Multidrug-resistant tuberculosis imported into low-incidence countries—a GeoSentinel analysis, 2008–2020

Abstract

Background:

Early detection of imported multidrug-resistant tuberculosis (MDR-TB) is crucial, but knowledge gaps remain about migration- and travel-associated MDR-TB epidemiology. The aim was to describe epidemiologic characteristics among international travellers and migrants with MDR-TB.

Methods:

Clinician-determined and microbiologically confirmed MDR-TB diagnoses deemed to be related to travel or migration were extracted from GeoSentinel, a global surveillance network of travel and tropical medicine clinics, from January 2008 through December 2020. MDR-TB was defined as resistance to both isoniazid and rifampicin. Additional resistance to either a fluoroquinolone or a second-line injectable drug was categorized as pre-extensively drug-resistant (pre-XDR) TB, and as extensively drug-resistant (XDR) TB when resistance was detected for both. Sub-analyses were performed based on degree of resistance and country of origin.

Results:

Of 201 patients, 136 had MDR-TB (67.7%), 25 had XDR-TB (12.4%), 23 had pre-XDR TB (11.4%) and 17 had unspecified MDR- or XDR-TB (8.5%); 196 (97.5%) were immigrants, of which 92 (45.8%) originated from the former Soviet Union. The median interval from arrival to presentation was 154 days (interquartile range [IQR]: 10–751 days); 34.3% of patients presented within 1 month after immigration, 30.9% between 1 and 12 months and 34.9% after ≥1 year. Pre-XDR- and XDR-TB patients from the former Soviet Union other than Georgia presented earlier than those with MDR-TB (26 days [IQR: 8–522] vs. 369 days [IQR: 84–827]), while patients from Georgia presented very early, irrespective of the level of resistance (8 days [IQR: 2–18] vs. 2 days [IQR: 1–17]).

Conclusions:

MDR-TB is uncommon in traditional travellers. Purposeful medical migration may partly explain differences in time to presentation among different groups. Public health resources are needed to better understand factors contributing to cross-border MDR-TB spread and to develop strategies to optimize care of TB-infected patients in their home countries before migration.

Introduction

Tuberculosis (TB) continues to threaten public health globally. Although the reported incidence of drug-sensitive TB is declining in most regions of the world, TB remains one of the top 10 overall causes of death in low- and low-middle income countries.¹ Multidrug-resistant (MDR) TB is among the major obstacles to elimination of the disease. Despite recent advances in diagnostics and shortened treatment regimens, difficulties in providing universal access to diagnostics, drug susceptibility testing, effective treatment, considerable treatment-related toxicity and excessive treatment costs contribute to inferior outcomes for drug-resistant TB.^2–4 Consequently, mortality in individuals with MDR-TB is increased compared to drug-sensitive TB.¹

Migrants, in particular refugees and asylum seekers, are a difficult-to-reach subpopulation of travellers to middle- and high-income countries who may import TB and who have important implications for clinical care.⁵ Among patients with MDR-TB, the proportion of migrants is higher than among patients with drug-susceptible TB. In fact, in many high-income countries, most patients presenting with MDR-TB are foreign-born.^6,7 In low-incidence countries of Western Europe, for example, 57 to 100% of patients with MDR-TB were foreign-born in 2017. Nevertheless, the proportion of MDR-TB among all TB cases in this population remains low at 0.7 to 4.5%.⁸ Domestically acquired MDR-TB occurs less frequently than imported cases, and clusters are seen mainly within groups who migrated from the same country.^9,10 Asylum seeking for MDR-TB healthcare has been studied, but no robust evidence exists on the prevalence of this putative pattern of migration.⁶ Tuberculin skin test conversion rates have been studied in long-term travellers (travel longer than 3–6 months).^10,11 However, evidence of acquisition of MDR-TB during long-term travel such as tourism, business travel and volunteerism remains anecdotal.^12,13

The crucial importance of the prevention of cross-border spread as well as the early detection and efficient management of imported MDR-TB is recognized by stakeholders in low-incidence countries, but considerable knowledge gaps remain about the epidemiology of migration and travel-associated TB.^14,15 The aim of this analysis was to describe epidemiologic characteristics and migration history among international travellers and migrants with MDR-TB reported to the GeoSentinel Surveillance Network.

Methods

Patient records were extracted from GeoSentinel, a global surveillance network of 68 travel and tropical medicine sites in 28 countries on six continents.¹⁶ Patient records from January 2008 through December 2020 were reviewed. Records were included when there was a clinician-determined and microbiologically confirmed diagnosis of MDR-TB deemed related to travel or migration from a high-incidence country (annual TB incidence of ≥20/100 000 population). Immigration was defined as any entry of a foreign-born individual to a reporting state with the goal to remain (as opposed to short- or long-term travellers and expatriates). This definition thus comprised immigrants holding a visa, refugees and asylum seekers. Standard WHO definitions for drug-resistant TB were used: MDR-TB was defined as resistance to at least isoniazid and rifampicin. MDR-TB with additional resistance was categorized as (1) pre-extensively drug resistant TB (pre-XDR), defined as MDR-TB plus resistance to either any fluoroquinolone or second line injectable and (2) extensively drug resistant (XDR-) TB, defined as MDR-TB plus resistance to any fluoroquinolone and any second-line injectable drug (capreomycin, kanamycin and amikacin). This definition was determined for all cases after 2015 and for most cases before 2015 using supplemental antimicrobial susceptibility testing data. Traveller or migrant demographics, trip details and migration details (if available) were analysed. For foreign-born patients (those not born in the country of the GeoSentinel site they presented to), descriptive analyses of the time interval from immigration to the date of the first presentation to a GeoSentinel site and the date of symptom onset were performed. Sub-analyses were performed for three groups based on country of origin due to a large proportion of patients from the Republic of Georgia (Georgia) in the data especially from the Paris GeoSentinel site¹⁷: Georgia, the former Soviet Union (FSU) excluding Georgia, and other countries (countries not in the FSU). Temporal analyses related to the time point of travel were performed in relation to the degree of resistance by grouping pre-XDR and XDR-TB cases into a high-grade resistance group referred to as pre-XDR/XDR, comparing it with the remaining MDR patients. Drug-sensitive (DS) TB cases reported to the GeoSentinel database for the same time period were analysed for length of time from immigration to presentation at a GeoSentinel site and compared by frequency to MDR-TB cases.

Data analysis was descriptive. For data analysis, STATA^® 16 (StataCorp, College Station, TX, USA) was used. GeoSentinel’s data collection protocol has been reviewed by a human subjects advisor for the Centers for Disease Control and Prevention’s National Center for Emerging and Zoonotic Infectious Diseases and is classified as public health surveillance and not human subjects research. For sites located in countries where regulations require it, ethical clearance was obtained.

Results

Demographics

Of 275 778 records in the database during January 2008 to December 2020, 4067 had a diagnosis of active TB; 3865 (95.0%) with DS-TB and 202 (5.0%) with MDR-TB, the latter of which this report focuses on. One record (1/202) was excluded a priori as the patient’s diagnosis was not deemed to be related to travel or migration and the only reported cross-border travel was to another low incidence Western European country. In total, 184 of 201 (90.0%) patients’ TB isolates could be definitively classified by their resistance profiles: 136 were MDR-TB (67.7%), 23 were pre-XDR-TB (11.4%) and 25 were XDR-TB (12.4%) (Table 1). A total of 17 patients did not have detailed isolate resistance profiles reported. Patients were reported from 19 GeoSentinel sites, with centres in France (Paris, Bordeaux and Marseille) and Norway (Oslo) accounting for 143 (70.8%) of the patients. Other countries that reported more than five cases were Sweden (Stockholm), Australia (Melbourne), Italy (Brescia), Canada (Vancouver and Ottawa) and the UK (Cambridge). Most patients were male (123; 61.2%). There were six pregnant patients among the 78 females (7.7%). An extrapulmonary focus as part of the clinical syndrome was reported in 28 individuals (13.9%).

Table 1.

Demographics of patients with multidrug-resistant tuberculosis reported to GeoSentinel, 2008–2020 (n = 201)

Patient characteristics	n (%)
Male sex	123 (61.2%)
Median age, years [IQR]	31.5 [26–39]
HIV	8 (4.0%)
Insulin-dependent diabetes mellitus	3 (1.5%)
Pregnancy	6 (3.0%)
Degree of TB resistance
MDR	136 (67.7%)
Pre-XDR	23 (11.4%)
XDR	25 (12.4%)
Unspecified MDR/XDRa	17 (8.5%)
Localization of disease
Pulmonary	173 (86.1%)
Extrapulmonary	28 (13.9%)
Foreign-born (immigrant)	196 (97.5%)
VFR last 12 months	14 (7.1%)
WHO region and country of origin b
African	31 (15.4%)
Congo (DRC)	2 (1.0%)
Côte d’Ivoire	2 (1.0%)
Eritrea	15 (7.5%)
Ethiopia	4 (2.0%)
Mali	3 (1.5%)
Senegal	2 (1.0%)
Eastern Mediterranean	22 (10.9%)
Pakistan	4 (2.0%)
Somalia	15 (7.5%)
European	107 (53.2%)
Belarus	3 (1.5%)
France	4 (2.0%)
Georgia	47 (23.4%)
Lithuania	15 (7.5%)
Moldova	7 (3.5%)
Romania	7 (3.5%)
Russia	12 (6.0%)
Ukraine	4 (2.0%)
Uzbekistan	3 (1.5%)
South-East Asia	11 (5.5%)
India	6 (3.0%)
Myanmar	2 (1.0%)
Nepal	2 (1.0%)
Western Pacific	28 (13.9%)
China	6 (3.0%)
Mongolia	5 (2.5%)
Philippines	8 (4.0%)
Viet Nam	7 (3.5%)

Note. Data are numbers (n) and percentages (%) unless indicated otherwise. Abbreviations. n: number of patients; IQR: interquartile range; HIV: human immunodeficiency virus; TB: tuberculosis; MDR: multidrug-resistant; XDR: extensively drug resistant; VFR: visiting friends and relatives; DRC: Democratic Republic of Congo.

^a17 of 201 cases recorded from 2008–2014 could not be further classified.

^bOnly countries with two or more cases are shown. The denominator is the total number of patients (n = 201). When countries with only one case are included all WHO regions are represented.

Of 201 patients, 196 (97.5%) were immigrants born outside the reporting country (foreign-born). Five patients (2.5%) were born in one of the reporting low incidence countries—three reported potential risk exposure during long-term business-related travel, one while studying abroad, and one while visiting friends and relatives (VFR). Birthplaces of patients included all WHO regions. However, only two patients originated from the Americas (Guadeloupe (French West Indies) and Peru) (Table 1). A total of 47 patients were born in Georgia (47 of 201, 23.4%), 41 (87.2%) of which were reported by sites in France. An additional 45 patients (22.4%) originated from other FSU states. Among patients with detailed information on drug susceptibility, pre-XDR and XDR-TB were seen more frequently among patients from Georgia (26 of 45, 57.7%) and other republics of the FSU (12 of 41, 29.2%) when compared to patients from other countries (7 of 82, 8.5%) (Table 2). Among patients with pre-XDR and XDR-TB, 81% (21 of 26) of Georgian patients were male compared to 54 and 56% (7 of 13 and 5 of 9) of patients from the FSU and other countries, respectively.

Table 2.

Drug resistance profile and time interval from immigration to presentation analysed by region of origin among 175 foreign-born individuals with drug-resistant TB reported to GeoSentinel, 2008–2020

Region of origin			Time to presentation from immigration
			Median	IQR	<1 month		1–12 months		1–5 years		>5 years
	n	%	[days]	[days]	n	%	n	%	n	%	n	%
FSU (excl. Georgia)	41	23.4	227	23–825	13	31.7	10	24.4	10	24.4	8	19.5
MDR	29		369	84–827	6	20.7	8	27.6	9	31.0	6	20.7
Pre-XDR/XDR	12		26	8–522	7	58.3	2	16.7	1	8.3	2	16.7
Georgia	45	25.7	5	1–17	37	82.2	5	11.1	1	2.2	2	4.4
MDR	19		2	1–17	15	78.9	2	10.5	1	5.3	1	5.3
Pre-XDR/XDR	26		8	2–18	22	84.6	3	11.5	0	0	1	3.8
Other	89	50.9	307	94–1266	10	11.2	39	43.8	25	28.1	15	16.9
MDR	82		290	94–1266	8	9.8	38	46.3	21	25.6	15	18.3
Pre-XDR/XDR	7		485	4–1266	2	28.6	1	14.3	4	57.1	0	0
Total	175	100	154	10–751	60	34.3	54	30.9	36	20.6	25	14.3
MDR	130		220	43–827	29	22.3	48	36.9	31	23.8	22	16.9
Pre-XDR/XDR	45		10	4–67	31	68.9	6	13.3	5	11.1	3	6.7

Note. Only foreign-born patients with specified drug resistance status and known arrival date were included (n = 175): The median time to arrival was calculated for patients with a known arrival date (n = 175). Arrival date missing for one patient from the FSU (pre-XDR/XDR) and three from other regions (two MDR, one pre-XDR/XDR); drug resistance not further classified into MDR or Pre-XDR/XDR-TB in 17 of 201 cases from 2008 to 2014.

Abbreviations. n: number of patients; IQR: interquartile range; FSU: former Soviet Union; XDR: extensively drug resistant; MDR: multidrug-resistant.

Fourteen and 11 patients were reported to have travelled back to their country of birth as VFRs during the last 12 months and 1–5 years before presentation, respectively. An additional three had a potential exposure in a high-incidence country other than their country of birth: a Syrian patient spent 5 years in Russia for studies before immigrating to Northern Europe; a patient from Ivory Coast lived in China immediately before immigrating to Europe; and a Moldovan patient spent 2 months in Russia before immigrating to Central Europe.

Analyses related to the time point of immigration

The arrival date in the country of the GeoSentinel site was known for 175 of 196 (89.3%) foreign-born individuals. The median time interval from arrival to presentation to a GeoSentinel site was 154 days (interquartile range [IQR]: 10–751 days); 60 (34%) presented within 1 month after entering the country, 54 (31.8%) presented between 1 month and 1 year, and 61 (35%) presented more than a year after immigration (Table 2). In contrast, DS-TB cases had a median interval of 1292 days (IQR: 292–3583 days) from immigration to presentation at a GeoSentinel site. When analysed by region of origin, median time to presentation was 5 days (IQR: 1–17 days) for MDR-TB patients from Georgia, 227 days (IQR: 23–825 days) for patients from the FSU other than Georgia and 307 days (IQR: 94–1266 days) for patients from other countries. Among patients from FSU countries other than Georgia, the median time to presentation for patients with pre-XDR/XDR-TB was shorter than for patients with MDR-TB (26 days [IQR: 8–522] vs. 369 days [IQR: 84–827], respectively). In patients from Georgia, median time to presentation was very short in both MDR and pre-XDR/XDR groups (Table 2, Figures 1 and 2).

Figure 1.

Dropped-line plot depicting the individual temporal relation between symptom onset, immigration and clinical presentation among foreign-born patients with known symptom onset (n = 87). One outlier from the group ‘Other’ with a presentation almost three decades after immigration was removed for better scaling. Every horizontal bar represents an individual patient. The intercept (y = 0) represents the time point of immigration. The red dots depict the time point of symptom onset, and the black dots represent the time point of presentation at the reporting site. High colour intensity of the bar represents ongoing symptoms, and low colour intensity represents the asymptomatic interval since immigration

Figure 2.

Correlation of time intervals from immigration to presentation (x, logarithmic) and from symptom onset to immigration (y, linear) for patients with MDR-TB (left) and higher degree of resistance (pre-XDR/XDR, right). Intercept (y = 0) = time point of immigration

Among foreign-born individuals with detailed resistance information, 87 of 175 (49.7%) had a recorded date of symptom onset. Of these, 42 (48.3%) reported symptom onset before immigration, 34 of which (81%) presented to the clinic within 1 month. Patients originating from Georgia clustered as early presenters with long-standing symptoms in relation to the time point of immigration and 37 of 45 (82%) presented within 1 month (Figures 1 and 2). No similar subcohort of early presenters with long-standing symptoms was observed among DS-TB patients. When analysed by resistance, 22 of 29 (75.8%) patients with pre-XDR and XDR-TB reported symptom onset before immigration, while 19 of 58 (32.8%) patients with MDR-TB reported symptom onset before immigration.

Discussion

This descriptive analysis of patients presenting with MDR-TB at GeoSentinel travel and tropical medicine sites showed that almost all cases occurred among foreign-born individuals. Given the known median incubation period of a few months to 2 years for active TB, the observed lag time between arrival and presentation in our cohort suggests that at least half of our patients were likely exposed in their country of origin.¹⁸ However, it cannot be ascertained whether exposure occurred before immigration, during transit, while living in the new country of residence or while visiting friends or relatives back in the country of origin. More than half of TB cases in migrants occur five or more years after immigration, underlining the importance of latent TB infection screening and treatment programs.¹⁹ A recent comprehensive whole-genome-sequencing-based outbreak analysis highlights the complexity of transmission dynamics and the potential use of whole genome sequencing to narrow down the region of acquisition via phylogenetic markers.²⁰ Only 7.1% of patients reported VFR-travel to high-incidence countries in the past year, which is lower than the 20% to 55% of all TB cases in VFRs reported in the literature.^21,22 This difference may be explained by different travel patterns of patients attending our GeoSentinel clinics.

None of the cases in our analysis were short- or long-term tourists or missionaries/volunteers; three occurred among business travellers and one in a student. This finding is in line with previous reports of a relatively low absolute risk of acquiring tuberculosis during short-term and long-term travel.^10,23 Specifically, a prospective trial from The Netherlands reported an increased relative risk of infection with a tuberculin skin test conversion rate of 2.8 per 1000 person-months of travel in non-healthcare workers spending 3 to 12 months in high-endemicity countries (incidence >1000 cases per 100 000), thus approaching the estimated infection rate of the host country.^11,24 The US Centers for Disease Control and Prevention and others have issued detailed recommendations for the prevention of TB exposure for international travellers anticipating prolonged TB exposure during their travels.^25,26

Our analysis included a large number of patients from Georgia, who mainly presented to French sites with active TB and high levels of resistance (pre-XDR and XDR-TB). Median time from immigration to presentation was extremely short for this group, and almost all patients had symptom onset preceding immigration. In 2018, 31% of all MDR-TB cases in Georgia were reported among patients with previous anti-TB treatment as per the WHO.²⁷ The high proportion of Georgian patients with pre-XDR and XDR-TB in our cohort suggests a high rate of previous treatment. In fact, 67% (42 of 63) of all MDR/XDR-TB patients treated at Pitié-Salpêtrière Hospital in Paris between 2011 and 2016 reported previous treatment and 63.5% were from Georgia (40 of 63).²⁸ High rates of person-to-person spread of XDR-TB strains in Georgia could also explain this finding, but there is no evidence in the literature to date to support this. A common perception among Georgian patients is that the tailored treatment regimens available in France are more efficient and less toxic than the treatment offered by the national treatment program in Georgia, which may explain the focused migration to France for primary or, in most cases, repeated courses of treatment (Eric Caumes, personal communication—9 June 2020).

Differences were identified in temporal presentation patterns between patients from different regions of origin, with a short time to presentation among XDR-TB patients originating from the FSU. Similar to our findings, in a study from France reporting data from 2006–2012, 34.4% of patients from the FSU (for whom the date of arrival to France was known) were hospitalized within a month of arrival and subsequently diagnosed with MDR-TB.¹⁷ In another case series of 20 XDR-TB patients arriving from the FSU to France, the median time from arrival to hospitalization was 2 days (IQR: 1–7 days).²⁹ It is unclear why XDR-TB patients from the FSU and specifically Georgia present for care within such a short time interval in comparison to other groups, but this may be due to differences in TB screening protocols in destination countries or purposeful medical migration for TB care. Of note, DS-TB patients reported to GeoSentinel presented much later after immigration compared to MDR-TB (median 1292 days [IQR: 292–3583 days] vs. 154 days [IQR: 10–751 days]) and no subcohorts of patients with long-standing symptoms before immigration were identified. However, no statistical comparisons can be made between MDR-TB and DS-TB since GeoSentinel data are not systematically reported and the observed difference may partly be a result of referral bias to GeoSentinel’s specialized infectious disease centres, e.g. in the event of difficult-to-treat TB.

National treatment programs in high-burden countries may fail to provide true patient-centred care, leading to loss of faith in treatment success, fear of catastrophic costs and persistent stigma, which may create a potential driving force for medical migration.³⁰ The international public health community should continue to work with high TB-burden countries to identify patients at high risk and provide both robust treatment regimens and directly observed treatment to reduce both the incidence of MDR-TB and medical migration.

This study has several limitations. The majority of cases were reported from only two countries, France and Norway, and few cases were reported from sites in North America and Asia, thus decreasing generalizability. The few cases reported from US sites are likely a reflection of the use of specialized TB health department clinics upon migration that are not affiliated with GeoSentinel, and the same may apply in other locations. A general limitation of GeoSentinel is that the majority of the network’s sites are found in Western Europe and North America (Northern Hemisphere dominance). Although a third of patients in this study show a pattern suggestive of medical migration, we were unable to determine the relative contribution of medical migration to the MDR-TB incidence in low-burden countries. By virtue of migration patterns, populations originating from specific countries might be over-represented at certain GeoSentinel sites. Given the long incubation period of tuberculosis, the causal link to travel or migration cannot be ascertained with certainty, as previously mentioned. In addition, time to presentation after immigration may have been overestimated in patients who might have been seen at another clinic before presentation at the GeoSentinel site. This may apply to DS-TB patients in particular. Finally, GeoSentinel is not designed to collect detailed clinical data such as molecular epidemiological data, or outcome information. It also does not collect past treatment received, which is an important risk factor for MDR-TB acquisition.

In conclusion, traditional travel was very rarely associated with the acquisition of MDR-TB in this multinational study. Purposeful medical migration for MDR-TB treatment and retreatment may partly explain the variation in time to presentation among different groups of migrants. Public health resources are needed to better understand factors contributing to the cross-border spread of MDR-TB and to develop strategies to optimize care of TB-infected patients in their home countries before migration.