Severe efavirenz associated neurotoxicity: A retrospective cohort study

Background Efavirenz (EFV) is associated with neuropsychiatric symptoms. Severe neurotoxicity has been reported but the clinical phenotype and risk factors are poorly defined. Objectives To characterise clinical presentations, risk factors and outcomes to help clinicians recognise severe neurotoxicity earlier. Method The authors retrospectively identified adults with supratherapeutic EFV concentrations (> 4 mg/L) obtained during routine clinical care in Cape Town, South Africa. Clinical and laboratory data at the time of EFV quantification were extracted from medical records. Logistic regression was performed to identify associations with neuropsychiatric symptoms, and with severe neurotoxicity. Results Eighty one patients were included; 62 with neuropsychiatric manifestations (most frequently ataxia [n = 20] and psychomotor slowing [n = 24]); and 19 with hepatotoxicity. Overall, 28 (34.6%) were male, 49 (60.5%) had concomitant isoniazid exposure, and median EFV concentration was 12.1 mg/L (interquartile range [IQR]: 6.6–20.0). Neuropsychiatric symptoms were associated with longer duration of EFV therapy, adjusted odds ratio (aOR) 1.3/180-day increment (95% confidence interval [CI]: 1.0–1.7); higher EFV concentrations, aOR 1.2/1 mg/L increase (95% CI: 1.0–1.4) and isoniazid exposure, aOR 8.2 (95% CI: 2.5–26.7). Severe neuropsychiatric symptoms occurred in 47 (75%) patients at a median of 5.9 months (IQR: 2.1–40.8) after EFV initiation. Severe symptoms odds were 1.2-fold higher (95% CI: 1.1–1.4) per 1 mg/L increase in EFV concentration. Symptoms resolved completely within 1 month in 25 (76%) patients with severe neurotoxicity who discontinued EFV. Conclusion A concentration–effect relationship for severe neurotoxicity exists, which occurred late and resolved in most patients after EFV discontinuation. Contribution The authors highlighted clinical heterogeneity and morbidity of EFV-associated neurotoxicity.


Introduction
Efavirenz (EFV), a non-nucleoside reverse transcription inhibitor, has been a backbone of antiretroviral therapy (ART) for the last 15 years.Daily dosing, inclusion in fixed drug combinations and lack of significant drug-drug interaction with rifampicin have made it a widely used first-line drug in combination with two nucleoside reverse transcriptase inhibitors (NRTIs), 1 especially in resource-limited countries with high tuberculosis (TB) burdens. 2 Efavirenz is metabolised in the liver by cytochrome P450 isoenzyme 2B6 (CYP2B6).Singlenucleotide polymorphisms (SNPs) in CYP2B6, present in up to 20% of sub-Saharan African populations, 3,4 confer 'slow metaboliser' genotypes and lead to a risk of increased EFV concentrations, particularly in patients on concomitant TB treatment, 5 as alternate pathways of EFV metabolism are also inhibited by isoniazid (INH).
Two major adverse effects of EFV include hepatotoxicity, secondary to an immunoallergic pathway, 6 and central nervous system (CNS) toxicity possibly because of direct glial toxicity. 7,8linical features of CNS EFV toxicity range from sleep and mood disturbances in milder forms 9 to psychosis, cerebellar ataxia, encephalopathy 10 and rarely, death 11 ; long-term toxicity may promote Background: Efavirenz (EFV) is associated with neuropsychiatric symptoms.Severe neurotoxicity has been reported but the clinical phenotype and risk factors are poorly defined.

Severe efavirenz associated neurotoxicity:
A retrospective cohort study

Read online:
Scan this QR code with your smart phone or mobile device to read online.
http://www.sajid.co.zaOpen Access the development of HIV-associated neurocognitive disorder (HAND). 12Apart from host genetics and lower body weight (leading to higher EFV exposures), clinical risk factors for EFV neurotoxicity have not been established.Previous cohort studies have only included women, did not separate severe from milder symptoms and only small case series have documented severe EFV neuropsychiatric sequelae. 10,11,13,14gh rates of HIV-associated TB place patients in South Africa at particular risk of EFV neurotoxicity, especially given the relatively high background prevalence of 'slow metaboliser' genotypes and ongoing use of EFV for ART by some clinicians despite dolutegravir availability.However, the diagnosis may be missed because of overlap with other common neurological syndromes, and pharmacogenetic risk stratification may not be feasible in resource-limited settings.
We conducted a retrospective cohort study to describe the clinical phenotype of severe EFV-induced neurotoxicity and explore risk factors for its development.

Study population
We searched the University of Cape Town Clinical Pharmacology laboratory database for EFV concentrations measured in routine care at five public sector Cape Town hospitals between February 2008, when the current database was started, and July 2017; monitoring of EFV concentrations is conducted for clinical suspicion of toxicity or for adherence checks.Medical records of patients with elevated concentrations > 4 mg/L (normal range 1 mg/L -4 mg/L) were retrieved and reviewed.We included data from all patients over the age of 18 years with available records.

Clinical data
The following data were extracted from medical records, national laboratory services and picture archiving and communications system: biometrics including age, weight, and gender; treatment history relating to ART, TB therapy and isoniazid preventive therapy (IPT) and clinical manifestations at the time of, and subsequent to, EFV toxicity.Results of blood, cerebrospinal fluid (CSF) and radiological investigations were recorded to exclude other causes of neuropsychiatric syndromes including: neurosyphilis; bacterial, TB or, cryptococcal meningitis; neurological TB immune reconstitution inflammatory syndrome (IRIS); stroke and electrolyte abnormalities and hypoglycaemia.Data were captured using unique participant identifiers onto paper case report forms and entered into an electronic database (REDCap).
Efavirenz-associated neurotoxicity was defined as the presence of known neuropsychiatric manifestations of EFV toxicity without an alternative clinical or radiological explanation.Indicative clinical features included ataxia or cerebellar signs, psychomotor slowing (including slowed speech, decreased movement, impaired cognitive function and catatonia), mood disorders, psychosis, sleep disorders and confusion. 9,12,13,15Severe EFV-associated neurotoxicity was defined as a Division of Allergy and Infectious Diseases (DAIDS) altered mental status Grade 3 or more (Confusion, memory impairment, lethargy and somnolence causing inability to perform usual social and functional activities; or delirium, obtundation or coma), 16 and/or ataxia of DAIDS Grade 3 or more.A pre-specified case definition of EFV toxicity was applied to information extracted from clinical records.Cases were then classified according to DAIDS criteria.Patients fulfilling eligibility criteria (which included having supratherapeutic EFV concentrations) but no neuropsychiatric effects were included as 'nonneuropsychiatric' cases.

Analysis
Descriptive statistics were used to summarise the demographic and clinical characteristics of the study population.Univariable logistic regression was performed to determine associations between pre-specified predictors and the primary outcome of severe EFV-associated neurotoxicity.Independent variables included age, weight, EFV concentration, duration of EFV therapy, isoniazid exposure (either as TB therapy or IPT) and gender.Data completeness were used to determine a candidate set of variables for inclusion in a multivariable model with only variables with < 20% missing data included.This set was reduced by a backward step-wise model elimination using the Akaike Information Criterion (AIC) as the optimising criteria.We also used logistic regression to explore factors associated with the presence of neuropsychiatric symptoms of any severity, using cases with hepatitis (and no neurological manifestations) as a comparator.We checked for multicollinearity by testing the correlation between clinically linked variables and quantifying effects on model parameters -predictors that resulted in increased variance greater than or equal to 10% without an impact on coefficient size were dropped from the final model to avoid collinearity.The Hosmer-Lemeshow statistic was used to assess the calibration of the final model; discriminative ability was quantified by the area under the receiver operating characteristic (ROC) curve.Survival was represented using Kaplan-Meier plots with censoring at 3 months after initial supratherapeutic EFV concentration.Time to development of EFV toxicity after ART initiation was represented as an empirical cumulative distribution function, stratified by severity.Statistical analysis was conducted using R software, version 3.6.1.

Ethical considerations
Ethical approval for this study was obtained from Human Ethics Research Committee of the University of Cape Town (843/2016).

Clinical phenotype and outcomes
A total of 109 patients with supratherapeutic EFV concentrations were identified over the study period; data from 81 patients were included in the analysis, distributed as follows: Brooklyn Chest Hospital (n = 1), DP Marais (n = 16), Groote Schuur Hospital (n = 54), Mitchell's Plain District Hospital (n = 8) and New Somerset Hospital (n = 2) (See consort diagram Appendix 1, Figure 1-A1).Patients were excluded if their clinical records were missing or if they were under the age of 18 years at the time of EFV sampling.One patient had chronic diarrhoea and no other apparent indication of having had an EFV concentration performed and was also excluded.All patients were HIV-positive and were prescribed EFV as part of their ART.
A total of 62 patients had a neuropsychiatric syndrome and 19 had hepatitis as a reason for EFV sampling.Overall, 28 (34.6%)patients were male and the median age was 37.  1).Median EFV concentration was 20.0 mg/L (the upper limit of assay detection, IQR: 13.0-20.0) in those with severe versus 7.0 mg/L (IQR: 4.5-11.8) in those with mild neurotoxicity (p < 0.01).
At the time of EFV concentration sampling, 12 (19%) patients with neuropsychiatric symptoms had been noted in their clinical records as having suspected HAND and 6 (10%) were documented to have depression.
Efavirenz was stopped in all those for whom supratherapeutic concentrations were detected.A total of 34 patients were started on a protease inhibitor (PI) regimen and 18 were changed to nevirapine.Only one patient was re-challenged with EFV.Fourteen patients in the hepatitis group (74%) and 46 (74%) of those with neuropsychiatric presentations improved on withdrawal of EFV; the condition remained unchanged at 1 month in 6 (8%) patients.Among those with severe neurotoxicity, there was complete resolution of symptoms within 1 month in 25 out of 33 (76%) who are recorded to have discontinued EFV.
Fourteen (17%) patients died within the 3-month follow-up period, nine (64%) were associated with neuropsychiatric presentations, five (55.6%) of which were categorised as severe.Four patients had drug-resistant TB, two patients had sepsis, and three had generalised seizures.Five (36%) patients without neuropsychiatric symptoms died in hospital, all secondary to fulminant liver failure, two of whom were less than 3 months postpartum.Median time to death after diagnosis of EFV toxicity was 21 days (IQR: 10.0-36.0)overall; 20 days (IQR: 9.0-42.3)for neuropsychiatric presentations and 22 days (IQR: 12.5-35.5)for non-neurological presentations (Figure 2).

Discussion
We describe the clinical phenotype and factors that may contribute to higher risk of neurotoxicity among 81 patients with elevated EFV concentrations.There were durationand concentration-dependent effects, as well as higher risk with concomitant INH exposure and those with lower    CD4 count.Most patients with severe neurotoxicity had symptom resolution within 1 month after stopping EFV, although overall 3-month mortality was high in this population.
Prior adult studies involved smaller case series of neuropsychiatric symptoms in female patients. 13,14Our study is one of the few to include men in the neuropsychiatric cohort.Female gender has been associated with higher EFV concentrations compared with men, possibly because of a higher dose for weight with standard fixed drug combinations. 17,18However, more recent epidemiological data show a shift towards higher body weight in women in South Africa, potentially resulting in lower EFV concentrations. 19en did have a higher median EFV concentration (15.7 mg/L; IQR: 6.9-20 vs 10.8 mg/L in women; IQR: 6.6-20, p = 0.27, supplementary Figure 2-A1), but we could not ascertain weights from many of the records, so we cannot conclude this difference is because of weight alone.
The protective effect of a higher CD4 count against neuropsychiatric symptoms may also be linked to weightlower CD4s are associated with lower weights. 20Lower CD4 counts are also associated with higher risk of TB, 3 placing patients at a higher chance of being exposed to INH, either as part of TB treatment or prophylaxis, which would also increase their risk of EFV toxicity and neuropsychiatric symptoms.
The association between longer duration of EFV therapy and neuropsychiatric symptoms may be linked to direct toxicity of glial cells from EFV metabolites, 6 with a cumulative exposure threshold necessary for symptom appearance.Although the exact mechanism of EFV hepatitis symptoms remains unclear, suggested mechanisms include an EFV-induced mitochondrial dysfunction pathway. 21This potential difference in toxicity mechanism may explain why patients with hepatitis appeared to have poorer survival outcomes in the earlier part of their illness, with fulminant liver failure being the cause of death, although the difference was not statistically different (p = 0.38).Decreased survival from neurotoxicity occurred later, coinciding with the later onset of toxicity with prolonged course.A higher EFV concentration was associated with neuropsychiatric but not nonneuropsychiatric symptoms, supporting the hypothesis that there may be different pathological mechanisms in the clinical manifestations of EFV toxicity.
Higher EFV concentration was the only independent risk factor for severe neuropsychiatric symptoms.Symptoms of neuropsychiatric toxicity can be vague and non-specific.Patients may have presented later as they only sought medical assistance once symptoms were severe enough to impede daily functioning, and similarly, clinicians may have only requested EFV concentrations when they thought symptoms were serious enough to change to alternative ART.The median time to EFV neurotoxicity was 6 months in the current study, but other case series have noted much longer delays of up to 2 years, 14 which might reflect a greater awareness of the presentation of EFV toxicity during our study, as the lack of clinician awareness was cited previously as a reason for delayed diagnosis. 14nsistent with other case reports, cerebellar signs and neuropsychiatric symptoms often co-exist or pre-date each other. 13,14Nineteen per cent of patients with a pre-existing diagnosis of suspected HIV-associated encephalopathy is also in keeping with the finding of long-term neurocognitive depression associated with EFV toxicity. 15,22hibited by Isoniazid is a known risk factor for increased EFV toxicity in those with SNPs 23,24 for slow metaboliser genotypes, and in our population INH exposure was a significant risk factor for the development of EFV toxicity and neuropsychiatric symptoms.Pharmacokinetic and pharmacogenetic studies have shown that clearance of INH in patients taking both EFV and INH is highly dependent on the NAT2 or CYP2B6 polymorphisms, with presence of the NAT2 and CYP2B6 mutations having as much as a five-fold difference in EFV clearance between 'slow' and 'normal' metabolisers. 23With an estimated 30% -52.5% of South Africa's population having a slow NAT2 mutation, 25 and our data suggesting concentration-dependent toxicity, genotyping should ideally be performed prior to coadministering EFV and INH is known to be a risk. 26The association with INH co-administration may also explain why some patients also have a prolonged asymptomatic period -patients became toxic once INH was co-administered and symptoms took a month to resolve after EFV had been stopped.
Inability to obtain accurate, complete data from medical records is a limitation of all retrospective studies.Prespecified predictors, such as weight, were excluded from multivariable models because of missing data, potentially influencing outputs.Statistical analysis was also further complicated by certain subsets within the cohort being much larger than others, which also likely influenced the outputs of both univariate and multivariate models.Clinical notes detailing the neuropsychiatric condition of patients were not recorded in a standardised manner so that symptom severity may have been misclassified.Interobserver variation was, however, limited by one researcher collecting all the data.Although patients with neuropsychiatric symptoms had CSF and neuroradiological abnormalities, on reviewing the totality of the clinical information, we are fairly certain (acknowledging the limitations of retrospective analysis) that EFV was an important contributor.We excluded other conditions to the best of our ability, but as the patient cohort was hospitalised inpatients with multiple co-existing comorbidities, an undiagnosed or untreated contributor cannot be excluded.Potential cases with neurotoxicity but within the therapeutic EFV concentration range would not have been identified within this cohort, but the finding of a dose-dependent relationship suggests that supratherapeutic http://www.sajid.co.zaOpen Access concentrations are more likely to lead to neuropsychiatric symptoms.
Ethnicity, a surrogate for metaboliser phenotype, was not included as a parameter in this study as this variable was not consistently recorded and could not be inferred from the available demographic information in the hospital records.We were unable to collect samples for genotyping because of difficulties contacting patients.

Conclusion
This study highlights the clinical heterogeneity of EFVassociated neurotoxicity.Efavirenz toxicity is a reversible condition and recognition is critical to avoid misdiagnosis with potentially fatal outcome with continuation of EFV.Our findings support replacement of EFV by the integrase inhibitor dolutegravir as a first-line drug in ART programmes. 27,28

FIGURE 1 :
FIGURE 1: Empiric cumulative distribution function showing probability of efavirenz toxicity over time.

TABLE 1 :
Clinical characteristics at time of supratherapeutic efavirenz concentration.

TABLE 3 :
Associations with severe neuropsychiatric symptoms.