Toxicity of Bupropion Overdose Compared With Selective Serotonin Reuptake Inhibitors

This was a retrospective observational study using data from NPDS. These data are collected by member poison centers and maintained in a near real-time, de-identified database by the American Association of Poison Control Centers. Demographics, clinical effects, therapies provided, and medical outcome are coded from phone calls in a standardized manner by trained Certified Specialists in Poison Information at individual poison centers.

All human exposure cases reported to US poison centers involving an adolescent (10–19 years of age) with single-substance exposure to an SSRI or bupropion and coded as suspected suicide were included. Cases with any co-ingestion were excluded. The SSRIs studied were citalopram, escitalopram, fluoxetine, fluvoxamine, paroxetine, sertraline, and, as categorized by NPDS coding, “other types of SSRI[s],” which include vilazodone and vortioxetine. Bupropion was included in “other types of TCA” in NPDS coding until it was given its own substance code on January 30, 2013. Analysis of cases reported in the months after this date revealed that poison centers were consistently coding bupropion exposures using its specific substance code by the end of May 2013. Therefore, we elected to consider data from June 1, 2013, through December 31, 2017. Each case was considered an independent occurrence because it was not possible to link the de-identified data to potential repeat patients. Analyses of demographic characteristics, clinical effects, highest level of care required, therapies provided, and severity of medical outcome were performed. Specific therapies of interest were intubation, mechanical ventilation, vasopressors, benzodiazepines, sedation, oxygen, anticonvulsants, activated charcoal, cardiopulmonary resuscitation (CPR), and extracorporeal membrane oxygenation (ECMO). This study was deemed by the Indiana University Institutional Review Board to be quality improvement research using de-identified data and thus not required to undergo review from the Indiana University Institutional Review Board.

According to American Association of Poison Control Centers guidelines, clinical effect is stratified as follows:

• Minor effect: The patient develops minimally bothersome symptoms that resolve rapidly without lasting effects. This category includes gastrointestinal symptoms, drowsiness, oral irritation, and sinus tachycardia without hypotension.

• Moderate effect: Effects of an exposure are more pronounced, prolonged, or systemic and often require treatment but are not life-threatening and resolve without lasting effects. This includes acid-base disturbance, disorientation, hypotension, isolated seizures, and conduction disturbances without hypotension.

• Major effect: The patient experiences life-threatening symptoms or has significant permanent dysfunction as a result of the exposure. This includes status epilepticus, symptomatic ventricular tachycardia, coma with hypotension, cardiac or respiratory arrest, ventricular fibrillation, renal failure, and rhabdomyolysis.

• Death: This includes patients who die as a direct result of the exposure, from a complication of the exposure, or for whom the exposure was a contributing factor in the death.¹ 

Differences between the 2 groups in clinical effects in cases with any outcome, clinical effects in cases with moderate or major outcome or death (MMD), selected therapies as listed above, the highest level of care required, chronicity of exposure, and sex were assessed by using χ² or Fisher’s exact tests. The percent of patients per month who were intubated, received vasopressors, experienced a single seizure, experienced multiple seizures, or who had an MMD were analyzed by logistic regression. All-severity outcome frequencies were assessed with Poisson regression. Logistic regression was used to assess trends in exposures over time. An interaction term was used to assess the differences in reported clinical effects, outcomes, and therapies between SSRIs and bupropion ingestions over time. A level of significance of .05 was used for all statistical tests. All statistical analyses were performed in R and were completed using RStudio version 1.1.383 (RStudio, Inc, Boston, MA).

There were 30 026 poison center case records that met inclusion criteria. Of these, 3504 were exposures to bupropion and 26 522 were exposures to an SSRI. The average age was 15.8 years, and girls constituted the majority of both groups (Table 1). The most common SSRIs implicated were sertraline and fluoxetine, which accounted for 59% of the cases (Table 2). Proportionally, there were more boys in the bupropion case group than in the SSRI group (25.1% vs 16.2%; Fig 1). Mortality was rare; all 8 fatalities (0.22%) occurred in the bupropion cases. Poison center cases involving an SSRI were much more likely to result in either minor or no effects from the exposure than those involving bupropion (68% vs 33.2%). By contrast, cases involving bupropion were more likely to have an MMD (58.1% vs 19%) compared with those involving an SSRI (Table 3).

Considering the 10 most commonly reported effects among cases with an MMD, bupropion exposure was significantly more likely to cause tachycardia (83.7% vs 59.9%), vomiting (24.8% vs 20.6%), cardiac conduction disturbances (20.0% vs 17.1%), agitation (20.2% vs 11.7%), seizures (27.0% vs 8.5%; Figs 2 and 3), and hallucinations (28.6% vs 4.3%). All were significant at the level of P < .001 except for conduction disturbance (P = .005). SSRI cases were significantly associated with the development of hypertension (25.3% vs 17.6%; P < .001; Table 4). Proportions of drowsiness, nausea, tremor, and other effects were not significantly different between case groups.

Considering all-severity outcomes, poison center cases involving bupropion were associated with significantly higher rates of each of the 10 most common reported effects (P < .001, except P = .002 for abdominal pain; Table 5). Bupropion cases were also associated with significantly higher rates of medical therapy, including intubation (4.9% vs 0.3%; Fig 4), vasopressor use (1.1% vs 0.2%; Fig 5), administration of a benzodiazepine (34.2% vs 5.4%), supplemental oxygen requirement (8.2% vs 0.8%), and CPR (0.5% vs 0.01%), with P < .001 for all. Three patients required treatment with ECMO; they were all in the bupropion group (Table 6). Poisson regression indicated that the number of reported poison center cases increased independently and significantly throughout the study period by an average of 807 cases per year for SSRIs and an average of 134 cases per year for bupropion. There was not a significant difference between cases for either substances over time (Fig 6). The percentage of poison center cases involving bupropion that resulted in MMDs also significantly increased in each year (Fig 7).

Bupropion’s pharmacologic actions and potential for serious toxicity in overdose are well known within toxicology circles but may be under-recognized in the wider pediatric and primary care communities. This study revealed that poison center cases involving adolescents who attempt self-harm with bupropion had significantly higher rates of serious outcome or death than those involving SSRIs. Patients in the bupropion cases, compared with patients in the SSRI cases, were 2.5 times more likely to have a moderate or major outcome or die and required hospitalization at nearly triple the rate, including a fourfold higher risk of ICU admission. These risks are particularly concerning given the rising rates of total attempts and poor outcomes after ingestion. Overdoses have been shown to vary directly with sales of each agent;²⁴ therefore, decreases in prescription of bupropion could be expected to decrease the number of serious outcomes in antidepressant overdoses.²⁵ Downstream effects would include decreased health care use, fewer ICU admissions, and decreased total costs to the health care system.²⁶ 

Relative to poison center cases with reported SSRI exposure, patients with reported bupropion exposure had higher rates of acute neuropsychiatric sequelae, including a nearly sevenfold risk for developing hallucinations and an approximately twofold risk for becoming agitated. Bupropion is well known for its ability to lower the seizure threshold, sometimes even precipitating new-onset seizures at therapeutic doses.²¹ In our study, seizures occurred >3 times as often with poison center cases involving bupropion than with SSRIs. This analysis of NPDS data reveals that neuropsychiatric effects are not the only toxicity of concern with bupropion; in fact, this agent poses a number of additional substantial risks, including tachycardia and malignant dysrhythmias. These adverse effects likely account for the greater requirement for use of aggressive management such as intubation, vasopressors, CPR, and ECMO.

Sheridan et al⁸ conducted a study in which they also used NPDS data to compare adolescent poison center cases involving exposure bupropion to those involving TCAs although during a slightly different time period and allowing co-ingestion of ethanol. Their study had many similar findings to ours, including higher rates of hallucinations, seizures, tachycardia, and vomiting in the bupropion cases. Another similarity was the overall finding that bupropion cases were more likely to have MMDs and need for hospitalization when compared with either TCA or SSRI cases. We did find that hypertension was more frequent in the SSRI cases than in the bupropion cases, whereas Sheridan et al⁸ found that hypertension was more common in the bupropion cases than in the TCA cases. They also found that TCA cases had a higher rate of intubations than bupropion cases, whereas we found that bupropion cases were more likely to be associated with intubation compared with SSRI cases. Both the bupropion and TCA cases had a 0.3% mortality rate in the Sheridan et al⁸ study. We had a similar mortality rate of 0.22% in the bupropion cases but had no deaths reported in the SSRI cases.

Adolescents who survive a first overdose are at a greater risk of later successfully completing suicide. Authors of a large population-based cohort study estimated this risk at 32-fold compared with matched controls without a history of self-poisoning at 1 year after the index attempt. The effect was also durable, with a risk of at least 10-fold persisting out to 12 years after the index attempt.⁶ It would therefore seem advisable to limit access of adolescents to agents that are less toxic in overdose, particularly for patients who have a history of attempting self-harm.

There are a number of limitations to our study. First, it is a retrospective study of voluntarily reported data; therefore, it is likely that there are qualifying exposures that do not appear in our data set. This may be because patients did not seek medical attention, died before receiving it, did not develop symptoms from their exposure, or because treating medical providers did not deem it necessary to consult with a poison center. Second, although poison centers strive to create complete records, it is often difficult to precisely quantify the size of an exposure, and confirmatory testing is often not performed. Third, a number of case records were coded as “lost to follow-up/left [against medical advice],” which imprecisely describes, and possibly obscures, a range of possible outcomes. Fourth, it is possible that our data are biased toward more symptomatic patients because they are more likely to seek or require medical attention for their exposure. Fifth, poison centers are often limited to the medical history and treatment details provided by the treating personnel because they are call center–based and often do not have access to the electronic medical record system of every hospital for which they provide coverage. These factors may create inaccuracies in case records; however, all cases undergo review by medical toxicology faculty to ensure quality and consistency. In addition, we selected the major outcomes of intubation, seizure, CPR, and ECMO and therapies including oxygen and benzodiazepines, because they are objective measures that are more likely to be documented consistently. Sixth, the source of medications used in overdose is often not known; however, some inferences may be made from the chronicity of exposure as recorded in the poison center case. If it is known at presentation or learned during the clinical course that the patient’s own prescribed medication was involved, the exposure will be coded as “acute-on-chronic” in NPDS data; by contrast, exposures to medications prescribed to another person or acquired for the purpose of misuse or abuse would be coded as “acute.” Finally, a higher proportion of asymptomatic bupropion exposures may be hospitalized because of the well-known risk for delayed seizures up to 18 hours postingestion and the common poison center recommendation for at least 23 hours of observation. By comparison, the recommended observation time for SSRIs is often ≤8 hours.

In our analysis, we were unable to determine if any individual agent was responsible for a disproportionate amount of a given clinical effect, therapy, or outcome. We considered SSRIs as a class when determining their clinical effects in overdose, although there are nuances in the toxicities of individual agents. For instance, paroxetine is the most anticholinergic, whereas citalopram and escitalopram are most likely to cause QT prolongation.^27,28 Vilazodone and vortioxetine are new members of the SSRI class that received FDA approval for use in adults with depression in 2011 and 2013, respectively; they are not approved for pediatric use. Authors of 1 study of single-agent vilazodone ingestion using NPDS data found considerable toxicity with this agent; 2% of patients required intubation, 2% had ≥1 seizures, 6% had reported confusion, 4% had tremor, and 1% had hyperthermia and muscle rigidity.²⁹ Both substances are presently coded as other types of SSRIs in NPDS data, and it is therefore highly likely that our data contain some cases with exposure to these medications. It is difficult to determine their overall contribution to the studied clinical effects and outcomes; however, it would most likely have the effect of biasing our data toward worse morbidity with SSRIs.

This study of poison center data reveals that in reported cases of adolescent self-harm via overdose, bupropion cases had significant increases in rates of serious outcomes, such as seizures, respiratory failure, agitation, conduction disturbance, need for ICU admission, and death when compared with cases involving SSRIs. Suicidal ingestions are increasing steadily, as are the numbers of adolescents treated with medication for depression. In light of bupropion’s disproportionately significant morbidity and mortality risk, it would be prudent for practitioners to avoid the use of this medication in adolescents that are at risk for self-harm.

CPR

cardiopulmonary resuscitation

ECMO

extracorporeal membrane oxygenation

FDA

Food and Drug Administration

MMD

moderate or major outcome or death

NPDS

National Poison Data System

SSRI

selective serotonin reuptake inhibitor

TCA

tricyclic antidepressant