardous

Page 1 of 27

Environmental Science & Technology

1

RESEARCH ARTICLE

2

Does the choice of NOEC or EC10 affect the hazardous concentration for 5% of the

3

species?

4 5

Yuichi Iwasaki*,† Kensuke Kotani‡, Shosaku Kashiwada†, and Shigeki Masunaga†

6 7

†Research Center for Life and Environmental Sciences, Toyo University, 1-1-1 Izumino,

8

Itakura, Oura, Gunma 374-0193, Japan

9

‡Graduate School of Environment and Information Sciences, Yokohama National

10

University, 79-7 Tokiwadai, Hodogaya, Yokohama 240-8501, Japan

11

†Faculty of Environment and Information Sciences, Yokohama National University, 79-7

12

Tokiwadai, Hodogaya, Yokohama 240-8501, Japan

13 14

Corresponding Author

15

*E-mail: [email protected]

16

Tel: +81-276-82-9337

17

Fax: +81-276-82-9337

ACS Paragon Plus Environment


18

Abstract

19

We evaluated if the choice of no observed effect concentration (NOEC) or 10% effect

20

concentration (EC10) affects the hazardous concentrations for 5% of the species (HC5s)

21

estimated from species sensitivity distributions (SSDs). By reviewing available literature

22

reporting NOECs and re-analyzing original toxicity data to estimate EC10s, we developed

23

two SSDs for five chemicals (zinc, lead, nonylphenol, 3,4-Dichlorobenzenamine, lindane)

24

based separately on 9 to19 EC10s and NOECs. On average, point estimates of HC5s based

25

on EC10s were 1.2 (range: 0.6–1.9) times higher than those based on NOECs. However,

26

both EC10-based and NOEC-based HC5s estimated for five substances were of the same

27

order of magnitude, and their 95% confidence intervals overlapped considerably. Thus,

28

although EC10 was chosen a representative of ECx in this study, our results suggest that

29

the choice of ECx (e.g., EC5, EC10, or EC20) or NOEC does not largely affect the

30

resulting HC5s. Therefore, use of NOECs would be acceptable particularly in regulatory

31

contexts, although the NOEC has important shortcomings and should be used with caution.

32

Keywords

33

EC5, EC20, No observed effect concentration, Species sensitivity distribution, ecological

34

risk assessment


Page 2 of 27

Page 3 of 27


35

INTRODUCTION

36

No observed effect concentration/level (NOEC or NOEL), which is the highest

37

concentration that does not cause a statistically-significant adverse effect in a toxicity test,

38

is one of the commonly-used toxic measurements in ecological risk assessments (ERAs).

39

However, because of substantial shortcomings, the NOEC and a related toxicity

40

measurement (i.e., lowest observed effect concentration, LOEC) have been heavily

41

criticized for more than 30 years1-3. Reasons include that (1) statistical significance (i.e.,

42

estimation of NOEC and LOEC) depends on experimental design, data variability, sample

43

size, effect size, and statistical analysis used (including the significance level chosen); (2)

44

statistical insignificance does not guarantee ecological, biological, or ecotoxicological

45

insignificance; and (3) the magnitude of the effect at the NOEC or LOEC is not explicitly

46

defined. Indeed, several studies demonstrated that effect sizes at NOECs are mostly

47

between 0 and 20%4, 5.

48

Recently, Landis and Chapman1 called for the ban of using NOECs and LOECs

49

and for more emphasis on concentration–response approaches. This editorial stimulated a

50

line of discussion that both agrees and disagrees with the original call6-10. The most

51

frequently-applied alternative is to estimate the x% effect concentration (ECx) based on the



52

concentration-response relationship. Values of 5–20% have been commonly suggested or

53

used as the x% (e.g.,

54

EC10) will likely be carried out at least gradually, it is important to determine if and how it

55

affects the outcome of ERAs.

11-13

). Because the replacement of the NOEC with an ECx (e.g.,

56

Species sensitivity distributions (SSD) have been frequently applied to estimate

57

the hazardous concentration for 5% of the species (HC5), which is used as a “safe”

58

concentration (e.g., for environmental water quality criteria) and a predicted no effect

59

concentration (PNEC) in ERAs14-16, mostly by applying a safe (or assessment) factor. The

60

SSD is typically estimated by fitting a statistical distribution (e.g., a log-normal

61

distribution) to multiple NOECs. Although the relationship between NOEC and ECx has

62

been evaluated in a few studies4, 5, it is uncertain how the use of ECx instead of NOEC

63

changes the resulting HC5.

64

In this study, we evaluated how the choice of EC10 or NOEC affects the resulting

65

HC5s. Although EC10 was operationally selected as a representative ECx, discussion on

66

using other values such as EC5 and EC20 can be found below. To this end, we first

67

reviewed published literature that employed the SSD approach using NOECs (rarely ECx)

68

and re-analyzed the original toxicity data to quantify concentration-response relationships


Page 4 of 27

Page 5 of 27


69

that allow us to estimate EC10. Then, we estimated two SSDs for five chemicals (zinc, lead,

70

nonylphenol,

71

[lindane]) based separately on EC10s and NOECs, and compared the resulting HC5s.

72

Through this procedure, we also estimated the magnitude of effect (in this study, predicted

73

percent reduction compared to controls) at NOECs and LOECs and the EC10/NOEC ratios

74

to investigate relationships between NOECs an EC10s. Determining whether and how the

75

use of EC10 changes the resulting HC5 provides essential knowledge for performing ERAs

76

and setting environmental quality criteria. Even if there is little impact, such information

77

can help optimize risk assessment processes.

3,4-dichlorobenzenamine

[3,4-DCA],

gamma-hexachlorocyclohexane

78 79

METHODS

80

Data

81

We referred to 15 risk assessment documents that include ecological risk assessments

82

performed by the Research Center for Chemical Risk Management, National Institute of

83

Advanced Industrial Science and Technology, Japan (https://unit.aist.go.jp/riss/crm/

84

mainmenu/e_1.html) and Versteeg et al.17 that estimated SSDs for 11 substances. From

85

these sources, we chose five substances based on two criteria: (1) the original SSD was



86

estimated based mostly on NOECs, and (2) EC10s can be determined for more than 4 to 8

87

species, which is the minimum database size that is generally accepted for regulatory

88

contexts18. Furthermore, only ecotoxicity tests for which the data could be obtained from

89

original literature, and that included more than four concentration levels including controls

90

were used to estimate EC10s. The numbers of toxicity tests analyzed for the five chemicals

91

were zinc19 (17), lead20 (19), nonylphenol21 (9), 3,4-DCA17 (16), and lindane17 (13).

92

From the original literature used for estimation of NOECs, measured

93

concentrations (if unavailable, nominal concentrations) and response variables (e.g.,

94

survival) were collected to estimate the corresponding EC10s. If requisite data were only

95

available in figures, we obtained numerical data using GSYS2.2 (Japan Nuclear Reaction

96

Data Centre (JCPRG) Digitizing software; http://www.jcprg.org/gsys/gsys-j.html). Note

97

that if the raw data were unavailable and the EC10 was estimated and used in the original

98

study, we used the reported EC10 for the present study. Although bioavailability affects a

99

metals toxicity22 we did not consider bioavailability in this study because our aim was to

100

compare HC5s based on EC10s and NOECs.

101 102

Data analysis


Page 6 of 27

Page 7 of 27


103

Concentration-response relationships were modeled with the “drc” package (version

104

2.3-96) in R version 3.0.2

105

models (log-logistic, Weibull-1, and Weibull-2 functions with 2–4 parameters), and the best

106

model with the smallest value of Akaike’s information criterion (AIC) was selected. A

107

smaller AIC value indicates a more parsimonious model that makes better prediction (see

108

e.g., ref 25).

23, 24

. Individual toxicity datasets were fitted by a total of 9

109

The best models selected for most of the datasets were log-logistic, Weibull-1, and

110

Weibull-2 models with two and three parameters. The three-parameter log-logistic model

111

(LL.3) is

112

y=

d 1 + exp(b(log(x) − log(e)))

Equation 1,

113

where x is the substance concentration, b is the slope at the concentration e, and d is the

114

maximum response value. In the two-parameter log-logistic model (LL.2), the value of d is

115

set to 1. Similarly, the three-parameter Weibull-1 model (W1.3) and three-parameter

116

Weibull-2 model (W2.3) are

117

y = d exp( − exp(b(log( x) − log(e))))

Equation 2

118

y = d (1 − exp( − exp(b(log( x ) − log(e)))))

Equation 3

119

respectively, and the value of d is 1 in the two-parameter Weibull-1 model (W1.2) and



120

two-parameter Weibull-2 model (W2.2). Further details on other models can be found in

121

Ritz et al.23 and Ritz24. By using the best models selected, for individual toxicity datasets,

122

we estimated the EC5, EC10 and EC20 and the percent reductions at the NOEC and LOEC

123

compared to the control responses.

124

We then fit the NOECs and EC10s separately to a log-normal distribution (i.e.,

125

estimated SSD), and estimated HC5s and their 95% confidence limits26. If there were

126

multiple NOECs or EC10s for a species, the geometric mean was calculated and used for

127

deriving the SSD.

128 129

RESULTS AND DISCUSSION

130

Details about the species, endpoints, test durations, response concentrations, and best-fit

131

concentration-response models in the 74 toxicity data set for the five chemicals are

132

provided in Tables S1–S5, Supporting Information. For each chemical, 9 to 19 NOECs

133

were available, although one and three EC10s were used as surrogate NOECs for zinc and

134

lead, respectively, because the NOEC and LOEC were not reported for those studies

135

(Tables S1 and S2). Major endpoints were survival (34%), reproduction (23%), growth

136

(18%), and population growth (14%; Tables S1–S5). Other endpoints were fish deformity,


Page 8 of 27

Page 9 of 27


137

algal biomass, insect emergence, time to hatch, and survival time. The numbers of species

138

ranged from 8 to 16, and fish (47%), crustaceans (28%), and algae (9%), which are

139

common taxon groups in general toxicity tests, dominated the datasets.

140 141

Percent reduction predicted at NOEC and LOEC

142

Among the five chemicals, no significant difference in the estimated magnitude of effect

143

(i.e., percent reduction) at the NOECs or LOECs was observed (Kruskal-Wallis test; p =

144

0.31 and 0.72, respectively; Figure S1 in Supporting Information). After pooling all the

145

data, the median (and range) of percent reductions estimated at the NOECs and LOECs

146

were 5.3 (0.0–67.6) and 34.0 (2.7–99.8)%, respectively (n = 69; Figure 1). This result

147

indicates that, on average, the NOEC approximately equals the EC5. The distribution of

148

percent reductions at the NOECs was not uniform, with 50% of reductions at the NOECs

149

being ≤5%, 70% of reductions being ≤10%, and 85% of reductions being ≤20%. It is rather

150

surprising that, in approximately 15% of cases, the actual effects at NOECs were larger

151

than 20%. In contrast, the percent reductions at the LOECs were distributed almost

152

uniformly from 0 to 100%. Given the reductions at the NOECs, this is simply because any

153

percent reductions can be observed at the LOECs depending on the shape of



154

concentration-response curves. These results are generally consistent with previous studies4,

155

5

156

(algae, cladocerans, and zebrafish; n = 22) were 4.7% and 3.0%, respectively5, which is

157

close to our estimate despite different toxicity datasets used.

. For example, the mean percent reductions that were observed and predicted at NOECs

158 159

Relationship between NOEC and EC10

160

The median (and range) of the distribution of EC10/NOEC ratios was 1.3 (0.1–6.5; Figure

161

S2), indicating that EC10s were usually higher than NOECs, as with obtained results above.

162

Among all the toxicity data, the EC10 was lower than the NOEC in approximately 30% of

163

the cases (Figure 2). These results are similar to Isnard et al.5, who reported the median

164

EC10/NOEC ratio was 1.32.

165 166

Influences of using EC10 on species sensitivity distributions

167

On average, HC5s estimated based on EC10s (hereafter referred to as the EC10-based

168

HC5) were 1.2 times higher than those estimated based on NOECs ((range of ratios of

169

(EC10 based HC5)/(NOEC based HC5): 0.6–1.9; Table 1 and Figure 3); only for lindane,

170

the EC10-based HC5 (0.17 µg/L) was lower than the NOEC-based HC5 (0.27 µg/L).


Page 10 of 27

Page 11 of 27


171

However, both EC10-based and NOEC-based HC5s estimated for five substances were of

172

the same order of magnitude (Table 1). The smaller range of (EC10-based

173

HC5)/(NOEC-based HC5) ratios is somewhat surprising given that the range of the

174

EC10/NOEC ratios was relatively large (0.1–6.5). This is because HC5s are less affected

175

by individual toxic values since SSDs are derived from multiple NOECs or EC10s and we

176

fitted the log-normal distribution to the data (in other words, log-transformed values of

177

NOECs and EC10s were used to derive SSDs). Overall, the close overlap of the 95%

178

confidence intervals for EC10-based HC5s with those for NOEC-based HC5s suggests that

179

the choice of EC10 or NOEC has little influence on the resulting HC5s.

180 181

Implications and limitations

182

This study provides empirical evidence that the choice of EC10 or NOEC does not largely

183

affect the resulting HC5s estimated from SSDs. If EC5s or EC20s were used instead of

184

NOECs to calculate HC5s, the resulting point estimates of HC5 would have decreased or

185

increased marginally, but the confidence intervals of HC5s derived from EC5s and EC20s

186

would have still overlapped those derived from NOECs. Indeed, because NOECs generally

187

corresponded to EC5s in the current datasets, the EC5-based HC5 was closer to the



188

NOEC-based HC5 (data not shown).

189

Other factors affect SSDs and thus HC5s: (1) how to estimate the SSD (selection

190

of statistical distributions, and use of bootstrap-based and regression-based estimations)14,

191

27, 28

192

proportions of multiple taxonomic groups included in an SSD (e.g., how to adjust the

193

proportions of algae, invertebrates, and vertebrates)29, 30. By deriving chronic SSDs for 15

194

substances, Duboudin et al.29 demonstrated that the latter two factors had more influence on

195

estimation of HC5s than the first factor. Also, the selection of toxicity data used and acute

196

versus chronic definitions can be very important31. Furthermore, there still remains

197

considerable uncertainty in the extrapolation from results of simplified laboratory toxicity

198

tests to field effects32. Therefore, we suggest that the overall influence of the choice of

199

NOEC or EC10 on the estimation of HC5 is trivial.

, (2) how to deal with multiple toxicity data for a species29, and (3) how to determine

200

We conclude that the use of NOECs would be practically acceptable for

201

calculation of HC5s, particularly in regulatory contexts, since our results indicate that

202

HC5s derived from EC10s or NOECs do not differ much. However, we emphasize that

203

NOECs have important shortcomings as discussed elsewhere (see the Introduction section).

204

We suggest that actual magnitudes of effect (e.g., percent reduction of survival,


Page 12 of 27

Page 13 of 27


205

reproduction, etc.) be provided for NOECs in case of using or reporting them. This is

206

essential for transparent ERAs, as the NOEC should not be thought of as the concentration

207

where there are no effects. In addition, sensitive traits in terms of NOECs might not

208

correspond to traits critical for population-level consequences33 that are determined

209

through nonlinear interaction among effects on individual-level traits such as survival and

210

reproduction. Thus, information on actual effect size at the NOEC is unquestionably

211

important in interpretation of the test results. Note that the use of ECx has a similar issue

212

and the x value must be carefully chosen7. Likewise, reporting uncertainty in estimates of

213

ECx is also important (see e.g., Tables S1–5) to assess the reliability although such

214

information for NOECs cannot be generally estimated. Finally, we recommend that

215

concentration-response relationships be quantitatively modeled to the extent possible1, and

216

that model parameters as well as raw data be reported so readers can generate ECx values

217

such as the EC10.

218 219 220

■ASSOCIATED CONTENT

221

Supporting Information



222

Table S1: Summary table for zinc. Table S2: Summary table for lead. Table S3: Summary

223

table for nonylphenol. Table S4: Summary table for 3,4-dichlorobenzenamine.

224

Table S5: Summary table for lindane. Figure S1: Percent reduction of endpoint response

225

estimated at the no observed effect concentration (NOEC) and the lowest observed effect

226

concentration (LOEC) for the five chemicals evaluated in this study. Figure S2:

227

Relationships between the no observed effect concentration (NOEC) and the 10% effect

228

concentration (EC10) for the five chemicals evaluated in this study. This material is

229

available free of charge via the Internet at http://pubs.acs.org

230 231

AUTHOR INFORMATION

232

Corresponding Author

233

*Dr. Yuichi Iwasaki; phone: +81-276-82-9337; e-mail: [email protected]

234

Notes

235

The authors declare no competing financial interest.

236 237

ACKNOWLEDGMENT

238

Useful comments and edits by J.S. Meyer and four anonymous reviewers are greatly


Page 14 of 27

Page 15 of 27


239

appreciated. This study is partly supported by the New Project Fund for Risk Assessments,

240

from the Ministry of Economy, Trade and Industry, Japan and a Grant-in-Aid for Strategic

241

Research Base Project for Private Universities, which is funded by the Ministry of

242

Education, Culture, Sport, Science, and Technology, Japan, 2014–2018 (Grant number

243

S14111016). YI was funded by the Japan Society for the Promotion of Science (JSPS)

244

Research Fellowship for Young Scientists.

245



246

REFREENCES

247

(1) Landis, W. G.; Chapman, P. M. Well past time to stop using NOELs and LOELs. Integr.

248

Environ. Assess. Manag. 2011, 7 (4), vi–viii.

249

(2) Chapman, P. M.; Caldwell, R. S.; Chapman, P. F. A warning: NOECs are inappropriate

250

for regulatory use. Environ. Toxicol. Chem. 1996, 15 (2), 77–79.

251

(3) Crane, M.; Newman, M. C. What level of effect is a no observed effect? Environ.

252

Toxicol. Chem. 2000, 19 (2), 516–519.

253

(4) Moore, D. R. J.; Caux, P. Y. Estimating low toxic effects. Environ. Toxicol. Chem. 1997,

254

16 (4), 794–801.

255

(5) Isnard, P.; Flammarion, P.; Roman, G.; Babut, M.; Bastien, P.; Bintein, S.; Esserméant,

256

L.; Férard, J. F.; Gallotti-Schmitt, S.; Saouter, E.; Saroli, M.; Thiébaud, H.; Tomassone, R.;

257

Vindimian, E. Statistical analysis of regulatory ecotoxicity tests. Chemosphere 2001, 45

258

(4–5), 659–669.

259

(6) Sánchez-Bayo, F. Should we forget NOECs? Integr. Environ. Assess. Manag. 2012, 8

260

(3), 564–565.

261

(7) Iwasaki, Y.; Hanson, N. Using population level consequences as a basis for determining

262

the “x” in ECx for toxicity testing. Integr. Environ. Assess. Manag. 2013, 9 (2), 344–345.


Page 16 of 27

Page 17 of 27


263

(8) Fox, D. R. Response to Landis and Chapman (2011). Integr. Environ. Assess. Manag.

264

2012, 8 (1), 4–4.

265

(9) Fox, D. R.; Billoir, E.; Charles, S.; Delignette-Muller, M. L.; Lopes, C. What to do with

266

NOECS/NOELS—prohibition or innovation? Integr. Environ. Assess. Manag. 2012, 8 (4),

267

764–766.

268

(10) Green, J. W.; Springer, T. A.; Staveley, J. P. The drive to ban the NOEC/LOEC in

269

favor of ECx is misguided and misinformed. Integr. Environ. Assess. Manag. 2013, 9 (1),

270

12–16.

271

(11) Warne, M.; van Dam, R. NOEC and LOEC data should no longer be generated or used.

272

Australasian Journal of Ecotoxicology 2008, 14 (1), 1–5.

273

(12) U. S. Environmental Protection Agency 1999 Update of ambient water quality criteria

274

for ammonia; EPA-822-R-98-008; U. S. Environmental Protection Agency: Washington,

275

DC, 1999; p 160.

276

(13) Van Der Hoeven, N.; Noppert, F.; Leopold, A. How to measure no effect. part I:

277

Towards a new measure of chronic toxicity in ecotoxicology. Introduction and workshop

278

results. Environmetrics 1997, 8 (3), 241–248.

279

(14) Newman, M. C.; Ownby, D. R.; Mézin, L. C. A.; Powell, D. C.; Christensen, T. R. L.;



280

Lerberg, S. B.; Anderson, B. Applying species-sensitivity distributions in ecological risk

281

assessment: Assumptions of distribution type and sufficient numbers of species. Environ.

282

Toxicol. Chem. 2000, 19 (2), 508–515.

283

(15) Schipper, A. M.; Posthuma, L.; de Zwart, D.; Huijbregts, M. A. J. Deriving field-based

284

species sensitivity distributions (f-SSDs) from stacked species distribution models

285

(S-SDMs). Environ. Sci. Technol. 2014, 48 (24), 14464–14471.

286

(16) Posthuma, L.; Suter, G. W. I.; Traas, T. P. Species Sensitivity Distributions in

287

Ecotoxicology. CRC Press: Boca Raton, FL, 2002.

288

(17) Versteeg, D. J.; Belanger, S. E.; Carr, G. J. Understanding single-species and model

289

ecosystem sensitivity: Data-based comparison. Environ. Toxicol. Chem. 1999, 18 (6),

290

1329–1346.

291

(18) SETAC Aquatic Dialogue Group: Pesticide Risk Assessment and Mitigation. Society

292

of Environmental Toxicology and Chemistry, SETAC Press: Pensacola, FL, USA, 1994.

293

(19) Nakanishi, J.; Naito, W.; Kamo, M. Risk Assessment Document for Zinc (Japanese

294

only). Maruzen Company: Japan, 2008.

295

(20) Nakanishi, J.; Kobayashi, N.; Naito, W. Risk Assessment Document for Lead. Maruzen

296

Company: Japan, 2006.


Page 18 of 27

Page 19 of 27


297

(21) Nakanishi, J.; Tokai, A.; Lin, B. L.; Miyamoto, K.; Ishikawa, Y. Risk Assessment

298

Document for Nonylphenol. National Institute of Advanced Industrial Science and

299

Technology: Japan, 2004.

300

(22) Di Toro, D. M.; Allen, H. E.; Bergman, H. L.; Meyer, J. S.; Paquin, P. R.; Santore, R.

301

C. Biotic ligand model of the acute toxicity of metals. 1. Technical basis. Environ. Toxicol.

302

Chem. 2001, 20 (10), 2383–2396.

303

(23) Ritz, C.; Streibig, J. C. Bioassay analysis using R. J. Stat. Softw. 2005, 12 (5), 1–22.

304

(24) Ritz, C. Toward a unified approach to dose-response modeling in ecotoxicology.

305

Environ. Toxicol. Chem. 2010, 29 (1), 220–229.

306

(25) Burnham, K. P.; Anderson, D. R. Model Selection and Multimodel Inference: A

307

Practical Information-Theoretic Approach. second ed.; Springer-Verlag: New York, 2002.

308

(26) Aldenberg, T.; Jaworska, J. S.; Traas, T. P. Normal species sensitivity distributions and

309

probabilistic ecological risk assessment. In Species Sensitivity Distributions in

310

Ecotoxicology, Posthuma, L.; Suter, G. W.; Traas, T. P., Eds. Lewis Publishers: Boca Raton,

311

FL, USA, 2002; pp 49–102.

312

(27) Hickey, G. L.; Craig, P. S. Competing statistical methods for the fitting of normal

313

species sensitivity distributions: recommendations for practitioners. Risk Anal. 2012, 32 (7),



314

1232–1243.

315

(28) Grist, E. P. M.; Leung, K. M. Y.; Wheeler, J. R.; Crane, M. Better bootstrap estimation

316

of hazardous concentration thresholds for aquatic assemblages. Environ. Toxicol. Chem.

317

2002, 21 (7), 1515–1524.

318

(29) Duboudin, C.; Ciffroy, P.; Magaud, H. Effects of data manipulation and statistical

319

methods on species sensitivity distributions. Environ. Toxicol. Chem. 2004, 23 (2), 489–

320

499.

321

(30) Hayashi, T. I.; Kashiwagi, N. A Bayesian method for deriving species-sensitivity

322

distributions: Selecting the best-fit tolerance distributions of taxonomic groups. Hum. Ecol.

323

Risk Assess. 2010, 16 (2), 251–263.

324

(31) Hahn, T.; Diamond, J.; Dobson, S.; Howe, P.; Kielhorn, J.; Koennecker, G.; Lee-Steere,

325

C.; Mangelsdorf, I.; Schneider, U.; Sugaya, Y.; Taylor, K.; Dam, R. V.; Stauber, J. L.

326

Predicted no effect concentration derivation as a significant source of variability in

327

environmental hazard assessments of chemicals in aquatic systems: An international

328

analysis. Integr. Environ. Assess. Manag. 2014, 10 (1), 30–36.

329

(32) Iwasaki, Y.; Ormerod, S. J. Estimating safe concentrations of trace metals from

330

inter-continental field data on river macroinvertebrates. Environ. Pollut. 2012, 166, 182–


Page 20 of 27

Page 21 of 27


331

186.

332

(33) Iwasaki, Y.; Hayashi, T. I.; Kamo, M. Comparison of population-level effects of heavy

333

metals on fathead minnow (Pimephales promelas). Ecotoxicol. Environ. Saf. 2010, 73 (4),

334

465–471.

335 336



337

Figure captions

338

Figure 1. Cumulative probability distribution of percent reductions (i.e., effect magnitudes)

339

observed at the no observed effect concentration (NOEC) and lowest-observed effect

340

concentration (LOEC). Data were pooled across the five chemicals evaluated in this study

341

(n = 69). Tests in which only EC10s were available or raw data was unavailable were

342

excluded.

343 344

Figure 2. Cumulative probability distribution of ratios of the 10% effect concentration

345

(EC10) to the no observed effect concentration (NOEC). Data were pooled across the five

346

chemicals evaluated in this study (n = 70). Tests in which only EC10s were available were

347

excluded.

348 349

Figure 3. NOEC- and EC10-based species sensitivity distributions for zinc, lead,

350

nonylphenol, 3,4-dichlorobenzenamine (3,4-DCA), and lindane.

351

NOEC = no observed effect concentration, EC10 = 10% effect concentration

352


Page 22 of 27

Page 23 of 27


353

Table 1. Estimated hazardous concentration for 5% of species (HC5) based on no observed

354

effect concentrations (NOECs) and 10% effect concentrations (EC10s). NOEC-based

EC10-based

Substance HC5

95% CI

HC5

95% CI

Zinc

21.5

4.2–46.3

22.3

3.8–50.3

Lead

6.61

0.99–16.9

7.99

1.34–19.61

Nonylphenol

2.75

0.10–8.78

3.48

0.11–11.38

3,4-DCA

1.12

0.01–4.59

2.12

0.03–8.74

Lindane

0.27

0.00–1.31

0.17

0.00–0.90

355

CI = confidence interval.

356

3,4-DCA = 3,4-dichlorobenzenamine



1

Page 24 of 27

TOC/Abstract Art

Fraction affected

Zinc Question: Does the choice of NOEC or EC10 affect the hazardous concentration for 5% of the species?

Lead

NOEC-based EC10-based

Nonylphenol

3,4-DCA

2 3


Lindane

Page 25 of 27

Figure 1

Cumulative probability

4


NOEC

Percent reduction 5


LOEC


Figure 2

Cumulative probability

6

EC10/NOEC 7


Page 26 of 27

Page 27 of 27

8


Figure 3

Zinc

Lead

NOEC-based

Fraction affected

EC10-based

Nonylphenol

3,4-DCA

Concentration (μg/L) Lindane

Concentration (μg/L) 9


Does the Choice of NOEC or EC10 Affect the Hazardous

Recommend Documents