Transmission of fluoroquinolones resistance among multidrug-resistant tuberculosis in Shanghai, China: a retrospective population-based genomic epidemiology study

ABSTRACT

Fluoroquinolones (FQ) are essential for the treatment of multidrug-resistant tuberculosis (MDR-TB). The FQ resistance (FQ-R) rate in MDR-TB in China and its risk factors remain poorly understood. We conducted a retrospective, population-based genomic epidemiology study of MDR-TB patients in Shanghai, China, from 2009 to 2018. A genomic cluster was defined as strains with genetic distances ≤ 12 single nucleotide polymorphisms. The transmitted FQ-R was defined as the same FQ resistance-conferring mutations shared by ≥ 2 strains in a genomic cluster. We used multivariable logistic regression analysis to identify the risk factors for drug resistance. Among the total 850 MDR-TB patients included in the study, 72.8% (619/850) were male, the median age was 39 (interquartile range 28, 55) years, 52.7% (448/850) were migrants, and 34.5% (293/850) were previously treated patients. Most of the MDR-TB strains belong to the Beijing lineage (91.7%, 779/850). Overall, the genotypic resistance rate of FQ was 34.7% (295/850), and 47.1% (139/295) FQ-R patients were in genomic clusters, of which 98 (33.2%, 98/295) were presumed as transmitted FQ-R. Patients with treatment-naïve (aOR = 1.84; 95% CI: 1.09, 3.16), diagnosed in a district-level hospital (aOR = 2.69; 95% CI: 1.56, 4.75), and streptomycin resistance (aOR = 3.69; 95% CI: 1.65, 9.42) were significantly associated with the transmission of FQ-R. In summary, the prevalence of FQ-R among MDR-TB patients was high in Shanghai, and at least one-third were transmitted. Enforced interventions including surveillance of FQ drug susceptibility testing and screening among MDR-TB before initiation of treatment were urgently needed.

KEYWORDS:

• Multidrug-resistant tuberculosis
• fluoroquinolones resistance
• whole-genome sequencing
• drug resistance transmission
• China

Introduction

Multidrug-resistant tuberculosis (MDR-TB) is an infectious disease that is resistant to at least rifampin (RIF) and isoniazid (INH). In 2021, it was estimated that there were 450,000 cases of MDR-TB/rifampin-resistant TB (RR-TB) worldwide [1]. The high prevalence, long duration of treatment, low cure rate, and high economic burden make MDR/RR-TB a significant public health challenge that needs to be addressed globally [1,2].

Fluoroquinolones (FQ), such as moxifloxacin or levofloxacin, is one of the three group A drugs strongly recommended for treating MDR-TB, along with bedaquiline and linezolid, according to the World Health Organization (WHO)[3]. Studies showed that the FQ resistance (FQ-R) MDR-TB was associated with a higher treatment failure rate [4]. Besides, FQ susceptibility (FQ-S) is one of the prerequisites for initiating short-course chemotherapy regimens (9-month all-oral) that can improve the cure rate of MDR-TB[5]. Furthermore, the drug susceptibility testing (DST) result of FQ is also crucial to the optimal selection of the recently recommended 6-month treatment regimen (BPaL-M/BPaL) for MDR/RR-TB patients by the WHO [5]. In addition, FQ-R is also relevant to the evolution of MDR-TB to extremely drug-resistant TB (EDR-TB, MDR-TB additional resistance to any FQ and to any of three second-line injectable drugs) [6] or extensively drug-resistant tuberculosis (XDR-TB, MDR/RR-TB additional resistance to any fluoroquinolone and at least one additional group A drug) [1], which has an even lower cure rate and higher mortality than MDR-TB.

China has the fourth highest MDR-TB/RR-TB burden worldwide, making it one of the key countries in the global control of MDR/RR-TB [1]. In the past decades, FQ has been widely used in China for the treatment of bacterial infections of the respiratory, gastrointestinal, and urinary tracts [7], as well as for the treatment or empirical treatment of patients with suspected TB [8]. Gradually, it became a core drug in the treatment regimen of patients with drug-resistant TB [9]. The prevalence of FQ-R TB in China increased significantly from 2000 to 2013 [10–12]. Widespread or inappropriate use of FQ may lead to acquired and transmitted FQ-R, which could seriously threaten the effective treatment and control of MDR-TB. Here, we retrospectively analysed the prevalence of FQ-R, its risk factors, and the transmission of FQ-R strains among MDR-TB patients in Shanghai, China, based on the whole-genome sequencing (WGS) data of strains and epidemiological data.

Methods

Study population and data collection

Shanghai is one of the most populous cities in China, with an estimated population of 24 million, and it has a relatively well-functioning TB control program. This study included all patients ≥ 15 years old with culture-confirmed MDR pulmonary tuberculosis who were reported in Shanghai between January 1, 2009, and December 31, 2018. All Mycobacterium tuberculosis (M. tb) isolates from sputum cultures were collected at diagnosis before treatment initiation and were sent to the Tuberculosis Reference Laboratory in the Shanghai Municipal Centre for Disease Control and Prevention for DST and strain preservation. DST was performed using the proportion method on the Löwenstein-Jensen medium for RIF and INH as previously described [13]. Epidemiological information was collected from the National Tuberculosis Surveillance System, including demographic, clinical, and microbiological records.

Patient delay was defined as the time elapsed from the onset of symptoms to the date of a healthcare-seeking visit. Health system delay was defined as the time elapsed from the date of that healthcare-seeking visit to diagnosis. Previously treated indicates a history of previous TB treatment, while treatment-naïve indicates no history of TB treatment. Residents indicate household registration in Shanghai, and migrants indicate internal (in-country) migrants whose household registration is not in Shanghai.

WGS analysis and phylogenic analysis

DNA extraction used the cetyl trimethyl ammonium bromide method, as previously reported [13]. Paired-end DNA sequencing was performed on a Hiseq 2500 platform (Illumina, San Diego, CA, USA) with an expected coverage of 100 times. The WGS analysis, including mapping and SNP calling, followed a previously validated pipeline [13]. Briefly, we used Bowtie2 (v2.3.1) to map the sequencing reads to the reference genome H37Rv (NC_000962.3). Single nucleotide polymorphisms (SNPs) were called using SAMtools (v1.6), with a minimum mapping quality of 30. We identified fixed mutations (frequency ≥ 75%) using VarScan (v2.3.6), with at least ten reads supporting and the strand bias filter option enabled. PPE/PE-PGRS family genes, phage sequences, mobile genetic elements, insertions, and deletions were excluded from our SNP analysis [13].

A genomic cluster was defined as isolates with pairwise genetic distances ≤ 12 (SNPs), which were considered likely to be consistent with transmission [14,15]. A phylogeny tree was constructed using RAxML software (v1.0.2) and the maximum likelihood method with 500 bootstraps. The phylogeny tree was visualized and annotated with the Interactive Tree of Life (https://itol.embl.de/). The classification of lineages/sublineages was identified from the clade-specific SNPs [16,17]. Genotypic drug-resistance profiles were predicted using a TB-Profiler tool (v4.1.0) [18]. EDR-TB was defined as MDR-TB strain plus additional resistance to any FQ and any second-line injected drug (amikacin, AM; kanamycin, KM; capreomycin, CM) [6]. XDR-TB was defined as an MDR-TB strain with additional resistance to any FQ and at least one additional Group A drug (bedaquiline, linezolid) [1]. Pre-extensively drug-resistant TB (pre-XDR-TB) was defined as MDR-TB strain additional resistance to any FQ, but not resistance to second-line injectable drugs or group A drugs simultaneously [1]. See the Appendix for further details.

Definitions of transmitted FQ-R

The transmitted FQ resistance (transmitted FQ-R) was defined as ≥ 2 strains in a single genomic cluster sharing the same FQ resistance-conferring mutation type. The remaining were non-transmitted FQ resistance (non-transmitted FQ-R), including the strains of the same FQ resistance-conferring mutation type sharing by < 2 strains in a single genomic cluster, and non-clustered strains.

Statistical analysis

We used R software (v4.1.2, Vienna, Austria) for all statistical analyses. Continuous variables were presented as medians (interquartile range [IQR]), and categorical variables were presented as numbers and percentages. Differences among groups were compared using the chi-square test for categorical variables and the Mann–Whitney U test for continuous variables (all not normally distributed). We used the phi coefficient (r) to measure the degree of association between two categorical variables. We used logistic regression analysis to identify the risk factors associated with FQ-R, transmitted FQ-R, EDR-TB, and second-line anti-TB drugs. The statistical results were expressed as odds ratios (OR) and 95% confidence intervals (CI). Variables with p-values less than 0.2 in the univariable analysis were included in the multivariable analysis and used a backward step-wise approach to estimate the adjusted odds ratios (aOR). A p-value less than 0.05 was considered statistically significant.

Results

Characteristics of patients and strains

A total of 1,236 patients were culture-confirmed with MDR-TB in Shanghai during the study period, of which 265 patients had no available strains or matched records. After excluding 81 strains that failed to reculture or were contaminated, 890 strains were sequenced, of which 850 (95.5%, 850/890, one strain per patient) were finally included (Figure 1). Among them, 72.8% (619/850) were male, the median age was 39 (IQR 28, 55) years, and 52.7% (448/850) were migrants. More than one-third (34.5%, 293/850) were previously treated patients, and the median of the total diagnostic delay was 26 (IQR 12, 45) days (the patient delay was 10 [IQR 3, 28] days, and the health system delay was 7 [IQR 4, 17] days). The first diagnosis unit of 66.1% (562/850) of the patients was in a district-level hospital (Supplementary Table S1).

Figure 1. Flow chart of the study. Abbreviations: MDR-TB, multidrug-resistant tuberculosis; FQ, Fluoroquinolones; M. tb, Mycobacterium tuberculosis.

Phylogenetic analysis showed that the majority of strains belonged to the Beijing lineage (L2, 91.7%, 779/850), among which 0.3% (2/779) belonged to proto-Beijing (L2.1), 7.2% (56/779) belonged to ancient Beijing (L2.2.2), and 92.5% (721/779) belonged to modern Beijing (L2.2.1). Among the remaining strains, 70 (8.2%, 70/850) were the Euro-American lineage (L4), and one belonged to the East African Indian lineage (L3, 0.1%, 1/850) (Figure 2). In total, 39.4% (335/850) isolates were clustered into 115 genomic clusters, each containing 2–12 strains (Figure 2 and Figure 4, A).

Figure 2. Phylogeny, clustering, and resistance profile of 850 Mycobacterium tuberculosis strains isolated in Shanghai. The different colours on the branches indicate different lineages and sublineages. A total of 850 MDR-TB strains included 779 L2 strains (including 2 of L2.1 [orange], 56 of L2.2.2 [purple], 721 of L2.2.1 [fuchsia]), 1 L3 strain (green), and 70 L4 strains (blue). The first outer circle indicates genomic-clustered strains differing by ≤ 12 SNPs. The second outer circle indicates the distribution of fluoroquinolone resistance and whether it is transmitted. The small rectangles of different colours on the outer middle ring indicate second-line drug resistance. The outermost circle indicates the household registration of patients. Abbreviations: FQ, Fluoroquinolones; EDR-TB, extremely drug-resistant tuberculosis.

Genotypic drug-resistance prediction profiles

All isolates were genotypically resistant to RIF, 94.8% (806/850) to INH, 41.2% (350/850) to pyrazinamide (PZA), 61.9% (526/850) to ethambutol (EMB), and 76.2% (648/850) to streptomycin (SM). Among the second-line anti-TB drugs, FQ had the highest resistance rate (34.7%, 295/850), followed by ethionamide (ETO, 23.5%, 200/850), KM (9.5%, 81/850), para-aminosalicylic acid (PAS, 9.2%, 78/850), CM (8.0%, 68/850), AM (7.6%, 65/850), cycloserine (0.1%, 1/850), and linezolid (0.1%, 1/850). No resistance-conferring mutations were identified for bedaquiline, clofazimine, or delamanid in our study; 28.0% (238/850) of the patients were pre-XDR-TB, 6.7% (57/850) were EDR-TB, and one patient had XDR-TB. The proportion of treatment-naïve patients accounted for more than half (56.9%, 168/295) among patients with FQ-R (Supplementary Table 2).

Of 295 FQ-R strains, 94.6% (279/295) had gyrA mutations and 10.5% (31/295) had gyrB mutations. Notably, 11.2% (33/295) of FQ-R strains had double-locus mutations, 1.4% (4/295) had triple-locus mutations, and 4.7% (14/295) had both mutations of gyrA and gyrB (Supplementary Table S3). The most frequent mutation sites of other anti-TB drugs and their proportion in FQ-R patients are provided in Supplementary Table S4 and Table S5. There was no significant difference in the distribution of all mutations in FQ-R strains between treatment-naïve and previously treated patients (all p > 0.05, Supplementary Figure S1).

Correlation between FQ-R and resistance to other anti-TB drugs

FQ-R strains may also exhibit resistance to other drugs, which can complicate treatment options. Our analysis showed that PZA, EMB, SM, AM, KM, CM, ETO, and PAS were significantly associated with FQ-R (all p < 0.01), but the correlation coefficient was not high (all r < 0.30) (Figure 3, A).

Figure 3. Correlation analysis between FQ and resistance to other anti-TB drugs. (A) The heatmap of the correlation between each drug pair. (B) The heatmap of the correlation between each of the most frequently drug-resistant mutation pair, and their correlation between FQ-R and FQ-S strains. p values were calculated using the chi-square test and Fisher’s exact test (when the minimum theoretical frequency was <1). Use the phi-coefficient (r) to measure the degree of correlation between two categorical variables and numerical values in the figure indicate the degree of correlation. Abbreviations: RIF, Rifampicin; INH, Isoniazid; PZA, Pyrazinamide; EMB, Ethambutol; SM, Streptomycin; SLID, Second-Line Injectable Drugs (amikacin; capreomycin; kanamycin); ETO, Ethionamide; PAS, Para-aminosalicylic acid; FQ, Fluoroquinolones.

We further surveyed the most frequent drug-resistant mutation sites of other anti-TB drugs among the FQ-R strains. The correlation analysis showed that embB M306 V, rpsL K43R, fabG1 −15C > T, and rrs 1401A > G mutation were significantly associated with FQ-R (all p < 0.05), but only rrs 1401A > G mutation was more frequent in FQ-R (n = 46) than FQ-S (n = 15) strains (r = 0.24, p < 0.001) (Figure 3, B and Supplementary table S5).

Risk factors of FQ-R and other second-line anti-TB drugs in MDR-TB patients

The multivariable logistic regression analysis showed that patients who were being residents (aOR = 1.49; 95% CI: 1.08, 2.06), previously treated (aOR = 1.56; 95% CI: 1.13, 2.16), unfavourable treatment outcomes (aOR = 2.00; 95% CI: 1.43, 2.80), PZA resistance (aOR = 2.46; 95% CI: 1.79, 3.41), and EMB resistance (aOR = 3.15; 95% CI: 2.20, 4.57) were associated with FQ-R (Table 1).

Table 1. Univariable and multivariable logistic regression on the risk factors of FQ-R patients (N = 850).

Characteristics	FQ-S (n = 555)	FQ-R (n = 295)	OR [95% CI]	aOR [95% CI]
Demographic factors
Gender
Female	150 (27.0)	81 (27.5)	Ref
Male	405 (73.0)	214 (72.5)	0.98 [0.71, 1.35]
Age
≤ 24	90 (16.2)	40 (13.5)	Ref
25–44	252 (45.4)	104 (35.3)	0.93 [0.60, 1.45]
45–64	164 (29.5)	110 (37.3)	1.51 [0.97, 2.37]^a
≥ 65	49 (8.8)	41 (13.9)	1.88 [1.08, 3.30]^a
Household registration
Migrants	318 (57.3)	130 (44.1)	Ref
Residents	237 (42.7)	165 (55.9)	1.70 [1.28, 2.27]^a	1.49 [1.08, 2.06]^b
Occupation
Commercial service	29 (5.2)	15 (5.1)	Ref
Student	15 (2.7)	13 (4.4)	1.68 [0.63, 4.46]
Office worker	37 (6.7)	13 (4.4)	0.68 [0.28, 1.65]
Labor worker/Farmers	158 (28.5)	64 (21.7)	0.78 [0.40, 1.59]
Retirement	79 (14.2)	61 (20.7)	1.49 [0.74, 3.09]
Unemployed/unknown/Others	237 (42.7)	129 (43.7)	1.05 [0.55, 2.08]
TB treatment history
Treatment-naïve	389 (70.1)	168 (56.9)	Ref
Previously treated	166 (29.9)	127 (43.1)	1.77 [1.32, 2.38]^a	1.56 [1.13, 2.16]^b
Total diagnostic delay
≤ 2 weeks	178 (32.1)	82 (27.8)	Ref
2–4 weeks	137 (24.7)	62 (21.0)	0.98 [0.66, 1.46]
> 4 weeks	240 (43.2)	151 (51.2)	1.37 [0.98, 1.91]^a
First diagnostic unit
District-level hospitals	382 (68.8)	180 (61.0)
Tertiary hospitals	173 (31.2)	115 (39.0)	1.41 [1.05, 1.89]^a	1.36 [0.97, 1.90]
Sputum smear positive
No	157 (28.3)	86 (29.2)	Ref
Yes	391 (70.5)	207 (70.2)	0.97 [0.71, 1.32]
Unknown	7 (1.2)	2 (0.6)	0.52 [0.08, 2.21]
Severe cases
No	443 (79.8)	225 (76.3)	Ref
Yes	112 (20.2)	70 (23.7)	1.23 [0.87, 1.72]
Treatment outcomt
Favorable outcomes	420 (75.7)	175 (59.3)	Ref
Unfavourable outcomes	135 (24.3)	120 (40.7)	2.13 [1.58, 2.89]^a	2.00 [1.43, 2.80]^b
Bacteriological factors
Genotype
Non-Beijing	54 (9.7)	17 (5.8)	Ref
Beijing^c	501 (90.3)	278 (94.2)	1.76 [1.02, 3.19]^a
Pyrazinamide
Susceptible	382 (68.8)	118 (40.0)	Ref
Resistant	173 (31.2)	177 (60.0)	3.31 [2.47, 4.45]^a	2.46 [1.79, 3.41]^b
Ethambutol
Susceptible	270 (48.6)	54 (18.3)	Ref
Resistant	285 (51.4)	241 (81.7)	4.23 [3.03, 5.98]^a	3.15 [2.20, 4.57]^b
Streptomycin
Susceptible	149 (26.8)	53 (18.0)	Ref
Resistant	406 (73.2)	242 (82.0)	1.68 [1.19, 2.40]^a
Main mutation of MDR-TB
rpoB S450L (RIF)
No	203 (36.6)	109 (36.9)	Ref
Yes	352 (63.4)	186 (63.1)	0.98 [0.73, 1.32]
katG S315T (INH)
No	132 (23.8)	74 (25.1)	Ref
Yes	423 (76.2)	221 (74.9)	0.93 [0.67, 1.30]
Genomic-Clustered (12SNPs)
No	359 (64.7)	156 (52.9)	Ref
Yes	196 (35.3)	139 (47.1)	1.63 [1.22, 2.18]^a

^a Variables with p values < 0.2 in the univariable regression analysis.

^b Denotes statistical significance.

^c Include L2.1 (n = 2) and L2.2 (n = 777).

Abbreviations: RIF, Rifampicin; INH, Isoniazid; FQ-S, fluoroquinolones-susceptible; FQ-R, fluoroquinolones-resistant; OR, odds ratio; CI, confidence interval; Ref, reference.

Meanwhile, multivariable logistic regression analysis showed that male (aOR = 2.45; 95% CI: 1.20, 5.57), sputum smear-positive (aOR = 0.47; 95% CI: 0.26, 0.85), PZA resistance (aOR = 3.02; 95% CI: 1.63, 5.85), and EMB resistance (aOR = 4.02; 95% CI: 1.77, 10.85) patients were associated with EDR-TB (Supplementary Table S6). We also conducted a univariable analysis for other second-line anti-TB drugs, and none of the demographic or clinical characteristics were significantly associated with AM, KM, CM, ETO, or PAS resistance (Supplementary Table S7).

Transmission of FQ-resistant MDR-TB strains

In total, 139 (47.1%) of the 295 FQ-R strains were in genomic clusters, of which 98 (70.5%, 98/139) had the same FQ resistance-conferring mutation type sharing by ≥ 2 strains in a single genomic cluster, defined as transmitted FQ-R, and accounted for about one-third of (33.2%, 98/295) of the total FQ-R strains (Figure 1 and 2). Most of the transmitted FQ-R strains (89.8%, 88/98) belong to the modern Beijing sublineage (L2.2.1). A total of 38 transmitted FQ-R genomic clusters were identified, most of which were in small clusters (2–3 strains), and the largest one contained six FQ-R strains (Figure 4, A).

Figure 4. The proportion of transmitted FQ-R in clustered MDR-TB (A) and cumulative rate in FQ-R patients (B). (A) The bar indicates the number of clusters for each cluster size, and shaded areas refer to MDR-TB clusters with transmitted FQ-R in the clusters. Numbers on the bars indicate the cluster number of transmitted FQ-R/MDR-TB clusters in the same cluster size. (B) The bar indicates the cumulative number of FQ-R patients, the triangle indicates the FQ-R cumulative clustering rate (FQ-R clustered strains/FQ-R strains) and the circle indicates the cumulative transmission rate (transmitted FQ-R strains/FQ-R strains). Abbreviations: FQ-R, FQ resistance.

The cumulative clustering rate and transmission rate of FQ-R strains showed a persistent increase year by year during the study period (Figure 4, B; Supplementary Figure S2). To further describe the dynamics of the FQ-R rate and its transmission between 2009 and 2018, we calculated the FQ-R rate and transmitted FQ-R rate using a sliding 3-year window. There were no significant changes in the FQ-R rate and transmitted FQ-R rate for the study period (p > 0.05, Supplementary Figure S3).

Of the 38, 11 clusters (28.9%) contained patients that were all residents, and seven (18.4%) clusters were all migrants, while the remaining 20 clusters (52.6%) were mixed with both the resident and migrant population groups. In addition, nearly half (49.0%, 25/51) of patients in the mixed clusters were migrants (Figure 5).

Figure 5. Transmitted FQ clusters. (A) ≥ 3 strains in a cluster; (B) 2 strains in a cluster. The orange branches refer to the migrant, and the gray refers to the resident. The outermost label is the mutation type of FQ and second-line injectable drugs (including AM, KM, and CM). Abbreviations: C, cluster; EDR-TB, extremely drug-resistant tuberculosis.

As shown in Figure 5, nine clusters (26 strains) had 23 EDR-TB strains. Among them, six clusters contained 19 strains that were all EDR-TB strains (Cluster-01[C01], C04, C06, C18, C30, and C38), indicating the transmission of EDR-TB and further supporting the significant correlation between the FQ-R and the mutation of rrs 1401A > G.

Table 2 shows the multivariable logistic regression analysis of factors associated with the transmission of FQ-R patients. Patients who were treatment-naïve (aOR = 1.84; 95% CI: 1.09, 3.16), diagnosed in a district-level hospital (aOR = 2.69; 95% CI: 1.56, 4.75), and SM resistance (aOR = 3.69; 95% CI: 1.65, 9.42) were significantly associated with the transmission of FQ-R.

Table 2. Univariable and multivariable logistic regression on the risk factors of transmitted FQ-R patients(N = 295).

Characteristics	Acquired FQ-R (n = 197)	Transmitted FQ-R (n = 98)	OR [95% CI]	aOR [95% CI]^a
Demographic factors
Gender
Female	57 (28.9)	24 (24.5)	Ref
Male	140 (71.1)	74 (75.5)	1.26 [0.73, 2.21]
Age
≤ 24	24 (12.2)	16 (16.3)	Ref
25–44	67 (34.0)	37 (37.8)	0.83 [0.39, 1.77]
45–64	79 (40.1)	31 (31.6)	0.59 [0.28, 1.27]^a
≥ 65	27 (13.7)	14 (14.3)	0.78 [0.31, 1.92]
Household registration
Migrants	88 (44.7)	42 (42.9)	Ref
Residents	109 (55.3)	56 (57.1)	1.08 [0.66, 1.76]
Occupation
Commercial service	8 (4.1)	7 (7.1)	Ref
Student	8 (4.1)	5 (5.1)	0.71 [0.15, 3.22]
Office worker	5 (2.5)	8 (8.2)	1.83 [0.41, 8.71]
Labor worker/Farmers	46 (23.4)	18 (18.4)	0.45 [0.14, 1.44]^a
Retirement	43 (21.8)	18 (18.4)	0.48 [0.15, 1.55]
Unemployed/unknown/Others	87 (44.1)	42 (42.8)	0.55 [0.19, 1.67]
Clinical factors
TB treatment history
Previously treated	93 (47.2)	34 (34.7)	Ref
Treatment-naïve	104 (52.8)	64 (65.3)	1.68 [1.02, 2.80]^a	1.84 [1.09, 3.16]^b
Total diagnostic delay
≤ 2 weeks	57 (28.9)	25 (25.5)	Ref
2–4 weeks	41 (20.8)	21 (21.4)	1.17 [0.57, 2.37]
> 4 weeks	99 (50.3)	52 (53.1)	1.20 [0.68, 2.15]
First diagnostic unit
Tertiary hospitals	89 (45.2)	26 (26.5)	Ref
District-level hospitals	108 (54.8)	72 (73.5)	2.28 [1.36, 3.92]^a	2.69 [1.56, 4.75]^b
Sputum smear positive
No	60 (30.5)	26 (26.5)	Ref
Yes	135 (68.5)	72 (73.5)	1.23 [0.72, 2.14]
Unknown	2 (1.0)	0 (0.0)	—
Severe cases
No	150 (76.1)	75 (76.5)	Ref
Yes	47 (23.9)	23 (23.5)	0.98 [0.55, 1.72]
Treatment outcomt
Favorable outcomes	118 (59.9)	57 (58.2)
Unfavourable outcomes	79 (40.1)	41 (41.8)	1.07 [0.65, 1.76]
Bacteriological factors
Genotype
Non-Beijing	15 (7.6)	2 (2.0)	Ref
Beijing	182 (92.4)	96 (98.0)	3.96 [1.09, 25.43]^a
Pyrazinamide
Susceptible	85 (43.1)	33 (33.7)	Ref
Resistant	112 (56.9)	65 (66.3)	1.49 [0.91, 2.49]^a
Ethambutol
Susceptible	35 (17.8)	19 (19.4)	Ref
Resistant	162 (82.2)	79 (80.6)	0.90 [0.49, 1.69]
Streptomycin
Susceptible	46 (23.4)	7 (7.1)	Ref
Resistant	151 (76.6)	91 (92.9)	3.96 [1.82, 9.94]^a	3.69 [1.65, 9.42]^b
Main mutation of MDR-TB
rpoB S450L (RIF)
No	80 (40.6)	29 (29.6)	Ref
Yes	117 (59.4)	69 (70.4)	1.63 [0.98, 2.76]^a	1.52 [0.88, 2.67]
katG S315T (INH)
No	52 (26.4)	22 (22.4)	Ref
Yes	145 (73.6)	76 (77.6)	1.24 [0.71, 2.22]

^a Variables with p values < 0.2 in the univariable regression analysis.

^b Denotes statistical significance.

Abbreviations: RIF, Rifampicin; INH, Isoniazid; FQ-R, fluoroquinolones-resistant; OR, odds ratio; CI, confidence interval; Ref, reference.

As we used a strict definition of the transmitted FQ-resistant MDR-TB strains and excluded 41 FQ-R strains that did not share the same resistant mutation type in a single genomic cluster, we conducted a sensitivity analysis with the inclusion and exclusion of these 41 strains in the risk factor analysis. Similar results were identified in both multivariable models, except that the mutation of rpoB S450L appeared significant in the sensitivity analysis (Table 2 and Supplementary Table S8).

Discussion

There is an urgent need for related localized data to understand the FQ-R situation among MDR-TB in China to guide clinical treatment and implement more efficient interventions. Our findings show a high resistance rate of 34.7% for FQ among MDR-TB patients in urban China, with at least one-third of them likely caused by direct transmission of drug-resistant strains. Additionally, MDR-TB patients who were residents, previously treated, PZA resistance, and EMB resistance were associated with FQ-R.

The high proportion of FQ-R among MDR-TB patients in China is an obstacle to MDR-TB treatment and the implementation of short-term regimens that include FQ [9,19]. More than one-third of the MDR-TB in this study had FQ-R, which was significantly higher than the global rate reported by the WHO (20.0%) [20]. In China, FQ accounts for 49% of all antibiotics used in outpatient visits related to pneumonia and urinary tract infections [21], which has been reported to increase the rate of FQ-R [22]. Many provinces or cities, including Shenzhen, Hunan, Chongqing, and Beijing in China, also reported high FQ-R rates among MDR-TB, ranging from 28.6 to 43.6% [23–26]. In addition, in the 2013 national drug resistance survey, the resistance rates for ofloxacin and moxifloxacin in MDR-TB were 40.8% and 41.4%, respectively[11]. The high rate of FQ-R in MDR-TB patients in China indicates that including FQ could lead to ineffective treatments of MDR-TB and unfavourable outcomes. These findings also suggested an urgent need for FQ-R testing before initiating MDR-TB treatment regimens.

Of note was that at least one-third of the FQ-R was due to the transmission of FQ-R strains. The transmission of drug-resistant strains is considered a significant driver of the high prevalence of MDR-TB in China [15]. Our study found that FQ-R was significantly associated with second-line injectable drugs (KM, AM, and CM), especially the mutation of rrs 1401A > G. Furthermore, six clusters with transmitted FQ-R were all EDR-TB strains, indicating the direct transmission of EDR-TB strains (MDR-TB additional resistance to any FQ and to any of three second-line injectable drugs). It suggests that FQ-R transmission is serious in Shanghai, which may lead to more patients with FQ-R MDR-TB. Thus, the focus to mitigate the transmission of FQ-R should be the timely diagnosis of drug resistance including both MDR and FQ profiles. However, the current diagnosis of drug resistance to FQ is mainly based on DST. It is necessary to vigorously promote rapid diagnostic tools to provide evidence for the rational use of FQ in treating MDR-TB as soon as possible, to control the transmission of FQ-R strains.

Besides, the FQ-R patients’ first diagnostic unit in a district-level hospital was significantly associated with the transmission of FQ-R. In our study, the days of the health system delay were shorter in tertiary hospitals than in district-level hospitals (Supplementary Table S9). This may be because the timeliness of diagnosis in district-level hospitals is worse than those in tertiary hospitals, leading to delayed diagnosis and treatment and even transmission. In all, more tools and research are needed to predict further the high-risk factors and populations of FQ-R strains transmission and inform possible targeted interventions.

In our study, MDR-TB patients with a history of TB treatment were at increased risk for FQ-R. Similar situations have also been reported in the Beijing, Jiangsu, and Zhejiang provinces in China, including a history of FQ exposure or anti-tuberculosis drug use, retreatment, and rifampicin resistance that were associated with FQ-R [25,27,28]. This may be due to the use of FQ in the previous TB treatment. Empirical administration of FQ before the DST can lead to FQ-R, and even a short-term FQ use after the DST could also contribute to the development of FQ-R [8,29,30]. To understand how previous FQ use impacts the FQ-R patterns in MDR-TB patients, we used the Shanghai Health Information Network to find FQ medication records for the 295 FQ-R patients. Among the 68 patients with available records, the proportion of patients with previous FQ medication history was similar among genomic-unique FQ-R (23.7%, 37/156), acquired FQ-R (24.4%, 10/41), and transmitted FQ-R patients (21.4%, 21/98) (supplementary material part 2).

Also, resident patients were at high risk of FQ-R MDR-TB. In China, medical insurance is tied to household registration [31]. Individuals with Shanghai household registration are more likely to receive medical insurance support from healthcare institutions in Shanghai, and there is a broader array of designated medical facilities and pharmacies where drugs can be obtained [32]. Therefore, we cannot disregard the possibility that residents are more likely to be exposed to FQ during medical treatment. Besides, previous studies have shown that FQ is a commonly used antibiotic drug in hospitals or pharmacies in Shanghai [32], and even the FQ-R rate in RIF-susceptible TB is as high as 6% in the rural district of Shanghai [33]. Overall, easy access and inappropriate use of FQ can increase drug resistance risk, highlighting the importance of regulated use and strict management of FQ in China’s local medical institutions. In our study, PZA and EMB resistance were associated with FQ-R MDR-TB and EDR-TB patients. PZA and EMB are essential components of TB treatment regimens in China’s TB treatment guidelines [34,35]. PZA, in particular, is recommended as a component in the short- and long-term treatment regimens for MDR-TB treatment by WHO [9] and China [19]. Therefore, it is crucial to obtain resistance results before starting treatment for MDR-TB, including FQ, PZA, and EMB, and more importantly, the key drugs such as bedaquiline and linezolid before the initial of treatment.

Our study has several limitations. Firstly, due to the retrospective study design, not all of the patients have strains available for the final analysis, potentially underestimating the rate of FQ-R transmission. Secondly, data on previous FQ use before an MDR-TB diagnosis were not available, we cannot determine whether or not the FQ-R in the current study has resulted from previous treatment exposure. Finally, Shanghai has a high proportion of internal migrants. Some migrants may be lost after diagnosis. A prospective cohort study is necessary to further determine the source of FQ-R MDR-TB patients and quantify resistant strains’ transmission.

In conclusion, our study revealed a concerning prevalence of FQ-R in MDR-TB patients in Shanghai, China. One-third of FQ-R was presumed to be transmitted. These findings emphasize the urgent need for FQ-R surveillance and application management in newly diagnosed TB patients and to be vigilance against the transmission of drug resistance.

Authors’ contributions

M.L., Y.Z., X.S., and C.Y. conceived, designed, and managed the study. M.L., Y.Z., Z.W., R.S., Y.J, J.Y., J.L., H.L., R.Z., L.W., X.W., and F.Y. contributed to the data collection. M.L. and Y.Z. conducted data cleaning and data analysis. M.L. drafted the manuscript. M.L., Y.Z., X.S., C.Y., and Q.J. revised the manuscript. All coauthors reviewed and approved the final manuscript before submission.

Ethical approval

The study was reviewed and approved by the Ethical Review Committee at the Shanghai Municipal Center for Disease Control and Prevention (2011630).

Acknowledgments

We thank Lixin Rao from the Shanghai Municipal Center for Disease Control and Prevention and Jiazhen Liu from the Shanghai Municipal Health Statistics Center for their generous support and cooperation in searching the FQ medication record of FQ-R patients.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability

The datasets and sequencing data used and analyzed during the current study are not accessible online, but may be made available to the corresponding author upon written request.

Additional information

Funding

This work was supported by the National Key Research and Development Program of China [grant number 2023YFC2307305], National Natural Science Foundation of China [grant number 81872679 and 82373650], Science and Technology Innovation Plan of Shanghai Science and Technology Commission [grant number21DZ2202400], Shanghai Municipal Health Commission [grant number 202340298], Shenzhen Nanshan district San-Ming project [grant number SZSM202103008], “Pearl River Talent Plan” Innovation and Entrepreneurship Team Project of Guangdong Province [grant number 2021ZT09Y544], Shanghai three-year (2023-2025) action plan to strengthen the public health system [grant number GWVI-11.1-05], and the Top Young Talents in Shanghai [X.S.].