Infusion reactions to infliximab in children and adolescents: frequency, outcome and a predictive model

Authors


Correspondence to: Dr W. Crandall, Columbus Children's Hospital, Section of Pediatric Gastroenterology and Nutrition, 700 Children's Drive, Columbus Ohio, 43205-2696, USA. E-mail: crandallw@pediatrics.ohio-state.edu

Summary

Background : Crohn's disease commonly affects children and adolescents, however the majority of research into the safety and efficacy of therapies for inflammatory bowel disease, including infliximab, has occurred only in adults.

Aim : To determine the rate of reactions in children following infliximab infusions, and to identify variables that might be predictive of those reactions.

Methods : We performed a retrospective review of all infliximab infusions performed at Columbus Children's Hospital from December 1998 through September 2001.

Results : Fifty-seven children received 361 infusions. Three hundred and fifty-five of the 361 infusions (98.3%) were completed. Fifty children had 304 repeat infusions. There were a total of 35 infusion related reactions. Female gender and the use of immunosuppressive medications for less than 4 months were risk factors for a reaction to infusion number 2. A reaction to infusion 2 and immunosuppressive use for less than 4 months were risk factors for infusion number 3.

Conclusions : The rate of infusion reactions in children receiving infliximab is similar to that in adults. Female gender, immunosuppressive use for less than 4 months and prior infusion reactions may be risk factors for subsequent infusion reactions in children.

Introduction

Crohn's disease affects approximately 400 000–600 000 people in North America,1 with 20–30% of cases occurring in people under 20 years of age.2 Despite the fact that this disease often presents in the paediatric population, the majority of research into the safety and efficacy of various therapies for inflammatory bowel disease occurs in adults.

Infliximab, a chimeric antibody to tumour necrosis factor-α, has become an important modality in the treatment of moderate to severe Crohn's disease, as well as in the treatment of rheumatoid arthritis. Large trials in adults, and smaller paediatric studies have shown it to be efficacious for both Crohn's disease3–10 and rheumatoid arthritis.11

Infusion reactions, thought to be secondary to the formation of antibodies directed against infliximab, are a well-described complication of infliximab therapy, occurring in 17–24% of adult patients.8, 9, 12 Both immediate and delayed reactions have been described.8, 9 Infusion reactions range from relatively mild symptoms such as flushing, to severe anaphylactic/anaphylactoid reactions. Slowing the infusion is often sufficient to treat patients with infusion reactions, and the adult experience suggests that most infusions can be completed despite a reaction.8–10 It has also been suggested that concomitant use of immunosuppressive medications and shorter intervals between infusions may decrease the incidence of infusion reactions.9, 10,12

Little is known about the rate of infliximab-related infusion reactions in children,6,13, 14 although one group of authors has suggested that it may be lower in children than in adults.15 Similarly, there are little data regarding the percentage of infusions that are completed, or regarding what factors may be either predictive of, or protective against infusion reactions in children. We sought to begin to answer these questions through a review of our use of infliximab.

Methods

We reviewed the records of all children who received infliximab in the infusion centre at Columbus Children's Hospital from December 1998 through September 2001. This study was limited to infusions performed in the infusion centre in order to have as much uniformity as possible in the observation, reporting and treatment of reactions. The infusion centre has a dedicated team of four nurses trained and experienced in the administration of infliximab and other intravenous therapies. All infusions were given per the following protocol. The infliximab was reconstituted as instructed on the package insert. Infusions were started within 3 h of preparing the infliximab. Patients were premedicated as per individual physician orders (see Results). Infusions were started at 10 mL/h and advanced every 15 min to 20 mL/h, 40 mL/h, 80 mL/h and to a maximum rate of 125 mL/h according to predetermined institutional protocol. Patient interview and vital signs were performed every 15–30 min.

Seven infliximab infusions (1.9% of the total infusions) occurred outside of the infusion centre, all in an in-patient setting. Three of these infusions (0.8% of the total infusions) involved patients reported in this study. Although infusion reactions for these three infusions were not evaluated in the current study, they were considered in calculating the variable ‘time interval from the last infusion’. In-patient infusions occurred due to either disease exacerbation that resulted in hospitalization, or according to insurance companies reimbursement guidelines; none occurred in the in-patient setting due to a perceived increased risk of a reaction, therefore lessening the potential bias of excluding these infusions.

The infusion centre records were reviewed, and all symptoms that occurred during an infusion were recorded, regardless of severity. The gastroenterology or rheumatology clinic chart was also reviewed, and complaints consistent with a delayed hypersensitivity reaction occurring within 30 days of the infusion were noted. With regards to delayed symptoms such as fever or arthritis, the patient's primary gastroenterologist was asked to make a determination as to whether the complaints were related to the infliximab if those complaints predated the infusion. All medications were recorded for each patient. Immunosuppressive medications were defined as 6-mercaptopurine, azathioprine and methotrexate. Immunosuppressive use was analysed as greater than or less than 4 months duration, due to some evidence for a delayed response to 6-mercaptopurine and azathioprine.16, 17 No patients were on cyclosporin or mycophenolate.

Institutional review board approval for this study was obtained prior to all data collection. No financial support was received for this study.

Statistics

Descriptive statistics were used to examine rates of reactions and trends in the data across all infusions. Minor reactions that did not require treatment were felt to be less clinically important than reactions that necessitated medical intervention, therefore Chi-square analysis, Fisher's exact tests and stepwise logistic regression analyses were conducted to investigate predictors of the occurrence of ‘medically significant reactions’, defined as any reaction resulting in any intervention by the nurse or physician (e.g. medication, slowing the infusion). Individual predictors in the regression models were assessed with the Wald statistic when appropriate. When the Wald statistic was inappropriate due to large standard error,18 the Hauck and Donner19 method of comparing models with and without the variable in question was used (to assess the contribution of the individual variable to the overall model). Separate logistic regression analyses were conducted to examine predictors of reactions on the second and third infusions. These infusions were chosen because both the likelihood of having a reaction and the sample sizes were greatest on these infusions, thus increasing power.

Results

We administered 361 infliximab infusions in 57 children from December 1998 through September 2001(Table 1). The majority of infusions were for the treatment of Crohn's disease (Table 1). This is a retrospective clinical series, therefore indications/treatment criteria for the use of infliximab were based on the prescribing physician's clinical judgement. Most children were premedicated with acetaminophen (70% of infusions) and diphenhydramine (87.3% of infusions), while steroids were given prior to 25.7% of infusions (range of 20–100 mg of methylprednisolone). The number of infusions per child ranged from one to 32, with a mean of 6.3. Thirty-four (60%) patients had four or more infusions. Three hundred and four repeat infusions were performed. Infusions were completed 98% (355/361) of the time.

Table 1.  Patient demographics
 MaleFemale
N%Mean (s.d.)RangeN%Mean (s.d.)Range
Diagnosis
 Crohn's disease2388.5  2787.1  
 Ulcerative colitis00.0  26.5  
 Unspecified inflammatory bowel disease27.7  00.0  
 Juvenile rheumatoid arthritis13.8  26.5  
Age at first infusion (years)  15.00 (2.47)11.08–20.58  15.60 (3.84)5.15–22.71
No. infusions per child   6.58 (8.15)1–32  6.13 (4.92)1–23
 1 infusion311.5  412.9  
 2–4 infusions1350.0  1032.3  
 5–7 infusions519.2  929.0  
 ≥ 8 infusions519.2  825.8  

Thirty-one (8.5%) immediate infusion reactions and four (1.1%) delayed reactions were identified. Twenty-two children (38.6%) had infusion reactions, with a range of zero to four immediate reactions per child, and zero to one delayed reactions per child. Eight children had more than one reaction. One child had both an immediate and a delayed reaction with the same infusion (Figure 1). Infusion reactions occurred in seven per cent of first infusions. Most reactions occurred on the second and third infusions (Figure 2).

Figure 1.

The number of total and medically significant infusion reactions per child.

Figure 2.

The rate of infusion reactions per infusion number (i.e. first infusion, second infusion, etc.).

The most common symptoms of immediate infusion reactions were flushing and shortness of breath (Table 2). Symptoms of immediate reactions were most commonly treated with diphenhydramine and/or stopping and restarting the infusion, usually at a slower rate. Delayed symptoms occurred an average of 6 days after infusion, with a range of 2–10 days. The most common delayed symptoms were joint pain and swelling, which were treated with diphenhydramine, ibuprofen and prednisone.

Table 2.  Infusion reaction symptoms and treatments
Immediate reactionsNDelayed reactionsN
  • *

     In addition to joint pain.

  •  In addition to rash.

  •  In addition to fever.

Symptoms Symptoms 
 Flushed12 Joint pain/swelling6
 Shortness of breath11 Rash3
 Hot6 Fever2
 Dizzy/light-headed4 Abdominal pain*1
 Hives3 Fatigue*1
 Nausea/vomit3 Flushed face*1
 Numbness2 Headache1
 Rash3  
 Leg/knee pain2  
 Hypotension1  
 Red/puffy eyes1  
 Cold/chills1  
 Difficulty swallowing1  
 Headache1  
 Heart racing1  
 Pale1  
 Stomach pain1  
 Tingling1  
 Dysphoria1  
 Cold throat1  
Treatment Treatment 
 Diphenhydramine14 Diphenhydramine3
 Stopped then   restarted infusion11 Ibuprofen3
 Steroid9 Prednisone3
 No treatment6  
 Aborted infusion4  
 Epinephrine2  
 Ondansetron1  

Twenty-four (6.6%) of all infusions required medical intervention for an infusion reaction, and were classified as ‘medically significant infusion reactions’, with 20 being immediate, and four delayed. Thirteen (23%) patients had medically significant reactions, ranging from zero to three reactions per patient. Seven patients (12.3%) had more than one significant infusion reaction. One (1.8%) medically significant reaction occurred during the first infusion. Most significant reactions occurred on the second and third infusions (Figures 1 and 2). The number of subsequent infusions after a significant reaction ranged from 0 to 14. Two patients required subcutaneous adrenaline (epinephrine) due to severe reactions. One of these children had no further infusions; the other had 13 subsequent successful infusions prior to the end of this series (Table 3).

Table 3.  Summary of medically significant infusion reactions
PatientGender Diagnosis*Age at
first infusion
(years)
Date of
first
infusion
Date of infusion
with reaction
Infusion
number
Premedications TypeSymptomsTreatmentNo. of infusions
after last reaction
  • CD = Crohn's Disease; JRA = juvenile rheumatoid arthritis.

  • † 

    † I = immediate reaction; D = delayed reaction.

 1FCD17.7212/983/992DiphenhydramineIShortness of breath;
 dizzy/lightheaded
Diphenhydramine;
Prednisone; aborted
 infusion
7
     11/993Acetaminophen
Diphenhydramine
Methylprednisolone
IDizzy; dysphoriaStopped and
 restarted infusion
 
 2FCD15.581/995/992Acetaminophen
Diphenhydramine
IFlushed; shortness of
 breath; hot
Diphenhydramine;
Methylprednisolone;
 stopped and
 restarted infusion
7
     8/993Acetaminophen
Diphenhydramine
Methylprednisolone
DJoint pain/swellingDiphenhydramine;
Ibuprofen
 
 3FCD10.485/996/992NoneIFlushed; shortness
 of breath
Diphenhydramine;
Methylprednisolone;
 aborted infusion
0
     9/013Diphenhydramine
Methylprednisolone
DJoint pain/swelling;
 rash; fever; flushed
Diphenhydramine;
Ibuprofen;
Prednisone
 
     9/014Diphenhydramine
Methylprednisolone
I

Flushed; shortness of
 breath; numbness;
 puffy eyes
Diphenhydramine;
Aborted infusion
 
 4FCD16.975/999/99
10/99
2
3
None
Acetaminophen
Diphenhydramine
Prednisone
I
I
Nausea/vomit
Flushed; shortness of
 breath; leg/knee  pain
Ondansetron
Diphenhydramine;
Methylprednisolone;
Epinephrine;
7
       Ondansetron  Aborted infusion 
     8/009Acetaminophen
Diphenhydramine
Methylprednisolone
Ondansetron
IShortness of
 breath; numbness
Diphenhydramine;
 aborted infusion
 
 5FCD16.589/9910/992NoneIFlushed; shortness
 of breath
Diphenhydramine;
Prednisone; stopped
and restarted infusion
0
     11/993Acetaminophen
Diphenhydramine
Methylprednisolone
IFlushed; shortness
 of breath
Diphenhydramine;
Methylprednisolone;
 stopped and
 restarted infusion
 
     1/004Acetaminophen
Diphenhydramine
Methylprednisolone
IShortness of breathEpinephrine 
 6FCD22.719/9911/992Acetaminophen
Diphenhydramine
IFlushed; hot;
 heart racing
Methylprednisolone;
 stopped and
 restarted infusion
8
     12/993Diphenhydramine
Methylprednisolone
IFlushed; hotDiphenhydramine;
 stopped and
 restarted infusion
 
     2/004Acetaminophen
Diphenhydramine
Methylprednisolone
IFlushedDiphenhydramine;
 stopped and
 restarted infusion
 
 7MCD12.559/0011/003Acetaminophen
Diphenhydramine
IHives; rash;
 red eyes
Diphenhydramine;
Methylprednisolone;
 stopped and
 restarted infusion
0
 8MJRA14.258/0011/004Acetaminophen
Diphenhydramine
IFlushedDiphenhydramine1
 9MCD13.161/0011/006Acetaminophen
Diphenhydramine
Methylprednisolone
IFlushed; shortness of
 breath; cold throat
Diphenhydramine0
10FCD19.742/013/012DiphenhydramineIHot; tinglingStopped and
 restarted infusion
3
11FCD14.0010/993/015Diphenhydramine
Ibuprofen
DJoint pain/swelling;
 rash; fever
Prednisone3
12FCD18.617/017/011Acetaminophen
Diphenhydramine
IHypotensionStopped and
 restarted infusion
1
13FCD16.913/019/012Acetaminophen
Diphenhydramine
Methylprednisolone
I
D
Difficulty swallowing;
hives
Rash
Diphenhydramine;
 aborted infusion
Diphenhydramine
0

Chi-square analyses or Fisher's exact tests were performed on potential predictor variables (Table 4). Variables that remain constant across time (e.g. gender) were analysed with the variable ‘reaction ever’ across all infusions. Variables that change from infusion to infusion (e.g. days since prior infusion) were analysed separately for the second and third infusions as noted in the methods section. Immunosuppressive medication use less than 4 months previously and having had a medically significant reaction on the prior infusion were significantly associated with having a medically significant reaction. Gender approached statistical significance.

Table 4.  Chi-square and Fischer exact analysis of potential predictor variables of infusion reactions
 ReactionNo reaction 
 %M (s.d.)%M (s.d.)P
All infusions
Gender    0.060
 % Female76.9 47.7  
 % Male23.1 52.3  
Infusion schedule
 % PRN56.5 50.0  
 % 0/2/6 weeks43.5 50.0 0.416
Age at first infusion 16.1 (3.3) 15.1 (3.27)0.339
Total number of infusions 6.7 (4.4) 6.2 (7.1)0.824
Second infusion
Infusion schedule
 % PRN50.0 45.2  
 % 0/2/6 weeks50.0 54.8 0.552
On immunosuppressive medication    0.024
 ≥ 4 months62.5 95.2  
 < 4 months37.5 4.8  
Days since prior infusion 82.9 (58.4) 62.0 (73.8)0.453
Third infusion
Infusion schedule
 % PRN42.9 44.1  
 % 0/2/6 weeks57.1 55.9 0.642
On immunosuppressive medication    0.031
 ≥ 4 months42.9 85.3  
 < 4 months57.1 14.7  
Medically significant
  reaction on second infusion
    0.000
 Yes85.7 2.9  
 No14.3 97.1  
Days since prior infusion 178.3 (280.2) 56.4 (36.91)0.294

Stepwise logistic regression analyses were conducted on the second and third infusions to identify predictors of a medically significant reaction. For both infusions, the variables of gender, the amount of time since the last infusion and use of an immunosuppressive medication (> or < 4 months) were examined. For the third infusion, a prior reaction on the second infusion was added to the set of variables. Gender, immunosuppressive medication use (> or < 4 months) and a history of a prior reaction were included as a result of clinical observation and statistical significance in the initial analysis. Time since last infusion was included in the analysis due to the opinion of some experts that a longer interval between infusions is a risk factor for infusion reactions.

For the second infusion, the time since last infusion did not enter into the model. Examining overall model statistics, a model with the remaining variables (gender and immunosuppressive medications) was significant, χ2 (2, N = 50) = 15.88, P < 0.001. This indicates that the predictors as a set reliably predict the occurrence of a medically significant reaction, and the model accurately classified 88.0% of cases (Table 5). For the individual predictors, the Wald statistic was significant for immunosuppression medication (z = 4.33, P < 0.001), indicating that it is significantly associated with a lower risk of reaction. A large standard error for gender (S.E. = 53.01) suggested that the Wald statistic was not appropriate for this variable. The Hauck and Donner19 method for determining significance was therefore employed: a model with immunosuppression medication only was analysed and compared to the previous model with both immunosuppression medication and gender. The model with immunosuppression medication only was not significant, χ2 (1, N = 50) = 2.10, P > 0.05. Thus, female gender is a significant predictor because adding gender to the model with immunosuppression medication only significantly improves overall prediction.

Table 5.  Logistic regression results for predicting medically significant reactions
Second infusionThird infusion
 Model 1
Significant predictors:Significant predictors:
 ? No immunosuppressive medication or
 immunosuppressive medication < 4 months.
? No immunosuppressive medication or
immunosuppressive medication < 4 months.
 ? Female gender? Female gender
This model accurately predicted 88% of outcomesThis model accurately predicted 88% of outcomes
 Model 2
 Significant predictors:
 
 
? No immunosuppressive medication or
immunosuppressive medication < 4 months.
 ? Medically significant reaction on second infusion.
 This model accurately predicted 95% of outcomes.

For the third infusion, a model with time since last infusion, gender and immunosuppressive medication for more than 4 months was tested first. Time since last infusion did not enter into the model, and overall model statistics indicated that the model with the remaining variables was significant, χ2 (2, N = 41) = 10.44, P < 0.01. This model accurately classified 87.8% of cases, and both individual predictors were significant (immunosuppression medication z = 5.69, P < 0.05; gender z = 4.01, P < 0.05). A second model with the variable prior reaction on second infusion added was tested. In this model, gender no longer entered the model, leaving only reaction on second infusion and immunosuppression medication in the model. Overall statistics for this model were significant, χ2 (2, N = 41) = 27.24, P < 0.001. This model accurately classified 95.1% of cases, making it a better model for the third infusion (Table 5). Both variables in the second model had large standard errors (S.E. = 81.26 for immunosuppression medication; S.E. = 81.27 for prior reaction), therefore Hauck and Donner's19 method was employed to determine significance. Adding the variable of prior reaction on second infusion results in better prediction than a model with immunosuppression medication and gender (see chi-square results above for the first model for infusion number three). Occurrence of a prior reaction on second infusion is associated with a higher risk of significant reaction on the third infusion, and immunosuppressive medication > 4 months is associated with a lower risk.

Discussion

There are an estimated 100 000 children with Crohn's disease in the United States, making this one of the most common chronic gastrointestinal disorders treated by paediatric gastroenterologists.1, 2 Infliximab appears to be an important therapeutic option for moderate to severe Crohn's disease in children. As with many drugs, however, there are limited data regarding the efficacy, safety and side-effects of this medication in the paediatric population. This retrospective series begins to clarify some of the issues related to the safety and side-effects of infliximab in children.

The rate (8.5% of infusions) of infusion reactions in our series was comparable to that reported in adults (approximately 4–13%),8, 9 while the rate of infusion reactions per individual (38.6%) was higher than often reported in adults (17–24%).8, 9,12 A small study by Kugathasan et al. suggested a decreased rate of infusion reactions in children as compared to adults,15 while a much larger cohort reported by Stephens et al. had a rate ‘similar to that of adults’.13 Small sample sizes of paediatric studies, variable reporting of the rate of infusion reactions per patient versus the rate per infusion, as well as variable definitions of what events constitute a reaction contribute to the variability in the rates of infusion reactions in paediatric and adult studies, making comparisons difficult. Larger studies with a consistent definition of an infusion reaction would be needed to determine whether or not the rate of infusion reactions in children is different from that in adults, however the available data do not suggest a clear difference.

This was not a controlled trial, therefore no placebo reaction rate was available for comparison. However, seven per cent of children experienced reactions on the first infusion, presumably prior to any sensitization to the infliximab antigens. Infusion reactions are thought to be a result of antibodies to infliximab, thus the rate of infusion reactions prior to the formation of any such antibodies (i.e. the rate with the first infusion) may give a rough estimate of the placebo rate of infusion reactions in our population. This is supported by the fact that only one ‘medically significant’ reaction occurred during the first infusion.

Delayed reactions were an uncommon event (1.1%) in our population, with a rate equivalent to that seen in adults.8, 9 Furthermore, the rate in our study may have been overestimated because symptoms of fever, joint pain and rash may have been due to the Crohn's disease itself, to other medications, or to intercurrent infections. As these reactions were uncommon, we had insufficient power to comment on what factors may be predictive of delayed infusions in children.

Only two (immediate) reactions (severe shortness of breath in both) resulted in the use of adrenaline (epinephrine). Both children had been pretreated with steroids. Although one child was not re-infused, the other went on to have multiple subsequent infusions, suggesting that even a prior, severe reaction is not necessarily an absolute contraindication to subsequent infusions.

The formation of antibodies to infliximab has been associated with an increased risk of infusion reactions,9, 20 although the presence of antibodies does not ensure a reaction. Concomitant use of immunosuppressive medications (for more than 4 months) was found to be associated with a lower risk of infusion reactions for both the second and third infusions, presumably by inhibiting the production of such antibodies. A prior significant reaction was indicative of an increased risk for subsequent infusion reactions, perhaps indicating the presence of an existing, significant titre of antibodies to infliximab. It is not routine practice to measure antibody levels at Columbus Children's Hospital, thus this supposition cannot be confirmed. Female gender was also associated with an increased risk of reactions. Although gender has only recently been identified as a risk factor for infliximab related infusion reactions21 there is a well established gender dimorphism in immune function. Females are known to have enhanced immunoreactivity as compared to males, mediated in part by sex hormones, resulting in an increased resistance to some infections, stronger immune responses to some immunizations, and an increased risk for multiple autoimmune disorders.22, 23 It is not difficult then to hypothesize that females might be predisposed to the formation of anti-infliximab antibodies and the subsequent development of infusions related reactions.

It has been hypothesized that a decreased interval between infusions and more frequent initial dosing (i.e. 0, 2 and 6 weeks) is associated with a decreased risk of infusion reactions. However, we did not observe significant differences with either of these variables in our population. Several explanations for this discrepancy are possible. First, these factors may not be significant in children and adolescents. As this series only had sufficient numbers to detect large effects, there may have been insufficient power to detect smaller, yet real differences. Finally, the relative contribution to the risk of an infusion reaction for these variables may be minor compared to the variables of immunosuppressive use, gender and prior infusion reactions, resulting in a failure to reach significance in a regression model.

We limited this review to infusions performed in a dedicated infusion clinic in order to have consistency in the way the drug was administered, the treatment of infusion reactions (infusion reactions were managed by attending physicians only), and the recording of infusion reactions. However, because time from the last infusion was a variable that we considered, we reviewed the in-patient infusions as well. Seven in-patient infusions were identified during the time of this study. Of the seven, only three involved patients who were reported in the study. The time from last infusion variable was adjusted to reflect their in-patient infusions. We were unable to make corrections for any infusions that may have occurred at other institutions prior to our assuming the care of patients, however, to our knowledge, this was only an issue in one patient.

In conclusion, the rate of infliximab infusion reactions in children is similar to that in adults. Even in the presence of a reaction, most infusions can be completed. Severe reactions may not be an absolute contraindication to future infliximab therapy. Further study is needed to determine whether models using gender, immunosuppressive use and prior infusion reactions are predictive of infusion reactions in children.

Acknowledgements

We would like to thank Dr John Barnard for his expert review and commentary.

Ancillary