CLINICIAN’S CAPSULE
What is known about the topic?
Most children who sustain a concussion suffer from post-concussion symptoms for many weeks following the accident.
What did this study ask?
Is there an association between ondansetron administration and the presence of persistent post-concussion symptoms at 1 week and 1 month in children?
What did this study find?
This observational study found that the use of ondansetron among children presenting to the emergency department with an acute concussion was associated with a higher risk of persistent symptoms at 1 month.
Why does this study matter to clinicians?
The treatment of short-term symptoms of a concussion should be balanced with the potential long-term harm of ondansetron.
INTRODUCTION
A concussion is a problem commonly evaluated in the emergency department (ED). Recent studies suggest that children may represent as much as 90% of all concussionsReference Amoo-Achampong, Rosas and Schmoke 1 with an annual incidence for teenagers, varying between 10.5/1,000 to 16.5/1,000.Reference Zhang, Sing and Rugg 2 Between 55% and 90% of patients who sustain a concussion suffer from post-concussion symptoms at 1 week following the accident.Reference Carroll, Cassidy and Peloso 3 - Reference Zemek, Barrowman and Freedman 5 These symptoms can be cognitive (memory loss, attention deficit, etc.), somatic (headache, fatigue, nausea), or psychological (depression, irritability, etc.) in nature. Moreover, studies report that 40% of concussed patients suffer persistent symptoms at 1 month,Reference Ingebrigtsen, Waterloo and Marup-Jensen 6 and as many as 15% remain symptomatic at 1 year post-injury.Reference Alexander 7 , Reference Reitan and Wolfson 8
Many patients requiring medical attention following a concussion are initially evaluated in the ED.Reference Zemek, Grool and Rodriguez Duque 9 As soon as proper assessment is completed, treatments offered for patients suffering from a concussion are limited and mainly directed at alleviating symptoms. This includes medication given in the acute phase to decrease pain, nausea, or sometimes dehydration secondary to vomiting. Aside from guidelines recommending a period of activity restriction (physical and cognitive rest),Reference McCrory, Meeuwisse and Aubry 10 - Reference Giza, Kutcher and Ashwal 14 there is little evidence to inform management of the concussion. Four systematic reviews confirm the paucity of studies evaluating treatment for a concussion,Reference Borg, Holm and Peloso 15 - Reference King, Brughelli, Hume and Gissane 18 concluding that no acute intervention has been shown to improve the recovery and outcome of concussed patients.
Over the past few years, there is a growing trend among pediatric emergency physicians to prescribe ondansetron for children with a concussion who present with vomiting.Reference Sturm, Pierzchala, Simon and Hirsh 19 , Reference Freedman, Uleryk, Rumantir and Finkelstein 20 Although being given mainly to relieve symptoms in the acute setting, ondansetron may improve recovery from a concussion by decreasing early symptoms of nausea and vomiting, decreasing energy demands, and enhancing brain rest. However, the clinical impact of ondansetron on time to recovery following a concussion has been poorly evaluated. The primary objective of this study was to evaluate the association between ondansetron administration and the presence of persistent post-concussive symptoms (PPCS) at 1 week and 1 month in children.
METHODS
Design
This was a secondary analysis of the data of the 5P study,Reference Zemek, Barrowman and Freedman 5 a prospective cohort study of children with a concussion evaluated in multiple pediatric EDs in Canada. The idea of this sub-analysis emerged during the recruitment of patients but before the analysis.
Setting
Nine Canadian pediatric hospital EDs were involved in the project and recruited patients in 2013-2015. These sites are members of the Pediatric Emergency Research Canada (PERC) group. All of their institutional review boards approved the study.
Participants
Study participants were children and adolescents ages 5 to 17.99 years who presented in the first 48 hours following a head trauma and met the Zurich concussion criteria.Reference McCrory, Meeuwisse and Aubry 10 Patients with a language barrier, abnormal neuroradiological findings, neurosurgical operative intervention, intubation or pediatric intensive care unit (PICU) care required, multisystem injuries with treatment requiring admission to the hospital, operating room or procedural sedation, neurological developmental delay, or intoxication were excluded. All participants/families provided written informed consent to be included in the study.
Participants of the sub-study reported nausea and/or vomiting at recruitment (based on an item from the self-reported version of the Post-Concussion Symptom Inventory [PCSI]).
Outcomes
The primary outcome was PPCS at 1 week following the index traumatic event, as defined in the 5P study.Reference Zemek, Barrowman and Freedman 5 This was defined as an increase from the pre-concussion baseline of at least three symptoms from the PCSI. The PCSI is a symptom scale that queries symptoms reflecting physical, cognitive, emotional, and sleep domains.Reference Sady, Vaughan and Gioia 21 The decision to evaluate the outcome at 1 week was based on the observation that more than 50% of children who sustain a concussion suffer from persistent symptoms at 1 week,Reference King, Brughelli, Hume and Gissane 18 , Reference King 22 , Reference King, Crawford and Wenden 23 and that most guidelines recommend rest for at least 1 week.Reference Harmon, Drezner and Gammons 12 , Reference Giza, Kutcher and Ashwal 14 Because the outcome may be more important at 1 month for some individuals, persistence of symptoms for 1 month was also included as a primary outcome.
Secondary outcomes included the presence of nausea at 1 week and 1 month.
Treatment
The exposure of interest was a dichotomous variable representing those who received at least a single dose of intravenous or oral ondansetron in the ED. The comparison (reference) group included all other children with nausea and/or vomiting at baseline but did not receive ondansetron.
Demographic and clinical characteristics
Because this was not a randomized trial, other characteristics were considered as covariates in the analyses to minimize the risk of confounding bias when estimating the treatment effect of ondansetron. We considered a broad range of 37 pre-treatment variables measured in the original 5P study as the initial list of candidate confounders, which included demographic characteristics (e.g., age, sex), pre-morbid history (e.g., patient previous history of concussion, migraines, learning disability, depression), initial clinical presentation at the ED (e.g., loss of consciousness, mechanism of injury), extent of a wide range of acute concussion-related symptoms, as well as performance on standardized tests in the ED (e.g., Balance Error Scoring System tandem stance test). The full list of candidate confounders is presented in Table 1.
Table 1. Participants’ baseline demographic characteristics (n=1,805)
All PCSI-P scores are delta scores (difference between current symptom and pre-injury symptom; if a negative value then a score of 0 is given).
PCSI=Post-Concussion Symptom Inventory.
Analysis
Univariate and multivariable logistic regression analyses were used to estimate the relationship between treatment (receiving or not receiving ondansetron in the ED) and each of the four study outcomes. For each multivariable model, we maximized the number of covariates that were included based on a 15:1 event-per-variable ratio where “events” referred to the number of observations in the outcome level with the smaller proportion of the sample. Because the number of events differed for each of four study outcomes, the number of covariates ultimately included in the respective regression models also varied (see Table A1 in the Online Appendix). To prioritize the inclusion of the maximum number of covariates for regression adjustment, the non-linearity of continuous variables was not considered, and variables with greater than two categories were collapsed into two levels considered the most meaningful or treated as a continuous variable.
Because there was a non-trivial amount of missing values for covariates and outcomes (24.2% of cases had at least one missing value), we initially performed multiple imputation using additive regression, bootstrap, and the predictive mean matching method.Reference White, Royston and Wood 24 Variables in the imputation model included all 38 candidate covariates, the four study outcomes, plus the study site. The imputation procedure was repeated 25 times (i.e., 100x the fraction of cases with incomplete covariates). As soon as these imputed data sets were generated, the proposed multivariable logistic regressions were fitted separately for each of the imputed data sets, where parameter estimates are then averaged across these data sets, and imputation-adjusted variance estimates are computed. Lastly, clustered standard errors for the treatment effect were estimated to account for potential intra-site correlation. All analyses were performed using R version 3.3.2 (rms, Hmisc packages).
RESULTS
Among the 3,063 children included in the 5P study, 1,805 children were eligible for the current study (Figure 1), the PPCS outcome at week 1 was available for 1,549 children, and it was available for 1,516 children at 4 weeks. Table 1 demonstrates that baseline characteristics of the participants were similar between the two groups with the exception that patients in the ondansetron group received other medications more frequently than controls. Among these participants, 132 (7%) received oral (120 ) or intravenous (10 ) ondansetron (2 received it by both routes). PPCS were present in 969/1,549 (63%) at 1 week and 524/1,516 (35%) at 1 month. With respect to the presence of nausea, 480/1,546 (31%) reported this symptom at week 1, and 212/1,518 (14%) at week 4.
Figure 1 Flow diagram of participant recruitment and selection.
In the unadjusted analysis, the use of ondansetron was not statistically associated to the PPCS at 1 week (OR: 1.08 [95% CI: 0.73-1.59]) or 1 month (OR: 1.18 [95% CI: 0.80-1.74]) (Table 2). Once adjusted for covariates, there remained no statistical association between ondansetron use and the risk of PPCS at 1 week (adjusted OR: 1.17 [95% CI: 0.68-2.04]); however, an increase association was observed at 1 month (OR: 1.33 [95% CI: 1.05-1.68]).
Table 2. Association between use of ondansetron and outcomes (n=1,805)
* The denominator changes depending on how many missing values on the outcome (due to loss to follow-up or participant simply not completing question); values reflect raw data prior to performing multiple imputations.
† To prevent from overfitting, the number of model covariables included depended on sample size considerations (i.e., number of observations in a smaller level of binary study outcome). See Table 1 for specific covariables included in each model.
Use of ondansetron in the ED was not statistically associated to the presence of nausea at 1 week or 1 month following the ED visit on univariate logistic regression and after adjusting for other predictors (see Table 2).
DISCUSSION
This study failed to identify a decrease in post-concussion symptoms at 1 week and 1 month following a concussion among children ages 5 to 17.99 years who received ondansetron in the ED. However, the use of ondansetron was associated with an increase in the risk of PPCS at 1 month on multiple regression.
There is a paucity of literature specifically describing the use of ondansetron for patients with head trauma and, to our knowledge, no study has evaluated other antiemetics for long-term concussion symptoms. A retrospective cohort study involving 6,311 children with head concussion reported a lower risk of revisit associated with the use of ondansetron in children.Reference Sturm, Simon, Khan and Hirsh 25 However, major differences between the two study populations (use of ondansetron, 25% v. 7%), the proportion of a computed tomography (CT) scan (100% v. 4% ) and the number of sites involved may have led to confounding by severity – the sicker patients being more at risk of receiving ondansetron and having PPCS. Another small pilot study conducted to assess the feasibility of a randomized controlled trial of the impact of ondansetron in concussed teenagersReference Gravel, Carriere and D’Angelo 26 reported a trend towards a lower proportion of children with PPCS at 1 week (OR: 0.60 [95% CI: 0.08-4.4]) and 1 month (OR: 0.20 [95% CI: 0.02-1.70]) following oral ondansetron.Reference Gravel, Carriere and D’Angelo 27
From a theoretical perspective, ondansetron holds promise for potentially minimizing post-concussion symptoms. The most commonly cited hypothesis for the pathophysiology of concussion suggests that shear forces experienced at the time of injury lead to ionic fluxes, the releases of excitatory neurotransmitters, and a spreading depression-like phenomenon.Reference Giza and Hovda 28 - Reference Shrey, Griesbach and Giza 32 The decreased cerebral blood flow after injuryReference Maugans, Farley and Altaye 31 limits the influx of adenosine triphosphate (ATP) required to restore homeostasis. The mismatch between demand and supply for ATP may prolong concussion symptoms. As such, physical and cognitive rest may improve recovery through conservation of limited ATP supplies following injury to the brain. Ondansetron may improve recovery from concussion by decreasing early symptoms of nausea and vomiting, which has been associated with PPCS at 3 months.Reference Babcock, Byczkowski and Wade 33 Also, this could improve rest and improve recovery because it was previously demonstrated that PPCS is inversely related to rest quality.Reference Boutis, Cogollo and Fischer 34 , Reference Moser, Glatts and Schatz 35 On the other hand, ondansetron could also be deleterious to recovery for multiple reasons. By minimizing the initial symptoms, it could hasten return to play and limit brain rehabilitation, as well as increase the risk of a second impact on brain cells. However, a recent study showed that early activity was associated with a lower risk of PPCS.Reference Grool, Aglipay and Momoli 36 Also, decreasing initial symptoms could modify medication consumption or feeding. Finally, ondansetron’s effect on 5HT-3 receptors may interplay with an energy homeostasis of the brain and slow recovery process.
The clinical impact of this study is limited by its exploratory design. However, it raises the question as to whether ondansetron may have a deleterious effect on brain recovery following concussion. Consequently, the treatment of short-term symptoms like nausea and vomiting should be balanced with the potential long-term harm of ondansetron.
There are several limitations inherent to the current study. Among them, the use of an observational design is prone to bias by indication. It is possible that the baseline risk of PPCS was different for children who received ondansetron than for the control group. They may have suffered from a more severe concussion. To account for this, we adjusted the analysis according to multiple factors identified in the 5P study. To balance for unknown risk factor, it would be important to conduct a randomized clinical trial. Missing data limited the precision of our findings. To account for this, we used multiple imputation techniques. Although this permitted to include more patients, there is a risk of unknown biases related to missing data. Another limitation is the absence of a standard dose, number, or route of administration of the ondansetron. This may impact effectiveness of the medication and could be addressed by another study. Also, we did not collect information regarding home medication. Finally, the inclusion of patients who received other antiemetic medication in the control group may have biased the result towards the null hypothesis. However, there were very few patients for whom other antiemetic medication administration was documented.
In conclusion, this exploratory sub-analysis failed to identify a decrease in the risk of PPCS at 1 week among children ages 5 to 17.99 years who suffered a concussion and received ondansetron in the ED. Moreover, as a group, patients who received ondansetron had a worse outcome at 1 month. While this may be related to the design of our study, it highlights the importance of evaluating the impact of this medication on recovery from concussion. Future studies should continue to explore potential therapies to mitigate ongoing concussion symptoms and sequelae.
Competing interests
None declared.
SUPPLEMENTARY MATERIALS
To view supplementary material for this article, please visit https://doi.org/10.1017/cem.2018.384
