Jump to main content.


Guidance for Classifying Studies Conducted Using the OECD Test Guideline 223 (TG223) (Acute Avian Oral Sequential Dose Study)

PDF Version

Guidance for Classifying Studies Conducted Using the OECD Test Guideline 223 (TG223) (Acute Avian Oral Sequential Dose Study) (PDF) (12 pp, 209K, About PDF)

On this page


Memorandum

November 9, 2011

SUBJECT: Guidance for Classifying Studies Conducted Using the OECD Test Guideline 223 (TG223) (Acute Avian Oral Sequential Dose Study)

FROM: Donald J. Brady, Director, Environmental Fate and Effects Division (7507P)

TO: Environmental Fate and Effects Division (7507P), Office of Pesticide Programs

This guidance addresses how to classify studies that were conducted using the OECD test guideline 223 - Acute Avian Oral Sequential Dose Study and is effective immediately.

Test Guideline 223 was approved for use by OECD in July 2010. While extensive computer simulations have been conducted and reported in the validation report, the Terrestrial Biology Technical Team (TBTT) conducted an evaluation focusing on the results from in vivo validation studies that were conducted using northern bobwhite quail (Colinus virginianus). Five separate laboratories tested two of the same chemicals (isazophos and MCPA) following TG223 [data from only four labs are reported in the OECD validation report (OECD 2010); however, data from all five labs were included in the TBTT analyses]. The TBTT compared the results from the TG223 validation studies to results from avian acute oral studies that have been previously submitted to the Agency for the two test chemicals following EPA's 850.2100 (public draft) guidelines.

Based on their analyses, the TBTT determined that TG223 performed generally as well as 850.2100 regarding information on the LD50 value, slope, and observation of sublethal effects (see APPENDIX A for details). The TBTT did identify some considerations regarding the current TG223 guideline that might impact the adequacy of the study for use in ecological risk assessment:

Therefore, a study conducted using TG223 may be classified as "acceptable" (i.e., adequate for use in risk assessment and fulfills a data requirement for avian acute oral toxicity data) if it is scientifically valid, is conducted using the "LD50- slope test" or "limit dose test" guidelines (the "LD50 - only test" is not adequate for fulfilling data requirements) (see APPENDIX A for details), and meets all of the following criteria:

Studies that deviate from the above criteria may be classified as "acceptable" or "supplemental" if they are considered adequate for use in risk assessment or "invalid" if they are not scientifically sound; however, consultation with the TBTT will need to confirm the classification for any TG223 study that deviates from the above criteria. If you have questions regarding this policy you may contact any of the project workgroup members listed at the end of Appendix A to this guidance memo.


Footnotes

  1. The TG223 states that the LD50-only test can be used to estimate a percentile of a species sensitivity distribution. For species sensitivity distributions, TG223 states that the following species such as mallard duck (Anas platyrhynchos), feral pigeon (Columba livia), zebra finch (Poephila guttata) and budgerigar (Melopsittacus undulatus) may be used (see APPENDIX A for details).

Reference

Top of page


Appendix A
Terrestrial Biology Tech Team Project

Project Name: Validation of TG223 (Acute Avian Oral Sequential Dose Study)

Workgroup: TG223 Workgroup (inter-divisional OPP/TBTT workgroup)

  1. Project Objective
  2. Three Guideline Options for Use Under TG223
  3. Study Validation Discussion
  4. Additional Comments
  5. Potential Impacts to Ecological Risk Assessments if TG223 Data are Used
  6. Recommendations
  1. Project Objective

    Test Guideline 223 (TG223) has been approved by OECD (July 2010). Before EFED can decide whether or not endpoints from studies conducted using the TG223 can be used in risk assessment for risk estimation, an EPA evaluation of the OECD validation effort needs to be conducted. The objective of this project is to

    1. evaluate the OECD validation effort, and

    2. produce a document to inform management's decision on the acceptability of studies conducted using the TG223 protocol in risk assessment.

    While extensive computer simulations have been conducted and reported in the validation report, our evaluation focused on the results from in vivo validation studies that were conducted using northern bobwhite quail (Colinus virginianus). Five separate laboratories tested two of the same chemicals (i.e., isazophos and MCPA) following TG223 [data from only four labs are reported in the OECD validation report (OECD 2010), however, we included data from all five labs in our analyses]. The only portion of the TG223 that was validated was the "LD50- slope test" portion of TG223 (see below). In order to evaluate the OECD validation effort and provide a recommendation(s) to management, EFED compared the results from the TG223 "validation studies" to results from avian acute oral studies that have been previously submitted to the Agency for the registration of isazophos (PC Code 295900) and MCPA (PC Code 030501) following the 850.2100 guidelines (based on the unpublished draft guidelines from Feb. 2011). The results of this comparison are discussed below.

    Top of page


  2. Three Guideline Options for Use Under TG223

    • 2.1 LD50 - slope test

      OECD recommends this "... when regulatory or other requirements determine that the slope of the dose-response curve and/or the confidence interval is required in addition to the LD50". POTENTIALLY MEETS OUR NEEDS

    • 2.2 Limit dose test

      POTENTIALLY MEETS OUR NEEDS

    • 2.3 LD50 - only test

      OECD recommends this when only the LD50 value is required and it is not recommended for use as a regulatory endpoint in risk assessments (e.g., it can be used to estimate a percentile of a species sensitivity distribution of LD50's and to provide information for labeling purposes - hazard). DOES NOT MEET OUR NEEDS FOR SCREENING-LEVEL RISK ASSESSMENT.

    Our evaluation only considers the "LD50 - slope test" and "limit dose test" portions of the TG223.

    For a side-by-side comparison of the study designs for 850.2100 and TG223, see Table 1.

    Top of page


  3. Study Validation Discussion

    • 3.1 LD50

      The LD50 is the single most important parameter that avian acute oral toxicity tests provide. 850.2100 uses at least five doses surrounding the estimated LD50 (often estimated in a range-finding test). The doses are set at intervals of 2-3X, but where the LD50 actually falls may be anywhere between the highest and lowest dose. TG223 utilizes an initial stage of four or five birds at widely spaced doses and determines dosing in subsequent stages based on results from the initial stage. In general, TG223 places doses on both sides of the estimated LD50 as determined in each stage. Thus, the LD50 resulting from TG223 is roughly centered among the test doses.

      In the validation studies, TG223 resulted in LD50s within 2-3X of the single LD50 value generated under 850.2100 available for each chemical (i.e., isazophos and MCPA). For MCPA, the LD50s in the TG223 studies ranged from 333 to 554 mg ai/kg-bw, compared to 377 mg ai/kg-bw for the 850.2100 study. For isazophos, the TG223 estimates ranged from 13.8 to 27.4 mg a.i./kg-bw, compared to 11.1 mg a.i./kg-bw for the available 850.2100.

      Confidence intervals (around the LD50) in the TG223 studies were fairly comparable across most studies for MCPA, although one lab had somewhat wider confidence intervals (CI). Results from TG223 isazophos studies varied somewhat more than those for MCPA, although the magnitude of the values is considerably lower. Two labs had appreciably wider CIs than the 850.2100 study, and one lab was unable to provide a confidence interval surrounding its estimate at all (this is the same study in which a slope could not be estimated).

      An open literature paper (Hill et al. 1984), reviewed by EFED, using the 850.2100 design is available for diazinon. In that paper, eight side by side studies were conducted using bobwhite quail resulting in fairly consistent LD50 values. LD50 values across the eight studies ranged from 13 mg a.i./kg-bw to 17 mg ai/kg-bw.

      Bottom line

      Based on these examples, TG223 performs within the general performance parameter of 850.2100 regarding the LD50 endpoint. For isazophos (which is more toxic than MCPA acid), the LD50 estimate from the 850.2100 study was lower than the LD50 estimates from all of the TG223 studies.

      Top of page


    • 3.2 SLOPE

      Slopes in the TG223 tests varied from 3.30 to 7.75 in the isazophos studies and from 3.21 to 13.2 in the MCPA studies. The 850.2100 study for isazophos reports a slope of 4.68; the MCPA study reports a slope of 11.6. The ECOFRAM document (USEPA 1999) reports that historically (39 reports across all species) the standard error as a percentage of the slope for the 850.2100 design, as an indicator of variability, ranges from 22 to 65% (median value 27%). The standard error as a percent of the slope for the 850.2100 study for isazophos is 25.3% and for MCPA is 29.6%. The standard error as a percent of the slope in the TG223 studies for isazophos range from 30.4% to 42.2% and for MCPA the range is from 31.0% to 36.7%. Of the 10 total TG223 studies, one isazophos study did not achieve a slope; the other nine TG223 studies report a slope.

      Unfortunately, side by side studies using both the TG223 and 850.2100 study designs were not performed as part of the validation effort and reports of multiple studies using the 850.2100 design on the same chemical and species are uncommon. However, the Hill et al. (1984) study reports slope estimates for each of the eight diazinon studies. Dose-response slopes ranged from 4.0 to 9.0 and standard error as a percent of slopes ranged from 26.3% to 34.5%.

      Bottom line

      TG223, when it reports a slope, performs within the general performance parameter of 850.2100. One of 10 TG223 studies did not report a slope. However, 850.2100 studies often do not report a slope.

      Top of page


    • 3.3 SUB-LETHAL EFFECTS

      Neither test is designed to quantify sub-lethal effects. Based on a comparison between the results from 850.2100 and the TG223 validation studies, TG223 provides information on sublethal effects which are generally similar to the information provided by 850.2100. Differences between the tests are described below.

      Only one or two birds were observed at any one dose for the TG223 studies. The reviewer, therefore, must make calls of where to place the NOAEL- and LOAEL-based effects observed on just one or two birds. Effects may be missed if a bird tested at a dose near the LOAEL happens to be not sensitive. Additionally, because only one or two birds are tested at each dose, you cannot do hypothesis testing (ANOVA) to identify statistical significance of body weight changes and food consumption at specific doses. The data can be visually inspected to determine where these measurements seem to begin to become significantly different than the control, but this method is very subjective and potentially inaccurate. A regression analysis could be used to determine statistically significant differences in body weight across treatment levels; however, it should be noted that this approach would be different than the one currently used, where, if the data allows, a statistical NOAEL is determined. However, this ability to statistically determine a NOAEL is not frequently available. In addition, if a regression analysis were to be performed to establish biological significance for body weight change and food consumption, the ECx value that would be used to signify biological significance would have to be established (since one has not been established to date).

      Because birds are tested at many different doses, however, there may be more information to determine at what dosage sublethal effects first begin to take place. The sequential dosing results in testing many doses that are generally evenly placed around the LD50 (i.e., approximately one-half of the tested birds will receive a dose less than the estimated LD50 value and one-half of the tested birds will receive a dose greater than the estimated LD50 value). This results in having several doses where the birds survive, allowing for observations of sublethal effects. The placement of doses in 850.2100, on the other hand, may not be centered evenly around the LD50 determined in the study. If the doses happen to be mostly below the LD50, the study may provide good information on sublethal effects. However, if the doses happen to be mostly above the LD50, the study provides little information on sublethal effects because mortality occurring at lower doses may confound results.

      Another concern with both the TG223 and 850.2100 guidelines regarding sub-lethal effects is that observations and terminology of behavioral signs of toxicity vary greatly from lab to lab. It would be good if guidance was provided to make these observations more consistent. Also, for TG223 it would be helpful if the results of sublethal effects were sorted by dose level, rather than grouped by stages, to make interpretation easier.

      Botttom line

      TG223 provides information on sublethal effects which are generally similar to the information provided by 850.2100. There are fewer birds observed in TG223 than in 850.2100, and, therefore, there is a greater chance of missing rare effects. This makes visually determining the rate at which sublethal effects first occur somewhat more difficult. However, because birds are tested at many different doses in TG223, there is potentially more information to hone in on where sublethal effects first begin to take place.

      Top of page


    • 3.4 LIMIT TEST

      The limit test for TG223 utilizes five birds which are tested up to 2,000 mg a.i./kg-bw. By contrast, 850.2100 recommends the use of 10 birds tested up to 2,000 mg a.i./kg-bw or the environmentally relevant concentration (whichever is greater). The TG223 limit test is aborted if there is any mortality, while the 850.2100 limit may continue if there is only a single mortality (in the control). Using a five bird limit test with no mortality, there is 95% probability that the true mortality at the limit dose is no more than 45%. By contrast, in a 10 bird limit with no mortality, there is a 95% probability that true mortality at the limit is no more than 26%. A ten bird limit test with one mortality provides a 95% probability that true mortality at the limit is no more than 39%.

      Bottom line

      For both studies, a successful limit test provides a high degree of certainty that the LD50 is above the limit dose tested. However, 850.2100 provides a greater margin of error. If the birds are not tested up to environmentally relevant concentrations (in either study design), the potential for effects at the relevant concentrations will not be known.

    Top of page


  4. Additional Comments

    • As stated in TG223, species with low background mortality and a low propensity to regurgitate are preferred. Species that frequently match these criteria are bobwhite quail and Japanese quail (Coturnix japonica)1. Because the background mortality of other laboratory species may not be well known, the TG223 testing design may not be appropriate for use with other species. This is less of an issue for the 850.2100 guidelines because more birds are tested. Due to these issues, and because the validation studies were only conducted using bobwhite quails, our current evaluation can only be based on the use of TG223 using bobwhite quail.

    • As stated in the TG223, TG223 is easiest to apply to chemicals that produce death within a few days. If the time until starting the next stage needs to be extended beyond three days due to delayed mortality (or effects), that can create situations where birds involved in the testing are weeks apart in age and may be even from different hatches.

    • Under TG223, only 5 control birds are required. If there is one incidental death in the initial five-bird control group, five more control birds are added to the test. EFED has concerns about this since the TG223 statement, "... five more control birds are added to the test" (TG223 paragraph 12) is vague and there is a potential for adding additional "control" birds any time during the study (even on the last day of the study, rendering the added "controls" essentially useless).

    • Because of the complexity of the TG223 testing design, training of EFED science staff and contractor staff will be required before studies using the TG223 can be adequately reviewed.

    • Because of the sequential design of TG223, the data cannot be analyzed using our current methods. A computer program, SEquential DEsign Calculator (SEDEC) (an Excel calculator), was developed and validated by the OECD/EPA TG223 team to determine the placement of doses during testing and to estimate the LD50, slope and confidence limits for TG223 studies. SEDEC is available on the OECD website and requires Microsoft® Excel for Windows 2000 (or later versions) to run.

    • Additionally, because of the complexity (and as stated by the labs that did the validation studies), there is a greater chance of making an error during the testing. However, as the testing labs become more familiar with the design, this problem is likely to attenuate.

    Top of page


  5. Potential Impacts to Ecological Risk Assessments if TG223 Data are Used

    • The LD50 values should be comparable to LD50 values obtained under the 850.2100 guidelines. For isazophos, the LD50 estimate from the 850.2100 study was lower than the LD50 estimate from all of the TG223 studies. Whether or not LD50 values for TG223 will be consistently higher than LD50 values from 850.2100 studies across chemicals cannot be known without additional data.

    • Regarding sublethal effects, since there are fewer birds tested under TG223 there is a greater chance of missing rare effects and visually determining the results of the data is more difficult. This may increase inconsistencies across study reviewers.

    • Regarding Limit Tests, both TG223 and 850.2100 provide a high degree of certainty that the LD50 is above the limit dose tested. However, with the fewer number of birds included in TG223 limit tests (5 vs 10 in the 850.2100), true mortality could be as high as 45% at the tested concentration if no birds die in the study, based on a one-tailed binominal analysis (95% probability level).

    • Unlike in the 850.2100 guidelines, TG223 does not specify that the highest dose needs to be at least as high as the environmentally relevant concentration. If birds are not tested up to the environmentally relevant concentration in TG223, the potential effects at the expected concentrations will not be known.

    • Because TG223 deviates substantially from the 850.2100 guidelines, EFED and contractor staff will need to be trained on the study design and the use of new statistical analyses and tools before being able to adequately review studies submitted under TG223.

    Top of page


  6. Recommendations

    • If a study conducted using TG223 is scientifically valid, the results from the "LD50-slope test" conducted using bobwhite quail could be used to fulfill the data requirement for an avian acute oral study for chemicals. Studies conducted using other test species may also be classified as acceptable if the laboratory conducting the study can demonstrate that the background mortality for the test species is ≤ the background mortality for bobwhite quail in their laboratory and the test animals do not regurgitate during the study.

    • TG223 should not be used with chemicals that cause delayed mortality. (Criteria for determining which chemicals have "delayed mortality" will be determined in a separate, follow-up project).

    • If a study conducted using TG223 is scientifically valid, the results from the "limit test" conducted using bobwhite quail could be used to fulfill one of the data requirements for an avian acute oral study if it is conducted at 2,000 mg a.i./kg-bw or the environmentally relevant concentration (whichever is greater).

    • We recommend using the same validity requirements as those outlined in the 850.2100 guidelines. For example, a study following the TG223 or 850.2100 guidelines should be invalidated if >10% of the controls die (the results from the study would be questionable if additional control birds are added during the course of the study). It should be noted that with a control of 5 birds, > 10% mortality occurs with one mortality.

    • All of the TG223 LD50's for isazophos were higher than the 850.2100 LD50 value for this chemical. However, because we only have limited comparative data, a potential bias toward higher LD50 values for TG223 when compared to 850.2100 cannot be confirmed or rejected at this time. Therefore, we recommend an ongoing comparison of LD50 values using both guidelines (if possible) to explore the potential for a bias toward higher or lower LD50 values using TG223 when compared to 850.2100.

Top of page


Table 1
A General Comparison Between 850.2100 and TG223
General Design (Definitive test) 850.2100TG223COMMENT(S)
Test species Northern bobwhite quail (Colinus virginianus); mallard duck (Anas platyrhynchos); passerine (TBD) Preferred species: Northern bobwhite quail (Colinus virginianus); Japanese quail (Coturnix japonica) For TG223, captive bred species with low background mortality and a low propensity to regurgitate are preferred.
Number of Birds/Dose ≥ 10 birds/dose1 or 2/dose, depending on effects observed Test more birds at fewer concentrations using 850.2100
Number of Doses ≥ 5≥ 10 depending on the number of stages and effects observed Test more birds at fewer concentrations using 850.2100
Dosing Regime All dosed on day 0 and observed for 14 days Sequentially dosed at three day intervals (up to 4 stages); 9 days between first dose and last dose administered (if 4 stages) - may be longer (see below); 14-day observation period for each dose group Dosing regime more complex with TG223
Duration of Test 14 days ≥ 23 days (if reach 4th stage; based on a 3-day dose interval) In the TG223: "In some cases, it may be necessary to wait up to 14 days before moving to the next stage. If the observation period prior to calculating the working LD50 is extended for this stage, it is extended for all subsequent stages. If there is evidence of delayed effects after 14 days, the length of the stage may be further extended."
Age of Birds ≥ 16 weeks (same hatch) Mature plumage (may be from different hatch) In the TG223: "if, upon the start of a new stage, birds from a hatch different from the one used to initiate the test are used, control birds from the second hatch should be included on the day that these birds are started on test;
Controls 10 5 (If there is one incidental death in the initial five-bird control group, five more control birds are added to the test); birds are also added if birds from a second hatch are used In the TG223 there is no guidance for when the 5 "additional control" birds can be added; the value of the control birds is decreased if they are added late in the study.
LD50 Value At least 5 doses at geometric intervals. LD50 estimate may be towards one end of the distribution. Due to this, probit not always possible, resulting in use of moving average or binomial estimates Doses tend to be placed around the initial estimate of the LD50 resulting in an LD50 estimate roughly centered among the doses. Probit determines LD50 estimate. TG223 at least as reliable as 850.2100
Slope Probit analysis is not always possible due to lack of partial mortalities. If probit cannot be used, no slope is available. Should result in an estimate of slope more frequently than 2100 TG223 more likely than 850.2100 to calculate a slope.
Sublethal Effects Not designed to evaluate sublethal effects therefore does not normally provide a statistically valid NOEL for sublethal effects. Not designed to evaluate sublethal effects therefore does not normally provide a statistically valid NOEL for sublethal effects. Effects on body weight can more easily be visually determined under the 850.2100 guideline.
Signs of Intoxication Report signs of intoxication and cessation of abnormal behavior by dosage level and by day; estimate number of birds exhibiting such signs for each dosage level. Report signs of intoxication and cessation of abnormal behavior by dosage level and by day; estimate number of birds exhibiting such signs for each dosage level. None
Body Weight Individual body weights of birds are recorded at the time of calculating the dosage and weekly thereafter until the test is concluded. Birds are weighed before dosing and then at 3, 7 and 14 days after dosing. More measurements recommended under TG223
Food Consumption Measure and record food consumption at least weekly in each pen throughout the test. Measured daily until day 3, then for periods 3-7 and 7-14 days after dosing. More measurements recommended under TG223
Regurgitation Birds are closely monitored for 60 to 120 min. after dosing Birds are observed during the first two hours after dosing for regurgitation None
Validity Requirements A test is not acceptable if > 10 % of the control birds die during the test period The test is invalid if there is one non-incidental death or more than one death from any other cause. If there is one incidental death in the initial five-bird control group, five more control birds are added to the test Under TG223, additional control birds can be added at any time during the study
Complexity None Study is more complex than 850.2100 Due to the complexity of TG223 when compared to 850.2100, there is a greater chance of errors in testing and analysis with the TG223 studies; TG223 will require training of EPA staff.

Top of page


Footnotes

  1. The TG223 states that the LD50-only test can be used to estimate a percentile of a species sensitivity distribution. For species sensitivity distributions, TG223 states that the following species such as mallard duck (Anas platyrhynchos), feral pigeon (Columba livia), zebra finch (Poephila guttata) and budgerigar (Melopsittacus undulatus) may be used.

Top of page


References

Top of page


Workgroup members

Brian Anderson (EFED), Management Lead
Amy Blankinship (EFED), Tech Team Lead
Shannon Borges (BPPD)
Kevin Costello (PRD)
Mary Frankenberry (EFED)
Brian D. Kiernan (EFED), Co-chair
Nick Mastrota (EFED)
Kim Nesci (RD)
Ed Odenkirchen (EFED), Senior Technical Consultant
Melissa Panger (EFED), Co-chair
Michael Wagman (EFED), Tech Team Lead

Top of page


Publications | Glossary | A-Z Index | Jobs


Local Navigation


Jump to main content.