ISO TS 20281-2006 Water quality - Guidance on statistical interpretation of ecotoxicity data《水质 生态毒性数据的统计说明指南》.pdf

1、 Reference number ISO/TS 20281:2006(E) ISO 2006TECHNICAL SPECIFICATION ISO/TS 20281 First edition 2006-04-01 Water quality Guidance on statistical interpretation of ecotoxicity data Qualit de leau Lignes directrices relatives linterprtation statistique de donnes cotoxicologiques ISO/TS 20281:2006(E)

2、 PDF disclaimer This PDF file may contain embedded typefaces. In accordance with Adobes licensing policy, this file may be printed or viewed but shall not be edited unless the typefaces which are embedded are licensed to and installed on the computer performing the editing. In downloading this file,

3、 parties accept therein the responsibility of not infringing Adobes licensing policy. The ISO Central Secretariat accepts no liability in this area. Adobe is a trademark of Adobe Systems Incorporated. Details of the software products used to create this PDF file can be found in the General Info rela

4、tive to the file; the PDF-creation parameters were optimized for printing. Every care has been taken to ensure that the file is suitable for use by ISO member bodies. In the unlikely event that a problem relating to it is found, please inform the Central Secretariat at the address given below. ISO 2

5、006 All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying and microfilm, without permission in writing from either ISO at the address below or ISOs member body in the co

6、untry of the requester. ISO copyright office Case postale 56 CH-1211 Geneva 20 Tel. + 41 22 749 01 11 Fax + 41 22 749 09 47 E-mail copyrightiso.org Web www.iso.org Published in Switzerland ii ISO 2006 All rights reservedISO/TS 20281:2006(E) ISO 2006 All rights reserved iii Contents Page Forewordxii

7、Introduction xiii 1 Scope 1 2 Normative references 1 3 Terms and definitions .1 4 General statistical principles8 4.1 Different statistical approaches .8 4.1.1 General8 4.1.2 Hypothesis-testing methods 8 4.1.3 Concentration-response modelling methods .10 4.1.4 Biology-based methods 11 4.2 Experiment

8、al design issues .11 4.2.1 General11 4.2.2 NOEC or EC x : Implications for design.12 4.2.3 Randomization .12 4.2.4 Replication13 4.2.5 Multiple controls included in the experimental design13 4.3 Process of data analysis.14 4.3.1 General14 4.3.2 Data inspection and outliers.14 4.3.3 Data inspection a

9、nd assumptions .15 4.3.3.1 Scatter .15 4.3.3.2 Heterogeneous variances and distribution .15 4.3.3.3 Heterogeneous variances and true variation in response.16 4.3.3.4 Consequences for the analysis 16 4.3.4 Transformation of data16 4.3.5 Parametric and non-parametric methods .17 4.3.5.1 General 17 4.3

10、.5.2 Parametric methods.17 4.3.5.3 Generalized linear models (GLMs) .18 4.3.5.4 Non-parametric methods.18 4.3.5.5 How to choose?18 4.3.6 Pre-treatment of data.19 4.3.7 Model fitting19 4.3.8 Model checking20 4.3.8.1 Analysis of residuals .20 4.3.8.2 Validation of fitted dose-response model .21 4.3.9

11、Reporting the results.21 5 Hypothesis testing.21 5.1 Introduction21 5.1.1 General21 5.1.2 NOEC: What it is, and what it is not.25 5.1.3 Hypothesis used to determine NOEC25 5.1.3.1 Understanding the question to be answered 25 5.1.3.2 One-sided hypothesis26 5.1.3.3 Two-sided trend test 26 5.1.3.4 Tren

12、d or pair-wise test.26 5.1.4 Comparisons of single-step (pair-wise comparisons) or step-down trend tests to determine the NOEC28 ISO/TS 20281:2006(E) iv ISO 2006 All rights reserved5.1.4.1 General discussion . 28 5.1.4.2 Single-step procedures. 28 5.1.4.3 Step-down procedures 29 5.1.4.4 Deciding bet

13、ween the two approaches . 30 5.1.5 Dose metric in trend tests 31 5.1.6 Role of power in toxicity experiments 31 5.1.7 Experimental design . 32 5.1.8 Treatment of covariates and other adjustments to analysis 33 5.2 Quantal data (e.g. mortality, survival). 34 5.2.1 Hypothesis testing with quantal data

14、 to determine NOEC values . 34 5.2.2 Parametric versus non-parametric tests 35 5.2.2.1 Basis . 35 5.2.2.2 Single-step procedures. 36 5.2.2.3 Step-down procedures 36 5.2.2.3.1 Choice of step-down procedure. 36 5.2.2.3.2 Test for monotone dose response 36 5.2.2.3.3 Analysing the monotonic response for

15、 quantal data Step-down procedure . 37 5.2.2.3.4 Possible modifications of the step-down procedure. 37 5.2.2.4 Alternative procedures . 37 5.2.2.4.1 Parametric and non-parametric procedures. 37 5.2.2.4.2 Pair-wise ANOVA-based methods . 38 5.2.2.4.3 Jonckheere-Terpstra trend test38 5.2.2.4.4 Poisson

16、tests . 38 5.2.2.5 Assumptions of methods for determining NOEC values 38 5.2.3 Additional information 39 5.2.4 Statistical items to be included in the study report. 40 5.3 Hypothesis testing with continuous data (e.g. mass, length, growth rate) to determine NOEC 40 5.3.1 General . 40 5.3.2 Parametri

17、c versus non-parametric tests 41 5.3.3 Single-step (pair-wise) procedures . 42 5.3.3.1 General . 42 5.3.3.2 Dunnetts test. 42 5.3.3.3 Tamhane-Dunnett test. 42 5.3.3.4 Dunns test . 42 5.3.3.5 Mann-Whitney test. 43 5.3.4 Step-down trend procedures . 43 5.3.5 Determining the NOEC using a step-down proc

18、edure based on a trend test 43 5.3.5.1 General . 43 5.3.5.2 Preliminaries 43 5.3.5.3 Step-down procedure 43 5.3.5.3.1 Preferred approach . 43 5.3.5.3.2 Williams test 44 5.3.5.3.3 Jonckheere-Terpstra test 44 5.3.6 Assumptions for methods for determining NOEC values 44 5.3.6.1 Small samples Massive ti

19、es. 44 5.3.6.2 Normality 45 5.3.6.3 Variance homogeneity 45 5.3.7 Operational considerations for statistical analyses 46 5.3.7.1 Treatment of experimental units 46 5.3.7.2 Identification and meaning of outliers 46 5.3.7.3 Multiple controls 46 5.3.7.4 General . 47 5.4 Statistical items to be included

20、 in the study report. 47 6 Dose-response modelling 48 6.1 Introduction . 48 6.2 Modelling quantal dose-response data (for a single exposure duration) . 49 6.2.1 General . 49 6.2.2 Choice of model 50 ISO/TS 20281:2006(E) ISO 2006 All rights reserved v 6.2.2.1 General 50 6.2.2.2 Probit model .51 6.2.2

21、.3 Logit model.53 6.2.2.4 Weibull model.54 6.2.2.5 Multi-stage models.55 6.2.2.6 Definitions of EC 50and EC x .55 6.2.3 Model fitting and estimation of parameters 56 6.2.3.1 Software and assumptions .56 6.2.3.2 Response in controls.56 6.2.3.3 Analysis of data with various observed fractions at each

22、dose group57 6.2.3.4 Analysis of data with one observed fraction at each dose group 58 6.2.3.5 Extrapolation and EC x .58 6.2.3.6 Confidence intervals58 6.2.4 Assumptions 59 6.2.4.1 General 59 6.2.4.2 Statistical assumptions .59 6.2.4.3 Evaluation of assumptions .59 6.2.4.3.1 Evaluation of basic ass

23、umptions .59 6.2.4.3.2 Evaluation of the additional assumption.59 6.2.4.4 Consequences of violating the assumptions60 6.2.4.4.1 Consequences of violating basic assumptions60 6.2.4.4.2 Consequences of violating the additional assumption .60 6.3 Dose-response modelling of continuous data (for a single

24、 exposure duration) 60 6.3.1 Purpose.60 6.3.2 Terms and notation60 6.3.3 Choice of model.61 6.3.3.1 First distinctions 61 6.3.3.2 Linear models.62 6.3.3.3 Threshold models 62 6.3.3.4 Additive versus multiplicative models.63 6.3.3.5 Models based on “quantal” models.63 6.3.3.6 Nested non-linear models

25、 .64 6.3.3.7 Hill model 67 6.3.3.8 Non-monotone models 67 6.3.4 Model fitting and estimation of parameters 68 6.3.4.1 Software and assumptions .68 6.3.4.2 Response in controls.68 6.3.4.3 Fitting the model assuming normal variation .68 6.3.4.4 Fitting the model assuming normal variation after log-tra

26、nsformation .68 6.3.4.5 Fitting the model assuming normal variation after other transformations69 6.3.4.6 No individual data available69 6.3.4.7 Fitting the model using GLM.69 6.3.4.8 Covariates .70 6.3.4.9 Heterogeneity and weighted analysis71 6.3.4.10 Confidence intervals73 6.3.4.11 Extrapolation

27、73 6.3.4.12 Analysis of data with replicated dose group.73 6.3.5 Assumptions 74 6.3.5.1 General 74 6.3.5.2 Statistical assumptions .74 6.3.5.3 Additional assumption 74 6.3.6 Evaluation of assumptions .75 6.3.7 Consequences of violating the assumptions .75 6.3.7.1 Basic assumptions 75 6.3.7.2 Additio

28、nal assumption 76 6.4 To accept or not accept the fitted model? 77 6.4.1 Can the fitted model be accepted and used for its intended purpose?.77 6.4.2 Is the model in agreement with the data? .77 6.4.3 Do the data provide sufficient information for fixing the model? 77 ISO/TS 20281:2006(E) vi ISO 200

29、6 All rights reserved6.5 Design issues 81 6.5.1 General . 81 6.5.2 Location of dose groups 81 6.5.3 Number of replicates 81 6.5.4 Balanced versus unbalanced designs82 6.6 Exposure duration and time. 82 6.6.1 General . 82 6.6.2 Quantal data. 82 6.6.3 Continuous data 83 6.6.3.1 General . 83 6.6.3.2 In

30、dependent observations in time . 83 6.6.3.3 Dependent observations in time 85 6.7 Search algorithms and non-linear regression . 85 6.8 Reporting statistics. 86 6.8.1 Quantal data. 86 6.8.2 Continuous data 87 7 Biology-based methods . 87 7.1 Introduction . 87 7.1.1 Effects as functions of concentrati

31、on and exposure time 87 7.1.2 Parameter estimation 89 7.1.3 Outlook. 89 7.2 Modules of effect-models. 90 7.2.1 General . 90 7.2.2 Toxico-kinetic model 91 7.2.3 Physiological targets of toxicants. 91 7.2.4 Change in target parameter . 92 7.2.5 Change in endpoint. 93 7.3 Survival 93 7.3.1 Relationship

32、 between hazard rate and survival probability . 93 7.3.2 Assumptions of survival probability at any concentration of test compound . 94 7.3.3 Summary 94 7.4 Body growth 97 7.4.1 Routes for affecting body growth 97 7.4.2 Assumptions 97 7.4.3 Von Bertalanffy growth curve 98 7.5 Reproduction. 99 7.5.1

33、Routes that affect reproduction. 99 7.5.2 Assumptions 100 7.5.3 Implication . 100 7.6 Population growth. 101 7.6.1 General . 101 7.6.2 Assumptions 101 7.7 Parameters of effect models 103 7.7.1 General . 103 7.7.2 Effect parameters 103 7.7.2.1 Toxicity and dynamic parameters . 103 7.7.2.2 Killing rat

34、e, b k . 104 7.7.3 Discussion . 105 7.7.4 Eco-physiological parameters. 107 7.8 Recommendations 109 7.8.1 Goodness of fit 109 7.8.2 Choice of modes of action . 110 7.8.3 Experimental design . 110 7.8.4 Building a database for raw data. 110 7.9 Software support. 110 7.9.1 General . 110 7.9.2 DEBtox .

35、 111 7.9.3 DEBtool 111 ISO/TS 20281:2006(E) ISO 2006 All rights reserved vii 8 List of existing guidelines with references to the subclauses of this Technical Specification.112 Annex A (informative) Analysis of an “acute immobilization of Daphnia magna” data set (OECD GL 202 ISO 6341) using the thre

36、e presented approaches115 A.1 Data set (see Table A.1) .115 A.2 Examples of data analysis using hypothesis testing (NOEC determination) .115 A.3 Example of data analysis by dose-response modelling120 A.4 Example of data analysis using DEBtox (biological methods).125 Annex B (informative) Analysis of

37、 an “algae growth inhibition” data set using the three presented approaches.127 B.1 General127 B.2 Examples of data analysis using hypothesis testing (NOEC determination) .128 B.3 Example of data analysis by dose-response modelling135 B.4 Examples of data analysis using DEBtox (biological methods).1

38、39 Annex C (informative) Analysis of an “Daphnia magna reproduction” data set (OECD GL 211 ISO 10706) using the three presented approaches142 C.1 Examples of data analysis using hypothesis testing (NOEC determination) .143 C.2 Example of data analysis by dose-response modelling148 C.3 Examples of da

39、ta analysis using DEBtox (biological methods).155 Annex D (informative) Analysis of a “fish growth” data set (OECD GL 204/215 ISO 10229) using the three presented approaches 160 D.1 Data set .160 D.2 Examples of data analysis using hypothesis testing (NOEC determination) .162 D.3 Example of data ana

40、lysis by dose-response modelling172 D.4 Examples of data analysis using DEBtox (biological methods).177 Annex E (informative) Description and power of selected tests and methods.180 E.1 Description of selected methods for use with quantal data .180 E.2 Power of the Cochran-Armitage test .189 E.3 Des

41、cription of selected tests for use with continuous data .198 E.4 Power of step-down Jonckheere-Terpstra test 218 Annex F (informative) Annex to Clause 7 “Biology-based methods”231 F.1 General231 F.2 Effects on survival.231 Bibliography 237 Figure 1 Conceptual illustration of accuracy and precision.

42、2 Figure 2 Illustration of a concentration-response relationship and of the estimates of the EC xand NOEC/LOEC . 5 Figure 3 Analysis of quantal data: Methods for determining the NOEC . 23 Figure 4 Analysis of continuous data: Methods for determining the NOEC 24 Figure 5 Analysis of continuous data:

43、Methods for determining the NOEC (continued) 24 Figure 6 Flow-chart for dose-response modelling. 50 ISO/TS 20281:2006(E) viii ISO 2006 All rights reservedFigure 7 Probit model fitted to observed mortality frequencies (triangles) as a function of log-dose .52 Figure 8 Logit model fitted to mortality

44、dose-response data (triangles) 53 Figure 9 Weibull model fitted to mortality dose-response data (triangles) 54 Figure 10 Logit model fitted to mortality dose-response data (triangles), with background mortality .57 Figure 11 Two members from a nested family of models fitted to the same data set.66 F

45、igure 12 Cholinesterase inhibition as a function of dose at three exposure durations71 Figure 13 Relative liver masses against dose, plotted on log-scale .72 Figure 14 Dose-response model fitted to the data of Figure 13, showing that the heterogeneous variance was caused by males (triangles) and fem

46、ales (circles) responding differently to the chemical 73 Figure 15 Model fitted to dose-response data with and without an outlier in the top dose .76 Figure 16 Two different models (both with four parameters) fitted to the same data set resulting in similar dose-response relationships79 Figure 17 Tw

47、o data sets illustrating that passing a goodness-of-fit test is not sufficient for accepting the model.80 Figure 18 Observed biomasses (marks) as a function of time, for nine different concentrations of Atrazine.84 Figure 19 Growth rates as derived from biomasses observed in time (see Figure 18) at

48、nine different concentrations (including zero), with the Hill model fitted to them84 Figure 20 Estimated growth rates as a function of (log-)concentration Atrazine 85 Figure 21 Fluxes of material and energy through an animal, as specified in the DEB model.92 Figure 22 Time and concentration profiles

49、 of the hazard model, together with the data of Figure 2795 Figure 23 Time and concentration profiles for effects on growth of Pimephalus promelas via an increase of specific maintenance costs by sodium pentachlorophenate (data by Ria Hooftman, TNO-Delft)98 Figure 24 Time and concentration profiles for effects on growth of Lumbricus rubellus via a decrease of assimilation by copper chloride (data from Klok and de Roos 1996) 99 Figure 25 Effects of cadmium on the reproduction of Daph