ORIGINAL ARTICLE Notable Minimum Value of Recognized by Japanese Epidemiologists and Rule of Proof in Civil Trial: Questionnaire Survey

Minoru Sugita, Shaw Watanabe1, Masao Kanamori and Takashi Izuno

Department of Environmental and Occupational Health, Toho University School of Medicine and 1Division of , National Cancer Center Research Institute, Tokyo, Japan

(Received for publication on March 11, 1993)

Abstract. Members of The Japanese Society of Cancer Epidemiology were questioned about relative risk in cancer epidemiology. The notable minimum value of relative risk recognized by the epidemiologists distributed from 1.1 to 5.0, and the mean and the standard deviation of the notable minimum value were 1.92 and 0.760. In civil litigation of United States a plaintiff must prove one's case by a "preponderance of evidence," which means that a causal relationship between risk factors and a health disturbance to a plaintiff is proved when relative risk is 2.0 or more. The mean value 1.92 of the notable minimum value of relative risk nearly corresponded to the relative risk 2.0, which indicates that opinion of Japanese epidemiologists did not disagree with the rule of the proof in civil litigation of United States. It is mentioned that the standard of the proof should be "high probability" based on "a high level of conviction in the judge's own mind" in Japanese civil trials, and that the "high probability" means that relative risk is 5.0 or more if we dare to quantify the vagueness. This value corresponded to the maximum value of the notable minimum relative risk in the questionnaire to the Japanese epidemiologists. In civil litigation of Japan, the standard of the proof based on epidemiological results may be more conservative than the opinion of Japanese epidemiologists. (Keio J Med 42 (3): 115-118, September 1993)

Key words: epidemiologic data, health disturbance, civil litigation

Introduction especially in United States.4 Relative risk is used in civil trials as a measure to judge whether alleged factors In analytical epidemiology we try to reveal quantitative caused health disturbances of plaintiffs in United States.1 relation between health disturbances and their risk We shall have to pay attention to notable minimum factors. Relative risk is an indicator which expresses value of relative risk for the civil courts in the near future strength of association between a health disturbance in Japan. and its risk factor. is an indicator which Meaning of relative risk is varied by the attributable expresses magnitude of a social problem with an increased risk value from the viewpoint of social medicine . In order rate of a health disturbance due to its risk to detect the notable minimum value of relative risk factor. It gives a good idea of the impact that a success generally recognized by epidemiologists in Japan, mem ful prevention program of removing the risk factors bers of The Japanese Society of Cancer Epidemiology might have. were questioned about relative risk in cancer epidemiology In civil trials the role of epidemiology has become without consideration of the attributable risk. The results increasingly important,1 and there are cases which are of the survey were reported in the present study . influenced by evidence from epidemiological studies.2,3 Tools in the interface between law and epidemiology needed for resolving the issue have been developed

杉 田 稔,渡 辺 昌,金 森稚 夫,伊 津 野孝 Reprint requests to: Dr Minoru Sugita, Department of Environmental and Occupational Health , Toho University School of Medicine, 5-21-16 O mori nishi, Ota-ku, Tokyo 143, Japan

115 116 Sugita M, et al: Notable Minimum Value of Relative Risk

Table 1 Translation of the Questionnaire Items about Relative Risk Recognized by Epidemiologists in Cancer Epidemiology

Materials and Methods 1.1 to 5.0, and the mean and the standard deivation of the notable minimum value were 1.92 and 0.760. The Members of The Japanese Society Cancer Epide range of (Q2) the maximum relative risk was an area from miology were questioned about relative risk in cancer 10 to infinity, and the mean and the standard deviation epidemiology. Translation of the questionnaire items is of the maximum relative risk were 630 and 2240, and shown in Table 1. The subjects were questioned by mail those of (Q2') the common logarithm conversion value of from August through September in 1988. the maximum value were 1.93 and 0.673 (anti-logarithm conversion values to be 85.1 and 4.71), where infinity Results was calculated as 9999.9. The differences of means and the standard deviations between (Q1) the notable The questionnaires were distributed to 75 subjects minimum value and (Q2') the common logarithm con by mail, and 45 subjects (60.0%) responded to the version value of the maximum were small. The extent of questionnaire. (Q3) the notable minimum value of relative risk in a Minimums, maximums, means, standard deviations, sensitive sub-group distributed from 1.0 to 5.0, and the and numbers of subjects of (Q1) the notable minimum mean and the standard deviation of the value were 2.13 value of relative risk in cancer epidemiology recognized and 0.915. The mean of (Q1) the notable minimum value by epidemiologists, (Q2) the maximum value of relative of relative risk was slightly smaller than that (Q3) in risk which is known, (Q2') the common logarithm (base a sensitive sub-group. of logarithm si 10) conversion value of the above maxi Figure 1 shows histogram of (Q1) the notable minimum mum value, and (Q3) the notable minimum value of value of relative risk in cancer epidemiology, (Q2) the relative risk in a sensitive sub-group are shown in Table 2. maximum value of relative risk, or (Q3) the notable Table 2 indicates that (Q1) the notable minimum value minimum value of relative risk in a sensitive sub-group. of relative risk in cancer epidemiology distributed from Logarithm scale (breadth of each class to be •ã10 times)

Table 2 Minimums, Maximums, Means, Standard Deviations, and Numbers of Subjects of (Q1) Notable Minimum Value of Relative Risk in Cancer Epidemiology, (Q2) Maximum Value of Relative Risk, (Q2') Common Logarithm Conversion Value of the Above Maximum Value, and (Q3) Notable Minimum Value of Relative Risk in a Sensitive Sub-group

SD: standard deviation No.: number of subjectsa : •‡ was assigned to 9999.9 for calculation. b: •‡ was assigned to log109999 .9 for calculation. 117 Keio J Med 42 (3): 115-118, 1993

Q1. Notable Minimum Value of Relative Risk was used for (Q2) the maximum value of relative risk. (Total Number of Subjects: 40) Each histogram shows positive skewness of the distri bution, i.e. longer tail at the side of the large value. The histogram of (Q1) the notable minimum value of relative risk resembled closely that of (Q2) the maximum value of relative risk expressed by logarithm scale. As main cancer sites and their risk factors for maximum relative risk, lung cancer (9) and larynx cancer (6) caused by smoking and hepatocellular carcinoma (9) due to hepatitis B virus were listed by many epidemiologists. The numbers in the parentheses indicate numbers of epidemiologists.

Discussion

In epidemiology, relative risk is used as an indicator which expresses strength of association between a Q2. Maximum Value of Relative Risk and its risk factor, and attributable risk is used as an (Total Number of Subject: 38) indicator which expresses magnitude of a social problem with an increased incidence rate of a disease due to its risk factor. Significance of relative risk is affected by value of the attributable risk from the viewpoint of social medicine. In the questionnaire survey of the present study, the notable minimum value of relative risk in cancer epidemiology generally recognized by Japanese epidemiologists was investigated without consideration of the attributable risk. Attributable risk percent (AR%) is an indicator usually used in epidemiology. The meaning of AR% is the proportion of patients with a disease in the group that could be prevented by eliminating its risk factor. The calculation method of AR% value is

AR%=(RR-1)/RR, where RR is relative risk value. Q3. Notable Minimum Value of Relative Risk in a Sensitive Sub-group The range of the notable minimum value of relative (Total Number of Subjects: 31) risk was 1.1-5.0 and the standard deviation was 0.760, which does not show that large difference of recognition about the notable minimum value of relative risk in cancer epidemiology among Japanese epidemiologists was detected. The mean of the notable minimum valae was 1.92, this value is slightly smaller than 2. The mean 1.92 for the notable minimum relative risk is corresponds to AR% 47.9%. Let us inquire these values from view point of civil court. In civil litigation of United States, a plaintiff must prove one's case by a "preponderance of evidence" rule.1,4.5 This criterion is a basic trend of the United States legal system.1,4,5 It is also termed the "more likely than not" rule,1,4,5 the "50.1 percent or more probability" rule,1,4,5 or the "reasonable medical certainty" rules "Preponderance of evidence" is d efined as "evidence of greater weight or more convincing than the evidence Fig 1 Histograms of the questions. which is offered in opposition to it ... 'preponderance' denotes a superiority of weight ."5 Therefore, in civil 118 Sugita M, et al: Notable Minimum Valueeof Relative Risk

litigation of United States causal relationship between their recognition as a social problem. Therefore, it is risk factors and a disease to plaintiffs is proved when mentioned that useful information for civil trial was AR% is 50% or more i.e. relative risk is 2.0 or more,1 provided by the questionnaire survey of the present whereas scientific proof and statistical significance test study. by convention on demonstration of a 95% or greater The mean of (Q1) the notable minimum value of probability that the results of an or an investi relative risk was slightly smaller than that (Q3) in a gation were not due to chance.4,5 In criminal cases, sensitive sub-group, which may express that the Japanese the usual standard of proof is "beyond a reasonable epidemiologists responded to these questions with slightly doubt."4,5 Such a probabilistic standard is clearly more different values because of difficulty in finding the sen easily satisfied with epidemiologic evidence than the sitive sub-group in medical studies. criminal law test.4 It is interesting that (Q1) the notable minimum value The mean 1.92 of the notable minimum value of relative of relative risk in cancer epidemiology recognized by risk in the questionnaire to the Japanese epidemiologists epidemiologists almost coincided with (Q2') the common nearly corresponded to relative risk 2.0, which indicates logarithm conversion value of the maximum of relative that their opinions did not disagree with the rule of proof risk which is known. in civil litigation of United States. As and their risk factors of the maximum The minimum 1.1 of the notable minimum value relative risk, Japanese epidemiologists listed lung and of relative risk indicates a view that epidemiologists laryngeal cancers due to smoking and hepatocellular should pay attention to even a slight increase of a cancer carcinoma due to hepatitis B virus. The attributable risks incidence rate due to its risk factor. When attributable of these diseases due to their risk factors are high, and risk is large, the notable minimum value of relative risk the mortality rates of these diseases are increasing in is small because of the magnitude of the social problem Japan, which is a significant problem from a viewpoint of due to the increased incidence rate of the disease due to social medicine. its cause. Epidemiological study on relation between exposure to environmental tobacco smoke and risk of Acknowledgement:The authors thank Dr NobuyukiHamajima for lung cancer is a good example. Hamajima6 reported adviceof lawterms. This is a modifiedversion of a paperpresented at The15th Meeting of the Japanese Society of CancerEpidemiology, relative risk under an effect modifier and legal causality. 1992,Osaka. The significance of relative risk is varied also by effect modifiers. References It is mentioned that a standard of proof should be "high probability" based on "a high level of conviction . Hamajima N: Epidemiology in civil trials. Nihon Koshu Eisei in the judge's own mind" in Japanese civil trials.1,7 Zasshi (Jpn J Public Health) 1991, 38: 541-545 (in Japanese) If we dare to quantify the vegueness of the standard 2. Henderson TW: Legal aspects of disease clusters -- toxic tort of proof in Japanese civil trials, the "high probability" litigation: medical and scientific principles in causation. Am J means an 80% of conviction or an 80% of AR%, i.e. Epidemiol 1990, 132: S69-S78 3. Black B: Matching evidence about clustered health events with relative risk is 5.0.1,8,9 It is interesting that this value tort law requirements. Am J Epidemiol 1990, 132: S79-S86 corresponded to the maximum value of the notable 4. Lilienfeld DE, Black B: The epidemiologists in court: some com minimum value of relative risk in the questionnaire ments. Am J Epidemiol 1986, 123: 961-964 to the Japanese epidemiologists. In civil litigation of 5. Hoffman RE: The use of epidemiologic data in the courts. Am J Japan, the standard of proof based on epidemiological Epidemiol 1984, 120: 190-202 6. Hamajima N: Attributable risk percent under effect modifier and results may be more conservative than the opinion of legal causality. Nihon Koshu Eisei Zasshi (Jpn J Public Health) Japanese epidemiologists. 1992, 39: 302-306 (in Japanese) The questionnaire survey of the present study was 7. Kasuga I: A principle of a high level of conviction in the judge's not performed in order to investigate recognition of own mind In: Mikazuki A, Aoyama Y, eds, Issues in the Code of relative risk in civil trial by Japanese epidemiologists. Civil Procedure, Tokyo, Yuhi-kaku, 1988, 236-239 (in Japanese) A civil trial, in which relative risk has been used as a 8. Morishima A: Jijitsuteki ingakankei no shomei. In: Morishima A, ed, Text Book of Laws of Illegalities, Tokyo, Yuhi-kaku, 1987, measure to judge whether alleged factors caused health 288-305 (in Japanese) disturbances of plaintiffs, has not held public attention as 9. Niimi I: Proof of causal relationship by a social problem in Japan. If the questionnaire survey is (a second series). Jurist 1986, 871: 89-94 (in Japanese) conducted when Japanese epidemiologists pay attention to such a civil trial, the result may be affected because of

1