LETTERS TO THE EDITOR
These issues are critical when the public is already skeptical of risk analysis. If an analysis makes claims beyond the data, the public may view later analyses even more suspiciously. In perspective, the research from Seattle-King County should be important to all environmental health professionals. It is useful to find reported associations between restaurant inspections and foodborne illness. However, Irwin, et al, were too zealous in interpreting their results. REFERENCES 1. Hatfield TH: Restaurant inspections may not predict outbreaks of foodbome illness. Am J Public Health 1989; 79:1678. 2. Irwin K, Ballard J, Grendon J, Kobayashi J: Results of routine restaurant inspections can predict outbreaks of foodborne illness: The Seattle-King County Experience. Am J Public Health 1989;79:586-590. 3. Irwin K, Ballard J, Kobayashi J, Veazie M, Grendon J, Yuen G: Response from Irwin, et al. Am J Public Health 1989; 79:1678-1679. 4. Rothman K: Modem Epidemiology. Boston: Little, Brown and Co, 1986; 87. 5. Bryan FL, Lewis KH, Anderson HW, et al: Procedures to Investigate Foodbome Illness. Ames, IA: International Association of Milk, Food, and Environmental Sanitarians, 1976; 11.
Thomas H. Hatfield, RS, DrPH Assistant Professor, Department of Health Sciences, California State University, Northridge, CA 91330
Response from Irwin, et al. As noted in our article,' the findings of our study are best used as an example of why certain violations should be emphasized in inspectionsbecause they are epidemiologically associated with reported outbreaksrather than to dictate in a detailed way the training of sanitarians of food handlers or restaurant regulation. In this small study, the first of its kind, we did not intend to definitively establish cause-effect relationships. As we discussed earlier,',2 misclassification, confounding, and statistical imprecision may have distorted our risk estimates, including those for violations 5 and 22. We therefore conducted a larger matched case-control study involving 70 restaurants outbreaks in 1986-88.3 The odds ratio for violation 22 decreased from 14.9 (95% CI 2.6, 85.4) in the original study to 2.9 (95% CI 1.4, 6.1) in the second study. The odds ratio for violation 5 increased from 1.4 (95% CI 0.4, 5.2) to 2.4 (95% CI 1.1, 5.0). Most of the outbreaks in our first study' involved only a few persons for at least three reasons. First, our outbreak definition required only one perAJPH August 1990, Vol. 80, No. 8
son to define toxin-related outbreaks (e.g., botulism, scombroid, chemical) because each type is relatively rare and the disease symptoms fairly specific. Second, our definition required that illness in affected individuals must be epidemiologically related to restaurant food consumption. When only two persons report illness, especially two who have eaten together at the same restaurant, it is more difficult to establish an epidemiologic link because other common causes of illness must be considered and ruled out, if possible. Thus, criteria for an epidemiologic association are usually stronger (e.g., laboratory confirmation) when only two or nonindependent disease reports are involved. Thus, an outbreak that affects only two individuals is not necessarily "less" of an outbreak than one that affects more. Third, because SeattleKing County's surveillance system for reporting of foodborne illness is passive, outbreaks associated with only two reported cases may often involve more persons who do not report their symptoms. Given limited resources to investigate outbreaks, active efforts to identify other ill persons may not always be justified once an association with restaurant food has been established for two persons. As an internationally recognized standard and one that had been used for decades as the basis for outbreak surveillance by the Seattle-King County Health Department, this definition4 was the best choice for this study. A definition which required pathogen confirmation would not reflect the true spectrum of reported outbreaks because the cause of many outbreaks, (e.g., viral) cannot be readily confirmed with laboratory tests. We invite Dr. Hatfield to test the sensitivity and specificity of other definitions which can be applied to the routine surveillance of foodbome illness. REFERENCES 1. Irwin K, Ballard J, Grendon J, et al: Results of routine restaurant inspections can predict outbreaks of foodborne illness: The Seattle-King County experience. Am J Public Health 1989;
79:586-590. 2. Irwin K, Ballard J, Kobayashi J, et al: Response to letter from TH Hatfield. Am J Public Health
1989; 79:1678-1679. 3. Crofts N, Grendon J, Irwin K, et al. Routine restaurant inspections and foodborne illness outbreaks, Seattle-King County, 1986-1988. Report to the Communicable Disease Epidemiology Section, Washington State Department of
Health, Seattle, Washington. 4. Bryan FL, Lewis KF, Anderson HW, et al: Procedures to investigate foodborne illness. Ames, IA: International Association of Milk, Food and Environmental Sanitarians, 1987.
Kathleen L. Irwin, MD, MPH Department of Epidemiology, University of Washington, School of Public Health and Community Medicine, Seattle; and John Grendon, DVM, MPH John Kobayashi, MD, MPH Mark Veazie, RS Washington State Department of Health, Seattle.
Kettering Laboratory: A Pioneer in Lead Research In the December 1989 issue of the Journal, Richard Rabin is accurate about the tardiness and procrastination of the paint industry in eliminating lead from all paints both for exterior and interior use.' He is inaccurate in characterizing all of the distinguished scientists investigating lead intoxication as being pawns of the paint and lead industry. His description of the pioneering scientific effort on lead intoxication in the period from the 1920s to the present is grossly incomplete. I was struck by the author's description of the role of Dr. Robert A. Kehoe and the Kettering Laboratory and the history of lead poisoning in general and childhood lead poisoning in particular. In 1924, Robert A. Kehoe developed a University-based research program for the study of both organic and inorganic lead compounds, which by 1930 became the Kettering Laboratory. He developed a clinical and laboratory program devoted to a wide range of environmental issues which was unique for its time and remains so today.2 The Kehoe program required an accurate analytical method for measuring lead in environmental as well as biological materials including urine and blood. This was first developed by Jacob Cholak,3,4 then a young chemical engineer, and Frederick Thamann, a chemist. The Kettering Laboratory was the first university research unit created for the scientific investigation of occupational and environmental health problems. Starting in the 1930s Dr. Kehoe and a group of physicians housed in the Kettering Laboratory carried out the continuous surveillance of workers and their environments in the lead industries. He developed a collaborative program with the Children's Hospital of Cincinnati which became the center for the diagnosis and treatment for children at risk for the entire region. The collaborative efforts have continued to grow, and Kettering Laboratory in conjunc1001
LETTERS TO THE EDITOR
tion with the Children's Hospital comprises a national center for the study of lead toxicity of children supported by the National Institutes of Environmental Health Sciences, with additional support from the lead industry. Dr. Kehoe conducted a study over a period of 34 years of the absorption, metabolism and excretion of lead in humans. It is the only study of its kind in the toxicology of lead. Human subjects were exposed to low levels of lead-via ingestion and others by inhalation. It should be noted that all of the subjects lived active, normal lives after their participation in the study. All of them are still living; most are now in their 70s and 80s. This work was published as a monograph in Food and Chemical Toxicology in June 1987.5 Dr. Robert A. Kehoe is now 96 years old and would probably be as unhappy about these omissions as I am. A historian of science should recognize that in the twenties and thirties there was little or no federal support for toxicological research at universities. The support from the private sector, including automotive, petroleum and even the lead industries, was welcomed by scientists such as Robert Kehoe. REFERENCES 1. Rabin R: Warnings unheeded: A history of child lead poisoning. (Public Health Then and Now) Am J Public Health 1989; 79:1668-1674. 2. Campbell IR: The house that Robert A. Kehoe
built. Arch Environ Health 1966; 13:143-150. 3. Cholak J: The quantitative spectrographic determination of lead in urine. J Am Chem Soc 1935; 57:104-107. 4. Cholak J, Hubbard DM, McNary RR, Story RV: Determination of lead in biological materials. Comparison of spectrographic, dithizone and s-diphenylcarbazide methods. Ind Eng Chem, Anal Ed 1937; 9:488-490. 5. Kehoe RA: Studies of lead administration and elimination in adult volunteers under natural and experimentally induced conditions over extended periods of time. Food Chem Toxicol 1987; 25:428487 (suppl).
Raymond R. Suskind, MD Director Emeritus, Institute of Environmental Health and the Kettering Laboratory, University of Cincinnati Medical Center, 3223 Eden Avenue, Cincinnati, OH 45267-0056 © 1990 American Journal of Public Health
Response from Mr. Rabin Dr. Suskind's letter misses the point of my references to scientists who were associated with the lead industry. ' The purpose of identifying those researchers was not to discuss their ac-
complishments or to consider the effect of industry funding on their research. 1002
The purpose of describing the ties of Robert Kehoe and other scientists to the lead industry was simply to make clear that what such researchers knew could not have been unknown to the Lead Industries Association or its members. When Dr. Kehoe stated that". strenuous efforts must be devoted to eliminating lead from [children's] environment,"2 the lead pigment manufacturers could not claim that they were unaware of the dangers of lead paint to young children. Indeed, one of the reasons that one can refer to the "procrastination" of the paint industry in removing lead from paint is that industry-funded researchers were well aware of the dangers of lead paint long before the industry took action. REFERENCES 1. Rabin R: Warnings unheeded: A history of child lead poisoning. (Public Health Then and Now) Am J Public Health 1989; 79:1668-1674. 2. McKhann C, Vogt E: Lead poisoning in children. JAMA 1933; 101: 1131-1135. Richard Rabin 11 Cogswell Avenue, Apt. 8, Cambridge, MA 02140 © 1990 American Journal of Public Health
On Borderline Statistical Significance The title of the recent article by Maclure, et al, "Elevated Blood Levels of Carcinogens in Passive Smokers,"' misrepresents the data presented by the authors. While the design of the study is commendable, the conclusions drawn are not supported by the data presented. The data on 4-aminobiphenylhemoglobin (4-ABP-Hb) adducts are highly variable; the three highest levels of 4-ABP-Hb adducts are found in subjects with no detectable urinary cotinine. Additionally, the difference between cotinine-positive and cotininenegative subjects is not statistically significant if the commonly accepted criterion of p < 0.05 is used. Additionally, it does not appear that 3-aminobiphenyl-hemoglobin (3-ABP-Hb) adducts can be considered a consistent marker of 4-ABP exposure as the subject with the highest level of 4-ABP adducts had no detectable level of 3ABP adducts. The changes in 3-ABP adduct levels between cotinine-positive and cotinine-negative subjects may have been statistically significant, but the large number of values below the limit of
detection and the low quantitative values ofthese adducts make the biological significance of the difference unclear. More importantly, there is little, if any, definitive evidence that 3-ABP is carcinogenic. The reference cited by the authors in support of this claim is a review article published in 19662 in which no data are presented nor citation provided on 3-ABP carcinogenicity. Furthermore, the authors, in a recent paper3 state that 3-ABP is reported to be noncarcinogenic. The rationale for studying 3-ABP-Hb adducts therefore appears to be lacking. The variability and small differences among groups could be explained by non-tobacco sources of 4-ABP and 3-ABP, a possibility raised by these authors in a previous study.4 This previous study reported as much as a 300-fold difference among four animal species in their background levels of 4-ABP-Hb adducts (their Table 1), supporting the probability that, in the present study, these adducts may arise from non-tobacco sources. While it is important to accurately assess the carcinogenic risk of passive exposure to tobacco smoke, this goal is not advanced when data are overinterpreted and a provocative, yet unsupported title is used. Unfortunately, this paper may be cited in the future as supportive of a cause and effect relationship which is not scientifically justified by the data presented. REFERENCES 1. Maclure M, Katz RB, Bryant MS, Skipper PL, Tannenbaum SR: Elevated blood levels of carcinogens in passive smokers. Am J Public Health 1989; 79:1381-1384. 2. Weisburger JH, Weisburger EK: Chemicals as causes of cancer. Chem Eng News 1966; 44(6): 124-142. 3. Skipper PL, Bryant MS, Tannenbaum SR: Determination of human exposure to carcinogenic aromatic amines from hemoglobin adducts in selected population groups. In: King CM, Romano LJ, Schuetzle D (eds): Carcin Mut Responses Aromatic Amines Nitrosamines. New
York: Elsevier Science Publishing Co., 1988; 65-71. 4. Bryant MS, Skipper PL, Tannenbaum SR, Maclure M: Hemoglobin adducts of 4-aminobiphenyl in smokers and nonsmokers. Cancer Res 1987; 47:602-608. Mark J. Reasor, PhD Professor, Department of Pharmacology and Toxicology, West Virginia University, Morgantown, WV 26506
Response from Maclure, et al. The difference of opinion between Reasor and us' concerns how to deAJPH August 1990, Vol. 80, No. 8
These issues are critical when the public is already skeptical of risk analysis. If an analysis makes claims beyond the data, the public may view later analyses even more suspiciously. In perspective, the research from Seattle-King County should be important to all environmental health professionals. It is useful to find reported associations between restaurant inspections and foodborne illness. However, Irwin, et al, were too zealous in interpreting their results. REFERENCES 1. Hatfield TH: Restaurant inspections may not predict outbreaks of foodbome illness. Am J Public Health 1989; 79:1678. 2. Irwin K, Ballard J, Grendon J, Kobayashi J: Results of routine restaurant inspections can predict outbreaks of foodborne illness: The Seattle-King County Experience. Am J Public Health 1989;79:586-590. 3. Irwin K, Ballard J, Kobayashi J, Veazie M, Grendon J, Yuen G: Response from Irwin, et al. Am J Public Health 1989; 79:1678-1679. 4. Rothman K: Modem Epidemiology. Boston: Little, Brown and Co, 1986; 87. 5. Bryan FL, Lewis KH, Anderson HW, et al: Procedures to Investigate Foodbome Illness. Ames, IA: International Association of Milk, Food, and Environmental Sanitarians, 1976; 11.
Thomas H. Hatfield, RS, DrPH Assistant Professor, Department of Health Sciences, California State University, Northridge, CA 91330
Response from Irwin, et al. As noted in our article,' the findings of our study are best used as an example of why certain violations should be emphasized in inspectionsbecause they are epidemiologically associated with reported outbreaksrather than to dictate in a detailed way the training of sanitarians of food handlers or restaurant regulation. In this small study, the first of its kind, we did not intend to definitively establish cause-effect relationships. As we discussed earlier,',2 misclassification, confounding, and statistical imprecision may have distorted our risk estimates, including those for violations 5 and 22. We therefore conducted a larger matched case-control study involving 70 restaurants outbreaks in 1986-88.3 The odds ratio for violation 22 decreased from 14.9 (95% CI 2.6, 85.4) in the original study to 2.9 (95% CI 1.4, 6.1) in the second study. The odds ratio for violation 5 increased from 1.4 (95% CI 0.4, 5.2) to 2.4 (95% CI 1.1, 5.0). Most of the outbreaks in our first study' involved only a few persons for at least three reasons. First, our outbreak definition required only one perAJPH August 1990, Vol. 80, No. 8
son to define toxin-related outbreaks (e.g., botulism, scombroid, chemical) because each type is relatively rare and the disease symptoms fairly specific. Second, our definition required that illness in affected individuals must be epidemiologically related to restaurant food consumption. When only two persons report illness, especially two who have eaten together at the same restaurant, it is more difficult to establish an epidemiologic link because other common causes of illness must be considered and ruled out, if possible. Thus, criteria for an epidemiologic association are usually stronger (e.g., laboratory confirmation) when only two or nonindependent disease reports are involved. Thus, an outbreak that affects only two individuals is not necessarily "less" of an outbreak than one that affects more. Third, because SeattleKing County's surveillance system for reporting of foodborne illness is passive, outbreaks associated with only two reported cases may often involve more persons who do not report their symptoms. Given limited resources to investigate outbreaks, active efforts to identify other ill persons may not always be justified once an association with restaurant food has been established for two persons. As an internationally recognized standard and one that had been used for decades as the basis for outbreak surveillance by the Seattle-King County Health Department, this definition4 was the best choice for this study. A definition which required pathogen confirmation would not reflect the true spectrum of reported outbreaks because the cause of many outbreaks, (e.g., viral) cannot be readily confirmed with laboratory tests. We invite Dr. Hatfield to test the sensitivity and specificity of other definitions which can be applied to the routine surveillance of foodbome illness. REFERENCES 1. Irwin K, Ballard J, Grendon J, et al: Results of routine restaurant inspections can predict outbreaks of foodborne illness: The Seattle-King County experience. Am J Public Health 1989;
79:586-590. 2. Irwin K, Ballard J, Kobayashi J, et al: Response to letter from TH Hatfield. Am J Public Health
1989; 79:1678-1679. 3. Crofts N, Grendon J, Irwin K, et al. Routine restaurant inspections and foodborne illness outbreaks, Seattle-King County, 1986-1988. Report to the Communicable Disease Epidemiology Section, Washington State Department of
Health, Seattle, Washington. 4. Bryan FL, Lewis KF, Anderson HW, et al: Procedures to investigate foodborne illness. Ames, IA: International Association of Milk, Food and Environmental Sanitarians, 1987.
Kathleen L. Irwin, MD, MPH Department of Epidemiology, University of Washington, School of Public Health and Community Medicine, Seattle; and John Grendon, DVM, MPH John Kobayashi, MD, MPH Mark Veazie, RS Washington State Department of Health, Seattle.
Kettering Laboratory: A Pioneer in Lead Research In the December 1989 issue of the Journal, Richard Rabin is accurate about the tardiness and procrastination of the paint industry in eliminating lead from all paints both for exterior and interior use.' He is inaccurate in characterizing all of the distinguished scientists investigating lead intoxication as being pawns of the paint and lead industry. His description of the pioneering scientific effort on lead intoxication in the period from the 1920s to the present is grossly incomplete. I was struck by the author's description of the role of Dr. Robert A. Kehoe and the Kettering Laboratory and the history of lead poisoning in general and childhood lead poisoning in particular. In 1924, Robert A. Kehoe developed a University-based research program for the study of both organic and inorganic lead compounds, which by 1930 became the Kettering Laboratory. He developed a clinical and laboratory program devoted to a wide range of environmental issues which was unique for its time and remains so today.2 The Kehoe program required an accurate analytical method for measuring lead in environmental as well as biological materials including urine and blood. This was first developed by Jacob Cholak,3,4 then a young chemical engineer, and Frederick Thamann, a chemist. The Kettering Laboratory was the first university research unit created for the scientific investigation of occupational and environmental health problems. Starting in the 1930s Dr. Kehoe and a group of physicians housed in the Kettering Laboratory carried out the continuous surveillance of workers and their environments in the lead industries. He developed a collaborative program with the Children's Hospital of Cincinnati which became the center for the diagnosis and treatment for children at risk for the entire region. The collaborative efforts have continued to grow, and Kettering Laboratory in conjunc1001
LETTERS TO THE EDITOR
tion with the Children's Hospital comprises a national center for the study of lead toxicity of children supported by the National Institutes of Environmental Health Sciences, with additional support from the lead industry. Dr. Kehoe conducted a study over a period of 34 years of the absorption, metabolism and excretion of lead in humans. It is the only study of its kind in the toxicology of lead. Human subjects were exposed to low levels of lead-via ingestion and others by inhalation. It should be noted that all of the subjects lived active, normal lives after their participation in the study. All of them are still living; most are now in their 70s and 80s. This work was published as a monograph in Food and Chemical Toxicology in June 1987.5 Dr. Robert A. Kehoe is now 96 years old and would probably be as unhappy about these omissions as I am. A historian of science should recognize that in the twenties and thirties there was little or no federal support for toxicological research at universities. The support from the private sector, including automotive, petroleum and even the lead industries, was welcomed by scientists such as Robert Kehoe. REFERENCES 1. Rabin R: Warnings unheeded: A history of child lead poisoning. (Public Health Then and Now) Am J Public Health 1989; 79:1668-1674. 2. Campbell IR: The house that Robert A. Kehoe
built. Arch Environ Health 1966; 13:143-150. 3. Cholak J: The quantitative spectrographic determination of lead in urine. J Am Chem Soc 1935; 57:104-107. 4. Cholak J, Hubbard DM, McNary RR, Story RV: Determination of lead in biological materials. Comparison of spectrographic, dithizone and s-diphenylcarbazide methods. Ind Eng Chem, Anal Ed 1937; 9:488-490. 5. Kehoe RA: Studies of lead administration and elimination in adult volunteers under natural and experimentally induced conditions over extended periods of time. Food Chem Toxicol 1987; 25:428487 (suppl).
Raymond R. Suskind, MD Director Emeritus, Institute of Environmental Health and the Kettering Laboratory, University of Cincinnati Medical Center, 3223 Eden Avenue, Cincinnati, OH 45267-0056 © 1990 American Journal of Public Health
Response from Mr. Rabin Dr. Suskind's letter misses the point of my references to scientists who were associated with the lead industry. ' The purpose of identifying those researchers was not to discuss their ac-
complishments or to consider the effect of industry funding on their research. 1002
The purpose of describing the ties of Robert Kehoe and other scientists to the lead industry was simply to make clear that what such researchers knew could not have been unknown to the Lead Industries Association or its members. When Dr. Kehoe stated that". strenuous efforts must be devoted to eliminating lead from [children's] environment,"2 the lead pigment manufacturers could not claim that they were unaware of the dangers of lead paint to young children. Indeed, one of the reasons that one can refer to the "procrastination" of the paint industry in removing lead from paint is that industry-funded researchers were well aware of the dangers of lead paint long before the industry took action. REFERENCES 1. Rabin R: Warnings unheeded: A history of child lead poisoning. (Public Health Then and Now) Am J Public Health 1989; 79:1668-1674. 2. McKhann C, Vogt E: Lead poisoning in children. JAMA 1933; 101: 1131-1135. Richard Rabin 11 Cogswell Avenue, Apt. 8, Cambridge, MA 02140 © 1990 American Journal of Public Health
On Borderline Statistical Significance The title of the recent article by Maclure, et al, "Elevated Blood Levels of Carcinogens in Passive Smokers,"' misrepresents the data presented by the authors. While the design of the study is commendable, the conclusions drawn are not supported by the data presented. The data on 4-aminobiphenylhemoglobin (4-ABP-Hb) adducts are highly variable; the three highest levels of 4-ABP-Hb adducts are found in subjects with no detectable urinary cotinine. Additionally, the difference between cotinine-positive and cotininenegative subjects is not statistically significant if the commonly accepted criterion of p < 0.05 is used. Additionally, it does not appear that 3-aminobiphenyl-hemoglobin (3-ABP-Hb) adducts can be considered a consistent marker of 4-ABP exposure as the subject with the highest level of 4-ABP adducts had no detectable level of 3ABP adducts. The changes in 3-ABP adduct levels between cotinine-positive and cotinine-negative subjects may have been statistically significant, but the large number of values below the limit of
detection and the low quantitative values ofthese adducts make the biological significance of the difference unclear. More importantly, there is little, if any, definitive evidence that 3-ABP is carcinogenic. The reference cited by the authors in support of this claim is a review article published in 19662 in which no data are presented nor citation provided on 3-ABP carcinogenicity. Furthermore, the authors, in a recent paper3 state that 3-ABP is reported to be noncarcinogenic. The rationale for studying 3-ABP-Hb adducts therefore appears to be lacking. The variability and small differences among groups could be explained by non-tobacco sources of 4-ABP and 3-ABP, a possibility raised by these authors in a previous study.4 This previous study reported as much as a 300-fold difference among four animal species in their background levels of 4-ABP-Hb adducts (their Table 1), supporting the probability that, in the present study, these adducts may arise from non-tobacco sources. While it is important to accurately assess the carcinogenic risk of passive exposure to tobacco smoke, this goal is not advanced when data are overinterpreted and a provocative, yet unsupported title is used. Unfortunately, this paper may be cited in the future as supportive of a cause and effect relationship which is not scientifically justified by the data presented. REFERENCES 1. Maclure M, Katz RB, Bryant MS, Skipper PL, Tannenbaum SR: Elevated blood levels of carcinogens in passive smokers. Am J Public Health 1989; 79:1381-1384. 2. Weisburger JH, Weisburger EK: Chemicals as causes of cancer. Chem Eng News 1966; 44(6): 124-142. 3. Skipper PL, Bryant MS, Tannenbaum SR: Determination of human exposure to carcinogenic aromatic amines from hemoglobin adducts in selected population groups. In: King CM, Romano LJ, Schuetzle D (eds): Carcin Mut Responses Aromatic Amines Nitrosamines. New
York: Elsevier Science Publishing Co., 1988; 65-71. 4. Bryant MS, Skipper PL, Tannenbaum SR, Maclure M: Hemoglobin adducts of 4-aminobiphenyl in smokers and nonsmokers. Cancer Res 1987; 47:602-608. Mark J. Reasor, PhD Professor, Department of Pharmacology and Toxicology, West Virginia University, Morgantown, WV 26506
Response from Maclure, et al. The difference of opinion between Reasor and us' concerns how to deAJPH August 1990, Vol. 80, No. 8
