TESS-based dose–response using pediatric clonidine exposures

Toxicology and Applied Pharmacology 213 (2006) 145 – 151 www.elsevier.com/locate/ytaap

TESS-based dose–response using pediatric clonidine exposures Blaine E. Benson a,⁎, Daniel A. Spyker b , William G. Troutman c , William A. Watson d a

New Mexico Poison and Drug Information Center and University of New Mexico College of Pharmacy, Albuquerque, NM 87131, USA b Alexza Pharmaceuticals, Palo Alto, CA 94303, USA c University of New Mexico College of Pharmacy, Albuquerque, NM 87131, USA d American Association of Poison Control Centers, Washington, DC 20016, USA Received 10 August 2005; revised 3 October 2005; accepted 17 October 2005 Available online 15 December 2005

Abstract Objective: The toxic and lethal doses of clonidine in children are unclear. This study was designed to determine whether data from the American Association of Poison Control Centers Toxic Exposure Surveillance System (TESS) could be utilized to determine a dose– response relationship for pediatric clonidine exposure. Methods: 3458 single-substance clonidine exposures in children b6 years of age reported to TESS from January 2000 through December 2003 were examined. Dose ingested, age, and medical outcome were available for 1550 cases. Respiratory arrest cases (n = 8) were classified as the most severe of the medical outcome categories (Arrest, Major, Moderate, Mild, and No effect). Exposures reported as a “taste or lick” (n = 51) were included as a dose of 1/10 of the dosage form involved. Dose ranged from 0.4 to 1980 (median 13) μg/kg. Weight was imputed based on a quadratic estimate of weight for age. Dose certainty was coded as exact (26% of cases) or not exact (74%). Medical outcome (response) was examined via logistic regression using SAS JMP (release 5.1). Results: The logistic model describing medical outcome (P b 0.0001) included Log dose/kg (P = 0.0000) and Certainty (P = 0.045). Conclusion: TESS data can provide the basis for a statistically sound description of dose–response for pediatric clonidine poisoning exposures. © 2005 Elsevier Inc. All rights reserved. Keywords: Dose–response; TESS; Clonidine; Poisoning

Introduction There are more than 2 million poisonings reported by poison centers in the United States each year (Watson et al., 2004). Poison centers assist the public and health professionals by providing immediate professional guidance during poisoning emergencies. Poison center information specialists determine the potential harm associated with each exposure and develop an appropriate patient management plan. Exposures involving minimal risk are usually managed at home, while those associated with substantial risk are referred to healthcare facilities for further evaluation and treatment. In the absence of published treatment thresholds, poison center information ⁎ Corresponding author. Fax: +1 505 272 5892. E-mail address: [email protected] (B.E. Benson). URLs: http://hsc.unm.edu/pharmacy/poison/index.shtml (B.E. Benson), http://www.aapcc.org/ (W.A. Watson). 0041-008X/$ - see front matter © 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.taap.2005.10.009

specialists use their intuition or poison centers develop their own guidelines to assess the risk for common exposures. Previous work has shown that treatment thresholds vary among poison centers and that this variability influences healthcare facility referral patterns (Benson et al., 2001). Much of the current practice in clinical toxicology, including risk assessment, has been based on limited clinical evidence such as case reports and small case series (Buckley and Smith, 1996; Tenenbein, 1998). There are significant limitations to this approach. Small case series may fail to show a dose–response response relationship when one really exists (type II error). There can be a substantial delay between the time a product is first marketed and the development of a rational data-based treatment strategy. Finally, it can be difficult to update clinical datasets in order to refine risk assessment because investigators usually conduct their research independently. Ideally, the risk assessment of every poisoned patient would be based on the best available dose–response relationships and

146

B.E. Benson et al. / Toxicology and Applied Pharmacology 213 (2006) 145–151

derived from relevant population data. Poisoning dose–response relationships would be available soon after product introduction, and the dose–response curves would be refined in real time as new data became available. The Toxic Exposure Surveillance System (TESS) is a population-based data collection system owned by the American Association of Poison Control Centers (AAPCC). Data are collected in a standardized format by all poison centers in the United States and are submitted to TESS in real time (Litovitz, 1998). Since the year 2000, all poisoning cases submitted to TESS have included both exposure amount (dose) and clinical effect (response) information. This information, along with estimates of the certainty of the dose exposure, medical outcomes, and patient descriptors, provides the components that could be utilized to define a poisoning dose–response relationship. The purpose of this study was to determine whether TESS data could be used to describe dose–response relationships in pediatric clonidine ingestions. Clonidine is an α2-adrenergic agonist used to treat a variety of medical conditions including essential hypertension, migraine headache, pain, attentiondeficit disorder, Tourette’s syndrome, substance withdrawal, post-anesthetic agitation, spasticity, akathisia, and menopausal flushing (Patterson and Drugdex Editorial Staff, 2005). In 2003, there were 1736 clonidine exposures involving children less than 6 years of age reported to TESS (Watson et al., 2004). Clonidine exposure in children has been associated with apnea, bradycardia, coma, and hypotension (Nichols et al., 1997; Stein and Volans, 1978; Heidemann and Sarnaik, 1990; Wiley et al., 1990; Artman and Boerth, 1983; Olsson and Pruitt, 1983; Conner and Watanabe, 1979; Seger, 2002; Fiser et al., 1990). The dose that produces life-threatening clinical effects is unclear (Conner and Watanabe, 1979; Seger, 2002). Fiser and colleagues suggested that doses of more than 0.01 mg/kg produce apnea and bradycardia, based on observations in nine patients (Fiser et al., 1990). Methods All unintentional, single-agent clonidine exposures in children less than 6 years of age with known medical outcome reported to TESS from January 2000 through December 2003 were eligible for inclusion in this study (Cases Screened). The TESS data file included the following fields: exposure site, route (s) of exposure, quantity, quantity units, product formulation, dose certainty, age, age units, weight, weight units, clinical effects, and medical outcome. Cases without interpretable exposure amount data were excluded (Cases Excluded). The dose for each remaining case (Cases Analyzed) was calculated by multiplying the reported amount missing (e.g., 2 tablets) by the strength of the product (e.g., 100 μg). If the amount was reported as a “taste” or “lick,” the dose was assumed to be 1/10 the clonidine content of the dosage form. Dose per kilogram of body weight and Log10 dose per kilogram of body weight were calculated for each case. Due to the large number of missing data, weights for cases were imputed from a quadratic fit for age based on TESS pediatric patients (n = 338). This analysis is available upon request (ask for Supplementary Table 1). The equation used was: Weight ðkgÞ ¼ 7:2701 þ 0:25610 Age ðmonthsÞ  0:0015657ðAge ðmonthsÞ  23:677Þ2

ð1Þ

The clinical response for each case was obtained from the TESS medical outcome field (none, minor, moderate, major, death). A summary of TESS

medical outcome definitions is given in Table 1. A more detailed explanation of TESS data coding can be found in the document “Instructions for The American Association of Poison Control Centers Toxic Exposure Surveillance System (TESS)” available upon request from the AAPCC. Respiratory arrest was added as a separate medical outcome category ranked between major and death in severity (none, minor, moderate, major, arrest, death). These six categories were mutually exclusive and comprehensive in scope. Cases Analyzed were sorted based on dose into five groups of equal size (quintiles). The medical outcome frequencies were examined by quintile. Logistic regression was used to estimate the probability of each medical outcome category (response) across dose range and to identify additional variables (covariates) that contributed to the dose–response model. Since the response variable was ordinal (none, minor, moderate, major, arrest), the cumulative response probabilities to the logistic distribution function of a linear model were fit using maximum likelihood. Likelihood-ratio test statistics were calculated for the whole model, lack of fit, main effects, confidence limits, and odds ratios for the maximum likelihood parameter estimates. All calculations were performed using JMP v 5.1.1 (SAS, Carey, NC) running under Windows 2000. JMP was also used to create a cumulative logistic probability plot for ordinal outcome and to calculate doses and confidence intervals associated with selected response probabilities.

Table 1 AAPCC TESS medical outcome definitions No effect: The patient developed no symptoms as a result of the exposure. Follow-up is required to make this determination unless the initial poison center call occurs sufficiently long after the exposure that you are reasonably certain no effects will occur. Minor effect: The patient exhibited some symptoms as a result of the exposure, but they were minimally bothersome to the patient. The symptoms usually resolve rapidly and usually involve skin or mucous membrane manifestations. The patient has returned to a pre-exposure state of well-being and has no residual disability or disfigurement. Follow-up is required to make this determination unless the initial poison center call occurs sufficiently long after the exposure that you are reasonably certain that the clinical effects(s) will not worsen. Symptomatic patients must be followed until symptoms have resolved or nearly resolved, unless the residual symptoms are anticipated to be long-term and of minimal clinical significance. Moderate effect: The patient exhibited symptoms as a result of the exposure which are more pronounced, more prolonged, or more of a systemic nature than minor symptoms. Usually, some form of treatment is or would have been indicated. Symptoms were not life-threatening, and the patient has returned to a pre-exposure state of well-being with no residual disability or disfigurement. Follow-up is required to make this determination unless the initial poison center call occurs sufficiently long after the exposure that you are certain the clinical effect(s) will not get worse. Symptomatic patients must be followed until symptoms have resolved or nearly resolved, unless the residual symptoms are anticipated to be long-term and of minimal clinical significance. Major effect: The patient has exhibited symptoms as a result of the exposure which were life-threatening or resulted in significant residual disability or disfigurement. Follow-up is required to make this determination unless the initial poison center call occurs sufficiently long after the exposure that you are certain the clinical effect(s) will not get worse. Symptomatic patients must be followed until symptoms have resolved or nearly resolved, unless the symptoms are anticipated to be long-term or permanent. Death: The patient died as a result of the exposure or as a direct complication of the exposure where the complication was unlikely to have occurred had the toxic exposure not preceded the complication. Only include those deaths which are probably or undoubtedly related to the exposure. Code outcome for unrelated or probably unrelated deaths as unrelated. Abstracted from “Instructions for the American Association of Poison Control Centers Toxic Exposure Surveillance System (TESS)”. Used with permission from the American Association of Poison Control Centers.

B.E. Benson et al. / Toxicology and Applied Pharmacology 213 (2006) 145–151 The study protocol was reviewed by the University of New Mexico Human Research Review Committee and granted exemption from Health and Human Services regulations for protection of human subjects as defined in 45CFR46.101(b).

Results Cases Screened included 3458 single-agent clonidine exposures in children less than 6 years of age with known medical outcomes reported to TESS from January 2000 through December 2003. Cases Excluded numbered 1908. Most were excluded because the product strength was unknown or because the quantity ingested could not be established during case review (Fig. 1). For 324 cases, the reason for exposure did not match entry criteria. In 60 cases, the patient ingested a combination product containing multiple active ingredients. The qualifying exposures (Cases Analyzed) numbered 1550. Table 2 summarizes the characteristics of the Cases Analyzed. Most patients were between 1 and 3 years of age and had ingested 100 to 300 μg of clonidine. The exact dose was known in 26% of the cases. Most patients (86%) were treated in healthcare facilities. Forty-four percent were treated with gastrointestinal decontamination (ipecac syrup, gastric lavage, and/or activated charcoal), and 60% were symptomatic. The most common clinical effects were drowsiness, bradycardia, hypotension, respiratory depression, and miosis (Table 3). None of the patients died, but eight developed respiratory arrest. Fig. 2 shows the distribution of the medical outcomes among the Cases Analyzed. Fig. 3 shows the medical outcome as a function of mg/kg dose on a log scale (dose–response relationship). TESS medical outcomes were related to dose (P b 0.0001). The weightadjusted dose–response Log10 dose/kg (P b 0.0001) was stronger than the unadjusted dose–response. Details of this

Fig. 1. Relationship of analyzed cases to screened cases for pediatric clonidine dose–response study. All Cases Analyzed (n = 1550).

147

Table 2 Demographics of all Cases Analyzed (n = 1550) Age in months (mean ± SD) Median dose ingested (25th and 75th percentiles) Median μg/kg dose ingested (range) Dose certainty (% of total cases)

Treated with gastrointestinal decontamination

26 ± 12 200 μg (100–300 μg) 13 μg/kg (0.4–1980) 26% Exact 74% Not exact 43% Estimate 21% Max dose possible 10% Unknown 44%

analysis are available upon request (ask for Supplementary Table 2). Table 4 shows the relationship of TESS medical outcomes to quintile dosage ranges. The percentage of patients with more severe outcomes increased as quintile dose range increased. Moderate effects were reported at doses as low as 0.65 μg/kg. The logistic model provides a dose–response probability of each of the medical outcomes over the range of observed doses. Table 5 shows the estimates and associated 95% confidence intervals (CIs) for selected response probabilities. For example, the ED50% value [and CI] for minor or worse was 4.28 [2.63, 6.01] μg/kg. These “inverse solutions” were calculated based on present-or-absent (binary) logistic regression. Table 3 Clinical effects associated with clonidine exposure, sorted by frequency, all Cases Analyzed (n = 1550) Clinical effect

% of Patients

Drowsy/lethargic Bradycardia Hypotension Respiratory depression Miosis Agitated/irritable Coma Tachycardia Ataxia Hypertension Vomiting Respiratory arrest Hypothermia Dermal pallor Dysrhythmia (other) Confusion Muscle weakness Dizziness/vertigo Cyanosis Oropharyngeal irritation Mydriasis Hyperglycemia Diaphoresis Bronchospasm Peripheral neuropathy Slurred speech Hyperventilation/tachypnea Ileus/no bowel sounds Nausea Cough/choke Chest pain Cardiac arrest

51.10 8.97 7.68 3.74 2.32 1.87 1.68 1.55 1.48 1.42 0.77 0.52 0.45 0.45 0.32 0.26 0.19 0.19 0.19 0.13 0.13 0.13 0.13 0.06 0.06 0.06 0.06 0.06 0.06 0.06 0.06 0.06

148

B.E. Benson et al. / Toxicology and Applied Pharmacology 213 (2006) 145–151

Fig. 2. Distribution of medical outcomes in pediatric clonidine exposures. All Cases Analyzed (n = 1550).

The model did not improve when the gastrointestinal decontamination status of the patients was added as a covariate. Certainty of dose assessment, however, did provide a statistically significant (P = 0.0069) improvement in the ordinal logistic fit for medical outcome as summarized in Table 6. Discussion This is the first attempt to use TESS data to determine a dose–response relationship. We chose to examine an agent that

was likely to show a dose–response relationship (a large number of cases with a broad range of medical outcomes) and where dose–response analysis would likely be useful for patient triage in poison centers. We chose to examine clonidine ingestions in young children because triage decisions for asymptomatic children are based on dose rather than intent. Adult clonidine ingestions are often intentional and are referred by poison centers to healthcare facilities regardless of dose ingested. Published case series of pediatric clonidine overdose have shown a spectrum of clinical effects including apnea, bradycardia, hypotension, and death (Nichols et al., 1997; Stein and Volans, 1978; Heidemann and Sarnaik, 1990; Wiley et al., 1990; Artman and Boerth, 1983; Olsson and Pruitt, 1983; Conner and Watanabe, 1979; Seger, 2002; Fiser et al., 1990). We used the medical outcome categories used by U.S. poison centers as “effect” since these ordered categories are designed to capture the overall severity of a patient’s clinical course and are mandatory data elements of each case submitted to TESS. A disadvantage of using TESS outcomes is that a patient’s clinical signs and symptoms are melded into a single outcome measure, which necessarily loses specific clinical effect discrimination. We added respiratory arrest as a separate outcome category to provide a simple “splitting” of the “major” medical outcome category into non-life-threatening and life-threatening clinical subcategories. Ordinal logistic regression estimates the probability of choosing one of the response levels, with the conditions: each response is a smooth (sigmoidal) function of the dose, each response probability must be between 0 and 1, and the response probabilities must sum to 1 across the response levels for a given value of dose. In a logistic probability plot, the y axis represents

Fig. 3. Nominal logistic fit of medical outcome by dose (mg/kg) on a log scale. All Cases Analyzed (n = 1550). Arrows labeled E and N–E indicate the contribution of Certainty = Exact and Certainty = Not Exact relative to a 0.1 mg/kg dose.

B.E. Benson et al. / Toxicology and Applied Pharmacology 213 (2006) 145–151

149

Table 4 Frequency of doses versus medical outcome by quintiles, all Cases Analyzed (n = 1550) Clinical outcome

Dose quintile [μg/kg]

None Minor Moderate Major Arrest Mod–Arrest

Total

[b7.8]

[7.8–9.9]

[9.9–16]

[16–31]

[N31]

[Any]

168 (11%) 109 (7%) 28 (2%) 5 (0.3%) 0( 33 (11%)

152 (10%) 108 (7%) 47 (3%) 3 (0.2%) 0( 50 (16%)

122 (8%) 122 (8%) 58 (4%) 8 (0.5%) 0( 66 (21%)

104 (7%) 120 (8%) 75 (5%) 10 (0.7%) 1 (0.1%) 86 (28%)

70 (5%) 106 (7%) 104 (7%) 23 (1.5%) 7 (0.4%) 134 (43%)

616 (40%) 565 (36%) 312 (20%) 49 (3%) 8 (0.5%)

probability of that event. For k response levels, k − 1 smooth curves partition the total probability (1) among the response levels. For the ordinal logistic regression shown in Fig. 3, the equation describing the four curves (logit models) separating 5 response levels is Pk ðLog10 DoseÞ ¼

1 1þ

eðInterceptk  Slope  Log 10 DoseÞ

ð2Þ

where k are the response level intercepts (2.3145 for minor, 0.6473 for moderate, −1.5260 for major, and −3.5410 for respiratory arrest) and the slope is 1.0033. The maximum likelihood fitting principle employed for this logistic regression minimizes the sum of the negative logarithms of the probabilities fitted to the observed response events. Details of this analysis are available upon request (ask for Supplementary Table 2). The dose–response graphic ordinal logistic regression (Fig. 3) generates “pharmacologically familiar” sigmoidal log dose– response probability curves. The graph displays individual cases for an observed dose (x axis) plotted against their probabilistic (made-up) outcomes (y axis). The graphics algorithm is simply distributing (jittering) the y location within the appropriate outcome category on the graphic. Despite this “made-up” y value, this display does provide a useful display of the highest and lowest probability values for a given dose and an immediate feel for the amount of data contributing to the probability estimates. For a dose of 0.1 mg/kg (100 μg/kg) as an example, the probability of respiratory arrest is 1.1%. For the same dose, the probabilities of major, moderate, minor, and no effect outcome responses are 6.3%, 34%, 38%, and 21% respectively. These probabilities sum to 100%. This logistic regression approach can also support the calculation of the dose and confidence interval for a particular outcome category and a given response probability (the inverse

Table 5 Dose by outcome, effect doses, and confidence intervals quartiles and medians, all Cases Analyzed (n = 1550) Clinical outcome

n

Dose quartile 25%

None 616 6.15 Minor or greater 934 7.45 Moderate or greater 369 9.87 Major or greater 57 14.9 Arrest 8 32.1

Median 75% 8.65 14.9 20.1 30.5 83.1

17.2 30.8 48.9 107 141

Effect dose

solution). The Effect Doses (EDx%) shown in Table 5 were completed using present-or-absent (binary) logistic regression. For example, the moderate or greater ED25% value of 18.9 [14.6, 24.7] was based on the binary regression of the moderate, major, or arrest outcomes (n = 369) versus none or minor outcomes (n = 1181). SAS JMP calculates the estimates from

Dose [95% CI] (μg/kg)

ED50% 4.28 [2.63, 6.01] ED25% 18.9 [14.6, 24.7] ED5% 37.0 [23.2, 60.2] ED2% 144 [63.8, 595]

Table 6 Ordinal logistic fit for medical Outcome + Certainty, all Cases Analyzed (n = 1550) Whole model test Model

−Log likelihood

Difference 23.7292 Full 1826.4082 Reduced 1850.1374 RSquare (U) Observations (or sum weights)

DF

Chi-square

Prob N ChiSq

2

47.45843

b0.0001

0.0128 1550

Lack of fit Source

DF

−Log likelihood

Chi-square

Lack of fit Saturated Fitted

2014 2016 2

753.4559 1072.9523 1826.4082

1506.912 Prob N ChiSq 1.0000

Parameter estimates Term

Estimate

Std Error

Chi-square

Prob N ChiSq

Intercept [Arrest] Intercept [Major] Intercept [Minor] Intercept [Moderate] Log10 dose/kg Certainty [Exact]

−4.3556089

0.3906352

124.32

b0.0001

−2.3488703

0.213997

120.48

b0.0001

0.53885127

0.1821653

8.75

0.0031

1.37213455

0.1849648

55.03

b0.0001

0.55458214

0.0924827

35.96

b0.0001

−0.1465093

0.0542089

7.30

0.0069

Source

Nparm

DF

Wald Chi-square

Prob N ChiSq

Log10 dose/kg Certainty

1

1

35.959

0.0000

1

1

7.304

0.0069

Effect Wald tests

150

B.E. Benson et al. / Toxicology and Applied Pharmacology 213 (2006) 145–151

the solution equation and the confidence limits using Fieller’s theorem (SAS, Carey, NC). Inverse solutions for the effect doses in Table 5 and the larger family of inverse solutions are available on request (ask for Supplementary Table 3). A limitation of Fig. 3 is that it cannot show the additional dimension of the covariates – in this case, certainty of exposure – for the displayed model. Since certainty is a binary endpoint, one could create a separate graph for each certainty subset, but such an approach would rapidly become unwieldy as the number of covariates increases. A calculation of the probability of each medical outcome can easily be carried out for an individual patient using our dose–response model (including covariates). Logistic regression identified a clear dose–response relationship for clonidine exposure in young children using TESS data. Decontamination status did not improve the model. This is not a surprise because decontamination is a time-dependent therapy that would only confound medical outcome if used early in the course of a poisoning. It was impossible to determine time of decontamination from our dataset so it is unclear whether confounding occurred. Furthermore, medical outcome classification is a “final” assessment intended to integrate all clinical effects that cannot be detected. Subtle variations within a medical outcome category will be underappreciated. Finally, medical outcome could have been impacted by other treatments such as intubation, fluids, vasopressors, or naloxone. For these reasons, our dose– response results may not reflect the true impact of gastrointestinal decontamination on the medical outcomes of pediatric clonidine poisonings. The model was improved when the certainty of ingestion was considered. The parameter estimate ± standard error of the mean (SEM) for Log10 dose/kg was 0.555 ± 0.0925 and for Certainty [Exact] was −0.147 ± 0.0542 (Table 6). Thus, the probability of each severity was shifted to the “left” (lower dose) by 71% (10−0.147) where the Certainty = Exact and to the “right” by 140% when Certainty = Not Exact. If we assume Exact to better reflect true exposure, then this finding is consistent with an overestimate of dose by 197% (10−0.294) in the Not Exact cases. The impact (or lack of impact) of this “conservative” 2-fold overestimate of exposure is indicated by the arrows on either side of the 0.1 mg/kg dose in Fig. 3. The statistical significance of this finding reflects, in part, the “ease” of detecting small effects with a large dataset. There are limitations associated with using TESS data to model dose–response. First, reported doses are frequently based on caregiver histories. The dose was not confirmed by thirdparty tablet counts or by measurement of serum concentrations of clonidine. Second, the clinical effects reported to TESS were also unconfirmed. In a comparison between poison center charts and hospital charts, Hoyte and colleagues showed that poison center records might under-represent cardiovascular, gastrointestinal, and electrolyte abnormalities (Hoyt et al., 1999). It is possible that the TESS data do not adequately reflect changes occurring in the hospital or home setting. There may be transcription mistakes, factual inaccuracies, and errors of omission or commission associated with the transfer of information into TESS. It is likely that subtle, non-life-

threatening clinical effects such as cardiac conduction blocks, early neurological changes, or unexpected laboratory abnormalities do not get incorporated into poison center records. Such inaccuracies could limit the ability of these data to identify subtle dose–response relationships. Third, our analysis was limited to 44% of the eligible clonidine exposure cases, so it possible that the subset does not adequately reflect the full range of exposures. Finally, the model utilized data from young children so the dose–response relationship may not apply to other age groups. These preliminary results are promising. With additional refinement, it may be possible to use TESS to describe dose– response relationships for other agents. Since TESS is a realtime data collection system, analyses could be applied to newly approved medications and could be updated in real time, leading to enhanced dose–response definition over time. In addition, TESS dose–response analyses could fill an important void. Much of the published literature in clinical toxicology consists of case reports or small case series describing massive overdoses. TESS data probably better represent the more common low or mid-dose exposures absent from case reports and case series. The broader range of exposure should lead to better characterization of dose–response. These analyses could be important for clinicians, poison centers, and for the development of national triage guidelines for poison centers.

Appendix A. Supplementary data Supplementary data associated with this article can be found in the online version at doi:10.1016/j.taap.2005.10.009.

References Artman, M., Boerth, R.C., 1983. Clonidine poisoning. Am. J. Dis. Child. 137, 171–174. Benson, B.E., Smith, C., McKinney, P., Litovitz, T., Tandberg, D., 2001. Do poison center protocols affect health care facility referrals? J. Toxicol. Clin. Toxicol. 39, 433–438. Buckley, N.A., Smith, A.J., 1996. Evidence-based medicine in toxicology: where is the evidence? Lancet 347, 1167–1169. Conner, C.S., Watanabe, A.S., 1979. Clonidine overdose: a review. Am. J. Hosp. Pharm. 36, 906–911. Fiser, D.H., Moss, M.M., Walker, W., 1990. Critical care for clonidine poisoning in toddlers. Crit. Care Med. 18, 1124–1128. Heidemann, S.M., Sarnaik, A.P., 1990. Clonidine poisoning in children. Crit. Care Med. 18, 618–620. Hoyt, B.T., Rasmussen, R., Giffin, S., Smilkstein, M.J., 1999. Poison center data accuracy: a comparison of rural hospital chart data with the TESS database. Toxic Exposure Surveillance System. Acad. Emerg. Med. 6, 851–855. Litovitz, T., 1998. The TESS database: use in product safety assessment. Drug Safety 18 (1), 9–19. Nichols, M.H., King, W.D., James, L.P., 1997. Clonidine poisoning in Jefferson County, Alabama. Ann. Emerg. Med. 29, 511–517. Olsson, J.M., Pruitt, A.W., 1983. Management of clonidine ingestion in children. J. Pediatr. 103, 646–650. Patterson, L.E., Drugdex Editorial Staff, 2005. Clonidine (Drugdex Drug Evaluation). In: Klasco, R.K. (Ed.), Drugdex System. Thomson Micromedex. Greenwood Village, Colorado. Seger, D.L., 2002. Clonidine revisited. J. Toxicol. Clin. Toxicol. 40, 145–155.

B.E. Benson et al. / Toxicology and Applied Pharmacology 213 (2006) 145–151 Stein, B., Volans, G., 1978. Dixarit overdose: the problem of attractive tablets. BMJ 2, 667–668. Tenenbein, M., 1998. Good reasons to publish in Clinical Toxicology. J. Toxicol. Clin. Toxicol. 36, 137–138. Watson, W.A., Litovitz, T.L., Klein-Schwartz, W., Rodgers, G.C., Youniss, J.,

151

Reid, N., Rouse, W.G., Rembert, R.S., Borys, D., 2004. Annual report of the American association of poison control centers toxic exposure surveillance system. Am. J. Emerg. Med. 22, 335–404. Wiley, J.F., Wiley, C.C., Torrey, S.B., Henretig, F.M., 1990. Clonidine poisoning in young children. J. Pediatr. 116, 654–658.