Neurological and psychiatric safety of tafenoquine in Plasmodium vivax relapse prevention: a review

Neurotoxicity is a complex area for preclinical investigation. In vitro studies of cytotoxicity against human neuroblastoma cells or similar models are suggestive of neurotoxic potential, but the relevance of such results depends on drug penetration into the CNS, as well as pharmacokinetic and other factors, for example, metabolic status. In animal models, cytotoxicity and morphological damage are obvious outcomes, but may be restricted to very specific brain areas. Neurological effects can also derive from purely functional disruption, with no morphological damage, for example, convulsions induced by fluoroquinolones [14‐16]. Thus, in modern drug development, neurobehavioural testing is used to detect any functional effects of drugs. This uses both observational and more complex methods, and multiple tests must be used to objectively evaluate different potential functional and behavioural impacts of the drug.

For older anti-malarial drugs, comprehensive neurotoxicological preclinical testing may not have been conducted. However, for new agents, supra-therapeutic or lethal drug doses are administered in animal models to attempt to induce observable lesions in CNS tissues. If lethal doses do not induce neurotoxicity, then the toxicities that caused death are deemed dose limiting. In such circumstances, a “safety factor” (typically > 10-fold) is applied to the pharmacokinetic exposures in animals in order to predict the “highest achievable exposure” for humans. Consequently, the absence of neurotoxicity findings in pre-clinical animal models at supra-therapeutic or lethal doses is reassuring evidence that neurotoxicity in humans will be unlikely if the plasma concentrations in animals are at least tenfold greater than those achieved in humans following therapeutic or prophylactic dosing.

Even if drugs have the potential for neurotoxicity in animal models at supra-therapeutic doses, if the human dosing regimen does not achieve pharmacodynamically relevant exposures, and if the safety margins with CNS toxicity in preclinical species are acceptable, then NPAEs should not be expected. For example, although artemisinins are neurotoxic in animal models [17‐21], the rate of NPAEs with these agents in patients is very low (1.1%) and despite extensive use of artemisinins globally, NPAEs have not emerged as a significant safety concern [5]. The pre-clinical studies were conducted at supra-therapeutic doses with long durations of exposure and in some cases an oil vehicle was used, which further increased artemisinin drug exposure [22, 23]. There is also evidence that higher exposures are needed to elicit neurotoxicity in primates versus rodents and dogs [22]. In contrast, artemisinins for malaria treatment are given orally, for short durations and at low doses, generally resulting in human drug exposures that would not be expected to cause neurotoxicity [24, 25]. However, concerns still apply to certain populations where intrinsic factors may affect the pharmacokinetics of the drug. For example, oral dispersible artemether-lumefantrine at a dose of 20/120 mg twice daily for 3 days could not be recommended to treat infants weighing < 5 kg with P. falciparum malaria, as the artemether and dihydroartemisinin exposures exceeded the preclinical neurotoxicity safety margins [26]. Neurotoxicity may also limit the opportunity to increase artemisinin doses to overcome drug resistance [24]. Note also that following frequent dosing with artesunate in severe malaria, dihydroartemisinin levels in cerebrospinal fluid increased with time while levels in plasma declined, indicating accumulation [27]. Thus, NPAE risk depends on both the neurotoxic potential of a particular agent documented in preclinical studies and the pharmacokinetic/pharmacodynamic profile of the dosing regimen used.

The MedDRA preferred term classification system is the standard method for assigning adverse events (AEs) to drug treatments used in clinical trials. Two system organ classes are relevant for neuropsychiatric risk: neurological disorders and psychiatric disorders (Table 1).

Numerous papers of neuropsychiatric risk following anti-malarial treatment or during prophylaxis have been published, reaching various conclusions and promoting conflicting hypotheses [28]. It is important to note that, although case reports may act as flags for further investigation, on their own they cannot be accepted as evidence of a general drug-related neuropsychiatric effect, particularly for subjective symptoms.

A recent systematic meta-analysis of the published literature on ‘mental and neurological manifestations’ associated with anti-malarial drugs by Bitta et al. [5] provides a rigorous evaluation of the available clinical evidence. Specific psychiatric and neurological symptoms were examined for six anti-malarial drug classes across studies of treatment and prophylaxis (Fig. 1).

It can be seen immediately that there is no relationship between the prevalence of pooled NPAEs and the class of anti-malarial, except perhaps for the artemisinins, where prevalence is low for both artemether and artesunate. The lowest pooled prevalence of NPAEs was with the 8-aminoquinoline primaquine and the antibiotic dapsone and the highest was noted for the antibiotic minocycline and the 4-aminoquinoline amodiaquine. It can also be seen that for some drugs (mefloquine, pyrimethamine, sulphadoxine), the prevalence of events is far higher in prophylaxis studies than during malaria treatment, whereas for other drugs (chloroquine, atovaquone, artemether, artesunate) there is little difference between prophylaxis and treatment, and for quinine, prevalence is higher during treatment than for prophylaxis. Thus, we can conclude that there is no clear pattern of predicting NPAEs based on class or clinical indication, but that the risk depends on the specific neurotoxic potential and pharmacokinetic/pharmacodynamic properties of the drug (i.e. whether toxicity is related to maximum drug concentrations or to cumulative exposure) under the different dosing regimens for treatment and prophylaxis.

Malaria itself can produce varied and sometimes severe neuropsychiatric symptoms, the most serious being associated with cerebral malaria, including abnormal behaviour, impairment of consciousness, seizures, coma, or other neurologic abnormalities [29]. However, even in milder disease, symptoms commonly include high fever, headache, fatigue and confusion [29, 30]. Previous malaria episodes can also lead to long-term cognitive, behavioural and psychological symptoms [31‐39]. Thus, evaluating the relative contributions of drug and disease to NPAEs following anti-malarial treatment is very difficult, and the same data can be interpreted differently, leading to a range of opinions.

Prophylaxis studies remove the confounding effect of malaria symptomatology and allow a placebo-controlled comparison with a reference group. However, the dosing regimens and drug pharmacokinetics in prophylaxis are very different to those for treatment. In treatment, higher doses are generally given over a shorter time period, usually no more than 3 days, with the possible risk of adverse events related to maximum concentrations achieved. In contrast, prophylaxis generally involves lower doses regularly administered over a protracted time period, often months, with the risk of adverse events secondary to drug accumulation. For example, in the Bitta et al. meta-analysis outlined above, the difference between the NPAE risk for mefloquine in prophylaxis (25%) versus treatment (1%) was stark (Fig. 1). The increased frequency of NPAEs with mefloquine in prophylaxis is likely derived from the specific pharmacokinetics of the drug during the prophylactic dosing regimen, i.e. its ability to cross the blood–brain barrier, distribute and accumulate into cerebral tissue [40, 41], combined with its underlying potential for neurotoxicity at high doses, as suggested in pre-clinical studies [4, 42‐44]. Drugs that do not accumulate following multiple dosing in prophylaxis may actually have a lower frequency of adverse events than during treatment, for example, quinine (Fig. 1). Thus, data from prophylaxis studies cannot be extrapolated to predict NPAE risk during malaria treatment or vice versa.

The population that receives malaria treatment or prophylaxis may not have the same risk for neuropsychiatric symptoms as the general population. For example, as mentioned above, previous malaria can lead to ongoing psychiatric and behavioural issues [31‐39]. Exposure to insecticides used for malaria prevention, including dichlorodiphenyltrichloroethane (DDT), and pesticides for crop protection in rural areas, have also been suggested as a source of neurotoxicity [45‐47]. Poverty and malnutrition [48], as well as population displacement and conflict can also predispose to neuropsychiatric events [49], and are also risk factors for malaria [50].

In the case of prophylaxis, travel poses a number of potential neurological risks. Changes to routine and situation can trigger neuropsychiatric symptoms, for example by disruption of normal circadian rhythms or separation from support networks [51], as can certain foods and infections that travellers may be exposed to [52, 53].

The most controversial example of data generated on malaria prophylaxis is from studies in military personnel. Malaria is a serious risk to the operational effectiveness of a military force. Thus, there is a compelling motivation to identify the most suitable prophylactic agents for use in military deployments to malaria endemic regions. Moreover, malaria is a potentially fatal disease in non-immune individuals and long-term sequelae may afflict survivors. Even where studies have compared a non-immune military population deployed to a malaria endemic region against a ‘control’ of either an untreated resident population, or a non-deployed military population [54], it is important to understand the different underlying risk for neuropsychiatric findings in the compared populations.

To evaluate NPAE risk in soldiers given anti-malarial prophylaxis (deployed or non-deployed, with or without combat), we must first understand the underlying neuropsychiatric risk in military personnel versus civilians or in deployed versus non-deployed military personnel. If military personnel have a greater underlying risk for neuropsychiatric events than the general population, then trying to determine which NPAEs are drug-related, potentially over several months of treatment, becomes problematic, particularly if the underlying risk increases with deployment, i.e. over the same period that anti-malarial prophylaxis is taken.

In general, examination of large military health databases indicates that serving military personnel are at greater risk of neuropsychiatric symptoms, particularly depression, post-traumatic stress disorder and other internalizing conditions than the civilian population [55‐67]. This may result from a higher frequency of pre-enlistment mental health issues in military personnel, particularly soldiers, versus those who choose alternative occupations [56, 61, 62]. Deployment itself, particularly if associated with combat, affects neuropsychiatric risk [58, 61, 63‐66, 68‐71]. Thus, it might be expected that NPAEs occur more frequently in military personnel than the general population because of a higher underlying incidence of neuropsychiatric disorders. However, it is also possible that a drug which is associated with rare NPAEs in the general population may elicit such events more frequently in a more vulnerable population with a greater prevalence of pre-existing mental health issues, and/or a population subjected to severe stress, such as war-like military deployment [2, 72]. It seems reasonable, therefore, that NPAE risk should be evaluated prospectively and separately in specific populations (travellers, malaria endemic residents, deployed military, non-deployed military), as the results cannot necessarily be extrapolated between these groups.

Tafenoquine was previously investigated as monotherapy in P. vivax relapse prevention [80]. In a randomized, active-control, double-blind trial (Bangkok, Thailand), P. vivax patients were randomized to tafenoquine 400 mg once daily for 3 days (N = 46) or standard 1500 mg (base) total dose chloroquine given over 3 days plus primaquine 15 mg daily for 14 days (N = 24) [80]. This high monotherapy dose was abandoned as it had slow clearance of malaria parasites coupled with an unacceptable rate of drug-induced methaemoglobinemia. Hence, the further clinical development of tafenoquine was conducted in combination with a blood schizonticide. In terms of NPAEs, headache and dizziness were more common in the tafenoquine monotherapy group versus chloroquine/primaquine (Fig. 2). Although this could conceivably be a drug-related nervous system adverse effect, the headache and dizziness were more likely caused by slow parasite clearance under tafenoquine monotherapy: malaria symptoms persisted for longer in this group versus the chloroquine/primaquine group (Fig. 2) [80].

Initial studies of tafenoquine/chloroquine in P. vivax relapse prevention examined tafenoquine doses of 300 mg/day for 7 days, 600 mg/day for 3 days and 600 mg single dose compared to chloroquine plus primaquine 15 mg/day for 14 days (Table 2) [81]. Even at the highest tafenoquine dose (2100 mg cumulative), NPAEs occurred less frequently overall than with primaquine/chloroquine and were confined to vertigo and headache (Table 2). There was a trend for both vertigo and headache to occur more frequently at the higher tafenoquine doses. However, at the tafenoquine 600 mg single-dose, the incidence of vertigo was similar to that of primaquine/chloroquine and the incidence of headache was lower at all tafenoquine doses versus primaquine/chloroquine (Table 2).

Based on the need to co-administer a blood schizonticide, three comparative clinical studies have been conducted with single-dose tafenoquine plus standard 3-day chloroquine in P. vivax relapse prevention in centres across Asia, South America and Africa.

There are some key aspects that need to be highlighted in the assessment of NPAE risk with tafenoquine in P. vivax relapse prevention in these studies. Firstly, patients were infected with P. vivax malaria and symptomatic. Secondly, tafenoquine was co-administered with standard 3-day chloroquine, and so evaluation of tafenoquine NPAEs is hampered by the absence of a true placebo arm with comparisons having to be made against either chloroquine alone or primaquine/chloroquine. Lastly, in all three clinical trials, P. vivax recurrence after day 29 resulted in retreatment with primaquine/chloroquine and so adverse events recorded after this point include adverse events associated with recurrence of malaria as well as those associated with the rescue therapy. As relapse was more frequent in the chloroquine alone group than with either of the 8-aminoquinolines, adverse events after day 29 were consequently more frequent with placebo/chloroquine than for the 8-aminoquinoline arms. Thus, only adverse events occurring before day 29 should be considered for comparison to the placebo/chloroquine arm.

To further investigate whether an NPAE signal might be present in the safety database from these studies, an expanded definition of NPAEs was applied. As well as including the standard terms included under ‘nervous system disorders’ and ‘psychiatric disorders’, the expanded definition added labyrinthitis, vertigo, vestibular disorder, asthenia, fatigue, and alcohol intolerance to include terms with possible reference to vestibulocochlear function, which has its seat in the brainstem. This was considered important, considering the eighth cranial nerve appeared to have been affected adversely by some earlier quinoline molecules [82, 83].

In Table 5, 1445 patients received prophylaxis at the highest tafenoquine doses of > 300 mg > 3 days. However, 34.0% (492/1445) of subjects in the highest dose group were Australian Defence Force (ADF) personnel deployed on United Nations peacekeeping duties in East Timor, many of whom experienced a ‘war-like’ scenario [54]. A recent examination compared data for ADF subjects deployed in East Timor to non-ADF resident subjects who received the same tafenoquine regimen (200 mg loading dose for 3 days, then 200 mg weekly prophylaxis) [54]. In this analysis, all adverse events were more common in the deployed ADF group (467/492; 94.9%) than in resident non-ADF subjects (225/333; 67.6%), and psychiatric adverse events occurred in 5.1% (25/492) of ADF soldiers versus 2.1% (7/333) of resident non-ADF subjects [54, 79].

A possible explanation for this increased reporting of NPAEs in ADF personnel is that soldiers experienced traumatic events, including danger of being killed or injured (71%); seeing dead bodies (49%); fear of exposure to a toxic agent, contagious disease, or injury (31%); and having a friend/associate killed or injured (30%) [86]. Additionally, around 40% of soldiers who received tafenoquine sustained physical injuries as a result of their deployment while other illnesses, such as gastroenteritis and heat-related illnesses, were also common [54]. Thus, over the 6 months of tafenoquine administration, soldiers were exposed to many factors, other than tafenoquine, that could have contributed to their reporting of NPAEs and which confound causality assessment. Thus, the higher incidence of NPAEs at higher tafenoquine exposures not only reflects the long duration over which adverse events were recorded, but also the circumstances in which the drug was taken.