The incidence of adverse reactions to drugs among paediatric patients was recently reported to be 15.1 reactions per 1000 children (1). This represents 2% of the admissions to a paediatric hospital, a rate similar to that for adult patients (2,3), of whom only a small proportion (<1%) present hepatic disorders, which may range from mildly elevated transaminases to acute liver failure (2).
Amoxicillin-/clavulanic acid (AC) is a semisynthetic penicillin associated with a β-lactamase inhibitor, which has been used in Spain since 1981. The most common secondary effects of this drug are of a gastrointestinal or cutaneous nature. The first case of hepatotoxicity was described in 1988.
The frequency of hepatotoxicity owing to AC is estimated at approximately 1 case per 78,000 prescriptions, although recent data published report 1.7 cases per 10,000 prescriptions (4), whereas among elderly patients with repeated exposure to the drug (3) the rate may rise to 1 case per 1000 exposures. The frequency may be lower among the paediatric population (<16 years), although little information in this respect is available.
The most frequent type of lesion is cholestatic acute hepatitis (50%), which is assumed to be produced by a mechanism of an immunoallergic nature (5). In general, the pattern of lesion appears to be influenced by age, with hepatocellular injury predominating among younger subjects, and mixed cholestatic/hepatocellular injury among older patients (3,5).
In the case of AC, the clavulanic acid molecule appears to be the main factor responsible for hepatic injury, as amoxicillin alone presents an incidence of only 0.3/10,000 prescriptions (3). The incubation period between the beginning of treatment and the appearance of symptoms or of analytical alterations ranges from 2 to 45 days (6).
Among the few published references to this health problem, a noteworthy report is that of the appearance of a case of cholestasis rapidly evolving to cirrhosis in a child aged 3 years, who developed portal fibrosis following exposure to AC prescribed for otitis media (7).
The aims of the present study were to determine the importance (frequency, forms of presentation, severity, and tendency or otherwise to chronicity) of hepatotoxicity caused by AC among a paediatric population, to characterise episodes identified by type of treatment used, dose, exposure time, pharmacological associations, clinicopathological characteristics (type of injury), or evolution (severity and time to resolution), and to identify possible specific factors in the host.
The subjects included in the present study were all paediatric patients—newborns and older children—with suspected hepatotoxic reaction caused by AC (n = 11), observed from January 1, 2008 to May 8, 2011, as part of a prospective observational study involving 8 Spanish hospitals. The following signs were taken as grounds for suspecting drug-induced hepatotoxicity by AC:
1. ALT > 2N (more than 2 × upper normal values for age)
2. Conjugated bilirubin > 2N
3. Combined elevation of aspartate aminotransferase, bilirubin, and γ-glutaryl transpeptidase, provided that at least 1 exceeds 2N
The existence of any one of the above was considered sufficient for hepatotoxic reaction to be suspected.
We excluded from the study patients who, despite meeting causality criteria, presented a concomitant hepatic disease (viral, alcoholic, metabolic, fat deposit, or autoimmune) or any other nonhepatic disease that might alter the hepatic profile (thyroid disease, cardiac insufficiency, HIV infection, obesity, among others).
For each patient included in the study, a structured codified data collection protocol was completed, including the following items: demographic variables for the patient, characteristics of the treatment to which the reaction is related, concomitant medication, and evolution of the episode; laboratory variables during the reaction, determined at 6-month intervals from the first examination until eventual resolution of the case (Supplemental Table 1, http://links.lww.com/MPG/A157) (8).
When a possible case was detected, the causality criteria were examined, applying the Council for International Organizations of Medical Sciences (CIOMS) scale (9). According to the total score obtained, each case was included in 1 of the following suspected categories: definite or highly probable, probable, possible, improbable, or excluded. With respect to the type of hepatic injury, we followed the criteria published by the 1990 International Consensus Conference (10), defining it as hepatocellular, cholestatic, or mixed. The mechanism of the lesion was defined, by classifying it as either intrinsic or idiosyncratic and the latter as immunoallergic or metabolic. We also evaluated the time to complete resolution, following cessation of treatment, for each case, or chronification.
Since the pharmacovigilance protocol was introduced in January 2008, 24 cases of possible hepatotoxicity have been reported, of which 11 (45.8%) correspond to AC. Of these 11 patients, 9 were boys and 2 were girls, with a mean age of 5.25 ± 4.75 years (1–11 years). In every case, medication was administered orally. The dose ranged from 25 to 50 mg · kg−1 · day−1 (of amoxicillin) with a mean value of 42 mg · kg−1 · day−1. The concentration of clavulanic acid was provided in an amoxicillin proportion of 8:1 in 7 cases and of 4:1 (500 mg amoxicillin/125 mg clavulanic acid) in 4 cases.
Table 1 shows the pathology suggestive of AC hepatotoxicity, treatment duration, the time of diagnosis of suspected hepatotoxicity, clinical signs warranting the analysis of liver enzymes, laboratory parameters examined, and the time elapsed until normalisation of transaminase levels, for each patient. 54.5% of the boys presented digestive symptoms (abdominal pain and vomiting), 45% had anorexia and asthenia, and 18% had diarrhoea. Only 1 of the patients studied presented jaundice. Seven of the patients (64%) required admittance to hospital, with a mean stay of 7.9 days (5–12 days). The reason for hospital admittance in all 7 cases was the presence of incoercible vomiting. All of the patients received occasional concomitant medication, mostly antipyretics: 67% paracetamol and 44% ibuprofen.
An abdominal ultrasound examination was performed on all of the patients, and only in 1 case were pathologic signs recorded, with evidence of hepatosplenomegaly and intra-abdominal free fluid. Nevertheless, the clinical signs were self-limiting, and analytical values returned to normal after 4 to 36 weeks, with a mean duration of 14 weeks. In 4 cases, liver enzymes returned to normal values in <8 weeks. Only 1 patient maintained elevated transaminase levels for a period of 36 weeks. None of the children were biopsied.
Taking into account the clinical and analytical findings, cases 2 and 8 were classified as mixed (18%), case 4 as cholestatic (9%), and cases 1, 3, 5, 6, 7, 9, and 10 as hepatocellular (64%). No case presented clinical or analytical evidence suggesting acute liver failure. In all cases, viral serology was negative in at least 2 determinations, and was negative in all the tests included in the study protocol.
Finally, according to the CIOMS scale, 3 of the cases were characterised as highly probable for hepatotoxicity, 3 as probable, and 5 as possible.
Hepatotoxicity is an important challenge in pharmacotherapy today, and continues to be the main cause of drugs being withdrawn from the market. Nevertheless, it may be considered a relatively “orphan” disease, because of its rarity and the difficulty of establishing causality, owing to the absence of a criterion standard for diagnosis. In consequence, it is necessary to aggregate cases, with reliable information, to shed further light on its epidemiology and pathogenesis (2,11,12).
In this respect, we believe the information made available to date, derived mainly from the communication of suspected cases of hepatotoxicity to regulatory agencies using the system for the spontaneous notification of adverse reactions (the yellow card system) or the publication of isolated cases or series in medical journals, represents probably no more than the tip of the iceberg (11,13). This circumstance, which we have detected in our analysis, was highlighted in a prospective population study carried out in France, in which the incidence of toxic hepatitis was found to be 16 times greater than that recorded by the yellow card system (14). Therefore, it is a matter of urgency that action be taken to improve the identification of adverse hepatic reactions and to raise the quality of diagnosis.
It is no simple matter to estimate the prevalence of AC toxicity. According to the Spanish Institute of Statistics (INE) (14), some 1,600,000 AC treatment programmes are prescribed for children each year (the child population is approximately 6,800,000 according to the same source). On the basis of the children ages younger than 15 years treated at the hospitals participating in the present study, a prevalence of approximately 1 in 150,000 prescriptions could be calculated; however, we believe these data could be highly distorted, for 2 main reasons:
1. In the present study, several “target” clinical services were examined, to compare the number of reactions identified with the real prevalence; however, identifying such reactions is no easy matter, even for the members of the multicentre team, and thus imprecise reporting is almost inevitable. The number of hepatotoxic reactions reported each year and included in the official record is now approximately 25% higher than during the first year of this study (ie, the real prevalence was previously underestimated).
2. These target services are provided as part of hospital attention. Therefore, children treated at primary health care centres, rather than in hospital, are not recorded in the present system. For this reason, in a second phase of the study, we intend to include primary health care.
For these reasons, we believe that the real prevalence may be considerably higher than the theoretical figure. Accordingly, the present record should be actively maintained, together with a large-scale campaign to raise awareness of the problem to obtain accurate data on prevalence.
During the study period, we observed only 1 hepatotoxic reaction caused by amoxicillin alone (unpublished data) although, in principle, the latter should have been much more frequently prescribed than AC, for which indications are more restricted. This fact leads us to believe that although its lesion-provoking mechanism remains unclear, clavulanic acid is responsible for a significant proportion of liver injuries, as has been reported (5).
Most of the indications for AC in our study were related to respiratory diseases. With the exception of the cholestasis attributable to bacterial pneumonia, no associations have been reported between respiratory and hepatic diseases. This fact limits the confounding factors in the interaction between AC and liver damage. For all of these reasons, we believe, as reported in the literature, that the AC combination is more likely to be associated with acute liver damage than the use of amoxicillin alone (3,15).
It has been reported that half of the drugs given to patients lack scientific justification for their formulation and prescription. The use of AC was only fully justified for 18% of our patients, with the remaining 82% presenting doubtful indications; thus, amoxicillin alone could have been prescribed instead. This circumstance leads us to believe that the physician can take significant action to prevent drug-induced hepatic toxicity, by avoiding the prescription of unnecessary drugs (15–17).
The most characteristic clinical manifestations reported of hepatotoxicity owing to AC consist of asthenia, jaundice, and general malaise. The presence of fever, cutaneous exanthema, and peripheral blood eosinophilia is believed to be related to the immunoallergic mechanism that produces liver injury (18). In our study, the most common clinical sign was that of vomiting. Asthenia and general malaise tend to be data obtained from the patients’ own description of their symptoms, and could be difficult to determine in paediatric practice, in which the subjects may not report such clinical signs correctly.
All of the cases reported in our study evolved favourably; in no case was a biopsy required, nor did the condition progress to a severe state, contrary to data published regarding studies of adults (6,19,20). In our study, 73% of the patients required hospital admission. According to published reports, resolution of patients’ analytical parameters ranges from 4 to 16 weeks. In our study, 10 of the 11 patients achieved normal values for hepatic enzymes within 8 weeks. The other patient was not discharged from the paediatric hepatology ward until 36 weeks after the condition was first recognised.
Unlike previous reports, we were unable to show that previous liver disease, drug dose, comorbidity, or form of drug administration influenced paediatric hepatotoxicity as a result of AC. With respect to the form of administration, it is logical to believe that the oral administration of this drug is much more frequent than the intravenous route, and therefore the probability of a toxic reaction appearing must also be much greater. This leads us to reject, a priori, the theory that the oral route of administration is positively related to hepatotoxicity by AC. Nevertheless, owing to the small sample size in this study, the possible effect of this factor cannot be totally discounted.
The doses given in our cases were the normal ones in paediatric clinical practice. Treatment with AC provides a therapeutic index and margin of safety equivalent to 40 to 100 mg · kg−1 · day−1 of amoxicillin, according to the pathology. The mean dose of amoxicillin used was 42 mg · kg−1 · day−1. Clavulanic acid was given in a proportion of 250 mg amoxicillin/62.5 mg of clavulanic acid in 7 of the 11 cases. To date, there is no information on the possible importance of the proportion of clavulanic acid medication, with respect to liver disease (19).
The duration of treatment may be related to the appearance of hepatotoxicity by AC. None of the patients in the present study (Table 1) received AC for a longer period than that recommended in guidelines to clinical practice and/or published protocols for respiratory diseases, as presented in our cases and which consider an acceptable treatment period to be 7 to 21 days.
In our study, 3 cases were detected in which the treatment duration was abnormally short, at 5, 3, and 5 days, respectively. The sample was small for reliable conclusions to be drawn regarding treatment duration. It is noteworthy that in the only case in which jaundice was presented, together with evident clinical manifestations, liver injury developed after only 3 days’ treatment. In other cases in which clinical signs were less apparent, treatment periods of 7 to 10 days were completed, and elevated transaminases were detected some time later.
Hepatotoxicity as a result of AC may be detected up to 6 weeks after medication is given (18). In our series, all of the cases were detected within 3 to 29 days. Therefore, later-occurring elevated transaminases may be underestimated at present, with physicians not considering this diagnostic possibility.
During childhood, comorbidity is much less frequent than in adulthood, and so among the paediatric population this aspect does not seem to be related to the frequency of liver injury. In our study, none of the patients presented comorbidity, contrary to the data published for adults.
Distribution of the patients by ages did not reveal any characteristic pattern. We believe further studies are required to determine whether there exists any particular susceptibility to hepatotoxicity. With respect to gender distribution, of the 11 cases studied, 9 were boys and 2 were girls. This is in line with published data, which also reflects a proportion of 4:1 in favour of boys (21).
The cases observed in the present study presented mixed or hepatocellular patterns, which agrees with the data published concerning young populations, in contrast to the predominance of cholestatic injuries among elderly adults (6,22).
We believe the main limitation of the method described in the present study is that it does not detect hepatic reactions owing to AC that are clinically insignificant in nontarget services and in outpatient attention. On the contrary, it does have the potential to detect all of the hepatotoxic reactions of a moderate to severe nature occurring in the hospital environment and all of those generated in target services, whatever their degree of severity.
In conclusion, further investigation of hepatotoxicity is needed, concerning issues such as genetic markers of susceptibility, markers of specificity, and simplification of the micromethod determination of transaminases (taking samples from a fingertip puncture); however, in the meantime, the introduction of data registration systems and support to those being developed will provide a good means of improving epidemiological knowledge about drug-induced liver injury (23). The present study demonstrates the relatively frequent occurrence of paediatric hepatotoxicity by AC; therefore, paediatricians should include transaminase analysis in routine analytical examinations and perform in-depth studies when normal levels are exceeded.