(b) Probit 9, dose-mortality tests and confirmatory tests

75. The United States noted that dose-mortality tests were a critical tool to determine what commercial treatment might be effective. The dose-mortality test result established a range of treatments for varieties from which scientists could compare and estimate a final treatment for the product as a whole. The highest minimum dose observed in the dose-mortality tests that scientists believe would achieve the level of protection required by Japan (probit 9) was supplemented by 10-20 per cent in the second stage of testing, the confirmatory tests, to account for all sources of variation in the dose-mortality tests. Thus it was the confirmatory test that was the relevant indicator of efficacy of treatment.
76. The primary reason for the 10-20 per cent (buffer) increase of dose in confirmatory tests, as well as the reason why scientists did not rely on dose-response results to establish a commercial quarantine treatment, was that not every replicate of a dose-mortality test would exactly mirror another. Such factors as experimental error, physical condition of the fruit, sorption of the fumigant by packing material, and load of fruit in the chamber could account for differences in results. Phytotoxic ⁶⁴ effects and effects on residue levels would ordinarily establish the upper bound for a proposed buffer.
77. The United States noted that dose-response data could vary from test to test with the same species of insect in the same variety. It was well established that insect susceptibility to insecticidal treatment differed greatly among individual insects, and that the response of insect populations varied. Since each test necessarily required new insects, the response of test insects would differ naturally among tests and that accounted in part for the observed variability in dose-mortality test results. ⁶⁵ With small scale dose-mortality tests, using limited numbers of pests, it was unlikely that variability of insects relative to susceptibility to the fumigant would be fully represented. Confirmatory tests, however, contained sufficient numbers of insects (10,000 to 30,000) to ensure representation of insects at all levels of MB susceptibility.
78. Thus, because of natural variation of the test insect population, as well as other factors that ensured that no one dose-mortality test would be exactly the same as another, it was impossible to conclude that differences in variety were the explanation for variations in dose-response results. The United States argued that Japan's own acceptance of varieties that exhibited differences in dose-response results established that Japan recognized effective quarantine treatment always accommodated some differences in dose-mortality testing.
79. Confirmatory tests established the efficacy of treatment for a product because they took into consideration all of the sources of variation that could be attributed to the smaller scale tests - most notably the variation in a pest population, and the experimental error that was bound to occur from one small-scale test to the next. A confirmatory test would show if a treatment were too low because there would be an unacceptable rate of survival of the pest. However, confirmatory tests would not indicate if a treatment were too high.
80. The United States noted that although confirmatory tests were necessary in the initial development of a quarantine treatment, they were unnecessary for every new variety of the same product. The confirmatory test administered to the initial variety, with a target pest population of 30,000 insects, was of such a design and order of magnitude to include a "real world" range of the pest population (and therefore relative levels of tolerance to methyl bromide) to detect an inadequate treatment; in other words, such a large-scale test would capture the range of susceptibility of insect population. The United States noted that 30,000 codling moths was a substantially larger number of insects than had ever been encountered in either dose-mortality tests or the actual conditions of US products proposed as candidates for export to Japan. ⁶⁶ The United States noted that the experts advising the Panel had affirmed that the incidence of codling moth on commercially marketable US products was low. Pest population levels for US products were measured, at most, in extremely low numbers, not hundreds per fruit, as encountered in the confirmatory test. Any treatment that could kill 30,000 codling moths in one application would not have difficulty killing the rare individual codling moth that might appear in a particular commercial shipment. Dr. Heather had further emphasized the robustness of methyl bromide treatment. The United States further claimed that no literature or scientific data supported an inherent variability in the products to indicate that a successful confirmatory test on 30,000 insects would not be uniformly successful for all varieties of the host product. The buffer of treatment used by scientists accounted for the natural variability of the pest and the experimental variability seen in small scale dose-mortality tests. Due to these factors, a confirmatory test would account for the variability in small scale tests and establish a treatment that was appropriate for all varieties of a product.
81. Japan claimed that empirical dose-response results confirmed the proposition that varietal differences could affect the mortality effect of MB fumigation. Data which indicated the presence of such statistically significant differences was contained in:

1. a 1987 study on American nectarines, in which Summer Grand was found significantly more susceptible to MB fumigation ⁶⁷,

Table 4 Susceptibility of Codling Moth Eggs to MB Fumigation (2 hours, 21 oC) in 6 Nectarine Varieties (American Nectarines)
Variety	No. of Sample Insects	LD50(95%CL) (g/m3)
Summer Grand	2,210	6.3 (2.2 - 9.1)
May Grand	2,458	14.2 (11.1 - 17.0)
Firebrite	1,880	18.8 (14.7 - 22.8)
Spring Red	1,019	17.7 (14.0 - 20.9)
Fantasia	1,548	17.6 (14.0 - 20.0)
Red Diamond	1,445	18.4 (17.1 - 19.8)

2. a 1987/88 test on New Zealand cherries in which LD₅₀ showed a significantly lower value for Bing than for Rainer and Sam⁶⁸, and

Table 5 Susceptibility of 1-day-old Codling Moth Eggs to MB Fumigation (2 hours, 12oC) in 5 Cherry Varieties (New Zealand Cherries)
Variety	LD50 (95%FL) (g/m3)	LD99 (95%FL) (g/m3)
Dawson	33.6 (31.8 - 35.1)	61.1 (55.9 - 69.8)
Bing	30.0 (28.9 - 30.9)	46.8 (44.7 - 49.6)
Rainier	33.8 (32.1 - 35.3)	62.2 (57.2 - 70.3)
Sam	35.4 (33.8 - 36.6)	52.9 (49.6 - 58.6)
Lambert	32.3 (29.9 - 34.0)	52.9 (48.4 - 61.7)

3. a 1983/84 study on disinfestation of New Zealand nectarines in which Fantasia showed a significantly lower LD₅₀ than Redgold. ⁶⁹

Table 6 Susceptibility of 1-day-old Codling Moth Larvae to MB Fumigation (2 hours at 12oC) (New Zealand Nectarines)
Variety	LD50 (95%FL) (g/m3)	LD99 (95%FL) (g/m3)
Fantasia	10.96 (10.56 � 11.33)	21.15 (19.82 - 22.93)
Redgold	12.09 (11.54 � 12.59)	32.86 (29.83 - 37.12)

82. Japan reaffirmed that it used the results of the dose-mortality test (which was part of the "basic test" paragraph 2.23) in the screening process to select a representative variety. ⁷⁰
83. Probit analysis ⁷¹ typically estimated LD₅₀ by measurement of the mortality rate in different doses and regression analysis of the statistically processed data (e.g., probit-conversion). The use of LD₅₀ value was justified on the basis of relative accuracy of estimation; the confidence level diminished as the value departed from the 50 per cent. A leading textbook of the analysis stated:

"As will become apparent in later chapters, by experiment with a fixed total number of subjects effective doses in the neighbourhood of ED₅₀ [effective dose 50%] can usually be estimated more precisely than those for more extreme percentage levels, and this is characteristic of the stimulus; its chief disadvantage is that, especially in toxicological work, much greater interest may attach to doses producing nearly 100% responses than to those producing only 50%, in spite of the difficulty of estimating the former." ⁷²

84. The use of the LD₅₀ value (estimated on the basis of the dose-mortality test results) in comparing efficacy of insecticide agents was a generally accepted scientific method of analysis. While the United States claimed that the dose-mortality test was effective only for estimating the ultimate quarantine treatment, it was typically utilized for the comparison of susceptibility of developmental stages of the treatment, in developing treatment schedules. ⁷³
85. Japan noted that the United States had pointed to "experimental error, physical condition of the fruit, sorption of the fumigant by packing material, and load of fruit in the chamber", as well as "natural variation of the test insect population", and claimed that "[t]he condition of any particular fruit could affect dose-response results". The conclusion drawn was that small scale dose-mortality tests were unable to correct for these variations. Of these other exogenous variables, "experimental error, physical condition of the fruit, sorption of the fumigant by packing material, and load of fruit in the chamber" were the factors which scientists who conducted these tests were responsible for controlling and consciously made equal. ⁷⁴ Indeed Japan noted that scientists endeavoured to set conditions that resembled each other as closely as possible and test insects were taken from artificially reared groups. For example, Yokoyama and other authors of the 1987 experiment on nectarines described in detail conditions of the fumigation chamber, wrapping of fruits, the load factor, time of fumigation, conditions of the codling moth and their rearing, and, from these descriptions, one could recognize that the scientists exerted efforts to ensure as much similarity between test samples. What would not be acceptable - if these factors affected the results - was the data they generated, and not Japan's hypothesis. In Japan's view, any responsible scientist would have to ensure that such exogenous factors did not falsify the results. It was not scientifically correct to reject a statistical conclusion on the grounds of experimental errors. If there were such exogenous factors as the United States claimed, their presence would mean that it was dangerous to draw any kind of conclusion from the results of such an experiment. None of the experts advising the Panel had concluded that experimental errors explained all of observed differences in the LD₅₀ values across varieties.
86. Japan claimed that another possible explanation for these factors was that they resulted from inevitable sampling error. If this was the case, the US position that confirmatory tests were the relevant indicator would be justifiable. However, Japan noted again that finding ways to alleviate, if not eliminate, problems of this kind belonged to the scientists of the government of exporting countries. Japan claimed that while there were always variables other than varietal differences, this was a statement of truth for any natural phenomenon. It was the responsibility of the exporting government to identify the variables and establish a treatment which would satisfactorily incorporate them so that Japan's level of protection would be achieved despite natural variation. The reason Japan chose to address the issue of varietal difference was because one could reasonably assume the presence of a route by which fruit characteristics of a particular variety might affect the outcome of a disinfestation treatment through their impact on CxT values. On the other hand, it was an established practice to ignore crop-to-crop or other natural variations in fruits or insects. The issue was the level of protection which an importing Member had the authority to choose.
87. In respect of the issue of the "buffer", Japan noted that while the United States claimed that the buffer was likely to cover conceivable varietal differences, the experts advising the Panel had not made any definitive arguments in this regard, nor had the United States explained the scientific grounds for their conviction. The United States seemed to rely intuitively on the past record of the efficacy of the treatment for additional varieties. However, as Dr. Heather had pointed out, this was a question of "risk management". In other words, the risk of inefficacy existed, and the rest was to be determined by the policy of the importing country.
88. Further, in regard to the buffer dose, Japan could not share the view of Dr. Heather and Mr. Taylor who had stated that the 10-20 per cent additional buffer dose might possibly cover the presupposed varietal differences. Japan argued that the buffer dose proposed by the United States had been established on the basis of laboratory scale dose-mortality tests. However, conditions for large-scale commercial application were different and the effect of gas sorption far from negligible. The amount of gas equivalent to the buffer dose would be sorbed to warehouses and containers. This coupled with inevitable gas leakage could mean the CxT values might be lowered to risky levels. Japan maintained that according to US data, as much as 20 per cent of the methyl bromide dose could be sorbed by picking bins, warehouses and possible leakage. On that basis, Japan could hardly assume that the 10-20 per cent buffer dose would cover possible differences of the magnitude of varietal variations.

To continue with Nectarines

⁶⁴ Toxic or injurious to plants.

⁶⁵ Finney, D.J., (1964) "Statistical Methods in Biological Assay," Griffin, London, pp. 91-92, Cited in Robertson, Preisler, Hickle and Gelernter, "Natural Variation: A Complicating Factor in Bioassays with Chemical and Microbial Pesticides", 88(1) J. Econ. Entomol. 1-10, 4-6. (1995). Finney states, "In general the assumption that a response curve once determined can be used in future assays is inadmissible. Because of natural variation, responses of a group of insects tested at any one time will therefore never be exactly the same as responses of another group tested at either the same time or at a different time, regardless of the extent to which bioassay techniques are standardized". In examining responses of Colorado potato beetle, diamondback moth, and western spruce budworm to three chemical and one microbial pesticides, Robertson et al., concluded that , "Studies of resistance, control of product quality, and tests of treatment efficacy are complicated by variation in response to a pesticide (whether chemical or microbial) that occurs within generations of a particular strain or within cohorts of a population: such variation is a natural phenomenon when any bioassay is repeated". (US Exhibit 11)

⁶⁶ The United States referred, in general, to US Exhibits 7, 8, 9, 10, 16 and 30 as well as US Exhibit 22: Work Plans already in place for the export to Japan of certain varieties of US products.

⁶⁷ Yokoyama, Miller and Hartsell, "Methyl Bromide Fumigation for Quarantine Control of Codling Moth (Lepidoptera: Tortricidae) on Nectarines,"80 J. Econ. Entomol. 840-842, 1987. (US Exhibit 14)

⁶⁸ Waddell, Birtlex, Dentener and Wearing, "Disinfestation of New Zealand Cherries Cultivar Comparison Test 1987/88", New Zealand. Department of Scientific and Industrial Research, Entomology Division, 1988. (Japan, Exhibit 17)

⁶⁹ Batchelor, Wearing, O'Donnel, "Disinfestation of New Zealand Nectarines 1983/1984", 1984. (Japan, Exhibit 18)

⁷⁰ Japan claimed that the estimation of LD₅₀ by probit analysis was commonly used for evaluation of toxicity of agricultural chemicals or for comparison of tolerance of pests to disinfestation treatment. Japan referred to e.g., OECD Guidelines for Testing of Chemicals, adopted 17 July 1992 (Japan, Exhibit 19) for the use of LD₅₀ value to assess toxicity of industrial chemicals; Knowles, C. (1988) J. Econ. Entomol.81;1586-1591 for comparison of LD₅₀ values between different insect groups. (Japan, Exhibit 20).

⁷¹ Definition in paragraph 2.14.

⁷² Probit Analysis (3rd ed.) Finney, D.J. (1971).

⁷³ Japan referred to a letter of 9 May 1997 from Robert G. Spade, Assistant Deputy Administrator, PPQ, which stated that: "During the development of quarantine treatment schedules, does-response data is used only to identify the least susceptible life stage of a target insect and to estimate effective treatment dosages". (Japan, Exhibit 34).

⁷⁴ Japan noted that statistical analysis, significance testing in particular, was precisely the tool to correct for experiment error.