American Journal of Epidemiology Copyright © 2002 by the Johns Hopkins Bloomberg School of Public Health All rights reserved Vol. 155, No. 2 Printed in U.S.A. Belgian Coca-Cola-related Outbreak Gallay et al. Belgian Coca-Cola-related Outbreak: Intoxication, Mass Sociogenic Illness, or Both? A. Gallay,1,2 F. Van Loock,1 S. Demarest,1 J. Van der Heyden,1 B. Jans,1 and H. Van Oyen1 carbonated beverages; disease outbreaks; hydrogen sulfide; poisoning An epidemic of health complaints, including nausea, vomiting, abdominal pain, dizziness, and headache, potentially related to consumption of Coca-Cola Company soft drinks occurred in June 1999 in Belgium. The epidemic started on June 8 in one secondary school (school A). Two to 6 days later, students in four other secondary schools (schools B–E) complained of the same symptoms. During the same period, several complaints were reported in the Belgian and the French populations (1, 2). Between June 8 and June 20, the Belgian Poisoning Call Centre recorded over 1,400 telephone calls; 55 percent were complaints related to consumption of Coca-Cola soft drinks, and 45 percent of the callers asked for information about the quality of the soft drinks (1). The CocaCola-related calls constituted one third of all calls the Poisoning Call Centre received during this period. On June 15, The Coca-Cola Company announced that chemical analysis of the incriminated beverages had revealed very low concentrations of hydrogen sulfide in the glass bottles of Coca-Cola supplied to school A, and that 4chloro-3-methylphenol, applied to transport pallets, had contaminated the exterior surface of the cans delivered to schools B–E. In both cases, the company concluded that the very low concentration of these two substances could not have caused any toxicity. Still, The Coca-Cola Company withdrew 15 million crates of its soft drinks across Belgium, France, and Luxembourg and temporarily closed three of its factories in Europe. On June 23, the Belgian Ministry of Public Health commissioned the Unit of Epidemiology of the Scientific Institute of Public Health (Brussels) to investigate this outbreak and to identify the cause and mode of transmission. Epidemiologic and clinical information was collected on cases in the affected schools, and a case-control study was performed to determine the weight of evidence on both competing hypotheses—consumption of Coca-Cola Company products and mass sociogenic illness—as a risk factor for illness. MATERIALS AND METHODS It was decided to consider the outbreaks in school A and in schools B–E as two distinct incidents because 1) school A was supplied with glass bottles from an Antwerp (Belgium) plant, whereas the bottles and cans for schools B–E were supplied by plants in Gent (Belgium) and Dunkerque (France); 2) The Coca-Cola Company had identified a different toxicologic substance in the soft drinks delivered to school A and schools B–E; and 3) the events at school A and schools B–E occurred at different times. Received for publication February 26, 2001, and accepted for publication October 12, 2001. Abbreviations: CI, confidence interval; OR, odds ratio; RR, relative risk. 1 Unit of Epidemiology, Scientific Institute of Public Health, Brussels, Belgium. 2 European Programme for Intervention Epidemiology Training (EPIET), Brussels, Belgium. Reprint requests to Dr. Herman Van Oyen, Unit of Epidemiology, Scientific Institute of Public Health, J. Wytsmanstraat 16, 1050 Brussels, Belgium (e-mail: [email protected]). Coca-Cola, Coca-Cola light, and Fanta are manufactured by The Coca-Cola Company, Atlanta, Georgia. Descriptive epidemiology To obtain contextual information regarding the outbreaks, qualitative and open-ended telephone interviews were conducted with the school directors, and a self-administered 140 Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on March 5, 2014 An epidemic of health complaints occurred in five Belgian schools in June 1999. A qualitative investigation described the scenario. The role of soft drinks was assessed by using a case-control study. Cases were students complaining of headache, dizziness, nausea, vomiting, abdominal pain, diarrhea, or trembling. Controls were students present at school on the day of the outbreak but not taken ill. An analysis was performed separately for school A, where the outbreak started, and was pooled for schools B–E. In school A, the attack rate (13.2%) was higher than in schools B–E (3.6%, relative risk = 3.6, 95% confidence interval (CI): 2.5, 5.3). Exclusive consumption of regular Coca-Cola (school A: odds ratio (OR) = 29.7, 95% CI: 1.32, 663.6; schools B–E: OR = 7.3, 95% CI: 2.9, 18.0) and low mental health score (school A: OR = 16.1, 95% CI: 1.3, 201.9; schools B–E: OR = 3.1, 95% CI: 1.5, 6.6) were independently associated with the illness. In schools B–E, consumption of Fanta, consumption of Coca-Cola light, and female gender were also associated with the illness. It seems reasonable to attribute the first cases of illness in school A to regular Coca-Cola consumption. However, mass sociogenic illness could explain the majority of the other cases. Am J Epidemiol 2002;155:140–7. Belgian Coca-Cola-related Outbreak 141 questionnaire was distributed to the physicians in the emergency rooms. Cases were defined as students in school A or in schools B–E who suffered from at least one of the following complaints on the first day or on the second day after the onset of the outbreak in each school: headache, dizziness, nausea, vomiting, abdominal pain, diarrhea, or trembling. Cases were identified according to a school register of illnesses filled out by a nurse. Emergency room and hospital medical records were then checked for the symptoms and results of physical examinations. Results of biologic and toxicologic tests of blood and urine samples were collected. Case-control study Analysis Crude and gender-specific attack rates in school A and schools B–E were calculated by dividing the number of cases by the number of students. The mean ages of cases in school A and schools B–E were compared by using the Kruskal-Wallis test. To compare exposures between cases and controls, odds ratios and 95 percent confidence intervals were computed by using Epi Info software (version 6.04; Centers for Disease Control and Prevention, Atlanta, Georgia). Exposures found by univariate analysis to be associated with the illness (p value < 0.2) were included in a Am J Epidemiol Vol. 155, No. 2, 2002 RESULTS Scenario of the outbreaks On June 8, 1999, students in school A complained of gastrointestinal symptoms with dizziness and headache shortly after consuming Coca-Cola from the school restaurant during the midday break. Between 12:30 and 1:00 p.m., three students reported to the secretarial office with health complaints. At 1:10 p.m., the courses started; by 2:00 p.m., six more students from different classes complained of feeling ill. Following advice from the medical school inspector, all ill students were taken to the local hospital. Of the 33 students who went to the emergency unit on June 8, 12 were hospitalized overnight for observation. Six other students were taken to the same hospital on June 9. With students reporting a “rotten” smell from Coca-Cola bottles, a possible link was made with consumption of this beverage. In the afternoon of June 8, the supplier of Coca-Cola removed most of the remaining crates from the school. On June 10 (school B), June 11 (school C), and June 14 (schools D and E), students complained of similar symptoms. In school E, the chief of police ordered all students reporting complaints to be sent to the hospital. In school C, a physician came on site to evaluate the situation. On the basis of the number of students with health complaints, several ambulances and a medical emergency team were sent to the school. In schools B–E, all ill students were transported to the local hospital either by ambulance or in staff members’ cars. Following the hotline instructions of The CocaCola Company, the staff of school E had removed all cans stamped on the bottom with specific codes early on the morning of June 14 before the courses started. Because of extensive media attention given to the outbreak in school A, the events were assumed to be related to consumption of Coca-Cola Company soft drinks. Descriptive epidemiology In school A, two potential cases did not meet the eligibility criteria. The overall attack rate was 13.2 percent. Thirtyone students became ill on the first day and six on the following day (figure 1). In schools B–E, 72 cases occurred on the first day of the outbreaks and three on the following day (figure 2). The attack rate (3.5 percent) was lower in schools B–E than in school A (relative risk (RR) 0.28, 95 percent confidence interval (CI): 0.19, 0.40; table 1). In addition, the attack rate was higher for girls than for boys in both school A (RR 1.8, 95 percent CI: 0.9, 3.6) and schools B–E (RR 5.7, 95 percent CI: 1.8, 17.9). In school A, the cases were younger than those in schools B–E (p < 0.001, table 1). In school A, with onset of illness occurring before 2:10 p.m.—the first break period after the incident started—the first nine cases of illness occurred among students in six different classrooms (one to three cases per class); 20 (71.4 percent) ill students of the remaining 28 cases were grouped in four classrooms (four to six cases per class). In schools Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on March 5, 2014 Controls were students from the same class as the cases and were next on the alphabetic list; these students were present at school on the first and the following day of the outbreak and had not been ill during the 2 weeks before the day of the outbreak up to the following day. A 50 percent prevalence of exposure to Coca-Cola Company soft drinks among healthy students and a 1:2 ratio of cases to controls were assumed; therefore, a sample size of 49 cases and 98 controls in each school group was required to detect an odds ratio of 3 or greater with 95 percent confidence and a power of 80 percent. Exposure to Coca-Cola Company soft drinks was defined in two different ways. The first definition was consumption of regular Coca-Cola compared with any other consumption (other Coca-Cola Company products, non-Coca-Cola Company products, water, or no consumption at all). According to the second definition, the exposure was exclusive consumption of a Coca-Cola Company product compared with water or no consumption at all. A structured questionnaire was used, and information was collected on demographics (gender, age), food consumption (place, time) and beverage consumption (place of purchase, place and time of consumption, package, particular taste or smell) on the day of the outbreak, illness among friends, mental health status, and symptoms (type, time of occurrence). A mental health score was calculated according to responses to the questions on mental health status on the SF36 Health Survey (3). Information was gathered during a face-to-face interview in each school between June 25 and June 27. In one of the B–E schools, students were grouped and were assisted by an interviewer when completing the questionnaire. logistic regression model by using a forward stepwise selection strategy (SPSS 8.0; SPSS Inc., Chicago, Illinois). 142 Gallay et al. FIGURE 1. Number of cases of Coca-Cola-related illness, by gender and time of onset of illness, school A, Belgium, 1999. Coca-Cola is manufactured by The Coca-Coca Company, Atlanta, Georgia. Case-control study Because of practical constraints in schools to interviewing during the end-of-school examinations, fewer than one control per case in school A and fewer than two controls per case in schools B–E could be interviewed. In school A, the age and gender of the 37 cases and 34 controls were similar. In schools B–E, the 75 cases and 130 controls were similar in age, but cases were more likely to be girls (odds ratio (OR) 4.3, 95 percent CI: 1.2, 23.6). The proportion of students that had bought and consumed regular Coca-Cola at school was higher among cases than controls in both school A (OR 36.8, 95 percent CI: 7.8, 220.1) and schools B–E (OR 3.5, 95 percent CI: 1.7, 7.0) (table 3). In school A, cases also were more likely to consume regular Coca-Cola exclusively (OR 23.2, 95 percent CI: 3.7, 235.5). Exclusive consumption of other beverages (other Coca-Cola and non-Coca-Cola products) was similar among cases and controls (table 4). In schools B–E, cases were more likely to exclusively consume regular Coca-Cola (OR 5.5, 95 percent CI: 2.4, 12.9), Fanta (OR 3.5, 95 percent CI: 1.1, 10.9), or Coca-Cola light (OR 12.8, 95 percent CI: 2.8, 77.9) (table 4). In both school groups, cases were more likely to have a low mental health score; the odds ratio (2.4) was similar in both school groups but was statistically significant for only schools B–E (table 5). In school A, cases were more likely to report an off-odor (OR 43.2, 95 percent CI: 8.0, 407.4) or a bad taste (OR 28.0, 95 percent CI: 3.7, 1,206.7) to the regular Coca-Cola. About one third of the cases described the smell as nasty or rotten. In schools B–E, few cases and no controls noted a bad smell, and few cases and one control reported a bad taste. In both school groups A and B–E, having a friend become ill was not associated with the disease. Because school A did not provide food, the majority of students ate a homemade lunch. In schools B–E, cases were less likely than controls to have eaten the food provided by the school (table 5). The following variables were included in the multivariate model: age; gender; exclusive consumption of beverages (regular Coca-Cola, Fanta, Coca-Cola light, other soft drinks) as a set of dummy variables, with consumption of water or no consumption at all as the reference; mental health status; the reporting of an off-odor; and the reporting of a bad taste to soft drinks. In school A, exclusive consumption of regular Coca-Cola (OR 29.7, 95 percent CI: 1.32, 663.6) and having a low mental health score (OR 16.1, 95 percent CI: 1.3, 201.9) remained independently associated with the illness. In schools B–E, Am J Epidemiol Vol. 155, No. 2, 2002 Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on March 5, 2014 B–E, two thirds (51/75) of the cases were clustered in 11 classrooms. The number of cases per class ranged from three to nine. In school A, time of beverage consumption and time of onset of symptoms were available for 31 and 37 cases, respectively; in schools B–E, this information was available for 50 and 75 cases, respectively. In school A, all but three cases drank beverages between 12:00 and 12:30 p.m., with onset of symptoms 30 minutes to 24.5 hours (median, 3 hours) later. In schools B–E, the delay between consumption of soft drinks and occurrence of symptoms ranged from 30 minutes to 7.5 hours (median, 1.5 hours). There was no difference in time of onset of symptoms between girls and boys in both schools. Medical records could be checked for 32 of the 37 cases from school A and for 62 of the 75 cases from schools B–E. Headache, nausea, and dizziness were the main clinical symptoms reported by the first nine cases in school A. Abdominal pain, headache, nausea, and respiratory troubles were reported more frequently by the later cases. In schools B–E, headache, abdominal pain, nausea, and dizziness were the main clinical symptoms reported on the medical charts (table 2). Physical examination was normal for 27 (84.4 percent) and 56 (90.3 percent) of the patients from school A and schools B–E, respectively. Extreme pallor was noted for some of the first cases from school A, and flushed skin and/or red eyes were noted for six (9.7 percent) students from schools B–E. All symptoms disappeared spontaneously within several hours for the majority of patients. In each of the school groups A and B–E, 12 students were hospitalized for a period of 1–3 days. Six students from school A and two students from the other schools relapsed within a couple of days. Although blood and urine samples were collected from students in school A, no results from the routine biologic and toxicologic analysis could be obtained. The results of a range of routine biologic tests performed on the 56 (74.6 percent) blood samples and the seven (9.3 percent) urine samples taken from the 75 students from schools B–E were normal. Belgian Coca-Cola-related Outbreak 143 Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on March 5, 2014 FIGURE 2. Number of cases of Coca-Cola-related illness, by gender and time of onset of illness, schools B–E, Belgium, 1999. Coca-Cola is manufactured by The Coca-Coca Company, Atlanta, Georgia. exclusive consumption of regular Coca-Cola (OR 7.3, 95 percent CI: 2.9, 18.0), Fanta (OR 5.8, 95 percent CI: 1.7, 19.5), and Coca-Cola light (OR 15.7, 95 percent CI: 3.1, 78.2) remained independently associated with the illness. Girls (OR 4.2, 95 percent CI: 1.0, 16.5) and students with a low mental health score (OR 3.1, 95 percent CI: 1.5, 6.6) also were more likely to have reported the illness. Am J Epidemiol Vol. 155, No. 2, 2002 DISCUSSION The epidemiologic investigation suggested that consumption of regular Coca-Cola was a strong determinant of illness in school A. The short interval between exposure to the soft drink and occurrence of symptoms favored a toxicologic cause (table 6). No other soft drink or food item was associated with becoming ill. Regular Coca-Cola was the 144 Gallay et al. TABLE 1. Descriptive epidemiology of Coca-Cola*-related illness, school A and schools B–E, Belgium, 1999 Schools B–E School A Attack rate (%) Among girls Among boys Relative risk of being ill according to female gender, by school TABLE 3. Exposure to regular Coca-Cola*,† in school A and schools B–E, Belgium, 1999 13.2 (37/280) 15.6 (28/179) 8.9 (9/101) 3.5 (72/2,060) 4.3 (72/1,672) 0.7 (3/388) 1.8 5.7 (95% CI†: 0.9, 3.6) (95% CI: 1.8, 17.9) Range, 13–15; Range 13–19; Age (years) median, 13 median, 15 * Coca-Cola is manufactured by The Coca-Cola Company, Atlanta, Georgia. † CI, confidence interval. School A Symptom Cases with onset of illness before 14:10 (first cases) (n = 9) No. % Later cases (n = 23) No. % Schools B–E (n = 62) No. % 13 56.5 48 77.8 7 Headache 77.4 7 30.4 36 66.7 6 Nausea 58.1 3 13.0 19 44.4 4 Dizziness 30.6 Abdominal 15 65.2 39 33.4 3 pain 62.9 5 21.7 0 22.2 2 Asthenia Respiratory 7 30.4 3 11.1 1 troubles 4.8 6 26.0 18 11.1 1 Trembling 29.0 2 2.7 3 11.1 1 Weakness 4.8 2 8.7 8 0 Vomiting 12.9 1 4.3 7 0 Diarrhea 11.3 Heart rate ≥100 2 10.0 5 33.4 3 /minute† 12.5 1 7.1 3 0 Fever ≥38˚C‡ 8.1 * Coca-Cola is manufactured by The Coca-Cola Company, Atlanta, Georgia. † Pulse rate was known for the first cases and 20 of the later cases in school A and for 40 cases in schools B–E. ‡ Fever was known for 7 of the first cases and 14 of the later cases in school A and for 37 cases in schools B–E. only soft drink that the students characterized as having a rotten smell, typical of carbonyl sulfide and hydrogen sulfide contaminating the carbon dioxide used in the beverage. In the sensory analysis conducted by The Coca-Cola Company’s Northwest Europe Division, a clear off-odor was established. The sulfur-containing compound responsible for this off-odor was detected in the regular Coca-Cola consumed in school A by gas chromatography in combination with a sniffing technique (GC-SNIFF) (4). Moreover, the main symptoms observed (headache, nausea, and dizziness), particularly among the first cases, were compatible Cases Controls 95% CI‡ OR‡ No. % No. % School A Yes No Total 34 3 37 91.9 8.1 8 26 34 23.5 76.5 36.8 Schools B–E Yes No Total 31 44 75 41.3 58.7 22 108 130 16.9 83.1 3.5 7.8, 220.1 1.7, 7.0 * Coca-Cola is manufactured by The Coca-Cola Company, Atlanta, Georgia. † Regular Coca-Cola bought and consumed at school on the day of the outbreak. ‡ OR, odds ratio; CI, confidence interval. TABLE 4. Exclusive exposure to specific beverages* in school A and schools B–E, Belgium, 1999 Cases Controls (no.) (no.) Beverage School A (glass bottles) Regular Coca-Cola Fanta Coca-Cola light Other Coca-Cola Company products Non-Coca-Cola Company products Water/no drink OR† 95% CI† 31 0 0 8 9 0 23.2 0 3.7, 235.5 0, 8.3 1 3 2.0 0, 50.6 0 2 2 12 0 Reference 0, 47.2 Schools B–E (cans and bottles) 26 20 5.5 Regular Coca-Cola 2.4, 12.9 9 11 3.5 Fanta 1.1, 10.9 9 3 12.8 Coca-Cola light 2.8, 77.9 Other Coca-Cola Company 5 5 4.3 products 0.9, 20.3 Non-Coca-Cola Company 1 5 0.8 products 0, 8.5 19 81 Reference Water/no drink * Beverages bought and consumed at school on the day of the outbreak. Coca-Cola and Fanta are manufactured by The Coca-Cola Company, Atlanta, Georgia. † OR, odds ratio, CI, confidence interval. with the expected symptoms of exposure to carbonyl sulfide and hydrogen sulfide (4, 5). Brief exposure to hydrogen sulfide can be sufficient to induce such symptoms (5, 6). The first cases may have been exposed to a higher concentration of carbonyl sulfide and hydrogen sulfide, while the later cases may have paid attention when opening and drinking the soft drink. It is very unfortunate that, although blood and urine samples were collected from school A students at the local hospital, there was no evidence that any analyses were conducted; if there were, the results were unobtainable. In schools B–E, a similar but weaker association with regular Coca-Cola consumption was observed. Other CocaCola Company soft drinks (Fanta, Coca-Cola light) were also identified as risk factors. A sensory analysis of the outside of the cans from the production site supplying schools B–E detected a “medicine-like” odor, and the GC-MS technique (gas chromatography in combination with mass specAm J Epidemiol Vol. 155, No. 2, 2002 Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on March 5, 2014 TABLE 2. Symptoms of Coca-Cola*-related illness abstracted from the medical records, school A and schools B–E, Belgium, 1999 Consumption of regular Coca-Cola Belgian Coca-Cola-related Outbreak 145 TABLE 5. Exposure to other risk factors in school A and schools B–E, Belgium, 1999 Cases Controls (no.) (no.) Risk factor School A† Food provided by school Having a friend be ill Mental SF-36 Health Survey score§ <median Bad smell Bad taste —‡ 36 —‡ 30 4.8 20¶ 27 17 11¶ 2 1 2.4 43.2 28.0 95% CI* 0.4, 242.8 0.8, 7.2 8.0, 407.4 3.7, 1,206.7 6 65 26 101 0.4 1.9 0.1, 0.9 0.8, 4.6 47 6 11 54 0 1 2.4 1.3, 4.4 22.2 3.1, 961.5 * OR, odds ratio; CI, confidence interval. † Cases, n = 37; controls, n = 34. ‡ No food was provided at school A. § Refer to Ware et al. (3). ¶ 1 missing value among the cases and 2 missing values among the controls. # Cases, n = 75; controls, n = 130. trometry) revealed the presence of a very low concentration of chlorocresol (4-chloro-3-methylphenol) on the external TABLE 6. Evidence* for a causal toxicity connection versus evidence for a mass sociogenic illness with Coca-Cola†-related illness in school A and schools B–E, Belgium, 1999 Evidence for a Causal toxicity Consumption of soft drink associated with the illness Short interval between exposure to soft drink and occurence of symptoms Typical odor Toxicologic analyses Compatibility between observed and expected symptoms Mass sociogenic illness Classic risk factors of mass sociogenic illness (occurence among adolescents, girls, clusters; unusual mental stress) Benign morbidity, no clinical or laboratory evidence, relapse of illness Rapid spread and dissolution Identification of a trigger: a bad odor No readily apparent environmental cause Person-to-person transmission by line-of-sight or audiovisual cues Role of the media in transmission School A (Regular Coca-Cola) (Several beverages) (Regular Coca-Cola: typical of carbonyl sulfide and hydrogen sulfide) (of exposure to (of exposure to carbonyl sulfide and chlorocresol) hydrogen sulfide) * , strong evidence; , evidence; , no evidence. † Coca-Cola is manufactured by the Coca-Cola Company, Atlanta, Georgia. Am J Epidemiol Vol. 155, No. 2, 2002 Schools B–E Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on March 5, 2014 Schools B–E# Food provided by school Having a friend be ill Mental SF-36 Health Survey score§ <median Bad smell Bad taste OR* surface of the cans from the French plant (Dunkerque) (4). However, in schools B–E, the observed symptoms were not compatible with the expected symptoms (either eye and skin irritation as a result of dermal contact or severe effects on mucous membranes caused by ingestion) of the p-chloro-mcresol exposure (4, 5). None of these symptoms was reported by the students. In addition, all laboratory results and physical examinations were normal (table 6). Bacterial, viral, or parasitologic investigations carried out by The Coca-Cola Company laboratory were all negative. The company submitted the results of chemical analyses to a laboratory for toxicologic advice (4). According to the toxicologic reports, it is unlikely that such low concentrations of carbonyl sulfide, hydrogen sulfide, and 4-chloro-3methylphenol could have caused any toxicity. However, in 1962, Hall noted that water carbonated with carbon-dioxidecontaining carbonyl sulfide produced detectable hydrogen sulfide some hours later, becoming stronger and disappearing after a few days (7). The beverages consumed in school A on the day of the outbreak had been produced just 4 days earlier; all toxicologic analyses were performed only several days after the day of the outbreak. Since carbonyl sulfide and hydrogen sulfide concentrations are known to decrease over time (7), the fact that these test results showed low or undetectable levels of gases may not be a meaningful indicator of a potential exposure. It is unclear whether the beverages produced on June 4 were supplied to other places within the same time delay and whether the samples taken for analyses came from the lots suspected of causing the illness (4). 146 Gallay et al. transmitting the outbreak from school to school and from school to the general population (11, 23–25). Moreover, lack of transparency about the safety of the Coca-Cola product and controversial information from officials intensified the community’s concern. The high awareness and anxiety about the safety of modern food products, combined with a very strong symbolic image of the incriminated product, may have contributed further to the psychosocial distress of the general population (1, 2, 22). When a possible outbreak of mass sociogenic illness is investigated, a number of difficulties arise in balancing competing requirements. These include ensuring, as quickly as possible, that no toxicologic cause exists; identifying the existence of an outbreak of mass sociogenic illness; and communicating the diagnosis in order to stop the spread of the illness. However, because of the lack of pathognomonic indicators of mass sociogenic illness and the difficulty in proving the presence of a toxicologic substance, investigations often have become extensive before a diagnosis can be stated, which can increase stress (13, 16, 17). Even anxiety can cause real symptoms; if cases are told that their source of anxiety is a false belief, this explanation will not reduce their embarrassment and can exacerbate their condition (26). In this outbreak, interrupting the transmission of symptoms by separating exposed groups and suppressing audiovisual transmission seemed practically impossible (16). Because of the great distribution of the soft drinks, and because the symbolic image of The Coca-Cola Company is so well known and is highlighted by extensive media coverage, it would have been difficult to stop the process without a quick and complete analysis of the incriminated product and clear information about the safety of the soft drink. Such outbreak investigations need transparency, objective features, and clear information from the different actors implicated. In the present study, several deficiencies in crisis management were identified. For example, the Ministry of Public Health was already very involved in the dioxin crisis, public measures for withdrawing the implicated products were insufficient, and The Coca-Cola Company performed nearly all soft drink analyses. In such a situation, credibility of the results and the official information depends on the absence of a conflict of interest. The findings of this study are subject to a number of limitations. First, the survey was performed 2 weeks after the outbreak, during the last week of end-of-school examinations, which could have introduced recall errors that particularly affected time of beverage consumption and time of onset of illness. This information was probably more precise in school A, where all soft drinks were sold at noon break only. Furthermore, it could be argued that in school A, a selection of cases on the exposure may have induced the association between consumption of regular Coca-Cola and symptoms. This possibility seems highly unlikely. The proportion of cases that consumed regular Coca-Cola was 100 percent among the first nine cases compared with only 87 percent among cases whose onset of symptoms occurred after the afternoon break between classes. If selection of cases would have occurred based on the exposure, one would expect the opposite, with the highest proportion of Am J Epidemiol Vol. 155, No. 2, 2002 Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on March 5, 2014 Previous papers have attributed, without investigation, the outbreaks to episodes of mass sociogenic illness (9, 10). In this study, classic risk factors for mass sociogenic illness were identified in both school groups A and B–E (table 6). The outbreak was characterized by occurrence among adolescents or preadolescents in a school setting, a preponderance of illness among girls, clustering of cases in classrooms, evidence of unusual mental stress among those reporting illness, benign morbidity and no clinical or laboratory evidence of illness, relapse of illness, and rapid spread and dissolution of the outbreak (11–15). In both school groups, cases more often than controls had reported a friend being ill on the day of the outbreak (16), although the association was not statistically significant. However, it is likely that friends consume similar products or even share them. Similar to several outbreaks of mass sociogenic illness described previously, a bad odor of a “gas” was identified by students and could have been a trigger (11–14, 16–19). Jones et al. propose that mass sociogenic illness be considered in any outbreak of acute illness thought to be caused by exposure to a toxic substance but with minimal physical findings and no environmental cause readily apparent to the investigator (17). Nonetheless, regular Coca-Cola was clearly identified in school A as having a “rotten egg” odor, which is typical of the odor of sulfide. This was not usual in previous outbreaks of mass sociogenic illness in which airborne substances were commonly incriminated with unspecific sources (14, 18–20) or, conversely, many vectors were incriminated (21). In schools B–E, the odor described varied. In both school A and schools B–E, belonging to a group having a low mental health score was independently associated with illness and could highlight a stressful situation being experienced by the ill students. First, the outbreak took place during the end-of-school examination period (11). Second, this outbreak occurred within the context of the recent Belgian general election and a dioxin crisis in Belgium 2 weeks earlier, which had heightened anxiety in the population about food safety (22). The SF-36 Health Survey scores addressed the feelings of the students during the previous 4 weeks and provided only an indicator of mental health status. Many students reported that it was difficult to answer this question, because they felt bad since the incident had occurred. Nonetheless, since mental health status could easily modify symptom severity or a person’s behavior, the associations with low mental health scores do not disprove chemical contamination as a cause of the outbreaks. More complex instruments have been proposed to evaluate mass sociogenic illness (14). Under the hypothesis of a mass sociogenic illness, several features enhanced contagion of the outbreak (table 6). In schools B–E, as in many previous outbreaks, the arrival of ambulances with emergency personnel and the police could have increased the general excitement and anxiety, resulting in the spread of symptoms (11, 16). In addition, propagation of the illness accelerated with person-to-person transmission when students were grouped (during break periods, school lunch) by line-of-sight or audiovisual cues (16). Extensive nationwide radio and television media coverage of the first incident in school A probably played a substantial role in Belgian Coca-Cola-related Outbreak 147 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. ACKNOWLEDGMENTS 16. The authors acknowledge the directors of the five schools (Bornem, Brugge, Harelbeke, Kortrijk, and Lochristi) for allowing the investigation to be conducted in the school setting and for providing the information on the scenario. They also thank the physicians of the hospitals for providing the medical information. 17. 18. 19. 20. REFERENCES 1. De Schuiteneer B, Goossens E. Affaire Coca-Cola Belgique— Juin 1999. Observations du Centre Antipoisons portant sur la période du 08/06/1999. (In French). Brussels, Belgium: Centre Antipoisons, 1999. 2. Institut de Veille Sanitaire. Etude descriptive et surveillance des manifestations symptomatiques rapportées par les consommateurs de boissons commercialisées par la marque CocaCola en juin 1999. (In French). Paris, France: Institut de Veille Sanitaire, 1999. 3. Ware J, Snow KK, Kosinski E, et al. SF-36 Health Survey: manual and interpretation guide. Boston, MA: The Health Institute, New England Medical Center, 1993. 4. Dansk Toksikologi Center. Report on the analysis of the prod- Am J Epidemiol Vol. 155, No. 2, 2002 21. 22. 23. 24. 25. 26. ucts of The Coca-Cola Company, June 1999. Copenhagen, Denmark: Dansk Toksikologi Center, 1999. Online carbonyl sulphide—HSDB-Hazardous Substances Data Bank. October 1, 1999. (http://sis.nlm.nih.gov/sis1). Goode D. Mass psychogenic illness attributed to toxic exposure at a high school. (Letter). N Engl J Med 2000;342:1673–5. Cox J. Determination of carbonyl sulfide concentration in carbon dioxide for quality control. Fort Lauderdale, FL: International Society of Beverage Technologists, 1997. Radford-Knoery J, Cutter G. Biogeochemistry of dissolved hydrogen sulfide species and carbonyl sulfide in the western North Atlantic Ocean. Geochimica et Cosmochimica Acta 1994;58:5421–31. Nemery B, Fischler B, Boogaerts M, et al. Dioxins, CocaCola, and mass sociogenic illness in Belgium. (Letter). Lancet 1999;354:77. Coke adds life, but cannot always explain it. (Editorial). Lancet 1999;354:173. Small GW, Borus JF. Outbreak of illness in a school chorus: toxic poisoning or mass hysteria? N Engl J Med 1983;308: 632–5. Boss LP. Epidemic hysteria: a review of the published literature. Epidemiol Rev 1997;19:233–43. Small GW, Feinberg DT, Steinberg D, et al. A sudden outbreak of illness suggestive of mass hysteria in schoolchildren. Arch Fam Med 1994;3:711–16. Goh KT. Epidemiological enquiries into a school outbreak of an unusual illness. Int J Epidemiol 1987;16:265–70. Levine RJ. Epidemic faintness and syncope in a school marching band. JAMA 1977;238:2373–6. Cole TB, Chorba TL, Horan JM. Patterns of transmission of epidemic hysteria in a school. Epidemiology 1990;1:212–18. Jones TF, Craig AS, Hoy D, et al. Mass psychogenic illness attributed to toxic exposure at a high school. N Engl J Med 2000;342:96–100. Philen R, Kilbourne E, McKinley T, et al. Mass sociogenic illness by proxy: parentally reported epidemic in an elementary school. Lancet 1989;2:1372–6. Levine RJ, Sexton DJ, Romn FJ, et al. Outbreak of psychosomatic illness at a rural elementary school. Lancet 1974;2: 1500–3. Smith HC, Eastham EJ. Outbreak of abdominal pain. Lancet 1973;2:956–8. Desenclos JC, Gardner H, Horan M. Mass sociogenic illness in a youth center. Rev Epidemiol Sante Publique 1992;40:201–8. Van Oyen H. Dioxin in feed and food: is public health running behind? J Epidemiol Community Health 1999;53:744–5. Seydlitz R, Spencer JW, Laska S, et al. The effects of newspaper reports on public’s response to a natural hazard event. Int J Mass Emergencies Disaster 1991;9:5–29. Small GW, Borus JF. The influence of newspaper reports on outbreaks of mass hysteria. Psychiatr Q 1987;58:269–78. Hefez A. The role of the press and the medical community in the epidemic of “mysterious gas poisoning” in the Jordan West Bank. Am J Psychiatry 1985;142:833–7. Rifkin A. Mass psychogenic illness attributed to toxic exposure at a high school. (Letter). N Engl J Med 2000;342: 1673–5. Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on March 5, 2014 exposure among the cases identified later, once regular Coca-Cola was suspected as a cause of symptoms. Factors that contribute to the possible existence and role of mass sociogenic illness are difficult to discern. Mental health status at the time was influenced by many factors, not the least of which was the end-of-year examinations. There also may have been an overestimation of a low mental health score among cases. Finally, extensive media coverage of the incident in school A could have introduced an information bias and resulted in overidentification of cases in schools B–E. In conclusion, an association was observed between consumption of regular Coca-Cola and illness in school A. Exposure to carbonyl sulfide and hydrogen sulfide could explain the observed symptoms, particularly in early cases. Nonetheless, classic factors of mass sociogenic illness were present and could explain the majority of the later cases in both schools A and B–E. Some limitations of this investigation were related to deficiencies in managing the crisis. This problem underscores the need for appropriate independent structures able to analyze readily identified causes and to react quickly.
