Cycloserine for treatment of multidrug-resistant tuberculosis: a retrospective cohort study in China.

PURPOSE: Cycloserine has been used in multidrug-resistant tuberculosis (MDR-TB) treatment since the 1950s. We evaluated the efficacy and safety of cycloserine and sought to clarify the role of cycloserine for treatment of simple MDR-TB, pre-extensively drug-resistant tuberculosis (pre-XDR-TB), and extensively drug-resistant tuberculosis (XDR-TB).
MATERIALS AND METHODS: A retrospective observational study was performed in Zhejiang Province, China. We enrolled 144 cycloserine-treated and 181 cycloserine-nontreated patients consecutively and determined the treatment outcome as the primary outcome. The proportion of patients with sputum culture conversion and the frequency of adverse drug reactions were also assessed.
RESULTS: One-hundred (69.4%) out of 144 patients in the cycloserine group successfully completed treatment. The HR of any unfavorable treatment outcome after the introduction of cycloserine was 0.58 (95% CI: 0.38-0.86, P=0.008). Subgroup analysis showed that cycloser-ine could benefit simple MDR-TB cases reducing the risk of unfavorable treatment outcomes (HR: 0.43, 95% CI: 0.24-0.76, P=0.004), but not pre-XDR-TB (HR: 0.65, 95% CI: 0.30-1.38, P=0.263) or XDR-TB (HR: 0.73, 95% CI: 0.22-2.37, P=0.589). The culture conversion rate at the intensive phase was similar whether cycloserine was administered or not (P=0.703). Of the 144 patients treated with cycloserine, a total of 16 (11.1%) patients experienced side effects attributed to cycloserine.
CONCLUSION: Cycloserine is an attractive agent for the treatment of MDR-TB, and its safety profile warrants its use in most MDR-TB cases. Cycloserine significantly improved the chance of a favorable outcome for patients with simple MDR-TB but not pre-XDR-TB and XDR-TB. Thus, more aggressive regimens might be required for pre-XDR-TB or XDR-TB patients.

Introduction

Tuberculosis (TB) has been a continuing threat throughout the ages. Since the early 1990s, multiple outbreaks of multidrug-resistant tuberculosis (MDR-TB), defined as TB caused by a strain of Mycobacterium tuberculosis that was resistant to at least isoniazid and rifampicin, have been reported, and it is generally accepted that resistance to these two potent anti-TB agents is associated with an increased probability of catastrophic treatment costs and poorer treatment outcomes. Furthermore, extensively drug-resistant tuberculosis (XDR-TB), defined as MDR-TB plus resistance to a fluoroquinolone and an injectable second-line drug, has recently emerged and threatened public health on a worldwide scale.1 In 2017, there were an estimated 460,000 new cases of MDR-TB, and China is a major contributor of drug-resistant TB (DR-TB). The latest data from the WHO reported a treatment success rate of 55% for MDR-TB and only 34% for XDR-TB.2

Cycloserine, a cyclic analog to D-alanine, could target alanine racemase and D-alanine ligase, thus blocking the formation of the bacterial cell wall.3 Cycloserine has been used in TB therapy since the late 1950s.4 Years later, Somner et al reported its potential for therapeutic intervention in chronic pulmonary TB.5 However, neurological toxicity associated with cycloserine concerns clinicians and limits its use. An earlier report described that symptomatic seizures were encountered in ~10% of patients.6 With the discovery of more effective drugs (eg, rifampicin), cycloserine has been applicable only in the treatment of apparent or proven DR-TB.

To implement TB control, the WHO published treatment guidelines for DR-TB in 1997 and cycloserine has been suggested since then, as it shares no cross-resistance with other agents and might be valuable to prevent resistance to other active drugs.7 In the 2008 recommendations, cycloserine was classified as a Group 4 oral bacteriostatic second-line medication.8 Recently, cycloserine has been recommended as one of the Group B drugs and should be generally included in the starting line-up in the longer regimen for the treatment of MDR-TB.9 However, the clinical studies that focused on cycloserine are scarce, particularly in East Asian patients, as the use of cycloserine was not approved in China until recently. Also, the role of cycloserine in the treatment of XDR-TB or pre-XDR-TB (defined as resistance to isoniazid and rifampicin plus any fluoroquinolone or one of the injectable drugs) is unclear. To address these uncertainties, we evaluated the efficacy, tolerability, and safety of cycloserine in MDR/pre-XDR/XDR-TB treatments using a sizable cohort of patients with MDR-TB from China.

Materials and methods

Study design and procedures

This cohort study was performed at two hospitals located in Zhejiang Province, China, and initiated by the Zhejiang Disease Control and Prevention Center (CDC) which has set up routine drug resistance monitoring for TB since 1999.10 The two clinical studies were completed in both the hospitals: one singer-arm study was supported by the Global Fund MDR-TB Project,11 and another cohort study entitled “Optimization of MDR-TB Treatment Regimen Based on the Molecular Drug Susceptibility Results of Pyrazinamide” was registered on ClinicalTrial.gov with number NCT02120638.12 To acquire complete follow-up information, we screened the patients in the two studies regarding eligibility consecutively. Approval for the collection of datp regimen. Culture and sputum conversion and chest X-rays were performed periodically for the evaluation of treatment outcomes. Moreover, adverse drug reactions (ADRs) were monitored and promptly managed during the entire treatment course.

Definitions

In the present study, we use the term simple MDR-TB to refer to those with resistance to just isoniazid and rifampicin and complicated MDR-TB to refer to those with additional resistance beyond isoniazid and rifampicin, including pre-XDR-TB and XDR-TB.

Standard treatment outcome definitions were applied according to the definitions and reporting framework for TB from WHO in 2013.13 Cured was defined as treatment completed without evidence of failure and three or more consecutive cultures were negative after the intensive phase. If bacteriological results were lacking, the case was defined as treatment completed. Treatment failure was defined as treatment terminated or need for permanent regimen change of at least two anti-TB drugs because of lack of conversion by the end of the intensive phase, or bacteriological reversion in the continuation phase after conversion to negative, or ADR. The default was defined as interruption of treatment for at least 2 months not meeting the criteria for failure. This study used the following brief outcomes: favorable outcome was defined as cured and treatment completion; and unfavorable outcome was defined as any failure, default, or death while on treatment.

When assessing the ADR, we distinguished two types of side effects: major side effects and minor side effects.14 The former refer to any adverse reaction that resulted in temporary or permanent discontinuation of anti-TB drugs, while the latter refer to those that only required a dose adjustment, addition of concomitant treatment, or both.

Drug susceptibility testing

Sputum culture on Löwenstein–Jensen medium or MGIT 960 were applied routinely. Phenotypic drug susceptibility testing to two first-line drugs (rifampicin and isoniazid) and two second-line drugs (ofloxacin and kanamycin) (Baso, Zhuhai, Guangzhou Province, China) was performed from the first positive M. tuberculosis culture with the use of the proportion method, and the result was compared with the standardized strains. The critical drug concentrations of rifampicin, isoniazid, ofloxacin, and kanamycin were 40, 0.2, 2, and 30 µg/mL, respectively.15

Data management and statistical analysis

The clinical data were collected through questionnaires and medical records by trained health workers. For the analysis, patients were divided into two cohorts according to the presence or absence of cycloserine in the background regimen (cycloserine cohort vs non-cycloserine cohort). Continuous variables were calculated as mean with SD and median with IQR and were further compared by the Mann–Whitney U test. Categorical data were presented as numbers (percentage) and were compared using chi-squared test.

The primary outcome was the proportion of favorable treatment in each treatment cohort. All patients’ treatment outcomes were identified according to the definitions described earlier by two clinicians blinded for the background regimen. Considering the potential confounders, we investigated the effect of cycloserine upon the treatment outcome by using a Cox proportional-hazards model among the factors with a P-value less than 0.05 in the univariate Cox regression analysis. Furthermore, we did a specific subgroup analysis of patients with different drug resistance patterns.

Secondary outcome included the efficacy of cycloserine measured by the proportion of conversion within the intensive phase and safety and tolerability of cycloserine measured by the frequency of major and minor reactions.

A two-tailed P-value of less than 0.05 was considered statistically significant. All statistical calculations and analyses in this study were performed using SPSS Statistics, version 22.0 (IBM).

Results

Study population

Between March 2012 and December 2015, a total of 582 patients were recruited to the Global Fund Project and the clinical trial NCT02120638 and then initiated the treatment for MDR-TB. Among 582 patients who were assessed for eligibility, 241 patients were excluded because their background regimens were not adapted to the inclusion criteria. Eleven patients were excluded because the strains from their isolates were identified as nontuberculous mycobacteria. Moreover, three HIV-positive patients and two patients with mental illness in the control group were also excluded. Consequently, 325 patients confirmed to have an organism resistant to both rifampicin and isoniazid were enrolled, of which 144 were treated with cycloserine in their background regimen according to WHO guidelines for designated dosages of 500 mg or 750 mg per day (500 mg for 38 patients weighing less than 50 kg; 750 mg for 96 patients weighing more than 50 kg). All patients’ background regimen included one of the later-generation fluoroquinolones and only two patients in the cycloserine group had not been treated with aminoglycosides as the initial treatment.

Most of the demographic and baseline clinical characteristics were comparable among two treatment cohorts except that fever and TB cavity being more frequent in the cycloserine group. The mean age was 44.0 and 41.7 years in the cycloserine-treated and non-cycloserine-treated group, respectively. Approximately, 70% of patients were male in both groups. More than a quarter (27.4%, 89/325) of patients were treated with at least one of the fluoroquinolones or aminoglycosides more than 30 days before. More details are given in Table 1.

Table 1

Characteristics of multidrug-resistant tuberculosis cases treated with or without cycloserine

Characteristics	Cycloserine (N=144)	No cycloserine (N=181)	P-value
Age (years)
Mean ± SD	44.0±12.7	41.7±13.1	0.067
Median (IQR)	45 (35–54)	40 (31–53)
Female sex	45 (31.3%)	51 (28.2%)	0.546
Bodyweight (kg)
Mean ± SD	54.2±8.8	53.8±7.8	0.541
Median (IQR)	54 (48–60)	52.5 (49–60)
Medical history of diabetes	21 (14.6%)	19 (10.5%)	0.265
Tuberculosis symptoms
Fever	28 (19.4%)	19 (10.5%)	0.023
Fatigue	26 (18.1%)	42 (23.2%)	0.257
Hemoptysis	26 (18.1%)	35 (19.3%)	0.769
Dyspnea	2 (1.4%)	6 (3.3%)	0.309
Cough	127 (88.2%)	151 (83.4%)	0.225
Chest radiograph
Presence of cavity	113 (78.5%)	117 (64.6%)	0.006
Bilateral involvement	103 (71.5%)	139 (76.8%)	0.279
Previous TB medications
Fluoroquinolones	36 (25.0%)	41 (22.7%)	0.694
Aminoglycosides	22 (15.3%)	25 (13.8%)	0.709
Drug-resistant patterns			0.569
Multidrug-resistant tuberculosis	85 (59.0%)	117 (64.6%)
Pre-extensive drug-resistant tuberculosis	48 (33.3%)	51 (28.2%)
Extensive drug-resistant tuberculosis	11 (7.6%)	13 (7.2%)
Treatment regimen			0.102
Individualized	14 (9.3%)	30 (16.5%)
Standardized	130 (90.7%)	151 (83.4%)
Most frequently used anti-TB drugs in the background regimen
Fluoroquinolones	144 (100%)	181 (100%)	NA
Aminoglycosides	142 (98.6%)	181 (100%)	0.196
Pyrazinamide	140 (97.2%)	162 (89.5%)	0.008
Prothionamide	136 (94.4%)	180 (99.4%)	0.012
Para-aminosalicylic acid	8 (5.6%)	152 (84.0%)	<0.001

Note: Data are presented as n (%), unless otherwise stated.

Abbreviations: TB, tuberculosis; NA, not applicable.

Treatment outcome assessment

The introduction of cycloserine to the standard regimen resulted in significantly less risk of unfavorable treatment outcomes (HR: 0.58, 95% CI: 0.38–0.86, P=0.008). The treatment outcomes are listed in Table 2. One hundred out of 144 (69.4%) cycloserine-treated patients and 108 out of 181 (59.7%) non-cycloserine-treated patients achieved treatment success (P=0.081). The absence to sputum conversion at 6 months and severe ADRs resulting in two or more drug stoppages were the main reasons for treatment failure; the relative responsibilities were 35.1% and 43.2% in the cycloserine group and 43.3% and 41.7% in the non-cycloserine group, respectively. One patient was complicated by pulmonary infection and died in the 18th month of treatment.

Table 2

Treatment outcomes of multidrug-resistant/extensively drug resistant tuberculosis cases treated with or without cycloserine

Treatment outcomes	Cycloserine (N=144)	No cycloserine (N=181)
Treatment success	100 (69.4%)	108 (59.7%)
Cure	94 (65.3%)	106 (58.6%)
Treatment completion	6 (4.1%)	2 (1.1%)
Treatment failure	37 (25.7%)	60 (33.2%)
Fail to conversion at 6 months	13 (9.0%)	26 (14.4%)
Reversion	8 (5.6%)	9 (5.0%)
Adverse drug reactions	16 (11.1%)	25 (13.8%)
Death	0 (0.0%)	1 (0.5%)
Default	7 (4.9%)	12 (6.6%)

Treatment outcomes stratified by resistance patterns

The treatment outcomes were further compared between the two groups stratified by resistance patterns (Figure 1). Among simple MDR-TB patients, the proportion of treatment success in the cycloserine group was higher than in the non-cycloserine group, reaching statistical significance (68/85, 80.0% vs 73/117, 62.4%, P=0.007). Cox regression analysis showed that cycloserine could provide benefit to simple MDR-TB patients by reducing the risk of unfavorable treatment outcomes (HR: 0.428, 95% CI: 0.240–0.761, P=0.004).

Figure 1

Proportions of favorable treatment outcome and the HR of any unfavorable treatment outcome after the introduction of cycloserine, according to the resistance pattern among patients treated with and without cycloserine.

Abbreviations: MDR-TB, multidrug-resistant tuberculosis; XDR-TB, extensively drug-resistant tuberculosis; pre-XDR-TB, pre-extensively drug-resistant tuberculosis.

For other strata, the treatment success rate in the cycloserine group was almost similar to patients who were not treated with cycloserine. Or rather, among pre-XDR-TB patients, the proportion achieving favorable outcome was 56.3% (27/48) in those receiving cycloserine vs 56.9% (29/51) in those treated without cycloserine, and among XDR-TB patients, the proportion of favorable outcome was 45.5% (5/11) and 46.2% (6/13), respectively. Cox regression analysis found that cycloserine might fail to improve the treatment outcome for pre-XDR-TB (HR: 0.650, 95% CI: 0.305–1.383, P=0.263) and XDR-TB patients (HR: 0.723, 95% CI: 0.223–2.374, P=0.589). Moreover, a downward trend in favorable treatment outcome rate was observed with the increase in the extent of drug resistance in both groups.

Efficacy of end point assessment

Efficacy was mainly measured by sputum culture conversion and proved to be roughly similar between the two groups. There was no difference in the proportion achieving sputum culture conversion at the end of the intensive phase (117/144, 81.3% vs 144/181, 79.6%, P=0.703) or the end of treatment (127/144, 88.2% vs 149/181, 82.3%, P=0.142) when comparing the cycloserine group and non-cycloserine group. The median (IQR) time to culture conversion was similar between the two groups (31 days vs 61 days), for a HR for conversion in the cycloserine group of 1.057 (95% CI: 0.81–1.37, P=0.679). We also calculated the sputum conversion rate at 6 months and observed no significant difference between the two groups regardless of the resistance patterns (data not shown).

Safety assessment

Overall, 132 of 144 patients (91%) in the cycloserine group and 161 of 181 patients (89%) in the non-cycloserine group had clinically significant ADRs. The most frequent adverse events were gastrointestinal effects (nausea and vomiting), arthralgia, liver injury, and hypokalemia in both treatment groups (Figure 2). Among the 132 patients reporting adverse events in the cycloserine group, 37 (28%) experienced major adverse effects and 95 (72%) patients experienced minor side effects. The adverse events attributed to cycloserine are shown in Table 3. Side effects that were possibly or probably related to cycloserine appeared after a median of 71 days (range 10–331 days) of cycloserine treatment. A total of 16 patients reported 17 episodes related to cycloserine, including nine patients who discontinued cycloserine temporarily or permanently. We observed eight episodes of headache, and cycloserine was permanently withdrawn from the treatment regimen in two patients. Moreover, two cases of seizure, one case of depression, and two cases of anxiety were observed, with these events resulting in cycloserine discontinuation within the first 6 months of treatment. No suicidal ideation was observed.

Figure 2

Adverse drug reactions in patients with and without cycloserine treated for multidrug-resistant tuberculosis in China.

Notes: Adverse drug reactions which were associated with cycloserine are marked in sky blue (minor adverse effects) or navy blue (major adverse effects).

Table 3

Side effects associated with cycloserine or requiring to withdraw cycloserine

Patient ID	Age (years)	Daily CS dosesa	Interval timeb (days)	Side effects	Relationship to CS	Dose adjustment or stoppage
P9	48	500 mg	119	Abdominal distension	Unlikely related	De-escalation to 250 mg
P80	52	750 mg	115	Headache	Probably related	Temporarily stopped (2 days)
P89	51	500 mg	132	Depression	Probably related	Permanently stopped
P90	39	750 mg	10	Seizures	Possibly related	Permanently stopped
P96	28	750 mg	164	Headache	Probably related	Permanently stopped
P106	56	500 mg	127	Seizures	Possibly related	Permanently stopped
P109	57	500 mg	1616	HeadachePeripheral neuropathy	Probably relatedPossibly related	No adjustmentNo adjustment
P110	61	500 mg	80	Tremor	Possibly related	Permanently stopped
P112	63	500 mg	41	Rash	Possibly related	No adjustment
P115	45	750 mg	331	Headache	Possibly related	De-escalation to 500 mg
P117	49	500 mg	50	Headache	Probably related	Permanently stopped
P119	45	750 mg	28	Anxiety	Possible related	Temporarily stopped (1 month)
P123	38	750 mg	58	Headache	Probably related	No adjustment
P124	35	750 mg	154	Headache	Probably related	De-escalation to 500 mg
P127	37	750 mg	71	Headache	Probably related	No adjustment
P143	39	750 mg	21	Anxiety	Probably related	Permanently stopped

Notes:

Daily CS doses refer to the doses in the background regimen.

Interval time from start of therapy to appearance of side effects (days).

Abbreviation: CS, cycloserine.

Risk factors to unfavorable treatment outcomes

The associations between unfavorable treatment outcomes and each baseline variable were first assessed using a univariate Cox regression model (Table S1). Using a Cox regression analysis (Table 4), we found that a significant risk of unfavorable treatment outcomes was related to age greater than 60 years (HR: 2.40, 95% CI: 1.36–4.22; P=0.002), the presence of a cough before starting treatment (HR: 2.48, 95% CI: 1.14–5.11; P=0.022), resistance to fluoroquinolones proven by drug-susceptibility testing (HR: 2.05, 95% CI: 1.24–3.38; P=0.005), and the individualized treatment regimen (HR: 1.66, 95% CI: 1.01–2.63; P=0.048).

Table 4

Cox regression analysis of potential independent variables associated with unfavorable treatment outcome in multidrug-resistant tuberculosis cases

Variables	Univariate Cox regression		Multivariate Cox regression
	Crude HR (95% CI)	P-value	Adjusted HR (95% CI)		P-value
Age ≥60 years	2.23 (1.32–3.66)	0.003		2.40 (1.36–4.22)	0.002
Cough	3.11 (1.45–6.68)	0.004		2.48 (1.14–5.139)	0.022
Fever	1.83 (1.17–2.87)	0.008		1.52 (0.95–2.44)	0.080
Bilateral involvement	1.92 (1.18–3.10)	0.008		1.50 (0.92–2.47)	0.105
Previous fluoroquinolones treatmenta	1.80 (1.20–2.70)	0.004		1.61 (1.00–2.60)	0.051
Previous aminoglycosides treatmenta	1.71 (1.10–2.69)	0.019		1.07 (0.63–1.83)	0.801
Resistance to fluoroquinolones	2.55 (1.57–4.14)	<0.001		2.05 (1.24–3.38)	0.005
Resistance to aminoglycosides	2.61 (1.37–4.99)	0.004		1.36 (0.69–2.72)	0.375
Cycloserine treatmentb	0.66 (0.45–0.96)	0.030		0.58 (0.38–0.86)	0.008
Standardized treatment regimenc	0.46 (0.30–0.72)	0.001		0.61 (0.38–0.99)	0.048

Notes:

Treated with fluoroquinolones or aminoglycosides more than 30 days before.

Treated with cycloserine, pyrazinamide, clarithromycin, high-dose isoniazid, or amoxicillin-clavulanate as the baseline regimen.

Standardized treatment regimen included a later-generation fluoroquinolone, pyrazinamide, prothionamide, an aminoglycoside, and cycloserine, or else para-aminosalicylic acid.

Discussion

The weaknesses and intolerability of current regimens brought more than half of DR-TB patients to treatment failure. The arrival of novel drugs such as delamanid and bedaquiline has offered fresh opportunities.16,17 However, up to now, there are insufficient new drugs to compose an entirely new regimen, so that the effective use of existing tools is urgently needed to combat MDR-TB.

The evaluation of cycloserine is greatly hampered by the paucity and weakness of the existing evidence. To our knowledge, this is the first study with a control group aimed at defining or optimizing the role of cycloserine in MDR-TB treatment. Our study suggested a significant trending toward improved treatment outcome after the introduction of cycloserine. When exploring the role of cycloserine for patients with different resistance patterns, we observed a significant improvement in treatment outcomes related to cycloserine in simple MDR-TB patients. However, for complicated MDR-TB patients, cycloserine alone was of less benefit without more effective drugs, such as linezolid and bedaquiline,14,18 indicating the requirement for reprioritization of cycloserine and new or repurposed agents when managing highly resistant forms of TB. Another interesting fact was that no association was found between the resistance to injectable agents and unfavorable treatment outcome, while the resistance to the fluoroquinolones may well give rise to treatment failure, which may suggest that fluoroquinolones are more critical for the treatment of MDR-TB (Table 4).

Our data reported an overall treatment success rate of 69.4% within 24 months in the patients treated with cycloserine. Previous studies reported that the successful outcome rate ranged from 67.5% to 77.0%,19–21 which was in accordance with our findings. There are several possible reasons to explain these subtle differences. First, the definition of treatment outcome has been updated and further emphasizes the tolerability of the regimens which was likely to be underappreciated before. Second, some studies combined adjunctive therapy such as surgical resection, which resulted in improved treatment outcomes.20,21 Moreover, the accelerated development of pre-XDR-TB and XDR-TB probably reduces the treatment success rate.

Currently, MDR-TB treatment outcome assessments need to integrate efficacy end points and safety end points. Efficacy end points in this study were mainly measured by time to and proportion of sputum culture conversion. Unlike drug-susceptible TB, failure to sputum conversion rather than relapse or sputum reversion accounted for a greater proportion of treatment failure,23 suggesting that the regimen in this study might not show strong sterilizing activity. Our study did not provide the evidence that cycloserine could confer a benefit to culture conversion. A possible explanation for this finding is that treatment of MDR-TB includes multiple drugs, and an observational study without strict placebo controls hardly assesses the efficacy of a single agent. Furthermore, a recent study has shown that more than half of patients with the recommended dosage of 10 mg/kg of cycloserine prescription had peak serum concentrations lower than the minimum inhibitory concentrations of the strains isolated from the corresponding patients, suggesting the need for adjusting each patient’s dosages depending on the clinical pharmacokinetic and pharmacodynamic assessments.24,25

ADR remains problematic during the treatment course of MDR-TB patients. By contrast with other anti-TB agents, ADR attributed to cycloserine was relatively uncommon, with a frequency of 11.1%. Consistent results were found in a meta-analysis that estimated the frequencies of any ADR from cycloserine at 9.1% (95% CI: 6.4–11.7).26 Neuropsychiatric reactions, as expected, were representative of adverse effects of cycloserine, since its central active mechanism as a partial NMDA-agonist and high brain–blood barrier permeability.27 In this study, headache was one of the most common side effects of cycloserine, although most headaches resolved quickly while seizures were rare, mainly associated with high dosages,28 co-administration of fluoroquinolones, and alcoholism,27 but all led to the withdrawal of cycloserine. Psychiatric disturbances were also described on rare occasions in our study but were more complicated to manage for clinicians. In greater detail, depression or anxiety might be partly attributable to the inadequate social support and lack of confidence owing to previous poor treatment outcomes, such as the patient (P89), who was suffering from depression and had been infected with M. tuberculosis for more than 8 years and suffered complications post-TB destroyed lung and complained of unbearable arthralgia during the treatment. A major challenge is the lack of reference standard against which to evaluate drug-induced psychiatric events. However, the current psychiatric reactions to cycloserine are mainly based on case reports,29 and further controlled studies are now needed.

Limitations

This retrospective study has several limitations. First, the major limitations derive from the observational study design, which precluded us from controlling confounding bias and other drugs in the background regimen and addressing some important topics, especially pharmacokinetic and pharmacodynamic assessments of cycloserine. Second, some strains isolated from patients were missing and thus we did not perform the drug susceptibility testing to cycloserine. Moreover, as cycloserine had not been approved in China until 2014, its availability and affordability require further evaluation.

Conclusion

The introduction of cycloserine improved the overall favorable outcome of MDR-TB patients. Cycloserine is considered a better-tolerated agent with infrequent adverse side effects characterized by neuropsychiatric reactions. For simple MDR-TB patients, we believe our results support the use of cycloserine in the setting of correct patient assessment and monitoring. For complicated MDR-TB patients, more effective treatment options should be considered.

Data availability

Individual participant data after deidentification that underlie the results reported in the article could be shared from 12 months following article publication for 2 years. Study documents could be available immediately after publication. The data set could only be shared with investigators who provide a methodologically sound proposal for individual participant data meta-analysis. Proposals should be directed to y_li11@fudan.edu.cn. Data requestors will need to sign a data access and proposals do not necessarily mean it will be assented.

Table S1

Univariate Cox regression analysis of potential independent variables associated with unfavorable treatment outcome in multidrug-resistant tuberculosis cases

Variables	n/N (%)	Univariate Cox regression
		Crude hazard ratio (95% CI)	P-value
Age (years)
<40	40/135 (29.6%)	Reference	–
40–59	60/157 (38.2%)	1.32 (0.88–1.98)	0.179
≥60	17/33 (51.5%)	2.23 (1.32–3.66)	0.003
Sex
Male	85/229 (37.1%)	Reference	–
Female	32/96 (33.3%)	0.87 (0.58–1.31)	0.505
Bodyweight	–	0.99 (0.97–1.01)	0.412
Medical history of DM
No	99/285 (34.7%)	Reference	–
Yes	18/40 (45.0%)	1.32 (0.80–2.18)	0.276
Fatigue
No	86/257 (33.5%)	Reference	–
Yes	31/68 (45.6%)	1.48 (0.98–2.24)	0.068
Cough
No	7/47 (14.9%)	Reference	–
Yes	110/278 (39.6%)	3.11 (1.45–6.68)	0.004
Fever
No	93/278 (33.5%)	Reference	–
Yes	24/47 (51.1%)	1.83 (1.17–2.87)	0.008
Hemoptysis
No	91/264 (34.5%)	Reference	–
Yes	26/61 (42.6%)	1.34 (0.86–2.07)	0.192
Dyspnea
No	113/317 (35.6%)	Reference	–
Yes	4/8 (50.0%)	1.66 (0.61–4.49)	0.322
Presence of cavity
No	34/95 (35.8%)	Reference	–
Yes	83/230 (36.1%)	1.02 (0.68–1.51)	0.940
Bilateral involvement
No	20/83 (24.1%)	Reference	–
Yes	97/242 (40.1%)	1.92 (1.18–3.10)	0.008
Previous fluoroquinolones treatment >30 days
No	79/248 (31.9%)	Reference	–
Yes	38/77 (49.4%)	1.80 (1.20–2.70)	0.004
Previous aminoglycosides treatment >30 days
No	93/278 (33.5%)	Reference	–
Yes	24/47 (51.1%)	1.71 (1.10–2.69)	0.019
Resistance to fluoroquinolones
No	31/104 (29.8%)	Reference	–
Yes	44/75 (58.7%)	2.55 (1.57–4.14)	<0.001
Unknown	42/146 (28.8%)	0.99 (0.61–1.59)	0.961
Resistance to aminoglycosides
No	65/175 (37.1%)	Reference	–
Yes	11/18 (61.1%)	2.61 (1.37–4.99)	0.004
Unknown	41/132 (31.1%)	0.68 (0.58–1.30)	0.480
Fluoroquinolones treatment in the baseline regimen
Levofloxacin	83/242 (34.3%)	Reference	–
Moxifloxacin	34/83 (41.0%)	1.18 (0.79–1.76)	0.408
Aminoglycosides treatment in the baseline regimen
Capreomycin	22/56 (39.3%)	Reference	–
Kanamycin	46/148 (31.1%)	0.75 (0.45–1.25)	0.270
Amikacin	39/119 (41.2%)	1.08 (0.66–1.80)	0.747
Prothionamide treatment in the baseline regimen
No	2/9 (22.2%)	Reference	–
Yes	115/316 (36.4%)	1.66 (0.41–6.72)	0.477
Cycloserine treatment in the baseline regimen
No	73/181 (40.3%)	Reference	–
Yes	44/144 (30.6%)	0.66 (0.45–0.96)	0.030
Pyrazinamide treatment in the baseline regimen
No	15/23 (65.2%)	Reference	–
Yes	102/302 (33.8%)	0.38 (0.22–0.66)	0.001
Para-aminosalicylic acid treatment in the baseline regimen
No	59/165 (35.8%)	Reference	–
Yes	58/160 (36.3%)	1.03 (0.71–1.47)	0.888
Ethambutol treatment in the baseline regimen
No	110/310 (35.5%)	Reference	–
Yes	7/15 (46.7%)	1.40 (0.65–3.01)	0.387
Clarithromycin treatment in the baseline regimen
No	15/23 (65.2%)	Reference	–
Yes	102/302 (33.8%)	2.32 (1.44–3.72)	<0.001
High-dose isoniazid treatment in the baseline regimen
No	106/306 (34.6%)	Reference	–
Yes	11/19 (57.9%)	2.23 (1.20–4.15)	0.012
Amoxicillin-clavulanate treatment in the baseline regimen
No	114/322 (35.4%)	Reference	–
Yes	3/3 (100%)	4.11 (1.30–12.99)	0.016
Standardized regimen
No	25/44 (56.8%)	Reference	–
Yes	92/281 (32.7%)	0.46 (0.30–0.72)	0.001

Abbreviation: DM, diabetes mellitus.