We retrospectively studied all patients with confirmed MDR TB who had received >1 dose of linezolid during January 2011–December 2016 and had been followed up for >12 months after the end of treatment. All patients had been admitted to Pitié-Salpêtrière Hospital, Paris, France, and most were discharged to Bligny Sanatorium and then followed up as outpatients. The TB drug regimen was adapted for each patient according to the results of drug susceptibility testing and discussion with the French Consilium team (20) in keeping with recommendations from the WHO and the Haut Conseil de la Santé Publique (14,21).

To monitor adverse effects of treatment, physicians performed daily neurologic and ophthalmologic examinations of hospitalized patients and monthly examinations of outpatients. Neurologic adverse effects included motor and sensory deficits, tendon reflex abnormalities, cranial nerve abnormalities, and pallesthesia. Ophthalmologic adverse effects included visual acuity loss, visual color abnormalities, or scotoma. If peripheral neuropathy was suspected, electromyography (EMG) was performed, including nerve conduction velocity (NCV) testing, to assess motor and sensory injuries to confirm the diagnosis. If a patient reported visual impairment indicating optic neuropathy, an ophthalmologist performed a specific clinical examination including Snellen best-corrected visual acuity measurement, relative afferent pupillary defect assessment, slit lamp examination, funduscopic examination, and color vision (Farnsworth D-15 Hue Test) and visual fields (Humphrey 24–2 SITA [Swedish Interactive Threshold Algorithm] standard or Goldmann) assessments. Nerve fiber layer thickness was evaluated by using optical coherence tomography for optic nerve abnormalities. If indicated, electrophysiologic measurement of visual evoked potential was also performed by an ophthalmologist. If an adverse effect was identified, follow-up frequency was determined by the physician, depending on severity of the effect. Neurologic and ophthalmologic adverse effect severity was classified by using the Common Terminology Criteria for Adverse Events (CTCAE) criteria (22).

Therapeutic drug monitoring for linezolid was performed at least once at the beginning of treatment (within the first month) for all patients except a few hospitalized at the beginning of 2011 when this analysis was not yet available at Pitié-Salpêtrière Hospital. If adverse effects occurred or if the dosage of linezolid was modified, serum concentration of linezolid was measured by using accredited ultra high–performance liquid chromatography–electrospray ionization/high-resolution mass spectrometry (European Medicines Agency, https://www.ema.europa.eu) and using stable isotope-labeled linezolid as an internal standard and a calibration range of 0.01–10.00 mg/L. We were able to follow patients for 12 months after end of treatment. Hence, we defined irreversibility of signs/symptoms as the persistence of signs/symptoms 12 months after the end of treatment.

We collected epidemiologic, clinical, and biological information from patients’ medical files. We investigated whether some variables were associated with neurologic toxicities by using a logistic regression model. Linezolid trough concentration was considered high if it was >2 mg/L at least once at baseline or during treatment. The data fulfilled the confidentiality criteria of the French National Data Protection “Commission Nationale de l'Informatique et des Libertés” (CNIL 2085894).

In our study, 58% of MDR/XDR TB patients reported a neurologic or ophthalmologic disorder possibly associated with linezolid exposure; neuropathies were confirmed for 47%. Linezolid-related peripheral neuropathy was confirmed (by EMG/NCV) for almost 32%. Neuropathy and demographic characteristics did not seem to be associated with trough concentration, or linezolid duration.

Linezolid-related optic neuropathies in patients receiving treatment for Staphylococcus aureus infections have been described (24) and can be asymptomatic or lead to reduced visual acuity, blurred vision, central scotoma, and dyschromatopsia (19,25). We found a high prevalence of confirmed optic neuropathy (25% of our cohort) that was fully reversible (meaning complete recovery of ophthalmologic disorder 12 months after the end of TB treatment) for only 2 of our patients after treatment withdrawal. We therefore suggest testing visual acuity and performing a sensory color vision test and a Goldmann visual field test before initiating linezolid with monthly checkups, as has been recently recommended (19).

Our finding of a peripheral neuropathy rate of 32% is similar to the neuropathy rate of 30% reported in a recent meta-analysis (12). In our study, peripheral neuropathy diagnosis was confirmed after a median duration of 8 months of treatment. In contrast, a study in China found that peripheral neuropathy occurred 2 to 4 months after treatment start (11). This difference can be explained by the high daily doses of linezolid received by patients in that study (1,200 mg/d loading dose for 6 weeks, then 600 mg/d) compared with the 600 mg/day received by all patients in our study. Various studies report irreversible neuropathy (10,12,26,27). In our study, 78% of patients with confirmed peripheral neuropathy did not fully recover despite linezolid withdrawal. Those patients were followed up for 12 months after the end of their TB treatment. Some may have recovered after that follow-up period. Thus, this result should encourage closer monitoring of neurologic complaints with detailed and regular clinical neurologic examinations. Linezolid should also be immediately discontinued at the onset of neurologic signs/symptoms.

Questions persist about the 2 patients who died. First, both patients had concurrent conditions known to increase the risk for neurologic disorders. One patient had hepatitis C, diabetes, and high blood pressure; the other had hepatitis C and a history of substance abuse. Second, linezolid treatment was continued for both patients despite development of neurologic signs/symptoms. At that time, in 2013, TB treatment was continued in the context of a highly resistant strain of M. tuberculosis, lack of an effective drug (e.g., delamanid) for treatment, and severe disease, despite daily in-patient monitoring. In hindsight, stopping any neurotoxic treatments as soon as neurologic signs/symptoms occurred may have been wise. Third, the second patient received several psychotropic drugs, which can potentiate the linezolid effect and induce serotonin syndrome. Linezolid-associated serotonin syndrome has been described in association with use of various drugs such as metoclopramide, paroxetine, or amitriptyline (28,29). Hence, when long-term use of linezolid is unavoidable because of highly resistant TB, drugs that induce serotonin syndrome must not be prescribed.

Another way to monitor adverse effects of antimicrobial drugs is to systematically measure serum trough concentrations. The role of linezolid dosage, treatment duration, and cumulative dose on neuropathy occurrence is still debated. Bolhuis et al. stated that the linezolid area under the curve over 24 hours does not affect adverse effect occurrence but that long exposure duration and cumulative doses do (18). Song et al. demonstrated a direct correlation between linezolid dose, trough concentrations, and development of clinical toxicity. They found that adverse effects developed in all patients for whom mean linezolid trough was >2 mg/L, whereas they developed in >50% of those for whom mean linezolid trough was <2 mg/L (17). In our study, linezolid trough concentration was not associated with development of neurologic side effects. The fact that our study was retrospective could explain why our findings were not consistent with those of Song et al. (17).

When we compared patients with and without clinical or confirmed peripheral or optical neuropathy, we found no demographic characteristics associated with neuropathy occurrence (Table 5). However, in a study of MDR TB patients in India, the high rate of neuropathy was explained by a higher proportion of malnourished patients (27). For specific TB parameters, we found no association between the onset of neurologic adverse effects and trough concentration or linezolid treatment duration. Neuropathy seems to be independent of linezolid trough concentration (88% for >2 mg/L, 82% for <2 mg/L, 75% for those without measurement) or linezolid treatment duration. In addition, linezolid treatment duration was shorter among patients who did not experience neurologic adverse effects than among patients who did, but this finding was not statistically significant, probably because of small sample size.

Although the 2018 WHO recommendations state that linezolid should be offered to all patients, the optimal treatment duration has not yet been established (15). The high prevalence of linezolid toxicity in our cohort, together with the lack of reversibility for more than half of the patients, is not in favor of long linezolid treatment durations for MDR TB patients. Furthermore, recent findings regarding the short duration of treatment regimens for MDR TB (30–32) have led the WHO to recommend 9–12 months of therapy if particular patient criteria apply (33). Hence, because of the high level of linezolid neurotoxicity and a general trend to decrease MDR TB treatment duration, we suggest that exposure to linezolid should be as short as possible to avoid development of irreversible side effects (34). Close monitoring should also be offered to all patients receiving linezolid. In our study, 72% of patients with signs of neuropathy at clinical examination were found to have EMG/NCV abnormalities. Using clinical examination to screen for peripheral neuropathy is therefore effective, even if EMG/NCV is unavailable. Ophthalmologic examination is more complicated. Ophthalmologic abnormalities were found in 9 of 13 patients without signs/symptoms and in 5 of 6 patients with signs/symptoms. We are unable to make clear recommendations regarding ophthalmologic testing for this patient population. However, we felt it necessary to raise awareness about linezolid toxicity and highlight the need to evaluate visual acuity as often as possible for patients receiving long-term linezolid.

Our monocentric study is limited by its retrospective design. The data were incomplete for several key variables, such as pharmacokinetics and pharmacodynamics, which could explain the difference between our findings and those of the study by Song et al. (17). The overall sample size in each patient group was small and may have been insufficiently powered to show differences between groups. The fact that 42% (24/57) of patients had hepatitis C and 14% (8/57) had alcohol dependency might increase the risk for peripheral neuropathy, independent of linezolid exposure. Moreover, among the 17 patients who received concomitant ethambutol, 8 had confirmed optic neuropathy. Although the ophthalmologic injuries resulting from linezolid differ from those resulting from ethambutol, completely excluding ethambutol toxicity in our findings is difficult. Thus, whenever possible, ethambutol use in a linezolid-based regimen must be avoided. Last, clinical examinations regarding ophthalmologic adverse effects were performed according to physician discretion without a systematic algorithm for recording adverse effects, which can bias the exact number of patients affected. This limitation, linked to the retrospective nature of the study, raises the issue of systematic ophthalmologic monitoring for patients receiving linezolid. One minor limitation is that only patients followed up for 12 months were included in the study. Two patients stopped coming to follow-up visits, possibly because of adverse linezolid events; thus, adverse events may have been underreported. However, these patients left during the first 3 months, probably before any neurologic adverse effects had developed.

Our study does, however, reflect the real difficulties of medical care and treatment surveillance of patients with highly resistant TB and serves to warn clinicians about long-term linezolid adverse effects. It also raises the crucial question of linezolid treatment duration for MDR TB patients. The patients in our study were in precarious economic and social situations; language barriers might explain the high rate of neurologic complications because patients might not have been able to explain their symptoms soon enough or might not have been sufficiently aware of the potential adverse effects of linezolid despite detailed explanations delivered by translators (by phone or internet). Overall, management was challenging for those patients, a situation that may occur elsewhere.

In conclusion, our study illustrates that MDR TB patients receiving linezolid should be monitored because neurologic adverse effects are severe, frequent, and often irreversible (at least at 12 months after treatment termination). Further studies are needed to evaluate the risk factors associated with linezolid toxicity and to evaluate the best treatment duration to decrease the rate of neurologic adverse effects without affecting MDR TB outcomes. However, systematic clinical examination should be implemented for all patients before treatment and monthly thereafter so that linezolid withdrawal can be discussed if neuropathy develops.