Adverse Drug Reactions Causing Hospital Admissions in Childhood: A Prospective, Observational, Single-Centre Study

Authors

  • Petra Langerová,

    1. Department of Pharmacology, Faculty of Medicine and Dentistry, Palacký University and University Hospital Olomouc, Olomouc, Czech Republic
    Search for more papers by this author
  • Jiří Vrtal,

    1. Department of Pharmacology, Faculty of Medicine and Dentistry, Palacký University and University Hospital Olomouc, Olomouc, Czech Republic
    Search for more papers by this author
  • Karel Urbánek

    Corresponding author
    1. Department of Pharmacology, Faculty of Medicine and Dentistry, Palacký University and University Hospital Olomouc, Olomouc, Czech Republic
    • Author for correspondence: Karel Urbánek, Department of Pharmacology, Faculty of Medicine and Dentistry, Palacký University and University Hospital Olomouc, Hnevotinska 3, 77515 Olomouc, Czech Republic (e-mail urbanek@fnol.cz).

    Search for more papers by this author

Abstract

Adverse drug reactions (ADRs) are common problems in both paediatric and adult medicine. The aim of this study was to prospectively identify the ADRs causing hospital admission of children and identification of the risk factors and involved drugs. The study was performed at the University Hospital in Olomouc, Czech Republic. All patients aged 19 years or under admitted to hospital were included in the study, and all admissions for ADRs were prospectively screened for a period of 9 months. Suspected ADRs were subsequently evaluated in detail, and causality assessment was undertaken to determine whether each suspected reaction was possible, probable or definite. The assessment of ADR causality was performed using the Naranjo algorithm, the Liverpool ADR Causality Assessment Tool and the Edwards and Aronson causality assessment method. During the study period, 2903 admissions were identified; of these, there were 143 admissions of patients with an oncological disease. Sixty-four admissions (2.2%) were caused by an ADR. Anticancer chemotherapy accounted for 35% of the cases, followed by antibiotics (18%), immunosuppressants and vaccines (9% each). The use of different scoring systems does not lead to the differences in the numbers of ADR-diagnosed patient but may result in differences in the determination of the level of certainty. ADRs cause a substantial proportion of children's hospital admissions. The majority of the ADR-diagnosed patient affected the hematopoietic and gastrointestinal systems; the drugs most frequently involved were cytotoxic agents and antibiotics. The most important risk factors identified were female sex and oncological disease.

Specialist knowledge of adverse drug reactions (ADRs) is crucial for safe use of medications. ADRs are a common clinical problem in both paediatric and adult medicine that can lead to significant morbidity. ADRs are a frequent cause of hospitalization and one of the leading causes of morbidity and mortality [1]. The toxicity of many medicines in children is different to that seen in adults. Indeed, there are some groups of paediatric patients, such as neonates, in whom drug toxicity appears to be relatively common. However, most ADRs in older children are usually similar to those in adults [2]. On the other hand, mainly young children are the largest sector of the population subjected to medication errors [3].

The safety of drug prescribing has become a very important topic in adult and geriatric medicine. Although children are also vulnerable to ADRs, much less attention has been focused on them. Paediatric patients constitute a sensitive group with regard to rational drug therapy because many new drugs are released onto the market without the benefit of even limited experience in this age group. This often causes that paediatricians prescribe drugs for children in an off-label or unlicensed manner [4]. Adequate controlled clinical trials in children are lacking, mainly because of the issues of cost and responsibility. Spontaneous reporting systems are considered to be subject to under-reporting of ADRs, even of those which are severe. More prospective studies are needed to obtain reliable information about the real incidence of ADRs [5, 6].

The lack of reliable data in the paediatric population is associated with specific problems including limited availability of safety data due to the lack of clinical trials in the paediatric population; under- or over-dosing in some age groups due to the lack of pharmacokinetics data; and maturation, growth and development of the paediatric population susceptible to drug-induced growth and development disorders as well as to delayed ADRs not findable in adults. Pre-marketing trials are able to provide information about the benefits of drugs but do not manage to establish a safety profile [7].

Based on this, there was a need for a legal obligation for pharmaceutical companies to perform studies if they intended to develop medicines for the use in the paediatric population. In 1997, the European Commission discussed problematic issues of paediatric medicines. One of the conclusions was that there was a need to strengthen the legislation. In 1998, the commission supported the need for international discussion on the performance of clinical trials in children in the context of the International Conference on Harmonization (ICH) – an organization working on the harmonization of pharmaceutical regulatory requirements between the EU, Japan and the USA. An ICH guideline was therefore adopted. The goals were to encourage and facilitate timely paediatric medicinal product development internationally and to provide an outline of critical issues in paediatric drug development and approaches to a safe, efficient and ethical study of medicinal products. Subsequently, the ICH guideline became the European guideline. The Directive on Good Clinical Practice for Clinical Trials came fully into force in 2004. This directive takes into account some specific concerns about performing clinical trials in children, and it lays down criteria for their protection in clinical trials [8].

A systematic review of seventeen observational studies of ADRs in paediatric patients performed between 1976 and 1996 (five of them dealt with ADRs in children leading to hospital admission) found the rate of ADR admissions to be 2.09% (from 0.59% to 4.1%). The studies were conducted in seven different countries, mainly in the United States, United Kingdom and Spain (four each) [6]. Another review of prospective paediatric studies published between 2001 and 2007 failed to identify any large studies on the incidence of ADRs causing hospital admissions [9]. Some results were included in a recent systematic review by Smyth et al. in 2011. The authors reviewed prospective studies investigating ADRs in three settings: ADRs in inpatients, those causing acute admission to hospital and those occurring in outpatients. Seventy-one per cent (72/102) of the studies assessed causality, and 34% (34/102) performed a severity assessment. Nineteen studies (19%) assessed avoidability. The incidence rates for ADRs causing hospital admission ranged from 0.4% to 10.3% of all children [10].

The aim of this study was to ascertain the incidence of ADR-related hospital admissions in children and to determine the proportion of the drug groups involved and the types of syndromes. The secondary aim was to compare the results obtained by the use of three different scoring systems for ADR causality.

Methods

All patients younger than 19 years of age admitted to the Department of Pediatrics were included in the study. Surgical cases were included as well. All admissions to the Department of Pediatrics of the University Hospital Olomouc, Czech Republic, during a 9-month period, including weekends and holidays, were prospectively screened. This department is the only paediatric facility in the region. Admissions possibly caused by ADRs were further evaluated. The study period was from 1 March 2012 to 30 November 2012. The definition of ADR used was that of Edwards and Aronson which is ‘an appreciable harmful or unpleasant reaction, resulting from an intervention related to the use of a medical product, which predicts hazard from future administration and warrants prevention or specific treatment, or alteration of the dosage regimen or withdrawal of the product’ [11]. This definition was chosen because it describes only clinically significant adverse reactions that cause harm and includes the concept of preventive action. Patients who were admitted because of intentional drug overdose were excluded.

Admissions were assessed in the hospital information system on a daily basis. The study team collected the following information from the case notes: age, sex, presenting complaint, clinical history, diagnosis and medication including over-the-counter drugs taken previously.

Suspected ADRs were subsequently evaluated in detail, and causality assessment was undertaken to determine whether each suspected reaction was unlikely, possible, probable or definite. The reactions classified as unlikely were excluded from the analysis. The assessment of causality in each patient admitted due to an ADR was performed using the Naranjo algorithm [12], the Liverpool ADR Causality Assessment Tool, an algorithm developed by Gallagher et al. [13], and the Edwards and Aronson causality assessment method [11]. Two investigators (PL and JV) independently assessed causality for all ADR cases using the above-mentioned algorithms. Agreement on the classification result between the two investigators was taken as accepted consensus. Where the investigators did not achieve consensus, a third investigator (KU) assessed the cases to decide on causality. The categories obtained by all three algorithms were compared.

Analyses of ADR rates were based on the number of admissions with the rate expressed as the number of admissions caused by ADR per 100 admissions, together with 95% confidence intervals. Multivariate logistic regression was used to calculate odds ratios (ORs) for possible risk factors for ADRs. Univariate statistical analysis was performed using the Mann–Whitney U-test for the assessment of age-related ADR distribution. Frequency data were analysed by chi-square test.

Results

Over the study period, there were 2903 admissions (1664 boys and 1239 girls). Some patients were admitted repeatedly and some of them had more than one ADR. The average age of the patients admitted was 7.1 ± 5.7 years (for demographic data, see table 1). Of the total, there were 143 admissions of patients with an oncological disease (4.9%).

Table 1. Demographic data of the study cohort, age expressed in years as a mean value ± S.D. (standard deviation)
 N%
Number of admissions2903100
Male166457.3
Female123942.7
Number of patients2405100
Male137457.1
Female103142.9
Average age7.1 ± 5.7 
Male7.0 ± 5.5 
Female7.2 ± 5.9 

Sixty-four admissions (2.2%) were caused by the ADR (28 girls and 36 boys). This gives an incidence of 2.2 ADRs per 100 admissions. ADRs were divided into subgroups of drugs causing ADRs (fig. 1) and organ systems affected by ADRs (fig. 2). Twenty-eight admissions (43.1%) caused by the ADR were found in children suffering from oncological diseases. Sixteen of them were due to an infectious complication. Among the effects of anticancer chemotherapy, the most common ADRs were febrile neutropenia (12 admissions) and mucositis (five admissions) (figs 1 and 2).

Figure 1.

Groups of drugs causing adverse drug reactions (ADRs) leading to admission.

Figure 2.

Organ systems affected by adverse drug reactions (ADR).

The second largest group of ADRs was vomiting and allergic reactions (exanthema) caused by antibiotics. The total number of such admissions was eleven (fig. 2). Six of them were caused by beta-lactam antibiotics, two of them by clarithromycin, one of them by co-trimoxazole and in two cases, it was not possible to identify the antibiotic taken.

Both vaccines and immunosuppressants were equally responsible for 9% of ADRs. Six admissions were caused by five different types of vaccines and the ADRs were heterogeneous (enlarged lymph nodes, fever, muscle pain, headache and others). Among the ADRs caused by immunosuppressants, cyclosporine was considered to cause 50% of them (three cases). These ADRs were different in each case (nephropathy, pancreatitis and vomiting caused by toxic cyclosporine plasma levels).

The most serious ADRs recorded in this study were acute pancreatitis and pulmonary embolism. In two cases, the pancreatitis was assessed as probable and considered to be caused by immunosuppressants (cyclosporine and azathioprine). In one case, the pancreatitis was assessed as definite/certain and was caused by mesalazine. All three cases occurred in patients with inflammatory bowel disease. One case of pulmonary embolism was due to combined oral contraception as were three other cases of deep vein thrombosis. Glucocorticoids probably caused diabetes mellitus decompensation (one case) and a pathological fracture of the tibia (one case). Steroid diabetes was definitely induced by high doses of glucocorticoids in another patient. Two patients were admitted for ADRs after biological therapy (infliximab and imatinib); both assessed as probable according to all the algorithms. No deaths were attributable to an ADR, and none of the patients died during the study period.

The other aim of this study was to assess the causality (possible/probable/definite) of ADRs using three different algorithms. The results are shown in table 2. There were 16 cases assessed as definite/certain and 33 as probable according to all three algorithms. In 15 cases, there was a different result in the final category according to different algorithms. None of them was assessed as possible in all three algorithms.

Table 2. The probability of adverse drug reactions (ADRs) as assessed by the Naranjo algorithm [7], the Liverpool ADR Causality Assessment Tool [8] and the Edwards and Aronson causality assessment method [9]
 NaranjoLiverpoolEdwards
Definite/certain212017
Probable404439
Possible308
Total646464

By means of multivariate analysis (table 3), female sex was identified as a statistically significant risk factor for ADRs (p < 0.05) as was oncological diagnosis which had an OR of 9.8 (p < 0.001). The effect of age was not significant. Univariate analysis showed that patients with identified ADRs were significantly older than those without, 6.3 and 5.7 years, respectively (p < 0.05); this difference was more pronounced in the non-oncological diagnosis group (8.7 versus 5.5, respectively, p < 0.05), but it was non-significant in the oncological patient group (6.1 versus 5.8, NS).

Table 3. Multivariate logistic regression analysis for risk factors for occurrence of adverse drug reactions (ADR) admission
 Odds ratio95% CI for ORp-Value
Female sex1.5839(1.37, 1.80)0.0364
Oncology9.8002(5.77, 16.65)0.0001
Age1.0308(0.99, 1.07)0.1014

Discussion

Adverse drug reactions in children can contribute to substantial morbidity [14]. Our current study shows an incidence of ADR-related admissions of 2.2%, which is similar to a pooled estimate of 2.9% from a recent comprehensive systematic review [10]. The incidence of ADRs in this study was lower than that (3.96%) in a large US study published in 1988 [15]. According to other previously published data, the incidence of ADR-related admissions increases with age from 0.8% in patients aged <18 years to 3.2% in patients aged ≥80 years [16]. A similar age-related increase was also found in this study. Indeed, not all ADRs in our study were assessed as definite (table 2), most of them were assessed as probable and only few of them as possible. In all larger studies, it is usual to involve probable and possible ADRs into the total number of drug-related diseases [11]. All reactions classified as unlikely were excluded from our analysis.

In a previously mentioned study, the top three drugs causing ADRs were phenobarbital, aspirin and phenytoin, all of which are now used in children much less because of safety concerns and better available alternatives [10]. On the contrary, in the most recent study by Gallagher et al. [5], the absolute majority of ADRs were caused by anticancer drugs (44.2%). Substantially fewer ADRs were caused by corticosteroids (41%), NSAIDs (12.4%) and vaccines (8.8%) [4]. Our study shows a somewhat different sequence: anticancer chemotherapy caused 35% of the cases, followed by antibiotics (18%), immunosuppressants and vaccines (both 9%). Glucocorticoids caused 5% of the cases. NSAIDs were involved in two cases only; another two were caused by paracetamol, with all of them classified as antipyretics and accounting for 6%.

The majority of ADRs seen in our study occurred in patients with an oncological disease, a finding similar to that in the study by Gallagher et al. [5]. Oncological patients are very often exposed to medications that cause ADRs, including febrile neutropenia, nausea, vomiting, thrombocytopenia and others, all of which require admission. These ADRs may be unavoidable. Although several studies have evaluated a potential preventative strategy for neutropenia [17], no definite evidence exists regarding the use of granulocyte colony-stimulating factors to prevent such ADRs [18].

Antibiotics account for 18% of all admissions due to an ADR; 55% of them were caused by beta-lactams. Antibiotics are the most frequent drugs prescribed in children worldwide, and beta-lactam antibiotics are the most commonly prescribed group of antibiotics in children. In children treated with beta-lactams, skin rashes are often reported. Such rashes are frequently assumed to be a drug-related allergy, although viral infection is also often considered in the differential diagnosis. It has been suggested that most of these rashes are actually not allergic in origin. However, in clinical practice, the large majority of these children are labelled ‘penicillin-allergic’ without appropriate testing, mostly for fear of a more severe allergic reaction. This diagnosis often persists until adulthood. As a result, they may be denied the optimal antimicrobial treatment. Alleged penicillin allergies are likely to be treated with more toxic, broad-spectered and more expensive antibiotics, with effects on microbial resistance patterns and public economy as a consequence [19, 20].

The diagnosis of a beta-lactam allergy is usually determined using skin tests (to exclude an IgE-mediated allergy), and in the case of a negative skin test, an oral challenge test (OCT) is occasionally performed [19]. One of the limitations of our study is that none of the patients underwent a skin test to confirm beta-lactam allergy. None of them underwent an OCT either. None of the skin ADRs caused by beta-lactams was assessed as definite/certain, so we admit the possibility that some of the skin reactions may not have been due to beta-lactams.

Cyclosporine may be associated with a number of potentially serious ADRs. In this study, three different patients were admitted due to probable cyclosporine toxicity. One of the patients was treated for myelodysplastic syndrome (MDS), the treatment of which is based on immunosuppressive therapy that includes cyclosporine [21]. This patient was admitted because of nausea and vomiting probably caused by toxic plasma levels of cyclosporine. The treatment with cyclosporine had been started 11 days before the admission, and the plasma levels were measured regularly. When the patient was admitted, the levels were higher than recommended (396 μg/l). Cyclosporine treatment was temporarily stopped and then administered in lower doses.

It is well known that cyclosporine has a narrow therapeutic window. In clinical practice, the pharmacokinetic profile can provide an indicator of the appropriate dose to obtain an optimal effect and to try to avoid an ADR. In children, there is a lack of evidence regarding the best schedule that should be adopted. Optimal strategies with the least toxicity remain to be determined [22]. Cyclosporine plasma levels should be measured regularly, and therapeutic drug monitoring in cooperation with clinical pharmacologists might be the best strategy [23].

Steroid-induced hyperglycaemia, insulin resistance, diabetes mellitus, osteoporosis, anxiety, depression, etc., are among the most frequently reported corticosteroid side effects. In the paediatric population, corticosteroids as well remain the key therapy responsible for medication-induced growth impairment [24]. Although they can cause a number of severe ADRs, they are still considered as one of the most potent and consistently effective long-term treatments for a variety of conditions.

The assessment of causality in each ADR-related case was performed using the Naranjo algorithm, the Liverpool ADR Causality Assessment Tool and the Edwards and Aronson causality assessment method. Gallagher et al. compared one of the most widely used causality assessment tools (the Naranjo tool) with the Liverpool ADR Causality Assessment Tool, which is the only comprehensive assessment of ADRs in children. One of their results was that, in the Naranjo tool, almost all cases were categorized as possible or probable. With the Liverpool tool, the range of categorizations was broader with some cases judged as being definite [13].

Our study is the only one to compare the results obtained by the use of all three above-mentioned scoring systems. In 15 cases, there was a different result in the final category according to different algorithms. In eight cases, the ADRs were assessed as possible according to the Edwards and Aronson causality assessment method and as probable according to the Liverpool algorithm. The reason is that when using the Liverpool algorithm, the question ‘What is the probability that the event was due to an underlying disease?’ may be answered ‘high’ and it may as well come out as ‘probable’ [13], while using the Edwards algorithm, there is only a single-choice answer: the syndrome could be explained by concurrent disease or not [11]. It is also probably the reason why using the Edwards algorithm the final score was one level lower in 13 cases of 15 than using either of the other two algorithms (Table 2).

In conclusion, the Edwards algorithm seems to be the strictest of the three tools used, but on the other hand, it is more dependent on the personal view of the investigator than the other two. Nevertheless, the most important conclusion of this comparison of the three available tools is that there is no difference in the total number of ADRs of the three highest levels of probability.

Female sex has been shown to be a risk factor for ADRs. Pharmacoepidemiological studies have shown that there are approximately 30% more ADR reports on women than on men [25]. Other studies show that female patients have a 1.5- to 1.7-fold greater risk of developing an ADR, compared with male patients [26]. The precise reason for this sex difference is unknown. Is the increased risk due to sex differences in pharmacokinetics, in pharmacodynamics or did females receive more medications and higher milligram/kilogram doses than males? Recent studies suggest that all of the above may play a role [27, 28].

In our study, female sex was identified as a statistically significant risk factor for ADRs. Reasons for this increased risk are not entirely clear but include gender-related differences in pharmacokinetic, immunological and hormonal factors as well as differences in the use of medications by women compared with men. Women generally have a lower lean body mass, a reduced hepatic clearance, have differences in activity of cytochrome P450 enzymes and metabolize drugs at different rates compared with men. Other important factors include conjugation, absorption, protein binding and renal elimination, which may all have some gender-based differences. However, how these differences result in an increased risk of ADRs is not clear [26]. Indeed, all these conditions can be present in child age. Studies focusing on these differences in children are almost entirely lacking.

Conclusion

We have demonstrated that ADRs cause a small but substantial proportion of children's hospital admissions, with some of them being serious and potentially avoidable. The majority of the ADR-diagnosed patient affected the hematopoietic and gastrointestinal systems; the drugs most frequently involved were cytotoxic agents and antibiotics. The most important risk factors identified were female sex and oncological disease. The use of different scoring systems does not lead to the differences in the numbers of ADR-diagnosed patient but may result in differences in determination of the level of certainty. Preventing avoidable ADRs requires careful adherence to good prescribing practice.

Acknowledgement

The study was supported by Internal University Grant IGA UPOL LF_2014_008.

Ancillary