Skip Navigation

Signs and Symptoms Predictive of Death in Patients with Foodborne Botulism—Republic of Georgia, 1980–2002

  1. Jay K. Varma1,2,a,
  2. Guram Katsitadze4,
  3. Maia Moiscrafishvili4,
  4. Tamar Zardiashvili4,
  5. Maia Chokheli4,
  6. Natalia Tarkhashvili4,
  7. Ekaterina Jhorjholiani4,
  8. Maia Chubinidze4,
  9. Teimuraz Kukhalashvili4,
  10. Irakli Khmaladze4,
  11. Nelli Chakvetadze4,
  12. Paata Imnadze4,
  13. Mike Hoekstra3, and
  14. Jeremy Sobel2
  1. 1Epidemic Intelligence Service, Epidemiology Program Office, Atlanta, Georgia
  2. 2Foodborne and Diarrheal Diseases Branch, Atlanta, Georgia
  3. 3Biostatistics and Information Management Branch, Division of Bacterial and Mycotic Disease, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, Georgia
  4. 4National Center for Disease Control, Tbilisi, Republic of Georgia
  1. Reprints or correspondence: Dr. Jay K. Varma, CDC/HIV, Box 68, American Embassy, APO, AP 96546 (jvarma{at}cdc.gov).

  • ↵a Present affiliation: Thailand MOPH–US CDC Collaboration, Nonthaburi, Thailand.

 
Next Section

Abstract

Foodborne botulism is a severe, potentially fatal disease characterized by cranial nerve palsies and descending paralysis. Little is known about signs and symptoms predictive of death from botulism. We studied patients with botulism in the Republic of Georgia, which has the highest reported rate of foodborne botulism in the world. After abstracting medical records of patients with botulism who were hospitalized during 1980–2002, we performed classification-and-regression-tree analysis to identify clinical syndromes predictive of survival and death. We identified records for 706 patients hospitalized for foodborne botulism from 1980–2002. Trivalent antitoxin was administered to 623 patients (88%). Fifty-four (8%) died. Patients with shortness of breath and impaired gag reflex and without diarrhea were 23 times more likely to die than were patients without this syndrome. Validating this clinical prediction rule may help reduce mortality from botulism in Georgia. Validation in other settings could help public health preparations for large outbreaks of naturally occurring or bioterrorism-related botulism.

Botulism is a severe paralytic disease caused by neurotoxins produced by the gram-positive, anaerobic, spore-forming bacterium Clostridium botulinum. Four naturally occurring botulism syndromes exist: foodborne, wound, infant, and adult intestinal colonization. Of these, foodborne botulism is the greatest public health concern, because of its epidemic potential [1].

Foodborne botulism is caused by consumption of food contaminated with botulinum toxin, the most potent biological toxin known to man [2, 3]. Three types of neurotoxins (A, B, and E) are responsible for most cases of foodborne botulism [4]. Spores of C. botulinum are ubiquitous in soil and readily contaminate foods; however, they germinate and elaborate toxin only under a rare confluence of conditions, including anaerobic environments with low concentrations of salt, sugar, and acid [2]. Paralysis in cases of botulism may extend to respiratory muscles, with 60% of patients in the United States requiring mechanical ventilation [4, 5]. Case fatality rates in the United States range from 5% to 15%; worldwide, the case fatality is estimated to be 10% [4, 6–8].

The clinical syndrome of botulism has been well characterized, yet little is known about clinical features that predict severe outcomes, such as death. This is because most of what we know about botulism is from nonuniform, aggregate data culled from small outbreak investigations in the United States [6]. Identifying risk factors for severe outcomes could help triage large numbers of ill persons in a massive outbreak. This is especially critical, because botulinum toxin is categorized as a class A biological agent by the Centers for Disease Control and Prevention (CDC; Atlanta, GA); the toxin has been weaponized and could be deliberately disseminated by aerosol or by contamination of foods or beverages [2, 9].

The Republic of Georgia, a small, mountainous country of 4.4 million persons in the southern Caucasus, has the world's highest reported incidence of foodborne botulism: 0.9 cases per 100,000 population [10]. To learn more about the clinical features of foodborne botulism, we reviewed medical records of patients with botulism who were hospitalized in Georgia during 1980–2002.

Previous SectionNext Section

Methods

We visited hospitals in cities and regions of Georgia that reported at least 1 botulism case from 1980–2002. At each hospital, medical records were sorted by discharge diagnosis. Because there is no formal case definition used by all clinicians, a patient was considered to have botulism if medical records indicated that this was the final diagnosis. For each patient, a trained epidemiologist completed a standardized data abstraction form. This form asked for data about patient demographic characteristics; medical history; history of present illness; physical examination findings at hospital admission; time course of admission, diagnosis, treatment, and hospital discharge; clinical course of disease, including adverse reactions, complications, and death; suspected source of botulism; and results of laboratory tests for botulism.

Patients were classified into those who survived and those who died. Because Georgian physicians rarely discharge patients with botulism before achievement of complete recovery, we did not attempt to ascertain patient outcomes after hospital discharge. Outcome data was not available for 1 patient.

A staff physician at the Tbilisi Infectious Pathology Center, which supplied data on most of the patients in this study, compared a random sample of data from 94 patient records (13%) with data on corresponding completed abstraction forms to insure accuracy of data abstraction. After entering data into an electronic database, we compared all electronic and paper records for all patients to ensure accuracy of data entry.

Diagnostic testing of food or human specimens was performed at Georgia's National Center for Disease Control (NCDC; Tbilisi, Georgia) using the standard mouse bioassay for detection of botulinum toxin [4].

For continuous variables, we performed the Wilcoxon rank sum test to compare medians and Student's t test to compare means. For categorical variables, we performed χ2 analysis or, when appropriate, Fisher's exact test to compare proportions. All P values were 2-sided, and values <.05 were considered to be significant. Analyses were conducted with SAS, version 9.0 (SAS Institute).

We performed classification and regression tree (CART) analysis to identify clinical syndromes at the time of initial presentation that were predictive of death or survival. CART analysis progressively classifies patients into subgroups on the basis of patient attributes and the proportion within each subgroup that has the outcome of interest [11]. In the first step of CART analysis, a single best predictor of the outcome is selected, and the study sample is divided into subgroups on the basis of the presence or absence of this predictor. The process is repeated for each subgroup, with a best predictor of the outcome chosen for each initial subgroup and subsequent subgroups until further partitioning does not add to the predictive value of the model—a process known as recursive partitioning. In this analysis, we included age as a categorical variable (<18, 18–64, and ⩾65 years) and included all symptoms and physical examination signs that were recorded at admission to the hospital. We did not include variables related to prior medical conditions, because of difficulty in verifying the accuracy and assessing the severity of these conditions. We did not include variables associated with in-hospital care, such as antitoxin treatment or mechanical ventilation, because our goal was to assess clinical predictors of survival and death that were measurable at the time of diagnosis, rather than to assess factors associated with optimum patient care. Analyses were conducted with JMP, version 5.0 (SAS Institute). Partitions were chosen on the basis of large values of the likelihood-ratio χ2 (i.e., G2) statistic. We set the minimum partition size at 10. We used cross-validation as a model-fit diagnostic test [12].

The studies reported in this article were reviewed by the human subjects committee at the CDC and at the NCDC.

Previous SectionNext Section

Results

Patient characteristics. We identified medical records for 706 persons hospitalized with botulism during 1980–2002 at 8 hospitals located across Georgia. During this same period, 879 botulism cases were reported to the NCDC. Even though patients identified through medical records resided across 8 of 10 different regions of Georgia, 90% were initially hospitalized at or transferred to the Tbilisi Infectious Pathology Center, which serves as the national referral hospital for botulism.

The median age of patients was 34 years (range, 1–90 years). Three hundred and fifty-five patients (50%) were female. Patient ethnicities included Georgian (73%), Azerian (11%), Armenian (10%), and Russian (5%). The most common previous medical condition was high blood pressure (9% of patients).

All botulism cases were suspected or confirmed to be foodborne. The median incubation period, defined as the interval between exposure to suspected contaminated food and symptom onset, was 1 day (n = 682; range, 0–12 days); among the 199 patients for whom the exact interval in hours was known, the median incubation period was similar (median, 20 h; range, 1–218 h). When restricted to the 70 patients known to have type B intoxication, the median incubation period was also 1 day, whether measured in days (for 70 patients) or in hours (for 16 patients).

Among all patients, the median interval between symptom onset and hospital admission was 2 days (n = 697; range, 0–19 days). The most common symptoms at admission to the hospital were fatigue (90% of patients), muscle weakness (89%), and difficulty swallowing (81%) (figure 1). The most common physical examination findings were ophthalmoplegia (79% of patients), ptosis (76%), and slurred speech (58%). Four hundred and eighty-one patients (68%) had at least 3 of the 5 symptoms that are part of the so-called botulism diagnostic pentad (nausea and vomiting; dysphagia; diplopia; dry mouth; and dilated and fixed pupils); only 13 (2%) had all 5 symptoms [8].

Figure 1
View larger version:
Figure 1

Frequency of signs and symptoms at admission among 706 patients hospitalized with botulism in Georgia, 1980–2002

Six hundred and twenty-three patients (88%) received trivalent botulinum antitoxin. Antitoxin was administered a median of 3 days (n = 610; range, 0–11 days) after exposure to contaminated food and 2 days (n = 616; range, 0–10 days) after symptom onset. Allergic reactions to antitoxin occurred in 68 patients (11%). The most common reaction was rash, which occurred in 63 patients (10%).

Mechanical ventilation was initiated in 93 patients (13%). The median interval between exposure and mechanical ventilation was 3 days (n = 90; range 0–17 days), and the median interval between symptom onset and mechanical ventilation was 2 days (n = 93; range: 0–14). Among 51 patients who underwent mechanical ventilation and survived, the median length of intubation was 14 days (range, 0–82 days). Pneumonia developed in 96 patients (14%). Among 705 patients for whom a final outcome was known, 54 (8%) died. The median interval between exposure to contaminated food and death was 7 days (n = 50; range, 1–93 days), and the median interval between symptom onset and death was 4 days (n = 54; range, 0–92 days). The most frequently listed causes of death were cardiac arrest (39 patients [74%]) and respiratory failure (11 patients [21%]).

Clinical factors associated with survival and death. In univariate analysis, a large number of patient characteristics were associated with increased odds of death. Twelve patients (4%) aged 18–39 years died, compared with 27 patients (11%) aged 40–64 years (OR, 3.2; 95% CI, 1.6–6.4) and 10 patients (25%) aged ⩾65 years (OR, 8.4; 95% CI, 3.3–21). Neither ethnicity (P = .75) nor sex (P = .12) was associated with death. The odds of death were greatest for patients who had a history of congestive heart failure (OR, 18.0; 95% CI, 8.4–38.8), complained of shortness of breath at admission (OR, 22.1; 95% CI, 7.9–62), or required mechanical ventilation (OR, 36.9; 95% CI, 18.6–73.3) (table 1).

Figure 2
View larger version:
Figure 2

Classification and regression tree analysis of clinical syndromes predictive of death and survival among patients hospitalized with botulism in Georgia, 1980–2002. Arrows pointing to the left, factors predictive of survival; arrows pointing to the right, factors predictive of death.

Table 1
View larger version:
Table 1

Characteristics at hospital admission and risk of mortality for patients with botulism in Georgia, 1980–2002.

Patients who died had the same incubation period as did those who survived (median duration of incubation, 1 day for both groups; P = .71) but were admitted to the hospital sooner after symptom onset (median time to admission, 1 day for those who died vs. 2 days for those who survived; P < .01). Patients who died were equally likely to receive antitoxin (OR, 1.1; 95% CI, 0.4–2.6) and, in fact, received antitoxin therapy sooner after exposure (median time after exposure, 2 days for those who died vs. 3 for those who survived; P < .01) and sooner after symptom onset (median time after symptom onset, 1 day for those who died vs. 2 days for those who survived; P < .01).

In a CART analysis restricted to patient age, signs, and symptoms, we identified a clinical syndrome strongly predictive of survival (figure 2). All patients survived if they presented at hospital admission with no history of shortness of breath or vomiting and had normal facial muscle strength on physical examination (0 [0%] of those who died vs. 206 [32%] of those who survived [OR for death, 0; P < .01]). For identifying patients who subsequently survived, this clinical prediction rule had a sensitivity of 32% (95% CI, 29–36), a specificity of 100% (95% CI, 92–100), and a positive likelihood ratio of infinity. The clinical syndrome most predictive of death was shortness of breath, impaired gag reflex, and no history of diarrhea. Patients who died were 22.6 times more likely to have this clinical syndrome than were patients who survived (13 [24%] of those who died vs. 9 [1%] of those who survived [95% CI, 9.1–56.0]). For identifying patients who subsequently died, this clinical prediction rule had a sensitivity of 33% (95% CI, 22–48), a specificity of 99% (95%, CI 97–99), and a positive likelihood ratio of 24.1 (95% CI, 11.4–51.1).

Previous SectionNext Section

Discussion

In this large case series, we found that patients with botulism receive prompt diagnosis and treatment; that overall mortality in Georgia is comparable with that in the United States; that patients with advanced symptoms are likely to die, even if they receive prompt therapy; and that definable clinical syndromes may predict survival or death.

One clinical syndrome—no history of shortness of breath or vomiting and normal facial muscle strength—was 100 percent predictive of survival. Using this clinical syndrome to identify low-risk patients could help in a resource-poor setting, such as Georgia. For example, patients with botulism are routinely transported to Tbilisi regardless of where they live; even though most hospitals stock antitoxin, few outside of Tbilisi are staffed and equipped to provide advanced medical care. Prospective validation of this clinical prediction rule could help reduce the need for costly, time-consuming transportation for a subgroup of low-risk patients. Similarly, patients presenting with a clinical syndrome predictive of death—shortness of breath, impaired gag reflex, and no diarrhea—could be triaged immediately to higher-level facilities. Application of this prediction rule would be most useful in massive outbreaks in which intensive care and antitoxin may need to be rationed. This study also suggests that shortness of breath is the single best predictor of identifying persons at high risk of death. Public health officials in the United States are increasingly concerned about how they would respond to a massive outbreak of naturally occurring or intentionally occurring botulism [9]. If validated outside of Georgia, these prediction rules could help guide patient triage in a mass-casualty scenario. This model was not developed, however, to predict which patients need ventilatory support or antitoxin and should not be used to predict how many ventilators or doses of antitoxin might be needed.

We found that the classic diagnostic pentad, which has previously been proposed for the diagnosis of botulism, was of limited sensitivity for diagnosing botulism in the patients we studied [8]. We found that the incubation period was defined for a large proportion of patients and was consistent with previous estimates that persons exposed to contaminated food usually develop symptoms of botulism 1–2 days after exposure [13]. Further studies should attempt to define the exact sequence of how these neuroparalytic signs and symptoms develop after exposure, because techniques for using history and physical examination to determine when exposure occurred have not been developed.

The Georgian public health system is challenged by extremely limited resources [14–17]. In this context, we were surprised that the mortality rate was not >8%. One explanation is that most cases in this series were probably caused by neurotoxin type B, which is believed to cause less severe botulism than other types [6]. Second, cases were diagnosed and patients were hospitalized promptly. Third, and most important, patients received antitoxin immediately after diagnosis. The brief interval between symptom onset and antitoxin administration in Georgia may have an enormous impact on outcome, given that antitoxin improves survival when administered soon after symptom onset [18]. By comparison, diagnosis and treatment are frequently delayed by days to weeks in the United States and Canada [19–21].

Our study has several important limitations. First, we only included diagnosed cases, meaning we can only make conclusions about the sensitivity, but not the specificity, of clinical findings for the diagnosis of botulism. Second, our population of patients who received a diagnosis and were hospitalized could be biased, limiting how much we can generalize our results. Because the national surveillance data consisted only of summary reports and lacked personal identifiers, we could not determine whether the 706 patients identified through hospitals represented a subset of the total number of cases reported during surveillance or the totality of patients with botulism who were hospitalized and received a diagnosis in Georgia during this time. The diversity of clinical outcomes among hospitalized patients, however, suggests that patients from both ends of the clinical severity spectrum were evaluated. Third, most cases were not laboratory confirmed, and there is no formal case definition used in Georgia by all clinicians. We believe that misclassification was unlikely, because the clinical syndrome of botulism is highly distinctive; all patients had characteristic presentations, and most reported consuming high-risk foods. Fourth, outcomes were not assessed after discharge. We suspect that few, if any, patients died from botulism after discharge, because there is little incentive for early discharge in Georgia. Fifth, clinical prediction rules based on CART analysis require prospective validation in other populations before they can be recommended for widespread use.

We have separately described the increasing incidence of botulism in Georgia [10]. In brief, a unique constellation of factors—including cultural and culinary tradition, financial hardship, food scarcity, and the collapse of the industrial canning, power, and municipal water infrastructure—have helped increase the frequency of vegetable preservation in the home and have helped decrease the safety of the process. It is possible that similar forces may be at work in other former Soviet republics and that botulism may be an underrecognized public health problem. In Georgia, we are developing a multifaceted approach for botulism prevention. Improving clinical management of this uncommon, but important, disease will help both Georgia and the world.

Previous SectionNext Section

Acknowledgments

Financial support. This study was funded by the US Department of Health and Human Services' Biotechnology Engagement Program. The funding source played no role in the design of the study; the collection, analysis, and interpretation of the data; the writing of the report; or the decision to submit for publication.

Conflict of interest. All authors: No conflict.

  • Received December 1, 2003.
  • Accepted March 5, 2004.
Previous Section
 

References

  1. ↵
    1. Shapiro RL,
    2. Hatheway C,
    3. Becher J,
    4. Swerdlow DL
    . Botulism surveillance and emergency response: a public health strategy for a global challenge. JAMA 1997;278:433-5.
    Abstract/FREE Full Text
  2. ↵
    1. Arnon S,
    2. Schechter R,
    3. Inglesby TV,
    4. et al
    . Botulinum toxin as a biological weapon: medical and public health management. JAMA 2001;285:1059-70.
    Abstract/FREE Full Text
  3. ↵
    1. Doyle MP,
    2. Beuchat LR,
    3. Montville TJ
    1. Dobbs KL,
    2. Austin JW
    . Food microbiology: fundamentals and frontiers. Doyle MP, Beuchat LR, Montville TJ, editors. Washington, DC: ASM Press; 1997. p. 288-304.
  4. ↵
    1. Centers for Disease Control and Prevention (CDC)
    . Botulism in the United States, 1899–1996: handbook for epidemiologists, clinicians, and laboratory workers. Atlanta: CDC; 1998. Available at: http://www.cdc.gov/ncidod/dbmd/diseaseinfo/botulism.pdf. Accessed 14 July 2004.
  5. ↵
    1. Cherington M
    . Clinical spectrum of botulism. Muscle Nerve 1998;21:701-10.
    CrossRefMedlineWeb of Science
  6. ↵
    1. Hughes JM,
    2. Blumenthal JR,
    3. Merson MH,
    4. et al
    . Clinical features of types A and B food-borne botulism. Ann Intern Med 1981;95:442-5.
    Abstract/FREE Full Text
    1. Peck MW
    . Micro Today. 2002. Clostridia and food-borne disease. 9–12 February. Available at: http://www.socgenmicrobiol.org.uk/pubs/micro_today/pdf/020204.pdf. Accessed 14 July 2004.
  8. ↵
    1. Wainwright RB,
    2. Heyward WL,
    3. Middaugh JP,
    4. et al
    . Food-borne botulism in Alaska, 1947–1985: epidemiology and clinical findings. J Infect Dis 1988;157:1158-62.
    Abstract/FREE Full Text
  9. ↵
    1. Centers for Disease Control Prevention
    . Biological and chemical terrorism: strategic plan for preparedness and response. Recommendations of the CDC Strategic Planning Workgroup. MMWR Mor Mortal Wkly Rep 2000;49(RR-4):1-14.
    Medline
  10. ↵
    1. Varma JK,
    2. Kasitadze G,
    3. Moiscrafishvili M,
    4. et al
    . Emerg Infect Dis. 2004. Fooborne botulism in the Republic of Georgia. (in press).
  11. ↵
    1. Breiman L,
    2. Friedman JH,
    3. Olshen RA,
    4. Stone CJ
    . Classification and regression trees. Belmont, CA: Wadsworth International; 1984.
  12. ↵
    1. Stone M
    . Cross-validatory choice and the assessment of statistical predictions. J Roy Statist Soc Ser 1974;36:111-47.
  13. ↵
    1. Woodruff B,
    2. Griffin PM,
    3. McCroskey LM,
    4. et al
    . Clinical and laboratory comparison of botulism from toxin types A, B, and E in the United States, 1975–1988. J Infect Dis 1992;166:1281-6.
    Abstract/FREE Full Text
  14. ↵
    1. International Monetary Fund
    . Country notes: debt sustainability and policy priorities to ensure growth in the CIS7 countries. Available at http://www.imf.org/external/np/eu2/2002/edebt/eng/040202a.pdf. Accessed 14 July 2004.
    1. Gzirishvili D,
    2. Mataradze G
    . Discussion paper #5: healthcare reform in Georgia. Tbilisi, Georgia: United Nations Development Program; 1998.
    1. Khetsuriani N,
    2. Imnadze P,
    3. Dekanosidze N
    . Diptheria epidemic in the Republic of Georgia, 1993–1997. J Infect Dis 2000;181(Suppl 1):80-5.
    CrossRef
  17. ↵
    1. Barwick RS,
    2. Uzicanin A,
    3. Lareau S,
    4. et al
    . Outbreak of amebiasis in Tbilisi, Republic of Georgia, 1998. Am J Trop Med Hyg 2002;67:623-31.
    Abstract
  18. ↵
    1. Tacket CO,
    2. Shandera WX,
    3. Mann JM,
    4. Hargrett NT,
    5. Blake PA
    . Equine antitoxin use and other factors that predict outcome in type A foodborne botulism. Am J Med 1984;76:794-8.
    CrossRefMedlineWeb of Science
  19. ↵
    1. Hayes MT,
    2. Soto O,
    3. Ruoff KL
    . Case 22-1997: a 58-year old woman with multiple cranial neuropathies. N Engl J Med 1997;337:184-90.
    CrossRefMedlineWeb of Science
    1. St Louis ME,
    2. Peck SH,
    3. Bowering D,
    4. et al
    . Botulism from chopped garlic: delayed recognition of a major outbreak. Ann Intern Med 1988;108:363-8.
    MedlineWeb of Science
  21. ↵
    1. Shapiro RJ,
    2. Hatheway C,
    3. Swerdlow DL
    . Botulism in the United States: a clinical and epidemiologic review. Ann Intern Med 1998;129:221-8.
    Abstract/FREE Full Text
| Table of Contents

This Article

  1. Clin Infect Dis. (2004) 39 (3): 357-362. doi: 10.1086/422318
  1. AbstractFree
  2. » Full Text (HTML)Free
  3. Full Text (PDF)Free

Classifications

Share

  1. Email this article

Navigate This Article

Current Issue

  1. March 1, 2012 54 (5)
  1. Clinical Infectious Diseases
  1. Alert me to new issues

Most