Definition of AE and study design

We defined an AE as the appearance of any unwanted adverse effect temporarily associated with the use of an antimicrobial compound at usual doses, whether for diagnostic or prophylactic use¹⁰. A retrospective design was applied using the reports of notifications (by passive methodology) of AE made to the National Pharmacovigilance System from years 2014 to 2017 in our institution. Events involving antimicrobial compounds of oral or parenteral use among hospitalized or ambulatory patients were included.

Case analysis

For each reported AE, we collected patient’s demographic data, comorbidities, immunosuppressive drugs use, allergic symptoms, previous adverse reactions, hospitalization unit where the event occurred, system or organ involved, antimicrobial compound suspected, route of administration and time elapsed until appearance. Management of the AE and its outcome was also recorded. Causal analysis was made using the Karch-Lasagna method and preventability by a previous described method^10,11. The work only includes AE where the causal analysis assigned a score in the definitive (≥ 8 points), probable (6-7 points) or possible (4-5 points) categories. Events were also categorized as mild (digestive disorders, headache, tiredness, myalgia, compromised general condition, sleep disturbances), moderate (skin lesions, visual disturbances, tremors, urination problems, noticeable variation on mind or mental functions) or serious/lethal (life-threatening or fatal, AE leading to severe disability or invalidity or causing or prolonging hospitalization and birth defects).

Ethical aspects

This work was approved by the Comité de Ética Científico, Servicio de Salud Valdivia, Chile.

AE reported and patient characteristics

Between years 2014 and 2017, 84 AE associated with antimicrobials were recorded. In 14 cases (16.7%) an alternative cause for another drug or the underlying disease was identified and in 10 (11.9%) there were incomplete or absent data, leaving 60 final events in which clinical information could be obtained (71.4%). These 60 AE involved 56 patients; one patient had 3 events in the same hospitalization, 2 patients had a double event and the remaining 53 only one (94.6%). Patients were associated with a wide age range (2 months to 96 years) with a balanced distribution by gender and different comorbidities (mainly heart disease, diabetes mellitus, chronic kidney disease and leukemia/lymphoma), but a third did not present any of them (Table 1). Additionally, another third of patients presented immunosuppression either by drugs (5 patients, 25% of the group) or due to an underlying pathology, and 16.1% had a history of previous AEs (Table 1). Near 10% has an allergic history. Patients affected by these AE were hospitalized in Units of different complexity, including the Emergency Unit and 2 cases were reported as outpatients (Table 1).

Type of adverse reactions and eosinophilia

The most frequent AE reported was skin involvement with rash in approximately half of the cases and most of the time of generalized type (Table 2). In 3 of the 34 cases, the rash was associated with mucosal involvement or bullous skin lesions, and in 2 of these 3 cases, the condition was classified as Stevens-Johnson syndrome (SJS) (3.3% of all events) (Table 2). The second system affected was the liver with hepatitis, cholestasis, or mixed type. CNS followed in frequency mainly with seizures. The series also includes hematological disarrangements in all series and two anaphylactic events. Renal, respiratory, or gastrointestinal involvement was occasionally observed (Table 2). In total, 7 events were associated with eosinophilia (> 500/ μL), 5 of them in skin lesions (16.7% of the cutaneous group), one in an anaphylactic shock and the other with anemia.

Antimicrobials involved

The compounds mostly associated with adverse reactions were beta-lactams in about half of them, followed by anti-tuberculosis drugs, then glycopeptides (vancomycin) and quinolones. Other compounds were less frequently reported and in some cases by association of 2 or more antibiotics (excluding anti-tuberculosis drugs, Table 3).

Causal analysis

This analysis revealed a definitive association in 5 events (8.3%), probable in 42 (70%) and possible in the remaining 13 cases (21.7%). Table 4 presents the causal analysis for skin lesions, showing the predominance of beta-lactams in the definitive and probable categories. Skin lesions were also associated with glycopeptides, quinolones, isoniazid, doxycycline, and different combinations of antimicrobials (Table 4). The 2 cases with SJS were related to ceftriaxone and the combination of vancomycin with ceftazidime (both possible). Hepatobiliary disorders were mostly associated with anti-tuberculosis drugs (Table 5). Hepatitis events were associated with isoniazid and/or pyrazinamide and cholestasis with rifampicin. Mixed disorders (cholestasis with hepatitis) were assigned to pyrazinamide or drug combinations (Table 5). No patients had history of liver damage and in any case doses were higher than those recommended per weight (data not shown). Age was not significantly higher in the group with this complication compared to the rest of the patients (Mann-Whitney non-parametric test, data not shown). For AE involving the CNS (Table 6), compounds associated with seizures, the most frequent manifestation in this group, were carbapenems, especially ertapenem and in one case metronidazole. Ciprofloxacin was associated with an agitation-anxiety event.

AE compromised all hematological series. Thrombocytopenia (nadir 3.000-53.000/μL) was associated with linezolid (1 probable case), ceftriaxone (possible in 1 case) and vancomycin with ciprofloxacin (possible in 1 case). Neutropenia events (nadir 400-900/μL) also observed in 3 cases were associated with vancomycin, cloxacillin, and cefadroxil, all 3 as probable. The only reported case with anemia was assigned to linezolid (probable). Renal events with streptomycin-associated insufficiency was also observed in one probable case, anaphylactic shock in 2 cases (cotrimoxazole and ceftazidime, both as probable), one gastrointestinal event with nausea and vomiting due to metronidazole (definitive category), and an event of respiratory depression due to colistin (possible).

Adverse effects onset

Skin lesions presented within a median of 4 days (interquartile range [IQR] 1-6 days; range 0-20 days) with 41.2% in the first 72 hours. The 2 SJS events occurred at 5 days. The median onset for skin events was slightly higher than the elapsed time for neurological manifestations (median 2.5 days, IQR 1-5 days, range 1-12 days) although these differences were not significant (Mann-Whitney non-parametric test). On the other hand, hepatobiliary AE presented with a median of 10.5 days (Figure 1, IQR 8-17 days, range 2 to 28 days), a figure significantly higher than the onset time for skin or neurological lesions (p = 0.025 by Mann Whitney test). For hematological conditions, the three thrombocytopenia events appeared between 1 and 9 days after antimicrobial initiation with their nadir 4-6 days later (Figure 2). Neutropenia presented between 5 and 65 days after initiation.

Figure 1 Box plot showing onset in days for cutaneous, central nervous system (CNS) and hepatobiliary adverse events. Median values are denoted by a square inside each box. Upper and lower borders of each box indicate the 25-75% interquartil range. Horizontal bars outside boxes show non-extreme maximal and minimal values, and open circles, extreme values. Onset of hepatobiliary adverse events were significantly later than values for skin and neurological events (see text). 

Figure 2 Platelet counts in 3 events of thrombocytopenia associates to antimicrobial compounds. Values were adjusted by the least square method. Zero indicates the day antibiotic was suspended. 

The only anemia event was detected 8 days after starting linezolid. AE from other systems were observed at 24 hours in a gastrointestinal reaction, at 29 days for streptomycin and at 24 hours and 4 days in anaphylactic shock events.

AE duration

The longest events corresponded to hepatobiliary reactions that had a median duration of 24.5 days (range 6 to 37), followed by thrombocytopenia (median of 12 days, range 10 to 13), and then cutaneous events with a median of 6 days (range 0-39). In turn, the median for 3 neutropenia events was 4 days (range 4 to 18) and for the 2 anaphylactic events was 5.5 days (3 and 8 days). Shortest symptoms were associated with CNS with a median of 0 days (range 0-2 days). The isolated event of anemia lasted 9 days, that of renal failure 30 days, the gastrointestinal event 1 day and the respiratory depression 12 days.

AE management, severity, outcome and therapy change

AEs were managed in different ways in 59 of the 60 reported events (98.3%), mostly with suspension of involved drugs (n = 54 events; 91.5% of the subgroup; 83,3% of the whole group). Additionally, 27 events received corticosteroids (45.8% of the treated subgroup) and 29 antihistamines (49.2% of the subgroup). On the other hand, anticonvulsant therapy was indicated in the 5 events with seizures and connection to mechanical ventilation in the event of respiratory depression associated with colistin. Anemia linked to linezolid required transfusion as one of the events of thrombocytopenia. Corticosteroid therapy was used in patients with skin lesions but also in events of thrombocytopenia and anaphylactic shock.

AE were categorized as mild (n = 2, 3.3%), moderate in most cases (n = 45, 75%) and severe in 21.7% (n = 13). In all but one case there was a recovery from the adverse reaction (98.3%). However, one event with SJS had a fatal outcome (1.7% of events). Overall, AE forced to change the antimicrobial regimen in two thirds of cases (n = 39; 65%).

Preventability analysis

In 15 AEs (25% of the total), their appearance could have potentially been prevented. Six cases of rash could potentially have been avoided: one with a previous history of allergy to the same family of antimicrobials (beta-lactams), two cases without bacterial infection that justified their use, one with beta-lactams therapeutic duplication and one by vancomycin where the low prevalence in Chile of community-acquired infections by methicillin-resistant Staphylococcus aureus discourage its use. Finally, an event associated with penicillin in combination with other drugs for a possible streptococcal pharyngitis had already received adequate treatment. For 4 cases with seizures (80% of the 5 reported cases) this complication could have been avoided after considering preexistence brain lesions or strokes in 2 cases, kidney failure without dose adjustment and redundant anaerobic coverage (1 each). Furthermore, the event of agitation/anxiety associated with quinolones could have been avoided by choosing another option for the patient’s folliculitis. All of these conditions represent potential avoidable events linked to medical prescription.

Two anaphylactic shock events observed in this series, could also have been avoided by choosing another option to treat acne in one case (cotrimoxazole) or simply avoid its use in a case of imaging-ruled out pneumonia (ceftazidime). A lethal SJS case associated with ceftazidime and vancomycin, the use of the latter in community-acquired pneumonia had no epidemiological justification in Chile. Finally, anemia secondary to linezolid in a context of nosocomial pneumonia could have been avoided by choosing to keep tigecycline in use.