Do antiepileptic drugs increase the risk of infectious diseases? A meta‐analysis of placebo‐controlled studies

Abstract

Aims

Experimental studies show that some antiepileptic drugs (AEDs) may modify natural immune defences, thus influencing the risk of developing infectious diseases. The aim of this meta‐analysis was to explore whether AEDs as a class of drugs or singularly may increase risk of infectious diseases.

Methods

A meta‐analysis of all randomized, double‐blind, placebo‐controlled trials (RCTs) investigating any AED in any condition was performed. All terms that could be coded in the System Organ Classes (SOCs) of infections and infestations using the Medical Dictionary for Regulatory Activities were recorded. Additional subanalyses were performed also pooling together AEDs sharing similar mechanisms of action.

Results

Two hundreds and sixty‐nine double‐blind, placebo‐controlled studies were identified and, among them, 127 RCTs with 16 AEDs (brivaracetam, gabapentin, lacosamide, levetiracetam, lamotrigine, oxcarbazepine, perampanel, pregabalin, phenytoin, remacemide, retigabine, rufinamide, tiagabine, topiramate, valproate, zonisamide) reported at least one of 19 symptoms or diseases that could be included in the Medical Dictionary for Regulatory Activities SOC term infections and infestations. These terms were singularly recorded and then pooled together in the SOC term infection and infestation. Topiramate was significantly associated with an increased risk of infection (risk difference = 0.04; 95% confidence interval = 0.01/0.06), while oxcarbazepine was significantly associated with a lower risk (–0.005; –0.09/–0.01). Risk difference of all studies with all AEDs showed a slight, but significantly increased risk of infection (0.01; 0.00/0.002). Levetiracetam and brivaracetam RCTs, when pooled together, were associated with a significantly increased risk of infection (0.03; 0.01/0.05).

Conclusions

Some AEDs are associated with a mild increased risk of infection.

Table of Links

LIGANDS
Brivaracetam	Levetiracetam	Pregabalin	Topiramate
Gabapentin	Oxcarbazepine	Retigabine	Valproate
Lacosamide	Phenytoin	Rufinamide	Zonisamide
Lamotrigine	Perampanel	Tiagabine

This Table lists key ligands in this article that are hyperlinked to corresponding entries in http: //www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY 1.

Introduction

Antiepileptic drugs (AEDs) might exert effects different from those primarily responsible for their anticonvulsant properties, which may involve immune system.

In vitro studies have demonstrated that several AEDs decrease several cytokines and tumour necrosis factor (TNF)‐α and have anti‐inflammatory properties 2. These effects may be related with the main mechanism of action of a drug or due to other effects. For example, binding of synaptic vescicle protein 2A (SV2A) constitutes the main antiepileptic mechanism of levetiracetam 3, 4, and it has been observed that this protein is also expressed in CD8+ T lymphocytes, whose function is affected by this drug 5. An inhibitory effect on the immune system can also be caused by several AEDs through modulation of ion channels and γ‐aminobutyric acid receptors, which are expressed by immune cells 2. In other cases, immune changes may be related to mechanisms different from the antiepileptic ones. For example, topiramate induces changes in proliferative activity of splenocytes 6 as well as anti‐inflammatory effects through modulation of lipid peroxidation and glutathione metabolism in the cell system 7.

These effects may increase the risk of developing infectious diseases and some clinical data suggest a higher frequency of upper respiratory infectious diseases in subjects under treatment with specific drugs 8, 9.

Meta‐analyses of randomized controlled studies have been proven to be a powerful tool in identifying the relationships between a certain treatment and the associated risk of infection in patients with other diseases 10, 11.

With the aim to explore whether chronic AED treatment increases the risk of infectious diseases, a meta‐analysis has been performed of all infectious diseases reported as adverse events (AEs) during the double‐blind phase of all placebo‐controlled randomized clinical studies (RCTs) with all AEDs in all indications in which they were studied.

Methods

We performed a systematic review of placebo‐controlled double‐blind RCTs assessing AEDs in any condition. Studies were identified through Medline (PubMed interface) and EMBASE up to December 2015. The study was done according to the preferred reporting items for systematic reviews and meta‐analyses guidelines 12. See PRISMA checklist (Supporting information Data S1).

Eligibility criteria

We selected all randomized, double‐blind, placebo‐controlled trials investigating any AED in any condition, with a parallel or cross‐over design and a duration of double‐blind phase ≥4 weeks and which reported infectious diseases among AEs. For details on search strategy and inclusion/exclusion criteria, see S2. Cochrane Collaboration's tool for assessing risk of bias 13 was used to ascertain the validity of eligible RCTs.

Data abstraction

All identified RCTs were divided in five groups. For each group of trials, two of the authors (S.G., F.S.G., F.G., V.F. and G.Z.) assessed eligibility and extracted data. AEs were usually reported in a table or in a section of results. In a few cases, both all causes AEs and treatment‐related AEs were reported. In such cases only all causes AEs were considered.

Investigators recorded, from the AEs section, all terms that could be coded in the System Organ Class (SOC) named infections and infestations of the Medical Dictionary for Regulatory Activities (MedDRA). In the hierarchy of MedDRA, there are five levels of terms arranged from very specific to very general. Preferred terms are descriptors of a single medical concept (for example, in our case pharingitis or influenza) which are grouped together in high level terms and then in higher terms. The highest term infections and infestations is the SOC that includes all infectious diseases.

From each study, we extracted the number of patients randomized to active drug or placebo and the number of patients in each group with AEs that could be coded as infections or infestations. Terms that could not be coded in this SOC were excluded. For each of the selected terms, only patients from those studies in which this term was reported were included. In no case additional unpublished information from companies or authors of studies was obtained.

Data analysis

The meta‐analysis was conducted using the software RevMan version 5.3.5 14. The risk differences (RDs) with 95% confidence intervals (CIs) between active drug and placebo were calculated for all studies included. Heterogeneity between studies was assessed by I² and Cochrane Q test. Because of the heterogeneity among studies, data were analysed using a random effects model.

Results

From all the articles identified through PubMED and EMBASE (for the process of selection, see S3), 269 double‐blind, placebo‐controlled studies exploring the effect of 20 AEDs in any condition for >4 weeks, were identified. For 16 AEDs (brivaracetam, gabapentin, lacosamide, lamotrigine, levetiracetam, oxcarbazepine, perampanel, phenytoin, pregabalin, remacemide, retigabine, rufinamide, tiagabine, topiramate, valproate and zonisamide), a total of 127 studies reported symptoms or diseases (preferred terms) that could be included in the MedDra SOC term infections and infestations and were selected (main characteristics of selected studies are reported in S4).

Nineteen different terms were identified. Each of these preferred terms was singularly recorded (some preferred terms considered identical were pooled). All identified preferred terms are reported in Table 1.

Table 1.

List of all identified MedDRA preferred terms which have been coded in the System Organ Class infection and infestation

Bronchitis, cellulitis, conjunctivitis, cold, fever, flu‐like syndrome, flu syndrome, influenza, nasopharingitis, othitis media, pharyngitis, pyrexia, respiratory infection, rhinitis, upper respiratory infection, urinary tract infection, sinusitis, viral infection, viral illness

All preferred terms in each RCT were subsequently summed to produce only one measure (SOC term infection or infestation). The majority of trials had a low risk of bias (see S5).

Meta‐analysis of the SOC term infections and infestations

In Figure 1A, RD (95% CI) of the SOC term infection and infestation is reported aggregated by AED and for all AEDs summed. Full forest plot is given in S6 and additional data are reported in S8, Table IS.

Figure 1.

Risk difference (95% confidence interval) of Medical Dictionary for Regulatory Activities System Organ Class term infections and infestations. (A) The risk difference aggregated by drug (full forest plot is given as supplementary material S6). (B) The results of a subanalysis performed in three groups of antiepileptic drugs (AEDs) arbitrarily subdivided according to similar mechanisms of actions. In particular, these groups included: (i) brivaracetam and levetiracetam; (ii) gabapentin and pregabalin; and (iii) AEDs affecting ion channels (lacosamide, lamotrigine, oxcarbazepine, phenytoin, remacemide, retigabine, rufinamide). A low but significantly increased risk of infection is shown for the sum of all AEDs, for topiramate, and for brivaracetam and levetiracetam pooled. A significantly lower risk is shown for oxcarbazepine. For some AEDs such as phenytoin, remacemide, and retigabine only a nonsignificant trend for an increased risk is observed. For these drugs, note the wide confidence intervals, which indicate that only few subjects have been evaluated

It can be observed that topiramate is significantly associated with an increased risk of infection (RD, 95% CI: 0.004, 0.01/0.06) while oxcarbazepine was significantly associated with a lower risk (RD, 95% CI: –0.005, –0.09/–0.01). All other AEDs were not significantly associated with a different risk although in selected cases some trends towards an increased risk of infection could be observed (Figure 1). Overall, percentages of patients with infections were similar among those with the active treatment and among placebo‐treated patients (15% vs. 14%). However, RD of all studies showed a slight, but significantly increased risk of infection between patients treated with the active drug (RD, 95% CI: 0.001, 0.00/0.02).

Meta‐analyses of all preferred terms identified in RCTs

RD (95% CI) of all preferred terms identified in RCTs, are reported in S7. Briefly, the preferred term upper respiratory infection was significantly more frequent with topiramate and with valproate (RD, 95% CI: 0.04, 0.02/0.06 and 0.19, 0.03/0.35, respectively). Fever was associated with gabapentin (0.07, 0.01/0.13). It should be noted that both for valproate and gabapentin there was only one RCT generating these results.

Subanalyses

In a first subanalysis, some AEDs were grouped according to the mechanism of action. In particular, brivaracetam and levetiracetam, gabapentin and pregabalin, and all drugs affecting ion channels (lacosamide, lamotrigine, oxcarbazepine, phenytoin, remacemide, retigabine, rufinamide) were grouped. Results are reported in Figure 1B and S8, table IIS. In the group of patients treated with levetiracetam or brivaracetam, percentages of patients with infections were significantly higher in the active drug arm than in the placebo arm (12% and 10%, respectively; RD, 95% CI: 0.03, 0.01/ 0.05). Summing data from all RCTs performed with gabapentin or pregabalin (0.00, 0.001/0.002) and all RCTs exploring drugs whose main mechanism of action is on ion channels (either sodium or potassium channels, or both; 0.00, –0.01/0.02) gave identical risk of infections between patients treated with active drug and those with placebo.

A second subanalysis concerned only preferred terms, which we arbitrarily considered as expression of a disease affecting upper respiratory infection (URI). Results of this analysis failed to show new information (S8, Table IIIS). RD (95% CI) of 118 RCTs was 0.01 (0.00/0.02). Also in this case, topiramate was the only AED significantly associated with higher percentages of patients with URI (0.04, 0.01/0.07). A nonsignificant trend for an increased risk of URI was observed for levetiracetam.

Discussion

We found a small, but significant difference between placebo‐ and AEDs‐treated patients in percentages of patients who had some infectious diseases, which may be included in the MedDRA SOC term infection and infestation. It is unlikely that the increased risk of infections is a class effect of AEDs. Most probably, this effect has several different mechanisms and may be typical of selected AEDs. It should be acknowledged that in the MedDRA SOC term infection and infestation are also included unspecific terms that may have weakened results of this analysis but that had to be included to avoid unwarranted selections.

Topiramate was associated with a significantly increased risk of infection, while several other AEDs showed only nonsignificant trends or a risk comparable to that found in placebo‐treated patients.

A subanalysis in which AEDs with similar mechanisms of actions were grouped, provided additional interesting data. While gabapentin and pregabalin, when analysed either singularly or together, were not associated with an increased risk, levetiracetam and brivaracetam were associated with a significantly increased risk when grouped together, although this risk did not reach significance when each of them were singularly assessed. These findings suggest that this effect may be in some way associated with the mechanism of action that is almost identical for the latter two AEDs 15.

Different results emerge from grouping together data on drugs whose main mechanism of action is the modulation of ion channels (sodium or potassium channels, or both). In this case, no significant trends for an increased risk were observed for each of them when singularly considered, with the exception of oxcarbazepine, which was significantly associated with a decreased infection risk. However, it is worth noting that for oxcarbazepine, only three studies could be included in the analysis. Even when grouping these drugs, no differences between the active drug and placebo were observed. Thus, although we cannot exclude that some of these AEDs might facilitate infections, this effect does not seem to be associated with the main mechanism responsible for their anticonvulsant properties. For this group of AEDs we should acknowledge that mechanisms of actions of these drugs are very heterogeneous.

Concerning the reasons why some AEDs are associated with an increased risk of infections, we can hypothesize that anti‐inflammatory properties found for several AEDs may be responsible for this effect.

While excessive and/or dysregulated inflammation has been described as a mechanism which can worsen infections 16, 17, 18, other lines of research suggest that inflammation is associated with the induction of protective host responses 19, 20.

Topiramate bears antihyperalgesic and anti‐inflammatory properties 7 through several potential mechanisms, which include an effect on γ‐aminobutyric acid‐A and opioid receptors in the central nervous system, as well as peripheral mechanisms 21. Experimental studies have also shown that topiramate decreases the production of TNF 22: this might facilitate infections, as it is known that TNF‐α blockers increase the risk of opportunistic infections in patients with rheumatoid arthritis 23.

Anti‐inflammatory effects have also been observed in experimental studies with other AEDs such as valproate 24, primidone, carbamazepine, oxcarbazepine, lamotrigine and phenobarbital, which decrease the production of some proinflammatory cytokines 2.

The case of levetiracetam and brivaracetam is particular. It has been observed that levetiracetam has protective effects in animal models of inflammation, modulating the production of plasma TNF‐α and antioxidant capacity 25. Anti‐inflammatory properties of this drug have even been claimed to contribute to its anticonconvulsant effect 26. Interestingly, the main mechanism of action of both levetiracetam and brivaracetam is related with their binding with SV2A protein in central nervous system neurons 4, 15. However, SV2A protein is also expressed in other cell types, including human CD8+T lymphocytes 5 and experimental studies have shown that levetiracetam has inhibitory effects on the function of these lymphocyte subtypes, which might be mediated by interaction with SV2A protein 5.

Even though previous clinical observations suggested an increased risk of upper respiratory infections with levetiracetam 8, 9, our subanalysis failed to show a specific effect on these infections, thus suggesting that this facilitating effect concerns all infectious disease and is not specific of any particular infectious disease.

Finally, the protective effect which has been observed for oxcarbazepine has not a direct explanation. It is known that mechanisms of anti‐inflammatory effects of oxcarbazepine are at least in part different from those described for topiramate or levetiracetam/brivaracetam 2. Carbamazepine, which has similar effects to oxcarbazepine has particular anti‐inflammatory effects 26. Since excessive inflammation may in some cases worsen infections 16, 17, it is possible to speculate that oxcarbazepine has a modulatory effect on inflammation, which may be protective against infections. Alternatively, this protective effect might be the consequence of a still unknown mechanism.

We are aware that we made a meta‐analysis of the observed infectious diseases based on a classification of events reported in clinical studies, which were not designed to specifically target frequency of infectious disease, and in which validated measures for the assessments of these events had not been adopted. For this reason, we did not assess risk of bias since both companies and investigators considered this outcome as an AE probably nondrug‐related (by definition an AE is any untoward medical occurrence associated with the use of a drug in humans, whether or not considered drug related). We could predict that in the case of a bias, this should have influenced results in the opposite direction. One possible alternative explanation for these findings might be that patients treated with the active drug might complain more frequently of AEs, and that patients or investigators might misdiagnose as infectious diseases, symptoms that are caused by other effects of drugs. However, the latter potential explanation is unlikely to be true, as AEDs with several AEs during double‐blind studies, such as pregabalin 27 or oxcarbazepine 28, were not associated with an increased frequency of infections, while this was the case for AEDs with less frequent AEs, such as levetiracetam 29 and brivaracetam 30. We cannot exclude that in those RCTs in which the experimental drug had been added to a previous therapy, a pharmacodynamic interaction might have modulated the effect of the new drug on the immune system.

Finally, this increased risk for infectious diseases seems to be small since the overall difference between active drug treated patients and placebo treated patients is of the order of 1% while it is of the order of 4% for topiramate and 3% for brivaracetam and levetiracetam summed. This finding suggests that this effect is negligible in the vast majority of patients and does not have any clinical consequence although it cannot be excluded that in some subpopulations of patients at higher risk for infections, this increased risk from specific AEDs or combinations of them (for example, topiramate, levetiracetam or brivaracetam), might become clinically relevant.

Conclusions

In conclusion, due to several possible biases, caution should be adopted in accepting these findings and further studies should be undertaken to verify and analyse in more detail this phenomenon also evaluating periods of time longer than few months, which is the typical duration of a controlled clinical trial. Since the amplitude of this effect seems to be very small and it has been observed in a selected population of patients (patients recruited in clinical studies) which limit the external validity of this finding, large epidemiological studies might be the only way to assess whether patients under chronic treatment with topiramate, levetiracetam or brivaracetam have an higher risk of infection in respect to patients treated with gabapentin pregabalin or oxcarbazepine.

In the meantime, the process of selection of the best treatment for the single patient, particularly for those subjects with an higher risk of infections, should take into account that some AEDs might have a weak facilitating effect for infectious diseases.

Competing Interests

G.Z. has received speaker's or consultancy fees from EISAI and UCB Pharma. F.G., F.S.G., S.G. and F.M.T. report no disclosures. V.F. is a former employee of Eisai s.r.l., Italy.

We thank Flavio Moroni for his valuable and very helpful comments. The authors received no funding for this study. FG is supported by a grant by Ente Cassa di Risparmio di Firenze.

Contributors

G.Z. proposed the project, wrote the protocol, carried out the review and wrote the article. F.G. made all analyses. V.F. performed literature search. F.M. classified terms in MedDRA system. G.Z., F.G., F.S.G., V.F. and S.G. extracted data for a subset of included articles, cross checked the results, assessed eligibility, and commented on or edited sections of the article.

Supporting information

Data S1 PRISMA checklist

Data S2 Inclusion and exclusion criteria

Data S3 Search strategy and flow diagram of antiepileptic drugs studied

Data S4 Main features of the 127 randomized, double‐blind, placebo‐controlled trials included in the analysis

Data S5 Risk of bias of the included studies

Data S6 Full forest plot of the main analysis

Data S7 Meta‐analyses of preferred terms

Data S8 Tables of results

Zaccara, G. , Giovannelli, F. , Giorgi, F. S. , Franco, V. , Gasparini, S. , and Tacconi, F. M. (2017) Do antiepileptic drugs increase the risk of infectious diseases? A meta‐analysis of placebo‐controlled studies. Br J Clin Pharmacol, 83: 1873–1879. doi: 10.1111/bcp.13296.